Why Default Software Material Cards Fail for Hyperelastic Rubber Simulations

Modeling Rubber and Viscoelasticity with Abaqus

Why Default Software Material Cards Fail for Hyperelastic Rubber Simulations

Every FEA engineer working with elastomers has been tempted by it at some point: you are setting up a complex assembly in Abaqus, ANSYS, or LS-DYNA, you need to model an O-ring or a vibration isolator, and there it is a built-in material library offering a pre-calibrated material card labeled simply: “Rubber (Mooney-Rivlin)” or “Neoprene.”

You click assign, hit solve, and wait.

What happens next usually falls into one of two categories:

  1. The solver fails to converge, throwing negative eigenvalues, distorted element errors, or severe volumetric locking at 15% strain.
  2. The solver converges, but the physical prototype deflects twice as much as predicted, seals leak, or the part fails in endurance testing.

Why do default material cards and generic library constants fail so spectacularly when it comes to hyperelastic materials?

The short answer: Rubber is not steel, and a generic strain energy density function is an engineering illusion.

Here is a deep dive into the physics and material realities behind why default rubber cards fail, and what you need to do instead.

1. “Rubber” Is Not a Material Specification

When you select generic Structural Steel (e.g., ASTM A36) or Aluminum 6061-T6 from a software library, the modulus and yield strength are reasonably predictable across suppliers.

Elastomers, however, are complex chemical formulations. What an engineer colloquially calls “EPDM 70 Shore A” or “NBR” is not a singular material. Its mechanical response is dictated by:

  • Polymer grade and molecular weight distribution
  • Filler loading (carbon black or silica type, particle size, and volume fraction)
  • Plasticizers and processing oils
  • Cure system (sulfur vs. peroxide vulcanization)
  • Crosslink density

Two compounds moulded to the exact same nominal durometer (e.g., 70 Shore A) can exhibit up to a 50% difference in stress-strain behaviour at moderate to large strains. Using a generic library card means you are simulating a hypothetical, chemically ambiguous rubber that likely bears zero resemblance to the compounded stock on your factory floor.

2. The Uniaxial Trap and Multiaxial Blindness

Most built-in material constants (or default fitting wizards) are calibrated exclusively against simple uniaxial tensile test data.

A hyperelastic model calibrated only on uniaxial tension is mathematically blind to other modes of deformation.

If you fit a 2-term Mooney-Rivlin or a 3-term Ogden model strictly to uniaxial data, the model might look like a perfect fit on your 2D curve. But when the FEA solver applies a triaxial or shear stress state to an element, the mathematical surface outside the uniaxial domain frequently produces unphysical behavior, including negative stiffness or severe over-stiffening.

3. The Math Behind the Crash: Drucker Stability Violations

Commercial FEA packages evaluate material stability using Drucker’s stability criterion, which requires that the work done by an incremental stress over an incremental strain must be positive. Default material cards especially higher-order polynomial models or unconstrained parameters often violate Drucker stability at strain levels just beyond the calibrated data.

When your simulation hits a strain state where the material derivative turns negative:

  1. The tangent stiffness matrix loses positive definiteness.
  2. The solver encounters material instability.
  3. The simulation diverges with a non-convergence error or localized mesh distortion.

Default cards like in Abaqus rarely warn you about the exact strain limits where their underlying coefficients become unstable.

4. Incompressibility and the Volumetric Penalty Faux Pas

Elastomers are nearly incompressible. Their bulk modulus (K) is typically 1,000 to 10,000 times larger than their shear modulus (G). Without measuring the actual bulk modulus based on proper confined volumetric testing, your contact force and groove-fill predictions are essentially guesswork.

5. Ignoring Real-World Physical Phenomena

Default hyperelastic material cards assume a conservative, frictionless, rate-independent energy potential. Real rubber, however, violates almost every assumption of pure hyperelasticity:

  1. The Mullins Effect (Stress Softening): Virgin rubber behaves stiffly on its first pull. On subsequent cycles, it softens dramatically. If your default card represents virgin data, but your actual component undergoes cyclic loading, your simulation will drastically overestimate operating stiffness.
  2. Viscoelasticity & Strain Rate Sensitivity: Default cards represent a single, unspecified strain rate. In dynamic applications (e.g., engine mounts, bump stops), rubber stiffens noticeably as frequency/strain rate increases.
  3. Permanent Set & Creep: Hyperelastic cards snap back to zero strain upon unloading with zero hysteresis loss. Real elastomeric parts exhibit residual strain and relaxation.

Summary: What Default Cards Miss

ParameterDefault Library CardReality in Physical Elastomers
Material ChemistryGeneric nominal valuesFormulations, fillers, and cure dependent
Deformation StatesTypically calibrated for 1 modeCoupled multiaxial modes (Tension, Shear, Compression, Biaxial)
Stability LimitsOften untested; prone to Drucker instabilityPhysically stable across realistic deformation ranges
Bulk ModulusArbitrary default assumptionCrucial for confined seals and contact pressure
ViscoelasticityAssumes purely elastic recoveryViscoelastic force decay and stress relaxation

The Engineer’s Checklist: Moving Beyond the Default Card

If default material cards are fundamentally unreliable, how should you simulate hyperelastic rubber accurately? Follow this workflow:

  1. Test Your Specific Compound:
    Never rely on a generic handbook card for production-intent designs. Obtain physical samples vulcanized from the exact compound and process used in production.
  2. Characterize all Deformation Modes:
    • Uniaxial Tension
    • Equibiaxial Tension
    • Unaxial Compression
    • Planar Shear
    • Volumetric Compression
  3. Select the Right Potential for Your Strain Range:
    • Strains: Less than 30-50%: Neo-Hookean or Mooney-Rivlin
    • Strains up to 150-200% Yeoh (reduced polynomial) or Arruda-Boyce
    • Strains Greater than 300%, High Non-linearity: Ogden (3-term or higher)
  4. Run Single Element Checks:
    Before running your component model, run your fitted coefficients through the software’s built-in stability evaluator across all deformation modes up to the maximum expected strain.
  5. Use Hybrid / Mixed-Formulation Elements:
    Always pair your hyperelastic material definition with elements designed for incompressibility (e.g., Abaqus C3D8H, ANSYS KEYOPT(6)=1, or similar mixed U-P formulations) to avoid volumetric locking.

Summary

Default software material cards are useful for one thing only: sanity-checking that your boundary conditions, contacts, and solver setup execute without syntax errors.

The moment your simulation is expected to guide tooling, predict seal integrity, or assess fatigue life and durability, step away from the built-in library. In hyperelastic FEA, garbage in does not just mean garbage out, it usually means the solver won’t even run.