2. LS-DYNA material failure modeling landscape

Before calibrating any failure parameter, you must decide which failure formulation in LS-DYNA actually represents the physical mechanism you are trying to capture. LS-DYNA offers a layered set of options: stress- or strain-based failure criteria, phenomenological damage models, continuum damage mechanics (CDM), and element-deletion schemes. Choosing the wrong layer is the most common root cause of a calibration that "converges numerically" but predicts the wrong crash response.

Three Modeling Layers

It helps to separate the LS-DYNA failure landscape into three functional layers that are often combined in a single material card:

• Layer 1 — Failure criteria (onset triggers). Scalar conditions evaluated from stress, strain, or a combination. Examples: *MAT_ADD_EROSION with MXEPS, MAXPS, EPSSH, or a GISSMO-style equivalent plastic strain threshold. These do not degrade stiffness; they only flag or delete.
• Layer 2 — Damage evolution models. Variables that grow from 0 to 1 and reduce the effective stress or stiffness. Examples: *MAT_DAMAGE_1, *MAT_DAMAGE_2, GISSMO (*MAT_ADD_DAMAGE_GISSMO), and DIEM (*MAT_ADD_DAMAGE_DIEM).
• Layer 3 — Element deletion / regularization. Controls how failed integration points or elements are removed, and how mesh-size dependence is handled. Examples: EROSION in *CONTROL_ cards, EFG, and the regularization options inside GISSMO/DIEM.

The critical engineering point: a criterion alone (Layer 1) gives you a brittle on/off switch; a damage model (Layer 2) gives you a gradual loss of load-carrying capacity that better matches ductile metals. For automotive crash, Layer 2 is almost always required for aluminum and advanced high-strength steel (AHSS).

Applicability to Automotive Materials

Automotive materials fall into distinct failure regimes, and each regime maps to a preferred LS-DYNA formulation:

Notice that GISSMO appears repeatedly. This is not accidental: GISSMO (Generalized Incremental Stress State dependent damage MOdel) is the de-facto standard for ductile metals in automotive crash because it decouples the onset of damage from the fracture point and accounts for stress-state history. DIEM (Damage Initiation and Evolution Model) is a newer alternative that adds Lode-parameter dependence and is preferred when shear-dominated loading is significant.

How the Layers Interact — A Signal-Flow View

The following diagram shows how a stress state flows through the LS-DYNA failure stack for a ductile metal, from the constitutive model through damage accumulation to element deletion. It is the mental model you should carry into every calibration task.

Choosing Between GISSMO and DIEM

Both models accumulate damage incrementally, but they differ in how the fracture locus is defined and how regularization is applied:

• GISSMO uses a fracture strain curve as a function of triaxiality η, with an optional Lode parameter dependence. It is mature, widely validated, and supported by most automotive OEM material databases. The damage exponent DMGEXP controls the nonlinearity of damage accumulation.
• DIEM uses a fracture surface defined by both η and the Lode parameter ξ, making it more accurate for shear-dominated and plane-strain states. It is newer and requires more calibration data (at least tension, shear, and plane-strain coupon tests).

For a first calibration on DP780, GISSMO is the pragmatic choice. For a shear-critical aluminum extrusion where the Lode parameter varies significantly through the thickness, DIEM is worth the extra effort.

Practical Calibration Workflow

The following sequence reflects how an engineer actually calibrates a ductile failure model in LS-DYNA. It is not a rigid template — the number of iterations depends on the material and the target accuracy.

1. Characterize the stress state. Run coupon simulations (tension, shear, plane-strain, biaxial) with the calibrated plasticity model and extract η and ξ histories at the critical element. These are your calibration inputs, not the nominal test values.
2. Fit the fracture locus. Plot experimental fracture strains against the simulated η (and ξ for DIEM). Fit a curve — often a power law or a piecewise linear table. For GISSMO, this becomes the LC table in *MAT_ADD_DAMAGE_GISSMO.
3. Set damage evolution parameters. Choose DMGEXP (typically 1–4) and the critical damage DCRIT (typically 0.9–1.0). A lower DMGEXP makes damage accumulate more linearly; a higher value delays damage until late in the loading.
4. Apply regularization. Set the characteristic length L and the regularization exponent FADEXP. For shell elements, L is typically the square root of the element area. This step is what makes the result mesh-insensitive.
5. Validate against a component test. Run a hat-section or axial crush simulation and compare the force-displacement curve and the fracture location against the physical test. Adjust the fracture locus or DMGEXP if the crack initiates too early or too late.

Key Parameters at a Glance

The landscape is broad, but the decision path is narrow: identify the dominant failure mode, select the matching model layer, calibrate the plasticity first, then fit the fracture locus, and finally validate on a component. The next chapter moves from this landscape overview into the specific coupon tests and data reduction methods that feed the calibration.