4. Mesh quality and its impact on PAM-CRASH crash simulation

In crash simulation, the finite element mesh is the medium through which the solver perceives geometry, stiffness, and contact. A model can have perfect material cards and boundary conditions, yet still produce non-physical results if the mesh is distorted, warped, or too coarse in critical regions. PAM-CRASH, like all explicit solvers, is particularly sensitive to element quality because the stable time step is governed by the smallest element characteristic length, and because shell elements dominate automotive crash models. This section examines the four quality metrics that matter most — element distortion, aspect ratio, warpage, and mesh convergence — and shows how to measure, judge, and correct them in practice.

Why Explicit Solvers Punish Poor Mesh Quality

PAM-CRASH advances the solution with a central-difference scheme whose stability limit is approximately:

dt_crit = L_c / c
c = sqrt(E / rho)          (bar wave speed)
L_c = characteristic element length

For a shell element, L_c is derived from the element area and its longest edge. A single badly distorted element with a very short edge can reduce the global time step for the entire model, increasing wall-clock time dramatically. Worse, distorted elements produce inaccurate strain fields, hourglassing, and premature element deletion, which corrupt the energy balance that engineers use to judge result validity.

Element Distortion

Distortion describes how far a quadrilateral deviates from a rectangle. The most common metrics are the Jacobian ratio and the minimum interior angle. In PAM-CRASH preprocessing (typically via ANSA, HyperMesh, or the built-in PAM-CRASH preprocessor), a Jacobian ratio below 0.6 or an interior angle below 30° signals a problem element.

Distortion arises from three main sources: mapping a coarse mesh onto curved geometry, large plastic deformation during the crash itself, and manual node editing during model assembly. The first two are unavoidable to some degree; the third is entirely controllable. A practical rule is to keep the initial Jacobian ratio above 0.7 in all regions that will experience large deformation, such as the front rail, bumper, and B-pillar.

Aspect Ratio

Aspect ratio is the ratio of the longest edge to the shortest edge of a shell element. A perfect square has an aspect ratio of 1.0. In crash models, aspect ratios up to 3:1 are generally acceptable in non-critical areas, but should stay below 2:1 in energy-absorbing structures.

High aspect ratios cause two problems. First, they bias the bending stiffness: a long, thin element resists bending differently along its two axes, introducing artificial anisotropy. Second, they reduce the accuracy of the contact algorithm, because the element's normal and its penetration depth are computed from a single integration point. In a frontal impact model, a row of 5:1 elements along the front rail can delay the predicted folding pattern by several milliseconds — enough to change the occupant injury criteria.

Warpage

Warpage measures how far a quadrilateral deviates from planarity. It is typically expressed as the angle between the two diagonals of the element, or as the out-of-plane distance of one node relative to the plane of the other three. A warpage angle above 15° is generally flagged; above 20° the element's bending behavior becomes unreliable.

Warpage is especially common when meshing curved panels such as door skins, roof rails, and wheel arches. The standard remedy is to split the warped quadrilateral into two triangles, or to re-mesh the region with a finer, better-aligned pattern. In PAM-CRASH, warped shell elements can also trigger spurious hourglass modes because the element's local coordinate system becomes ill-defined.

Mesh Convergence Strategy

Mesh convergence answers a simple question: is the mesh fine enough that further refinement no longer changes the result? In crash simulation, the answer is rarely "refine everything" — that would make the model computationally infeasible. Instead, convergence is applied selectively to the structures that dominate the response.

A practical convergence study for a frontal impact model proceeds as follows. First, identify the critical energy-absorbing components: front rail, crash box, bumper beam, and shotgun. Second, establish a baseline model with a nominal element size of 8 mm in these regions. Third, create two refined variants at 5 mm and 3 mm, keeping all other parameters identical. Fourth, run all three models and extract the peak force, intrusion, and internal energy of the critical components.

The results are then compared. If the peak force changes by less than 5% between the 5 mm and 3 mm models, the 5 mm mesh is considered converged for that component. If the change exceeds 5%, further refinement or a different element formulation is required. In practice, most automotive crash models converge at 5–6 mm in the front structure and 8–10 mm in the rear and floor, provided the element quality metrics are satisfied.

Practical Quality Thresholds and Workflow

The following table summarizes the quality thresholds commonly applied in PAM-CRASH crash models. These are starting points; the acceptable values depend on the component's role and the maturity of the program.

In the PAM-CRASH workflow, quality checks are typically run in the preprocessor before the solver is invoked. A useful practice is to export a quality report and compare it against the thresholds above, then iterate on the mesh until all critical regions pass. After the simulation, the .out and .er files should be inspected for element inversion warnings and negative Jacobian messages. If these appear in a region that was flagged as marginal during preprocessing, the mesh — not the material model — is the likely culprit.

Balancing Quality Against Cost

Every refinement reduces the stable time step and increases the number of elements, both of which raise computational cost. A model with a 5 mm front rail and a 10 mm floor may run in 6 hours on 32 cores; refining the floor to 5 mm could double the element count and halve the time step, pushing the run to 20 hours or more. The convergence study tells you where refinement actually changes the answer, allowing you to spend the budget only where it matters.

A pragmatic strategy is to maintain a "quality map" of the vehicle: critical regions at 5 mm with strict quality thresholds, secondary regions at 8 mm with relaxed thresholds, and non-structural regions at 12–15 mm with minimal quality requirements. This map should be reviewed at each program milestone, because the critical regions change as the design evolves — a component that was non-structural in the concept phase may become a load path in the final design.

Ultimately, mesh quality is not a preprocessing checkbox. It is a modeling decision that directly controls the accuracy, stability, and cost of every PAM-CRASH crash simulation. Engineers who understand distortion, aspect ratio, warpage, and convergence — and who can quantify their effect on the results — are far better equipped to defend their predictions and to diagnose the non-physical results that inevitably appear in complex crash models.