You are on page 1of 50

Introduction to

FRANC2D, FRANC2D/L and FRANC3D


Prof. Tony Ingraffea
Dr. Wash Wawrzynek
Cornell University Fracture Group
ari1@cornell.edu
www.cfg.cornell.edu

ASTM Workshop on Computational Fracture Mechanics for Composites


Salt Lake City
March 22-23, 2004
1
Outline of Presentation
Computational Fracture Mechanics (CFM) 2004
A Taxonomy of Approaches

CFM the Cornell Way


Adaptive FEM/BEM Approaches with FRANC2D/3D

Examples Applications
Fatigue cracking in fuselage structure
Ductile tearing in fuselage structure
Fatigue cracking in transmission gear

A Look into the Near Future in CFM


Multiscale simulation of initiation and propagation

2
The Lay of the Land In Computational Fracture Mechanics
Cracks are either explicitly represented in the geometry model of the structure, or
they are implicitly acknowledged in the numerical model of the structure.

GEOMETRICAL NON-GEOMETRICAL
REPRESENTATION REPRESENTATION

Constrained Arbitrary Constitutive Kinematic


Shape Methods Shape Methods Methods Methods

Prescribed Methods Meshfree Methods Smeared Crack Enriched Element


Methods Methods

Analytical Geometry Adaptive FEM/BEM


Methods Methods Element Extinction X-FEM
Methods
Known Solution Lattice Methods
Methods Computational Cell
Methods
Particle Methods

Atomistic Methods

3
FRANC2D: A Crack Propagation Simulator
for Plane and Axisymmetric Structures
Finite-element-based, mixed-mode, LEFM, stress intensity factor calculator.
Interactive, menu-driven, integrated pre/post processing and FEM analysis.
Automatic re-meshing to accommodate arbitrary crack growth:
-- no limit on number of cracks, automatic propagation under LEFM.
Interface and contact mechanics.
Nonlinear fracture mechanics via cohesive zone models.
Newton and dynamic relaxation solvers.
Mid-80s origin with continuous development to present.

30

Stress Intensity (ksi sqrt(in))


25

20 KI
15

10

5 KII
0
0 0.5 1 1.5 2 2.5
-5
Crack Length (in.)

4
FRANC2D/L: A Crack Propagation
Simulator for Plane Layered Structures
Finite-element-based, mixed-mode, LEFM, stress intensity factor calculator.
Interactive, menu-driven, integrated pre/post processing and FEM analysis.
Automatic re-meshing to accommodate arbitrary crack growth:
-- no limit on number of cracks, automatic propagation under LEFM.
Interface and contact mechanics.
Nonlinear fracture mechanics via cohesive zone models.
Newton and dynamic relaxation solvers.
Mid-90s major variant of FRANC2D, continued development to present.

PLUS
Layering of plane regions, using adhesive and/or riveted connections.
Any number of cracks in any layer.
Out-of-plane loads for plate bending analysis.
EPFM/CTOA with state mapping for tearing simulations.

5
FRANC2D/L 2-Layer Analysis of Effect of Patch on
Crack Growth from Hole


45000

40000
FRANC2D/L: No Patch
35000 ANALYTIC: No Patch

Stress Intensity (psi-in1/2)


FRANC2D/L: With Patch
P a t ch over 30000
pla t e wit h h ole
25000
0.5" R
E =10.0E 6 psi 8.0"
=0.25 20000
t h ick =0.04 in
Adh . G=10.0E 3 psi 15000
Adh . t h ick =0.005 in
10000

5000

4.0"
0
0 0.2 0.4 0.6 0.8 1 1.2

Crack Length (in)

Schematic of Plate with Hole Effect of Patch on Reducing Stress


Intensity
6
Automatic Simulation of Arbitrary Crack Growth in the
FRANC Family of Simulators

A simple, 2D, crack growth simulation using FRANC2D

By way of a simple,
2D simulation
of crack growth in
a gear

7
FRANC3D: A Crack Propagation Simulation
System for 3D Shell and Solid Structures
Designed as a value-added crack growth simulation environment for use
with your FEM/BEM analysis engine.
Interactive, windows-driven, integrated pre/post processing.
Automatic re-meshing to accommodate arbitrary crack growth:
-- no limit on number of cracks, automatic propagation under LEFM.
Late 80s origin, with continuous evolution to present.

8
Overview of Manual Input
finite element
the Simulation model:
OSM
ANSYS,
Process with PATRAN,
NASTRAN,
FE
Translation
IDEAS,
FRANC3D BLADE-GT, or
ABAQUS Solid
Model
Boundary
Rep. File
format

FRANC3D
Boundary
Crack Conditions
Finite element model Propagation
translation using OSM is
Fracture Introduce Mesh
described in the Analysis Flaw(s)
documentation on the CFG yes no
web site: Existing Superposition
Flaw? yes ?
www.cfg.cornell.edu no
Life Crack Face
Prediction Loads

BES
FEM/BEM
Results (Stress Analysis
Analysis
Analysis)
File File

9
Outline of Presentation
Computational Fracture Mechanics (CFM) 2004
A Taxonomy of Approaches

CFM the Cornell Way


Adaptive FEM/BEM Approaches with FRANC2D/3D

Examples Applications
Fatigue cracking in fuselage structure
Ductile tearing in fuselage structure
Fatigue cracking in transmission gear

A Look into the Near Future in CFM


Multiscale simulation of initiation and propagation

10
Fuselage Fatigue
Crack Growth
Simulation
Global-Local
Hierarchical Modeling

11
Predicted Curvilinear Fatigue Crack Growth:FRANC3D/STAGS

Potyondy, Wawrzynek, Ingraffea, Int. J. Num. Meth. Engr., 38, 1611-1633 1995.
12
Plastic Zone Evolution in KC-135 Fuselage Panel:FRANC3D/STAGS

20 in.

10 in.

Effective Stress (ksi)

60 yld
53
46

39

Chen, C.-S., Wawrzynek, P. A., Ingraffea, A. R., AIAA Journal, 40, 566-575, 2002.

Chen, C.-S., Wawrzynek, P. A., Ingraffea, A. R., AIAA Journal, 40, 1644-1652, 2002.
13
Mesh Model of SH 60 Seahawk
Power Transmission Spiral Bevel Gear
Teeth Hub

JMESH Application
14
Initial Flaw Size and Location

Problem Demands
214,000 -
ELEMENTS 327,000
920,000 -
DOF 1,400,000

15
Comparison: Simulated versus Observed

Observed

Simulated

Crack Trace on the


Face of Tooth

16
Comparison: Simulated versus Observed
Fracture Surfaces

Simulated

Observed
17
Comparison: Simulated vs Observed
Crack Trace on Gear Hub

Observed
Crack

Simulated
1 cm

18
Mesh Detail on Tooth Surface

Later Stage of Simulation

Initial Flaw/Mesh

19
Outline of Presentation
Computational Fracture Mechanics (CFM) 2004
A Taxonomy of Approaches

CFM the Cornell Way


Adaptive FEM/BEM Approaches with FRANC2D/3D

Examples Applications
Fatigue cracking in fuselage structure
Ductile tearing in fuselage structure
Fatigue cracking in transmission gear

A Look into the Near Future in CFM


Multiscale simulation of initiation and propagation

20
Current Research in the CFG
Continuing advances in computational power (eg. //HPC) and laboratory
observations (eg. EBSP) are enabling progress in two main thrust areas:

1. Understanding basic mechanisms for crack initiation and growth.


2. Performing very high resolution simulations of initiation and growth
at appropriate length/time scales.

1 mm

With support from: NSF-ITR EIA-0085969, NSF RI EIA-9972853, NASA


LaRC NAG-1-02051, NASA Institute for Future Space Transport,
DARPA, and the Cornell Theory Center
21
Multiscale Modeling Concept
Structure-Scale

~1 m

~1 mm

Meso-Scale:
Polycrystal
~1 nm

Atomic Scale:
Grain/Particle Boundaries
22
Cyclic Loading Simulation in FRANC2D/L
n

Elastic-plastic
0.5 mm

orthotropic
grains; cohesive
grain boundaries y

x
yy (MPa)
Average applied strain

6.9e-3
500

100
Time, Cycle 23
Scaling Tests of 3D Polycrystal

50 grain polycrystal Surface mesh

24
Detail of Meshing of Grains

> 7,000,000 tets

> 3,000,000 nodes

Using JCrystal mesher

25
Simple Scaling Test on a Target Polycrystal
Geometry:
50 m cube
100 Grains
6,271,419 DOF
1,519,816 10-noded
tetrahedra
Solution:
64 Dell 1550 servers
Using 1 CPU each
3669.28 s (61 minutes)
1 GHz PIII CPUs
2GB RAM
Giganet interconnect
Windows 2000 Advanced
Server
Wireframe displaced shape
MPI Pro 1.6.3 Displacement contour - ||u||

26
Illustrative, Small 3D Problem
1 mm cube
5 grains
Elastic grains
E = 72 GPa
= 0.3
CCZM
c = 500 MPa
c = 10 m
k0 = 200 MPa
2% applied xx
Mesh:
3174 Tetrahedral elements
201 Interface elements
932 Nodes
15,744 DOF

27
3D Polycrystal: xx Contours

(GPa)
28
3D Polycrystal: Grain Boundary Separation

Cracked

Decohering

Closed

29
3D Polycrystal: Interface Normal Traction
on Grain Boundaries

(GPa)
30
Summary

Wide variety of CFM approaches available


identifiable by their crack representation schemes.

The FRANC family of simulators uses geometrical representation, allows arbitrary


crack numbers and shapes, via automatic, adaptive FEM/BEM remeshing.

Ever-improving high-performance computing, and techniques for observing cracking


processes in finer detail, DEMAND that CFM simulations include better physics
and mechanics, within more realistic, statistically meaningful geometries.

therefore

Fracture mechanics does not apply to composite materials, Single, dominant cracks
do not occur in composites: damage is diffuse and multi-phenomenal: in my opinion,
these are no longer meaningful or useful statements.

31
FRANC2D/L 2-Layer Analysis of Effect of Patch on
Crack Growth from Hole

Plate Layer Patch Layer


32
Empirical Criterion for Crack Tearing in
Thin Aluminum Sheet
When Crack Tip Opening Angle, CTOA, Reaches a
Specified Value, the Crack Can Tear.

10 m

Micrograph of Crack Tip Region in 2024-T3 Aluminum Sheet


Micrograph Courtesy of NASA LaRC 33
Predicted KC-135 Residual Strength
With a 10.0 inch Long Lead Crack, With and Without
Multi-Site Damage (MSD) and Corrosion
16
No MSD (RS = 15.3 psi)
Operating Pressure (psi)

14
required residual strength (12.5 psi)
12 MSD = 0.025 in. (RS = 11.3 psi)

10
MSD = 0.025 in., Corrosion (RS = 10.8 psi)
8

0 1 2 3 4
Total Crack Extension (inch) 34
Fatigue Loading Simulation
0.7

Applied Strain (%)


0.6
0.5
n 0.4
0.3
0.2
0.1
0
0.0 1.0 2.0 3.0
0.5 mm

Cycle

Hill (MPa) CZZM


E1 72,000 5% G 250 5% Nm/m2
E2 40,000 5% tp 500 5% MPa
y G 10,000 5% k0 4e8 GPa
xx yld 500 5% c 1m
x yy yld 450 5%
xy yld 400 5%
35
Inclusion Particles:
insoluble, undissolved or precipitated
coarse particles
formed and distributed heterogeneously

Crack Initiation Observations


cracks form at broken inclusion particles.
cracks form at decohered particle-matrix
interface of inclusions and inclusion
AA 2024-T3 sheet, 500X
clusters.

Simulations
discretely and statistically
represent particles
Perform simulations to
investigate the influence of
inclusions on fatigue crack
initiation and propagation.

100 m 36
y, 0.5 GPa

0 GPa Effective Stress at


0.4% Applied Strain
37
Particle/Matrix
Interface
Damage

debonded

Interface elements
debonded,
void left behind
Not
debonded
at 0.4% applied strain

38
Major Issues in 3D Meso-Scale
Creating grain geometry
Meshing for FEM
simulations
Assigning individual
material properties for
each grain
Determining appropriate
boundary conditions
Selecting solution
techniques
Determining necessary
computational power
39
Validation Study #1

2219-T851 aluminum beam laboratory test specimen in four-point bending

p 11.54 kips, R = 0.214


3" projection onto Section A - A
1.5"

Crack
Tip 1 .508"
A
Crack
2.412" 2" Tip 2
side view
a1 a3
A

12"

0.25R" EDM
view 2 flaw a2
initial notch

view 1 45 angle
bottom view crack

40
Initial Boundary Element Mesh

Geometric Model

Flaw Mouth
Initial Flaw

Full Beam
41
Predicted and Observed Crack Fronts

Final Fatigue Crack Front


a
Predicted Transition into Corner a
Predicted Transition into Corner b

a b
b Observed Fatigue Crack Shape

42
Final Predicted Crack Shape

Deformed solid

Experiment
FRANC3D
Comparison of computed initial
crack
and observed crack trace
on the surface
side bottom face side

43
Computed SIF History and Comparison to Observations

a1 a2

Event Observed FRANC3D %


Difference
a-Corner (cycles) 106, 800 140,000 32
b-Cor ner (cycles) 171, 000 170,000 -0.5
Last Fr ont (cycles) 175, 000 190,000 8.5
a1 (in.) 1. 26 1.42 12.7
a2 (in.) 1. 38 1.34 -2.8
44
Validation Study #2

Allison 250-C30R Engine

U.S. Army OH-58 Kiowa

Fatigue Cracks in Spiral Bevel


Power Transmission Gear

U.S. Army Vehicle Technology Center


NASA/Lewis Research Center
45
Initial, Tetrahedral Finite Element Mesh
(only element faces on the outer surfaces are shown)

Initial Flaw Location

46
Deformed Solid During Simulated Crack Growth

initial after 10
steps

after 20
steps
47
Deformed Solid During Simulated Crack Growth (cont.)

after 30 steps after 39 steps

48
Comparison Among
Observed and
Simulated Crack
Trajectories
BEM

FEM
observed

BEM

FEM

observed
49
The Current Cost of Realism
Item FRANC3D
Computer power 128 processors
PC-cluster, 2.4 GHz
Number of unknowns ~ 1 to 3 million
Max. solution time per load cycle ~ 2 hours
(1 load cycle=15 load steps)
Typical time for one crack step ~ 5 hours

50

You might also like