Linear and Nonlinear Analysis of High Dynamic Impact

Events with Creo Simulate and Abaqus/Explicit

SAXSIM 2015, TU Chemnitz
Rev. 1.1 | 31.03.2015
Dr.-Ing. Roland Jakel
 At a glance: Altran, a global leader

20+
Countries

EUROPE

23,000+ AMERICA

ASIA
Innovation Makers

5
Industries

1,756m
2014 Revenues

2
 Our Customers

AUTOMOTIVE, INFRASTRUCTURE
& TRANSPORTATION

AEROSPACE,
DEFENCE & RAILWAYS

ENERGY, INDUSTRY
& LIFE SCIENCES

TELECOMS & MEDIA

FINANCIAL SERVICES
& PUBLIC SECTOR

3
 Table of Contents
Slide:
1. Description of the Example Problem 5

2. Analysis of the Problem in Creo Simulate 6-15

2.1 Solution method coded in Creo Simulate 6
2.2 Model setup 8
2.3 Modal analysis 9
2.4 Dynamic time analysis with impulse function applied 11
2.5 Dynamic time analysis with half sine wave applied 13
2.6 Conclusions 15

3. Solution of the Problem in Abaqus/Explicit 16-31

3.1 The explicit solver for dynamic analysis 16
3.2 Framework for modeling damage and failure in Abaqus 17
3.3 Damage initiation criteria for ductile materials 18
3.4 Damage evolution 20
3.5 Definition of the ideal material response curve 21
3.6 Iterative procedure for defining a rough material response curve 24
with only very limited tensile test data available
3.7 Transfer to the example problem 29

4. References 32

4
Rev. 1.1 | 31.03.2015
 1. Description of the Example Problem
steel
Goal and given data protective
panel
Goal of the study
Show by finite element analysis that a steel
protective panel withstands the impact of
an idealized fragment
Given Data idealized
Impact fragment with m=65 kg and impact impact
energy 500 kJ fragment
v 2 Ekin / m 124,03 m / s 450 km / h
m
I m v 65kg 124 8062 Ns
s
Worst case scenario: The sharp edge of the
fragment (edge length approx. 85 mm) v
bangs in the panel
Protective panel dimensions:
thickness t=45 mm, height1,5m,
width1 m
It is assumed that the panel is simply
supported at all its edges
The impact takes place at the geometric
panel center
5
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.1 Solution method coded in Creo Simulate

Basic Equation for Dynamic Systems

Creo Simulate can only solve for dynamic problems which can be described with the
following linear differential equation of second order:
M u Cu K u F (t )
Herein, we have [M]=mass matrix, [C]=damping matrix, [K]=stiffness matrix,
{F}=force vector, {u}=displacement vector and its derivatives
Solution Sequence:
Before a dynamic analysis is performed in Simulate, a damping-free modal analysis,
M u K u 0 , is carried out to obtain the modal base for the modal transformation
The system is then transformed from physical to modal space by replacing the physical
coordinates u
Herein, is the eigenvector matrix, and modal coordinates; has a number of rows
equal to the DOF in the model, and columns equal to the number of modes;
has one column and rows equal to the number of modes
In a subsequent dynamic analysis, in which modal damping C=2M and a forcing
function is added, we have M, C and K as diagonal matrices now in modal coordinates
After the solution is performed, the solution is transformed back into physical space
for post-processing
Remark: This solution method is used in many FEM codes for linear, small damped
dynamic systems because of its computational efficiency!

6
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.1 Solution method coded in Creo Simulate

Limitations of the solution method

We have only a linear system (all matrices are constant), that means no nonlinearities
can be taken into account like
contact
change of constraints (boundary conditions)
nonlinear material
Only modal damping can be applied (max. 50 % of critical damping, =1) to keep the
damping matrix diagonal and therefore run times short

Special challenge when solving the described problem

Contact between panel and idealized fragment cannot be modelled, therefore the
unknown impact force-vs-time curve cannot be computed
The impact force must therefore be applied as external force, using some
assumptions
Most conservative approach is to model the impact as impulse function (Dirac
impulse), that means the impulse of 8062 Ns is applied in an infinite short time span
Later, conservatism may be removed by assuming the force-vs.-time curve to be a
half sine function

7
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.2 Model setup

Plate is meshed with p-bricks (mapped mesh)

To reduce the singularity of the edge
constraint (simple support), an Isolate for
Exclusion AutoGEM Control (IEAC) is used
(exclusion of stresses > p-level of 3)
Linear steel material applied with yield
limit 405 MPa
We analyze the failure index
for the distortion energy
criterion

8
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.3 Modal analysis

50 modes requested
Modal stresses requested (to speed up
later dynamic analysis)
Mass normalization requested (config-
option sim_massnorm_modes), allows to
compare modal stresses (always output for
mass normalization) and Eigenvector
displacement (usually unit normalized)
Single pass convergence with advanced
controls

9
uiT M ui 1
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.3 Modal analysis

10
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.4 Dynamic time analysis with impulse function applied

Impulse of 8062 Ns applied

11
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.4 Dynamic time analysis with impulse function applied

Time step t=0,1 ms with max. Movie of impact event

failure index (in scale) (in scale, duration 1 ms)

12
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.5 Dynamic time analysis with half sine wave applied

Results show max. von Mises stress >50 times

higher than yield limit at time t=0,1 ms
Approach is much too conservative, need to
remove conservatism by assuming the impact is
in form of a half sine wave with duration TI=T/2
of the first fundamental plate Eigenfrequency f0
The force-vs-time curve of the half sine impact
can be described as follows:
F (t ) Fmax sin(t )
The impulse then becomes:
TI
I Fmax sin(t )dt
t 0

I 0
Fmax I f 0 5,538 106 N
2
This has been coded into the Simulate
form sheet for dynamic time analysis,
see right

13
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.5 Dynamic time analysis with half sine wave applied

Time step t=1,7 ms with max. Movie of impact event (duration 10 ms;
failure index at impact location in scale; max failure index at singular
(in scale) edge constraint)

14
Rev. 1.1 | 31.03.2015
 2. Analysis of the Problem in Creo Simulate

2.6 Conclusions

With the less conservative approach, max. loading at the impact location is still
factor 8.7 above yield
With the computational approach provided in Creo Simulate therefore no strength
proof is possible
The plate had to be thickened significantly in order not to leave the linear domain of
validity
Anyway, the foreseen steel has a large ductile region that could take significant
kinetic energy of the fragment we need a computational approach that can take
this into account
engineering stress [MPa]

engineering strain [%]

15
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.1 The explicit solver for dynamic analysis

The explicit dynamics analysis procedure in Abaqus/Explicit is based upon the

implementation of an explicit integration rule together with the use of diagonal or
lumped element mass matrices for computational efficiency [1]
The explicit dynamic procedure requires no iterations and no tangent stiffness matrix:
M ui C ui R int i R ext i
Herein, {Rint}, {Rext} are the internal and external load vectors; i: time increment
The equations of motion for the body are integrated using the explicit central
difference integration rule
1
(i )
1
(i ) t ( i 1) t ( i ) ( i )
u 2
u 2
u
2
1
(i )
( i 1) ( i 1)
u u t
(i )
u 2

with u=displacement DOF with derivatives, respectively, and superscripts (i) increment
number; (i-1/2), (i+1/2) midincrement values
The central difference integration operator is explicit in that the kinematic state can
be advanced using known values from the previous increment
The stability limit for each integration time step is given by
t 2 / max

where max is the highest element frequency in the model

An analysis typically has some 100,000 increments
16
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.2 Framework for modeling damage and failure in Abaqus

To specify the material failure in Abaqus,

four distinct steps are necessary:
1. The definition of the effective (or d'
undamaged) material response
(points a-b-c-d)
2. A damage initiation criterion (c) c
3. A damage evolution law (c-d)
b
4. A choice of element deletion whereby
elements can be removed from the
calculations once the material stiffness
is fully degraded (d)

Note: a (D=1) d
D is the overall damage variable (D=01).
After damage initiation, the stress tensor
in the material is given by the scalar
damage equation
(1 D )
Note: Image shows true stress vs. log strain, the curve
with =stress in the material with should not be mixed up with a classical tensile test
diagram (engineering stress vs. engineering strain)!
absence of damage

17
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.3 Damage initiation criteria for ductile materials

Ductile criterion: Two main mechanisms can cause the fracture of

a ductile material:
A phenomenological model for predicting the
onset of damage due to nucleation, growth, 1) ductile fracture
and coalescence of voids
The model assumes that the equivalent plastic
strain at the onset of damage, Dpl, is a function
of the stress triaxiality and equivalent plastic 2) shear fracture
strain rate:
Dpl ( , pl )
p 1 ( 1 2 3 )
3
vonMises vonMises
with p=hydrostatic or pressure stress
(stress just responsible for volume change)
Damage initiation then takes place if the
following condition is satisfied:
d pl
D 1
Dpl ( , pl )
Herein, D is a state variable that increases
monotonically with plastic deformation
Its incremental increase D is computed at
each increment during the analysis

18
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.3 Damage initiation criteria for ductile materials

Shear criterion:
A phenomenological model for predicting the onset of damage due to shear band
localization
The model assumes that the equivalent plastic strain at the onset of damage, S ,
pl

is a function of the shear stress ratio S and equivalent plastic strain rate:
Spl ( S , pl )
S ( vonMises k S p ) / max
Herein, kS is a material parameter (for example 0,3 typically for Aluminum)
Damage initiation then takes place if the following condition is satisfied:
d pl
S pl 1
S ( S , pl )
Herein, S is a state variable that increases monotonically with plastic deformation
Its incremental increase S is computed at each increment during the analysis

Local necking:
This is an instability problem which is computed automatically in nonlinear
elastoplastic analysis if volume elements are used
Just if shell elements are used, special damage initiation criteria local necking
(instability)
for sheet metal instability have to be defined

19
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.4 Damage evolution

The damage evolution capability for ductile metals

assumes that damage is characterized by the
progressive degradation of the material stiffness,
leading to material failure;
uses mesh-independent measures (either plastic
displacement upl or physical energy dissipation) to
drive the evolution of damage after damage
initiation;
takes into account the combined effect of different
damage mechanisms acting simultaneously on the
same material and includes options to specify how
each mechanism contributes to the overall material
degradation; and
offers options for what occurs upon failure,
including the removal of elements from the mesh.

20
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.5 Definition of the ideal material response curve

Intuitively we may think a simple tensile test is sufficient, but

The classical uniaxial tensile test is not well suited to measure the
ideal material response of ductile materials, since it shows necking, a
superimposed instability problem
The data can therefore just be used until the maximum value of the
engineering stress vs. engineering strain curve is reached
Refined test methods, for example in [4], or other specimen types [5]
must be used to obtain reliable data

What happens during a uniaxial tensile test of a ductile specimen [3]?

Until the maximum value of the engineering stress/strain-curve is
reached, the stress state is uniaxial
Further elongation leads to instability (necking), what locally creates a
two-dimensional stress state at the necked surface and a three-
dimensional stress state within the specimen
The uniaxial stress outside the necked region then decreases, the
strain rate in the necked region increases!
The real uniaxial fracture strain can therefore not easily be measured
with this test (see [4] for a refined approach)
A better orientation for the fracture strain out of this test may be the
percentage reduction of area after fracture, Z

21
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.5 Definition of the ideal material response curve

The uniaxial tensile test

Lu L0
Fracture strain: A (Lu=rupture length, L0=initial length)
L0
S Su
Reduction of area: Z 0 (Su=smallest cross section at rupture location;
S0 S0=initial cross section)

As longer the tensile test rod, as smaller the engineering strain until failure:

Uniform elongation
(without necking)
Non-uniform elongation with necking
Ag

Engineering stress-strain curve of soft steel for different ratios of the tensile rod A=L0/d0 [2]

22
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.5 Definition of the ideal material response curve

Material Laws
To fit test data, plasticity laws may be used from literature
For example, Creo Simulate offers three material laws for describing plasticity:
linear plasticity, power (potential) law, exponential law [3]
The laws may be used especially to extrapolate the true stress-strain curve to higher
strains, if for example just tensile test data is available
However, Abaqus requires any tabular input for the curve (true stress vs. log. plastic
strain) and interpolates between the data points
Quantity Conversion
Note that the data from the uniaxial testing machine usually has to be converted in
the following way [3]
For stresses:
(true <-> engineering)

For strains:
(logarithmic or true <->
engineering)

23
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.6 Iterative procedure for defining a rough material response curve with only very limited
tensile test data available
For the material foreseen for our example problem, just data from 5 tensile test
specimens is available, so the following assumptions & simplifications where done:
No temperature and strain rate dependency
No stress state dependency (just values for triaxiality =1/3=pure tension entered)
Damage initiation only according to the simple ductile criteria
Damage evolution linear for just a small plastic fracture displacement u fpl 0,3 mm
Then, the following steps have been performed:
1. A 2D axial symmetric FEM-model of the tensile test specimen is created in Abaqus.
A small imperfection (diameter reduction) is used to have the start of necking at the
specimen center and not at the constraints
2. The ideal material response curve in the region of uniform elongation is simply
calculated out of the uniaxial test results (see equations on previous slide)
3. The ideal material response curve after onset of local necking is iteratively trimmed
by comparing the measured force of the test (engineering stress) with the reaction
force of the FEM analysis (since the analysis delivers only true stresses). Initially,
damage is not taken into account in the FEM model
4. After sufficient curve fit, the damage parameter is activated (with element removal for
D=1). Start value for the fracture strain is taken from Z and trimmed iteratively until
the calculated curve also fits on its right end within the measured curve
Note: This is no recommended procedure! It shall only give listeners new to this topic an
impression about the difficulties of a simple tensile test!

24
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.6 Iterative procedure for defining a rough material response curve with only very limited
tensile test data available
Comparison of test results with FEM analysis - forces

Entered curve is without necking!

Drop of all the curves (except

the input curve) always comes
from local necking, not from
damage initiation!

Here, damage initiation

takes place

25
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.6 Iterative procedure for defining a rough material response curve with only very limited
tensile test data available
Ideal material response curves

D 1
D 0 D0

D0

Dpl ( 1 / 3) 0,9
Response curve expressed in the
required Abaqus input format u fpl 0,3 mm

26
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.6 Iterative procedure for defining a rough material response curve with only very limited
tensile test data available
Von Mises stress within the 2D axial symmetric FEM of the tensile test specimen

27
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.6 Iterative procedure for defining a rough material response curve with only very limited
tensile test data available
Equivalent plastic strain within the 2D axial symmetric FEM of the test specimen

28
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.7 Transfer to the example problem

Abaqus finite element model

Uses half symmetry
Eight linear elements with reduced integration
over the protection panel thickness
Thickness 45 mm, in addition for comparison
models with 40 mm and 35 mm thickness
Same material model for the protection panel as
used for the test specimen
Fragment uses steel with just simple linear
behavior
Contact with friction, friction coefficient 0,1
(conservative)

29
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.7 Transfer to the example problem

Animation of the impact event, von Mises stress;

from left to right: 45, 40 and 35 mm panel thickness

30
Rev. 1.1 | 31.03.2015
 3. Solution of the Problem in Abaqus/Explicit

3.7 Transfer to the example problem

Animation of the impact event, von Mises stress;

from left to right: 45, 40 and 35 mm panel thickness

31
Rev. 1.1 | 31.03.2015
 4. References

[1] Without further notice, many Abaqus related information of this presentation is
taken from the Abaqus 6.12 documentation manuals (Dassault Systmes Simulia)

[2] Domke: Werkstoffkunde und Werkstoffprfung, 9. Auflage 1981,

Girardet, Essen, ISBN 3-7736-1219-2

[3] Roland Jakel: Basics of Elasto-Plasticity in Creo Simulate, Theory and Application,
Presentation at the 4th SAXSIM, 17.04.2012, Rev. 2.1

[4] Sren Ehlers, Petri Varsta: Strain and stress relation for non-linear finite element
simulations, Thin-Walled Structures 47 (2009), pp. 1203-1217

[5] Antonin Prantl, Jan Ruzicka, Miroslav Spaniel, Milos Moravec, Jan Dzugan, Pavel
Konopk: Identification of Ductile Damage Parameters, 2013 SIMULIA Community
Conference (www.3ds.com/simulia)

A good overview about simulating ductile fracture in steel may be found in:
Henning Levanger: Simulating ductile fracture in steel using the finite element method:
Comparison of two models for describing local instability due to ductile fracture; thesis
for the degree of master of science; Faculty of Mathematics and Natural Sciences,
University of Oslo, May 2012

32
Rev. 1.1 | 31.03.2015
Dr.-Ing. Roland Jakel
Senior Consultant, Structural Simulation

Phone: +49 (0) 421 / 557 18 4111

Mobile: +49 (0) 173 / 889 04 19
[email protected]

Henning Maue
Global aircraft engineering architect ASDR group
Head of Engineering ASDR Germany
Technical Unit Manager Airframe Structure

Phone: +49 (0) 40 35 96 31 96 122

Mobile: +49 (0) 172 83 63 335
[email protected]