Lecture5 All

IFB
A.K. Pickett, 2013-2014

Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
FE composites and failure/damage
FE composites: Ply modelling for orthotropic elastic

and failure-damage behaviour
Outline of lecture:
Composites constitutive laws:
1. Linear orthotropic elasticity
2. Composites classical failure criteria (for limit load design)
3. Advanced Ply damage and failure laws (for impact and crash)
Calibration of a PAM-CRASH Bi-Phase ply damage model

Some basic differences between Solid and Shell elements
Tutorials:
1. Shell problem having damage and failure (PAM-CRASH Explicit)
2. Solid problem having damage and failure (PAM-CRASH Explicit)
3. Comparison of classical ply failure and damage models (PAM-CRASH Explicit)
1
IFB
A.K. Pickett, 2013-2014


Outline of lecture:
Some theory for composites orthotropic elasticity matrices
Necessary transformations for element stiffness calculations
Further transformation for element stresses and strains (Fibre frame)
Failure modes in composites
Distinction of failure and damage models
Typical classical failure models for composites
PAM-CRASH models for damage and failure
Calibration of a composites damage model (using PAM-CRASH)
IFB
A.K. Pickett, 2013-2014


Outline of lecture:
IFB
A.K. Pickett, 2013-2014

Element stiffness matrix: Remember the derivation for an isotropic material

The element stiffness matrix is derived using the following integral (=stiffness according to
the principle for minimisation of potential energy) which requires the strain-displacement
matrix [B] and elasticity matrix [E].
f = B T EBdV u = [k ] u
V
Element stiffness
matrix [k]
y23 0 y31 0 y12 0

y23 0 y31 0 y12 0
1
1
[k ] = t A 0 x32 0 x13 0 x21 [E] 0 x32 0 x13 0 x21
2A
2A
x32 y23 x13 y31 x21 y12
x32 y23 x13 y31 x21 y12
Elasticity matrix [E]
x
E

y =
1 2
Beware [B] is in global frame; [E] must also be in global xy

The elasticity law for the element depends on material

type; elastic, orthotropic (composite), elasto-plastic,
For a 2D plane element (e.g. plane stress) this is a 3*3
matrix relating the 3 stress components to strains.
frame (important for anisotropic materials).
1
1
0 0
0 x

0 y
1
2 xy
IFB
A.K. Pickett, 2013-2014

Element stiffness matrix: Comparison of isotropic and orthotropic elasticity

For a composite element the stiffness derivations are the same; except that the material
elasticity [E] changes from an isotropic to an orthotropic definition.
The elasticity matrix [E] is usually denoted [Q] for a ply in composites notation.
E
Q11 =
T
1 2 Isotropic Case
f = B EBdV u = k u
E
E
Q22 =
V
1 2
E
Stiffness [k]
E
E
Q12 = Q21 =
=
0 1
1 Q11 Q12
2
1
1 2
2 = Q21 Q22
0
0
12
[]
Q11 =
E1
For an orthotropic material the

elasticity [E] is defined in the
principal (fibre) frame (1,2).
E1
(1 12 21 )
Q12 = Q21 =
Orthotropic Case
21 E1
12 E2
Q66 = G12
A.K. Pickett, 2013-2014

Rather like the transformations we did for the 1D

Bar element stiffness (Local-to-Global frame) it is
necessary to transform the elasticity matrix from the
Fibre to the Global frame.
After this the element stiffness can be computed (in
the Global frame) and assembled for the Global
stiffness matrix [K].
c2 s2
2cs
2
2
[T ] = s c 2cs
cs cs c 2 s 2
(1 12 21 ) (1 12 21 )
The transformed orthotropic elasticity matrix
E2
Q22 =
(1 12 21 )
) (
Q66 = G
E2
IFB
0 2
Q66 12
It will have to be transformed

to the global frame (x,y) before
computing [k].
[Q ] = [T ]
Q11
[Q] [T ] = Q16
Q16
Global frame
c = Cos
s = Sin
Q12
Q22
Q26
Q16
Q26
Q66
Local (Fibre) frame
As with any composites analysis it is necessary to define fibre directions. The can be done
in a number of ways:
1. For example from a fixed vector direction in the Global frame, or from a vector relative
to the Local element frame.
2. Or from a process simulation of the draping and mapping of results to the mechanical
analysis.
IFB
A.K. Pickett, 2013-2014

Accumulation of [E] for many plies
A composite laminate (usually) consists of many plies.

Each ply has its own fibre direction, mechanical/
failure properties and thickness.
A summation is done to compute [E] for all plies in the Global frame:
[ ]
1. [Q] is computed for each ply and converted to Q
2. These are then summated (in the common Global frame) to give
the so-called ABD matrix
[E] =
A B
B D
This is usually a 6*6 matrix having in plane stiffness [A], bending stiffness [D] and a
stretch-bending coupling matrix [B]; each 3*3.
It will have reduced size for 2D or other problems.
The [ABD] matrix is used in the previous slide for stiffness [k].
We shall derive the [ABD] matrix in the lecture on laminate analysis.
IFB
A.K. Pickett, 2013-2014

The full 3D orthotropic elasticity matrix
Note the usual notation

is [C] for the full system
The same operations as before would be required for a solid element:

1. Define a fibre direction relative to the Global frame and input fibre frame mechanical data
(E1, E2, G12)
2. Compute the 6*6 orthotropic elasticity matrix (above) and transform it from Fibre to Global
frame.
3. Compute the element stiffness in the Global frame, etc
8
f = B T EBdV u = [k ] u
V
IFB
A.K. Pickett, 2013-2014

Composite element stresses and strains
{u} = [K ] {P}
1
{ x , y } = [B ]{ue }
{ue }
{ x , y } = [E ]{ x , y }
1
{ 1, 2 } = [T ]{ x , y }
{ 1, 2 } = [T ]{ x , y }
IFB
A.K. Pickett, 2013-2014


Outline of lecture:
10
IFB
A.K. Pickett, 2013-2014

Composite failure modes
Mechanical Loading
2.Compressive failure in the fiber direction.

3.Tensile, or compressive, failure in the
transverse directions.
4.Fiber debonding at the fiber-matrix
interface (often occurs under shear loads).
5.Matrix crazing (microscopic cracks) due to
loading, aging, curing, thermal stresses...
Fractographs of the Failure Zone
2
11
Fibre tensile
rupture
11
3
Fibre compression
kink band
1
22
Matrix Failure
1.Tensile failure in the fiber direction; this

may start at stress levels well below
ultimate failure.
Intra-laminar Failure
Failure of composites can be due to:
Fibre Failure
22
Transverse tension /
compression
Matrix transverse tensile cracking
3
1
12
12
In-plane shear
6.Delamination: Separation between plies at

the resin rich interface.
23
23
7.Others.
Matrix shear cracking
In addition there may be other causes for

structural failure; e.g.
This may also be due to other causes, such
as excessive deflections or buckling.
Inter-laminar
Failure
Out-of-plane shear
Delamination
(Mode-I)
Ply separation
11
IFB
A.K. Pickett, 2013-2014

Methods to handle composites failure

Essentially two methods are most commonly used in FE codes:
1.
Use a conventional composites failure criteria and either:

a. Eliminate (or partially damage) the Finite Element once the criteria is
reached. Better if this is done ply-by-ply.
This would change the structure stiffness matrix and a non-linear Implicit or
Explicit FE analysis is required. This will be necessary if the structure ultimate
failure loads is required.
b. Keep the element, with no change in elastic mechanical properties, and
output for visualisation purposes only when (at what load) the failure criteria
is reached. Useful for the design of initial failure load.
2.
Use a damage and failure criteria:

This allows individual, or coupled, failure of the different possible failure modes.
Used for problems where the post-failure behaviour is needed; e.g. composites
impact and crash analysis. Problems are highly non-linear and most applications
12
use an Explicit FE code.
IFB
A.K. Pickett, 2013-2014

Classical Ply failure versus impact/crash damage models
Damage and failure criteria
Failure criteria
Classical ply failure models

include:
Useful for:
Poor estimations between

known coupon test points
No material non-linearities
For impact/crash we need other

criteria:
Identification of each failure mode
Independent damage/failure of each
failure mode:
o Fibre failure in tension
o Fibre buckling in compression
o Transverse or shear matrix
Max. stress
Max. strain
Tsai-Hill/Wu
Puck + many others..
First (or maybe last) ply failure

Limit load design
1
22
22
Transverse tension /
compression
Matrix transverse tensile cracking
13
IFB
A.K. Pickett, 2013-2014

Failure criteria
Failure criteria
14
IFB
A.K. Pickett, 2013-2014

Typical conventional composites failure criteria: For design type

analysis
Usually empirical formulae are used to approximate the combination of the different
modes of failure that may occur. The most popular failure criteria for composites are:
1. Maximum stress criteria.
2. Maximum strain criteria.
3. Tsai-Hill failure criterion.
Quadratic criteria
4. Tsai-Wu failure criterion.

5. BUT there are many others (Puck, Hoffman, ).
Failure criteria are empirical formulae used to describe failure of a single composite ply;
they have constants that must be adjusted to fit experimental (coupon) failure data.
It is hoped, sometimes dangerously, that the criteria is valid over a wide range of loadings
often well outside the regions used in testing to calibrate the criteria failure constants.
Under complex loading the law may be invalid as another failure modes could initiate first.
Prediction of failure is risky and should preferably be supported with appropriate testing.
15
IFB
A.K. Pickett, 2013-2014

Maximum stress criterion (1/2)

Failure is assumed to occur if any one of the stresses in the material (fibre) system (1, 2,
6, 4, 5) exceed the material strength. That is, any one of the following conditions is true:
1 > F1t
if 1 > 0
abs ( 1 ) > F1c if 1 < 0
2 > F2 t
if 2 > 0
abs ( 2 ) > F2 c if 2 < 0
6 > F6
4 > F4
shears
5 > F5
6= F6
2
6
design
space
Zone of safe
combinations
of 2 and 6
abs(2)= F2c
2= F2t
The above shows graphically the stress condition (2 0, 6 0, all other stresses = 0).
The safe design space is within the limiting failure conditions F6 , F2c and F2t .
This failure envelop may easily be extended to include the additional stress dimensions.
16
Note: The quantities F1t, F1c, F2t, etc. are ultimate (failure) values. If these are used in
design a factor of safety must be included to protect against failure. Alternatively,
they may be allowable values to limit material damage.
IFB
A.K. Pickett, 2013-2014

Maximum stress criterion (2/2)

The Puck criteria
models this part
well
Experimental
failure envelope
6
Note:
1.The maximum stress criterion may be over
conservative due to compressive loading helping
to resist shear failure.
2.BUT more importantly it may over predict failure
loads for mixed mode loading due to the
combined action of loads on failure.
3.Generally the failure model is good if failure is

dominated by one stress component (which is
measured by a test!).
17
IFB
A.K. Pickett, 2013-2014

The Tsai-Hill Criterion

As previously mentioned a disadvantage of the Maximum stress and Maximum Strain
criteria is that they do not consider the interaction of stress/strain components on failure.
These criteria are not conservative when two or more principle stresses/strains are close
to their ultimate value.
The Tsai-Hill criteria tries to account for interaction of stress/strain components. The
proposed fitting law is:
Inplane shear failure term
( ) + (
f 2
1
2
1
(F )
f
1
2f )
2
(F1 )
( ) + ( ) + ( ) + ( )
f 2
2
2
2
(F )
f 2
6
2
6
(F )
Failure strength values for the ply

Conditions of stress that causes
failure on the failure envelope
f 2
4
2
4
(F )
(F )
1 = 0
eg failure for F2
(note symmetry)
Interaction term for fibre and

transverse (matrix) failure
18
f 2
5
2
5
Failure
region
(F 1)
2
( )
f 2
2
2
2
(F )
=1
No fail
region
(F< 1)
Tsai-Hill for 2- 6
1= 0 stress space
Fibre failure term
IFB
A.K. Pickett, 2013-2014

Comparing failure criteria and safe design

2
Interactive quadratic
criteria e.g. Tsai-Wu
1
F2T
F1C
1
Safe design
stress space
F1T
Maximum stress
F2C
Maximum strain
In real design some people work on the safe side and test several selected failure criteria.
Then use the within the bounds of the selectedl criteria as the safe loading area.
Examples of potentially dangerous predictions!!!!
19
IFB
A.K. Pickett, 2013-2014

Failure criteria available

in PAM-CRASH
Visualisation only
Failure activated
Available
failure
criteria
20
10
IFB
A.K. Pickett, 2013-2014

Example of laminate failure prediction and failure envelopes
Failure data for a ply (in

the fibre frame)
Lay-up
Mechanical data
http://tu-dresden.de/die_tu_dresden/fakultaeten/
fakultaet_maschinenwesen/ilr/aero/download/laminatetheory/index_html
Dresden
notation:
RIIt = F1t
RIIc = F1c
Rt = F2t
Rc = F2c
RII = F6
21
IFB
A.K. Pickett, 2013-2014

TU Dresden program: Plot of example failure envelopes
22
11
IFB
A.K. Pickett, 2013-2014

TU Dresden program: Plot of example failure envelopes
Imposed load
on laminate
Stresses (or strains) per ply

23
IFB
A.K. Pickett, 2013-2014

Failure
criteria
Scale factor
for failure
Failure
type
24
12
IFB
A.K. Pickett, 2013-2014


Outline of lecture:
25
IFB
A.K. Pickett, 2013-2014

Some differences of the Bi-Phase and Global ply model

G0
The Global ply and Bi-phase are fundamentally different. The Global ply has a
damage+plasticity model and Bi-Phase uses damage only (but also other differences)
The unloading is not important for continuous progressive loading cases; but possibly
important if residual deformations and springback are required (Global ply is then better)
Global ply is strictly more correct; but some data/parameters are difficult to get (or
understand !)
26
26
13
IFB
A.K. Pickett, 2013-2014

Typical input for failure data for PAM-CRASH composite models

Basic structure of PAM- composites cards (these are linked and reference each
other):
Parts cards
Reference fibre direction data (this can
be simple vectors, or an external file
from a separate draping simulation)
Materials cards
Mat 31 for solid
Mat 130/131/132 for shell
Specific output for Contour/THP plots
For shells laminate stacking sequence
Ply cards
Elastic data for orthotropic behaviour
Damage data for composite models
And/or ply failure data (for shells only)
Information on layups,
orientations (relative to the
directions in Part cards),
thicknesses, ply types
Actual ply data (modulus and

damage); failure model to be
used and data
27
IFB
A.K. Pickett, 2013-2014

Damage type methods suitable for impact and crash: Allows treatment
of individual composite damage modes
The FE code PAM-CRASH contains two damage models suitable for impact and crash
analysis of composites. They are typical state-of-the-art models in this field. The main
features are:
1.
The Bi-Phase model for UD and woven composites (shell and solid elements):
a. Considers fibre and matrix separately for mechanical stiffness and failure.
b. Used simple damage mechanics to degrade mechanical stiffness properties.
c. A useful model if only limited or composites manufacturers data is available.
2.
The Global Ply model (Ladeveze model) for UD and woven composites (shells only):
a. Considers the composite as a homogenised orthotropic material.
b. Uses damage and plasticity laws to degrade mechanical properties.
c. Probably more accurate, but does require some special (non standard) mechanical
testing for shear behaviour.
28
14
IFB
A.K. Pickett, 2013-2014

Ply Model 1: The Bi-Phase ply damage model

The model separates matrix and fibre phases
(Bi-phase). For each phase separate matrix and
fibre damage functions can be defined.
Calibration against coupon tests is not straightforward.
FIBERS (f) +
=
UD-COMPOSITE (UD)
3
MATRIX (m)
2
=
UD= unidirectional
f = fiber
m = matrix
stress
Simple damage functions are used to degrade

mechanical properties for fibres and matrix phases,
E() = (1 - d ()) Eo
Eo
E()
Residual
strength
Damage function: this is based on the state

of fibre strain, or J1/J2 in the matrix.
Various (standard) techniques
are available for definition of the
fibre directions.
OR information may be
transferred from a pre- draping
analyses (V2010).
strain
Z
8
n2
l
l
l
l
l
l
l
l
y
4
4
l
l
n1
global
system
2
1
local system
for bricks
1
2
local system
for shells
29
IFB
A.K. Pickett, 2013-2014

Ply Model 2: The Global Ply damage model

Three damage variables, d1, d2 and d12, characterise ply damage:
1. Axial (Fibre breakage):
E1 = E1o ( 1 - d1 )
2. Transverse (Matrix microcracking):
E2 = E2o ( 1 - d2 )
3. Shear (Fibre matrix debonding):
G12= G12o (1 - d12 )
E1
2
1
G12
Gd = Go (1 - d12 )
Go
12
23
22
E2
21
13
12
11
Ed = Eo (1 - d)
Fibre direction response
12
Diffused matrix
microcracking
due to
transverse
loading
22
12
Shear direction response
Model needed for

damage growth
Model needed for damage

and plasticity growth
11
Fibre/matrix
interface
debonding
due to
shear
loading
1
Fibre failure due

axial loading
30
15
IFB
A.K. Pickett, 2013-2014

Some features required by composite ply damage and failure models

Shear and transverse directions
Fibre Direction
250
200
e.g. residual
damage =0.9
150
E=131 GPa
100
50
E=118 GPa
experiment
simulation
0
0.000
0.050
0.100
0.150
0.200
0.250
31
Non-symmetric modulus in tension-compression
Non-symmetric failure in tension-compression
Non-linear modulus (especially in compression)
Residual damage
Shear is symmetric but non-linear with

coupled modulus damage and plasticity
Criteria for failure (e.g. Critical shear

damage)
Similar methods are required for

transverse matrix loading and failure
31
IFB
31
A.K. Pickett, 2013-2014

Typical fitting simulation to test (woven fabric - Global Ply model)

-0.009
-0.008
-0.007
-0.006
-0.005
-0.004
-0.003
-0.002
-0.001
1000
0
0
900
M46J Experimental
M46J Sim
T1000J Experimental
T1000J Sim
-50
Tension
Simulation T1000J
Experiment T1000J
Simulation M46J
Experiment M46J
800
700
-100
xx (MPa)
600
xx
-150
Compression
500
400
-200
300
-250
200
100
-300
0
0
-350
0.002
0.004
0.006
xx
0.008
0.01
0.012
0.014
0.016
xx
180
160
140
xx (MPa)
120
100
80
Simulation 7
Specimen 7
Simulation 11
Specimen 11
60
40
Shear
20
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
xx
32
16
IFB
A.K. Pickett, 2013-2014

Example calibration strategy using PAM-CRASH model 131 ply

type 0 (Bi-Phase)
Example for a UD composite material covering elastic, damage and failure
Tests needed to calibrate the model are:
1. Tension/Compression in the fibre direction
2. Tension/Compression in the transverse fibre direction
3. Shear in the 45 direction
The model is most easily

checked using the three
single element test cases
shown below
2
1
Case 1: Axial fibre
Case 2: Transverse matrix
Case 3: Shear
33
IFB
A.K. Pickett, 2013-2014

Calibration description and datasets

4
2
1
Case 1: Axial fibre
Case 2: Transverse matrix
Case 3: Shear
http://www2.ifb.uni-stuttgart.de/FEM/TutorialsAndMaterialCalibration.htm
Note:
All documentation and datasets are on the web at the above link
The documentation and datasets use material model 131 - Ply type 0 (=Bi-Phase); both
shell and solid elements are covered, and extension to woven materials are included.
34
17
IFB
A.K. Pickett, 2013-2014

Calibration strategy
First assign and check the elastic data:
1. Fiber direction (x) - tension
2. Fiber direction (x) - compression
3. Shear (x-y) use normal loadings with fibers rotated at 45 to the loading (x and y)
directions
4. Transverse direction (y)
Second assign and check the failure data:

1. Fibre direction (x) - tension fiber failure
2. Optional fibre direction (x) - compression fiber failure
3. Shear (x-y) matrix failure (loading as for item 3 above)
4. Fibre direction tension fiber and shear failure
5. Transverse direction (y)
35
IFB
A.K. Pickett, 2013-2014

Datasets from link
36
18
IFB
A.K. Pickett, 2013-2014

Ply Mechanical data: Elastic Case 1

no failure
Ply identification no.
E = slope = that given

by the law of mixtures
Density.
Ply type.
Cross plot of aux4
(ordinate) versus
aux1 (absisca)
Example of tension x (in the fibre direction
(no failure case)
Matrix data tension.
Fibre data tension.

Matrix data compression
Fibre data compression

37
IFB
Matrix and fibre modulus
Fibre volume ratio
A.K. Pickett, 2013-2014

Ply Failure data:

Shear Case 3
Example of shear xy (Case

3) with failure.
Crossplot of aux6 (stress12)
ordinate versus aux3
(strain12) absisca
Intermediate
Initial
Gi = di*Go
G=stress/12 where 12= 212
IMPORTANT: Input and output
are tensor shear strain 12
NOT engineering shear strain 12
Final
Matrix failure (shear)

Matrix failure (hydrostatic)
Fibre failure.
tension
Data for element under
hydrostatic compression
38
19
IFB
A.K. Pickett, 2013-2014

Ply Failure data:

Transverse tension
Case 2
Example of tension y in the

transverse direction (with
failure data).
Crossplot of aux5 (stress2)

versus aux2 (strain2)
Intermediate
strain
Initial
strain
Ei = di*Eo
E = slope = E22 : Note
for the transverse
direction the input (not
the law of mixtures) is
used.
Final
strain
Note:
The model does not allow control of failure for transverse tension (Case 2 loading).
For example, the above curve is obtained for transverse tension using the previous
slide (Case 3) data: The failure observed is actually due to shear shear input data.
39
IFB
A.K. Pickett, 2013-2014

Material cards
Number of plies in
the laminate
This line repeats NOPLY times:

Ply cards
Thickness of ply
Angle of fibres w.r.t the
reference axis (see PARTS
cards)
Special auxilary outputs for the composite plies:

Number of ply in the stack
The reference number if the required auxilary variable
40
20
IFB
A.K. Pickett, 2013-2014

Ply cards
Connected material
Type of fibre axis definition

and the vector direction for
the fibres
ANGPL
Vector
41
IFB
A.K. Pickett, 2013-2014

An additional exercise
Take the single element test case (loaded
in the fibre direction) and make sure the
elastic and failure data is complete.
In Visual-Crash open the materials cards
and convert this to a laminate having 4
plies. Each ply is 0.25mm thick; use a
layup of 0/45/45/0.
Add a section force so that the boundary
reactions can be stored/monitored. Option
Crash > OutPut > Section Force;
select the type Support and add the
two reaction nodes to the nodal selection
list.
Store the dataset under a suitable name
and run it using PAM-CRASH.
0 plies break
matrix and 45 plies break
Start Visual-Viewer and load the .THP

file. For the reaction support (force
magnitude) you should get the adjacent
42 plot.
21
A.K. Pickett, 2013-2014

IFB
Finite Elements: Composites damage and failure

with Shell and Solid elements
Outline of lecture:
Brief discussion on differences between shell and solid elements for

stress and failure analysis
Tutorial examples:
o
Shell element - Example composites damage and failure

analysis of a plate with a hole
Solid element - Example composites damage analysis of a

plate with a hole
43
A.K. Pickett, 2013-2014

IFB
Comparing solids and shell elements for FE failure analysis

Essentially:
Shell elements:
Through thickness strains are assumed to be a function of the in plane strains only
There is no stress in the thickness direction (plane stress is assumed for thin shells)
Therefore proper tri-axial stress prediction is not possible (even with a fine mesh)
The element is best for analysis of thin walled problems that are essentially determined
by in plane loading.
Solid elements:
A fine 3D solid mesh will predict accurate tri-axial stress distributions
This element type should therefore be used in problems having complex 3D stress
distributions
Or for problems where in planar loadings cause complex 3D stresses; e.g. around
notches.
These differences should be borne in mind during this exercise; the
following few slides illustrate some of these points.
44
22
IFB
A.K. Pickett, 2013-2014

Solid elements: Typical applications

3D solid elements are needed for problems requiring (accurate) 3D stress distributions
The number of elements (and CPU costs) increase rapidly with the number of elements
Usually shells are used for thin sheets, but solids may be needed if necking and failure
has to be modeled (in these cases high tri-axial stresses occur)
Instability point
Void growth (in the

constrained center
Void nucleation
Central fracture and

edge shear band failure
(in ductile metals)
45
Damage and
failure
Only a solid 3D model can properly analyse

3D stresses through the thickness and
around the notch. Prediction of these
stresses in a specialised failure law like
Gurson can properly simulate metal failure.
Assumes plasticity strain softening due to void growth
IFB
A.K. Pickett, 2013-2014

Example using 3D solid elements: Analysis of a tufted composite laminate

bonded to Aluminium
46
23
IFB
A.K. Pickett, 2013-2014

Shell elements
True
stress
Engineering
stress
Instability
point
x
y
Conventional shell and similar 2D elements:
Assume zero stress in the thickness direction
Strain is approximated using a constant volume assumption
z = -(x + y)
A shell element cannot predict proper stress tri-axiality and therefore cannot predict
failure (very) accurately:
1. For metals the Forming Limit Diagram can be reasonably used with shell elements
to estimate failure (at least to the instability point).
2. For composites a damage law or classical failure criteria (e.g. Max stress, TsaiWu) can similarly be used.
47
IFB
A.K. Pickett, 2013-2014

A new development in this field: Adaptive shell-solid remeshing to take

advantages of shells (CPU fast) and solids (expensive but accurate)
Test
48
Shells only
Crack propagation of a 3 point loaded beam using Shell-to-Solid adaptive remeshing
24
IFB
A.K. Pickett, 2013-2014

Exercise 1
Converting the metal hole-in-a-plate
example to a composites problem
having damage and failure:
1. For shell elements (damage and failure)
2. For solid elements (damage and failure)
3. For shell elements (classical failure criteria only)
Note:
These damage and failure analyses are all to be done with the Explicit FE method.
Handling progressive damage with an Implicit code requires a non-linear solution since [K]
changes continuously with damage. It could be done, but would be difficult and CPU expensive.
49
IFB
A.K. Pickett, 2013-2014

Getting started: Converting the metal 'Hole in a Plate' problem to

composites
1. Use the full model of the elastic (Aluminium) plate with the hole problem
HoleInPlate_FullModel_Elastic.pc
http://www.ifb.unistuttgart.de/FEM/TutorialsAndMaterialCalibration.htm
2. Make a copy and re-name it, e.g.

e.g. HoleInPlate_FullModel_ExplicitComposite131.pc
3. The only things to be changed (see next slides) are:
131 is the material

model UD ply
Replace the existing Aluminium part and material cards with a composite
elastic-damage model.
Some controls and loading to make it run faster (mainly for teaching purposes)
Some useful additional outputs (e.g. support force)
50
25
IFB
A.K. Pickett, 2013-2014

Part, Material and Ply data: Additional ply cards are needed
Metals
Composites
Parts
Geometric data: e.g. thickness
Parts
Geometric data: e.g. thickness
ALSO the reference fibre direction
Materials
Mechanical data: This is mainly given in
the ply cards see below
Materials
Mechanical data: e.g.modulus,
yield, strain rate.
The laminate stacking sequence
Ply data
Mechanical data: e.g.modulus, damage and
failure data in Tension and Compression
New see next slides
51
IFB
A.K. Pickett, 2013-2014

Some basics: The compliance law and input for mechanical data:
The Bi-phase composites models is used in PAM-CRASH for this study:

1. For elastic analysis:
The above data must be given for shells (selected) or solids (all)
2. For damage analysis:
Modulus reduction (damage) is strain controlled and requires

additional damage data
52
26
IFB
A.K. Pickett, 2013-2014

Some basics: The damage law and input parameters

(i) fracturing damage function :
damage function d(e)

e- ei
d1
e1 - e i
e- e1
du- d1 + d1
eu- e1
1- (1- du) eu/e
d(e)
1
du
d1
ei
e1
eu
(ii) modulus damage :

E(e)
e = equivalent strai n
ei = initial threshold strai n
e1 = intermediate
d1 = intermediate damage
range
ei < e < e 1
e 1 < e < eu
eu < e <
E 0 = initial elastic modulus

s u = resi dual constant stress level
eu = ulti mate strain
du = ulti mate damage
modulus damage
Eo
E(e) = 1- d(e) E 0
E1
Eu
ei
e1
eu
(iii) stress-strain diagram :

range
stress-strain relations (e)
s (e)
s
1
s
E(e)
E 0 e 1-
ei
e1
eu
ei < e < e1
d -d
E 0 e 1- d1 u 1 (e- ei )
eu- e1
s u = E 0 eu 1- du = const
const
su
d1
(e- ei )
e1- ei
e1 < e < eu
eu < e <
(iv) equivalent scalar strain measures (e = strain tensor ; s = unit tensor) :

volume damage
e = ev = I1 = tr (e)
53
shear damage
e = es =
IFB
A.K. Pickett, 2013-2014

(tr = trace)
J2 = 1/2 tr (e - 1/3 ev d) 2
1/2
Compression
Tension
1. Add a new composites ply

and define the data
54
e.g. Matrix part

mechanical data
e.g. Fibre mechanical data
(NB uses law of mixtures to
combine Fibre and matrix)
Eo
(1-di)Eo
(1-du)Eo
i 1 u
27
IFB
A.K. Pickett, 2013-2014

Failure data used for study

See data sets in directory: Example_CalibrationElementM131_BiPhase/Failure
Shear failure data (hyd.tension)

Fibre failure data (tension)
Shear failure data (hyd. compression)

Fibre failure data (compression)
Run: TensionY_TransvFailure
Crossplot: Aux2 v. Aux5
Run: ShearXY_MatrixShearFailure
Run: TensionX_FibreFailure
55
IFB
A.K. Pickett, 2013-2014


composites

131 is the material

model UD ply
56
28
IFB
A.K. Pickett, 2013-2014

Use material type 131
2. Redefine the material type to 131 (= Composite shell)
e.g. set to 4 for a four ply laminate

This opens set the ply number,
thickness and fibre angles for the 4 plies
Plies are
referred
to here
57
IFB
These are outputs selected I have output auxs 1-6 = strain 11,22,12
and stress 11,22,12 respectively in the fibre frame for the first ply only
A.K. Pickett, 2013-2014

3. Parts: Make sure this is linked to the new material
3
4
YL
1
YG
XG
Global frame
Options here are:
=0 for a reference vector in the global frame
Local frame
this is w.r.t
the element
nodes (e.g.
xL lies along
nodes 1-2)
=1 for a reference vector in the local element frame

58
29
IFB
A.K. Pickett, 2013-2014


composites

131 is the material

model UD ply
59
IFB
A.K. Pickett, 2013-2014

4. Some other changes to make the analysis suitable for explicit loading
a. Place all the lower nodes in an OUTPUT > SECTION FORCE type support this will give
a time-history of the reaction force at the base in the results file, so we can see the
value of maximum load at failure.
b. Make sure the analysis type is EXPLICIT and the

Solver is CRASH
c. Set a suitable termination time (RUNEND) to 1 msec
60
d. In option OCTRL select

suitable time history (x-y)
and deformed plot output
intervals
30
IFB
A.K. Pickett, 2013-2014

5. Final step before running

Finally, give the top
nodes a velocity
loading (CRASH >
LOADS > 3D BC)
type VELBC. Give
the velocity in the ydirection increasing
from 01msec at
10mm/msec and
then held constant.
Ramping the load
limits dynamic
inertial effects.
Note that for 1msec

of simulation this
gives an end
displacement of
5mm.
61
IFB
A.K. Pickett, 2013-2014

6a. Example results: Damage and other contour plotting (from the .DSY)
file)
Example plot of total damage at the final
state (T=0.6msec). This is approximately
the final damage at full section failure.
Example plot of AUX 1 (fibre

stress 11) at time = 0.6msec.
This is just after the point of
breaking (see next slide).
62
31
A.K. Pickett, 2013-2014

IFB
6b. Example results: Reaction force time history (from the .THP file)
Below is a plot of reaction force (SECTION direction y) from the .THP file. The coupon
breaks at about 100kN and dynamically oscillates about 10-20kN (this is a typical dynamic
mass effect). Slower imposed loading over a longer duration would reduce this post failure
oscillation.
Partial failed section and

oscillating residual load
Max load ca. 100kN
63
A.K. Pickett, 2013-2014

IFB
Studies with the shell model: Varying the layup

It is a simple exercise to try different laminate lay-ups: We shall try to compare failure for
laminates having layups:
1.
2.
3.
4.
[0,0,0,0]
DONE
[45,45,45,45]
[90,90,90,90]
[45,-45,-45,45]
Copy the previous dataset to: Name_45deg.pc, Name_90deg.pc, Name_45S.pc
For each (either with an editor in the .pc file, or with Visual-PAM):
1.
2.
3.
Edit the layup to have the required sequence (variation 2, 3 or 4 above)

Change RUNEND (e.g. increase to 1.5 msec) to make sure of failure
Modify the title (TITLE /- card in the controls) to reflect the model; e.g name_45deg.
Then run each model and compare results with the previous study [0,0,0,0]
Note there is a simple way to overlay results as shown:

1. Generate the first required plot in the format you would like
2. Click on the curve (left mouse key) and it will show black
3. Click the right mouse key and select Tools > Chase curve(s)
4. Finally, find the .THP file of the next plot you want; select it,
and the curve will be automatically plotted.
64
32
IFB
A.K. Pickett, 2013-2014

Results: Reaction force time histories
0
Definition used for
the 0 direction
Do not worry about the oscillations in the post- failure phase. Its is due to mass (inertia)
effects caused by the rapid loading used here to reduce CPU time.
The non-linear initial loading (to failure) is due to the velocity control used for loading.
65
IFB
A.K. Pickett, 2013-2014

Results: Damage contour plots

Results from similar models may be compared in one window (as shown below):
1. Open the first model (.DSY
file)
2. In Window > Page
Layout select the layout
option for 4 plots.
3. Then click on each new
empty window (it will show
green and open a new
results (.DSY) file
4. The 4 windows can be
synchronised using
Results > Syncronise;
Each will then have the
same time state
5. Activate in contours the
shell Total Damage option
for each window.
66
33
IFB
A.K. Pickett, 2013-2014

Mesh dependancy
Damage type problems can be very mesh

dependent.
A fine mesh computes greater stresses around
the notch and damages quicker at a lower load.
One work-around is to calibrate using a certain
mesh size and use this in structural problems.
See directory:
Example_HoleInPlate_MeshDependency
67
IFB
A.K. Pickett, 2013-2014

Consistent results: Dont always expect explicit codes to give exactly the same results!
Small differences can occur depending on precission (single/double),
version, machine, operating system
2008
2007
2006
2008
2007
2006
68
34
IFB
A.K. Pickett, 2013-2014

2008
2007
2006
2008
2007
2006
69
IFB
A.K. Pickett, 2013-2014

Converting the shell model to an equivalent solid model

The shell mesh can be used to develop a 4 layered solid model (1 layer of solids is needed
per ply); that is, four layers of solids will be needed to have the same representation as the
previous multi-layered shell model.
This requires more CPU due to four times more elements; also, there is a much smaller
simulation timestep associated with the small solid element size in the thickness direction.
Briefly the steps to follow are:
1. Make a copy of the previous composites shell model:
e.g. HoleInPlate_FullModel_SolidComposite30.pc
30 is the solid
material model
for a UD ply
2. Start Visual Mesh and load in the dataset

3. The following 2 slides show generation and modification of the shell mesh to get
the equivalent solid model mesh.
70
35
IFB
A.K. Pickett, 2013-2014

Build the solid mesh from the shell mesh

1. Use the option 3D > layer mesh. Specify a distance of 2.0mm (representing the
laminate thickness), select all the shell elements (faces) and then update faces and
press mesh to open the second panel.
2. Use 4 for the number of layers. Then press Mesh and Apply.
Save the dataset.

Perform the next steps
using Visual Crash for
PAM for all the entities
(loads, boundary
conditions...), materials,
etc.
71
IFB
A.K. Pickett, 2013-2014

Delete unwanted parts

(e.g. the old shells) and convert each layer of solids to
have its own PID number. This is needed here so that
each part (PID) can have a different fibre direction
Open the Parts in the explorer panel; press the part to
be deleted (e.g. old shells) and open the new panel with
the right mouse key. Press delete to remove it.
Renumber the PIDs for each layer of elements
1. Select the second layers of solids (drag a box over

the elements with the left mouse key) then with
the right mouse key open the panel to change
PID, e.g:
Change the second layer to PID 21
Change the third layer to PID 22
Change the forth (top) layer to PID 23
72
36
IFB
A.K. Pickett, 2013-2014

Material cards:
Assigning the
material model, ply
mechanical data
and output
(auxiliary)
information
In the material
editor set material
type to 30 (= solid
UD composite) and
the other data as
shown in rings.
Save this (Apply
and close).
The selected output for contour and time history plotting are chosen via auxiliary reference
numbers (e.g. the forth auxiliary is set to 9 for output of total damage dt)
73
IFB
A.K. Pickett, 2013-2014

Parts cards: These link to the material/ply cards and also set the fibre directions
In this case each part represents one layer. Link the material model
(IMAT) and set the fibre directions, e.g.:
+45 ply x_local = 1,1,0 and y_local = 1,-1,0
-45 ply x_local=1,-1,0 and y_local = 1, 1,0
Note: Solid elements need 2 vectors to be defined; representing the main fiber and the
transverse directions. The third direction is computed as the cross product
74
37
IFB
A.K. Pickett, 2013-2014

Other things to be defined:
1. Redefine all the base nodes to have a suitable boundary condition (either fully fixed,
or free with fixed only in the loading direction). Similarly, the top nodes could be
constrained from out-of-plane movements.
2. Load the nodes at the top using imposed velocities in the y-direction (VEL3D). The
same loading function as for the shell problem should be used.
3. Put all the base nodes in a section force time history so we can get the reaction force
time history.
4. The control data should be the same as for the shell model (actually it should already
exist from the previous shell model that has been adapted).
5. Save the dataset and run it.
6. Results (shown on the next slides) are visualised with Visual-Viewer.
75
IFB
A.K. Pickett, 2013-2014

Example output: Aux4 (for total damage at 1msec)
76
38
IFB
A.K. Pickett, 2013-2014

Comparing solid and shell models (total damage at failure 0.6msec)
77
IFB
A.K. Pickett, 2013-2014

Further Exercise:
Using a classical composites failure criteria
(e.g. Maximum Stress)
1. Use the previous shell model and copy the input dataset to a new
file having an appropriate name.
2. Open this new file using Visual Crash for PAM.
3. Make the following changes; run the model and examine the
results.
78
39
IFB
A.K. Pickett, 2013-2014

Typical composites failure criteria

Interactive quadratic
criteria e.g. Tsai-Wu
2
F2T
F1C
F1T 1
Safe design
stress space
Maximum stress
F2C
Maximum strain
Select a criteria (most codes have several)

The safe area is stress/strain states inside the envelope, failure is stress/strain states
on or outside the chosen criteria (envelope).
The complete material is assumed to fail once the failure criteria is reached.
79
IFB
A.K. Pickett, 2013-2014

Ply card modifications:

This parameter opens
options to activate
classical composites
failure criteria
Set all damages to

zero (making them
inactive)
New data for maximum

stress criteria. This data is
similar to the previous
Biphase damage model
80
Flag for element elimination

(=1) or visualisation only (=0)
Type of criteria (e.g.

= 5 Max stress
40
IFB
A.K. Pickett, 2013-2014

Material card
modifications:
Use 130; or better is 131
The IFAIL parameter allows

choice of the criteria for
element elimination:
first ply,
ratio of plies,
all plies
Mostly new auxiliary
variables are set for the
chosen failure criteria
(see table below
extracted from the PAMmanual)
81
IFB
A.K. Pickett, 2013-2014

Comparing Maximum Stress and Biphase damage results
Maximum stress criteria
Biphase criteria
82
41
IFB
A.K. Pickett, 2013-2014

Example contour plots
Max Stress: Damage

at 0.6msec
Biphase: Damage
at 0.6msec
Beware: These comparisons look good,

but a damage based criteria is better!
Separate damage modes are properly

treated, rather than simple element
elimination once a criteria is reached.
83
IFB
A.K. Pickett, 2013-2014

Some questions:
1. Can the shell model capture and model the effect of delamination ?
2. Can the solid model capture and model the effect of delamination ?
84
42
IFB
A.K. Pickett, 2013-2014

Methods for delamination

Solid: Thin resin only layers of solids could be added between the fibre plies
(computationally expensive)
CPU expensive (small timestep and
high number of elements), but
potentially good accuracy.
Stacked Shell: A rather new approach of modelling shells for each ply and
delamination interfaces to tie the plies. These are sometimes
called cohesive zone elements, or interface elements
Stacked shells (for ply failure) with special
tied interfaces (for delamination failure).
In this case solid elements are used to
model the core in this sandwich composites.
CPU costs are reasonable even for large
structures.
85
43

Lecture5 All

Uploaded by

Copyright:

Available Formats

Lecture5 All

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture5 All

Uploaded by

Copyright:

Available Formats

IFB

A.K. Pickett, 2013-2014

FE composites: Ply modelling for orthotropic elastic

Calibration of a PAM-CRASH Bi-Phase ply damage model

A.K. Pickett, 2013-2014

FE composites: Ply modelling for orthotropic elastic

Some theory for composites orthotropic elasticity matrices

Necessary transformations for element stiffness calculations

Further transformation for element stresses and strains (Fibre frame)

Failure modes in composites

Distinction of failure and damage models

Typical classical failure models for composites

PAM-CRASH models for damage and failure

Calibration of a composites damage model (using PAM-CRASH)

A.K. Pickett, 2013-2014

FE composites: Ply modelling for orthotropic elastic

Some theory for composites orthotropic elasticity matrices

Necessary transformations for element stiffness calculations

Further transformation for element stresses and strains (Fibre frame)

Failure modes in composites

Distinction of failure and damage models

Typical classical failure models for composites

PAM-CRASH models for damage and failure

Calibration of a composites damage model (using PAM-CRASH)

A.K. Pickett, 2013-2014

Element stiffness matrix: Remember the derivation for an isotropic material

y23 0 y31 0 y12 0

Elasticity matrix [E]

Beware [B] is in global frame; [E] must also be in global xy

The elasticity law for the element depends on material

frame (important for anisotropic materials).

A.K. Pickett, 2013-2014

Element stiffness matrix: Comparison of isotropic and orthotropic elasticity

For an orthotropic material the

A.K. Pickett, 2013-2014

Rather like the transformations we did for the 1D

The transformed orthotropic elasticity matrix

It will have to be transformed

Local (Fibre) frame

A.K. Pickett, 2013-2014

Accumulation of [E] for many plies

A composite laminate (usually) consists of many plies.

1. [Q] is computed for each ply and converted to Q

A.K. Pickett, 2013-2014

The full 3D orthotropic elasticity matrix

Note the usual notation

The same operations as before would be required for a solid element:

A.K. Pickett, 2013-2014

Composite element stresses and strains

A.K. Pickett, 2013-2014

FE composites: Ply modelling for orthotropic elastic

Some theory for composites orthotropic elasticity matrices

Necessary transformations for element stiffness calculations

Further transformation for element stresses and strains (Fibre frame)

Failure modes in composites

Distinction of failure and damage models

Typical classical failure models for composites

PAM-CRASH models for damage and failure

Calibration of a composites damage model (using PAM-CRASH)