LS DYNA® Aerospace Working Group MAT - 224 User Guide
LS DYNA® Aerospace Working Group MAT - 224 User Guide
LS DYNA® Aerospace Working Group MAT - 224 User Guide
LS‐DYNA® Aerospace Working Group
*MAT_224 User Guide
A User Guide for *MAT_TABULATED_JOHNSON_COOK in LS‐DYNA®
May 8, 2017
Report Updates to Tom Vasko
vaskothj@ccsu.edu
i
Table of Contents
1 Introduction ............................................................................................ 1
2 Theoretical Overview .............................................................................. 2
3 LS‐DYNA® Input Parameters.................................................................... 5
4 Stress‐Strain Relation Input Generation ................................................. 9
4.1 Elastic‐Plastic Modeling ..................................................................... 9
4.2 Stress‐Strain Relationship after Necking ......................................... 10
4.3 High Strain Rate Sensitivity .............................................................. 13
4.4 Conversion of Plastic Work into Heat .............................................. 15
5 Failure Surface Generation ................................................................... 17
5.1 Temperature Scaling Function Creation .......................................... 21
5.2 Strain Rate Scaling Function Creation ............................................. 21
5.3 Regularization Curve Creation ......................................................... 22
6 Material Model Validation .................................................................... 23
7 Test Program ......................................................................................... 24
7.1 Stress‐Strain Behavior Tests ............................................................ 24
7.2 Specimen Failure Tests .................................................................... 24
7.3 Validation Tests ............................................................................... 25
8 References ............................................................................................ 26
Appendix A Stress‐Strain Relationship Modeling Procedure Flow Chart ..... 27
List of Figures
Figure 1 True Stress‐True Strain Curve Before and After Trimming ............ 11
Figure 2 Extrapolated Curves after Necking Point ....................................... 12
Figure 3 Compression Test Stress at 5% Strain ............................................ 15
Figure 4 Stress States for a Material Testing Program ................................. 17
Figure 5 Grid and Control Point Example ..................................................... 18
Figure 6 Surface Plots Generated by MATLAB Subroutine .......................... 19
Figure 7 Discretized 3D Failure Surface for LS‐DYNA® ................................. 20
Figure 8 Flow Chart of Stress‐Strain Relationship Modeling Procedure ...... 29
List of Tables
Table 1 LS‐PrePost® History Variables ........................................................... 8
ii
1 Introduction
*MAT_TABULATED_JOHNSON_COOK (*MAT_224) is a new constitutive
model for metals that was developed in a collaborative effort of the Federal
Aviation Administration (FAA), National Aeronautics and Space
Administration (NASA), George Washington University (GWU), Ohio State
University (OSU), and George Mason University (GMU). This collaboration
resulted in the new physics‐based constitutive model for metals that can be
used to accurately predict the mode of failure and extent of damage in
aerospace impact events. Previous to the development of this model,
material failure parameters were tuned to match a specific test with a
specific failure mode. In order to obtain accurate results for a different
mode of failure, a different set of material parameters was required and,
since the failure mode had to be known ahead of time, its use as a
predictive tool in design and analysis was precluded.
This new *MAT_TABULATED_JOHNSON_COOK model is an elastic‐
viscoplastic material model with the option to define strain rate and
temperature dependent stress versus strain curves. In addition, plastic
failure strain can be defined as a function of any combination of triaxiality
and Lode parameter (both are functions of the 3D state of stress), strain
rate, temperature, and element size. The model also gives the user the
option to use the temperature or strain rate dependencies with or without
failure.
This document is a user guide for developing the material model input
parameters for *MAT_224. This user guide includes sections on the
theoretical overview, LS‐DYNA® input parameters, mechanical property
input generation, failure surface generation, material model validation, and
test program.
As a summary user guide, this document references various technical
reports that were written during the development of the material model,
all of which are listed in the reference section.
1
2 Theoretical Overview
In LS‐DYNA®, *MAT_TABULATED_JOHNSON_COOK or *MAT_224 is an
isotropic elastic thermo‐viscoplastic constitutive relation that states stress
is a function of strain, strain rate, and temperature:
, ,
Where σij is stress, εij is strain, is strain rate, and is temperature.
Tensile tests conducted at various strain rates and temperatures are used
to derive the LS‐DYNA® input stress‐strain curves and tables.
In the plastic region, the material response is determined by a von Mises
type yield surface that expands or contracts due to strain hardening, rate
effects, and thermal softening:
, , , ,
Where vm is the von Mises stress, eff is the equivalent plastic strain, and
p
effp is the equivalent plastic strain rate.
To model accumulated damage and element failure, *MAT_224 uses four
input parameters. The first parameter is either a load curve for use with
shell elements or a table of curves for use with solid elements. The load
curve defines the plastic failure strain as a function of triaxiality and the
table of curves defines the plastic failure strain as a function of triaxiality
and Lode parameter, which results in the definition of a failure surface. The
second parameter is a load curve that defines the plastic failure strain as a
function of plastic strain rate. The third parameter is a load curve that
defines the plastic failure strain as a function of temperature. The last
parameter is a load curve (or table of curves) that defines that plastic
failure strain as a function of element size (and triaxiality).
Triaxiality is defined by the equation:
2
where is the pressure defined by
3
and is the Von Mises stress defined by
3
2
The Lode parameter is defined by the equation:
27
2
where , , and are the principal deviatoric stresses. The Lode
parameter ranges between plus and minus one, which provides a
convenient boundary for the definition of the failure surface. The stress
triaxiality, however, ranges between plus and minus infinity and, therefore,
it is necessary to define reasonable boundaries. A triaxiality range of ‐1.0 to
+1.0 should be sufficient for most impact applications.
The plastic failure strain is defined by:
,
where is the triaxiality, is the Lode parameter, is the plastic strain
rate, and is the element size. When more than one of the failure
parameters are used, the net plastic failure strain is the product of the
functions defined in the above equation.
The failure criterion is based on an accumulated damage parameter defined
by:
3
where is the plastic strain rate and is the plastic failure strain, which
is continually changing based on triaxiality and Lode parameter. When this
damage parameter is greater than or equal to one, the element has failed
and is deleted.
Failure prediction using shell elements requires the plastic strain failure to
be defined as a function of only triaxiality and, therefore, a failure curve is
defined. Failure prediction using solid elements requires the plastic failure
strain to be defined as a function of both triaxiality and Lode parameter
and, therefore, a failure surface, specified in a table, is defined.
To determine the failure surface, various types of material specimens are
created and tested with each specimen having a unique triaxiality and Lode
parameter representing a specific point on the failure surface. To generate
an accurate and complete failure surface, therefore, requires as many
material specimens as possible. Using the triaxiality, Lode parameter, and
failure strain data from the specimen testing, a complete failure surface can
then be created using a three‐dimensional curve fitting tool.
After a failure surface is generated, the remaining parameter load curves
can be created. A strain rate testing series and a temperature testing series
must be completed to determine the next two failure parameters. Finally,
by varying the size of the elements in the mesh analytically, a load curve
can be created that accounts for many different sizes of discretization.
4
3 LS‐DYNA® Input Parameters
The material model input parameters are listed here with the variable
name from the LS‐DYNA® input manual, the variable symbol used in
equations, and a brief description.
RO: Mass density, ρ
E: Young’s modulus, E Temperature dependent Young’s moduli may be
specified using a load curve using a negative modulus with the absolute
value pointing to the specific load curve.
PR: Poisson’s ratio, µ
CP: Specific heat, Cp
TR: Room Temperature, TR
BETA: Amount of plastic work converted into heat, β (Taylor‐Quinney
Coefficient)
Temperature increase is caused by plastic work
The value of β may be a single constant value or, if input as a negative
number, it is the curve ID for a curve of β vs plastic strain for various strain
rates.
LCK1: A load curve ID that defines the effective stress as a function of
effective plastic strain OR a table ID that defines a load curve ID for each
plastic strain rate that gives the (isothermal) effective stress versus
effective plastic strain for that plastic strain rate. This load curve should be
defined at room temperature as specified above with the TR input
parameter.
5
LCKT: A table ID that defines a load curve ID for each temperature that
gives the effective stress versus effective plastic strain for that
temperature. These load curves should be defined under quasi‐static
conditions.
Plastic failure strain is defined as a function of triaxiality, Lode parameter,
plastic strain rate, temperature, and element size by
27
,
2
using the load curves and tables LCF, LCG, LCH, and LCI defined below. If
more than one of these four variables LCF, LCG, LCH, and LCI are defined,
the net plastic failure strain is the product of the above multiple functions.
In that case, only one of the variables LCF, LCG, LCH, and LCI can point to
curve(s) with plastic strain along the curve ordinate. The remaining non‐
zero variables must point to curve(s) that have a dimensionless scaling
factor along the curve ordinate. If a particular scaling curve is not used, the
scaling factor defaults to a value of 1.0, however, best practices
recommend that a scaling curve with ordinates of 1.0 should be defined
and referenced.
LCF: A load curve that defines the plastic failure strain or scale factor as a
function of triaxiality OR a table ID that defines a load curve ID for each
Lode parameter that gives the plastic failure strain versus triaxiality for that
Lode parameter. The table option is only available for solid elements, i.e., if
shell elements are used, only a curve of plastic failure strain as a function of
triaxiality can be input.
LCG: A load curve ID that defines the plastic failure strain or scale factor as
a function of plastic strain rate. To use logarithmic rather than linear plastic
strain rates make the first abscissa in the curve negative.
LCH: A load curve ID that defines the plastic failure strain or scale factor as a
function of temperature.
6
LCI: A load curve ID that defines the plastic failure strain or scale factor as a
function of element size OR a table ID that defines a load curve ID for each
triaxiality that gives plastic strain versus element size for that triaxiality.
NUMINT: Number of integration points which must fail before the element
is deleted.
FAILOPT: Flag for an additional failure criterion F2.
NUMAVG: Number of time steps for the running average of plastic failure
strain in the additional failure criterion.
NCYFAIL: Number of time steps that the additional failure criterion must be
met before element deletion.
The default failure criterion for the material model depends on plastic
strain evolution and on plastic failure strain and is obtained by
accumulation over time:
where element erosion takes place when 1, which provides load‐path
dependent failure.
An additional, load‐path independent, failure criterion, , can be invoked
by setting FAILOPT=1, where the current state of plastic strain is used:
Using this failure criterion, two additional parameters are available as
countermeasures against stress oscillations. With NUMAVG defined, the
plastic strain is averaged over NUMAVG time steps. NCYFAIL defines the
number of time steps that 1 must be met before element deletion
takes place.
Additional *MAT_224 history variables for post‐processing d3plot and
d3thdt database files are available using the optional keyword
7
*DATABASE_EXTENT_BINARY where NEIPS and NEIPH indicate the number
of additional history variables for shells and solids, respectively. The
available history variables are tabulated in Table 1. It should be noted that
to obtain the correct triaxiality from *MAT_224, the additional history
variables must be requested because the triaxiality calculated in LS‐
PrePost® using the d3plot files is different from that calculated in
*MAT_224.
Table 1 LS‐PrePost® History Variables
8
4 Stress‐Strain Relation Input Generation
Appendix A contains a flow chart of the stress‐strain relationship modeling
procedure that is presented this section.
4.1 Elastic‐Plastic Modeling
To generate the input curves for the effective stress as a function of plastic
strain, a series of standard tension tests is performed at various strain rates
and temperatures. In the tension tests, a dog‐bone specimen under a
constant grip is pulled in a tensile machine. During the test, the force
versus time is measured with the tensile machine, and the displacement
versus time relation is measured by an extensometer fixed to the specimen.
The force versus displacement curve is then generated by cross plotting
these two curves. Engineering stress and engineering strain are calculated
using
where eng t is the engineering stress, is the engineering strain,
F(t) is the axial force, Ao is the original cross section area, L(t) is the
instantaneous gauge length, L0 is the original gauge length, and D(t) is the
displacement. True stress and true strain are then calculated from
1
ln 1
Once the stress‐strain curve is obtained, Young’s Modulus is determined
from the slope of the curve in the elastic region and the yield strength, σy, is
determined at the end of the linear portion of the curve. If several strain
rate tests were performed, Young’s Modulus should be determined from
the lowest strain rate test. It should also be noted that while Young’s
9
modulus does not typically vary with temperature, the ability to input this
temperature variation in LS‐DYNA® is available by defining a negative
modulus with the absolute value of the modulus pointing to a curve
defining this variation. In the plastic region, the plastic strain, εp is
computed from
/
and the effective plastic strain, effp is a monotonically increasing scalar value
which is calculated incrementally as a function of the plastic strain rate
tensor, ijp as
2/3 ∙
This procedure for determining the elastic‐plastic material behavior is valid
only to the point of necking of the test specimen. Once necking begins, the
test section cross‐sectional area is no longer constant in the necked region
and a uniaxial stress state no longer exists with the development of
transverse stresses in the necked region. As a result, using the previous
equations to create the elastic‐plastic material input data after necking
occurs is invalid and a more accurate method is now described.
4.2 Stress‐Strain Relationship after Necking
The recommended method to create the post‐necking portion of the stress‐
strain input curves is a trial‐and‐error process. Simulations of the tensile
tests that explicitly model post‐necking behavior must be performed using
several different input stress‐strain curves. The input stress‐strain curve
which then most closely matches the test results will be the best
approximation of the material’s actual stress‐strain relationship.
To accurately simulate the tensile test in LS‐DYNA® after necking has
occurred, start by estimating several post‐necking plastic stress‐strain
curves for input into the tensile test simulation by extrapolating the curve
before necking. Then compare the simulation force vs. displacement curve
10
output with the test results and pick the input curve that gives the closest
match to test. At this stage, the strain rate effect is not considered in the
simulation. This process of determining a single input stress‐strain curve to
match each strain rate tested is repeated for all the tested strain rates,
which makes each individual analysis strain rate independent. Because of
this strain rate independence, the simulation can be performed with an
artificial (higher) loading speed instead of the actual loading speed that was
used in the test, allowing for a much shorter simulation time.
The necking point is given by the intersection between the true strain
versus true stress curve and its own derivative. As shown in Figure 1 only
the part of the strain‐stress curve before necking is used for further
processing.
Figure 1 True Stress‐True Strain Curve Before and After Trimming
The hardening curves are extrapolated after necking using the formula:
where for a given plastic strain εp the stress σ is calculated. The exponent n
is expected to vary between 0 and 1 as the hardening curve should be
monotonically increasing and have a monotonically decreasing tangent.
Thus, given a specific n value, k and εe are determined from
11
where A is the stress at necking, B is the plastic strain at necking, and C is
the slope (hardening modulus) at necking. After assuming different n
values, a cluster of curves can be generated as shown in Figure 2.
Figure 2 Extrapolated Curves after Necking Point
Several candidate curves are then chosen from the generated plastic strain
vs. stress curves and these curves are input into the material model and
tensile test simulations are completed. Force‐displacement curves
obtained from the simulations are then overlaid with the actual tensile test
results to select the curve that is the best estimation of the plastic strain vs.
stress relation. The input data developed using this process will match the
actual test data as closely as possible while removing high frequency
oscillations and negative slope areas that could cause convergence
problems.
Matching each test individually is required because each specimen has a
unique geometry. The exact geometry from measurement is used in the
finite element model. If the exact test geometry is not used in the analysis,
significant errors can result. Although the input curve, because it is a
function of stress and strain, can be created from an average of all the test
results, only a single geometry can be used in each simulation. The force‐
displacement results from the simulations will vary from each other, while
the stress‐strain results will not.
12
This procedure is valid for all constant strain rate (including quasi‐static)
isothermal uniaxial tensile tests, independent of the temperature at which
they were performed. If the quasi‐static tensile tests are performed at low
enough strain rates, the process is isothermal and no thermo‐mechanical
coupling occurs during the test. As a result, the yield curves obtained using
this procedure can be allocated to the specific temperature at which the
test was conducted. A table of temperature dependent yield curves can
then be created by individually simulating all temperature tensile tests for a
given strain rate without considering any thermal coupling in the
simulations.
The series of generated stress‐strain input curves from the constant strain
rate tensile tests can then be bundled together as tabulated input for
*MAT_224. Using these temperature and rate dependent yield curves
allows for the simulation of dynamic experiments where thermal and rate
effects occur simultaneously and interact with each other. However, the
procedure described above cannot be used to simulate many higher strain
tests, because they are not isothermal and do not maintain a constant
strain rate.
4.3 High Strain Rate Sensitivity
Depending on the material being tested, in many higher strain rate tests
the strain rate is not constant. This is because after necking occurs the
strain localizes, the strain rate increases in the necked area, and both the
strain and the strain rate are no longer uniform across the specimen. In
addition, since the duration of higher rate tests is short, the heat produced
by the plastic work may not conduct enough to maintain a constant
temperature. As a result, the procedure required to create isorate and
isothermal stress‐strain curves requires an additional trial‐and‐error
process.
At extremely high strain rates, the strain rate in the region of localization
may reach values significantly above the nominal strain rate for the
specimen and it is also not isothermal. As a result, the stress‐strain input
generated using isothermal and isorate tension tests may be supplemented
13
with synthetic isothermal curves generated using rate sensitivity trends
from compression tests. Compression tests maintain a more constant
strain rate than tension tests due to relatively small localization.
Compression tests can also reach much higher strain rates than the tension
tests, but complete stress‐strain curves are difficult to derive from them.
As a result, synthetic curves can be created using a combination of
information from compression and tension tests. These synthetic curves
can then be combined with the stress‐strain curves derived directly from
isorate and isothermal tension tests.
The synthetic curves can be generated from log‐log plots and curve fits of
the stress at a particular strain of the compression tests for various strain
rates as shown in Figure 3. Typically, strain rate sensitivity conforms to the
theory that the yield strength increases as a logarithmic function of strain
rate between rates of ~10‐4 sec‐1 and ~103 sec‐1, and as a linear function of
the of strain rate at strain rates greater than ~103 sec‐1 [6]. Higher strain
rate curves are created by translating a previously created isorate and
isothermal stress‐strain curve so that their stresses pass through the stress
values at the defined strain and strain rate.
14
Figure 3 Compression Test Stress at 5% Strain
4.4 Conversion of Plastic Work into Heat
During high strain rate events in the localized necking region, there is not
sufficient time for conduction to carry away the heat generated by the
plastic deformation, and, as a result, the process becomes adiabatic. This
adiabatic process causes a significant increase in the temperature in the
specimen locally and, as a result, the simulation of the tension test is
sensitive to the amount of energy generated by the plastic work that is
converted into thermal energy. The percentage of plastic work that is
converted into thermal energy is defined by the Taylor‐Quinney coefficient,
β. This coefficient is not a constant, but is a function of both strain rate and
plastic strain.
Users have the option of defining a single constant value for β, effectively a
rate independent β, or a curve of β vs plastic strain for various strain rates.
With this capability in *MAT_224, a fully coupled thermal solution that
simulates the conduction of thermal energy away from highly strained
15
elements can be performed. Without including this conduction,
temperatures will rise more in the simulations than in tests, leading to non‐
physical analytical results. This non‐physical result will be very small in high
rate, short duration simulations, and can be safely ignored. However, in
simulations where both high and low rate loading takes place
simultaneously and significant strains are introduced into elements (e.g.,
longer duration simulations, such as full engine blade loss events) including
this conduction may be required for accurate analysis.
16
5 Failure Surface Generation
To develop the failure surface, various material specimens, each with a
unique triaxiality and Lode parameter, are created and tested to failure. To
generate an accurate and complete failure surface, there should be as
many material specimens as possible covering a broad range of triaxiality
and Lode parameter. Using the specimen test results, a three‐dimensional
curve fitting tool can then be used to create a complete failure surface.
The fundamental strategy of the curve fitting tool is to use points
representing triaxiality, Lode parameter, and failure strain from the
specimens tested to develop a complete failure surface. MathWorks
MATLAB software was used in the development of the failure surface that
is presented in this example.
Suppose a material testing program provides twenty different material
tests with these states of stress:
Figure 4 Stress States for a Material Testing Program
17
Since it is challenging to design specimens representing a broad range of
triaxiality and Lode parameters, there are areas of the failure surface that
are not covered by physical tests and, as a result, a “grid and control point
strategy” was developed to define the entire failure surface. It should also
be noted that the complete failure surface does not have to be defined, as
long as the range of triaxiality and Lode parameter covers the range for the
application in which the material model is being developed.
An orthogonal grid structure can be overlaid onto the existing plot. While
these grid lines can be arbitrary, they should be based on existing physical
tests where possible. If there are several specimens that are close together,
one grid line can represent those points. Each test point or group of group
of test points creates a grid line. An example of a grid structure for the data
in Figure 4 is shown in Figure 5.
Figure 5 Grid and Control Point Example
Using these grid lines, a new set of points is defined based on the points of
grid line intersection. For each grid line intersection, a point on the
triaxiality‐Lode parameter plane is created. In this example, there are a
total of 66 “control points” and, as a result, three 66 point 1‐D vectors can
be created using: 1) the triaxiality of the control point, 2) the Lode
parameter of the control point, and 3) the failure strain of the control
18
point. The failure strain is determined manually by the user of the tool. At
control point locations where there is a single physical test, that specimen’s
failure strain defines the control point failure strain. If there are a group of
points that represent one control point, those points are averaged using
equal weights. If there is no physical test for a given control point, the
nearest point is used, or the average of many nearby points. Using this
method, all three of the 66 point 1‐D vectors are full of values.
After the input vectors are defined, the program can be executed. MATLAB
begins by implementing a subroutine that was developed using the Curve
Fitting Toolbox, which uses a cubic interpolation routine that takes the
three 66 point 1‐D vectors and generates a fit 3‐D surface. The user is then
shown two plots of the surface as those shown in Figure 6.
Figure 6 Surface Plots Generated by MATLAB Subroutine
The next step of the failure surface generation tool is to discretize the 3‐D
surface into individual points for input into LS‐DYNA®. The 3‐D surface is
sampled into 2814 total points (134 triaxiality points by 21 Lode angle
points). The tool will also show the user a visual representation of the
discretization of the 3‐D surface as shown in Figure 7.
19
Figure 7 Discretized 3D Failure Surface for LS‐DYNA®
The final step for the failure surface generation tool is to automatically
generate a keyword file to use in LS‐DYNA®. This file includes a
*DEFINE_TABLE which has 21 values from ‐1 to 1 that represent the line for
each Lode parameter shown in Figure 7. Also included is the 21
*DEFINE_CURVE keywords that represent the failure strain as a function of
the triaxiality. Each one of these curves has 134 total points. This keyword
file can then be directly included into the simulation model. This is very
valuable to the analyst because there are typically many iterations of failure
surface generation that must be completed to get a representative surface
for a given material. After each specimen is simulated, the displacement at
failure is checked against the experimental test result. If the simulation
reaches the failure criteria too late (or too early) in the simulation, the
control point(s) near that specimen’s state of stress can be adjusted and
the surface can be immediately regenerated.
It should be noted that these failure surface iterations are required because
the triaxiality and Lode parameter vary during the tests and, as a result,
their exact point location in Figure 4 is not definitive. In addition, the
20
relatively empty areas in Figure 4 where specimen tests have not been
defined are used in the simulations that are compared to dynamic impact
tests and, therefore, those points require adjustments to produce a good
match.
5.1 Temperature Scaling Function Creation
The temperature scaling factors for the failure surface are derived from a
series of tension tests at various temperatures for a constant strain rate. It
is assumed that temperature scaling is independent of the strain rate
because in the theoretical development the strain rate and temperature
scaling are multiplicative. As a result, the effect of a change in temperature
is the same at every strain rate.
The scaling function is created by separately running each temperature
yield curve without failure to determine the failure strain when the analysis
specimen reaches the displacement at which the test specimen failed. The
scaling factors are then calculated by normalizing the failure strain to the
room temperature (RT) failure strain as
Using the scaling factor and the single room temperature yield curve, the
simulations are then re‐run using the failure surface to make sure the
analysis specimens fail at the proper displacement with minor adjustments
made to the scaling factors as necessary. The material melting point
temperature should also be added at the end of the curve so that high
temperatures will not result in extremely high scaling factors due to
extrapolation.
5.2 Strain Rate Scaling Function Creation
If a series of tests was performed at various strain rates, a strain rate
scaling function for the failure surface can be developed. The analytical
specimens should be simulated using the displacement‐time curves from
21
the actual specimen tests with the exact specimen geometries. A scaling
curve for strain rate can be developed and adjusted until all the specimens
fail within the experimental spread of the test data.
5.3 Regularization Curve Creation
Failure strain as a function of element size is input into the material model
using a mesh size regularization scaling function that is input as a load curve
or table. The element size is calculated as the element volume divided by
the maximum side area. The regularization curve is generated by analytical
simulations of one of the test specimens using several different mesh sizes.
These simulations should be performed with the single strain rate yield
curve for the specific test specimen and the failure surface with no other
scaling factors.
To generate the regularization scaling curve, several mesh sizes for the test
specimen should be run initially with a regularization scale factor of 1.0. If
the analytical simulation fails before the displacement from the physical
test specimen is reached (at a smaller displacement), the scale factor is
increased; and, if the analytical simulations fails after the displacement
from the physical test specimen is reached (at a larger displacement), the
scale factor is reduced. Several iterations of the scale factor may be
required until the analytical simulation fails at the displacement from the
physical test. Additional data points should then be created so that there is
no LS‐DYNA® load function extrapolation for very small and very large mesh
sizes.
It should be noted that this procedure is based on a single test specimen
with a specific triaxiality. The regularization curve can also be input as a
table where a separate load curve giving the plastic failure strain versus
element size for each triaxiality is specified. Since there is much more
sensitivity to mesh size in tension failures versus shear failures, this option
can improve results crossing many failure modes.
22
6 Material Model Validation
To validate the effectiveness of the material model input parameters
including both the failure surface and scaling curves, dynamic impact tests
using a rigid projectile should be completed. Results of the dynamic impact
tests can then be compared directly to analytical simulations of those tests
to evaluate the material model robustness and serve as validation for the
material model inputs. Both dynamic punch tests and ballistic impact tests
are recommended so that a range of strain rates will be evaluated for
material model validation.
23
7 Test Program
The test program needed to develop the LS‐DYNA® inputs for *MAT_224
requires mechanical property tensile tests to generate temperature and
strain rate dependent yield curves, specimen failure tests for various
triaxiality and Lode parameters to generate the failure surface, and
dynamic impact tests to validate the model.
7.1 Stress‐Strain Behavior Tests
Tensile tests need to be performed at various strain rates with a constant
temperature and at various temperatures with a constant strain rate. The
range of strain rates and temperatures should encompass the areas of
interest for the specific impact applications being modeled. For the
aluminum 2024 model that is posted on the Aerospace Working Group
Webpage under the LS‐DYNA® Material Parameter Sets Menu the strain
rates tested included room temperature, 1E‐4 sec‐1, 1E‐2 sec‐1, 1.0 sec‐1,
500 sec‐1, and 1500 sec‐1 and the temperatures tested included ‐50 °C, 150
°C, 300 °C, and 450 °C. For the titanium Ti‐6Al‐4V that is in development,
the strain rates tested included 1E‐4 sec‐1, 1E‐2 sec‐1, 1.0 sec‐1, 500 sec‐1,
and 1000 sec‐1 and the temperatures tested included room temperature, ‐
50 °C, 200 °C, 400 °C, and 600 °C.
7.2 Specimen Failure Tests
A set of twenty specimens have been designed to generate an effective
failure surface. These specimens were designed to cover as broad a range
of triaxiality and Lode parameter as possible in order to develop the most
complete failure surface possible. Efforts are also continuing to design
specimens that will enable the full failure surface to be developed. The
specimen designs used in the aluminum 2024 tests are documented in
Appendix B of Reference 2 and the specimen designs used in the titanium
Ti‐6Al‐4V tests are documented in Section 5 of Reference 4. While it may
not be feasible to complete all twenty tests, it should be recognized that
24
the more failure tests that are completed, the more accurate and general
the failure surface will be.
7.3 Validation Tests
Dynamic impact tests such as dynamic punch tests and ballistic tests using
rigid projectiles on the target material are required for final validation of
the input parameters, failure surface, and scaling curves. These tests allow
for a direct comparison of measured test parameters with analysis results
to validate the effectiveness of the material model.
25
8 References
[1] Development of a New Material Model in LS‐DYNA®, Part 1: FAA, NASA
and Industry Collaboration Background. DOT/FAA/TC‐13/25, P1. William
Emmerling, Donald Altobelli, Kelly Carney, and Mike Pereira. April 2014.
[2] Development of a New Material Model in LS‐DYNA®, Part 2:
Development of a Tabulated Thermo‐Viscoplastic Material Model with
Regularized Failure for Dynamic Ductile Failure Prediction of Structures
Under Impact Loading. DOT/FAA/TC‐13/25, P2. Murat Buyuk. July 2014.
[3] Development of a New Material Model in LS‐DYNA®, Part 3: Plastic
Deformation and Ductile Fracture of 2024 Aluminum Under Various
Loading Conditions. DOT/FAA/TC‐13/25, P3. Jeremy Seidt. April 2014.
[4] Development of a Titanium Alloy Ti‐6Al‐4V Material Model Used in LS‐
DYNA®. DOT/FAA/AR‐15/11. Sean Haight. May 26, 2015.
[5] LS‐DYNA® Keyword User Manual Volume II ‐ Material Models. LS‐
DYNA® R8.0, March 23, 2015. Livermore Software Corporation.
[6] Carney K., Pereira M., Revilock D., and Matheny P., “Jet Engine Fan
Blade Containment using an Alternate Geometry and the Effect of Very
High Strain Rate Material Behavior,” International Journal of Impact
Engineering, 36 (2009) pp 720‐728.
26
Appendix A Stress‐Strain Relationship Modeling Procedure Flow
Chart
Isothermal Tensile Test
Data
‐50 °C
Room Temp
200 °C
400 °C
Force vs. Time
Displacement vs. Time
Quasi‐Static Strain Rate
Plastic Strain vs. Stress
Find the Proper Artificial Loading
Find Necking Point
Cut at Necking
Before Necking:
Keep the Curve
Simulate the Tensile Test
N Force vs. Displacement and
Strain Field in Simulation
Close to that of test?
Y
Cont’d on next page
27
Cont’d from previous page
The Set of Isothermal Input Curves
Input Displacement vs. Time as B.C. for all Simulations of Different Strain Rates
Find the Proper Artificial Loading
Find Necking Point
Cut at Necking
Before
Necking: Keep
the Curve
After Necking: Extrapolate a
Candidate Plastic Curve
Simulate the Tensile Test
N Force vs. Displacement and
Strain Field in Simulation
Close to that of test?
Y
Cont’d on next page
28
Cont’d from previous page
The Set of Curves at Different Strain
Create Strain Rate Sensitivity Curve
Using Compression and Tension Test
Create Synthetic Higher Rate Stress‐Strain
Curves by Translating a Lower Rate Curve
(Curves must be defined higher than the
highest rate seen in testing)
Assemble isothermal curves of
different temperatures
Modify Strain
Rate Sensitivity
Curves and Beta Asemble isorate curves of
different strain rates
Simulate and Compare the Force vs.
Displacement Curve between Test and
Simulation
N Does the Simulation Force and
Strain Rate vs. Displacement
Give Close Match to all Tests?
Y
Document the Final Set of Curves
Figure 8 Flow Chart of Stress‐Strain Relationship Modeling Procedure
29