Why Your Material Card Is Probably Wrong (And What to Do About It)
Every finite element simulation of a plastic component rests on one assumption so fundamental it is rarely examined: that the material card accurately represents the material’s behavior under the actual loading conditions. In most cases, it does not. This is not a software problem or a modeling problem. It is a data problem. And it starts earlier than most engineers realize.
The gap between a material and its description
1
Static and dynamic mechanical characterization
2
Modelling and parameter identification
3
Validation and implementiation to bigger models
Physical materials are complex. A thermoplastic under dynamic loading does not behave the way it does under quasi-static tension. Its stiffness increases. Its failure strain decreases. It yields differently in compression than in tension. Its response depends on loading history, stress state, temperature, and the speed at which force is applied.
Numerical simulation cannot reproduce this in full. It represents selected aspects of material behavior within a defined mathematical framework. That framework requires explicit constitutive relationships, a fixed number of parameters, and stable formulations, none of which can capture the full richness of what a physical material actually does.
This creates a structural gap: not a failure of the tools, but an inherent difference in what reality contains and what a simulation can represent. The engineering task is not to eliminate this gap, but to manage it consciously, to know what is being captured, what is being simplified, and what the implications are.
A simulation result is never an answer. It is an argument.
Questions about material characterization?
Our team is happy to help – reach out with your questions.
What standard data actually measures
Standard material data is frequently treated as objective truth. In practice, it is a narrow snapshot of behavior under specific conditions.
A typical tensile test measures material response at a single strain rate, at ambient temperature, in a single loading direction, without damage or failure characterization. For many materials, this snapshot is taken under conditions that have little to do with the actual service scenario.
For a plastic automotive component in a crash event, the strain rates involved are several orders of magnitude higher than in a quasi-static tensile test. Stiffness and strength both change. Failure can occur at strains far below the static value. A material card derived from that tensile test will systematically misrepresent the component’s behavior, not because the test was wrong, but because it measured the wrong thing.
The problem compounds for polymers specifically. Unlike metals, plastics exhibit tension-compression asymmetry: they are stiffer and stronger in compression than in tension. A tensile test, by definition, samples only one side of this asymmetry. A material card calibrated from tensile data alone will consistently underestimate component stiffness under bending-dominated loading, which is the dominant load case in the vast majority of plastic structural components.
This was precisely the finding that triggered 4a engineering’s original development work over 20 years ago. An automotive OEM found that material cards based on high-speed tensile testing systematically underpredicted component response under impact. The correction factor they arrived at empirically, approximately 1.25, depended on the specific polymer and had no physical basis. It was a symptom of a deeper issue: the test method did not sample the material behavior that mattered.
Why bending is the better starting point
Three-point bending places a specimen simultaneously into tension on one face and compression on the other. The outer fiber on the tension side and the outer fiber on the compression side are both loaded to high strains while the neutral axis carries minimal stress.
This makes bending sensitive to tension-compression asymmetry in a way that pure tension never can be. A material card calibrated against bending data inherently incorporates the behavior of both loading sides. When that card is used to simulate a component under bending-dominated loading, as most plastic structures are, the prediction transfers reliably.
Additional rate dependency information is equally accessible from bending tests. Our testing systems, LINOVIS® and IMPETUS®, are well suited to perform these tests, with IMPETUS® covering velocities from 0.5 to 4.4 m/s and LINOVIS® extending the range further into intermediate and lower velocities as well as higher forces. Across this combined velocity and force range, full dynamic force and displacement measurement as well as local strains are captured for impact and crash applications.
The result is a data set that contains what the simulation model actually needs, not what a standard procedure happens to produce.
Use Case:
Pedestrian Head Impact on a Bumper
A pedestrian head impact on a plastic bumper (≈40 km/h) is a bending-dominated load case. Both a tensile-based and a bending-based material card capture the first impact peak reasonably well, but only the bending-based card reproduces the second peak accurately, where the tensile-based card underpredicts the response.
If the component bends in the real application, start with bending data.
From test design to model selection
Basic Characterization
Static and dynamic 3-point bending tests form the foundation for material characterization, enhanced by additional tensile tests. These results are used to capture the deformation behavior of the material across varying strain rates.
Advanced Characterization
Additional static and dynamic tests (e.g. puncture tests, tensile tests with local strain evaluation, etc.) with variation of the stress state in order to identify stress state dependent yield and hardening, as well as damage and failure properties.
Validation
Simulation of all performed tests with the identified set of parameters in order to prove the correlation between test and simulation results with a given discretization on coupon level. Validation of the material card based on a small component test.
Database
All data is consolidated into a single database, organized into general information, test setup, sample data evaluation settings and simulation models and results.
Even with the right test data, a critical engineering decision must be made before any parameter is identified: which material model to use.
This is not a fitting decision. It is a conceptual commitment. The model determines which physical mechanisms are represented and which are simplified away. Choosing the model after fitting data to it, rather than before, leads to the most common failure mode in material card generation: excellent local agreement with the calibration data, and unreliable predictions everywhere else.
A model calibrated against bending data using a von Mises formulation, for example, will accurately reproduce bending behavior, but may systematically over-predict stiffness in puncture tests where the stress state is biaxial and volume dilation begins to play a role. The deviation is not a calibration error. It reflects a fundamental limitation of the chosen formulation.
A model is not chosen because it fits the data. Data is interpreted because a model was chosen.
The most physically complete models for unreinforced thermoplastics, such as the SAMP-1 formulation (MAT_187 in LS-DYNA), support user-defined yield surfaces, viscoelasticity as a function of strain rate, viscoplasticity, volume dilation, and tension-compression asymmetry in a unified formulation. The computationally efficient variant, MAT_187L, retains most of these features with only approximately 20% higher CPU cost than the industry-standard MAT_24, while providing substantially better predictive accuracy across combined tension, bending, and puncture load cases.
The appropriate model is not always the most complex one. Where MAT_24 captures the dominant behavior within acceptable accuracy, additional complexity introduces cost without benefit. The selection criterion is minimum complexity sufficient for the application, not maximum physical detail.
Table: Comparison of commonly used LS-DYNA material models for unreinforced plastics. The material model defines which physical mechanisms can be represented by the material card.
| Material model | Yield surface | Viscoelasticity | Viscoplasticity | Compression / tension asymmetry | Plastic Poisson’s ratio |
|---|---|---|---|---|---|
| *MAT_024 | von Mises | – | ✓ | – | 0.5 |
| Fast baseline model for standard plastic simulations. | |||||
| *MAT_124 | 2× von Mises | ✓ Prony series | ✓ | ✓ | 0.5 |
| Intermediate model with tension/compression asymmetry. | |||||
| *MAT_187 SAMP-1 |
Linear, parabolic or piecewise linear | ✓ E(ε̇) | ✓ | ✓ | ✓ νp(ε) |
| Physically complete model for demanding plastic applications. | |||||
| *MAT_187L SAMP-1 Light |
Linear | ✓ E(ε̇) | ✓ | ✓ | ✓ νp(ε) |
| Efficient SAMP-1 variant for broad production use. | |||||
The appropriate model is the minimum level of complexity that still represents the relevant physics of the application.
Failure is not a special case
One of the most persistent gaps between material card capability and simulation reality is failure prediction. Deformation behavior is routinely calibrated. Failure is frequently an afterthought, assigned as a single equivalent plastic strain value or omitted entirely.
The practical consequence appears in front-grille impact simulations, pedestrian safety assessments, and drop tests: the simulation tracks the force-displacement response reasonably until fracture begins, and then diverges completely. With a properly constructed failure model, the same simulation can capture the onset and progression of damage closely.
Failure characterization for thermoplastics is simpler than it appears, because two engineering observations allow significant simplification. Compression failure is not observed in ductile thermoplastics (buckling precedes material failure). And shear failure in these materials is tension-dominated rather than pure-shear-dominated, making the shear failure strain approximately equal to the uniaxial tensile value. These two assumptions reduce a complex multi-axial failure characterization to a tractable test program: bending, tensile, and puncture tests cover the relevant triaxiality range from 0.33 to 0.66. These three test setups provide a solid baseline for a well-conditioned failure model, with additional testing available to refine it further where needed.
What the framework looks like in practice
The path from a raw material sample to a validated, solver-ready material card follows three stages that must remain technically aligned: purposeful testing designed around the simulation requirements, structured parameter identification that evaluates behavioral consistency across multiple load cases rather than curve-fitting to individual tests, and validation against independent data to confirm where the model is reliable and where its limits are.
Each stage informs the others. Validation results drive adjustments in test programs. Test data quality constrains which model complexity is justified. Model selection determines which tests are relevant.
VALIMAT®, the software environment that ties these stages together, implements this logic through automated workflows: automated FEM model generation based on actual test data and geometry, successive stepwise identification of material parameters, and a documented link from every parameter in the final card back to the test data it was derived from. That traceability is not an audit formality. It is what makes a material card reusable, revisable, and defensible.
The outcome of a properly executed material card generation program is not a file. It is a documented engineering artifact with known applicability limits, traceable to its physical origins, ready for validation, and trustworthy within its defined range.
You can stay up to date by following us on social media: