📚 Course ContentsChapter 1 Diagnose Free Edges in a Shell Mesh

1. Diagnose Free Edges in a Shell Mesh

Why free edges matter in a shell mesh

In a shell mesh, every element edge is shared by exactly two elements when the mesh is conformal and closed. A free edge is an edge that belongs to only one element. HyperMesh reports these through the Edges panel (or the free edges display mode) by comparing element connectivity: an edge whose two node pair is not matched by a second element is flagged as free. Free edges are not automatically errors — a legitimate boundary (an open shell, a trimmed panel, a symmetry cut) will always produce free edges. The diagnostic task is to separate intended boundaries from unintended discontinuities such as gaps, missing elements, or mismatched node numbering along a seam.

Reading the edge state before you touch anything

Load the model and switch the display to edges. In HyperMesh the quickest route is the Edges panel: set the edges to find selector to free edges, pick the component or the whole model, and click find. The result is written to a temporary set you can isolate. Two numbers matter:

Count — how many free edges exist. A closed solid shell should report zero; an open panel reports a number consistent with its perimeter.
Location — where they sit. Free edges clustered along a straight seam between two components usually mean a connectivity break, not a real boundary.

Before trusting the count, confirm the model is in the state you expect: no hidden elements, no suppressed components, and the correct element type (shell vs. solid). A common false alarm is running edge detection on a model where one component is masked — the masked elements are excluded from the search and their shared edges appear free.

Run edge detection twice: once on the full model, once per component. Edges that are free in the full model but shared inside a component are almost always inter-component seams; edges free in both are genuine boundaries or real gaps.

Classifying each free edge

Not every free edge needs fixing. Work through the reported set and assign each edge to one of three classes:

Class	Typical signature	Action
Legitimate boundary	Runs along the outer perimeter of an open shell, a symmetry plane, or a trimmed edge; geometry edge coincides with the mesh edge	Leave as-is; record it as an expected boundary
Unintended gap	Two components meet but their edges are offset by a small distance; free edges run parallel on both sides	Equivalence or stitch the nodes; re-run detection
Missing / mismatched element	Free edge terminates mid-surface with no opposing edge; a hole in the mesh	Fill the hole or correct the element; re-run detection

The decisive check is geometric: does a free edge lie on a real geometric boundary? Use the geometry edges as the reference. If a free edge follows a surface edge in the CAD, it is a boundary. If it cuts across a continuous surface, it is a defect.

Quantifying the discontinuity

For suspected gaps, measure rather than guess. Two practical measurements:

Node-to-node distance across the seam. Pick a node on each side and read the coordinates; a gap of one element length or less that repeats along the seam indicates unmerged coincident nodes.
Edge count ratio. If a seam should be fully shared, the number of free edges on each side should match. A mismatch means elements are missing on one side.

Record the free-edge count before and after any repair. The delta is your verification metric: a successful equivalence should reduce the count by the number of merged edge pairs, and the remaining free edges should map one-to-one onto legitimate boundaries.

Worked example: a two-panel assembly

Consider a simplified model of two shell panels that should share a common seam. The left panel is a 4×2 grid of quad elements, the right panel a 4×2 grid, and the two are intended to be welded along a vertical line. After import, edge detection reports 48 free edges. The perimeter of the combined outline accounts for 28 of them (top, bottom, and the two outer vertical sides). The remaining 20 sit along the central seam — 10 on the left panel's right edge and 10 on the right panel's left edge, offset by a small distance.

Classification: the 28 perimeter edges are legitimate boundaries. The 20 seam edges are an unintended gap. The offset is consistent with unmerged coincident nodes, so the repair is an equivalence operation on the seam nodes with a tolerance slightly larger than the measured offset. After equivalence, re-running edge detection should report 28 free edges — the perimeter only. If it reports more, some seam nodes were outside the tolerance and remain unmerged; if it reports fewer than 28, an intended boundary was accidentally collapsed and must be restored.

Do not raise the equivalence tolerance blindly to "make the count drop." A tolerance larger than the smallest intended gap will merge legitimate boundaries and destroy the model's design intent. Measure the actual offset first, then set the tolerance just above it.

Checklist for a clean diagnosis

Confirm no components are masked or elements hidden before running edge detection.
Run detection on the full model and per component; compare the two sets.
Classify every free edge as boundary, gap, or missing element using the CAD surface edges as reference.
Measure the offset for suspected gaps; do not guess the tolerance.
Record the free-edge count before and after repair; the delta must equal the number of merged edge pairs.
Re-run detection after repair and confirm the remaining free edges map one-to-one onto legitimate boundaries.

Once the free-edge set is reduced to the intended boundary, the mesh is ready for the next connectivity check — for example, verifying that no T-junctions or duplicate nodes remain along the seams you just repaired.