Professional Documents
Culture Documents
Norrnar kjarnryggisrannsknir
Pohjoismainen ydinturvallisuustutkimus
Nordisk kjernesikkerhetsforskning
Nordisk krnskerhetsforskning
Nordic nuclear safety research
NKS-126
ISBN 87-7893-188-6
April 2006
Abstract
The simulation and analysis of data for enhancing low cycle fatigue test procedures is discussed in this report. The analysed materials are an austenitic
stainless piping steel and an austenitic weld material. This project continues the
work performed in 2003 and 2004. The fatigue test data treatment application
developed within the project in 2004 for the preparation of the fatigue data has
been developed further. Also, more fatigue test data has been analysed with the
application than in 2004. In addition to this numerical fatigue simulations were
performed with FEM code ABAQUS. With the fatigue test data treatment application one can e.g. both calculate cyclically certain relevant characteristic values,
e.g. elastic range, and form a set of certain cyclical parameter values needed as
a part of ABAQUS analysis input files. The hardening properties of metals were
modelled with both isotropic and kinematic hardening models.
The further development of the application included trimming of the analysed
data, and consequently trimming of resulting hardening parameters. The need for
the trimming arose from the fact that the analysed fatigue test data presents
some scatter caused by the limited accuracy of the test equipment and the sampling rate. The hardening parameters obtained from the application analysis results were used in the subsequent ABAQUS analyses, and then the fatigue test
data were compared with the ABAQUS simulation results. After finding a procedure to trim result data to get smooth curves for cyclic hardening, hardening and
softening could be reproduced in ABAQUS analysis with a reasonable accuracy.
The modelling of the fatigue induced initiation and growth of cracks was not
considered in this study. On the other hand, a considerable part of the fatigue life
of nuclear power plant (NPP) piping components is spent in the phase preceding
the initiation and growth of cracks.
Key words
low cycle fatigue, steel, stainless, austenitic, material model, finite element model
NKS-126
ISBN 87-7893-188-6
Electronic report, April 2006
The report can be obtained from
NKS Secretariat
NKS-775
P.O. Box 49
DK - 4000 Roskilde, Denmark
Phone +45 4677 4045
Fax
+45 4677 4046
www.nks.org
e-mail nks@nks.org
RESEARCH REPORT
TUO72-056604
4.4.2006
Confidentiality:
Public
Reports title
Simulation and Analysis of Data for Enhancing Low Cycle Fatigue Test Procedures
Customer
Order reference
VYR
Project name
INPUT VSYMA
G5SU01043
Author(s)
Pages
32 /
Key words
TUO72-056604
Summary
The simulation and analysis of data for enhancing low cycle fatigue test procedures is discussed in this report.
The analysed materials are an austenitic stainless piping steel and an austenitic weld material. This project
continues the work performed in 2003 and 2004. The fatigue test data treatment application developed within the
project in 2004 for the preparation of the fatigue data has been developed further. Also, more fatigue test data has
been analysed with the application than in 2004. In addition to this numerical fatigue simulations were performed
with FEM code ABAQUS. With the fatigue test data treatment application one can e.g. both calculate cyclically
certain relevant characteristic values, e.g. elastic range, and form a set of certain cyclical parameter values needed
as a part of ABAQUS analysis input files. The hardening properties of metals were modelled with both isotropic
and kinematic hardening models.
The further development of the application included trimming of the analysed data, and consequently trimming
of resulting hardening parameters. The need for the trimming arose from the fact that the analysed fatigue test
data presents some scatter caused by the limited accuracy of the test equipment and the sampling rate. The
hardening parameters obtained from the application analysis results were used in the subsequent ABAQUS
analyses, and then the fatigue test data were compared with the ABAQUS simulation results. After finding a
procedure to trim result data to get smooth curves for cyclic hardening, hardening and softening could be
reproduced in ABAQUS analysis with a reasonable accuracy.
The modelling of the fatigue induced initiation and growth of cracks was not considered in this study. On the
other hand, a considerable part of the fatigue life of nuclear power plant (NPP) piping components is spent in the
phase preceding the initiation and growth of cracks.
Confidentiality:
Public
Espoo 04.04.2006
Signatures
Pentti Kauppinen
Deputy Technology Manager
VTTs contact address
P.O. Box 1000, FI-02044 VTT
Urpo Sarajrvi
Research Scientist
Arja Saarenheimo
Checked
STUK: Registry (4), Keskinen R. (1), Hytnen Y. (1); FNS: Neuvonen A. (1);
TVO: Pulkkinen E. (1 kpl); NKS: Kierkegaard J. (1); VTT: Archive (1)
The use of the name of the Technical Research Centre of Finland (VTT) in advertising or publication in part of
this report is only permissible by written authorisation from the Technical Research Centre of Finland.
Preface
This report has been prepared under the research project INPUT VSYMA. The project is a
part of SAFIR, which is a national nuclear energy research program. In the structure of
SAFIR, this research project is a subproject of project INPUT, which is a part of a larger
project system INTELI. INPUT stands for Reactor circuit piping, and INTELI stands for
Integrity and lifetime of reactor circuits. The work was carried out at VTT Industrial Systems.
VSYMA project was funded by the State Nuclear Waste Management Fund (VYR), Nordic
nuclear safety research (NKS; NKS-R Framework) and the Technical Research Centre of
Finland (VTT).
Espoo
Authors
Contents
1 Introduction
2 Goal
11
11
12
13
15
15
15
17
22
23
24
26
27
29
References
32
Introduction
The simulation and analysis of data for enhancing low cycle fatigue test procedures is
discussed in this report. The analysed materials are an austenitic stainless piping steel and an
austenitic weld material. This project continues the work performed in 2003 and 2004, the
results of which are reported in references /1,2/. The fatigue test data treatment application
developed within the project in 2004 for the preparation of the fatigue data has been developed
further. Also, more fatigue test data has been analysed with the application than in 2004. In
addition to this numerical fatigue simulations were performed with FEM code ABAQUS. With
the fatigue test data treatment application one can e.g. both calculate cyclically certain relevant
characteristic values, e.g. elastic range, and form a set of certain cyclical parameter values
needed as a part of ABAQUS analysis input files.
In the numerical analyses performed in this study with ABAQUS the type of considered fatigue
data was low-cycle, and consequently only strain rate independent models were applied. The
applied yield function was that of von Mises. The hardening properties of metals were modelled
with both isotropic and kinematic hardening models.
The main emphasis in this study, however, was on the analyses performed with the fatigue test
data treatment application, and further development of it. In 2004 only part of the fatigue data
available to the project were analysed with the application. To obtain more accurate and
representative estimates for kinematic hardening and cyclic hardening parameters needed in the
ABAQUS analyses, considerably more constant amplitude fatigue test data were analysed with
the application in 2005. The further development of the application included trimming of the
analysed data, and consequently trimming of resulting hardening parameters. The need for the
trimming arose from the fact that the analysed fatigue test data presents some scatter, which is
also typical for any fatigue test data. This scatter is caused by the limited accuracy of the test
equipment and by the duration of the time step between the measurement time instants (i.e. the
cyclic stress-strain curves are not continuous, but are composed of a finite number of
measurement points). The hardening parameters obtained from the application analysis results
were used in the subsequent ABAQUS analyses, and then the fatigue test data were compared
with the ABAQUS simulation results.
The modelling of the fatigue induced initiation and growth of cracks was not considered in this
study. On the other hand, a considerable part of the fatigue life of nuclear power plant (NPP)
piping components is spent in the phase preceding the initiation and growth of cracks.
The structure of this report is the following. After this introduction the report continues with
presenting the goal of this study. A brief description of the characteristics of hardening and
softening is presented then. The processing of the fatigue test data and procedures to estimate
kinematic and cyclic hardening parameters are presented after that. Then the report moves on to
describe the applied approaches in trimming the fatigue test data and the hardening parameters.
A comparison of the ABAQUS simulation results to the fatigue test data is presented then. The
report ends with conclusions and suggestions for future plans.
Goal
The main challenge in the project is to analyse data acquired in fatigue low cycle tests in a
way to determine parameters for elastic-plastic analysis in ABAQUS FE-software. Fig. 2.1
illustrates fatigue test data as a stress-strain curve. Figs. 2.2 and 2.3 show illustrations of
parameters required in FE-analysis to describe plastic behaviour of the material. Finally, Fig.
2.4 shows an example of results, i.e. the difference between test data and calculated data.
Another challenge is to develop a procedure to determine reliable parameters with an
optimum amount of work. The task includes finding ways e.g. to optimise data acquisition
during fatigue tests, to minimise processing of test data and to minimise validation analysis.
450
400
350
300
250
200
150
Stress [MPa]
100
50
0
-50
-100
-150
-200
-250
-300
-350
-400
-450
-0.80%
-0.60%
-0.40%
-0.20%
0.00%
0.20%
Strain [mm/mm]
Figure 2.1
0.40%
0.60%
0.80%
425
400
Stress [Mpa]
375
350
325
300
275
250
0.000
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.010
Strain [m m /m m ]
Figure 2.2
325
300
275
250
225
200
175
150
125
100
0.010
0.100
1.000
10.000
100.000
Equivalent strain [m m /m m ]
Figure 2.3
Stress [MPa]
150
100
50
0
-50
-100
-150
-200
-250
-300
-350
-400
-450
-0.80%
-0.60%
-0.40%
-0.20%
0.00%
0.20%
0.40%
0.60%
Strain [mm/mm]
Test data
Figure 2.4
ABAQUS data
0.80%
In commonly applied and generally accepted plasticity theory the basic concept with which
material hardening and softening can be approached is the yield surface. More precisely, the
yield surface is defined as the boundary of elastic range for rate-independent plastic material,
in either stress or strain space. The shape of the yield surface depends on the entire history of
deformation from the reference state. During plastic deformation the states of stress or strain
remain on the subsequent yield surfaces. The yield surfaces for actual materials are
experimentally found to be mainly smooth, although they may develop pyramidal or conical
vertices, or regions of high curvature. If elasticity within the yield surface is linear and
unaffected by plastic flow, the yield surfaces for metals are convex in the Cauchy stress space
/3/.
In addition to stress distribution the yield surface is dependent of temperature and so called
internal variables /4/.
The yield surface can change its size and shape in the stress space. When the yield surface
expands it is said that material hardens, and when it contracts it is said that material softens.
These two phenomena can be illustrated by looking at a case of uniaxial stress in two
specimens of metal alloy analysed in this study, the stress amplitude curves of which are
shown as a function of experienced load cycles in Fig. 3.1. Besides the steeply descending
short end parts, during ascending curve parts the specimens harden and during descending
parts they soften, i.e. during the former parts the yield limit rises, and during the latter parts it
lowers. As mentioned above, both of these curves have abruptly descending, almost vertical
end parts, during which macroscopic cracks first initiate, and then grow and coalesce, which
finally leads to rupture of the specimens.
675
650
625
600
575
550
525
500
475
450
425
1
10
100
1000
Cycle [-]
Figure 3.1. Samples of hardening and softening. Besides the steeply descending short end
parts, during ascending curve parts the specimens harden, and during descending parts they
soften.
Another and more detailed example of cyclic metal hardening and softening is presented in
Fig. 3.2. In the enlarged detail figures of the various stages of hardening and softening the
limits of accuracy of the test equipment start to show too, as most of the points forming the
curve in question deviate slightly from the smooth curve path.
425
400
400
380
360
340
375
410
410
405
405
320
300
350
280
800 900 1000 1100 1200 1300 1400 1500
325
400
400
395
395
300
390
390
0
10
10 20 30 40 50
60 70 80 90 100
275
1
Figure 3.2
10
100
Cycle [-]
1000
10000
The dependence of the yield function on the internal variables describes usually the
hardening/softening properties of the material /4/. In the simplest plasticity model ("perfect
plasticity") the yield surface acts as a limit surface and there are no hardening parameters at
all: no part of the model evolves during the deformation. However, complex plasticity models
usually include a large number of hardening parameters.
Stress states that cause the yield function to have a positive value cannot occur in rateindependent plasticity models, although this is possible in a rate-dependent model.
Drucker defines a work-hardening, or stable, plastic material as one in which the work done
during incremental loading is positive, and the work done in the loading-unloading cycle is
nonnegative; this definition is generally known in the literature as Druckers postulate /5/.
The hardening described by the expansion of the yield surface in the stress or strain space is
called isotropic, while that described by the translation is called kinematic. Many materials
can show mixed behaviour, i.e. to a varying extent combined isotropic and kinematic work
hardening/softening /4/.
In the following Figs. 3.3 to 3.5 are for the metal alloy analysed in this study examples of:
cyclic hardening during the first load cycles, gradual softening during the consequent load
cycles, and final hardening and breaking down.
425
400
410
400
375
405
375
350
400
325
350
395
325
390
300
10
300
0.0076
0.0077
1st
2nd
0.0078
5th
0.0079
0.0080
Strain [m m /m m ]
10th
275
1
10
100
1000
Cycle [-]
Figure 3.3.
425
400
410
400
375
405
380
350
360
400
340
395
325
320
390
10
300
20
30
40
50
60
70
80
90
100
300
0.0076
0.0077
10th
30th
0.0078
50th
0.0079
100th
0.0080
Strain [m m /m m ]
275
1
10
100
Cycle [-]
Figure 3.4.
1000
425
400
400
360
-180
320
-220
280
-260
240
-300
200
375
350
325
-340
160
0.0055
-380
300
100
1000
0.0065
0.0070
1300
0.0075
1400
0.0080
1480
-420
-0.0080
100
-0.0075
1000
-0.0070
-0.0065
1300
-0.0060
1400
-0.0055
1480
275
100
Figure 3.5
0.0060
1000
Cycle [-]
10000
4
4.1
In order to describe the behaviour of the material for ABAQUS analysis in forms of
parameters illustrated in Fig. 2.2 and 2.3, the test data, i.e. stress-strain curves, have to be
processed. A code to process the data all through the test duration was written as reported in
reference /2/. Below (Fig. 4.1) there is a short description of the procedure. For each cycle the
yield stress in compression is determined as the point where stress curve crosses a straight
line corresponding the selected plastic strain (strain offset parallel to slope of the curve start).
The elastic range is the difference between the peak stress and the yield stress. The yield
stress in tension is the valley stress with the addition of the elastic range.
Strain offset e.g. 0.2-limit
500
450
400
350
300
250
Stress [MPa]
200
150
100
Parallel lines
50
0
-50
-100
-150
-200
-250
-300
0.010
0.012
0.014
0.016
0.018
0.020
Strain [m m /m m ]
Test cycle #10
Offset stress
Compr. yield
300
250
200
150
100
50
Stress [MPa]
0
-50
-100
-150
-200
-250
-300
-350
-400
-450
-500
-0.020
-0.018
-0.016
-0.014
-0.012
-0.010
Strain [m m /m m ]
Test cycle #10
Figure 4.1
Offset stress
Tens. yield
4.2
Kinematic hardening
One of the parameters for ABAQUS analysis is the kinematic hardening. It describes the form
of the stress-strain curve in the plastic zone. The tenth cycle is assumed to represent a
stabilized cycle and is selected to describe the kinematic hardening. In reference /2/ there are
a couple of results reported, but here all of the test data is considered. Fig. 4.2 shows the
kinematic hardening of all the test data for austenitic base material. Tests were done with five
different strain amplitudes, two of which having two or three tests. Lower strain amplitude
tends to produce lower hardening, though the start point, i.e. yield stress, appears to vary so
that the curves cross each other. The phenomenon is discussed later in Chap. 5.2 and Chap. 6,
where test data and results from ABAQUS analyses are compared.
550
500
450
400
400
350
Stress [MPa]
350
300
300
250
250
200
200
150
150
100
100
0.000
50
0
0.000
0.3%
Figure 4.2
0.005
0.4%
0.010
0.6%
0.015
0.020
Strain [m m /m m ]
0.8%
0.8%
1.2%
0.001
0.025
2.0%
0.002
0.030
2.0%
0.003
0.035
2.0%
4.3
Cyclic hardening
Another parameter for ABAQUS analysis is the cyclic hardening. It defines the progress of
the elastic range when loading is repeated. The cyclic hardening is described as the elastic
range in stress units (MPa in SI units) per equivalent plastic strain in strain units (mm/mm). In
reference /2/ a couple of cases are introduced. In this project all test data is gathered together.
Fig. 4.3 illustrates elastic ranges per number of loading cycles and Fig. 4.4 the final cyclic
hardening of an austenitic base material. As previously in Chap. 4.2, tests were done with five
different strain amplitudes, two of which have two or three tests. The material has a hardening
tendency, especially with high strain amplitudes. Fig. 4.5 shows the cyclic hardening of an
austenitic weld metal, which tends to soften.
475
450
425
400
375
Stress [MPa]
350
325
300
275
250
225
200
175
150
125
100
1
0.3%
Figure 4.3
10
0.4%
100
0.6%
0.8%
1000
Cycle #
0.8%
1.2%
10000
2.0%
100000
2.0%
2.0%
The diagrams of cyclic hardening for both materials seem to have more or less of scattering of
data. In some cases the elastic range seems to vary smoothly, whereas other diagrams are
strongly fluctuating. The next chapter deals with the need of trimming data and procedures
how to trim.
475
450
425
400
375
Stress [MPa]
350
325
300
275
250
225
200
175
150
125
100
0.001
0.3%
Figure 4.4
0.010
0.4%
0.100
1.000
10.000
Equivalent plastic strain [m m /m m ]
0.6%
0.8%
0.8%
1.2%
100.000
2.0%
2.0%
1000.000
2.0%
525
500
475
450
425
Stress [MPa]
400
375
350
325
300
275
250
225
200
0.001
0.010
0.4%
Figure 4.5
0.100
1.000
Equivalent plastic strain [m m /m m ]
0.6%
0.8%
0.8%
1.2%
10.000
2.0%
100.000
2.0%
5.1
The developed code is able to process all cyclic data of a fatigue test with symmetrical strain
amplitude. The output is, however, either smooth or fluctuating, as illustrated in Figs. 4.4 and
4.5. As for the stress amplitudes, the diagrams are smooth, which means that hardening and
softening do not progress in a fluctuating manner. Therefore, there is a need to take a closer
look at the data to reach smooth diagrams for the cyclic hardening. Two approaches have
been taken into consideration, the first of which requires pre-processing of results but the
latter one is written in the code.
5.2
There are two kinds of diagrams of cyclic hardening: smooth and fluctuating. Fig. 5.1 (left)
illustrates a sample of smooth diagrams for elastic range. There are also the stress amplitude
and ratios of the stress amplitude and elastic range. Fig. 5.1 (right) is a similar chart of a
sample with a strongly fluctuating elastic range. In Fig. 5.1 (left) the ratio remains rather
stable, whereas in Fig. 5.1 (right) there is a large deviation. The approach to trim the diagram
of the elastic range is to define an average value for the ratio and to define a modified diagram
for the elastic range. The approach includes an attempt to simplify the trimming. There are
statistical tools to process all data, but the attempt was to limit the processing to the first ten
cycles. The basis for the limitation is the idea that only the first ten cycles would be studied
thoroughly and the rest could be processed using the average ratio and stress amplitudes,
which data is easily obtained from fatigue tests.
450
3.5
450
3.5
400
3.0
400
3.0
350
2.5
350
2.5
300
2.0
300
2.0
250
1.5
250
1.5
200
1.0
200
1.0
150
0.5
150
0.5
0.0
100
100
1
Figure 5.1
10
Elastic range
100
1000
Stress A mplitude
10000
Ratio
0.0
1
10
Elastic range
100
1000
Stress A mplitude
10000
Ratio
Ratio (Stress amplitude/Elastic range) with a small deviation (left) and a large
deviation (right).
Fig. 5.2 shows ratios with average values of the first and tenth value. In Fig. 5.3 there is an
example of a modified elastic range. The form of the diagram is the same as the stress
amplitude. In Chap. 6.1 there is a sample ABAQUS analysis that shows the accuracy of this
trimming procedure.
2.00
1.80
1.60
1.40
1.20
1.00
0.001
0.010
0.6%
2.0%
0.8%V2 avg 1/10
Figure 5.2
0.100
1.000
0.8%
0.6% avg 1/10
1.2% avg 1/10
10.000
100.000
Strain [m m /m m ]
1.2%
0.8% avg 1/10
0.8%
0.8% avg 1/10
2.0% avg 1/10
425
400
375
Stress [MPa]
350
325
300
275
250
225
200
1
10
100
Stress amplitude
Figure 5.3
Elastic range
1000
Modified
Cycle [-]
10000
In Fig. 5.4 there are modified diagrams for cyclic hardening of an austenitic base material
with various strain amplitudes. The two diagrams for strain amplitude 0.8% are close to each
other whereas in Figs. 4.4 and 5.1 there is a gap of ca. 100 MPa between them. They are
supposed to lie close because their stress amplitudes are nearly the same and the kinematic
hardening is supposed to be similar. Therefore the cyclic hardening should not differ. In fact,
these two tests were cases that lead to consider trimming procedures.
Cyclic hardening modified, Base
material
475
700
450
650
425
600
550
375
Stress [MPa]
Stress [MPa]
400
350
325
300
275
500
450
400
350
250
300
225
200
0.001
0.010
0.100
1.000
10.000 100.000
250
0.001
0.010
0.100
1.000
Strain [m m /m m ]
Strain [m m /m m ]
Figure 5.4
5.3
0.6%
0.8%
1.2%
2.0%
0.8% V2
10.000 100.000
0.6%
0.8%
1.2%
2.0%
0.8% V2
When taking a closer look at the cycles of the two tests with the same strain amplitude, one
can see the difference between the resolutions of the data. Fig. 5.5 is a magnification of the
peak and valley of the tenth cycle of a test where the strain rate was low. Scattering is absent,
peak and valley points are accurate and slopes can be defined easily. In Fig. 5.6 data
scattering is prominent at peaks and valleys due to effects caused by a higher strain rate. Note,
that a different strain rate had no effect on material behaviour, as the stress amplitude and
lifetime were close nearly the same in the two tests. In Fig. 5.6 there are illustrations of
various possibilities to define the slope - and possibilities to get various inaccurate ones. The
first versions of the code defined the slope like the blue line causing narrow elastic ranges. By
adding calculation points various options were obtained, but like the green line shows the
result was not adequate. By fixing the peak point to the absolute maximum stress and strain
values the accuracy could be enhanced but still a variation of slopes could be reached: the red
lines illustrate sample values.
-150
450
-200
400
-250
350
Stress [MPa]
Stress [MPa]
-300
300
-350
250
-400
200
-450
-0.81%
-0.76%
-0.71%
150
0.66%
-0.66%
0.71%
Strain [mm/mm]
Test cycle #10
Figure 5.5
410
400
-360
390
Stress [MPa]
Stress [MPa]
-350
-370
-380
380
370
-390
360
-400
350
-0.795%
-0.785%
-0.775%
Strain [mm/mm]
Test data #10
Figure 5.6
0.81%
Strain [mm/mm]
-340
-410
-0.805%
0.76%
340
0.770%
0.780%
0.790%
0.800%
Strain [mm/mm]
Test data #10
Scattered data points at cycle valley and peak illustrated with various slopes.
Fig. 5.7 shows a case where the peak point is fixed to maximum values and the slope is
calculated with an average value of the first five points of the falling edge of the curve. The
differences between cycles reveal that the scattering of data points is irregular. The result is
that the progress of elastic range is vague. In Fig. 5.7 there is also the elastic modulus as
defined from the start of the test data. Most of the slopes are greater than the elastic modulus
but there are some lower values as well.
5.0E+05
4.5E+05
4.0E+05
3.5E+05
3.0E+05
2.5E+05
2.0E+05
1.5E+05
1.0E+05
5.0E+04
0.0E+00
1
10
100
Slope
Figure 5.7
1000
10000
Cycle [-]
Elastic modulus
Fig. 5.8 illustrates the form of cycles from ABAQUS analysis. Within the elastic range
ABAQUS uses the elastic modulus for the material. Hence, a reasonable approach is to define
the cyclic hardening with a constant slope, i.e. with the elastic modulus. Fig. 5.9 shows that
the curve for cyclic hardening is smooth when elastic range is defined with the elastic
modulus. The values are also higher in this case. The differences to test data are represented
in Chap. 6.1.
450
400
350
300
250
200
150
Stress [MPa]
100
Equal slope
50
0
-50
Elastic range
-100
-150
-200
-250
-300
-350
-400
-450
-0.80%
-0.60%
-0.40%
-0.20%
Test data
Figure 5.8
0.00%
ABAQUS
0.20%
0.40%
ABAQUS start
0.60%
0.80%
Strain [mm/mm]
400
375
350
325
300
275
250
225
200
175
150
125
100
0.01
0.10
Variable slope
Figure 5.9
1.00
Average ratio
10.00
100.00
Constant slope Equivalent plastic strain [m m /m m ]
The curves for cyclic hardening of an austenitic base material with various strain amplitudes
are gathered in Fig. 5.10. In some cases there may appear a deviating point in the curve. In
that case the combination of the two approaches is helpful: first, the elastic ranges are defined
using constant slope and subsequently the curve is smoothed using an average ratio.
Cyclic hardening, Base material
550
700
500
650
450
600
550
350
Stress [MPa]
Stress [MPa]
400
300
250
500
450
400
200
350
150
100
0.001 0.010 0.100 1.000 10.000 100.00 1000.0
0
00k
0.3% Constant k
0.6% Constant
300
250
0.001
0.010
0.100
0.8% Constant k
0.8% Constant k
0.3%
0.6%
1.2% Constant k
2.0% Constant k
0.8% V2
1.2%
Figure 5.10 Cyclic hardening using stress amplitudes and constant slopes.
2.0%
When using the elastic modulus as a constant slope the kinematic hardening had to be
redefined, while the yield stress was shifted. Naturally, there is no need for redefinition when
data is processed using a constant slope from the start. Fig. 5.11 shows the curves for
kinematic hardening with constant slopes. Compared to Fig. 4.2 the yield stresses are higher
and the deviation between the strain amplitudes is smaller. Furthermore, the curves do not
cross each other like in Fig. 4.2 where the slopes are somewhat indeterminate.
The curve for strain amplitude 0.3% deviates obviously, because the yield stress was defined
using a different yield criteria viz. 0.1% strain instead of 0.2% (strain offset in Fig. 4.1). This
was done because of the low degree of plastic deformation with strain amplitude 0.3%.
The fact that ABAQUS requires a throughout ascending curve for kinematic hardening leads
to a need to filter the scattering illustrated in Fig. 5.6. The code omits a value lower than the
previous one when proceeding from the yield point to the peak (visible in Fig. 5.11 as larger
gaps between data points). Moreover, if the data point with the highest strain value has a
stress value lower than the latest one in kinematic hardening curve, the code adds a data point
calculated with the largest strain value and the absolute maximum stress value of the cycle.
550
500
375
450
Stress [MPa]
350
400
325
300
350
275
300
250
250
200
0.000
0.005
0.010
0.015
0.3% Constant k
0.6% Constant k
1.2% Constant k
2.0% Constant k
225
0.000
0.001
0.020
0.025
0.8% Constant k
0.002
0.003
0.030
0.035
Strain [m m /m m ]
0.8% Constant k
A stress-strain curve from an ABAQUS analysis contains two parts: linear elastic and plastic.
The form of the plastic part is defined by kinematic hardening and the level of the highest
stress by cyclic hardening. The transition from elastic to plastic part is fit according to the two
parameters. Fig. 6.1 illustrates a sample stress-strain curve together with the test data. The
actual linear elastic part of the test data is very limited: the diagram is curved soon after the
reversal of the load. In order to follow the curve accurately the elastic range should be limited
as well, i.e. the offset strain should be very small. There is also an opposite effect: it takes a
longer time for ABAQUS solver to compute the plastic range than to compute the elastic
range. The larger the elastic range, the faster the solving is.
In this project, the offset strain is set to the so called 0.2-limit, which often stands for the yield
criterion for austenitic steels. The deviation within cycles was not defined. Instead, the
deviation of maximum and minimum stresses was studied in order to define the accuracy of
both the kinematic hardening and the cyclic hardening.
400
350
300
250
200
Stress [MPa]
150
100
50
0
-50
-100
-150
-200
ABAQUS ends elastic range here
-250
-300
-350
0.00%
0.20%
Test cycle #1
Slope
Figure 6.1
0.40%
ABAQUS
0.2% offset
0.60%
Strain [mm/mm]
0.80%
Elastic modulus
Yield stress
6.1
-5.0%
440
-5.5%
-400
-5.5%
430
-6.0%
-410
-6.0%
420
-6.5%
-420
-6.5%
410
-7.0%
-430
-7.0%
400
-7.5%
-7.5%
Cycle [-] 100
390
-440
1
10
ABAQUS min
Figure 6.2
Test min
dmin
ABAQUS max
0.0%
-400
-1.0%
-410
-2.0%
-420
1
10
ABAQUS min
Figure 6.3
10
Cycle [-]
Test max
-8.0%
100
dmax
ABAQUS data vs. test data; strain amplitude 0.8%, both kinematic and cyclic
hardening defined from strain amplitude 0.8% using average ratio (Chap. 5.1).
-390
Stress [MPa]
Stress [MPa]
-390
-3.0%
Cycle [-] 100
Test min
dmin
Stress [MPa]
Stress [MPa]
In the first comparison it is studied if the trimming approaches are adequate. In the case
illustrated in Fig. 6.2 the stress amplitude of ABAQUS analysis (cyclic hardening produced
with average ratio) is higher than that of the test data, the deviation altering from 5 to 10 %. In
Fig. 6.3 the cyclic hardening is produced using elastic modulus as a constant slope. The stress
amplitude of ABAQUS analysis is slightly higher than that of the test data. On the other hand,
the deviation is reduced being 1 to 4 %. Fig. 6.4 shows the reduction of the deviation. All in
all, constant slope seems to be a more accurate trimming factor than average ratio of stress
amplitude and elastic range.
420
0.0%
415
-1.0%
410
-2.0%
405
-3.0%
400
-4.0%
395
-5.0%
390
1
10
ABAQUS max
Cycle [-]
Test max
-6.0%
100
dmax
ABAQUS data vs. test data; strain amplitude 0.8%, both kinematic and cyclic
hardening defined from strain amplitude 0.8% using constant slope (Chap. 5.2).
0.0%
-1.0%
Difference [%]
-2.0%
-3.0%
-4.0%
-5.0%
-6.0%
-7.0%
-8.0%
1
Cycle [-]
10
dmin 0.8% AR
dmin 0.8% CS
dmax 0.8% AR
100
dmax 0.8% CS
Figure 6.4 Comparison between differences in minimum and maximum stresses (AR=
average ratio, CS= constant slope)
6.2
As Fig. 5.11 shows, the curves for kinematic hardening are close to each other regardless of
the strain amplitude. A study was carried out to assess how accurate it is to use a common
kinematic hardening. In Fig. 6.5 there are minimum and maximum stresses of both test data
with strain amplitude of 0.8% and of ABAQUS analysis results where the kinematic
hardening is taken from the strain amplitude of 2.0% and cyclic hardening from the strain
amplitude 0.8% itself.
-390
0.0%
440
0.0%
-400
-2.0%
430
-2.0%
-410
-4.0%
420
-4.0%
-420
-6.0%
410
-6.0%
-430
-8.0%
400
-8.0%
-440
1
10
ABAQUS min
Figure 6.5
-10.0%
Cycle [-] 100
Test min
dmin
Stress [MPa]
Stress [MPa]
Further, it was studied whether the cyclic hardening of some other strain amplitude could be
used. Fig. 6.6 shows a sample from ABAQUS analysis results for strain amplitude of 0.8%
using both the kinematic and the cyclic hardening of strain amplitude 2.0%. As Fig. 5.10
shows, the higher the strain amplitude, the stronger the hardening is with equal equivalent
plastic strain. Therefore the result presented in Fig. 6.6, i.e. an ascending deviation from test
data, is predictable.
390
1
10
ABAQUS max
-10.0%
Cycle [-] 100
Test max
dmax
ABAQUS data vs. test data; strain amplitude of 0.8%, kinematic hardening
defined from strain amplitude of 2.0%, cyclic hardening defined from strain
amplitude of 0.8%; constant slope (Chap. 5.2).
-380
0.0%
520
0.0%
-400
-5.0%
500
-5.0%
-420
-10.0%
480
-10.0%
-440
-15.0%
460
-15.0%
-460
-20.0%
440
-20.0%
-480
-25.0%
420
-25.0%
-500
-30.0%
400
-30.0%
-35.0%
Cycle [-] 100
380
-520
1
10
ABAQUS min
Test min
Stress [MPa]
Stress [MPa]
dmin
10
ABAQUS max
-35.0%
Cycle [-] 100
Test max
dmax
Figure 6.6 ABAQUS data vs. test data; strain amplitude 0.8%, both kinematic and cyclic
hardening defined from strain amplitude of 2.0% using constant slope (Chap. 5.2).
Fig. 6.7 illustrates that the best result is achieved when using both kinematic and cyclic
hardening produced from test data while the strain amplitude is the same as in ABAQUS
analysis. On the other hand, kinematic hardening can be defined with high strain amplitude
without losing accuracy. It is not recommended, however, to use cyclic hardening of different
strain amplitude.
0.0%
Difference [%]
-5.0%
-10.0%
-15.0%
-20.0%
-25.0%
-30.0%
1
dmin 0.8%/0.8%
dmax 0.8%/0.8%
Figure 6.7
10
dmin 2.0%/0.8%
dmax 2.0%/0.8%
Cycle [-]
100
dmin 2.0%/2.0%
dmax 2.0%/2.0%
6.3
-450
10.0%
650
5.0%
-500
8.0%
600
3.0%
-550
6.0%
550
1.0%
-600
4.0%
500
-1.0%
-650
2.0%
450
-3.0%
0.0%
Cycle [-] 100
400
-700
1
10
ABAQUS min
Test min
Stress [MPa]
Stress [MPa]
A test run up to the end of symmetric cycles could be carried out with strain amplitude of
2.0%. In one of the tests cracking appeared after 90 cycles. By using coarse steps (20 steps
per load cycle), the ABAQUS run could be carried out within a reasonable time. Fig. 6.8
illustrates the result that the difference between maximum stresses remained within 3%. The
difference between minimum stresses ascended but did not exceed 10%. In Fig. 6.9 there are,
together with the kinematic hardening based on test cycle #10, curves for test cycle #80 as
well as for cycles #10 and #80 from the ABAQUS analysis results. The curves were defined
in the same way as in case of kinematic hardening. The shift due to cyclic hardening is visible
as well as the accuracy of the ABAQUS analysis results with coarse steps.
dmin
-5.0%
Cycle [-] 100
10
ABAQUS max
Test max
dmax
Figure 6.8 ABAQUS data vs. test data; strain amplitude of 2.0%, both kinematic and cyclic
hardening defined from strain amplitude of 2.0% using constant slope (Chap. 5.2).
650
600
Stress [Mpa]
550
500
450
400
350
300
250
0.000
0.005
Test #10
Figure 6.9
0.010
0.015
ABAQUS #10
0.020
Test #80
0.025
0.030
ABAQUS #80
0.035
Strain [-]
6.4
-450
2.0%
-475
0.0%
-500
-2.0%
-525
-4.0%
-550
1
10
ABAQUS min
-6.0%
Cycle [-] 100
Test min
Stress [MPa]
Stress [MPa]
The samples in Chaps. 6.1-6.3 deal with austenitic base metal, which tends to harden in low
cycle fatigue tests. In this chapter there are two samples of a hardening material, viz.
austenitic weld metal without heat treating. Fig. 6.10 shows the elastic range up to 100th cycle
and a comparison between minimum and maximum stresses with strain amplitude of 1.2%.
The diagrams show also the accuracy of modelling softening tendency with kinematic and
cyclic hardening as well as defining the cyclic hardening using elastic modulus as constant
slope.
525
2.0%
505
0.0%
485
-2.0%
465
-4.0%
445
-6.0%
425
1
dmin
10
ABAQUS max
-8.0%
Cycle [-] 100
Test max
dmax
420
410
400
390
380
370
360
350
340
1
10
Cycle [-]
100
-460
10%
-480
8%
-500
6%
-520
4%
-540
2%
-560
1
10
ABAQUS min
0%
100 Cycle [-] 1000
Test min
dmin
Stress [MPa]
Stress [MPa]
Fig. 6.11 shows a case where ca. 80% of the specimen lifetime (200 cycles) has been
analysed. In this case the accuracy of minimum and maximum stresses improves towards the
end of the lifetime. The maximum deviation is, however, 7.5% (minimum stress) vs. 4%
(maximum stress).
540
6%
520
4%
500
2%
480
0%
460
1
10
ABAQUS max
-2%
100 Cycle [-] 1000
Test max
dmax
340
330
320
310
300
290
280
1.00
10.00
100.00
After finding a procedure to trim result data to get smooth curves for cyclic hardening,
hardening and softening can be reproduced in ABAQUS analysis with a reasonable accuracy.
Fig. 7.1 shows the deviation of minimum and maximum stresses between test data and
ABAQUS analysis for three strain amplitude levels. The best accuracy is reached by using
parameters generated from tests with the same strain amplitude as in the analysis. Kinematic
hardening defined from high strain amplitude can be used for lower strain amplitudes. Cyclic
hardening, however, depends on strain amplitude, which can be seen e.g. in Fig. 5.10. Not
only the elastic range is higher with higher strain amplitude, but the difference grows with
growing equivalent plastic strain. Figs. 7.2 and 7.3 illustrate cycles with different strain
amplitudes and the same equivalent plastic strain.
10.0%
Difference [%]
5.0%
0.0%
-5.0%
-10.0%
1
Figure 7.1
10
100
dmin 0.8%
dmin 1.2%
dmin 2.0%
dmax 0.8%
dmax 1.2%
dmax 2.0%
Cycle [-]
Stress [MPa]
200
100
0
-100
-200
-300
-400
-500
-600
-0.020
-0.015
-0.010
-0.005
0.000
0.005
Strain [m m /m m ]
Figure 7.2
0.010
0.015
0.020
Stress-strain curves of cycles having the same equivalent plastic strain 1.4.
Stress [MPa]
200
100
0
-100
-200
-300
-400
-500
-600
-700
-0.020
-0.015
-0.010
-0.005
0.000
0.005
Strain [m /m m ]
Figure 7.3
0.010
0.015
0.020
Stress-strain curves of cycles having the same equivalent plastic strain 4.5.
The elastic ranges and stress amplitudes from Figs. 7.2 and 7.3 are collected in Fig. 7.4. The
ratio between different strain amplitudes changes with growing equivalent plastic strain. In
many analysis cases cyclic loading contains variable amplitudes instead of constant
amplitude. Therefore the usability of parameters for constant amplitudes will be tested when
strain amplitudes are variable. There is test data available for comparisons to be run during
2006. Interest will be set on the possibilities to model the progress of the cyclic hardening
with a range of strain amplitudes, i.e. the progress of CSSC (=cyclic stress strain curve).
700
600
Stress [MPa]
500
400
300
200
100
0
0.0%
0.5%
ER/1.4
ER/4.5
1.0%
SA/1.4
1.5%
SA/4.5
2.0%
Strain am plitude [m m /m m ]
Figure 7.4 Elastic ranges and stress amplitudes with various strain amplitudes and
equivalent plastic strains (ER= elastic range, SA= stress amplitude, 1.4=equivalent plastic
strain 1.4, 4.5=equivalent plastic strain 4.5).
The future research plans and needs are discussed in the following.
By simulating with ABAQUS the fatigue data of particular stainless steels under particular
cyclic loadings, the overall aim is not only to be able to repeat very accurately the original data,
but moreover to find through the analyses such material and work hardening models together
with suitable model parameter values which would be transferable to actual piping components.
It is necessary to analyse more austenitic stainless piping steel fatigue test data of more varying
nature before the full capabilities of the applied numerical models can be assessed. Only then it
is possible to decide if the applied numerical models are capable enough for describing the
fatigue behaviour of austenitic stainless piping steels under low-cycle loading, or whether it
would be necessary to develop a separate material model for this purpose. Through such
analyses it would be possible to find those material and work hardening models together with
suitable model parameter values which would best suit to be transferable to actual piping
components.
In particular it would be of interest to simulate with ABAQUS the fatigue behaviour of
austenitic stainless piping steels under fast loading transients, and under variable amplitude
loading.
The data reading procedure of the fatigue test data treatment application should be extended
to the capability of reading continuous data. Also, the analysis procedure should be upgraded
with cycle analysis, i.e. the capability to identify individual loading cycles from a data stream.
Further, the application should be developed to be capable of recognizing variable amplitude
loading - this far the considered data has been limited to output from constant loading
amplitude tests. The analysis procedure for variable amplitude loading is more complicated
than the one for constant amplitude loading. Still, there is a growing amount of interest and
need for studies of effects of variable amplitude loading.
References
1. Cronvall, O. Numerical modelling of low-cycle fatigue behaviour of austenitic stainless
piping steel. Technical Research Centre of Finland (VTT), Research Group Structural
Integrity, Research Report BTUO72-031200. Finland, 2004. 42 p.
2. Cronvall, O., Sarajrvi, U. Treatment and numerical simulation of fatigue data of
austenitic stainless piping steel. Technical Research Centre of Finland (VTT), Research
Group Structural Integrity, Research Report BTUO72-041317. Finland, 2005. 47 p.
3. Lubarda, V. Elastoplasticity Theory. CRC Press. U.S., 2002.
4. Lubliner, J. Plasticity Theory. New York 1990, Macmillan Publishing Company. 495 p.
5. Drucker, D. J. of Appl. Mech. 26, 101. 1959.
6. ABAQUS Theory manual, Version 6.4. ABAQUS Inc., 2003. Pawtucket, Rhode Island,
U.S.A.
NKS-126
Title
Simulation and Analysis of Data for Enhancing Low Cycle Fatigue Test
Procedures
Author(s)
Affiliation(s)
ISBN
Date
April 2006
Project
NKS_R_2005_40 CorroisionFatigue
No. of pages
32
No. of tables
No. of illustrations
40
No. of references
Abstract
The simulation and analysis of data for enhancing low cycle fatigue test
procedures is discussed in this report. The analysed materials are an
austenitic stainless piping steel and an austenitic weld material. This
project continues the work performed in 2003 and 2004. The fatigue test
data treatment application developed within the project in 2004 for the
preparation of the fatigue data has been developed further. Also, more
fatigue test data has been analysed with the application than in 2004. In
addition to this numerical fatigue simulations were performed with FEM
code ABAQUS. With the fatigue test data treatment application one can
e.g. both calculate cyclically certain relevant characteristic values, e.g.
elastic range, and form a set of certain cyclical parameter values needed as
a part of ABAQUS analysis input files. The hardening properties of metals
were modelled with both isotropic and kinematic hardening models.
The further development of the application included trimming of the
analysed data, and consequently trimming of resulting hardening
parameters. The need for the trimming arose from the fact that the
analysed fatigue test data presents some scatter caused by the limited
accuracy of the test equipment and the sampling rate. The hardening
parameters obtained from the application analysis results were used in the
subsequent ABAQUS analyses, and then the fatigue test data were
compared with the ABAQUS simulation results. After finding a procedure
to trim result data to get smooth curves for cyclic hardening, hardening
and softening could be reproduced in ABAQUS analysis with a reasonable
accuracy.
The modelling of the fatigue induced initiation and growth of cracks was
not considered in this study. On the other hand, a considerable part of the
fatigue life of nuclear power plant (NPP) piping components is spent in
the phase preceding the initiation and growth of cracks.
Key words
Available on request from the NKS Secretariat, P.O.Box 49, DK-4000 Roskilde, Denmark.
Phone (+45) 4677 4045, fax (+45) 4677 4046, e-mail nks@nks.org, www.nks.org