4. Check Element Quality Against a Custom C

Why a Custom Criterion Set Matters

HyperMesh ships with a default qualityindex template that reflects a generic automotive/sheet-metal bias: warpage ≤ 5°, aspect ratio ≤ 5, Jacobian ≥ 0.6, and so on. Those numbers are a starting point, not a specification. When you are checking a thin-walled injection-molded bracket, a cast aluminum housing, or a composite panel, the acceptance thresholds come from the solver deck, the material model, and the manufacturing process — not from the template. The task in this case is to replace the defaults with a project-specific criterion set, run the quality index, and read the resulting flagged-element distribution correctly.

The three criteria chosen here are deliberately complementary: warpage is a shell-only, out-of-plane distortion metric; aspect ratio is a shape metric that affects stiffness and wave-speed accuracy in both shells and solids; Jacobian is a mapping-quality metric that detects elements whose isoparametric transformation is close to singular. Together they cover geometry distortion, shape distortion, and mapping distortion — three failure modes that a single metric cannot catch.

Step 1 — Define the Criterion Set Before Touching the Model

Open the quality criteria panel (Quality → Criteria in the HyperMesh Quality ribbon, or the *criteria command in a Tcl script). The panel exposes one row per criterion with an enable flag, a target value, and a failure value. HyperMesh classifies each element into three bands: pass (better than target), warn (between target and failure), and fail (worse than failure). Only the fail band drives the quality index percentage; the warn band is advisory.

For this example, assume a stamped steel closure panel with a 1.2 mm gauge, checked against a solver deck that requires:

Enter these values in the criteria panel, enable only the three rows, and leave the remaining criteria (skew, taper, min/max angle) disabled so the index isolates the three metrics under investigation. Disabling unused criteria is not cosmetic: the quality index is a weighted composite, and leaving defaults active will mask the distribution you are trying to read.

Step 2 — Run the Quality Index and Read the Distribution

Launch the check with Quality → Quality Index, select the component set to be evaluated (all shell components in this case), and apply the saved criteria file. HyperMesh returns a single composite index plus a per-criterion breakdown. The composite number alone is not actionable; the breakdown and the spatial distribution are.

Suppose the run reports a composite index of 87.4, with the per-criterion fail counts shown below. These are illustrative values for a deliberately imperfect first-pass mesh, not measured results:

• Warpage: 214 elements fail (> 8°), concentrated along the curved rear-quarter transition.
• Aspect ratio: 61 elements fail (> 6.0), clustered at the flange run-out where the mesh was projected onto a narrow strip.
• Jacobian: 9 elements fail (< 0.50), isolated at two corner fillets.

The interpretation rule is straightforward: a high fail count in one criterion with a tight spatial cluster points to a meshing decision, not a random defect. Warpage failures along a curved transition usually mean the quads were generated on a doubly-curved surface without a projection correction; aspect-ratio failures at a flange mean the seeding density was too coarse in the narrow direction; a handful of Jacobian failures at fillets mean the corner elements were collapsed or over-skewed during the transition.

Step 3 — Verify the Fix Without Moving the Goalposts

After correcting the warpage cluster (re-projecting the quads and splitting the worst offenders), re-run the index with the same criteria file. The verification criterion is not "the composite index went up" — it is that the specific fail counts you targeted dropped while the other criteria did not regress. A useful check is to compare the per-criterion fail counts before and after:

# Tcl snippet: export per-criterion fail counts for comparison
# (run after each quality index pass)
set critList {warpage aspect jacobian}
foreach c $critList {
    set n [hm_getqualityindexfailcount $c]
    puts "criterion=$c fail_count=$n"
}

If warpage fails drop from 214 to a handful but aspect-ratio fails jump from 61 to 90, the re-projection pushed elements into a narrower strip — the fix traded one distortion for another. That is the signal to revisit the seeding, not to relax the aspect-ratio threshold. Relaxing a threshold to make the index look better is the single most common misuse of this workflow; the criterion set is a contract with the solver, and it should only change when the solver deck or the manufacturing process changes.

Deliverable — Criterion Set Record

Keep a short record with each mesh release so the index is auditable. The fields below are the minimum needed to reproduce a check:

The key discipline is that the criterion set is defined first, from the solver and process requirements, and the index is read as a distribution — not a single score. Once you can point to a cluster on the model and name the meshing decision that produced it, the quality index has done its job.