Developments in Civil Engineering

Vol. 1 The Dynamics of Explosion and its U s e (Henrych)

Vol. 2 The Dynamics of Arches and Frames (Henrych)
Vol. 3 Concrete Strength and Strains ( A v r a m e t a l . )
Vol. 4 Structural Safety and Reliability (Moan and Shinozuka, Editors)
Vol. 5 Plastics in Material and Structural Engineering (Bares, Editors)
Vol. 6 Autoclaved Aerated Concrete, Moisture and Properties (Wittmann, Editor)
Vol. 7 Fracture Mechanics of Concrete (Wittmann, Editor)
Vol. 8 Manual of Surface Drainage Engineering, Volume II (Kinori and Mevorach)
Vol. 9 Space Structures (Avram and Anastasescu)
Vol. 10 Analysis and Design of Space Frames by the Continuum Method (Kollar and Hegedus)
Vol. 11 Structural Dynamics (Vertes)
Vol. 12 The Selection of Load-Bearing Stuctures for Buildings (Horvath)
Vol. 13 Dynamic Behaviour of Concrete Structures (Tilly, Editor)
Vol. 14 Shells, Membranes and Space Frames (Heki, Editor)
Vol. 15 The Time Factor in Transportation Processes (Tarski)
Vol. 16 Analysis of Dynamic Effects on Engineering Structures (Baia and Plachy)
Vol. 17 Post-Buckling of Elastic Structures (Szabo, Gaspar andTarnai, Editors)
Vol. 18 Fracture Toughness and Fracture Energy of Concrete (Wittmann, Editor)
Vol. 19 Pavement Analysis (Ullidtz)
Vol. 20 Analysis of Skeletal Structural Systems in the Elastic and Elastic-Plastic Range (Borkowski)
Vol. 21 Creep and Shrinkage of Concrete Elements and Structures (Smerda and Kfistek)
Vol. 22 Theory and Calculation of Frame Structures with Stiffening Walls (Pubal)
Vol. 23 Time Effects in Concrete Structures (Gilbert)
Vol. 24 Stresses in Layered Shells of Revolution (Kovaf ik)
Vol. 25 River Intakes and Diversion Dams (Razvan)
Vol. 26 Analysis of Dimensional Accuracy of Building Structures (Vorlicek and Holicky)
Vol. 27 Reinforced-Concrete Slab-Column Structures (Ajdukiewicz and Starosolski)
Vol. 28 Finite Models and Methods of Dynamics in Structures (Henrych)
Vol. 29 Endurance of Mechanical Structures ( N e m e c and Drexler)
Vol. 30 Shells of Revolution (Mazurkiewicz and Nagorski)
Vol. 31 Structural Load Modeling and Combination for Performance and Safety Evalution (Wen)
Vol. 32 Advanced Analysis and Design of Plated Structures (Kfistek and Skaloud)
Vol. 33 Regular Lattice Plates and Shells (Sumec)
Vol. 34 Combined Ultrasound Methods of Concrete Testing (Galan)
Vol. 35 Steel-Concrete Structures for Multistorey Buildings (Kozak)
Vol. 36 Analytical Methods in Bin-Load Analysis (Drescher)
Vol. 37 Design of Welded Tubular Connections - Basis and U s e of AWS Code Provisions (Marshall)
Vol. 38 Fresh Concrete - Properties andTests (Bartos)
Vol. 39 Stability, Bifurcation and Postcritical Behaviour of Elastic Structures (Pignataro, Rizzi and Luongo)
Vol. 40 Cable-Stayed Bridges - Recent Developments and their Future (Ito et al., Editors)
 DESIGN OF WELDED
TUBULAR CONNECTIONS
Ba
ss
i and Use of AWS Co
dePr
ovs
io
in
s
PETER WILLIAM M A R S H A L L
Civil Engineering Consultant, Shell Oil Company, Houston, Texas, U.S.A.

ELSEVIER
A m s t e r d a m - L o n d o n - N e w York - T o k y o
1992
E L S E V I E R S C I E N C E P U B L I S H E R S B.V.
Sara Burgerhartstraat 25
P.O. Box 211, 1000 A E Amsterdam, The Netherlands

ISBN: 0 444 88201 4

1992 Elsevier Science Publishers B.V. All rights reserved

N o part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or
by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written
permission of the publisher, Elsevier Science Publishers B.V., Copyright & Permissions Department,
P.O. Box 521, 1000 A M Amsterdam, The Netherlands.

Special regulations for readers in the U . S . A . - This publication has been registered with the Copyright
Clearance Center Inc. (CCC), Salem, Massachusetts. Information can be obtained from the CCC about
conditions under which photocopies of parts of this publication may be made in the U . S . A . All other
copyright questions, including photocopying outside of the U . S . A . , should be referred to the copyright
owner, Elsevier Science Publishers B.V., unless otherwise specified.

N o responsibility is assumed by the publisher for any injury and/or damage to persons or property as a
matter of products liability, negligence or otherwise, or from any use or operation of any methods,
products, instructions or ideas contained in the material herein.

Printed in The Netherlands



PREFACE

Although tubular structures are reasonably well understood by designers of offshore

platforms, onshore applications often suffer from "learning curve" problems, particularly in the
connections, tending to inhibit the wider use of tubes. This book was written primarily to help
remedy this situation by the principal author of the AWS D l . l Code provisions for tubular
structures.
The intended audience is users of the Code: designers of offshore platforms, designers of
significant onshore tubular structures, and engineers involved in formulating company guidelines
for these applications. Writers of other codes and graduate students and researchers in the area of
tubular structures will also find it useful as a source of background material.
This book is intended to be used in conjunction with the AWS Structural Welding Code -
Steel, AWS Dl.1-90, published by the American Welding Society, Miami. It relies on the use of
Code material which is not reproduced herein.
The manuscript was prepared as a PhD dissertation for the Department of Architecture,
Kumamoto University, Kumamoto, Japan. The author is grateful to his committee chairman,
Professor Yoshiaki Kurobane for inspiring this effort, and to Professor Joseph A. Yura,
University of Texas, and Professor Jaap Wardenier, Delft University of Technology, for their
input and guidance during the preparation of the manuscript. Charles Spitzfaden and Yolanda
Estrello assisted with drafting and word processing, respectively, and Joop Paul proofread the
completed work.

RECOMMENDED COMPANION REFERENCES

1. AWS Structural Welding Code - Steel, AWS D l . 1 - 9 0 , American Welding Society,

Miami, 1988.
2. AISC Manual of Steel Construction, Eighth Edition, American Institute of Steel
Construction, Chicago, 1980.
3. D. R. Sherman, Tentative Criteria for Structural Applications of Steel Tubing and Pipe,
AISI Committee of Steel Pipe Producers, August, 1976.
Chapter 1

INTRODUCTION TO TUBULAR STRUCTURES

1.1 ATTRIBUTES OF TUBES

Tubular members benefit from an efficient distribution of their material, particularly in
regard to beam bending or column buckling about multiple axes. For architecturally exposed
applications, the clean lines of a closed section are aesthetically pleasing, and minimize the
amount of surface area for dirt, corrosion, or other fouling. Simple welded tubular joints can
extend these clean lines to include the structural connections. With circular tubes, reduced drag
forces also apply for wind, waves, and blast loadings.

1.2 ARCHITECTURAL AND STRUCTURAL FORMS

1.2.1 Onshore Applications

Tubular columns are extensively used in high-clearance single story buildings, such as
shopping malls and warehouses. Here radius of gyration is more important than section area, and
the connections are simple and straightforwardfillet welded base plates and shear plates for
bolting to beam webs.
Tubular designs are also widely used for lightweight long span structures, such as
expressway overhead signs, pedestrian bridges, booms for construction cranes and mining
draglines, drilling derricks, radio masts, and the like. They have also been proposed for orbiting
space stations.
Tubular space frames are increasingly finding use in such dramatic and monumental
architectural applications as long-span roofs, atrium skylights, radio-telescope dish antennas,
Olympic ski-jumps, space-shot launching towers, and spectacular looping amusement park rides.
Like other rolled shapes, rectangular tubes offer simple welded connections in orthogonal planes.
However, for the truly unusual structure, circular tubes offer simple welded connections in any
orientation desired.
Unfortunately, the potential elegance of these structures is often spoiled because of
problems with the connections. The designer may lack confidence in simple direct welded
connections, and devise an awkward, ugly gusseted joint to do the same job. The fabricator may
be unprepared for the specialized layout, cutting, fitting, welding, and inspection tasks involved.
The erector may require bolted field connections. Finally, the project may become embroiled in
a dispute with officials who are also not fully prepared to deal with the specialized technology
involved.
Solutions to these problems are covered by the "Tubular Structures" section of the
American Welding Society D l . l Structural Welding Code - Steel. Much of the technology from
which this part of code evolved was developed by the offshore oil industry, as reflected in the
parallel provisions of API RP 2A, Recommended Practice for Planning, Designing, and
Constructing Fixed Offshore Platforms.

1.2.2 Offshore Applications

Thousands of large tubular structures have been built for offshore oil drilling and
production in the last forty years. The typical structure consists of a tubular space frame, or
jacket, which extends from the seafloor to just above the sea surface. This is usually fabricated
in one piece onshore, transported by barge, launched at sea, and upended on site by partial
flooding. Tubular piling are driven through the jacket legs to resist vertical gravity loads and
2

Fig. 1.1. Onshore applications of tubular structure, (a) Firth of Forth railway bridge, Scotland,
1880's. (b) Atrium space frame, Houston, 1980's.
 3

lateral storm loads. To complete the structure, a working deck section is added, usually a
composite of tubular members and conventional rolled sections (ref. 1).
Tubular construction is also used for the lattice legs of jack-up mobile drilling units, and
for the interconnecting space frame of column- stabilized semisubmersibles, a class of floating
drilling rigs.
Early development of offshore technology was largely a trial and error experience.
Structural design was not so much governed by official regulations as it was by the desire on the
part of offshore operators to protect their own considerable investment. The collapse of even a
small drilling/production platform involved a loss of tens of millions of dollarsincluding, in
addition to the structure itself, equipment, wells, clean-up costs, and loss of income. For today's
deepwater structures, the loss can exceed $1 billion. Because a degree of uncertainty exists in
both the strength of structures and the magnitude of applied loads, the risk of structural failure is
not totally eliminated by the inclusion of a safety factor. Rather, an attempt is made to select
design criteria on a rational economic basis; that is, to minimize the sum of first cost plus
deferred future risks (ref. 2).
In making such trade-offs, the optimum point is not sharply defined; thus calculation of
the probability of failure need not be absolutely precise in order to serve its purpose.
Furthermore, the reliability viewpoint provides a useful rationale, in that it forces one to examine
the bias and uncertainty at each step of the way. This rationale has proven useful in interpreting
research results and defining the design criteria we now use. Finally, there are social constraints
present which make it unpalatable to make trade-offs between dollars and human safety or
environmental pollution. The safety index is a useful measure of structural reliability for this
purpose, without the legal, social, and psychological impact of probabilities of failure.
We can define the safety index as the expected value of the margin between real load
and real resistance, expressed in units of the standard deviation of total uncertainty. For onshore
public structures, the safety index ranges from 2.5 to 4.0, and failures are so rare that their
statistics are not well defined. For new offshore platforms designed for the 100-year storm, the
safety index ranges from 2 to 3 in terms of the lifetime risk of overload failure; the corresponding
average annual loss rate is on the order of 0.1% or less. This is low enough that overload is not
the dominant risk; blowouts, fires, and collisions account for more of the catastrophic losses.
Offshore structures were not always this reliable. Early joint design consisted of the
instruction: "cope to fit and weld solid". Tubular braces were simply welded to the jacket legs,
which served as the main member at the tubular connection without any reinforcement. After
several hurricanes, recurring failure modes became apparent in these simple connections. As will
be discussed in subsequent chapters, these include local punching-shear/pullout failure in the
main member, general collapse of the main member, progressive failure of the weld, and lamellar
tearing. Materials problems were also experienced, including poor weldability and brittle
fracture. Although fatigue failure has been an ongoing concern of research over the last 20 years,
this geriatric mode of failure has only recently begun to be observed in actual structures.

1.3 T H E NEED FOR AN INTEGRATED APPROACH

Despite the availability of codes of practice like AWS D l . l , welded structural

connections in tubular space frames have developed a certain mystique. This is no doubt
enhanced by a number of spectacular problems which have occurred. A few have resulted in
structural collapse, while many others spelled financial disaster for the contractor involved.
Often, when a welded tubular connection fails, the fracture is in the heat affected zone at the toe
of the weld joining a branch member or attachment to the main tube. The designer involved may
seize upon this fact to attribute the failure to faulty materials or welding, and elaborate
metallurgical witch hunts may be staged to bolster this claim. Never mind that the weld toe is
4

Fig. 1.2. Offshore applications of tubular structure, (a) Topsides of self-contained drilling and
production platform, (b) Space frame of semi-submersible drilling rig. (c) Fish-eye view of
8-leg platform for 100m water, (d) Bullwinkle jacket for 400m water.
 5

also the site of stress concentrations which are so high that most practical connections experience
localized plastic straining before reaching their design load. The lawyers and their expert
witnesses get rich, and the mystique grows.
Perhaps to a larger degree than with other structural forms, welded tubular connections
require an integrated approach to fracture control. Design, material selection, fabrication,
welding, and inspection must all be consideredand they are interrelated. Responsible design
includes more than using stress analysis calculations to dimension the main structural elements.
Connections require equal attention, if not more. The designer must understand the demands he
implicitly places on the materials to be used, e.g., ductility as well as yield strength and
availability; and he must anticipate the methods of fabrication and welding, their limitations, and
their effects on service performance. The designer who blindly uses the code formulas is a
failure waiting to happen. If only to protect themselves, the practical materials and welding
people who follow in executing his design should also understand what demands are being
placed upon their part of the overall fracture control picture.

1.4 AUTHOR'S VIEWPOINT FOR THIS MONOGRAPH

The architecture of tubular structures has fascinated the author through his career as a
structural engineer. "Architecture" is defined as the art and science of designing and successfully
executing structures in accordance with aesthetic considerations and the laws of physics, as well
as practical and material considerations.
Onshore, where tubular structures are often exposed for dramatic effect, it has often
been painful to see grand concepts fail in execution due to problems in the tubular joints, or
structural connections. Such "failures" range from awkward detailing, to "learning curve"
problems during construction, to excessive deflections or collapse.
Offshore, the oil industry went through the painful stage about 20 years ago. Research,
testing, and practical applications have progressed to the point where tubular connections are
about as reliable as the other structural elements which engineers normally deal with. The author
participated in the resolution of the problem areas, synthesizing and putting into practice the
research of such pioneers as Toprac, Bouwkamp, and Pickett. His joint designs and design
procedures are part of most of Shell's large Gulf of Mexico platforms, including the world record
Bullwinkle jacket in 1350-ft. water depth, as well as the Brent "A" platform offshore from
Scotland (famous for its widely quoted "North Sea Brent" crude oil price marker).
The art and science of welded tubular connections which emerged from this effort has
been codified in AWS D l . l (ref. 3). This Monograph will describe, from the viewpoint of a
primary participant, the conceptual basis and historical development of the code, including recent
revisions. It draws heavily on the author's previous work, notably the 1984 Houdremont lecture
(ref. 4), and on his three chapters in McClelland's book on offshore platforms (ref. 5).
Although there will be updating and expansion upon the previous work, and an effort to
compare the Code with some of the v o l u m i n o u s new data coming forth, no claim of
comprehensiveness in this regard is made. Recent, more exhaustive reviews of the worldwide
data base can be found in Wardenier (ref. 6) and in Billington, Tebbett, and Lalani (ref. 7).
Similarly, this work will focus on tubular connections, rather than design of tubular
members, save for the broad remarks which follow. Fully detailed background and justification
for these would take up another book.

1.5 TUBULARS AS STRUCTURAL MEMBERS

API Recommended Practices for the Planning, Designing, and Constructing of Fixed
Offshore Platforms, API RP 2A, (ref. 8), gives detailed guidance for tubular structures as used
6

offshore. With few exceptions, structural steel design follows the basic allowable stresses of the
AISC Specification for the Design, Fabrication and Erection of Structural Steel for Buildings,
extending these criteria to tubular members.
The AISC Steel Construction Manual (ref. 9) lists dimensions and section design
properties for a number of tubular sections. Standard weight, extra strong, and double extra
strong circular sections from half-inch to 12-inch nominal diameter are widely available from
stock, particularly in mild steel grades, 35 to 36 ksi yield strength (246 to 253 MPa). In the U.S.,
commonly used larger sizes include diameters and wall thicknesses as listed in Table 1.1.
In offshore practice, still larger sizes are custom fabricated from plate, typically in 6-
inch (152mm) increments of diameter and 0.125-inch (3mm) increments of wall thickness.
Diameter/thickness ratios commonly range from 20 (a limit for cold-straining) to 60 (a limit for
local buckling).

TABLE 1.1 PROPERTIES OF COMMONLY USED SIZES OF STRUCTURAL PIPE

WALL MOM. OF SECTION RADIUS

O.D. THICK. AREA WEIGHT INERTIA MODULUS GYRATIC
INCHES IN. SQ. IN. LB/FT IN.-4TH IN.-3RD IN.

6 5/8 .280 5.58 19.0 28.1 8.4 2.24

6 5/8 .432 8.40 28.6 40.4 12.2 2.19
6 5/8 .562 10.70 36.4 49.6 14.9 2.15

8 5/8 .322 8.39 28.6 72.4 16.8 2.93

8 5/8 .406 10.48 35.6 88.7 20.5 2.90
8 5/8 .500 12.76 43.4 105.7 24.5 2.87
8 5/8 .718 17.83 60.6 140.5 32.5 2.80

10 3/4 .365 11.90 40.5 160.7 29.9 3.67

10 3/4 .500 16.10 54.7 211.9 39.4 3.62
10 3/4 .593 18.92 64.3 244.8 45.5 3.59

12 3/4 .375* 14.57 49.6 279.3 43.8 4.37

12 3/4 .500 19.24 65.4 361.5 56.7 4.33
12 3/4 .687 26.03 88.5 475.1 74.5 4.27

14 .375* 16.05 54.6 372.7 53.2 4.81

14 .438* 18.66 63.4 429.4 61.3 4.79
14 .500 21.20 72.1 483.7 69.1 4.77
14 .750 31.21 106.0 687.3 98.1 4.69

16 .375* 18.40 62.6 562.0 70.2 5.52

16 .438* 21.41 72.8 648.7 81.0 5.50
16 .500* 24.34 82.8 731.9 91.4 5.48
16 .656 31.62 108.0 932.3 116.5 5.42

18 .375* 20.76 70.6 806.6 89.6 6.23

18 .500* 27.48 93.4 1053.1 117.0 6.18
18 .625 34.11 116.0 1289.0 143.2 6.14

20 .375* 23.12 78.6 1113.4 111.3 6.93

20 .500* 30.63 104.0 1456.8 145.6 6.89
20 .593* 36.15 123.0 1703.7 170.3 6.86
20 .812 48.94 166.0 2256.7 225.6 6.79

24 .375* 27.83 94.6 1942.3 161.8 8.35

24 .500* 36.91 125.0 2549.3 212.4 8.31
24 .687* 50.31 171.0 3421.2 285.1 8.24
24 .750* 54.78 186.0 3705.4 308.7 8.22
24 .968 70.04 238.0 4652.6 387.7 8.15
24 1.000 72.25 246.0 4787.0 398.9 8.13

NOTE: 1 INCH = 25.4mm

*D/t of 30 to 60; semi-compact section (limited plastic rotation capacity)

 7

The AISC manual also lists a large number of square and rectangular sections and their
design properties. However, some of the sections listed have only limited availability. Again,
larger sections can be fabricated from plate.

1.5.1 Columns
Realistic design for axial compression must reflect the fact that the strength of actual
columns is significantly below both of the two theoretical bounds yield and elastic buckling.
This departure is due to variations in material properties (static yield strength versus the
conventional rapid tension test) and imperfections (centerline crookedness, out-of-roundness, and
misalignment of adjacent material at butt joints), as well as residual stress.
The AISC design curve, and the original CRC column curve upon which it is based,
reflect such considerations and are based on a large number of column tests, representing a
variety of sections-hot rolled and welded shapes; open, closed, and solid sections; and both mild
and high strength steel; as shown in Figure 1.3(a).
Large tubular columns were not well represented in the original data base. Welded
tubes differ from hot rolled sections in possessing significant residual stresses, which promote
earlier yielding and lower column strengths. Figure 1.3(b) shows the pattern of residual stresses
in a welded box column and a fabricated tube (ref. 10). In addition to the mean longitudinal
stresses shown, circumferential residual stresses due to cold forming of the plate also exist,
varying through the thickness in a pattern typical of plastic bending followed by springback, for
the circular tube.
Column behavior for the fabricated box sections falls significantly on the unsafe side of
the CRC curve as shown in Figure 1.4. Tests on small cold formed circular tubes also suggested
a lower design curve (ref. 11). Faced with this, the author prevailed upon API to sponsor a series
1.4r

0 .2 .4 .6 .8 1.0 1.2 1.4 1.6

(a) DESIGN CRITERIA

TENSION

(b) RESIDUAL STRESS

Fig. 1.3. Column stability considerations for tubular structures (from ref. 10).
8

of tests on fabricated pipe columns at Lehigh University, results of which are also shown (ref.
12). The large range covered by each data plot indicates the range of ambiguity in test
interpretation, due to differences between static and conventional dynamic yield strengths, and to
friction in the spherical end bearings affecting the effective column length.

DIMENSIONLESS SLENDERNESS RATIO

Fig. 1.4. Column buckling curves.

Using advanced analytical methods, Chen et al were able to match experimental test
results within a few percent (ref. 13), when actual imperfections and residual stresses in the test
specimens were taken into account. Chen then used this same analytical method to produce
curve "A" in Figure 1.4, for members just meeting code fabrication tolerances. Since this falls
remarkably close to the 1.67 times the AISC design criteria, offshore design practice continues to
follow AISC.
The author has not had a similar degree of involvement with criteria development for
square and rectangular hollow sections. Most such sections currently available in the U.S. are
cold finished. This raises the tensile yield strength, but produces a "round house" stress-strain
curve and complex residual stress patterns, so that the relative column behavior is less favorable.
American (ref. 14) and European (ref. 15) sources suggest the use of lower column design curves
for this application, as indicated by the AISI and ECCS curves in Figure 1.4. A Canadian review
of over 300 tests (ref. 16) also suggests the use of multiple column curves, depending on the
method of tube manufacture.
Tubular struts with welded end connections enjoy a degree of end fixity which permits
the use of effective length factors "k" less than unity. For example, API RP 2A recommends "k"
of 0.8 for primary bracing which frames into the larger, stiffer legs of offshore jackets, using
connections which substantially match the strength of the sections joined. For other types of
tubular structures, applicability of "k" factors less than unity will largely offset the penalty of
having a lower column design curve. See Table 1.2.
 9

Although the AISC code permits columns with slenderness ratios, kL/r, up to 200,
circular tubular members subject to wind action should observe lower limits in order to avoid
vortex induced vibrations. The traditional limit for offshore jackets is kL/r of 120; this
corresponds to a critical wind speed of 18 mph (8m/s) and suffices for short construction periods
at sites that are not too windy. Members violating this limit frequently vibrate, and some have
suffered fatigue cracks. Theoretically, dense members with a lot of damping should be able to
withstand wind speeds above critical, without excessive vibration. However, welded members
have very low damping, as low as 0.1% of critical, so that only members having D/t ratios less
than 16 would be dense enough to avoid the problem. For windy construction sites, with
consistent winds of 30 mph (14m/s), a few members with kL/r greater than 90, and D/t of 30 to
60, have encountered vibration problems. Slenderness ratios, kL/r, of 60 or less would be
required for lifetime exposure to winds having sustained speeds of up to 70 mph (60m/s),
especially for members having low density (high D/t).

TABLE 1.2 EFFECTIVE LENGTH FACTOR k

SITUATION AMERICAN ( R E F . 8 ) OVERSEAS ( R E F . 1 5 )

CHORD OF TRUSS IN-PLANE 1 . 0 TO NODES MAY BE < 1 . 0 CONSIDERING RESTRAINT

CHORD OF TRUSS 0UT-0F-PLANE 1 . 0 TO BRACING POINTS PROVIDED BY WEB MEMBERS ( R E F . 2 8 )

WEB MEMBERS IN-PLANE 0.8 0.7

WEB MEMBERS 0UT-0F-PLANE

TUBULAR CHORDS 0.8 0 . 7 W/0VERLAP, > 0 . 6 ( R E F . 2 9 )
OPEN SECTION CHORDS 1.0
X-BRACES 0 . 9 OF SHORTER HALF, COUNTER I N TENSION
SECONDARY BRACING 0.7

PORTAL SIDESWAY COLUMNS > 1 . 0 USE AISC ALIGNMENT CHART

1.5.2 Bending
(i) Circular. In the range where structural pipe may be treated as a compact section-
-that is, no local buckling occurs well into the plastic rangewe can take advantage of the
favorable plastic bending shape factor, Z/S, for tubes (ref. 17).

= (1 + (1.1)
S D

Typical values for tubes listed in the AISC manual range from 1.30 up. About 96% of the fully
plastic moment is developed at only twice yield strain. Thus, on the surface, the bending
allowable of 0.75 Fy, corresponding to a shape factor of 1.25 seems quite reasonable, consistent
with a bending allowable of 0.66 Fy for compact wide flange shapes. A difficult problem,
however, lies in the definition of a D/t ratio below which members may be considered as
compact.
Let us consider the range of behavior in bending for tubes with various D/t ratios, as
shown in Figure 1.5 (ref. 18). For very stocky sections, we do not have to worry about local
buckling. The moment-curvature (M-phi) behavior is fairly linear up to the yield moment. A
modest amount of plastic curvature brings us to the fully plastic moment. With strain hardening,
ultimate tensile failure is reached at a moment of about twice the yield moment, and at curvatures
beyond the range of most practical applications.
10

MOMENT

ULTIMATE STOCKY

< 25 t o 50
FULLY PLASTIC M p- |

YIELD J

CURVATURE

Fig. 1.5. Inelastic behavior of tubular section with different D/t ratios.

The behavior of plastic design sections is shown next. These can reach the fully plastic
momentand, beyond this, possess sufficient rotation capacity to redistribute moments and form
a plastic mechanism, e.g., as required to satisfy the ductility requirements of earthquake design.
The mode of section failure is plastic collapse, not classical buckling; the tension-compression
couple combined with large curvature act to cause a flattening of the section.
The upper D/t limit for this type of behavior depends on the kind of loading. It is about
50 for combinations of tension, bending, and hydrostatic pressure, as encountered in deep water
pipelaying operations. It is lower for combinations of compression, bending, and shear -- e.g.,
beam-columns as used in structures. Hydrostatic pressure severely reduces the bending
performance. Under certain conditions, the collapse can propagate far beyond the region of
severe loading which initiated itinfamous propagating buckle (ref. 13).
Bending behavior of merely compact tubular sections is not quite as favorable. These
can still develop the full plastic moment. However, only limited curvature and rotation capacity
is exhibited, before local buckling leads to a fairly rapid degradation of capacity to about half the
peak. Their applicability in earthquake design requires a detailed analysis which considers this
degradation. The buckle can be outwards or inwards. Filling the member with cement grout is
not particularly effective in suppressing the outward buckle.
For non-compact members which fail in the plastic buckling range, the bending strength
is somewhere between yield and fully plastic, with essentially negligible plastic rotation capacity.
The upper D/t limit for this class is 190 per the API design equations, and about 100 at the
experimentally observed onset of local buckling with yielding.
For members which fail in the elastic buckling range, the capacity is less than yield, and
very sensitive to imperfections. There is a very sudden, catastrophic drop at the onset of local
buckling, to perhaps 1/5 of the peak capacity. The classical diamond pattern of local buckling
may be observed.
(ii) Non-Circular Tubes. Applicability of plastic design to square and rectangular
tubes has also been studied, in terms of their ability to reach the fully plastic moment, and to
undergo sufficient plastic rotation to redistribute bending moments, as required to develop a
collapse mechanism. Data from Korol (ref. 19) and Graff (ref. 20) are shown in Figure 1.6,
along with compact section criteria from the AISI Guide (ref. 14). Once again, we see the lower
performance of cold formed sections, particularly in plastic rotation capacity.
 11

1.5.3 Local Buckling

The AWS Code states:

" 10.3.1 For circular sections having D/t greater than 3300/F ,the possibility of
local buckling at axial compressive stresses less than the yield strength shall be
considered (where F y is the yield strength of the base metal, ksi)."

For the most part, the problem of local buckling in tubular compression members is
avoided in structural design, by simply using relatively compact sections. All of the circular
sections listed in AISC and Table 1.1 meet the foregoing criterion, for steels up to 50 ksi yield
strength.

\ KEY

v HOT ROLLED (GRAFF)

\ COLD FORMED (KOROL)
: F NI IKN S0 T E
\ co v
\ AISI LIMITS FOR F y = 50
o

\ 8
8
COMPACT
WACT I \
^ ^V Z2I1O0/ /VVFFy 7 \

J I' -

10 20 30
WIDTH/THICKNESS b/t

PLASTIC MOMENT

f > 1.15
REDUCED EFFECTIVE
LU WIDTH
oa I 245/yT"

10 20 30
WIDTH/THICKNESS b/t

Fig. 1.6. Ultimate moment & plastic rotation of capacity of square tubes.

While a classical elastic theory for local buckling of axially loaded circular tubes is
available, its results are notoriously on the unsafe side of reality. This discrepancy has been
largely traced to the effect of initial imperfections. Even small deviations, with depths on the
order of 0.001 times the diameter, lead to a drastic reduction in strength, as indicated by both
theory and test results. In tests, failure by local buckling often occurs suddenly and
catastrophically with little or no advance warning. The fabrication tolerances for fabricated
12

structural pipe permit somewhat larger deviations, typically .003 to .010. During field
fabrication and erection, structural members are occasionally dented (under conditions which
may preclude practical repair) leading to still larger imperfections and prompting the author (ref.
17) to adopt a rather conservative design curve, as shown in Figure 1.7.

KEY T O D A T A P O I N T S
W I L S O N A N D NEWMARK
6. I t
t
= 1/32
1/8
NO R E D U C T I O N FOR L O C A L BUCKLING
WILSON
WHERE D / t < 3 3 0 0 / F y
48, t = i to i in.
34
12
CONWAY B R I D G E 1 8 4 6
PLANTEMA 1 9 4 6
X-52 ) B E N D I N G DATA FROM
C O N S I D E R LOCAL B U C K L I N G GRADE I S C H I L L I NMG
AS PLOTTED
WHERE D / t > 3 3 0 0 / F y W fp
SUBSTITUTE ULTIMATE WRINKLING STRESS F un
FOR F y I N A P P R O P R I A T E A I S C F O R M U L A E

(FOR = 3 0 , 0 0 0 K S I )

J l_
10 15 20

DIMENSIONLESS THICKNESS PARAMETER

Fig. 1.7. Historical local buckling criteria for circular tubes (ref. 17).

The current approach is to adopt more optimistic performance criteria and then re-assess
members which get dented. Criteria from API RP 2A are shown in Figure 1.8. For axial
compression, API finds no influence of F in the range of 35 to 60 ksi (245 to 420 MPa), in
contrast to both theory and earlier criteria. For bending in the range of plastic buckling, one can
cite Schilling's (ref. 21) and Sherman's (ref. 22) tests to justify retention of the shape factor in
this range. In the range of elastic buckling, one may cite the observation that the buckling stress
in bending is 30% higher than for axial load, yielding a moment capacity increase similar to that
obtained from the shape factor in the plastic range.
In the design of beam-columns, it seems reasonable that the effects of initial curvature,
amplification of bending moments, etc., would have no less an effect on failure by local buckling
than they would on a failure by yielding. Thus, API's recommended design practice is to
substitute the ultimate local buckling stress for yield stress in the AISC code formulae.
Local buckling for square and rectangular tubes is adequately treated elsewhere, in terms
of the effective width concept (e.g., Appendix C of the AISC Code, ref. 9). The corners of box
sections retain their strength long after the mid-side regions have buckled, giving a less
catastrophic form of post buckling behavior than for circular tubes.
Various D/t and width/thickness limits for tubes are shown in Table 1.3. These are an
amalgam of AWS, API, AISC, and AISI criteria; and designers working to a specific code should
refer to that code for guidance. The limits for connection design refer to the criteria to be
 13

API CRITERION

1.8

1.6
c o n s t a nt m o m e nt
1.4

1.2

1.0
Mult
0.8

0.6

0.4F
ROTATIONAL CAPACITY, /
,_ u y
0.2 - (b)
0
2000 4000 6000 8000 10,000

F vD / t ( F v in ksi )

Fig. 1.8. API RP 2A criteria for local buckling of fabricated steel cylinders, (a) Axial compression,
(b) Bending (ref. 22).

presented in subsequent chapters; rather stocky sections are required to develop the full material
strength in the failure modes indicated in the first three columns. For most practical tube
proportions, reduced connection strength applies. The limits for member design reflect the
spectrum of behavior from full plasticity to local buckling, as described earlier. Terminology
varies; for example, AISC-LRFD recognizes the following ranges of behavior: plastic design,
compact, non-compact, and slender members.

1.5.4 Beam Columns

Due to the reversible nature of storm forces, most of the lateral bracing members in an
offshore structure must be designed for compressive axial loads. These members are also subject
to localized wave pressures, buoyancy, and gravity forces, for which they act as beams. Thus,
they are designed as beam-columns, using the interaction formulae for combined axial
compression and bending stress spelled out in the AISC code. For biaxial bending in circular
14

sections, which have equal bending strength about any axis, it is advantageous and proper to use
the resultant bending moment, rather than the linear sum of the two bending terms as is usually
done for wide-flange shapes.

TABLE 1.3 D/t AND WIDTH/THICKNESS LIMITS FOR TUBES

FOR AWS CC)NNECTION DESIGN FOR MEMBER DESIGN

YIELD
APPLIC- PLASTIC MOMENT LIMIT OF
LOCAL GENERAL CONE- ABILITY FULL MOMENTS OR LIMIT FULL LOCAL
FAILURE COLLAPSE CYLINDER OF RULES PLASTIC LIMITED OF ELASTIC YIELD BUCKLING
c
ULT V p= . 5 7 F yo F
eo
= F
yo 1 : 4 FLARE IN 1 0 . 5 DESIGN ROTATION BEHAVIOR AXIAL FORMULAE

16
API RP2A

FOR K- 1300 1500 6000

CIRCULAR TUBES

F F F
CONNECTION 60 300
Y Y Y

12
FOR T 6 Y
30 3300
F
Y 2070 8970 3300 13000
F F F F
AISC

9
Y Y Y Y
FOR X

8 210 35 190 210 238 238

K
BOX SECTIONS

CLASS A

F
VFy VFY V ( F Y- 1 0 ) VFY
AISI

Y
22 20 FOR GAP @ M = S ( F y- 1 0 ) NO
FOR K&N
CONNECTIONS LIMIT
190 150 238
CLASS
AISI

7 * Y
FOR T&X
F in ksi ( 1 ksi - 7 MPa)
AISI Class A - hot formed
AISI Class = cold formed and welded

For detailed analysis of the inelastic behavior of beam-columns, the plastic deformation
of tubular sections may be described by moment-thrust-curvature relations (M-P-phi curves).
Residual stresses cause early departure from linear behavior, consistent with the observed
buckling strength of centrally loaded columns. Such a procedure was first used by Fowler (ref.
23), using computer programs developed at the University of Texas. A more general
representation of the inelastic behavior requires four parameters (moment, thrust, curvature, and
axial deformation) with behavior is described in terms of an interaction surface and a flow rule.
The added dimension is important in understanding the behavior of struts, particularly the
phenomenon of column growth which occurs during cyclic buckling and straightening.
Professor Sherman (ref. 24) used M-P-phi curves with a 20-segment inelastic beam-
column model to study the ultimate strength behavior of tubular struts. These are members
carrying primarily axial load, in the presence of lateral loads due to wave force, self weight,
buoyancy, local acceleration, etc. Typical results are shown as the solid lines in Figure 1.9.
These are closely followed by the arc-sine equation shown in the figure, and indicate ultimate
capacities well in excess of API and AISC first yield criteria. For typical struts with mostly axial
load, and L/D in the range of 20 to 50, the capacity is 20-25% greater. Where there is more
bending, the difference is even more dramatic.
Sherman also conducted a series of over 100 tests of strut and portal type beam-columns
at the University of Wisconsin-Milwaukee (refs. 25-27). His results are compared with the arc-
sine ultimate strength equation in Figure 1.10. The proposed equation is generally conservative,
except for the "D" series and the dark points. The "D" series has tubes with D/t of about
 15

80 which failed by sudden local buckling, before achieving a plastic mechanism. The dark
points are as-received cold formed manufactured tubes with yield strength in excess of 50 ksi, a
rounded stress strain curve, and a low UTS/yield ratio not typical of larger fabricated tubes
used in offshore platfonns.

Fig. 1.9. Beam-column interaction of tubular struts - analysis.

Fig. 1.10. Beam column interaction of tubular struts and portals - tests.
16

1.5.5 Shear and Torsion

Compared with other common types of structural members, tubes of uniform wall
thickness have very high capacity in shear. For circular tubes, the effective area for beam shear
is half the gross area, and moment failure will take precedence over shear failure for all but the
shortest spans less than four diameters. Even for these short spans, local stress problems in
getting the load into and out of the cross section (e.g., punching shear as will be described for the
tubular connection problem) will often be more important than beam shear.
As closed sections, tubes also have very high capacity for torsion. Circular tubes can
take torsional moments comparable to their capacity in bending. Except for a caveat about local
stresses at junctions and elbows, this frees the designer to make some very fanciful and novel
designs in tubular structures.

1.6 SIMPLE WELDED JOINTS

Although many different schemes for stiffening tubular connections have been devised,
the simplest is to simply weld the branch member to the outside surface of the main member (or
chord). Where the main member is relatively compact (D/t less than 15 or 20), and the branch
member thickness is limited to 50% or 60% of the main member thickness, the connection will
be sufficiently strong to develop the full static capacity of the members joined, without
performing detailed engineering calculations (Sherman, ref. 14). Where these conditions are not
met, e.g., with large diameter tubes, a short length of heavier material (or joint can) is inserted
into the chord to locally reinforce the connection area. Here, the design problem reduces to
selecting the right combination of thickness, yield strength, and toughness for the joint can.
The detailed considerations involved in this design process are the subject of the
following chapters. More complex forms of joint reinforcement are also discussed.

REFERENCES

1 Marshall, P.W., Fixed Pile-Supported Steel Offshore Platforms, Journal of Structural Engineering,
ASCE, Vol. 107, No. ST6, June 1981.
2 Marshall, P.W., Risk Factors for Offshore Structures, Proc. Civil Engineering in the Oceans, San
Francisco, ASCE, September 1967.
3 AWS Structural Welding Code - Steel, AWS Dl .1-88, American Welding Society, Miami, 1988.
4 Marshall, P.W., Connections for Welded Tubular Structures, 1984 Houdremont Lecture, Proc.
2nd International Conference on Welding of Tubular Structures, Boston, July 1984, Pergamon,
1984.
5 Bram McClelland, Reifel, M.D. (editors), Planning and Design of Fixed Offshore Platforms, van
Nostrand Reinhold Co., New York, 1986.
6 Wardenier, J., Hollow Section Joints, Delft University Press, 1982.
7 Design of Tubular Joints for Offshore Structures, UEG, London, 1985.
8 API Recommended Practice for the Planning, Designing, and Constructing Fixed Offshore
Platforms, API RP 2A, 17th Edition, American Petroleum Institute, Dallas, 1987.
9 AISC Manual of Steel Construction, 8th Edition, American Institute for Steel Construction,
Chicago, 1980.
10 Marshall, P.W., Stability Problems in Offshore Structures, Proc. Column Research Council, 1970
(longer version available from author).
11 Wolford, D.S. and Rebholz, M.J., Beam and Column Tests of Welded Steel Tubing with Design
Recommendations, Bulletin 233, ASTM, Philadelphia, October 1958.
12 Chen, W.F. and Ross, P.A., Tests of Fabricated Tubular Columns, Journal of Structural
Engineering, ASCE, V. 103, No. ST3,1977.
13 Chen, W.F. and Han, D.J. Tubular Members in Offshore Structures, Pitman Press, Boston, 1985.
14 Sherman, D.R., Tentative Criteria for Structural Applications of Steel Tubing and Pipe, AISI
Committee of Steel Pipe Producers, August 1976.
 17

15 Wardenier, J., Static Design of Hollow Sections in Steel Structures, Van Leeuwen Technical
Information No. 6, Zwijndrecht, Holland, 1987.
16 Berkemoe, P.C. and Bjorhovde, R., Limit States Design of HSS Columns, Canadian Structural
Engineering Conference, 1978.
17 Marshall, P.W., Design Criteria for Structural Steel Pipe, Proc. Column Research Council, 1971
(longer version available from author).
18 Marshall, P.W., An Overview of Recent Work on Cyclic Inelastic Behavior, Panel Discussion on
Stability of Offshore Structures, Proc. Structural Stability Research Council, 1982.
19 Korol, R.M., The Plastic Behavior of Hollow Structural Sections with Implications for Design,
Canadian Structural Engineering Conference, 1972.
20 Graff, W.J. and White, A.H., Bending Strength of Rectangular Hollow Sections, ASCE National
Structural Engineering Conference, Madison, August 1976.
21 Schilling, C.G., Buckling Strengths of Circular Tubes, ASCE Structural Journal, Vol. 91, No. ST3,
October 1965.
22 Sherman, D.R., Bending Capacity of Fabricated Pipes at Fixed Ends, Report to API, University of
Wisconsin-Milwaukee, December 1985.
23 Fowler, D.W. and Erzurumlu, H., Ultimate Strength of Round Tubular Beam-Columns, draft
ASCE paper, August, 1970.
24 Sherman, D.R., Ultimate Capacity of Tubular Members, Shell Oil Company, CE-15 Report,
August 1975.
25 Sherman, D.R., Experimental Study of Post Local Buckling in Tubular Portal Type Beam
Columns, University of Wisconsin-Milwaukee report to Shell Oil Company, October 1979.
26 Sherman, D.R., Post Local Buckling Behavior of Tubular Strut Type Beam Columns: An
Experimental Study, University of Wisconsin-Milwaukee report to Shell Oil Company, June 1980.
27 Sherman, D.R., Interpretative Discussion of Tubular Beam Column Test Data, University of
Wisconsin-Milwaukee report to Shell Oil Company, December 1980; with supplement July 1981.
28 Matsui, C , et al, Lateral Torsional Buckling of Truss with Rectangular Tube Section, Proc. 2nd
International Conference on Welding of Tubular Structures, IIW, Boston, July 1984.
29 Mouty, J., Effective Lengths of Lattice Girder Members, CIDECT Monograph No. 4,1981
Chapter 2

CONCEPTUAL BASIS FOR THE DESIGN RULES

2.1 DEFINITIONS

A number of definitions will be useful in our discussion of tubular connections, which

we now formalize. These are as used in the AWS codes.
T u b u l a r connections - a connection in the portion of a structure that contains two or
more intersecting members, at least one of which is a tubular member; the structural node as a
subassemblage. Connection geometry may be described in terms of the topology of the
intersecting memberstheir size, shape, position, and orientation.
Parts of a tubular connection are described in Figure 2.1. Much of the design attention
is focused on the m a i n m e m b e r (sometimes called the c h o r d , or jacket-leg in offshore
applications), which receives loads from attached branch members (sometimes called braces),
thereby incurring high localized stresses. A short length of thicker material inserted into the
main member for the purpose of reducing these localized stresses is called a joint-can.

Heel

(A) Circular s e c t i o n s (B) B o x s e c t i o n s

Fig. 2.1. Parts of a tubular connection.

T u b u l a r joint - a welded joint at the interface created between members in a tubular

connection, consisting of the weld deposit, heat-affected zone, and immediately adjacent base
metal. Joint geometry and welding requirements may be described with reference to the
thicknesses of the intersecting parts and the local dihedral angle. In AWS terminology, there is a
hierarchy: connection/joint/weld, with the latter denoting the weld deposit itself. In the
literature, "tubular joint" is often used loosely to refer to tubular connections as well.
Local dihedral angle - the angle between tangent lines of the intersecting base metal
surfaces, measured in a plane perpendicular to the weld line. In a tubular connection, the local
dihedral angle, and therefore the joint geometry, usually changes continuously as one
proceeds around the intersection. Acute angles occur in the heel (or crotch) position, with more
generous angles in the side (or saddle) and toe portions of the connection (Figure 2.1). The
British refer to heel and toe as the "crown" position.
 19

() Geometric parameters

Parameter Circular Sections Box sections

r b/ / ? o r d / D b / D
b

a x / D

* / t D / 2 t c
7 c
l
r t b/ c

Angle between member center lines

Local dihedral angle at given point

on w e l d e d joint

g/D i n K-connections
(see below)

Note: Relevant g a p is b e t w e e n
braces w h o s e loads a r e
essentially balanced

Fig. 2.2. Non-dimensional parameters of a tubular connection.

20

Non-dimensional parameters for describing the geometry of a tubular connection are

given in Figure 2.2. , , 0, and $ describe the surface topology. 7 and r are two very
important thickness parameters, a (not shown) is an ovalizing parameter, depending on load
pattern as discussed later, is the local dihedral angle.
In the AWS Structural Welding Code, the term "-, Y-, and K-connection" is used
generically to describe structural nodes, as opposed to butt or lap joints between co-axial tubes.
A letter of the alphabet (, , K) is used to evoke a picture of what the node subassemblage
looks like. The relevant Code provisions also apply broadly to other "alphabet" connections,
such as X (double-T or cross) and N-connections (K with one of the two branches perpendicular
to the chord), as well as to more complex and multi-planar forms.
Usually, the "Design" provisions of a Welding Code are restricted to such subjects as
allowable unit stresses in welds, selection of weld size or effective throat, and to joint detailing
such as groove angle, root gap, etc. Member sizing is usually left to the governing design or
building Code (e.g., AISC), as is the sizing of special connection material such as gusset plates.
However, in commonly used types of tubular connections, the members themselves make up the
connection, and many designers are tempted to take the simplistic view that a full penetration
weld will take care of everything. However, the welded joint is often not the limiting factor in
determining the useful capacity of the connection, even though welding often gets the blame
when failures occur. Comprehensive design requires that a number of other failure modes must
be considered, in addition to the usual checks provided for in most design codes. As discussed in
the following sections, these include local failure (punching shear), general collapse (ovalizing),
progressive failure (unzipping), and various forms of material failure (brittle fracture, fatigue
cracking, lamellar tearing, etc.).
We will follow (in Section 2.3) with methods of analysis for the tubular connection as a
structure on its own merits, using elastic analysis, limit state methods, and model tests.
Two very useful design simplifications will then be introduced (whose AWS definitions
are elaborated upon in Section 2.4):
Hot spot s t r a i n - the cyclic total range of strain which would be measured (or
calculated with best available theory) at the point of highest stress concentration in a welded
connection (on the outside surface of intersecting members, at the toe of the weld joining them).
Punching shear - average or nominal stress on a potential failure surface as shown in
Figure 2.3. ^

BRANCH
MEMBER

PUNCHING
SHEAR
STRESS V p Nr

F i g . 2 . 3 . Simplified
concept of punching
shear.

MAIN MEMBER 77
 21

A discussion of stresses in welds completes the chapter. A complete listing of symbols

and notation used herein can be found in Appendix I. These follow that of AWS, except that
italicization is not always observed in this text.

2.2 FAILURE MODES OF TUBULAR CONNECTIONS

The AWS Structural Welding Code describes a number of failure modes for tubular
connections which must be considered by designers (or fabricators stuck with the job of making
incomplete designs work). The somewhat unusual inclusion of design criteria in the welding
code was made necessary by the realization that materials and welding considerations alone
would not solve the problems alluded to in the preceding section (ref. 1), and by the absence of
appropriate guidance in other American codes. Although the headings which follow correspond
to the AWS criteria as originally proposed, adopted, and published (refs. 2-5), the discussion has
been broadened to take advantage of more recent thinking, as indicated by more recent
references. The exposition here is largely heuristic; detailed historical and technical development
of the criteria will follow in subsequent chapters.

2.2.1 Local Failure

Figure 2.4 shows an example of local failure of a tubular connection, in a small offshore
structure which had been through several storms and at least one collision. The large hole is
where an attached brace has pulled out a chunk of the jacket leg; in this outdated design, there
was no joint-can, so that the main member was not much thicker than the branch member.
Partial failure at a second brace may also be seen. The metallurgist in the picture is conducting a
post-mortem examination, which indicated that the failure was by ductile overload. Even though
the fracture closely followed the heat affected zone of the intersection weld, there was visible
plastic deformation in this vicinity. In the writer's experience, this has been the most commonly
observed mode of failure for large-scale tubular structures in offshore service.

Fig. 2.4. Example of local failure in service.

22

While tubes are generally efficient in carrying membrane stresses in the plane of their
material, they are quite inefficientboth weak and flexiblein their response to radial loads. In
AWS terminology, the radial component of branch member loads, expressed as an average or
nominal stress on the potential failure surface, is termed PUNCHING SHEAR, as shown in
Figure 2.3. The actual load-carrying mechanism involves a complex combination of shell
bending, warping, arching, and large deflection membrane effects, in response to radial line loads
and the corresponding punching shear. Tubular connections generally show gross plastic
deformation before separation; failure does not occur in the manner of punching out bolt holes,
and the p u n c h i n g shear stress rarely reaches the full shear strength of the m a t e r i a l .
Internationally, "plastic (flexural) failure of the chord face" is the preferred terminology for the
general case, with "punching shear" reserved for the rare exceptions which do reach the full shear
strength of the material (e.g., as used in ref. 6). Where design criteria are derived empirically
from tests, the distinction may be largely academicrelated to one's choice of format for the
equations, and to the kind of understanding one wishes to convey to designers for their use in the
practical extrapolations which seem to be inevitable.
The behavior of stepped box connections (a tubular connection of square or rectangular
hollow sections with the branch narrower than the chord) closely parallels the behavior of
circular sections. Figure 2.5 shows a yield line mechanism for plastic flexural failure of the
chord face. Axial load in the branch member may be represented by a radial line load applied at
the locus of the intersection weld, or the corresponding by punching shear (ref. 7).

Fig. 2.5. Limit state for

small beta.

As beta (branch width/chord width) approaches unity, higher line load capacities are
achieved along the sides of the connection, while a different yield line pattern supports reduced
line loads at the heel and toe of the connection, as shown in Figure 2.6. Limits on the capacity
along the sides are punching shear at the chord material shear strengthand, for beta equal to
unity (e.g., matched box connections), sidewall yielding or web crippling of the main member
(ref. 8). The yield line pattern in Fig. 2.6(b) assumes a uniform line load instead of uniform
displacement of a rigid branch. The associated deformation of the branch and its effect on the
actual load distribution is discussed further in Chapter 5. International codes (e.g., ref. 9) treat
 23

the reduced capacity along the heel and toe of the connection in terms of a reduced "effective
width" concept, with the load concentrated near the main member sidewall.

Fig. 2.6. Limit states for large beta > 0.8.

2.2.2 General Collapse

The first-proposed general collapse criteria for AWS D l . l simply stated that "strength
and stability of the main member in a tubular connection, acting as a cylindrical shell together
with any reinforcement, should be investigated using available technology, in accordance with
the applicable design code."
The principal concern here was circular tubes and ovalizing of the main member ~ e.g.,
a joint-can of inadequate length, or an offshore jacket launch-leg subjected to crushing loads in
a way not adequately covered by the initial local failure criteria, which was largely based on tests
of T-connections. "Applicable technology" could be taken to mean closed ring solutions, e.g.,
those of Roark (ref. 10).
Over the years, ovalizing has tended to become incorporated into the local failure
criteria, i.e., consideration of capacity based on connection type or load pattern. The concept of
general collapse has been retained, and extended to include any other failure mode that, given a
thorough design investigation of the connection subassemblage (particularly the main member)
as a structure on its own merits, would fall within the purview of ordinary structural engineering
and the provisions of other design codes. Among others, such failure modes would include the
following:

(a) ovalization of circular members, unreinforced or with stiffening rings

(b) beam bending in tee-connections, occasionally an embarrassment in test set-ups
(c) beam shear in the gap region of K-connections, especially with single-web W or
shapes as the chord (ref. 11)
(d) web crippling under the concentrated branch member reactions in matched box
connections (ref. 12)
24

(a)

(b)

(c)

(d)

(e)

Fig. 2.7. General collapse failure modes, (a) Ovalizing. (b) Beam bending, (c) Beam shear, (d)
Web crippling, (e) Longitudinal distress.
 25

(e) local distress of the main member (e.g., local buckling), due to concentrated
longitudinal loads being delivered at the branch member footprint; possible in thin-
wall chords whose radial inadequacies have been dealt with by stiffening or by
direct load transfer between overlapping braces.

Figure 2.7 shows these general collapse failure modes. Cross-hatching denotes yielding;
contours, buckling.

2.2.3 Uneven Distribution of Load

While simplified representations of punching shear and the corresponding radial line
loads are useful for understanding local failure modes in the main member, we must not try to
carry them too far. Due to differences in the relative flexibilities of the main member loaded
normal to its surface, and the branch member loaded in the plane of its material, load transfer
across the welded joint at their intersection is highly non-uniform, as shown in Figure 2.8. Local
yielding and re-distribution of load must occur before the connection reaches its ultimate
capacity. Punching shear and shell bending in the chord wall are inherently ductile mechanisms,
and can easily accommodate this. However, an undersized weak-link weld at the intersection can
fail progressively, in an "unzipping" mode.

(b)

Fig. 2.8. Uneven distribution of load, (a) Circular sections, (b) Box sections.

To prevent progressive failure of the weld and insure ductile behavior of the tubular
joint, the AWS Code requires that the minimum weld provided in simple -, Y-, or K-
connections NOT be the weak link in the system. Pre-qualified groove-welded joint details,
capable of yielding the adjacent branch material before the weld ruptures, are provided to meet
this requirement.
Fillet welds do not automatically match the line load yield strength of the branch
member material. IIW accomplishes a match by requiring a minimum effective throat equal to
the branch thickness, which is economical only for thicknesses up to 5/16-in. (8mm), and hard to
achieve without bevelling for local dihedral angles greater than 120-degrees (refs. 13 & 14).
AWS relaxes this requirement, providing a minimum effective throat of 0.7 times the lesser
26

thickness, but this only provides matching strength when E70XX welds are used on mild steel.
For heavier thicknesses, higher strength steel, or a desire to economize the welding on lightly
loaded members, further investigation by the designer is required, to make sure the breaking
strength of the weld at least matches the lesser of branch or chord line load capacity, or to
explicitly take the uneven distribution of load into account in his calculations.
AWS has traditionally taken the former approach; while, for box connections, Packer
has suggested the latter, using an "effective width" approach to discount the most weakly
(flexibly) supported part of the intersection weld (ref. 15). Weld stress and line load design
calculations based on geometry and statics (e.g., using Section 10.8 "Effective Weld Area and
Length" of AWS-D1.1) are somewhat misleading in this regard, as they do not reflect uneven
load distribution.
Similar considerations of uneven load distribution apply to the branch member itself, as
well as to the weld. Under compression loading, local buckling can lead to a loss of ductility and
premature failure in the branch member. More restrictive branch member width/thickness or
diameter/thickness ratios appear to be needed for avoiding this problem versus what is required
for uniformly compressed compact members 27 vs. 34 for box sections, and 37 vs. 66 for
circular sections, at the 50-ksi yield strength level (refs. 9 & 16 vs. AISC). In line with Makino's
finding that local buckling of the branch is affected by connection efficiency (ref. 16), the author
has observed progressive local buckling of branch members having D/t of 43 and 48 at nodes
where uneven distribution of load would be expected.
Furthermore, for matched box connections and those with large beta, the branch member
may not be able to match the high local chord capacity which applies along the sides of the
connection. For the 1975-90 AWS criteria, Marshall's design guide (ref. 17) specifies a
minimum tau ratio (branch thickness/chord thickness) required for developing the full chord
capacity. International rules handle this problem more directly via "effective width" calculations
applied to the branch member (see refs. 6 & 9, and Chapter 5).

2.2.4 Materials Considerations

Materials considerations appear in a section of the Code written for designers for the
same reason that a section on design appears in a book used primarily by welding engineersthe
need for an integrated approach to structural integrity and fracture control. Reference 1 discusses
a number of materials problems which are either under the control of the designer or influenced
by his decisions. As these are discussed at length in Chapters 4 and 7, the following introductory
remarks will be kept brief.
Historically, fatigue has been the second most common failure mode in the tubular
connections of offshore structures. With the static strength problem having been solved two
decades ago, and an aging fleet of structures in hostile environments like the North Sea, fatigue is
likely to become the predominant failure mode of the future. For structures in the ocean, random
loading and environmental influences on corrosion fatigue must be considered (ref. 18). Since
the mid-1970's, these problems have been the subject of intensive, long-term research programs,
sponsored by the American Petroleum Institute (API), the UK Department of Energy, and the
European Coal and Steel Community; and a workable understanding of them is now in hand (ref.
19).
As previously described, tubular connections depend upon ductility and the ability to
undergo localized yielding in order to reach the ultimate limit state upon which their design is
based. This yielding must take place in the "hot spot" region adjacent to the intersection weld,
where there are stress concentrations and triaxial stresses due to the overall geometry of the
connection, notch effects due to the shape of the weld profile, and crack-like defects in the heat
affected zone at the toe of the weld (overlap, undercut, microscopic slag intrusions, shrinkage-
induced tearing, incipient fatigue cracks, and worse). Thus, extraordinary demands are placed on
 27

the notch toughness of the joint-can material. For offshore structures, this has been addressed via
design guidelines for toughness testing and material selection (API RP 2A), and via special steel
specifications (API Spec 2H). Since onshore applications of tubular structures have tended to be
less critical and smaller in scale (with thinner material), the need for expensive notch toughness
requirements has been less obvious, and the AWS committee has been reluntant to impose them.
On the other hand, the unique importance of the lamellar tearing failure mode in tubular
connections has been recognized from the very first in AWS D l . l . Under the "footprint" of
attached braces, the joint-can is subjected to thru-thickness ("Z" direction) tensile stresses, not
only from tensile service loads, but also from weld shrinkage at the point of attachment. Due to
their non-homogeniety, ordinary steels have reduced strength and very little ductility in the "Z"
direction. This was high-priority problem in the late 1960's and early 1970's, with failures
occurring both during fabrication and in service (Figure 2.9). Eventually, the AWS Commentary
developed a practical combination of design guidelines (e.g., low strength attachments with soft
weld metal) and fabrication "fixes" (e.g., shrinkage control and surface buttering). In offshore
practice, the problem has been virtually eliminated with the use of cleaner, low-sulfur steels,
subjected to thru-thickness tensile test ductility requirements.

Fig. 2.9. Lamellar tearing failures, (a) Failure during fabrication, (b) Metallographic examination.
(c) Failure in service, collision overload, (d) Schematic of failure mode due to large
delamination.
28

Since the designer is often the one who specifies the steel to be used, he also needs to be
aware of weldability issues. Hard, brittle weld heat affected zones can adversely affect fracture
behavior and lamellar tearing, as well as being susceptible to shrinkage cracks due to the
combination of high restraint (virtually a given in -, Y-, and K-connections) and hydrogen
(from the welding environment or from cathodic protection). Traditional high strength steels,
which achieve their strength through alloy additions, are particularly susceptible to these
problems. New low alloy thermo-mechanically processed steels have been developed for the
offshore market, along with CTOD (crack tip opening displacement) and weldability
performance specifications (e.g., API RP 2Z), by which they can be pre-qualified at the steel mill
to demonstrate an ability to deal with these problems.
Hicks' listing of materials problems (ref. 20) includes one or two more, like weld metal
solidification cracking. However, these are more clearly the responsibility of the welding
engineer, who chooses the welding processes and consumables, rather than the designer. The
author (and the AWS code) prefers to deal with this class of problem with performance
specifications on weld quality, using welding procedure qualification tests and non-destructive
inspection to make sure the goals are met.

2.2.5 Summary
A listing of all the failure modes discussed in the foregoing sections can be found in
Table 2.1. AWS Code treatment is in Section 10.5, "Limitations on the Strength of Welded
Tubular Connections". The reader may wish to peruse those provisions at this point.

TABLE 2.1 LIMITATIONS ON THE STRENGTH OF WELDED TUBULAR CONNECTIONS

LOCAL FAILURE OF THE CHORD

plastic failure of chord face at radial lines loads
punching shear at the material strength
sidewall yielding or crippling
reduced line load capacity/effective width

GENERAL COLLAPSE OF THE CHORD

ovalization
beam bending
beam shear
sidewall web buckling
longitudinal distress

UNEVEN DISTRIBUTION OF LOAD

weld unzipping/effective width
branch member local buckling/effective width

MATERIALS CONSIDERATIONS
fatigue/corrosion fatigue
brittle fracture/notch toughness
lamellar tearing
weldability
 29

2.3 GENERAL PROCEDURES FOR ANALYSIS AND DESIGN

As described in the following sections, many of the methodologies of modern structural

mechanics have been applied to the challenging problem of analyzing and understanding tubular
connections as structures in their own right. Elastic stress analysis includes shell theory and
finite element methods. Limit analysis methods include both the method of cutting sections
(loosely based on the lower bound theorem of plasticity) and plastic yield line analysis (an upper
bound method); non-linear finite element methods are also beginning to be used. Experimental
methods cover the full range of behavior from elastic to the ultimate limit state. In some cases,
these analysis methods were applied for research purposes; once they yielded up their measure of
understanding, more simplified procedures were adopted for routine design, such as punching
shear and hot spot stress. In other cases, special design analyses of critical connections (whose
importance could justify the effort and expense) not only solved the particular problem at hand,
but also gave insights for wider applications.

2.3.1 Elastic Analysis

The local stresses in tubular connections are extremely complex, and closed form
solutions do not exist for most cases of practical interest. However, a few general principles will
serve to bring our understanding into focus.
In the simplest tubular connections, the branch members are simply welded to the main
member or chord. Although relative flexibilities influence the exact distribution of load, the
branch members deliver their reactions to the chord in the form of line loads. Localized shell
bending stresses in the chord wall reach a peak at these line loads, with steep local gradients that
are related to punching shear in much the same way that moment gradients are related to shear in
beams.
Since the present section deals with elastic stresses, we shall defer consideration of
practical ultimate strength until later.
(i) Shell T h e o r y . Closed form solutions for elastic stresses in cylindrical shells
subjected to radial line loads are available for simple two-dimensional cases, two of which are
quite instructive, as shown in Figure 2.10.

(a) For the axisymmetric case (ref. 21), the line load is carried initially by punching
shear V on both sides (double shear), with longitudinal shell bending stress
incurred in the task of transferring the concentrated load to its eventual support,
hoop stresses within a region of roughly 1.56 V(R/T) centered upon the load.

(b) For the two parallel line load cases (ref. 10), the line loads are carried initially in
punching shear (expressed this time in single shear, reflecting consideration of a
free body containing both line loads), with transverse shell bending stresses
incurred in the task of transferring the concentrated load to its eventual support,
global beam shear.

In both cases, for shells of typical proportions, the highest stress is the shell bending
which peaks at the locus of the line load. Formulas in the figure give the punching shear and line
load capacities at shell bending first yield; these are seen to be a function of shell yield strength,
F G, shell thickness, T, and shell gamma ratio, R/T. For case (b), the capacity is also a function
or load pattern, expressed by the geometric beta ratio (defined in the figure). Considering that
appears both in the conversion from line load to punching shear, and again in the gamma ratio,
total line load capacity is seen to be proportional to the 1.5 to 2.0 power of T.
30

AXISYMMETRIC
LINE LOAD

Fig. 2.10. Closed Form Solutions for cylindrical Shells, (a) Axisymmetric line load, (b) Parallel line
loads.
Fig. 2.11. Cylindrical shell equations of equilibrium. F i g . 2.12. Cylindrical shell equations of compatibility.
31
32

Although reference 21 describes a valiant attempt to apply the previous cylindrical shell
solutions of Bijlaard (ref. 22) to the tubular joint problem, it remained for Dundrova (refs. 23, 24,
25) to develop the first workable solution. Parameter studies using this solution provided our
first broad understanding of the trends of tubular connection behavior (ref. 26). Unfortunately,
widespread commercial application of Dundrova's work has been hampered by proprietary
restrictions, and it was soon supplanted in the USA by more computer intensive finite element
solutions; however, it has continued to be used in eastern Europe and the USSR.
Dundrova approached the analysis of tubular joints by coupling a membrane
representation of the branch member to a full shell theory representation of the chord. Both
substructures are reduced, at their common boundary, to a system of compatibility and
equilibrium equations representing the transfer of line load across the weld joining the branch
member to the chord. The solution gives displacements along the common boundary, and the
actual (uneven) distribution of the boundary load, which are back substituted to obtain stresses
throughout the chord, represented by Fourier series in the two shell surface coordinates.
Figures 2.11 and 2.12 show the five equations of equilibrium and eight equations of
compatibility used within the chord shell. Here Q is the internal punching shear, expressed as a
line traction (units of kips/in). Note how the gradient of shell bending moment (along with
warping moment) is related to the corresponding punching shear, in the two Equations (5) and
(6) for moment equilibrium. We also see how punching shear can be absorbed (i.e., develop a
gradient) through the arching action of hoop stress, N2, in Equation (3) for equilibrium of normal
forces, and how it also appears in Equation (2) for circumferential forces.
Punching shear Q transmits the effect of external load throughout the shell, but does not
appear in the equations of compatibility, which relate the other forces and moments to
displacements U, V, and W. However, we do see the pervasive appearance of the gamma ratio in
every equation.
For the simple examples of Figure 2.10, case (a) would only have non-zero terms in Q l ,
M l , and N2 varying along a generatrix; case (b) would only have Q2, M2, and N2 varying
around the circumference; while most practical tubular connections would have all ten terms,
varying in both directions. Some of Dundrova's equations get extremely long. However, they
can be readily solved on a small computer, using a program called FRAMETI.
(ii) Thin Shell Finite Elements. To analyze tubular joints with the finite element
method, intersecting cylindrical shells are subdivided into a mesh of elements which approximate
the in-plane (membrane) and out-of-plane (punching shear, shell bending, etc.) behavior of the
actual tubes. The mesh lies at the mid-plane of the shells.
Most finite element computer programs use the stiffness method, in which element
forces and displacements (strains) are formulated in terms of the unknown node displacements,
in a manner which is consistent with the behavior of the continuum for a given set of
assumptions, e.g., constant strain, constant curvature, or linear variation of strain and curvature.
Clough and Greste (refs. 27 & 28) developed the first finite element program to be
specialized for tubular connections, 20 years ago. Mesh generation is automatic but limited to
simple K-connections with equal diameter braces in one plane. Figure 2.13 shows a medium
mesh for the chord, along with stress contours at a 10 ksi (70 MPa) interval; this is the same
geometry as a K-connection of studied with shell analysis and experimental stress analysis by
other investigators. The steep gradients adjacent to the brace-to-chord intersection peak out at
hot spot stresses of 2.5-2.7 times the nominal branch member stress.
 33

(a)

(b)
(c)

Fig. 2.13. Mesh and stress contours for K-connection. (a) Overview of entire joint, (b) Detail of gap
region, (c) Detail of branch member (developed view).
34

Visser (ref. 29) described another early finite element program, SATE, which utilized
mixed-mode thin-shell elements, in which membrane behavior is formulated in terms of
displacements, while out-of-plane behavior is formulated in terms of shell bending moments.
With these more sophisticated elements, a coarser mesh could be used. Semi-automatic mesh
generation, accomplished with user-supplied FORTRAN subroutines, permitted the analysis of a
variety of configurations, including the stiffened connections.
A number of commercially available finite element computer programs may also be
used to analyze tubular connections - for example, STARDYNE, MARC, ABACUS, ANSYS,
NASTRAN, SAP, SNAP, STRUDL, and SESAM-80. Some of these are supported by service
organizations and mesh generation software adaptable to a variety of tubular joint configurations.
Defining the common intersection line in the coordinate systems of both branch and chord is a
necessary starting point. Without such help, the problem of mesh generation becomes so
formidable that the cost of finite element analysis can exceed that of strain gage experimental
stress analysis of an accurately scaled model joint.
Bouwkamp has developed an approach which eases the problem of mesh generation,
using his computer program TOJO (ref. 30). Tubular connection substructures are automatically
generated from data already available in the space frame analysis file (plus joint can thickness
and length). Then, the space frame and its connection substructures are analyzed together to get
a fully consistent solution to both global and local stresses. The automatically generated mesh is
somewhat unconventional, as shown in Figure 2.14. Interpolation algorithms are used to take
care of nodal mismatches at the brace-to-chord intersection and at changes in the chord grid
density. Yet benchmarking against other programs showed the results to be surprisingly
accurate.

30 h

Fig. 2.14. TOJO substructure for multi-planar tubular connection, (a) Example mesh (fineness 4). (b)
Benchmarking comparison, circumferential stress in chord at normal brace of Bouwkamp K-
joint.
 35

(iii) Three-Dimensional I s o p a r a m e t r i c Finite Elements. Isoparametric elements

assume a polynomial displacement pattern along the element boundaries, which are mapped onto
the actual element shape (which may be curved) and integrated numerically to obtain element
stiffness properties. This same polynomial interpolation function is used again to obtain stresses
from the displacement solution; hence, the name isoparametric (ref. 31). The formulation is
quite general and computationally efficient, permitting the use of solid elements to model the
finite thickness of the shells ~ and, more importantly, the weld geometry at their intersection
with about the same effort as with a finer grid of less sophisticated elements. This avoids the
paradoxical results that are sometimes obtained for "surface" stresses at the mid-plane
intersection, using thin-shell analysis.
Reimer has described application of isoparametric finite elements to tubular joints (ref.
32). The author helped him organize a joint-industry consortium to develop this methodology
into the computer program PMBSHELL (ref. 33), which includes automated mesh generation for
cross, X, tee, , (both gap and overlap), and K-T (3-brace) connections. Branch and chord are
both modeled with 16-noded curved shell elements. The weld and weld footprint are modeled
with 3-D "brick" elements, with weld geometry conforming to the parameters shown in Figure
2.15 (ref. 34). Note that the weld "brick" includes a portion of the branch member parent metal,
in order to improve the aspect ratio of the element. This gets the weld toe "hot spot" in the
correct position, and models the effect of weld stiffness and gross geometry, but is not detailed
enough to capture very localized or microscopic notch effects at the toe of the weld.

Fig. 2.15. Weld geometry parameters for PMBSHELL based upon AWS Dl.1-75.
36

Figure 2.16 shows a 3-D isoparametric "thick shell" mesh for the much studied K-
connection (previously seen in Figure 2.13). Despite its relative coarseness, it yielded
satisfactory numerical accuracy. Brute force manual mesh generation requires about two man-
weeks, versus about $200 in computer time when automated. PMBSHELL also provides semi-
automatic mesh generation, in which it defines the intersection weld geometry, while the user
fills in the rest of the mesh and adds his non-standard features of interest (stiffeners, out-of-plane
braces, etc.). This mesh geometry can also be converted to a "neutral file" and used with other
finite element computer programs.

Fig. 2.16. Isoparametric thick shell finite element model of K-joint (medium mesh).

Theoretical shell analysis, thin shell and thick shell finite element analyses all reproduce
the overall pattern of stresses in the chord, However, near the weld, which is the region of
interest for hot spot stress, the thick shell modelling is more realistic.
Section A-A of Figure 2.16 details the author's suggestions for mesh modelling near the
weld. The Gauss-point surface stress (GPSS) is often taken to be the most accurate stress within
the element. With judicious choice of element size adjacent to the weld (as shown), the GPSS
corresponds to the location of the strain gage which measures hot spot stress (American
definition). The European definition of hot spot stress (linear extrapolation to the toe of the
weld) is also entirely consistent with this kind of modelling, in which the shell elements assume a
linear variation of internal strains.
Figure 2.17 compares results from each of the three analytical methods with
experimental strain gage measurement, all for the same familiar K-connection geometry. The
pattern of circumferential stress, section C-D, retains the same characteristic shape as seen in the
shell theory results (ref. 26). However, in addition to the high circumferential stresses at the
chord saddle position, there are now equally high longitudinal stresses in the chord at the crown
position, in the gap region between the two branch members (section A-B, near end A). Since
punching shear is proportional to the gradient of shell bending stresses in the chord, we see
 37

indications that much of the load is being transferred into the gap region. This load transfer
across the gap, with the inward punching load of brace offsetting the outward punching load of
brace A, is responsible for the reduced chord ovalizing tendency, lower chord stress, and greater
efficiency of the K-connection. The other crown (heel) position, near point B, has lower chord
stresses and stress gradients, with less load being transferred across the weld, more like a T-
connection.

WELD FOOTPRINT

() LONGITUDINAL STRESS, CHORD SURFACE

420

350

280

210

< 140

co -70 # TOP SIDE BRACE A

A BOTTOM SIDE BRACE A
-M0 COMPARABLE LOCATION
ON BRACE
CIRCUMFERENTIAL STRESS, CHORD SURFACE

Fig. 2.17. Comparisons of analytical results with experimental results for ungrouted K-connection
(tested at 1:2.5 scale).
38

Comparison of the various methods of analysis is also of interest. They all give
essentially the same picture of circumferential chord stresses, which arise locally due to the
overall geometry of the connection and the loading pattern imposed on the branch members.
Indeed, the scatter between replicate experimental measurements is larger than the scatter
between various methods. Experimental stress analysis can be very sensitive to small
eccentricities and indeterminacies in the loading arrangements and to small differences in strain
gage placement.
For longitudinal stress, the Dundrova shell theory has some difficulty reproducing the
complex pattern and sharp peaks with its trigonometric series expansion. Including additional
boundary element terms, which peak sharply at the intersection, would probably improve this
approach. Shell theory retains an advantage over finite element methods in that the time-
consuming effort of mesh generation is eliminated; this effort becomes particularly burdensome
for complex multi-planar connections, for which automated mesh generation schemes are not yet
widely available.
A "re-zone" (ref. 35) or "zoom" (ref. 36) mesh refinement technique can be used to
obtain local strains across the weld in a tubular connection, as shown in Figure 2.18. Results
from both thin-shell and 3-D isoparametric finite element analysis are shown. For the latter,
successive mesh refinement was used to get microscopic detail in the weld area. The coarse

-2000 I P= K Ton

Fig. 2.18. Local Weld Strains by

Re-Mesh Technique, (a) Von
Mises equivalent string along
outer surface of tee connection
at plane of symmetry (ref. 35).
(b) Yoshida's results showing
effect of weld profile (ref. 36).

 39

mesh thin-shell analysis of the overall connection provides forces or displacements to be applied
at the boundaries of the fine mesh 3-D modelling of the weld zone; alternatively, the weld zone
can be treated as a substructure of the overall model. It is instructive to compare the thin-shell
peak strains at mid-plane intersection with the results which would be measured at the hot spot
strain gage locations. The two are of comparable magnitude for the chord hot spot; however,
they differ by a factor of 1.7 on the brace side. It appears that this discrepancy is systematic, and
that thin shell results should be adjusted by a reduction factor when hot spot stresses in the brace
end are computed (see Section 4.3 herein).
Note that very high peak stresses at the toe of the weld are revealed by the analytical
rezone "microscope". Indeed sharp-angle notches at the toe of the weld can become singularities,
in which the peak stress is unbounded, with the finite element result depending on the degree of
mesh refinement, in much the same way that analytical notch stress results depend on the radius
at the root of the notch. Since these microscopic effects are built into the empirical S-N curve
used for fatigue design, however, it is usually not necessary (or even desirable) to include them
again in the engineering stress calculation.
Isoparametric finite elements can also be extended to grouted tubular connections.
Where piles are grouted inside the jacket leg of an offshore structure, the composite chord
becomes a three-layer sandwich of solid elements. Unfortunately, linear solutions indicate high
bond stresses at the grout-to-steel interface, which would likely cause separation or slippage. A
non-linear analysis is required to properly treat these phenomena (see 5.3 herein).

2.3.2 Limit State Methods of Analysis

While elastic analysis provides valuable insights into the general behavior of tubular
connections, and is indispensable in finding stress concentration factors to be used in fatigue
calculations, ultimate strength methods are needed to establish practical working capacities. The
methods to be discussed in this section are: cutting sections (also known as cut-and-try), yield
line methods (based on the upper bound theorem of plasticity), and inelastic finite element
analysis. As these methods involve a degree of approximation and simplification (particularly
the first two), experimental calibration is also desirable.
(i) Cutting Sections. In the days before widespread finite element computer usage, the
method of choice for design of connections was "cut-and-try", particularly for riveted gusset and
angle connections of wide-flange or shapes. Although this type of construction is less
common now, the discipline of the design methodology remains useful, even if a finite element
analysis is also performed, helping the designer to focus on load paths and anticipate failure
modes at a detailed level.
The cut-and-try approach (as outlined in Section 2.5.5c.5 of API RP 2A)consists simply
of cutting sections and taking free bodies. For connections, this is done repetitivelyfor the
connection as a whole, for major subdivisions, for individual member ends, and finally for
smaller and smaller parts of the reinforcement. For each free body taken, the six equations of
statics must be satisfied. Assumptions made about the distribution of boundary stresses acting on
the free body should be reasonably compatible with the relative stiffnesses involved. The
location of cutting planes often affects how well the stress distribution may be intuitively
perceived, and hence the reasonableness of the solutions. Many different cutting planes must be
tried. The success of the method depends on the perceptiveness and thoroughness with which
each part of the connection is investigated.
Static strength analysis and design of complex connections by the cut-and-try method
may be justified on the basis of a loose restatement of the lower-bound theorem of plasticity, as
follows:
40

If the designer can find an assumed distribution of load between the various
elements of a connection, which everywhere satisfies equilibrium with stresses at or
below the material strength, then the actual ultimate strength of the connection will be at
least as much as he calculatesprovided material selection and detailing are such that
yielding can occur without premature failure by local overstraining, brittle fracture, or
buckling.

Reference 37 describes an early attempt to apply the "cut and try" method to the design
of complex tubular connections. Complex gusseted designs were being used for connections in
major tubular structures at the time (early 1960's) because of a general distrust of the capacity of
thin-wall unstiffened shells to carry radial loads. The high local stresses and low capacities being
given by theoretical shell analysis, and service failures with thin-wall unstiffened joints, only
served to heighten this concern. The designer wanted a load path he could understand and trust.
Examples of this philosophy can be found even today.
Figure 2.19 shows a class of connections known as "filler gusset" style, where each
brace end is reinforced by four external gussets, placed radially at 90 degrees apart. These are
less efficient than penetrating gussets, but are much easier to fabricate. Using a "building
block" approach, the axial capacity for each brace end is taken as the sum of the ungusseted
tube-on-tube connection, plus the capacity of the four attached gussets (two type A, one type B,
and one type C), with the gusset contribution being delivered as a shear load where the gusset is
welded to the brace. (Note that if the and C gussets have substantially different capacities, it
would be appropriate to either consider the eccentric reaction in brace design, or under-utilize the
stronger in order to balance the axial load.) The numerical results tabulated in the figure are
intended to only illustrate the kind of building blocks a designer might derive and accumulate for
his repetitive applications; they are specific to mild steel members of the sizes shown, as used in
small offshore platforms of the time (e.g., ref. 38).

ULTIMATE C A P A C I T I E S *
LOAD PERPENDICULAR TO LEG
(KIPS)
BRACE S I Z E ( I N )

CHORD

150 150 348 348 39.750

76 76 176 176 39$.500
122 122 20.500

Fig. 2.19. "Building Block"

capacities for a class of "Filler
Gusset" connections, using mild
steel.
 41

Connection efficiencies (connection capacity/brace yield) for the "filler gusset" designs
range from 44% to 100%, compared with 16% to 50% for the corresponding unstiffened designs.
Their adequacy has been further demonstrated in connections which survived severe overloads in
service (e.g., Figure 14 of ref. 1). Scalloped tips for the gussets (ref. 39) mitigate a "hard spot"
and potential tearing in the brace wall at this location (shown dashed in the figure). Although
"filler gussets" are effective for the smaller compact members, efficiencies drop for larger
members of the same thickness, necessitating a different design approach for larger deepwater
structures.
More specifics on the building blocks of "filler gusset" connections are shown in the
next two figures. Figure 2.20 shows some historical examples of the line load capacities used for
the cylindrical shells. The original attempt tried to use the elastic ring solution, with an effective
width calibrated to match shell theory. However, the resulting capacities generally were too low
to be practical indicating that a lot of existing structures should have already collapsed. With
the advent of Marshall's "punching shear" approach in 1966 (ref. 40), an empirically derived,
more realistic, line load capacity became available, with ultimate capacity depending only on
chord gamma ratio and thickness. Its simplicity and apparent universality made it well suited,
within the context of the naive enthusiasm of the time, for use as a building block in the cut-and-
try approach. It is the basis of the results presented here, which may be termed "optimistic lower
bound" in that possible interactions between the various line load segments are neglected. More
recent punching shear criteria (e.g., AWS since 1984) are specific to connection geometry and
loading pattern, and are not as well suited for this purpose.

ORIGINAL PRESENT

Fig. 2.20. Historical estimates of line load capacity of tubular members. Capacities are for mild steel.

To illustrate the level of detail which is still required for the cut-and-try method to
succeed, Figure 2.21 sets up the constraints for limit state analysis of type C filler gussets. The
careless designer might be tempted to only check for horizontal shear through the gusset.
However, vertical shear at the chord is limited by the small contact length across the gap, and
punching line loads on the braces are required for vertical equilibrium with the inclined load of
the diagonal brace; these in turn create moments about point "0". Thus, in all, there are three
equations of equilibrium, and five limiting conditions of plasticity (inequalities with unknown
forces or moments in the numerator, and membrane shear or line load limits in the denominator),
which must be satisfied as we try to find the maximum horizontal shear which the gusset can
carry. While formal optimization might have been applicable, trial and error solution was used
in practice, to produce the results shown earlier. Somewhat different results would be produced
with Akiyama's building blocks for gusset capacity, as shown in Table 2.2 (ref. 41).
42

Fig. 2.21. Limit state analysis for Filler Gusset Type "C". and are limiting weldline or
membrane shear (kips/inch) capacities of weld or adjoining base metal at gusset-chord and
gusset-brace interfaces, respectively. The w's are corresponding punching line load
capacities. Forces and moments are defined by capitalized symbols as shown in the figure.

TABLE 2.2 AKIYAMA'S RESULTS FOR GUSSET CAPACITY "BUILDING BLOCKS".

cm ID
- id
3
= t\ Fy [ 7 . 0 ] = t\ F y [ 5 . 3 + 2.33] = t F y [21 ]

115 kips 143 kips

AKIYAMA 1560 k - i n

MARSHALL 360 k - i n 120 kips 120 kips

7
v p = F y/ 0 . 5 y
 43

Considerations for the moment capacity of a brace end with filler gussets are illustrated
in Figure 2.22. Referring to part (a) of the figure, in-plane moment capacity is estimated by
taking moments about point "o", with building block contributions from the gussets (V) and
punching line loads on the chord (w). The latter is conservatively approximated by taking the
capacity of the shaded 90-degree segments applied at the tube extremities, as shown by the
dashed vectors. Interaction between axial load, in-plane bending, and out-of-plane bending is
shown in part (b) of the figure, with ultimate strength interaction of the gussets-and-quadrants
model (solid lines) enveloping that of the circular section (dashed).

Fig. 2.22. Lower bound limit

analysis for moments, (a)
Schematic of forces involved,
(b) Interaction diagram.

Another frequently used building block is the "knife-edge" crossing, with intermediate
tube shell or beam flange, as shown in Figure 2.23. Equations expressing the ultimate capacity
for either of two behavior modes are given in the figure, along with guidance on the range of
some of the parameters. Consistent with these being lower bound estimates of ultimate capacity,
it would seem appropriate to take the larger capacity of either the plate bending model or the
spreading model. However, Reference 42 suggests that only the spreading model (with 1:2.5
slope) be used for compression loads, where local buckling of the attachment material might
preclude full development of the plastic mechanism anticipated with the larger K-values in the
plate bending model.
In addition to checking the local capacity for each brace end at a connection
individually, the cut-and-try method also requires that the connection be checked for general
44

AT EACH CROSSING

BENDING OR

t 2[ t 1+ 2 f 1+ 2 a ( t c+ f 2) ] F y

WHERE t 2 IS THINNER ATTACHMENT

f 2 IS FILLET SIZE
= SPREADING SLOPE
RANGE 1 TO 2 . 5
= CONTINUITY FACTOR
1 1 NEAR FREE EDGES (AS SHOWN)
2 WITH CONTINUITY PARALLEL
I TO THINNER ATTACHMENT Fig. 2.23. "Building Block"
4 AT INTERIOR CROSSINGS
c a p a c i t i e s for k n i f e - e d g e
crossings with intermediate
shell plate or flange.

failure modes involving combinations of members. Several relevant cutting planes and load
combinations are shown in Figure 2.24 (refs. 11 and 37). As this is not necessarily an exhaustive
set for all conceivable situations, the designer should spend some time thinking about other
possibilities.
Application to a gusseted joint in a large-diameter leg of an ice-resistant offshore
platform for Cook Inlet, Alaska, is illustrated in Figure 2.25. Because of the high D/t ratios
involved, this design provides for load transfer via membrane stress, rather than radial line loads
as follows:

1. Brace loads are transferred into the gusset via shear in welds (a).
2. Equilibrium of the gusset is established by shear transfer of vertical forces into the
leg, and of horizontal force into the diaphragm, via welds (b).
3. The horizontal force in the diaphragm is transferred into tower leg beam shear
(c).
4. Strength checks are also made by cutting sections as indicated by the dash-dotted
lines.
 45

EH 4
H

( a ) LONGIT. SHEAR = F-, + F , 3

(SEC. BB & CC)

(b) TRANSVERSE SHEAR = F n + F- &U

(SEC. AA)

( c ) IN-PLANE MOMENT = M, + M ?
(SEC. BB & CC)
LOCALLY
APPLIED (d) OUT-OF-PLANE MOMENT = M. + M 4
FORCES (SEC. BB & CC)

( e ) BEARING = Fi + F Q = F 4 - F 2
(SEC. BB & CC)

Fig. 2.24. Cutting sections for

combined load design checks.

DIAPHRAGM
(a)
TOWER LEG?

BRACES

CUTTING
P L A N E S FOR
STRENGTH
= 0 CHECKS

Fig. 2.25. Decomposition of

load paths for large diameter
g u s s e t e d c o n n e c t i o n , (a)
Elevation, (b) Gusset showing
loads, reactions, and cutting
p l a n e s , (c) Plan view of
horizontal diaphragm, (d)
(d) Failure mode at cutting plane A-
POSSIBLE FAILURE A . S e e t e x t for f u r t h e r
MODE AT HOLES explanation.
46

Elastic shell theory may also be used as a very conservative lower bound ultimate
strength. The conservatism may be alleviated somewhat by using the ultimate tensile strength
(F u j t ) instead of yield strength in evaluating capacity. This is justified when there is only a small
zone in which stresses above yield are being tolerated, so that local strains of several percent do
not create unacceptable overall deflections; provided the mode of yielding, e.g., shell bending,
does not create a potential for local buckling; and where the material selection and weld detailing
assures ductile behavior.
Figure 2.26 shows application of shell theory to the cone-cylinder transition in a flared
connection. Part (c) shows how the radial load created by the angle change in member axial
stresses can be handled by hoop stress in a stiffening ring. The limiting stress concentration
factor (SCF) of 1.6 applies to shell stresses adjacent to a rigid ring, with higher values for a
flexible ring. This secondary (deflection-induced) shell bending is not required for static
equilibrium, and may be neglected for ultimate strength considerations.

(a)

(b)

Fig. 2.26. C o n e - c y l i n d e r
(c) transition, (a) Overview, (b)
Unstiffened. (c) With stiffening
ring.

Part (b) shows the unstiffened transition, where the missing radial reaction from the
missing ring creates a load case identical to that of Figure 2.10(a). Here, the local shell bending
stress is required to spread out the missing reaction and cannot be neglected. Two formulas are
shown, one for stress concentration factor (SCF) and the second for static strength efficiency
(EFF), ultimate capacity of the transition versus full yield of the member.
 47

In the SCF formula, the first term is the membrane stress in the cone, while the second
term is shell bending response to the radial line load of t^ fR trnxp.
In the EFF calculation, note that we are using shell bending and membrane traction from
the elastic solution, together with fully plastic section behavior and stress up to F u j r Calculated
limiting angles for 100% EFF with steels up to 60 ksi (Fy/F u j t up to 0.75) are comparable to the
API limits (ref. 43) in Table 2.3. In setting the API limits, an upper bound solution and the
effects of hoop stress were considered.

TABLE 2.3 LIMITING ANGLES FOR UNSTIFFENED CONE-CYLINDER TRANSITIONS)

API ANGLE PER
DIAMETER/THICKNESS ANGLE FIG. 2.26

60 10.5 8.1
48 11.7 9.2
36 13.5 10.8
24 16.4 13.8
18 18.7 16.7
(ii) Yield Line Analysis. The upper bound theorem of plasticity provides a rational
approach to the ultimate strength of stepped box connections with respect to plastic flexural
failure of the chord face. The general method is to assume a rigid-plastic failure mechanism, and
compute the load level at which internal work due to yielding equals external work due to the
loads. Since possible points of overstress may be overlooked (i.e., not chosen as regions of
yielding), the results form an upper bound to the true theoretical limit load; thus the results
depend on the reasonableness of the assumed mechanism, and the analyst should be diligent in
trying different patterns. Fortunately, this search for the minimum need not be totally
exhaustive, as comparisons with test data (or "optimistic" lower bound solutions) often show that
strain hardening and large deflection behavior bring reserve strength which is not being
accounted for.
As applied to box connections, the regions of yielding are lines along which the plastic
bending of the shell is assumed. These yield lines divide the chord face into rigid facets which
make up a kinematically possible (and hopefully reasonable) failure mechanism. Typically, the
branch member is also assumed to act as a rigid body, with no yielding of its own (this is not a
requirement of the lower bound theorem, and failure mechanisms involving partial yielding and
deformation within the side wall of main or branch member would be appropriate for matched
box connections, for example).
Figure 2.27 shows how the yield line calculation would be formulated. External work
due to applied loads or moments appears in the left-hand side of the equation. Internal work due
to plastic bending rotation along all the yield line segments is summed up at the right. Segment
lengths and rotations are just a matter of geometry (one finds the angle by cutting a section
perpendicular to the yield line). Terms are defined in the figure. The quantity before the
summation sign is the plastic bending capacity, e.g., kip-inches per inch, of the chord wall, and is
often used as a scaling parameter when expressing the final results in total load format, with all
other terms dimensionless (this may take some getting used to, but it does have the proper
dimension of load, and it does really work).
The author's early application of the method (ref. 44), to derive the ultimate capacity of
a T-connection in square tubes, is shown in Figure 2.28. The 45 pattern was assumed without
making a rigorous search for alternatives which might give lower capacity. We start by showing
48

AXIAL LOAD BENDING MOMENT

c yo Fig. 2.27. Formulation of yield

ALL line analysis based on upper
YIELD
LINES bound theorem: external work
equals internal work.
= SPECIFIED MINIMUM YIELD STRENGTH OF MAIN MEMBER
= ANGULAR ROTATION OF YIELD LINE SEGMENT i IS
DETERMINED BY GEOMETRY OF MECHANISM
= LENGTH OF YIELD LINE SEGMENT i
= WALL THICKNESS OF CHORD

4(b+2c) 4b 1 + 4 J T c J 7 -
C C
I- 0> (2) (3) J

2r f b +
g i v i n g = 4 t c Fy o [ ~Tc~

For b = BD
2c = ( l - e ) D Fig. 2.28. Yield line derivation
and = for T-connection in square
V = tubes.
P 4b t c

2
r
tc Fy o
h
I -A- 0.25 . yo
6(1 - ) TL5T
 49

the internal work terms for all the line segments in detail, then simplifying to give the total load
capacity, P. Substituting the dimensionless terms beta and gamma, we then get an even simpler
expression for total load, as well as the corresponding expression for ultimate punching shear.
Here the basic punching shear term (Fy Q/0.5 ) is modified by a term expressing the influence of
beta.

= NON-DIMENSIONAL GAP PARAMETER FOR K-JOINTS

(c)

Fig. 2.29. Yield line patterns, (a) T-connection, square tubes, with "fan" corners, (b) Y-connection or
rectangular T-connection. (c) K-connection. (d) In-plane bending, (e) Out-of-plane
bending.

Yield line patterns for several different box tube connection configurations are shown in
Figure 2.29. Part (a) of the figure shows an alternate yield line pattern for the square tube T-
connection, with "fan" corners (ref. 45); this produces 10% to 20% lower capacity than the
simpler pattern of the previous figure. However, Figure 2.30 shows that the two analytical
results are closer to each other than they are to the test data (refs. 46 & 47). The tests show that
there are additional sources of reserve strength, so that the "upper bound" method may actually
err on the low side. However, for beta approaching unity, yield line analysis gives unrealistically
high capacities; here material shear failure, or some of the other failure modes discussed earlier,
must be considered.
Yield line solutions of Redwood and Jubb (ref. 45), reduced to punching shear format,
are given in Table 2.4. Note that the right hand part of the expression above the table is similar
to that in Figure 2.28, with the basic punching shear term modified by a term which depends
on connection geometry and load pattern. Equivalent expressions for total load (or moment)
format are also shown. The following equations show how total load and moment are
related to punching shear for box connections having footprint dimensions T?D )3D and chord
wall t c :
50

. .
D

7 =
2tC

YIELD
LINES

o.z Q4 0.6 1.0

- RATIO

Fig. 2.30. Ultimate strength analysis of square tubes vs. Graffs test results.

= 2 { + r ? ) D t cV p (2.1)

2
= ( + ^)D t V (2.2)
C

V = Q y = 4 F Q (2.3)
D
Q 0.57 Y Q

TABLE 2.4 ULTIMATE CAPACITIES OF - & Y-BOX CONNECTIONS

Results from yield line analysis reduced to punching shear format.

ultimate V p = Qq

f
Qq f o r Qq r Qq f o r E X P R E S S I O N FOR U L T I M A T E
L O A D OR MOMENT
0<3<0.635 0.635<3<1.0* =3=0.5

2
.+ + 3
AXIAL * +.
0.89 sine =
2 1- - /
LOAD 2
4(+) 4(3+) F yT . 8(3+n)Qq

1
1 . , + + n 2 =
IN-PLANE + +
2, 2 1- 1.52 2
BENDING
F yT n D '4(3 + f ) Q q
4(3 + ) 4(3 + $)

alternate failure modes may g o v e r n as beta approaches unity.

Note that for scaling the bending results, an additional length term is needed, here taken as the
width of the branch member footprint in the plane of bending.
 51

The Redwood and Jubb solutions for axially loaded T- and Y-connections of rectangular
tubes are plotted in Figure 2.31. Note the importance of the footprint length parameter, eta,
especially at small beta. Decreasing punching shear capacity for larger eta indicates that the total
load capacity is increasing more slowly than the footprint length.

Fig. 2.31. Yield line results for

a x i a l l y l o a d e d T- and Y-
connections.

Mouty's solution for K-connections (ref. 48), the equations in both punching shear and
total load formats, as well as a plot, can be found in Figure 2.32. Note the high capacity which is
indicated for small gap between the braces. For gap approaching zero (as for beta approaching
unity), unrealistically high capacities are indicated by the yield line solution. For gap less than
chord thickness, punching shear exceeds the shear strength of the material, and a scissors failure
rather than bending failure can occur in the gap region (provided the branch member or
attachment weld does not fail first).
L
yo
ultimate V D = Q a
0.5

ultimate sin - yo'-c 8 ( + )

+ 2

1- 2 nJ 1 -

8 ( + )

.6

.8
Fig. 2.32. Yield line solution
for K-connections (see text for
Q.A-
gap parameter caveat).
52

Figure 2.33 part (a) shows that this type of K-connection behavior involves uneven
distribution of load in the branch member, even for gaps larger than one chord thickness. In the
yield line solution, even though pure axial deflection of the branch member is assumed, the
uneven distribution of load creates a bending moment in the branch member. Load tests of
isolated connections with imposed pure axial displacement would show similar behavior. This
moment does no work and drops out of the analysis. In the real world, however, one may
question whether the branch member is strong enough or stiff enough to sustain this moment.
Joint deformation or rotation capacity is also important here, and undersized welds could fail
prematurely by "unzipping". Weaker branch members or long flexible braces may lead to the
alternative failure mechanism shown in part (b) of the figure. Here a pure axial load results in
end rotations of the branch member, again doing no external work, with the yield line pattern and
capacity equation which results given in the figure. For beta and eta of 0.3, and zeta (gap
parameter) of 0 . 1 , the alternative failure mechanism is 22% weaker, with the difference
becoming larger for smaller gaps. Tests of connections with pure axial load (e.g., with a Lehigh
load following mechanism) would be expected to show similar behavior. Testing of truss
subassemblages with realistic full length braces was eventually required to resolve these
questions (ref. 49). As a result, design codes do not rely directly on yield line analysis of K-
connections, but on simpler empirical expressions (see Chapter 5 and Appendix ).

Fig. 2.33. Alternative modes of behavior for K-connection with small gap and small . (a) Pure axial
deflection causes moments in short rigid braces, (b) Pure axial load causes rotation in long
flexible braces.
 53

(Hi) Inelastic Finite Element Methods. Inelastic finite element analysis can overcome
some of the difficulties of using fictitiously high local stresses from an elastic analysis as a basis
for design.
Clough (ref. 50) has described the finite element stiffness method in terms of the
following steps: (1) Express element internal displacements in terms of assumed deformation
patterns (e.g., constant strain, linearly varying curvature, higher order spline functions, etc.),
which approximate behavior of the continuum, are more-or-less compatible at element
boundaries, and whose magnitude is given by generalized coordinates, one for each degree of
freedom. (Steps 2-4) Express both nodal displacements and internal strains in terms of the same
generalized coordinates and deformation patterns, and vice versa. (5) Evaluate internal stresses
from the internal strains, with material characteristics represented by the stress-strain matrix; in
general these may be isotropic, orthotropic, elasto-plastic, or any other specified characteristics.
(6) In generalized coordinates, integrate over the element volume to compute internal virtual
work due to internal stresses and strains, compute external work due to nodal forces and
displacements; and equate these to extract the element stiffness. (7) Transform to the desired
nodal point stiffness matrix for each element.
The process is repeated for all the elements to assemble the global stiffness matrix for
the whole structure. For linear structure, this is solved by matrix inversion or numerically
equivalent methods. The two principal methods of treating non-linear frame structures are also
applicable for inelastic finite element analysisincremental loading and iteration. Where only
non-linearity due to material yielding is involved, this may be represented by successive
modification of the stress-strain relationship during the analysis of element internal virtual work
and stiffness (step 5). However, defining a general non-linear flow rule (e.g., outward normal
strain tensor based on the level of von Mises stress and orientation of the failure surface) can be a
challenge, especially if there are history effects, e.g., unloading, are involved. Also, the
concentration of strain in regions of local yielding is not fully reflected in the assumed
deformation patterns within elements and must often be approximated by taking a finer mesh.
Large deflection geometric non-linearities can be handled by re-formulating the stiffness in terms
of displaced nodal coordinates.
Pedro Marcal (ref. 51) has described the element stiffness matrix in terms of linear
(fixed) terms, and incremental terms which vary linearly and quadratically with the nodal
displacements. He then derived expressions for potential energy, total stiffness, incremental
stiffness, and buckling stability. These behavioral models, taken collectively, afforded the
opportunity to develop a single automated systemMARCfor handling a broad class of non-
linear structural analysis problems.
We would like to be able to say that inelastic finite element methods are capable of
finding the correct ultimate limit state, in between the results of upper bound and lower bound
methods. Unfortunately, this is not always the case. Finite element results can err on the high
(unsafe) side if the mesh is too coarse, if missing deformation patterns turn out to be important
(e.g., material failure in punching shear), or if local buckling possibilities and the limited
ductility of material at the toe of welds are ignored. They can err on the low side if strain
hardening is understated, if large deflection membrane effects are ignored, or if neglected effects
of finite weld size (including locally h a r d e n e d heat affected zones) are significant.
Approximations, simplifications, and programming "bugs" in element formulation, as well as
poor convergence in the solution procedure, may result in undetected errors in either direction.
Thus the analysis procedures (mesh size and layout, element selection, material
descriptions, and solution strategy) should be carefully calibrated and benchmarked against
reliable experimental results. Once this is done, however, finite element analysis can be used as
a research tool to discern trends in ultimate strength behavior which might otherwise be obscured
54

by scatter in experimental programsfor example, the interaction envelope for various combined
loads on a given connection geometry (ref. 52).
Reference 53 gives an additional example of such benchmarking exercises by research
organizations, using the computer programs MARC/PATRAN and ADINA/PMBSHELL (the
second-named program in each case is the mesh generator). Figure 2.34 shows some
comparisons between computed and experimental results. The analysis is as close to the
experiments as they are to each other (three "identical" specimens). It required a CRAY
supercomputer.

D = 838mm
= 25.4

d = 406
t = 8

3 4 0 MPa AT F I R S T Y I E L D
4 4 0 MPa AT 7 . 5 % S T R A I N
4 5 0 M P a U L T I M A T E A T 22%

(a) ONE - QUARTER OF THE - JOINT

F i g . 2.34. Inelastic finite

element analysis of T-joint (ref.
52). (a) Mesh, MARC thin shell
element type 72. (b) Three
"identical" benchmark
200 250 D IOO 150 200 250 experiments, (c) Comparable
RAM DISPLACEMENT ( m m ) AXIAL BRACE DISPLACEMENT ( m m )
analyses (ref. 53).
(b) (c)
Even though the fundamentals were set forth over 20 years ago, and a number of
applicable computer programs exist today, design applications of inelastic finite element analysis
have been hampered by the same mesh generation problems as for elastic analysis, by the huge
amounts of computer time involved, and by the difficulty of maintaining calibrated expertise in
view of infrequent applications. Designers whose primary need is getting their project done on
schedule, rather than conducting quasi-research studies, usually opt for the use of empirical
design equations instead.

2.3.3 Model Tests

Model tests can be used to study the elastic stresses, ultimate strength limit state, and
fatigue behavior of tubular connections. They provided for our first insights into these issues,
and are often used as the standard against which other proposed methods of analysis are
compared.
In the early days of the offshore industry, model tests provided the only practical way to
prove out connection designs; insistence on being able to analyze them theoretically would have
prohibited the use of otherwise attractive schemes. While the cut and try methodology could
sometimes be used to demonstrate adequate static strength, at best it gave only a crude measure
of the stress concentration factors needed for fatigue analysis. Even today, for complex
connections beyond the range of automated finite element mesh generation, a model test is
competitive in cost with the numerical analysis.
 55

Nevertheless, some attempt at prior analysis is always recommended for the connections
to be tested. This will (1) save expensive testing of grossly inadequate designs, (2) help in the
selection of critical load patterns for the test, and (3) enhance the designer's insight, increasing
the usefulness of information gained from the exercise.
Early strain gage studies of steel models of various tubular connections were performed
by Toprac (refs. 21 & 54), Bouwkamp et al (refs. 55 & 56) and Grigory et al (refs. 57 & 58), in
addition to specific design studies (e.g., ref. 59). In many cases, these experimental stress
analyses were followed by a fatigue test to failure, providing the original data base from which
the AWS-X fatigue curve for hot spot stress was developed. A typical model test set-up is shown
in Figure 2.35.

Fig. 2.35. Test set-up for Cook Inlet model connection.

While unfortunately there was considerable variation in technique between the different
studies, a "standard" technique which is consistent with the early American data base may be
described according to the following recommendations (ref. 60). These are more elaborate than
would be required for a simple static strength test with axially loaded branch members.

1. The model should be constructed of steel having the same stress-strain properties as
the prototype. Caution: thin gage material of the same specification is typically
stronger.
56

2. The model should be as large as possible, subject to the limitations of jacks and
reaction frames. Typical scale factors have ranged from 1:6 to 1:2.5.

3. Geometric scaling should be rigorously observed for all dimensions and details of
connections which model a particular prototype design. Actual (rather than
nominal) thickness should be used in data interpretation.

4. Weld sizes should be scaled down appropriately to the model thickness. Oversize
welds in the model can lead to unrealistically favorable behavior (e.g., oversized
fillets which move the hotspot to an area of lower stress). Gas-metal arc welding
can often be used to advantage on thin-gage model material.

5. Applied loads should be scaled so as to produce the same nominal axial and
bending stresses in the members for both model and prototype. Selected critical
load patterns may be desired for testing specific design; while loads scaled to
branch member capacity would be appropriate for generic research tests. Loads
scale down as the square of the dimensional scale factor, moments as the cube.
Gravity loads due to connection self-weight are not properly scaled, but are usually
negligible. Pressure (e.g., hydrostatic) is the same for both model and prototype.

6. The loading system should be statically determinate in three dimensions so that no

unintended loads get into the model. Particular care is required for compressive
loads to avoid eccentricities due to model displacements. Design of the test
loading system can be a challenge in itself. Preferably, the test frame should be
sized to out-perform the object being tested by factors of 2 on load, 8 on life.

7. A redundant system for applying and monitoring loads (e.g., both jack pressures
and strain gages to measure loads and nominal stress) should be used, with
discrepancies between the two resolved.

8. Shakedown refers to the process by which the load-strain behavior stabilizes into
closed loops after a few cycles of load. On the virgin loading of as-welded
specimens, residual stresses may cause premature yielding, which may obscure or
exaggerate the strains due to applied loads. See Figure 2.36.
Q
LOAD

LOAD

<C

T R = 3324M e TR = 2352M 2180

|

2 2 2
Cpeak = M

/
/ STRAIN / / STRAIN / STRAIN
/ /
/ /
/
^ (b) (c)
/
(a)
Fig. 2.36. Three estimates of hot spot strain range from the same model tubular connection,
Bouwkamp Phase II (ref. 56), showing the importance of "Shakedown", (a) Initial loading
and unloading, as used in Bouwkamp Phase I; dashed line shows assumed symmetry about
strain axis, (b) Actual full cycle shows lower strain range, (c) 70th cycle shows stabilized,
closed, symmetric hysteresis loop.
 57

9. Stress-coat is a brittle lacquer which cracks with tensile strains of about 700 in/in
(20 ksi). Above this threshold, crack density can be used to read strains with a
tolerance of 200 in/in (6 ksi), providing the principal direction and rough
magnitude of major tensile stress. A stress-coat test after shakedown is suitable for
determining strain gage locations and orientation.

10. Strain gages should be located to capture the worst hot spot strains, and, where
feasible, give additional information about the stress distribution. Hot spot gages
are centered within 0.25 inch (6mm) to 0.1 VRt of the weld, and typically have a
gage length of 0.125 inch (3mm). Because of the steep stress gradient in this area,
larger gages further from the weld will yield stresses which are too low. The
surface stress gradient may be measured by a series of gages leading away from the
weld.

11. The worst hot spot strain range measured after shakedown, e is the basis of the
empirical fatigue design curve. If the orientation of maximum principal strain is
known to be perpendicular to the weld, uniaxial gages may suffice, and the stress
(or strain) concentration factor may be defined as

SCF = e T R / A f n (2.4)

where is Young's modulus and Afn is the variation in nominal brace stress. The
stress perpendicular to the weld is what is further amplified by notch effects at the
weld toe. However, to determine the true biaxial surface stresses, strain gage
rosettes are required.

Depending on the design purpose for which a particular connection is being tested, the
experimental stress analysis may be followed by a proof load and/or fatigue test. Static strength
may be verified by a proof load which consists of the worst simulated design load, increased by
the nominal safety factor used for member design (usually 1.67, or 1.25 where the 1/3 increase
applies). This ensures that the connection will deliver a similar margin of safety as being used
for the rest of the structure, but does not indicate whether the connection is statically over-
designed. Most connections should be able to withstand the proof load with little or no distress,
so that a fatigue test can still follow. Caution: Application and release of a tensile proof load
may leave residual compression at notches and flaw tips, leading to unrealistically favorable
fatigue test results. Thus, compressive proof load, followed by tensile fatigue loads, is
preferable.
The fatigue test may be thought of as a way to ferret out critical hot spot locations that
might have been missed by the stress analysis. For this purpose, a relatively high cyclic load, to
produce failure in about 10,000 cycles, is appropriate. This would involve a hot spot strain range
of about 0.25% (2500 in/in).
Crack initiation and growth should be monitored with (1) visual observation against a
white painted surface, or magnetic particle inspection for surface appearance (initiation) and (2)
ultrasonics, dye marking, or potential drop measurements to trace the growth in depth. The
model should be inspected at geometric multiples of the expected cycles to failure e.g., .05,
.07, .10, .14, .20, etc.so that unexpectedly early initiation can at least be approximately
determined. Crack growth typically occupies the major part of fatigue life. During crack
growth, the model should be monitored continuously; a load counter and automatic shut down
upon failure is advisable. Failure is defined as complete joint separation, or inability to develop
58

the test load within available jack stroke. Through cracking (e.g., as determined by leakage) is
also widely reported as the failure event.
Models constructed and tested as described in the foregoing will reproduce the elastic,
yield, and ultimate strength behavior of the prototype, at comparable levels of applied nominal
stress. Fracture behavior does not scale up as well, particularly when the model is in the plane
stress region and the thicker prototype in the plane strain or transition region, so that the model
may exhibit relatively more ductile crack propagation behavior that can be realized in the
prototype.
European research (ref. 19) also indicates a size effect in fatigue, such that experimental
fatigue life should be divided by a factor of F ^ - ^ , where 1:F is the scale factor, in order to
estimate the corresponding fatigue life of the prototype. Further, a somewhat different definition
of hot spot stress has been adopted (refs. 61 & 62), as discussed further in Sections 2.4.1 and
Chapter 4
The fatigue life of the prototype cannot be extrapolated directly from that of the model,
but must be determined by calculations which reflect the applied stress spectrum and other
service conditions of the structure. Here, environmental effects create the most uncertainty. The
endurance limit for crack initiation is lost in seawater under freely corroding conditions. Fatigue
crack propagation in salt water can be 3 to 10 times as fast as in air. Although the assumption is
often made that cathodic protection restores the in-air fatigue performance, this is not always
true, particularly in the case of excessive current levels. See Figure 2.37. Because of the
inconvenience of conducting elaborate tubular connection tests in anything but air, it seems more
appropriate to evaluate environmental effects with correlative materials tests, and to modify the
air-based fatigue test results accordingly.

RECOVERY POSSIBLE WITH

CATHODIC PROTECTION

POSSIBLE LOSS
FROM EXCESSIVE
CATHODIC PROTECTION]

HY 80 IN NaCl SOLUTION

| - * * INCREASING CATHODIC PROTECTION

Fig. 2.37. Effect of environment on low cycle fatigue of a heat treated steel, from NRL Report 6167.

A number of recent experimental stress analyses, using arrays of micro strain gages and
photoelastic stress-freezing techniques (refs. 63-65) have exceeded the sophistication of the
techniques used earlier. These reveal microscopic peak stresses at the root and toes of welds,
which are higher than the standard hot spot measurements, and as might be expected
inconsistent with the AWS-X fatigue curve. While the ratio of these microscopic peak stresses to
standard hot spot measurements is not invariant in the scientific sense, the latter remains a useful
engineering approximation. While we may someday have fatigue criteria which make use of
these microscopic stresses (ref. 70), today's designers must take care to select stress analysis
results which are compatible with the fatigue criteria they are using.
Reference 67 describes an alternative model testing procedure, in which acrylic plastic
models are strain-gaged in order to measure the hot spot stresses. The advantage of this method
is easier model fabrication than for steel, and lower test loads. Loads are scaled to produce
comparable nominal strains as the prototype, rather than comparable stresses, as the modulus of
 59

elasticity is much lower for plastic than for steel. The effect of the different Poisson's ratio
appears to be negligibly small. These plastic models cannot be used directly for proof load or
fatigue tests; hence the measured hot spot strains must be used with an empirical design curve
appropriate to the welded steel prototype.

2.4 DESIGN SIMPLIFICATIONS

At the beginning of this chapter, we defined two very useful design simplifications
hot spot strain and punching shear. These terms were used, sometimes implicitly, in discussing
failure modes and methods of analysis for tubular connections. Now, we shall undertake a more
detailed development of these concepts.
Let us begin by reviewing the several scales at which analysis of a tubular structure
should be considered. A global stress analysis of the overall structure resolves applied gravity
and environmental loads into nominal axial and bending stresses in the various members. A
typical level of nominal stress for tubular bracing would be 20 ksi (or 140 MPa), for the lifetime
extreme wave load. If we now focus on the connection as a structure, we might find punching
shear on the order of 7 to 10 ksi (50 to 70 MPa). An experimental or finite element stress
analysis of the connection would typically give hot spot stresses around of 60 ksi (420 MPa); that
is, the geometric stress concentration factor (SCF) is about 3.0 at potential fatigue sites, even in
well designed connections.

2.4.1 Hot Spot Stress

Among the various levels of detail in stress analysis which may be adopted as a basis for
fatigue calculations, the hot spot stress has evolved as the most practical basis for design
purposes. Hot spot stress places many different structural geometries on a common basis,
ranging from butt welds to nozzles in pressure vessels, to tubular joints in offshore platforms. In
AWS practice, the reference stress (or strain) is the total range which would be measured by a
strain gage placed adjacent to the toe of the weld, and oriented perpendicular to the weld so as to
reflect the stress which will be amplified by weld toe discontinuities. This is used with an
empirical S-N curve based on measured hot spot stress and cycles to failure, in tests of realistic
as-welded hardware. The effect of representative local/microscopic discontinuities at the weld
toe are presumed to be built into the data base of realistic as-welded hardware.
Techniques used in the early tests were described in Section 2.3.3. These constitute the
early de facto definition of hot spot stress.
(i) Development. American development of the concept of hot spot strain as a useful
design parameter for low cycle fatigue and fracture is described more fully in Chapter 4. Much
of this early work was proprietary.
Iida (ref. 70) discusses the application of the hot spot strain methodology to non-tubular
applications as well. However, his chronology of the historical development lags that presented
herein, because of reliance upon what was archived in the open literature.
(ii) Attributes. To be of practical use to designers, it was recognized that the hot spot
strain should have the following desirable attributes: (1) It should place different connection
geometries on a common design basis. (2) For repetitive designs, the results should be
generalizable in terms of stress concentration factors; and be invariant for a given connection
geometry, depending only on load pattern. (3) The design reference stress or strain should be
equally derivable from model tests or analysis (e.g., thin shell finite element) of the connection.
(4) Weld notch effects, residual stresses, etc., which are not amenable to such measurement or
calculation, are implicitly included in the empirical design curve. (5) Hot spot strain should
provide a measure of safety consistent with other design methods being used.
60

The first attribute was demonstrated by having a wide variety of practical as-welded
hardware geometries in the data base. The full data set, as of mid-1973 (ref. 71) is shown in
Figure 2.38. They all fall in a common scatter band. Te fact that the scatter band is rather broad
indicates that hot spot strain is a useful design approximation, not a precise research tool.
TUBULAR JOINTS FATIGUE DATA

SRI PVRC
.E SRI SHELL PAN A M
*^ MGS-C
SEDCO

Fig. 2.38. Database and hot spot strain design curve, about the time of first inclusion in AWS Dl .1-72.

The second attribute was the motivation behind all the attention given to stress
concentration factors (SCF), with shell analysis and finite element work leading towards
parametric SCF formulas, both in the early work and in many later efforts. In a practical sense,
designers need SCF which are uniquely defined in terms of connection type (, , , X, etc.),
geometry (tau, gamma, beta, eta, theta, zeta, etc.), and load pattern (axial, in-plane bending, out-
of-plane bending, and degree of chord ovalizing). In its simplest terms, the SCF is the ratio of
hot spot stress to the corresponding nominal stress in the adjoining branch member.
The third attribute of hot spot stress, derivability from model tests or analysis of the
connection, forms the basis of the parametric SCF used in general practice and puts more
accurate determination of hot spot stress for a specific design within reach of designers willing to
spend the time and money required.
Consistency with the original design curve, would mean consistency with the strain gage
techniques which produced the data base. Hot spot strains were measured as near as practical,
but not exactly, at the toe of the weld. In the steep stress gradients of typical hot spot regions,
strain gage size and corresponding placement greatly affect the results obtained. Pickett's
original design curve (ref. 68) was based on full-size strain gages (0.25-in. gage length) on full-
size prototype hardware. Early model testing at SWRI (ref. 57) attempted to keep approximate
proportionality of weld size (using small short-arc welding passes) and strain gage size (e.g.,
0.125-in. gages for 1:2.5 scale models of 50-inch Gulf of Mexico joint-cans). More recent
understanding and alternative definitionse.g., European WG3, the SAE rule, and even more
detailed consideration of weld toe notch effectsare given in Chapter 6.
 61

The fourth attribute of hot spot strain is that items not subject to measurement or
calculation are built into the design curves. One such item is the weld shape notch effect and the
closely related "size effect", as will be discussed in Chapter 7. For low cycle fatigue of as-
welded structures, total range of stress or strain becomes the relevant parameter, without
reference to mean stress or stress ratio.
The fifth attribute, reliability, can be appreciated in connection with Figure 2.38, in
which we see the design curve falling on the safe side of roughly 97% of the data, with the
median fatigue life being 5 to 8 times that at the design curve.
Consistency with other design codes was achieved as illustrated in Figure 2.39. AWS
curve C-C for butt welds was drawn to match the corresponding AISC criteria (ref. 72), as shown
in part (a) of the figure. A continuous curve, rather than tabulated step function allowables, was
used because it is more appropriate to offshore structures subjected to a continuous spectrum of
cyclic loads. In this simple situation, the nominal stress (axial plus beam bending) is the same as
what a strain gage adjacent to the weld would measure. Curve X-X for hot spot stress in tubular
connections is an extension of curve C-C, and entirely consistent, since the local transverse stress
adjacent to the weld is considered in both cases. In order to achieve this consistency, curve X-X
had to be drawn lower than both the author's original design curve (ref. 69) and the AS ME
design curve (ref. 73). In Chapter 7, we shall examine new data which has resulted in today's
AWS design criteria being defined even more conservatively, to take care of weld profile and
size effects.

2 3 4 5 6 7 8
(a) Fig. 2.39. The original AWS
10 10 I0 I0 I0 I0 10 hot spot design curve, (a) Curve
CYCLES
C-C for butt welds, compared to
A I S C . (b) Curve X-X for
tubular joints, compared to
ASME.

102 103 10^ 105

CYCLES
62

(iii) Strength. Relatively little work has been done on the use of hot spot stress or strain
as an indicator of ultimate capacity of tubular connections, as there are alternative design
methods which address the problem empirically, requiring less analytical effort. In section
2.3.2(i), we discussed the use of elastic finite element analysis as a lower bound ultimate
capacity; the caveats given therein regarding our ability to take full advantage of plastic section
properties in shell bending and strain hardening in the hot spot region are repeated in the ASME
treatment of peak strains. Ideally, a model test or inelastic finite element analysis would put
questions of excessive deflection, plastic collapse, or local buckling to rest. Practically, we
would like some simple guidelines to be used in lieu of testing.
Smooth steel test coupons can take strains of 40-50% in a standard bend test, and 70-
100% (logarithmic definition) in the necked-down fracture region of a tension test. As-welded
and notched geometries have a lower tolerance limits for strain; a Ship Structure Committee
research project is currently seeking to define these limits. The original design curve X had a
peak strain limit of 5% for fatigue failure in one cycle, and the Betero & Popov displacement
controlled tests of welded beams (ref. 74) show lives of 10-1000 cycles at peak strains of 5-15
times yield. The most severely loaded tubular connection tested at SWRI failed by brittle
fracture after a few hundred cycles at a strain amplitude of three times yield; the terminal failure
occurred after development of a through-wall fatigue crack several inches long, when the
laboratory cooled overnight to a temperature below the Charpy transition temperature of the steel
(refs. 57 & 58).
Figure 2.40 compares several strength and fracture limits on peak hot spot strain, plotted
non-dimensionally. The CTOD (crack tip opening displacement) design curve from PD 6493
(ref. 75) is shown for several levels of residual stress. For a given material toughness (expressed
by the ratio of critical CTOD to yield strain), the critical (or terminal) flaw size which can be
tolerated gets smaller as the hot spot strains get larger. Auxiliary scales give actual flaw sizes for
specific combinations of CTOD and yield strength.

Fig. 2.40. Strength and fracture limits on peak hot spot strain.
 63

Qualitative results from the Norwegian A5.2 project show a similar trend; except that
using elastic SCF to estimate a fictitious hot spot strain much beyond yield can lead to
unconservative predictions in fracture, just as it did in fatigue. The Norwegian results also
indicate the ultimate limit state (failure by yielding) well on the safe side of the lower bound
prediction based on strain hardening and plastic section modulus. Open disclosure of the A5.2
results in more than qualitative form is eagerly awaited.
The design point for the author's "Big MACS" connections (ref. 76) for Bullwinkle
(Figure 1.2(e)) is also shown: for joint-can steel having 6-inch toughness parameter (/e ) in 4-
inch thickness, yield level hot spot strains limit the terminal flaw depth to only 30% of the
thickness. Fortunately, the major loads on this connection are in compression.

2.4.2 Punching Shear

Punching shear has been defined earlier, at the beginning of this chapter, as nominal
stress on a potential failure surface in the chord wall, around the branch member footprint as
shown in Figure 2.41, reflecting the influence of radial line loads at that locus. It was used to
describe the local failure mode of tubular connections, whether such failure involved local shear
failure only, plastic failure of the chord face, or combinations of the two (Section 2.2.1). In
discussing shell theory (Section 2.3.1(i)), punching shear was related to the gradient of shell
bending stress, and used as a plotting parameter for displaying the elastic stress efficiency of T-
connections. Punching shear, and the corresponding radial line load capacity, was used as a
building block in cut-and-try analysis (Section 2.3.2(i)). The results of yield line analysis were
also expressed in punching shear format (Section 2.3.2(H)). For a simple-minded concept, it ends
up being remarkably versatile; Hicks may have been right after all, when he remarked (somewhat
facetiously, in discussion of ref. 76) that we might end up doing more with punching shear than
with hot spot stress.

F i g . 2 . 4 1 . Definitions of
punching shear, (a) Local
failure in main member at toe of
weld, (b) Geometry and statics
of overall connection.

Although Johnston's first publication of the punching shear, or shear-area, method was
over 25 years ago (ref. 77), the method had already been in use by designers of offshore
platforms for several years. This method considers that axial load in the brace is carried by direct
64

shear in the chord; if the shear area (chord thickness times brace circumference) is inadequate to
carry the radial load at a reasonable stress level (17% to 25% of yield, ref. 78), then gusset plates
or additional chord thickness is used to give additional shear area. Apparently, designers were
already aware that the full material shear strength (53% of yield allowed by AISC, with the 1/3
increase) was not a reasonable design limit. Nevertheless, punching shear is a convenient
intermediate checkpoint, giving designers a feel for the relative severity of loading on different
connections and providing a basis for extrapolating prior success/failure experience. It is
interesting to note that today's design criteria still gives the same allowable capacity, for chords
having D/T of about 20.
As first published by the author (ref. 1), strength design criteria for tubular connections
used punching shear, calculated on the basis of local failure in the main member at the toe of the
weld, as shown in Figure 2.41(a). The equation for the calculation of the acting punching shear,
Vp, due to applied loads is:

a c t i n g V p = r fa sin0 (2.5)

where tau ( r ) is the ratio of branch thickness to chord thickness, fa is the branch axial stress
(beam bending stress is treated the same way), and theta is the angle between branch and chord
member centerlines. Tau reflects the overriding importance of chord thickness. The inclusion of
sin0 indicates that it is the radial component of line loads which influences shell stresses and
ultimate failure in the chord wall.
Although the simple expression of Equation 2.5 is back in use today, both AWS and
API Code Committee initially wanted a more rigorous calculation, based on geometry and statics
of the overall connection, as shown in Figure 2.41(b). Equations for this calculation are given in
Table 2.5 Axial load, in-plane bending, and out-of-plane bending are treated separately. Beam

TABLE 2.5 EXPRESSIONS FOR NOMINAL PUNCHING SHEAR AND WELD STRESS, BASED
ON OVERALL CONNECTION GEOMETRY AND STATICS, FOR THIN-WALL
CIRCULAR SECTIONS

LOAD TYPE P U N C H I N G SHEAR V p WELD S T R E S S f w

K-FACTOR

1 +
K sine
^ sine
AXIAL \ t / Ka r
a " 2
w w
OR S E E BELOW

2
IN-PLANE
BENDING
r 1*
K
fib \ ! b y _ / ^ \ /
by
_
4
sine
sine
by

OUT-OF-PLANE r bz sin
f
1.4-
BENDING
f + K sine
(4) bz[(t^f) {wt) bz 4
K t o= r l/sine
f f r 2
COMBINED tb [ a rm + b / m \ K a A S BEFORE
K
f a +
LOADS * ( r ) t w [ a *w K b \ r w / K K
b " by

fa = A X I A L S T R E S S I N BRANCH = t b/ t c MORE A C C U R A T E K fl PER B S 4 4 9 :

f bv = I N - P L A N E BENDING tb = BRANCH T H I C K N E S S 2 2
f b2 = OUT-OF-PLANE BENDING t_ = CHORD T H I C K N E S S X + + 3>/x + Y WITH
= f2 +
U = WELD E F F E C T I V E T H R O A T 2
fb bz rm = MEAN R A D I U S O F BRANCH 1 3 - B
X =
Kb I S LESSER OF K by OR K bz
r" = BRANCH R A D I U S T O C E N T R O I D O F 25 3ff -2 2
WELD E F F E C T I V E T H R O A T
tb
2 *
ft) -(St)'
 65

shear and torsion are usually neglected for punching shear calculations; they are usually small in
the long slender members of offshore platforms, and cause no punching in the typical beam-and-
column framing of buildings. Sin0 appears in the axial expression, but not in-plane bending,
because the footprint takes the full moment, regardless of brace angle. For out-of-plane bending,
the moment in the brace resolves into a moment and a torsion on the footprint; only the moment
causes radial line loads and punching. In early code treatment of combined axial load and
bending, the two bending moments were resolved vectorially, and a worst-case K-factor used
with the resultant.

/ fa s i n e 9 + \
acting V

Fig. 2.42. Plot of K-factors for
use in calculating the acting
punching shear stress Vp, based
Brace i n t e r s e c t i o n angle, on geometry and statics of the
Notes:
1. Solid curves are for circular and square box sections having small (J. Shaded areas overall connection.
indicate higher values possible using Table 2.5.

The K-factors reflect the increased length, or bending section modulus, of the
intersection footprint, compared to a normal cross section of the branch tube. The equations in
the table were derived for an elliptical footprint (ref. 79), corresponding to small beta
(branch/chord diameter ratio); they are plotted in Figure 2.42. For larger beta, a more accurate
expression may be used (ref. 80). This is plotted, in an inverted format, in Figure 2.43. The
thinking was, if the British are going to all this trouble, it must be worthwhile. Unfortunately,
subsequent research indicated that the actual strength of tubular connections does not reflect the
trend of the K-factors, so we ended up having to take this "refinement" back out in the
expressions for allowable punching shear (ref. 81).
1.00

v = T f B f |

Fig. 2.43. Punching shear in

axially loaded CHS connection,
using K a according to BS 449.
66

More recent editions of both AWS and API Codes have reverted to the simpler form of
Equation 2.5, with no K-factors. In deriving equivalent strength criteria for both punching shear
and total load formats (e.g., ref. 82), the following conversions have been used, based on
idealized thin-wall, K-free cylinders:

Pu sin* = db t c Vpu (2.6)

2
Mu s i n * = d b t c Vpu (2.7)

where P u and Mu are ultimate load and moment, respectively, and d^ is branch diameter (other
terms are as defined at Equation 2.5). For practical tubes, the effect of the branch members
having a finite wall thickness causes a small discrepancy between the two formats (ref. 83). In
translating the ASCE committee report (ref. 82) into Code provisions (ref. 43), API fixed this by
adjusting the coefficients slightly (3% for axial load and 9% for bending, average values for a
representative selection of bracing sizes), in addition to an unrelated change in safety factor.
Alternatively, we could use more accurate punching shear equations as follows:

for axial load... V = r f sin* (2.8)

a

for bending. V = r f sin* 1 - - 2 (2.9)

b

This permits the formulation of design equations in both punching shear and total load formats
having a large degree of commonality, as well as numerical equivalence.
Most of the foregoing has focused on circular hollow sections (CHS). Several different
punching shear formats were discussed, some now mostly of historical interest. Empirical
strength criteria (allowable punching shear and ultimate total load capacities) for CHS are
developed in the next chapter. As with any empirical design criteria, the corresponding design
calculations must maintain a consistent format with the way rules were calibrated.
The subject of square and rectangular sections is deferred to Chapter 6. Although
historical AWS criteria have been in terms of punching shear, and the expressions in Table 2.5
are theoretically correct for the geometry and statics of square tubes, more recent international
criteria (refs. 6 and 9) are in total load format, and use effective width concepts in a way which
makes it very difficult, if not impossible, to state equivalent criteria in both formats.

2.5 STRESSES IN WELDS

Our discussion of stresses in welds in tubular structures will focus separately on (a)
welds in structural members, and (b) welds in structural connections.

2.5.1 Welds in Tubular Members

Stresses in welds in tubular members, away from connections, can be readily calculated,
using widely understood principles of elementary engineering mechanics. These will not be
repeated here (see Chapter 1). Such stresses include axial stress, beam bending stresses (about
two axes), beam shears (two axes), and torsional shear. In offshore structures, hoop stress due to
differential hydrostatic pressures would also be added to the list. Local stresses, shell bending
and punching shear, as occur in connections, are considered later.
 67

Table 10.4.1 of the AWS Code describes the rules for stresses in tubular members.
These provisions are similar to those in AWS Code Section 8, Statically Loaded Structures
(a.k.a. Buildings), and the AISC Specification (ref. 84). The principal addition is some
descriptive words on typical tubular applications of the various kinds of stress. The rules reflect
a number of American design practices and conventions, as described below. They have
remained substantially unchanged since 1970 (ref. 3).
While various material failure theories give the shear flow stress as 50% to 58% of the
tensile yield stress, the allowable stress design value for shear in base metal (0.4 Fy) is given as
67% of the corresponding tensile allowable (0.6 Fy). This dates back to the first edition of the
AISC Specification in 1923. The AISC Commentary justifies the apparent reduction in safety
factor by "the minor consequences of shear yielding". This anomaly has been largely eliminated
for beam shear in AISC-LRFD. More appropriate for stresses adjacent to welds, may be an
implicit reliance on strain hardening, especially if the area of high stress is localized.
This reliance on strain hardening is more explicit in the case of fillet welds, where
allowable stresses are indexed to the tensile strength, rather than yield. Yielding in welds is of
less consequence than in the gross section of the member, as the size of the yielded zone is close
to zero. The allowable weld stress is taken as 30% of the ultimate tensile strength. This is based
upon physical tests, including shears both parallel to and transverse to the weld axis (ref. 85),
where the ultimate strength of fillet welds was found to be 66% to 80% of the uniaxial tensile
strength (the larger number for 70-ksi and lower strength weld metals). Perhaps coincidentally,
these weld ultimate strengths correspond to 50-60% of the tensile rupture stress at 25% reduction
of area.
Stresses in fillet welds are treated the same as shear regardless of the actual
direction of loading. Stresses on the effective throat of fillet welds are calculated as P/A (total
load/least area) or Q / T w (line load/effective throat). Stresses on immediately adjoining base
metal are calculated essentially the same way, ignoring for the time being local moments needed
to prevent "log-rolling" of the weld; presumably, local hardening of the heat affected zone can
accommodate what we do not calculate, at least in lower strength conventional steels. Base
metal further removed from the weld must be checked for the eccentric loads involved except
for tubular members, in which the eccentricities are balanced out when the member as a whole is
considered, and localized shell bending beyond yield is not objectionable (see 2.3.2(i) and 2.4.1
herein).
Butt splices with complete joint penetration groove welds and matching weld metal are
governed by the capacity of the adjoining base metal. The requirement for matching weld metal
is waived for compression joints, where the weld is confined by adjacent stronger material and
localized yielding is of no consequence. For joints designed to bear in compression, the weld
metal is only needed to resist incidental shear loads.
Longitudinal seams in tubular structural members just "go along for the ride" for the
principal axial and bending stresses, and do not have to be checked for these stresses if they are
parallel to the weld axis. They usually have rather minor load transfer demands, in shear and
hoop stress, for which design provision is required. Theoretically, rather small under-matched
welds would be possible, but full penetration welds of matching strength are almost universally
used in manufacturing tubular products. In structural connection areas, fully matching complete
penetration seams are required for high local shell bending and punching shear stresses not
only in the main member joint-can, but also in the attached branch member brace-ends, where
shear-lag stresses accompany uneven load transfer across the intersecting weld.
68

2.5.2 Welds in Tubular Connections

Stresses in welds rarely govern the ultimate strength of tubular connections, but they
often need to be checked anyway, for completeness. The other failure modes which DO govern
the strength of connections are described in Section 10.5 of the code (see 2.2 herein); and fatigue
is discussed in Section 10.7 of the Code (and Chapter 4 herein).
Weld design checks are not required for complete joint penetration groove welds using
weld metal of matching strength. Under-matched weld metal may be used to mitigate lamellar
tearing problems, and partial penetration or fillet welds may be used for economy (or to ease the
rigors of welder qualification). These latter situations require that the weld be checked for design
loads and/or for unzipping (uneven distribution of load).
There are several ways to address the unzipping problem. The AWS pre-qualified
partial joint penetration groove welds for -, Y-, and K-connections(Code Figure 10.13.2), and
the IIW minimum fillet welds both provide an effective throat equal to the branch member
thickness. When matching weld metal is used on steels up to 50-ksi yield (i.e., E70XX), the
breaking strength of such welds exceeds the yield strength of the adjoining branch member. For
higher strength levels, a LOT weld (at 0.66 F g ^ ^ ) o n l y provides 74% to 90% of the matching
branch member yield strength, and IIW increases the required weld size (effective throat) to
1.07T. The AWS pre-qualified fillet details (effective throat of 0.7T) only match the branch
member strength when E70XX electrodes are used with mild steel tubes.
For welds failing to pass the foregoing check, unzipping can also be prevented if chord
face plastic failure (or punching shear) occurs before the weld reaches its breaking strength. This
comparison should be made in terms of local line load capacities, rather than taking advantage of
K-factors based on geometry and statics of the overall connection. Even so, the adverse effects
of uneven distribution of load are not fully accounted for in the calculation, as the empirical
punching capacity is really an average value.
Fillet welds also require a check for design loads, even if they pass the unzipping check.
For circular sections, formulas are given in Table 2.5 herein for calculating the applied weld
stress; these are based on geometry and statics of the overall connection, ignoring uneven load
distribution. Note that "sin0" does not appear in the axial load term; the equation given in 10.8.3
of the 1975-1988 Code was in error, as the weld must transfer the whole load, not just the radial
component. The formulas given herein also account for differences in branch member mean
radius and the radius to the centroid of weld effective throat, as encouraged by 10.8.1 and 10.8.2
of the Code; this correction can be particularly helpful for large fillet welds on small tubes.
The resulting stresses are compared with the allowable stress of 30% of the weld metal
tensile strength. Even though this allowable takes a huge safety factor (2.67) on breaking
strength, safety norms can still be compromised by uneven load distribution. For example, if the
load transfer across the weld has 50% efficiency, with peak line load of twice the average, the
safety factor is locally reduced to 1.33, versus 1.8 used for sizing the connection. When the one-
third increase is used, the safety margin is locally used up, and the weld could begin to fail
progressively.
Explicit guidelines for accounting for uneven load distribution in connections of circular
tubes are not well developed. The example just cited comes from FRAMETI, a shell theory
solution in which load transfer across the weld is treated explicitly. Defining load transfer
efficiency as nominal/peak line load, data appendices to the parameter study of reference 26 have
been used to construct Figures 2.44 and 2.45 for T-connections and K-connections, respectively.
For T-connections, load transfer peaks at the saddle position. Improved load transfer
efficiency is seen for smaller gamma (tending towards a solid chord) and small beta, with worse
distribution for connections having very high gamma ratio and high tau (branch/chord thickness
 69

Fig. 2.44. FRAMETI load transfer Fig. 2.45. FRAMETI load transfer efficiency

efficiency across the weld across the weld of a 60 -60 -60
of a 90 T-connection in CHS. K-connection.

ratio, i.e., a relatively stiffer branch on a very flexible chord. Evidence on the effect of large
values of diameter ratio, beta, is mixed. Based on experimental evidence and thin shell finite
element results, we might expect poor load distribution for a beta of unity, such that the most
effective load transfer is at the saddle, where loads are delivered tangentially to the chord; but
FRAMET1 does not show this.
K-connections often have better load distribution than T-or cross-connections, with load
transfer peaking in the gap region. Overlap K-connections have different mechanism of load
transfer, beyond just radial line loads on the chord, and cannot be analyzed using FRAMETI
shell theory.
Thus, it is difficult to formulate a simple but general rule for unevenness factor in
circular sections at this point in time. IIW rules for box connection explicitly account for uneven
distribution of load, through the effective width concept, as described in Chapter 5.

REFERENCES

1 Carter, R. M., Marshall, P. W., et al, Materials Problems in Offshore Structures, Proc. Offshore
Technology Conference, OTC 1043, May 1969.
2 Minutes of the AWS s/c on Welded Tubular Structures, October 9,1969, New Orleans.
3 Marshall, P. W., et al, Report of team "K", Design Stresses, AWS s/c on Welded Tubular
Structures, February 6,1970.
4 Report of Subcommittee 10, Welded Tubular Structures, minutes of the AWS Structural Welding
Committee, December 14-15,1971, Pittsburgh.
5 AWS Structural Welding Code, AWS D 1.1-72, first edition, American Welding Society, 1972.
6 Packer, J. ., et al, Canadian Implementation of CIDECT Monograph No. 6, University of
Toronto, July 1984.
7 Marshall, P. W. and Toprac, . ., Basis for Tubular Joint Design Codes, ASCE Preprint 2008,
San Francisco, April 1973.
8 Marshall, P. W., proposed revision to Dl.1-72, November 13, 1973 (AWS committee
correspondence); also presented at the Structural Welding Code Seminar, New York, October
1973.
9 Wardenier, J., Modified Eurocode 3 Design Recommendations for Hollow Section Lattice Girder
Joints, IIW Doc. SC-XV-E-87-120; also in Bjorhoyde, R., et al, eds., Connections on Steel
Structures, proceedings of the Paris workshop, Elsevier, May 1987.
70

10 Roark, R. J., Formulas for Stress and Strain, McGraw-Hill, New York, 1954.
11 Marshall, P. W., et al, Limit State Design of Tubular Connections, Proc. ASCE Conference
Methods of Structural Analysis, Madison, WI, 1976 (also BOSS-76, Trondheim).
12 Davies, G., Packer, J. ., et al, The Behavior of Full Width RHS Cross Joints, Welding of
Tubular Structures, Proc. 2nd International Conference, IlW/Pergamon, Boston, July 1984.
13 IIW Committee XV, Design rules for arc-welded connections in steel submitted to static loads,
IIW Doc. XV-358-74 (revised), International Institute of Welding, 1974.
14 Wardenier, J., Design and calculation of predominantly statically loaded joints between square
and rectangular hollow sections, Technical Information Bulletin #7 (in English), Van Leeuwen,
Holland, 1988.
15 Packer, J. A. and Frater, S. G.,Performance of Fillet Weldments in Hollow Section Connections,
Proc. International Conference, Safety Criteria in Design of Tubular Structures, IIW/AU, Tokyo,
July 1986.
16 Kurobane, Y., et al, Local Buckling in Tubular K-Joints, IIW Doc. XV-E-85-088; also summary
paper ALT annual conference, September 1983.
17 Marshall, P. W., Designing Tubular Connections with AWS Dl.1, agenda for the AWS Structural
Welding Committee, San Francisco, October 1987; also Welding Journal, March 1989.
18 Marshall, P. W., Problems in Long-Life Fatigue Assessment for Fixed Offshore Structures, ASCE
Preprint 2638, San Diego, April 1976.
19 Noordhoek, C. and de Back, J., Steel in Marine Structures (SIMS '87), Proc. 3rd International
ECSC Conference, Delft, June 1987, Elsevier.
20 Hicks, J. G., A Study of Material and Structural Problems on Offshore Installations, research
report E/55/74, The Welding Institute, Cambridge, January 1974.
21 Toprac, . ., et al, Welded Tubular Connections: An Investigation of Stresses in T-Joints,
Welding Journal, January 1966.
22 Bijlaard, R. P., Stresses from Radial Loads in Cylindrical Pressure Vessels, Welding Research
Supplement, Welding Journal, December 1954 (also see December 1955).
23 Dundrova, V., Stresses at the Intersection of Tubes - Cross and Tee Joints, University of Texas
SFRL Technical Report P-550-5 (1966).
24 Dundrova, V., Stress and Strain Investigation of General Joints in Tubular Structures, Report TL-
A-03-67, University of Texas, July 1967 (proprietary).
25 Dundrova, D., Stress Concentration in Joints Subjected to Axial Loads, Bending Moments, and
Shears, Report TL-A-01-68, University of Texas, March 1968 (proprietary).
26 Caulkins, D. W., Parameter Study for FRAMETI Elastic Stress in Tubular Joints, Shell Oil
Company, CDG Report 15, September 1968.
27 Greste, O. and Clough, R. W., Finite Element Analysis of Tubular Joints: Report on a Feasibility
Study, University of California, Structures & Materials Research Report No. 67-7, April 1967.
28 Greste, O., A Computer Program for the Analysis of Tubular K-Joints, University of California,
Structures and Materials Research Report No. 69-19, November 1969.
29 Visser, W., On the Structural Design of Tubular Joints, Proc. Offshore Technology Conference,
OTC 2117, May 1974.
30 Bouwkamp, J. G., et al, Effect of Joint Flexibility on the Response of Offshore Structures, Proc.
Offshore Technology Conference, OTC 3901, May 1980.
31 Dovey, . H., Extension of 3-D Analysis to Shell Structures Using Finite Element Idealization,
University of California, Report UC-SESM-74-2, January 1974.
32 Reimer, R. B., et al, Improved Finite Elements for Analysis of Welded Tubular Joints, Proc.
Offshore Technology Conference, OTC 2642, May 1976.
33 PMB Systems Engineering, PMBSHELL User Input Specification & Programmers Manual (3
vols.), January 1981.
34 Marshall, P. W., private communication to PMB, October 20,1975, and November 20,1979.
35 Morgan, E. F., Solid Finite Elements and the Re-Zone Technique Applied to the Analysis of
Grouted and Ungrouted Tubular Joint Intersections, Southwest Research Institute, 1978.
36 Yoshida, K., et al, Behavior Analysis and Crack Initiation Prediction of Tubular T-Connections,
Proc. Offshore Technology Conference, OTC 2854, May 1977.
 71

37 Marshall, P. W., Some Design Considerations for Welded Tubular Structural Joints, Shell Oil
Company, preliminary November 1964, revised April 1967; Part "A" of State-of-the-Art Tubular
Joint Design Guide, Shell Internationale Petroleum Mij., EP 42245, October 1970.
38 Marshall, P. W., West Delta Block 30 Quarters Platform, Shell Oil Company, drawing series
WD-30-P-D-64-1 thru 6, and design binder, May 1964.
39 Bouwkamp, J. G., Considerations in the Design of Large Size Welded Tubular Truss Joints, AISC
Engineering Journal, July 1965.
40 Marshall, P. W., Design of Simple Tubular Joints, Shell Oil Company, CDG Report 12, January
1967.
41 Akiyama, H., Contribution to the Monograph - TC43 - Connections, Joint Committee on Tall
Buildings, 1974.
42 Wardenier, J., Plate to I-Beam Connections, Doc. IIW XV 88-83, Delft, 1983; also see Welding
in the World, V.28, No. 3/4,1985.
43 API Recommended Practice for Planning, Designing, and Constructing Fixed Offshore Platform,
API RP 2A, 17th edition, American Petroleum Institute, Dallas, 1987.
44 Marshall, P. W., ASCE Team 2 - Punching Shear for Square and Rectangular Tubes, internal
working document, ASCE Committee on Tubular Structures, December 20,1970.
45 Jubb, J. . M. and Redwood, R. G., Design of Joints to Box Sections, Institute of Structural
Engineering Conference on Industrialized Building, May 1966.
46 Graff, W. J., Welded Tubular Connections of Rectangular and Circular Hollow Sections, ASCE
Texas Section, El Paso, October 1970.
47 Graff, W. J., et al, Punching Shear Characteristics of RHS Joints, ASCE preprint No. 1963, April
1973, San Francisco.
48 Mouty, J., Calcul des charges ultimes des assemblages soudes de profils creux carres et
rectangulares, Construction Metallique, June 1976.
49 deKoning, C. . M. and Wardenier, J., Tests on Welded Joints in Complete Girders Made of
Square Hollow Sections, Stevin report 6-79-4, Delft, 1979.
50 Clough, R. W., The Finite Element Method in Structural Mechanics, in O. C. Zienkiewicz et al
(1st edition), Stress Analysis, John Wiley & Sons, London, 1965.
51 Marcal, P., et al, Finite Element Analysis of Non-Linear Structures, Journal of the Structural
Division, ASCE, Vol. 94, No. ST9, September 1968.
52 Hoadley, P. W., Ultimate Strength of Tubular Joints Subjected to Combined Loads, PhD
Dissertation, University of Texas at Austin, 1984.
53 van der Valk, C. A. C , Factors Controlling the Static Strength of Tubular T-Joints, Proc. 5th
International Conference on Behavior of Offshore Structures, BOSS-88, Trondheim, June 1988.
54 Beale, L. A. and Toprac, . ., Analysis of In-Plane -, Y-, and K-Welded Tubular Connections,
WRC Bulletin 125,1968.
55 Bouwkamp, J. G., Tubular Joints under Static and Alternating Loads, University of California
Structures and Material Research Dept., No. 66-15, June 1966.
56 Bouwkamp, J. G. and Steven, R. M., Tubular Joints Under Alternating Loads (Phase II),
University of California Structures and Material Research Dept., No. 67-29, November 1967.
57 Grigory, S. C , A Study to Develop a Design Procedure for Analysis of Plastic Fatigue Life of
Tubular Joints in Offshore Structures, Southwest Research Institute final project report 03-1882,
May 1969 (proprietary).
58 Holliday, G. H., Welded Tubular Joint Research, quarterly progress reports, August 1966 -
September 1968, Shell Development Company (proprietary).
59 Marshall, P. W., Middle Ground Shoal Platform "C" Structural Joint Model Test, progress reports
January 1967 and November 1967, Shell Oil Company, CDG, New Orleans (proprietary).
60 Marshall, P. W., Recommendations for Model Testing Specific Tubular Joint Designs, 4/24/67;
Part "C" of State-of-the-Art Tubular Joint Design Guide, Shell Internationale Petroleum Mij., EP
42245, October 1970.
61 Snedden, N. W., Background to Proposed New Fatigue Design Rules for Steel Welded Joints in
Offshore Structures, report of the Department of Energy Guidance Notes Revisions Drafting
Panel, May 1981.
72

62 Fatigue in Offshore Steel Research, proceedings of the conference held at Westminster, London,
February 24-25,1981, Institution of Civil Engineers.
63 Holliday, G. H. and Graff, W. J., Three-Dimensional Photo-Elastic Analysis of Welded T-
Connections, Proc. Offshore Technology Conference, OTC 1441, April 1971.
64 European Offshore Steel Research, proceedings of the select seminar, Cambridge, November
1978.
65 Steel in Marine Structures, Proc. 2nd International ECSC Conference; Institut de Recherches de
la Siderurgie Francaise, Paris, October 1981.
66 Marshall, P. W., State of the Art in the USA, Steel in Marine Structures - SIMS '87, Proc. 3rd
International ECSC Conference, Delft University/Elsevier, June 1987.
67 Wordsworth, A. C. and Smedley, G. P., Stress Concentrations at Unstiffened Tubular Joints,
European Offshore Steel Research, Cambridge, November 1978.
68 Pickett, A. G., et al, Full-Size Pressure Vessel Testing and its Application to Design, ASME
Paper 63-WA-293,1963.
69 Marshall, P. W., Risk Factors for Offshore Structures, CDG Report 5, Shell Oil Company, New
Orleans, October 1965.
70 Iida, K., Application of Hot Spot Strain Concept to Fatigue Life Prediction, University of Tokyo,
Dept. of Naval Architecture, NAVT Report No. 9036, IIW Doc. -1103-83, June 1983.
71 Marshall, P. W., General Considerations for Tubular Joint Design in Offshore Structures, Shell-
Arco-Mobil Gulf of Alaska Criteria & Platform Study, ODC Report No. 49, August 1973.
72 Appendix B, Fatigue, AISC Specifications for Design, Fabrication, and Erection of Structural
Steel for Buildings, American Institute of Steel Construction, New York, 1969 (also 1980).
73 Rules for Construction of Nuclear Power Plant Components, Division 1 of Section III (Pressure
Vessels) of the ASME Boiler and Pressure Vessel Code, 1973.
74 Betero, V.V. and Popov, E.P., Effect of Large Alternating Strains on Steel Beams, Journal of
Structural Engineering, Proc. ASCE, Vol. 91, No. ST1, February 1965.
75 Guidance on some methods for the derivation of acceptance levels for defects in fusion welded
joints, PD 6493, British Standards Institute, 1980.
76 Marshall, P.W., Design of Internally, Stiffened Joints, Proc. International Conference Safety
Criteria in Design of Tubular Structures, Y. Kurobane et al ed., IIW/AIJ, Tokyo, July 1986.
77 Johnston, L. P., The Welded Tubular Joint Problem in Offshore Oil Structures, First University of
Texas Conference on Drilling & Rock Mechanics, Shell EPR Publication 326, January 1963.
78 Lee, G. C , in minutes of API mini-group on Welded Tubular Joints, October 2,1969.
79 Blodgett, O., AWS Dl s/c 10 communication, January 1975.
80 Determination of the Length of the Curve of Intersection of a Tube with Another Tube or with a
Flat Plate, Appendix C, BS 449:1959.
81 Marshall, P. W., A Review of American Criteria for Tubular Structures - and Proposed Revisions,
IIW Doc. XV-405-77, International Institute of Welding, Copenhagen, 1977.
82 Graff, W. J., et al, Review of Design Considerations for Tubular Joints, progress report of the
committee on tubular structures, ASCE Preprint 81-043, New York, May 1981.
83 Roussel, H. J., API mini-group communication, September 1982.
84 Section 1.5.3, Allowable Stresses-Welds, AISC Specification for the Design, Fabrication, and
Erection of Structural Steel for Buildings, November 1978.
85 Higgins, T. R. and Preece, F. R., Proposed Working Stresses for Fillet Welds in Building
Construction, Welding Journal, Research Supplement, October 1968.
Chapter 3

STATIC STRENGTH OF CIRCULAR SECTION JOINTS

3.1 EARLY W O R K ON -, Y-, AND K-CONNECTIONS

Although the first modern offshore platform was built in the Gulf of Mexico in 1947,
and tubular steel space frames (or templates) became the dominant form for permanent platforms
in the mid-1950s, the supporting research came a bit later (ref. 1). In the spirit of the old-time
wildcatter, adventuresome engineers just went out there and did it, with only some rather
unsophisticated approximations backing up their design calculations especially in the areas of
wave forces, pile-soil interaction, and tubular connections. Sometimes the designs worked and
sometimes they did not; fortunately a policy of de-manning the platforms and securing the wells
in advance of hurricanes prevented loss of life and environmental damage during this trial-and-
error process. Eventually, workable design practices evolved, e.g. punching shear, as discussed
earlier. Some major projects included experimental stress analysis or "proof load" strength tests
of representative tubular connections as an adjunct to the engineering effort (refs. 2 and 3).
Isolated research efforts were also undertaken in the U. S., some of which are outlined below.
However, it remained for Johnston in 1963 to review the available information on
tubular connections (ref. 4) and organize a coordinated joint-industry attack on the problem. The
ensuing Texas effort, and a parallel effort in California, resulted in our first comprehensive
understanding of the problem and the first American design Codes for tubular structures. Much
of the emphasis was on theoretical methods of analysis, despite their difficulty and shortcomings,
in order to gain a level of understanding which would extend beyond the specific connection
geometries tested. A large body of prior work done in Japan was also brought to the attention of
engineers in the U. S. (ref.5).

3.1.1 Non-Dimensionalization
A certain convention for non-dimensionalizing the parameters was already established,
following earlier work with cylindrical shells. This has proved to be indispensable in
generalizing results from one situation to another, and in scaling up the results of scale model
tests. Where results are not presented in non-dimensional or in dimensionally consistent form,
(e.g., expressing total load in terms of yield stress times thickness without specifying the
intended units), great potential is created for error, confusion, and misinterpretation.
Alpha is traditionally taken as the chord length, divided by radius or thickness. This is
related to two effects: beam bending and ovalizing in the chord. These are important in the test
set-up for model tests, e.g. the span and distance from end diaphragms in T-connections, but
difficult to relate to design situations. Beam bending is usually accounted for in member design,
separately from local stresses in the structural connection details. Diaphragms far removed from
the connection are not very effective and rarely a feature of practical designs; thus ovalizing is
more related to load pattern. In the first parts of this chapter, the traditional definition of alpha is
retained; however, it will later be redefined to represent ovalizing as a function of load pattern.
Beta is the ratio of branch to chord radii or diameters. Its effect is similar to the
difference between concentrated midspan loads and distributed loads in the design of beams or
arches, so it is a very important parameter, again related to ovalizing in the chord.
Gamma is chord radius divided by thickness. Large gamma values denote more
flexible, weaker chords. Gamma appears in the equations of compatibility for cylindrical shells
(e.g., see Figure. 2.12). Although third in order, it is probably the most important parameter of
the chord.
74

Tau is branch thickness, divided by chord thickness. It is so important to understanding

the efficiency of tubular connections, that it is included in the definition of punching shear,
before we even get to the chord parameters.
Like beta, theta, eta, and zeta define the surface topology of the connection. These were
previously defined in Figure 2.2.

3.1.2 Theoretical Studies

The two areas of theoretical stress analysis which most influenced the early
development of design criteria for tubular connections can both trace their roots back through the
work of Roark (refs. 6 and 7). They are:

(1) radial loads on cylindrical shells, and

(2) closed ring solutions.

(i) Shells. Bijlaard (refs. 8 and 9) developed analytical solutions for the stresses in a
cylindrical shell under the influence of "patch" loads, radial pressure applied uniformly over a
discrete square or rectangular region. Combinations of inward and outward pressure patches were
used to approximate bending moments. These equations have been reduced to parametric charts
(e.g., ref. 9), and have been widely used to approximate the effect of loads and moments on
attachments to pressure vessels, e.g., nozzles.
Johnston compared Bijlaard's solution with some early experimental stress analysis of
T-connections, as shown in Figure 3.1. This favorable comparison encouraged the development
of a computer program to generate a number of parametric charts for cases of interest in the
design of tubular T-connections. Since the uniform pressure patch loading is a crude
approximation of the line load transfer in a tubular connection, the need for some kind of
adjustment was anticipated. As part of the Texas joint-industry project (ref. 11), the Bijlaard
solution was systematically compared to the available experimental stress analyses, and a second
series of charts was developed, correction terms to make theory match the experiments.

SPECIMEN C-l

THEORETICAL YIELD STRESS

" D MEMBER -
33.000 pit

, INCHES

(a) (b)
Fig. 3.1. Comparison of Bijlaard theory with experimental stress analysis of T-connections (ref. 4).
(a) Chord shell longitudinal stresses along the crown; note that Bijlaard's assumed patch
loading causes stresses to peak at the member centerline, rather than at the weld line, (b)
Load-deflection behavior; note that yielding is indicated at a small fraction of the ultimate
load.
 75

Unfortunately, the form of the regression equations was not based on physical considerations,
and the empirical correction terms were often as large as the basic Bijlaard part. Also, the charts
gave what seemed at the time to be incredibly high stresses, well above yield for connections
w h i c h were k n o w n from practical e x p e r i e n c e to have w o r k e d adequately in service.
Understandably, these results led to a certain lack of confidence in the method as a basis for
improved design procedures.
An alternative method for calculating shell stresses in the chord wall of a tubular
connection was developed at the University of California (ref. 12). Prof. A. C. Scordelis
developed a solution for cylindrical shells under localized line loads applied along a segment of a
line running parallel to the cylinder axis. By superposition, results from a number of these
segments were combined to represent the load transfer across the weld in a tubular connection,
with a computer program keeping track of the arithmetic (using the Duhamel integral, analogous
to the way a time varying load is treated as a series of impulsive loads in dynamics). Figure 3.2
shows results for a T-connection with beta of 0.5, in which the branch member is assumed to be
rigid (i.e., all the line segments are displaced the same amount). The uneven transfer of load
across the weld, and sharp peaking of chord shell stresses at the locus of the intersection weld,
can be clearly seen. These features represented a substantial improvement in realism over the
uniform pressure patch loading of Bijlaard.

p=1000

Fig. 3.2. Scordelis shell theory solution for a T-connection (ref. 12). (a) Connection geometry, (b)
Load transfer across the weld, represented by a series of line segments, (c) Internal shell
forces and moments: non-dimensionalized longitudinal "x" and circumferential "phi"
direction, unit membrane tension (N, kips/in) and moment (M, kip-in/in).

Concurrently, Dundrova (refs. 13 to 15) developed a more general solution, which

considered the actual membrane stiffness of the branch member rather than having to assume it
to be rigid. She extended the method to cover multi-brace and multi-planar connections, and
moments as well as axial loads, and automated the solution in computer programs GENTUBE
and FRAMETI. This approach was described earlier in Section 2.3.l(i). The reader may wish to
76

compare the results of Figure 3.2 with those of Figure 3.4, obtained with Dundrova's program.
The comparison is fairly close, suggesting a degree of reasonableness for Scordelis' assumption
that the branch is essentially rigid in comparison with radial behavior of the chord.
Unfortunately, these promising shell theory methods were so closely held by their
creators (and sponsors), that they were supplanted by the finite element method before they could
be widely disseminated for direct use in design. However, they were both extremely useful in
giving a more comprehensive view of chord shell stresses than could reasonably be obtained
from strain gage measurements, and for the parametric studies in which they were used.
Holliday (ref. 16) first used GENTUBE to study stress concentrations and elastic strength of T-
connections, covering the range of proportions then in use by designers. Stress concentration
factors as high as 30 were found, along with punching shear efficiencies (punching shear at first
chord yield, divided by chord material shear strength) of only 10% to 20%.
Reference 17 contains the results of a more systematic FRAMETI parameter study.
Results for an axially loaded tee-connection are shown in Figure 3.3. The geometry is
reasonably typical, with beta (branch/chord diameter ratio) and tau (branch/chord thickness ratio)
both being about 0.5, while gamma (chord radius/thickness) is 20. The peak shell stress in the
chord is 160-ksi, or 7.3 times the nominal stress of 22-ksi in the branch member. This is often
referred to as the "hot spot stress" (a more rigorous definition will be given later) and its location
is generally on the chord surface, where the line load is delivered by the intersection weld. Its
particular location on the intersection, at the "saddle" position, is typical of tee and cross
connections. Circumferential stresses can be seen to exhibit a pattern similar to the ring
ovalizing case of Figure 2.10(b), being largest around the middle of the branch member footprint,
and decaying as one moves along the chord axis, away from the attached brace. Longitudinal
stresses also peak at the saddle position, as does the load transfer across the weld. Indeed, we
have a triaxial stress condition here, with the following stress components:

circumf: 160-ksi, or 7.3 times nominal

longit: 102-ksi, or 4.7 times nominal
radial: 44-ksi, or 2.0 times nominal

Figure 3.4 shows comparable results for a K-connection having the same member sizes.
We notice several differences from the previous T-connection. The peak shell stress is much
lower, 55-ksi, or only 2.5 times nominal; and the load transfer across the weld is more uniform.
Deflections (not shown) and circumferential stresses indicate that there is less ovalizing of the
chord, with the inward load from one branch offsetting the outward load from the other. Radial
line loads are also reduced by the branches being at 45-degrees rather than perpendicular, and by
the slightly larger footprint.
We saw in an earlier comparison with model test data (Figure 2.17) that the magnitudes
of the calculated peak hot spot stresses were confirmed by the experimental stress analysis, in
terms of measured strains where theoretical stresses are above yield. However, the trigonometric
series used in FRAMETI had difficulty in reproducing the more complex stress patterns of multi-
member connections, with the longitudinal stress failing to peak as sharply as it should have
when crossing the brace footprint.
Figure 3.5 shows the results of a parameter study on the elastic stress efficiency of
axially-loaded tee connections (ref. 17). Connection efficiency may be defined as the nominal
branch stress at which first yield occurs, divided by chord yield stress. For perpendicular branch
members and tau of unity, this is the same as punching shear efficiency. Efficiency is presented
as a function of the chord radius-to-thickness ratio, gamma, and of the brace-to-chord diameter
ratio, beta. For example, for gamma of 20 and beta of 0.5, the punching shear efficiency is only
7%, which is consistent with Figure 3.3. Efficiency passes through a broad valley for beta in
 77

(a)

(b)

(c)

Fig. 33. FRAMETI solution for T-connection. (a) Circumferential surface stress in chord, (b)
Longitudinal surface stress in chord, (c) Load transfer across intersection weld.
78

(a)

(b)

(c)

Fig. 3.4. F R A M E T I s h e l l t h e o r y r e s u l t s f o r a K - c o n n e c t i o n . ( a ) C i r c u m f e r e n t i a l s u r f a c e stress i n t h e

c h o r d , ( b ) L o n g i t u d i n a l s u r f a c e stress i n t h e c h o r d , ( c ) L o a d t r a n s f e r a c r o s s t h e w e l d .
 79

(a)

(b)

Fig. 3.5. FRAMETI parameter study - connection efficiency of axially loaded tee connection, (a)
Effect of 7 and . (b) Effect of 7 ring.
80

range of 0.3 to 0.7, which covers many practical applications (especially for the orthogonal truss
frames of offshore jackets). With this range, efficiency is more or less independent of beta, but
varies inversely as the 0.5 to 1.0 power of gamma. Conservatively, the 0.7 power could be used
(dashed lines, Figure 3.5(b)) with the overall capacity of the connection being proportional to
brace perimeter and the 1.7 power of chord thickness. However, one should be careful not to
generalize the behavior of tubular connections too much based on the behavior of only T-
connections.
(ii) Closed rings. Figure 3.6 shows the closed ring solution (refs. 7 and 18) for a
symmetrical pair of loads, whose reaction is resisted by beam shear in the cylinder to which the
ring is attached. In the plots, moment peaks at point of load application; also note the strong
tendency towards higher moments for smaller beta (more closely spaced loads). This solution is

(a)

OF CURVATURE ON
FLANGE STRESS.

Fig. 3.6. Closed ring solution. Fig. 3.7. Ring stiffened tubular
(a) Statement of the connection, (a) General view,
problem, (b) Circumferential (b) Limit analysis mechanism
bending moments. for deriving flange effective
width.

extremely useful for the design of ring-stiffened connections; superposition can be used to build
up a large number of load cases, using the tabulated results of Table 3.1. Although this is
identical with the closed-form shell theory of Figure 2.10(b), and the stress pattern is
qualitatively similar to circumferential stresses obtained from shell theory solutions for simple
tubular connections, a problem arises in selecting the effective width of shell to be mobilized in
resisting loads of finite length.
Bryant (ref. 19) first applied this solution to the analysis of ring-stiffened connections of
the type shown in Figure 3.7. The effective width of shell to be used as a flange on the ring was

TABLE 3.1 TABULATED SOLUTION FOR ROARK CASE 25 (overleaf)

Circular ring loaded along two chords by parallel loads W/2, with the load removed by tangential
shear. Total load = W. Radius = R. Angles are measured from plane of symmetry, on the
loaded size; theta is the load position and X is the position for which results are given.
Circumferential moment, thrust, and shear in the ring are , T, and V, respectively.
 81

THETA X FM FT FV THETA X FM FT FV
DEC. DEC M=(FM)WR T=(FT)V V=(FV)W DEC DEC M=(FM)WR T=(FT)W V=(FV)W
0 0 .2386 -.2387 .0000 50 0 -.0014 -.1453 -.0000
0 10 .1554 -.3171 -.4512 50 10 .0007 -.1383 .0249
0 20 .0818 -.3763 -.3904 50 20 .0071 -.1175 .0474
0 30 .0196 -.4150 -.3210 SO 30 .0170 -.0842 .0652
0 40 -.0299 -.4328 -.2467 50 40 .0295 -.0399 .0762
0 50 -.0663 -.4300 -.1711 50 50 .0432 .0129 .0786
0 60 -.0897 -.4080 -.0977 50 60 .0064 -.3613 -.1786
0 70 -.1007 -.3687 -.0297 50 70 -.0192 -.3368 -.1174
0 80 -.1006 -.3150 .0301 SO 80 -.0348 -.2988 -.0618
0 90 -.0908 -.2500 .0795 50 90 -.0413 -.2500 -.0138
0 100 -.0735 -.1773 .1169 50 100 -.0401 -.1936 .0249
0 110 -.0507 -.1010 .1412 50 110 -.0331 -.1330 .0535
0 120 -.0249 -.0249 .1522 50 120 -.0221 -.0716 .0713
0 130 .0016 .0470 .1502 50 130 -.0088 -.0129 .0786
0 140 .0267 .1114 .1362 50 140 .0047 .0399 .0762
0 ISO .0486 .1650 .1119 SO 150 .0172 .0841 .0652
0 160 .0654 .2053 .0794 50 160 .0271 .1175 .0474
0 170 .0760 .2302 .0411 50 170 .0335 .1383 .0249
0 180 .0796 .2387 .0000 50 180 .0357 .1453 .0000
10 0 .1590 -.2339 .0000 60 0 -.0102 -.1193 -.0000
10 10 .1626 -.2255 .0403 60 10 -.0084 -.1127 .0204
10 20 .0887 -.3718 -.3920 60 20 -.0032 -.0931 .0385
10 30 .0262 -.4109 -.3234 60 30 .0048 -.0617 .0522
10 40 -.0238 -.4291 -.2498 60 40 .0146 -.0200 .0595
10 50 -.0608 -.4269 -.1748 60 50 .0251 .0296 .0587
10 60 -.0849 -.4056 -.1019 60 60 .0346 .0846 .0488
10 70 -.0967 -.3671 -.0342 60 70 .0047 -.3279 -.1418
10 80 -.0973 -.3141 .0254 60 80 -.0151 -.2942 -.0874
10 90 -.0884 -.2500 .0747 60 90 -.0260 -.2500 -.0397
10 100 -.0719 -.1782 .1122 60 100 -.0294 -.1981 -.0005
10 110 -.0500 -.1027 .1367 60 110 -.0268 -.1419 .0291
10 120 -.0249 -.0273 .1480 60 120 -.0198 -.0846 .0488
10 130 .0009 .0439 .1465 60 130 -.0103 -.0296 .0588
10 140 .0255 .1077 .1331 60 140 .0001 .0200 .0595
10 ISO .0468 .1609 .1095 60 150 .0099 .0617 .0522
10 160 .0633 .2008 .0777 60 160 .0180 .0931 .0386
10 170 .0736 .2255 .0403 60 170 .0232 .1127 .0204
10 180 .0772 .2339 .0000 60 180 .0250 .1193 .0000

20 0 .0956 -.2201 .0000 70 0 -.0126 -.0981 -.0000

20 10 .0990 -.2119 .0379 70 10 -.0111 -.0918 .0167
20 20 .1087 -.1878 .0730 70 20 -.0069 -.0732 .0313
20 30 .0452 -.3989 -.3303 70 30 -.0004 -.0433 .0416
20 40 -.0062 -.4185 -.2587 70 40 .0072 -.0038 .0459
20 50 -.0449 -.4181 -.1854 70 50 .0151 .0432 .0425
20 60 -.0710 -.3987 -.1138 70 60 .0216 .0952 .0305
20 70 -.0850 -.3624 -.0472 70 70 .0252 .1491 .0092
20 80 -.0879 -.3117 .0117 70 80 .0017 -.2906 -.1082
20 90 -.0814 -.2500 .0609 70 90 -.0128 -.2500 -.0609
20 100 -.0673 -.1806 .0986 70 100 -.0199 -.2018 -.0214
20 110 -.0477 -.1074 .1237 70 110 -.0208 -.1491 .0092
20 120 -.0248 -.0342 .1361 70 120 -.0172 -.0952 .0305
20 130 -.0009 .0350 .1359 70 130 -.0107 -.0432 .0425
20 140 .0219 .0971 .1243 70 140 -.0028 .0037 .0459
20 150 .0418 .1489 .1026 70 150 .0048 .0433 .0417
20 160 .0573 .1878 .0730 70 160 .0113 .0732 .0313
20 170 .0670 .2119 .0379 70 170 .0155 .0918 .0167
20 180 .0704 .2201 .0000 70 180 .0170 .0981 .0000
30 0 .0488 -.1989 .0000 80 0 -.0120 -.0843 -.0000
30 10 .0518 -.1910 .0342 80 10 -.0107 -.0782 .0143
30 20 .0606 -.1679 .0658 80 20 -.0071 -.0602 .0266
30 30 .0744 -.1306 .0920 80 30 -.0017 -.0314 .0347
30 40 .0209 -.4023 -.2723 80 40 .0046 .0067 .0370
30 50 -.0204 -.4045 -.2016 80 50 .0107 .0521 .0319
30 60 -.0495 -.3881 -.1322 80 60 .0153 .1021 .0185
30 70 -.0668 -.3551 -.0671 80 70 .0167 .1538 -.0037
30 80 -.0733 -.3081 -.0090 80 80 .0135 .2041 -.0350
30 90 -.0705 -.2500 .0397 80 90 -.0035 -.2500 -.0747
30 100 -.0600 -.1843 .0777 80 100 -.0129 -.2042 -.0350
30 no -.0440 -.1146 .1038 80 110 -.0162 -.1538 -.0037
30 120 -.0245 -.0448 .1177 80 120 -.0148 -.1021 .0185
30 130 -.0036 .0214 .1197 80 130 -.0102 -.0521 .0320
30 140 .0166 .0809 .1107 80 140 -.0041 -.0067 .0370
30 150 .0344 .1306 .0920 80 150 .0022 .0314 .0347
30 160 .0483 .1679 .0658 80 160 .0076 .0602 .0266
30 170 .0571 .1910 .0342 80 170 .0113 .0782 .0143
30 180 .0601 .1989 .0000 80 180 .0125 .0843 .0000
40 0 .0172 -.1729 .0000 90 0 -.0113 -.0795 -.0000
40 10 .0198 -.1655 .0297 90 10 -.0101 -.0735 .0135
40 20 .0274 -.1435 .0569 90 20 -.0067 -.0557 .0249
40 30 .0394 -.1081 .0790 90 30 -.0016 -.0272 .0323
40 40 .0546 -.0610 .0939 90 40 .0042 .0104 .0339
40 50 .0100 -.3878 -.2215 90 50 .0098 .0552 .0283
40 60 -.0227 -.3751 -.1547 90 60 .0136 .1045 .0144
40 70 -.0441 -.3462 -.0915 90 70 .0143 .1554 -.0082
40 80 -.0550 -.3036 -.0346 90 80 .0102 .2050 -.0397
40 90 -.0566 -.2500 .0138 90 90 -.0000 .2499 -.0795
40 100 -.0507 -.1888 .0521 90 100 -.0103 -.2050 -.0397
40 110 -.0390 -.1235 .0794 90 110 -.0143 -.1555 -.0082
40 120 -.0236 -.0578 .0953 90 120 -.0136 -.1045 .0144
40 130 -.0064 .0047 .0998 90 130 -.0098 -.0552 .0283
40 140 .0106 .0610 .0940 90 140 -.0042 -.0104 .0339
40 150 .0258 .1081 .0790 90 150 .0016 .0272 .0323
40 160 .0378 .1435 .0569 90 160 .0067 .0557 .0250
40 170 .0454 .1655 .0297 90 170 .0101 .0735 .0135
40 180 .0480 .1729 .0000 90 180 .0113 .0795 .0000
82

derived by analogy with the axisymmetric hoop loading of Figure 2.10(a), in which a ring of
width 1.56<J(RT) reproduces the load-point deflection and elastic peak hoop stress. This same
effective section was then used for combinations of thrust and moment from an arch solution or
closed ring solution, using allowable stress design from the AISC rules for beam-columns.
Bryant's method is sometimes referred to as "column analogy", from the type of arch solution he
used.
Limit analysis, using the mechanism shown in Figure 3.7(b), yields an effective width of
3J(RT), able to achieve gross yielding in the effective flange in the circumferential direction,
when strain hardening up to 1.5 times yield is allowed for local shell bending stresses in the
longitudinal direction.
An attempt was made to extend this same approach to loads of finite length on
unstiffened shells, using the form

B E FF = 77D + VRT (3.1)

where is the footprint length as measured along the chord axis. Based on a limited number of
tests, K=6 was selected. The resulting effective width is plotted in Figure 3.8, along with several
other proposed criteria.

Fig. 3.8. Comparison of several

criteria for effective width to be
used with ring bending solution.
 83

Marshall's criterion (1964) was obtained by matching the closed ring solution to elastic
stresses from Bijlaard's shell theory results (EPR 738), for gamma of 16, at the two beta values
indicated. A very conservative extrapolation to smaller beta values was taken. This resulted in
low initial estimates of chord line load capacity, as discussed in Chapter 2.
At the other extreme, we have Washio's proposed effective width of three chord
diameters (ref. 20) used for limit analysis of cross connections. Limit analysis of a closed ring
for opposed load pairs yields

B e ff 2
sin* = 1 T F (3.2)
u R 1 - yo

where P u is the total load, theta is the angle of inclined loads (as in Y-connections), B e f f is
effective width of ring, R is chord radius, beta is ratio of load spacing (or branch diameter) to
chord diameter, is chord thickness, and F y Q is chord yield stress. An effective width of two
chord diameters gives equivalent ultimate load capacity to that of box connections; i.e., Equation
3.2 reduces to the limit analysis expression derived in Figure 2.28. The ability of circular
connections to mobilize longer effective widths is taken as a reflection of favorable curved shell
effects and other sources of reserve strength.
A more recent analysis of this type is given by Kurobane, et al (ref. 21). By comparing
the empirical ultimate strength formula for cross connections with an empirical modification of
the above ring equation, they derive effective widths which peak out at 2.6 to 3.3 chord
diameters. The modified ring equation, based on non-linear frame analysis of rings, is given by

ine
S = ?eff . _ ^ 2 5 . -.201 2
T F 3 3) (

u R 1 - .917 yo

where gamma is chord radius/thickness ratio. The generality of results obtained by comparing
two empirical expressions may be debatable, but Kurobane's effective width is comparable to
that used by AWS for general collapse calculations Q$eff = 2.5 D) for beta in the practical
range of 0.4 to 0.8.
Circular shells show a surprising ability to mobilize an effective length much greater
than the loaded area, in resisting the ovalizing mode of failure which ring analysis treats.
Marshall's original assessment, based on local peak stresses from Bijlaard's elastic theory, now
seems to be somewhat over-conservative, particularly for assessment of ultimate strength;
nevertheless, such lower-bound solutions permitted designs of world record water-depth offshore
platform projects to proceed while the researchers were catching up.

3.1.3 Experimental Studies

A number of early experimental programs played a key role in the formulation of the
basic AWS punching shear criteria for designing tubular connections. They are described in this
section. In many cases, the data have been re-interpreted and presented in a different format than
that used by the original investigators, in order to build a common picture of the behavior of
these connections. All the early test specimens were of mild steel, having specified minimum
yield strength of 33, 35, or 36-ksi (230-250 MPa).
Small scale tests conducted in 1957 as part of an offshore platform owner's engineering
effort were belatedly published in Reference 3. These were simple T-connections loaded in
compression. Ultimate punching shear ranged from 8-ksi at gamma of 26, to 16-ksi at gamma of
10 (the latter chord being stockier than prevailing design practice of the time). Electric resistance
strain gages were used, but large gages on small specimens failed to reveal the high local stresses
at the hot spot which we would now expect to find.
84

TABLE 3.2 SPECIMENS USED IN THE SOUTHERN METHODIST UNIVERSITY TESTS

1 _ ^ - ~ - *- COVER

4 1/2" O.D. PIPE p=l

FOR A LL TESTS pM

<*.i/b
,5 e/o

SHEAR SHEAR
MARK O . D. t I y 8 LOAD AREA STRESS

C - 12 8 5 / 8" .188 39" 2 2 . 45 0 . 5 34 37 2.65 14.0

C - 18 8 5 / 8" .281 39" 1 4 . 78 0 . 5 40 109 3.95 27.6
C - 26 8 5 / 8" .406 39" 1 0 . 11 0 . 5 45 196 5.75 34.0
E - 16 12 3 / 4 " .250 45" 2 5 . 00 0 . 3 60 80 3.55 22.4
E - 24 12 3 / 4 " .375 45" 1 6 . 50 0 . 3 64 130 5.30 24.4
E - 32 12 3 / 4 " .500 45" 1 2 . 21 0 . 3 67 215 7.05 30.4
G - 19 18" .297 45" 2 9 . 80 0 . 2 54 87 4.20 20.7
G - 24 18" .375 45" 2 3 . 50 0 . 2 55 115 5.30 21.6
G - 32 18" .500 45" 1 7 . 51 0 . 2 57 205 7.05 29.0

Professor S. Thompson of SMU directed a series of tests in 1958 (ref. 22) which gave
the first broad picture of tubular joint behavior to American engineers. Test parameters are
shown in Table 3.2; these are T-connections which would fit on the table of a universal testing
machine, and loaded in compression. By keeping the branch size constant, and systematically
varying chord radius and thickness, the tests focused on the important parameters gamma and
beta for the cylindrical shell which must carry the loads. Results are plotted in punching shear
format in Figure 3.9. Failure occurred at stresses much lower than the shear strength of the
material, which would have been about 20-ksi. Failure was defined as the point at which
deflection continued to increase without additional load.

10 0 2 0 3 0 4
Fig. 3.9. SMU test results
= a /t
p l o t t e d in punching shear
format.
 85

. . Gonzales of Pittsburgh Testing Laboratory, a commercial laboratory in New

Orleans, performed a series of 16 welded tubular connection tests in 1959, which are summarized
in Table 3.3. These were double-T or cross connections, configured to accommodate a universal
testing machine and loaded in compression; however, the far ends of the chord were held circular
with rings or diaphragms, so that the mode of failure was local plastification of the chord face
rather than general collapse ovalizing (compare specimens 1 and 2). The tests were nearly full
scale replicas of configurations being used by the sponsoring design organization. Some had
ring stiffeners as shown in Figure 3.7, and did not fail at the 300-kip capacity of the test machine.
Results for simple joints, those without rings, were reduced to punching shear format by the
present author. For some of these tests, "excessive" deflection, rather than peak load, was used
as the definition of failure.

TABLE 3.3 PTL PIPE JOINT TESTS

no r i n g s w i t h i n j o i n t
e x c e p t as n o t e d

F A I L LOAD SHEAR SHEAR

NO. BRACE CHORD (KIPS) AREA STRESS 8

1 8-5/84>i 2803/8 32.5+ ENDS OF CHORD NOT HELD OPEN

2 8-5/8* 283/8 65.0 10.1 6.4 37 .31
3 8-5/8* 263/4 199.5 20.2 9.8 17 .33
4 1 0 } 283/8 77.5 11.8 6.6 37 .37
5 263/8 232.5 ONE RING STIFFENER 0 J O I N T
8-5/8
1 2 - 3 / 4 * 15.0 5.4 37 .46
6 27-3/43/8 80.0
7 8-5/8* 263/8 82.5 10.1 8.2 35 .33
8 20} 283/8 120.0+ 23.4 5.2+ 37 .71
9 16 J 283/8 87.5+ 18.9 4.6 37 .57
10 10-3/4*1 263/4 NO FKA I L (? T R I P L E RING S T I F F E N E R 1? J O I N T
300
11 6-5/8* 163/8 47.5+ 7.8 6.2 21 .42
12 6-5/8* 263/4 KIL 9
NO F A RING STIFFENER
300
13 6-5/8*1 203/8 79.5 7.8 10.2 26 .26
14 8-5/8* 263/4 NO F A I L RING S T I F F E N E R , EXTENSIVE GAGES
15 6-5/8* 243/8 64.5 7.8 8.2 32 .28
16 8-5/8* 263/4 KIL 9
NO F A FLAT RING REINFORCING
300
+ = Test stopped, excessive deflection.

The same design organization also sponsored the full scale proof test of a grouted
double-K-connection as described in Reference 2. The chord was 4000.75-in mild steel with a
36-inch pile grouted inside. The 90-degree 16^0.50-in compression branches were loaded by a
rod threaded through the connection. The 35-degree 2000.50-in tension branches were loaded by
a universal testing machine, developing a load of 1400-kips when the tension branch member
failed at the test fixture. Nevertheless, branch member had passed its specified minimum yield
stress, and the corresponding punching shear stress of 21.2-ksi was high enough to raise
considerable interest, being more than twice as high as that achieved in the PTL tests.
Early work by Professor Bouwkamp at the University of California is described in
References 24-26. He first popularized the use of negative-eccentric or overlapping K-
connections, which doubled the capacity of the corresponding gap connection. The second
reference tested a series of 13 relatively stocky tubular K-connections with 6-inch main
members, as might be used for small onshore structures, with the objective of defining for AISI
the range of connection parameters for which 100% of the branch member capacity could be
developed; the one test which failed in the chord (#2-thin) developed 23-ksi punching shear
(computed at the toe of the weld) for a gamma of 11. Finally, fatigue cracking occurred after
only 60 cycles in the 12.750.33- chord of a 35-90-degree K-connection with 5.56-in branches,
at a peak punching shear stress of 10.8-ksi; this plots in the same scatter band as the static tests
discussed above.
Professor Toprac's Phase I research at the University of Texas is reported in References
11 and 27. His series of seven T-connection tests, loaded in tension, was designed to explore the
parameter space of alpha, beta, and gamma, as displayed in Table 3.4 and Figure 3.10. Alpha is
86

TABLE 3.4 TOPRAC PHASE I TEST SPECIMENS

Pipe 3-

P,pe A

= L / a , = c / a , = a / t where a and c a r e the

mean r a d i i o f Pipes A and B, r e s p e c t i v e l y

CHORD BRANCH
PIPE A PIPE
MARK O.D. h O.D. L

1 12.75" 0.500" 2.875" 0.203" 48" 7.84 0.218 12.25

2 12.75" 0.250" 2.875" 0.203" 48" 7.68 0.214 25.00
3 16.00" 0.250" 3.500" 0.216" 72" 9.14 0.209 31.50
4 12.75" 0.250" 5.563" 0.258" 48" 7.68 0.425 25.00
5 8.625" 0.250" 5.563" 0.1875" 24" 5.74 0.643 16.75
6 12.75" 0.250" 5.563" 0.258" 96" 15.36 0.425 25.00
7 12.75" 0.250" 5.563" 0.258" 72" 11.51 0.425 25.00

varied in specimens 4-6-7, with the longest chord failure involving beam bending rather than just
punching shear. Beta varied from 0.21 to 0.64. Gamma varied from 12 to 32, with specimens 1-
2 permitting the cleanest comparison. Additional specimens from Phase II, including Y- and K-
connections, are documented in References 28-30.

DEFLECTION

Fig. 3.10. Load deflection plots for Toprac's tests.

Toprac's tests showed the tremendous reserve strength of simple tubular connections.
Stages of this behavior are illustrated in Figure 3.11, considering a section through the chord, at
its intersection with the branch member. The distribution of circumferential stress is shown in
stage 1; at low loads, this follows elastic theory and shows the very high stress concentration
factor (low connection efficiency) discussed earlier. As the connection is loaded beyond yield
(stage 2), it deforms plastically, while the applied load continues to increase. Finally, at loads
1.6 to 6.5 times that at first yield (ref. 31) the connection fails by pullout as shown for tension
loads, or by ovalizing collapse of the chord for compression loads. Either way, the ultimate
punching shear can be taken as the radial component of shear stress on the potential failure
surface.
 87

STAGE 1 STAGE 2 STAGE 3 I ^_

DEFLECTION
Fig. 3.11. Stages in the overload behavior of a simple tubular connection.

Figure 3.12 gives a more general description of load-deflection behavior of tubular

member-connection assemblies, and the various possible definitions of failure, adapted from
Toprac and Kurobane (refs. 32 and 33). The behavior of a 30-diameter long strut with "perfect"
end connections is also shown for comparison. Localized first yielding in the "hot spot", as

DEFLECTION

Fig. 3.12. Load-deflection behavior and failure definitions. (1) Luder's yield lines appear. (2) Non-
linear load-deflection. (3) First crack. (4) Maximum load in compression. (5) Deflection
limit. (6) Maximum load in tension. (7) Second maximum.

indicated by strain gages or the appearance of Luder's lines, is indicated by point 1; this may
occur even earlier than predicted by elastic theory, due to the presence of residual stresses. At
point 2, more widespread yielding causes noticeable non-linearity in the load deflection curve.
For tension loading, a crack or ductile tearing may appear at the notch at the toe of the weld at
point 3. If the behavior is ductile, and it usually is in statically loaded small scale specimens of
mild steel, the specimen goes on to a much higher ultimate load (point 6). If the behavior is
brittle, as observed in service failures of large scale joints involving high strength steel operating
below the crack arrest transition temperature, the load drops suddenly; in non-redundant
structures with no alternate load paths, this can trigger catastrophic collapse. Sometimes the
connection itself mobilizes a secondary load path or residual strength (point 7); in redundant
structures this combines with load redistribution to prevent collapse. In compression, the
88

connection also unloads after passing its peak load, point 4, with similar considerations of
residual strength and redundancy applying thereafter. Reference 34 (and later ref. 69) also
imposes a deflection limit as a definition of failure, point 5; at this limit, a 30-diameter long strut
with flexible connections at each end would have five times the axial deflection of the member
alone (considerably distorting the results of most frame analyses). Not shown is the case where
the loading equipment runs out of capacity, but the loads achieved are high enough to be useful
anyway.
The original data base described in the foregoing contained 36 tests of simple tubular
connections, in which radial loads from branch members were delivered directly to the chord via
the welds. Toprac compared the ultimate punching shear from his seven tension tests to the
material shear strength as shown in Figure 3.13(a). However, when the larger data base is
considered, Figure 3.13(b), and first crack is taken as the limit for tension tests, the trend is
towards much lower punching shear stress at failure, more like the early "rules of thumb" cited in
Chapter 2. However, at this point in time (1965), the variation of capacity with geometric
parameters of the connection was not yet understood.

NOMINAL
MATERIAL
SHEAR
STRENGTH

(a)
7EHL
10 15 20
ULTIMATE PUNCHING SHEAR - KSI

"RULE OF THUMB"
DESIGN VALUES
F i g . 3 . 1 3 . H i s t o g r a m s of
punching shear capacity, (a)

Ultimate load from Toprac's

/
/
/
s / /
(b) tension T-connections only, (b)
* / / / Failure load from all tests in the
s / /
* . original data base, with first
0 5 10 15 20 25 30
PUNCHING SHEAR AT FAILURE - KSI crack taken as failure for
tension.

Only 13 of these early tests were retained in later data bases (refs. 31 and 34) ~ four of
Toprac's and all of the SMU series. The others were eliminated because of proprietary
considerations, because they were too small to be relevant to offshore structures, because of
inadequate documentation (e.g., actual yield strengths not reported), or because of some other
deficiency perceived from the perspective of later sophistication. However, they were all we
had at the time, and they provided the first breakthrough, as well as an enduring understanding of
the strength behavior of tubular connections. From a structural reliability viewpoint, this may be
viewed as a data base in which all the variables are lumped together, rather than having separate
bias and uncertainty for dimensional variations and yield strength variations, as well as variations
due to the engineering model (ref. 35).
 89

3.2 LESSONS FROM FIELD FAILURES

It will be instructive to look at a few case histories in which the static strength of tubular
connections played an important role. Figure 3.14 shows a small single-well jacket which
collapsed during hurricane Hilda in 1964. The structure came apart in the connections and was
reduced to a pile of scrap pipe, which was recovered. Lying in the junk pile can be seen the well
with internal tubulars, the jacket legs with holes where braces used to be attached, and relatively
undamaged but now separate braces, still sealed at the ends with coupons pulled from the jacket

Fig. 3.14. Well jacket failure in

H u r r i c a n e H i l d a , Gulf of
Mexico, (a) Before, (b) After
salvage.

A post mortem analysis indicated that Hilda's waves in this location were two-thirds of
the design wave, with the corresponding stress in the most highly loaded diagonal brace being
only 13.2-ksi. However, the jacket leg was only 0.375-inch thick, with no joint-cans or other
reinforcement at the connections. Based on the author's punching shear criteria (to be presented
shortly), the ultimate strength of the connection was only 65% of the applied loads. At some of
the connection failures, the main member was dimpled outward in the vicinity of the break. One
of these was preserved and exhibited as "found art" sculpture (ref. 36). At other locations, there
was not much apparent distortion, yet the failure surface showed ductile shearing type of failure
rather than brittle fracture. The primary cause of collapse was failure of the chronically under-
designed connections.
90

Industry-wide, a total of 24 offshore platforms failed hurricane Hilda (refs. 37 and 38).
Examination of the salvaged wreckage revealed many connection failures similar to those just
described. Some showed partial failures, suggesting the possibility of progressive damage as
low-cycle fatigue. Spurred by this experience, the next few years saw intensification of research
efforts already underway, into the static strength and low cycle fatigue behavior of tubular
connections.
Thus, when hurricane Camille occurred five years later, structures designed with the
newly developed punching shear criteria fared considerably better. Figure 3.15 shows a platform
in 322 feet of water just after the storm. Notice the extensive damage to the equipment on the
cellar deck, 45-feet above sea level, indicating a wave crest 4 to 6 feet higher. From this, an
extreme wave height of 75 to 80 feet has been estimated, considerably in excess of the design
wave in use at that time. Shell had six major platforms in the area of peak storm intensity, one of
them being the world water depth record holder. Underwater visual inspection of these structures
after the storm revealed no damage to the tubular connections. Although one structure collapsed
as a result of seafloor movements (ref. 39), its tubular jacket structure remained intact.

(a)
(b)

Fig. 3.15. SP 62 platform after Hurricane Camille. (a) Model of jacket, which was undamaged, (b)
Photo showing damage to topside equipment.

Collision damage also provides some instructive examples. Figure 3.16 shows a single
well structure which was struck by a freighter off the coast of Texas. While severely damaged,
the structure did not collapse. Although a number of connections ultimately failed, for the most
part they held together long enough that a great deal of energy was absorbed by bending and
buckling of members. These connections were reinforced by the addition of "filler" gussets, as
described in Section 2.3.2(i) herein, and come close to developing the full strength of the
members joined. Ductile tearing failures did start at points of stress concentration, at the tips of
gussets. No design or materials problems were evident in this case; this was simply a matter of
gross overload. In other cases, these types of connections have experienced complete severance
of the braces, from brittle fractures starting at the tips of gussets.
 91

Shear failure in brace

at edge of gusset

Fig. 3.16. Well jacket hit by ship.

In the early 1960s, a general distrust of the shell capacity of tubular connections led to
the use of elaborately reinforced designs (ref. 40). Figure 3.17 shows an example of this type,
used by Shell from 1963 to 1968. In addition to using a joint-can, overlap keeps most of the
brace axial load transfer out of the chord, and partial ring gussets take care of out-of-plane
bending. During hurricane Betsy in 1965, the world's first semi-submersible collided with the
the world's deepest jacket, destroying both. Underwater examination of the jacket (using the
Star-II submarine) showed structural failures occurring in the braces about one diameter away
from the connections, Figure 3.17(b). Thus, in these over-reinforced designs, we finally had
enough static strength.

(a) (b)

Fig. 3.17. Gusseted overlapping connection, (a) Brittle fracture of topside connection exposed to low
air temperature, (b) Connection intact, brace severed after collision; inset shows submarine
used for underwater inspection.
92

However, in the years 1964-65, partial failures were found in five of these connections.
All the observed failures were in braces above the waterline, as shown in Figure 3.16(a). None
of them resulted in platform collapse, due to structural redundancy. All were repaired and have
served well for twenty more years. Investigation indicated the following contributory factors:
material having low notch toughness at winter-time air service temperatures, high stress
concentrations at the gusset tips, and the presence of a short weak link between the connection
and corrosion doubler wrap. The principal lesson here is that awkward reinforcement, while
appearing to solve the static strength problem, may create new ones.

3.3 OVERCOMING THE BETA PARADOX

Reference 27 describes the Kellogg method used by process designers to estimate local
stresses at tubular intersections in pressure piping. This is based on analogy with the behavior of
a circular cylinder subject to the axisymmetric radial line load case of Figure 2.10(a). While it is
a rather large extrapolation from the axisymmetric case to a tubular intersection, the resulting
expression for hot spot stress has all its terms in the right place and in approximately the right
proportions:

^ (l.5 t 2. + t ^ ) (3.4)
hotspot \ b A b S /

Here P/A and M/S are nominal axial and bending stresses, respectively, in the branch member.
The expression in parentheses is the analogous line load and contains an empirical adjustment
factor of 1.5 for the tendency of axial loads to have a more severe effect on chord ovalizing than
in-plane (longitudinal) bending moment in the branch member. No mention is made of the
diameter ratio, beta. While this may at first seem a serious defect, we shall see that this omission
is what makes the method so workable.
Consider the case of a jacket launch leg, loaded by two line loads (Q = W/2) from a
continuous launch cradle and carrying the load as a beam. The failure locus can be pictured as
running along the outside of the two line loads, with the nominal punching shear given by Vp =
Q/T. Failure occurs when the circumferential shell stress reaches yield, Fy Q . Roark's closed ring
solution for this case, shown in Figure 3.6, may be written in terms of B e a ' s tabulated
coefficients (FM and FT, Table 3.1) as

F
W = 2T V = (3.5)
6FMR + FT
2

For unstiffened chords in the usual range of gamma, 10 to 20, the thrust term FT is insignificant,
and the expression can be simplified to a shell bending form with only one coefficient, F(0) =
1/(24-FM) as follows:

ult V = f () (3.6)
0.5

That is, strength, in terms of punching shear efficiency, is inversely related to the chord thinness
ratio (gamma) and shows a marked increase with diameter ratio (beta), as shown by the plot of
f(0) in Figure 2.10.
 93

For the case of tubular T-connections, calculations with Bijlaard's equations show a
similar strong effect of beta. For example, Johnston (ref. 4) showed a theoretical increase in
permissible punching shear stress (at constant hotspot stress) of 70% in going from beta of 0.3 to
beta of 0.7. That is, the connections get stronger as beta increases, even faster than would be
accounted for by the increase in branch perimeter and shear area. For a long time, many design
engineers were convinced this was the natural order of things.
Unfortunately, such an advantageous effect of beta is not borne out by tests of ultimate
strength. The 1958 Southern Methodist static test series, as re-plotted in Figure 3.9, shows a
40% decrease in punching shear efficiency in going from beta of 0.25 to beta of 0.54. Similarly,
Toprac (ref. 11) found in his elastic stress measurements that there was a 30% increase in his
dimensionless stress factor (i.e. a decrease in efficiency) in going from beta of 0.20 to beta of
0.42.
So here was the beta paradox: The best available theory and opinion indicating a
favorable beta-effect, and test data indicating quite the opposite.
The Kellogg formula, with its omission of the beta-effect, avoids the paradox. It may be
re-written for the case of axial brace loads only as

F
ult V = ^ (3.7)
1.8 . 0.5
7

In comparing this formula with the closed ring solution, for which the exponent of gamma is
unity, one might think that the exponent of 0.5 is too low. Also, the coefficient of 1.8 might be
suspect, since no plug is cut out of the main member in as tubular connection, as it is in piping
intersections. Thus, it seems reasonable to state this equation in a more general form

U lt = (3.8)
E
7

and use available test data to obtain a best fit for and E.

3.4 EMPIRICAL PUNCHING SHEAR DESIGN EQUATION

The author's original empirical design criteria for simple tubular connections i.e., gap
connections without overlap, rings, gussets was first formulated and used in 1966 (refs. 41 and
42). It considered two aspects of the problem and their interaction: (1) punching shear or local
failure in the vicinity of the attached brace, and (2) general collapse or ovalizing of the entire
joint can.

3.4.1 Local Failure

Test results for 35 simple tubular connections loaded to static failure were available in
the original data base. These came from various sources as discussed earlier: the California Co.
(refs. 2 and 23), Freeport Sulfur Co. (ref. 3), the University of Texas (refs. 27-30), the University
of California (refs. 24 and 25) and Southern Methodist University. All the tests plotted here were
set up so that failure occurred in the chord, locally around the attached brace, rather than in
general collapse of the entire chord (i.e., PTL test # 1 , Table 3.3, is not plotted). Tension and
compression tests, mostly tee and cross connections with a few K-connections (N-type) are
represented. Failure was defined as plastic collapse for the compression tests, and as first crack
for the tension tests.
94

All this data is plotted in Figure 3.18, as ultimate shear stress on the potential failure
surface, versus chord thinness ratio, gamma. Taking the nominal yield stress Fy for these tests as
36-ksi, it was possible to draw a conservative "best fit" ultimate strength design curve as shown
in the figure. The corresponding coefficients in Equation 3.8 are = 0.5 and = 0.7.

KELLOGG fn=ief*J
ALL C U R V E S F O R F yo = 36 KSI

20 30 40 50 60
<t *)4 C H O R D THINNESS RATIO

CALCO TESTS TOPRAC

PTL . . .
COMPRESSION J\ \ 1\ COMPRESSION FAILURES ,0 TENSION .
YIELD FAILURE U ' YIELD FAILURE & IN C H O R D I J PHASE I T E E JOINTS I

=
=
= .26 T O .37 .25 AISI T H I N W A L L # 2 FIRST FULL
ffl .42 T O .71 - .36 <B> - .36 CRACK FAILURE
- .54
SHELL CALCO 1960 0= .21



0-
POS. E C C E N T R I C = .44
FREEPORT SULPHUR m = .52 .64
A
0 PILEGROUTED [ _J PHASE 2
CALCO 1965-CYCLIC TESTS

7 T
V / ? = . 5 2 , .33 A .44

Fig. 3.18. Effect of gamma on static strength - the original data base for simple tubular connections.

Introducing tau as the ratio of branch to chord thickness, the resulting design equations for 90-
degree connections are then:

a c t i n g V p = . fn (3.9)

F
ultimate V = ^ (3.10)
0 7
0.5 .
These express the resistance to local plastic collapse or progressive tearing of the cylindrical
shell which serves as the chord, including large deflection, plastic bending, membrane action,
strain hardening, etc. Note that for gamma less than 7, the Equation 3.10 would exceed the
material shear strength of 0.57 Fy Q . Although there is no data in this region, this limit is also
observed in the design criteria.
 95

Figure 3.18 also shows some of the design criteria existing prior to 1966. Marshall's
effective width approach was discussed in connection with Figures 3.8; the vertical spread is due
to the effect of beta, not represented here. Johnston's criteria (ref. 43) was based on the shell
theory of Bijlaard; its extreme conservatism led to the elaborately reinforced connection designs
prevalent in the early 1960's.
During the preparation of this book, a couple of numerical errors were discovered in the
original data reduction. The data in Tables 3.2 and 3.3 have already been corrected. However,
Figure 3.18 is the original historical plot, including the errors, as it shaped our thinking. The
corrected data are re-plotted in Figure 3.19; symbols are as in the previous figure. This does not
fit the design curve nearly as well. Although the mean bias on the safe is slightly higher than
before, there is considerably more scatter. The corresponding lower reliability was eventually
discovered as larger data bases were examined, as will be discussed in Section 3.6.4.
to

Oil

CURYE 1 F O R Fyo = 36 KSI

\ A
\ 1 MARSHALL'S BEST F I T 1966

2o
V . %
<
<
I
to
10 20 30 40 50

<t CHORD THINNESS RATIO

Fig. 3.19. Effect of gamma on static strength - corrected data.

For connections other than 90-degrees, the load perpendicular to the chord is reduced in
proportion to sin0. This effect is introduced into the design procedure by restating the acting
punching shear as

a c t i n g Vp = s i n 0 (3.11)

This approach conservatively neglects the increase in perimeter for angle connections, but also
neglects stresses on the failure surface due to the longitudinal component of load (parallel to the
chord axis).
These equations were used with an additional safety factor of 1.8, following the ASCE
recommendations for plastic design (ref. 44). The equations are even more conservative for the
few K-connections tested than they are for the bulk of tee and cross connections. However, it
was also recognized that they could be unconservative (err on the unsafe side) for very short
unstiffened joint cans in which general collapse of the entire cylinder is a problem. In offshore
practice, the lowermost node of the jacket, where space frame lateral loads are transferred to the
piles, fall in this category, as do the launch legs which are subject to crushing loads during
construction. In onshore practice, the support points of a truss may have the same problem.
General collapse is discussed in the next section.
96

During the initial euphoria after finding the ultimate punching shear as given above, Q =
ult Vp was taken as a characteristic and nearly universal expression of the radial line load
capacity of cylindrical shells. While extrapolations based on this view were considered too
speculative for codification, they were extremely useful in the author's own design work. One
example is the use of this line load capacity as a "building block" in the design of complex
gusseted connections, as discussed earlier in Section 2.3.2(i). Another is illustrated in Figure
3.20, in which the beneficial effects of tangential load transfer into a cylindrical shell is evaluated
by substituting the local dihedral angle, psi, for theta in Equation 3.11. For the equal diameter
cross connection shown, numerical integration around the brace circumference indicated a 1.9-
fold increase in axial load capacity, and 2.3-fold increase in out-of-plane bending, compared to
the usual method of calculation. Similar advantages also accrue for tangential gusset plates and
flared connections, as shown in part (b) of the figure. These types of design enjoyed a brief
popularity for platforms in Cook Inlet, Alaska.

Fig. 3.20. Tangential load transfer, (a) Equal diameter cross connection, (b) Flared connection, (c)
Tangential gusset.

3.4.2 General Collapse

General collapse failure of a tubular connection may be defined as gross flattening or
distortion of a large part of the cylindrical shell which forms the chord. It is in contrast to the
more localized punching shear failure emphasized in the previous section.
Perhaps this type of failure can be best illustrated by an example taken from a full-size
connection test (PTL test # 1 , ref. 23). The test specimen consisted of a short length of chord
material, with a perpendicular brace on each side, i.e., a cross connection, Figure 3.21(a). The
braces were both loaded in compression until the connection failed by general flattening of the
chord. Because of this unusual mode of failure, this connection was not included in the data
from which the design criteria for local failure was derived. The test load was only 56% of that
predicted for local failure, Figure 3.21(b).
 97

General collapse behavior can be treated analytically with the available closed ring
solutions (refs. 7 and 18). The solution for this particular loading, Roark's case #8, was also
tabulated by Bea and is shown in Figure 3.21(c). Since we are trying to compute an ultimate load
which can be compared to test results, the plastic section modulus for circumferential shell
bending of the chord wall is used, and the entire length (just over two diameters) of the joint can
is taken as being an effective part of the ring. Unfortunately, this method, taken alone, calculates
a failure load which exceeds the test result by 26%.

;ST L O A D ( ^ )
32.5

2 8 * 3 / 8 w. A - 7 S T E E L
X 60' LONG

Fig. 3.21. Interaction between local failure and general collapse, (a) PTL specimen #1, Table 3.3. (b)
Local failure, (c) General failure, (d) Interaction.

The actual mode of failure involves interaction between local shear and general
ovalizing. For the case of direct stress and shear on a plane, with no direct stresses on the other
axes, the constant work of deformation criterion for yielding reduces to circular interaction. By
analogy, we can write for the problem at hand,

/_!uit_y + ( ^ i t - 1 ^ 1.0 (3.12)

p P ocal
\ generalJ I l
p
where P u j t is the ultimate load for combined local and general failure, and P g e n e r ia and local
are the computed ultimate loads for each mode of failure taken alone. While strictly speaking,
we cannot prove this analogy to be valid, Equation 3.12 checks well with the test result, Figure
3.21(d). It also passes the test of reasonableness, since it indicates that local failure limits the
strength of connections in remotely stiffened or long chords of uniform thickness, while it
suggests that general collapse would be a limiting factor in very short shells (e.g., heavy wall
joint cans).
Reference 42 carries this analogy even further, with a very elaborate and speculative
procedure for K-connections. In practice, this was eliminated in favor of the practical design
requirement that the joint can extend at least D/4 beyond the branch member footprint. D/4 may
98

STRENGTH
PHENOMENON
INCREASE
FACTOR

PLASTIC

(a) VS
ELASTIC S
1.5

(b)

| - ^ - - A V E R A G E YIELD STRENGTH
.2
SPECIFIED MINIMUM

(c)
STRENGTH

STRAIN HARDENING

-TENSILE STRENGTH
UTILIZED
VS 1.6
MINIMUM
YIELD STRENGTH
(d) SPECIFIED
TENSILE RANGE
IN F O R M U L A

STRENGTH

Fig. 3.22. Sources of post-yield reserve strength, (a) Plastic section, (b) Triaxiality. (c) Yield bias,
(d) Strain hardening.
 99

not suffice where the load pattern produces severe chord ovalizing (see AWS paragraph
10.5.2.1(2)).

3.5 UNDERSTANDING THE SOURCES OF RESERVE STRENGTH

Since the ultimate strength used in the design of tubular connections is substantially
more than that obtained elastically, it will be useful to dwell briefly on the sources of this extra
strength which we rely upon. See Figure 3.22. Explaining reserve strength factors of up to 6.5
times first yield was one of the more interesting challenges of the early work.
Plastic section - One source of the increase is the difference between the plastic section
modulus and the elastic section modulus S, a factor of 1.5.
Triaxiality - A unit element of material under the intersection of the brace and the chord
will be subject to triaxial stress as shown in Figure 3.22(b). The point illustrated corresponds to
the stress at the hot spot, on the chord surface at the intersection in thin shell analysis, Figure 3.3.
For the example shown, the distortion energy theory of failure predicts an effective increase in
the material flow stress of 60%.
Yield bias - For mild steel, the average yield strength is about 1.2 times the specified
minimum, Figure 3.22(c). This is included in the original empirical correlation being discussed
here, which was done in terms of specified minimum or nominal yield. In subsequent
correlations, the actual yield strength of the test specimens was used, in order to calibrate the
engineering strength model by itself. In these latter cases, the yield bias must be re-introduced
when evaluating reliability of the resulting connection designs.
Strain hardening - Many of the tubular connection tests were carried to the point where
the material parted. In these specimens, it seems apparent that the full tensile strength of the
material was at least locally utilized in developing the observed ultimate strength. This ability of
the material to strain harden is an integral part of the behavior on which empirical design is
based. Connections constructed with steel having a high yield/tensile ratio would not be
expected to show as much post-yield strength increase as connections constructed of the mild
steels tested here. This situation makes it a good practice to require the joint-can steel to have a
tensile strength of at least 1.5 times yield; or failing this, to use an effective yield of 2/3 the
tensile strength.
Load redistribution - The factors mentioned so far, taken cumulatively, could account
for a post-yield strength increase to 4.6 times the load at first yield. To explain the behavior of
some of the connections tested, still further increases in strength must be called upon. Consider
the tubular connection shown in Figure 3.23. In simplified terms, the chord may be visualized as
a grid of rings and stringers. First yielding occurs at the hot spot, point A. At this point, a yield
line forms (indicated by the cross hatched area), analogous to the formation of a plastic hinge in a
continuous frame. The full strength of ring A-B is not reached until a considerable angle change
has taken place at A and the ring deforms sufficiently for yielding to occur at B. Ring A-B then
continues to deform at constant load, while the load and deflection in ring C-D catch up. This
catching-up results in a more nearly uniform distribution of load transfer across the weld joining
brace to chord. The small deflection limit load of the assemblage is reached when ring C-D
deforms enough to develop its ultimate load, and the longitudinal stringer C-D yields in its
attempt to distribute the load further along the axis of the chord.
Large deflection behavior - The deformed shape of the assemblage as it reaches its
ultimate load is indicated, without much exaggeration, by the dashed lines in the figure. At this
point, changing geometry of the connectionthat is, large deflection tensile membrane action of
the chord shellresults in still further increase in strength.
100

Fig. 3.23. Redistribution of load. See text for explanation.

Inelastic finite element analysis of the connection as a continuous shell structure should
give more rigorous results, but perhaps less understanding, than the foregoing heuristic
explanation.
All these - plasticity, triaxiality, strain hardening, load redistribution, and large
deflection behavior - place extraordinary demands on the ductility of the chord material,
particularly since it is required to undergo all this distortion at the toe of the intersection weld, at
the hot spot. It is here that materials that are less than ideally ductile first give up. Thus,
complete fulfillment of the designer's responsibility requires that he pay close attention to the
materials being used at tubular connections, especially in the chord. This subject is discussed
further in Chapters 7 and 8.

3.6 FURTHER EVOLUTION OF THE A.W.S. CODE

In this section, we shall follow the development of the American design codes for
tubular connections during the years 1966-1983. This was, and continues to be, very much an
evolutionary process. Changes in the design practice came incrementally, building on what
existed before, rather than basing everything on the latest piece of work.

3.6.1 Compromise and Consensus

The status of tubular connection research in the USA in 1966, at the time of the author's
initial breakthrough on punching shear, is summarized by the chart in Figure 3.24 (heavy blocks
denote research then being proposed for funding by Shell). Most of this work has been described
in the earlier parts of this chapter, and in Chapter 2, providing the conceptual basis and initial
data base for the Code. During the next three years, ongoing research and practical applications
were used to confirm the design procedure, and to resolve in important issues in the areas of
fatigue, fracture, materials, fabrication, and inspection. Among the practical applications were
five offshore platforms which exceeded the previous world water depth record, including that
shown in Figure 3.15. The results were initially documented in References 45 and 46 and
published at the first Offshore Technology Conference in 1969 (ref. 47).

Fig. 3.24. Tubular connection research and development as of 1966 (overleaf).

 1955

SHER
A A R AE M E T H ]O D
1 K EGL L MO EGT HD O 1
| J O I TN T E SST ( S M) U 1
1
J J
C AOL CX J O I TN T E SST ( P LT & U )l
I1 1 I1
I 1 1960
F I L L RE G U S S SE T 1
IFN E G A T EI V E C C E N T R IYC I T1
| J O I TN T E SST ( F R E E P TO RS U L P H )U R 11
1961
I 1
1962
C O L U NM A N A L OY G ( B R Y A) N T 1 B I J L A AD R
1 1 S H E LL
T H E O YR
COMPUR TE
P R O G RMA I
1 1963
I J O H N S TNO a K t Y l tw I
1
, DE J O GN J O I NS T
N E G A T EI V
1 E C C E N TCR I I
R E I N F O RDC E
1 I 1964
' 1 T O P R CA P H AES I CO - OP
P H A ES 1 F A T I GE U 1 M A R S H LA L J O I TN T E SST
T E ST 9 S P E C ICF I "SOM E O E S ING
C H E V RNO D E S I GSN C O N S I D E R A TSI O N " A D J U S T "E DT H E OY R
F OR T U B U LRA J O I N T" S
1 | j 1 HILDA
I 1
1 T A B U L DA T E 1965
I
]
1 1 F I NEI T
1 R I GN S O L U T ISO N T O P RCA P H AES CO - OP
L A R EG . | E L E M TE N '
1
A I SI R O FO T R USS J O I NS T JOIN T 4 DUHAM L E' | S TMO R
1 I T E S ST OF A - 63 T. Y. J O I TN
1 1
( B O U W K PA M) S T U YD SH E LL 1 1 TR
F O REC S P E C A F A I L U RS E
| 1 1 ( E M E R Y V I L) L E T t a ia
T H E O YR I P H AES
ic
1 L* '1 ei T 1i r1i\jUT
O P T I M I Z A NT I OF
O S T A ET OF T HE A RT 1966
I 1 1 M A R S H LA L
JOIN T D E S I GS N V AI ON T HE D E S ING OF P H A ES M
| C O N C R EE T F I N I ET E M E R Y V I L! L E S WI R
T H E O YR & P H O T- O S I M PEL P H A ES m C O - P O
1 1 F I L LD E I 1 T U B U LRA J O I NST " E L E M ET N F R A C T UE R | P A T I EG U
U . S. S T ELE C H A RY P T E SS T C O - OP M O RE T E SST
J J O I NS T || E L A S TC I S T U D I .E S 1 A N A L Y SS I N OW
F A T I GEU DUNDRO AV
T E ST DE J O GN J O I TN E L A S TC I A - 4 14 I N S P E C T I! O N
J T E S ST | S T A N D A RD j T H E OYR
M O RE F A T I GE U T E SST 1 T E M P L AE T 1 T H E OYR
B E H A V RI O S I M PEL
1 \t |1 I T E ST G U S S EDT EJ O ITN I & T E SST ON J O I NST
( ID
X NOT T E M P L AE T S I M PEL 1967
R
L O NG T E M S T UYD J O I NST
AVAILABLE 1 ' C L HO U G I T E ST G U S S E DT E J O ITN 1
F I N I ET E L E M TE N T O W RE
1 S H LE ^L [> 1 E M E R Y V IEL L& E PR G E N E R A L DI Z E
S H E LL
(17) F A I L UE R A N A L YSS I
A N A L Y SS I
COMPUR TE 1968
P L A S TC I (18) BAS
C I M E C H A N LI C A P R O G RMA
1 I P R O P E R TS I E
P H A ES JE B E H A V RI O
(12)
1969

"RATION L AD E S ING OF
1 T U B U LRA J O I N -T S 1970
I
O T H RE R E S E A HR C I K A IRS E B O U W K AP M 1 M .I S C
S H ELL C O N S T R U C NT I O S H E LL M A T E R I SA L& S W RI T O P R CA
101

J B O U W KPA M J C H E V RNO J O U ET S I D D E S I NG G R OPU & PR F A B R I C A TN I O DUNDRO AV

102

Three more years were required to incorporate this material in the AWS Structural
Welding Code. During this period, The AWS Subcommittee on Welded Tubular Structures
functioned under the chairmanship of Professor A. A. Toprac of the University of Texas; while
the committee secretary, Barney DeGeorge of Continental Emsco, imposed a strong sense of
order and kept things organized through a rapid succession of drafts. Committee members
represented mostly designers and fabricators, rather than researchers, offering a variety of
practical viewpoints on what worked and what didn't. The following list indicates their areas of
responsibility as team captains:

Paul Masters, American Bridge - Dl liaison

Lowell Johnston, Shell Development - format, general provisions, materials
Peter Marshall, Shell Oil - design stresses (ref. 48)
Roland Graham - U. S. Steel
Omer Blodgett, Lincoln Electric - welding processes, qualification, workmanship
Don Sprow, Southwestern Labs - prequalified joints, filler metal
Griff Lee, McDermott (Jim Palmer, alternate) - inspection
Lane Phillips, Lawrence-Allison - editing
Wilfred Lee, Chevron - groove design for welded joints
Harold Phenix, Conoco - terminology, structural details
. E. Cummings, Joe Stine Inc. - joint preparation, tolerances
Joe Nutt, M. W. Kellogg Co.
Tom Ferrel, Belmas Co.

Concurrently, the author and Professor Toprac also served on the ASCE Task
Committee on Tubular Structures (Profs. W. J. Graff, D. W. Fowler, D. R. Sherman, J. G.
Bouwkamp, and H. Erzurumlu), again with Toprac as chairman. This provided a parallel
research focus for the technical aspects of static strength and fatigue of tubular connections. For
example, ASCE added the cyclic punching shear provisions. On at least one occasion, a quick
research project was conducted at the University of Houston to resolve an AWS code question
which had come up in the ASCE committee. They were also sticklers for detail, e.g., the
connection/joint/weld hierarchy defined earlier. The fact that AWS design stress provisions had
been reviewed by ASCE was instrumental in gaining their acceptance by the main structural
welding committee (ref. 49).
A third committee which functioned during this time was a task group of the author,
Griff Lee, and Arthur Guy, part of the API Committee on Standardization of Offshore Structures
(L. P. Johnston, chairman), which developed the tubular connection provisions of API RP 2A.
These provisions were initially identical to those of AWS, understandable given the overlap
between the two committees, yet they were passed by a substantially different set of reviewers.
Finally, the Welding Research Council's committee on tubular structures provided
funding for Professor Toprac and published the author's summary of work during this period
(ref. 50).
The Team "K" Report on Design Stresses (ref. 48) initially pulled together the design
provisions for tubular structures. It followed the format and adapted many of the provisions of
AWS Dl.0-69, Code for Welding in Building Construction, and paralleled the developing API
provisions. All three documents anticipated that the primary design effort would be governed by
the AISC Specification for the Design, Fabrication and Erection of Structural Steel for Buildings
(ref. 51), including use of the 1/3 increase for rare storm loadings.
 103

The provisions for butt welds and fillets were previously discussed in Section 2.5.1; they
are similar to those of the building codes, except for the addition of material describing particular
tubular applications. More radical at the time was a new section, "Limitations on the Strength of
Tubular Connections", which dealt with failure modes not covered by traditional weld stress
calculations. The conceptual basis for this was discussed at length in the previous chapter. Its
manifestation in the Code is recapitulated below:
The local shear failure mode adopted the author's ultimate punching shear design
formula (eq. 3.10) for axial loads, and his safety factor of 1.8. The general feeling was that this
was terribly conservative, compared to the survival experience of structures which did not meet
this criterion. A number of "refinements" were then added in committee to lessen the degree of
conservatism. The shear area was modified to include the BS 449 effective intersection length
factor, and a bending term was added, so that the acting punching shear became:

acting V = sin0 ( + | (3.13)

k k
^ a b '

where fa and f^ are nominal axial and bending stresses, and K a and are the corresponding
effective length and section factors (see Table 2.5). Shear area was also allowed to be calculated
at the toe of the weld, rather than at the nominal branch O.D., a further relaxation. These
reasonable extrapolations were accepted at the time, even though they were not covered by the
o r i g i n a l c a l i b r a t i o n , in the spirit of c o m p r o m i s e . A g r o s s e r r o r , and their general
inappropriateness, were not recognized until later.
The provisions for general collapse covered ovalizing, but not the rest of the failure
modes discussed in Section 2.2.2. "Strength and stability of the main member in a tubular
connection, acting as a cylindrical shell (and any reinforcement) should be investigated using
available technology, in accordance with the applicable design code." Since the closed ring
solution was a widely used design method at the time, it was initially felt that more specific
guidance was unnecessary. Also missing was a clear mandate to always consider both local
failure and general collapse, as well as their interaction. API RP 2A did include some detailing
guidelines, Figure 3.25, which generally avoided collapse problems in balanced K-connections,
along with a specific warning for "cross joints, launch leg joints, and other joints where load is
transferred through the chord." Unfortunately, in the AWS Code, punching shear came to be
used as a stand-alone design method, leading some designers to unsatisfactory results (ref. 52).
Similar to the elaboration of punching shear calculations, the committee wanted weld
stress calculations which incorporated the effective length and section K-factors, simplistic
provisions based on overall connection geometry and statics, which survive to this day. In
recognition of the uneven load distribution in tubular connections (see sees. 2.2.3 and 2.5.2), the
following check of local weld capacity was also specified for partial penetration and fillet welds,
to make sure that such welds would not become the weak link:
fc F
0.67 b y
or F E XX . t w e ld > or (3.14)
0.80 ult V

where the 0.80 factor applies to weld nominal tensile strengths, ^ of 60 or 70-ksi, and the
lower factor applies to higher strengths.
104

SPECIAL STEEL IN BRACE

(OPTIONAL)

Fig. 3.25. API detailing for a simple K-connection.

Finally, the initial provision for lamellar tearing was a 20-ksi limitation on radial or
thru-thickness stress in the chord, a working stress at which weldments with mild steel
attachments seemed to have satisfactory history of performance in service, having survived the
initial shrinkage stresses. Fortunately, this provision was soon replaced with a more appropriate
commentary on the metallurgical factors involved, and practical counter-measures, so that higher
strength steels could be effectively utilized.

3.6.2 Publication
The 1972 Structural Welding Code (ref. 53) combined the AWS provisions for
Buildings and Bridges, which had previously existed as separate documents, as well as Tubular
Structures, which were being covered for the first time ever. The Code remains similarly
structured today, although it has been greatly expanded.
Chapters 1 through 7 contain a body of provisions applicable to all three kinds of
structures, as outlined below:

(1) general provisions

(2) design of welded connections

general requirements
structural details
details of welded joints
drawings of prequalified weld details
 105
(3) workmanship
assembly tolerances
profile requirements

(4) technique
filler metal requirements
preheat
other general requirements
specific requirements for various welding processes
(including limitations for prequalified status)
manual SMAW
submerged arc
GMAW and FCAW
electroslag and electrogas
stud welding

(5) qualification
welding procedures
limitations on essential variables
required tests and testing methods
welders, operators, and tackers
special test for or Y connections (6 GR)

(6) inspection
duties of the inspector
obligations of the contractor
radiographic testing of welds
ultrasonic testing of groove welds (except tubular)

(7) strengthening and repair of existing structures

Chapter 8 covers the requirements for the Design of New Buildings, or statically loaded
structures in general. It covers general requirements, design stresses, structural details, and
workmanship provisions for this class of structures.
Chapter 9 covers the Design of New Bridges, or similar dynamically loaded, critical
structures. In addition to specific provisions for fatigue design, it features lower design stresses
and tighter workmanship tolerances.
Finally, Chapter 10 covers the Design of New Tubular Structures. The Code
requirements as they appeared in the initial publication of D 1.1-72, are as discussed below.
P a r t I - General Requirements - made it clear that this was intended for structural
applications, not pressure piping. There was an academically precise classification of , , K,
K-T, T-Y and offset connections, complete with many drawings, harmless distinctions not used
in the ensuing design provisions (leaving this in got another "yes" vote). The list of acceptable
base metals was expanded to include pressure vessel steels being used in joint cans for their
improved notch toughness.
P a r t II - Unit Stresses in Welds - had separate provisions for ordinary weld seam
applications as opposed to , K, and Y connections. Matching filler metals were defined. The
unique set of Limitations on the Strength of Welded Tubular Connections has already been
discussed. Provisions for weld areas, effective length, and effective throat, as well as shear area,
106

included the not-so-harmless BS formula for K f t. Fatigue provisions were also included, with a
family of S-N curves based on nominal stress, cyclic punching shear, and hot spot stress (section
2.4.1 herein), as discussed more fully in Chapter 4.
P a r t I I I S t r u c t u r a l Details - included fillet weld details, e.g., lap joints, and
transitions of thickness, including single-sided butt welds. The special demands and reduced
performance of some of these details were not initially recognized.
P a r t IV - Details of Welded J o i n t s - described the special provisions for welded
tubular connections between intersecting members. Weld details for complete penetration welds
in generic , Y and connections, which vary as a function of local dihedral angle, were based
on Shell's codification of McDermott's practice; this subject is more fully discussed in Chapter 8
herein.
P a r t V - W o r k m a n s h i p - includes assembly tolerances and requirements for visual
inspection. For non-destructive testing, radiography-based standards were invoked. A consensus
on ultrasonic testing had not yet developed, and the early Code merely provided that: "When
ultrasonic testing is required, the testing procedure and acceptance criteria shall be specified in
the contract."

3.6.3 Further Refinements

The first two editions of the combined Structural Welding Code were issued in 1972 and
1975, in loose leaf form so that interim revisions could be issued and inserted during the "off"
years (refs. 54-58). During this same period, continuing research, committee activities, inquiries
from users, and feedback from AWS educational seminars on the Code, generated a number of
changes in the chapter on tubular structures. These are described below.
Fixing the "wandering sin0" error in Equation 3.13 results in:

acting V = r [^ sin0 + ) (3.15)

KV K
a b '

As discussed in Section 2.4.2, for in-plane bending the entire moment acts as radial
loads on the intersection footprint, making use of the sin(theta) term unconservative for this case.
For out-of-plane bending, the plotted values of (e.g., Figure 2.42), included the effect of
sin(theta), making it inappropriate to include it again in the equation.
More recently, a similar embarrassing error was discovered in the equation for weld
stress, Code section 10.8.3, which has retained the K-factors for effective length and section.
Instead of keeping sin# for axial load as in Equation 3.15, the correct formula, as it appears in the
1990 edition of the Code, is:

f = ^> . / a + b\ )

weld tw \Ka K b/

This reflects the fact that the full axial load must be transferred across the intersection
weld, as opposed to only the radial component of line load affecting punching shear and chord
shell stresses. More rigorous treatment of the weld stress problem was discussed in Section 2.5
herein.
Basis for a number of the other incremental changes made in the early 1970s can be
found in Reference 50, first presented at the ASCE convention in San Francisco, April 1973.
Two modifiers were introduced into the allowable punching shear expression, as follows:
 107

allowable V = Q Q 22 (3.17)
f 0 7
0.9

where Q ^ e t a reflects the influence of beta, and Qf reflects the influence of concurrent chord
stresses, on the capacity of the connection. The Texas translations of early Japanese research
(refs. 5 and 20) provided source material for both of these.
Insight into the influence of beta came from Togo's simplified limit analysis of cross
connections which employs the physical model shown in Figure 3.26(a). Referring to the
equation on the figure, which has been reduced to punching shear format by the author, if the
chord effective length is taken as equal to its circumference (i.e., just over three diameters), the
last term becomes unity and the expression becomes similar in form to that derived for box
connections on the more rigorous basis of yield line theory (see Figure 2.28), but with twice the
capacity. The leading term is analogous to the Q f c e ta modifier, and the second term analogous to
the basic punching shear capacity. Thus, based on the limit analysis of both box and circular
connections, as well as the reasoning of Figure 3.20, an increase in capacity as beta approached
unity was anticipated.

Fig. 3.26. Japanese source data for Q^. (a) Yield line analysis, (b) Test results and empirical
equation.

Calibration on the Japanese data yielded an empirical modification of the limit analysis
expression shown in Figure 3.26(b). In this expression, the Q ^ e t a term has the following
properties and implications:
108

(1) A value of 1.0 for beta of 0.6; this brings us back to the basic punching shear
expression on the mid-range of beta, for which it was originally derived.

(2) Increasing connection efficiency for larger beta ratios, up to a limiting increase of
1.8-fold for beta of unity.

Note that for the mid-range of diameter ratios, the assumption of constant punching
shear capacity also provides a reasonable fit to the data, as shown in the figure. For very small
beta, the reasoning of Figure 3.20 would not indicate any increase in capacity, nor does the
experimental data overwhelmingly indicate the huge increases given by the expression.
Accordingly, the Code uses a constant Q ^ e t a of unity for beta less than 0.6. With this proviso,
the Q b e t a expression has proved to be remarkably durable, having been retained in three
subsequent major revisions of U.S. criteria, and in several more abroad.
This incremental adaptation of Q ^ e t a to the prior basic punching shear expression was
done in a way which filtered out differences with the corresponding Japanese function of F y Q and
gamma. With different and (in the sense of eq. 3.8), the two expressions are equal at a
gamma of 5.24. At higher gammas, the Japanese expression gives lower capacity, being 70% of
the author's ultimate at gamma of 17. In retrospect, this difference is consistent with the
ovalizing tendencies having been suppressed in the author's original local failure data base
(chords with end diaphragms), compared to the Japanese test set-ups.
The Japanese data was also used to understand interaction effects i.e., the extent to
which axial load in the chord member reduces its capacity to carry radial loads and punching
shear from attached branch members at a connection. Figure 3.27 shows both the data base and
the empirical expression used in the Code. In design, I U I would be taken as the AISC
utilization ratio of the chord member (with respect to its own axial and bending stresses versus

l-2f

Fig. 3.27. Data and criteria for interaction.

criteria based on yield). Where heavy wall joint cans are used at tubular connections, IU I will
often be less than 0.44, corresponding to no reduction for interaction effects. Note that the Code
 109

expression is symmetrical, as indicated by the solid line; while the data suggest that the
combination of tension in the chord and compression in the branches does not suffer any
interaction, concern over biaxial stresses in the saddle region of equal diameter cross connections
(which are often used in secondary cross-bracing without joint cans) led to the more conservative
approach. This expression for circular connections only lasted nine years in the Code (1975-
1983), being subsequently revised to account for the effects of gamma and loading type, as
described in Section 3.6.5 herein.
During the evolutionary period of the first two editions, several other incremental
revisions were incorporated into the Code:

(1) Flared connections, spherical connections, and overlapping connections, as

discussed in Chapter 6 herein.

(2) Design provisions for box connections (square and rectangular tubes) in 1975
punching shear based on yield line analysis for beta up to 0.8, and special rules for
larger beta and matched connections, in a format similar to circular sections,
including the same Qf as discussed in Chapter 5 herein.

(3) An extensive Commentary (ref. 59), first published as a separate document in 1977,
including extensive discussion of yield line theory, reserve strength factors, and
safety factors; also first time coverage of lamellar tearing and ultrasonic inspection
(see Chapters 7 and 8).

Fatigue provisions, as discussed in Chapter 4, did not change much during this early
period, but have changed more rapidly during the last ten years, as a large amount of research
related to North Sea applications became available.
f sine f.
_a b
acting Vp = +
~

Fig. 3.28. Summary of original
a b
allowable V Q f Q ft b a s i c Vn AWS criteria.
b u t n o t more t h a n 0 . 4 F w

CHS RH S

F
7
0.6
0.9

0.3 0.25 / S e e AWS \

large 3
3 d - .8333) 3(1 - 3) 10.5.1.3 J
\ f o r 3>0.8 /
1.0 f o r 1.0 f o r
small 3
3<0.6 3<0.5

1.22-0.5 U f o r U>.44
1.0 f a U<.44
Qf
Where U i s c h o r d u t i l i z a t i o n factor
due t o i t s own f a a n d f ^

l+sine
axial
2 sine

n i l ae n b e n c ,
1+3 sine
*Vy P ing by
use l e s s e r KD with resultant moment
+3 s9 i
f .b z o u t o f p l a n e b e n d i n g - f e6 - = " s i n
4 sin^e
110

3.6.4 Specialization by Connection Type and Load Pattern

Through 1977, the provisions for the design of circular tubular connections in both the
AWS Code and API RP 2A were virtually identical, both being based on the author's modified
punching shear criteria. Figure 3.28 summarizes the AWS Code provisions, for both circular and
box connections, as they existed prior to this time.
The reliability of the rules for circular sections may be evaluated by comparing the
allowable design load P-allow with the ultimate static test loads P-max, as shown in Figure 3.29.
The earlier correlation, based on the original 38-test data base, yields a safety index B (safety
factor expresses in standard deviations) of 2.7, given loads exactly at the design allowable. The
complete reliability analysis would of course include variability in the loads, as well as
dimensional errors.

TOTAL = 311 NON-OVERLAPPING JOINTS

CHARACTERISTIC STRENGTH
. S . F .

ASSUMING L0GN0RMAL DISTRIBUTION

MEDIAN SF = 2 . 2 8
55
s

P
.E5
/P
I.O
_ P R ETS E N

MAX ALL0W API/AWS RULES

Fig. 3.29. Reliability of the original criteria in comparison to test data.

When the corresponding notional probability of failure is cpnsidered in the overall

process of criteria selection for unmanned, redundant template-type offshore platforms in the
Gulf of Mexico, this leads to a reasonably optimum trade-off between first cost and future risk of
storm damage (ref. 35). For critical members whose sole failure would be catastrophic, or for
architectural applications where the local deformations which preceded the attainment of ultimate
capacity would be objectionable, the Code recommended that the allowable punching shear be
REDUCED by one-third. When this is done, the indicated safety index for tubular connections
increases to about 4, which compares favorably with that of other types of structural components.
After ten years of additional research worldwide, a much wider data base became
available (ref. 34), consisting now of 311 tests for non-overlapping simple tubular joints.
However, when the original criteria are compared to this wider range of test data, a less favorable
comparison results. If we take the new data at face value, the indicated risk is increased by an
 Ill
order of magnitude from the earlier correlation, as shown in Figure 3.29. However, yield
strength was included in the original correlation, but filtered out of the more recent ones by using
tensile coupons representing the actual test specimens. If we re-introduce this material bias (1.1)
and scatter (8% COV) into the structural reliability evaluation [see section 3.7.l(i)], the safety
index improves to 1.9, corresponding to a 3% failure probability. This is still worse than the
original correlation and was undesirable enough to prompt a re-examination of the criteria (ref.
61).
The approach taken was evolutionary rather than revolutionary. Rather than simply
applying a blanket increase in safety factor, an attempt was made to improve the reliability by
reducing the scatter. This was done by lowering the allowables for chronically weak situations
(e.g., cross connections in compression) and also by raising the criteria for systematically high
cases (tension, in-plane bending, small-gap K-connections, etc.). Finally, the test data itself was
critically re-examined.
To aid in the work, the OTC 2644 data base was fed into a computer file, along with
programs which would apply existing and various trial criteria to the test cases. After several
trials, new proposed criteria evolved, as discussed below. A new punching shear modifier
was introduced, in place of the old Q ^ e t a reflecting the influence of connection type and loading
pattern, as well as geometry.
(i) Basic Vp. The prevailing strong influence of the chord thinness ratio, gamma, on the
basic ultimate punching shear capacity, is shown in Figure 3.30. The exponent of 0.7 still
seemed as good as any. In making this plot, all other parts of the proposed design criteria were
applied to the test data, with the value of basic Vp required to match the test data being obtained.
A similar procedure was also applied to study the other parameters.

( -joints

T-joints
TEST
ULTIMATE X-joints
s o l i d symbols = tension

2fc i n v a l i d data point

\oo

chord thinness r a t i o R/T

Fig. 3.30. Effect of chord thinness ratio, gamma, with exponent of 0.7.

For the purposes of applying the Code rule that effective chord yield strength as used for
ultimate punching shear should not exceed two-thirds the tensile strength, a few of the high-yield
data points were adjusted as described in ref. 61.
112

Fig. 3.31. Data vs. criteria for cross connections.

(ii) Cross a n d Tee Connections. Values of reflecting the test data for cross
connections are shown in Figure 3.31. The safety factor of 1.8 has been backed out here, so data
should fall along the design curve. For compressive loading, the data provide an excellent fit to
the proposed criteria, which have been reduced to 70% of the original. For tension loads, the
existing criteria remain adequate.
Values of for T-connections are shown in Figure 3.32. As the original criterion were
derived mainly from compressive tests of T-connections, it is not surprising that it still fits the
data. For tension loads, the flat part of the design curve is increased by 40%. Although the
median of the data might have suggested a larger increase, there was concern over the ability of
heavy wall prototype joint cans to deform, and gain load after initial cracking, in the same way as
the small scale connections tested. The low test points denoted by asterisks failed by beam
bending of the chord, rather than local failure at the connection.
(iii) K-Connections The effect of gap on the strength of K-connections is shown in
Figure 3.33. Again the safety factor has been removed. The proposed gap effect a 30%
increase as the gap parameter zeta decreases from 0.15 to zero ~ is similar to that of several other
authors (refs. 34 and 62), and falls on the safe side of that given by Gibstein (ref. 63).
The more favorable behavior of K-connections depends on the inward load of one
branch offsetting the outward pull of the other, to offset the ovalizing effect. Similarly, the lower
performance of cross connections results from their double dose of ovalizing as compared to T-
connections. Following this reasoning, load pattern is just as important as connection
configuration in deciding which criteria to apply. API classification guidelines are shown in
Figure 3.34. Interpolation for the in-between cases is also indicated.
The large scatter shown in Figure 3.33 has provided incentive for continuing
refinements in the criteria. The effect of y (0.7 exponent instead of 1.0), prestress (Qf) and
yield/tensile ratio (2/3 rule) anticipate the later work of Kurobane (see Table 3.12).
 113

\ ^ Q without lower l i m i t

3\ o
, \

PROPOSED (API - 7 8 ) 0
^

"
EXISTING (AWS-75)
EXISTING = PROPOSED

TENSION COMPRESSION

-J 1 I ' I l i t ! - J I I l I ' . I
.5 1.0
Fig. 3.32. Tee connections.
A SMALL g<0.5
LARGE >0.5

t
r j

.2 .3 A

= GAP PARAMETER g/D

Fig. 3.33. Effect of gap in K-connections API-78 criteria.

114

5 0 % K , 507. T8Y
( (Interpolote)

^ r-

Fig. 3.34. API classification

guidelines.

(iv) Bending. Akiyama's criteria indicate that, in terms of punching shear, the strength
of a typical T-connection for in-plane bending is twice that for axial load. The proposed criteria
yield a similar increase and are compared to Gibstein's test data in Figure 3.35. For this load
pattern, it is hard to picture any beneficial effect of beta approaching unity.

5 *h
proposed
API - 78

existing
7"
AWS - 75

Fig. 3.35. In-plane bending of

.5 T- and K-connections.

For out-of-plane bending of T-connections, the original criteria were retained, although
Akiyama's criteria would indicate a slight increase. The bending criteria given for other types of
connections were largely based upon extrapolation ~ higher for K-connections and lower for
cross, similar to axial load behavior.
 115

The fully plastic interaction between axial load and bending moment is more favorable
than the linear interaction built into the formula for acting punching shear (eq. 3.15). This is well
documented for the behavior of tubular members (see Chapter 1) and has been extended by
inference for punching shear (ref. 65). A new punching shear modifier, Qp, was introduced to
account for this favorable plasticity effect. Figure 3.36 shows equation used and the intended
effect.

Fig. 3.36. Plastic versus linear interaction between axial load and bending.

Unfortunately, in practice, when this Qp is combined with interpolation on for the

different types of load involved, then an unintended synergistic bulge develops in the computed
failure envelope, as shown in Figure 3.37. Surprisingly, although Stamenkovic's test data (ref.
67) does confirm that the interaction is more favorable than linear, it is ambiguous as to whether
the intended or unintended failure envelope should apply. Physical reasoning and conservatism
would seem to favor the intended one.

On
1 9 7 8 - 8 2 API RP2A CRITERIA

Fig. 3.37. Stamenkovic' s results.

(v) I n t e r s e c t i o n L e n g t h K - F a c t o r . Figure 3.38 compares the K a derived from

geometry and statics with the residual effect of brace angle theta coming from the test data. The
two show opposite trends. Taking a hint from the adverse effect of footprint length (eta = beta/
sin(theta)) in box connections, Figures 2.31 and 2.32, the inappropriate trend of K a was corrected
by reducing for eta larger than unity. This reflected the long-enshrined status of K a and a
certain hesitancy about going overboard in an area not yet fully understood. Subsequent
revisions to the Code have removed K a altogether, eliminating the need for a correction to undo
116

its inappropriate effect and simplifying the computation of acting Vp, as well as puncturing the
illusion that punching shear gives an accurate, detailed picture of connection behavior.

zo
1X3

15

u_

1.0

o I 1 1 I I I I I 1 1

9* S O* 7 0* feO* 3 0" 20 10

Fig. 3.38. Effect of theta versus that given by K &.

f f
.. .. / a sin \ , b
a c t l gn Vp . -
allowable V p = Q f Q q Q p basic V p
but not more than 0.4 F y

Q as defined below
q
T Y PE OF L O AD

A X I AL A X I AL I N P L A NE OUT OF P L A NE
T E N S I ON C O M P R E S S I ON B E N D I NG B E N D I NG

O v e r l ap 1.5 p l us s ee 2 . 2 2 . c .2 2.5 1.0

0 < < 15 1.3 - 2 2.25 1.0

J O I N TS > . 15 1.0 2.25 1.0

>.6 u s e Qg i f l a r g er 2.25
%
> 1 .0 use l / o f a b o ve 1.5 m in 1 .0 m in
TYPE OF JOINT & GEOMETRY

e<.6 1.4 1.0 2.0 1.0

T &Y l a r g er o f 2.0
>6
1 . 4 o r Qg
J O I N TS
use 1 / f a b o ve 1.5 m in 1 .0 m in
> 1 .0

<.6 1.0 0.7 QB 1.4 0.7

but n o t o v er 1 . 0
CROSS &>.6 1.4 0 . 7 Qg
0 . 7 Qg
J O I N TS > 1 .0 use 1 /n 0 . 7 m in
1.0 m in
f a b o ve
V / D I A P H R A G MS s a me a s Qg
2.0
r 2 . 2 2 . c . 4 ( a ) t e n s i on but n o t o v er 1 . 4

For biaxial bending, interpolate based on contribution of inplane f.

y
and D
out of plane f. to bf. = f / . + 2f .f 2
bz y b y bz
For combined axial load and bending, interpolate based on the contribution
of each to acting V p.

Fig. 3.39. Summary of proposed criteria (API-78).

 117

(vi) Implementation. The new criteria proposed in 1977 are shown in Figure 3.39.
They were published into API RP 2A the following year (ref. 68). These were intended to be
"interim" criteria, as radical improvements by Yura, et al (ref. 69) were already being formulated.
However, they survived for seven years in RP2A and served the industry even longer in joint-
checking computer programs.
The reliability of these interim criteria is indicated by Figure 3.40. The changes were
successful in restoring the reliability of connections to what we previously thought we had.
Again, if the bias and scatter on material yield strength are included, the safety index improves to
3.2.

NOM 306 N o n - o v e r l a p p i n g
SF-1.8 + 34 Overlapping
Joints

.sinning l o g n o r m a l distributior
median SF = 2 . 5

* Points deleted:
3 Novikov K - j o i n t s
erroneously reported
2 Toprac t e e j o i n t s
w h i c h f a i l e d by
bending o f chord
1 o v e r l a p j o i n t which
violates 50 rule

hJ Fy adjusted for 2/3

yield/tensile rule

* * *

. ... J . . . A l L
.1 0.1 a* l,o 2
PMAX(test)/API 78(allowable)

Fig. 3.40. Reliability of the proposed criteria.

However, these changes were not incorporated into the AWS Code. They were judged
to be unnecessarily complex and would have upset the parity between circular and box section
criteria. There was also skepticism that raising the allowable stress really improved reliability.
The one area where previous criteria were seriously on the unsafe side ~ cross connections in
compression was fixed by strengthening the general collapse provisions to include a check of
total branch load, P, based on Yura's results (ref. 69):

2
sine = Tc F y o ( 1 . 9 + 7.2/3) Qp Q f (3.18)

This capacity applies to chords having a large extent of the same size wall thickness as
well as diameter. As shown in Figure 3.41, about 2.5 chord diameters are required for the
general collapse mechanism to develop this empirical capacity. For connections reinforced by a
"joint can" of increased thickness, but having a length less than 2.5 diameters, the Code provides
an interpolation formula which in effect adds the capacities of the joint can and the remaining
portion of thinner chord, within the 2.5-diameter region.

3.6.5 Comparison With W R C Data Base

Although AWS did not adopt the full scope of the API-78 criteria described in the
preceding section, it nevertheless formed an important stepping stone in the evolution of our
thinking. Rodabaugh (ref. 70) reviewed these criteria at the time they were first proposed.
Rodabaugh's WRC data base on static strength is essentially that of Reference 34, as shown
118

Fig. 3.41. Effective length of chord for general collapse.

previously in Figures 3.29 and 3.40, except that it has been edited to eliminate duplications. He
also notes that some of the tests include failure mechanisms other than the tubular connection
(e.g., branch yielding or chord beam bending); however, the connections would be at least as
strong as indicated, so the data was retained.
Rodabaugh also compared the API-78 criteria allowables with empirical ultimate
strength formulae of Kurobane (ref. 71) and Gibstein (ref. 63), which are given in Table 3.5.
These formulae are in the international total load format:

2
sine = T F y Q F(0, 0, J , e t c . ) (3.19)

TABLE 3.5 CORRELATION EQUATIONS FOR TUBULAR CONNECTIONS WITH AXIAL

LOADS IN THE BRANCH MEMBERS
2
REFERENCE & Limits
E q u a t i o n s f o r P / ( F y oT / s i n 0 )
JOINT TYPE
5 ( ? .25 < < .85
1.79 r a* ..iam - .> '#?
GIBSTEIN 10 < < 55

0 < g/D < 0.6

30* < < 90'

2
.19 < < .82
7.14 [1 + 3 . 7 9 ( | ) ] { l + 3.34(1 - 20.9 - 0.776(|) J
10 < < 51
KUROBANE _ 1 2
-21 < g/T < 66
[1 + -
u
tan (0.4 - 0.2 J)]
1
\ (1 + 0.0392 cos6 -
)
0.187 cos 8)
30" < e < 90"
2
-.69 < < .94
(1 + 0.254n - 0.339n )

.25 < < .85

7.5 6*^ (Compression)
GIBSTEIN
5 10 < < 20
T,Y (2.3 + 6) (Tension)
30* < 6 < 90*

2
.19 < < 1.00
KUROBANE 6.43(1 + 4.60 ) (Compression)

T,Y 8 < < 47

30* < < 90"

(2.4 + 183)sin6 (Compression) .25 < < 0.85

GIBSTEIN
X (2.4 + 18B)C osin6 (Tension) 10 < < 25

C - 1.7 for ultimate load, C - 1.3 for first crack

KUROBANE .19 < < 1.0

X [6.57/(1 - 0.810)]1 (Compression)
9 < < 47
 119

where "etc." includes such factors as chord stress utilization ratio ("n" in the table) and gamma
(whose very inclusion belies the basic assumption that the t-squared plate bending point load
model is universally applicable to line loads on cylindrical shells).
These ultimate strengths are compared to API-78 allowables in Figures 3.42 and 3.43.
If the two were in perfect agreement, we would expect to see the intended safety factor of 1.8
across the plot. For K-connections, we see that the API criteria are more conservative in the area
of small gap, preferring to be a lower bound rather than a best fit to widely scattered data (see
also Fig. 3.33). Comparing parts (a) and (b) of Fig. 3.42, we see that the effects of branch angle,
theta, are consistently treated. There are no overwhelming discrepancies due to gamma, as the
international criteria have been "fixed up" with a gamma correction to back out the excesses of
their basic thickness-squared formulation. T- and Y- connections tell a similar story, at least
according to Gibstein.

Fig. 3.42. Comparison of the correlation equations in Table 3.4 with API-78 allowable for K-
0
connections. (a) Gamma of 20, theta of 45 . (b) Gamma of 20, theta of 60 (ref. 70).

However, for cross connections, ultimate capacity does indeed follow the thickness-
squared trend, and the API-78 criteria becomes increasingly unconservative as gamma increases
beyond about 20, as shown in Fig. 3.43.
Nevertheless, despite these apparent discrepancies, Rodabaugh concludes that the API-
78 criteria as used for offshore platforms have indicated margins of safety which "equal or
exceed those used in other high-risk structures, e.g. bridges, high-rise buildings, and nuclear
power plant piping systems", provided the sampling of non-dimensional parameters (beta, theta,
gamma, etc.) is representative of the application.
120

API vs GIBSTEIN

.| l I
"0 .25 .50 .75 1.0

Fig. 3.43. Comparison of the correlation equations in Table 3.4 with API-78 criteria for cross-
connections.
3.7 FINAL FORM

Both the AWS and API standards published their present design criteria for circular
tubular connections in 1984. Although these retained the basic punching shear format of the
original criteria, the Q-modifiers which evolved during the early years, and the specialization by
connection type of the 1977 revisions, there were substantial changes in the details. API also
adopted alternative total load criteria in the international thickness-squared format. This
followed the work of Yura, as described below, as well as a second round of intensive committee
effort and debate covering several years.

3.7.1 Yura'sWork
The work of Yura, Zettlemoyer, and Edwards (ref. 69) provided the basis for this effort.
Indeed, access to their work in progress also influenced the 1977 revisions, which were originally
intended just as an interim fix.
(i) Ultimate Load Capacity. The first task was to re-examine the existing data base.
The WRC list of over 300 tests was pared down to only 137 simple , Y, cross, and gap
connections. Small specimens, with chord diameters less than 5.5-in (140-mm), were eliminated
because their relatively large weld fillets, and absence of restraint effects in fracture-controlled
tensile behavior, were not representative of the larger tubes used in monumental structures like
offshore platforms. Tests which did not report the actual yield strength of the specimens were
also eliminated. A deflection limit was also imposed such that the joint deformations at each end
would not increase the overall member axial deflection by more than a factor of five; thus, tests
without deflection data could not be used. Finally, tests with failure mechanisms other than the
connection were dropped. New data were added. The final data set came from references 24, 25,
27, 72, and 73-82.
 121

The resulting criteria, in the international thickness-squared format, are shown in Table
3.6.

TABLE 3.6 YURA' S ULTIMATE CAPACITY EQUATIONS FOR TUBULAR CONNECTIONS

LOAD JOINT EQUATION

- I T.Y
AXIAL
COMPRESSION "*"! DT.X p
- " i s r
( 3 4
-
+ 1
^
v , a

AXIAL ^. T.Y. Fy T
2

TENSION

AXIAL
1 ' DT,X f=,

COMPRESSION
AND
TENSION - p
- - - i s T
a 4 +
1
^
Q

dsnt (3.4 + 19/3)

2
IN-PLANE Fy T
BENDING [ 3 ALL JOINTS M - 0.8 d

M. - 0.8 d
2

sni (3.4 Q
OUT-OF-PLANE a T Fv
BENDING ALL JOINTS -4-^- + 70) FL
9

Data base, correlation means, and coefficients of variation are shown in Table 3.7, for
each connection type. In each case, the predicted ultimate falls on the safe side of roughly 85%
of the data. The safety indices given include adjustment for material strength bias and scatter,
1.10 and 8%, respectively (ref. 83). They also include a safety factor of 1.7, corresponding to the
traditional AISC safety factor for plastic design. The conditional safety index is for known loads
at static allowable stresses, as described earlier. Except for the X tension connections, this
represents a tremendous reduction in scatter, and consequent improvement in reliability, over the
earlier criteria. The Moses safety index (beta in ref. 83) reflects the large uncertainties in lifetime
storm loads on offshore platforms, as well as the use of the one-third increase in allowable
stresses for these loads; the result is an overall decrease in the indicated reliability, somewhat
obscuring the differences between connection types.

TABLE 3.7 YURA DATABASE AND SAFETY INDEX CORRELATIONS

-DATABASE CONDITIONAL MOSES
CONNECTION TYPE TESTS MEAN COV SAFETY INDEX BETA

X compression 37 1.067 7. 1% 6.5 2.3

X tension 19 1.411 42.7% 2.2 1.9
gap (excl Nakajima) 31 1. 161 17.7% 4.0 2.3
gap (incl Nakajima) 48 1.310 26.0% 3.3 -
in-plane bending 16 1.227 13.3% 5.4 2.6
out-of-plane bending 17 1.171 15.3% 4.5 2.4

The Q g factor and effect of gap in K-connections are shown in Figure 3.44. Nakajima's
data includes 45-90-degree K-connections, some loaded so that the gap region is tension. Large
deflection membrane effects here generally increased the ultimate capacity, but this was largely
offset by increased scatter in the resulting fracture-controlled failures. Neglecting chord prestress
in data reduction, and forcing the effect of gamma to conform to thickness-squared notions,
might also have contributed to the large scatter.
International criteria for K-connections and in-plane bending often apply a gamma
correction to the thickness-squared format (e.g., see Table 3.4). However, Yura, et al concluded
from a parametric study of the experimental data for cross connections, which truly do fit the
thickness-squared format without any gamma correction, that such a format is equally applicable
122

to all connection types. This weakness in the criteria was examined by the API committee (ref.
84), and is still being re-examined. Meanwhile, its use depends on the gamma range of the
application being similar to that of the data base.

4.0

3.0 h

NAKAJIMA
2.0 ALL OTHERS

(1.8 - 0.8 j )

2.0 3.0 4.0

g/d (gap/diameter)

Fig. 3.44. Effect of gap on axially loaded K-connections (ref. 69).

Figure 3.45 shows this range to be more representative of heavy wall joint cans in
offshore platform designs when the Nakajima data is excluded. For larger gamma the criteria
errs on the safe side; however, it becomes unconservative for gamma less than 10.

Fig. 3.45. Range of gamma for

K-connections in Yura database.
 123

(ii)Chord Stress Effects. Following the publication of Yura's initial work, a joint
industry project was formed to conduct additional experiments on chord stress effects,
represented in the Code by the Qf parameter. Figure 3.46 shows the results of these experiments
(ref. 85). Tests included axial load, in-plane bending, or out-of-plane bending in the branch
member (designated by the letters A, I, O, respectively), as well as axial loads or in-plane
moments in the chord (letters and M).
Al.17,08

BRANCH LOADING
- AXIAL
- OPB
- IPB
I I
-1.0 -0.5 0

Fig. 3.46. Yura's data for chord stress reduction effect, gamma = 25.

For the relatively large gamma ratio (25) of these experiments, the new data fell on the
unsafe side of the old Qf criteria, as shown by the solid line in the figure. A new parabolic form
was derived from the ultimate strength interaction of membrane thrust and shell bending
moment, representing the chord's own loads and local connection stresses in the chord,
respectively. As shown by the dashed line in the figure, this includes an empirical factor of 0.75
to fit the test data, as strain hardening, load redistribution, etc. mitigate the effects relative to
localized first yield.
It can also be seen that in-plane bending capacity of the connection is more severely
affected by chord stress than axial branch load capacity, while out-of-plane bending is less
affected. Out-of-plane bending results in high circumferential shell bending moments in a small
region near the saddle position; early yielding here re-distributes chord loads to other parts of the
circumference, leaving most of the local shell bending capacity available. In-plane bending
causes longitudinal shell bending along a broad front in the crown region, more closely fitting
the assumptions of the parabolic model and leaving less opportunity for load re-distribution.
Axial branch loads affect both crown and saddle regions and experience intermediate interaction
effects.
In Figure 3.47, the effect of gamma ratio is examined, to reconcile the new data with
earlier findings. Note that 0.030 gamma is the same as the 0.75 factor in the previous figure,
with tests having gamma of 25. Washio's data (ref. 20) at gamma of 16, basis for the old Qf, is
124

now consistent with this plot. Data of Kurobane (ref. 86) and Nakajima extend the correlation to
K-connections as well as cross connections.

O WASHIO DT / = I6.0 X=5I.6

WASHIO y = l 6 . l * = 35.9

KUROBANE /=ll.3* Y - 25.8

% Q( BASED ON MAXIMUM
CHORD COMPRESSION

I I I I I I I -
4 8 12 16

U 2
7

Fig. 3.47. Effect of gamma ratio on chord stress interaction, for branch axial loads (ref. 85).

In implementing these findings, the lambda factor (0.030 for axial loads) was adjusted
upwards (to 0.044), reflecting the more severe interaction for in-plane bending, and reduced (to
0.018) for out-of-plane bending, reflecting the influences seen in Fig. 3.46.
(iii) Load Interaction. Yura subsequently studied the effects of load combinations in
the branch member (ref. 87). In addition to the old Qp factor discussed in connection with
Figure 3.36, various equivalent forms of interaction between axial load and resultant bending
moment which can be derived from the plastic behavior of circular sections are given below:

= cos /\ (3.20)
M P
ult \ u l t /

+ 1 - cos \ = 1.0 (3.21)

M P
ult \ ult /

+ ^ arcsine / )= 1.0 (3.22)

p M
ult \ ult /

where

M + 3 2 3
= ^ i n - p l a n e ^out-of-plane < >

Yura's data are shown in Figure 3.48. Although Yura found a better fit to the empirical
expression of Fig. 3.48(a), the data also fit the arc-sine expression reasonably well, as shown in
 125

Fig. 3.48(b). API has continued to use the arc-sine expression as discussed earlier in connection
with Stamenkovic's data and Figure 3.37. Note that the interpolation difficulties discussed
earlier are avoided by keeping axial and bending calculations completely separate until the
interaction ratios are combined in the arc-sine equation, rather than mixing the two in both acting
and allowable punching shear.

(*-.(_. 1
\MU/OPB \ M U / I BP

0. 2 0.4 0.6
(M/Mu^(M/Mu)|p3

Fig. 3.48. Yura's data for combined branch loads: A = axial; I = in-plane bending; = out-of-plane
bending, (a) Comparison with Yura's empirical expression, (b) Comparison with arc-sine
equations.

The AWS Code uses the numerically similar expression of Equation 3.24, after Chen,
which is less likely to cause computer errors (as happens when the argument of arc-sine exceeds
unity).

1.75 = 1.0 (3.24)

p E
( u i j ult

3.7.2 ASCE Review

In 1981, the ASCE Committee on Tubular Structures sponsored a session at the
society's spring convention. Their progress report (ref. 88) reviewed both the API-78 criteria
and the proposed new criteria based on Yura's work, although the committee avoided taking a
position on its adoption into Codes. In addition, papers from around the world (refs. 89-92) were
brought before an audience of American engineers. A meeting of IIW s/c XV-E (Design of
Welded Tubular Structures) at the same venue further stimulated the exchange of ideas. These
two events influenced the development of a new consensus on design criteria.
An Appendix to the ASCE report compared various data bases of tubular connection
tests against the static strength criteria of Marshall and API (refs. 61 and 68), Gibstein and DNV
(refs. 63 and 93), Akiyama (ref. 64), Wardenier (ref. 72), and others, as well as the new proposals
of Yura, et al. Figure 3.49 shows the comparison of Yura's criteria against his data base, for all
connection types. A reliability calculation, following the load and resistance factor design
(LRFD) concepts of Galambos (ref. 94), is included in the figure. This leads to a partial safety
factor on resistance of 0.85 (the original paper neglected material bias and uncertainty, and got a
factor of 0.79). The target safety index of 3.0 was chosen to be consistent with the practice for
member sizing, rather than the higher values usually recommended for connection material (e.g.,
bolts or welds), in that selecting the joint-can thickness does not include the same workmanship
 126

issues as the latter. For statically-loaded onshore structures having 25% dead load and 75% live
load (occupancy), the corresponding allowable stress safety factor would be 1.76, using the load
factors of the draft AISC-LRFD code (ref. 95).
NOMINAL
DATA M A T E - COMB- S F = I.O
BASE RIAL INED
50 MEAN B I A S 1.23 1.10 1.75
COV .27 .08 .28
FOR 0 S= 3 . O
0= 0.85

j 40
2
or
< = e x p ( - a / ? sV R) ^ .
"fl
=37
INCLUDING NAKAJIMA RESULTS
&
all k
30

= 1.2

20 = 1.6

10

I I I I I
O.I 0.2 10
0.5P 1.0 2
TEST / YURA

Fig. 3.49. Correlation of Yura's design criteria with his data base.

Table 3.8 shows the criteria as presented by the ASCE report. Yura's criteria has been
reduced to the traditional American punching shear format, using the K-free thin wall
approximation discussed earlier in Section 2.4.2. This format has also been made consistent with
the existing AWS criteria for box connections. It features a common basic allowable punching
shear, which includes a safety factor of 1.8, modified by several Q-factors whose values tend to
start out close to unity. In terms of punching shear or line load capacity, the strength for in-plane
bending is seen to be about three times that for axial loads (the use of the factor was just a
convenient numerical approximation, and has no theoretical basis).

TABLE 3.8 ASCE CRITERIA IN PUNCHING SHEAR FORMAT

P l u s 1/3 i n c r e a s e
A c t i n g V p = s i n ( f f l + t " b) allowable p = Q pQ qQ f ^
where appl Icable

Note: Q p, Q q. Q g a n d Q p Plastic reserve factor Q

may conservatively

be t a k e n as u n i t y Chord utilization factor Qf

.*. a l l o w a b l e V n
0.6
Load type & geometry factor Qq i s given in table below

Notes to table: TABLE FOR V A L U E S T Y P E O F LOAD I N BRANCH MEMBER

"1.8 - 0 . 8 J f o r g < d
OF Q q AXIAL
TENSION
AXIAL
COMPRESSION
IN-PLANE
BENDING
OUT-OF-PLANE
BENDING

Qg
TYPE OF JOINT & GEOMETRY

1.0 f o r g > d CIRCULAR

K-JOINTS
f or >6 0
B ( l - 0 . 38 3 3 g ) CIRCULAR
li* ) .(! ) (0.37 + ^ )
1.0 f o r < 0.6 & JOINTS

CIRCULAR
( 0 . 7 + ^ i ) q 6
CROSS J O I N T S

S T E P P E D BOX 1.0 f o r < 0.5

CONNECTIONS 2
? : ^ f o r > 0.5 ( s e et e x t f o r >0.8
^ 1 - /
 127
3.7.3 API Implementation
API implementation of the Yura criteria presented both the traditional American
punching shear format and the international thickness-squared total load format (ref. 96). In the
punching shear format, the acting punching shear is:

acting = fn sin0 (3.25)

where fn is the nominal stress in the branch member, either axial or bending (treated separately).
This should not exceed the allowable punching shear,

F 0
allow V = Q Q . V- (3.26)
f q .67

with Qf being the reduction factor for chord stress effects as discussed previously [section
3.7.1(H)], and Q Q representing the effects of connection geometry and load pattern as defined in
Table 3.9.

TABLE 3.9 API-84 VALUES OF Q q

TYPE OF LOAD I N BRACE MEMBER

Axial Axial In-Plane Out-of-Plane

Tension Compression Bending Bending
EOMETRY]

overlap 1.8 plus see 2.5.5c.2

gap (1.10+OL20) QK

0
T&Y (1.10 + 0.20) (3.72 + 0.67) (1.37 + 0.67) (fy
3
E-

OF JOtt

w/o diaphragms (1.10 +0.20) (0.75 + 0.20) (fy

CROSS
w/ diaphragms (1.10 + 0.20)
>- per 2.5.5c.4

Q 21 forfl>0.6 Q = 1.8-0.1g/Tfory 2 0
( 1 3 3 >
-* Q, = . 18 - 4 K / D f r 7 > 2 0
= 1.0 for < 0 6 but in no case shall Q g be taken as less than 1.0.

Note that the old Qp interaction between axial load and bending in the branch member
has been eliminated, being replaced by the arc-sine interaction equation, as discussed earlier
[Section 3.7.l(iii)].
Differences in the API and ASCE tables for reflect further deliberations by the API
working group. Zettlemoyer (ref. 97) revised the treatment of gap effects in K-connections to
provide a more all-encompassing lower bound to the data, following suggestions of Kurobane
and Marshall. He also examined the residual effect of gamma, as shown in Figure 3.50. Roussel
(ref. 98) discovered unexpectedly large discrepancies between the punching shear and total load
formats, traced to the ASCE having used the thin-wall approximation. However, rather than use
a more accurate calculation of acting punching shear (i.e., Equations 2.8 and 2.9), the allowables
were increased to reflect typical branch member t/d ratios. Finally, the safety factor was reduced
from 1.8 to 1.7, further increasing the values; this change was primarily justified on the basis
of a "brute force" calibration: making the average joint design by the new lower
128

bound criteria close to the average of the old API-78 criteria, thereby avoiding a blanket increase
over designs which had been giving satisfactory service.

P T = Test capacity

(a) ' (b) sin

governs governs
(a) Q 0 = 1 . 8 - 0 . 1 g/T 7 < 20
(b) = 1.8-4g/D7 >20
but in no case shall Q g be taken
at less than 1.0.

.5 I I I I I
10 20 30 40 50
7
Fig. 3.50. Variation in K-connection axial capacity, tests versus API-84 criteria, as a function of chord
flexibility parameter gamma (ref. 96).

While the logic for these fixes now seems obvious, they were initially the source of
much confusion and controversy, as the problems often showed up as unexplained discrepancies
between trial design calculations being done with the various proposed formats. Other
controversies, having little to do with the substance of the Yura criteria, also worked to retard
their adoption; these included the traditional rule (adopted from AISC) that connections should
develop at least 50% of the member capacity, and concurrent work on allowable cyclic stresses
for fatigue in shallow water (ref. 100). The 50% rule was substantially relaxed, consistent with
the notion that plastic deformation of the joint-can provides a source of ductility not present in
traditional non-tubular structures. A controversy involving equal diameter cross connections
(i.e., the paradox that API criteria shows them to be weaker in tension than in compression)
continues to this day.

3.7.4 AWS Implementation

Yura-based design criteria were published in the 1984 edition of the AWS Code. Table
3.10 shows the 1984 AWS punching shear design criteria, alongside the ASCE format from
which it was derived. Instead of a separate equation for each type of connection, there is a
unified equation covering all the axial load cases, and another covering bending. The unifying
parameter is alpha, now defined as an ovalizing parameter, supplanting the original usage, and
also shown in the Table. For the time being, we shall regard alpha as somewhat arbitrary,
following a convention established for elastic stress concentrations in References 99 and 100. In
Chapter 6, we shall see how alpha was derived, and how it allows the extension of these criteria
to multi-planar tubular connections (ref. 101).
 129

TABLE 3.10 AWS-84 PUNCHING SHEAR CRITERIA

ACTING V = s i n 0 (f or f v)
a b
F
ALLOWABLE V = Q xQ
x
6
q f 0,

TYPE VALUE OF Q q FOR AXIAL LOADS

OF
JOINT ASCE AWS

/1.7 .18 \ 0.7(o - 1)

T/Y

X
( ' *

IPB ( 3,14 + 0.57/ )

( M + M J q 1.2(a-.67)
H
\ J P
OPB / 1.16 + 0.57/ ^Qp

ALPHA
l+0.7g/d
1.8

1.8
as g - * - 0

as . 0
T/Y
X
1.7
2.4
o
IPB 0.67
OPB 1.50

In Figure 3.51, we see how the strength of K-connections increases as the gap between
branch members decreases. The relevant gap is between branch members whose radial loads and
chord ovalizing effects tend to offset each other. The strength decreases to that of T- and Y-
connections as the gap gets as large as one branch diameter. Like API, AWS also sought a more
all-encompassing lower bound than the original Yura criteria (compare with Fig. 3.44).

3.Q
YURA DATA BASE
EXCL. NAKAJIMA

2.0U
BP
.oo
9>
1.0
USING a=l+0.7g/d
(ALSO DEPENDS ON )
Fig. 3.51. Effect of gap on the
strength of K-connections
(AWS criteria).
1.0 2.0
g/d
130

Trends of connection efficiency, versus alpha, are shown in Figure 3.52 for various
American criteria, as well as the IIW design strength (ref. 109). E y is defined as the ratio of
allowable punching shear to allowable tensile stress. Note that the horizontal axis has dual
labelling, connection type and the corresponding alpha. Connection efficiency decreases as the
connection type progresses to ones with more ovalizing; i.e., increasing alpha. All the criteria
are quite similar in their treatment of DT or cross connections as opposed to T- and Y-
connections: capacity diminishes as the ovalizing effect is doubled.
to

7= 14

= 0.5

1.0 1.14 1.7 2.4

OVALIZING PARAMETER 0.

Fig. 3.52. Variation of efficiency with connection type.

The criteria differ in their treatment of K-connections. Although both API-84 and
AWS-84 both took a more conservative approach than the original Yura (ASCE), this is partly
obscured in API's case by their reduced safety factor increasing the nominal efficiency across the
board. The earlier API-78 criteria was even more timid about giving increased capacity to K-
connections, with the plateau corresponding to the original 1972 criteria (note this comparison is
for a particular gamma ratio).
Another cut through the criteria is given in Figure 3.53. is the punching shear
modifier for geometry and load pattern, with ASCE, API, and AWS all using the format of
Equation 3.26. Note that starts out with a minimum value of approximately unity, and
increases as the connection type or load pattern becomes more favorable. The more conservative
treatment of K-connections by AWS is again apparent. Also note the paradox created by the
sudden introduction of Q ^ e t a for cross connections in the ASCE (also API) criteria: at beta of
unity, they become stronger than T-and Y-connections! The AWS avoids this paradox by
 131

Fig. 3.53. Effect of geometry

and load pattern.

gradually phasing in Q b e t a, using an exponent which varies from zero for zero-gap K-
connections, to unity for cross connections. This results in equal-diameter connections, which
deliver loads tangentially to the chord, being relatively immune to ovalizing effects. Similarly, in
bending, the exponent on Q ^ e t a is zero for in-plane bending and unity for out-of-plane bending
(Table 3.10).
A histogram comparison of the AWS-84 criteria with the WRC data base (ref. 70) is
given in Figure 3.54. The correlation is reasonably tight. The mean safety factor and B s (safety
index for known loads at the static allowable) were initially calculated without considering bias
and scatter in material yield strength. The WSD (working stress design) B , including material,
is comparable to the 2.7 obtained with the original criteria and its original data set, but inferior to
the 3.0 obtained for the API-78 criteria in Figure 3.40. However, this calculation seems to have
been penalized by the bimodal data set, as separate comparisons with the compression and
tension subsets of data are more favorable (i.e., higher B values). For LRFD (load and
resistance factor design), the partial safety factor on resistance comes out higher than for the
original Yura calibration (Fig. 3.49) for two reasons: First, the AWS criteria are more
conservatively drawn, at least for K-connections. Second, the larger WRC data set includes a
larger number of more favorable small scale tests.
The AWS criteria do not distinguish between tension and compression loads, using the
same capacity for both. The apparent large safety factor and safety index shown for tension tests
are biased by the large number of small tubes in the WRC data base. If only connections with
chord thickness of 0.25-in (6-mm) or larger are considered, the mean safety factor drops to 3.7;
for of 0.5-in (13-mm) or more, the safety factor is only 2.2. Considering the singularity (sharp
notch) at the toe of typical welds, and the unfavorable size effect in fracture-controlled failures,
the AWS approach seems more prudent than API's bonus for tension loads.
132

3 0 6 JOINTS IN WRC DATA BASE

(NON-OVERLAP)

MEAN S F = J . 9 5 J E XU C ^

WSD B" = 2.1 @ SF = 1.

LRFD
(COV =*.43)

JOINTS
T&Y COMPRESSION
X COMPRESSION
1
MEAN SF = 2 . 4 4 |
MAT 1
= 3.4
SF = 1.

LRFD = 0 . 9 5 @ B s= 3 . 0
(COV .27)

I s T&Y TENSION
\
1 / X TENSION
MEAN SF = 5 . 1 4 1
EXCL. MAT 1
= 5.0
4 WSD B5 = 5.2 1 SF = 1.8

LRFD > 1 . 0 @ 5= 3.0

(COV * .33)

/
test allowable

Fig. 3.54. Correlation of AWS-84 criteria with WRC data base.

The AWS Qf reduction factor for chord stress effects follows Yura's work and the API
criteria, as discussed previously. Treatment of the interaction between axial load and bending
moments in the branch members also closely follows the work of Yura and API, with Equation
3.24 approximating the arc-sine expression. Due to concerns raised by the scatter in Figure
3.48(b), in which the end point are experimental single-load capacities, a direct comparison
between the Yura tests and the AWS criteria was made, as shown in Table 3.11. As before, the
data are tightly clustered, and mostly on the safe side of the nominal ultimate (safety factor
removed). The Phase One tests (ref. 85) support the Qf expressions. The Phase Two tests (ref.
87) address the arc-sine interaction and vectorial combination of bending moments. While test
A 0 4 is a bit worrisome, test A I 0 1 6 , which includes the same loads plus in-plane bending, is
conservatively predicted.

3.7.5 Comparison With Kurobane's International Data Base

There are two formats in general use for the design of tubular connections. One is the
punching shear format, as extensively discussed herein. The second is the Limit State Design
(LSD) format. It is based on the the theoretical expression for localized shell bending failure
under the action of a point load, P, given by:
 133

ultimate sine (3.27)

yo

TABLE 3.11 COMPARISON OF AWS-84 CRITERIA WITH YURA'S INTERACTION TESTS

P-TEST/
TEST I . D . * P-AWS (ULTIMATE)

PHASE ONE:
Al 1.24
AP2 ( 1 ) 1.15
AP2 ( 2 ) 1.19
AP5 ( 1 ) 1.21
AP5 ( 2 ) 1.20
AM6 ( 1 ) 1.26
AM6 ( 2 ) 1.19
17 1.52
IP12 1.44 * KEY :
IM11 1.41
08 1.00 A BRANCH AXIAL LOAD
0P9 0.98 I BRANCH IN-PLANE BENDING
0M11 1.03 0 BRANCH OUT-OF-PLANE BENDING
CHORD AXIAL LOAD
PHASE TWO: CHORD MOMENT
AI3 1.43
A I 3 (PARENTHETICAL) 1.62
AI17 1.71
A04 0.92
A013 1.16
1014 1.20
1015 1.07
AI016 1.07
AI018 1.36
AI019 1.26

The theoretical value for Q u is 4.0 for flat plates; however, the actual value is found to
be a function of connection type (or load pattern), connection topology (as defined by non-
dimensional parameters beta, eta, zeta, and theta), as well as chord thinness ratio (gamma) and
the yield-to-tensile ratio of the chord material. Qf is the de-rating factor for the effect of self
loads in the chord, as discussed previously.
Table 3.12 gives several different design criteria in LSD format, for K-connections in
circular hollow sections. The API load format is derived from the work of Yura, as are the AWS
expressions. The latter are converted back from the punching shear format, using the K-free
thin-wall approximation. American criteria are compared to the 1981 criteria of Kurobane (refs.
89 and 102). These are similar to (but not exactly the same as) those being included in the IIW
s/c XV-E recommendations (ref. I l l ) , which have also been adopted by Eurocode 3.

TABLE3.12 EMPIRICAL EXPRESSIONS FOR THE STRENGTH OF CIRCULAR K-

CONNECTIONS: ULTIMATE AXIAL CAPACITY AS * SIN (THETA).

AWS (1984) API (198A) KUROBANE (1981)

2 2 2
T F y ' Q f * [ 6 n 0 Q q] T F y ' Qf * [ q u] T F y ' Qf (f(B)-f(g,T)-f(Y)-f(8)*f(a)]

2 2 2
Q f - 1 - 0.03 Q f - 1 - 0.03 Q f - 1 + 0.3U - 0.281U

Q u - (3.A + 193) Q g f(0) - 5.0(1 + A.688)

1 7
O.OOSKf) '*
n ( - 1 + 0.7g/d ( 1.8-0.lg/T for 1 20
f(g.T) - 1 +
Q6 ) - 1.8-4g/D for > 20
1 + exp(0.37| - 0.853)
)>_ 1.0

Y 0.182

f(Y) - o>
1.0 for 8 <. 0.6
2
f(8) - 1 - 0.326 cos 6
/F \ -0.730
use F
2
< -
y - 3 ult
use F < 4

y - 3 ult
134

In all these criteria, the bracketed terms at the top of the Table correspond to Q u in
Equation 3.27, and are detailed in the lower parts of the Table. The expressions on line (1)
reflect the influence of diameter ratio (beta), with the values of Q u ranging from 3.4 to 5.0 for
small beta, up to around 20 as beta approaches unity. The expressions on line (2) produce
strength increases up to 1.8-fold for the effect of very small gap. Kurobane's line (3) indicates
that the strength of K-connections is not really proportional to chord thickness-squared, as
implied in Equation 3.27, but varies as the 1.8 power of thickness. Other authorities also indicate
exponents less than two, ranging from 1.5 (refs. 27 and 82) to 1.7 (refs. 47 and 103). Line (4) is
a minor correction to the notion that only loads perpendicular to the chord need to be considered,
less than 15% for practical brace angles. Line (5) shows a significant detrimental effect for very
high yield-to-tensile ratios in the chord, reflecting the importance of strain hardening as a source
of reserve strength.
In all, Kurobane shows separate design expressions for 24 different categories of
connections involving circular tubes. Figure 3.55 (ref. 104) shows comparisons of his line (1)
expressions for X-connections, T&Y-connections, and K-connections, versus the data base from
which they were derived. This is a very large database, with 581 tests. AWS-84 criteria are also
plotted in comparison to Kurobane's criteria, with typical gap and gamma assumptions as stated
in the figure. The AWS criteria are intended to provide a lower bound to the test data, while
Kurobane provides the unbiased best fit and scatter factor suitable for deriving a Level-II
reliability-based design code (e.g. LRFD). Aside from this difference, both sets of criteria follow
the trends of the data, as a function of diameter ratio, beta. All three connection types show
similar strengths for very small beta, and for beta of unity. For the mid-range of beta, where
chord ovalizing is most influential, the strength of X-connections sags the worst, K-connections
do not sag at all, and T&Y-connections exhibit intermediate behavior. This confirms the gradual
phasing-in of the Q ^ e t a effect in the AWS criteria. The treatment of outliers was previously
described in Figure 3.40.
The effect of gap variation is not treated in Figure 3.55(c), but can be seen in the
comparison of AWS vs. IIW in Figure 3.52. Here, the IIW design strength incorporates a safety
factor of 1.1 (resistance factor of 0.9) relative to the 95% confidence characteristic ultimate
strength. As a result, the AWS criteria appear to be less conservative than they really are.

3.7.6 Summary
Earlier sections of this chapter traced the evolution of American design rules for
structural connections of circular tubes. The present Section 3.7 describes the development of
their present form, as well as validation against the international data base. Although the
American rules do not match the international rules exactly, they are reasonably consistent with
the underlying test data. These rules cover a wider range of design situations with a simpler set
of equations, apparently providing greater economy (Figure 3.52) with acceptable safety (Figure
3.54), and there is no overwhelming incentive for American designers to conform to the IIW
hegemony.

3.8 DESIGN CHARTS

This section reviews present-day AWS punching shear criteria (ref. 105) for the design
of tubular connections, using circular sections. A procedure for using the charts in the design of
simple tubular trusses is described. Practical suggestions for the overall design strategy are
included.
 135

KUROBANE
1/(1-0.8l2d/D)

6
e
2 s
/(1-0.8l2d/D)

w 5

Q in OBSERVATION
ZE

CO k
CO
LE

3
= 2.4 6.54

CO
2
w 76 TEST RESULTS FOR

1
X-JOINTS COMPARED TO (a)
CRITERIA

0.2 O.A 0.6 0.8 1.0

d/D
KUROBANE
l+A.9Md/D)
SC
6 2 2s


[l+i*.9Md/D) ]e
W
3

5 OBSERVATION
CO

> CO OUTLIER
t CO
J w 6II3Q
3
H
ON

NDS = 1
6.36
CO

2
W
7 5 TEST RESULTS FOR
1 T&Y-JOINTS COMPARED TO (b)
CRITERIA

0.2 0.1* 0.6 0.8 1 .0

d/D
KUROBANE
l+i*.67d/D
COMPARISONS FOR
SCI
(l+^.67d/D)e SMALL GAP g=T
TYPICAL 7=20

W OBSERVATION
C^i

Q CO
W OUTLIER
CO
CO
CD
s2 6II6Q
NDS

co

5.0 f(g,T)


430 TEST RESULTS FOR
K-JOINTS COMPARED TO
CRITERIA (c)
0.2 Q.k 0.6 0.8 1.0
d/D

Fig. 3.55. Comparisons of Kurobane data base with Kurobane ultimate strength and AWS-84 design
criteria, (a) X-connections. (b) T&Y connections, (c) K-connections.
136

3.8.1 Introduction
Since the publication of design charts for hollow structural section trusses (ref. 106) by
Packer, et al, which was based on CIDECT and IIW criteria, the author received numerous
requests for similar charts based on the AWS D l . l Code, for use by American designers. The
present section does just this, following the logical step-wise design procedure of Packer.
However, the actual format of the charts follows that of Wardenier (refs. 107-109) because his
format is conceptually compatible with punching shear, permits direct comparison of AWS and
international criteria, and requires fewer charts.
The charts give the maximum punching shear efficiency, E y , in terms of the above-
described non-dimensional parameters alpha, beta, and gamma, where...

a l lw o v
= maximum a l l o w a b l e p u n c h i n g s h e a r s t r e s s = p ^ 28)
m a i n member a l l o w a b l e t e n s i l e s t r e s s 0.6 F y o

Where self-loads in the main member are present, this maximum allowable punching shear must
be de-rated by the Qf factor, for which charts are also given. Note that safety factors have been
included in both the numerator (1.8) and denominator (1.67) in the above equation.
The design procedure, presented below, uses E y from one chart, Qf from another, and a
very simple calculation. For those interested in making comparisons, E y is equivalent to
Wardenier's non-dimensional strength parameter zeta, which he uses in a total-load limit state
design format, rather than in terms of punching shear and working stress design. For the
procedures given herein, however, the choice of LSD or WSD makes surprisingly little
difference.
In either format, the connection efficiency, Ej, is simply:

Ev Qf F
= ' -l (3.29)
F
J t . in0
S y

where F y o is the specified minimum yield strength of chord (main member), and F is that of
branch member, and (t/T) is the branch/main thickness ratio (tau). Where branch and chord are
of the same material, the ratio of yield strengths may be omitted. In limit state design, joint
efficiency is the design ultimate capacity of the tubular connection, as a fraction of the branch
member squash load. In working stress design, joint efficiency is the branch member nominal
stress corresponding to the tubular connection reaching its allowable punching shear (or other
measure of capacity), as a fraction of the tensile allowable stress. Connections with 100% joint
efficiency develop full yield capacity of the attached branch members.

3.8.2 Charts for Circular Sections

Punching shear efficiency, E y , is shown as a function of alpha, beta, and gamma in
Figures 3.56 through 3.60. The choice of which figure to use depends on the connection
configuration and load pattern, as reflected in the ovalizing parameter alpha. For comparison,
API criteria for gamma of 14 are superimposed on the charts. The corresponding HW-based
charts can be found in References 107-109, and were compared to AWS in Figure 3.52 (for D/T
of 28).
For all the c o n n e c t i o n t y p e s , s t r e n g t h d e c r e a s e s as g a m m a ( m a i n m e m b e r
radius/thickness) increases. For very stocky members, gamma less than 8, punching shear
strength can approach the shear strength of the material. Corresponding to a shear allowable of
 137

Fig. 3.56. Punching shear

efficiency of axially loaded K-
connections.

Fig. 3.57. Punching shear

efficiency for axially loaded
T&Y-connections.

Fig. 3.58. Punching shear

efficiency for axially loaded
cross-connections.
138

40% of yield, versus a tensile allowable of 60%, the punching shear efficiency reaches a plateau
of 0.67. Members this stocky include standard weight pipe smaller than 3-inch nominal
diameter, extra strong of 6-inch and less, and all the double extra strong sizes listed in the AISC
Manual (ref. 110, page 1-89).
In axially loaded K-connections, Figure 3.56, the inward load from one branch member
is balanced by the outward load carried to another nearby branch member in the same plane, so
that the net ovalizing effect is cancelled out. Increasing the gap, g, between branches causes them
to act more like isolated members, and the efficiency decreases towards that of T- and Y-
connections.
For gaps between 0.2 and 0.7 times the branch member diameter, interpolate between
Figures 3.56 and 3.57. The relevant gap is between branch members whose loads balance.
Similarly, load patterns which do not balance out increase the ovalizing and decrease the strength
towards that of T- and Y- connections.
In isolated T- (perpendicular) and Y- (angle) connections, the radial loads delivered by
the branch member end up being carried as beam shear in the main member, and a full dose of
ovalizing is felt. As shown in Figure 3.57, this decreases the efficiency, particularly in the mid-
range of beta (ratio of branch diameter to chord diameter).
In X- (cross) connections, branch members pushing on opposite sides of the main
member produce a double dose of ovalizing, leading to very low strengths, as shown in Figure
3.58. Connections which may not look like cross joints can also have this severe crushing load
pattern for example, where the end post of a truss delivers load to the support bearing. In
space structures, it is possible for the same main member to have a cross joint in tension in one
plane, and a cross joint in compression in another plane at the same location, producing a
quadruple dose of ovalizing and strengths substantially lower than shown here. For such multi-
planar joints, designers should refer to the AWS Code, and the Commentary, which gives a
formula for computing alpha (also see Chapter 6).
Figures 3.59 and 3.60 give punching shear efficiency for connections with bending in
the branch member, for all geometric configurations. In-plane bending causes deflections in the
plane defined by branch and main member centerlines; circular section connections are strongest
for this type of loading. Out-of-plane bending in the branch member produces torsional loads in
the main member, and lower joint efficiencies.
Limits of applicability for Figures 3.56 thru 3.60 include the full plotted range of
parameters alpha, beta, and gamma, and the following:

- gap joints (see Chapter 6 for overlapping joints)

- uniform thickness circular sections for both branch and main members
- compact sections, both branch and main members (D/T < 3300/Fy)
- ductile mild steel with tensile/yield ratio of 1.5, or notch-tough high strength steel
with effective yield taken as 2/3 tensile strength
- matching weld metal and prequalified weld details of AWS Dl.1-90 Figures 10.9 to
10.13 (also fillet welds, which match the strength of the members joined, as discussed
in 2.5.2).

3.8.3 De-rating Factor

In most structures, the main member (chord) at tubular connections must do double
duty, carrying loads of its own (axial stress fa and bending f^) in addition to the localized
loadings (punching shear) imposed by the branch members. Interaction between these two
causes a reduction in the punching shear capacity, as reflected in the Qf de-rating factor.
 139

F i g . 3.59. Punching shear

efficiency for in-plane bending.

18

22

26

30

F i g . 3.60. Punching shear

efficiency for out-of-plane
bending.

0 .25 .50 .75 1.0

DIAMETER RATIO /3

Figure 3.61 gives Qf for circular connections; this was updated in 1984 based on Yura's
tests of T- and X-joints. In-plane bending experiences the most severe interaction, as localized
shell bending stresses at the joint are in the same direction and directly additive to chord's own
nominal stresses over a large part of the cross section. For chords with high gamma (very high
D/T), P-delta effects due to nominal stress further reduce the capacity for localized shell stresses.
Out-of-plane bending is less vulnerable to both these sources of interaction, as high shell stresses
only occupy a localized part of the cross section, and are transverse to P-delta effects. Axially
loaded joints of the types tested so far exhibit intermediate behavior (although the gap region in
K-joints might be expected to behave more like in-plane bending, i.e., it will be fairly sensitive to
chord prestress).

3.8.4 Other Failure Modes

Several possible failure modes other than punching shear (synonymous with local
failure in AWS terminology) must be considered in a comprehensive design check. In the present
work, most of these are either included in the E y charts, or covered by limits on applicability, as
discussed below.
140

General Collapse The appropriate limits are included in the charts, for ovalizing
failure of circular sections, for members of uniform thickness. Where "joint can" reinforcement
of limited length is provided, further guidance can be found in the Code. Other forms of
reinforcement should be investigated as structures in their own right.

(a)
0 .25 .50 .75
CHORD UTILIZATION 0~2

(b)

(c)
Fig. 3.61. De-rating factor Q f .

Uneven Distribution of Load -- Welds which develop the strength of sections joined are
required to prevent "unzipping" or progressive failure of the joint. The AWS prequalified groove
welds meet this requirement. Under-sized fillet welds are vulnerable; the 1988 Code will
upgrade the prequalified fillet weld details so that they will at least be adequate when E70XX
electrodes are used to join mild steel.
Local buckling in the branch member Due to differences in the relative flexibilities of
the members at a tubular connection, the actual distribution of axial stress in the branch member
is not like the uniform nominal stress we calculate. Localized yielding and re-distribution of load
may be required to develop the full capacity. Compact sections provide for this.
 141

Local buckling in the main member Due to concentrated delivery of branch loads, the
actual distribution of axial stress in the main member is not like the uniform nominal stress we
calculate. In addition, there are high localized shell bending stresses which accompany punching
shear. Localized yielding and re-distribution of load may be required to develop the full
capacity. Compact sections provide for this.
Beam shear in the main member Although rarely found to govern in offshore practice,
this is a potential failure mode for gap joints when the product of Ej tau beta exceeds 0.13 to
0.33. The lower number applies where the chord is fully stressed axially, but thus usually does
not occur at points of high shear loading. The AISC interaction between axial and shear stress in
a compact section is applicable here.
Lamellar Tearing See Code Commentary.
Fatigue ~ This is extensively covered in the Code. Mild steel joints with 100% strength
efficiency can be expected to safely withstand about 3,000 applications of load equal to their
allowable static capacity (but only a few hundred full reversals). Fatigue performance does not
increase in proportion to yield strength, so joints attempting to exploit high strength steel have
even shorter lives.

3.8.5 Design Procedure

What follows is a step-by-step design procedure for simple tubular trusses, using the
charts presented in the foregoing, subject to the stated limitations.
Step 1 Lay out the truss and calculate member forces using statically determinate pin-
end assumptions. Flexibility of the connections results in secondary bending moments being
lower than given by typical rigid-joint computer frame analyses. Secondary moments may be
neglected for connection design, provided welds match the line load capacity of members joined,
and the connection is otherwise detailed to provide ductile deformation capacity.
Step 2 ~ Select members to carry these axial loads, using the appropriate governing
Code, e.g., AISC. While doing this, consider the architecture of the joints, considering the
following guidelines:

(a) Keep compact sections, especially low D/T or width/thickness, for the main
member (chord).

(b) Keep tau (branch/main thickness ratio) less than unity, preferably about 0.5.

(c) Select branch members to aim for large beta (branch/main diameter ratio), subject
to avoidance of large eccentricity moments.

(d) In K-joints, use a minimum gap of tj + t2 (where t^ and t2 are the two branch
thicknesses) or 2-inches between large diameter members, for welding access.
Reconsider the truss layout if this gets awkward.

Step 3 Calculate and distribute eccentricity moments. These are not secondary
m o m e n t s and must be provided for. They may be allocated entirely to the chord for
eccentricities less than 25% of the chord diameter, but should be distributed to both chord and
branches for larger eccentricities. Re-check members for these moments and re-size as
necessary.
Step 4 For each branch member, calculate utilization against member-end yield at the
joint,
142

fa + fb
A = , ^ , or 32 (3.30)
y 0 . 6 Fy 0 . 6 Fy 0 . 6 Fy
where fa is nominal axial and fb bending in the branch, with the 1/3 increase applicable to the
denominator, where used. Also calculate chord utilization, using chord nominal stresses and the
formula given in Figure 3.61 for circular sections. At gap K-joints, also calculate chord shear
utilization against the allowable 0.4 Fy, using half the gross area for round tubes. Then check
the gap region of the chord for combined shear and axial interaction, using AISC criteria.
Step 5 For each end of each branch member, calculate connection efficiency, Ej,
using Equation 3.29 and the appropriate charts:

E y for circular tubes

axial K-connections Fig. 3.56
axial T&Y connections Fig. 3.57
axial X-connections Fig. 3.58
in-plane bending Fig. 3.59
out-of-plane bending Fig. 3.60
de-rating factor Qf Fig. 3.61

For unusually critical situations, where AWS 10.5.1.7 applies, take 2/3 of the calculated value of

Step 6 For axial loading alone, or bending alone, the connection is satisfactory if
member-end utilization is less than joint efficiency, i.e.

* 1.0 (3.31)
E j

For combinations of axial load and bending, the check for circular section joints is...

1 .75
+
bL ^ 1.0 (3.32)
L j J bending
D J axial

Step 7 ~ To redesign unsatisfactory connections, go back to step 2, and

(a) increase the chord thickness, or

(b) increase the branch diameter, or
(c) all of the above.

Consider overlapped or stiffened joints only as a last resort. Overlapped joints increase
the complexity of fabrication, but can result in substantial reductions in the required chord wall
thickness. Large internally stiffened truss joints are further described in Chapter 6.
Step 8 When you think you are done with the design, go talk to potential fabricators
and erectors. Their feedback could be invaluable for avoiding unnecessarily difficult and
expensive construction headaches. Also make sure they are familiar with, and prepared to
follow, Code requirements for special welder qualifications, and that they are capable of coping
the brace ends with sufficient precision to apply AWS prequalified procedures. See Chapter 8.
 143

REFERENCES

1 Lee, G.C., Twenty Years of Offshore Platform Development, OFFSHORE, June 5,1968.
2 Stallmeyer, J. E., Static Test of a Full Scale Pipe Joint, . M. Newmark/Talbot Lab (Urbana IL)
report to the California Co., June 1959.
3 Zumwalt, . H., Power Plant at Sea, Civil Engineering, June 1960.
4 Johnston, L. P., The Welded Tubular Joint Problem in Offshore Structures, Shell EPR Pub. 326,
presented at 1st University of Texas Conference on Drilling and Rock Mechanics, Austin, TX,
January 1963.
5 Toprac, . ., et al, Studies on Tubular Joints in Japan, Part I, report to Welding Research
Council, September 1968.
6 Roark, R. J., The Strength and Stiffness of Cylindrical Shells Under Concentrated Loading, Trans,
of ASME, v.57 pp 147-152,1935.
7 Roark, R. J., Formulas for Stress and Strain, McGraw-Hill, NY, 1954.
8 Bijlaard, P. P., Stresses from Local Loadings in Cylindrical Pressure Vessels, The Welding
Journal, Welding Research Supplement, December 1954 (also see Trans, of ASME v.77 no.6,
1955).
9 Bijlaard, P. P., Stresses from Radial Loads and External Moments in Cylindrical Pressure Vessels,
The Welding Journal, Welding Research Supplement, December 1955.
10 Bijlaard, P. P., Additional Data on Stresses in Cylindrical Shells under Local Loading, WRC
Bulletin 50(2), May 1959.
11 Toprac, . ., et al, An Investigation of Stresses in Welded T-Joints, Univ. of Texas S.F.R.L.
Tech. Rept. P-550-3, March 1965.
12 Bouwkamp, J. G., Recent Trends in Research on Tubular Connections, presented at SPE Offshore
Technology Symposium, New Orleans, May 1966; also Journal of Petroleum Technology
November 1966.
13 Dundrova, V., Stresses at the Intersection of Tubes - Cross and Tee Joints, University of Texas
SFRL Rept. P-550-5, (1966).
14 Dundrova, V., Stress and Strain Investigation of General Joints in Tubular Structures, Univ. of
Texas Rept. TL-A-03-67, July 1967 (proprietary).
15 Dundrova, V., Stress Concentration in Joints Subjected to Axial Loads, Bending Moments and
Shears, Univ. of Texas Rept. TL-A-01-68, March 1968 (proprietary).
16 Holliday, G. H., Elastic Strength of T-Joints, Shell Development Co. R&D Note, January 1967.
17 Caulkins, D. W., Parameter Study for FRAMETI Elastic Stress in Tubular Joints, Shell CDG Rept.
15, September 1968.
18 Bea, R. G., Tabulated Values for Circular Rings and Arches, Shell CDG Report, January 1966.
19 Bryant, J. E. Jr., Circular Tubular Joint Design, MS Thesis, Tulane Univ., New Orleans, April,
1962.
20 Togo, T, Experimental Study on Mechanical Behavior of Tubular Joints, doctoral dissertation,
Osaka University, 1967; also Washio, K., et al, Cross Joints of Tubular Members, report to Kinki
branch of AIJ, May 1966 (in Japanese; also see ref. 5).
21 Makino, Y. and Kurobane, Y., Recent research in Kumomoto University in Tubular Joint Design,
IIW Doc. XV-615-86.
22 Adrian, L. E., Sewell, . ., and Womack, W. R., Partial Investigations of Directly Loaded Pipe T-
Joints, Unpublished Theses, Southern Methodist University, Dallas, 1958.
23 Pittsburgh Testing Laboratory, Pipe Joint Tests, report to the California Co., New Orleans,
December 1959.
24 Bouwkamp, J. G., Research on Tubular Connections in Structural Work, Welding Research
Council Bulletin 71A, August 1961.
25 Bouwkamp, J. G., Behavior of Tubular Truss Joints under Static Loads, Univ. of Calif, report to
A.I.S.I., July 1963.
26 Bouwkamp, J. G., Report of Progress on Tubular Fatigue Program, memorandum, 1964.
27 Toprac, . ., et al, Welded Tubular Connections: an Investigation of Stresses in T-Joints,
Welding Journal Research Supplement, January 1966.
144

28 Toprac, . ., et al, An Experimental Investigation of Tubular T-Joints, Univ. of Texas SFRL

Tech Rept No. P550-8, January 1966.
29 Toprac, . ., et al, Stresses in Steel Tubular Y-Joints, Univ. of Texas SFRL Tech Rept No. P550-
7, January 1966.
30 Toprac, . ., et al, Investigation of Elastic Stresses in Welded Tubular Steel K-Joints, Univ. of
Texas SFRL Tech Rept No. P550-6.
31 Rodabaugh, E. C , Review of data Relevant to the Design of Tubular Joints for use in Fixed
Offshore Platforms, Welding Research Council Bulletin 256, January 1980.
32 Toprac, . ., An Investigation of Welded Steel Pipe Connections, Welding Research Council
Bulletin 71B, August 1961.
33 Kurobane, Y., Makino, Y. and Mitsui, Y., Ultimate Strength Formulae for Simple Tubular Joints,
IIW Doc. XV-385-76.
34 Pan, R. B., et al, Ultimate Strength of Tubular Joints, Proc. Offshore Tech. Conf. OTC 2644, May
1976.
35 Marshall, P. W., Risk Factors for Offshore Structures, Proc. 1st Conf. on Civil Engineering in the
Oceans, ASCE, San Francisco, September 1967.
36 Dominique de Menil, Made of Iron, catalogue item 322 p. 196, Univ. of St. Thomas, Houston,
1981.
37 Lee, G. C , Review of Offshore Platform Failures During Hurricane Hilda, presented at Delta
section SPE, New Orleans, Feb. 9,1965.
38 Marine Board Committee on Safety of OCS, Safety and Offshore Oil, National Academy Press,
Washington DC, 1981.
39 Sterling, G. H., et al, Failure of South Pass 70 Platform "B" in Hurricane Camille, Proc. Offshore
Tech. Conf. OTC 1898, May 1973.
40 de Jong, F., Tubular Joints, informal Shell report, August 1963.
41 Marshall, P. W., VE 257 design calculations, Shell Oil Co. CDG files, Jan-Mar 1966.
42 Marshall, P. W., Design of Simple Tubular Joints, Shell Oil Co. CDG Report 12, January 1967.
43 Johnston, L. P., A Review of Welded Tubular Joint Design Methods, Shell Development Co. EPR
Report 738, May 1963.
44 CE Manual No. 41, Commentary on Plastic Design in Steel, 1961.
45 Marshall, P. W., Considerations for the Selection of Structural Steels for use in Tubular Joints of
Offshore Platforms, Royal Dutch/ Shell Group Production R&D Conference, Agenda Item 92,
Rijswijk, Holland, March 1969.
46 Marshall, P. W., Ultrasonic Inspection Applied to Tubular Joints in Offshore Structures, Royal
Dutch/ Shell Group Production R&D Conference, Agenda Item 93, Rijswijk, Holland, March
1969.
47 Carter, R. M., Marshall, P. W., et al, Materials Problems in Offshore Structures, Proc. Offshore
technology Conf., OTC 1043, May 1969.
48 Marshall, P. W., et al, Report of Team "K", Design Stresses, AWS s/c on Welded Tubular
Structures, Feb. 6 1970.
49 Report of s/c 10, Tubular Structures, minutes of AWS Structural Welding Committee, Pittsburgh,
Dec. 14-15, 1971.
50 Marshall, P. W. and Toprac, . ., Basis for Tubular Joint Design, Welding Research Supplement,
May 1974.
51 AISC Manual of Steel Construction, Sixth Edition, New York, 1963.
52 Wardenier, J., personal communication, c. 1977.
53 AWS Structural Welding Code, First Edition, Dl.1-72, American Welding Society, Miami,
September 1972.
54 i.b.i.d. Revision 1, September 1973 (pink pages).
55 i.b.i.d. Revision 2, April 1974 (blue pages).
56 i.b.i.d. Second Edition, Dl.1-75.
57 i.b.i.d. 1976 Revisions (yellow pages).
58 i.b.i.d. 1977 Revisions (green pages).
59 Commentary on the Structural Welding Code, AWS Dl.2-77, American Welding Society, Miami,
1977.
 145

60 API Recommended Practice for the Planning, Designing, and Constructing of Fixed Offshore
Platforms, API RP 2A.
61 Marshall, P. W., A Review of American Criteria for Tubular Structures~and Proposed Revisions,
IIW Doc. XV-405-77, International Institute of Welding annual assembly, Copenhagen, 1977.
62 Wardenier, J., Testing and Analysis of Truss Joints in HSS, Proceedings Int'l Symposium on
Hollow Structural Sections, Toronto, May 1977.
63 Gibstein, . B., Static Strength of Tubular Joints, DNV Report 73-86-C, May 1973.
64 Akiyama, H., contribution to Monograph TC 43, Joint Committee on Tall Buildings, 1976.
65 Cheng, A. P., et al, Plastic Consideration on Punching Shear of Tubular Joints, Proc. Offshore
Tech. Conf. OTC 2641, May 1976.
66 Stamenkovic, A. and Sparrow, K. D., Load Interaction in T-joints of Steel Circular Hollow
Sections, ASCE Journal of Structural Engineering, September 1983.
67 Marshall, P. W., discussion of ref. 66, ASCE Journal of Structural Engineering, November 1984.
68 API Recommended Practice for the Planning, Designing, and Constructing of Fixed Offshore
Platforms, API RP 2A, 8th Edition, 1978.
69 Yura, J. ., et al, Ultimate Capacity Equations for Tubular Joints, Proc. Offshore Tech. Conf.,
OTC 3690, May 1980.
70 Rodabaugh, E. C , Review of Data Relevant to the design of Tubular Joints for Use in Fixed
Offshore Platforms, WRC Bulletin 256, January 1980.
71 Kurobane, Y., et al, Ultimate Strength Formulae for Simple Tubular Joints, IIW Doc. XV-385-76.
72 Wardenier, J., Design Rules for Predominantly Statically Loaded Welded Joints in Circular Hollow
Sections, IIW annual assembly, Bratislawa, 1976.
73 Kanatani, H., Experimental Study on Welded Tubular Connections, Memoirs of the Faculty of
Engineering, v.14 n.12, Kobe Univ., Japan, 1966.
74 Japanese Institute of Steel Construction, Study on Tubular Joints used for Marine Structures,
March 1972 (in Japanese).
75 Sammet, H., Die Festigkeit Knotenblechloser Rohrverbindungen im Stahlbau, Schweisstechnik
v.13,1963 (in German).
76 An Investigation of Welded Tubular Joints Loaded by Axial and Moment Loads, Offshore Job No.
ER-0169, Feb. 1976.
77 Beale, L. A. and Toprac, . ., Analysis of In-Plane , Y, and Welded Tubular Connections,
WRC Bull. 125, October 1967 (based on refs. 28, 29, 30).
78 Grigory, S. C , Experimental Determination of the Ultimate Strength of Tubular Joints, Southwest
Research Research Institute, Proj. No. 03-3054, San Antonio, Texas, Sept.1971.
79 Nakajima, T., Experimental Study on the Strength of Thin Wall Welded Tubular Joints, IIW Doc.
XV-312-71, London 1971.
80 Yura, J. ., et al, Ultimate Load Tests on Tubular Connections, Civil Engineering Structural
Research Lab. Rept. No. 78-1, University of Texas, Austin, Sept. 1978.
81 Zimmermann, W., Tests on Panel Point Type Joints for large Diameter Tubes, Otto Graf Institute
report to CIDECT, Sept. 1965.
82 Gibstein, M., The Static Strength of T-joints Subjected to In-Plane Bending Moments, Det Norske
Veritas Rept. No. 76-137, Oslo, July 1976.
83 Moses, F., Development of Preliminary Load and Resistance Design Document for Fixed Offshore
Platforms, API PRAC Project 85-22, Case Institute of Technology, January 1986.
84 Notes of API offshore structures committee meeting, June 18,1981.
85 Yura, J. ., et al, Ultimate Strength of Tubular Joints: Chord Stress Effects, Proc. Offshore Tech.
Conf. OTC 4828, May 1984.
86 Kurobane, Y., Welded Truss Joints of Tubular Structural Members, Memoirs of the Faculty of
Engineering, Kumamoto University, Japan, 1964.
87 Hoadley, P. W. and Yura, J. ., Ultimate Strength of Tubular Joints Subjected to Combined Loads,
Proc. Offshore Tech. Conf. OTC 4854, May 1985.
88 Graff, W. J., et al, Review of Design Considerations for Tubular Joints, ASCE Preprint 81-043,
New York, May 1981.
89 Kurobane, Y., Recent Developments in Tubular Joint Design, ASCE Preprint 81-002, New York,
May 1981.
146

90 Taylor, R. G., British Development and Practice of Long Span Tubular Construction, ASCE
Preprint 81-027, New York, May 1981.
91 Wardenier, J. and de Back, J., Considerations in Static and Fatigue Design of Tubular Joints,
ASCE Preprint 81-048, New York, May 1981.
92 Furnes, O., Design and Future Aspects of Offshore Tubular Structures, ASCE Preprint 81-133,
New York, May 1981.
93 Det Norske Veritas, Rules for the Design, Construction, and Inspection of Offshore Structures,
Appendix C, Steel Structures, 1977.
94 Galambos, . V. and Ravindra, . K., Load and Resistance Factor Design for Steel, Journal of the
Structural Division ASCE, Sept. 1978.
95 AISC Proposed Load & Resistance Factor Design Specification for Structural Steel Buildings,
American Institute of Steel Construction, Chicago, Sept. 1983.
96 API Recommended Practice for Planning, Designing, and Constructing Fixed Offshore Platforms,
15th Edition, October 1984.
97 Zettlemoyer, N., API committee correspondence, March 22,1982.
98 Roussel, H. J., API committee correspondence, Sept. 1982, Jan. 1983, March 1983, etc., etc.
99 Marshall, P. W., Review of SCF in Tubular Connections, Shell Oil Co. CE-32 Report, April 1978.
100 Marshall, P. W. and Luyties, W. H., Allowable Stresses for Fatigue Design, Proc. 3rd Intl. Conf.
on the Behaviour of Off-Shore Structures, BOSS-82, vol.2, MIT Cambridge MA, 1982.
101 Marshall, P. W., The Design of Multiplanar Joints, presented at ASCE Structures Congress, New
Orleans, October 1982.
102 Kurobane, Y., New Developments and Practices in Tubular Joint Design, IIW Doc. XV-81-010,
1981.
103 Garf, E. F., Engineering Methods of Calculating Tubular Welded Assemblies in Deep-Sea
Foundation Structures, Automatic Welding, vol.33 no.2,1980.
104 Marshall, P. W., Connections for Welded Tubular Structures, I.I.W. Houdremont Lecture, Boston,
July 1984.
105 American Welding Society, Structural Welding Code - Steel, AWS D 1.1-86,1986 edition, AWS,
Miami.
106 Packer, J. ., Berkemoe, P. C , and Tucker, W. J., Design Aids and Design Procedures for H.S.S.
Trusses, A.S.C.E. Journal of Structural Engineering, vol.117, no.7, pp.1526-43, July 1986.
107 Wardenier, J., Hollow Section Joints, Delft University Press, 1982.
108 Reusink, J. H. and Wardenier, J., Simplified design charts for axially loaded joints of circular
hollow sections, IIW doc XV-671-88, T.U. Delft/Stevinlab, February 1988.
109 Wardenier, J., Ontwerp en berekening van overwegend statisch belaste verbindingen van ronde
buisprofielen, Van Leeuwen Technische Informatie Nr. 8 (in Dutch; English version also
available).
110 American Institute of Steel Construction, Manual for Steel Construction, 8th Edition, 1980.
111 IIW s/c XV-E, Design Recommendations for Hollow Section Joints, 2nd Edition, IIW Doc. XV-
701-89, September 1989.
Chapter 4

FATIGUE DESIGN

Earlier chapters have been concerned mainly with issues of static strength. While
strength is a most fundamental requirement, once it has been satisfied we must then examine
alternative modes of failure. Because of the high local stresses in the "hot spot" regions of a
tubular connection, fatigue and fracture need to be considered, even though localized yielding
may be tolerated for static loading and ductile materials.
Fatigue may be defined as damage that results in fracture after a sufficient number of
stress fluctuations. Performance may be characterized as a plot of stress range versus number of
cycles to failure (S-N curve).
A fatigue analysis for an offshore structure must include the following elements:

1. Long-term wave climate is the starting point of fatigue analysis. This is the
aggregate of all sea states occurring yearly (or for longer periods of time).
Obtaining this data often requires a major effect, with significant lead times.
2. Global-scale space-frame analysis is performed to obtain structural response in
terms of nominal member stress for each sea state of interest.
3. Geometric stress concentrations at all potential hot spot locations within the tubular
connections must be considered, since fatigue failure initiates as a local
phenomenon.
4. Accumulated stress cycles are then counted and applied against suitable fatigue
criteria to complete the analysis of fatigue damage.
5. In view of the scatter and uncertainty in fatigue, the choice of a target calculated
fatigue life requires careful evaluation of the economic and risk factors involved.
Typically, the target life is a multiple of the required service life.

An example of a deterministic fatigue life calculation for a shallow water well jacket
with adequate joint-cans, is shown in Table 4.1. For offshore structures, Ref. 1, elaborates on the
various alternative formats in which wave climate and dynamic structural analysis may be

TABLE 4.1 EXAMPLE FATIGUE DAMAGE CALCULATION

DEEP WATER HOT SPOT
WAVE HEIGHT AVERAGE STRESS
RANGE NUMBER RANGEg AWS CURVE
(FT.) PER YEAR T R KSI
X-MOD DAMAGE RATIO
9
0-5 3,060,000 3.0 1.0 1 0 .0031
8
5-10 410,000 5.6 1.1 1 0 .0037
7
10-15 130,000 12.4 5.0 1 0 .0260
6
15-20 4,790 19.0 1.0 1 0 .0048

5
I 20-30 810 30.0 1.8 1 0 .0045
5 4
2.0 1 0 .0018
~ 30-40 37 47.0
600 .0033
I 40-50 2 71.0

ANNUAL DAMAGE RATIO .0442

CALCULATION LIFE 23 YR.

148

considered. These structures are subject to cyclic wave loads, from a variety of seastates, which
have been discretized in terms of significant wave height (and corresponding stress range) in
order to construct the table. Using the RMS stress directly with an ordinary S-N curve results in
errors on the unsafe side (refs. 3, 4), as shown in Figure 4.1. Also note that the endurance limit,
apparent for constant stress range, is eliminated under random loading, and the high cycle end of
the S-N curve merely flattens in slope.

Fig. 4.1. Various

representations of fatigue results
for n a r r o w b a n d r a n d o m
loading.

CYCLES TO FAILURE

For other types of structures, the sources of cyclic loads to be considered may include
turbulent wind loads (e.g., towers and sign structures), traffic loads (e.g., bridges), and duty cycle
loads (e.g., cranes). The global analysis of these loads is a challenging subject and a laborious
part of the design process, but beyond the scope of a book on tubular connections. However,
some useful generalizations may be stated:
Key parameters of the long tenn stress distribution are the total number of cyles, N j , the
extreme stress range A a m a x, and the Weibull shape parameter . These are illustrated in Figure
4.2. Lifetime total cycles can range from 10,000 (for a once-a-day duty cycle) to 100-million
(for offshore structures). The extreme stress range can be varied by the designer (e.g., by varying
member thickness), as required, to achieve satisfactory structural perfomiance.

Fig. 4.2. Long term cyclic

stress distributions.

CUMULATIVE CYCLES LARGER

 149

The shape of the long term stress distribution is often a characteristic of a given class of
structure and its service environment, and the results of laborious prior analyses or in-service
measurements can be generalized in terms of to aid in the preliminary design of subsequent
structures of the same class (refs. 5-7). For offshore structures, is typically in the range of 0.5
to 1.0. The Rayleigh distribution corresponds to of 2, with the typical distributions used for
bridge design (ref. 8), yielding somewhat higher values. Constant cycles all at the design stress
range (e.g., industrial duty cycles) correspond to of infinity.
For the purposes of fatigue damage calculations, the cumulative stress distribution is
discretized into "bins", each with its corresponding incremental number of cycles, and the
cumulative damage ratio D is computed according to Miners rule

D =

where is the number of cycles applied at a given stress range, and is the allowable number of
cycles at that stress range. Where A a m a x, N j , and are known, a convenient closed form
expression for D exists (ref. 5). The allowable fatigue life is exhausted at a damage ratio of
unity, with a corresponding safety index of about 2 ( 3 % failure rate). Fractional damage ratios
are used to achieve higher reliability (ref. 6).
Few members and connections in conventional building frames need to be designed for
fatigue, since most load changes occur infrequently, and wind loads produce only minor cyclic
stresses in comparison to gravity loads. Generally, the full design wind or earthquake loads are
sufficiently rare that fatigue need not be considered (ref. 9). An exception to the foregoing
statement would be elements of a structure which are expected to act as energy absorbers under
extreme earthquakes, and which may fail in low-cycle fatigue. Under exceptional circumstances,
e.g., elevated structures on slender legs, wind-induced fatigue may also become significant (ref.
10).

4.1 LEVELS OF ANALYSIS

In Chapter 2, two very useful design simplifications, hot spot stress and punching shear
were defined. These terms were used, sometimes implicitly, in discussing failure modes and
methods of analysis for tubular connections. Now, a more detailed development of the hot spot
concept will be presented.
Let us begin by reviewing the several scales at which analysis of a tubular structure
should be considered, using offshore structures as an example (ref. 11). See Figure 4.3.
A global stress analysis of the overall structure resolves applied gravity loads, wind,
wave, and current into nominal axial and bending stresses in the various members. A typical
level of nominal stress for tubular jacket bracing would be 20 ksi (140 MPa), for the one-time
extreme wave load.
If one were to focus on the connection as a structure, one might find punching shear on
the order of 7 to 10 ksi (50 to 70 MPa). An experimental or finite element stress analysis of the
connection would typically give peak hot spot stresses around of 50 ksi (420 MPa); that is, the
geometric stress concentration factor (SCF) is about 2.5 at potential fatigue sites A (in the joint
can) and (in the end of the branch member), even in well designed connections. The
corresponding once-in-a-lifetime stress range would be 80 ksi, as wind and current produce a
non-zero mean.
Finally, in the right hand part of the figure, we consider the localized weld cross section.
In between the strain gage location where hot spot stress is measured, and the weld toe where
cracks initiate, there is additional stress concentration on the local scale (reflecting overall weld
150

profile, e.g., concave vs. convex) and the microscopic scale (reflecting the sharp notch, undercut,
and crack-like defects at the toe of the last pass). It is in this region where the size effect arises,
as discussed in Chapter 7.

Fig. 4.3. Levels of analysis for offshore structures, (a) Global, (b) Connection as a structure, (c)
Local or microscopic.

4.2 HOT SPOT STRESS

Among the various levels of detail in stress analysis which may be adopted as a basis for
fatigue calculations, the hot spot stress has evolved as the most practical basis for design
purposes. Hot spot stress places many different structural geometries on a common basis,
ranging from butt welds to nozzles in pressure vessels, to tubular joints in offshore platforms. In
AWS practice, the reference stress (or strain) is the total range which would be measured by a
strain gage placed adjacent to the toe of the weld, and oriented perpendicular to the weld so as to
reflect the stress which will be amplified by weld toe discontinuities. This is used with an
empirical S-N curve based on measured hot spot stress and cycles to failure, in tests of realistic
as-welded hardware. The effect of representative local/microscopic discontinuities at the weld
toe are presumed to be built into the data base of realistic as-welded hardware. In the plastic
range, strain is used instead of stress.

4.2.1 Development
The concept of hot spot strain as a useful design parameter for low cycle fatigue and
fracture was initially developed during the mid-1960's. It appears to have been an idea whose
time had come, as the development was along several parallel fronts. Although there may be
many paths to enlightenment, the author can only retrace the one he followed. To properly
define hot spot strain, we must do so in the context of how it is to be used in design.
Following the conceptual work of Peterson, Neuber, and Manson (refs. 12, 13, 14), and
pressure vessel fatigue tests at Ecole Poly technique, a fatigue design curve for full size pressure
vessels and other practical welded hardware was proposed by Pickett et al at the Southwest
Research Institute (SWRI), San Antonio, Texas, in 1963 (ref. 15). The empirical design curve
 151

was based on the worst measured or peak local strain range. It was similar in shape, but lower
than, median fatigue S-N curves for smooth polished specimens (e.g., Manson), reflecting
scatter, notch effects, and metallurgical effects at the toes of welds. The focus was originally on
low-cycle fatigue, involving failure in less than 10,000 cycles (one duty cycle per day for 27
years) and peak strains in the plastic range. This work eventually became part of the ASME
Code for pressure vessels design by analysis (ref. 16), as opposed to the more traditional
"cookbook" rules. The ASME Code uses hot spot strain in terms of a fictitious "peak stress"
(strain single amplitude times elastic modulus), which is further defined as "the highest stress in
the region under consideration... (having) the basic characteristic that it does not cause any
noticeable distortion, and is only objectionable as a possible source of a fatigue crack or a brittle
fracture". In pressure vessels, the peak stress is not always at a weld; it may be at the bore of a
cut-out, or at a geometric discontinuity in a forging. For pressure loading, zero to peak also
defines the working range.
Pellini et al recognized the role of plastic strains in the local region containing a notch or
flaw, in constructing their empirical Fracture Analysis Diagram (refs. 17 & 18). For small flaws
in ordinary constructional steels, local plastic strains are required for brittle fracture initiation
under static or quasi-static loads, even with material having marginal dynamic notch toughness.
However, at service temperatures below the Nil Ductility Transition Temperature (NDTT), brittle
fractures can propagate catastrophically, once initiated. Increasing margins on the safe side of
the NDTT increase the amount of local plasticity required for fracture initiation at small flaws,
increase the flaw size that can be tolerated at yield, and increase the likelihood that propagating
fractures can be arrested. Brittle fracture is discussed further in Chapter 7; the key point here is
the role of local plastic strains, those at the hot spot.
The author began using these concepts in his work designing offshore platforms for
Shell Oil Company as early as 1965 (refs. 19 & 20), allowing peak hot spot strains of 0.2% for
the design wave load, well into the plastic range for the materials being used. Recognizing
differences between structural and pressure vessel practice such as the tendency of peak
stresses to occur at welds, and the general absence of weld dressing and stress relief he
constructed his own design curve for local strains in tubular connections. The first such curve is
shown in Figure 4.4. In addition to Pickett et al, early data sources included Munse (ref. 21),
Betero & Popov (ref. 22), and Bouwkamp Phase I (refs. 23 & 24). As more data became
available, it was added to the plot (refs. 25, 26, 27, 28, & 29); it was gratifying to see the new
data falling in the same scatter band, given the more rigorous attention to experimental technique
(smaller strain gages closer to the weld toe, strain range taken after shakedown, etc.) being
practiced in the newer work.
Work at SWRI merged with that of the University of Texas at Austin, during Phase III
of a joint industry project organized by Shell Development Company (ref. 26). Previous work at
Austin (Phases I and II) had covered static strength testing, as well as experimental and
theoretical stress analysis (refs. 30, 31, 32, & 33), referring to the point of highest shell stress and
most severe plastic deformation in the chord wall as the "hot spot". Besides being "where the
action is", and the focus of all the stress analysis work, plastic work at the hot spot can actually
generate heat under rapid cyclic loading. In simple -, Y-, and gap K-connections, the hot spot
was found on the outside surface of the intersecting members, at the toe of the weld joining them.
Typically, the largest strains were perpendicular to the weld, involving shell bending of the
chord, with large plastic strains parallel to the weld being suppressed by the stiffening effect of
the intersecting surfaces. Typical hot spot locations are shown in Figure 4.5
When Toprac first visited SWRI in 1967, fatigue testing at Austin had already been
started by Kurobane (ref. 34), initially using small scale T-connections which could be bench
tested on a Sonntag fatigue testing machine. Larger scale tests followed, with the special loading
frame, hydraulic jacks, and testing being completed with support from Welding Research
152

AVG. r
M U N S E - B U T T WELDED CARBON S T E E L
MIN i
B E T E R O & P O P O V S T E E L BEAMS
P I C K E T T E T A L - F U L L SIZE P R E S S U R E V E S S E L S
SWRI-SHELL & JOINT SERIES
SINGLE V B U T T WELD - A 242
A S - W E L D E D S U P E R L O TEMP ARMCO DATA
c^3C^> M G S - C MODEL (INITIAL & FINAL F A I L U R E )
HY 8 0 - I N D E X O F S T R U C T . F A T I G U E L I F E - N R L
BOUWKAMP-CHEVRON TUBULAR JOINTS
A L A B O R A T O R Y DATA R E P . BY D U N L O P

6 7
10 5 10 10
CYCLES OF LOAD

Fig. 4.4. Marshall's original CDG-5 fatigue design curve for tubular connections.

(a) (b)
Fig. 4,5. Views of branch member
\ footprint on chord with shaded regions
/ in which the hot spot is typically
U f o u n d , (a) A x i a l l y l o a d e d T-
(qJ ( connection, (b) Out-of-plane bending.
(c) In-plane bending, (d) Axially
loaded Y-connection, > 40. (e)
0
Axially loaded Y-connection, < 3 0 .
(f) Gap K-connection, with balanced
axial loads, (g) K-T connection,
() (f) central brace not loaded.

(g)
 153

Council and the U.S. Navy. The "larger" specimens (8.6-in. dia, 0.2-in. wall chord) showed
significantly lower strength, more in line with the other data being generated for offshore
structure usage (refs. 35 & 36). Although the Austin tests were initially reported in terms of
cyclic amplitude of punching shear, strain gages at the hot spot were also recorded, so that these
results could eventually be added to the statistical data base in the latter format, nearly doubling
the amount relevant data available in the open literature by the end of the 1960's.
The original CDG-5 design curve underwent further evolution between its first open
publication (ref. 37) and its incorporation as curve X-X in the AWS Structural Welding Code in
1972, under the subcommittee chairmanship of Professor Toprac (refs. 38-41). As shown in
Figure 4.6, early published data has been re-plotted in full log-log format, and given a
probabilistic interpretation by displacing the original curve along the fatigue life scale until it
encompassed 15% to 85% of the original data (plus/minus one standard deviation). When
additional data is considered (e.g., Toprac's), and the shape of the design curve is changed to be
more conservative in the high-cycle region, it falls on the safe side of 97% of the data (safety
margin of two standard deviations).

Fig. 4.6. Evolutionary form of the hot spot strain design curve, showing University of Texas data.

4.2.2 Attributes
We shall now elaborate on the desirable attributes of hot spot strain. They are: (1) It
should place different connection geometries on a common design basis. (2) For repetitive
designs, the results should be generalizable in terms of stress concentration factors; and be
invariant for a given connection geometry, depending only on load pattern. (3) The design
reference stress or strain should be equally derivable from model tests or analysis (e.g., thin shell
finite element) of the connection. (4) Weld notch effects, residual stresses, etc., which are not
amenable to such measurement or calculation, should be implicitly included in the empirical
design curve. (5) Hot spot strain should provide a measure of safety consistent with other design
methods being used, despite its somewhat unprecedented foray into the realm of cyclic plasticity.
154

(i) Commonality. The first attribute was tested by having the following variety of
practical as-welded hardware geometries in the data base (including unpublished data): butt
welds (refs. 21 & 29); stiffened and unstiffened cone-cylinder transitions (refs. 42 & 43);
pressure vessel nozzles (ref. 15); many simple T-connections (refs. 25, 35, 36, & 44); cross
connections (ref. 45); thin-wall gap K-connections (refs. 24, 25, & 46); heavy wall gap K-
connections, overlap (negative-eccentric) K-connections, gusseted connections, ring-stiffened
connections, grouted connections, and even an early cast node (ref. 47). Full size nodes from a
Gulf of Mexico offshore platform (ref. 48), service strains from the failure site in a semi-
submersible (refs. 49 & 50), and corrosion-fatigue data (refs. 24, 28, 29, 51, & 52) were also
included. These 70-plus data points, from 20-odd references, were added to the plot as they
became available, two decades ago. The full data set, as of mid-1973 (ref. 53), is shown in Figure
4.7; as originally published (ref. 9), the data points were not identified, as some of them were
proprietary at the time. The degree to which the data all falls in a common scatter band
demonstrates the desired attribute; the fact that the scatter band is rather broad indicates that hot
spot strain is a useful design approximation, not a precise research tool. Unfortunately, the
corresponding tabular data set does not exist; however, Rodabaugh's WRC data base (ref. 54),
includes many of the same tests, those covering simple , , K, and cross connections.

TUBULAR JOINTS FATIGUE DATA

SRI PVRC

Fig. 4.7. Database and hot spot strain design curve, about the time of first inclusion in AWS Dl.1-72.

(ii) Invariance. The second attribute was the motivation behind all the attention given
to stress concentration factors (SCF) and shell analysis at Austin, finite element work at
Berkeley, and parametric SCF formulas, both in the early work and in many later efforts. In a
practical sense, designers need SCF which are uniquely defined in terms of connection type (T,
, , X, etc.), geometry (tau, gamma, beta, eta, theta, zeta, etc.), and load pattern (axial, in-plane
bending, out-of-plane bending, and degree of ovalizing). In its simplest terms, the SCF is the
ratio of hot spot stress to the corresponding nominal stress in the adjoining branch member.
 155

Elements of the problem which are not uniquely defined by the linear member-by-
member SCF approach, e.g., non-linear plasticity, stiffening from out-of-plane branch members,
and the effect of weld shape, must either be ignored (i.e., accounted for in the scatter) or
explicitly taken into account (increasing the complexity of the design procedure).
After the onset of plasticity, strains tend to be more highly concentrated in the regions
which are yielding. Using elastic SCF will tend to underestimate hot spot strains in the plastic
range, especially in regions which are load controlled, with little opportunity for load shedding.
This effect is mitigated if the zone of yielding is displacement controlled by surrounding elastic
material. Pickett's early treatment of the peak strain concept recommended using Neuber's rule
(ref. 13) to estimate the inelastic strain concentration factor (SNCF) from the theoretical elastic
stress concentration factor (Kp, and the actual stress concentration factor ( , derived iteratively
from the peak local strain range and the inelastic cyclic stress strain curve) as follows:

SNCF = (4.2)

The effect of plasticity is to reduce the actual hot spot stress (and ) , thereby increasing plastic
strains and SNCF. Using the approximation SNCF = SCF for strains beyond yield is
inconsistent with the experimental basis of the original design curve, particularly the steeply
sloping part at fewer than 10,000 cycles to failure. However, for offshore structures, where most
of the loading cycles are elastic and only rarely go into the plastic range, this approximation has
been used successfully, but with a modified design curve, as shown by the dashed lines in Figure
4.7.
While the early work focused on cyclic strains in the low-cycle plastic range, the more
demanding fatigue environments of deep water and the North Sea refocused attention on the
long-life elastic range (ref. 11). Here the dashed line extension of the AWS-X-modified design
curve accounts for the effect of random loading, as previously shown in Figure 4.1. While in the
plastic range all we can really measure is strain, in the elastic range stresses become a more
meaningful measure of what is going on. Elastic stresses are no longer treated as uniaxial, and
under multi-axial conditions an interesting question arises as which stress to use with the failure
criteria. Where we are dealing with smooth specimens, or situations in which all notch effects
have been fully accounted for in the calculated stresses, the von Mises stress is generally
acknowledged to be the appropriate one to use. However, for welded tubular connections, the
author favors using the normal stress, perpendicular to the weld, as this is what gets amplified by
notch effects and flaws at the toe of the weld, and this is what drives the fracture mechanics
stress intensity factor (mode I) once fatigue cracks start to grow. Where this is a principal stress
(and it often is, at least approximately, for shell bending in the chord), hot spot stresses and hot
spot strains are related as follows (ref. 55):

1 + v
WSN/
SCF = SNCF z (4.3)
1 - v
where ej-iSN is hot spot strain normal to the weld axis
2 is strain parallel to the weld axis
is Poisson's ratio

Since the second strain typically ranges from 25% to 50% of the first, the actual SCF is 1.2 to 1.3
times the SNCF. Most design calculations are based on elastic SCF, even though the design
curve and much (but not all) of original data base were based on SNCF. In the elastic range,
156

using elastic SCF to enter the design curve would appear to be conservative, in contrast to
potential errors on the unsafe side in the plastic range.
Parametric SCF which relate the hot spot stress to a linear combination of nominal axial
and bending stresses in the branch member are discussed in detail in Section 4.3. Treatment of
SCF for multi-planar braces will be discussed in Chapter 6.
The effect of weld shape is a topic of high current interest, as it is a source of significant
ambiguity in current SCF-based design procedures. A premise in the original hot spot design
curve, stated in both the AWS and API codes, is that the weld profile should merge smoothly
with the adjoining base metal, so that the unmeasured notch effect in a tubular joint is about the
same as for a butt weld, placing these two geometries on a common basis. Where this is not the
case, we must clearly distinguish between the last two levels of analysis of Figure 4.3, which
give two kinds of local stress, sigma-G (OQ) and sigma-L (^). The definitions of Radenkovic
et al (ref. 56) may be restated as follows:
Sigma-G is the geometrical hot spot stress, which should be invariant given relative
diameters, thicknesses, and angles of the intersecting members. It presumes that a linear
variation of shell bending stress is dominant in the critical regions of a tubular connection. It can
be determined experimentally by extrapolation from measurements on two suitably disposed
strain gages, as defined by ECSC working group 3 and others (refs. 57, 58 & 59) in Figure 4.8.
It can be determined analytically by isoparametric finite elements which reproduce the linear
variation of stress adjacent to the weld; stress at the mid-plane intersection of thin shell analyses
will not always satisfy this definition. In design, parametric formulas derived from the foregoing
methods may be used.
Further concentrated (micro-notches
and f l a w s a t w e l d t o e )

Rapidly rising stress (Notch stress

Brace due t o o v e r a l l weld shape)
Wflli.

Stress linearity
(Geometric shell stress)

b, . b 2 b3
111
D i j k s t r a - d e Back 0.2/rT 0.65/FT 0.5/RT

in 4,
UK Guidance 0.2 /rT 0.65/FT 0.4 / r 1 R

Gurney.Van D e l f t 0.4 t 0.65/rT 0 . 4 VTTrT

Gibstein branch JU
chord .25
x 7
( 1 ) Not less than 4 mm
Fig. 4.8. European definitions of hot spot strain, sigma-G. (a) Linear extrapolation procedure (ref.
57). (b) Parameters for strain gage location. See Fig. 4.9(c) and (d) for actual examples of
strain gage placement.

Sigma-L is a more localized stress, which includes effects of weld profile shape and
size. Experimentally, it can be evaluated by strain gages placed as close as possible to ~ or even
straddling ~ the toe of the weld. Analytically, it corresponds to finite element analyses which
have been re-meshed to zoom in on the weld toe. Since we may be working in the vicinity of a
 157

notch or stress singularity, some care is required in order to maintain a consistent definition.
S AE-oriented designers of heavy mining and earth moving equipment use strains averaged over
a 0.25-inch (6mm) gage length straddling the weld toe (ref. 60), together with an empirical S-N
curve for welded specimens.
Yoshida and Iida (ref. 61) use a notional 0.6mm (.025 in) gage length, with a smooth
specimen S-N curve. Similarly, notch stress theory for the weld-toe heat affected zone (ref. 62)
is used with a worst-case notch-tip radius of 0.2mm (0.008"); the resulting K t is reduced before
entering the fatigue S-N curve, becoming

Kt + 1
4 )
Kf = 2 '

This level of analysis may also be referred to as "microscopic".

Although sigma-L requires more work on the part of the stress analyst than sigma-G,
uncontrolled local perturbations above the reference stress tend to be minimized.
It is indeed an unfortunate source of confusion that the AWS definition of hot spot stress
falls loosely somewhere between these two. However, in the author's treatment of weld profile
and size effects in terms of notch stress analysis and fracture mechanics, sigma-G is clearly
intended as the reference stress (ref. 63 and Chapter 7 herein). When the stress analysis stops at
sigma-G, weld profile and size effects must be addressed elsewhere in the design process, as
these effects can be quite important. Thus we have the "size effect" adjustments to the S-N curve
in the British D.O.E. rules (ref. 57), and the even more elaborate "size and profile" provisions
published in AWS D 1.1-86. Such elaborate corrections are symptomatic of a methodology
which is being pushed beyond its limits of fundamental applicability.
The importance of distinguishing between sigma-G and sigma-L is illustrated in Figure
4.9. Fig. 4.9(d) shows a large scale French test with a very abrupt weld profile (cited in ref. 96).
The strain concentration factor SNCF is 3.3 for sigma-G and 6.6 for sigma-L. When the
doubling of stress within the circled region is ignored, and the French test is plotted in terms of
sigma-G, the fatigue strength of this connection falls below the original American design curve.
In the American data base, in which the tubular joints were either small scale or had welds
profiled so as to achieve a smooth transition, Figs. 4.9(a) and (b), the difference arising from the
looser definition of hot spot stress disappears in the scatter band, as the unmeasured notch effects
within the circled regions are similar.
Dijkstra (ref. 64) and deBack describe test results for two large scale tubular connections
having the same overall geometry, with and without a specially improved weld profile which
merges with the adjoining base metal. A typical unimproved weld is shown in Fig. 4.9(c). The
connection with the improved profile had a three-fold longer fatigue life. This was explained in
terms of extrapolated hot spot stress at the actual weld toe location, which was reduced by a
factor of 1.33 for the improved profile, by virtue of weld reinforcement placing the weld toe
further down the stress gradient. However, this "situational" explanation is inconsistent with the
concept of sigma-G as invariant for a given connection geometry, and tends to obscure the good
news that weld profiling can improve the fatigue performance of a given connection geometry.
The invariance of physical laws, and the analytical predictability of nature, is
fundamental to the philosophy of modern science (refs. 65 & 66), and to the success of
engineering design predictions. Thus, the proposed "situational sigma-G" does not seem to be
particularly attractive, when variations in weld toe location are neither calculated by the designer
nor under his control. However, when improved profiles of the type shown in Figure 4.9(e) have
been specified to prolong the fatigue life of selected nodes, de Back's results are a welcome
confirmation that such a strategy is indeed effective. In some cases, it may be prudent to
investigate whether the stiffer weld has attracted larger shell bending moments.
158

(a)
(b)

(d) (c)

(e)

Fig. 4.9. Weld profiles in tubular joints (dimensions in mm), (a) Early American test, (b) American
production weld with profile control, (c) Dutch test, unimproved profile, (d) French test,
(e) Improved profile, Cognac platform (1977).
 159

(hi) Derivability. The third attribute of hot spot stress, derivability from model tests or
analysis of the connection, forms the basis of the parametric SCF used in general practice, and
puts more accurate determination of hot spot sigma-G for a specific design within reach of
designers willing to spend the time and money required. Although accorded equal standing, the
two methods do not necessarily produce identical results. Model tests are the original basis of
the design curve, and the benchmark against which other methods are usually compared; yet they
often exhibit scatter (and bias on the side of under-estimating peak strain) from the following
sources: welds too large in relation to the scale factor (local reinforcement not reflected in the
prototype), inconsistent strain gage technique (gages too large or too far from the weld toe), and
thickness variations (small pipe is typically over nominal). Finite element analysis requires
careful benchmarking of element type, mesh size, interpretation of stress results (e.g., weld toe
vs. mid-plane intersection), and consistency with the design curve; but once calibrated, can be
used for parametric SCF studies free of experimental scatter.
Consistency with the original design curve, taken literally, would mean consistency with
the strain gage techniques of the mid-1960's which produced the data base. Hot spot strains
were measured as near as practical, but not exactly, at the toe of the weld. In the steep stress
gradients of typical hot spot regions, strain gage size and corresponding placement greatly
affected the results obtained. Pickett's original design curve was based on full-size strain gages
(0.25-in. gage length) on full-size prototype hardware. Early model testing at SWRI (ref. 25)
attempted to keep approximate proportionality of weld size (using small short-arc welding
passes) and strain gage size (e.g., 0.125-in. gages for 1:2.5 scale models of 50-inch Gulf of
Mexico joint-cans). This technique became the de facto American standard, as described in
Section 2.3.3, and was extended to smaller scale tests (1:4, ref. 42; 1:6, ref. 36), using the same
strain gage size and placement as a matter of practical convenience. It was even extended to full
size specimens (ref. 48). On the smaller specimens, this "standard" strain gage placement would
be further down the stress gradient, while on the larger ones the strain gages would be in the
influence of the notch stresses due to weld shape ~ certainly not our ideal of an invariant sigma-
G for design, but tending to give a certain consistency in the unmeasured notch effect in
experiments.
Further appreciation of the ambiguity which can exist between different methods of
obtaining hot spot strain experimental, thin shell finite element, and others may be gained
from studying Figure 4.10.
(iv) Empiricism. The fourth desired attribute of hot spot strain is that items not subject
to measurement or calculation be built into the design curve. One such item, weld shape notch
effects, has already been discussed, along with the difficulties it creates. Other items are more
easily neglected in the low-cycle plastic fatigue range (the original Gulf of Mexico focus) than
they are in the long-life elastic range; these include residual stress and stress ratio. For as-welded
structures, it was taken for granted that cycling would be against full tensile yield stress for at
least one end of the hysteresis loop. For significant yielding in the hot spot region, the original
or nominal mean stress would be changed during shakedown, so that it ends up being unknown.
Fortunately, Goodman-type diagrams for such welds (ref. 67) show very weak dependence on
mean stress, so it could be neglected. Thus, total range of stress or strain becomes the relevant
parameter, without reference to mean stress or stress ratio.
(v) Reliability. The fifth attribute, reliability, was previously discussed in connection
with Figure 4.6, in which we saw the design curve falling on the safe side of 97% of the data.
References 5, 6, 7, & 68 discuss fatigue reliability issues in more detail, as they relate to ships
and offshore structures, including consideration of random loadings, extrapolation errors in going
from laboratory to ocean, SCF inaccuracies, and other sources or bias and scatter as well as the
beneficial effects of structural redundancy and inspection.
160

(a)

(c)
(b)

Fig. 4.10. Various representations of hot spot stress, (a) In chord (see key), (b) In branch member, (c)
Finite element mid-plane intersection.

4.3 STRESS CONCENTRATION FACTORS (SCF)

Nominal member axial and bending stresses, as obtained from the global structure
analysis, do not suffice for fatigue analysis of tubular connections. Their fatigue behavior is
governed by the higher localized stresses occurring near the interesection welds. Nominal
punching shear, Vp, comes one step closer to the relevant local stresses, and we shall later see
empirical fatigue criteria based on punching shear. However, the most generally useful criteria
are based on hot spot stress, tfjjg, which has been defined as the worst stress (or strain) range at
the toe of a weld, as measured by an adjacent strain gage in a model test, or calculated with
comparably accurate theory. The stress concentration factor (SCF) is the ratio of this hot spot
stress, to nominal stress. By this definition, the SCF of a straight butt weld is 1.0, and the SCF
for other connections would depend upon the loading pattern (e.g., the ovalizing parameter a),
and upon the particular geometric configuration, as reflected by the thickness parameters r and
7, and the topology parameters j3, , $, and . The reader may wish to refresh his memory on
these parameters by reference to the beginning of Chapter 2.
Methods for local stress analysis have been discussed earlier in the book, e.g., Section
2.3.1. Here we shall summarize results from such analysis in formats useful to the designer.
Figure 4.11 presents the results of a parameter study for stresses in the chord of K-connections.
m us r e l e c t m
Results are given as plots of <Jjjs/Vp, l g effects beyond those already included in the
simple punching shear concept. The empirical AWS category criterion is shown as o^g/V^ =
7, with a 20% scatter band, representing experience. In addition, the following are plotted:

1. Finite elementresults from a Shell Oil parameter study (ref. 69) using the Clough-Greste
computer program (ref. 70) with medium mesh.
 161

(a)

(b)

(c)

Fig. 4.11. K-joint parameter study, (a) Effect of chord wall thinness ratio . (b) Effect of diameter
ratio . (c) Effect of angle .
162

2. Kuangempirical formulas derived from Exxon Production Research parameter studies

using a modification of the same computer program, with fine and superfine mesh (ref.
71, revised from the original OTC paper).
3. FRAMETIresults of a parameter study (ref. 22) based on Dundrova's analytical shell
theory (ref. 73).
4. The venerable Kellogg equation (ref. 74):

W V
P = 1 8
<>
4 5

Figure 4.11(a) indicates a strong effect of chord wall thinness ratio 7 beyond what is
included in the simple punching shear concept. Reduced joint efficiency at high 7 is reflected
here in terms of higher SCF, just as we saw it reflected earlier, Section 3.4, in reduced allowable
stress for static loads. Overall, SCF's are inversely proportional to the 1.5 to 1.8 power of chord
wall thickness. The Kellogg formula falls in the midst of the various criteria, and is closely
similar to the Kuang KT results (a KT joint has a third branch member, as indicated by the
dashed lines).
Figure 4.11(b) indicates that the influence of diameter ratio is fairly weak. For the
Shell finite element results, the arrowheads denote the influence of having = 3 0 ; instead of
0
6 0 as for the rest of the plots. Similarly, the effect of brace angle, beyond the sin# contribution
already accounted for by punching shear, appears mixed, Figure 4.11(c).
The effect of the gap between braces can be significant and is responsible for some of
the variation seen in Figure 4.11(b). For the parameter studies, all joints were nominally
concentric, so that those with small have a large gap. For larger where concentric joints
might have overlapped, the gap was maintained at .04D for the Shell study, with 2-inch
minimum gap for Kuang. Kuang's parametric design equations indicate a beneficial effect for
small gap.
Chord hot spot SCF for bending loads in the branch are plotted in Figure 4.12. The
AWS criteria are plotted to reflect the use of alpha cyclic punching shear:

acting = r sin0
f
a
+
^
6 7 f
b y )
2 +
(^Sfbz)
2
(4.6)

for fatigue. The .67 factor for in-plane bending compares well with the theoretical results. For
out-of-plane bending, the theory confirms that this is a more severe case than axial loading (in K-
connections), as reflected by the 1.5 factor.
Until very recently, Kuang's parameter study has been the most extensive available.
The resulting correlation equations for the SCF in the chord are presented in Table 4.2 for , Y,
and connections. Use of these equations is subject to the following limits:

30 < D/T < 70, 0.1 < g/D < 1

0.2 < t/T < 0.8, 3 < L/D < 20 (for and Y)
0.3 < d/D < 0.8, 30 < < 90.

For cross joints, Table 4.2 gives design equations based on the work of Smedley (ref.
75), modified to reflect the effect of sin# in a conservative manner. Smedley's equations are
derived from physical considerations as well as correlation with acrylic models, and remain valid
for limiting cases, e.g., d/D = 1, provided the modification in Note 1 is followed.
For SCF in the branch, Kuang's equations have been modified by a factor, Q r , to
account for the fact that midplane intersect stresses given by thin-shell theory systematically
overestimate the brace end hot spot stress for welded joints with finite thickness. Using Q r = 1.0
gives Kuang's original equations. The modified equations should be used with a lower bound of
 AXIAL LOAD IN-PLANE BENDING OUT-OF-PLANE BENDING

1 5 12
( t / T ^ ^ V D r ^ s i n e ) - 1 1 00 ( t / T ) - 9 ( 4d / D ) - 0 ( 6s i n e ) - 9
K-l S EE & Y
( T / D ), 6 ( d6 / D6 ) ' 0 9 5
( T / D ) ' 38

K-2 1 O6 ( t / T ) K 0 ( d 6/ D 8) - 1 2 s i ne
S EE K - l S EE &
ZXL
& ( t / D ) - 54
<_> K-3
LL-
t-J 9 87 ?
C O _ (t,T,-Seld/0)-' is.nu)'-55 . 3 r
r

( t / T )1- 3 ^3 ) 13- 6 4 9 (
T D
/1.014
)
. 4 63 . ( t / T ) -8 ( 6 1 n e ?^
& Y J 5
( T . /8 0 D8 e 1) . 2 ( d / D ) ( [ ). 0 /5 7 L ) ( T / D ) - 6 ( d / D ) - 04 2 92 (t/T, ^(s,ne,'- " 0 . 6f5o r
D
' (TjD),014
ia/D)-619

X /D) ^ ^_
3 87 lt/T)(d/D)(s(ne)<^'2^-H.4d ( t /,-8
T (. ,1.5-1.6(d/D) D
>?<15
(t/T)(d/D)(s,n,,ltl/D -,4-W) _ . s.
2T/D)(1 F/T)' (2T/D)'b
(l F/T) 3
S ee (2/0)(1 - f-T;-JJ
n o te ( 1 )

fwn-5*fn/ni-0M
J-4*8 sn
ie
K-L 1.0 + Qr f 2. 8 2 7 ( t / T ) -3 ( 5s J n 8 ) - 5 - l.o] S EE & Y
1.0 .Qr .
8 25 (VT) (9 D 3B
e 1.0 1 L ( d / D ) 35
" J
L 1
57 4
41
(T/D)- (d/0)- J

10 +Q
r < 4 5
(T/D) (d/D)' ^ 1- J -
K-2 S EE K - l S EE & Y
=c

<
K-3 S EE K - l S EE 4 y
LL

CO

f,,T\-S43,.,.S01, ,2.033 1
,.0.Qr [.8.3 i i m _ l ^n( 5 _ i i - i l.oj forjf. .55
& Y 1.0 + .109 ( t / T l f f i l n e ) - 1* - l.ol
Qr
L ( T / D ) ' ^ ( d / D ) 3-8 J

X 1.0 + 0 . 6 3 ( S C F )C HD( ) R 1.0 + 0 . 6 3 ( S C F )C HD Q R 1.0 +0.63 ( S C F )C HD Q R

( 1) A s d / D a p p r o a c h es 1 . 0 , u s e I . D . o f b r a n ch for d
9
164

SCF > 1.8. Where enlarged (conical or thickened) brace-ends are used, both the nominal stress
and the SCF should be based on the actual brace-end dimensions in a consistent fashion.

(a) (b)

L
BRACE
TbO W A L L

CHORD CHORD ,

108 2.7B 108 2.75

3-

DIAMETER RATIO

Fig. 4.12. Parameter study for T-connection with bending, (a) In-plane bending, (b) Out-of-plane
bending.

The value of Q r remains to be specified. Figure 4.13 compares several reduction factors
proposed to reflect the influence of finite thickness and weld size (refs. 75, 76, & 77). Note that
Marshall's expression for Q r would be fairly well represented by a constant value of 5/8. In
Figure 4.14, the result of using this value is compared with isoparametric and experimental
results.

4.3.1 Design Applications

In fatigue design, using allowable hot spot stress for offshore structures, API RP 2A
(ref. 2) recommends the Alpha-Kellogg method for SCF in the chord (ref. 5). In effect, they
combine Equations 4.5 and 4.6 for computing the worst hot spot stress around the circumference
of the tubular joint. For the standard planar connections, alpha values for axial load are assigned
as follows:

connections 1.0
T&Y connections - 1.7
cross connections 2.4 (reduced to 1.7 for of unity)

For multi-planar connections, alpha may be computed as described in Section 6.2 hereinafter.
 165

SMEDLEY

MARSHALL Q r
FOR
A , D/T = 4 0
CLAYTON e " c o s \ F d/D = . 5
WHERE = 1 . 2 8 / / F t F = t

.8 1.0 1.2

FILLET SIZE ( F / T )

Fig. 4.13. Various proposed reduction terms for brace-end SCF (in the branch member).

HOT SPOT =
ISOPARAMETRIC

HOT SPOT =
EXPERIMENTAL

ARROWHEAD DENOTES
ESTIMATED STRESS
AT TOE OF MINIMUM
WELD.
4 6 8 10

THEORETICAL MIDPLANE INTERSECTION STRESS

Fig. 4.14. Empirical reduction term for brace-end SCF (in the branch member).

For grouted connections, Section 6.3 herein recommends computing the cyclic punching
shear (eq. 4.6) using only the outer shell of the chord, but with an enhanced effective gamma
ratio, 7 e ff in Equation 4.5. This recommendation is based on tests of K-connections, which have
minimal chord ovalizing. API notes that the benefits of grouting may be greater for other cases
with more chord ovalizing, i.e., higher alpha values. In analyzing a grouted skirt pile connection
with an unusually severe alpha of 3.8, the author found a good match to disbonded-grout finite-
element results, using an effective alpha and gamma given by

?eff
*eff = 1 + ( a 1 ) (4.7a)

7eff = R/T e f f (4.7b)

166

3 3
ccaann +

= / pile (4.7c)
eff / ,

to reflect this greater benefit in the Alpha-Kellogg format.

For SCF in the branch member, API uses

S C F
branch = Qr
1

6 +
- Jf
6 S C F
chord> >
1 8
-
4
< >
8

with Q r of 5/8. The expression in the brackets is derived in Reference 53 as shown in Figure
4.15. The base SCF value of 1.6 comes from axisymmetric rigid restraint to Poisson's ratio
breathing, as with a solid chord. The horizontal axis parameter was suggested by Toprac (ref.
78), with the radical term representing relative stiffness of the brace-end vis-a-vis the chord: a
/ ratio of unity has been suggested as a "balanced" design (ref. 79). The coefficient of 0.6 was
chosen to fit the data in the figure, which comes from thin shell finite element analysis of axially
loaded K-connections.

0 1 2 3 4 5 6 7 8 9 TO

< S C F ) c hd o r

Fig. 4.15. Correlation plot for hot spot stress in branch members.

The foregoing discussion of API practice is entirely consistent with the intent of AWS
Code rules for hot spot stress; however, API goes into more explicit detail for SCF.

4.3.2 Detailed Analysis

In order to do a reasonably complete fatigue analysis (e.g., ref. 80), it is necessary to
obtain 24 SCF's for each member-end connection, i.e.:
 167

- three kinds of load ~ axial, ; in-plane bending, My; and out-of-plane bending, M z ,
or the branch involved
at least four clock positions around the joint (one might prefer 8 or 12, but
interpolation based on sines and cosines and the assumption of plane sections does
not necessarily represent the actual behavior of tubular connections),
chord and branch side weld toes treated separately.

This results in an array of SCF such as that shown in Figure 4.16.

STRESS CONCENTRATION FACTORS y

1.1 1.7 2.1

1.3 0 1.6
0 2.3 0

1.4 2.2 2.5

1.3 0 1.5
0 3.6 0

80 2.00 ^

Fig. 4.16. Array of SCF for one branch in a simple K-connection, derived from the thin shell finite
element analysis. The Q r reduction has been applied to the branch side SCF.

Rather than doing a finite element analysis of each connection, the Alpha-Kellogg or
Kuang-Smedley parametric equations, previously presented, can be used. Other more recent sets
of parametric equations may also be cited, e.g., Smedley-Wordsworth (ref. 81) and Efthymiou
(ref. 82). They are based on strain-gaged acrylic models and isoparametric thick shell finite
element analysis, respectively, and do not require the Q r reduction. UEG (ref. 83) has also
published a comprehensive set of SCF formulae.
The one-branch-at-a-time approach taken in Figure 4.16 and the parametric equations
suffers from a certain lack of generality, in that hot spot stresses (particularly on the chord size)
are affected by chord loads and the ovalizing effect of other branches present. Thus, the SCF
depend on load pattern and are not unique. Conversely, loads may depend on connection
flexibility. Gibstein's ultimate solution (ref. 84) is to include a shell model of each node as a
substructure of the space frame, and to compute the hot spot stress directly. Cheap super
computers (or deep pockets) and automated mesh generation make this approach feasible.

4.4 S-N CURVES

From a designer's point of view, procedures to predict fatigue behavior from first
principles, using notch factors on data from smooth specimens, or applying fracture mechanics,
168

Fig. 4.17. AWS fatigue curves for redundant structures in atmospheric service. For critical members
whose sole failure would be catastrophic, the allowable number of cycles is one-third of the
values shown.
 169

TABLE 4.3 STRESS CATEGORIES FOR TYPE AND LOCATION OF MATERIAL FOR
CIRCULAR SECTIONS
Stress 1 Limiting weld size or 6
category Situation K i n d s o f stress attachment thickness.

Plain unwelded pipe. TCBR no limit

Pipe with longitudinal scam. TCBR

B u t t splices, c o m p l e t e j o i n t p e n e t r a t i o n g r o o v e w e l d s , g r o u n d flush TCBR no limit

a n d inspected b y R T o r U T (Class R ) .

M e m b e r s w i t h c o n t i n u o u s l y welded longitudinal stiffeners. TCBR

c, B u t t s p l i c e s , c o m p l e t e j o i n t p e n e t r a t i o n g r o o v e w e l d s , as w e l d e d . TCBR 2in (51mm)

c2 M e m b e r s w i t h transverse ( r i n g ) stiffeners. TCBR lin (25 mm)

D M e m b e r s w i t h m i s c e l l a n e o u s a t t a c h m e n t s s u c h as c l i p s , b r a c k e t s , TCBR lin (25 mm)

etc.

D C r u c i f o r m a n d T-joints with complete joint penetration welds TCBR lin ( 2 5 mm)

(except at tubular connections).

DT C o n n e c t i o n s d e s i g n e d as s i m p l e - , Y - , o r K - c o n n e c t i o n s w i t h T C B R in branch member. (Note: l i n ( 2 5 mm)

c o m p l e t e j o i n t p e n e t r a t i o n groove welds c o n f o r m i n g to F i g . 10.13.1 M a i n m e m b e r must be checked for p r o f i l e merging
(including o v e r l a p p i n g connections i n w h i c h the m a i n m e m b e r at s e p a r a t e l y p e r c a t e g o r y K , o r K 2. ) smoothly w i t h adjoining
each intersection meets punching shear requirements). See N o t e 2 bare metal.

Balanced cruciform a n d T-joints with partial joint penetration T C B R i n m e m b e r ; w e l d m u s t also

be checked p e r category F . lin (25 mm)
g r o o v e w e l d s o r fillet w e l d s ( e x c e p t a t t u b u l a r c o n n e c t i o n s ) .

M e m b e r s w h e r e d o u b l e r w r a p , cover plates, l o n g i t u d i n a l stiffeners, T C B R i n m e m b e r ; w e l d m u s t also

be checked p e r category F . lin ( 2 5 mm)
gusset p l a t e s , e t c . , t e r m i n a t e ( e x c e p t a t t u b u l a r c o n n e c t i o n s ) .

ET Simple - , Y - , a n d K-connections with partial joint penetration T C B R in branch member. (Note:

g r o o v e w e l d s o r fillet w e l d s ; a l s o , c o m p l e x t u b u l a r c o n n e c t i o n s i n M a i n m e m b e r in simple - , Y - , o r
w h i c h t h e p u n c h i n g shear capacity o f the m a i n m e m b e r c a n n o t K-connections must be checked
s e p a r a t e l y p e r c a t e g o r y K , o r K 2; 1.5in (38 mm)
c a r r y t h e e n t i r e l o a d a n d l o a d t r a n s f e r is a c c o m p l i s h e d b y o v e r l a p
( n e g a t i v e e c c e n t r i c i t y ) , gusset p l a t e s , r i n g s t i f f e n e r s , etc. S e e N o t e 2 w e l d m u s t also b e c h e c k e d p e r
category F T a n d 10.5.3.)

F E n d w e l d o f c o v e r p l a t e o r d o u b l e r w r a p ; w e l d s o n gusset p l a t e s , Shear in weld.

s t i f f e n e r s , etc.
0 . 7 i n ( 1 8 mm)
S h e a r i n weld (regardless o f direc- throat
F C r u c i f o r m a n d T - j o i n t s , l o a d e d i n t e n s i o n o r b e n d i n g , h a v i n g fillet
or partial j o i n t penetration groove welds (except a t tubular con- t i o n o f loading). See 10.8
nections).

FT S i m p l e - , Y - , o r K-connections loaded i n tension o r bending, S h e a r i n weld (regardless o f direc- lin ( 2 5 mm) throat
h a v i n g fillet o r p a r t i a l j o i n t p e n e t r a t i o n g r o o v e w e l d s . tion o f loading).
. 6 2 5 i n ( 1 6 mm)
x2 Intersecting members at simple - , Y - , a n d K-connections; a n y Greatest total range o f h o t spot
stress o r s t r a i n o n t h e o u t s i d e s u r -
for basic profile;
c o n n e c t i o n w h o s e a d e q u a c y is d e t e r m i n e d b y t e s t i n g a n a c c u r a t e l y
no l i m i t f o r c o n c a v e
s c a l e d m o d e l o r b y t h e o r e t i c a l a n a l y s i s ( e . g . , finite e l e m e n t ) . face o f intersecting m e m b e r s at
as-welded p r o f i l e
the toe o f the weld j o i n i n g t h e m
(AWS a l t , 0 2 ) .
m e a s u r e d after s h a k e d o w n i n m o d e l
or prototype connection or calcu-
l a t e d w i t h best a v a i l a b l e t h e o r y .
l i n ( 2 5 mm)
X| A s for X 2, p r o f i l e i m p r o v e d p e r 10.7.5 a n d 1 0 . 7 . 6 . of AWS D l . l As for X 2 for concave p r o f i l e ;
no l i m i t f o r f u l l y
U n r e i n f o r c e d cone-cylinder intersection. H o t - s p o t stress a t a n g l e c h a n g e ; ground.
X|
calculate per N o t e 4.

S i m p l e - , Y - , a n d K - c o n n e c t i o n s i n w h i c h t h e g a m m a r a t i o R / t co f P u n c h i n g shear f o r m a i n m e m - as f o r X 2
K2
m a i n m e m b e r does not exceed 24. (See N o t e 3) b e r s ; c a l c u l a t e p e r N o t e 5.
as t o r
A s f o r K 2, p r o f i l e i m p r o v e d p e r 1 0 . 7 . 5 a n d 1 0 . 7 . 6 . o f AWS D l . l

Notes to Table
1. = tension, C = compression, = bending, R = reversal i.e., total range of nominal axial and bending stress.
2. Empirical curves based on "typical" connection geometries; if actual stress concentration factors or hot spot strains are k n o w n , use of
curve X ( o r X 2is preferred.
3. Empirical curves based on tests with g a m m a ( # / l c) o f 18 to 24; curves on safe side for very heavy chord members (low / ? / t c) ; for
chord members (R/\.C greater than 24) reduce allowable stress in proportion to
7 0
Allowable fatigue stress / 24 \

Stress f r o m curve 1 fl/lc J

W h e r e actual stress concentration factors or hot-spot strains are k n o w n , use o f curve X , or X 2is preferred.
1
-

4. Stress concentration factor S C F = + 1 . 1 7 tan V Vh

Cos*
where
is angle change at transition
YB is radius to thickness ratio of tube at transition

5. Cyclic range o f punching shear is given b y

J 2
VP = sin [ AFA* V ( 0 . 6 7 / 6, ) + ( l . 5 / 6 ) l ]

where,
and are previously defined, and
FA is cyclic range of nominal branch member stress for axial load.
FBY is cyclic range o f in-plane bending stress.
FHT is cyclic range of out-of-plane bending stress.
A a s d e f i n e d i n s e c t i o n 6.31 I n t h i s b o o k .
6. F o r t h i c k n e s s t e x c e e d i n g l i m i t i n g t 5
allowable fatigue streas I mt V - ^ *
s t r e s s from given curve " \ t r feJ
170

have not been entirely successful for practical as-welded structures, although such studies
continue to provide useful insight and will be discussed in Chapter 7. In the region at the toe of
the weld, there are geometric notches, crack-like flaws, and metallurgical discontinuities of
which the designer has little knowledge and over which he has little control. AWS accounts for
these factors by including them in the data base, developing empirical S-N curves which reflect
their typical effect, and in the process eliminating unnecessary complexities for the designer.
Stress fluctuations are defined in terms of total stress range, peak to trough. Mean stress
is usually unknown, due to the presence of residual stress from the heat of welding. When there
is localized plastic deformation during the initial application of high loads, a new set of residual
stresses develop. What is usually measured is the total strain range, with the zero point
undefined, and this can also be used beyond the elastic range.
AWS fatigue criteria are presented as a family of S-N curves, with the situations for
which each is applicable categorized in an accompanying table. These are reproduced herein as
Figure 4.17 and Table 4.3, respectively.
Curves A, B, C j , D, E, F, and G are consistent with AISC criteria (Appendix to the
Design Specification) except that a continuous curve is drawn rather than using tabulated step-
function allowables. In these simple situations, the nominal member stresses fa and f^ represent
the actual stress which would be measured in the vicinity of the weld, but free of stress
concentrations.
Category X reflects current American design practice for offshore structures. The
relevant stress for fatigue of tubular connections is the hot spot stress measured by a strain gage
adjacent to the weld, as shown in Figure 4.18, or obtained using the SCF discussed previously.
Category X j is consistent with Category C j in that we are dealing with the local transverse stress
which would be measured adjacent to the weld in both cases, with the weld merging smoothly
with the adjoining base metal. Because the localized stresses in tubular joints are frequently in
the inelastic range, it is more appropriate to deal in terms of hot spot strain range rather than
stress.

-
(a)
\.v.
.../,

b AWS

STRAIN
GAGE
API X

102
1
103
1
I0
1
105
N0-CYCLES TO FAILURE
L_
IQ6 \Q7 \QB
(c) (b)
Fig. 4.18. Original data base, various as-welded geometries, (a) Nominal stress adjacent to weld, (b)
Hot spot strain adjacent to weld, (c) S-N curves and data. N3 is at joint separation.

The original data base for AWS Category X was previously discussed in Section 4.2.
The AWS S-N curves were "modernized" in 1980 to conform to the style used in the bridge
design section of the codei.e., with an endurance limit or cut-off stress, rather than a flattening
of slope in the high cycle region.
Within the range of the data, ordinary structural-quality steel was found to fall within
the same scatter band, independent of yield strength in the range of 36 to 100 ksi (250 to 700
2
N/mm ).
 171

Relatively little data is available for the fatigue behavior of welded steel subjected to
non-stationary random loading in the seawater environment with varying degrees of cathodic
protection. Hartt's five-year project illustrates the difficulty of such an undertaking (refs. 85-87).
Considerations of corrosion fatigue and cumulative damage under variable amplitude loading
suggest that the API fatigue criteria, with the endurance limit deferred to 2 10^ cycles, the
AWS X-modified with no endurance limit, or a flattening of the slope as shown by the dashed
line in Figure 4.1 would be a more appropriate extension into the low-stress high-cycle range.
This subject is reviewed in more detail elsewhere (refs. 11, 88-94).
When using curve X fatigue criteria, a compatible set of stress concentration factors are
needed. Using the SCF formulas of Table 4.2 to estimate hot spot stress (instead of strain gage
measurements), available fatigue test data for tubular K-joints (ref. 54) have been replotted in
Figure 4.19. Concentrating first on the circular data points for non-overlapping joints, we see
fairly tight correlation with the design curve, with bias and scatter compatible with the original
experimental correlation of Figure 4.18. For overlapping joints (triangular data points) these
criteria still seem workable, although they fall rather far to the safe side and are strictly speaking
outside the range of the SCF formulas.

crCL&5 TO

Fig. 4.19. WRC data base: K-connections with chord dia. > 8" (200mm) and thickness > 3/16"
(5mm).

AWS criteria have, since 1986, included explicit considerations for fatigue size effect.
Thickness limitations are shown in the right-hand column of Table 4.3, with reductions in fatigue
strength for heavier thicknesses as given in Note 6 to the table.
Since microscopic effects at the toe of the weld are build into the S-N curve, the
designer need notindeed should notinclude them again in his SCF. However, these effects do
vary with the quality of weld, especially the surface profile, and do affect the choice of S-N
curve. Figure 4.20 sets forth the traditional classification of surface quality and weld profiles
(ref. 53), corresponding to fatigue curves A, C, and D, appearing in AWS. Curve A is for
smooth plate. For ordinary tee welds (class D), there is typically a very abrupt transition at the
172

FULLY GROUND
PROFILE CLASS
A

APPLICATION

Kf3

MODEL MILL SCALE

AWS AWS C a X AWS A

LLOYDS-BASIC IMPROVED PROFILE BSI53A.C
BSI53F BSI53D

Fig. 4.20. Original AWS profile classification (1973).

toe of the weld, for which the local scale stress concentration factor, Kf, and the fatigue strength
reduction factor relative to curve A in the high cycle region, is approximately 3. In order to be
compatible with AWS fatigue criteria (i.e., curve X), API RP 2A requires that a special effort be
made to achieve an as-welded surface which merges smoothly with the adjoining base metal,
corresponding notionally to Kf of 1.7. Grinding the weld (class A) further improves the fatigue
performance. For welds without profile control, the more abrupt (and larger) notch at the toe of
the weld lowers the fatigue performance of the connection, as indicated by the lower AWS curve
X2 and API curve X ' , which appeared in the 1980 editions of these standards. Figure 4.21
compares tests in seawater (as of 1979, refs. 95-99) with API fatigue curves X and X \ These
design curves provide a reasonable lower bound to the data, although the free corrosion data
crowd a bit closer to the curve than the air data shown in the earlier figures.

Fig. 4.21. Tests in sea water (as of 1979).

 173

Integrated treatment of weld profile and fatigue size effects can be found in today's
AWS Code. The background of theory and testing is described more fully in Chapter 7.
On a somewhat lower level of sophistication than using explicit SCF, empirical S-N
curves are provided by AWS for various classes of tubular joints, using readily calculated
punching shear or nominal stresses as their basis. For checking the chord in T-connections and
K-connections, punching shear S-N curves are presented in Figures 4.22 and 4.23, respectively,
together with their supporting data. This data base was originally collected by Toprac (ref. 44).
It covers a rather narrow range of chord thinness ratio, gamma, as shown in the figures. For
higher values of gamma, the correction given in note 3 to Table 4.3 is required to avoid errors on
the unsafe side.

lOOh

Fig. 4.22. Punching shear ~ T-connections.

I00h

a5

Fig. 4.23. Punching shear - K-connections.

In addition, the brace end should be checked using nominal stress S-N curves DT and
ET, as shown in Figures 4.24 and 4.25. These curves imply brace-end SCF's of 2.5 for clean
simple tubular connections and 4.0 for messy gusseted and overlapping joints, respectively. It is
presumed that these curves would govern only when the chord joint can or other reinforcement is
designed so as not to be the weak link. Punching shear criteria for checking the chord in simple
tubular connections have already been presented.
 174

lOOh
".
<
cr

CO
CO
LU rr
rr
h-
co

<
2 SIMPLE
JOINT

CYCLES
Fig. 4.24. Nominal stress in simple joints.

Lu OVERLAP
-.
<
rr
CO
CO
lu
rr
-

-j
<
2: GUSSET

H
10* 10 \0 10 10 10'
CYCLES
DATA I N ASCE
GUSSETED JOINTS \
N EW D A AT
* OVERLAPPING J O I N T S /

INTERSECTING JOINTS
If W H I C H F A I L E D BY
P U N C H I N G S H E A R IN
CHORD
Fig. 4.25. Nominal stress in externally stiffened and overlapped joints.

For complex joints, Figure 4.26 may be of some use (refs. 37, 100, 101). Here the
sustainable cyclic load amplitude is related to the static strength as calculated by methods
presented elsewhere. Tubular connections which are detailed with an efficient transfer of load
and which develop 100% of the brace strength have generally been found upon finite element or
experimental stress analysis to have an SCF of about 3. Tubular connections with less than
100% efficiency would be expected to have proportionately higher SCF. This approach was
 175

confirmed for ring stiffened circular nodes (ref. 102). However, for other types of connections, it
should be regarded as an approximation and more accurate determination of SCF for the actual
configuration is preferred.

1 I I I I I I
2 3 4 5 6
0 10 10 10 10 10 10

CYCLES TO FAILURE
Fig. 4.26. Fatigue criteria based on static strength efficiency.

4.5 COMPARISONS WITH INTERNATIONAL RULES AND DATA

In the last 15 years, a large amount of research on the fatigue behavior of tubular
connections has been conducted in Europe, spurred mostly by offshore developments in the
North Sea. Research in Japan has been going on over an even longer time span, covering a
diverse range of applications ranging from mobile drilling rigs to onshore structures. There is
surprisingly little overlap between this work and the data base presented earlier as the basis of
AWS provisions. Thus, comparisons of the author's earlier work and AWS criteria against
European and Japanese rules and data can be viewed as an opportunity for independent
confirmation.
Figure 4.27 compares the upper and lower AWS design curves for hot spot stress with
two other criteria.

(1) DEn-T is the most widely accepted European design curve (ref. 57). It is based strictly
on geometric hot spot stress, sigma-G. The regulation includes the same thickness
correction as AWS, and is shown at the reference thickness of 30-mm. The log-log
slope of -3.0 is consistent with fracture mechanics crack growth laws. The flattening of
slope, instead of an endurance limit, in the high cycle region is in consideration of
random loading, similar to Figure 4.1.

(2) The NK curves are based on Japanese mobile rig rules (refs. 103, 104) and the work of
Iida (refs. 6 1 , 105). The rules themselves are based on very localized (microscopic)
stresses, including specific consideration of weld toe notch effects, but are plotted here
176

in terms of geometric hot spot stress for purposes of comparison. Fatigue crack
initiation considerations and the effects of local yielding result in a log-log slope similar
to AWS criteria.

FATIGUE LIFE
Fig. 4.27. Various S-N curves, based on geometric hot spot stress OQ = Ee t r.

The next several figures compare various hot spot design curves with the European data
base as summarized by van Delft (ref. 106). This data base covers a wider range of thicknesses
than its American counterpart (up to 3-in or 75-mm), but a narrower range of geometries (being
mostly simple and double-T connections). Measured hot spots stresses are strictly geometric
sigma-G, as defined in Figure 4.8. The welding in these specimens generally does not follow the
concave "improved" profile of AWS, but tends to have flat profiles as shown in Figures 4.9(c)
and (d). (Some early UKOSRP results are misleading, as their "acceptable" profiles do not meet
the current AWS disc test.) The combined result of large scale, sigma-G, and flat profile is larger
unmeasured notch effect, leading as would be expected to lower fatigue performance.
Comparisons of design curves versus the data base, based on the use of parametric SCF
formula to calculate hot spot stress, is consistent with the way they are used in practice, and is in
the same spirit as the earlier comparison of Figure 4.19. Two sets of SCF formulae are compared
here: Kuang's (EPR) and Efthymiou (SHELL).

CYCLES TO CRACK THROUGH IN3) CYCLES TO CRACK THROUGH ( N 3 )

Fig. 4.28. Design curves compared to the European data on the basis of calculated hot spot stress
range. API is lower curve X'. DNV is similar to API upper curve X. (a) Using Kuang SCF
formulas, (b) Using Efthymiou SCF formulas. Data have not been corrected for size
effects; symbol size is proportional to thickness (ref. 106).
 177

(a)

(b)

4.29. Selection of design S-N curve based on unmeasured local SCF. (a) AWS requirements for
"improved" weld profile, (b) Design S-N curves. Upper curve is for weld profiles merging
smoothly with the adjoining base metal.
178

Figure 4.28 shows comparisons for DNV and API criteria. Since these criteria do not
include a size effect correction, the data are treated likewise. The scatter band is large, covering
two log cycles. Some of the data for the thicker specimens clearly falls below the upper design
curve, as might be anticipated. At the 1978 Cambridge conference (ref. 89), this was alarming
news. The scary data points are safely contained by the lower curve, which was based on the
evidence presented in Figure 4.29, and appeared in American standards in 1980. However, the
"alanning news" continued to make headlines in the trade journals for many years.
In 1981, the UK Department of Energy (DEn) proposed the "T" curve, (ref. 57) as well
as a size effect correction on fatigue strength, based on a reference chord thickness of 30-mm
(1.2-in):

fatigue strength thickness - 0 . 25

(4.9)
design curve reference thickness

Alternative (generally more severe) values of the exponent have been suggested, e.g., -0.40 (ref.
109) and -0.075 log (ref. 110). However, these have not yet been adopted into any design
codes.
Using leads from the 1981 Paris Conference (ref. 90) and the 1984 Houdremont lecture
(ref. 63), AWS adopted an integrated treatment of weld profile and size effects in the 1986 Code.
When the basic flat weld profile is used for all thicknesses, this reduces to Equation 4.9, with a
reference branch member thickness of 0.625-in (16-mm), using the lower S-N curve. With
typical design practice yielding of 0.5, this corresponds to chord thickness of 1.25 in. (32mm),
which is similar to DEn.
Both DEn and AWS criteria are compared with the European data base in Figure 4.30.
The size effect correction has been applied to the data, for thicknesses exceeding the reference
thickness; van Delft's correction to 30-mm chord thickness approximates correction to 16-mm
branch thickness, for typical r ratios. This correction slightly reduces the scatter and increases
the conservatism of the lower design curve. Much of the data drops out of the EPR plot, because
of connection geometries outside the validity limits of the Kuang formulae.

(a) (b)
;j m i n \^ itiii r - 1 1 1 III!
:
EPR
ivAtio cnsES) ;

\
\ * \
* \
\
\\ \^

\ JX A
X \ A
\ :
A

^ \ V>

T-J01NT
\ \ \ a
V
- C P .
1 1 1 1 11111 , l i f t m i l
2 3 15 7 9 52

5 3 IS 7 9 . 2 3 MS 7 9
Sfcfm 12 1 31 HS1 Mil7 9 ,
7 u 2 3 H5 7 9 52s 3 IS 7 9 2 3*579
6 2773 MS 7 9 .
1CT 10 10 ' 10* 1 0 10* '
CYCLES TO CRACK THROUGH I N 3 ) CYCLES TO CRACK THROUGH ( N 3 )

Fig. 4.30. Design curves compared to the European data on the basis of calculated hot spot stress range.
DEn is U.K. Dept. of Energy curve "T". AWS is lower curve X2. (a) Using Kuang SCF
formulas, (b) Using Efthymiou SCF formulas. Data for thicker specimens have been corrected
to 30mm chord (~ 16mm branch) using -0.25 power rule; symbol size is proportional to
thickness (ref. 106).
 179

Comparisons based on actual measured hot spot stresses and strains might, at first, be
considered superior to those based on design approximations to the SCF. However, experimental
measurements of SCF are also subject to error, as demonstrated in Figure 2.20 and discussed in
Section 4.2.2(iii), especially in the case of in-plane bending where a linear stress gradient is hard
to find.
Figure 4.31 and 4.32 show comparisons of measured hot spot stresses and strains,
respectively, with DNV and API criteria, again without size effect correction. The comparison is
less favorable than before, especially in the case of strains. That the use of strains and SNCF
would be less conservative in the high cycle region was previously discussed in Section 4.2.2(ii).
Here we see roughly a 15% effect.
"I I llll 1I

T-JOINT T-JOINT
Y-JOINT Y-JOINT
X-JOINT X-JOINT
K-JOINT K-JOINT
%C.P. * C P .
n m ' m i \ i IJS.IH1I ^ 1 I I HljJ
2 311579 253 45 7 9 2fi 3 1 5 7 9 2 73 4 5 7 9 u 2 3 4 5 7 9 c 2 3 4 5 7 9 2 3 4 5 7 9 2
7 3 45 7 9 o
8
10 10 10 10 10 10 10 10 10 10
C Y C L E S TO CRACK THROUGH ( N 3 ) CYCLES 0 CRACK THROUGH ( N 3 )

Fig. 4.31 The measured hot spot stress Fig. 4.32 The measured hot strain range
range compared with the API compared with the lower API
and DNV design curve. Data design curve. Data not
not corrected for size effect corrected for size effect
(ref. 106). (ref. 106).
1 1 1 11 llll 1I 1 1 1 llll 1 1 1 II III
.
- \ MEASURED ;

(b) ^ \ \

500 k . \
: <o\ /tvr\\.
400

' T-JOINT
' <!> Y - J O I N T v \
- X-JOINT
K-JOINT
* C P .
1 1 I I I Mil 1 TVVjiii mi
2 3 4S 7 9 c 2 3 45 7 9 27 3 4 5 7 9 8 2 3 4 5 7 9 2 3 4 5 7 9 c 2 3 4 5 7 9 2 73 4 5 7 9 .
10" 10 10 10 10 10 10 10 10
CYCLES TO CRACK THROUGH ( N 3 ) CYCLES TO CRACK THROUGH ( N 3 )

Fig. 4.33. Design curves compared to the European data on the basis of measured hot spot stress
range. DEn is U.K. Dept. of Energy curve "T" (30mm reference chord thickness). AWS is
lower curve X2 (16mm reference branch thickness). Thickness correction applied to data
using -0.25 power rule, (a) Thicker specimens corrected to reference thickness, (b) All
specimens corrected to reference thickness. Symbol size is proportional to specimen
thickness (ref. 106).
180

Figure 4.33 shows satisfactory comparisons of with DEn and AWS criteria, with the size
effect correction. In the AWS rules, branches less than 0.375-in (10-mm) thick are permitted to
use the upper design curve, in effect making a size effect correction for thinner as well as thicker
material, as shown in part (b) of the figure. This brings the thinner specimens in line, reduces the
"shotgun" scatter, and finally yields correlation lines having a sensible slope.
Finally, we re-visit criteria relating fatigue strength to ultimate strength, in Figure 4.34.
Kurobane's data (refs. 103 and 107) are compared to Marshall's criteria, which anticipates
effects of cycle ratio R (strengh based on peak load) at the low cycle end, and fatigue strength
independent of yield strength (thus varying the ratio to ultimate load) at the high cycle end. The
recent work of deBack (ref. 108) confirms the beneficial effect of increasing yield strength, as
well as a flattening of the S-N slope and reduced size effect, at the low cycle end.
ioi-

Fig. 4.34. Fatigue test results for tubular T-joints. Fatigue strength is represented as a ratio to static
strength according to Japanese rules, similar to IIW. The dashed lines represent the
Marshall criteria from Fig. 6.26 (refs. 103 and 107).

4.6 SUMMARY

In this chapter, we have reviewed the American practice for fatigue design and analysis
of tubular connections. Although this evolved from offshore structures, the basic approach is
applicable to other situations.
Since most design codes treat fatigue in terms of nominal stress, AWS also provides
criteria for this format. Cyclic punching shear is one step closer to appropriate understanding of
tubular connections. However, hot spot stress is the most useful criterion, bringing many
different connection geometries to a common basis. Stress concentration factors (SCF),
derivable from finite element analysis or strain gages, provide the designer the means to calculate
the hot spot stress.
Ideally, the designer would prefer to deal with the geometric hotspot stress, sigma-G, for
which the SCF are invariant for a given connection geometry. Weld notch effects, profile
variations, residual stress, size effect, etc., issues which are not amenable to design calculations,
should be included in the empirical design criteria.
While American and European practices differ somewhat in the degree of invariance
they provide, and in their treatment of the abovementioned issues, the utility of the hotspot stress
approach has been reaffirmed by two decades of international research since it was first
published.
 181

REFERENCES

1 Marshall, P. W., Tubular Joint Design, Chapter 18 in B. McClelland, et al, Planning and Design of
Fixed Offshore Platforms, van Nostrand Reinhold Company, New York, 1986 (see pp. 652-687).
2 American Petroleum Institute, Recommended Practice for Planning, Designing Fixed Offshore
Platforms, API RP 2A, 18th ed., 1989.
3 Marshall, P. W., Basic Considerations for Tubular Joint Design, WRC Bulletin 193, April 1974.
4 Strating, J., Fatigue and Stochastic Loadings, doctoral dissertation, . H. Delft, May 1973.
5 Marshall, P. W. and Luyties, W. H., Allowable Stresses for Fatigue Design, Proc. 3rd International
Coinference on Behavior of Off-Shore Structures, BOSS-82, MIT Cambridge, 1982.
6 Wirsching, P. H., Probability-Based Fatigue Criteria for Offshore Structures, final report 2nd year,
API PRAC project 80-15,1981.
7 Munse, W. H., Fatigue Criteria for Ship Details, Proc. Extreme Loads Response Symposium,
SNAME/Ship Structure Committee, Arlington, VA, 1981.
8 Haibach, E., The Allowable Stress under Variable Amplitude Loading of Welded Joints, in
Fatigue of Welded Structures, The Welding Institute, Cambridge, July 1970.
9 Marshall, P. W. and Toprac, . ., Basis for Tubular Joint Design, Welding Journal, Research
Supplement, May 1974.
10 Kurobane, Y., et al, Fatigue Design of an Offshore Structure, Proc. Offshore Technology
Conference, OTC 2607, May 1977.
11 Marshall, P. W., Problems in Long-Life Fatigue Assessment for Fixed Offshore Structures, ASCE
Preprint 2638, San Diego, April 1976.
12 Peterson, R. E., Fatigue Metals in Engineering and Design, ASTM Marburg Lecture, 1962.
13 Neuber, H., Theory of Stress Concentration for Shear Strained Prismatical Bodies with Arbitrary
Non-Linear Stress-Strain Law, Proc. ASME Journal ofApplied Mechanics, December 1961.
14 Manson, S. S., Fatigue: A Complex Subject - Some Simple Approximations, Experimental
Mechanics, July 1965.
15 Pickett, A. G., et al,Full-Size Pressure Vessel Testing and its Application to Design, ASME Paper
63-WA-293, 1963.
16 Rules for Construction of Nuclear Power Plant Components, Division 1 of Section III (Pressure
Vessels) of the ASME Boiler and Pressure Vessel Code, 1973.
17 Pellini, W.S. and Puzak, P. O., Fracture Analysis Diagram Procedures for the Fracture-Safe
Engineering Design of Steel Structures, WRC Bulletin 88, May 1963.
18 Pellini, W. S., Status and Projections of Developments Hull Structural Materials for Deep
Ocean Vehicles and Fixed-Bottom Installations, U.S. Naval Research Laboratory Report 667,
Washington, DC, 1964.
19 Marshall, P. W., Risk Factors for Offshore Structures, CDG Report 5, Shell Oil Company, New
Orleans, October 1965 (also see ASCE Journal Structural Division, December 1969).
20 Marshall, P. W., Design of Simple Tubular Joints, presented at May 1966 Shell Offshore
Engineering Conference, New Orleans (also see ref. 9).
21 Munse, W. H. and Grover, L., Fatigue of Welded Steel Structures, Welding Research Council,
New York, 1964.
22 Betero, V.V. and Popov, E. P., Effect of Large Alternating Strains on Steel Beams, ASCE Journal
Structural Division, ST1, February 1965.
23 Bouwkamp, J. G., Report of Progress on Tubular Fatigue Program, memorandum September 1964
(proprietary).
24 Bouwkamp, J. G., Tubular Joints under Static and Alternating Loads, University of California,
Structures and Material Research Report, No. 66-15, June 1966.
25 Grigory, S. C , A Study to Develop a Design Procedure for Analysis of Plastic Fatigue Life of
Tubular Joints in Offshore Structures, Southwest Research Institute final project report 03-1882,
May 1969 (proprietary).
26 Holliday, G. H., Welded Tubular Joint Research, quarterly progress reports, August 1966 -
September 1968, Shell Development Company (proprietary).
27 Marshall, P. W., Middle Ground Shoal Platform "C" Structural Joint Model Test, progress reports
January 1967 and November 1967, Shell Oil Company, CDG, New Orleans (proprietary).
182

28 Dunlop, A. R., et al, Appraisal of Corrosion and Materials Problems Associated with Bottom-
Supported Structures in Deep Water, series of papers presented at Shell's 1966 Chemical
Engineering Conference.
29 Kampschafer, G. E., et al, Engineering Data for the Design and Fabrication of Offshore Drilling
Platforms with Heat-Treated Steels, ASME paper, Houston, TX, July 1965 (also see OTC 1046).
30 Toprac, . ., et al, Welded Tubular Connections: An Investigation of Stresses in T-Joints,
Welding Journal, January 1966.
31 Brown, R. C , An Experimental Investigation of T-Joints, MS Thesis, University of Texas, January
1966.
32 Chen, J., Stresses in Steel Tubular Y-Joints, MS Thesis, University of Texas, January 1966.
33 Horton, S. B., Investigation of Elastic Stresses in Welded Tubular Steel K-Joints, MS Thesis,
University of Texas, January 1966.
34 Kurobane, Y. (in charge), Fatigue Tests of Tubular T-Joints, progress report up to December 31,
1966, University of Texas.
35 Toprac, A. A. and Natarajan, M., The Fatigue Strength of Tubular T-Joints, University of Texas,
Structures Fatigue Research Lab Report, November 1968.
36 Toprac, A. A. and Natarajan, M., An Investigation of Welded Tubular Joints: Progress Report,
IIW Doc. XV-265-69, June 1969.
37 Marshall, P. W., et al, Materials Problems in Offshore Structures, Proc. Offshore Technology
Conference, OTC 1043, May 1969.
38 Minutes of the AWS s/c on Welded Tubular Structures, October 9,1969, New Orleans.
39 Marshall, P. W., et al, Report of team "K", Design Stresses, AWS s/c on Welded Tubular
Structures, February 6, 1970.
40 Report of Subcommittee 10, Welded Tubular Structures, minutes of the AWS Structural Welding
Committee, December 14-15,1971, Pittsburgh.
41 AWS Structural Welding Code, AWS Dl.1-72, first edition, American Welding Society, 1972.
42 Grigory, S. C , Experimental Stress Analysis and Fatigue Test of a Cook Inlet Offshore Platform
Joint, final report SWRI project 03-2004, February 1970 (also see ref. 27).
43 Nibbering, J. K. W., Testing of Two Variations of a Pipe Connection Applied in a Drilling Rig,
Ship Structures Lab Report 113, Delft.
44 Toprac, . ., Design Considerations for Welded Tubular Connections, draft prepared for WRC,
December 1970.
45 F. Brink, et al, Stress Analysis of a Tubular Cross Joint Without Internal Stiffening for Offshore
Structures, Proc. International Conference Welding in Offshore Constructions, Newcastle-upon-
Tyne, February 1974, The Welding Institute.
46 Bouwkamp, J. G., et al, Tubular Joints Under Alternating Loads (Phase II), University of
California Structures and Material Research Dept., No. 67-29, November 1967.
47 Bouwkamp, J. G., et al, Tubular Joints Under Static and Alternating Loads (Phase II, Part 2),
University of California, Structures and Materials Research, Report No. 70-4, March 1970.
48 Bouwkamp, J. G. and Becker, J. F., Fatigue Failure of Welded Tubular Joints, Proc. Offshore
Technology Conference, OTC 1228, May 1970.
49 Marshall, P. W., Sedco Binder, May 1968 (investigation of Bruyard loss and redesign of H-unit).
50 Bell, A. O. and Walker, R. C , Stresses Experienced by an Offshore Drilling Unit, Proc. Offshore
Technology Conference, OTC 1440, May 1971.
51 Kochera, J. W., Basic Mechanical Properties of Construction Steels - Low Cycle Fatigue Properties
of Structural Steels for Offshore Drilling and Production Platforms, Shell Development Company,
Technical Progress Report 281-70, February 1971, Emeryville, CA.
52 Kochera, J. W. and Marshall, P. W., Fatigue of Structural Steel..., Proc. Offshore Technology
Conference, OTC 2604, May 1976.
53 Marshall, P. W., General Considerations for Tubular Joint Design in Offshore Structures, Shell-
Arco-Mobil Gulf of Alaska Criteria & Platform Study, ODC Report No. 49, August 1973 (also see
ref. 3).
54 Rodabaugh, E. C , Review of Data Relevant to the Design of Tubular Joints for use in Fixed
Offshore Platforms, WRC Bulletin 256, Welding Research Council, New York, January 1980.
 183

55 Gibstein, . ., et al, Fatigue of Tubular Joints, in Fatigue Handbook for Offshore Structures,
A. Almar-Ness ed., Tapir, Trondheim, 1985.
56 Radenkovic, D., et al, Une Remarque sur la Conference Houdremont 1984 du point de vue de
L'analyse des Contraintes (Amicus Plato...), IIW Doc. XIII-1163-85 (in French, English and
Latin).
57 Snedden, N. W., Background to Proposed New Fatigue Design Rules for Steel Welded Joints in
Offshore Structures, report of the Department of Energy Guidance Notes Revisions Drafting Panel,
May 1981.
58 Dijkstra, O. D. and de Back, J., Fatigue Strength of Welded Tubular T- and X-Joints, Proc.
Offshore Technology Conference, OTC 3639, May 1980.
59 Gibstein, M. G., Stress Concentration in Tubular Joints, its Definition Determination and
Applications, Proc. International Conference, Steel in Marine Structures, Paris, October 1981.
60 SAE Fatigue Design Handbook, AE4, Soceity of Automotive Engineering, Warrendale, PA, 1968.
61 Yoshida, K., Iida, K., et al, Behavior Analysis and Crack Initiation Prediction of Tubular T-
Connections, Proceedings Offshore Technology Conference, OTC 2854, May 1977.
62 Lawrence, F. V. and Burk, J. D., Estimating the Fatigue Crack Initiation Life of Welds, Fatigue
Testing of Weldments, ASTM STP 648, Philadelphia, 1978.
63 Marshall, P. W., Assemblages pour Structures Tubulaires Soudees, Conference Houdremont,
presentee a la Conf. Int'l. de la Soudage des Structures Tubulaire, IIW/IIS, July 1984, Pergamon
Press, Boston (also in English).
64 Dijkstra, O. D. and Noordhoek, C , The Effect of Grinding and Special Weld Profile on the Fatigue
Behavior of Large-Scale Tubular Joints, Proc. Offshore Technology Conference, Houston, May
1985.
65 Descartes, R., Discourse on Method, Leiden, 1637.
66 Sherr, G. H. (ed.), Journal of Irreproducible Results, Workman Publishing Co., New York,
1983.
67 Munse, W. H. and Grover, L., Fatigue of Welded Steel Structures, Welding Research Council,
New York, 1964.
68 Marshall, P. W., Failure Modes of Offshore Platforms - Fatigue, Proc. 1st International Conference
on Behavior of Off-Shore Structures, BOSS-76, Vol. II, NTH Trondheim, 1976.
69 Marshall, P. W., A Review of SCF in Tubular Connections, Shell Oil Company, CE-32 Report,
1978.
70 Greste, O., A Computer Program for the Analysis of Tubular K-Joints, University of California,
Structures and Materials Research Report No. 69-19, November 1969.
71 Kuang, J. G., et al, Stress Concentration in Tubular Joints, SPE Journal, August 1977.
72 Caulkins, D. W., Parameter Study for FRAMETI Elastic Stress in Tubular Joints, Shell Oil
Company, CDG Report 15, September 1968.
73 Dundrova, V., Stresses at the Intersection of TubesCross and Tee Joints, University of Texas
SFRL Technical Report P-550-5 (1966).
74 Toprac, . ., et al, Welded Tubular Connections: An Investigation of Stresses in T-Joints,
Welding Journal, January 1966.
75 Smedley, G. P., Peak Strains at Tubular Joints, presented at the Institute of Mechanical Engineers,
seminar on Corrosion Fatigue in Offshore Installations, September 1977.
76 Marshall, P. W., et al, Limit State Design of Tubular Connections, Methods of Structural
Analysis, proceedings of the ASCE Conference, Madison, WI, August 1976.
77 Clayton, A. M., Effect of Weld Profile on Stresses in Tubular T-Joints, UKOSRP report 2/03,
August 1977.
78 Toprac, . ., et al, Analysis of In-Plane -, Y- and K-Welded Tubular Connections, University of
Texas SFRL Technical Report P-550-3, March 1965.
79 Bryant, J. E., Circular Tubular Joint Design, Tulane University, MS Thesis, 1962.
80 Kinra, R. K. and Marshall, P. W., Fatigue Analysis of the Cognac Platform, Journal of Petroleum
Technology, SPE, March 1980. '
81 Wordsworth, A. C. and Smedley, G. C , Stress Concentration at Unstiffened Tubular Joints,
European Offshore Steels Research Seminar, The Welding Institute, Cambridge, November 1978;
184

also A. C. Wordsworth, Stress Concentration Factors at and KT Tubular Joints, paper 7, Fatigue
in Offshore Steel, Institute of Civil Engineers, London, 1981.
82 Efthymiou, M., et al, Stress Concentration in T/Y and Gap/Overlap K-Joints, Proc. BOSS-85,
Delft, July 1985.
83 Billington, C. J., et al, Design of Tubular Joints for Offshore Structures, UEG Offshore
Research/CERIA, London, 1985.
84 Gibstein, . B., et al, Refined Fatigue Analysis Approach and its Application in the Veslefrikk
Jacket, International Symposium on Tubular Structures, Lappeenranta, Finland, September 1989.
85 Hartt, W. H., et al, Spectrum Fatigue Properties of Welded Steel as Applicable to Offshore
Structures, Report to API, April 1983.
86 Qian, Dong-Shi and Hartt, W. H., Evaluation of Spectrum Fatigue Data Under Conditions
Applicable to Welded Steel Offshore Structures, Proc. Offshore Technology Conference, OTC
4773, May 1984.
87 Hartt, W. H. and Lin, N., Variable Deflection Fatigue Properties of Welded Steel as Applicable to
Offshore Structures, Final Report to API, March 1985.
88 Broek, D., et al, Interpretive Report on Corrosion Fatigue of Welded Carbon Steel for Application
to Offshore Structures, Battelle Memorial Institute Report to API, Columbus, OH, February 1977.
89 Proceedings of the European Offshore Steels Research Seminar, The Welding Institute,
Cambridge, November 1978.
90 Steels in Marine Structures, Proceedings of the International Conference, Paris, 1981.
91 Burnside, . H., et al, Long Term Corrosion Fatigue of Welded Marine Steels, Report SSC-326,
Ship Structure Committee, Washington, DC, 1984.
92 Hudak Jr., S.J., et al, Analysis of Corrosion Fatigue Crack Growth in Welded Tubular Joints, Proc.
Offshore Technology Conference, OTC 4771, May 1984.
93 Noordhoek, C. and deBack, J. (editors), Steels in Marine Structures, Proceedings of the
International Conference SIMS-87, Elsevier, Delft, June 1987.
94 Wardenier, J. and Reusink, J. H. (editors), Fatigue Aspects in Structural Design, Delft University
Press, September 1989.
95 Hartt, J. H., et al, Influence of Seawater and Cathodic Protection upon Fatigue of Welded Plates, as
Applicable to Offshore Structures, Final Report, first two-year effort, API PRAC Project 12,1980.
96 Solli, D., Corrosion Fatigue of Welded Joints in Structural Steels and the Effect of Cathodic
Protection, in Reference 89.
97 Berge, S., Constant Amplitude Fatigue Tests Performed on Welded Steel Joints in Seawater, in
Reference 89.
98 Booth, G. S., Constant Amplitude Fatigue Tests Performed on Welded Steel Joints in Seawater, in
Reference 89.
99 DeBack, J., et al, Fatigue Behavior of Welded Joints in Air and Seawater, in Reference 89.
100 Akiyama, H., Contribution to the MonographTC43, Connections, Joint Committee on Tall
Buildings, 1974.
101 Marshall, P. W., Tubular Joint Design, lecture at WEGEMT, Trondheim, January 1979.
102 Marshall, P. W., Design of Internally Stiffened Tubular Joints, Safety Criteria in Design of Tubular
Structures, proceedings of the IIW/AIJ Conference, Tokyo, July 1986.
103 Kurobane, Y., Recent Developments in the Fatigue Design Rules in Japan, in Reference 94.
104 Nippon Kaiji Kyokai (NK), Rules for the Survey and Construction of Steel Ships, Part P, Mobile
Drilling Units, February 1988 (in Japanese).
105 Iida, K., State of the Art in Japan, in Reference 93.
106 van Delft, D. R. V., et al, The Results of the European Fatigue Tests on Welded Tubular Joints
Compared with SCF Formulas and Design Lines, in Reference 93.
107 Kurobane, Y., et al, Some Simple S-N Relationships in Fatigue of Tubular K-Joint, Trans, of AIJ,
No. 212, October 1973.
108 de Back, J., et al, Low Cycle Fatigue of Tubular T- and X-Joints, International Symposium on
Tubular Structures, IIW, Lappeenranta, Finland, September 1989.
109 de Back, J., Size Effect and Weld Profile Effect on Fatigue of Tubular Joints, in Safety Criteria in
Design of Tubular Structures, IIW/AIJ, Tokyo, July 1986.
 185

110 van Wingerde, A. M., et al, The Fatigue Behavior of T- and X-Joints Made of Square Hollow
Sections, International Symposium on Tubular Structures, Lappeenranta, September 1989.
Chapter 5

TUBULAR JOINTS INVOLVING NON-CIRCULAR SECTIONS

In this chapter, we shall examine tubular connections involving non-circular sections,

principally square and rectangular hollow sections (box sections). Both the existing AWS Code
rules, and proposed revisions based on IIW work will be examined. We will then consider
hybrid connections ~ circular/box and tubular/non-tubular ~ as well as other special topics.
Direct connections between rectangular tubes have not been investigated in the USA as
extensively as circular tubes. However, the behavior of the flat wall of the rectangular main
member is more amenable to analysis, especially in the inelastic range. Also, the sizes of
manufactured rectangular tubes are limited, so that design-relevant parameters fall in a well-
defined range. Box connections can be categorized as "matched" when the members have the
same nominal widths or as "stepped" when the branch has a smaller width than the main
member. It should be emphasized that available research concerns manufactured tubes with
rounded corners. Although the principles should apply to fabricated box sections, judgement
should be exercised in extrapolating empirical criteria and standardized details to this type of
member.

5.1 BOX CONNECTIONS - EXISTING AWS RULES

5.1.1 Evolution of the Code

Mass-produced square and rectangular structural hollow sections first became available
in the USA about 30 years ago. Standard sizes are generally manufactured by hot or cold
forming circular sections into the desired shapes. Special sizes can be made by press-forming
plates into channels which are then welded together, or by welding flat plates together at the
comers.
Such members enjoyed immediate and widespread success as columns in architectural
applications. They were also used 25 years ago in the deck sections of several world-record
offshore platforms (ref. 1), as secondary lateral bracing and grating supports. Connections were
designed by the lower bound cut-and-try method as described in Section 2.3.2(i), with load
transfer by membrane stress and shear in welds, but no reliance on punching shear.
One of the first offshore platforms built for the British sector of the northern North Sea
(Auk "A", ref. 2) used a welded box tube truss for the main superstructure, or module support
frame. Here, the upper-bound yield-line method was used to design the connections, as
described in Section 2.3.2(H). In those days, it was customary for engineers to critically review
each other's work, rather than referring everything to the computer, supervisor, certifying
authority, or verification agent. In the spirit of "fearless pursuit of truth" (ref. 3), there was often
a friendly contest to see who could find the most conservative solution which would stand up to a
vigorous counter attack by the about-to-be vanquished peer. The prevailing solution (which
came after much discussion of strain hardening, load redistribution, and other sources of reserve
strength) was fore-runner to the AWS punching shear criteria for full-width box sections, as will
be described shortly. As a result of the reduced capacity indicated by the punching shear
solutions, most of the major nodes in Auk ended up being designed as stiffened connections.
The original tubular provisions in the AWS Code (D 1.1-72) only covered circular
sections, as the subcommittee membership was then focused mainly on offshore structures.
However, the ASCE Committee on Tubular Structures harbored a lively interest in box sections,
and some of the committee members were doing the basic research in this area. Much of the
early literature which formed the basis for American box connection design codes was
 187

exchanged and discussed at meetings of this committee (refs. 4 to 12). Many of these papers
include valuable experimental data on elastic SCF (ref. 5), and on the ultimate strength of simple
axially-loaded truss-type connections (refs. 5, 8, 9, 11) and moment-loaded beam-column
connections (refs. 4, 6, 7). Particular note should also be made of Redwood et al, who introduced
the use of yield line analysis for stepped box connections, which provided an independent check
for the abovementioned work on Auk.
The reader may find himself frustrated by the limited availability of some of the
references cited. Nevertheless, these are the ones which historically influenced development of
the AWS code, and the author cannot rewrite that history, however parochial it may seem from
today's international viewpoint.
This activity provided the basis for AWS Code provisions for static strength design of
axially-loaded box connections, which were introduced in the 1975 edition. Except for adding a
Commentary (in 1977) and elaborating on the intent of the check for general collapse (in 1985),
these have remained virtually unchanged until now. A proposed Code revision (for the 1992
edition) is discussed in Section 5.2 herein.

5.1.2. Failure Modes Considered

Most of the failure mode considerations for circular connections apply in parallel to box
connections. These failure modes were discussed earlier in Section 2.2 herein. The AWS Code
coverage is in terms of the following, all in working stress format:

local failure
general collapse
uneven distribution of load
materials considerations

Local failure involves plastic failure of the chord face, with the Code provisions for
punching shear being derived from analysis of yield line plastic mechanisms, subject to material
shear strength limits and special provisions for matched (equal width) connections, as discussed
below.

(i) Yield Line Mechanisms. The upper bound ultimate strength methodology of yield
line mechanisms, applicable to stepped box connections, was described earlier in connection with
Figure 2.27. Marshall's derivation of the strength for square T-connections was displayed in Fig.
2.28, with the results given in in Figure 5.1 (ref. 12). This derivation assumed a trapezoidal yield

Fig. 5.1. Ultimate strength

analysis of square tube T-
connections.
- RATIO
188

line pattern with plain 45-degree corners, and results in the simplified expression shown in the
figure. Following the developing practice for circular tubes, the strength expression was given in
terms of punching shear, with the basic gamma term being modified by a geometry term Q^, as
shown in the figure.
The solution of Jubb & Redwood (ref. 5) is more rigorous in seeking the minimum
capacity, as one should properly do with the upper bound procedure, finding alternate corner
angles (arctan(l-/3)) and fanned corners (for < 0.635) which result in slightly less capacity as
shown in the figure. However, there are now two expresssions required to describe the ultimate
strength, each more complex than Marshall's (see Table 2.4).
Both solutions give unrealistic ally high capacity as approaches unity; here the material
shear strength of the chord wall in punching would limit the strength, as shown by the dashed
line. Also, for very small , the theoretical solutions converge on a non-zero capacity for zero
shear area, and the material limit again comes into play.
Both theoretical solutions are also compared with early tests (Jubb & Redwood as
reported by Graff, ref. 9) in Fig. 5.1. The tests generally show considerable reserve strength over
the theoretical solution, as will be discussed in the next subsection. However, in this limited data
set, a more uniform safety margin appears to result for small < 0.5 if the term reverts to
unity, again consistent with the practice for circular tubes.
A somewhat larger home-grown American data set (refs. 11 and 13 as reviewed in ref.
14) is given in Table 5.1. Both hot-formed and cold-formed tubes are included. Specimen
design and mode of loading were chosen to study plastic failure of the loaded chord face, rather
than general collapse (e.g. sidewall buckling). In either case, a sharply defined mechanism or

TABLE 5.1 COMPARISON OF AWS ALLOWABLE LOADS WITH TEST RESULTS FOR
RECTANGULAR TUBES
AWS ALLOW- PEAK LOAD %
ABLE LOAD LOAD SAFETY AT .02D OF
d/D / (KIPS) (KIPS) FACTOR DEFLC. AWS
,l
6 x2"x3/16" .083 8 7.0 15.0 2.13 6.7 95
.167 4 7.7 17.0 2.20 6.8 88
.333 2 9.1 24.0 2.63 9.5 104
.500 1.33 10.5 27.3 2.59 13.1 124
.667 1.0 13.5 37.4 2.78 24.0 178

6"x4"x3/16"
1.0
.083
1.0
1.0
+*
2.11
101
11.5
-
5.45
96
4.0 190
_

.167 1.0 3.5 14.7 4.18 6.0 171

.333 1.0 6.3 18.6 2.94 9.2 145
.500
.667
1.0
1.0
9.1
13.5
-
27.9
-
2.07
13.5
25.5
148
189
*Shear area based on l / 4 " w e l d around
p e r i p h e r y o f t h e branch and minimum
6"x6"x3/16"
1.0
.667
1.0
.125
**
7.9
-
21.8
-
2.75
_
11.2 141
_
specified yield f o r the material
.667 .25 8.7 22.1 2.54 13.4 154 grade.
.667 .50 10.3 23.8 2.31 15.7 153
.667 .75 11.9 26.2 2.21 18.0 152
.667 1.0 13.5 26.2 1.95 19.2 143 **Matched connection,punching shear
1.0
.17
1.0
1.0
**
4.0
97
14
-
3.46
97
3.6 89
- does n o t govern.

.25 1.0 5.8 14.8 2.53 4.7 80

.25 1.0 5.8 19.0 3.25 7.0 120
.33 1.0 7.6 18 2.36 6.6 86
.42 1.0 9.4 29 3.07 9.7 103
.42 1.0 9.4 24 2.54 10.2 108
.50 1.0 11.2 30 2.67 15.8 141
.58 1.0 13.4 34 2.54 22.0 165
.67 1.0 16.7 40 2.40 21.8 131
.67 1.0 16.7 36 2.16 25.2 151
.83 1.0 32.6 74 2.26 66 202
.17 2.0 5.8 16 2.74 6.2 106
.33 0.5 5.8 20 3.42 7.7 132
.33 2.0 11.2 19 1.69 11.0 98
.67 0.5 12.6 40 3.16 21.6 171
.50 1.33 13.0 29.8 2.29 20.6 158
.67 .75 14.7 40 2.73 25.6 174
ultimate load was not observed. As shown in Fig. 5.2, the best agreement with yield-line theory
is generally the load determined by the intersection of the linear elastic and post-plastic portions
of the load-deflection curves. is a reserve strength factor to account for strain hardening, large
deflection effects, etc. Additional points from this data set (for square branch members) are
compared with both the Jubb-Redwood solutions and the Marshall/AWS solution in Fig. 5.3.
 189

C O L D - F O R M ED ( R E F. 13)
MAIN 12 x 2 x 3 / 1 6"
BRANCH 8 x 8 x 1 / 2 "'
HOT - FORMED ( R E F. 11)
" MAIN 6 x 3 x 3 / 1 6"
BRANCH 4 x 4 x 1 / 4"

Fig. 5.2. Typical load-

AWS WITH K= 1.5 AND SF = 1.8
d i s p l a c e m e n t c u r v e s for
rectangular tubes with axial
AWS WITH 1/3 DECREASE ( 1 0 . 5 . 1 .7 )
force in the branch.

V E R T I C AL DISPLACEMENT OF BRANCH ( i n .

1 4.72
A I 4.15
* 3.45

ESTIMATED TENSION
. F I E LD EFFECT FOR

J/ FDA CE/ T -DEFL.

20
-.02D

Fig. 5.3. Width variation in

punching shear equations for
m e c h a n i s m in the face of
rectangular main tube with
square branch (Sherman, ref.
14).

For non-square footprints where the branch intersects the chord in or Y connections,
the parameter (footprint length/ chord diameter) becomes significant, as well as |3. If the
importance of corner zones is discounted (e.g. for large and 45-degree trapezoidal pattern), it
can be argued that the length of yield line, and hence the total internal work and capacity of the
connection, is proportional to branch member perimeter, justifying the use of Marshall's
190

punching shear expression, which only considers . However fallacious, this is how the AWS
Code developed. The solution of Redwood & Jubb rigorously considers both parameters and
as shown earlier in Fig. 2.31 (ref. 15). When is much greater than (i.e. the footprint is
elongated along the chord axis), they theoretically show substantially lower strength than
Marshall/AWS. Nevertheless, the test data of Table 5.1, which covers the a range of 0.125 to 8.0
for the branch tube aspect ratio, /, shows no such unconservatism; indeed, the AWS safety
factor seems to be remarkably consistent across this broad range.
At the time of development of the original AWS box section criteria (ref. 16), there was
little information available to the committee to distinguish behavior of K-connections from T-
and Y-connections. Some evidence from yield line theory indicated increased capacity for K-
connections with small gap, but the reliability of this effect was questionable (see the earlier
discussion of Figs. 2.32 and 2.33). An early empirical expression for K-connections with box
tube chords (refs. 8 and 17) is shown in Fig. 5.4. Unfortunately, this is based in part on small
tests of tubes whose thickness is defined by SWG (sheetmetal and wire gauge) numbers, and is
not dimensionally consistent as would be required for extrapolation to larger tubes (the vertical
load scale has the dimensions of inches). Fortunately, the results could be bracketed by yield line
solutions for T- and Y- connections of the sizes represented. Thus, no distinction for connection
type or load pattern was made for box connections in the Code when they were formulated in
1973, and they did not share in the intensive review that led to updated provisions for circular
tubes in 1984.

12 r INCHES
d ^ 4 "

F i g . 5 . 4 . Lower bound of
u l t i m a t e b r a n c h f o r c e in
rectangular K-connections with
gap.

Similarly, in cross connections, local failure of the chord faces on opposite sides of the
main member was viewed as independent, with any reduction in capacity due to chord sidewall
buckling to be treated as a general collapse problem.
In the AWS Code, the same "universal" allowable punching shear applies to bending, as
well as axial load. Theoretically, for in-plane bending, the Jubb-Redwood yield line solution
indicates increased ultimate capacity in terms of punching shear 70% more as shown
previously in Table 2.4. Figure 5.5 shows plots of moment versus joint rotation for several tests
(ref. 13), as well as the AWS allowable moment (using specified minimum yield strength) and
 191

Fig. 5 . 5 . Moment-rotation
relations for connections with
bending in the rectangular
branch.

the moment predicted by yield line theory (using actual yield strength). It can be seen that the
transition from linear elastic to post-plastic behavior is even less evident than for axial loads.
Although the connections attained their theoretical ultimate moments, and more, the associated
distortions are too large to be serviceable (except possibly as energy-absorbing mechanisms).
In analyzing the rotational stiffness of these stepped box tube connections relative to
branch members of reasonable proportions, it can be shown that the connections act essentially as
if they were pinned (ref. 13). For example, the end rotation of a 20-ft (6m) long, 8-inch (20cm)
deep member bent in double curvature to 30-ksi (210-MPa) is 0.01 radians, whereas the
corresponding connection in Fig. 5.5 would develop less than 10% of this bending at the same
rotation. In trusses, this means that secondary bending moments in branch members generally
can (and should) be neglected.
Similarly, in welded box tube building frames, beams should generally be designed as
simply supported for gravity loads. For lateral loads, the AWS punching shear capacity of
welded beam-to-column connections can be applied entirely to portal moments, as there is plenty
of ductility to accommodate the additional rotation due to gravity load. This approach is referred
to in the AISC Code (ref. 18) as Type 2 construction. Serviceability of the story drifts deserves
special consideration. The following limits may be cited (ref. 19):

cracking of plaster and stucco 0.2%

window glass in flexible gaskets 0.6%
typical 100-yr extreme dynamic response - 1.0%

For the connections shown in Fig. 5.5, taking the one-third increase over the AWS
allowable stresses (as is customary for wind loads) would result in about 2% story drift just in the
connections, in addition to that which would be calculated on the basis of member flexibility. A
remedy for such excessive drift is discussed in the next subsection.
192

Professor Sherman (University of Wisconsin - Milwaukee) gives additional guidance in

the design of beam connections to rectangular tubular columns in ref. 20.
For out-of-plane bending, no specific guidance is given, except to note that the yield line
pattern, and thus the expected behavior, is similar to in-plane bending (see Fig. 2.29), at least for
small- stepped connections.
(ii) Reserve S t r e n g t h and Safety F a c t o r . For connections in circular tubes, the
following sources of reserve strength were discussed earlier (Section 3.5):

plastic section
triaxiality
yield bias
strain hardening
load redistribution
large deflection behavior

These sources of reserve strength are implicitly included in the empirical design criteria
for circular tube connections. For box connections, plastic section and load redistribution are
already included in yield line solutions, so they cannot be invoked again. The mode of
deformation at yield lines would also seem to limit the opportunities for triaxiality. Thus, we
look to the remaining items to explain the observation of capacities above the "upper bound"
limit analysis.
Yield bias, actual strengths exceeding the specified minimum, can be considerable,
especially in cold formed box tubes of thin-gauge material. For example, the ASTM A-500 grade
tubes in Fig. 5.5 have 46-ksi (320 MPa) specified yield, versus 57-ksi (400 MPa) actual. The
data presentation shown in Table 5.1 and the related Figures counts yield bias as part of the
Code's reserve strength. This may be reasonable as long as the materials tested are representive
of the application, and we are careful not to invoke yield bias again elsewhere in the Code
calibration. However, this becomes potentially unconservative when the application involves
hot-formed tubes, heavier sections made from welded plates, or simply tubes from manufacturers
who work closer to spec (especially those who abuse the under-tolerance on thickness).
Large deflection behavior includes the membrane effect which comes into play as the
chord face is stretched and deflected out of plane. Unlike circular tubes, where large deflection
improvements to the arch shape only occur for tension branch loads, the membrane effect in box
connections is equally applicable to compression loads, insofar as plastic failure of the loaded
chord face is concerned. Stretching patterns investigated by Mouty (ref. 21) are shown in Fig.
5.6. He found that the transverse tensions of the original pattern could not be supported at the
edges of the chord, and developed the modified pattern shown, based on test observations.
However, he concluded that, if one accepted the common criterion limiting deformation to 1% of
the chord width, the membrane effect was negligible. Allowing larger deformation, Sherman et
al (ref. 14) found the membrane effect to be more significant, especially at large (see Fig. 5.3).
Finally, we come to strain hardening. If one observes the 2/3 rule (effective yield not
more than 2/3 the ultimate tensile strength, footnote 2 to Code Table 10.5.1), then a reserve
strength factor of 1.5 would eventually be available from this source. Whether this can be
developed at serviceable deflections is an open question.
Including the reserve strength factor (K) and safety factor (SF) explicitly in the
punching-shear design format, we get:

F
allowable V = Q ^ (5.1)
p
SF ^ 0.5 7
 193

(a) (b)

Fig. 5.6. Mouty's membrane stretching patterns, (a) Initial, (b) Modified (ref. 21).

where is the expression which results from directly yield line analysis of the connection (e.g.
see Table 2.4 or Fig. 5.1), and the other terms should by now be familiar.
Table 5.2 gives the suggested design factors, and SF. The reserve strength implicit in
circular tube criteria (1.6 to 6.5-fold beyond first yield) is shown explicitly for box sections; here,
we take a conservative value of 1.5. The AWS safety factor of 1.8 is consistent with that used in

TABLE 5.2 SUGGESTED DESIGN FACTORS

ASSUMED SF FOR SF WHERE
VALUE FOR STATIC 1/3 INCREASE
APPLICATION LOADS APPLIES

Where the ultimate breaking

through of the connection
including effects of strain
hardening, etc. -- can be
utilized

Redundant fail-safe structures

and designs consistent with
Section 10.5.1 1.5* 1.8 1.4

Critical members whose

sole failure would be
catastrophic 1.5* 2.7 2.0

Architectural applications where

localized deformation would be
objectionable 1.0 1.7 1.3

*Applicable where main member F is not taken to exceed 2/3 the specified
minimum tensile strength. ^

ultimate strength design of steel members for buildings. This and SF gives the basic AWS
Code formula, in which 0.5 in the above equation becomes 0.6. Using Marshall's yield line
solution, the 1975-1990 AWS design equation for box connections then becomes:
194
F
allowable V = Q Q ( 5 . 2 )
0.6 7 f

where Qa = ^25 r
f o o v r e 0 ^

(W)

and = 1.0 for smaller .

The foregoing equation is subject to further limitations for larger than 0.8, as described in the
next subsection (iii).
In tubular structures, the main member (chord) at tubular connections must do double
duty, carrying loads of its own (axial stress fa and bending f^) in addition to the localized
loadings (punching shear) imposed by the branch members. Interaction between these two
causes a reduction in the punching shear capacity, as reflected in the Qf de-rating factor. The
solid line in Fig. 5.7 shows the de-rating factor for box sections. This was originally extended
from circular to box section connections in the 1975 Code, in consideration of the similarity of
local plastic behavior in the two types of connections, and is now retained for lack of something
better. Dashed lines in the figure show how the new circular connection criteria (see section
3.7.1(ii) herein) bracket the old Qf, for the range of gamma values applicable to manufactured
box sections.

I I
F i 5 7
.25 .50 .75 i.o g - De-rating factor for
box section joints (solid line)
MAIN MEMBER UTILIZATION versus criteria for circular
sections.

The histograms in Figure 5.8 indicate the reliability of the AWS criteria for box
connections, as seen about the time the criteria were formulated (refs. 15 and 16). Up to 1975,
AWS Code development relied primarily upon North American research efforts, with only
secondary access to CIDECT work (e.g, ref. 17). Although some of the plotting positions appear
to have been shifted slightly, this represents the substantially same data set as Table 5.1 (ref. 14),
axially loaded T-connections. The criteria appear to provide ample margin against total failure of
the connections tested, although they do result in using design loads close to the 2% deflection
limit, with localized yielding at smaller loads to be expected.
 (a) ULTIMATE STRENGTH

Nominal Ultimate
W i t h S F = 1.8

Median SF=2.77
COV= 2 3 %
0 S= 4 . 4

1.0 2
1.0 2 5
p P @ . 0 2 D D e f l e c t i o n / P a | | ow
ult / Fallow
Fig. 5.8. Reliability of existing criteria for RHS. (a) Ultimate strength, (b) Deflections.

This approach, allowing substantial amounts of localized plasticity in tubular

connections at design loads, has proven satisfactory for redundant fail-safe structures consisting
of open tubular space frames in which resulting deflection could be tolerated. However, for
architectural applications where localized joint deformation would be objectionable, or for
critical connections whose sole failure would be catastrophic (e.g. at the fixed end of a cantilever
beam), a more conservative approach may be justified. Table 5.2, and the Code Commentary,
suggest using a factor of 1.0 (foregoing dependence on strain hardening, large deflection
behavior, etc.), or increasing the safety factor to 2.7 (consistent with that used against the
breaking strength of fillet welds), for these two situations, respectively. Either case would result
in a one-third decrease in allowable punching shear capacity, as provided in AWS clause
10.5.1.7.
(iii) Large and Matched Connection Limits. For stepped box connections with
larger than about 0.8, or larger than (i.e. a rectangular footprint with long axis running across
the chord), yield line solutions give very high punching shear capacities, which may become
unrealistic when failure modes other than plate bending occur. These alternative modes of
behavior are shown in Fig. 5.9; they were also discussed earlier in connection with Fig. 2.6. To
deal with them, additional checks for static load capacity were formulated (ref. 16), following the
"building block" approach described earlier in Section 2.3.2 (i). The corresponding AWS Code
requirements are summarized as follows:

I
Q
\
I
Til-

(d)
J4L --
(a) (b) (c)
Fig. 5.9. Modes of chord failure in the connection side region, (a) Yield line mechanism, plate
bending, for > 0.8. (b) Punching shear at the material limit, large . (c) Transition from
(b) to (d), shear and direct stress, (d) Local crippling in the side wall, direct stress at yield.
196

The allowable total load capacity normal to the main member, sin(0), shall not
exceed the sum of building blocks F ^ and F 2 , where:

(1) Along the sides of the connection, punching shear is limited to the material shear
strength limit (or corresponding working stress, 0.4 F y o ) , so that allowable

F = 2 a x (0.4 F y Q) (5.3)

where is chord thickness, and a x is branch footprint length along each side of the
connection. For r less than 0.67 F y ^ y , yielding in sides of the branch member would
take precedence, resulting in lower capacity. The Qf reduction factor is not applied
here, because the area of chord face involved in this mode of failure is very small, and
with limited deformation could easily shed self-load to other parts of the cross section.

(2) Along the heel and toe of the connection, F 2 is computed (in punching shear format)
using a special yield line solution for a pair of transverse line loads running across the
chord face, to be described shortly.

Five of the early American tests of Table 5.1 address the foregoing range. They indicate
a mean safety factor of 2.29 (vs. nominal 1.8), and a very low coefficient of variation, such that
the indicated safety index is better than for the data set as a whole.
For matched connections, the check for local failure, at or near the loaded chord face,
involves similar consideration of building blocks F^ and F 2 , where:

(1) Along the sides of the connection, as shown in Fig. 5.9(d), web crippling capacity of the
main member sidewall in direct yield (corresponding to 0.6 . F y Q allowable) gives:

F 2 T a 0 6 F ) Q 4 ) ( 5
l = x < yo f

Where r is less than F /Fy, yielding in sides of the branch member would take
precedence, resulting in lower capacity.

(2) Along the heel and toe of the connection, F 2 is computed using the same special yield
line solution for a pair of transverse line loads running across the chord face.

Attaining the deflections required to fully develop the F 2 line load mechanism requires
severe local plastic deformation in the Fj region of yielding. For compressive branch loads, a
general collapse mechanism, involving sidewall buckling, usually takes precedence over the
above-described local failure modes. This is discussed further in the next subsection (iv). For
tensile loads, tremendous demands are placed on the capacity of the sidewall and corners of the
intersection weld, and on the ductility of the materials used in the connection, as discussed in
subsections (v) and (vi).
The special yield line solutions which were historically used to compute F 2 , transverse
line loads at the heel and toe of the connection, will now be described. Their derivation is
detailed in Figs. 5.10 to 5.13.
 197

EXTERNAL WORK:

INTERNAL WORK:
Wj = 4 [ + 2 ]
2
Mp = Fyo/4
MINIMUM (-^ = ) AT = 55

SOLVING WE = W! YIELDS
2
ULT Q = 5.6 Fyo T / D

PUNCHING SHEAR: Q = 2T V D
2 5 I 0 E S
< , = D/2T

Fig. 5.10. Yield line derivation

for simply-supported chord face.

gnnnH-4 EXTERNAL WORK:

Hi-

INTERNAL WORK:
Wj = 4 [2 tarrt + 2 cot**]
2
Mp = T Fyo/4
MINIMUM AT = 45

SOLVING WE = Wj YIELDS
2
ULT Q = 8 Fyo T / D

PUNCHING SHEAR: Q
(2 SIDES)
D/2T
:
CT57

Fig. 5.11. Yield line derivation

for fixed-edge chord face.

SminS- EXTERNAL WORK:

WE DQA

INTERNAL WORK:
OUTER ZONES Wj- = 16 FOR = 45
INNER ZONES Wj = 4 n D Mp 2^

SOLVING Wr "= " "HIT YIELDS

2
ULT Q = 4 (1 + J) Fyo T / D

PUNCHING SHEAR: q = TV n p
(OUTSIDE ONLY CONVENTION)
ULT V p = ( 1 + \ ) j f a FOR < 2

ULT V = 2 JX- TWO SEPARATE PATTERNS

Fig. 5.12. Yield line derivation

for pair of transverse line loads.
198

Fig. 5.13. Effect of branch member strain

hardening and redistribution
of applied line load on reserve
strength. allow Vp = u l t Vp

SF = 1 .

Kc = 1 . 5

Q = 1.5Q
V
Q = Q Q = 2Q Q = 0.5Q

W
WORK = 0 . 7 5 T O 0 . 8 3 W
WORK = 43- W,E
KR = 1 . 5 USE K R = 1 . 2 5
R

UNIFORM BRANCH AT BRANCH AT

ASSUMPTION WORKING STRESS FULL Y I E L D

The first figure derives the theoretical upper bound limit load for a transverse line load
running across a simply supported chord face. This would be appropriate where the chord face is
substantially thicker than the sidewalls, as in fabricated box members, or where an insert plate
has been used in the connection area, analogous to a joint can in circular tubes. Instead of
assuming a rigid indenter, as is often done, we here take a uniform line load, Q, for the purpose
of computing external work. After a lot of geometry to determine hinge angles, the expression
for work done by internal yielding along yield lines is derived. The pattern (phi angle) giving the
minimum capacity is found in the usual way. Equating the two energies yields the solution, with
the total load capacity being Q times D. An equivalent punching shear expression is also
derived.
The second figure extends this derivation to the case where the full yield line plastic
moment can be developed at the lines of support. This would be the case for box tubes of
uniform wall thickness, or where the face plate is continuous over several stiffeners. The total
capacity is 40% greater than for the simply supported case, and twice that of a point load. In
punching shear format, the result is Marshall's "universal" basic punching shear expression,
applied on both sides of the loaded line.
The third figure treats the case of two transverse line loads. This corrects an error in the
original derivation (ref. 22), factors two on 17 and 25% on total capacity. Note that punching
shear here follows the "outside only" convention as for tubular connections.
The fourth figure deals with reserve strength factors. Kg of 1.5 represents strain
hardening, etc. in the chord face. Along with the safety factor, this is included in the AWS Code
allowable basic punching shear. Kj^ represents the effect of branch member strain hardening and
redistribution of the applied line load, relative to a value of 1.0 for the originally assumed
uniform distribution. When the branch member is at a nominal working stress of say 75% of
yield, and locally strain hardens to 1.5 times yield, the peak line load is twice the average Q-bar;
for the same total load and deflection, external work is two-thirds the reference case, internal
work is the same, so the reserve capacity is an additional factor of 1.5. If, in the spirit of
compatible strength, we assume the branch member to be at full gross section yield, the
opportunities for load redistribution are less; taking external work as 0.8 times the reference case,
a of 1.25 results. This appears explicitly in the special line load solution for F 2 , given in
Code clause 10.5.1.1(2).
 199

The last-mentioned pattern of line load is reminiscent of effective width behavior in the
branch member, similar to AISC design provisions for non-compact members. Indeed, we can
derive an equivalent effective width, for connection efficiency, from the following limit
state:

sin(0) = 1.8 (5.5)

(allow vp)

Using the 2 line load solution for allowable punching shear, we obtain

leff (3.75 Qg) __

= < 1.0 (5.6)
b sin(0)

where Q^ = 1 + 7 7 / 2 for < 2, and 2.0 for larger 77. Thus the hypothetical range of the numerical
constant (3.75 Q ) on the right hand side is 3.75 to 7.5, as compared to the empirical
IIW/CIDECT value of 5.0 (refs. 23 and 24).
(iv) General Collapse. This problem was discussed in general terms earlier in Section
2.2.2. Please refer to Fig. 2.7 for examples of various general collapse failure modes. Those
applicable to box connections include beam shear, sidewall buckling (web crippling), and
longitudinal distress. The AWS Structural Welding Code Code broadly requires that the strength
and stability of the main member in a tubular connection, together with any reinforcement, be
investigated in accordance with the applicable design code, e.g. AISC (ref. 25).
Beam Shear Beam shear in K-connections is a problem which may escape detection
by the usual code checks on the members involved; it requires that the designer give special
attention to load paths within the connection. However, once the proper cutting planes are
investigated, beam shear in the sidewalls (webs) of the chord may be treated according to AISC
rules for box girders. For typical compact sections, the allowable shear stress is 0.4 Fy Q. Like
punching shear, beam shear suffers from interaction with chord self stresses, as shown in Fig.
5.14, reflecting AWS Qf and AISC rules, respectively.

0 z z z z u z z <

UNACCEPTABLE
CHORD STRESSES

I-

Q
or
VALUE O F Qf F O R
X
BOX S E C T I O N S

Fig. 5.14. Interaction diagram

for b e a m s h e a r ( s h a d e d
boundary) and punching shear
(Qf lines) in chord.
CHORD SHEAR UTILIZATION U s
200

Sidewall B u c k l i n g -- This failure mode is analogous to ovalizing in circular

connections, and is particularly severe in cross connections, and in unstiffened connections
subject to crushing loads (e.g. at the end post or bearings of a truss). Inadequacy of the Code's
broad "motherhood statement" in this area was pointed out by Davies et al (ref. 26) in 1984, and
the next edition (1986) gave more specific guidelines, including the traditional 5T limit on load
spreading. For unreinforced matched box connections, the allowable limit on total compressive
branch load normal to the chord is:

sin(0) = 2 FeQ (a x + 5T) Qf (5.7)

where SL^ is the footprint length, and F e o is the web crippling or buckling working stress given by
the applicable design code, e.g. AISC eqn. 10.1-11:

FeQ = 0.67 (^) < 0.75 Fy (5.8)

where is elastic modulus and is web depth. This is the lower of two possible AISC formulas
from which one could choose, and reflects the expectation that the boundaries of the web will
rotate due to deformation of the loaded chord faces. Packer (ref. 27) also reports success in
correlating matched box connection test results to the AISC-LRFD limit state provisions. For
stiffened connections, the AISC formulas provide for higher values of F e Q depending on aspect
ratio of the stiffened web panel.
Longitudinal Distress No detailed guidance is given in the Code, but see Figure
2.7(e) herein. For simple (unstiffened) connections involving compact sections, this is generally
not expected to be a problem, as other failure modes will give lower connection strength.
(v) Uneven Distribution of Load. This problem area, and limitations on the use of
nominal section properties to compute stresses at welds in tubular connections have been
previously discussed in Sections 2.2.3 and 2.5.2. Experimental examples of uneven load
distribution (refs. 28 and 29) are shown in Fig. 5.15. In these matched box connections, the
sidewall takes the lion's share of the load; in addition, there is further concentration of load
transfer at the corners of the footprint. For stepped box connections, Redwood's early finite
difference analysis (ref. 4) also shows stresses peaking at the corners of the footprint. AWS
Code provisions for uneven distribution of load ~ that the weld should not be a weak link are
the same for both circular tube and box connections. In addition, provisions for making the
welded joint (continuous welding around the corners) stress the importance of these regions for
box connections.

Fig. 5.15. Elastic strain distribution at joint in box connections, (a) Axial force in branch, (b)
Bending in branch (ref. 29).
 201

(vi) Material Considerations. The generic issues of notch toughness, lamellar tearing,
and weldability were discussed previously in section 2.2.4, and will be detailed further in
Chapter 7. They all apply equally to box connections as to circular tubes.
One issue peculiar to box connections is the Code warning on the use of ASTM A-500
cold-formed tubing (footnote 38 to clause 10.2.1.9). Special investigation or heat treatment may
be required when this product is applied to box connections. The problem lies in the potential
for strain-aging and loss of ductility when the cold-formed corner regions are subjected to the
heat of welding. In , Y, and connections, the comer regions are typically the most highly
stressed, and plastic straining is expected at design loads, so this loss of ductility can become
significant. Heat treating, e.g. normalizing the chord section before incorporation into the
structure, generally wipes out the effects of cold forming, and improves notch toughness and
ductility; however, some loss of yield strength may occur. At least one structural failure has
been been attributed to strain-aging in cold-formed box connections, that of a long span roof
truss on a large aircraft hangar (ref. 30). On the other hand, hundreds of smaller structures have
been successfully built, using A-500 without any special precautions. Thus, one should not over-
react; the Code says "may be".

5.1.3. Simplified Design Charts.

Here we present a design chart for rectangular hollow structural section truss joints
based on the AWS Code, for use by American designers, following the logical step-wise design
format of Packer (ref. 31) and Wardenier (ref. 32), as previously described in Section 3.8 herein.
As before,the charts give the maximum punching shear efficiency, E y , in terms of the non-
dimensional parameters , 0, and 7 , where...

E = maximum a l l o w a b l e p u n c h i n g s h e a r s t r e s s
v
m a i n member a l l o w a b l e t e n s i l e s t r e s s (5.9)

Where self-loads in the main member are present, this maximum allowable punching
shear must be de-rated by the Qf factor (solid line in Fig. 5.7). Overall connection efficiency, Ej,
is simply:

F C H R )D
. = Of . Y Q < ( 5 . 1 0 )

3 (t/T) sin(0) fy (branch)

where F v o is the specified minimum yield strength of chord (main member), Fy that of the
branch member, and (T/t) is the branch/main thickness ratio ().
For box sections, punching shear efficiency, E y , is shown as a function of and 7 in
Fig. 5.16. Bear in mind that the overriding influence of r, branch/main thickness ratio, is already
included in the definition of punching shear. The AWS box section rules do not distinguish
between joint types and loading patterns, but tension and compression are treated differently.
As with circular sections, the strength of box section joints decreases as the main
member width/thickness ratio increases. For very stocky members, D/T less than 8, punching
shear equals the shear strength of the material, and the efficiency reaches a plateau of 0.67. Of
the standard AISC sizes, only the 4x4x0.50 and 2x2x0.25 achieve this limit (also the heaviest
8x4, 6x4, 4x2, and 3x2 when branch member attachments are limited to the narrow face).
Connection strength increases with , the branch/main width ratio. Fig. 5.17 shows the
yield line pattern associated with the behavior of joints in the mid range of . For stepped
connections with larger than about 0.8, the alternative limit states shown in Fig. 5.18 apply: (a)
localized failure limiting the achievable line load along the sides, (Fig. 5.9), and (b) a different
202
TENSION or
COMPRESSION TENSION COMPRESSION

Fig. 5.16. Punching shear efficiency chart for box section joints, all failure modes considered.

Fig. 5.17. Box connection

limit state for small .

hypothetical yield line pattern supporting line loads along the heel and toe (Fig 5.12). The first
localized failure is punching at the material yield strength, which requires r of at least 0.67 to
achieve. As we approach a matched (equal width) connection in tension, the localized failure
becomes side wall yielding of the main member, requiring r of 1.0 to achieve. References to the
governing AWS paragraphs are given in Fig. 5.16.
For compression loads, a general collapse mechanism involving web crippling of the
main member side walls applies, resulting in substantial reductions in capacity at large
height/thickness ratios. The capacities shown in the right-hand side of Fig. 5.16 are based on
AISC as well as AWS rules (as discussed in the previous section). Bear in mind that, for load-
balanced K-connections, one of the branch members will always be in compression, thus
governing the design.
 203

Fig. 5.18. Limit states for large > 0.8.

The limiting r ratios given in the figure are the minimum required for branch member
yield to match web crippling of the adjoining main member. For example, for a 90-degree
square-on-square joint with y of 22, of unity is required to achieve a joint efficiency of 62%.
This corresponds to the side faces of the branch being fully utilized, while only 24% of the
strength of the heel and toe faces are effectively utilized. The Code does not indicate what to do
for thinner branch members. Taking a proportionate reduction in capacity (i.e. keeping the same
effective width) would be on the safe side. Some of the international standards (e.g. refs. 23 and
24) treat this problem in terms of more explicit effective width criteria, also distinguishing
between K-connections and T&Y connections.
The bending capacity of rectangular section joints is also not explicitly treated by the
AWS Code. For stepped connections, yield line analysis (permitted under paragraph 10.5.1.4,
and described in the Commentary) indicates the ultimate punching shear efficiency for in-plane
bending to be at least equal to that for axial loads. However, unacceptably large joint rotations
may occur before this capacity is eventually reached. Designers wishing to utilize simple direct-
welded rectangular section joints in moment-resisting frames and Vierendeel trusses should
explicitly consider their moment-rotation characteristics. Some data on this can be found in the
Stelco Manual (ref. 34) along with various schemes of stiffening. Additional practical
suggestions for stiffened beam-to-column moment connections are given in Reference 35.
Somewhat stiffer and more effective moment-resisting joints result with matched and
nearly-matched connections ( larger than 0.85 at the toe of the weld). Taking "heel and toe"
limits for in-plane bending, and side-wall crippling (compression) limits for out-of-plane
bending, results in surprisingly similar capacities. Nearly 100% joint efficiency is achieved for
main member width/ thickness of 16 or less, with r not exceeding 0.5, even for worst-case Qf. A
large number of the available sections listed in AISC could be used with this criterion.
Limits of applicability for Fig. 5.16 include the full plotted range of the parameters , 0,
and 7 , and the following:

- tau limits as given in the figure (see effective width discussion)

- gap joints (see Code Section 10.5.1.5 for overlapping joints)

204

- uniform thickness square sections for branch members (this is not

a limitation of the Code, but was used in order to simplify the
derivation of the figure)

- uniform thickness square or rectangular sections for main member

(branches attached only to narrow face of rectangular, observing
indicated 2 limits for height/thickness as well as width/thickness)

- compact sections, both branch and main members (AISC width/thickness

limits are 40 for A-36, 34 for 50-ksi yield, e.g 12x12x0.375)

- ductile mild steel with tensile/yield ratio of 1.5, or notch-tough

high strength steel with effective yield taken as 2/3 tensile
(see AWS footnote 38 for A-500 and other cold-formed products)

- matching weld metal and prequalified weld details of AWS D l . l

Figures 10.13.IB or 10.13.3B (also fillet welds, Figure 10.13.5,
for the special case of E70XX electrodes joining mild steel)

Several possible failure modes other than punching shear (synonymous with local
failure in AWS terminology) must be considered in a comprehensive design check. In the present
work, most of these are either included in the E y charts, or covered by limits on applicability, as
discussed in the preceding section and summarized below:
General collapse The appropriate limits are included in the charts, for sidewall
buckling of box sections, with members of uniform thickness.
Uneven Distribution of Load Welds which develop the strength of sections joined
are required to prevent "unzipping" or progressive failure of the joint. The AWS prequalified
groove welds meet this requirement.
Local buckling Due to differences in the relative flexibilities of the members at a
tubular connection, the actual distribution of axial stress in the branch member is not like the
uniform nominal stress we calculate. Due to concentrated delivery of branch loads, the actual
distribution of axial stress in the main member is also not uniform. In addition, there are high
localized shell bending stresses which accompany punching shear. Localized yielding and re-
distribution of load may be required to develop the full capacity. Compact sections provide for
this.
Beam shear in the main member This is a potential failure mode for gap joints
when the product of Ej * r * exceeds 0.13 to 0.33, with square members. The lower number
applies where the chord is fully stressed axially, but this usually does not occur at points of high
shear loading. The AISC interaction between axial and shear stress in a compact section was
shown in Fig. 5.14. Rectangular sections turned sideways (i.e. loaded by branch members
attached to their broad face) and built-up box sections with thin webs are more vulnerable, and
outside the stated limits of applicability of the charts.
Lamellar Tearing ~ See Code Commentary and Chapter 7 herein.
Fatigue Fatigue of box joints is not extensively covered in the AWS Code. Figure
4.34 indicates that, in general, mild steel joints with 100% strength efficiency can be expected to
safely withstand about 3,000 applications of load equal to their allowable static capacity (but
only a few hundred full reversals). Fatigue performance does not increase in proportion to yield
strength, so joints attempting to exploit high strength steel have even shorter lives. More later
(see Section 5.5).
 205

A step-by-step design procedure for simple tubular trusses, using the chart format, was
given earlier in Section 3.8.5, subject to the stated limitations.
For axial loading alone, or bending alone, the connection is satisfactory if member-end
utilization is less than joint efficiency, i.e.

^ < 1.0 (5.11)

where Ay is the branch utilization; e.g., f n/0.6 Fy.

For combinations of axial load and bending, based on linear interaction, the check for
box section connections becomes...

Ay; 1.0
axial bending (5.12)

The AWS provisions for box section joints can be made to yield reasonable results;
although, as presently stated in the Code, they might not lead all designers to exactly the same
results as given in Fig. 5.16. Potential areas needing further work on box connections have been
identified, including a comprehensive examination of today's data base, effective width concepts,
moment-rotation capacities, and load interaction (Qf).
An alternate, even simpler, approach to the design of connections is to set limits on
certain key parameters, to insure that the full strength of the branch member can be developed.
This eliminates the need for more detailed calculations. A design chart for box connections in
this format has been presented by Sherman et al (refs. 14 and 36), as shown in Fig. 5.19. The
chart shows limiting parameters ~ t/, d/D, and D/T for which 100% connection efficiency is
provided with respect to the local failure mode (punching shear).

Fig. 5.19. 100% efficiency lines

for T-connections in rectangular
tubes with square axial branch
local failure mode (ref. 14).
206

5.1.4 Comparison With International Data Base

Internationally, there has been considerably more testing of box connections than that
represented in Table 5.1. Wardenier's book (ref. 33) covers the subject in detail. Thus, it seems
desirable to see how AWS design criteria compare internationally, even if only in hindsight.
D/Tc = 2 7

0
Fig. 5.20. Efficiency of and X joints between square hollow sections with = 9 0 , according to
IIW/CIDECT rules (ref. 32).

Figure 5.20 shows the IIW/CIDECT criteria for , Y, and X connections (ref. 32, based
on refs. 23 and 24). E y as given in the figure is related to connection efficiency (the fraction of
branch yield it can develop) as given previously by Equation 5.10. In the )3 range of 0.2 to 0.85,
the IIW criterion is based upon the trapezoidal yield line solution of Redwood and Jubb (Table
2.4 herein). Comparison of this criterion with test data is shown in Figs. 5.21 and 5.22, for
failure load and 1% deflection, respectively (ref. 37). The agreement shown is remarkably good,
given the ambiguity of yield stress (corner vs. face) and limit load (e.g., Fig. 5.2) which exists for
cold-formed box sections.

CALCULATED (TONNES) CALCULATED

Fig. 5.21. Test results (failure) v. Fig. 5.22. Delft test results for
yield line model (ref. 37). 1 % b Q deflection v. yield
line calculation (ref. 37).
 207

Since the international criteria have been presented in the same punching shear
efficiency format as Fig. 5.16, they can now be compared directly to AWS criteria, even though
one is limit state design and the other is working stress. This is done in Fig. 5.23. The trends of
strength versus and 7 parameters are similar, with AWS providing a more conservative
extrapolation to very small )3, especially in light of the data shown earlier in Figure 5.3.
Elsewhere, however, AWS criteria indicate considerably higher connection efficiencies. This
comes principally from the inclusion of the reserve strength factor of 1.5 and the acceptance of
larger local deformation, as previously discussed. While the IIW criteria, on the other hand,
includes the partial safety factor on strength, for use in limit state design (LRFD) format, this
factor is 1.0 for chord face plastification, in recognition of additional reserve strength. When the
1/3 decrease is applied (AWS clause 10.5.1.7), the comparison is much closer.
D/Tc = 2 7

0.8
AWS AWS

0.7 -
EFFI CIENO

0.6 27=10 2 7 = 10
IIW
27=10 aws

0.5
on AWS AWS
<C
30
UJ
co
0.4 -
C3 IIW
30
iCHU

0.3
ZD
Q.
0.2 30 AWS
IIW
>
LU joint X joint
0.1 IIW compress. compress,
tensile or compression tensile loading loading
loading , or X joint loading
. 1 1 1 1 l .. .
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.85 0.85 1.0

Fig. 5.23. Comparison of AWS and IIW/CIDECT criteria for square hollow sections.

For large and matched connections, the AWS criteria progress through a variety of
failure modes (material shear, web crippling, sidewall buckling, etc.), so that the AWS efficiency
curves include a number of plateaus and transitions. The IIW/CIDECT efficiency curves, as
presented here, simply use the yield line solution up to of 0.85, and a linear interpolation
between that point and the matched connection result; however, the criteria themselves include a
similar variety of failure modes.
For compression loading, sidewall buckling limits the strength of matched box
connections. AWS/AISC working stress allowables for this case are compared directly to the
relevant international test data (ref. 37) in Figure 5.24. There is disturbingly little reserve
strength indicated, especially in the vicinity of the knee, where the yield-based allowable meets
the buckling-based allowable. This is no doubt why some authorities (refs. 28, 29, 33) prefer to
use a column buckling type design curve for the s i d e w a l l s , e.g., E C S C , which more
conservatively rounds off the knee.
In the international criteria, and N-connections are differentiated from , Y, and X-
connections. For and N-connections with equal width square branch connections, the
IIW/CIDECT criteria can be reduced to punching shear efficiency, as previously defined,
permitting comparison with AWS/AISC criteria, as shown in Fig. 5.25; this should be read
together with the Qf reduction factor given in Fig. 5.26. The international criteria incorporate a
partial safety factor of 1.1 on the characteristic strength (ref. 39).
 208

Fig. 5.24. Test results for sidewall bearing and buckling, versus AWS/AISC allowable.
0.8

0.7

\ 1.0
0.5 AISC S
.
Fig. 5.25. Comparison of
\ B l AM SHEA
AWS/AISC versus
IIW/CIDECT efficiency of
and joints between square
hollow sections with gap of 0.5
.5
(l-) to 1.5 (1-/3) times chord
width, and branches having
equal widths.

25

2 7 = D/Tc

Fig. 5.26. Comparison of

reduction factor, IIW-CIDECT
(solid l i n e s ) v e r s u s AWS
(dashed line).
0 0.2 0.4 0.6 0.8 1.0
CHORD UTILIZATION
 209

At first glance, the two sets of criteria seem to be radically different. In K-free punching
shear format, the IIW criterion reduces to a very simple expression for the full range of
parameters shown in the figure:

ult V J0_ (5.13)

0.5
0.97

The reduced exponent of 7 in the IIW criteria reflects the influence of failure modes
other than yield line, e.g., membrane tension, beam shear, web crippling, etc., for which strength
varies linearly with chord thickness, rather than as thickness-squared.
The AWS rules are more complex, requiring explicit consideration of and alternative
failure modes. The "beam shear" limit shown for matched connections includes punching shear
across the loaded face (small gap), as well as membrane shear in the sidewalls.
When Figures 5.25 and 5.26 are combined, and we look at punching shear efficiency for
each set of criteria with their respective Qf reductions included (for fully stressed chord), as well
as the effect of 0, there is closer agreement between AWS/AISC and HW/CIDECT, as shown in
Figure 5.27. For the AWS/AISC beam shear limits, different reductions apply to membrane
shear and punching shear, as per Fig. 5.14. It is tempting to speculate that the empirically
derived IIW criteria may reflect the various modes of behavior which were hypothetically
addressed in the AWS "building block" approach.

0.6

^ IIV / C I D E C T
0.5

0.4

u
S 0.3

j3= l . o
0.2
mm*
= 0.8

- m m ]3= 0.5
0.1

AISC (l SHE \R) AWS *

( P I INCHING shear)

15 20 30 35

2 7 = D/Tc

Fig. 5.27. Comparison of AWS/AISC versus IIW/CIDECT efficiency of and joints between
square hollow sections with fully stressed chord.

5.2 BOX CONNECTIONS - PROPOSED NEW RULES

The complete set of IIW/CIDECT criteria for box connections are more extensive than
the simplified subset used in the comparisons of the foregoing section. For example, they
include additional terms for rectangular branch members and unequal branch widths in
connections.
210

TABLE 5.3 DESIGN STRENGTH AND VALIDITY RANGES OF WELDED JOINTS OF

RECTANGULAR SQUARE OR CIRCULAR BRACINGS AND A RECTANGULAR
CHORD SECTION (REF. 39)

TYPE OF JOINT DESIGN STRENGTH d-1.2)

- , Y- and X - J o 1 n t s
0.85 CHORD FACE YIELDING
2
F .T
VP 0 5
l
( - T V * HI /' ].F(N)
1 (1 - 0)sin sin 9j

4-1.0 CHORD SIDE WALL FAILURE 0.85 < i 1.0

. f..t 2h. linear Interpolation
n* - - H - i-rV +
I O T O] chord face yielding and
chord side wall criteria
1 sin $i 1
s1 $^
> 0.85
EFFECTIVE WIDTH

,/^ N
I "
F
Y R V
2 H
I "
4 T
I
+
2b e]

0.85 <, 1 - 1/7

PUNCHING SHEAR

* f .t 2h

1 1
- and N-GAP j o i n t s CHORD FACE YIELDING

. f .t b.+ b+ h,+ h.

1 0

CHORD SHEAR
f A ,V 2 0 ,5
. ; * S (Av - A J f + A u.f L[1 - (V/V
v )J ]
V yo V yo p'
1 73.sin 0
EFFECTIVE WIDTH
2 h - 4t. + b,+ b J1
V W 1 1 1 e

1 - 1/7 PUNCHING SHEAR

f 2 n
* Y9 i
. - .- . . . [l t " + b. + b J]
i /3.s1n 0.
1 s m 0, i e

- and N-0VERLAP joints * ) . 25% < Ov < 50% EFFECTIVE WIDTH

N
* - f ..t. [L (V ^ ) (
V 2 h . - 4t.) + b + b , JJ
yi 1 50' 1 ' e e(ov)

50% Ov < 8 0 % EFFECTIVE WIDTH

N* f ..t. [2h. - 4t. + b + b , JJ

yi 1 e e(ov)

u i Ov > 80% EFFECTIVE WIDTH

1 1

N* - f ..t. 1[2h.- 4t. + b. + b , ,] J

i yi i i i e(ov)
 211
TABLE 5.3 (continued)

FUNCTIONS

tension: f^- f compression: f^- (-,Y-jo1nts) f f c- 0 . 8 s 1 n 0 j . (X-joints)

f kn a c c o r d i n g t o ECCS b u c k l i n g c u r v e ' a ' with: - 3 . 4 6 ( h Q/ t o - 2 ) ( l / s 1 n 9^'*

A w - ( 2 hv + o . b ) . t - for a circular
f(n) - 1.0 (tension)
V ' brace s e c t i o n :
f(n) - 1.3 - ( 0 . 4 / 0 ) . | n | (compr.) f . A,. - 0
V . V 3r J
* 0
/3

10 f . . ti . J d
10 vo
b b b b b - - . b
e " b/t ' f ..t. ' 1 e " b ^ * i
' yi i 0 0 .(.) b./t. f y .) t , 1
s b
t

VALIDITY RANGES JOINTS WITH A RECTANGULAR CHORD AND HOLLOW SECTION BRACINGS

TYPE OF JOINT JOINT PARAMETERS (i 1 or 2, j = overlapped brace)

b./t.,h./t.,d./t. b b / t gap/overlap
1 1 V 1

compres. itension h / t b./b. t . / t .

' J j

T, Y, X >0.25 51.25/ 5 35

< 35 < 35
1
, GAP ^0.1+0.01- 0.5^ 52 5 35 0.5(1-0)5-51.5(1-/3)

0.35
g > tj + t 2

25% 5 Ov 5 1 0 0 %
t .1 b .1
, OVERLAP >0.25 5 40
7 51.0, >0.75
T. B.
J J

d.
CIRCULAR BRACE 0.4< 51.577- 5 50 limitations as above f o r d^= b.
b
yi

For C h e strength equations of joints of rectangular hollow sections

the following 7 ^ factors have been incorporated:

chord plastification: 7 . - 1.0 -, Y- and X-joints

: - 1.1 K- and -gap joints

side wall failure : 7 . - 1.25 X-joints compression

> J - 1.0 -, Y- and X-joints tension
m.l
effective width : 7 . -1.25 *) -, -, X-, K- and N-joints

chord shear at gap: 7 . - 1.0 K- and N-gap joints

punching shear : y^. -1.25 *) , , X, K- and N-gap joints

See also note Appendix C.4

Note *) 7 . only incorporated in the effective width terms b e

'mj j/ u
b , . and/or b
e(ov) e

212

The IIW criteria (refs. 38 and 39) are presented in Table 5.3. Nomenclature is that of
IIW, as defined in the table, with and y the same as used elsewhere herein. Failure modes
explicitly considered include chord face yielding, chord sidewall failure, punching shear at the
material limit, effective width failure in the branch, and beam shear failure in the chord. Separate
criteria are given for the following connection types:

and Y connections
X connections (more severe chord sidewall failure)
Gap and connections (primarily punching shear)
Overlap and connections (using effective width)

These criteria are supported by extensive data (e.g., refs. 33 and 37) and by a very active
committee, IIW s/c XV-E. Limits of applicability are basically limits of the data base, and the
partial safety factors of strength are chosen in consideration of demonstrated reserve strength and
scatter. Although physical models are invoked in formulating the criteria, there is little reliance
on extrapolation. It is the author's observation that things which are accepted on faith from
origins lost in obscurity are sometimes more difficult to change than things which have
demonstrable empirical roots. Thus, the "live" roots of the IIW provisions are part of their
attraction.
North American implementation of the IIW provisions has been pioneered by Profs.
Packer and Berkemoe at the University of Toronto (ref. 40). Both have been active on the ASCE
committee on tubular structures, and Prof. Packer recently joined the AWS subcommittee. In
parallel with the present work, comparable provisions in AWS format, and using AWS
nomenclature, have been drafted. These are given in Appendix .
The provisions for circular tubular connections continue to follow the author's alpha-
based criteria (Chapter 3), but have been stated in both allowable stress design (ASD, i.e.,
punching shear) and limit state (total load) formats. The latter is intended to be compatible with
the new AISC-LRFD design specification (ref. 41). The resistance factor of 0.8 for chord face
yielding is equivalent with the old total safety factor of 1.8, when used with the AISC-LRFD
partial safety factors, for structures having 40% dead load and 60% service loads. LRFD
provides greater safety for structures having a lower proportion of dead load. This choice of
safety factor is discussed further in the draft Commentary (see Appendix).
For box sections, the IIW criteria were given only in terms of limit state design (LRFD)
format and only cover axially-loaded truss connections. Thus, they had to be expanded
considerably to cover the dual formats and wider range of design application covered by AWS.
The choice of resistance factors closely follows the previously cited work of Packer et al for the
Canadian design code, which is similar to AISC-LRFD.
These proposed revisions are currently under consideration for inclusion in the 1992
edition of the AWS Code. They will likely see additional modification before being officially
adopted. The main committee has resolved that this should be part of a general overhaul of the
Code, to include parallel ASD and LRFD provisions throughout (ref. 42).

5.3 HYBRID CONNECTIONS

In many types of construction, it is advantageous to use different kinds of members for

the chord and web members of a truss. Box and wideflange (-section) members are
advantageous for the chord, because they provide for convenient framing of lateral beams, e.g. in
the floor system of bridges and the deck sections of offshore drilling platforms. Flat surfaces on
these types of chord also simplify end preparation for the attached branch members i.e. they can
simply be saw-cut (on an angle for diagonals) and fillet welded, rather than requiring a complex
 213

saddle-shaped cope as for circular tube-to-tube connections. Circular web members are often
preferred because of their L/r efficiency, appearance, or availability.

5.3.1. Circular and Box

Here we address connections of circular branches on box chords, and not the other way
around (which is so seldom used that there have been no studies).
Although the subject is not specifically addressed in the AWS Code, a reasonable
extrapolation of the rules would be to use the same allowable punching shear, as a function of 0,
etc., at least for stepped connections.
In IIW/CIDECT rules, and the proposed AWS revision, the equivalence of total load
capacity for box and circular branch members on box chords is based on the ratio of their
respective perimeters, /4. Tube diameter is used in lieu of the various box branch dimensions,
valid for the range of 0.4 to 0.8. This in effect applies the concept of punching shear to the
problem, even though these international criteria are always given in limit state format. The
results are warranted to be on the safe side of available test results (ref. 39). An additional check
for beam shear is suggested where a flat, wide chord is used.

5.3.2. Tubular and Non-Tubular

Fig. 5.28 shows some examples of composite tube-to-wideflange connections from a
heavily-loaded drilling platform truss (ref. 43). These connections were designed by the lower
bound, cut-and-try procedures described earlier in Section 2.3.2(i) herein. Particular reference is
made to the "crossings" building block shown in Fig. 2.23 and the cutting planes shown in Fig.
2.24.

Fig. 5.28. Examples of composite tube-wide flange connections. Member sizes are given as pounds-
per-foot notation.
214

TABLE 5.4 DESIGN STRENGTH AND VALIDITY RANGES OF WELDED JOINTS OF

RECTANGULAR SQUARE OR CIRCULAR BRACINGS AND AN I- OR H-SECTION
CHORD (REF. 39)

TYPE OF JOINT DESIGN STRENGTH (1-1.2)

-, Y- and X-joints CHORD WEB YIELDING

f . t . b
vo w m
sin $

EFFECTIVE WIDTH

N. - 2f .. t.. b
1 yl 1 e

K- and N-GAP joints CHORD WEB STABILITY

* f . t . b
u vo w m NO CHECK REQUIRED FOR
. - .
sin , EFFECTIVE WIDTH IN CASE:

EFFECTIVE WIDTH
g 5 20 - 28
5 1.0 - 0.037 and
- 2f .. t.. b
<V 2 l 2
d ,b /b
1 yi e 0.755 5 1.33
CHORD SHEAR

f . ..
vo V
73. sin

2 0 ,5
N* (A - A w).f + A H. f L[1 - V/V ) J]
o ^ o V yo V yo ' p'

K- and N-OVERLAP joints NOTE: * ) . 25% 5 0v < 50% EFFECTIVE WIDTH

* L V
/ V ; J
1 yi 50 i i e e(ov
50% <; Ov < 80% EFFECTIVE WIDTH

N* - f . t. [2h. - 4t. + b + b . JJ
yi e e(ov)

0v > 80% EFFECTIVE WIDTH

f ..t. [2h.- 4t. + b. + b . JJ

yi i i i e(ov)

FUNCTIONS

b A - A (2 - a)b .t + (t + 2r ).t
/ - - + 5(t v+ r )
Rj / m sin $ ' V
0 0 W 0 0

0.5
b < 2t. + 10(t v+ r ) for CHS-brace * 0
m '
CI d. 4fll
1 + V - (..sin . )
2
b - "7 + 5(t v+ r ) 3t
m sin , o' max
V - f . A W/ 7 3
yo

f ..t.
b - t + 2r + 7.-^. t J0_ Y.I ,1 . ub,
e w f .
5 b.
e(ov) b./t. * f ..t.* i * "1
yj J j yi
* ) . Only the overlapping bracing is to be checked. The bracing efficiency (i.e. the
design strength divided by the full yield capacity of the bracing) of the over-
lapped bracing is not to be taken higher than that of the overlapping bracing
 215
TABLE 5.4 (continued)

VALIDITY RANGES JOINTS WITH AN I- OR H- CHORD, AND HOLLOW SECTION BRACINGS

TYPE OF JOINT JOINT PARAMETERS (1 1 or 2, j overlapped brace)

h /t b./t., h./t . d./t. h./b. b./b.

r r r J '
Yf w compression tension

h
f \.Uf- *
1
X hw 400 0.5s r s2.0
b
w yo l

h b
T, Y i i

w yo 1 1 yl 1 1

N-GAP
h
w
<; 400 h^ yl
7
1

1
S 50
"
1

1
K-, N-OVERLAP 0.5S i2.0 ^ 0 . 7 5
b
i

For more lightly loaded, long span trusses, e.g. pedestrian walkway bridges, simple
unstiffened tube-to-H-section connections may suffice, despite their limited capacity. The
IIW/CIDECT rules (ref. 39) for these connections are given in Table 5.4. Failure modes
considered are: chord web yielding, chord web stability, chord beam shear, and effective width.
Comparisons between the effective width criterion and the "crossings" method were previously
discussed, with the latest IIW rule corresponding to a rather generous spreading slope of 3.5:1.
Where lateral bracing frames into the web of an -section chord, local capacity (e.g.
punching shear) of the connection would be similar to that given for chord face yielding in a box
section having the same thickness and width, provided the -section flanges are restrained
against rotation. Where the flanges are not so restrained, the yield line mechanism of Fig. 5.10
would suggest 30% lower capacity.

5.4 SPECIAL TOPICS

5.4.1 Overlapped Connections

By providing direct transfer of load from one branch member to the other in K- and N-
connections, overlapping joints reduce the punching demands on the main member, permitting
the use of thinner chord members in trusses. These are particularly advantageous in box sections,
in that the member end preparations are not as complex as for circular tubes.
In principle , the "building block" method of calculation contained in the existing AWS
Code (clause 10.5.1.5) should be extendable to box connections, even though the original
development was for circular tubes (see Section 6.1 herein). The AWS building blocks are
punching shear at the chord and weld shear at the overlap. The IIW/CIDECT provisions (Tables
5.3 and 5.4) address the problem differently, via effective width criteria. For matched box
connections, equivalency of AWS punching shear and IIW effective width has been previously
discussed.
Fully overlapped connections, in which the overlapping brace is welded entirely to the
thru brace, with no chord contact whatsoever, have the advantage of even simpler end
preparations. However, the punching problem that was in the chord for gap connections is now
transferred to the thru brace, which also has high beam shear and bending loads in carrying these
loads to the chord.
216

Most of the testing of overlapped connections has been for perfectly balanced load
cases, in which the compressive transverse load of one branch is offset by the tension load of the
other. In such overlapped connections, subjected to balanced and predominantly axial static
loading, tests have shown that it is not necessary to complete the "hidden" weld at the toe of the
thru member. In real world design situations, however, localized chord shear loading or purlin
loads delivered to the panel points of a truss result in unbalanced loads. In these unbalanced
situations, the most heavily loaded member should be the thru brace, with its full circumference
welded to the chord, and additional checks of net load on the combined footprint of all braces are
required.

5.4.2. Stiffened Box Connections

Given the loss of connection efficiency implied by effective width or equivalent
punching shear criteria for simple direct-welded box connections, heavily-loaded structures often
require stiffened connections. Figs. 5.29 and 5.30 show two examples, multi-planar connections
from a massive space-frame deck structure of a proposed offshore platform (ref. 44). Truss
chord members are 36x48-inch welded box members, with matching width 36x36-inch web
members. Lateral bracing is 24x24-inch box members, bottom-flush with the chord so as to
avoid interference with deck beams which are top-flush. The static strength design approach for
these connections will now be discussed:
Choice of load case (and design method) Alternatives considered are:

(1) design connections for many different specific load cases, or

(2) design connnections to develop full strength of members, using one or more
characteristic load patterns, as required for equilibrium.

The author recommends, and has extensively used, alternative number two. It is much
simpler, especially if the connection checks must be done manually. It is also more amenable to
checking future load changes; all the engineer must do is check the members, e.g. using the
computer analysis post-processor. Alternative number one would require the development of
such a post-processor for checking the connections, in order to be practical on large structures
with many different load cases. Automating these checks would be extraordinarily difficult,
given the nearly infinite variety of stiffening schemes which may be conceived.
The basic approach is the lower bound "building block" limit analysis as described
earlier, in Section 2.3.2.(i). Even where the rest of the design is based on working stress,
designing connections so that their characteristic ultimate strength matches the yield (or local
buckling) strength of the members is proposed.
Unstiffened box Where punching shear provides adequate capacity, or there are
alternative load paths within the connection (e.g. where gussets extract most of the load from a
member before it reaches the unstiffened intersection), matching internal diaphragms may be
omitted. Proposed revisions to AWS D l . l use effective width concept instead of punching
shear; this should be quite compatible with AISC treatment of local buckling. Other connection
failure modes must also be checked, e.g. general collapse of the side walls, etc.
Diaphragm The matching strength method will be described. Characteristic ultimate
strength of a solid diaphragm is given by:

F Q Q fc b F
Du = f a w y (5.14)

while the characteristic strength of an open diaphragm (with access hole) is:
 217

ft^4

PlA6a7rJAL &A66-

DETAIL C-5-U

Fig. 5.29. Example stiffened box connection with diaphragms.

Fig. 5.30. Example stiffened box connection, with gussets.

218

F 2h F 5 15
D u = Qf Qs *w y < ' >

where Q a and Q s are reduction factors for effective area and net section stress, respectively, as
defined in AISC effective width method (Appendix C to ref. 18); Qf is a reduction factor for the
presence of other loads in the diaphragm (e.g. chord stresses in intersecting trusses); b is the
width of the member; h is the width remaining on either side of the access hole; and t w is web
thickness. These characteristic strengths should equal or exceed the normal capacity of the
mcoming member face which it is intended to back up:

F fc b F s i n ( i 5 16
D u ^ 2a 2s b y < >

where is 90-degrees for a perpendicular branch member. Note that Eqs. 5.14 and 5.15 presume
that the diaphragm is perpendicular to the axis of the member in which it resides; if this is not the
case, these equations would also have a sin(0) term on the left hand side.
Gussets ~ For major intersecting truss nodes, a prefabricated assemblage of crossing
gussets may be used, as shown in Figure 5.30. These have been made 0.125 to 0.25-inch thicker
than the thickest member sidewall in each truss. Acute angles should be avoided where members
first encounter the gussets are to be avoided, with angles of 150-degrees or greater to be
preferred, for reasons of stress concentration. Scalloped gussets are even better. Load transfer
from the incoming members to the gusset is edge-to-edge membrane stress for the sidewalls
(very efficient) and by shear lag in the other faces (a shear length of at least one member-width is
required for full transfer). Internal cutting planes in the gusset should also be checked for
combinations of thrust, moment, and shear as required for equilibrium of the free bodies created
thereby (see Fig. 2.25).
T h r u material The side walls and face plate of the thru member at each member
intersection does double duty as connection material, and should be checked for this function, as
well as for interaction with stresses from the member's own loads. The characteristic ultimate
strength of a sidewall plate panel, between diaphragms is given by:

F fc a F 5 17
P u = Qf Qa w x y < >

This should equal or exceed the correspondingg branch capacity, i.e.,:

F Q fc a F s i n 5 18
Pu > ^ s b y <*> < - >

where a is the in-plane branch member dimension, and a^^ = a/sin(0).

Combined load effects These are reflected by Qf in the AWS Code, which takes a
value of 0.7 for fully loaded chords under the existing rules. Alternative forms of load interaction
considering biaxial membrane stress and shear can be found in API Bulletin 2V, Kinra's
more complete proposals, and SCI's fixes to Kinra (refs. 45,46,47).
Overlapping nodes Cutting planes for combined loads from groups of members are
shown in Fig. 2.24. The combined footprint can become critical when all the incoming members
are loaded in compression, as at support points.
Matching scantlings Since we are often dealing with a limited number of member
sizes, it will be useful to pre-calculate matching diaphragm and side panel sizes. This almost
reduces joint design to a table look-up and drafting exercise, at least for the simple repetitive
situations.
Additional comments relate to the two example connections shown:
Detail C-5-U (Fig. 5.29) ~ intersecting trusses, fully backed up by diaphragms, except
for 24-inch lateral bracing, which is not as fatigue sensitive as at the next detail.
 219

Detail B-2-L (Fig. 5.30) with crossed gusset subassemblage, and heavy tension chord.
Note use of partial penetration weld in shear load transfer, to save access window for back-weld.
Use of fillet weld at dead end, and lack of diaphragm back-up at this point, presume that most of
the load has already been taken out into the gusset via shear. Scalloped filler gussets on 24-inch
lateral bracing is for anticipated fatigue problems here; SCF at their notched termination would
need to be determined from finite element analysis. Finite element analysis has indicated that
scalloping the main vertical gussets, rather than having straight cuts as shown, is also
advantageous from the standpoint of stress concentration. For heavy sections, this is potentially
important from a brittle fracture standpoint, as well as for fatigue.

5.5 FATIGUE OF BOX CONNECTIONS

Fatigue of box connections has not been studied as much as their static strength, nor as
much as fatigue of circular sections. Much of the relevant work has only been reported within
the last decade and we are still trying to develop our understanding of the subject. The AWS
Code offers no specific guidance, leaving the designer on his own to extrapolate the provisions
for circular sections. As discussed below, fatigue criteria are given in two formats: (1) the
classification method, and (2) the hot spot method.

5.5.1 Classification Method

In this method, the designer searches a list of fatigue categories (e.g. Table 4.3 herein) or
browses through a picture gallery (e.g. in the bridge or building Codes), to find the category
which most closely fits his design situation, and then uses the corresponding member of a family
of S-N curves. Stress is usually defined as the peak-to-trough range of nominal stress (P/A +
M/S), or something equally simple like punching shear. All the S-N curves in the AWS code,
except XI and X2, belong to this method.
One of the simplest situations is an end-to-end butt weld, as shown in Fig. 5.31. For
complete joint penetration groove welds, the S-N curve of choice would be AWS category C.
However, in order to qualify for this level of performance in a single-sided closure weld, AWS
requires special welder testing and non-destructive testing (clause 10.12.6.1). Failing this, the
2
S r{ N / m m )

Fig. 5.31. Butt welded end-to-end connections (t = 4mm) (ref. 48).

220

weld classification reverts to partial penetration, with lower allowable stresses to account for
problems at the root of the weld, even where we assume the full cross section. In this case, AWS
curve F applies. As a lower bound, it appears to fit the test data (ref. 48), for which the
originators make reference to incomplete penetration at the root, even better than their own
correlation curves (labelled 50% and 95% survival).
Fillet-welded end connections involving a cross plate (but not the relative flexibility and
uneven load distribution problems inherent in plate/shell action of the chord wall in tubular
connections) are shown in Fig. 5.32. This situation is covered by AWS categories (for base
metal adjacent to the weld) and F (for the weld effective throat). These curves fall on the safe
side of data for circular sections (CHS, ref. 49). However, they fall on the unsafe side of data for
rectangular hollow sections. Using AWS curve XI with an SCF of 3.0 would be adequately
conservative, but there is nothing in the classification approach, or in previous experience with
circular sections, to suggest this to the designer. One can speculate on the reasons for this lower
performance of RHS (e.g. the lack of axisymmetric conditions making the effects of weld
eccentricity more serious than for CHS, and tending to concentrate load transfer in the corners,
where weld defects are also most likely), but it is basically a nasty surprise, especially for such
thin material.
2
f Sr (N/mm )

-f R = 0.1 F e 3 6 0 c.f
(100 100 6 )

X R =01 St 47
( B0 8 0 6 5 )

R=-1 St 47
(80.80.65)
R =0.5 S t 47
( 80 - 80 * 6 5 )

fillet welds
=t

CHS
St 35
90 5 i

UP >i 1 20
/icycles)
7
10

Fig. 5.32. Fillet welded end connections (t = 6-6.5mm) (ref. 48).

For tubular connections, things become even more difficult. Not only does the relevant
local stress pattern within the connection become quite complex and fundamentally different
from brace nominal stress, but the relationship between the two varies with connection type, and
with the geometric parameters within a given type. Nevertheless, until these relationships have
been understood (e.g. via parametric SCF formulae), it is possible, perhaps even necessary, to
take the empirical approach of testing many specimens of a given connection type and drawing a
characteristic (95% survival) design line, whose application is then strictly limited to the
parameter space covered by the data base. Ten years ago, this approach was taken by TH
Delft/Stevin Lab (refs. 33, 50), and the resulting design criteria are given in Fig. 5.33.
Focusing first on Delft curve A, for simple gap K-connections in RHS, we see the
reference cyclic stress range S r x defined as a kind of pseudo punching shear, i.e.:
 221

Sr = Sr for r > r l i m

tot
(5.19)
s S r f ro T r
r = r ' lim < lim
tot

where ^ r t Qt is the total nominal stress range in the brace, axial plus bending; and r is brace/chord
thickness ratio (t^A 0), with chord failure governing above r ^ m of 0.5. To avoid using the term
punching shear, some Europeans refer to this failure mode as "cratering" (ref. 49); this is also
appropriate oilfield terminology.
1
Sr (N/mm )

Recommended 9 5 % s u r v i v a l S r- N

curves for K - o n d N - t y p e joints

m a d e o f square h o l l o w sections

for - 1 R s +0.2

bracing c h e c M c u r v c s A.B.C):

f ( , f ( S
)
Sol" * V
< ^ rl i m i t
~ i

f* / c \ ( =
S 500 < 1.0
> b + 300

b i n mn)

Values f o r t Q/ t . l i m i t

gap joints 2.0

K-type overlap joints 1.2

N-type overlap joints 1.4

chord check (curve D )

5 r t to < S r x. f ( S )

Note:

(axial + bending)
nominal
7
2 3 4 5 6 7 8 9 3 4 5 6 7 8 9 10 (cycles)

Fig. 5.33. Delft 95% survival S f-N curves for K- and N-type joints made of square hollow sections
(ref. 50).

The supporting data base is displayed in Fig. 5.34. The corresponding limitations are:
square branches, of nearly equal size, intersecting the chord at 40 to 90-degree angle; chord
width/thickness up to 25 (gamma up to 12.5), widths up to 200mm (8 in.), steel grades Fe 360
and Fe 510 (50 to 72-ksi yield); betas of 0.5 to 1.0; gaps ranging from 100% to 220% of the side
step dimension; and stress ratios between -1.0 (full reversal) to 0.2. For chord widths less than
the limit, there is a beneficial size effect. For design against crack initiation, instead of failure as
shown, it is suggested to reduce the design fatigue strength by a factor of 1.2. Furthermore, if
secondary bending stresses and uneven joint compliance are not accounted for in the truss
analysis, performance is effectively knocked down by an additional factor of 1.5 (for diagonal
braces) to 2.2 (for the stiffer perpendicular braces in N-connections), according to Reference 51.
AWS criteria for punching shear are compared with Delft-A criteria in Fig. 5.35. The
upper curve, K l , is used because of the thinness of the joints to which these criteria apply. AWS
plots far to the safe side, even when adjusted for differences in limiting gamma ratios, using the
0.7 power rule, as indicated by the arrows. For diagonal braces, much of this apparent
 222

RHS K-and N-type

joints with gap 0.5 < -^1 ; 10
: ( failure)
bo

test results
plotted against
S r with

-joints
-joints

\.
_ k ' . .
8

>
95% s u r v i v a l line /

I I I I I I I ! I I 1 I 1 ! 1 I ' \ 1 1 1 1 1
5 6 7
10^ 10 10 10

NUMBER OF CYCLES

Fig. 5.34. Delft curve "A" and summarized test results for gap and connections at failure (ref. 51).

conservatism would be removed by the sin(0) term in the AWS punching shear expression.
Similarly, AWS curve DT for cyclic stress in the brace plots to the safe side of the corresponding
Delft curve, but the difference is largely offset by Delft's allowance for secondary bending stress,
cited above.

The Delft criteria show a beneficial effect for overlapping the braces in N- and K-
connections (the corresponding values of T y i m are raised to 0.7 and 0.83, respectively).
Comparing AWS curves DT and ET would suggest the opposite trend. Since the Delft criteria
are based on specific tests of box connectionswhile AWS curve ET is being extrapolated from
circular tubes, and is contaminated by messy gusseted jointsit should be obvious which one to
believe.
 223
5.5.2 Hot Spot Method
In applying this method, the designer must go beyond nominal stress in the brace or
punching shear stresses in the chord, and analyze the connection as a structure in its own right.
The geometric hot spot stress, sigma-G, has proven useful in bringing many different connection
geometries to a common design basis, and has long been established as the preferred method for
circular tubes in offshore structures. Its application to box connections has only recently begun
receiving attention, and a totally consistent picture is yet to emerge.
After a couple years' iteration, hot spot fatigue criteria were formalized by IIW Subcie.
XV-E in 1985 (ref. 51), as shown in Fig. 5.36(a). The earlier Delft-A criteria for simple gap
connections have been re-formatted as curve hot spot stress, together with the stress
c o n c e n t r a t i o n factors given in Fig. 5.37. T h e s e SCF have a m i n i m u m value of 3.0,
corresponding to branch member failure as in the "perfect" end connection of Fig. 5.32,
increasing to higher values for the "cratering" failure mode ( r > ri[m)> Ignoring the paradox
which arises from the notion that hot spot stresses in box K- and N-connections are somehow
different from other classes of welded hardware, the curve and SCF taken together do reproduce
the data base for small thin connections, including those with overlap.

NUMBER OF C Y C L E S NUMBER O F C Y C L E S

Fig. 5.36. Design S-N curves for hot spot stress, (a) IIW criteria, (b) Possible AWS adaptation.

< 12.5

- 0.5 Q <_ 1.0

> 40
-

-
GAP

- ^ = 0 . 5 to 1.1

F i g . 5 . 3 7 . IIW s t r e s s
OVERLAP ' concentration factor for various
50-100%
K-connections.
I I 1 1 1 1 1 1 1
0.5 1.0
r
224

Fig. 5.36(a) also shows a family of S-N curves for more rigorously derived values of the
geometric hot spot stress, sigma-G. These show the same size effect (exponent of -0.25 in eqn.
4.9) as previously adopted for offshore practice, after some committee fiddling with the position
of the knee of the curves and with flatter slope of the curves for thinner sections. Supporting
fatigue test data (ref. 52), with thicker sections than previously tested and more rigorous
determination of sigma-G, are also shown in the figure. These crowd uncomfortably close to the
proposed design curve, with some of the data actually falling on the unsafe side of the
corresponding offshore rules (U. K. Dept. of Energy curve T).
The critical hot spot stress locations in box connections are at the corners of the branch
tube, as shown in Fig. 5.38. For T- and X-connections, gage lines A and/or govern, depending
on r, while for gap K- and N-connections, gage lines D and in the gap region govern.

Fig. 5.38. Lines considered for

the measurement of hot spot
wall thickness of b r a c e 1
strain (ref. 53).
\ i

Local stresses along gage line in an X-connection tested and analyzed at Delft (ref.
53) are shown in Fig. 5.39. This reveals highly localized, steep, and non-linear strain gradients-
-raising some possible difficulties in applying the European definition of hot spot stress, sigma-
G. Conventionally, this is determined by linear extrapolation from a region outside the notch
effect at the toe of the weld. In connections of circular tubes, the relationship between shell
bending and punching shear which is built into the alpha-Kellogg formulation indicates a decay
distance for shell bending stress of approximately 0.3a chord thicknesses, or 3 thicknesses for
an axially loaded X-connection with 7 of 18, so the extrapolation is made from within the region
of representatively high shell stress. This is not the case for the box connection shown in the
figure, with the extrapolation having to be made from a region that barely gets above the nominal
stress. The investigators at Delft tried to cope with this situation by using a very fine mesh
(several elements within the extrapolation distance), using a quadratic extrapolation to the weld
toe, rather than linear, and using a multiplier of 1.1 to get from strains to stress (see eqn. 4.3).
The usual finite element modelling practice, with element size on the order of the thickness and
linear variation of strain across the element, is even more likely to underestimate the local
stresses in this kind of situation.
It is tempting to suggest that the less rigorous AWS definition of hot spot stress
(measured at about 3mm from the weld toe) or the SAE approach (6mm gage length straddling
the weld toe) might suffer from fewer of the aforementioned difficulties. However, they create
new ones, such as a lack of invariance of SCF with scale. Also, compensating errors in the
traditional AWS approach (missed gradient vs. included notch effect) seem likely to have
unpredictable results here. Sigma-G (adjusted for actual weld toe position, where appropriate),
and a conservative size effect, seem like a more defensible basis for practical designs. This last
approach was recently endorsed by IIW Subcie XV-E (ref. 54). Simply using the correlation
curves from small scale tests, without modification, is not tenable for large scale applications.
 225

range o f data
used f o r (7q
extrapolation

nominal stress
( as s t r a i n )
distance from
weld toe ( mm)

Fig. 5.39. Detail of local strains

at the line (ref. 53).

A possible adaptation of the AWS hot spot S-N curves for box connections can be found
in Fig. 5.36(b). These provide a more comfortable margin relative to data from the thickest
specimens tested thus far (still only 16mm chord, c. 10mm branch). For thicknesses beyond the
limit for curve X2 (assuming basic flat weld profiles), they also incorporate the modified size
effect suggested in the latest Delft work (ref. 52), with the exponent to be used for fatigue
strength in Eqn. 4.9 given by:

lexponentl = 0.075 log (5.20)

The size effect exponent is thus -0.30 for 10^ cycles and -0.45 for 10^ cycles, more severe than
the value of -0.25 given in current Codes. This has the desirable effect (from the standpoint of
matching recent data trends from ECSC fatigue research) of steepening the S-N curve for the
heavier thicknesses.
Figure 5.40 shows some initial results, strain concentration factors (SNCF), from the
ongoing Delft work (ref. 53). Consistent with IIW rules on presumed secondary bending, branch
nominal stress includes 33% in-plane bending; pure axial load would have 13% higher SNCF.
These results are also to be multiplied by 1.1 to get SCF. For of 0.6 and 2y of 25, we see
rough consistency with the earlier IIW rule: brace failure at SCF of 3.0 for r < 0.5, with higher
SCF and chord failure for higher values of r. In actual fatigue testing, brace failure occurs less
often than indicated by the raw SNCF, presumably because of the stress relaxation opportunities
and favorable size effect accompanying reduced brace thickness in typical connections.
Application of the foregoing methods, to matched K-connections of square branches
framing onto the narrow side of a rectangular chord, can be found in Fig. 5.41. Consistency of
the Delft-A curve (pseudo punching shear) with the IIW hot spot format (curve for thin
sections) can be seen , when the former is corrected for the effect of r as indicated by the upward
pointing arrows. These criteria fall well on the safe side of the test data (ref. 55).
Application of the AWS criteria is also shown, for the approach in which fatigue
strength is taken as a fraction of ultimate strength, as previously discussed for circular sections in
connection with Figs. 4.26 and 4.34. Static strength was estimated using the "building block"
approach for tensile loading in AWS box connection rules: sidewall yielding, heel & toe
punching, K a , sin(0), and shear in the overlap where present. Using the author's traditional
limiting SCF in the branch of not less than 1.8 (based on experience with circular sections),
 226

8
Li.

0 05 1.0 0 0.5 1.0

Fig. 5.40. Delft SNCF for concentric 4 5 K-connections of square tubes (gap varies with ) (ref. 53).

NUMBER OF CYCLES

Fig. 5.41. Fatigue data and design approaches for matched K-connections (ref. 55).

hardly any benefit accrues from the overlapping third brace; while the test data suggests that the
superfluous overlap might even be harmful. Comparing AWS (low SCF and conservative S-N
curve) and IIW (higher SCF and optimistic S-N curve), we see similar results. Although using
the limiting SCF of 1.8 with AWS curve XI produces results on the safe side for this particular
data set, it is more optimistic than IIW. Given the evidence of Fig. 5.32 (SCF of 3.0 for fillet
welded "perfect" end connections), caution in using SCF this low is advised. Using a limiting
SCF value of 2.5, as implied in AWS curve DT for connections with complete penetration
 227

groove welds, would bring the IIW and AWS approaches in line with each other, as indicated by
the downward arrows in Fig. 5.41. In the ECSC work, a lower limit SCF of 2.0 has been chosen
(ref. 56).
Wardenier gives additional practical design guidance for fatigue in box connections in
Reference 57.

REFERENCES

1 Marshall, P. W., Prospect 225 Deck (WD 133) design calculations and drawing set NO-D-64-21,
Shell Oil Company, 1964.
2 Godfrey, D. G., Auk A Deck Section, design calculations and drawing set NO-D-72-3, Shell Oil
Company, 1971-72.
3 Quotation from Rice University diploma.
4 Redwood, R. G., The Behavior of Joints Between Rectangular Hollow Structural Members, Civil
Engineering in Public Works Review, London, October 1965.
5 Jubb, J. . M. and Redwood, R. G., Design of Joints to Box Sections, Institute of Civil Engineers
Conference on Industrialized Building and the Structural Engineer, London, May 1986.
6 Cute, D., et al, Welded Connections for Square and Rectangular Structural Steel Tubing, Drexel
Institute of Technology, Philadelphia, November 1968.
7 Brockenbrough, R. L., Strength of Square-Tube Connections Under Combined Loads, Proceedings
ASCE, Journal of Structural Division, December 1972.
8 Eastwood, W. and Wood, . ., Recent Research on Joints in Tubular Structures, Canadian
Structural Engineering Conference, 1970.
9 Graff, W. J., Welded Tubular Connections of Rectangular and Circular Hollow Sections, paper
presented at Texas Section ASCE, El Paso, October 1970.
10 Graff, W. J. and DeGeorge, B. J., Review of Research on Welded Tubular Connections, ASCE
Specialty Conference on Steel Structures, Columbia, MO, June 1970.
11 Graff, W. J., et al, Punching Shear Characteristics of RHS Joints, ASCE Preprint No. 1963,
National Structural Engineering Meeting, San Francisco, April 1973.
12 Marshall, P. W. and Toprac, . ., Basis for Tubular Joint Design Codes, ASCE Preprint 2008,
National Structural Engineering Meeting, San Francisco, April 1973.
13 McCarthy, J. R., Welded Connections of Shaped Structural Steel Tubes, Design Project for Master
of Engineering Degree, University of Wisconsin-Milwaukee, 1976.
14 Sherman, D. R. and Marshall, P. W., Commentary on the Static Design of Tubular Connections,
ASCE preprint, Ocean Engineering Convention, San Diego, April 1976.
15 Marshall, P. W., A Review of American Criteria for Tubular Structures - and Proposed Revision,
IIW Doc. XV-405-77, Copenhagen, July 1977.
16 Marshall, P. W., proposed revision to Dl.1-72, November 13,1973; also presented at the Structural
Welding Code Seminar, New York, October 1973.
17 Cran, J. ., et al, Hollow Structural Sections - Design Manual for Connections, Steel Company of
Canada, Ltd., Hamilton, Ontario, 1971.
18 AISC Specification for the Design, Fabrication & Erection of Structural Steel for Buildings,
American Institute of Steel Construction, New York, 8th Edition, 1980.
19 Monograph on the Planning and Design of Tall Buildings, Vol. CL, Criteria and Loading, ASCE,
New York, 1980.
20 Sherman, D. R., et al, Beam Connections to Rectangular Tubular Columns, AISC National Steel
Construction Conference, Miami, June 1988.
21 Mouty, J., Calcul des charges ultimes des assemblages soudees de profils creux carres et
rectangulaires, Construction Metallique, June 1976.
22 Marshall, P. W., engineering calculations dated November 9,1973, Tubular Joints binder.
23 Giddings, T. W. and Wardenier, J., The Strength and Behaviour of Statically-Loaded Welded
Connections in Structural Hollow Sections, Section 4, CIDECT Monograph 6, British Steel
Corporation, Tubes Division, 1986.
228

24 Wardenier, J., Modified Eurocode 3 Design Recommendations for Hollow Section Lattice Girder
Joints, IIW Doc. XV-E-87-120, in R. Bjorhovde, ed., Connections in Steel Structures, Elsevier,
May 1987.
25 American Institute of Steel Construction, Manual for Steel Construction, 8th Edition, 1980.
26 Davies, C , et al, The Behavior of Full-Width RHS Cross Joints, Proceedings 2nd International
Conference on Welding of Tubular Structures, IIW, Boston, July 1984 (Pergamon Press).
27 Packer, J. ., Review of American RHS Web Crippling Provisions, technical note, ASCE Journal
of Structural Engineering, December 1987.
28 Mehotra, B. L. and Redwood, R. G., Load Transfer Through Connections Between Box Sections,
AISC Engineering Journal, August/September, 1970.
29 Mehotra, B. L., et al, Shear Lag Analysis of Rectangular Full-Width Tube Junctions, Journal of the
Structural Division, ASCE, Vol. 98, No. ST1, paper 8665, January 1972.
30 Dawson, T., remarks at AWS Dl Code Committee meeting, c. 1974.
31 Packer, J. ., Berkemoe, P. C. and Tucker, W. J., Design Aids and Design Procedures for H.S.S.
Trusses, ASCE Journal of Structural Engineering, July 1986.
32 Wardenier, J., Design and Calculation of Predominantly Statically-Loaded Joints Between Square
and Rectangular Hollow Sections, Van Leeuwen Technical Information No. 7, Zwijndrecht,
Holland, 1988.
33 Wardenier, J., Hollow Section Joints, Delft University Press, 1982.
34 Hollow Structural Sections - Design Manual of Connections, 2nd Edition, Stelco, Inc., Hamilton,
Ontario, 1981.
35 Ricker, D. T., Practical Tubular Connections, presented at the ASCE Structural Congress, Chicago,
September 1985.
36 Sherman, D. R., Tentative Criteria for Structural Applications of Steel Tubing and Pipe, AISI
Committee of Steel Pipe Producers, August 1976.
37 Wardenier, J. and Davies, C , The Strength of Predominantly Statically-Loaded Joints with a
Square or Rectangular Hollow Section Chord, TU Delft/TNO/IIW Doc XV-492-81, Oporto,
September 1981.
38 IIW s/c XV-E, Design Recommendations for Hollow Section Joints Predominantly Statically-
Loaded, IIW Doc. XV-491-81, Oporto, 1981 (revised).
39 IIW s/c XV-E, Design Recommendations for Hollow Section Joints ~ Predominantly Statically-
Loaded - 2nd Edition, 1989, draft of May 25,1989.
40 Packer, J. A. and Berkemoe, P. C , Canadian Implementation of CIDECT Monograph No. 6,
University of Toronto, Department of Civil Engineering, report 84-04, CIDECT report SAJ-84/9-
E, IIW Doc s/c XV-E-84-072, July 1984.
41 Proposed Load & Resistance Factor Design Specification for Structural Steel Buildings, American
Institute of Steel Construction, September 1983.
42 AWS agenda item 6-06-895, Revisions to Tubular Connection Design Rules, videotape
presentation for Walt Disney World, March 1990.
43 Marshall, P. W., 12-Leg Deck for VE-257, Shell Drawing NO-D-65-10A, 1965.
44 Marshall, P. W., et al, Shell TLP Study: Structural Design of Deck, Hull, and Template, Shell Oil
Company CE Report 85, May 1987 (proprietary).
45 API Bulletin on Design of Flat Plate Structures, API Bulletin 2V, American Petroleum Institute,
May 1987.
46 Kinra, R. K., Summary of proposed interaction formulas for plates under bi-axial loading, lateral
pressure, and edge shear, Shell memo, February 23,1987.
47 Dier, A. F., et al, Review of Proposed Plate Interaction Formulae, the Steel Construction Institute,
Ascot (UK), report to Shell Oil Company, doc. SCI/101/87, December 1987 (proprietary).
48 Wardenier, J. and Dutta, D., The Fatigue Behaviour of Lattice Girder Joints in Square Hollow
Sections, IIW Doc. XV-493-81, Oporto; also presented at Conference on Joints in Structural
Steelwork, Teeside, April 1981.
49 Dutta, D., Mang, F. and Wardenier, J., The Fatigue Behaviour of Hollow Section Joints, CIDECT
Mongraph No. 7, English version, Constrado Division of British Steel, Croydon, 1982.
50 Wardenier, J., et al, The Fatigue Behaviour of Welded Joints in Square Hollow Sections - Part II,
Final Report to ECSC, Delft/TNO-IBBC Stevin Report 6-80-3, May 1980.
 229

51 Subcie XV-E, Recommended Fatigue Design Procedure for Hollow Section Joints, IIW Doc. XV-
685-85, Strasbourg, 1985.
52 van Wingerde, A. M., et al, The Fatigue Behaviour of T- and X-Joints Made of Square Hollow
Sections, International Symposium on Tubular Structures, Lappeenranta, September 1989.
53 Puthli, R. S., et al, Numerical and Experimental Determination of Strain (Stress) Concentration
Factors of Welded Joints Between Square Hollow Sections, Heron, V. 33, No. 2, TH Delft/TNO-
IBBC, 1988.
54 Notes of IIW s/c XV-E meeting, June 22,1987.
55 Strommen, . N., Experimental Investigation of the Fatigue Capacity of K-Joints with Rectangular
Hollow Sections, IIW Doc. XV-E-83-045, September 1983.
56 Wardenier, J., personal communication.
57 Wardenier, J., Fatigue Design of Hollow Section Joints, van Leeuwen Technical Information No.
9, Zwijndrecht, 1988.
Chapter 6

SPECIAL TOPICS FOR CIRCULAR SECTION JOINTS

In this chapter, we shall examine the following specialized topics, as they apply to the
design of tubular connections for circular hollow sections:

overlapping connections
multi-planar connections
grouted connections
internally stiffened tubular connections

The first two are explicitly covered by provisions in the AWS D l . l Code. The last two
go beyond the Code, but are consistent with its usage.

6.1 OVERLAPPING CONNECTIONS

In overlapping connections, the branch members intersect each other as well as the
chord, and part of the load is transferred directly from one branch to the other through their
common weld. The principal advantage of such connections is that they make efficient use of
large diameter, thin wall tubular chords, without joint-cans, since the chord no longer is required
to transfer the entire load. This is particularly useful in long span roof trusses.
In trusses with wideflange -sections as chords, and tubular web members, a similar
advantage accrues from the direct brace-to-brace load transfer, circumventing the limited web
crippling and web shear capacities of the chord. This type of design may be seen in pedestrian
bridges, equipment modules, and deck trusses of offshore platforms.
In offshore jacket practice, the popularity of negative-eccentric joints peaked during the
period when the radial line load capacity of cylindrical shells was not fully understood, and
generally being underestimated as well. Their decline was prompted by the following
disadvantages: more complicated fabrication (with higher unit costs), and a potential for higher
stress concentrations in the branch members, particularly at their common weld where one
branch bears on the other. Nevertheless, overlap can still occur as a result of congestion, even in
nominally concentric connections.
The amount of overlap can be controlled by adjusting the eccentricity of the branch
member centerlines, as shown in Fig. 6.1. Negative eccentricity can be used to increase the
amount of overlap and the static load transfer capacity of the connection. Positive eccentricity

(a) (b) (c)

Fig. 6.1. Connection eccentricity, (a) Positive, (b) Zero, (c) Negative.
 231

can be used to maintain a gap. Moments due to eccentricity are primary moments, necessary for
the stability of the structure, and should generally be accounted for. In trusses where there is a
continuous chord, substantially larger than the branch members, it is reasonable (and consistent
with the lower bound theorem of plasticity) to allocate the entire moment to the chord. Some
Codes, e.g. API, allow this moment to be neglected if the eccentricity is "minor" (less than D/4);
however, this practice can result in unconservative designs, especially in the case of interior
space frame connections where all the members are of similar diameter, or where the enlarged
chord only extends for a short length (so-called "balloon joints").

6.1.1 Limit Analysis

Approximate limit analysis of these connections may be carried out using the cut-and-
try methodology and "building block" approach of Section 2.3.2(i) herein. Figure 6.2 shows a
historical example of such an evaluation for axial load capacity (using the prevailing low
estimates of shell line load capacity).

Fig. 6.2. Historical example of

"building block" approach to
limit analysis.

Building blocks, shown at their nominal or allowable capacities, include radial line load
bearing and membrane shear at brace-chord and brace-brace intersection welds, as well as the
somewhat superfluous contribution of the gussets. Note that the integrated membrane shear
capacity is proportional to the projected length of the weld, rather than the full arc length; using
the latter would involve radial loads in at least one of the intersecting thin-wall members (in
addition to what is already being accounted for separately). The directional strengths of these
building blocks are added vectorially as indicated in the figure. The various components are
initially assumed to act at full strength, except those which are extremely flexible and would not
take load until some other part of the connection had failed. Note that there is excess horizontal
capacity, perpendicular to the chord, so that some of the building block components would be
under-utilized in order to achieve equilibrium with axial brace loads.
Similarly, and pursuant to the common requirement that the centroid of elements of a
connection coincide with the axis of the member, elements that are all eccentrically on one side
of the member would not be fully effective. To be rigorous, moment equilibrium should be
investigated. However, evaluating this is made difficult by the unknown locus of the shell
bearing capacities. These are not really uniformly distributed as one might assume. For axial
branch loads, they tend to peak as we approach the saddle position or the triple-point (brace-
brace-chord) intersection.
A separate evaluation for in-plane moment capacity is shown schematically in Figure
6.3. The method conventionally applied to riveted and welded joints - polar moment of inertia -
232

is not applicable here because the strength elements are directional in effect. Again we apply the
building block approach. The various strength elements are assumed to act at full (or nominal)
capacity, except that some of them will have to be reduced in order to satisfy equilibrium of
forces (sum = 0). The moment is then computed about any convenient center of rotation.
STRENGTH OF CONNECTION FOR
MOMENT (IN-PLANE SHOWN)

WHEN I F = 0
* SUM OF MOMENTS OF ALL
FORCES ABOUT ANY
CONVENIENT CENTER OF
ROTATION

SHEAR AT WELD A
(REDUCED FOR I F = 0)

Fig. 6.3. Calculation of moment

capacity.

For combinations of axial load and bending moment, the plastic interaction is often
more favorable than linear, especially if strength elements are eccentrically deployed to favor the
direction of bending. However, the calculation procedure can be quite time consuming. For
members whose bending stress is a minor part of the total, and whose connection elements are
about as well distributed about the member axis as the material in the member itself, a reasonable
expedient is to design for an equivalent axial load,

= A ( f + f H) (6.1)

In addition to the foregoing checks for individual member ends, the adequacy of the
combined connection for groups of members should be checked as shown earlier in Fig. 2.24 for:

- total longitudinal shear (scratching shear in ref. 2)

- total transverse shear
- total longitudinal (in-plane) moment
- total circumferential (out-of-plane) moment
- transmittal of locally applied forces (e.g. at bearing points, or distributed loads on the
chord which come to the node as beam shear)

6.1.2 Simplified Code Approach

The basis for AWS and API Code provisions for overlapped tubular connections is
described in Reference 3. These have been very much simplified in comparison to the foregoing
discussion. Only two checks remain: (1) transverse capacity at individual member ends, and (2)
scratching shear for groups of web members.
The first check is illustrated in Fig. 6.4, along with the equation used. The principal line
load building blocks considered are bearing (punching shear) on the chord, and membrane shear
in the overlap weld. The latter is often limited by the membrane shear capacity of the adjacent
branch member base metal. Note, in comparison with Figure 6.2, that AWS neglects bearing on
the adjacent branch, and that the use of gussets is not contemplated.
 233

P Sin9 = ( T . i j ) + 2 ( v w . t w . i 2 )
ALLOW- PUNCHING SHEAR MEMBRANE SHEAR
ABLE ON MAIN MEMBER G> OVERLAP WELD

WHERE

ALLOWABLE PUNCHING SHEAR STRESS

EQUATION FOR THE MAIN MEMBER
= MAIN MEMBER WALL THICKNESS

CIRCUMFERENTIAL LENGTH FOR THAT

ii-
PORTION OF THE BRACE WHICH CONTACTS
THE MAIN MEMBER
AND
v., = ALLOWABLE SHEAR STRESS FOR THE
COMMON WELD BETWEEN THE BRACES
THROAT THICKNESS FOR THE COMMON
WELD BETWEEN BRACES
THE PROJECTED CHORD LENGTH (ONE Fig. 6.4. AWS simplified limit
SIDE) OF THE OVERLAPPING WELD, analysis - transverse load check.
MEASURED IN THE PLANE OF THE BRACES
AND PERPENDICULAR TO THE MAIN MEMBER

For convenience, the author's approximate equations for partial footprint and and
projected overlap shear length 2 I^> are given below for connections with equal branch
diameters d^.

Z1 = d b Ka ( - ) (6.2)

I = Iqapl +R { i - C Os [ a r c s i n ( 0 ) ] } (6.3)
2 c o t ( 0 Q L) + c o t ( 0 T H)

where overlap is expressed as negative gap, the thetas are defined in Fig. 6.1, other terms have
been previously defined, and

= arc cos (1 + J / T J ) (6.4)

where zeta is gap/D), eta is footprint length/D. These lengths can also be scaled from layout
drawings of the connection, a step which is always recommended.
A check for longitudinal shear at individual member ends is not mandatory in AWS.
Also, the specified combined check uses the full footprint length, rather than just the projected
length which would come from consideration of membrane shear only. The latter method would
become unnecessarily conservative in the case of axially loaded diagonal braces with full
footprints; the more liberal AWS methodology assumes the chord is adequate to handle the radial
loads involved. As a result, this check rarely governs; nevertheless, the designer should remain
alert for cases where it potentially becomes inadequate, e.g. when a diagonal brace is the
overlapping member, as shown in Fig. 6.2.
234

The Code also gives some advice on detailing practice. It recommends making the
thicker, more heavily-loaded branch member the through member, with its full circumference
welded to the chord; usually this is a diagonal. It also recommends that the overlap weld be
designed for at least 50% of the required transverse load. Because the overlap is a much stiffer
load path than the radially flexible chord shell, it will try to carry much of the load at elastic load
levels; here, a weak stiff element may lack the ductility required to avoid failure before the rest
of the connection catches up. Although this latter rule is frequently violated in the case of
incidental overlap in connections designed to be concentric, we seem to get away with it because
the chord has been designed for the full load, and is relatively stiff; however, a minimum overlap
of 6t or 3-in is recommended for this case.
The author considers fully overlapped connections (in which the perpendicular brace is
framed entirely onto the diagonal, never reaching the chord at all) to be poor practice, not
adequately covered by the foregoing Code checks. The connection between the braces would
need to be checked for punching shear as a simple joint, losing the benefit of load transfer by
membrane shear. The short length of diagonal between this joint and the chord carries a
combination of reduced axial load and massive beam shear, for which it is less efficient than with
the full axial load alone. At the intersection with the chord, the elimination of radial loads means
that projected length should be used to check longitudinal shear, rather than the full footprint
length. Nevertheless, Wardenier reports that fully overlapped joints with equal braces have given
good joint behavior.
It is unfortunate that design solutions which might be tempting from the standpoint of
easier fabrication, e.g. making the perpendicular braces the through braces or using fully
overlapped connections, result in potentially less efficient connections for which the Code is not
always conservative.
It is also unfortunate that the UEG guidelines (ref. 5), to which one might turn for more
definitive guidance, appears to confuse overlap arc length (their L^) with projected chord length
(2 L^, eqn. 4.3), and gives a formula which depends on the order of intersection, in a way
whose logic is difficult to fathom.

6.1.3 Comparison with Data

A comparison between computed capacities and the tests of Bouwkamp (ref. 4) can be
found in Fig. 6.5. The computed capacities combine ultimate punching shear (1.8 times
allowable) and yield membrane shear (1.67 times allowable), and are compared with the nominal
brace yield capacity to get connection efficiency. Two sets of computed results are shown: the
first for the original Code (AWS-72) and the second for the current version (AWS-84). Test
results are shown as connection efficiency relative to brace ultimate capacity. Either version of
the Code follows the efficiency trend of the tests. The absolute test results, load in the tension
diagonal, show an additional margin of safety. However, two tests do not make a very definitive
data base.
Billington, Tebbett, and Lalani (ref. 5) have more recently collected the available
worldwide data on overlapped tubular connections. They found a total of 86 load tests, of which
the 16 most relevant (failure in the connection and not elsewhere; full data reported, e.g. actual
F^; large scale, at least 6-in or 150-mm diameter) are plotted in Figure 6.6. Also plotted is the
trend of the AWS criteria, for the specific beta, gamma, and tau of the Bouwkamp tests. For gap
K-connections, we see a gradual increase in punching shear strength as the gap gets smaller. For
the overlap connections, a different load transfer mechanism comes into play (membrane shear in
the common weld), and the increase in strength is more dramatic. Bouwkamp's data were found
to be misplotted in Reference 5, requiring correction as indicated by the arrows. The rate of
strength increases with overlap depends on gamma and tau, as indicated by the dashed lines,
qualitatively consistent with the trend of f(7,g') in IIW criteria. Since tau was not reported for
 235

NEGATIVE ECCENTRICITY ZERO ECCENTRICITY POSITIVE ECCENTRICITY

SHEAR ON 9 "
SHEAR ON 6" SHEAR ON 2.5" VERT. WELD
OVERLAP WELD OVERLAP WELD BEARING ON LEG

COMPARISON O F JOINT EFFICIENCIES

CALCULATED TEST RESULTS
TYPE OF BASED ON 3 6 KSI BASED ON
JOINT . YIELD U L T IKM A T E
137 IN 6 ^ 8 255 IN 6 %
POSITIVE
ECCENTRICITY 41 % - 4 0 % 54%

ZERO
62% - 6 9 % 82%
ECCENTRICITY
NEGATIVE
8 6 % - 101% 108%
ECCENTRICITY
4
AWS-72< NAWS-84

Fig. 6.5. Comparison of AWS code rules with Bouwkamp's test results.

UEG d a t a base

A AWS-84 r u l e s 0=.52 7=25 r =.75

Fig. 6.6. Effect of gap parameter zeta on the strength (relative to that of or Y connection) for
uniplanar gap K-connections and overlapping connections. Data points from UEG ref. 5;
curve represents AWS criteria.
236

for the other data, direct 1:1 comparison with individual test points should not be taken too
literally. Nevertheless, it can be observed that the trend of the data follows AWS.
Other authorities, e.g., Kurobane (ref. 6) and IIW (ref. 7), have derived empirical design
criteria for overlapped connections which are continuous with the behavior of gap K-
connections. Although this approach does not explicitly consider any change in load carrying
mechanism, it can in principle still be used as long as the data base adequately covers the
parameter space of the applications, e.g. plastic design sections (d/t < 50). Capacities of most of
the connections in the full international data base are governed by plastification of the chord
walls. Shear failure along the overlap weld is observed only after the chord wall and brace end
have also deformed plastically. A significant failure mode, especially in overlapping
connections, is local buckling of the compression brace, in the area adjacent to the joint; as
discussed in Section 2.2.3, this leads to a more restrictive d/t limit of 37 (also see ref. 8).

6.2 MULTI-PLANAR CONNECTIONS

For design purposes, tubular connections are often classified according to their
configuration for example , , , X and N-connections and the other "alphabet" joints (so-
called because letters of the alphabet are used to evoke their configuration). Different strength
design and fatigue SCF (stress concentration factor) formulas are applied for each different type.
The research, testing, and analysis leading to these increasingly sophisticated criteria have for the
most part dealt only with connections having their members in a single plane.
However, many tubular space frames, including most of those used in offshore
structures, have bracing in multiple planes. Figure 6.7 shows a connection with 16 branch
members in the three orthogonal planes, as well as a suggested binary code for classifying the

ALPHABET JOINTS SPARSE K-JOINTS MUDLINE FULL K-JTS. HUB JOINTS

Fig. 6.7. Classification of

KT X DT
0 1 1 1 1 1 0 1 1 I 1 1 1 0 0 0 0 0
0
0
1
1
1
1
1
0
1
1
1
1 multi-planar connections.
0 0 0 0 0
s o0 o o o o o o
DIAG-BOT-N 0 0 0 0 0
E 0 0 0 0 0 0 0 0 0 0 0 0
W 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0
0 1 0 0 1
1 0 0 1 0 I
1 1 1 1 1
0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 I 1
NE 0 0 0 0
NW 0 0 0 0
SW 0 0 0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 1 0
 237

65,535 other possibilities by the presence (1) or absence (0) of a member in each position.
Examples are given in the figure for how the binary code would apply to the "alphabet" joints, as
well as to some common multi-planar connections of interest. For some loading conditions,
these different planes interact, and when they do, the sophisticated criteria developed for the
alphabet joints are no longer satisfactory. When one considers the number of possibilities, it
seems unlikely that parametric design formulas will ever be developed to cover each and every
one. Hence the attraction of approximate but general criteria which capture the major features of
these multi-planar interactions, even if they must be less precise than criteria for the much-
studied alphabet joints.

6.2.1 Ovalizing Parameter (Alpha)

Figure 6.8 presents the AWS formula for computing the ovalizing parameter (alpha), in
a way which recognizes that loading pattern, rather than just geometrical configuration, is
important to the behavior of tubular connections. Alpha is evaluated separately for each brace
for which punching shear is to be checked (the "reference brace"), and for each load case, with
the summation being taken over all braces present at the node for each load case. In the
summation, the circumferential cosine term and the axial exponential decay term express the
influence of braces on ovalizing stress at the reference position. This simple, but repetitive
calculation is suitable for a tubular connection design post-processor for a computerized
structural analysis of the space frame, and avoids the necessity of arbitrarily assigning one of the
alphabet classifications.

The influence function mimics the pattern of stress remote from "other" loaded braces,
as seen in theoretical shell analysis, Fig. 3.3(a). The additional 1.0 up front reflects the higher
stresses which occur in the vicinity of the loaded brace, which are of interest when this is the
"reference brace" being investigated.
238

(i) Alpha Punching Shear. The ovalizing parameter (alpha) was originally used in
connection with the cyclic punching shear fatigue provisions of the code. Reference 9 discusses
this in connection with allowable hot spot stress as used for Gulf of Mexico offshore structures.
Figure 6.9(a) and (b) show design S-N curves for and joints, respectively, along
with test data plotted on the basis of cyclic punching shear. Recent data from the UK-EEC
research effort have been added to the original data base, and we see results from large scale, low
stress, high cycle tests encroaching on the design curve. Note that punching shear does not bring
the two types of joints onto a common basis; the design curves are different by a factor of two.
Also, the cyclic punching shear criteria do not reflect the important influence of chord diameter-
to-thickness ratio (D/t), and should be adjusted downward for gamma ratios exceeding 25
(outside the range of test data).

ORIGINAL DATA BASE

* RECENT UK - EEC TESTS

Fig. 6.9. Fatigue design curves for cyclic punching shear, (a) T-connections. (b) K-connections.

The punching shear curve for joints is lower than hot spot design curve X by a factor
of 7, implying that hot spot stresses are typically 7 times the punching shear. For and Y joints,
the tolerance for punching shear is half that of joints for axial load.
Although the acting punching shear computed according to the rules for strength design
reflects the overall statics and geometry of the connection, a better reflection of the localized
cyclic stresses causing fatigue damage is given by the AWS expression for cyclic punching
shear.

cyclic V = rsin0 Off- + f (6.5)

CL <3 by>

where fa is axial nominal stress range, f^y represents in-plane bending, and f^z represents out-of-
plane bending. The term a is used to merge former design S-N curves and into the single
design curve K. In the 1980 AWS code, a was assigned a value of 1.0 for balanced joints and
2.0 for , Y, and cross joints. The 2/3 factor for in-plane bending is consistent with the original
Kellogg method, as discussed earlier in Section 3.3 herein. The larger factor for out-of-plane
bending reflects its more severe influence on local stresses (ref. 10).
(ii) Application to Planar Connections. Using the expression in Fig. 6.8 gives the
following results for the classical "alphabet" types of planar connections:
 239

TYPE OF JOINT VALUE OF COMPUTED ALPHA

balanced joint 1.0 to 1.4 (depends on footprint spacing)

T&Yjoints 1.7
cross joints 2.0 to 2.4 (spacing depends on angle)

These values are consistent with the values used for strength design (Table 3.10 herein, Table
10.2 in the Code, old 10.5.1). Intermediate load patterns on these same connection geometries
produce intermediate values of alpha. For , Y and cross joints, values in the range of 1.7 to 2.4
replace the old blanket value of 2.0.
(iii) Elastic Stresses in M u l t i - P l a n a r C o n n e c t i o n s . Figure 6.10 shows how the
Kellogg formula (Eq. 3.7), together with effective punching shear based on computed alpha,
reasonably well predicts the hot spot stress in a variety of multi-planar connections which have
been analyzed by more accurate means. Similar success has been achieved in a number of
similar design-specific comparisons (e.g., ref. 11).
chord - 5*.250 1 KSI 1 KSI 1 KSI

BRACES - 3*.125 /f 4

ZERO

SHEDLEY
/ u M , / ,
EXPERIMENTAL

BOUWKAMP

FINENESS 3 4 5

MARSHALL

COflPUTED ALPHA 3.5 6

KUANG

SCF FORMULA - 5.5

CHORD 2 0 * . 5 V
BRACES 10.750.25
- IS'
ZERO ECCENTRICITY
. . .
K-JOINT CORNER "
y*\ A CO^ER
. A
CAULKINS
- %* - go* - 180Fig. 6.10. Comparisons of hot
2 19
2.35 3.55 IAS
FRAHETI
spot stress in multi-planar
MARSHALL
COflPUTED ALPHA 2.9 3.3 2.5
2,5 connections.
KUANG SCF
FQRflULA K-l 2.87

CORNER

MIDSIDE
V = sine

WAVE

6.2.2 Ultimate Strength

The histogram of Fig. 3.54 showed how well AWS punching shear criteria, actually
based on computed alpha, match the ultimate capacity measured in tests of the Rodabaugh
240

(1980) data base. The compression test results cluster tightly on the safe side of the nominal
safety factor of 1.8. Reasons for not taking the higher tension results too literally have been
previously discussed (Sec. 3.7.4). These tests are all uni-planar connections.
(i) Evolution of Decay Term. The exponential decay term in the ovalizing influence
function went through several stages of evolution. It was originally modelled after the shell
theory solution for the axisymmetric load case of Fig. 2.10(a), which was in terms of the shell
parameter Z. Only the envelope function was retained, as the sine-cosine terms were not "well
behaved" in terms of having a predictable effect. Comparisons with shell theory, e.g. Fig.
2.13(a), and hand-held experiments with simple cylinders, e.g., rolled paper tubes and beer cans,
suggested the following: (1) Interplay between ovalizing (circumferential bending) and
membrane shell behavior is different from what is depicted by the axisymmetric solution (which
involves longitudinal bending and hoop stress, but no membrane shear). (2) The influence of
ovalizing extended over an order of magnitude greater distance. (3) The relative importance of
membrane shell action, versus ordinary ring bending, increased as the gamma ratio (R/T)
increased.
These observations led to the introduction of gamma in the denominator of the
exponent, and an empirical coefficient 1.6 to fit the available elastic stress data, as shown in Fig.
6.10. This version of the decay term could be equivalently stated as:

L/D
e x
exp
(6.6)
P " 1.67 4 VT/25

However, in comparisons with ultimate strength data on uni-planar gap K-connections,

it was noted that the foregoing decay term extended the beneficial effect of balanced loads over
unreasonably large distances, as shown in Fig. 6.11. It was also known that footprint spacing (L)
was inferior to gap as a correlation parameter for uni-planar connections (ref. 12).

4.0

O NAKAJIMA
AWS
ALL OTHERS
exponential
decay / 1.6,
= .3
- l
y 3 0

V
AWS
gap
formula y= io

0.0
0.0 1.0 2.0 3.0 4.0
g/d (gap/diameter)

Fig. 6.11. Gap K-connection data revisited.

 241

Thus, AWS made two last-minute fixes in adopting their criteria. First, a formulation in
terms of gap was applied to uni-planar balanced K-connections. Second, for the more general
case of intermediate load patterns and multi-planar connections, the 1.6 coefficient was reduced
to 0.6, so that the decay term became:

-2 & (6.7)
exp - 0. 6 T or exp - ^

A comparison of the resulting values of alpha is tabulated below (0=0.5, 7=25):

UNI-PLANAR OLD NEW

RULES DECAY DECAY

45 K 1.42 1.15 1.34

45 Y 1.70 1.70 1.70
45 X 2.40 2.25 2.06

We see that a fix that goes towards the safe side for K-connections may have quite the opposite
effect for X-connections. Clearly, the decay term deserves further study.
Part of the difficulty for planar K-connections can be traced to the slavish adoption of
thickness-squared strength formulations in the Yura-based American criteriawhereas, in the
author's original criteria and current IIW criteria, strength varies as the 1.7 to 1.8 power of
thickness. This was discussed earlier (section 3.7.5) and gives rise to a residual effect of gamma,
which increases the scatter in test-vs-code comparisons.
(ii) Japanese Data. Shortly after the the AWS criteria were formulated, a series of 20
ultimate strength tests were conducted at Kumamoto University, Japan (ref. 13). The
configuration has K-connections in two planes 60 apart, as would be found in the single bottom
chord of an inverted delta truss. The loading pattern is similar to comer "B" in Fig. 6.10.
Specimen geometry, dimensions, and tests results are given in Table 6.1.
AWS allowable capacity for these connections was computed with the BASIC computer
program shown in Table 6.2. As some of the connections are offset and eccentric, brace footprint
spacing (longitudinal L and circumferential phi) was computed from the corresponding gap data.
Results are given in Table 6.1, and are plotted as a histogram in Fig. 6.12.
The test results are clustered on the conservative side of the AWS nominal safety factor
of 1.8. Indeed, the correlation is tighter than for the uni-planar connections of Fig. 3.54. Some
of this conservatism comes from invoking the rule that effective F V Q should not exceed two-
thirds the tensile strength (footnote 2 of AWS Table 10.2).
Somewhat less satisfying are the comparisons of each multi-planar connection versus its
uni-planar counterpart. Makino, et al, observed an unusual failure mode on the transverse gap
region, as shown in Fig. 6.13(a). They found the strength of multi-planar connections to average
92.5% of the calculated strength (according to previous correlations) of the corresponding uni-
planar connections. The circumferential influence term in the AWS criteria, on the other hand,
predicts that ovalizing is suppressed by this pattern of loading, so that alpha approaches its
minimum value of 1.0, especially for a large transverse gap which places the brace footprints
about 90 apart. As the longitudinal and transverse gaps tended to increase together in these
experiments, and uni-planar K-connections are penalized by larger values of alpha for larger
gaps, the apparent benefit of suppressing the ovalizing increases with gap, contrary to the trend
of Makino's observations, as shown in Fig. 6.13(b).
 zvz

SPECIMEN CONFIGURATION A N D
GEOMETRICAL VARIABLES (ref.13)
xsHi wNvid-inniM ^ V9 aiavx

TEST RESULTS
E X P E R I M E NT
D d t g ( mm AWS A L L O W. C A P A C I TY
F yo F u lt PT E ST
S P E C I M EN M U LT I U N I - PT E ST PM U L TI
N O. ( m m) ( d e g r e e s) I n al u r ed i n al u r ed ( M P a) ( k N) P L A N AR P L A N AR
PA WS PU N I

D K- 1 2 17 1 4 . 41 48 9 3 2 60 60 60 5 4 .0 6 3 .3 3 52 4 72 8 2 .9 38 9 30 0 2 13 1 . 30
D K- 2 2 17 2 4 . 41 60 7 4 0 60 60 60 4 0 .9 5 1 .6 3 52 4 72 1 0 7 .9 45 1 37 2 2 39 1 . 21
D K- 3 2 17 1 4 41 76 6 4 0 60 60 60 2 3 .0 3 5 .1 3 52 4 72 1 4 9 .1 53 3 50 0 2 79 1 . 07
D K- 4 2 17 2 4 . 42 48 5 3 2 60 60 60 3 .6 4 .5 6 .3 1 1 .1 4 32 5 56 1 4 9 .1 44 1 47 0 3 38 0 . 94
D K- 5 2 17 0 4 45 60 1 4 0 60 60 60 7 0 .6 7 5 .0 8 6 .5 8 6 .4 4 32 5 56 1 1 3 .8 56 7 40 5 2 01 1 . 40
D K- 6 1 65 3 4 39 48 9 3 2 60 60 60 2 8 .4 3 6 .9 3 85 4 90 1 3 6 .3 48 0 41 1 2 84 1 . 17
D K- 7 1 65 0 4 29 48 3 3 2 60 60 60 3 .3 3 .5 7 .4 9 .3 2 78 4 02 1 2 6 .5 35 5 38 9 3 56 0 . 91
D K- 8 1 65 0 4 21 48 5 3 2 60 60 60 6 1 .6 6 2 .2 7 6 .2 7 6 .5 2 78 4 02 7 1 .0 39 0 27 2 1 82 1 . 43
D K- 9 1 65 2 4 21 60 2 4 0 60 60 60 3 8 .0 4 1 .5 5 2 .6 5 6 .4 2 78 4 02 9 7 .0 45 4 34 4 2 14 1 . 32
D K - 10 1 39 8 4 37 48 4 3 2 60 60 60 1 5 .7 1 6 .2 2 3 .5 2 4 .2 3 86 4 75 1 3 6 .3 51 3 48 0 2 66 1 . 07
D K - 11 2 16 2 4 48 60 7 3 8 90 41 60 7 8 .5 7 9 .4 1 0 2 .8 1 0 3 .9 4 72 5 21 1 0 5 .9 48 0 33 6 2 20 1 . 43
D K - 12 2 16 2 4 54 76 5 4 0 90 41 60 4 5 .8 4 7 .2 7 1 .4 7 2 .2 4 72 5 21 1 3 9 .3 57 7 45 0 2 41 1 . 28
D K - 13 2 16 0 4 49 76 6 4 0 90 41 60 5 9 .4 6 0 .0 8 8 .3 9 0 .0 4 72 5 21 1 3 1 .4 57 4 41 7 2 29 1 . 38
D K - 14 1 65 4 4 32 60 7 3 8 90 45 60 2 0 .9 2 1 .6 5 2 .4 5 2 .9 4 09 4 83 1 3 9 .3 50 0 42 7 2 79 1 . 17
D K - 15 1 65 4 4 42 60 6 3 8 90 45 60 3 0 .0 3 0 .3 6 5 .4 6 6 .3 4 09 4 83 1 2 5 .5 52 9 42 4 2 37 1 . 25
D K - 16 1 65 4 4 32 76 4 4 0 90 45 60 1 0 .8 1 1 .6 4 9 .1 5 0 .4 4 09 4 83 1 9 8 .1 60 8 55 9 3 26 1 . 08
D K - 17 1 65 3 4 41 76 3 4 0 90 45 60 1 1 .6 1 2 .0 5 0 .4 5 1 .1 4 09 4 83 1 7 9 .5 63 3 58 2 2 83 1 . 08
D K - 18 1 39 9 4 12 60 7 3 8 90 45 60 6 .9 7 .4 3 5 .2 3 6 .2 3 71 4 69 1 6 2 .8 50 2 47 9 3 28 1 . 05
D K - 19 HO 5 4 08 60 6 3 8 90 45 60 1 1 .4 1 2 .1 41 . 8 4 2 .2 3 71 4 69 1 6 2 .8 49 7 45 0 3 28 1 . 10
D K - 20 1 40 1 4 05 60 6 3 8 90 45 60 1 4 .6 1 5 .2 4 6 .6 4 6 .5 3 71 4 69 1 5 2 .0 49 1 43 3 3 09 1 . 13
 243

1 'AWS C A P A C I T Y FDR KUROEANE M U L T I P L A N A R JOINTS TABLE 6.2 COMPUTER PROGRAM FOR MULTI-
2 I N P U T ' H CHT * > HCr HT
4 HC=HC/57.3 PLANAR DK CONNECTIONS.
6 HT=HT/57.3
10 I N P U T *DDTT"JDD-TT LANGUAGE IS TANDY TRS-80/
12 INPUT - D " ; D
14 INPUT
20
* G T G ' G ? T G
L=0.5xD*(l/SIN(HC)+l/3IN(HT))+G
MICROSOFT BASIC.
22 P H I = ( D - T G > / ( 0.5XDD)
23 DPHI=PHI*57.3
24 PRINT ' L r P H I * LDPHI
30 INPUT 'FU'JFU
32 GM=0.5xDD/TT
34 B=D/DD
3 6 I F B > 0 . 6 T H E N QB = 0 . 3 / ( E x ( 1 - 0 . B 3 3 * B ) > ELSE QB=
40 FY=FU/1.5
2 TP=2*PHI
44 RG=-(L/DD>/<0.3x3GR<GM))
50 SUM=(l+COS<TP)>*<l-EXP(RG)>
52 A=l+-0.7xSUM
54 I F A < 1 , 0 THEN A = 1 . 0
RUN ' E X A M P L E D K - 4
60 FDR I = 1 TO 2
HC H T ? 60,60
62 PRINT 'ALPHA';A
DD-TT? 217,4.42
70 RG=C.7*(A-1)
D? 4 8 . 5
72 QQ=(1.7/A--0.18/E.)xQBLRG
G tTG? 4 . 5 1
74 BU=FY/(0.6xGM>
L rPHI 60.5C 31.4226
76 =&*
FU? 5 5 6
80 PC=3.14xD*TTxVP/SIN(HC>
ALPHA 1.17439
82 PC=PC/1000
COMPRESSION ERACE ALLOW. CAPY.
84 P R I N T "COMPRESSION BRACE ALLOW. C A P Y . ' PC
ALPHA 1.06495
87 ' REPEAT AS UNIPLANAR
COMPRESSION BRACE ALLOW. CAPY.
38 A=l+0.7xG/D
DDfTT? .
8? I F A > 1 . 7 THEN A = 1 . 7
90 NEXT I
1 0 0 GOTO 1 0

MEAN SF = 2.68
EXCL. MAT'L.
C0V = .18
VARIABILITY
NORMAL = 3 5
L0GN0RMAL = 5 6

X Fig. 6.12. Comparison of AWS

X X X X
m u l t i - p l a n a r criteria with
X X |X X
X X X X
Kumamoto test results.

P /P
TEST AWS

AWS

0.6 -
I 1 I I I I I I 1 I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
TRANSVERSE GAP / DIAMETER

Fig. 6.13. (a) Typical failure mode at compression braces, (b) Ultimate strength of multi-planar joint
compared with that of planar K-joint.
244

(iii) Inelastic Finite Element. Paul (ref. 14) describes the inelastic finite element
analysis of a multi-planar "hub" connection (0000000011110000, Fig. 6.7). The finite element
technique and mesh density, were similar to that described earlier (section 2.3.2(iii)), except that
the steel had UTS/YS ratio of 1.62 (reflecting specification values for mild steel). For
calibration, two uni-planar, double T-connections were also analyzed, with the results agreeing
closely with Kurobane's empirical best fit.
Ultimate loads correspond to plastic collapse of the chord, with local strains of about
15%. These are compared to the AWS lower bound (1.8 static allowable) in Fig. 6.14. As
compared to design rules which ignore multi-planar effects, AWS criteria more correctly reflect
both the adverse effects of increased ovalizing (a = 3.8) and the beneficial effect of suppressed
ovalizing (a = 1.0), as well as the greater importance of load pattern as opposed to connection
geometry (compare uni-planar DT and hub at alpha of 2.4).

AWS LOWER BOUND AWS LOWER BOUND FEM__ #

Fig. 6.14. Interaction plots of ultimate strength for multi-planar hub connection, comparing AWS
lower bound (1.8 allowable) with finite element method (FEM). (a) = 0.4. (b) = 0.6.
(ref. 14)

Although the match is not perfect, it is about as good as one gets for other classes of
connections. Finite element ultimate loads range from 1.55 to 3.37 times the AWS allowable,
which is comparable to the scatter shown in earlier correlations. For the case where ovalizing is
suppressed (a = 1.0), as for the Japanese double-K experiments, we see a strong effect of the
transverse gap, which is not reflected in AWS criteria based on angular footprint spacing.
The failure mechanism depends on load pattern as shown in Fig. 6.15. Mode (b) is the
ovalizing treated by AWS; when this is suppressed, mode (a) can occur, as observed on both the
Kumamoto experiment and the Delft finite element work. An attempt was made to adapt a
plastic ring analysis to these failure modes (ref. 14). Unfortunately, this approach continues to
suffer from ambiguity in the choice of effective ring width, as shown in Fig. 6.15(c). The data
points are effective widths required to match up the theoretical plastic ring analysis with the
inelastic finite element results. Although these appear to be consistent with observed stress and
deformation patterns, they do not provide a straightforward basis for extrapolation to other
design situations. Paul (personal communication) reports better results using an empirical
modification of the ring equation, matching the data as indicated by dash-dot lines in the figure,
with B e ff of 1.43D. An attempt to extract effective ring width as a function of transverse gap,
using Kurobane's earlier work with double-T connections, is also shown, yielding B e ^ in the
range of 1.4 to 1.7D in the range of the experiments.
 245

Fig. 6.15. Failure mechanism of a multi-planar hub connection, (a) Equally compression-loaded
braces (a = 1.0). (b) Unloaded or tension-loaded out-of-plane braces (a = 2.4 or 3.8). (c)
Effective ring width required for theoretical failure mechanism to match experimental
capacity.

Ring solutions also do not shed much light on problems involving the influence of
longitudinal footprint spacing, i.e., connections other than hub connections.
Clearly, the ultimate strength behavior of a wide variety of multi-planar connections
deserves further study. Cases of practical interest have been identified in Fig. 6.7. Meanwhile,
despite the present shortcomings, the real value of the AWS alpha criteria lies in preventing gross
undersign for the adverse loading patterns not yet covered by testing, such as those shown in Fig.
6.16. Computed alpha also continues to be useful as a means of automating the classification of
connection type based on load pattern.

Fig. 6.16. Adverse load patterns with a up to 3.8. (a) False leg termination, (b) Skirt pile bracing, (c)
Hub connection.
246

6.3 GROUTED CONNECTIONS

Among the advantages off grouting the pile-to-jacket-leg annulus in offshore structures,
is the strengthening and stiffening of tubular joints which results. For the smaller tubes used in
architectural applications, filling the entire tube is not unreasonable, and produces similar
benefits. Although there have been a number of projects dealing with grouted tubular joints (ref.
15-18), definitive design criteria have not been promulgated in the American design Codes.

6.3.1 Cognac Studies

In connection with the design of Shell's 1020-foot water depth platform for the Cognac
prospect, additional analysis and testing of grouted connections were carried out. For a critical
and repetitive complex multi-planar connection detail occurring at grouted skirt piles in the base,
a 1:4 scale model of that specific geometry was tested at South-West Research Institute (SWRI,
ref. 19). However, little beyond verification of that particular design was gained. Earlier tests of
Cook Inlet grouted joint designs (refs. 20, 21) suffer from a similar lack of generality.
For the Cognac jacket mid- and top-sections, however, there were to be large tubular
inserts grouted inside the legs to tie the sections together, resulting in a large number of grouted
K-connections. These were the subject of a more generic study. Rather than model a specific
Cognac K-connection, it was decided to study the same K-connection geometry as used in an
earlier project (ref. 22), and about which considerable information had already been developed.
This geometry is shown in Fig. 6.17. The study consisted of finite element analysis by PMB
Engineering (ref. 23), experimental stress analysis by SWRI (ref. 24), a fatigue test (ref. 25), and
the author's interpretive review (ref. 26).

20$.500
CHORD

7
18 . 5 0 0
PILE

Fig. 6.17. Grouted K-joint geometry.

A key focus of the study was the local scale stress as influenced by joint geometry and
the presence of grout. That these local "hot spot" stress concentration factors (SCF) provide a
common basis for treating fatigue in many different welded hardware configurations is taken for
granted. This basic approach to fatigue was introduced in Section 2.4.1 and further discussed in
Chapter 4.

6.3.2 Baseline: Behavior of Ungrouted K-Connections

Our understanding of grouted K-connections will be built consistently upon what is
already known for ungrouted connections, which is now briefly reviewed.
 247

(i) Empirical SCF Equations. One of the oldest design equations for tubular joints is
the Kellogg formula. The writer still considers it useful for estimating hot spot stress
concentration factors (SCF) in K-connections. As presented in Reference 10 and herein referred
to as alpha Kellogg criteria,

e ff n
(SCF) = 1 . 8 r s i n eJy (6.8)
chord * fn

(SCF) = 1.0 + 0 . 6 Q 1 + J ^ ~ (SCF) (6.9)

branch r [ 1 chord

where, analogous to AWS alpha cyclic punching shear (eqn. 6.5), the effective branch member
nominal stress is taken as

9 2 r> 2l 1 / 2
f = a f + (- f ) + (- f ) (6.10)
ef f a 3 by 2 bz J

and

0.5tc + t b
6 1 1
Q r = exp - *

Alpha was originally taken as 1.0 for K-connections, and other terms are as defined using AWS
nomenclature. In Equation 6.11, Q r is a correction for the difference between thin shell mid-
plane intersection and the actual weld toe location in the branch member, with t^ taken as an
estimate of weld fillet size.
SCF results for the ungrouted K-joint used in this study, from Alpha Kellogg criteria*,
are given in Table 6.3. Heel and toe locations are defined in Fig. 6.17, while top and bottom
refer to the model position as tested. The format shown is based on the needs of Shell's detailed
fatigue analysis program (refs. 27, 28), which does a separate cumulative damage calculation for
four locations around the weld at each end of each brace investigated. Different SCF's are
obtained for axial load, in-plane bending, and out-of-plane bending. The larger of branch side or
chord side factors are used.

TABLE 6.3 MATRIX FOR SCF'S - TABLE 6.4 MATRIX FOR SCF'S -
ALPHA KELLOGG CRITERIA - UNGROUTED GRESTE -
UNGROUTED AXIAL/FRAMETI MOMENTS
LOAD
F M F H
x z x 2

(12:00 2.8 1.9 toe

(12:00 2.5 1.1 toe
SIDE
SIDE

3:00 2.8 4.2 top 3:00 2.3 3.4 top

k 6:00 1.5 heel

1
i 6:00 2.8 1.9 heel 1.4
1
0

i, 9:00 2.8 4.2 bottoe \ 9:00 2.3 3.4 bottoa

3 f12:00 2.5 2.2 toe (12:00 2.0 M/A toe

H SID

j 3:00 2.5 3.1 top I 3:00 1.6 N/A top

OF 1
BRANCI

I 6:00 heel i 6:00 1.0 N/A heel

| 2.5 2.2

bottoa
{ 9:00 2.5 3.1 bottoe
_ [ 9:00 1.6 N/A
 248

(ii) Finite Element Analysis. SCF results for the Clough/Greste finite element analysis
(ref. 29) are shown in Table 6.4. Note that the SCF for axial load varies with position around the
weld. Table 6.4 also gives bending results from the Frameti analytical shell theory solution (refs.
30, 31), which incompletely models local stresses in the branch number.
A more realistic finite element model has been implemented by PMB (ref. 32). Using
isoparametric, curved, thick shell and solid elements, they obtain comparable accuracy with a
coarser mesh. In addition to modeling the actual thickness of the tubes, these elements permit
straightforward modeling of the weld geometry, which changes going around the intersection.
This permits a direct solution to the Q r effect, yielding hot spot stresses at appropriate points
adjacent to the weld, and reflecting the stiffening effect of the finite weld volume.
The PMB finite element approach also offers a direct extension to grouted joints, as
indicated in Fig. 6.18. The steel-grout-steel sandwich in the chord is modeled with three layers
of finite elements, each with the appropriate material properties.

OUTSIDE "SLEEVE"

GROUT ELEMENTS

INSIDE " P I L E "

Fig. 6.18. (a) Three-dimensional isoparametric, thin shell finite element, (b) Three-dimensional
modelling of grouted connections.

SCF's from PMB are given in Table 6.5. These are comparable to those given earlier in
Tables 6.3 and 6.4, for the Kellogg and Greste methods, respectively, showing the degree of
consistency which exists for ungrouted connections.

TABLE 6.5 MATRIX FOR SCF'S - TABLE 6.6 MATRIX FOR SCF'S -
PMB - UNGROUTED MEASURED - UNGROUTED
AVERAGE VALUES
LOAD
r

/12:00 2.4 2.0

r 12:00 2.8 toe
3:00 2.4 4.6
UJ

CO U J
3:00 2.3 top
6:00 1.1 0.8 heel

6:00 1.3 heel \ 9:00 2.4 4.6 bottom

9:00 2.8 bottom

12:00 1.7 1.2 toe

3:00 1.7 3.6 top

12:00 1.3 toe
6:00 0.9 1.2 heel
<
CO
UJ
I 3:00 2.0 top
9:00 1.7 3.6 bottom
C_5
z u_
*C 6:00 1.3 heel
CQ

9:00 2.0 bottom hot spot s t r a i n range

nominal stress range
 249

(iii) Experimental Stress Analysis. The earlier comparison of Figure 2.17 shows plots
of longitudinal and circumferential chord surface stresses for sections of the ungrouted K-joint.
Curves are from the two finite element analyses while the large data points are from the SWRI
experimental stress analysis. Circular points are from the brace A topside sections indicated,
while triangles are from the corresponding bottomside locations. Theoretically, these should be
identical, and the significant differences indicate an unintended eccentricity in the applied axial
load or small differences in strain gage placement in areas of steep stress gradient. The square
data points are from comparable locations at the opposite brace B; due to the asymmetric chord
support (one end only), these would be expected to be nearly the same as at brace A, but not
exactly.
Since test facilities were limited to 108 channels of data, uniaxial strain gages were
applied at most locations. These were oriented to pick up the major principal stress direction
indicated by prior analyses. The plotted points correspond to a = e. The small arrows on
some of the data points indicate the appropriate stress correction for biaxial effects (Eq. 4.3) for
those locations where multi-axial strain gages were applied.
Figure 2.17 shows good agreement between measured stresses and both finite element
analyses, except for the effects of unintended eccentricity in the test load.
Values from geometrically similar locations for brace A/brace and for top/bottom
should be approximately the same. These have been averaged to yield the results given in Table
6.6, filtering out the effects of unintended eccentricity in the applied load, or differences in strain
gage placement. These compare well with the analytical results given previously.

6.3.3 Grouted K-Connections

(i) E m p i r i c a l P r o c e d u r e s . The author's standing approach to grouted tubular
connections (ref. 33) may be stated as follows:

(1) Since for tensile loads the punching shear will be determined by the outer jacket
leg joint can or sleeve, this thickness should continue to be used in calculating
the acting Vp for static strength, and for r in the SCF equations.

(2) The effect of grout in strengthening the joint may be reflected by using an
increased effective thickness, T e ff in computing y for both allowable punching
shear and SCF equations. Strength (section modulus) rather than stiffness
(moment of inertia) is the appropriate parameter for dealing with stresses rather
than deflections. It is unfortunate that the literature is full of instances where the
latter has been mistakenly used.

Figure 6.19 gives two methods for arriving at T eff. The original method treats the steel-
grout-steel sandwich as a composite section, and computes an equivalent single thickness t e
which has the same local bending strength. However, early comparisons with test data indicated
this was too optimistic. To add a touch of conservatism and account for partial disbonding of the
grout, t is taken as the geometric average of t e and the original joint can thickness.
Method 2 (ref. 34) assumes essentially complete disbonding of the grout, so that it
functions only as a spacer between sleeve and pile. Local bending is split between sleeve and
pile, which act independently except for having the same curvature.
For the grouted K-joint being studied, Method 1 and Method 2 predict local chord
SCF's that are 79% and 84%, respectively, of the ungrouted case. Lesser reductions apply to
brace stresses. Specific results for Method 2 are given in Table 6.7.
250

Fig. 6.19. Two methods for computing effective chord thickness of grouted connections.

TABLE 6.7 MATRIX FOR SCF'S - TABLE 6.8 MATRIX FOR SCF'S -
GROUTED K-JOINT - GROUTED K-JOINT -
ALPHA KELLOGG & METHOD 2 PMB RED & GREEN CASE
- LOAD -
AXIAL
LOAD 1

toe FULL 0 to 0 to
2.H 1.6 REVERSAL TENSION COMPRESSION

2 3.5 t op
f12:00 1.8 2.1 1.5
CHORD SIDE
1posr0-

2.1 1.6 heel 1.3 1.7 0.8

OF V IELD

| 3:00

2.4 3.5 bottom I 6:00 0.3 0.1 0.1 heel

2.3 2.0 toe ^ 9:00 1.3 1.7 0.8

2.3 2.8 top

2.3 2.0 heel

2.3 2.8 bottom

(ii) Finite Element Analysis. The three-dimensional finite element mesh used by PMB
for the grouted connection is shown in Fig. 6.20. Note the three-layer sandwich used to model
the grouted chord (sleeve) and pile (pin). The steel was modeled as a linear elastic material for
all cases.

Fig. 6.20. Finite element model of K-joint with grout showing intial disbonding for red and green case.
 251

For the first grouted joint analysis, the grout was modeled as linear material, but having
different elastic constants than the steel. Peak local stresses in the chord were reduced to about
one-third those of the ungrouted case. However, grout bond stresses were unreasonably high,
1200 psi in tension and 1600 psi in shear.
Thus, subsequent analyses employed non-linear modeling of the grout bond, with a
failure envelope as given in Fig. 6.21. After disbonding, there is zero tensile capacity and zero
shear capacity. Compressive contact forces can still be transmitted, but any corresponding
frictional shear forces were not modeled. The transition is essentially brittle and instantaneous.

SHEAR S L I P P A GE

Fig. 6.21. Grouted tubular joint

criteria for normal and shear
failure.

The results obtained from this "base case non-linear" analysis show roughly a 30%
increase in local chord stresses over the linear grout case. After 10 iterations, the results seemed
to be converging on some meaningful answer. However, the rate of convergence had become
painfully slow, and the norm of out-of-balance force remained significant (60% of the external
force norm). Anomalous results were also indicated. Scattered elements under the tension brace
carried compressive contact forces. Also, under the compression brace we found bond shear
stresses which, though consistent with the stated failure criterion, appeared higher than one
should count on under reversed cyclic loading.
To help things along, we then specified large areas of initial disbonding as indicated in
Fig. 6.20. These were marked on a mesh layout in colored pencil ~ hence the name "red and
green case". The solution converged in three iterations, and yielded peak local stresses
intermediate between the "base case non-linear" solution and the ungrouted joint.
Since there are compressive contact forces under the compression brace, the grout more
effectively reduces the local hot spot stresses there than under tensile loading, as indicated by the
SCF values of Table 6.8. Intermediate SCF's apply for full reversal of load.
(iii) Comparison with Experimental Stress Analysis. Measured stress ( = e) on
the chord under tensile brace loading are plotted in Fig. 6.22, along with the following PMB
analytical results:

(1) linear grout

(2) non-linear grout "base case"
(3) non-linear grout "red and green case"
(4) ungrouted
252

Fig. 6.22. Effect of grout on

chord stress results near tension
branch, (a) Axial stress, chord
surface, (b) Circumferential
stress, chord surface.

30
- A . BOTTOM

The strain gage layout was selected so as to be diagnostic of which of the analytical results was
correct. The "red and green" case, as shown by the heavy line, appears to be most consistent
with the measured stresses.
To permit a direct comparison of analysis and experiment, free of unaccounted-for
biaxial effects, chord surface strains as would be read by the uniaxial gages were computed for
the "red and green" case. These are plotted in Fig. 6.23 along with the experimental strain data
points.

Fig. 6.23. Strain results for red-

green case, (a) Axial strain,
chord surface, (b)
Circumferential strain, chord
surface.

- 0 . 5 0 -J
 253

Stress patterns under the compressive brace are shown in Fig. 6.24. Both analysis and
experiment show lower stresses than at the tension brace. Again, the test data support the "red
and green" analytical results.

GROUTED-NONLINEAR BASE CASE'

Fig. 6.24. Effect of grout on

c h o r d s t r e s s r e s u l t s near
compression branch, (a) Axial
stress, chord surface, (b)
Circumferential stress, chord
surface.

Experimental SCF's for axial load are tabulated in Table 6.9. Note that these are after
100 cycles ~ i.e., stable hysteris loops were achieved, and progressive disbonding was given its
chance to occur. For compressive loading, the local chord stresses are significantly lower than
for tensile loading. Separate results are given for brace A and brace B; under the same type of
load, these should be similar but not necessarily the same. For full reversal, SCF results for
braces A and B , as well as for top and bottom, are averaged to filter out the influence of
unintended eccentricities in the applied axial loads.

TABLE 6.9 MATRIX FOR SCF'S - TABLE 6.10 MATRIX FOR SCF'S -
GROUTED K-JOINT - GROUTED K-JOINT -
EXPERIMENTAL CYCLE 100 EXPERIMENTAL AFTER
100 CYCLES

FULL 0 to 0 to
REVERSAL TENSION COMPRESSION

/12.00 2.1 2.27.8 1.5-7.0 to.

S o 11 : 00 1.5 3.1/1.4 2.0/0.8 top

~ s J 6 : 00
I ] 0.2 0.0/0.5 0.2/D.2 he*1

V 9:00 1.5 l.t/1.8 0.V0.9 botto.

/12:00 1.5 l.S/1.6 1./1.3 to.

3 0 1.0 1.V1.6 1.0/0.9 top

l i J
6 : 00 1.0 0.6/1.1 0.8/1.5 heel
1)
\ 9:00 1.0 0.9/0.6 -/0.5 bottc-

AVERAGE A/B A/B

hot spot s t r a i n range

hot spot s t r a i n range nominal s t r e s s range
nominal s t r e s s range
254

The SCF's in Table 6.10 include unit in-plane and out-of-plane bending load cases as
well as axial. In bending, chord hot spot stresses are significantly lower on the compression side,
while branch hot spot results are mixed.
These experimental SCF's may be compared with S C F ' s given by the empirical
procedure (Table 6.7) and the finite element results of Table 6.8. The empirical procedures fall
on the safe side of the experimental data for full reversal loading. For tensile axial load, some of
the experimental chord stresses are slightly higher. Brace stresses are conservatively predicted,
even when Q r is included. The "red and green" finite element results follow experimental trends
closely, but are not always conservative.
For full reversal, comparing averaged experimental SCF's of the grouted K-joints with
those of the ungrouted (Table 6.6), peak chord stresses are 47 to 88% as high, depending on type
of load. Brace stresses are 42 to 137% as high. The biggest improvement is where the ungrouted
SCF's were highest; i.e., for out-of-plane bending. For compressive loadings, the grouted chord
stresses range from 43 to 75% of ungrouted.

6.3.4 Fatigue
Figure 6.25 shows fatigue tests of grouted and ungrouted K-joints plotted in terms of
cyclic punching shear. The SWRI test done for Cognac is included. Comparing the two trends
of data, it appears that grouted joints would merit a 40% increase in allowable cyclic shear stress,
while maintaining a comparable safety factor on fatigue life.

18

16 -

14 -

12 AWS
CURVE
10

OPEN
BERKELEY TESTS
UNIV. OF TEXAS
SWRI TESTS

CRACK INITIATION
2
1 10 ID 103 104
NUMBER OF CYCLES
Fig. 6.25. Fatigue curve in terms of punching shear, for grouted and ungrouted K-connections.

However, when grouted and ungrouted fatigue test results are plotted on the basis of hot
spot stress, Figure 6.26, the two scatter bands of data are indistinguishable. This verifies our
initial assumption that hot spot stress concentration factors provide a common basis, and simply
leaves us with the task of estimating them. Either the empirical approach (Method 2) or the "red
and green" type of finite element analysis, or experimental stress analysis appear to be
satisfactory; indicating fatigue life increases by factors of 2.5 to 7. Placing a limit of T e ff <
e ,, o r s o n c
^sieeve ^ ^eff ^ ^ * ^ specimens is consistent with the observed 40% increase
in fatigue strength.
 255

100,000]

Fig. 6.26. Fatigue curve in

t e r m s of hot spot s t r e s s .
Grouted joints from Fig. 6.25;
AWS
X-MODIFIED others from early hot spot
FATIGUE CURVE database, Chapter 4.

GROUTED TUBULAR J O I N T S

UNGROUTED WELDED SPECIMENS

1 I 1 I t r
H
10' 10 10 10

CYCLES TO FAILURE

6.3.5 Static Strength

An early static test of a grouted full-size tubular joint is reported in Reference 15. As
shown in Fig. 6.27, initial cracking and reserve strength (in terms of punching shear) were
similar to other tension joints, up to the point where an end fixture failure terminated the test.
When the effect of K a is omitted, as it would be with today's Code, the grouted connection
would plot well above the other data.

TENSILE

Fig. 6.27. Static punching shear

test results, comparing early
grouted and ungrouted tests.

0 10 20 30 40

7
= R/T
CHORD THINNESS RATIO

If the lower measured chord hot spot stresses translate into a corresponding effect on
ultimate strength, the grouted joint would be 140% to 150% as strong as the ungrouted under
compressive axial loading, and almost twice as strong for out-of-plane bending. In-plane
bending would be virtually unchanged.
256

Applying these results in the context of the present AWS rules, suggested modifications
for grouted joints are as shown in Table 6.11. For the study K-joint, the allowable punching
shear would increase by a conservative 40%. For a compressive cross joint of comparable
dimensions, the doubling of strength is consistent with a closed ring analysis in which the sleeve
and pile deflect together.

TABLE 6.11 MODIFIED AWS PUNCHING SHEAR RULES FOR GROUTED CONNECTIONS
7 R
AXIAL e f f " ' T e ff
K-CONNECTIONS
unchanged
can
T
where
e f f " If T
can A X I A L , Y , AND X
CONNECTIONS
(e.g., for solid f i l l )
R T
eff / eff
T E N S I O N OR C Y C L I C LOAD

R T
COMPRESSION ONLY * eff * / eff
T = 1 T
eff *b eff

BENDING

IN-PLANE same as u n g r o u t e d
R
OUT-OF-PLANE T e f f * / T e ff

T T
- \ ' can

Although Tebbett and Billington have tested over 70 grouted tubular connections, much
of this data remains proprietary and unpublished. Reference 5 does, however, present sufficient
details from 11 tests, that they can be compared with the suggested application of AWS rules.
These are tabulated, along with the earlier test, in Table 6.12. Histograms of the comparison
between test and criteria are shown in Figure 6.28. The comparison shows a great deal of scatter,
more so than for other types of tubular connections. While this may be due to the brittleness and
variability of grout, there also appear to be trends in the data not reflected in the criteriafor
example, the increasing benefit of solid fill with increasing gamma ratio, and a stronger
beneficial effect of increasing beta ratios. Nevertheless, by taking a large mean safety factor (1.8
1.74), we achieve reasonable safety index values, 0 S , for known loads equal to the allowable.
When sleeve material bias and variability are included, the lognormal safety index becomes 3.6.

TABLE 6.12 TEST DATA FOR GROUTED CONNECTIONS

c
P sing / l Fy
SOURCE & TYPE 7 1.8 AWS TEST

UEG T - j o i n t s
axial sol i d 1.7 .382 20 21.2 37
axial t p = l . 5 t c 1.7 .382 20 21.2 36
axial tp = 0.6tc 1.7 .382 20 11.6 22
axial - solid 1.7 .331 20 19.2 29
axial - solid 1.7 .331 32 19.2 39
axial - solid 1.7 .331 43 19.2 43
axial - solid 1.7 .550 20 27.6 62
bending - s o l i d 1.5 .331 20 12.8 9
bending - s o l i d 1.5 .331 32 12.8 14
bending - s o l i d 1.5 .331 43 12.3 19
bending - s o l i d 1.5 .600 32 17.3 47

U. o f Illinois
multi-planar 1.5 = .50 27 20.1 23 crack
f u l l scale axial
39".75 chord = 35 40 failu

.75 p i l e
Fy = 3 6 a s s u m e d
 257
MEAN 1 . 7 4
NORMAL
.56
EXCL MATL VAR
0c 2.1

A
'[F| J B M A A AAA A A B | j

LOGNORMAL MEDIAN 1.64

EXCL MATL VAR COV .32
0s 3.4

Fig. 6.28 Reliability of grouted

fil A XBI n ^ T W T ^ A connection criteria.
.5 1.0 2.0 5.0

P T E /
S 1T . 8 P A W
S

The data base does not justify using an improved y c ^ for y less than 20, and its trend
suggests that doing so would be unconservative.
Also, since grouting improves the chord resistance to radial loads, load components
contributing to the longitudinal chord stress assume greater importance. This issue is discussed
in Section 6.4.2(H).

6.4 INTERNALLY STIFFENED TUBULAR CONNECTIONS

6.4.1 Introduction
For tubular space frames in offshore structures, the most popular style of node is the so-
called "simple" tubular connection, in which a large through member (the chord or main
member) has all the incoming branch members welded directly to it. Usually, there is a short
thickened section of the chord, or "joint can", provided to take care of load transfer through the
node, and to limit the localized "hot spot" stresses. Details and practices have evolved to permit
satisfactory welding to be done entirely from the outside of the tubes.
The use of internally stiffened tubular joints arises when this elegant "simple joint"
solution no longer works. Once we cross this threshold, the architecture of the entire structure
can be affected its appearance, its contraction sequence, its economy, and so on. Also, since
the variety of possible stiffening arrangements is limitless, practical design procedures have not
been codified to the same extent.
(i) Advantages and Disadvantages. Stiffened joints permit efficient connections to
large diameter, thin wall tubular members, such as the caisson legs of a semi-submersible drilling
unit, or the pontoon legs of a self-floating jacket.
In other applications, they permit a reduction in joint-can thickness, as compared to
unstiffened simple joints; this becomes important when the forming limits of fabricators become
a constraint.
Stiffening can be designed with clearly identifiable load paths, so that the designer is not
totally dependent on finite element analysis or empirical formulas, with their potential for latent
errors and mis-application. Being able to conceptualize also helps the formulation of design
strategies, as opposed to a trial-and-error approach (guess something and see if it gets past the
computer).
A stiffened joint can be a challenging structure in itself. A design approach which
consists of cutting sections and taking free bodies may be loosely justified by the lower bound
theorem of plasticity, provided material selection and detailing are such that yielding can occur
without premature failure by local overstraining, brittle fracture, or local buckling. Success of
258

the method depends on the perceptiveness and thoroughness with which each part of the
connection is examined (see Section 2.3.2).
It follows that stiffened joints require more engineering attention than simple joints.
Also, stiffened joints usually end up being more conservatively designed.
Among the disadvantages are higher unit cost, more complex fabrication, and higher
restraint during welding (possible need for lamellar tearing resistant Z-steel and stress relief).
Finally, internal stiffening may conflict with other functions of the main member, for example
pile driving through jacket legs.
(ii) Examples. We shall begin with some examples of stiffened connections for very
large diameter main members (ref. 35). All of the joints shown were designed to develop the full
axial capacity of the incoming branch members. In most cases, full bending capacity was also
developed. This approach is mandatory for earthquakes, and eliminates the need to check myriad
load cases for other applications.
Figure 6.29 shows a cutaway view of a node designed by the author in 1972, and used in
the North Sea Brent "A" Platform. Load transfer is mainly in shear, from brace to bulkhead
(which functions as a large gusset plate), with the vertical component going from bulkhead to leg
directly, and the horizontal component going via the diaphragms. A thickened stub end on the
brace is used to limit stress concentration where the stiffening is first encountered; finite element
analysis indicated a SCF of only 2.2. The stub also penetrates the leg wall, to eliminate lamellar
tearing as a failure mode. Finite element analysis was also used to evaluate the effect of the large
holes (for pile driving) in the diaphragms. All the plates are quite thin, and require extensive
stiffening for local buckling and hydrostatic loading.

HORIZONTAL
DIAPHRAGMS VERTICAL
ft 1.00 BULKHEADS
.625

4 8 . 8 7 5 BRACE

4 8 1.25
THICKENED
STUB ENDS Fig. 6.29. North Sea platform
joint (dimensions in inches).

288 .687
PONTOON/LEG
 259

Fig. 6.30. Two alternatives for large dimeter leg joints, (a) Gusseted, see Fig. 6.31. (b) Flared, see
Fig. 6.32.

Figures 6.30 to 6.32 show two different solutions proposed for another self-floating
structure. The gusseted connection is similar to the Brent design, except that the main vertical
gusset penetrates the leg wall instead of the brace, and a "wing" gusset must be provided for out-
of-plane bending. The gussets are tapered to prevent "hard spot" stress concentration where the
brace first encounters them. The flared connection transfers most of its load in direct membrane
stress. Although its geometry looks complicated, it is made up entirely of flat, cylindrical, and
conical surfaces, which can be laid out with elementary descriptive geometry, and cold-formed
from flat plates.
Figures 6.33 and 6.34 show leg connections from a proposed Gulf of Alaska platform.
These were being studied for fabrication in Japan, up until a series of disappointing dry holes
were drilled. Negative eccentricity and overlapping nodes were used to reduce the amount of
load being transferred in and out of the leg, as well as to reduce the length of joint reinforcement.
Internal rings, and a grillage, provide resistance to general collapse. Although the heavy brace
stubs and chord wall insert are shown as weldments, castings would also appear to be applicable
(ref. 36).
Some connections have no readily identifiable main member, such as the wye and cross
connections shown in Fig. 6.35. These are similar to the previously described flared connection,
in concept and construction except that for asymmetrical push-pull loading, the crotch of the
wye ends up with punching shear, since direct membrane stress alone no longer provides for
complete equilibrium. Other configurations which might also be considered are hemisphere for
the wye, and full sphere for the cross.
(iii) Type Considered Herein. Figure 6.36 shows a 50,000-ton, 1350-ft. (400m)
Bullwinkle jacket, which was constructed for Green Canyon Block 65 in the Gulf of Mexico.
Kawasaki Steel and IHI prefabricated members and joint cans, for assembly by Bullwinkle
Constructors in Texas. The jacket was launched in one piece from an 850-ft. barge, in 1988.
Launch leg joints carry heavy loads during this operation. The joints identified as "Big MACS"
are critical for in-place loadings, exceeding 24,000 kips in the comer legs, which must all be
transmitted to skirt piles via knee braces at this point. Early design calculations indicated joint
 260

Fig. 631. Gusseted joint details. Fig. 6.32. Flared joint details.

LU
261

Fig. 6.33. Tapered leg joint. Fig. 6.34. Large diameter leg joint, with internal grillage.
262

Fig. 6.35. Wye and cross connections.

Fig. 6.36. Bullwinkle platform at launch.

can thicknesses of nearly 7-in. would be required for simple unstiffened joints. Since this
exceeded the cold forming capacity of most fabricators, an alternative stiffened design was
investigated.
Typical stiffened joint designs of the type adopted for this project are shown in Figs.
6.37 and 6.38. Simple flat plate rings and diaphragms, at relatively wide spacing, were used.
This permitted the immediate reduction of joint can thickness to 5-in. maximum, as shown.
Subsequent design refinements reduced this further, to 4-in. maximum. This is heavy enough to
avoid "hard spots" where stiffener and brace footprints cross, and satisfies the design criteria
hereinafter described.
 263

Fig. 6.37. "Big MACS" joint.

6.4.2 Static Strength

The most basic requirement for satisfactory behavior of a tubular structure is the static
strength of its connections. This requires a practical design procedure, as described below.
(i) Punching Shear in Shell. As opposed to the connections described in Figs. 6.29 to
6.35, the ring-stiffened connections in Figs. 6.37 and 6.38 receive incoming brace loads via the
punching action of line loads acting on the outer shell, rather than by direct membrane stress or
in-plane shear. Punching shear has long been used to describe the capacity of the joint can in
simple unstiffened joints. Although it is an oversimplified representation of actual stresses in the
shell, we shall now extend this concept to stiffened joints.
Radial line Q may be calculated as

Q = fc f s i n9 ( T v
b n = p (6.12)

where t^ is branch thickness, is shell thickness, is angle between branch and main member,
and fn is nominal stress in branch. The summation () implies that double shear (inside and
outside) may be considered where appropriate.
264

54+2.375

(a)

(b)

(c) (d)
Fig. 6.38. Launch leg joints, (a) At intermediate member, (b) At tipping panel point, (c) At typical
launch diaphragm/ring detail, (d) At fail end of structure.
 265

Fig. 6.39 shows the estimated ultimate punching shear (Vp) capacity as a function of
ring stiffener spacing. This is a composite of several solutions, and should be considered
approximate. For very close stiffener spacing, the capacity is derived from yield line solutions
for flat plates, with higher capacity being shown for loads crossing the stiffeners than for loads
offset from the stiffeners. For stiffener spacings between 0.5 and 1.5 diameters, capacity is based
on the AWS Code rules for local punching shear (single shear), updated to the current edition.
For still larger stiffener spacings, the capacity for extensive line loads decreases due to general
collapse, with the trend shown being based on a heuristic analogy with the collapse behavior of
pressure vessels.

Fig. 6.39. Punching shear capacity of stiffened cylinders.

Stating the capacity as punching shear permits the total capacity of a joint to be
computed as the sum of its parts, using a "building block" approach, as suggested in Figure 6.40.
Each part has a different capacity, as governed by the parameter shown in parentheses.
Normally, in a tubular connection, the plug inside the brace footprint takes no load, and the joint
acts in single shear. However, when stiffener and brace footprints cross at frequent intervals,
opportunities for double shear arise. These depend somewhat on assumptions made by the
analyst.
Using the foregoing, total connection capacities have been derived for branch members
having the worst case diameter ratio (beta = 0.6), with the results shown in Figure 6.41. This is
provided as a check for those who prefer the total load format to punching shear. For very wide
stiffener spacing, as for unstiffened joints, the ovalizing parameter (alpha) becomes important.
For intermediate stiffener spacings, the effect is to reduce the ovalizing parameter alpha from that
of unstiffened double-cross joints (3.8), cross joints (2.4), and wye joints (1.7), towards the value
of unity which applies when ovalizing is suppressed. This benefit is most dramatic for shells
with high D/T.
266

Fig. 6.40. Punching shear at brace footprint as sum of its parts.

' .01 .02 .05 .1 .2 .5 1.0

sine / i t d Fy
AT U L T I M A T E

Fig. 6.41. Capacity of stiffened tubular joint.

For very close stiffener spacing, capacity depends on the matching of brace and stiffener
footprints. For the thinner shells, finite element studies typically show load transfer to be
concentrated where stiffener and brace footprints cross. Here, an altenative approach is to add up
the capacities of all the crossings, with the capacity of each "knife edge" crossing as described
earlier in Fig. 2.23.
 267

Fig. 6.42(a) gives a similar unambiguous derived capacity plot for line loads parallel to
the axis of the cylindrical shell (for example, a launch cradle). Figure 6.42(b) gives comparable
results from finite element analysis (ref. 3 7 ) . With the exception of the beam bending limit,
which is usually accounted for separately in design, these show the heuristic analogy drawn
earlier for large L/D to be on the safe side.

Q/TFy Q/TFy
ULTIMATE AT FIRST YIELD

Fig. 6.42. Line load on stiffened shell, (a) Empirical, (b) Finite element (ref. 37).

Nevertheless, there is clearly a need for further research on the radial line load capacity
of stiffened shells, using inelastic finite element analysis and load tests.
(ii) M e m b r a n e Loads in Shell. Because ring stiffeners increase the ovalizing and
punching capacity of the cylindrical shell, but do nothing for its capacity to carry axial membrane
stresses, the latter can assume increased importance. In addition to chord axial loads already
present, the branch contributes an additional axial stress, locally in the chord shell, of at least
(t/T) * cos(0) * fn. For the "Big MACS" joint, the contribution from the congested knee braces
is even higher, because the load does not have a chance to spread out, but stays confined to a
strip not much wider than the brace.
Interaction between axial membrane stress and punching in the shell should be
considered, using the worst case reduction factor, Qf, (see Section 3.7.1(ii)), with being
between 0.044 and I / 7 (the latter being the theoretical value for axial membrane vs. shell
bending interaction at the toe). The membrane stress contribution must also be included in hot
spot stresses for fatigue.
(iii) Demand/Capacity in Ring/Diaphragm. In stiffened connections, we must design
the stiffeners, as well as the shell. In the ring stiffened connections being considered here, the
rings serve primarily to prevent general collapse of the chord. There are several possible
approaches to this task.
Diaphragms, including those with access holes, are often sized to match the thickness
(or t^ sin0) of the heaviest incoming braces. They may conservatively be designed on the basis
of direct stress on the net section, as if they transmit radial loads without any contribution from
the more flexible shell. Another constraint to be checked includes minimum thickness to avoid
268

local buckling, keeping diameter/thickness less than 400/VFy~or 56, for 50 ksi (350 MPa) steel.
Local crippling usually is not a problem, if the radial capacities of stiffener and brace are
comparable, and the intervening shell meets the criteria described in 6.4.2(i) above.
Rings are designed as a composite tee section, with a plastic effective flange width in
the shell of 1.4V(DT), for the combination of moment, thrust, and shear which result from the
radial loads imposed on their part of the joint. Shear often dominates. Unbalanced loads are
transferred into membrane (beam) shear in the cylindrical shell. Elastically, this loading is
represented by Roark's case 25 (solution given in Table 3.1 herein), and combinations can be
built up using superposition. Plastically, some redistribution of moment can be allowed, as well
as ultimate strength interaction between moment and thrust. Safety factors similar to those used
in member design would be applicable. Height/thickness (h/t) ratios of 13 to 16 are appropriate
for steels up to 50-ksi, with the lower limit to be used when plastic behavior is being relied upon
(ref. 38).
An alternative approach to ring design is to provide an "equivalent gamma" section to
replace that of the required unstiffened shell in empirical design equations. Since we are dealing
here with strength, equivalency based on plastic section modulus seems appropriate. However, it
should be recognized that we will be working the material well beyond yield, up to the ultimate
strength, and limit h/t to values appropriate for extreme deformation (7 to 13 for steels up to 50-
ksi). Although there are references in the literature (repeated in ref. 5) to equivalency on the
basis of moment of inertia, this is not a measure of strength, and therefore potentially unsafe.

6.4.3 Stress Concentration

Stress concentration factors (SCF) are indispensible for fatigue analysis of stiffened
joints, as their performance is generally better than AWS category -prime, which applies in the
absence of known SCF. The results of elastic finite element stress analysis are also useful in
understanding the general behavior of tubular joints.
(i) Methods of Analysis. Fig. 6.43 shows the finite element mesh for one-eighth of a
stiffened double-cross joint. Geometry and severity of loading are similar to launch truss joints.
The author prefers finite thickness isoparametric shell elements for this analysis, as they give a
better representation of weld toe hot spot stresses than the mid-plane intersection of thin shell
analysis. Also thin shell analysis has the potential for exaggerating the hard spot "knife edge"
singularities where brace and stiffeners cross.

Fig. 6.43. PMBSHELL - undeformed structure plot of one-eighth double-K joint.

 269

Figure 6.44 shows a developed view of the mesh for one-half of the joint can of the "Big
MACS" joint. Footprints of 5 braces and 4 stiffeners can be seen. Chord, stiffeners, and braces
were all modeled with finite thickness shell elements. It took an engineer three months to
generate the mesh. The combined model had 61,000 degrees of freedom, and was solved in 44
hours on a VAX-785 computer. After reducing the wall thickness to 4-in., the worst knee brace
provided an SCF of only 2.83 in the chord. The analysis paid for itself by saving nearly 100-tons
of steel in the four replications of this connection (refs. 39, 40).

$$g&$m>to
nni inHa-ll-HI-'WIiii n n n n
to
: a i % 3

RING
FOOTPRINTS
KNEE BRACE
FOOTPRINTS

DETAIL
"D"

Fig. 6.44. Developed mesh

view of one-half "Big MACS"
joint can.
-'^V/^Sjii!53f jCNULI LI LI LI LI
Diz70DQDCMi:iLI CI
raaaizri ciizii
(ii) P a r a m e t r i c Formula. In Reference 4 1 , the author suggests using the following
formula to estimate the SCF due to radial loads on the chord of stiffened tubular connections:

SCF = 3/EFF (6.13)

where EFF is the connection efficiency, static ultimate strength divided by yield capacity of the
branch member. Here it is assumed that the designer will avoid abrupt discontinuities and
grossly mis-matched stiffness between load sharing elements.
270

Figure 6.45 shows how well this simple formula predicts the SCF of 21 ring stiffened
joints in various references. The data falls mostly with 20%, which is roughly the scatter band
of the other SCF formulae in use. This prediction errs on the safe side for joints with beta
(diameter ratio) of unity, for which alternative expressions are available (ref. 9).

DATA SOURCES:
AUK Fig. 6.45. Parametric SCF
SHIYEKAR
+ 80UWKAMP
V ECC-UKOSRP
formula for stiffened tubular
OTC 4109 MAGNUS
joints.

I 2
JOINT EFFICIENCY

The data base for this plot (refs. 42-45) is given in Table 6.13. Several items are
noteworthy. The weak link in several of the connections was at the toe of the brace, where the
line load was parallel to and offset from the stiffeners, rather than crossing. Shiyekar's joints
with one large ring have much higher SCF than similar joints with multiple rings intersecting
each brace footprint. Many of the Magnus joints had very closely spaced small rings, and were
governed by equivalent gamma of the composite section, rather than by the shell between
stiffeners; thus, they did not take full advantage of the 4-in. thick shell provided.
In addition to the contribution of radial loads to hot spot stress, we must also be
concerned with the contribution of longitudinal membrane stresses in the shell, as already
discussed. Also for brace ends, a lower bound SCF of 1.8 often governs (ref. 9).

REFERENCES

1 Marshall, P. W., Some Design Considerations for Welded Tubular Structural Joints, Shell Oil Co.,
November 1964.
2 Nakamura, K., et al, Structural Characteristics of Overlapped Nodes, Proc. BOSS-85, Delft,
Elsevier, July 1985.
3 Marshall, P. W. and Toprac, . ., Basis for Tubular Joint Design, Welding Research Supplement,
May 1974.
4 Bouwkamp, J. G., Research on Tubular Connections in Structural Work, WRC Bulletin 71A,
August 1961.
5 Design of Tubular Joints for Offshore Structures, UEG, London, 1985 (3 volumes).
 271

TABLE 6.13 RING STIFFENED JOINTS

VP / F y
JOINT
d/D t/T D/T L/D* ANGLE SCF EFFICIENCY

Auk-X (ref .7) 1.0 .75 32 .88 90 3.3 2 @ .18 .48

Auk-X (alt.) 1.0 .50 21 .88 90 1.8 2 @ .25 1.00

Auk-K 2-way stif. .90 .60 52 .17* 67 3.9 2 @ .18 .65

PMBSHELL d o u b l e X .5 .5 40 .5 45 2 @ .32 1.28

Shiyekar #1 .47 .71 41 one r i n g 90 2.5

.47 .71 41 .24 90 2.0 20.57 1.61

S h i y e k a r //2 .61 .82 41 one r i n g 90 2.8

.61 .82 41 .32 90 1,9 2 @ .56 1.36

S h i y e k a r 03 .77 .82 41 one r i n g 90 2.8

.77 .82 41 .39 90 2.3 2 @ .39 .95

S h i y e k a r //4 .91 .71 41 one r i n g 90 4.2

.91 .71 41 .46 90 3.7 2 < . 3 2 .90

Bouwkamp .44 .57 38 .40 90/95 2.7 2 @ .45 1.11

external rings .13,.26* est. @ .28+.35 plus d i r e c t shear
2 per brace S- curve transfer into rings

Brandi
with 2 rings .46 1.0 40 .53 60 4.0 2 (3 . 2 8 .65

Brandi
with 3 rings .46 1.0 40 .27 60 2.7 2 @ .57 1.32

Magnus DK if I .75 .8 24 . 12 46/43 2.75 2 @ . 5 7 (B)

23 s m a l l r i n g s 8.33(A) . 12* 1.17 @ .57(A) 1.16

Magnus DK # 1 .75 .8 24 .23 46/43 2.8 2 @ .35(B)*

13 r i n g s 9.2(A) .23* 57(A) .99

Magnus DK # 1 .75 .8 24 .35 46/43 2.8 2 ( . 3 5 ( A )

8 rings 10.5(A) .35* 2.8 2 @ .35(B) .90

Magnus DK # 1 .75 .8 24 .53 46/43 2.6 2 @ .42 1.22

6 diap .35* 2 @ .35*

Magnus h o r i z //2 OPB .53 .38 21.25 .14 90 1.6 2 @ .57(B) 1.89
16 r i n g s @ . 2 7 d 7.02(A) 1 . 2 5 (3 . 5 7 ( A )

Magnus h o r i z //2 OPB .53 .38 21.25 .29 90 2.0 2 @ .57(B) 1.68
8 rings 9.1(A) 1 . 1 3 (3 . 5 7 ( A )

Magnus //3 a x i a l .56 .70 24 .25 90 2.1 saddle 2 @ .57

3 b i g rings @ brace .25* 3.9 crown* 2 @ .35* 1.00

Magnus OPB .56 .70 24 .25 90 2.1 2 @ .57 1.63

Magnus # 4 .53 .50 33 .11 47.7 1.1 2 @ .57

axial .11* 2 0 .50* 2.70

Magnus H o r i z diag .31 .45 33 .11 86.9 1.1 2 @ .50*(B)*

8.2(A) 1.19 @ . 5 7 ( A ) 1.51

Magnus H o r i z OPB .31 .45 33 .11 86.9 1.9 2 @ . 5 7 (B)

8.2(A) 1.19 @ . 5 7 ( A ) 1.51

* Denotes e/R f o r l i n e loads p a r a l l e l to stiffeners

A denotes e f f e c t i v e gamma

based on s h e l l capacity
272

6 Kurobane, Y., et al, Ultimate Resistance of Unstiffened Tubular Joints, ASCE Journal of Structural
Engineering, vol, 110, no. 2, February 1984.
7 IIW s/c XV-E, Design Recommendations for Hollow Section Joints -
Predominantly Statically Loaded, IIW Doc. XV-709-89, Annual Assembly, Helinski, September
1989.
8 Kurobane, Y., et al, Local Buckling of Braces in Tubular K-Joints, in Thin Walled Structures, no.
4,1986, Elsevier Applied Science Publishers.
9 Marshall, P. W. and Luyties, W. H., Allowable Stresses for Fatigue Design, Proc. 3rd International
Conference on Behavior of Off-Shore Structures, BOSS-82, Boston, 1982.
10 Marshall, P. W., A Review of Stress Concentration Factors in Tubular Connections, Shell CE-82
Report, April 1978.
11 Marshall, P. W., Variable Mesh Finite Element Analysis of a Complex Multi-planar Joint, Shell
Head Office Engineering Report, Spring 1980.
12 Zettlemoyer, N., API Committee correspondence, March 22,1982.
13 Makino, Y., Kurobane, Y. and Ochi, K., Ultimate Capacity of Tubular Double K-Joints, Proc. 2nd
International Conference on Welding of Tubular Structures, IIW/Pergamon Press, Boston, July
1984.
14 Paul, J. C , The Static Strength of Tubular Multiplanar Double T-Joints, Shell Report
RKER.88.076/TU Delft Thesis Report/IIW Doc. XV-E-88-139, April 1988; also paper 9,
International Symposium on Tubular Structures, Lappeenranta, September 1989.
15 Stallmeyer, J. E., Static Test of a Full Scale Pipe Joint, report to Calco, June 1959.
16 Bouwkamp, J. G., Cement Grout Filled Tubular Joints Under Alternating Loads, University of
California Structural Engineering Lab. Report No. 68-16, November 1968.
17 Peterson, M. L., "Fatigue Tests of Two Full-Scale K-Joints for Offshore Structures, Conoco
Research Department report 106-3-4-1-73, Nov. 1973.
18 Bouwkamp, J. G., Cyclic Loading of Full Size Tubular Joints, Offshore Technology Conference,
OTC 2605, May 1976.
19 Briggs, M. J. and Burnside, . H., Experimental Stress Analysis of a Grouted Complex Joint,
Southwest Research Institute Project 03-4550, July 1977.
20 Grigory, S. C , Experimental Stress Analysis and Fatigue Test of a Cook Inlet Offshore Platform
Model Joint, SWRI Project 03-2004, February 1970.
21 Grigory, S. C , Experimental Stress Analysis of a Model Joint of Cook Inlet MGS Platform A,
SWRI Project 03-3547, February 1974.
22 Grigory, S. C , A Study to Develop a Design Procedure for Analysis of Plastic Fatigue Life of
Tubular Joints on Offshore Structures, SWRI Project 03-1882, May 1969.
23 Analysis of Grouted K-Joints, PMB Systems Engineering, Inc., September 1976.
24 Briggs, M. J. and Burnside, . H., Experimental Stress Analysis of A Grouted and Ungrouted K-
Joint, SWRI Project 03-4550, August 1977.
25 Briggs, M. J., Grouted K-Joint Fatigue Test, SWRI report in preparation, September 1977.
26 Kinra, R. K. and Marshall, P. W., Fatigue Analysis of the Cognac Platform, Proc. Offshore
Technology Conference, OTC 3378, May 1979.
27 Vugts, J. H. and Kinra, R. K., Probabilistic Fatigue Analysis of Fixed Offshore Structures, Offshore
Technology Conference, OTC 2608, May 1976.
28 Marshall, P. W. and Kinra, R. K., Dynamic and Fatigue Analysis for Deepwater Fixed Platforms,
presented at ASCE-EMD Specialty Conference, Raleigh, North Carolina, May 1977.
29 Greste, O., A Computer Program for the Analysis of Tubular K-Joints, University of California,
Structural Engineering Lab Report 66-19, November 1969.
30 Caulkins, D. W., Parameter Study for Frameti Elastic Stress in Tubular Joints, Shell Oil Company,
CDG, Report 15, September 1968.
31 Dundr, V. and Toprac, . ., Stress Concentration in Joints Subjected to Axial Loads, Bending
Moments and Shears, Report TL-A-01-68, for Shell Development Company, March 1969.
32 Liaw, C. Y., Reimer, R. B. and Litton, R. W., Improved Finite Element for Analysis of Welded
Tubular Joints, Offshore Technology Conference, OTC 2642, May 1976.
33 Marshall, P. W., General Considerations for Tubular Joint Design in Offshore Structures
(unabridged edition), Shell Oil ODC Report 49, August 1973.
 273

34 Reiff, C. ., Program for Joint Design (implemented as SEPOST*77), Shell H.O.C.E.

memorandum, January 16,1976.
35 Marshall, P. W., Design of Internally Stiffened Tubular Joints, paper presented at IIW/AIJ
International Meeting on Safety Criteria in Tubular Structures, Tokyo, July 1986 (long version).
36 Nakamura, K., et al, Static and Fatigue Strength Design of Overlapped Cast Steel Node, OTC 5138,
Houston, 1986.
37 Pacheco, L. N., An Analysis of Ring-Stiffened Cylindrical Shells Using Finite Element Methods,
GEM summer intern report to Shell Development Company, September 1986.
38 Sommers, P. B. and Nyman, D. J., Stress Analysis of Platform Bullwinkle Stiffened Tubular Joint at
Elevation -1190 Ft., report to Shell Oil Company, June 1986.
39 Sommers, P. B. and Marshall, P. W., Stress Analysis of Bullwinkle's Biggest Tubular Joint, Proc.
Offshore Technology Conference, OTC 6093, Houston, May 1989.
40 Sommers, P. B. and Nyman, D. J., Load Test for Internal Stiffener Rings, report to Shell Oil
Compapny, April 1986.
41 McClelland, B., et al, Planning and Design of Fixed Offshore Platforms, van Norstrand Reinhold,
New York, 1986 (three chapters on tubular joints, steel selection, and welding).
42 Wordsworth, A. C , et al, BP Magnus Internally Stiffened Bracing Node Studies, OTC 4109,
Houston, 1981.
43 Shiyekar, M. R., et al, Stresses in Stiffened Tubular T-Joint of an Offshore Structure, paper
submitted to ASME, 1981.
44 Bouwkamp, J. G. and Stephen, R. M., Tubular Joints Under Alternating Loads, University of
California, 1967.
45 Brandi, R., Behaviour of Unstiffened and Stiffened Tubular Joints, Steel in Marine Structures, proc.
EEC Conference, Paris, 1981.
Chapter 7

SPECIAL TOPICS IN FATIGUE AND FRACTURE CONTROL

In the preceding discussion of fatigue design, mention of weld profile and size effects
has been unavoidable. Sometimes this was done implicitly, without giving full information. In
this chapter, we will finally develop the rest of the picture.
Similarly, in conceptual development of the limitations on the strength of welded
tubular connections, materials problems were listed among the possible failure modes. In
addition to fatigue, an integrated approach to structural integrity requires consideration of brittle
fracture and lamellar tearing. Furthermore, inspection tolerances should logically be coordinated
with the performance levels which are assumed or required by the designer. These are also
discussed in the present chapter, as is the use of structural redundancy as a back-up line of
defense.
Finally, in the next chapter, we will complete our discussion of material selection and
weldability issues, in the context of the overall construction system.

7.1 FATIGUE SIZE AND PROFILE EFFECTS

In Chapter 4, the provisions for fatigue design of tubular structures in the AWS Code
(ref. 1) were broadly introduced. Here we will focus more closely on new rules for weld profile
and fatigue size effect. And, although the problem appears for practical purposes to have been
put to rest at the 1987 Delft conference on Steel in Marine Structures (ref. 2), some ongoing
research in this area is discussed.

7.1.1 Introduction
Weld profile effects were considered by writers of the AWS Code even before D l . l was
first issued in 1972. Relative to smoothly ground surfaces (category A), typical butt welds
(category C & X) exhibit a fatigue strength reduction factor (Kf) of 1.7, while fillet welds
(category D) have a more severe Kf of 3.0. This is reflected in the family of S-N curves given by
AWS.
Shell, Conoco, and several other oil companies have, for over 20 years, specified a
version of the "disc test" for weld profile control. Here the as-welded profile is required to be
concave, approximating the radius of a disc whose diameter is not less than the base metal
thickness, or 0.625 in., whichever is larger. The permitted gap relative to this disc is limited to
0.04 in. (1mm). Furthermore, any sharp crack-like notch (i.e., undercut) is limited to 0.01 in.
Beginning with the U.K. Offshore Steels Research Programme (UKOSRP) in the late
1970's, British and European laboratories began to produce fatigue data from fillet welds and
large scale tubular joints with fillet-like flat weld profiles. These welds were not consistent with
the foregoing profile control, although they met the industry standards under which they were
made. This new data fell below the original S-N curve for categories C & X. At that time
(1979), both American codes adopted lower fatigue design curves for welds without profile
controlAPI curve X-prime and AWS curve X-2.
However, API gave only qualitative guidance as to what kind of profile control would
be needed to avoid the performance penalty of the lower fatigue curve, and the original AWS
"dime test" was less stringent than the industry practice cited above. Certainly the 1979 revisions
would not be the last word.
In the years following the Cambridge Offshore Steels Research Seminar (ref. 3), much
of the ongoing research on fatigue of tubular joints was centered in Europe and Japan. The U.K.
 275

effort expanded into one coordinated by the European Coal and Steel Community (ECSC). The
American effort in which the author participated was primarily to digest this research and reduce
it to practice. An interim report on this effort was given in the 1984 Houdremont Lecture (ref. 4),
with the final results appearing in the 1986 AWS Structural Welding Code.
By the time of the Paris Conference on Steel in Marine Structures (ref. 5), continuing
European research had provided massive confirmation of an obvious trend of lower fatigue
strength for thicker specimens of the same geometry. In each of the test series, strict geometric
similarity was maintained, including the use of flat-faced weld profiles with sharp notches at
their toes. While this argueably did not represent the best of design and construction practice, it
did produce compelling and unambiguous experimental results and a challenge for the theoretical
analyst.
Explaining these results has required research-level analysis, beyond the hot spot stress,
and has greatly expanded our conceptual understanding of the problem. The effects of the
thickness on fatigue has been discussed by various authors as follows:

(1) The metallurgical size effect, including: coarser grain structure and lower yield strength;
higher residual stresses and likelihood of pop-ins; increased risk of hydrogen cracking
during fabrication, as well as lower notch toughness. These contributions to size effect
might be expected to "saturate" once the relevant parameters reach their worst case.

(2) The statistical size effect, the increased risk of having a significant initial flaw with
increase in stressed volume. This effect is well recognized in the design of castings,
forgings, and machine parts; however, in weldments without profile control, it seems to
be overwhelmed by the virtual certainty of a severe notch occurring at the weld toe.

(3) The geometrical size effect, which can be addressed by stress analysis. Notch stress
theory has been applied to the initiation phase of fatigue, and fracture mechanics to the
crack growth phase. The transition from plane stress to plane strain, as reflected in
relatively smaller terminal flaw sizes, may also be included in the latter. As elegant as
these theories are, we must remember that they do not address aspects (1) and (2) of the
problem, and thus may not account for the total size effect.

7.1.2 Notch Stress Approach

In going beyond hot spot stress, the local/microscopic perturbations at the toe of the
weld can be examined more closely, using the notch stress theory developed by Peterson,
Neuber, and others. This has been applied to welded joints by Lawrence, et al at the University
of Illinois (ref. 6) as shown in Figure 7.1. In the vicinity of a sharp notch at the toe of the weld,
the severity of the local stress gradient can be characterized by the parameter alpha-prime.
Theoretically, for zero radius, the stress at the notch becomes a singularity (not unlike a crack).
However, the actual reduction in fatigue strength is given by the factor Kf, modified by the
material constant "a" such that the limiting worst case is reached when the radius "r" is equal to
or less than "a", and the equation in Figure 7.1 reduces to Equation 4.4.
For butt welds in 0.5 inch (13mm) material, the weld reinforcement can be as much as
25% of the thickness (idealized weld toe angle up to 45 degrees), and the notch radius can also be
considered a random variable; a Kf of 1.7 is representative of such welds. If we were to scale
these welds up geometrically four fold, keeping the reinforcement at 25% of the thickness (45-
degree toe angle), Kf would increase to around 3 and a very strong size effect would manifest
itself. Fortunately, the AWS Structural Welding Code controls the weld profile so this does not
happen. Reinforcement on butt welds is limited to 0.125 in. (3mm), so that in 2-inch (50mm)
material, only welds represented by the idealized profile with 10-degree toe angle would meet the
276

+ 1 +
7
a = . 0 0 8 " FOR MILD STEEL HAZ

!f = WORST CASE FOR r = a

max

r = NOTCH T I P RADIUS

OLD (THIN) WRC

DATA BASE INCLUDES
TYPICAL RANGE OF
RANDOM a'AND r

AWS CLASS "C" BUTT WELD

1/8" MAX REINFORCEMENT

.25 .50 .75 1.0

(SEVERITY OF WELD PROFILE)

- -H
W
> U

Fig. 7.1. Notch stress theory.

-r-

4
/
to 1

code. With this profile control, a similar range of Kf is obtained and the size effect is effectively
suppressed for butt welds.
For thin test specimens, the differences between butt and fillet welds are small enough
that all the data could appear to fall in a single scatter band. However, as fillet welds are scaled
up geometrically, with no compensating profile control, a strong size effect would again be
expected.
The notch stress results are re-cast in Figure 7.2. The American comprehensive size-
effect format shows a family of curves with progressively lower fatigue performance for
progressively more severe notch conditions at the toe of the weld. However, each curve also
shows a size effect, with notch stress theory indicating log-log slopes of -0.25 to -0.40 for
geometric scale-up of a given geometry. The dashed line indicates again how the size effect
could be mitigated by profile rules in the AWS Code, i.e., the 1/8-inch (3mm) limitation on
reinforcement on butt welds.
The larger the unmeasured notch (perturbation above the reference stress) in a given
design approach, the larger the "size effect" adjustment which must be made. Notch stress
theory, which examines microscopic scale stresses, explains a great deal. Nominal stress or
geometric hot spot stress (sigma-G) require huge corrections.
 277

Fig. 7.2. Size and profile effects as indicated by notch stress theory, (a) American comprehensive
size-effect format, (b) European size-only format.

In Figure 7.2(b), the same results are plotted in European size-only format, widely used
in References 2, 3, and 5. Each geometry has been normalized on itself, at a reference thickness
of one inch. By filtering out everything else, this brings the size effect into sharper focus.
However, a great deal of useful information is lost. This format will not be used further herein.

7.1.3 Fracture Mechanics

Although many different authors have applied fracture mechanics to the problem of
tubular connections, much of what will be presented here is from the APTECH joint industry
project, in which the author was closely involved as project "godfather". Phase one was a
literature review by Hayes (ref. 7).
Gurney (ref. 8) and others have used fracture mechanics to justify a "universal" 1:3
slope for the S-N curve for welded joints, as well as the size effect. Cyclic crack growth data,
da/dN as a function of , is used to integrate from the initial flaw, a Q, to the terminal flaw size
at failure (effectively near infinity for ductile materials). Most of the early studies to date were
on flat 4 5 - d e g r e e fillet w e l d s , with the governing equations as shown in Figure 7.3.
Local/microscopic notch effects in the region adjacent to the weld toe are reflected in a geometry
correction term Y(a) in the equation for fracture mechanics stress intensity factor .
Accelerated crack growth through this region is responsible for the reduced fatigue performance
of welded joints.
For geometrically similar weldments (in which both the weld size and initial flaw size at
the weld toe increase with thickness and the terminal flaw size is very large), fracture mechanics
crack growth analysis indicates the fatigue strength to vary with the (l-m/2)/m power of
thickness, where m is the exponent in the Paris Law, i.e., the log-log slope of da/dN vs. . For
m of 2.5 to 4, the power of thickness becomes -0.1 to -0.25, respectively. Near the fatigue
threshold, as m approaches infinity, the power of thickness approaches -0.5.
In Figure 7.3, the results are presented in terms of the fatigue strength reduction factor,
Kf, relative to the performance of plain plate (or fully ground welds) with an 0.01 inch (0.25mm)
crack-like defect (e.g., permissible undercut). Results are given for welded joints having various
crack sizes and plate thicknesses, with the weld size scaled to thickness. We see that for a given
278

thickness, increasing the initial crack size makes things exponentially worse. For a given intial
flaw size (e.g., 0.01 in. or 0.25mm undercut), increasing the plate thickness makes things
moderately worse, but not as bad as when the initial flaw size also increases (e.g., 1% of
thickness), as shown by the dashed line.

Hi

Fig. 7.3. Fracture mechanics

equations and results for fillet-
welded joint.

It is also of interest to note that, for very small initial crack sizes, the curves seem to be
leveling offsuggesting that in the range of typical weld toe crack sizes found by NTH (ref. 9),
fatigue strength is not unduly sensitive to the assumed value of a Q. This is good news for those
wishing to establish fracture mechanics as a credible tool. However, it may be bad news for
those wishing to implement a fitness-for-purpose approach, in that the data base seems to reflect
initial flaw sizes much smaller than what is typically specified as inspection criteria.
Upon looking deeper, however, one finds that the size effect in welded joints is not
simply a function of plate thickness. Rather, it is related to the size of the unmeasured local
notch effect zone to the toe of the weld, through which the initial spurt of fatigue crack growth
occurs. In terms of characterizing the fatigue strength of welded joints, this local notch effect
zone (circled region in Figure 7.4) represents a perturbation of the stress field, where stresses are
locally above the reference stress (nominal stress for welded coupons, or hot spot stress for
tubular joints).
Using Gurney's BOSS-79 results (ref. 10) in Figure 7.4, when one goes from geometry
A (25mm plate with 10mm fillet and attachment) to geometry C three times as large in all
respects, a 15% reduction in fatigue strength is found. However, when one merely increases the
main plate thickness threefold, but keep the same perturbation and the same size fillet and
attachment (geometry B), then there is virtually no size effect.
This reasoning may be applied heuristically to tubular joints, referring to Figure 7.5.
Geometry A represents an actual weld cross section from one of the early tests in the WRC data
base; here the weld beads are small and the strain gage used to measure hot spot stress is located
 279

Fig. 7.4. Size effect in 45 fillet-welded joints.

close to the toe of the weld, where fatigue cracking occurred (dashed line). Geometry
represents a production weld from a U.S. Gulf Coast yard, where profile control was being
enforced and small stringer beads were used so that the weld merges smoothly with the adjoining
base metal. When Becker, et al (ref. 11) fatigue tested full-size tubular nodes removed from a
platform of this same vintage, measuring hot spot stress with strain gages close to the weld, the
data fell in the same scatter band as the earlier tests. Since the local perturbations above the
reference stress (circled regions) of A and were similar, no size effect would be expected.

Fig. 7.5. Weld profiles in tubular joints

(dimensions in mm), (a) Early
American test, (b) American
production weld with profile
control, (c) Dutch test,
unimproved profile, (d) French
test.
280

Geometry C is a tracing of the weld profile from one of the recent European tests. Here
everything was scaled up: the weld bead size, the notch effect zone, and the strain gage
placement (moved back specifically to avoid measuring any local notch effect). With this huge
perturbation above the reference stress, and some undercut too, a dramatic size effect (reduced
fatigue strength in terms of the measured hot spot stress) seems inevitable. An even more
dramatic size effect would be expected from the 75mm thick French tubular joint, in which they
went out of their way to limit the weld cap reinforcement. The result was a weld profile with a
large and sharp 70-degree notch at the toe.
A more rigorous illustration of size and profile effects may be found in Figure 7.6. If
we apply the present AWS "disc test" (radius = t/2) to the idealized flat-faced fillet weld profiles
for which fracture mechanics solutions are available, then we see in the upper part of the figure
that progressively flatter weld toe angles are required as the thickness (and disc radius) is
increased. However, for the original AWS version, a toe angle of about 60 degrees would be
permitted, regardless of thickness.

U 3
c

WELDS MEETING F i g . 7.6. Fatigue strength

SHELL CLASS "C"
PROFILE REQUIREMENTS
r e d u c t i o n as p r e d i c t e d by
AS SHOWN ABOVE fracture mechanics for fillet
welds with varying toe angles.
30 45" 60"

WELD TOE ANGLE

Results from the fracture mechanics work of Hayes (ref. 7) are also represented in
Figure 7.6 to show the combined effects of thickness and weld profile. Format is similar to
Figure 7.3 presented earlier, except that here the initial flaw size is kept constant at 0.01 in.
(0.25mm), varying toe angle and thickness.
We see that for a given toe angle (e.g., 60 degrees as permitted by the original AWS
dime test) there is indeed a moderate size effect, with thicknesses greater than 1.25 in. (32mm)
falling on the unsafe side of the presumed fatigue strength reduction factor of 1.7. However, for
a given thickness, improving the weld geometry improves the fatigue strength substantially
about a 10% improvement for 45 degrees toe angle instead of 60. Thus, when toe angle is
 281

varied with thickness, as required to meet the modified disc test, the dashed lines indicate that the
size effect can be mitigated by profile control, at least for the idealized conditions illustrated
here.
It is also of interest to note that results for zero degree toe angle, representing butt welds
with their profile ground flush. Here we see no adverse size effect at all; the weld toe
perturbation through which accelerated crack growth occurs has been removed. In fact, the
thicker sections last slighdy longer, as the crack has farther to grow.
These results are recast in the American comprehensive size-effect formatfatigue
strength versus thicknessin Figure 7.7. As such, it can be compared to the notch stress results
in Figure 7.2(a). A great deal of qualitative similarity can be seen. Each geometry (except the
plain plate) has an adverse size effect, even though we have improved the situation over strict
geometric similarity by not scaling up the initial flaw. Flatter angles at the weld toe produce
higher fatigue strength than more abrupt geometries, across the range of thickness variation.
Applying the rule that a 1mm wire should not pass under an 0.5T radius disc at the weld toe
would require progressive improvement of the weld toe angle as thickness increases, mitigating
the geometrical size effect as shown by the dashed line. Of course these fillet welds under
pulsating tension are a long way off from shell bending in tubular connections, and geometrical
size effect may not be the whole story, but the general insight these examples provide is valuable
nonetheless.

F i g . 7.7. Size and profile

effects as indicated by fracture
mechanics.

T H I C K N E S S

Phase Two of the APTECH project developed solutions specifically for the fracture
mechanics analysis of tubular connections (ref. 12). Cyclic stress intensity factor () is
computed as function of crack size (a), the reference cyclic hot spot stress range, and the
geometry correction term Y(a), which is derived from a refined 2-D mesh with specialized crack
tip elements, as shown in Figure 7.8(a). Note that the reference stress is the geometric hot spot
stress, as would come from a finite element shell analysis, so the effect of overall connection
geometry and loading is already accounted for. Thus the Y-factor only needs to account for weld
detail geometry. In AWS practice, weld details (A, B, C, D) are specified as a function of local
282

dihedral angle. Aptech developed Y-solutions for each of these, as well as a special case which
occurs at the saddle position in equal-diameter connections (beta of unity). In addition to the
basic flat weld profile, they also analyzed improved concave profiles and less desirable convex
profiles, representing a range of weld toe angles. The parameter space they covered is shown in
Figure 7.8(b). Separate solutions for shell bending, membrane stress, and punching shear, are
superimposed to match the 3-D shell conditions at the hot spot location. Notch effects elevate
in the first 5% of growth through the thickness, and levels off at 20% of thickness,
following the observed behavior of tubular joints in large scale fatigue tests.

D E T A I L A
<

UJ

<
1 ^ s m ^ i 1 1
jbi^ri I
1 f^^Pl 1
<


60
1
1 l/VI\AJ 1 1 1 1 1

(a)
J
D 60 90 120

(degrees) \J/

WELD TOE ANGLE

(b)
Fig. 7.8, Aptech fracture mechanics study for tubular connections, (a) Refined mesh at 2-D slice
used to develop Y(a) solutions, (b) Parametric range of fracture mechanics solutions.

These solutions are archived, interpolated, and superimposed in a PC diskette, TJLIFE,

which also integrates the crack growth from initial flaw to terminal failure, using repetitive
application of the annual stress range distribution. An example of this analysis is presented later.

7.1.4 Recent Large Scale Data

In Figure 7.9, British and European data as of 1984 (ref. 4) are presented in comparison
with the present AWS fatigue curves which were adopted in 1979. Curve X-2 is the new basic
standard, while curve X-l (similar to the original curve X) applies to welds with improved
profiles and/or limited thickness. Although the practical applicability of the hot spot stress
approach is reaffirmed, the data are segregated according to chord thickness. Most of the tests
are shown as cycles to failure (N-3), although N-l (detectable crack initiation) is also shown for
the thick in Dutch and French tests whose weld profiles were shown earlier. These specimens
cannot be considered to have improved profile. Thus, it should be no surprise that they fail near
or below the X-l curve. However, curve X-2 falls well on the safe side of even these extreme
cases of heavy section thickness and poor profile. Also, note that more than half the fatigue life
remains after detectable crack initiation.
The massive data produced by recent European Community research remains
unassailable as valid data. The adverse size effects they show are real. However, the fatigue
behavior of welds in tubular connections is not all size effect, as implied in the British Dept. of
Energy rules, or just weld profile as implied in the interim American codes, but a combination of
the two.
 283

Fig. 7.9. Experimental results

vs. American criteria (AWS
Dl.l).

7.1.5 Which hot spot stress?

As previously discussed, hot spot stress has evolved as the most practical basis for
design purposes, placing many different structural geometries on a common basis. In AWS
practice, the reference stress (or strain) is the total range which would be measured by a strain
gage placed adjacent to the toe of the weld, and oriented perpendicular to the weld. This is used
with an empirical S-N design curve, with the effect of representative local/microscopic
discontinuities at the weld toe presumed to be built into the data base of realistic as-welded
hardware.
Re-examination of the foregoing definition from a research viewpoint reveals that it is
full of ambiguities. Professor Radenkivic's classification of the different stress levels is more
rigorous, as discussed in Section 4.2.2(ii), and briefly reiterated below:

Sigma-G is the geometrical hot spot stress, which should be invariant given relative
diameters, thicknesses, and angles of the intersecting members.

Sigma-L is a more localized stress, which includes some notch effects of weld profile
shape and size.

To cite an example, for the large scale French tests (Figure 7.5(d), ref. 13), the strain
concentration factor SNCF is 3.3 for sigma-G, and 6.6 for sigma-L.
In the foregoing discussion of weld profile and size effects in terms of fracture
mechanics, sigma-G was clearly intended as the reference stress; and, whereas notch stress
analysis used sigma-L, results are presented in reference to sigma-G. This is what designers
using parametric SCF formulae relate to. When the stress analysis stops at sigma-G, weld profile
and size effects must be addressed elsewhere in the design processi.e., the "size effect"
adjustments to the S-N curves in the British Department of Energy rules, and the somewhat more
comprehensive "size and profile" provisions published in AWS D 1.1-86.
284

In OTC 4866 (ref. 14), Dijkstra, et al, explained away the beneficial effects of improved
weld profile in terms of reduced experimental hot spot stress at the extended weld toe location.
This situational interpretation is inconsistent with the design concept of sigma-G as invariant for
a given connection geometry, and would require an additional level of analysis to account for
weld profile.
If we must abandon the concept of invariant sigma-G, then it may be appropriate to re-
examine sigma-L as a basis for dealing with size and profile effects. At the very least, it seems
worthwhile to adopt a consistent standard definition of sigma-L, and to attempt an S-N
correlation of recent European data on this basis. This work remains to be done.

7.1.6. Design Application

In one of the author's recent design projects, following 3-D finite analysis of a large
ring-stiffened tubular connection (refs. 15 and 16), the analysis continued to examine 2-D slices
of weld details at critical locations. Figure 7.10 shows stress results from both the 3-D linear
shell analysis of the whole connection and the fine mesh analysis of a 2-D slice through the weld
at the heel, detail "D". The former indicated the brace hot spot to be the critical location.
However, because of the internal ring stiffener and concentrated transfer of load into the chord,
punching shear and shell stresses were higher inside the brace footprint, rather than outside as is
usually the case. 3 - 0 LINEAR V 4 I K S 2
J
SHELL Js. , " ,F'

Fig. 7.10. Results from refined

mesh finite element analysis.
STIFF. J
RING 2-D SLICE AT DETAIL "D*

The 2-D fine mesh analysis generally confirmed the 3-D analysis at the brace hot spot,
when the weld merges smoothly with the adjoining base metal and the actual weld toe location is
taken into account. However, at the root of the weld, there was a nasty surprise. Because of the
severe notch effect, local peak stresses are three times higher than the geometric hot spot stress
calculated in the full 3-D model.
The usual design method is based on hot spot stress (sigma-G) derived from shell
analysis or comparable parametric equations, with notch effects at the toe of the weld built into
the S-N curve. Since the analysis of 2-D slices allows us to take an explicit look at the notch
effects included in sigma-L, a different methodology must be used in translating this to fatigue
life. Here the author used what may be called the "SAE rule" (ref. 17), which has reportedly
been used successfully on welded earth-moving and mining equipment covering the thickness
range of .375 to 8 inches (10 to 200mm). It may be stated as follows:

(1) Local stress (sigma-L) is defined as that which would be measured by an 0.25-inch
(6mm) strain gage straddling the notch. For geometrically similar sharp notches, the
larger one will result in a higher measured stress, as more of the strain gage will be in
the peak region. Thus, some size effect is involved, and the SCF will vary with
thickness.
 285

(2) Fatigue life is calculated with an S-N curve which is quite similar to AWS curve X - l ,
for 1-inch reference thickness and long lives. It should be noted that the SAE method,
which used the higher sigma-L stress, will be generally more conservative than the
AWS method, which uses the sigma-G hot spot stress with the same curve.

(3) The fatigue strength is corrected for a small remaining size effect, in proportion to the
-0.034 power of stressed volume, or -0.1 power of thickness. The rationale is that this
reflects the larger probability of initial flaws being found when larger volumes are
involved.

Results of applying the SAE method to the ring-stiffened joint are shown in Figure 7.11 as
fatigue strength (the designer's conventional sigma-G) versus thickness, for both detail "A" at the
toe of the brace, and detail "D" as discussed above. Both details show a definite size effect, with
lower fatigue strength at heavier thicknesses. Furthermore, detail "D" with its more severe notch
shows much lower fatigue performance across the board than detail "A". In line with the A-to-B
comparison of Figure 7.4, the branch thickness is used as the reference for showing size effect.

160

Fig. 7.11. Results of applying

the SAE method.

Results of a fracture mechanics calculation of fatigue life for detail "A" are also shown
in Figure 7.11, indicating general agreement with the SAE method. Initial flaw size of 0.01-in. is
consistent with magnetic particle inspection of the finished weld toe. Terminal flaw size was
derived from CTOD calculations for the hot spot stress, which at this location includes an
unusually high proportion of membrane stress, in addition to shell bending and punching shear.
Calculations were done with the proprietary Aptech program, TJLIFE, as described earlier.
Crack growth rate used the DNV mean seawater curve with a threshold of 2.0 (ksi units). The
annual stress histogram came from the design spectral fatigue analysis for the platform.
286

7.1.7 New A.W.S. Size/Profile Rules

When AWS and API made an interim adjustment in their 1980 codes, the emphasis was
on weld profile effects, with both American codes including a lower fatigue design curve for
welds which do not merge smoothly with the adjoining base metal. The API curves are shown in
Figure 7.12. This reflected the point of view that it is the severity of the notch at the toe of the
weld, not thickness per se, that is responsible for the observed reduction in fatigue strength.
Despite this emphasis on weld profile, quantitative guidelines were lacking.

X1

10*

PERMISSIBLE CYCLES OF LOAD

Fig. 7.12. API/AWS fatigue design curves (lower API cutoff stress is shown).

Following leads from the Paris conference, the author examined the problem from the
standpoint that both size and profile are important, and their interaction ought to be quantified.
Both notch stress theory and fracture mechanics were examined, as previously discussed. Both
show progressively lower fatigue performance for progressively more severe notch conditions at
the toe of the weld. However, each weld geometry also shows a size effect, with notch stress
theory indicating log-log slopes of -0.25 to -0.40. The dashed lines in Figures 7.2 and 7.7
indicate how the size effect could theoretically be mitigated by profile rules in the AWS Code,
i.e, the 1/8-inch (3mm) limitation on reinforcement on butt welds, and the modified disc test
(radius of 1/2 the branch member thickness) for tubular joints. For thicknesses present in typical
Gulf of Mexico platforms, a U.S. dime or quarter is the disc of choice; for North Sea thicknesses,
a silver dollar may be required.
In 1986, the AWS D l . l Code defined a consistent set of weld profile requirements,
which represent a natural progression with branch member thickness for the welder to follow.
These vary from flat fillet-like profiles for the thin members, to concave radius-controlled
profiles for thick members, as shown in Figure 7.13. The applicable thickness range of these
profiles depends on the level of fatigue performance required by the designer, i.e., whether he has
used the upper fatigue curves (solid lines in Figure 7.12) or the lower ones (dashed lines). In this
regard, design must be integrated with QC/QA during construction to assure that the intended
weld profiles will be provided.
Corresponding changes have yet to be adopted by API.
 287

UPPER LIMITS FOR

UPPER CURVES LOWER CURVES
XI, X2, K2
BEVEL ROOT
THROAT "T"
ETC. PER AWS UNLIMITED NOT REQ'D
TABLE 10.7

AWS
FIG. 10.12
CONCAVE
UNLIMITED
DISK TEST

AWS
FIG. 10.11 5/8"
TOE FILLET (UNLIMITED FOR
STATIC COMPRESSION)

AWS
FIG. 10.9
BASIC 3/8" 5/8"

FLAT PROFILE

Fig. 7.13. AWS weld profile requirements and corresponding branch thickness limitations.

7.1.8 Confirmation Tests

(i) Rice University. As has been described, the AWS D 1.1 -86 rules for weld profile
and fatigue size effect were based largely on theoretical considerations (fracture mechanics and
notch stress analysis), with only second-hand access to the proprietary European research in this
area. Thus, before putting these proposed rules into practice on a major new offshore platform,
Shell commissioned a quick series of eighteen fatigue experiments, designed to explicitly follow
and test the AWS rules. These tests were conducted in air at Rice University (ref. 18). API and
several industry companies have jointly sponsored a more extensive three-year series of tests at
Florida Atlantic University, which is doing the same in a seawater corrosion fatigue
environment, as described in the next section.
Steel for both the Rice and FAU experiments is API-2W-Grade 42 (Kawasaki TMCP
type). Geometrically similar half-inch, one-inch, two-inch and four-inch thick specimens were
welded in various positions, so as to get a representative variation in quality of profile. A very
high level of profile workmanship was being achieved in heavy production welds by the Rice
series fabricator, Bullwinkle Constructors. This was reflected in the thickest test specimens,
while the thinner specimens were intended to reflect the more lenient profile requirements of the
new AWS Code, as shown in Figure 7.14. The 2-inch (50mm) thick specimens met the disc
profile test in all welding positions. However, despite a round of practice welds, some difficulty
was experienced in consistently achieving the alternative profiles permitted by AWS for thinner
sections. The 1-inch (25mm) overhead welds were worse than intended, and the 0.5-inch
(13mm) vertical welds were too good. Although a different Texas fabricator (Brown & Root)
made the FAU specimens, similar results were achieved, with some tendency to over-weld the
improved profiles.
288

BRANCH THICKNESS
1/2 INCH SPECIMEN
AWS BASIC-FLAT FACE Vf ' / 2" I"

<

>
<
LJ

UJ
CC

CHORD THICKNESS

Fig. 7.14. Interpreted results of Rice fatigue tests.

For fatigue testing, the stem of the tee-weld was loaded in tension, so as to produce a hot
spot stress range (sigma-G) of either 11 or 22-ksi (77 or 154 MPa) linear shell bending at the
specified weld toe position. After some of the thinner specimens survived the programmed
2,000,000 cycles at 22-ksi, they were retested at progressively higher stresses until they failed; in
plotting these results, cumulative fatigue damage versus AWS curve X-l was used.
Interpreted test results are plotted in Figure 7.14 as a function of thickness, fatigue
strength (sigma-G at the specified weld toe position) relative to the AWS X I design curve. As
the weld details represent AWS single-sided tubular joints for half the stem thickness, the upper
thickness scale interprets the stem as two branch thicknesses back-to-back; this represents a
conservative treatment of the data. The smaller, lighter symbols are runouts at 20,000,000 cycles
of 11 ksi; the heavy ones failures.
Also shown is a prediction line, plotted as sigma-G but based on SAE rule calculations
using sigma-L from Rice's detailed finite element analysis of the three weld profile types. Again
we see a family of curves, all showing a size effect, but with higher performance for the higher
quality weld profiles, as shown by the dashed lines. The solid line reflects the AWS intent, that
is to maintain a more-or-less constant level of fatigue performance over a range of thicknesses by
varying the weld profile. Extending the basic profile to heavy thicknesses beyond its intended
range would be expected to incur a significant size effect penalty.
The data all fall on the safe side of the AWS-X1 design curve, and even further on the
safe side of the SAE Rule prediction. Relaxation of the disc test profile requirements for thinner
sections appears to be fully justified. Indeed, the thinnest specimens show a greater scatter
towards fatigue lives on the high side, suggesting that the size effect in air may be somewhat
stronger than what the profile relaxation anticipated.
 289

(ii) Florida Atlantic. The API-joint industry program of size/profile tests at Florida
Atlantic University was conceived in 1984. By 1989, it had completed two of its three planned
years and had two years yet to go. The total program involves a large matrix of 80 tests to
develop S-N data covering four thicknesses, the three types of weld profile (each made in various
welding positions), and two natural seawater environments (free corrosion and minimal cathodic
protection at -0.8 SCE). Test methods and results to date have been described by Hartt (ref.
19). Some 45 tests have been completed for the analysis presented here.
Specimens are loaded with the stem of the tee in tension, at a stress ratio R of 0.1, with a
cycling rate of 0.3 Hz. Individual tests can last several months. To save machine time, several
specimens are being tested in parallel, necessitating stroke-controlled loading. Crack growth
behavior observed in tubular connections appears to be intermediate between load-controlled and
displacement-controlled solutions, with much of the life spent in propagating small cracks
unaffected by the difference. To avoid any question of unrealistically long lives due to the mode
of loading, failure was defined as cracking halfway through the thickness.
Results in S-N format are presented in Figures 7.15 and 7.16. These data are for free
corrosion, which show less variability than the cathodic protection data. Hartt's use of the
specimen centerline stress as hot spot stress is consistent with the use of thin shell finite element
methods in design, or parametric SCF equations based on such methods, e.g., Kuang's.
Cathodically-protected results and alternative stress intepretations will be presented later.
Welds meeting AWS Level I profile requirements, comparable to the Rice series,
generally fall on the safe side of the AWS-X1 (API-X) design curve, with the data having an S-N
slope comparable to the design curve. The 4-in. (100mm) thickness would require a size effect
reduction of fatigue strength according to Commentary Section 10.7.6 of the AWS Code, so its
lower performance is no surprise. The one premature failure in the 2-in. (50mm) thickness was
traced to a momentary lapse in meeting the profile quality of the rest of the welding, illustrating
one of the pitfalls of counting on weld profile for improved performance.
Flat-faced welds, made according to the AWS basic detail regardless of thickness, show
lower performance and a more severe size effect, also as expected. Their S-N slope is
comparable to the AWS-X2 (API X-prime) curve for welds without profile control. Recall that
curve X-2 is substantially lower than curve X-l (by as much as 32% in Figure 7.12, up to 38% as
plotted in the AWS Code).
A more direct comparison of size and profile effects, versus the AWS Code, can be
found in Figures 7.17 and 7.18. The fact that the S-N data agrees with the slope of the design
curves facilitates conversion to this size effect format. Both free corrosion and cathodically
protected data are now shown. The interpretation of invariant hot spot stress sigma-G is now at
the toe of the specified weld, consistent with treatment of the Rice data and with modern
definitions of hot spot stress (EEC-WG3, analysis with finite thickness isoparametric elements,
or use of Efthymiou SCF equations). For the specimen geometry tested, this is 90% of the
centerline stress, resulting in a more conservative treatment of the data. Although welders must
often over-weld in order to pass the disc test (especially the less artistic ones), this is to some
extent a self-compensating mechanism, and beyond the domain of design calculations.
For welds meeting AWS Level I requirements, the weld profile is varied with thickness
to keep substantially similar fatigue performance up to branch thicknesses of 1 in. (25mm).
Beyond this thickness, unless the weld surface is ground smooth, a size effect penalty is imposed,
following the -0.25 power of thickness. In the plot, the Code provisions are shown as a band,
depending on whether the effective branch thickness is interpreted as or T/2. Except for the
one premature failure discussed earlier, the seawater data fall on the safe side of AWS design
criteria. For non-redundant structures, AWS applies an additional safety factor of three on life,
corresponding to an additional 2 1 % reduction in design fatigue strengths beyond what is shown
in the figure, so the criteria would even be on the safe side of the outlier.
290

FREELY CORRODING

Fig. 7.15. Freely corroding fatigue data for AWS Level 1 profile specimens of all thicknesses (ref. 19).

N f - CYCLES TO HALF - CRACK

Fig. 7.16. Freely corroding fatigue data for Basic specimens of all thicknesses (ref. 19).
 291

BRANCH THICKNESS AS T/2 (mm)

WELDS MEETING FC CP
AWS LEVEL I BASIC
PROFILE REQ'TS
FILLET (ALT*I)

CONCAVE <ALT#2>

TTTJTT
Fig. 7.17. Invariant sigma-G
AWS CODE fatigue strength in size effect
PROVISIONS format, for welds meeting AWS
Level I profile requirements.

HOT SPOT S T R E S S = 0 . 9 U
INVARIANT AT S P E C I F I E D WELD T O E

CHORD THICKNESS (mm)

BRANCH THICKNESS AS T/2 (mm)

WELDS PER FREE CORROSION

AWS BASIC
PROFILE CATHOOIC PROTECTION

Fig. 7.18. Invariant sigma-G

BB
fatigue strength in size effect
format, for welds meeting AWS
basic profile requirements.
AWS CODE
PROVISIONS

HOT SPOT S T R E S S = O . S U ^ E
I N V A R I A N T AT S P E C I F I E D WELD TOE

25 50
CHORD THICKNESS (mm)
292

For flat-faced welds made according to the AWS "basic" details, all the test data fall on
the safe side of the lower design criteria which would apply - curve X-2 and earlier imposition of
the size effect reduction. The thinnest specimens (as T/2) qualify for Level I performance,
resulting in a step up in the design criteria. Both the data and the design criteria show the size
effect persisting over the full range of thicknesses tested for these welds.
Somewhat surprisingly, fatigue under cathodic protection does not show any consistent
improvement over that under free corrosion. Indeed, the CP data show much more erratic
variability, with it being possible to draw opposite conclusions from different subsets of the data
(perhaps this is why one finds so much confusion in the literature). The typical observation is
that CP improves fatigue performance up to crack initiation, but accelerates crack growth
thereafter.
The FAU test data are interpreted against European design criteria in Figure 7.19. Here
we apply the "de Back" interpretation of situational hot spot stress (ref. 14), using shell bending
stress projected to the actual weld toe location, with reductions from centerline stress as shown in
the Figure. Even so, we see a clear separation between the performance of flat-faced (basic)
weld profiles and concave (alt. #2) profiles, which would have been even greater in terms of the
designer's invariant hot spot stress sigma-G. The two trend lines are shown with different
slopes, as predicted by notch stress theory; they appear to merge in moderate thicknesses where
earlier investigators found little difference due to weld profile, but become more widely
separated at very large thicknesses.

FC CP
FAU TESTS BASIC t .91 U>
INTERPRETED
FILLET (.89 0^")
AS PER DE BACK
CONCAVE i .85 U^>

/ DENOTES RUNOUT

1981
D.EN
GUIDANCE

Fig. 7.19. Florida Atlantic data

interpreted against European
criteria.

THICKNESS (mm)

Interpretation against the U.K. Department of Energy "T" curve (ref. 20) is also shown.
In contrast with AWS practice, chord thickness is the only reference for size effect. Here the
shaded band represents the difference between rules for free corrosion and cathodic protection.
With a log-log slope of -3.0, the "T" curve is steeper than the AWS design curves and plots
higher in the range of the test data; the mismatch in slope also causes more scatter in converting
S-N data to size-effect format. Much of the CP data for flat-faced basic weld profiles falls on the
unsafe side of the CP design criteria. With profile control (i.e., concave weld details) in the
heavier thicknesses, performance more consistent with the design criteria would be obtained.
However, given all the criteria emphasis being on plate thickness per se, and the labor union
 293

problems which arise from making craftmanship distinctions, there seems to be little incentive
for specifying and enforcing the necessary profile control in practice. Perhaps the round of
revisions currently being considered will correct some of these problems.
(iii) Summary & Conclusions. AWS fatigue design criteria, along with recent size and
profile considerations, have been described. Test data have been presented which confirm these
criteria for thicknesses up to 4 in. (100mm), with a variety of weld profiles and service
environments. The geometric hot spot stress, invariant sigma-G, remains as the basis of today's
design practice. Other treatments of the subject may be found in References 21 and 22.

7.1.9 Fatigue Improvement Methods

AWS clause 10.7.5 discusses weld improvement methods which may be used to raise
the fatigue performance from the lower curve X2 to the upper curve X I . These are (1) improved
as-welded profile, (2) grinding, and (3) peening.
Bignonnet (ref. 23) and Haagensen (ref. 24) have presented recent overviews of the
subject. In a broad sense, weld improvement works by extending the fatigue crack initiation life,
by (a) improving the stress concentration (sigma-L) of the weld, (b) reducing the notch effect and
removing crack-like defects at the toe of the weld, and (c) reducing tensile shrinkage stresses or
inducing compressive residual stress.
The use of improved as-welded profile has been extensively discussed in the preceding
sections of this monograph. Figures 7.20 and 7.21 present yet another confirmation of the
validity of this approach (ref. 25). This is particularly gratifying, coming as it does from the
magis arnica veritasis of Reference 13.

2 3 4 56
CYCLES

Fig. 7.20. Influence of the improved weld profile on the fatigue resistance of fillet weld with improved
profile (ref. 23).
Last

Fig. 7.21. Control of the global geometry for improved weld profile (ref. 23). Shaded beads are those
present when the toe run is placed.
294

Figure 7.22 shows some of the early data from the UK Offshore Steels Research
Program (ref. 26), compared to today's AWS criteria. These results for flat-faced basic as-
welded profile were some of the data which led to the establishment of the lower fatigue curve
X2 for hot spot stress. The weld profiles are so severely notched that free corrosion in seawater
does not make their fatigue performance noticeably worse. If the plate bending is considered as a
nominal stress, AWS curve D would also be applicable; with its size effect adjustment, it falls
well to the safe side of all the data. Under free corrosion conditions, the weld improvements
wrought by grinding are eventually destroyed.
For fatigue in air, or in seawater with cathodic protection, Figure 7.23, grinding the weld
toe as shown in the inset greatly improves the fatigue performance. Grinding must extend below
the plate surface in order to remove microscopic weld toe inclusions. A brief (six week)
exposure to free corrosion without cyclic stresses does not destroy the beneficial effect. All the
data, as well as a characteristic design line (shown dashed), plots well above the upper AWS
design curve XI for hot spot stress. Indeed, based on this limited data set, one would seem
justified in using a still higher design criterion, e.g., AWS curve B.
When the full weld face is ground and geometric similarity is observed (i.e., radius
equal to half the branch member thickness), AWS allows the use of upper curve CI for unlimited
thickness, with no size effect. Other authorities (refs. 17 and 24) suggest a small residual size
effect, exponent of -0.1, when there are no unaccounted for notch effects. However, the data in
Figure 7.23 fall so far to the safe side of the CI design curve, that such a size effect would not
cause problems until the thickness reached 100 in. (2500mm).
Heavy hammer peening, with a blunt-nosed power tool, plastically reshapes the weld toe
(similar to toe grinding) and also introduces compressive residual stresses. The effect upon
fatigue performance is even better than grinding, as shown in Figure 7.24. However, there are
some caveats associated with this method. Since it can smear over weld toe defects, magnetic
particle inspection is mandatory. Also, consideration should be given to the possibility of
locally-degraded notch toughness due to the strain-aging effects which follow the heavy plastic
deformation involved.
Other methods of weld improvement, e.g., TIG remelting of the weld toe, also have
been found to be effective, but are not covered by AWS.

7.2 FRACTURE TOUGHNESS

As previously described, localized yielding, strain hardening, and triaxial stresses are
required as tubular joints mobilize the reserve strength upon which their design criteria depend,
placing extraordinary demands on the steel. These demands must usually be met in the presence
of notches, since the highest stresses almost always occur in the hot spot region, at the toe of the
weld joining the tubes where they intersect. Notch toughness is a most important material
property here, for without it the connection is subject to premature brittle fracture.
Brittle fracture is a type of catastrophic failure in structural steels that occurs with little
or no prior plastic deformation, and once initiated can propagate at extremely high speeds. The
failure is usually characterized by a flat fracture surface, having a coarse crystalline appearance
and chevron markings pointing back toward the point of fracture initiation. This flat fracture
may be accompanied by a small region of silky-appearing slant fracture near the surface, termed
shear lips.
Whether a fracture is premature or not is often debatable, depending on whether the
gross loading or deformation or other conditions of service (such as loading rate and
temperature) were within the envelope which the structure was supposed to withstand.
 295

9= 200

CYCLES TO FAILURE

Fig. 7.22. S-N curves in air and seawater for basic flat as-welded profile; also ground welds in
seawater, free corrosion (ref. 26).

2 3 4 5

CYCLES TO FAILURE

Fig. 7.23. S-N curves in air and seawater for ground specimen (ref. 27).

2 3 4 5

CYCLES TO FAILURE

Fig. 7.24. S-N curves in air and seawater for hammer-peened specimen (ref. 27).
296

The following factors must all be present in order for a brittle fracture to occur:

(1) Design, workmanship, or fatigue create a crack-like discontinuity or flaw.

(2) The flaw is not detected by inspection and eliminated.
(3) Limiting conditions of restraint, low temperature, and/or loading rate reduce the
effective ductility of the material to below some critical value for fracture initiation.
(4) High stress and/or inelastic deformation locally coincide with the foregoing conditions
in order to initiate a running crack. Residual stress in as-welded structures must be
included here.

Brittle fractures are not as common as other modes of structural failure and are mainly
feared because of the suddenness with which they occur. The probability of failure by brittle
fracture occurring somewhere in a complex steel structure may be expressed as the product of the
foregoing elements. Reduction of any of these factors correspondingly reduces the overall risk.
Reducing several of them, even modestly, has a compound benefit. Since structural steels
exhibit a transition temperature above which fracture toughness increases rapidly, this becomes
an obvious factor to control. It is not the only one, however.

7.2.1 Thermal Conditions of Service

The bulk of an offshore structure is submerged in seawater. By definition, temperatures
are above freezing, with minimum in the 28-50 F (-2C to 10 C) range, depending on
geographical location.
For the atmospheric zone, the Lowest Anticipated Service Temperature (LAST) is
subject to various definitions. These range from the 100-year lowest instantaneous air
temperature to the lowest 24-hour average temperature to be experienced once in ten years. The
latter definition recognizes that the structure is a heat sink and does not respond to instantaneous
temperatures. It also accepts some risk of the temperature criteria being violated, presumably
recognizing that low temperature alone is not a sufficient condition for brittle fracture, but only
one of the required chain of events.
The following values of the LAST have been suggested for various offshore operating
areas (ref. 28).

AIR WATER

Gulf of Mexico + 1 4 F (-10 C) 50 F (10 C)
0
Southern California +32 F ((0 C) 40 F (4 C)

Cook Inlet, Alaska -20 F (-29 C) 2 8 F (-2 C)

For the Gulf of Mexico, a somewhat higher temperature70 F (21 C)prevails for
both air and water during hurricane season, which corresponds to the occurrence of significant
loads approaching design level. At great depths, below 1000 ft., the Gulf of Mexico water
0
temperature is 40 F ( 4 C), like deep oceans worldwide.
For exposed onshore structures, temperature maps derived for bridge design are useful,
e
as shown in Figure 7.25 (ref. 29). For the interior structure of heated buildings, 5 5 F ( 1 3 C ) i s a
reasonable minimum.

7.2.2 Gulf of Mexico Experience

Early fixed platforms in the Gulf of Mexico were constructed of ordinary mild steel
ASTM A7 and A36 plate, and A53 Grade pipe. Because water temperatures rarely got below
60 F and section thicknesses were in the range of 3/8 to 3/4-inch (10 to 20mm), low-temperature
 297

Fig. 7.25. Isolines for 20-year extreme low hourly temperatures, with absolute minimums for each
state (from ref. 29).
298

brittle fracture was not anticipated as a major problem. Unfortunately, the transition
temperatures of these steels varied widely, with the high end of the scatter band falling on the
wrong side of environmental temperatures, as shown in Figure 7.26. In some of these cases,
brittle failure resulted, although there were often extenuating circumstances such as low air
temperatures, collisions, or nearby blasting. Because of the multiple load path redundancy of
typical fixed platform jackets, none of these failures led to catastrophic collapse of the structure
(ref. 30).
2 5 r

| J GULF O F MEXICO
u j j RANGE OF WATER
20h
I TEMPERATURES

3 to
UJ
U H- 15
Ou. ALL SAMPLES
u.

SAMPLES FROM

Si OFFSHORE FAILURES^

=>

2 5
I
-100 -50 50
C H A R P Y - V 15 FT-LBS TRANSITION TEMPERATURE F

Fig. 7.26. Transition temperature data for A36 plate and A53 Grade pipe.

7.2.3 Initiation Barrier

The notion that transition temperature is a fixed and universal property of the material ~
below which it behaves like glass, and above which nothing can make it fracture ~ is incorrect.
A more realistic picture of the transition behavior of structural steel is shown in Figure 7.27
along with the regions where various theories and strategies for dealing with fracture apply.

Fig. 7.27. Transition behavior of steel.

 299

Fracture toughnessenergy absorbed per unit area of fractureis a material property which can
be used to define the critical combination of stress (or deformation) and crack size which can
cause fracture. However, effective fracture toughness varies with specimen size, temperature,
and strain rate. It may be measured directly or indirectly in a variety of tests, such as the Charpy
V-notch, the NRL drop weight nil-ductility test, and more rigorous fracture mechanics tests.
These will be discussed in the following sections.
The problem with many of the common tests is that the testing conditions do not
necessarily represent those of the prototype structure whose failure we wish to predict and
prevent. Low-energy plane-strain cleavage fracture occurs when yielding near the crack tip is
suppressed by loading rate or restrained by heavy thickness. Above the nil-ductility transition
(NDT), as the temperature is increased, stress relaxation or yielding occurs more readily and the
size of the plastic zone with associated shear lips near the free surface increases. This increases
the energy absorbed during fracture. At high temperatures, the plastic zone extends through the
thickness and a high-energy ductile tearing mode of failure prevails.
For a given absolute size of plastic zone and shear lip, a heavy section (e.g., prototype)
will be relatively less ductile than a small specimen, since restraint to plastic flow in the interior
region means that a larger proportion of the fracture area is brittle. This shifts the transition
curve, i.e., the rise in average toughness, to higher temperatures.
The nil-ductility transition (NDT) temperature corresponds approximately to the first
rise in fracture toughness, as measured under limiting conditions of dynamic load on relatively
small laboratory specimens. With static or quasi-static loading of strain-rate-sensitive materials,
e.g., conventional structural steels, the entire pattern is repeated at a lower temperature. At a
given temperature, there is more effective fracture toughness under static conditions, because
stress relaxation and yielding have more time to occur than under dynamic conditions.
The rate of application of wave forces on an offshore platform, and of its dynamic
response, is intermediate between that of impact testing and static fracture testing. For a
platform subject to a typical design wave, the peak loading rate corresponds to a rise time of 2.3
seconds from zero to peak load. For a peak nominal stress of 20 ksi, a geometric SCF of 3.0 for
the tubular connection geometry, and a microscale SCF of 1.7 for weld toe notch effects, this
yields a loading rate of 0.0015 strain per second, compared to rates of 10-650 per second for
impact testsfive or six orders of magnitude slower.
In the sense of Figure 7.27, this difference in loading rates corresponds to a temperature
shift of 80-120 F for mild steel (refs. 31, 32), while static loading conditions would correspond
to an additional shift of 40 F. For high-strength steels, the shift is not as large, being less than
half as much at the 100 ksi yield strength level.
Taking the more conservative figure for mild steel, under the 70 F quasi-static
conditions of service for an offshore structure subject to hurricane loading, the steel should
exhibit notch toughness comparable to that of impact tests at 150 F. This higher toughness
constitutes an initiation barrier against initial propagation of fracture from small or moderate
crack-like flaws under static or quasi-static conditions. This initiation barrier explains the
satisfactory fracture performance of many older offshore structures, as well as that of most
buildings, bridges, and other onshore engineering structures.
However, once a fracture starts propagating, the strain rates at the advancing crack front
become more comparable to those obtained in impact tests. The initiation barrier is no longer
effective, and if there is inadequate dynamic notch toughness, the crack will continue to
propagate, with catastrophic results. Even when the overall structural loading is static, pop-in
can provide the initial dynamic conditions for fracture. In welded structures, the initial pop-in
can result from sudden failure of a locally embrittled weld or heat-affected zone. Even where the
welding procedure has demonstrated acceptable weld metal and heat-affected zone toughness,
300

variations in workmanship and technique can still provide isolated fracture initiation sitesfor
example, untempered toe beads having a poor profile.
Reference 29 describes a brittle fracture which occurred during construction of an
offshore platform, in which several adverse circumstances combined to overcome this initiation
barrier. The author was dismayed to discover how ubiquitous are the factors required for brittle
fracture can really beespecially for tubular connections, which require plastic behavior at the
hot spot, in the presence of weld toe notches and flaws.

7.2.4 Temperature-Shifted Charpy Criteria

There are today several viable approaches available to the designer for specifying notch
toughness. They cover a broad spectrum in the degree of protection they provide against
catastrophic brittle fracture, particularly in the contingencies for which they allow. However, no
reasonable level of specified base metal notch toughness can absolutely prevent all fractures
under all conditions of fabrication and service, for all thicknesses and applications.
A standard specification, which is often cited as a precedent for establishing minimum
levels of notch toughness, may be found in ASTM A709, Structural Steel for Bridges. For steels
having specified yield strengths below 50 ksi (350 MPa) and thickness up to 2 in. (50mm), the 15
ft.-lb. (20J) minimum average energy absorption in the charpy V-Notch test is specified as
follows:

SERVICE LOWEST ANTICIPATED SPECIFIED CHARPY

ZONE SERVICE TEMPERATURE TEST TEMPERATURE
op o C op c

T-l above 0 above-18 70 21

T-2 -lto-30 -18 to-34 40 4
T-3 -31 to-60 -35 to-51 10 -12

For grade 50 steels above 2 inches (50mm) thick, the energy requirement is increased to
20 ft-lb; furthermore, for errant heats of grade 50 steel whose actual yield strays above 65 ksi

(450 MPa), the Charpy test temperatue is reduced 15 F ( 8 C) for each increment or fraction of
10 ksi (70 MPa) above this limit.
The technical basis of Charpy testing shifted 70 C (40 C) on the unsafe side of the
service temperature is the initiation barrier which exists at quasi-static loading rates. For bridges,
30 ksi/sec (0.001 strain per sec) represents a conservative upper bound on loading rates observed
in the field. In Reference 28, it is observed that, for a given steel, the toughness indicated by a
dynamic fracture test, such as the Charpy test, would be duplicated at lower temperatures in
fracture tests conducted as slower loading rates. The magnitude of this temperature shift, for low
strength steels, is about 30 F (17 C) for each log cycle (tenfold) change in loading rate. Lesser
temperature shifts were observed for steels above 65 ksi (450 MPa) yield.
The overall fracture control approach in which these criteria have been successfully
applied should be borne in mind. Highway bridge girders are relatively clean structures, with the
tension flanges kept free of attachments which would cause stress concentration and unwanted
restraint. Welding is almost always done in a shop environment. Tension butt welds are ground
flush to eliminate stress risers, and non-destructively tested to acceptance standards considerably
higher than those cited for other classes of construction. The structure remains accessible for in-
service inspections.
Nevertheless, there have been a number of brittle fractures in heavy bridge girders built
of steels which met A709 temperature shift criteria. These were sufficiently dramatic to capture
the attention of the media and regulatory agencies; in most cases, there was no warning and
 301

complete collapse was prevented only by structural redundancy. Defenders of the A709
approach (ref. 33) correctly point out that most of these failures were preceded by a period of
fatigue crack growth, which began at defective or improperly repaired welds or at undesirable
attachment details. Tougher steels would not have prevented eventual failure, but only slowed
down the terminal stages.
The author's experience over the last 20 years has been that, since abandoning reliance
on the initiation barrier and temperature shift for protection, and adopting fracture safe design
practices described in the next section for his offshore platforms, the incidence of premature
brittle fractures has been reduced by two orders of magnitude.

7.2.5 Fracture-Safe Design

Several more conservative levels of fracture-safe design may be defined with reference
to the NRL fracture analysis diagram, Figure 7.28 (ref. 34). Derived empirically, this diagram
defines conditions of stress, temperature, and flaw size under which brittle fractures may initiate
or propagate. Symbols plotted on the diagram show how it may be related to various parts of an
offshore structure.
TENSILE
ST1INOTH
I N I T I A T I O N CURVES
( F R A C T U R E STRESSES
FOR SPECTRUM O F
F L A W SIZES) /
/ jpiNT/CAl^pl A S T,c
/ / / LOADS
J, yield
t/t STRESS f 4
/ 7/ \l INCREASING | STUB
STUB / /' ^ ~
ELASTIC
j/\FLAW SIZES / / LOADS
FRACTURES
DO NOT
"7ZZ2> PROPAGATE
< (TEMPERATURE L I M I T A T I O N )

i j/\ O LEG
,, m u in .
S-IKSI
(STRESS L I M I T A T I O N ) \

e
J I L
NDT NOT + 3 0 F NDT 4-60*F

LOWEST ANTICIPATED SERVICE TEMPERATURE

Fig. 7.28. Fracture analysis

diagram, illustrating offshore
application.
53(25.625
A-36
JACKET L E G

20)2$. 5 0 0
A - 5 3 BRACE

5401.500
NORMALIZED
A P I SPEC 2H
J O I N T CAN
302

The reference temperature for this diagram is the nil-ductility temperature, as defined by
the NRL drop-weight test (ref. 35). In this test, a brittle weld bead is applied to the surface of the
test plate with a saw cut notch to serve as a crack starter. The specimen spans as a simple beam
between supports and is loaded by a falling weight so that the brittle weld cracks in tension. This
provides pop-in dynamic fracture initiation as described earlier. A stop in the test jig limits
deflections and strains to a few percent. Thus, the test represents a realistic model for the hot
spot region of welded tubular joints. The test is evaluated on whether or not the initial pop-in
0
fracture propagates across the plate. The NDT temperature is defined as 10 F (6 C) below that
at which two consecutive samples exhibit no-break performance.
The nominal stress axis in the fracture analysis diagram refers to total tensile stress in
the region which contains the crack. For as-welded structures, this includes residual stresses,
which may approach yield. For flaws in the hot spot region of tubular joints, the effects of
geometric stress concentration factors must be included in calculating the relevant stress.
For mild and intermediate-strength steels up to 50 ksi (350 MPa) yield strength, yield-
level stresses are required to initiate brittle fracture from small flaws, even below NDT. The
dashed family of lines indicate that progressively larger flaws will initiate fracture at
progressively lower stress. The crack dimensions shown are crack lengths for half-thickness and
through cracks in material of limited thickness (up to 1 to 2 inches).
Above NDT, progressively higher stresses are needed to initiate brittle fracture for a
given size flaw, rapidly approaching the tensile strength in the case of small flaws.
The diagram also shows a crack-arrest (CAT) curve. Below NDT, brittle fractures will
continue to propagate as long as the stress is above 5-8 ksi, once initiated. At NDT +30 F,
fractures require stresses above half yield to continue propagation; at lower stresses crack arrest
occurs. Above the FTE temperature (fracture transition elastic, approx. NDT +60 F), brittle
fractures of any size will not propagate under static or dynamic elastic loading conditions.
Above the FTP temperature (fracture transition plastic, approx. NDT +120 F) failure is always
ductile, no matter what the loading, conditions of service, or prior crack size. Exceptions to the
foregoing occur for yield strengths over 50 ksi or thickness over 1 to 2 inches (25 to 50mm).
For higher strength steels, however, the amount of releasable strain energy may be so
great that even upper shelf toughness may not be sufficient to arrest dynamic fractures. Here
more modem approaches to crack arrest toughness should be applied (e.g., ref. 36).
The Fracture Analysis Diagram shown is for limited thickness, up to 1 or 2 inches (25 or
50mm). The crack arrest curve, FTE, and FTP shift 70-90 F further to the right for plate
thicknesses of 6-12 inches. Fracture initiation conditions are also affected: for example, where
8-inch through cracks in thin plate could be tolerated, for thick plates an 8-inch long by 1-inch
deep surface crack may become critical. Here more modem approaches to elasto-plastic fracture
mechanics, e.g., the CTOD design curve, (crack tip opening displacement, refs. 37 and 38),
should be used. However, the initiation curve for small surface flaws remains the same since
behavior here is related as much to flaw size as to section size.
Application of Figure 7.28 to offshore structure design will now be described. The
relevant conditions of stress, temperature and flaw size for each part of the example connection
are superimposed on the NRL diagram.
As previously discussed, the chord in tubular connections is subjected to triaxial stresses
and severe plastic deformation in developing its ultimate capacity. At NDT +28 F, or better,
typical normalized joint can material (e.g., ref. 39) would be able to arrest pop-in initiation from
small flaws and prevent initiation of brittle fracture from moderately sized fatigue cracks at
stresses well above yield.
The ends of the branch members are also subject to stress concentrations. However,
these hot spot stresses are a consequence of enforcing compatible displacements between
intersecting shells, and are not required for equilibrium; thus, plastic flow can limit the stress to
 303

yield level. For such applications below water, fully killed fine-grain practice mild steel may be
used. Operating at the no-break level of toughness, just above NDT, such steel should also be
able to withstand small and moderate flaws under its own less demanding conditions of service.
While use of special steel in the ends of braces may be desirable for important members, the
general use of such "stubs" is by no means universal. Their cost is disproportionately high in
relation to the modest improvement in notch toughness over ordinary mild steel, and the extra
welded connection is another potential source of problems.
Away from the tubular joint region, ordinary structural steel (A36 plate and A53 pipe) is
still used. Since these are operating below NDT, such a practice is relying on the initiation
barrier, as described earlier. For braces, the 20 ksi nominal stress is just above half yield.
Residual stresses will add to this as indicated by the arrow in the figure. Because minimum wall
thickness for local buckling usually governs for the jacket leg, this operates at even lower stress
levels.
The design and material selection practice presented in the foregoing will not prevent all
fractures. Rather, it is intended to eliminate premature, brittle fracture under conditions of
service (including flaw size) which represent reasonably credible extremes of what the structure
should be designed to resist. Here "flaw size" includes not only the built-in flaws that are
introduced during construction, but also are reasonable amount of growth in fatigue. In deriving
fatigue criteria for tubular joint design, fatigue failure was considered to occur when the crack
had grown virtually around the entire connection perimeter. For this fatigue life to exist, the
material has to tolerate very large cracks without fracturing. Fortunately, such large cracks will
usually have grown away from any embrittled weld heat affected zone, so that the higher base
metal toughness that prevails at slow strain rates may be invoked. If because of catastrophic
events, e.g., collision, or design blunders, the structure is severely overloaded and starts coming
apart, there may be some brittle fractures among the failures that were inevitable anyway.
The position of the nominal design relative to the crack arrest curve suggests that
individual braces may be completely severed, and that joint can fractures may propagate
completely around the localized high-stress region adjoining individual braces. However, since
such failures affect only the individual braces where they initiate, they are not likely in
themselves to cause catastrohpic collapse of redundant space frame structures under design
loading conditions. The role of redundancy is discussed in Section 7.4.

7.2.6 Charpy Criteria for Fracture-Safe Design

The Charpy V-notch impact test is one of the most widely used measures of notch
toughness. Validity of 15 ft-lb (20J) as a reliable indicator of above-NDT behavior for mild,
semi-killed, carbon steel has been well established by correlation with early welded ship failures.
The Charpy is also the test most often used, for historical and practical reasons, to control brittle
fracture in offshore platforms.
However, for intermediate and high-strength steels, the 15 ft-lb (20J) criterion becomes
less reliable. Comparison between Charpy tests and the actual NDT, as established by drop-
weight tests, for fully killed intermediate-strength steels of the 1960's (ref. 30) indicated
anywhere from 0 to 40 ft-lbs (14 to 55J) at the NDT. In API Spec 2H (ref. 39) and API RP 2A
(ref. 40), a median value of 25 ft-lbs (35J) is used. This approximately defined the no-break
level of toughness, with a tolerance of plus or minus 30 F. For today's more highly refined
node steels, however, the Charpy energy may be above 100 ft-lbs (130J) at no-break.
This lack of precision is one of the disadvantages of using the Charpy test, which is only
indirectly related to service fracture conditions. The measured energy comes from some
undefined combination of initial yielding at the blunt notch, eventual crack initiation, and
fracture propagation. The Charpy test can be fooled by high energy during the initial yielding
304

phase and may fail to detect steels which have unusually low initiation thresholds in the presence
of sharp natural cracks.
The temperature at which Charpy testing is to be done is not necessarily the same as the
lowest anticipated service temperature. In fact, there is an extremely wide range of criteria from
which to chooseranging from tests at LAST plus 70 F for full dependence on the initiation
barrier as per A709-to tests at LAST minus 120 F (or LAST minus 200 F for very thick plates)
to assure complete freedom from brittle fracture, no matter what, corresponding to FTP on the
fracture analysis diagram.
The FTP level of toughness is rarely needed for structures outside of military
applications where there is the possibility of explosive attack. At the other extreme, using impact
tests to control static fracture toughness relative to the initiation barrier is highly questionable.
Thus, the practical range of choice is somewhat narrower, ranging from tests at the lowest
0
anticipated service temperature to tests at approximately LAST minus 6 0 F, the latter criterion
corresponding to FTE.
The no-break level of notch toughness, as provided by Charpy testing at the lowest
anticipated service temperature, is appropriate where welding restraint and/or stress
concentration lead to the possibility of initial pop-in fractures occurring in a yield tension stress
field, and where the consequences of the resulting brittle fracture justify at least modest
preventive measures.
For subzero air temperatures, heat-treated (i.e., normalized or Q&T) steels may be
required to provide this same level of toughness.
For tubular joint cans, since the hot spot regions operate at stresses above yield, the
added toughness provided by testing to the no-break equivalent at 18 F (10 C) below LAST
(i.e., service at 28 F above NDT) is appropriate. In API Spec 2H, Charpy testing at -40 F
provides this for all non-Arctic conditions of service, plus a cushion for material degradation
during fabrication and welding, as described later.
Still higher levels of toughness may be needed for elements with little or no redundancy,
to avoid fracture of a critical element which could lead to catastrophic total collapse while the
structure was manned or capable of pollutionparticularly where adverse combinations of
thickness, cold work, restraint, stress concentration, and dynamic loading also exist. Such
critical applications may warrant Charpy testing at 36-72 F (20-40 C) below service
temperature, providing for arrest of large running fractures in material of limited thickness, at
stress levels ranging from nominal design allowables to above yield, respectively.
As an alternative to testing at temperatures below the service temperature, and to avoid
some of the previously mentioned difficulty with Charpy energy readings for high strength
levels, some specifications call for 50 percent shear fracture appearance as a more direct measure
of notch ductility. This may also be viewed as the no-break level of toughness (ref. 30).

7.2.7 Notch Toughness of Welds

Although it is readily recognized that notch toughness of welds is important, use of
Charpy tests for characterization of weld metal is often debatable. Fracture appearance-
-percentage shear fracture versus percentage brittle cleavage-is particularly difficult to interpret.
For some types of welds, e.g., some of the early flux-cored wire types, grossly overmatched yield
strength can result in welds which meet the 15 ft-lb energy criteria while still producing a flat,
basically brittle, fracture. Welding procedures which seem acceptable on the basis of Charpy
energy have been the cause of great consternation when evaluated subsequently on the basis of
fracture-mechanics-type testing, e.g., CTOD. At present, the most realistic tests involve
complete welded joints subject to static (wide plate) or dynamic (explosion bulge) loading.
Unfortunately, these are very expensive research-type procedures, and we end up falling back on
the familiar, but fallible, Charpy test for routine work.
 305

The Battelle classification (ref. 41) for weld metal impacts is shown in Figure 7.29.
Many of the AWS filler metals commonly used in welding mild and intermediate-strength steels
0
exhibit 20 ft-lbs at - 2 0 F as a basis of classification, corresponding to the Class 2 level of
toughness. Examples include E6010 and E7018 manual electrodes, F72-EXXX submerged arc
wire-flux combinations, and E60S-2 (now E70S-2) wire for gas-metal arc welding. Actual
results from contractors' procedure qualification test plates show considerable scatter, generally
on the safe side of the intended Class 2 trend line. Dilution of the weld alloy with melted base
metal, and the use of larger passes (high heat input) than in the AWS standard test, can
sometimes result in welding electrodes falling short of their as-classified level of toughness.

, 0 0
o = MANUAL 7018 a SHORT ARC 6 0 S - 2
= AUTOMATIC SUBMERGED ARC

Fig. 7.29. Battelle classification

for weld metal impacts, with
welding procedure qualification
test data from a U.S. Gulf Coast
fabricator, late 1960' s.

TEMPERATURE F

The Class 2 level of toughness for weld metal is appropriate for service at seawater
temperatures, even where criteria for base metal may be less stringent. Welds are almost always
the site of unintended and undetected initial defects, and are often at the site of stress
concentrations as well. In addition, a margin is provided for the fallability of the Charpy test in
this application.
For more severe conditions of service, e.g., subzero air temperatures, welding materials
approximating the Class 3 level of toughness are commercially available under standard AWS
specifications. Examples include E8018-C2 manual electrodes, F76-EXXXX submerged arc
wire-flux combinations, and other nickel-bearing materials. The Class 4 level of weld toughness
is generally reserved for military applications.
In addition to base metal, and weld metal, the weld heat-affected zone deserves scrutiny
in terms of notch toughness. Extensive areas of embrittled material can lead to catastrophic
fractures. Localized embrittled spots are often the result of unavoidable variations in
workmanship and technique. While these can lead to initial pop-in cracking, catastrophic brittle
fracture may be avoided if the base metal, weld, and heat-affected zones all have a generally high
level of toughness.
Different parts of the heat-affected zone see different thermal cycles. Right at the fusion
line, the material is heated almost to the melting point, and it is here that degraded properties are
most often found. At intermediate distances, the material is transformed to a normalized or
quenched microstructure, depending on cooling rates; in multipass welds, part of this zone is
subsequently tempered. This treatment can often improve the notch toughness over that of as-
306

rolled base plate; while for heat-treated plates, the weld heat input must be controlled in order to
break even. Finally, at greater distances from the weld, material which has been previously cold
worked may be subject to strain-aging embrittlement.
In order to evaluate all these areas, some user specifications call for Charpy tests with
the notch centered at the fusion line and at 1mm, 2mm, 5mm, and 8mm into the heat-affected
zone. Ideally, the notch should be parallel to the fusion line; to facilitate this, test plates may be
welded with one side of the groove prepared normal to the plate surface.
Target levels of heat-affected zone toughness, energy and/or percentage shear, are
subject to the same considerations as those for the base plate. However, it is often prudent to
reserve a "kitty" for possible degradation by ordering steel with more notch toughness than is
ultimately required after fabrication and welding. API provisions for tubular joint cans (ref. 40)
anticipate degradation of 18 F (10 C) for D/t down to 30 (forming strains up to 3%), and 36 F
(20 C) for D/t down to 20 (strains to 5%). Manufacturer's data indicate degradation of at least
this magnitude for the most popular heat-treated structural steels (ref. 42).

7.2.8 Code Provisions

The present (1990) AWS Structural Welding Code does not specifically address fracture
toughness requirements for tubular structures. Revisions proposed for the 1992 edition are given
in Appendix III. These are similar to provisions which have been in place for over a decade in
API RP 2A. Both are consistent with the foregoing discussion of historical Charpy-based criteria
for fracture-safe design. A more progressive CTOD-based approach to fracture control is
discussed by the author in Reference 43.

7.3 LAMELLAR TEARING

During the hot rolling of steel plates from ingots, microscopic inclusions tend to get
flattened out into planes of weakness. As a result, the through-thickness, or short-transverse,
properties are often degraded with respect to those measured in conventional longitudinal tests
(ref. 44). Tensile values can range as low as 20% of the usual mill test results, while impact
energy can be as low as 10%, even where there is no pre-existing lamination detectable by
normal inspection procedures.
At tee welds in simple tubular connections, through-thickness loads are imposed upon
the joint-can material by the incoming braces. Some of these have failed in the lamellar tearing
mode. Most of the failures have been during fabrication. However, lamellar tearing has also
been observed as the mode of failure in offshore collision damage and in laboratory fatigue tests.
Resistance to lamellar tearing demands a certain level of through-thickness ductility. In
welded structures, this demand increases with welding restraint, heat input, and the amount of
reworking. The available through-thickness ductility can be measured by a number of different
tests. The percentage reduction of area (RA) as measured by the Wold specimen, Figure 7.30,
has been correlated with practical experience as follows (refs. 45):

- Extensive tearing with RA of 1-6 percent

- Limited tearing with RA of 7-10 percent
- Tearing possible under adverse conditions with RA up to 20 percent
- Complete freedom from tearing with RA over 30 percent

Test results for ordinary quality structural steels are shown in the right-hand part of
figure 7.30. Although many of these steels appear to be fairly susceptible, the actual incidence of
lamellar tearing problems in platform construction is much lower than this suggests-less than 5
percent in the writer's experience.
 307

\ RANGE
MEAN VALUE
51 OF
VALUES

1 P R O B A B L E F R E E D O M FROM
L A M E L L A R T E A R I N G EVEN IN
HIGHLY R E S T R A I N E D S T R U C T .

POSSIBLE OCCURENCE TEARING|

TEST IN HIGHLY R E S T R A I N STRUCT
1 11 1
PLATE '

PREPARATION OF WOLD SPECIMEN

Fig. 7.30. Test for through-thickness reduction of area (RA), typical data, and interpretation (ref. 45).

Reduced sulfur content, vacuum degassing, and calcium-argon treatment are known to
produce cleaner steels, less susceptible to lamellar tearing.
A complementary approach to lamellar tearing combines a number of design and
materials measures. In designing complex stiffened joints, using pass-through elements for the
more critical members in effect bypasses the problem. In simple tubular joints, the brace stub
end and connecting weld can be made of low-yield material, providing a safety valve to limit the
local transverse stresses. Joint cans and other critical material can be ordered with through-
thickness RA target values of 20-25 percent, as in supplement S4 of API Spec 2H. Ultrasonic
inspection of the plate to ASTM A578 Level II can be used to reorient the joint can so as to avoid
having the branch member footprint fall in suspect areas with pre-existing laminations. Finally,
when occasional lamellar tears are found during fabrication, they are gouged out, repaired by
welding, and reinspected. In most cases, material which survives the rigors of fabrication by
welding (thermal strains well past yield) will not fail by premature lamellar tearing in normal
service with stresses below yield.
In dealing with lamellar tearing, one should be aware of the "Watergate effect": once
you start digging, all sorts of nasty things keep turning up. Efforts to repair a small flaw have
been known to trigger extensive additional cracking. Fabrication fixes are most successful when
applied in a preventive sense. Control of heat input and welding sequence to limit shrinkage
strains has been quite effective. A buttering layer of low-strength weld metal can be applied to
suspect surfaces prior to making the final joint weld, so that shrinkage strains occur in the
buttering layer rather than the more sensitive base metal. Finally, suspect material can be
completely gouged out and replaced with more nearly isotropic weld metal, along member and
stiffener "footprints" which are subjected to through-thickness loading; this is often more
efficient than repeated spot repairs after the fact.
As might be inferred from the dated references, lamellar tearing has essentially been a
solved problem for well over a decade. It occasionally makes news when someone fails to apply
the known technology.

7.4 ROLE O F REDUNDANCY

Clause 10.5.1.7 of the AWS Code calls for a 3 3 % reduction in allowable stresses
(design static strength) for critical connections whose sole failure would be catastrophic.
308

Similarly, clause 10.7.4.3 limits the fatigue damage ratio to one-third (corresponding to a 25%
reduction in the basic allowable hot spot stress) for critical members. Structural redundancy has
already been alluded to in the preceding discussion of fracture-safe design. Here we shall
attempt to bring the Design-Inspection-Redundancy Triangle (DIRT, ref. 46) into sharper focus,
with examples from the author's background in marine structures.

7.4.1 Structural Redundancy Concepts

With a few exceptions, the tubular space frame of a fixed offshore structure typically has
a multiplicity of load paths such that the sole failure of a single member does not lead
immediately to catastrophic failure. While the beneficial effect of redundancy has been
recognized for some time, this has largely been treated qualitatively, or in terms of specific
examples. We shall define a couple of useful terms with which to quantify the degree of
redundancyboth for simple systems with Nj^p identical parallel load carrying elements, and for
more complex structures in which the effect of member failure must be established by structural
analysis of intact vs. damaged structure. The damage being considered here is complete loss of a
member, as by brittle fracture, fatigue failure, or collision damage.
The redundancy factor RF ranges from zero for weakest link systems to very large
numbers for damage tolerant structures:

damaged s t r e n g t h
= N 1 1 ) ( 7
^ LP " = strength loss

Values of RF less than unity imply a high likelihood that initial failure will progress to total
collapse in the presence of nominal loads. Very high values of RF require extreme overloads for
total failure, assuming the intact design was adequate.
The damaged strength rating DSR expresses the remaining strength of the damaged
structure, as a fraction of its original intact strength:
N
damaged s t r e n g t h LP - 1
D SR = =
intact strength N L p (7.2)

Values of DSR range from zero for weakest link systems to unity for damage tolerant systems.
Figure 7.31 gives RF and DSR values for a typical Gulf of Mexico 8-pile offshore
platform, based on simplified strength analyses (both elastic and plastic) for the various damage
cases shown. The 10% label on the jacket leg relates its ultimate capacity as a portal frame to
that of an adjoining diagonal brace. Note that a plastic mechanism analysis with full load
redistribution and portal development, yields higher DSR and RF than elastic analysis. In
evaluating the risk of collapse due to overload of a damaged structure, plastic analysis seems
most appropriate. Computer programs are now available to perform such inelastic frame analysis
to the desired degree of accuracy (refs. 47, 48). However, for many purposes, classification
based on simplified analysis should suffice.
Results of plastic analysis may be applied in terms of a limit state design approach such
as that found in the DNV rules (ref. 49). In addition to the partial safety factor on strength (or
resistance factor in American LRFD practice), capacity is further modified by , a reserve
strength factor, originally intended to reflect unfavorable post buckling behavior with values less
than unity. If the post failure resistance is being used, e.g., from a collapse analysis, a value of
1.0 applies.
 309

7.4.2 Fail-Safe-While-Manned
An offsetting factor * Q may be established during initial design, as a function of
redundancy, to ensure that complete failure of a single member by fracture will not reduce the
ultimate strength of the structure below the acceptable limit as defined by the code, for periods
while the platform is manned. This approach will be referred to as the fail-safe-while-manned
fracture control strategy.
The lateral load required for collapse of an offshore platform jacket is typically about
1.75 times its design load, (ref. 50), indicating a plastic reserve of 40% over the nominal design
safety factor of 1.25. Using this in terms of DNV rules, which penalize plastic strength, the
corresponding usable reserve strength is reduced to 24%, or 0 of 1.24. This results in the
criteria shown in Figure 7.32 (refs. 51, 52). For Gulf of Mexico platforms, which are de-manned
for hurricanes and only see 40-50% of their design load in winter storms (i.e., while manned),
almost any conventional jacket would provide enough redundancy for this fail-safe-while-
manned status.
REDUNDANCY FACTOR

0 .5 1.0 2 3 4 5 10 o

= 1.24

Fig. 7.32. Fail-safe-while-

manned fracture control
strategy.
0 .25 .50 .75 1.0
DAMAGED STRENGTH RATING
310

For North Atlantic service, structures may be exposed to the design storm while
manned, and there is less margin for degradation. What exists is provided by the system plastic
reserve ( * Q ) or by designing for more severe conditions (longer return interval) than that
mandated by the regulations. Redundancy factors of 3.8 or better, or providing additional reserve
strength to offset lower redundancy factors (e.g., by using AWS clause 10.5.1.7) appears
necessary to achieve fail-safe-while-manned status. Where this has not been achieved, stringent
inspection and maintenance requirements would seem to be only prudent. Where fail-safe
redundancy has been provided, a reduction in inspection requirements is justified.

7.4.3 Progressive Fatigue Damage

The effect of fatigue failures on collapse risk may be illustrated by reference to a case
study (ref. 53) of an early North Sea steel structure. A typical target fatigue life for such a
structure is 200 years. As originally analyzed (neglecting the effect of directional wave
spreading), this structure did not meet the target, having the following distribution of fatigue
lives among six parallel members in the critical elevation within the structure (see Figure 7.33):

CALCULATED MEDIAN
D=l PER AWS TO FAILURE
1 member @ 20 yr 100 yr
2 members @ 60 yr 300 yr
3 members @ over 200 yr over 1000 yr

6' t o 9' .

2' to 4' -
Fig. 7.33. Example platform
showing members used in
progressive damage study.
 311

The large spread between calculated life (at a damage ratio of unity) and median time to
failure illustrates the uncertainty inherent in fatigue predictions. Although only 3 % of the test
data fall on the unsafe side of the design S-N curve, other uncertainties (e.g., applied loads, SCF,
environmental effects) have been estimated to increase the single member probability of failure
to 15% at the calculated life. Other authors (e.g., Wirsching) have been less pessimistic in this
regard.
Restricting our attention to braces in the critical elevation being studied, there are
several members competing to be the first to fail, and the median time for the first failure is 50
years instead of 100. One failure, because of multiple parallel load paths, still leaves the
structure with 80% of its original resistance to lateral loads. With the decreased mean resistance
and increased scatter, the risk rate (annual probability of failure by overload) has now increased
to about twice its original value. With further passage of time, the structure gets progressively
weaker and both the risk rate and the rate of fatigue damage accelerate (see Figure 7.34).
YEAR 0
Pj -1%

YEAR 50 V r
-2%
\
Pf J

-VARIATION IN
1
Y E A R 100 j FAILURE LOADS

/ h! /
p t- i o % ~ ^ v N '
VARIATION IN
Fig. 7 . 3 4 . Deteriorating
/
ANNUAL EXTREME
DYNAMIC BASE
/ J strength of redundant hybrid
SHEAR
\' structure-based on 20-year
\
/ *\ I calculated life.

V
1/50 YR DESIGN
I
S\ /\ sis
\

10 20 5 0 100 2 0 0
% OF DESIGN FORCE

In fail-safe redundant structures, with multiple parallel load paths, the failure of a single
brace does not lead immediately to collapse of the structure. Collapse occurs only when an
extreme value of the applied load exceeds the remaining strength of the structure, which is
reduced by progressive fatigue damage. Over the 20-year life, the added risk of collapse due to
fatigue is on the order of half the original risk of collapse due to overload.
So far the risk estimate assumes no in-service inspection; that is, any failures would be
allowed to progress to ultimate collapse without any kind of intervention. In practice this is not
what happens. For the example calculation of the expected interval between first failure and
subsequent failures is 11 years, with 95% probability that the interval will be at least one year-
-that is, long enough to permit the damage (complete failure of the one brace) to be found in a
periodic inspection and repaired. This interval is represented by a half-normal distribution as
shown in Figure 7.35.
PROGRESSIVE COLLAPSE
GIVEN INITIAL FAILURE

I N I T I A L FAILURE
COMPLEX STRUCTURE

0 .05 .10 .25 .5 1 2.5 5 10 25 50

Fig. 7.35.
INSPECTION
PERIOD TIME - A S FRACTION OF CALCULATED

Interval between initial failure and progressive collapse.

LIFE
312

For structures with less redundancy (lower DSR) than the example structure, progressive
m
damage is accelerated in proportion to ( D S R ) " , where m is the exponent of the S-N curve, as
shown in Figure 7.36. For a redundancy factor of 3, there is a reasonable chance that routine
inspection at intervals of 5-10% of the calculated life, and after each occurrence of the design
storm, would be able to detect complete loss of one brace before further progressive failure has
occurred. REDUNDANCY FACTOR

Fig. 7.36. Effect of redundancy

on interval between initial
failure and subsequent failures.

.25 ,50 .75 1.00

DAMAGED STRENGTH RATING

Given inspection opportunity, the lifetime risk of catastrophic total collapse due to
fatigue is reduced to less than 0.1%. In most cases of initial failure, detection would permit the
structure to be either repaired or abandoned in an orderly fashion. These outcomes are shown in
the event tree of Figure 7.37. As their economic impact was significant for the structure in
question, the risk was further reduced by upgrading the weld profiles (by grinding) at the critical

(3%)

1%)

(<0.1)

7.5 NON-DESTRUCTIVE INSPECTION CRITERIA

Because of the complex geometry of welded tubular connections, standard radiographic

testing methods are not applicable, and we are left with ultrasonic testing (UT) as the only viable
method of checking the internal quality of completed welds.
In this section, the basis for ultrasonic reject criteria appearing in API RP 2X (ref. 54)
and AWS D l . l is described. These were initially developed in the late 1960's (ref. 55) and early
1970's (refs. 56 and 57). In this section, various criteria will be examined from the standpoint of
previous codes, service experience with offshore structures, fatigue tests of tubular connections,
notch stress analysis, fracture mechanics, and practicality.
 313

AWS level R (API level A) represents traditional workmanship-based standards. AWS

level X (API level C) has been termed an experience-based fitness-for-purpose standard. Finally,
we shall discuss engmeering fitness-for-purpose assessments (API level F).

7.5.1 Introduction to Ultrasonic Testing

The concept of ultrasonic testing has been described elsewhere (refs. 58 to 61). It works
very much like radar. Typically, the system consists of the instrument (a black box containing
all the electronic elements and featuring a display screen along with a number of control knobs),
the transducer, and the human operator. In the instrument, a pulser generates intermittent short
bursts of electrical energy. In the transducer, a piezoelectric crystal converts the electonic burst
into the pulse of ultrasound; that is, mechanical vibrations above the range of human hearing.
Returning echos are converted back into an electrical signal. In the receiver circuits of the
instrument, this signal is amplified and displayed on the cathode ray tube. The vertical height of
the indication represents the intensity of the echo. The horizontal sweep represents time, and the
position of the indication on the screen can be interpreted as the length of the sound path, or
location of the reflector. Interpretation is provided by the all too human operator.
Shear wave inspection of welds is fairly complex, even for a simple butt-joint. The
piezoelectric crystal is usually mounted in a lucite wedge, such that the sound strikes the steel
surface at an angle. Continuity of sound transmission at the contact surface is provided by a fluid
couplant. Upon entering the steel, part of the sound energy undergoes mode conversion, from a
longitudinal compression wave, to a transverse shear wave. The shear wave is refracted so that it
0
travels at an angle to the plate surface, typically ranging from 45 to 8 0 (as measured from the
normal), and it tends to be concentrated in a beam. Should this poorly focused beam encounter a
small discontinuity in the metal, part of the sound scattered or reflected. Some of the reflected
energy will retrace its path to the transducer, where it is picked up and represented as a vertical
indication on the cathode ray tube. Provided other stray reflections do not confuse the picture,
and provided the operator notices and correctly interprets the indication, the discontinuity will
have been detected. Geometry provides the needed relationships between metal surface distance,
depth below the surface, beam angle, and length of the sound path.
Identification of a reflector or bona fide flaw requires considerable skill on the part of
the ultrasonic technician. By turning up the gain in the instrument, minor indications can be
made to appear large and rejectable. Unless the system has been properly calibrated, such false
echos may be interpreted as defects.
Despite its dependency on the human operator, ultrasonic inspection of welds is
significant for several basic attributes in which it is superior to other testing methods (e.g.,
radiography).
First, ultrasonic testing more nearly indicates the actual severity of the various types of
defects, because it depends upon signals reflected from areas. In descending order, the severity
of defects may be listed as:

cracks
incomplete fusion
inadequate penetration
slag
porosity

This effect is heightened if the testing can be arranged so that sound path during inspection is
nearly parallel to the stress path in service. Ultrasonic testing is more sensitive to the more
serious types of defects because they are typified by relative large surface areas transverse to the
stress path. Radiography, on the other hand, records volumetric differences in density. A tight
314

crack, unless it is nearly parallel to the rays, will often not be detected. Porosity shows up
prominently on radiographic film, although well-rounded gas pockets are not detrimental to the
performance of a welded joint until they exceed 8 to 12 percent of the cross section.
The second attribute of ultrasonic testing is its ability to locate defects in three
dimension. This permits more accurate detect assessment and removal.
The third attribute is the speed with which an ultrasonic test can be conducted, and its
simplicity as far as personnel hazards are concerned. As opposed to radiography, there are no
radiation hazards and the shock hazards associated with using the equipment are minimal.
Fourthly, and perhaps most importantly for our discussion here, ultrasonics is perhaps
the only method of below-the-surface, non-destructive inspection applicable to the complex
geometry and lack of back-side access, which occurs in tubular -, Y-, and K-connections.

7.5.2 Traditional Workmanship Basis

Radiography is the most widely used method for internal inspection of structural welds.
The initial consensus standards for radiographic reject criteria (e.g., ref. 62) were based in large
part on readily achievable standards of workmanship. Radiography is most sensitive to
volumetric defects, such as porosity and slag inclusions, so these receive the most attention.
Since crack-like defects are more difficult to find, and even harder to size, any detectable crack is
rejected.
Typically, radiography is applied to critical butt welds, e.g., girder splices in bridges, or
seams in pressure vessels. Often, the surfaces of these welds are ground smooth to enhance their
fatigue performance. With defect-free welds, this performance can approach that of plain
unwelded plate (S-N curve A in the AWS Code). Under these circumstances, tiny internal
defects are often the initiation site for fatigue failures, and a slightly lower performance is used
for design, e.g., S-N curve in AWS. This curve was established from testing welds which met
the radiographic workmanship standard (ref. 63). Thus, what was initially a workmanship
standard has taken on some aspects of fitness-for-purpose, in that it is associated with a defined
level of performance.
The first ultrasonic testing (UT) procedures in AWS were published in the 1969 Bridge
and Building Codes, and now appear in Section 6 of the combined D l . l Structural Welding
Code. They were developed to be the "equivalent alternative" of radiography, and the initial
field trials were in comparison to radiography (ref. 64). The amplitude method of flaw sizing
was adopted, including increased sensitivity for thinner sections to parallel radiography's
sensitivity of 2% of thickness. Additional bias on the safe side was included to compensate for
the inherent scatter in UT results. Elaborate "cookbook" procedures were codified, so that
different technicians and repeat inspections should produce similar findings.
It was recognized early on that difficulties would arise in the application of the AWS
"cookbook" procedures to the curved surfaces and more complex geometries of tubular joints
(ref. 65). Such applications have been excluded by a cautionary note in every edition of D l . 1.
However, by the late 1970's, development of an offshore industry consensus had progressed to
the point where UT procedures and criteria for tubular structures were drafted for inclusion in
Section 10 of the AWS Code (ref. 66).
Procedurally, the Code provides that a variety of different techniques may be used, if
they are...

(1) prepared by an ASNT Level III expert (ref. 67),

(2) and approved by the Engineer, provided
(3) technicians can qualify in practical trials using the procedure.
 315

The need for a transfer mechanism in calibrating the amplitude method, and for alternative flaw
sizing techniques, are recognized.
In the tubular structures section of AWS, level R ultrasonic testing is intended to be the
equivalent of radiography. For the body of the weld, the amplitude calibration standard and
acceptance criteria for defect length roughly parallel those of ASME (ref. 62). Where the welds
have been ground to enhance their fatigue performance, these tight workmanship based criteria
for internal defects are appropriate. Somewhat looser criteria are applied at the inaccessible root
of single-sided welds in tubular , K, and Y-connections, as described in the following sections
of this chapter; these are based on experience with simple unstiffened connections.

7.5.3 Inspection of Welds in Tubular , Y, and Connections

One of the requirements for satisfactory performance of tubular joints is that the welds
develop the strength of the material joined. Designers generally assume that a full penetration
weld will be provided. Due to the fact that access to the inside of small diameter branch
members is denied, the weld must be made from the outside, without backgouging. Many codes
prohibit groove welds made from one side only. Codes which do permit such welds recognize
the fact that incomplete penetration is the usual result, by specifying a deduction for effective
weld size and by using the reduced allowable stress applicable to fillet welds. Reinforcing such a
weld with an external fillet might increase its static strength to the point where the calculated
load could be carried; however, the undesirable notch at the root is not eliminated.
This is the motivation behind ultrasonic inspection of the critical tee welds in tubular
connections. Due to the complex geometry of tubular joints, such welds can not be inspected by
other methods, such as radiography. A paradox of early tubular construction was that the most
highly stressed welds in the space frame were also the most defect proneyet they could not be
inspected after the fact. Quality control was limited to certification of welders, checking the fit-
up of joints before welding, and spot visual observation of the work in progress. In desperation,
trepanning of completed welds was sometimes done. Ultrasonic techniques for inspection of
these welds during fabrication was developed to fill this need.
The complex geometry of a tubular joint is shown in Figure 7.38. For purposes of
ultrasonic inspection, it suffices to consider each short segment of the weld, one at a time, along
with a few inches of the attached material. To a first approximation, this restricted area can be
usefully thought of as an intersection of two flat platesthe primary variables being the local
dihedral angle (with the corresponding orientation of suspected fusion defects), and the
thicknesses involved. Effects of curvature are secondary, although they must eventually be
considered. In inspecting a weld, it is desired to have the probing sound beam nearly at right
angles to the weld axis; by moving the transducer towards and away from the weld, the beam can
be made to sweep the entire weld volume. When inspecting the end of a diagonal brace, this
means that the beam is at some anglethe cant angleto the brace centerline; curvature effects
cause the sound beam to deviate from the orientation of the transducer, as indicated by the skew
angle. Other curvature effects include a tendency for the skip distance AC to stretch out and
beam de-focusing effects which may occur at the convex contact surface.
Most practical testing organizations use some variation of a graphic overlay for keeping
track of the sound beam vis-a-vis local weld geometry in ultrasonic inspection. The beam angle,
the metal thickness and the beam path are plotted on the base card. A scale drawing of the weld
is made on a transparent slide. Moving one relative to the other simulates the back and forth
motion of the transducer used to scan the weld. If the cathode ray tube display on the instrument
is calibrated so as to read metal surface distance, defect location is readily obtained.
Calibration is a key step in ultrasonic weld inspection. The entire testing system-
-instrument, transducer (search unit), interconnecting apparatus, auxiliary apparatus, and the
human operatormust be checked out as a unit. Basic checks can be made using the
316

Fig. 7.38. Parameters associated with geometry of pipe intersection.

International Institute of Welding test block or similar blocks developed by the various testing
firms. Exit point of the sound beam, incident angle, instrument linearity and sweep calibration
can be checked out using standardized procedures, given in Section 6 of the AWS Code. For
tubular structures, amplitude calibration is less straightforward.

7.5.4 Significance of Discontinuities

Meaningful weld inspection requires that criteria for rejectable flaw sizes be established.
In the late 1960's, no codes were directly applicable to the welding in -, Y- and K-connections.
The closest analog was API 1104, covering pipeline welds which must also be made from the
outside of the tube. This code was reasonably tolerant of root defects.
Fabricators have been quick to point out that techniques for ultrasonic inspection are not
well established, and that interpretation is solely a function of the human operator, with no
permanent record being available for review. Accordingly, many contracts provide that
acceptance or rejection of a weld may be based upon destructive examination. After several
ultrasonic indications have been located, a cutting party is formed. This generally consists of a
workman to do the cutting, the fabricator's quality control representative, the ultrasonic
technician, and the owner's inspector. Metal is removed by arc-gouging or grinding; in the
vicinity of suspected defects, no more than 1/16" (1.5mm)of metal should be removed in each
pass, and all those present are given an opportunity to inspect the cut between passes. Defects
which are thus observed and which exceed the agreed limits confirm the flaw as rejectable.
 317

The calibration standard and reject criteria used for ultrasonic inspection should be
compatible with foregoing practicalities. In order to become a useful tool, ultrasonic inspection
must win the confidence of the people involved in fabrication; to this end, it is desirable to hold
the number of false alarms to a minimum. The goal is to find the really gross defects which
occur in the complete absence of non-destructive testing. The design procedures for tubular
joints and the selection of notch-tough materials for critical areas can accommodate the presence
of small flaws. Due to the redundancy of typical structures, a great deal of benefit is realized
from simply improving the general level of quality, without having to worry about finding every
last flaw.
The "Tentative Shell-SWRI Specification for Ultrasonic Inspection of Welded Tubular
Connections" (ref. 68) established the first such calibration standard in 1967. It was based on
field experience with Shell platforms, along with Southwest Research Institute's background in
ultrasonic testing. Before going to field trials, SWRI prepared mock-up tee welds with three
levels of artificial defects, as shown in Figure 7.39. Level 1 corresponds to the workmanship-
based reject criteria cited for bridges and buildings in the AWS Structurel Code. Level 3
corresponds to the large defect size deemed appropriate for underwater inspection and
engineering critical assessment of offshore structures (ref. 61). Level 2 is an intermediate
criterion. In the summer of 1967, some 25 joints on a platform being fabricated for 340 feet of
water (then a world record) were checked; 8 ultrasonic indications exceeding level 1 were noted,
of which two exceeded level 2. The indications exceeding level 2 were verified as rejectable
defects by the cutting party. However, excavation of the level 1 indications produced equivocal
results; rejectable defects were not positively verified, while the arc-gouging produced shrinkage
cracks and burn-thru condition which were difficult to repair. As a result of this trial and
subsequent experience, a reject standard of 1/8" height dimension corresponding to level 2 was
adopted as a workable basis for ultrasonic inspection.

OF DEPTH OF
LEV&L F U S I O N I D N1t *OOT
DEFE-CT ( I N ; OEFE-CT(IN)
1 ' / I Co'
2 . 0 8 3 3 /."
3 1/4."

Fig. 739. Tee weld with artificial defects.

The basic ultrasonic standard block for root defects was established as a 1/2-inch thick
plate (similar to typical brace thickness) which contains the standard 1/8" deep surface notch,
whose length (1") was chosen to be at least the width of the ultrasonic beam. The standard block
is made of the same steel and has the same surface preparation (sand-blasted) as the work to be
inspected. Attenuation losses and distance-amplitude corrections can be determined by peaking
318

the signal from the notch with the transducer successively at skips 1, 2, 3, etc. Some UT
technicians also use side-drilled holes as an alternative calibration standard, particularly for
evaluating internal flaws in the body of the weld, away from the root.

7.5.5 Accuracy and Repeatability

In an attempt to account for the effects of curvature, cant angle, and defect orientation, a
set of supplementary standard blocks was made, representing the practical range of these
variables. The original intent was to determine these effects as part of the calibration and apply
them as small corrections to the calibration obtained from the primary standard (the notched flat
plate). To cover all the permutations, some 450 corrections would have to be tabulated. To
make matters worse, initial measurements indicated that these secondary effects were large and
without a rational pattern, even when the results of multiple readings were averaged. Finally,
one of the testing firms made duplicate calibrations on successive days and found the results
were far from reproducible. See Figure 7.40. This non-reproducibility was large enough to
make the significance of apparent effects due to curvature, cant angle, and notch orientation open
to serious questions. As a result, the elaborate secondary calibration procedure was abandoned
and these secondary effects of curvature are generally being neglected.

P T L CALIBRATION
" - I IK. TRANSDUCER V/A MUZ

. : NONREPRODUCtfclLVTY
: - : : (DIFFERENCE OF I
. . READING^
* " " f* I
Dfc

: : :
I 2 I 1 SiLieVlO
EFFECT O F
C U ERVE&L TA UT IEVEE T W I CFKA NC ET O
<H\LT
? 8 324 ?
^R

I L LL r ' " r * LI I ( f c A * E D O N A \ i q . Or 7 )
-70 - +IO +
Dfb A T T E N U A T I O N Dfi> G A I N
(STEOMSER S I G N A L ) (WfcAKEK ^ G N A L )

EFFECT O F
B OO
O CANT. ANGLE ONLY

( < 5
-70 -10 + +70 ' V
DR> A T T E N U A T I O N D f t <3 c
C^TEONGE-R S I G N A L ) (WEAKER. blQNA.L)

EFFECT O F
NOTCH ORIENTATION ONLY

-7 -
(A.V<q. O F 7)
DR> A T T E N U A T I O N D 6 GAIN
(5TI?ONGEIZ DISMAL) (WEAKEN SIGNAL)

F i g . 7.40. Results with supplementary standard blocks.

The non-reproducibility limits the accuracy to which amplitude calibration can be used
to judge flaw size, even where calibration is on a realistic mock-up containing the reference level
flaw. Errors exceeded a factor of 2 almost 30% of the time. While beam boundary techniques
 319

can be used to trace larger flaws, the 1/8" flaw height is at the lower bound of its applicability.
Thus, a combination of flaw sizing techniques must be used and a "judgement area" between
clearly acceptable indications and clearly rejectable flaws should be anticipated in the criteria, as
shown in Figure 7.41.
ROOT DEFECTS (DETAILS A & ONLY) INTERNAL DISCONTINUITIES
D E F E C T S IN BACK U P W E L D S A L L O T H E R W E L D S
( D E T S C a D) A R E T O B E IGNORED

R E J E C T *

R E J E C T *
JUDGEMENT
LU X VU S E * Y J U D G E M E N T
s
NO DEFECT
2
CL NO DEFECT
<

' '> 2 4I 8 oo 0 l /c 2 I 2 4 co
INCHES INCHES
APPARENT LENGTH APPARENT LENGTH

S H E L L - S W R I L E V E L 2 CALIBRATION STANDARD
/" DEEP X I" L O N G NOTCH IN P L A T E SURFACE

'FLAW IN W E L D * SUBJECT T O CONFIRMATION DURING REPAIR

I INCH = 2 5 m m

Fig. 7.41. Guideline for ultrasonic weld inspection.

For tubular -, Y-, and K-connections, complete joint penetration groove welds have in
essence been redefined as welds which do not limit performance of the completed tubular
structure, from either static or fatigue standpoints. In simple connections and many reinforced
connections, the highest stress occurs at the outside surface of the tube due to the Poisson's ratio
effect and continuity of the branch member with the chord. Thus, a messy back-up weld or root
pass is not as critical as the external weld profile. Even welds with gross (50 percent) lack of
penetration at the root have failed elsewhere under static and fatigue loading. Figure 7.42 shows
an example from a small scale tubular joint fatigue tested in 1966 at SWRI. Similar results were
found later in full scale tests at Berkeley (ref. 11). Thus, less stringent reject criteria apply in the
root area of single-side welds of -, Y- and K-connnections, than apply elsewhere.

GUE
ACKS

FAILURE
SURFACE

Fig. 7.42. Sections taken from the brace to chord weldment of SRI T-joint after failure.

Considering the inaccuracies inherent in ultrasonic testing, qualified technicians do

reasonable well in being able to find 70% of the defects, with no more than 30% false alarms.
320

The need for some means of non-destructive testing of the critical tubular joint welds is so great
that, if 70% accuracy is all that can be expected, this must be lived with. Many of the errors are
false alarms; thus the expense of cutting out and repairing good welds simply increases the cost
of inspection.
UT operators generally seem to be better at identifying defective welds (including the
location of the defect along the weld, and whether it was above the reject standard or not) than
they are at telling the exact type, size, and location (in the weld profile) of the flaw. A difficulty
rating their performance may arise in trying to judge whether differences between the verbal
descriptions of the flaw and the results of excavation are a matter of semantics, intentional
vagueness, or errors.

7.5.6 Further Considerations

Although the Shell-SWRI level 2 reject criteria was initially established in 1967, it was
the subject of continuing review over the next several years. In addition to the practical
considerations which dominated the initial selection, these reviews have been conducted in terms
of notch stress analysis, fracture mechanics, and the emerging philosophy of Fitness-for-Purpose.
The philosophy here is to set reject criteria at the level where discontinuities begin to
adversely affect performance of the welded joint, reserving an allowance for possible
inaccuracies in the inspection. Emphasis is on fatigue performance.

TUBULAR JOINTS
:
- | - . 5 fOJjSlH

BOUWKAMP
THIN Ox

TEE FILLET WELDS

(MUNSE)

BOUWKAMI
THICK

TEE FILLET WELDS

FULL
REVERSAL

S / S u N E T SECTION S T R E S S
MILD S T E E L IN A I R
(C) (PETRSON)

9 7 s 2
I0 10 10 I0'
CYCLES TO FAILURE

Fig. 7.43. Notch stress analysis, (a) Geometry of internal notch, (b) Fatigue strength reduction factor
K f as a function of size, (c) S-N curve derived for K f of 4.
 321

In notch stress analysis (ref. 69), we start with elliptical flaws as shown in Figure
7.43(a). As the flaw becomes sharper, the theoretical stress concentration K t increases.
However, for mild and intermediate-strength steel having a normal grain structure, there is a
limiting value of notch acuity which the material can "feel". This phenomenon is reflected in a
notch sensitivity factor, which is a function of flaw tip radius (see the equations in Fig. 7.1).
This results in a strength reduction factor Kf for fatigue which is less than K t , particularly for
very small, sharp flaws. For notches in the size range of 1/16 to 1/8-inch (1.5 to 3mm), the
limiting value of Kf is around 3 or 4, as shown in Figure 7.43(b)
In the low cycle fatigue range, where joints are loaded close to their ultimate strength, it
is instructive to plot the S-N curve and test data on a relative load basis, as is done in Figure
7.43(c). The S-N curve shown was derived by Peterson for Kf of 4, using StoweH's equations
above yield (ref. 70). Test data for tubular connections and other welded joints come from
various sources as shown in the figure. The curve forms a lower bound to the data, similar to the
way the design S-N curve was drawn. This suggests that to permit flaws larger than 1/8 inch (Kf
larger than 4) in practice would be inconsistent with the design S-N curve and its data base, and
thus clearly undesirable.
TUBULAR JOINTS FATIGUE DATA
A U K - E E C L A R G E SCALE TESTS
SRI PVRC
SRI SHELL - PAN AM
MGS-C
SEDCO
RDM-NI8BERING
- SEA QUEST FAILURE
+ A537 BUTT WELDS IN SEA WATER
RANGE FROM MUNSE WELD TESTS
BOUWKAMP PHASE 1
BOUWKAMP PHASE 2
10000 x BOUWKAMP & BECKER
TOPRAC

CL W EMERYVILLE fCATHOOIC PROTECTED FLEXURAL COUPON
CO I IN SEA WATER
^ CORROSION FATIGUE DATA

1000
AWS X-MODIFIED
cr DESIGN \

<
or RANGE OF FRACTURE
V) MECHANICS FOR

I00j
VARIOUS STEELS Fig. 7.44. Fracture mechanics
3 6 - 5 3 7
consideration of flaws.
m = 3.0
50% M O R T A L I T Y

10 2 4 6 8
I0 I0 I0 I0 lO'O I0'2
CYCLES TO FAILURE

One suspects that the foregoing notch stress analysis dealt heuristically and empirically
with the phenomena which fracture mechanics can now explain more elegantly and rigorously.
Application of fracture mechanics to the problem at hand is shown in Figure 7.44, here
plotted in terms of hot spot strain. An expanded data base is plotted along with a band
representing S-N curves derived from various fracture mechanics characterizations (da/dN and
threshold) of various commonly used structural steels. These were derived by integrating crack
growth from an initial size a Q to a terminal size a^. The assumed Y ^ a Q of 0.1 inch has the same
initial as any of the following:
 322

(1) an internal flaw of 2a = 0.20 inches residing at the hot spot stress but remote from the
surface.
(2) a weld toe undercut .03 inch deep, residing in an unmeasured weld profile notch which
amplifies the hot spot stress by a factor of Y = 1.8.
(3) a root defect, 0.25" deep, residing in a stress not exceeding 60% of the worst hot spot
value (this lower stress is typical of what is found in tubular joint analyses, both
experimental and finite element).

These sizes allow a safety factor of 2 or 3 for inaccuracies of the inspection, when
compared to what the reject criteria specified as a target. As before, the derived S-N relationship
forms a lower bound to the data, suggesting that larger flaws would be clearly detrimental.
Indeed, for strains in the range of .013 to .070% (hot spot stresses of 4 to 20 ksi), the fracture
m e c h a n i c s results suggest that the original A W S - X - m o d i f i e d design curve might be
unconservative, given these crack sizes. The UK Department of Energy design curve (ref. 20)
has a log-log slope m of 3 and is more consistent with the fracture mechanics.
Both the notch stress analysis and fracture mechanics results are, to some extent,
hypothetical. For example, the fracture mechanics results presented ignore the time required for
pre-existing flaws to sharpen into growing cracks. Thus, it is perhaps more satisfying to draw
empirically on fatigue data in which the initial flaw sizes were measured directly.
A Japanese classification of flaw sizes (ref. 71) is shown in Figure 7.45. Flaws of JIS
class 5 would be accepted by the SWRI level 2 ultrasonic inspection criteria. Natural flaws of
this size and smaller are compatible with the AWS design curve "C" for butt welds, as shown in
the top of Figure 7.46; that is, they are no worse than the notch at the edge of the as-welded
surface profile. Only welds falling in JIS class 6 plot on the unsafe side of the design curve; this
class would include cracks larger than 1/8 inch by one inch (3mm 25mm). In the lower figure,
we see that porosity up to a 10 or 20 percent rate of defective area would be consistent with the
same AWS class "C" design curve.
ACCUMULATED LARGE
JIS SMALL DEFECT
DEFECTS SlOOi
CLASS NATURAL
<
DEFECTS
POROSITY
AND S L A G
3/8"
THICKNESS

CURVE C
6
I0
CYCLES TO FAILURE
i KEY TO W E L D QUALITY
] . JIS CLASS I
JIS CLASS 2 8.3
JIS CLASS 4&5
CD x JIS CLASS 6

0
i
; * o
la c
25 o

> 9 c CD CD
\ oo 00 AWS
Zen
CURVE C
E is!
0
WORSE
THAN
ABOVE 3 f

1% 10% 100%
SHOWN FOR 1/2" THICKNESS RATE OF DEFECTIVE AREA
APPLICABLE THICKNESS: OVER 3/8" THRU 3/4" ( POROSITY)

Fig. 7.45. Japanese classification Fig. 7.46. Japanese data - fatigue of

of flaw sizes. welds with defects.
 323

These results are consistent with British work (ref. 72). Welds containing slag
inclusions up to one inch (25mm) long all failed on the safe side of the AWS class "C" design
curve. See Figure 7.47. As an allowable defect size, this would exceed the limits set by most
radiographic codes (except API 1104).
r-40
35

30

1-25

4 9 6 7
I 0 I 0 I 0 I 0

ENDURANCE, CYCLES

Fig. 7.47. Welds containing slag inclusions up to 25mm long.

7.5.7 Experience-Based, Fitness-for-Purpose Code Development

Reference has already been made to the design S-N curves of the AWS Structural
Welding Code, D l . l . This code, and its predecessors, have contained detailed ultrasonic
inspection provisions for bridges and buildings since 1969. However, these were specifically
excluded from applying to tubular structures, being philosophically and technically incompatible
with the evolving offshore industry practice.
Philosophically, the bridge code approach was to provide an "equivalent alternative" to
radiographic inspection, using a "cookbook" procedure which would ensure repeatability of
results for the simple butt weld situations covered. In tubular structures, there was no valid
radiographic precedent, and the variable and complex weld geometry of , Y and connections
precluded defining a concise procedure.
Technically, the difficulty of obtaining a perfect weld root when welding tubular
connections from the outside only, coupled with indications that perfection was not really
required and that unnecessary repairs would likely be counter-productive, led to consideration of
larger permissible flaw sizes. Curvature effects tended to make amplitude measurements less
reproducible than on flat plate and arbitrary DB ratings particularly unreliable. These factors led
practitioners towards performance based criteria, in which a combination of techniques
(amplitude comparison and beam boundary) were used to estimate physical defect sizes, so that
engineering judgement could be applied to their acceptability.
This emphasis on performance rather than cookbook made technician selection
absolutely critical, and led to elaborate qualification trials involving full-scale mock-up joints
with built-in defects.
Throughout most of the 1970's, several industry, AWS, and API committees labored to
produce a consensus document covering ultrasonic inspection of tubular -, Y-, and K-
c o n n e c t i o n s . The end result is API RP 2X, R e c o m m e n d e d P r a c t i c e for U l t r a s o n i c
324

Examination of Offshore S t r u c t u r a l F a b r i c a t i o n and Guidelines for Qualification of

Ultrasonic Technicians. The first edition was published in March 1980 by the Production
Department of the American Petroleum Institute, Dallas, Texas. Contents of this voluminous (61
pages) document include:

QUALIFICATION OF PERSONNEL
General
Prequalifications
Qualification Examinations

TECHNICAL RECOMMENDATIONS
Applicability - Offshore Structures
Attributes and Limitations
Significance of Discontinuities
Equipment
Preparation for Examination
Scanning Techniques
Discontinuity Location
Discontinuity Evaluation
Reporting
Verification
Procedure Qualification and Approval

Since the adoption of API RP 2X, lists of qualified technicians have gradually (and
painfully) been developed by user companies, utilizing proprietary examination samples having
built-in flaws.
With this background available, the 1981 edition of AWS D l . l included condensed, but
similar provisions. After 14 years of practical application, the Shell-SWRI level 2 had become
the Experience-Based, Fitness-for-Purpose criteria (API level C, AWS class X). These criteria
are consistent with as-welded surface profiles where structural redundancy and a "fail safe"
design approach is used (as in typical template-type offshore platforms) and where a reasonable
level of notch toughness is provided for all parts of the welded joints.
These provisions have been reaffirmed in subsequent editions of both API and AWS
standards.

7.5.8 Engineering Fitness-for-Purpose

For situations in which the flaw sizes exceed the original inspection criteria or violate
the original design assumptions, acceptance or rejection may be on the basis of a more rigorous
Engineering Fitness-for-Purpose Analysis as outlined below. This analysis is based on the
equivalent fracture mechanics crack size a as defined in Figure 7.48 (ref. 73), and consists of the
following checks:

(1) The static strength based on remaining net section and specified minimum tensile
strength should be adequate for the applied loads, with the appropriate safety factor.

(2) Brittle fracture should be precluded by satisfying the following inequality for equivalent
crack size (ref. 72-74):
 325

Fig. 7.48. Equivalent crack size a.

. <. *(?) - 2
fa)
where the non-dimensional flaw size is given in Figure 7.49 and these fracture
mechanics, CTOD, or J-integral terms are respectively defined as follows:

critical stress intensity factor at failure (ksi in.)

critical crack tip opening displacement at failure (in.)
contour integral, total (elastic and plastic) energy expended per unit increase in
crack area, critical value at failure
yield strain
yield stress

The fracture mechanics resistance of weld metal, heat-affected zone, and base metal ( 6 C ,
J c , or Kj) should preferably be determined using crack depth, specimen thickness,
loading rate, and temperature representative of the application. (Valid plane strain
conditions are not required if the test thickness is at least that of the application.)

(3) The remaining fatigue life, required to grow a crack from a Q to a terminal crack size a^
(based on static strength or brittle fracture as outlined above) should be adequate.
Where the notch toughness throughout the weldment meets criteria which ensure that a^
will be very large, Figure 7.50 may be used to estimate median remaining fatigue life.
This figure is based on fracture mechanics crack growth calculations using published
"typical" data on da/dN versus , and threshold values, for tests in air (ref. 7). A
326
lower bound safe life comparable to API criteria would incorporate a safety factor of 5
to 8 on this expected median life. Where surface notches are exposed to a corrosive
environment, fatigue life may be further reduced by a factor of 3 to 6.

Fig. 7.49. Normalized flaw size

as a function of peak tensile hot
spot strain.

Engineering fitness-for-purpose flaw analysis is particularly appropriate for evaluating

large flaws discovered during in-service inspections. Although the AWS Code does not provide
for its application during initial construction, the Society has supported the concept by
sponsoring a joint conference on the subject with The Welding Institute (ref. 75). More recent
reviews of the subject can be found in References 76 to 81.

Fig. 7.50. Fracture mechanics

fatigue life for ductile material
in atmospheric service.

s
EXPECTED REMAINING LIFE C Y C L E S T O o f

REFERENCES

1 Structural Welding Code - Steel, AWS Dl.1-88, American Welding Society, Miami, 1988.
2 SIMS-87: Steel in Marine Structures, Proceedings of the International Conference, Delft, June
1987.
3 Proceedings of the Offshore Steels Research Seminar, Cambridge UK, November 1978.
4 Marshall, P. W., Houdremont Lecture: Connections for Tubular Structures, IlW/Pergamon
Press, Boston, July 1984.
5 Steel in Marine Structures, Proceedings of the International Conference, Paris, October 1981.
 327

6 Lawrence, F. V., et al, Estimating the Fatigue Crack Initiation Life of Welds, ASTM STP 648,
pp. 134-145,1978.
7 Hayes, D. J., Fracture Mechanics Based Fatigue Assessment of Tubular Joints: Review of
Potential Applications, Aptech Engineering Services, Palo Alto, CA, June 1981.
8 Gurney, T. R., Fatigue of Welded Structures, Cambridge University Press, 1979.
9 Berge, S., et al, Fatigue Crack Initiation in Weldments of a C-Mn Steel, Paper 6 in Reference 3.
10 Gurney, T. R., The Influence of Thickness on Fatigue Strength of Welded Joints, Proceedings
2nd International Conference on the Behaviour of Off-Shore Structures, BOSS-79, London, Vol.
l,pp. 523-534,1979.
11 Becker, J. G., et al, Fatigue Failure of Welded Tubular Joints, Proceedings Offshore Technology
Conference, OTC 1228, May 1970.
12 Grover, G. L., Fracture Mechanics Based Fatigue Assessment of Tubular Joints: Development
of Analysis Tools, Aptech Engineering Services, Palo Alto, CA, December 1984 (also see
OMAE-89, the Hague).
13 Radenkovic, D., et al, une remarque sur la Conference Houdremont 1984 du point de vue de
l'analyse des contraintes, IIW Doc. XIII-1163-85.
14 Dijkstra, O.D. and Noordhoek, C.,The Effect of Grinding and Special Weld Profile on the
Fatigue Behaviour of Large Scale Tubular Joints, Proceedings Offshore Technology
Conference, OTC 4866, May 1985.
15 Marshall, P. W., Design of Internally Stiffened Joints, Proceedings of the IIW-AIJ International
Meeting on Safety Criteria in Design of Tubular Structures, Tokyo, July 1986 (also see OTC
6093).
16 Marshall, P. W., Advanced Fracture Control Procedures for Deepwater Offshore Platforms,
Welding Journal, January 1990.
17 S.A.E. Fatigue Design Handbook, AE-4, Society of Automotive Engineers, Warrendale, PA,
1968.
18 Merwin, J.E., Fatigue in Welded Tees, Rice University report to Shell Oil, Houston, December
1986 (data also appears in SIMS-87 PS1).
19 Hartt, W. H., et al, Weld Profile and Plate Thickness Effects in Fatigue as Applicable to
Offshore Structures, Florida Atlantic University Report API-87-24 to American Petroleum
Institute, Dallas, May 1989.
20 Snedden, . V., Background to Proposed New Fatigue Design Rules For Steel Welded Joints in
Offshore Structures, AERE Harwell, May 1981.
21 Hudak, S. J., et al, Long Term Corrosion Fatigue of Welded Marine Steels, SSC-326, Ship
Structure Committee, Washington, DC, 1984 (also see OTC 4771).
22 Arockiasamy, M. and Reddy, D., Weld Profile and Thickness Effects on Fatigue of Tubular
Welded Joints, Florida Atlantic University, final report RP 86-19 to API, Dallas, January 1989.
23 Bignonnet, ., Improving the Fatigue Strength of Welded Steel Structures, paper PS4 in
Reference 2, SIMS-87.
24 Haagensen, P. J., Improvement Techniques, in Fatigue Aspects of Structural Design, Delft
University Press, 1989.
25 Lieraude, H. P. et al, The Influence of Cathodic Protection and Post Weld Improvement on the
Fatigue Welded Joints, paper TS40 in reference 2, SIMS-87.
26 Booth, G. S., Constant Amplitude Fatigue Tests on Welded Steel Joints, Performed in Air, in
reference 3.
27 Booth, G. S., Techniques for Improving the Corrosion Fatigue Strength of Plate Welded Joints,
paper TS41 in reference 2, SIMS-87.
28 McHenry, . I., et al, Fracture Control Practices for Metal Structures, NBSIR-79-1623, National
Bureau of Standards, Washington, DC, 1980.
29 Hartbower, C. E., Commentary on the FHWA Fracture Control Plan, Structural Engineering
series No. 5, Vol. Ill, Federal Highway Administration, Washington, DC, June 1978.
30 Marshall, P. W., et al, Materials Problems in Offshore sstructures, Proceedings Offshore
Technology Conference, Houston, OTC Paper 1043, May 1969.
31 Rolfe, S.T. and Barsom, J. M., Fracture and Fatigue Control in Structures, Englewood
Cliffs, NJ, Prentice Hall, 1977.
328

32 Hartbower, C , Reliability of the AASHTO Temperature Shift in Material Toughness Testing,

Federal Highway Administration, Structural Engineering Series No. 7, Washington, DC, FHWA
August 1979.
33 Barsom, J. M., et al, Fracture Control Considerations for Steel Bridges, January 1980.
34 Pellini, W. S. and Puzak, P. O., Fracture Analysis Diagram Procedures for the Fracture-Safe
Engineering Design of Steel Structures, Welding Research Council Bulletin 88, New York,
WRC, 1963.
35 American Society for Testing and Materials, Method for Conducting Drop-Weight Test to
Determine Nil-Ductility Transition Temperature of Ferritic Steels, ASTM E208.
36 Rolfe, S. T., et al, Proposed Fracture Control Guidelines for Welded Ship Hulls, Section 15.4 in
Fracture and Fatigue Control in Structures, Prentice-Hall, 1977.
37 PD 6493, Guidance on Some Methods for the Derivation of Acceptance Levels for Defects in
Fusion Welded Joints, British Standards Institute, 1980.
38 Marshall, P. W., et al, Advanced Fracture Control Procedures for Deepwater Offshore Towers
and Compliant Platforms, Proceedings Offshore Technology Conference, May 1990.
39 American Petroleum Institute, Specification for Carbon Manganese Steel for Offshore Platform
Tubular Joints, API Spec 2H.
40 American Petroleum Institute, API Recommended Practice for Planning, Designing, and
Constructing Fixed Offshore Platforms, 18th Edition API RP 2A, 1989.
41 Masubuchi, K., Interpretive Report on Weld Metal Toughness, Columbus, OH, Battelle
Memorial Institute, July 1965.
42 Armco Steel Corporation, Product Data Book (c. 1978).
43 Marshall, P. W., Advanced Fracture Control Procedures for Deepwater Offshore Towers,
Welding Journal, January 1990.
44 J. . M. Jubb, Lamellar Tearing, Welding Research Council Bulletin 168, New York, WRC,
1971.
45 Wold ., et al, Development of method for measuring susceptibility of steel plate to lamellar
tearing, Sveiseteknikk No. 3, 1972 (also see Proceedings 5th Annual Offshore Technology
Conference, Houston, OTC paper 1915, pp. 891-897,1973).
46 Faulkner, D., et al, The Role of Design, Inspection, and Redundancy in Marine Structural
Reliability, Proceedings of the International Symposium, November 14-16, 1983; National
Academy Press, Washington, DC.
47 Marshall, P. W., et al, Inelastic Dynamic Analysis of Tubular Offshore Structures, Proceedings
Offshore Technology Conference, OTC 2908, May 1977.
48 Marshall, P. W., An Overview of Recent Work on Cyclic Inelastic Behavior and System
Reliability, Proceedings Structural Stability Research Council, 1982.
49 DNV Rules for the Design, Construction, and Inspection of Offshore Structures, Det Norske
Veritas, Hovik, Norway, 1977.
50 Marshall, P. W., et al, Failure Modes of Offshore Structures, Proceedings 1st International
Conference on Behavior of Off-Shore Structures, BOSS-76, Vol. 2, Trondheim, August 1976.
51 Marshall, P. W., Strategy of Monitoring, Inspection, and Repair for Fixed Offshore Platforms,
Proceedings 2nd International Conference on Behavior of Off-Shore Structures, BOSS-79,
London, August 1979.
52 Marshall, P. W., Cost-risk tradeoffs in design, M.I.R.V. and fracture control for offshore
structures, Proceedings 7th ISSC, Paris, August 1979.
53 Marshall, P. W., Failure Modes of Offshore Structures, Part II Fatigue, Proceedings 1st
International Conference on Behavior of Off-Shore Structures, BOSS-76, Vol. 2, Trondheim,
August 1976.
54 API RP 2X, Recommended Practice for Ultrasonic Examination of Offshore Structural
Fabrication and Guidelines for Qualification of Ultrasonic Technicians, 2nd Edition, American
Petroleum Institute, Washington, DC, September 1988.
55 Marshall, P. W., Ultrasonic Inspection Applied to Tubular Joints in Offshore Structures, Shell
Oil OCD Design Note 1, February 1969; also presented at 1969 Group Production R&D
Conference, the Hague, April 1969.
 329

56 Tentative Specification for Ultrasonic Inspection of Welded Tubular Connections, Pecten Viet-
Nam, December 1974.
57 Marshall, P. W. (chairman), API Recommended Practice for Ultrasonic Inspection of Welded
Tubular Structures, Draft #2, May 1976.
58 Marshall, P. W., Experience-Based Fitness-for-Purpose Ultrasonic Reject Criteria for Tubular
Structures, in Welding of Tubular Structures, Proceedings 2nd International Conference,
Boston 1984, Pergamon Press; also IIW doc XV-514-82 in Reference 75.
59 McMaster, R. C , Nondestructive Testing Handbook (2 volumes), Society for Nondestructive
Testing, Ronald Press, New York, 1963.
60 Testing and Inspection of Welds, AWS School of Welding Technology, New Orleans, April
1968.
61 Robinson, G. C , Theory of Ultrasonic Testing, Prepared for Ocean Engineering Division,
Reading and Bates Offshore Drilling Company, 1967.
62 ASME Boiler & Pressure Vessel Code, Section III Nuclear Vessels, Appendix IX, 1969.
63 Grover, L. and Munse, W. H., Fatigue of Welded Steel Structures, Welding Research Council,
New York, 1964.
64 Olsson, D., Minutes of AWS Task Group on UT of Welds in Tubular Structures, Houston,
October 9,1972.
65 Robinson, G., protest letter to AWS on subject of UT, c. 1969.
66 Minutes of meeting AWS Structural Welding Committee, October 1979, and March 1980.
67 ASNT-TC-1A, Recommended Practice for Qualification of Nondestructive Testing Personnel,
American Society of NDT, Columbus, Ohio, 1968.
68 Lautzenheiser, C.E. and Whiting, A. R., Ultrasonic Techniques for Inspection of Weldments
in Offshore Drilling Platforms, Southwest Research Institute, Report to Shell Oil Company,
October 1967.
69 Peterson, R. E Stress Concentration Design Factors, John Wiley & Sons, New York, 1953.
70 Peterson, R. E., Fatigue of Metals in Engineering Design, ASTM Marburg Lecture, 1962.
71 Kihara, H., et al, Nondestructive Testing of Welds and Their Strength, Society of Naval
Architects of Japan, Tokyo, 1960.
72 Harrison, J. D., The Basis for a Proposed Acceptance Standard for Weld Defects, Welding
Institute Miscellaneous Report 27/3/72, March 1972.
73 Harrison, J. D., British Work on the Significance of Weld Defects, Proceedings Conference on
Significance of Defects, London, The Welding Institute, 1970.
74 Harrison, R. P., et al, Assessment of the Integrity of Structures Containing Defects, CEGB
Report R/H/R6-rev. 1, Central Electricity Generating Board, UK, 1977.
75 Fitness for Purpose in Welded Construction, Proceedings AWS/WRC/WI Conference,
Atlanta, May 1982.
76 AWS Dl Committee, UT task group, draft Appendix H, Ultrasonic Examination of Groove
Welds by Alternative Techniques, March 8,1989.
77 IIW Commission V, Evaluation of Ultrasonic Signals, IIW Doc. 850-86, also published by
The Welding Institute, 1987.
78 Cotton, H. C , ed., Fracture Toughness Testing: Methods, Interpretation, and Application,
The Welding Institute, Cambridge, 1987.
79 Houldcroft, P. T., ed., Rational Fabrication Specifications for the Offshore Industry,
Proceedings WI Conference, London, November 1985.
80 Garwood, S. J., et al, Crack Tip Opening Displacement (CTOD) Methods for Fracture
Assessments: Proposals for Revisions to PD 6493, The Welding Institute Report 371/1988.
81 Maddox, S. J., et al, Fatigue Analysis for the Revision of PD 6493,1980, The Welding Institute
Report 3873/1/86.
Chapter 8

CONSTRUCTION SYSTEM

The design process does not consist only of calculations against strength, fatigue, and
fracture criteria. To be comprehensively in control of his project, the designer must often be
responsible for material selection, tube manufacture, connection layout, welded joint design,
welder and procedure qualification, prefabrication, assembly, and inspection. These elements of
the construction system receive a great deal of attention in the AWS Code, and are briefly
reviewed in this chapter.

8.1 MATERIAL SELECTION

The AWS Code (Section 10.2) lists 32 material specifications which may be used in
tubular steel structures. Only eleven of these are tubular products per se. The remaining
specifications are for plate and strip which may be cold formed into structural pipe, using a
specification such as API Spec 2B. This Specification for Structural Steel Pipe (ref. 1) gives
welding and quality assurance requirements, as well as tolerances appropriate for structural
applicationssuch as column straightness and alignment at transverse girth welds. Inasmuch as
these members are not intended to carry internal pressure, pressure testing is not required.
However, out-of-roundness tolerances are specified, which are related to the collapse resistance
for external pressure.
At the risk of being repetitive, let us reiterate that a well engineered structure requires
that a number of factors be in reasonable balance. Factors relevant to the selection of structural
steel for tubular connections are: (1) static strength, (2) fatigue, (3) fracture toughness, and (4)
weldability. These are discussed by the author (ref. 2), and in the following sections.

8.1.1 Static Strength and Ductility

The fundamental requirement for satisfactory behavior of any structure is adequate static
strength in the connections. For tubular structures, this requires a practical design procedure, as
well as material with predictable strength and superior ductility, compatible with the design
procedure. Strength design is covered by Section 10.5 in the Code, and several previous chapters
in this book.
As discussed earlier, tubular connections depend upon plastic deformation and strain
hardening in order to develop the ultimate strength upon which their design is based. Yielding is
present in the hot spot region at design load levels. Even heavily reinforced connections may
have stress concentrations such that the ability to withstand localized yielding is a prerequisite to
satisfactory performance.
Caution is required when the design specifies high strength steel. Strength of the
connection may not actually increase in proportion to the yield strength. Often, the higher yield
strength is obtained at the expense of lower ductility and lower strain hardening capacity, and the
post-yield reserve strength which the design criteria implicitly count on may be compromised.
API and AWS codes limit the effective yield strength to 2/3 the tensile strength; Kurobane's
proposed rules (ref. 3) also de-rate steels with high yield-to-tensile ratios, but this feature has not
been retained in the latest IIW recommendations (ref. 4).

8.1.2 Fatigue
In developing the hot spot stress design criteria for fatigue, it was found that data for
steels having yield strengths in the range of 36 to 100 ksi (250 to 700 MPa) all fell in a common
 331

scatter band when tested with as-welded surface profiles; thus a single set of S-N curves is
applied to all grades of structural steel. In the fracture mechanics approach, crack growth
behavior (da/dN) is also remarkably similar between the various grades. However, fatigue does
affect material selection when one considers the trade-off between static strength and fatigue.
For example, if fatigue calculations indicate that a particular thickness is needed and the
static strength requirements are satisfied by a low or intermediate strength steel of this same
thickness, then there would be no advantage to using higher strength steel. Here the high
strength steel would incur higher material costs, more difficult forming, loss of ductility, and
possible welding problems, but permit no reduction in thickness.

8.1.3 Fracture Toughness

The extraordinary demands of localized yielding, strain hardening, and triaxial stresses
(required as tubular joints mobilize their full strength) must usually be met in the presence of
notches, since the highest stresses almost always occur in the hot spot region at the toe of the
weld joining the tubes at their intersection. Notch toughness is an extremely important material
property here, for without it the connection is subject to premature brittle fracture.
The author's historical approach to integrated fracture control solutionsfirst as a
necessary adjunct to his own design work, then in developing the U.S. national Codes for tubular
structureshas been the "conventional" Charpy-based approach described in Chapter 7 and
Appendix III. More advanced fracture control procedures for offshore towers are described in
References 5 and 6. As the CTOD test is a better analog of the behavior of a terminal fatigue
crack in the hot spot region of a tubular joint-can, and local brittle zones (LBZ's) are a topic of
current interest, further development in this area can be anticipated.
Brittle fractures are especially fearful to structural engineers, in that they occur suddenly
and prematurelyas opposed to gradual fatigue failure after many years of service, or strength
failure under conditions of overloadreflecting poorly on the engineer and his professional
liability insurance rating.
Illustrative of this "fear and loathing" is the cautionary footnote on A500 cold-formed
structural tubing (footnote to Code clause 10.2.1.9). Where the skelp from which these tubes are
made starts out with modest notch toughness, and the severely deformed corner regions are
degraded by strain aging and the heat of welding, they may not be up to the demands of yielding
in the presence of weld-toe notches, as occurs in the chord of tubular connections. Out of the
thousands of structures which have been made of this product, a few failures have been reported
to the AWS Structural Welding Committee. Although the apparent rate of failure is less than the
3-5% failure rate at the design S-N curve in fatigue, a caution flag has been raised. Some
committee members even wanted to make the "special investigation" mandatory. The special
investigation would presumably involve fracture toughness testing of the corner region, which is
subject to a broad range of interpretation, as previously discussed.
The need for such precautions would depend, at least in part, on whether the design of
the structure provides a degree of fail-safe redundancy or whether an isolated chord failure would
be catastrophic.

8.1.4 Weldability
When selecting steel for a welded structure, it is useful to have some technical basis for
judging the compatibilty of the material and the welding process. Factors to be considered
include (1) compositions of the base plate and welding materials, (2) thermal input due to preheat
and the welding arc, (3) thermal mass or heat sink, and (4) mechanical restraint (ref. 7).
Increases in carbon and alloy content make steel more hardenablethat is, more subject
to large increases in strength (and concurrently, loss of ductility) due to thermal cycles in the heat
affected zone of the weld. Steel with greater than about 0.35% carbon produces such hard, brittle
332

microstructures that it cannot be welded without special precautions. In weldable low allow
steels, the carbon content is limited to about 0.20%, and other elements such as Mn, Si, Co, Mo,
V, Ni, and Cu are added to increase the yield strength or to enhance other desirable properties
such as notch toughness or corrosion resistance.
Although these other elements are less spectacular than carbon in promoting hard
microstructures, they nevertheless have some effect. The "carbon equivalent" is an empirical
expression which combines carbon and the other elements so as to represent their total effect on
mechanical properties and hardenability. The following formula for carbon equivalent (CEQ)
may also be used to broadly predict the weldability of steels of the carbon-manganese and
structural low-allow variety:

C E Q = C + ^ + ^ + ^ + ^ + ^ + ^ (8.1)
6 15 15 5 5 5

An alternate formula has more recently gained popularity:

p C M = C + ^ + ^ + ^ + ^ + ^ + ^ + ^ - + 5B (8.2)
20 20 20 60 20 15 10

There is no critical value of carbon equivalent at which welding problems suddenly

appear. This is related to other factors such as preheat, welding arc energy input, joint geometry,
type of electrode (particularly whether or not of the low-hydrogen type), welding technique, and
post heat.
Mild steels generally have a carbon equivalent of less than 0.40%. These are the most
weldable of structural steels and are compatible with the widest variety of welding processes-
-including the use of cellulose-type stick electrodes (e.g., E6010). When used with low-
hydrogen welding processes, little or no preheat is required.
Intermediate-strength steels -42 to 55 ksi (300 to 400 MPa) yieldhave higher carbon
equivalents, ranging up to 0.45% and above when conventionally produced. These steels
generally require the use of low-hydrogen welding processes in order to avoid problems with
cracking in the heat-affected zone. Suitable preheats are listed in Table 4.3 of AWS Dl.l-90.
AWS Appendix "XI" describes an alternative approach to preheat, based on steel
chemistry, hydrogen level in the welding process, degree of restraint, as well as thickness. See
Figure 8.1. The lower "modified" preheats (MP) are applicable to new types of steel with "lean"
chemistries, which attain their strength via grain refinemente.g., thermo-mechanical control
process (TMCP) in Japan, and off-line accelerated cooling (Q&T) in the USA. These are
covered by API Specs 2W and 2Y, respectively.
Because the ASTM specifications for many of the conventional steels do not control
carbon equivalent, nor even require reporting all of the elements involved, problem heats are
occasionally produced which turn out to have excessively higher carbon equivalents, runaway
strength, and problems during welding. These situations often arise with electric furnace mills
using scrap metal as a source of iron. Residual elements derived from the scrap give the carbon
equivalent an unwanted boost. One such case history is described in Reference 2. As a result of
this and similar experiences, API Spec 2H limits the carbon equivalent to 0.45%.
Besides the delayed (hydrogen) cracking problem which carbon equivalent and preheat
address, the base metal composition can also be a factor in hot (solidification) cracking.
However, this risk has been greately reduced by the low sulfur and phosphorous content of
modern weldable steel specifications. This type of cracking is most likely when high-heat input
welding processes are used on thick plates with high restraint, or when contaminants (e.g.,
copper) are present.
 333

CEQ . 3 7 CEQ . 4 1
PCM ^ . 1 8

400
W E L D A B I L I T Y C L A S S MP PCM $ . 2 3
W E L D A B I L I T Y C L A S S C
400i
I 3 to 5 m l / I O O g eg. 7 0 1 8 HI
H2 7 to 10 m l / I O O g eg. INNERSHIELD H2 -

, HIGH RESTRAINT
Y-GR0VE TEST
300 HIGH RESTRAINT 300
Y-GROOVE T E S T

200 200 h
i-
<
UJ
I
or MODERATE RESTRAINT
LJ CTS TEST
-
00 - 100

MODERATE RESTRAINT !___!

CTS T E S T

I 2 I 2
T H I C K N E S S - IN. THICKNESS-IN.

Fig. 8.1. Modified (MP) and conventional (CP) preheats.

Lamellar tearing, which might also be considered a weldability issue, has been
previously discussed.

8.2 TUBE MANUFACTURE

8.2.1 Methods
There are several methods by which tubes for structural applications are produced.
They are briefly described below:
(i) Seamless. Pipe is produced by piercing a white-hot billet and subsequently
expanding it with rollers and drawing to the desired size and length. Uniform metallurgical
properties and low residual stresses are achieved, but there can be considerable variation in wall
thickness (typical tolerance of 12%).
(ii) Electric Resistance Weld (ERW). Skelp is formed into pipe and joined along the
longitudinal seam by electric resistance or flash welding, in which the pipe material is locally
heated close to the melting point and pressed together. Tight quality control, e.g., attention to
cleanliness, is required to avoid weld defects. Seam normalizing refines the undesirable grain
structure of the initial weld. Hydrostatic testing eliminates the most blatantly incomplete welds,
but corrosion-augmented failures of the weld in service are not uncommon. These are not as
serious in structural applications as they are in pressure piping, for which ERW pipe is
nevertheless widely used.
(iii) Submerged Arc Weld (SAW). The weld seams are made by the deposition of
metal from wire electrodes, with the molten weld covered by a flux or slag, which refines the
weld metal chemistry and protects it from the atmosphere. Welds of very high quality can be
consistently produced. SAW pipe is further distinguished by the method of forming, as follows.
(iv) Can Pipe. Plate is formed into short cylinders (or cans) of typically 10-ft. (3m)
length, using "pyramid" arrangement of three rollers, or a hydraulic press break. After making
the longitudinal seam, the cans are joined together by girth welds to make longer structural
members. A rather complex pattern of residual stresses is developed from the forming and
334

welding, both circumferentially as well as axially (the latter shown previously in Figure 1.3).
This method is widely used to custom fabricate the large variety of sizes typically found on
offshore structures. It is covered by API Spec 2B (ref. 1), which gives tolerances appropriate to
structural applications. Variations in thickness or steel grade along the length of a member can
also be accommodated. Very high quality steel may be incorporated, depending on the plate
specification used.
(v) UOE. Long sections of skelp are press formed on huge dies, first into a "U" shape,
M
then into a circle ( 0 " - i n g ) . After making the SAW long seam, the pipe is hydraulically
expanded ("E") beyond yield to dissipate residual stresses and produce a nearly perfect circle.
Very straight lengths of 40 to 60 feet (12 to 18m), with no girth seams, are produced. The
method is most economical for long runs (miles or kilometers) of the same size pipe.
(vi) Spiral Weld. Skelp from a steel coil is continuously formed into pipe with a spiral
steam, which may be either ERW or SAW. Based on poor early experience with the ERW spiral
weld pipe, structural use of this product is restricted by some Codes and user specifications,
perhaps unfairly.
(vii) Non-Circular Sections. Square and rectangular box sections are often made by
reshaping circular tubes. Depending on whether this is done hot (e.g., A501) or cold (e.g.,
A500), residual stresses and strain-hardened corner regions may be created, affecting
performance. Box sections are also made from plate by press forming two channels which are
then welded together; this process can easily accommodate custom size and steel grade
requirements. Very heavy box sections are made by welding four plates together at the comers.

8.2.2 Effect on Performance

Because of the extensive use of can pipe in offshore structures, its behavior has been
thoroughly studied (refs. 8 to 10). Both column buckling and local buckling are affected by the
shape of the stress-strain curve, residual stresses, and geometric tolerances. However, the
available test data are consistent with use of the AISC column buckling curve (ref. 11) and API
criteria for local buckling (ref. 12), using the yield strength of the virgin plate material. These
effects were discussed previously in Chapter 1.
Other kinds of manufactured pipe, e.g., seamless and ERW, are also used with the same
design criteria. Where the pipe material has been cold-strained during manufacture, and the
resulting higher yield strength is what gets reported on the mill certificate (e.g., UOE and some
of the higher strength grades of line pipe), then use of a lower column design curve should be
considered, e.g., AISI class "B" or ECSC curve "C" as shown earlier in Figure 1.4. Thus,
consistency with the design intent should be considered when substitute materials are proposed
during construction, even when the nominal yield strengths are the same.
Similar considerations apply to box sections with cold-strained corners, or those with
compressive residual stresses in mid-face (due to welding the corners), as discussed earlier in
connection with Figures 1.4 and 1.6.

8.2.3 Section Availability

The AISC Steel Construction Manual lists dimensions and design properties for a
variety of tubular sections up to 12-in. nominal diameter. Pipe sizes are typically larger than the
nominal diameter, e.g., nominal 2-in. pipe is actually 2.375-in. Standard weight, extra strong,
and heavier sections are widely available from stock, particularly in mild steel grades. When
ordered to some specifications, e.g., ASTM A-53, the pipe may be either seamless or ERW.
Commonly used larger sections, up to 24-in. diameter, were listed in Table 1.1 herein.
These may be either ERW or SAW pipe. Custom sizes of 24-in. diameter and up may be
fabricated as can pipe. Here the nominal and actual diameters are the same.
 335

The AISC manual also lists a large number of square and rectangular sections, along
with their properties. However, some of the sections listed have limited availability. Hot formed
box tubes (e.g., A-501) are no longer manufactured in the USA. Cold formed sections are more
widely available, but some North American suppliers exploit the tolerances and furnish less than
the specified minimum thickness. However, for the larger sizes, unavailable or custom sections
can be fabricated from plate (this is also possible, but not very practical, for small sizes).
It is generally advisable to check with local suppliers (or with importers whose quota is
not used up), before assuming that a given section will be available and that it actually meets
thickness tolerances.

8.3 CONNECTION LAYOUT

One of the unique aspects of structural connections for circular tubes is the geometry of
the welded joint where one tube joins another. In general, this follows a three-dimensional curve
in space, which may be determined graphically by using elementary descriptive geometry (ref.
13), as shown for a simple 90 T-connection in Figure 8.2. For angled tube-to-tube connections,
the locus of the intersection assumes an egg-shaped geometry (with the small end of the egg at
the acute angle heel of the connection), curved in space like the rim of a saddle.

Fig. 8.2. Graphical layout of

the intersection in a simple T-
connection. (a) Top view, (b)
Side view, (c) End view along
branch member aixs.

1 '*
A wrap-around paper template for cutting the end of each branch member at a tubular
connection may be constructed graphically as shown in Figure 8.3. The construction is made for
the intersection of branch member inside surface and main member outside surface, so that the
locus defined is that of the root of the weld. This is projected (stretched) to accommodate an
outside surface wrap-around. The distance from the work point at the intersection of member
axes, to a convenient reference line, must also be accurately determined.
In the early days of building tubular space frames for offshore structures, the descriptive
geometry and wrap-around template layout were done at full scale in a shipyard-style mold loft.
A working sketch for each member, showing centerline workpoint length, reference lines, and
long points, was also made. Reference lines were laid out on the structural pipe to be cut, the
template wrapped around, and a rough first cut made perpendicular to the local pipe surface, with
336

Fig. 8.3. Graphical construction and development of a wrap-around template, (a) End view, (b) Side
view, (c) Developed view of unrolled tube surface.

the cutting torch following the scalloped end of the template. Using the first cut to visualize the
position of the main member to which the branch will be welded, and anticipating the required
bevel angle, a highly skilled workman then made a second cut manually, also shortening the
member slightly so as to leave a small root opening when it is fit into place. Depending on the
degree of artistry with which this was accomplished, the bevelled end may or may not require
further touching up prior to fitting and welding.
The foregoing practice was able to compete successfully with early mechanical-analog
cutting machines for many years (until the late 1970*8) and may still be attractive for one-shot
jobs. Today, however, increasing degrees of automation have come into play. Working
sketches, and cutting templates where used, are computer generated. Getting a large enough
plotter is usually a bigger investment than the personal computer and software required. At a
higher level of capital investment, the entire process may be carried out by numerically
controlled cutting torches, which vary the distance "X" (Figure 8.3) and the bevel angle, as the
torch travels around the pipe, making the final cut in one pass.
For tubes intersecting flat surfaces, the intersection line falls in that plane and the 3-D,
egg-shape becomes an ellipse. Here, the first cut may be made by sawing the tube at an angle.
This approach does not work for circular tube-to-tube connectionswith the possible exception
of small diameter, small beta connections, where the deviation from planarity falls within the
root gap tolerance. However, it can be applied to most box-to-box, circle-to-box, and tube-to-
wide flange tubular connections.
In general, the geometry of the welded tubular joint (i.e., the weld itself and the
immediately adjoining base metal) varies continuously as one proceeds around the intersection,
as a function of the local dihedral angle. AWS defines local dihedral angle as follows: The
angle, measured in a plane perpendicular to the line of the weld, between tangents to the outside
surfaces of the tubes being joined at the weld. This is the exterior dihedral angle, where one
looks at a localized section of the connection, such that the intersecting surfaces may be treated
as planes. See Figure 8.4.
 337

Detail Applicable range of local dihedral angle,

A 180 t o 135
150 t o 50
C 75 t o 30 | N o t prequalified for

D (groove angles under

30

Area for Detail

Fig. 8.4.
Main member 7 Area for
Detail C or D

Weld joint detail selection as a function of local dihedral angle (AWS Dl. 1).

Figure 8.5 further defines details of the welded jointgroove angle, root opening, and
other key dimensionsas a function of local dihedral angle. The range of dimensions
corresponds to that observed in the manual cutting practices described earlier. These details are
applicable where complete joint penetration groove welds have been specified for tubular -, Y-,
and K-connections. Their technical background, and other alternatives, are described in the
following section.
Where the practice of manual cutting "by eye" has become a lost art, or where it never
existed, the required bevel angle may be determined by layout, as a function of local dihedral
angle. To aid in this, Reference 14 has derived local dihedral angle as a function of position
around the member for various member intersection angles and diameter ratios. This information
is also presented as a series of charts in an Appendix to the Code (ref. 15).

8.4 WELDED JOINT DESIGN

In previous chapters, we have discussed criteria by which the designer may select the
proper thickness and grade of steel for the joint can in simple tubular connections. Static
strength criteria have covered both local failure (punching shear) and general collapse
(ovalizing). Fatigue may also govern the thickness selection, particularly in offshore structures.
Critical material properties, such as ductility and homogeneity, were just mentioned. However,
the designer is not yet finished. In addition to bearing the ultimate responsibility for what
happens during welding and inspection, he still has some design decisions to make. As discussed
below, these fail in the areas of selection of weld type, weld sizing, and profile control. Joint
detailing, as prescribed by the code once these choices are made, is also discussed.

8.4.1 Selection of Weld Type

For tubular connections, the designer should carefully consider the selection of weld
type from among the following choices:

(a) complete joint penetration groove weld (Figure 8.5)

(b) partial penetration groove weld (Figure 8.6)
(c) fillet welds (Figure 8.7)

These joints differ in their performance (i.e., allowable stresses), ease of execution by the
fabricator, and degree of welder skill required. In tubular connections, these welds must usually
be made from one side only, as the welder does not have access to the inside of the tube.
Because of this difficulty, the designer should not automatically call for complete penetration
welds unless they are really needed for the service demands of his application.
338

Fig. 8.5. Weld details for complete joint penetration.

 339

Toe Side Heel

(stepped connections)

Fig. 8.6. Original AWS prequalified details for partial joint penetration welds on -, Y- and K-
connections.

Fig. 8.7. Original AWS prequalified details for fillet welds in -, Y- and K-connections.
340

Complete joint penetration groove welds may be presumed to develop the full static
strength of the sections joined, when matching filler metal (e.g., AWS Table 4.1) is used.
Although significant discontinuities at the root of the weld may occur (as discussed in Section
7.5 herein), the AWS prequalified details satisfy the foregoing definition in that the weld itself
does not limit the strength of the connection. Static strength is usually limited by the punching
shear strength of the joint can; and fatigue strength is limited by the hot spot stress concentration
factor and by the external notch at the toe of the weld.
Partial penetration groove welds in tubular connections can be executed with less
stringent brace end preparation and fit-up than complete penetration groove welds, and do not
require the use of highly specialized 6GR-qualified welders (see Section 8.5). However, despite
a sizeable loss factor (presumed lack of penetration at the root), these welds match the allowable
static strength of the sections joined, when using the prequalified details (AWS Fig. 10.13) and
slightly overmatched filler metals (e.g., 0.3 70 ksi versus 0.6 35 ksi using E70XX electrodes
with mild steel branch pipe). Somewhat lower cyclic stress allowables may apply (i.e., AWS
fatigue S-N curves ET and FT instead of curve DT). However, where fatigue is not of overriding
importance, the designer who is willing to perfonn the necessary weld sizing checks and specify
partial penetration welds can save a lot of grief and expense, particularly on small projects built
outside the areas of active tubular structure fabrication.
Fillet welds can be executed with the least stringent fit-up and welder qualification
requirements. They are particularly advantageous in that single-pass welds can be made for
small branch members having less than 0.21-in. (5.3mm) wall thickness, e.g., standard weight
pipe through 2.5-inch nominal diameter. However, when the minimum prequalified weld sizes
are used (AWS Fig. 10.14), they may develop as little as 50% of the allowable static strength of
the branch member. Thus, extensive weld sizing calculations by the designer may be required
for "unzipping" behavior (insufficient deformation capacity), as well as for applied loads.

8.4.2 Weld Sizing

There are two considerations for which weld sizing must be checked in tubular
connections: (1) sizing to carry the design loads at allowable stress levels, and (2) sizing to
prevent the weld from becoming a weak link, where the connection "unzips" prior to reaching its
intended capacity. Complete penetration groove welds always meet both requirements, so no
checks are required. The prequalified partial penetration welds may be presumed to meet (2)
when matching filler metals are used, but must be checked for (1). Fillet welds must generally
be checked for both (1) and (2).
In AWS, the calculations for sizing the weld to carry the design loads at allowable stress
levels are performed by taking a free-body boundary through the weld and resolving branch
member loads into weld stress using geometry and statics. Guidance for calculating weld
effective lengths may be found in AWS Section 10.8. Weld allowable stresses are given in AWS
Section 10.4.
Actual weld stress will in general be quite different from what the Code's simplified
approach indicates. Recall that due to the uneven transfer of load across the weld, the peak line
load may be twice that indicated by geometry and statics. A grossly undersized weld could fail
progressively, starting at the hot spot and unzipping around the connection. Hence, the second
check for uneven distribution of load.
As described in AWS clause 10.5.3, welds in tubular connections are prevented from
becoming a weak link by requiring that their breaking strength at least match the lesser of the
branch member yield strength or the main member punching shear capacity.
 341
8.4.3 Profile Control
In Section 7.1, we have seen how weld profile and thickness effects are inter-related
with the fatigue performance of welded tubular connections. The designer has the choice of two
levels of fatigue performance, corresponding to the upper X - l and lower X-2 hot spot S-N
curves. For each of these performance levels, the AWS Code defines a logical set of weld
profiles, as a function thickness, as shown in Figure 8.8. Once learned, they should become a
natural progression for welders to follow. They have evolved from the following experience
(refs. 16-19).
For very thin branch members in -, Y-, and K-connections, the basic flat profiles
shown for full penetration welds are also those which apply to partial penetration and fillet
welds, as used for onshore applications where fatigue is not critical. Here the entire weld cap is
made in one pass, with weaving as required. Using E6010 electrodes, the more artistic capping
specialist could make this a concave profile, merging smoothly with the adjoining base metal.
With the advent of higher strength steels and heavier sections, requiring low hydrogen
electrodes, with the introduction of high deposition rates, semi-automatic welding processes, this
seems to have become a lost art.
For heavier thicknesses, a definite fillet is added at the weld toe as required to limit the

weld toe notch effect to that of a 4 5 fillet weld. These fillet welds are scaled to the branch
member thickness so as to approximate a concave weld shape. However, we are also constrained
by the need to maintain minimum fillet weld sizes to avoid creating dangerously high hardnesses
in the heat affected zone at the weld toe (this is also the location of the "hot spot" which will
experience localized yielding at the design load levels). This alternative "standard" profile is
more straightforward to communicate to the welders and easier for them to achieve out-of-
position than the idealized concave weld profile shown in earlier editions of the Code. The
resulting weld profile is much like that observed on early Gulf of Mexico offshore platforms
whose fatigue performance over several decades of service has been consistent with Categories
X, K, and DT, with branch member thicknesses up to 0.625 in. (15.9mm) (typically associated
with chord thicknesses up to 1.25 in. (31.7mm)).
For thicknesses over these limits, a concave profile which merges smoothly with the
adjoining base metal is specified. Attempts to enforce this profile control in the field have
brought mixed results. Using small diameter stick electrodes with good wetting characteristics,
with capping specialists laying down small stringer beads, the desired profiles have been
achieved, although a temper bead technique is sometimes needed to avoid hard heat affected
zones. However, with the introduction of self-shielded FCAW~a semi-automatic process with
high deposition rates-one tends to get large, ropy beads in the vertical-up and overhead welding
positions. Vertical-down FCAW welding can improve the profile, but at the expense of less root
penetration and increased risk of cold-laps and slag inclusion. When corrective grinding is used
to meet the letter of the disc testwith just the tops of these beads ground off, leaving steep
canyons in betweenit is unlikely that the benefits derived justify the large amount of work and
inspection involved. Once it is necessary to resort to grinding, it is probably simpler, and
certainly more effective, to render the entire weld profile smooth, with particular attention to
eliminating undercut and other crack-like discontinuities at the toes of the weld.
Once again we see that design, welding, and inspection are not independent functions,
but must be inter-related in order to achieve a consistent overall fracture control strategy. While
joint detailing, welder qualification, and welding procedures are carried out by the fabricator, the
design engineer has a responsibility to follow through, and must understand what is going on, in
these areas. If he anticiapted good weld profiles and designed according to the upper fatigue
curve X - l , but got poor weld profiles instead, the fatigue life of his structure will be only a small
fraction of what he was counting on.
342
APPLICABLE BRANCH
THICKNESS t b FOR:

DESIGN TO DESIGN TO
UPPER LOWER
CURVES CURVES
, X2,K2

B
ACK UP
WEL
DM AD
E
F
ROM
B
A
W
F
R
C
E
O
K UP
L
DM
MO
AD
U
T
E
S
DIE
OUT
SD
IE ;
4
AND*
CONCAVE PROFILE:
over 5 / 8 " over 1 1/2"
LT
-
H E O
RET
C
IAL LT
HEO
RET
C
IAL LT
HEO
RET
K
(FATIGUE SIZE
- 75 - 30'
EFFECT PENALTY
DETAIL "
OVER 1")

-BUL
ID UP AS
REQURIED TO
MANTAIN l w
TOE FILLET:
3/8" t o 5/8" 5/8" t o 1 1/2"
(NO LIMIT FOR
U 3 S r
STATIC
- COMPRESSION)
. -135
DETAIL "A

BASIC FLAT:
O
F VARIES FROM
0TO ! b/2 AS
S thru 3 / 8 " thru 5/8"
* VARIES FROM tw _ F- tb/2
-
* - 90' 50

/ L L ^
/
l
B
Ack
upw ' U PLDWE L
ACKU PLD W
ELD i- BAC lb* 3
-75 F 5 30 * . 40'
" * .
FROM " TO "-

Fig. 8.8. AWS progression of weld profile requirements (dimensions in inches, 1 inch = 25.4mm).
 343
8.4.4 Joint Detailing
For butt welds and tee welds in which backing strips can be used, or where there is
access to the inside of the tube for welding from both sides, then the joint details developed for
other types of welded construction are equally applicable to tubular structures. However, for
tubular connections involving branch tubes intersecting the chord at various angles (generally
referred to as , Y, and connections in AWS), a number of complications arise:

(1) Often there is no access to the inside of the tube, and the complex geometry precludes the
use of backing rings.

(2) The welding position, as well as the geometry of the weld, may change continuously in
going around the intersection.

(3) The local dihedral angle and the weld groove angle extend to smaller values than
encountered in other types of construction, restricting accessibility.

Under carefully prescribed conditions, AWS permits "complete penetration" groove

welds in tubular connections to be made from one side without backing. The prequalified joint
details prescribed for this situation are shown in Figure 8.5. They were developed from
experience with all-position shielded metal arc welding, fast freezing short-circuiting gas metal
arc welding, and self shielded flux cored arec welding with similar fast freezing characteristics.
Local dihedral angle at a given point along the weld is pre-determined as a function of
the overall connection geometry. In turn, local dihedral angle determines which detail (A, B, C,
or D) is applicable at a given point. Several different details will be used at different points
around the connection weld.
Starting with the local dihedral angle as given, the left-hand graph in Figure 8.5 shows
the commonly used range of groove angles, as well as the required minimum weld size or throat
dimensions. Once the branch member has been cut, contour beveled, and fit into place for
welding, the middle graph shows the root opening required to get full penetration for details A
and B. For details C and D, dimension "W" in the right-hand graph shows the narrowest groove
for which sound welding can be assured; beyond "W" in the root region a "back-up weld" of
uncertain quality is presumed, and this is excluded from the theoretical weld. The wider grooves
shown for gas metal arc welding were found necessary to accommodate the shrouded tip of the
welding gun.
In addition to careful layout, skillful end bevelling, and exacting fit-up requirements,
extraordinary skills on the part of the welder are required in order for single-sided welds in
tubular connections to qualify as "complete penetration". The 6GR test for this specialist is
described in Section 8.5.

8.4.5 Joint Detailing for Box Connections

In general, the welded joint details developed for circular tubular connections are also
applicable to box connections, as a function of local dihedral angle. However, box connections
introduce two new situations: (1) a flare bevel groove weld at the side of matched box
connections, and (2) rapid transitions as the weld progresses around the corners of the branch
member.
(i) Flare Bevel. Manufactured box tubes typically have rounded corners. Where the
corner radius of the chord member is sufficiently generous, it will form a flare bevel when the
flat-cut end of a branch member is fit against the chord, sufficient for achieving a weld effective
throat equal to the branch member thickness. However, there are two ways of measuring the
comer: (1) with a concave radius gage, and (2) with a carpenter's square. Where the curved
344

corner terminates abruptly before becoming tangent to the flat face, the corner dimension of "c"
M
of method (2) is less than the radius "r of method (1). To avoid the kind of confusion which can
wreak havoc in a construction project, both should be considered in deciding if the corner is
generous enough. The AWS Code calls for "r" > 2t and "c" > t + .125-inch (3mm), where t is the
branch member thickness. If there is doubt, a sample joint should be prepared to verify adequate
effective throat (ref. 20).
(ii) C o r n e r Transition. At the corners of a matched box connection, the flare bevel
changes to either a V-groove (at the toe) or an external fillet (at an acute heel). Particular care in
detailing is required to maintain the weld effective throat throughout this transition. The AWS
prequalified joint details show examples of how this may be accomplished; the detailing is
surprisingly complex. As the corners are the most highly stressed part of the connection (see
Chapter 5), the Code also requires that welding be carried continuously around these corners,
with all starts and stops within the flat faces of the connection. Specialized welder skills are
needed to successfully accomplish this, and these are reflected in special qualification
requirements.

8.5 WELDER AND PROCEDURE QUALIFICATION

Welder and procedure requirements in AWS D l . l have evolved from a simple reference
to the 6GR test in the first edition (1972), to an elaborate logical structure and testing matrix in
the present Code. This was brought about by the extension of tubular construction from the U.S.
Gulf Coast (where everyone seemed to know what to do without having it written down) to more
formalized contractual and regulatory settings: North Sea platforms and nuclear power plants.
Following the author's tenure as chairman of the AWS Subcommittee on Tubular Structures
(1974-78), his successors Hubert Crick (1979-1984) and Jeffrey Post (1985-1990) represented
the viewpoint of experienced fabricators and struggled with these issues almost continuously.
The continuing stream of "helpful suggestions" from the AWS hierarchy and inquiries from the
user community suggests that they are still not completely clear. Rather than reiterate all the
Code provisions here, their logic and basis will simply be outlined.

8.5.1 Welder Qualification

Welder qualification tests form an essential part of the overall quality control scheme.
Where prequalified details and visual inspection only apply, these may be the only QC tests
performed.
Complete penetration groove welds in tubular connections present the following special
problems for the welder: The open weld root must be provided with well-fused initial pass, in
spite of restricted access and the tremendous heat sink provided by the heavy continuous chord.
Welding position changes continuously as one progresses around the weld; and in the crotch
(acute angle crown position) access is further restricted. A weld cap profile which merges
smoothly with the base metal in two different planes on either side of the weld is desired.
Figure 8.9 shows the 6GR test, which attempts to represent all these conditions to the
extent that this is possible in a single test. The pipe axis is inclined 45 degrees and fixed, so that
elements of all welding positions are included. A restriction ring simulates the access problem
posed by presence of the chord member, while the internal shoulder simulates the three-
dimensional heat flow and fusion problems at the root. Dashed lines show an optional external
shoulder which would test the welder's skill at merging the cap profile with adjoining base
metal. After visual inspection of weld cap and root, standard prismatic specimens are extracted
for mechanical testing (e.g., bend tests).
Welders inexperienced in tubular structures have difficulty passing this test. To some
extent, the skills required are also different from those for welding pressure piping. In the
 345

hierarchy of AWS welder qualifications, the 6GR test qualifies for all types of welded joints and
positions, except the open root pipe butt joint.

1/ c m m . ~*
(12 m m ) / ^

(3) 6GR T E S T SPECIMEN (b) DETAIL A T W E L D

Fig. 8.9. Welder qualification test specimen for complete penetration groove welds in tubular -, Y-,
and K-connections.

8.5.2 Welding Procedures

Under the AWS Structural Welding Code, many procedures are considered prequalified
on the basis of extensive prior testing and experience. These may be used by qualified welders
without further testing, provided all the pertinent provisions of the Code are followed exactly.
These provisions include prequalified groove geometry; surface preparation and workmanship
tolerances (Section 3); filler metal selection, preheat and interpass temperature, electrode/flux
moisture control, and limitations on electrode size, weld pass dimensions, welding position, etc.
(Technique, Section 4); as well as provisions specific to buildings, bridges, or tubular structures
(Sections 8, 9, or 10, respectively). Voltage, amperage, arc length, etc., are controlled within the
permissible range by the welder to suit the thickness of material, type of groove, welding
position, and other circumstances attending the work; thus, his qualification becomes a key
element in consistently producing sound welds.
Occasionally, one is faced with situations falling outside the range of the prequalified
proceduresfor example, new base metal or filler materials, different groove details, or a desire
to use welding parameters outside the previously established range. Procedure qualification tests
are then required in order to establish that the new procedure is capable of producing welds
having soundness and mechanical properties comparable to those which have previously proven
satisfactory. The Essential Variables listed in the Code serve as a check list of items which must
be:

(1) established during testing, and

(2) followed in production welding.
346

While the prescribed bench welds and laboratory tests do not guarantee successful
production and service experience, they are a necessary first step. For work involving tubular
connections, a full scale mock-up is also advisable (see AWS clause 10.12.6.3), in addition to the
6GR type butt welds made for the purpose of mechanical testing.
Whether the welding procedures are prequalified or established by testing, they should
always be documented by a written Welding Procedure Specification, so the various people
involved all understand how the work is to be done.

8.5.3 Outline of Code Provisions

The logical matrix of AWS Code provisions dealing with prequalified joint details,
procedure qualification, and welder qualification has been summarized in Table 8.1 (AWS Table
10.12, now 10.5). In the 1990 edition of the Code, the table and figure numbers have been
changed to conform to ANSI editorial format. It is hoped that confusion caused thereby will be
short lived.
Several specialized tubular applications are defined in which complete joint penetration
groove welds are permitted to be welded from the outside only, without backing:
(i) Pipe butt joints. In butt joints, complete joint penetration groove welds made from
one side are prohibited under the conventional provisions for dynamically-loaded structures and
statically-loaded structures, yet they are widely used in pressure piping applications. They are
now permitted for tubular structures, but only when all the special provisions of Code clause
10.12.3.1 are followed. The standard 60 pipe butt joint of AWS Figure 5.20A (now 5.23) is
recommended, as many pressure piping welders test on this configuration. However, it is not
prequalified in the sense that a complete set of tensile and bend tests is required for each
application.
(ii) -, Y-, and K-Connections. Prequalified joint details for both circular and box tube
connections are described in AWS Section 10.13. The situations under which these may be
applied are described in the table, along with the required procedure and welder tests. These
requirements are discussed further below.
Because of the special skills required to successfully execute a complete joint
penetration groove weld in tubular -, Y-, and K-connections, the 6GR level of welder
0
qualification for the process being used is always required. Where groove angles less than 3 0
are to be used, an acute angle sample joint test is also required for each welder.
Where groove details in -, Y-, and K-conections differ from the prequalified details, or
there is some question as to the suitability of the joint details for procedure, then a mock-up or
sample joint is required in order to validate the procedure.
Additional procedure qualification tests may be required on account of some essential
variable other than joint design. These c i r c u m s t a n c e s are described by the heading
"Metallurgical Compatibility" and include (but are not limited to) the following:

(1) The use of a process outside the prequalified range (e.g., short-circuiting
GMAW).
(2) The use of base metal or welding materials outside the prequalified range (e.g.,
the use of proprietary steels or a non-low-hydrogen root pass on thick material).
(3) The use of welding conditions outside the prequalified range (e.g., amps, volts,
preheat, speed, and direction of travel).
(4) The need to satisfy special owner testing requirements (e.g., impact tests).

The following prequalified joint details, referred to in Table 8.1 herein, are more fully
described in the Code:
 Partial joint penetration
Complete joint penetration tubular groove welds tubular groove welds Fillet welds

Single-welded butt joints

without backing Single welded, - , Y-, and K-connections
L/ouoie- Box - , Y - , All - , Y-,
Circular T - ,
welded or and It-
Standard Standard Standard Y,- and It- and i t -
single-welded connections
detail Other circular box Other connections connections
with backing
Joint detail
Prequalified Detail as
to which Detail as
joints per Fig. 10.13.2 Fig. 10.13.2 Fig. 10.13.3
procedure Fig. 5.20A qualified Fig. 10.13.1 A Fig. 10.13.1B qualified
Section 2 by test
applies by tests

Not required for S M A W

and self-shielding F C A W Not required for S M A W and self-shielding
When Always Always meeting Section 3 and 4 F C A W meeting Section 3 and 4 requirements.
Test for required required required requirements. Required for Required for G M A W and other processes
metallurgical G M A W and other processes outside prequalified limits.
compatibility outside prequalified limits.
and
mechanical Test joint Procedure 6G or 6G or 6GR or 6GR or 6G or 6G or Per 5.10.3
properties and testing qualification 2G + 5G 2G + 5G 6GR or 2G + 5G 2G + 5G 2G + 5G, 2G + 5G
for all (per Table per Section joint per specific 2G + 5G; Box joint typical joint per box joints
positions 5.10.1 or 5, Part * Fig. 5.20A; joints joint per per Fig. 5.21 A joint within Fig. 5.20A per Fig.
5.10.2, as to be Fig. 5.21 A + Fig. 5.21 range to or B; 5.20A or B;
applicable) qualified for corners be qualified

Sample joint When New procedure or variables

or tubular required outside prequalified limits; Required for
mock-up to Not Not always required for groove Always Not required corner radius Not
check suita- applicable applicable angle less than 30. required for S M A W under 2t required
bility of and F C A W Per 10.12.3.5
joint details Type of test Per 10.12.3.3 Per 10.12.3.3 Per 10.12.3.3 or 10.12.3.6

Minimum Welder and 2G + 5G or 2G + 5G or 6GR using 6GR using 6GR using 6G or 6GR or 3F + 4F;
welder welding 6G using 6G using Fig. 5.21 A Fig. 5.21 A + Fig. 5.21 A + 2G + 5G 2G + 5G for dihedral
qualification operator Fig. 5.20A specific (See Note 1) Fig. 5.21 Fig. 5.21 (See Note 1) round or under 60 deg
for all qualification joint detail (See Note 1) (See Note 1) Box tube use partial
positions per Section 5, or 3G + 4G penetration
Parts C & D Plate tests. requirements
TABLE 8.1 PROCEDURE AND WELDER REQUIREMENTS FOR TUBULAR JOINTS

Comments 100% N D T required for 1/8 in. diam max electrode

production for S M A W root pass

1 2 3 4 5 6 7 8 9 10 11

Note 1: Acute ar gle heel test required for groove angle under 30. See 10.12.6.4 and 10.12.6.5.
347

When welded utside prequlified status or by contract documents.

348

Fig. 10.13. - T-K-Y complete penetration details (dimensions in Table 10.7)

basicstandard flat profile (thin sections), now Fig. 10.9
toe fillet profile (intermediate), now Fig. 10.10
concave profile (thick sections), now Fig. 10.11
Fig. 10.13.IB - T-K-Y box connections complete penetration, now Fig. 10.12
Fig. 10.13.2 partial penetration T-K-Y, circular and box, now Fig. 10.13
Fig. 10.13.3 - fillet welded T-K-Y, circular and box, now Fig. 10.14

A number of test joint configurations are also referred to in the Table, for purposes of procedure
or welder qualification, as follows:

2G tube axis vertical, groove weld horizontal

5G tube axis horizontal; weld horizontal, vertical, o'head
6G tube axis inclined 45 , butt weld
6GR see Fig. 8.7(a) herein, for T-K-Y connections
0
Fig. 5.20A - 6 0 open root pipe butt joint (now Fig. 5.23)
Fig. 5.20B pipe butt joint with backing (now Fig. 5.24)
Fig. 5.21A - see Fig. 8.7(b) herein, for T-K-Y connections (now Fig. 5.25)
Fig. 5.21B alternate specimen, tube welded to plate (now Fig. 5.26)
Clause 10.12.3.3 sample joints or T-K-Y mock-up
Clauses 10.12.3.5 & 6 sample joint for flare bevel
Clauses 10.12.6.4 & 5 - acute angle heel test
3G and 3F plate test with weld axis vertical
4G and 4F plate test with overhead welding (G is groove, F is fillet)

Details of the test joint configuration; preparation of macro-etch, tensile, bend, and
impact specimens; and the required test results are generally spelled out in Section 5 of the Code.
A few of the tests, that are so specialized that nobody but tubular people wants to deal with them,
still reside in Section 10.
It bears repeating that the extent of qualification testing and required welder skills are
greatly reduced if partial penetration or fillet welds are used for -, Y- and K-connections.

8.6 PREFABRICATION

Prefabrication refers to the construction operations which take place before the steel is
incorporated into the final structure assembly. Traditionally, in Gulf of Mexico offshore
structure fabrication, prefabrication and assembly were done by the same contractor who started
with flat plates and produced a fully assembled jacket (tubular space frame weighing up to
30,000 tons), working in mill buildings and open-air yards at different parts of the same site. In
other geographical areas, and for larger one-of-a-kind projects, prefabrication and assembly may
be split between contractors at different sites. The latter would also normally be the practice for
buildings, bridges, and other onshore applications of tubular structures.
The form in which prefabricated material is to be furnished depends on the method of
erection and assembly. Two methods are widely used for tubular space frames: bent fabrication
and node prefabrication.

8.6.1 Bent Fabrication

Tubular members are shop prefabricated to full length, complete with stiffening rings,
thickness changes, and the saddle shaped end cope, ready for welding into the structure.
 349

Automated processes at fixed work stations can be brought to bear on this work. The complex
geometry problems are solved at a central location, which can maintain the necessary skill pool.
The field weld is in the T-Y-K-connection at the nodes, which is always a manual weld in any
case. The level of skill required of welders in the field depends on whether or not complete joint
penetration welds are required.
Reference 21 describes computerized prefabrication of 6320 pieces of pipe into 2346
members and subassemblies, for a 50,000-ton tubular structure. In principle, this system is also
applicable to smaller jobs. Items considered include:

(1) checking the designer's "approved for construction" drawings for consistency, and
resolving errors
(2) preparation of shop drawings
(3) inventory and quality control for many different grades or types of steel plate
(4) marking and cutting of individual pieces
(5) cold-forming
(6) fitting and welding
(7) member end profiling
(8) non-destructive weld inspection and dimensional tolerance checks
(9) scheduling and tracking members and subassemblies
(10) transportation to the erection site in shipload lots

In this case, the prefabrication subcontractor was also a steelmaker, with research and
engineering integrated with the production staff at each site, enabling him to optimize the total
system from blast furnace to delivery of the "giant Erector Set". For example, a lean steel
chemistry and thermo-mechanical control processing for the plate was designed to accommodate
high efficiency welding procedures during fabrication, while maintaining strength and notch
toughness; and the welding procedures incorporated "countermeasures" to avoid degrading the
steel properties (ref. 22).

8.6.2 Node Prefabrication

In this approach, the structural connections (or nodes) are prefabricated, complete with
all internal stiffening and stubs for each of the members to be attached. During erection and
assembly, these are attached to straight pieces of pipe with simple butt welds (or "closure"
welds). As compared to bent fabrication, even more of the specialized work takes place under
shop conditions, and the nodes can be stressed relieved if warranted by thickness and criticality.
Specailized welder skills are still required in the field if open root pipe butt welds are used for
closure; he're!, the use internal backing rings presents an interference problem in fitting straight
pipe members into place between the nodes (ref. 23).
Special care is required to maintain proper orientation of the member stub ends, so as to
avoid "dog-leg" members after assembly. This involves not only careful layout and thorough
dimensional checking, but also careful control of weld distortion. One fabricator used
continuous light-beam-and-target monitoring during welding to select a welding sequence which
balanced the distortion (ref. 24).
In the early days of North Sea platform fabrication, nodes were produced by pressure
vessel shops all over Scotlanda kind of modern-day "cottage industry"and inspectors used
Jaguar roadsters to make their rounds. This approach is rarely used in the USA.
350
8.7 ASSEMBLY

8.7.1 Erection Methods

Reference has already been made to the bent method of fabrication. In the construction
of offshore jackets, planar truss subassemblies of the space frame are fabricated at ground level,
then rolled up into position for incorporation into the space frame (Fig. 8.10). Infill members are
placed one-at-a-time with cranes, which hold onto each member until it has been fitted and
secured in place. Working at large heights, this method is quite intensive in its usage of large
crawler cranes. In shipyards, portal cranes and rail-mounted tower cranes have also been used
for this purpose. This traditional approach is described in Graff's textbook (ref. 25).

Fig. 8.10. Truss roll-up.

Alternative jacket erection methods are described in Reference 26. These include:
fabrication of "core block" subassemblies (which are in turn fabricated by the bent method and
then rolled up as a space structure); the use of previously erected core blocks to winch up large
bents; and the use of creeper platforms with stiff-leg derricks, atop the structure, for infill work.
Most of the foregoing erection methods also work equally well for node prefabrication,
except for the vulnerability of free stub ends to damage during roll-ups and similar operations.
The completed jacket is typically skidded onto a barge, transported to its offshore
installation site, and launched at sea. The jacket is upended and set on the seafloor by selective
flooding, often with the assistance of a large crane barge (barges with 250 to 1500-ton lift
capacity are widely available). Steel pipe piling (typical diameters of 30-in. to 84-in., 0.75m to
2.10m) are driven through the jacket legs, each developing 1000 to 8500 tons ultimate capacity.
In addition to offshore oil-drilling structures, this method has been used for docks, bridge piers,
and the like.
Large welded tubular roof structures have been assembled at ground level and then
winched or jacked into place as a single unit, e.g., the Chiangi hangar in Singapore. The lack of
a well-developed standard practice for quick bolted connections in tubular structural members
has tended to discourage the use of more conventional building erection techniques.
 351
8.7.2 Tolerances
Dimensional tolerances for tubular structures erected by the bent method must be fairly
tight in order to keep within the root gap tolerances for welding in -, Y-, and K-connections.
This tolerance range is typically only 1/8-inch (3mm). A properly fitting weld root is essential
for making an open-root, single-sided, complete-penetration weld. Mis-cut end preparations are
difficult enough to get right in the first place, but almost impossible to correct in-place. Diagonal
members in gap K-connections can sometimes be corrected by shifting the centerline slightly, but
this option is not available for overlapping connections. Conscientious inspectors often observe
the fit-up work in progress, subsequently concentrating ultrasonic testing on brace-ends which
have been "trimmed in" with the use of arc-air gouging.
Dimensional adjustments are somewhat easier with node prefabrication, as they involve
simpler butt weld end preparations on the infill brace members.
The fit-up tolerance for fillet welded -, Y-, and K-connections is somewhat looser,
with a working range of 3/16-inch (5mm)yet another advantage of not specifying complete
penetration welds where they are not really required.
Because the end cope on branch members wraps around the chord (especially in the case
of circular tubes with nearly equal diameters), interference problems may develop in swinging
members into place. Some of these problems can be avoided in planning the assembly and
erection sequence, or by leaving everything unwelded until all the members are in place. Where
interference is unavoidable, it may become necessary to install some members in two parts, with
a closure weld as in node prefabrication; or to cut a "construction window", temporarily
removing one side of one end. These windows should be engineered, rather than done on an ad-
hoc basis, so as to assure conditions for sound welding when the window is replaced.

8.8 INSPECTION

It is important for the Engineer and Inspector to understand each other. The Inspector
should understand the overall fracture control philosophy which has been applied during design
and material selectionas well as the practical details of welding, inspection, and paperwork with
which he will be most intimately involved. The Engineer should know what kinds of defects
might go undetected by inspectionin addition to the computer printout and stress calculations
with which he usually deals.
In general, inspection is divided into two separate functions.

(1) Quality Control (QC), usually performed by the fabricator and erector. The
purpose of QC is to make sure that the structure gets built right in the first instance.
QC functions may be performed by a separate organization on the contractor's
staff, by supervisory personnel, and by the workmen themselves.

(2) Quality Assurance (QA), or verification inspection, performed by the structure

owner, a government agency, or a third party inspector acting on their behalf.

All welds must be visually inspected. Often this is the only inspection they receive,
other than observation of work in progress (e.g., checking fit-ups). Such inspection assures that
welds of the proper size have been completed; that the weld profile is acceptable; that the weld
has not fractured completely in two; and that the weld is free from obvious cracks, porosity, or
fusion defects extending to the visible surface.
Magnetic Particle Inspection (MPI) may be useful as an adjunct to visual inspection,
particularly on fatigue-critical connections where weld profile and small surface flaws are
important. Most MPI indications can be eliminated with light grinding, and where this is done
352

routinely as part of the inspection, improved performance is obtained with a minimum of

disruption.
B e c a u s e of the complex geometry of welded tubular c o n n e c t i o n s , standard
radiographic techniques are not applicable, and we are left with ultrasonic testing (UT) as the
only viable method for inspecting the internal quality of completed welds in tubular connections.
This topic has been previously discused in Section 7.4.
Additional discussion of welding and inspection can be found in Reference 27. Duties
of the Inspector are outlined at the beginning of Section 6 of the AWS Structural Welding Code.
This is our ultimate reference.

REFERENCES

1 API Spec 2B, Specification for Structural Steel Pipe, American Petroleum Institute, Washington,
DC.
2 Carter, R.M., et al, Materials Problems in Offshore Platforms, Proceedings Offshore Technology
Conference, OTC 1043, May 1969.
3 Kurobane, Y., New Developments and Practices in Tubular Joint Design, IIW Doc. XV-81-010,
Oporto, 1981.
4 IIW s/c SV-E, Design Recommendations for Hollow Section Joints, International Institute of
Welding annual assembly, Helsinki, 1989.
5 Besuner, P. M., et al, Fracture Control for Fixed Offshore Structures, SSC-328, Ship Structure
Committee, Washington, DC, December 1984.
6 Marshall, P. W., Advanced Fracture Control Procedures for Deepwater Offshore Towers, Welding
Journal, January 1990; also see OTC 6387, May 1990.
7 Linnert, G. W., Welding Metallurgy (2 vols.), American Welding Society, Miami, 1965-67.
8 Chen, W. F., et al, Tubular Members in Offshore Structures, Pitman Press, Boston 1985.
9 Miller, C. D., Buckling of Axially Compressed Cylinders, Journal Structural Division, ASCE,
March 1977.
10 Sherman, D. R., Bending Capacity of Fabricated Pipe at Fixed Ends, report to API, University of
Wisconsin, December 1985.
11 AISC Steel Construction Manual, 8th edition, American Institute of Steel Construction, Chicago,
1980.
12 API Recommended Practice for Planning, Designing and Constructing Fixed Offshore Platforms,
API RP 2A, 18th edition, American Petroleum Institute, 1989.
13 Watts, E. F. and Rule, J. T., Descriptive Geometry, Prentice Hall, 1946.
14 Luyties, W. H., et al, Local Dihedral Angle Equations for Tubular Joints and Related Applications,
Welding Journal, April 1988.
15 Appendix G, Local Dihedral Angle, in AWS Dl.1-90, Structural Welding Code - Steel.
16 Marshall, P. W., et al, BATMIN/CDG Welding Studies, Shell Oil Construction Design Group,
New Orleans, December 1968.
17 Bennett, R. W., Design for Welding Offshore Structures, Proceedings Offshore Technology
Conference, OTC 1230, May 1970.
18 H.O.C.E. Hilites, September 7,1984, and September 19,1986.
19 Notes of joint meeting AWS s/c Tubular Sructures and API W/G Fatigue, November 14, 1984,
Houston.
20 Richard, J. R., Evaluation of ASTM A500 Tube Steel Comer Radii and Effective Throat of Flare
Bevel Groove Welds, Bechtel Technical Report 8409-05 EV, September 1984.
21 Takazawa, S., et al, Computerized Prefabrication and Materials for... Bullwinkle, Proceedings
Offshore Technology Conference, OTC 6094, May 1989.
22 Marshall, P. W., Design of Internally Stiffened Tubular Joints, Safety Criteria in Tubular
Structures, IIW/, Tokyo, July 1986..
23 Post, J. W., Gaining Confidence with Fabrication, Welding, and Inspection of Tubular
Connections, or All You Ever Wanted to Know About Tubular Structures, But Were Afraid to Ask,
presented at AISC Engineering Conference, Nashville, June 1989.
 353

24 Gorton, . ., et al, The Fabrication of Deck Nodes for North Sea Oil Drilling Rigs, Proceedings
International Conference Welding in Offshore Construction, The Welding Institute, Newcastle,
February 1974..
25 Graff, W. J., Introduction to Offshore Structures, Gulf Publishing Company, Houston, 1981.
26 Digre, K. A. and Rodrigue, M. J., Fabrication of the Bullwinkle Platform, Proceedings Offshore
Technology Conference, OTC 6051, May 1989.
27 Marshall, P. W., et al, Welding and Inspection, Chapter 27 in Planning and Design off Fixed
Offshore Platforms, van Nostrand Reinhold, New York, 1986.
 APPENDIX I

SYMBOLS & NOTATION

NOTE Distinctions between Roman and Italic type are made in the AWS Code, but generally
not herein, due to a limitation of the typeface used for equations. Corresponding IIW
nomenclature, where different, is shown in parentheses.

A area
A
o gross area of chord
Av effective area for chord beam shear (IIW)
branch member utilization (eqn. 3.30)
AF amplification factor
ASD allowable stress design
a(hi) width of rectangular hollow section in plane of truss; material notch sensitivity
a flaw size (depth)
a
f terminal flaw size
a
o initial flaw size
footprint length
@ at
B B effective width of ring
e> eff
safety index
b(bj) width of rectangular tubes (branch member)
b
gap effective width at gap of K-connection
b b
eo( e> branch effective width at chord
b b
eoi( ep) branch effective width for outside punching
b
et*e(ov)) branch effective width at thru member
b
m effective width or length of web in I or section chord (IIW)
CHS circular hollow section
COV coefficient of variation
CTOD crack tip opening displacement
c comer dimension; algebraic variable; other dimension as shown in context
D outside diameter OD (circular tubes) or width (box sections) of main member
DSR damaged strength rating
D cumulative fatigue damage ratio, n/N
d
ave average branch width in box K-connections
diameter of branch member
da/dN crack growth per cycle
modulus of elasticity; empirical exponent; effective throat
EFF, Ej static strength efficiency (connection/member)
E
v punching shear efficiency (Vp/Fy Q)
e offset or eccentricity
F toe fillet weld size; scale factor
FM,FT,FV dimensionless force, thrust, shear in Roark's ring solution, respectively
FTE fracture transition elastic
FTP fracture transition plastic
F
Du characteristic ultimate strength of diaphragm
 355

F
Pu characteristic ultimate strength of sidewall panel
F F F
0> 1' 2') forces, e.g., Figure 2.27
F 3, F 4, F 5
F
1
x Fy
, 1 forces in x,y direction, respectively
F
l total capacity developed along sides of box connection
F
2 total capacity developed along heel and toe of box connection
web crippling or buckling stress of main member
F
EXX
electrode tensile strength
F
ult ultimate strength
yield strength of base metal (in general, for member i)
1 v
yo ycr yield strength of main member
f
a axial stress in main member (also see / a )
f
b bending stress in main member (also see f^)
f
v beam shear
/"() function of
axial stress in branch member
u
fb bending stress in branch member
fby nominal stress, in-plane bending
fbz nominal stress, out-of-plane bending
fn nominal stress in branch member
ef/n alpha-modified effective nominal stress
/w stress in weld
A/2 fillet dimensions
GPSS Gauss point surface stress
g gap in K-connections
g' non-dimensional gap g/t Q (IIW)
transverse gap in multi-planar connection
weld height, Figure 2.18; web depth (box chord)
h width of diaphragm on either side of access hole
h
i height of member i in plane of truss (IIW)
ID inside diameter
J ratio of the out-of-plane loads to the in-plane loads
j denotes thru member at overlap
empirical coefficient; reserve strength factor
K- connection configuration
K
R reserve strength due to redistribution
K
S reserve strength due to strain hardening
k effective length factor
f
k fatigue strength reduction factor
*a relative length factor
*b relative section factor (bending)
^by for in-plane bending
*bz for out-of-plane bending
K
\ox for torsion
L size of fillet; length of column; weld dimension (Detail A of Fig. 2.18)
LAST lowest anticipated service temperature
LRFD load and resistance factor design
L
i length of line segment i
L length of joint can or shell
h actual weld length where branch contacts main member
356

I2 projected chord length (one side) of overlapping weld

^ summation of actual weld lengths
applied moment
Mc moment in chord
M M
1> 2'! moments
M 3 , etc \
plastic moment
ultimate moment
It

My yield moment
Mjpg ,M in-plane bending moment
Mopg,Mz out-of-plane bending moment
membrane traction
0
N- connection configuration (K-connection with 9 0 branch)
NDS normalized dimensionless stress
NDT nil ductility transition
Njj> number of parallel load paths
j lifetime total number of cycles
Nf cycles to failure; N j initiation; N 3 thru crack
N^* factored design axial capacity for branch i (IIW)
N Q* reduced chord axial capacity in presence of shear at gap (IIW)
number of cycles allowed
international U
cycles of load applied; number of tests
OD outside diameter
Oy fraction of overlap (IIW)
P S F j Q a (j partial safety factor for load (in LRFD)
P P
1> 2> I forces
P 3 ,etc. {
j o r capacity of or Y-connection
Pc axial load in chord
Pcr critical axial load
Pf probability of failure
P^ capacity of K-connection
Pn nominal resistance or capacity
P ul t, P u ultimate load
Pv yield load
projected footprint length of overlapping member
axial load in branch member
individual member load component perpendicular to main member axis
Q line load
Q average Q
Q&T quenched-and-tempered
Qa AISC reduction factor for effective area
(5 fully plastic Q
load or demand
Qs AISC reduction factor for net section stress
factor reflecting influence of spacing on capacity for transverse line loads
q amount of overlap
Qp plastic reserve factor
Qu geometry and load pattern modifier for total load format
 357
geometry modifier
Qf stress interaction term
branch member geometry and load pattern modifier
root opening (joint fit-up)
RF redundancy factor
RHS rectangular hollow section
RMS root-mean-square
R
m mean resistance or capacity
R
n nominal resistance or capacity
outside radius, main member
r comer radius of rectangular hollow sections as measured by radius gage; radius of
gyration; notch tip radius
r
b radius of branch
mean radius
r
w radius to weld centroid
S elastic modulus; membrane shear; surface distance
S r
tot total range of nominal stress in branch (IIW)
pseudo punching shear for box K-connection
SCE saturated columel electrode
SCF stress concentration factor
SNCF strain concentration factor, e ^ / M n
s,a standard deviation
T ( t 0) shell thickness of chord (also t c )
T- connection configuration
TCBR tension/compression or bending, or both, total range of nominal stress
TMCP thermo-mechanical controlled process of steelmaking
thickness of joint can or outer sleeve in grouted connection
can effective chord thickness for grouted connection
T
eff thickness of concentric pile inside chord
T
pite wall thickness of tube as defined by context; branch tube thickness
t(tp thicknesses at knife-edge crossing
h>*2 wall thickness of branch member; branch member for dimensioning of complete
joint penetration groove welds; thinner member for dimensioning partial
penetration groove welds and fillet welds
wall thickness of main member; joint can thickness (also T)
alternative formulations for (not recommended)
V
weld size (effective throat); web thickness (HW)
U utilization ratio of axial and bending stress to allowable stress, at point under
consideration in main member; x-axis deflection
2
special root-sum-squared form of U for combined axial and bending in chord
UT ultrasonic testing
UTS ultimate tensile strength
V y-axis deflection; shear (IIW)
V V
1> 2> 3
V shear force

W c I shear line load capacity

etc. ( punching shear stress; beam shear yield capacity (IIW, Table 5.3 only)
ultimate punching shear
allowable stress for weld between branch members
w backup weld width; z-axis deflection
358

Wg external work
Wj internal work
WP work point
w line load
w , w w
a b' c I line load capacity
etc. )
X- connection configuration (cross connection)
, axis; algebraic variable; circumferential angle in ring solution
Y(a) local geometry correction term for fracture mechanics
Y- connection configuration
y,y algebraic variable; axis; IPB
plastic section modulus; loss factor at root of weld; shell parameter
2, loss factor axis; OPB
a (alpha) chord ovalizing parameter; spreading slope; non-dimensional shell length
a non-dimensional factor for chord flange effectiveness in shear (IIW)
aQff effective alpha (eqn. 4.7a)
a' notch severity parameter
angular rotation at line i
(beta) diameter ratio of d to D; ratio of r^ to R (circular sections); width ratio of b
to D (box sections)
j3eff effective for k-connection chord face plastification
)3 e 0p dimensionless effective width for outside punching
0 g ap dimensionless effective width at gap of k-connection
7 (gamma) main member flexibility parameter ratio R to t c (circular sections); ratio
of D to 2 t c (box sections)
7^ radius to thickness ratio of tube at transition
7 ef f effective gamma R / T e ^
7 mj resistance factor (IIW)
7t thru member 7 (for overlap connection)
cyclic range of stress intensity factor (fracture mechanics)
, (delta) deflection;
cyclic range of reference stress
6c , crack-tip opening displacement
e (epsilon) applied strain (peak tensile)
e t o ts a t r am r a m e
TR * 5
y yield strain
$ (zeta) non-dimensional gap g/D
(eta) ratio of footprint length to chord diameter or width
(theta) acute angle between two member axes; angle between member center
lines; brace intersection angle
0QL angle of overlapping branch
^TH angle of through branch
0 c ,HC angle of compression branch
# t ,HT angle of tension branch
(lambda) interaction sensitivity parameter; dimensionless slendemess ratio
(nu) Poisson's ratio
(xi or squiggly) Weibull shape parameter
(pi) ratio of circumference to diameter of circle
(rho) angular location on branch member
partial safety factor on system and analysis effects
 359
system factor in LRFD
OQ geometric hotspot stress, excluding notch effects
local or microscopic stress, including some notch effects
a
HS> ( hot spot stress
a
hotspot)
a , ( 7 , i 7 3
K LL DL P ^ * safety factor on load, live or dead
0 residual stress
Oy yield stress at crack tip
A a m ax once-in-a-lifetime extreme stress range
r (tau) branch-to-main relative thickness; ratio of t^ to t c (also t/T)
r l im limiting ^ 0 for eqn. 5.19
r
t ^verlap/^hru
(phi) curvature or rotation; LRFD resistance factor
joint included angle; angle defining circumferential position; angle defining yield
line pattern
(psi) local dihedral angle; reserve strength factor
(psi bar) supplementary angle to the local dihedral angle; angle change at
transition
*0 initial reserve strength of a structural system, collapse load/nominal capacity
ratio of applied to ultimate axial load
ratio of applied to ultimate bending moment
(omega) end preparation angle; partial circumference angle (eqn. 6.4)
 APPENDIX II

DESIGN REVISIONS - TUBULAR STRUCTURES

AS APPROVED - OCTOBER 1. 1 9 9 0

Note - Tables and Figures to be renumbered when issued

in Dl.1-92. Numbers in this draft follow Dl.1-88.

10.2.5.3 The provisions for welded tubular connections are not intended for use with
circular tubes having a specified minimum yield, F y , over 60,000 psi (415
MPa) or for box sections over 52,000 psi (360 MPa).

Part

Allowable Unit Stresses

10.3 BASE METAL STRESSES

These provisions may be used in conjunction with any applicable design specifications in
either allowable stress design (ASD) or load and resistance factor design (LRFD) formats.
Unless the applicable design specification provides otherwise, tubular connection design
shall be as described in 10.5, 10.6, and 10.7. The base metal stresses shall be those
specified in the applicable design specifications, with the following limitations:

10.3.1 Limitations on diameter/thickness for circular sections, and largest flat

width/thickness ratio for box sections, beyond which local buckling or other
local failure modes must be considered, shall be in accordance with the
governing design code. Limits of applicability for the criteria given in 10.5
shall be observed as follows:

(a) circular tubes: D/t < 3 3 0 0 / F y

(b) box section gap connections: D/t < 210/VFy but not more than 35
(c) box section overlap connections: D/t < 190/VFy

10.3.2 Moments caused by significant deviation from concentric connections shall

be provided for in analysis and design. See figure 10.1.2 (H).
 361

10.4 UNIT STRESSES IN WELDS

10.4.1 Except as modified in 10.5, 10.6, and 10.7, the allowable stress in welds
shall be as shown in Table 10.4.1.

10.4.2 Fiber stresses due to bending shall not exceed the values prescribed for
tension and compression, unless the members are compact sections, (able to
develop full plastic moment) and any transverse weld is proportioned to
develop fully the strength of sections joined.

10.4.3 Plug or slot welds shall not be ascribed any value in resistance to stress other
than shear in the plane of the faying surfaces.

Type of Tubular Application
Weld
Longitudinal seam of
tubular members
Butt splices of tubular
members
Structural -, Y-, or K-
connections in structures
Complete Joint Penetration groo\re weld
designed for critical bading
such as fatigue, which
would normally call for
complete joint penetration
welds
Longitudinal seam of
tubular members
=8
[*|
Structural -, Y-, or K-
connection in ordinary
structures; lap splice of
tubular members
Table 10.4.1
Allowable Stresses in Welds 362
ASD LRFD
Kind of Stress Permissible Resistance Nominal Required
Unit Stress Factor^ Strength Metal
Strenqth Level *
Tension or compression parallel to axis of the
weld 2 Same as for base metal3 0.9 Weld metal with a
strength level equal to
or less than matching
weld metal may be
used.
Compression normal to the effective area 2 0.9
Matching weld metal
Base metal 0.6 Fy must be used. See
Same as for base metal 0.9 Table 4.1.1.
Shear on effective area Weld metal 6
0.8
Tension normal to the effective area. 0.9 Fy
Same as base metal or Same as base metal or as limited Matching weld metal
Tension, compression or shear on base metal as limited by connection by connection geometry (see 10.5 must be used. See
adjoining weld conforming to detail of Fig. geometry (see 10.5 provisions for LRFD) Table 4.1.1
10.13.1 (tubular weld made from outside only). provisions for ASD)
Tension compression, or shear on effective area
of groove welds, made conventionally from both
sides or with backing.
Tension or compression parallel to axis of the Same as for base metal 0.9 Fy Weld metal with a
weld. strength level equal to
less than matching
weld metal may be
Shear on effective area. 0 3 xF used.
EXX
0.75 0 6 F
EXX
0.75 O.OFEXX
Shear on effective throat regardless of direction 0 . 3 0 F j r x x or as or as limited by connection Weld material with a
of bading. (See 10.8 and 10.5.1.3) limited by connection geometry (see 10.5 provision for strength level equal to
geometry (see 10.5) LRFD) or less than matching
weld metal may be
used. ^
 Longitudinal seam of Tension or compression parallel to axis of Same as for base metal^ 0.9 F Weld metal with a strength level
y
tubular members the weld 2 equal to or less than matching
weld metal may be used.

0.50 F g x x , except that

Joint not stress on adjoining base
Compression designed metal shall not exceed
normal to to bear 0.60 Fy. Weld metal with a strength level
the effective 0.9 F equal to or less than matching
y
throat Joint weld metal may be used.
designed Same as for base metal
Butt splices of tubular to bear
members

Shear on effective throat 0.30 F x X , exept that 0.75 Weld metal with a strength level
stress on adjoining base equal to or less than matching
metal shall not exceed weld metal may be used.
Tension on effective area 0.50 F for tension, or base metal 0.9 F
y
0.40 Fy for shear. weld metal 0.8 0 6 F
EXX

0.30 F g x x or as base metal 0.9 Matching weld metal must be

limited by connection weld metal 0.8 o.eFlxx used. See Table 4.1.1

Partial joint
penetration
groove weld
Structural - , Y - , or K - Load transfer across the weld as stress on geometry(see 1 0 . 5 ) ,
connection in ordinary the effective throat (see 10.8 and 10.5.1.3) except that stress on an
structures adjoining base metal shall
not exceed 0.60 Fy for
tension and compression,
nor 0.40 Fy for shear.
or as limited b y c o n n e c t i o n
geometry (see 10 5 provisions for
LRFD)

1. For matching weld metal see Table 4.1.1.

2. Beam or torsional shear up to 0.30 minimum specified tensile strength of weld is permitted, except that shear on adjoining base metal shall not exceed 0.40 Fy (LRFD; see shear).
3. Groove and fillet welds parallel to the longitudinal axis of tension or compression members, except in connection areas, are not considered as transferring stress and hence may take the same
stress as that in the base metal, regardless of electrode (filler metal) classification. Where the provisions of 10.5.1 are applied, seams in the main member within the connection area shall be
complete joint penetration groove welds with matchingfillermetal, as defined in Table 4.1.1.
4. See 10.5.3.

r^EXX - specified minimum tensile strength of weld metal.

Fy = specified minimum yield strength of base metal.
363
364

Figure 10.5.1 - Punching Shear Stress

Parameter Circular Sections Box sections

tJR or d b / D b/D
a x/ D
y */tc D/2tc
b/e
Angle between member center lines
Local dihedral angle at given point

on welded joint
Corner dimension as measured to
the point of tangency or contact
C
with a 9 0 degree square placed on
the corner

Figure 10.1.2 (M) Geometric Parameters

 365

10.5 LIMITATIONS O F THE STRENGTH OF WELDED TUBULAR

CONNECTIONS

10.5.1. CIRCULAR - Y- and K- CONNECTIONS

10.5.1.1 Local Failure. Where a -, Y-, or K- CONNECTION is made by simply

welding the branch m e m b e r s ) individually to the main member, local stresses
at a potential failure surface through the main member wall may limit the
usable strength of the welded joint. The shear stress at which such failure
occurs depends not only upon the strength of the main member steel, but
also on the geometry of the connection. Such connections shall be
proportioned on the basis of either (1) punching shear or (2) ultimate load
calculations as given below. The punching shear is an allowable stress
design (ASD) criterion and includes the safety factor. The ultimate load
format may be used in load and resistance factor design (LRFD), with the
resistance factor to be included by the designer; see 10.6.2.

(1) Punching Shear Format - The acting punching shear stress on the potential failure
surface (see Figure 10.5.1) shall not exceed the allowable punching shear stress.

The acting punching shear stress is given by

acting V p = r / n s i n

The allowable punching shear stress is given by

allow V p = Q q Q r F y o / ( 0 . 6 T )

The allowable V p shall also be limited by the allowable shear stress specified in
the applicable design specification (e.g., 0.4 F y o) .

Terms used in the foregoing equations are defined as follows:

r, 0, , and other parameters of connection geometry are defined in

Figure 10.1.2(M).
/ n is the nominal axial (fa) or bending (f^ stress in the branch member
(punching shear for each kept separate)^ *
F y o = The specified minimum yield strength of the main member chord, but
not more than 2 / 3 the tensile strength.
Qq, Qf are geometry modifier and stress interaction terms, respectively,
given in Table 10.5.1.

4 1 . For bending about two axes (e.g., y and z), the effective resultant bending stress may be
taken as :
366

Table 10.5.1
Terms for Strength of Connections
(Circular Sections)

Branch member 1.7 0.181 0.7 ( - 1) For axial loads

Geometry and load Qp (see note 6)
modifier Qq <= L J
\2 061 1.2 (a-0.67) For bending
L < J

Qs = 1.0 For * 0.6

(needed 0.3
for Q q (1 - 0.833 ) For > 0.6

chord 1.0 + 0.7g/d b For axial load in gap k-connections having all
members in same plane and loads transverse to
ovalizing 1.0* < 1.7 main member essentially balanced (See Note 3)

parameter 1.7 For axial load in T- and Y-connections

2.4 For axial load in cross connections
OL =
(needed 0.67 For in-plane bending (see Note 5)
forQ q ) 1.5 For out-of-plane bending (see Note 5)

2
Main member stress 1.0-
interaction term = 0.030 For axial load in branch member
Qf = = 0.044 For in-plane bending in branch member
(See Notes 4 and 5) = 0.018 For out-of-plane bending in branch member

Notes:

3. Gap g is defined in Figs. 10.1.2.E, F, and H; d b is branch diameter.

4. U is the utilization ratio (ratio of actual to allowable) for longitudinal compression

(axial, bending) in the main member at the connection under consideration.

2
U ~
0.6 F J 0.6
/ W F
L , V J
yo * yo
5. For combinations of the in-plane and out-of-plane bending, use interpolated
values of a and .

6. For general collapse (transverse compression) also see 10.5.1.2.

 367

For combined axial and bending stresses, the following inequality shall be
satisfied:
1 7 5
[acting V 1 r acting V D 1 ^10
I _L| + j
Lallow Vp j a x i al Lallow Vp J 5 e n d i gn

(2) LRFD Format (loads factored up to ultimate condition - see 10.6)

Branch member loadings at which plastic chord wall failure in the main member
occurs are given by:
2
axial load: P u sin = t c F y Q [6*0Q q] Qf
2
bending moment: M u sin = t c F y Q [d^/4] [/JQq] Qf

with the resistance factor = 0.8.

2 2 2
Qf should be computed with U redefined as ( P c / A F y Q ) + ( M c / S F y Q ) where
P c and M c are factored chord load and moment, A is area, S is section modulus.

These loadings are also subject to the chord material shear strength limits of:

P u s i n 0 * * d b t c F y (/ V 3
2
M u sin * d 5 t c F y Q/ V 3

with = 0.95

where t c is chord wall thickness

d^ is branch member diameter and other terms are as defined in
10.5.1.1(1).
The limit state for combinations of axial load and bending moment is given
by:
1 75
(P/Pu) + M/Mu s l . O

10.5.1.2 G e n e r a l Collapse. Strength and stability of a main member in a tubular

connection, with any reinforcement, shall be investigated using available
technology in accordance with the applicable design code.

General collapse is particularly severe in cross connections and connections

subjected to crushing loads; see Figures 1 0 . 1 . 2 ( G ) and (J). Such
connections may be reinforced by increasing the main member thickness, or
by use of diaphragms, rings, or collars.
368

(1) For unreinforced circular cross connections, the allowable transverse

chord load, due to compressive branch member axial load P, shall not
exceed
2
P s i n 0 = t c F y( 1 . 9 + 7.20)Q0Qf

(2) For circular cross connections reinforced by a "joint can" having increased
thickness t c , and length, L, the allowable branch axial load, P, may be
employed as

= P ( 1 )+ [P ( 2) - P(i)l L/2.5D for L < 2.5 D

= P(2) for L * 2.5D

where is obtained by using the nominal main member thickness in the

equation in (1); and P ^ ) is obtained by using the joint can thickness in the
same equation.

The ultimate limit state may be taken as 1.8 times the foregoing ASD
allowables, with = 0.8.

(3) For circular K-connections in which the main member thickness required
to meet the local shear provisions of 10.5.1.1 extends at least D / 4 beyond
the connecting branch member welds, general collapse need not be checked.

10.5.1.3 Uneven Distribution of Load (Weld sizing)

(1) Due to differences in the relative flexibilities of the main member loaded
normal to its surface, and the branch member carrying membrane stresses
parallel to its surface, transfer of load across the weld is highly non-uniform,
and local yielding can be expected before the connection reaches its design
load. To prevent "unzipping" or progressive failure of the weld and insure
ductile behavior of the joint, the minimum welds provided in simple -, Y-,
or K-connections shall be capable of developing, at their ultimate breaking
strength, the lesser of the brace member yield strength or local strength
(punching shear) of the main member. ^ 2

(2) This requirement may be presumed to be met by the prequalified joint

details of Figures 10.13.1 A (complete penetration) and 10.13.2 (partial
penetration), when matching materials (Table 4.1.1) are used.

42. The ultimate breaking strength of fillet welds and partial joint penetration groove welds
shall be computed at 2.67 times the basic allowable stress for 60 ksi (414 MPa) or 70 ksi
(483 MPa) tensile strength and at 2.2 times the basic allowable stress for higher strength
levels. The ultimate punching shear shall be taken as 1.8 times the allowable V n of
P
10.5.1.1.
 369

(3) Compatible s t r e n g t h of welds may also be p r e s u m e d with the

prequalified fillet weld details of Figure 10.13.3, when the following effective
throat requirements are met:

(a) = 0.7 t^ for elastic working stress design of mild steel circular steel
tubes (Fy s 4 0 ksi or 2 8 0 MPa) joined with overmatched welds
(classified strength FEXX = 70 ksi).

(b) = 1.01^ for ultimate strength design (LRFD) of circular or box tube
connections of mild steel, Fy 40 ksi (280 MPa), with welds satisfying
the matching strength requirements of Table 4 . 1 .

(c) = lesser of t c o r 1.07 t^ for all other cases

NOTE: revise table in Figure 1 0 . 1 4 (old 1 0 . 1 3 . 3 ) follows

MIN L for

= 0.7t E=t E = 1.07 t

heel < 60 1.5t 1.5t larger of 1.5t

or 1.4t +

side * 100 t 1.4t 1.5t

side 100-110 l.lt 1.6t 1.75t

side 110-120 1.2t 1.8t 2.0t

t o e > 120 t 1.4t full bevel

bevel bevel 60-90 groove

(4) Fillet welds smaller than those required above to match connection
strength, but sized only to resist design loads, shall at least be sized for the
following multiple of stresses calculated per 1 0 . 8 . 3 , to account for
nonuniform distribution of load:

ASD LRFD
E60XX and E70XX - 1.35 1.5
Higher strengths - 1.6 1.8

10.5.1.4 Materials Considerations for Base Metal Selection

(1) Steel for Tubular Connections. Tubular connections are subject to local
stress concentrations which may lead to local yielding and plastic strains at
the design load. During the service life, cyclic loading may initiate fatigue
cracks, making additional demands on the ductility of the steel, particularly
under dynamic loads. These demands are particularly severe in heavy-wall
joint-cans designed for punching shear. See Commentary at CI0.2.6.2.
370

(2) Laminations and Lamellar Tearing. Where tubular joints introduce

through-thickness stresses, the anisoptropy of the material and the
possibility of base metal separation should be recognized during both design
and fabrication. See Commentary.

10.5.1.5 Overlapping J o i n t s , in which part of the load is transferred directly from

one branch member to another through their common weld, shall include
the following checks:

(1) The allowable individual member load component, P_L perpendicular to

the main member axis shall be taken as PJ_ (Vpt cl + 2V wt wl2> where V p
is the allowable punching shear as defined in 10.5.1.1, and

t c = the main member thickness

1-^ = actual weld length for that portion of the branch member which
contacts the main member
V p = allowable punching shear for the main member as K-connection
(a-1.0)
V w = allowable shear stress for the weld between branch members (Table
10.4.1)
= the lesser of the weld size (effective throat) or the thickness, t^, of the
thinner branch member
12 = the projected chord length (one side of the overlapping weld, measured
perpendicular to the main member.

These terms are illustrated in Figure 10.5.1.5.

The ultimate limit state may be taken as 1.8 times the foregoing WSD
allowables, with = 0.8.

Figure 1 0 . 5 . 1 . 5 - Detail of Overlapping Joint

(2) The allowable combined load component parallel to the main member
axis shall not exceed V w t w Z l 1 where I l j is the sum of the actual weld
lengths for all braces in contact with the main member.
 371

(3) The overlap shall preferably be proportioned for at least 50% of the
acting j . In no case shall the branch member wall thickness exceed the
main member wall thickness.

(4) Where the branch members carry substantially different loads, or one
branch member is thicker than the other, or both, the heavier branch
member shall preferably be the through member with its full circumference
welded to the main member.

(5) Net transverse load on the combined footprint shall satisfy 10.5.1.1 and
10.5.1.2.

(6) Minimum weld size for fillet welds shall provide effective throat of 1.0t b
for F y < 40 ksi, 1.2 t b for F y > 40 ksi.

10.5.1.6 Flared connections and tube size transitions not excepted below shall be
checked for local stresses caused by the change in direction at the
transition. (See note 4 to Table 10.7.3). Exception, for static loads:

Circular tubes having D/t less than 30, and

Transition slope less than 1:4.

10.5.1.7 O t h e r Configurations a n d Loads

(1) The term, -, Y-, and K-connections, is often used generically to

describe tubular connections in which branch members are welded to a
main member, or chord, at a structural node. Specific criteria are also
given for cross (X-) connections (also referred to as double-tee) in 10.5.1.1
and 10.5.1.2. N-connections are a special case of K-connections, in which
one of the branches is perpendicular to the chord; the same criteria apply.
See commentary for multiplanar connections.

(2) Connection classification as , & Y, or cross should apply to

individual branch members according to the load pattern for each load
case. To be considered a connection, the punching load in a branch
member should be essentially balanced by loads on other braces in the
same plane on the same side of the joint. In and Y connections the
punching load is reacted as beam shear in the chord. In cross connections
the punching load is carried through the chord to braces on the opposite
side. For branch members which carry part of their load as K-connections,
and part as & Y or cross connections, interpolate based on the portion of
each in total, or use computed alpha (see commentary).
 372

10.5.2 BOX -, -, AND K-CONNECTIONS

Criteria given in this section are all in ultimate load format, with the safety
factor removed. Resistance factors for LRFD are given throughout. For
ASD, the allowable capacity shall be the ultimate capacity, divided by a
safety factor of 1.44/. The choice of loads and load factors shall be in
a c c o r d a n c e with the g o v e r n i n g design specification; see 1 0 . 6 .
Connections shall be checked for each of the failure modes described
below. These criteria are for connections between square and rectangular
sections of uniform wall thickness, in planar trusses where the branch
members loads are primarily axial. If compact sections, ductile material,
and compatible strength welds are used, secondary branch member
bending may be neglected.^
Criteria in this section are subject to the following limitations:
E7

U. F\./j

^ -

0.55H s e * 0.25H
* 30
H / t c and D / t c s 35 (40 for overlap and N-connections)
a / t b and b / t b * 35
F * 52 ksi (360 MPa)
0 5 * H/D s 2.0
F F 0 8
yo/ ult* -

10.5.2.1 Local Failure - Branch member axial load P u at which plastic chord wall
failure in the main member occurs is given by:
2
P us i n f l = F tc r2a 4_ ] Q

for cross, T-, and Y-connections with 0.25 s < 0.85 and = 1.0.
2
Also, P u sin = F y o t c [9.8 0 e ff VT] Q F

with = 0.9 for gap K- and N-connections with least 0eff 0.1 + / 5 0
andg/D >0.5 (1-/3)

4 3 . Secondary bending is that due to joint deformation or rotation in fully triangulated

trusses. Branch member bending due to applied loads, sideway of unbraced frames, etc.
cannot be neglected and must be designed for. See 10.5.2.6.
 373

where F y o is specified minimum yield strength of the main member, t c is

chord wall thickness; is D / 2 t c (D = chord face width); 0, ?, , and $ are
connection topology parameters as defined in Figure 10.1.2M (0 eff is
equivalent defined below); and Qf = 1.3 - O.4U/0 (Qf ^ 1.0; use
Qf = 1.0 for chord in tension) with U being the chord utilization ratio.

U = I fa- I + I fb I
I Fyo I I Fyo
~ I
+ a + + a
^efP ^compression compression ^ tension tension) / ^ D
branch branch branch branch

These loadings are also subject to the chord material shear strength limits
of

P u sin = (F y o/V3) t c D [2 + 2 0 e o p]

for cross, T-, or Y-connections with > 0.85, using = 0.95, and

P u sin = (F y o/V3) t c D [2+ 0g ap + 0 e o p]

for gap K- and N-connections with 0.1 + / 5 0 , using = 0.95

(This check is unnecessary if branch members are square and equal width),
where:
= or K a nd
0gap 0 * " N-connections with <> 1.5 (1 - )
= or a o t n er
0gap 0eop* ^ connections

j3 e Qp (effective outside punching) = 5/3/7but not more than .

10.5.2.2 General Collapse. Strength and stability of a main member in a tubular

connection, with any reinforcement, shall be investigated using available
technology in accordance with the applicable design code.

(1) General collapse is particularly severe in cross connections and

connections subjected to crushing loads. Such connections may be
reinforced by increasing the main member thickness or by use of
diaphragms, gussets, or collars.

For unreinforced matched box connections, the ultimate load normal to

the main member (chord) due to branch axial load shall be limited to

P u sin0 = 2 t c F y o (a x + 5 t c)

with = 1.0 for tension loads,

and = 0.8 for compression.
and
_ _AL
8 2 0 0

P u sin =
H-4tc *YO ' Qf
374

with = 0.8 for cross connections, end post reactions, etc. in

compression
(ksi units)
2
or P u sin = 270 t c [ l + 3 a x / H ] VF y Q Q f

with = 0.75 for all other compression branch loads (ksi units)

(2) For gap K- and N- connections, beam shear adequacy of the main
member to carry transverse loads across the gap region shall be checked
including interaction with axial chord forces. This check is not required
for U ^ 0.44 in stepped box connections having + <, H / D ( is
height of main member in plane of truss).

10.5.2.3 Uneven Distribution of Load (Effective Width). Due to differences in

the relative flexibilities of the main member loaded normal to its surface
and the branch member carrying membrane stresses parallel to its surface,
transfer of load across the weld is highly non-uniform, and local yielding
can be expected before the connection reaches its design load. To
prevent progressive failure and insure ductile behavior of the joint, both
the branch members and the weld shall be checked, as follows:

(1) Branch Member Check - The effective width axial capacity P u of the
branch member shall be checked for all gap K- and N-connections,^
and other connections having > 0.85.

p
ue = Vb |2a + b
gap
+ b
eor b]
4t

with = 0.95

where

Fy is specified minimum yield strength of branch.

t b is branch wall thickness

a, b are branch dimensions (see Figure 10.1.2B)

b g a p = b for K- and N-connections with 1.5 (1-)

b g a p = b e o if o r all other connections
b 5b F
eoi= __ yo ^ b

Note: r ^ 1.0 and Fy < Fy Q are presumed.

44. This check is unnecessary if branch members are square and equal width.
 375

(2) Weld checks - the minimum welds provided in simple -, Y-, or K-

connections shall be capable of developing, at their ultimate breaking
strength, the lesser of the branch member yield strength or local strength of
the main member.

This requirement may be presumed to be met by the prequalified joint

details of Figures 10.13.IB (complete penetration and partial penetration),
when matching materials (Table 4.1.1) are used.

(3) Fillet welds shall be checked as described in 10.8.5.

10.5.2.4 M a t e r i a l s C o n s i d e r a t i o n s - The designer should consider special

demands which are placed on the steel used in box -, Y-, and K-
connections. See Commentary.

10.5.2.5 Overlapped Connections - Lap joints reduce the design problems in the
main member by transferring most of the transverse load directly from one
branch member to the other. The criteria of this section are applicable to
statically loaded connections meeting the following limitations:

(a) The larger, thicker branch is the thru member.

(b) > 0.25.

(c) The overlapping branch member is 0.75 to 1.0 times the size of the
thru member with at least 25% of its side faces overlapping the thru
member.

(d) Both branch members have the same yield strength.

(e) All branch and chord members are compact square or rectangular
tubes with width/thickness s 35 for branches, and ^ 40 for chord. .

The following checks shall be made:

(1) Axial capacity P u of the overlapping tube, using = 0.95 with

P u = F y t b [ Q 0 L (2a - 4t b ) + b e o + b e t) for 25% to 50% overlap, with

~
Q
% overlap
OL " -""50%

P u = F y t b [ ( 2 a - 4t 5 ) + b e o + b e t] for 50% to 80% overlap.

P u = F y t b [(2a - 4t b ) + b + b e t] for 80% to 100% overlap.

P u = F y t b l(2a - 4t b ) + 2b e tl for more than 100% overlap

376

where b e o is effective width for the face welded to the chord,

5 F
b yo
y
beo = - - *b

and b e t is effective width for the face welded to the thru brace.

b et = 5 b / ( 7 tT t M b

with:

7 t = b/(2t^) of the thru brace

T =
t Werlap^thru

and other terms are as previously defined.

 377

(2) Net transverse load on the combined footprint, treated as a T- or Y-

connection.

(3) For more than 100% overlap, longitudinal shearing shall be checked,
considering only the sidewalls of the through branch footprint to be
effective.

10.5.2.6 Bending Moments

Primary bending moment, M, due to applied load,, cantilever beams,

sideways of unbraced frames, etc., shall be considered in design as an
additional axial load, P:

- _!L
JD sin
In lieu of more rational analysis (see commentary), JD may be taken as
r/D/4 for in-plane bending, and as 0D/4 for out-of-plane bending. The
effects of axial load, in-plane bending and out-of-plane bending shall be
considered as additive. Moments are to be taken at the branch member
footprint.

10.5.2.7 O t h e r Configurations

Cross, -, Y-, gap K-, and gap -connections with compact circular branch
tubes framing into a box section main member may be designed using
78.5% of the capacity given in 10.5.2.1 and 10.5.2.2, by replacing the
box dimension "a" and "b" in each equation by branch diameter, d b (limited
to compact sections with 0.4 <> <> 0.8).

10.6 ALLOWABLE S T R E S S E S AND LOAD AND R E S I S T A N C E SAFETY

FACTORS

10.6.1 Allowable S t r e s s Design. Where the applicable design specifications permit

the use of increased unit stresses in the base metal for any r e a s o n , a
corresponding increase shall be applied to the allowable unit stresses given herein,
except for fatigue. The allowable stresses given herein are consistent with a
nominal base metal working stress of 0.6 F v .
378

10.6.2 Load & Resistance Factor Design. Resistance factors, , given elsewhere in
Part of this chapter, may be used in the context of load and resistance factor
design (LRFD) calculations in the following format:

( P u or M u) = I(LF Load)

where P u or M u is the ultimate load or moment as given herein; and LF is

the load factor as defined in the governing LRFD design code, e.g., AISC
Load and Resistance Factor Design Specification for Structural Steel in
Buildings.

10.7 FATIGUE - NO CHANGE

10.8 EFFECTIVE WELD AREA AND LENGTH

10.8.1 Groove Welds. The effective area shall be in accordance with 2.3.1 and the
following: the effective length of groove welds in structural -, Y-, and K-
connections shall be computed in accordance with 10.8.4 or 10.8.5, using the
mean radius r m or face dimensions of the branch member.

10.8.2 Fillet W e l d s . The effective area shall be in accordance with 2.3.2 and the
following: the effective length of fillet welds in structural -, Y-, and K-
connections may be computed in accordance with 10.8.4 or 10.8.5, using the
radius or face dimensions of the branch member as measured to the center line of
the weld.

10.8.3 S t r e s s e s in Welds. When weld allowable stress design calculations are required
for circular sections, the nominal stress in the weld joining branch to chord in a
simple -, Y-, or K-connection shall be computed as:
 379

, *b fa
f
m
r
^
f
b m ]
r 2

Wld
=
~-
*w
\JR
La w
K
r
b
K

r
w J
2

where:

t b is thickness of branch member

is effective throat of the weld

fa and fb are nominal axial and bending stresses in the branch

In ultimate strength or LRFD format the following expression for branch
axial load capacity shall apply for both circular and box sections:

P = ( L
u V eff

where Q w = weld line load capacity (kips/inch) and Leff = weld effective
length.

For fillet welds,

6 t F
Qw = ' w E X X
with = 0.8

where F^^x = classified tensile strength of weld deposit.

K a and are effective length and section factors given in 10.8.4 and
10.8.5.

10.8.4 Circular -, Y-, K-Connections. Length of welds and the intersection length
in circular -, Y-, and K-connections shall be determined as 2nr K a where r is
the effective radius of the intersection (see 10.8.1 or 10.8.2)

2 2
Ka = x + y + 3 V(x +y )

= 1/(2 sin )
2
y = 1 3-0
2
3 2-

Where:

- the acute angle between the two member axes

= diameter ratio, branch/main, as previously defined.

380

Note: the following may be used as conservative approximations:

1 + 1/sin
= for axial load

3 + 1/sin
Ku = for in-plane bending
4 sin

1 + 3/sin
= - for out-of-plane bending

10.8.5 Box Connections

10.8.5.1 The effective length of branch welds in structural, planar, gap K- and N-
connections between Box Sections, subjected to predominantly static axial
load, shall be taken as:

2 a x + b, for * 6 0 :

2 a x + 2b, for ^ 5 0 :

Thus for ^ 50 the heel, toe and sides of the branch can be considered
fully effective. For 6 0 , the heel is considered ineffective due to

uneven distribution of load. For 5 0 < < 6 0 , interpolate.

10.8.5.2 The effective length of branch welds in structural, planar, -, Y- and X-

connections between box sections, subjected to predominantly static axial
load, shall be taken as:

2 a x + 2b, for < 0.85.

2 a y, for > 0.85.

 381

TUBULAR DESIGN REVISIONS - COMMENTARY

APRIL 1 9 9 0

Part
Allowable Unit Stresses in Welds
... [no change to lead paragraphs] ...

NOTE: ALL EXISTING FIGURES TO BE RETAINED.

1 0 . 3 BASE METAL STRESS

10.3.1 Limiting diameter/thickness and width/thickness ratios depend on the

application. Referring to Table C 1 0 . 3 . 1 , the left hand side deals with
connection design issues covered by the AWS Code. The first three columns
delimit stocky members for which simplified design rules apply; beyond these
limits the more detailed calculations given in the Code must be performed.

The limits for designing members against local buckling at various degrees of
plasticity are shown on the right hand side. These are an amalgam of API,
AISC and AISI requirements. Naturally, requirements of the governing design
specification would take precedence here.

10.4 UNIT STRESSES IN WELDS

... [no change] ...

10.5 LIMITATIONS ON THE STRENGTH OF TUBULAR CONNECTIONS

A number of unique failure modes are possible in tubular connections. In

addition to the usual checks on weld stress provided for in most design codes,
the designer should check for:

CIRCULAR BOX

(1) Local failure * 10.5.1.1 10.5.2.1

(2) General collapse 10.5.1.2 10.5.2.2
(3) Progressive failure (unzipping) 10.5.1.3 10.5.2.3
(4) Materials problems 10.5.1.4 10.5.2.4

* Overlapping connections are covered by 10.5.1.5 and 10.5.2.5 respectively.

 TABLE C 1 0 . 3 . 1 382
SURVEY OF DIAMETER/THICKNESS AND FLAT WIDTH/THICKNESS LIMITS FOR TUBES

F O R AWS C O N N E C T I O N DESIGN F O R MEMBER DESIGN

YIELD
APPLIC- PLASTIC MOMENT L I M I T OF
LOCAL GENERAL CONE- ABILITY FULL MOMENTS OR L I M I T OF FULL LOCAL
FAILURE COLLAPSE CYLINDER OF RULES PLASTIC LIMITED ELASTIC YIELD BUCKLING
U L T V p = . 5 7 F y (3 F eo = cF D 1 : 4 FLARI3 in 10.5 DESIGN ROTATION BEHAVIOR AXIAL FORMULAE
y<

16 1300 1500 6000

FOR K - <
F F F 60 300

TUBES
y Y Y
CONNECTION Oi

12 30 3300 <

FOR T&T F 2070 8970 3300 13000

y
co

CIRCULAR
9 F F F F
Y Y Y Y <
FOR X

8 210 < 35 190 210 238 238

V F F F-10 >|F
1 CO
Y Y Y 10
22 20 @ 7 7 = 5 ( F y- 1 0 ) NO co <C
FOR GAP _q
FOR K&N

SECTIONS
<
LIMIT
CONNECTION.
19 F O R || 150 238

BOX
co
VF OVERLAP VF >|F CO
co <C
7 y 1 *4
y :
FOR T&X
II
Fy in ksi (1 ksi = 7 MPa)
A I S I Class A = hot formed
A I S I Class = cold formed and welded
Flat width may be taken as D-3t for box section member design
 383

C 1 0 . 5 . 1 CIRCULAR SECTIONS

C10.5.1.1 Local Failure

The design requirements ... [no change to this paragraph] ...

Treatment of box sections ... [delete this paragraph] ...

Alternatives to the punching shear approach ... [delete this paragraph] ...

In the 1984 edition ... [no change to this 6 paragraph] ...

Figure CI0.2. gives ... [no change to this 3 paragraph] ...

Figure C I 0 . 3 . shows ... [no change to this paragraph] ...

The apparently large safety factor ... [no change to this paragraph] ...

In the 1992 edition, the Code has also included tubular connection design
criteria in ultimate strength format, section 10.5.1.1 (2) for circular sections.
This was derived from, and intended to be equivalent to, the earlier punching
shear criteria. The thin-wall assumption was made (i.e. no t^/d^ correction),
and the conversion for bending uses elastic section modulus.

When used in the context of AISC-LRFD, with a resistance factor of 0.8, this is
nominally equivalent with the allowable stress design (ASD) safety factor of 1.8
for structures having 40% dead load and 60% service loads. The change of
resistance factor on material shearing was done to maintain this equivalency.

LRFD falls on the safe side of ASD for structures having a lower proportion of
dead load. AISC criteria for tension and compression members appear to
make the equivalency trade-off at about 2 5 % dead load; thus, the LRFD
criteria given herein are nominally more conservative for a larger part of the
population of structures. However, since the t ^ / d b correction to punching
shear is not made

acting V p = Tsin0 fn (1 - ^ / d ^

the ASD punching shear format also contains extra conservatism.

Figure C10.5.1C indicates a safety index of 3.45, appropriate for selection of

the joint-can as a member (safety index is the safety margin of the design
criteria, including hidden bias, expressed in standard deviations of total
uncertainty). For further comparison, the ASCE Committee on Tubular
Structures in reference 2 derived a resistance factor of 0.81 for similar Yura-
based tubular connection design criteria, targeting a safety index of 3.0.
384

Since local the failure criteria in 10.5 are used to select the main member or
chord, the choce of safety index is comparable to that used for designing other
structural members - rather than the higher values often cited for connection
material such as rivets, bolts, or fillet welds, which raise additional reliability
issues, e.g. local ductility and workmanship.

For offshore structures, typically dominated by environmental loading which

occurs when they are unmanned, the 1986 draft of API RP2A-LRFD proposed
more liberal resistance factors of 0.90 to 0.95, corresponding to a reduced
target safety index of 2.5 (actually, as low as 2.1 for tension members). API
also adjusted their allowable stress design criteria to reflect the benefit typical
t ^ / d b ratios.

In Canada (ref. 19), using these resistance factors with slightly different load
factors, a 4 . 2 % difference in overall safety factor results. This is within
calibration accuracy.

10.5.1.2 General Collapse

... [no change to this paragraph, old 10.5.2] ...

10.5.1.3 Progressive Failure (Unzipping)

The initial elastic distribution of load transfer across the weld in a tubular
connection is highly non-uniform, with peak line load (kips/inch) often being a
factor of two or three higher than that indicated on the basis of nominal
sections, geometry, and statics, as per section 10.8. Some local yielding is
required for tubular connections to re-distribute this and reach their design
capacity. If the weld is a weak link in the system, it may "unzip" before this re-
distribution can happen. The criteria given in the Code are intended to prevent
this unzipping, taking advantage of the higher safety factors in weld allowable
stresses than elsewhere. For example, the line load ultimate strength of an
0.7t fillet weld made with E70XX electrodes is 0.7t(2.67x0.3x70) = 39t,
adequate to match the yield strength of mild steel branch material. For another
example, if the peak line load is really twice nominal, designing for 1.35 times
the nominal line load will give a joint safety factor of 1.8, when the weld
strength is 2.67 times its allowable stress. IIW rules, and LRFD-based strength
calculations, suggest larger matching weld sizes are required, e.g. l.Ot or 1.2t
(1.07t in the draft Eurocode). Given this easy way out of the problem, there
has not been much testing to validate the foregoing AWS logic for smaller
welds.

10.5.1.4 (2) Laminations and Laminar Tearing

... [no change to these paragraphs, old 10.5.4] ...

 385

C10.5.2 BOX SECTIONS

In D l . 1 - 9 0 and earlier editions of the Code, treatment of box sections had

been made as consistent as possible with that of circular sections. Derivation of
the basic allowable punching shear V-sub-p for box sections included a safety
factor of 1.8, based on a simple yield line limit analysis, but utilizing the
ultimate tensile strength, which was assumed to be 1.5 times the specified
minimum yield. This is why F-sub-y in the design formula for punching shear
was limited to 2 / 3 times the tensile strength. A favorable redistribution of load
was also assumed where appropriate. Localized yielding should be expected to
occur within allowable load levels. Fairly general yielding, with connection
distortion exceeding 0.02 D, can be expected at loads exceeding 120-160% of
the static allowable.

A rational approach to the ultimate strength of stepped box connections can be

taken, using the upper bound theorem of limit analysis (see Figure C10.5.1.4A)
and yield line patterns similar to those shown in Figure C10.5.1.4.B. Various
yield patterns for plastic chord face failure should be assumed in order to find
the minimum computed capacity, which may be equal to or greater than the
true value. Fan corners (as shown for the T-connection) often produce lower
capacities than plain corners as shown for the other cases. Suggested design
factors, given in Table C10.5.1.4A, are consistent with the way we take
advantage of strain hardening, load redistribution, etc. in using tests to failure
as the basis for empirical design criteria. In general, the capacity will be found
to be a function of the dimensionless topology parameters beta, eta, and zeta
(defined in the figure) as well as the chord thickness-squared (corresponding to
tau and gamma in the punching shear format).

...[NOTE: figures and tables need to be re-numbered. For IPB and OPB, correct figure
to show panel which rotates with brace end.]

For IPB and OPB, correct figure to show panel which rotates with brace end.
For very large beta (over 0.85) and K-connections with gap approaching zero,
yield line analysis indicates extremely high and unrealistic connection capacity.
In such cases, other limiting provisions based on material shear failure of the
stiffer regions, and reduced capacity for the more flexible regions (a.k.a.
effective width) must also be observed and checked.

Although the old AWS criteria covered these considerations (ref. 16), for
bending as well as for axial load (ref. 17), more authoritative expressions
representing a much larger data base have been developed over the years by
CIDECT (Comite' International pour le Developpement et l'Etude de la
Construction Tubulaire, ref. 18) and by members of IIW subcommittee XV-E
(ref. 22). These criteria have been adapted for limit state design of steel
structures in Canada (Packer et al ref. 19). The Canadian code is similar to the
AISC-LRFD format. In the 1 9 9 2 edition, these updated criteria were
incorporated into the AWS Code, using the thickness-squared ultimate strength
format and Packer's resistance factors, where applicable.
386

C10.5.2.1 Local failure

Resistance factors vary from equation to equation to reflect the differing

amounts of bias and scatter apparent when these equations are compared to
test data (ref. 19). For example, the equation for plastic chord face failure of T-
Y-, and cross connections is based on yield line analysis, ignoring the reserve
strength which comes from strain hardening; this bias provides the safety
factor with a Phi of unity. The second equation, for gap K- and N-connections
was empirically derived, has less hidden bias on the safe side, and draws a
lower resistance factor.

C10.5.2.2 General Collapse

To avoid a somewhat awkward adaptation of column buckling allowables to the

box section web crippling problem (e.g. ref. 15), AISC-LRFD web yielding,
crippling, and transverse buckling criteria have been adapted to tension, one-
sided, and two-sided load cases, respectively. The resistance factors given are
those of AISC. Packer (ref. 20) indicates a reasonably good correlation with
available box connection test results, mostly of the two-sided variety.

CI0.5.2.3 Uneven Distribution of Load

For box sections, this problem is now treated in terms of effective width
concepts, in which load delivery to more flexible portions of the chord is
ignored. Criteria for branch member checks are given in 10.5.2.3(1), based
empirically on IIW/ CIDECT work. Criteria for load calculation in welds
(section 10.8.5) are based upon the testing of Packer (ref. 21) for gap K- and
N-connections; and upon extrapolation and simplification of the IIW effective
width concepts for -, Y-, and cross connections.

CIO.5.2.4. Materials Considerations

Tubular connections are subject to stress concentrations which may lead to

local yielding and plastic strains. Sharp notches and flaws at the toe of the
welds, and fatigue cracks which initiate under cyclic loading, place additional
demands on the ductility and notch toughness of the steel, particularly under
dynamic loads. These demands are particularly severe in the main member of
tubular -, Y-, and K-connections. Cold formed square and rectangular tubing
(e.g., ASTM A-500 and tubing fabricated from bent plates) is susceptible to
degraded toughness due to strain aging in the corners, when these severely
deformed regions are subjected to even moderate heat of nearby welding.
Suitability of such tubing for the intended service should be evaluated using
tests representing their final condition (i.e., strained and aged, if the tubing is
not normalized after forming). See Section 10.2.6.2 for a discussion of impact
testing requirements.
 387

CI0.5.2.5 Overlapping Connections

By providing direct transfer of load from one branch member to the other in K-
and N-connections, overlapping joints reduce the punching demands on the
main member, permitting the use of thinner chord members in trusses. These
are particularly advantageous in box sections, in that the member end
preparations are not as complex as for circular tubes.

Fully overlapped connections, in which the overlapping brace is welded entirely

to the thru brace, with no chord contact whatsoever, have the advantage of
even simpler end preparations. However, the punching problem that was in
the chord for gap connections, is now transferred to the thru brace, which also
has high beam shear and bending loads in carrying these loads to the chord.

Most of the testing of overlapped connections has been for perfectly balanced
load cases, in which the compressive transverse load of one branch is offset by
the tension load of the other. In such overlapped connections, subjected to
balanced and predominantly axial static loading, tests have shown that it is not
necessary to complete the "hidden" weld at the toe of the thru member. In real
world design situations, however, localized chord shear loading or purlin loads
delivered to the panel points of a truss result in unbalanced loads. In these
unbalanced situations, the most heavily loaded member should be the thru
brace, with its full circumference welded to the chord, and additional checks of
net load on the combined footprint of all braces are required.

C10.5.2.6. Bending

Since international criteria for bending capacity of tubular connections are not
as well developed as for axial loads, the effects of primary bending moments
are approximated as an additional axial load. In the design expression, JD
represents half the moment arm between stress blocks creating the moment,
analogous to concrete design - half, because only half the axial capacity lies on
each side of the neutral axis. Various ultimate limit states are used in deriving
the expressions for JD in Table C 10.5.2.6. For chord face plastification, a
uniform punching shear or line load capacity is assumed. For the material
shear strength limit, the effective width is used. General collapse reflects a side
wall failure mechanism. Finally, a simplified expression for JD is given, which
may conservatively be used for any of the governing failure modes.

Caution should be exercised where deflections due to joint rotations could be

important, e.g. sidesway of portal frames in architectural applications.
Previous editions of the Code provided a 1/3 DECREASE in allowable
connection capacity for this situation.
 388

TABLE C10.5.2.6 - VALUES OF JD

IN-PLANE OUT-OF-PLANE
BENDING BENDING

PLASTIC
CHORD WALL tjD ( + /2) (? + /2)
FAILURE 2( + ) 2( + )
UJ
Q

MO
CHORD
MATERIAL ^ ( + /2) ^ + & - ^ / 2 ) ]

SHEAR

.URE
1
STRENGTH 2 +
11 ( 0 e o p ) 2 ( + / 3 )
<C
LL_ GENERAL T?D + 5tc D
CD COLLAPSE 4 2
2: I
I
11 BRANCH MEMBER (] = /2) [ + i3 E O (L-b
I e o i/2^)]
EFFECTIVE

'ERN
o WIDTH 4 2(1 + ^EOI)
CD
CONSERVATIVE
APPROXIMATION
FOR A N Y M O D E 4 4
 389

C10.5.2.7. O t h e r Configurations

The equivalence of box and circular branch members on box chords is based
on their respective perimeters (0.785 is pi/4). This in effect applies the
concept of punching shear to the problem, even though these international
criteria are always given in ultimate strength format. The results are on the
safe side of available test results.

REST OF PART ... [NO CHANGE] ... C10.7.3

C10.7.4
C10.7.5
C10.7.6
C10.8.5 See commentary at 10.5.2.3.
C10.10.2

REFERENCES . . . A D D :

16. P. W. Marshall, Designing Tubular Connections with AWS D l . l , Welding

Journal, March 1989

17. D. R. Sherman and S. M. Herlache, Beam Connections to Rectangular Tubular

Columns, AISC National Steel Construction Conference, Miami FL, June 1988

18. T. W. Giddings and J. Wardenier, The Strength and Behaviour of Statically

Loaded Welded Connections in Structural Hollow Sections, Section 6, CIDECT
Monograph, British Steel Corp. Tubes Div., 1986

19. J. A. Packer, P. C. Birkemoe, and W. J. Tucker, Canadian Implementation of

CIDECT Monograph 6, CIDECT Rept. 5AJ-84/9E, IIW Doc. SC-XV-84-072,
Univ. of Toronto, July 1984

20. J. A. Packer, Review of American RHS Web Crippling Provisions, ASCE

Journal of Structural Engineering, December 1987

21. J. A. Packer and G. S. Frater, Weldment Design for Hollow Section Joints,
CIDECT Rept. 5AN-87/1-E, IIW Doc. XV-664-87, Univ. of Toronto, April
1987

22. IIW s / c XV-E, Design R e c o m m e n d a t i o n s for Hollow Section J o i n t s -

Predominantly Static Loading, 2nd edition, IIW Doc. XV-701-89, International
Institute of Welding Annual Assembly, Helsinki, Finland, September 1989.
390

ADDITIONAL NOMENCLATURE FOR AWS REVISIONS

web depth (Box Chord) in plane of truss

Feo yield strength of main member

Pu ultimate load

Mu ultimate moment

effective throat

rm mean radius to effective throat of welds

F
EXX classified tensile strength of weld deposit

j denotes thru member at overlap

projected footprint length of overlapping member

q amount of overlap

PSFj o a cj partial safety factor for load (in LRFD)

r
t * overlap^* thru

7t thru member (for overlap conn.)

b e t ( b e ( o vj branch effective width at thru member

^eo ) branch effective width at chord

b e o i ( b e p) branch effective width for outside punching

bgap effective width at gap of K-connections

/3g a p dimensionless effective width at gap of K-connections

0 e Op dimensionless effective width for outside punching

0eff effective for K-connection chord face plastification

applied moment

Mc moment in chord

Pc axial load in chord

 APPENDIX III

N O T C H T O U G H N E S S TASK G R O U P

D R A F T #5 - P R O P O S E D R E V I S I O N F O R AWS Dl.1-90

10.2 BASE METAL

10.2.1 Steel base metal to be used for welded tubular structures shall conform
to t h e r e q u i r e m e n t s of t h e l a t e s t edition of any specification listed
below. Combinations of approved steel base m e t a l s may be welded
together.

10.2.1.1 ASTM A36, Specification for Structural Steel

10.2.1.2 ASTM A53, Grade B, Specification for Pipe, Steel, Black and Hot Dipped
Zinc Coated, Welded and Seamless

10.2.1.3 ASTM A106, Grade B, Specification for Seamless Carbon Steel Pipe for
High-Temperature Service

10.2.1.4 ASTM A131, Specification for Structural Steel for Ships

10.2.1.5 ASTM A139, Grade B, Specification for Electric-Fusion (Arc) Welded

Steel Pipe (Sizes 4 in. and Over)

10.2.1.6 A S T M A 2 4 2 , H i g h - S t r e n g t h Low-Alloy S t r u c t u r a l S t e e l (if t h e

properties are suitable for welding).

10.2.1.7 ASTM A500, Specification for Cold-Formed Welded a n d S e a m l e s s

Carbon Steel Structural Tubing in Rounds and Shapes^

10.2.1.8 ASTM A 5 0 1 , Specification for H o t - F o r m e d Welded a n d S e a m l e s s

Carbon Steel Structural Tubing

3 8 . S p e c i a l i n v e s t i g a t i o n o r h e a t t r e a t m e n t s h a l l b e r e q u i r e d when
n o n - c i r c u l a r s h a p e s made t o t h i s s p e c i f i c a t i o n a r e a p p l i e d t o
t u b u l a r - , Y - , a n d K- c o n n e c t i o n s o r f o r d y n a m i c l o a d i n g , o r
both.
392

10.2.1.9 ASTM A514, Specification for High-Yield S t r e n g t h , Quenched and

3 4 0
Tempered Alloy Steel Plate, Suitable for Welding ^
10.2.1.10 ASTM A516, Specification for Pressure Vessel Plates, Carbon Steel, for
Moderate- and Lower-Temperature Service

10.2.1.11 ASTM A517, Specification for Pressure Vessel Plates, Aloy Steel, High-
39
Strength, Quenched and Tempered

10.2.1.12 ASTM A524, Specification for Seamless Carbon Steel Pipe for Process
Piping

10.2.1.13 ASTM A 5 3 7 , S p e c i f i c a t i o n for P r e s s u r e V e s s e l P l a t e s , C a r b o n -

39
Manganese-Silicon Steel, Heat Treated

10.2.1.14 ASTM A572, Specification for High-Strength Low-Alloy Columbium-

3 9 4
Vanadium Steels of Structural Quality ^

10.2.1.15 ASTM A573, Grade 65, Specifiaction for Structural Carbon Steel Plates
of Improved Toughness

10.2.1.16 ASTM A588, Specification for High-Strength Low-Alloy Structural Steel

39
with 50 ksi (345 Mpa) Minimum Yield Point to 4 in. (100 mm) Thick
40

10.2.1.17 ASTM A595, Specification for Steel Tubes, Low-Carbon, Tapered for
5 9 40
Structural Use

10.2.1.18 ASTM A618, Grades II and III (Grade I if the properties are suitable for
welding), Specification for Hot-Formed Welded a n d Seamless High-
39
Strength Low-Allow Structural Tubing

10.2.1.19 ASTM A633, Specification for Normalized High S t r e n g t h Low-Alloy

39
Structural Steel

10.2.1.20 ASTM A678, Quenched-and-Tempered Carbon-Steel and High-Strength

Low-Alloy Steel Plates for Structural Applications.
3 9 40
10.2.1.21 ASTM A709, Specification for Structural Steel for Bridges

39.Reduced effective yield shall be used as F y Q in the design of

tubular connections See note 2 to Table 10.5.1.

40.Caution: In the absence of a notch toughness requirement this

m a t e r i a l may be u n s u i t a b l e for use as the m a i n m e m b e r in a
tubular connection. See 10.2.6.2.
 393

10.2.1.22 ASTM A710, G r a d e A, Specification for Low Carbon Age-Hardening

Nickel-Copper-Chromium-Molybdenum-Columbium and Nickel-Copper-
9
Columbium Alloy Steels. ^

10.2.1.23 ASTM A 8 0 8 Specification for H i g h S t r e n g t h Low Alloy Carbon,

Manganese, Columbium, Vanadium Steel of S t r u c t u r a l Quality with
Improved Notched Toughness ^
4
10.2.1.24 API Specification 5L, Grade B, Specification for Line P i p e *

10.2.1.25 API^Specification 5L, Grades X42 and X52, Specification for Line Pipe

10.2.1.26 API Specification 2B (when m a d e from p l a t e s t e e l listed h e r e i n ) ,

Specification for Fabricated Structural Steel Pipe.

10.2.1.27 API Specification 2H, Grades 42 and 50, Specifications for Carbon-
Manganese Steel Plate for Off-Shore Platform Tubular Joints.

10.2.1.28 A P I Specification 2W, Specification for S t e e l P l a t e s for Offshore

S t r u c t u r e s , P r o d u c e d by T h e r m o - M e c h a n i c a l Control P r o c e s s i n g
(TMCP).

10.2.1.29 API Specification 2Y, Specification for Steel P l a t e s , Quenched and

Tempered, for Offshore Structures.

10.2.1.30 ABS G r a d e s A, B, D, E, DS, a n d CS, R e q u i r e m e n t s for O r d i n a r y -

42
Strength Hull Structural S t e e l

10.2.1.31 ABS G r a d e s AH32, DH32, EH32, AH36, DH36, and EH36,

Requirements for Higher-Strength Hull Structural Steel.

10.2.2 When an ASTM A709 grade of structural steel is considered for use, its
weldability shall be established by the steel producer, and the procedure
for welding it shall be established by qualification in accordance with
the requirements of 5.2 and other such requirements as prescribed by
t h e Engineer with the following exception: If the grade to be supplied
will meet the chemical and mechanical properties of ASTM A36, A572
Grade 50, or A588, the applicable prequalified procedures of this Code
shall apply.

10.2.3 B a s e M e t a l s N o t A p p r o v e d i n 10.2.1

41. A m e r i c a n P e t r o l e u m I n s t i t u t e (API), 1220 L. Street, N.W.,

Washington, D.C. 20005.

42.American Bureau of Shipping (ABS), 45 Eisenhower Drive, Paramus,

N.J. 07652
394

10.2.3.1 U s e of U n l i s t e d B a s e M e t a l s . When a steel other t h a n one of those

listed in 10.2.1 or 10.2.3.3 is approved u n d e r t h e provisions of t h e
general specification, and such steel proposed for welded construction
u n d e r t h i s C o d e , w e l d i n g p r o c e d u r e s s h a l l be e s t a b l i s h e d by
qualification in accordance with the requirements of 6.2. The fabricator
shall have the responsibility for establishing the welding procedure by
qualification.

10.2.3.2 W e l d a b i l i t y . The E n g i n e e r m a y prescribe a d d i t i o n a l weldability

t e s t i n g of the steel. The responsibility for determining weldability,
i n c l u d i n g t h e a s s u m p t i o n of a d d i t i o n a l t e s t i n g costs involved, is
assigned to the party who either specifies a material not listed in 10.2.1
or who proposes the use of a s u b s t i t u t e material not listed in 10.2.1.
The party proposing the use of a substitute material not listed in 10.2.1
shall assume the additional costs involved in 10.2.3.1.

10.2.3.3 The applicable prequalified procedures of this code shall continue to

apply to ASTM A7 and the following formerly listed steels, insofar as
this is consistent with satisfactory previous usage:

ASTM A381 grade Y-35

ASTM A441
ASTM A529
ASTM A570
ASTM A606 type 2
ASTM A607

10.2.4 B a s e M e t a l for Weld T a b s , B a c k i n g , a n d S p a c e r s

10.2.4.1 Weld T a b s . Weld tabs used in welding shall conform to the following
requirements:

(1) When used in welding with a n approved steel listed in

10.2.1, they may be any of the steels listed in 10.2.1.

(2) When used in welding with a steel qualified in accordance

with 10.2.3, they may be

(a) The steel qualified, or

(b) Any steel listed in 10.2.1.

10.2.4.2 B a c k i n g . Steel for b a c k i n g s h a l l conform to t h e r e q u i r e m e n t s of

10.2.4.1 (1) and (2), except t h a t 100 ksi (690 MPa) m i n i m u m yield
s t r e n g t h steel as backing shall be used only with 100 ksi (690 MPa)
minimum yield strength steels.

10.2.4.3 S p a c e r s . Spacers used shall be of the same material as the base metal.
 395

10.2.5 Base Metal Limitations

10.2.5.1 The provisions of this Code are not intended for use with steels having a
specified minimum yield point or yield strength over 100,000 psi (690
MPa).

10.2.5.2 All groove and fillet weld procedures for weld metal and base metal with
a minimum specified yield strength over 75,000 psi (515 MPa) shall be
qualified to t h e satisfaction of t h e E n g i n e e r prior to use by t e s t s as
provided in 5.2. E D I T O R I A L N O T E : T h i s a l s o r e q u i r e s r e v i s i o n
t o 5.1.1.

10.2.6 Base Metal Notch Toughness

10.2.6.1 Welded t u b u l a r tension members, having a thickness of 2-inches (50

mm) or greater AND a specified minimum yield strength of 40-ksi (280
M P a ) or g r e a t e r , shall be r e q u i r e d to d e m o n s t r a t e C h a r p y V-notch
absorbed energy of 20 ft.-lbs. a t 70 *F (27 J @ 2 0 ' C ) with t e s t i n g in
accordance with ASTM A673 Frequency (heat lot). For the purposes
of this paragraph, a tension member is defined as one having more than
10 ksi (70 MPa) tensile stress due to design loads.

10.2.6.2 Tubulars used as the main member in structural nodes whose design is
g o v e r n e d by cyclic l o a d i n g ( e . g . , t h e j o i n t - c a n i n - Y- a n d K-
connections), having a thickness of 2-inches (50 mm) or greater OR a
yield strength of 50 ksi (345 MPa) or greater in thicknesses of one-inch
(25 mm) or greater, shall be required to demonstrate Charpy V-notch
absorbed energy of 20 ft.-lbs. (27 J ) a t t h e lowest anticipated service
temperature (LAST). Testing shall normally represent the as-furnished
tubulars, and be in accordance with ASTM A673 Frequency (heat lot).
If LAST is not specified, or t h e s t r u c t u r e is not governed by cyclic
e
loading, testing shall be at 4 0 ' F ( 4 C) or lower.

10.2.6.3 Alternative notch toughness requirements shall apply when specified in

contract documents. The Commentary gives additional guidance for
designers. Toughness should be considered in relation to redundancy
vs. criticality of structure at an early stage in planning and design.
396

10.12.4 Weld N o t c h T o u g h n e s s

10.12.4.1 Welding procedures for b u t t j o i n t s (longitudinal or circumferential

seams) within 0.5D of attached branch members, in tubular connection
joint-cans requiring Charpy testing under 10.2.6.2, shall be required to
demonstrate weld metal charpy V-notch absorbed energy of 20 ft.-lbs.

(27 J) a t t h e LAST, or at 0* F (-18 C), whichever is lower. If AWS
specifications for t h e welding m a t e r i a l s to be used do not encompass
this requirement, or if production welding is outside the range covered
by prior testing, e.g., tests per AWS filler metal specifications, then weld
metal Charpy t e s t s shall be made during procedure qualification, as
described in Appendix III. Heat-afTected zone Charpy tests, if any, shall
be reported for information only, except as provided in 10.12.4.2.

10.12.4.2 Alternative notch toughness requirements shall apply when specified in

contract documents. If base m e t a l s with enhanced notch toughness
(beyond the minima of 10.2.6) have been specified, then an engineering
e v a l u a t i o n of weld m e t a l and h e a t affected zone t o u g h n e s s is also
r e q u i r e d . T o u g h n e s s s h o u l d be c o n s i d e r e d in r e l a t i o n to d e s i g n ,
inspection, criticality of service, and practical achievability, at an early
stage in project planning. The commentary gives additional guidance.

E D I T O R I A L N O T E : A d d r e f e r e n c e t o 'Svhen C h a r p y t e s t i n g is r e q u i r e d " i n
c o l u m n 3 of T a b l e 10.12 a s n o t e d i n S a v a n n a h h a n d o u t .
 397

COMMENTARY F O R AWS Dl.1-90

C.10.2.6 BASE METAL T O U G H N E S S

The steels permitted by 10.2.1 are listed by strength group and toughness
class in exhibits 1, 2, and 3. These listings are for guidance to designers,
and follow long-established practice for offshore structures, as described
in Ref. 1 and the following:

S t r e n g t h G r o u p s - Steels may be grouped according to s t r e n g t h level

and welding characteristics as follows (also see 4.1.1 and 4.2):

1. Group I d e s i g n a t e s mild steels w i t h specified m i n i m u m yield

s t r e n g t h s of 40 ksi (280 M P a ) or less. Carbon e q u i v a l e n t is
generally 0.40% or less*, and these steels may be welded by any
of the welding processes as described in the code.

2. Group II designates intermediate strength steels with specified

minimum yield strengths of over 40 ksi through 52 ksi (280 thru
360 MPa). Carbon equivalent r a n g e s u p to 0.45% and higher,
a n d t h e s e s t e e l s r e q u i r e t h e u s e of low h y d r o g e n w e l d i n g
processes.

3. G r o u p I I I d e s i g n a t e s h i g h s t r e n g t h s t e e l s w i t h specified
minimum yield strengths in excess of 52 ksi thru 75 ksi (360 thru
5 1 5 M P a ) . S u c h s t e e l s m a y be u s e d , p r o v i d e d t h a t e a c h
application is investigated with regard to:

(a) Weldability and special welding procedures which may be

r e q u i r e d . Low h y d r o g e n w e l d i n g p r o c e d u r e s w o u l d
generally be presumed.

(b) Fatigue problems which may result from the use of higher
working stresses, and

(c) Notch toughness in relation to other elements of fracture

control, such as fabrication, inspection procedures, service
stress, and temperature environment.

4. Group IV includes ultra-high strength steels in the range of over

75 ksi t h r u 100-ksi yield (515 t h r u 690 MPa). E x t r e m e care
should be exercised w i t h r e g a r d to hydrogen control to avoid
cracking and h e a t i n p u t to avoid loss of s t r e n g t h due to over-
tempering.

Carbon equivalent is defined in Appendix "" 6.1.1

398

EXHIBIT 4

C L A S S I F I C A T I O N MATRIX
FOR APPLICATIONS
 399

T o u g h n e s s C l a s s - Toughness classifications A, B, C may be used to cover various

degrees of criticality shown in the matrix of Exhibit 4, and as described below:

P r i m a r y (or fracture critical) structure covers elements whose sole

failure would be catastrophic.

Secondary structure covers elements whose failure would not lead to

catastrophic collapse, under conditions for which the structure could
be occupied and/or capable of major off-site damages (e.g., pollution).

For highly redundant tubular space-frame structures, fracture of a single brace or

its end connection is not likely to lead to collapse under normal or even moderately
severe loads. The strength is reduced somewhat, however, and the risk of collapse
under extreme overload increases correspondingly.

1. Class C steels are those which have a history of successful application

in welded structures at service temperatures above freezing, but for
which impact t e s t s are not specified. Such steels are applicable to
structural members involving limited thickness, moderate forming,
low restraint, modest stress concentration, quasi-static loading (rise
t i m e 1 second or longer) and s t r u c t u r a l r e d u n d a n c y such t h a t an
i s o l a t e d f r a c t u r e would not be c a t a s t r o p h i c . E x a m p l e s of s u c h
a p p l i c a t i o n s a r e piling, b r a c e s in r e d u n d a n t space frames, floor
beams, and columns.

2. C l a s s s t e e l s a r e s u i t a b l e for u s e w h e r e t h i c k n e s s , cold work,

r e s t r a i n t , s t r e s s c o n c e n t r a t i o n , i m p a c t l o a d i n g , a n d / o r lack of
redundancy indicate the need for improved notch toughness. Where
impact t e s t s are specified, Class steels should exhibit Charpy V-
notch energy of 15 ft.-lbs. (20J) for Group I, 25 ft-lbs. (34J) for Group
II, and 35 ft-lbs. (48J) for group III, at the lowest anticipated service
temperature. Steels listed herein as Class can generally meet these
charpy requirements at temperatures ranging from 50 * to 3 2 ' F (10"
to 0 C). Examples of such applications are connections in secondary
structure, and bracing in primary structure. When impact tests are
specified for Class steel, h e a t lot testing in accordance with ASTM
A673, Frequency H, is normally used. However, there is no positive
a s s u r a n c e t h a t Class toughness will be p r e s e n t in pieces of steel
t h a t are not tested.

3. Class A steels are suitable for use at sub-freezing temperatures and

for critical applications involving adverse combinations of the factors
cited above. Critical applications may w a r r a n t Charpy testing a t

36-54* F (20-30 C) below the lowest anticipated service temperature.
This e x t r a m a r g i n of notch toughness prevents the propogation of
brittle fractures from large flaws, and provides for crack a r r e s t in
thicknesses of several inches. Steels enumerated herein as Class A
c a n g e n e r a l l y m e e t t h e C h a r p y r e q u i r e m e n t s s t a t e d above a t
400

t e m p e r a t u r e s ranging from - 4 F to -40 F (-20 C to -40 * C). Impact
testing frequency for Class A steels should be in accordance with the
specification under which the steel is ordered; in the absence of other
requirements, heat lot testing may be used.

C10.2.6.1 S t e e l for T u b u l a r B r a c i n g ( B e t w e e n C o n n e c t i o n s )

T h e s e m i n i m a l notch t o u g h n e s s r e q u i r e m e n t s for h e a v y - s e c t i o n
tension members follow t h e provisions recently proposed by AISC.
T h e y r e l y to a c o n s i d e r a b l e e x t e n t on t h e t e m p e r a t u r e - s h i f t
phenomenon described by Barsom (Ref. 2), as 70* F is not a realistic
LAST (lowest a n t i c i p a t e d service t e m p e r a t u r e ) for a n y t h i n g b u t
enclosed heated building structures. The temperature-shift effect is
t h a t statically loaded materials exhibit similar levels of ductility as
dynamically loaded impact specimens tested at a higher temperature.
For higher strength steels, Groups III and IV, the temperature-shift
is less effective; also f r a c t u r e m e c h a n i c s s t r a i n e n e r g y r e l e a s e
considerations would suggest higher required energy values. Testing
as-rolled steels on a heat-lot basis leaves one exposed to considerable
variation within the heat; with impacts showing more scatter t h a n
strength properties; however, it is better t h a n no testing at all.

C 10.2.6.2 S t e e l for T u b u l a r C o n n e c t i o n s

The main members in Tubular connections are subject to local stress

concentrations which may lead to local yielding and plastic strains at
the design load. During the service life, cyclic loading may initiate
fatigue cracks, m a k i n g additional d e m a n d s on the ductility of t h e
steel, particularly under dynamic loads. These d e m a n d s are
particularly severe in heavy-wall joint-cans designed for punching
shear.

(a) U n d e r w a t e r C o n n e c t i o n s . For u n d e r w a t e r p o r t i o n s of
redundant template-type offshore platforms, API recommends
t h a t steel for joint cans (such as jacket leg joint cans, chords in
m a j o r X a n d j o i n t s , a n d t h r u m e m b e r s in c o n n e c t i o n s
d e s i g n e d as o v e r l a p p i n g ) m e e t one of t h e following notch
toughness criteria at the temperature given in the Table below.

1. NRL Drop-Weight Test no-break performance, (preferred)

2. Charpy V-notch energy: 15 ft.-lbs. (20 joules) for Group I

steels, 25 ft.-lbs. (34 joules) for Group II steels, a n d 35
ft.-lbs. (48J) for Group III steels (transverse test).
 401

IMPACT TESTING CONDITIONS

Diameter/thickness Test Temperature Test Condition

OVER 30 36 * F (20' C) below LAST* Flat plate

20-30 54 F (30' C) below LAST Flat plate
#
under 20 1 8 F (10 C) below LAST As fabricated

LAST = Lowest Anticipated Service Temperature

The preferred NRL crack arrest criteria follow from use of the
F r a c t u r e Analysis Diagram (ref 3), and from failures of heavy
connections m e e t i n g t e m p e r a t u r e - s h i f t e d Charpy initiation
criteria. For service t e m p e r a t u r e s a t 40* F (4* C) or higher,
these requirements may normally be met by using any of the
Class A steels.

(b) A t m o s p h e r i c S e r v i c e . For connections exposed to lower

temperatures and possible impact, or for critical connections at
a n y l o c a t i o n in w h i c h it is d e s i r e d to p r e v e n t all b r i t t l e
fractures, the tougher Class A steels should be considered, e.g.,
API Spec 2H, Gr. 42 or Gr. 50. For 50 ksi (345 MPa) yield and
h i g h e r s t r e n g t h steels, special a t t e n t i o n should be given to
welding procedures, in order to avoid degradation of t h e h e a t
affected zones. Even for the less demanding service of ordinary
s t r u c t u r e s , t h e following group/class b a s e m e t a l s a r e NOT
r e c o m m e n d e d for u s e a s t h e m a i n m e m b e r s i n t u b u l a r
connections: IIC, IIIB, IIIC, IV.

(c) C r i t i c a l C o n n e c t i o n s . For critical connections involving high

r e s t r a i n t (including a d v e r s e geometry, h i g h yield s t r e n g t h
and/or thick sections), and through-thickness tensile loads in
service, consideration should be given to the use of steel having
improved through-thickness (Z-direction) properties, e.g., API
Spec. 2H. Supplements S4 and S5, or ASTM A770.

(d) B r a c e E n d s . Although the brace ends at tubular connections

are also subject to stress concentration, the conditions of service
are not quite as severe as the main member (or joint-can). For
critical braces, for which brittle fracture would be catastrophic,
consideration should be given to t h e use of stub-ends in t h e
b r a c e s h a v i n g t h e s a m e class as t h e j o i n t - c a n , or one class
lower. This provision need not apply to t h e body of braces
(between connections).
402

REFERENCES

(1) A P I R P 2 A , R e c o m m e n d e d P r a c t i c e for P l a n n i n g , D e s i g n i n g , a n d
C o n s t r u c t i n g Fixed Offshore P l a t f o r m s , 17th ed., A m e r i c a n P e t r o l e u m
Institute,, April, 1987.

(2) S. T. Rolfe and J. M. Barsom, Fracture and Fatigue Control in Structures,

Prentice Hall, 1977.

(3) R. M. C a r t e r , P . W. M a r s h a l l , et. al., M a t e r i a l s P r o b l e m s in Offshore

Structures, Proc. Offshore Tech. Conf., OTC 1043, May 1969.
 EXHIBIT 2
STRUCTURAL STEEL P I P E & TUBULAR SHAPES
404

YIELD STRENGTH TENSILE STRENGTH

GROUP CLASS SPECIFICATION & GRADE k s i MPa KSI MPa
1 c API Spec 5L Grade B* 35 240 60 415
ASTM A53 Grade 35 240 60 415
ASTM A135 Grade 35 240 60 415
ASTM 139 Grade 35 240 60 415
ASTM A500 Grade A (round) 33 230 45 310
(shaped) 39 270 45 310
ASTM A501 (round and shaped) 36 250 58 400
I ASTM A106 Grade (normalized) 35 240 60. 415
ASTM A524 Grade I (thru 3/8" w.t.) 35 240 60 415
Grade II (over 3/8" w.t.) 30 205 55-80 380-550
I A ASTM A333Grade6 35 240 60 415
ASTM A334 Grade 6 35 240 60 415
C API Spec 5L Grade X42 (2% max. cold expansion) 42 290 60 415
API Spec 5L Grade X52 (2% max. cold expansion) 52 360 66 455
ASTM A500 Grade (round) 42 290 58 400
(shaped) 46 320 58 400
ASTMA618 50 345 70 485
API Spec 5L Grade X52 with SR5, SR6 or SR8 52 360 66 455
m C ASTM A595 Grade A (tapered shapes) 55 380 65 450
ASTM A595 Grades and C (tapered shapes) 60 410 70 480
*Seamless or with longitudinal seam welds
See list of Referenced Specifications for full titles of the above.
Note 2: Structural pipe may also be fabricated in accordance with API Spec 2B, ASTM A139+, ASTM A252+, or ASTM A671 using
grades of structural plate listed in Exhibit 1 except that hydrostatic testing may be omitted.
+ with longitudinal welds and circuniferential butt welds.
 EXHIBIT 3
STRUCTURAL STEEL SHAPES

ASTM YIELD STRENGTH TENSILE STRENGTH

GROUP CLASS SPECIFICATION & GRADE KSI MPa KSI MPa

I C A36 (to 2" thick) 36 250 58-80 400-550

A131 Grade A (to 1/2" thick) 34 235 58-80 400-550

I A709 Grade 36T2 36 250 58-80 400-550

C A572 Grade 42 (to 2" thick) 42 290 60 415

A572 Grade 50 (to 1/2" thick*) 50 345 65 450
A588 (to 2M thick) 50 345 70 485

A709 Grades 50T2, 50T3 50 345 65 450

A131 Grade AH32 46 320 68-85 470-585
A131 Grade AH36 51 360 71-90 490-620

To 2" Thick for Type 1 or 2 Killed, Fine Grain Practice

This table is part of the commentary on toughness considerations for tubular structures (or composites of tubulars and other shapes)
e.g., used for offshore platforms. It is not intended to imply that unlisted shapes are unsuitable for other applications.
405
406

C10.12.4 WELD N O T C H T O U G H N E S S

Weld metal and heat-affected zone toughness should be based on the

same engineering considerations as used to establish the base metal
toughness requirements. However, fracture avoidance, by increasing
toughness alone, is not cost effective. Fatigue cracking, hydrogen
induced cold cracking, and solidification hot cracking m u s t also be
dealt with. Other parts of the code address these other problems, via
design, qualification, technique, and inspection requirements. Notch
toughness just helps us live with imperfect solutions.

Weld M e t a l - Notch tough base metals should have compatible weld

m e t a l properties. The following t e s t t e m p e r a t u r e s and m i n i m u m
energy values are recommended, for matching the performance of the
v a r i o u s s t e e l g r a d e s a s l i s t e d i n e x h i b i t s 1-3. W h e n w e l d i n g
procedure qualification by test is required (i.e. when the procedure is
not prequalified, when comparable impact performance has not been
previously demonstrated, or when the welding consumables are to be
employed outside the range of essential variables covered by prior
testing), qualification should include Charpy V-notch testing of the
as-deposited weld metal. Specimens should be removed from the test
weld, a n d i m p a c t tested, in accordance w i t h Appendix III,
R e q u i r e m e n t s for Impact Testing. Single specimen energy values
(one of three) may be 5 ft-lbs (7J) lower without requiring re test.

STEEL STEEL IMPACT TEST WELD METAL AVG.

GROUP CLASS TEMPERTURE FT-LBS (JOULES)

I C 0'F(-18'C) 20 (27)
e
I 0 F(-18'C) 20 (27)
e e
I A -20 F(-29 C) 20 (27)

II c 0-F(-18*C) 20 (27)
II -20-F(-29-C) 20 (27)
II A -40'F(-40'C) 25 (34)

III c -20'F(-29'C) 20 (27)

e
III -40 F(-40'C) 20 (27)
e
III A -40'F(-40 C) 30 (40)

IV SPECIAL INVESTIGATION

Code r e q u i r e m e n t s r e p r e s e n t t h e lowest common denominator from t h e

foregoing table.

Since AWS welding procedure requirements are concerned primarily with

tensile strength and soundness (with minor emphasis on fracture toughness)
it is appropriate to consider additional essential variables which have a n
 407

influence on fracture toughness e.g. specific brand wire/flux combinations,

and the restriction of SAW consumables to the limits actually tested for AWS
classification. Note t h a t , for Class A steels, specified energy levels higher
t h a n the AWS classifications will r e q u i r e t h a t all welding procedures be
qualified by test, rather t h a n having prequalified status.

Charpy impact testing is a method for qualitative assessment of material

t o u g h n e s s . A l t h o u g h lacking t h e f r a c t u r e mechanics basis of crack tip
opening displacement (CTOD) testing, the method has been and continues to
be a reasonable measure of fracture safety, when employed with a definitive
program of nondestructive examination to eliminate weld area defects. The
r e c o m m e n d a t i o n s c o n t a i n e d h e r e i n a r e b a s e d on practices w h i c h h a v e
generally provided satisfactory fracture experience in structures located in
moderate temperature environments (e. g. 40-deg-F sea water and 14-deg-F
air exposure). For e n v i r o n m e n t s which a r e e i t h e r more or less hostile,
i m p a c t t e s t i n g t e m p e r a t u r e s s h o u l d b e r e c o n s i d e r e d , b a s e d on local
temperature exposures.

For critical welded connections, the more technical CTOD test is appropriate.
CTOD tests are run at realistic temperatures and strain rates, representing
t h o s e of t h e e n g i n e e r i n g a p p l i c a t i o n , u s i n g s p e c i m e n s h a v i n g t h e full
p r o t o t y p e t h i c k n e s s . T h i s yields q u a n t i t a t i v e i n f o r m a t i o n useful for
engineering fracture mechanics analysis and defect assessment, in which the
required CTOD is related to anticipated stress levels (including residual
stress) and flaw sizes.

R e p r e s e n t a t i v e CTOD r e q u i r e m e n t s r a n g e from 0.004-inch a t 40* F

#
(0.10-mm at 4*C) to 0.015-inch at 1 4 F (0.38-mm at -10-deg-C). Achieving
the higher levels of toughness may require some difficult trade-offs against
other desirable a t t r i b u t e s of the welding process for example, t h e deep
penetration and relative freedom from trapped slag of uphill passes, versus
the lower heat input and highly refined weld layers of downhill passes.

H e a t a f f e c t e d z o n e - In addition to weld metal toughness, consideration

should be given to controlling the properties of the h e a t affected zone (HAZ).
Although the h e a t cycle of welding sometimes improves as-rolled base metals
of low toughness, this region will often have degraded toughness properties.
T h e HAZ is often t h e site of h y d r o g e n induced u n d e r b e a d cracking. A
n u m b e r of early failures in welded tubular joints involved fractures which
either initiated in or propagated through t h e HAZ, often before significant
fatigue loading.
408

Appendix III gives requirements for sampling both weld metal and HAZ ,
with Charpy energy and temperature to be specified in contract documents.
T h e following average HAZ values h a v e been found by experience to be
reasonably attainable, where single specimen energy values (one of three)
5ft-lbs (7J) lower are allowed without requiring retest:

STEEL STEEL IMPACT HEAT AFFECTED ZONE

GROUP CLASS TEMPERATURE FT-LBS (JOULES)

I c 50-F(10'C) for information only

I 4 0 - F (4-C) 15 (20)
e
I A 14 F(-10'C) 15 (20)
e
II C 5 0 F (10'C) for information only
II 4 0 - F (4'C) 15 (20)
e
II A 14'F(-10 C) 25 (34)
e
III A 14 F(-10'C) 30 (40)

As criticality of t h e component's performance i n c r e a s e s , lower t e s t i n g

temperatures (implying more restrictive welding procedures) would provide
HAZ's which more closely match the performance of the adjoining weld metal
and parent material, r a t h e r t h a n being a potential weak link in the system.
The owner may also wish to consider more extensive sampling of the HAZ
t h a n the single set of Charpy tests required by Appendix III, e.g. sampling at
0.4-mm, 2-mm, a n d 5-mm from t h e fusion line. (These dimensions may
change with h e a t input.) More extensive sampling increases the likelihood
of finding local brittle zones with low toughness values.

Since HAZ toughness is as much dependent on the steel as on the welding

p a r a m e t e r s , a preferable alternative for addressing this issue is through
w e l d a b i l i t y p r e q u a l i f i c a t i o n of t h e steel. API RP 2Z spells o u t such a
prequalification procedure, using CTOD as well as Charpy testing. This
prequalification t e s t i n g is p r e s e n t l y being applied as a s u p p l e m e n t a r y
requirement for high-performance steels such as API Specs 2W and 2Y, and
is accepted as a requirement by some producers.

Caution: Section 5 of this code permits testing one 50-ksi steel to qualify all
other grades of 50-ksi and below. Consequently, selection of API-2H-50-Z
(very low sulfur, 200-fb-lb upper shelf Charpies) for qualification test plates
will virtually assure satisfying a HAZ impact requirement of 25-ft-lbs, even
when welded with high heat inputs and high interpass temperatures. There
is no reasonable way to extrapolate this test to ordinary A572 grade 50 with
t h e expectation of reproducing either the HAZ impact energies or t h e 8:1
degradation of the test on API-2H-50-Z. Thus, separate Charpy testing of
different steel grades, thickness ranges, and processing routes should be
considered, if HAZ toughness is being addressed via WPQ testing.
 409

Local Brittle Zones (LBZ) - Within the weld heat affected zones there may
exist locally embrittled regions. Under certain conditions, those LBZ's may
be detrimental. The engineer should consider the risk of LBZ's and
determine if counter measures should be employed to limit the extent of
LBZ's and their influence on structural performance. Some counter
measures and mitigating circumstances in offshore practice are listed below:

(1) the use of steels with moderate crack-arrest capabilities, as

demonstrated by no-break in the NRL drop-weight test (small flaw)

(2) overmatch and strain hardening in conventional normalized 42 to

50-ksi carbon-manganese steels in which the weld metal and HAZ
have higher yield strength than adjacent base metal, forcing plastic
strains to go elsewhere

(3) the tendency for fatigue cracks in welded tubular joints to grow out of
the HAZ before they reach appreciable size (assuming one avoids
unfavorable tangency of joint can weld seam with the brace footprint)

(4) prequalified limits on weld layer thickness in welding procedures,

which along with observing limits on heat input, promote grain
refinement in the HAZ and minimize the extent LBZ.

(5) Composition changes, e.g. reduced limits on vanadium and nitrogen,

and increased titanium.

Reference:

American Petroleum Institute, Recommended Practice for Pre-Production

Qualification of Steel Plates for Offshore Structures, API RP2Z, 1st edition, 1987.
 INDEX

API 1, 5,12,102,116,128,164, 303, 323 damaged strength rating 308

AISC 6,9,12 detailing 104,343
AISI 8 definitions 18
ASCE 102,125 design 3,5
amplification factor 15 charts 134f,201f
allowable stress design 5, 212, 362, 365, 377 strategy 141
alpha 31,73,162, 237f diaphragm 216,267
analysis dihedral 18,343
levels 149,156, 224 dimensions 19
methods 29f, 74f, 166, 268
architecture 1,5 eccentricity - see overlap
effective
backup weld 343 gamma 249
beam-column 13f length 8,9,82,107
beam shear 16,24,199 nominal stress 247
bending (moment) 232, 377, 387 thickness 166,250
beam 9,24, 267 throat 64,369
in-plane 45,114,139,164 width 26, 82,199
out-of-plane 45,114,139,164 efficiency
plate 47f connection 79, 136,175,205, 270
shell 29f, 74f punching shear 76,136
bent 348 electrode 103,345
beta 19,73,107,187f erection 350
paradox 92 errors 95,106, 220, 243
bevel 338f eta 19
Bouwkamp 71, 85,143, 181f, 270f
box members 7,10, 382 fabrication 348
box connections 47f, 186f, 343, 371, 385 failure
brace 18 definition of 57,87,188
branch member 18 field 21,89f
buckling modes 21f, 187f
column 8 probability 110,117
local 11,26,382 fatigue
building block 40f, 195, 231 design 147f, 284
Bullwinkle 4 improvement 293
strength 166,178f, 219f, 254, 277, 281, 330
characteristic strength 207 test procedure 57
Charpy 298, 303f, 395 fillet weld 220,339
chord 18 finite element 32f, 160,167, 244, 248, 268
CIDECT 206f, 209f fitness-for-purpose
column 7f engineering 324
complete joint penetration 338 experience-based 323
connection 18 flaw size 316f, 322, 325
corner 18,343f flexibility 191,195,206
crack growth 277f, 326 footprint 63,266
crack initiation 171, 275, 298 fracture 26
crack tip opening displacement 306, 325 control 294f
cross connection 85,107,112,137 critical 304,399
cumulative damage 149 mechanics 277f, 321
cutting sections 39f transition 298f
 411

gamma 19,31,73,94, 111, 187,192,197,202 modifier

gap 19, 113, 122,208,240 chord stress 108,123f, 140,194
transverse 24 If geometry and load pattern 49f, 107,11 If, 126f, 188f
Gauss point 36 moment - see bending
general collapse 23f, 96,118, 204, 367 multi-planar connections 236f
Gibstein 118,155, 156
Graff 11,71,102,227,353 N-connection 20,221
grouted connections 246f node 18,349
gussets 40f, 259 nominal stress 66,174, 219
Gulf of Mexico 73,296 notch
severity parameter 276
heat affected zone 305f, 408 stress 275f, 320
heel 18 toughness 294f, 391f
hot spot 59f, 150f,223 non-dimensional 19,73
attributes 153f NRL 301
definition 20,160
geometric 156 offshore If
local or microscopic 156f ovalizing 24,237
situational 157,292 overlap 215, 230f, 370

IIW 130, 206f, 209f, 223f, 236 partial penetration 339

interaction 15, 97,115,124,133, 244 partial safety factor - see LRFD or
inspection 312,351 plastic
isoparametric 35 interaction 115
moment 9
joint 18 reserve 99
joint can 16,18 profile 172, 177, 274f, 341
proof load 57
K-connection 19, 37, 52,104,112,122,137, projected length 233
K-factor 64,115 propagation 299,301f
K-free 66,133 punching shear 20, 22, 53f, 88, 93f, 129,173,
Kellogg formula 92,162, 239,247,250 187,194, 221,238,263
Kuang-Smedley 163
knife edge 44 qualification 344
Kurobane 70, 87,118,132f, 143f, 151,
180,181f,236, 272 radius 19
rectangular hollow section 7, 10, 186f
lamellar tearing 27, 306 reduction factor
layout 335 brace-end SCF 165
line load 22, 30, 75 chord stress 108,123f, 140,194
load and resistance factor design 126, 212, effective area 218
308, 362, 367, 378 fatigue strength 157
local failure 21f, 93, 383 net section stress 218
local brittle zone 409 redundancy 308
lowest anticipated service temperature 297,401 reserve strength 99,192, 309
resistance factor 126, 207, 362, 367f, 378, 383f
main member 18 ring 80f,261,267
Makino 26,241 root 319,343
manufacture 333
membrane S-N curve 60,167f
shear 31 SAErule 224,284f
stress 193,267 safety index 3, 110,117, 121,132,195
traction 31 shakedown 56
model tests 54f, 83 shear - see punching, beam, etc.
Shell 71, 72,101,143f, 160, 228, 272f
412

shell 29f, 74f zeta 19

Sherman 14,17,102,189,205, 227 Zettlemoyer 120,122
sidewall 23, 200,208, 218 zoom 38
simple joints 16
size effect 58,178, 225, 274f
static strength 79, 83f, 186f, 230f, 239f, 255, 263f
steel 26,201,330f,391
strain gages 57
strain concentration factor 60,155,223
strain hardening 9,99
stress
hot spot 20,59f
in branch member 68,166, 222
in main member 123
in welds 66f, 361f, 378f
nominal 66
residual 7
stress concentration factor 57,160f, 226, 250, 268
stress intensity factor 278,325
stresscoat 57

, Y, or K-connection 20, 315

tau 19, 73, 203
tee connection 84, 86,137
tee joint 337
tension 113,132
theta 19
thermal-mechanical controlled process 332
toe 18
Toprac 70, 85, 102,143f, 151,182f
torsion 16,65

ultimate strength 15, 83f, 94,121,187,239, 325

ultrasonic 313
unzipping 25, 200, 368, 384
utilization factor 123,139

vortex shedding 9

Wardenier 8,16f, 52,69f, 144f, 184, 219f, 228f

web crippling 24,196
Weibull distribution 148
weld size 340
weld toughness 304, 396,406
weldability 331,394
welder 344
welding procedures 345
workmanship 314

X-joint - see cross connection

yield line 47f, 187f, 197

yield strength 99,117, 330, 397
Yura 120f, 145