2. Mesh quality diagnostics for crash models

Crash models rarely fail because the physics is wrong. They fail because a handful of degenerate shell elements collapse the stable time step, trigger negative volume errors, or produce hourglassing that contaminates the whole energy balance. Mesh quality diagnostics is therefore not a cosmetic pre-processing step — it is a numerical stability control. This section covers the five metrics that matter most for LS-DYNA explicit crash models (aspect ratio, warpage, Jacobian, skew, and their combined effect on the time step), how to inspect them in LS-PrePost, and how to translate the findings into *CONTROL_TIMESTEP settings.

Why Explicit Solvers Are Sensitive to Mesh Quality

In an explicit integration, the stable time step for a shell element is governed by the element characteristic length and the wave speed:

dt_stable = L_c / c
c = sqrt(E / (rho * (1 - nu^2)))   (shell, plane stress)

where L_c is the characteristic length, E the Young's modulus, rho the density, and nu the Poisson ratio. LS-DYNA computes L_c from the element geometry, so a distorted element with a small minimum dimension produces a small L_c and a small dt. Because the global time step is the minimum over all elements, a single bad element can force the entire model to run at a fraction of its intended step size. The five diagnostics below are the practical proxies for "how small will L_c get, and how badly will the element behave before it does."

The Five Core Metrics

These are not independent. A highly skewed quad usually also has a poor Jacobian; a warped quad often has a large aspect ratio. The diagnostic workflow is to find the worst offenders on each metric, then check whether they are the same elements.

Inspecting Mesh Quality in LS-PrePost

LS-PrePost provides a dedicated mesh-quality panel. The practical sequence is:

1. Load the keyword deck (File → Open → Keyword) and select the part or the full model.
2. Open Meshing → Check → Element Quality (or the Check tab in the mesh toolbar).
3. Enable the metrics you want to visualize: Aspect Ratio, Warpage, Jacobian, Skew, Min Angle, Max Angle.
4. Set a threshold per metric (for example, aspect ratio > 5, Jacobian < 0.6, warpage > 10°). LS-PrePost highlights only the failing elements.
5. Use Save Failed to write the failing element set to a separate part or a *SET_SHELL, so you can inspect it in isolation.

A useful habit is to run the check on the undeformed mesh first, then again on a mid-crash state. Elements that pass at t = 0 can fail after large deformation, and those are the ones that cause late-stage negative volume errors.

Connecting Diagnostics to *CONTROL_TIMESTEP

The *CONTROL_TIMESTEP card controls how LS-DYNA handles the time step. The key fields are:

*CONTROL_TIMESTEP
$#  DTINIT   TSSFAC    ISDO     TSLIMT     DT2MS      LCTM     ERODE     MS1ST
       0.0     0.9        0       0.0        0.0         0         0         0

• TSSFAC — time-step scale factor (default 0.9). The actual step is TSSFAC × dt_stable. Lowering it to 0.6–0.8 adds robustness at the cost of runtime.
• DTINIT — initial time step. Leave at 0.0 to let LS-DYNA compute it from the mesh.
• TSLIMT — minimum time step. Elements below this are either deleted or mass-scaled, depending on ISDO.
• DT2MS — mass scaling time step. If positive, mass is added to small elements so they meet this step; if negative, mass is added only to elements below |DT2MS|.
• ISDO — controls whether the minimum step is based on the initial mesh or updated during the run.

The diagnostic loop is: run the mesh-quality check, identify the worst elements, and decide whether to fix the mesh or compensate with DT2MS. Fixing the mesh is always preferable; mass scaling is a mitigation, not a cure.

Worked Example: A B-pillar Model

Consider a simplified B-pillar shell model with roughly 120,000 elements. After loading the deck in LS-PrePost and running the element-quality check with thresholds of aspect ratio > 5, Jacobian < 0.6, and warpage > 10°, the following representative results are observed (these are illustrative values for teaching, not measured data):

The 52 low-Jacobian elements are the critical ones. If the model's nominal stable step is 0.8 µs, these elements may reduce it to 0.15 µs, increasing runtime by more than 5×. Two options:

1. Remesh the corner fillets to raise the Jacobian above 0.6. This is the correct fix and removes the root cause.
2. Apply selective mass scaling with DT2MS = -3e-7 (i.e., 0.3 µs) so only elements below 0.3 µs receive added mass. This keeps the global step near 0.3 µs but adds non-physical inertia to the fillets, which can affect local acceleration results.

Diagnostic Workflow at a Glance

The diagram below summarizes how the five metrics feed into the time-step decision and the two remediation paths.

Practical Checklist

• Run the LS-PrePost element-quality check on the undeformed mesh and on a mid-crash state.
• Record the worst aspect ratio, Jacobian, warpage, and skew, and the element count failing each threshold.
• Identify whether the worst elements on different metrics are the same elements — if so, remesh that region first.
• Estimate the nominal stable step from L_c and wave speed, then compare with the actual dt reported in d3hsp.
• If mass scaling is used, set DT2MS to the smallest value that brings the global step to an acceptable level, and verify added mass < 1%.
• Keep TSSFAC at 0.9 unless the model shows instability; lowering it to 0.6–0.8 is a last resort.

Mesh quality diagnostics is a feedback loop, not a one-time gate. The metrics tell you where the model is numerically fragile; *CONTROL_TIMESTEP tells you how much you are willing to pay in runtime or added mass to tolerate that fragility. The engineering judgment is in deciding when to fix the mesh and when to compensate — and in always verifying that the compensation did not change the answer.