3. Resolve Duplicate Nodes Along a Weld Line
Why duplicate nodes survive on a weld seam
In a shell model of a welded assembly, the two flanges are usually meshed independently. Each flange gets its own node row along the seam, so the seam ends up with two coincident node sets sitting at identical coordinates but belonging to different elements. Visually the mesh looks closed; topologically it is two free boundaries pressed together. HyperMesh will not transfer load across that seam, and any subsequent equivalence or quality check will report the seam as a discontinuity.
The trap is that a weld line is exactly where you do want nodes merged, while the same model often contains intentional gaps — a 0.5 mm paint gap, a 1.0 mm adhesive layer, a sliding joint — where coincident-looking nodes must stay separate. A blind equivalence with a large tolerance will close both. The task is therefore two-part: find the duplicate pairs on the seam, and merge only those, with a tolerance that is provably smaller than the smallest intended gap.
Establish the tolerance budget before touching anything
Work out the merge tolerance from the model, not from habit. Three numbers matter:
- Smallest intended gap in the model, e.g. 0.8 mm adhesive layer. The merge tolerance must be well below this — a common rule is one third or less, so ≤ 0.25 mm here.
- Actual node offset on the seam. Independent meshing rarely produces exactly coincident nodes; expect 0.01–0.05 mm of drift from projection and midside-node placement.
- Element size at the seam. A tolerance above roughly 5% of the local element edge length starts pulling in unrelated nodes on curved geometry.
With a 0.8 mm gap and 0.03 mm seam drift, a tolerance of 0.1 mm is safe: it captures the drift with margin and is eight times smaller than the gap. Record this number — it is the single parameter that decides whether the operation is correct.
Locate the duplicate pairs with find and equivalence preview
Do not run equivalence directly. First quantify what is there.
- Isolate the weld region: put the seam elements in a component or use a box selection so the operation cannot reach the rest of the model.
- Open Tool page → equivalence. Set
toleranceto 0.1 and leaveuse actualoff so the nominal coordinates are used for comparison. - Press preview. HyperMesh colours the nodes it intends to merge and reports the count of coincident nodes found. On a typical flange seam of ~400 mm length with 5 mm elements you should see roughly 80 pairs — one per element edge along the seam. A count far above that means the tolerance is reaching into the gap or into the flange thickness.
- Cross-check with find → nodes → by coincident using the same tolerance, restricted to the seam component. This gives you a node list you can inspect individually rather than a bulk colour map.
If the preview count is zero, the two flanges are not actually coincident — check for a modelling offset or a mid-surface shift before assuming the mesh is wrong.
Merge, then verify the seam is conformal and the gaps are intact
With the preview clean, execute the equivalence. HyperMesh merges each pair into a single node, keeping the position of the retained node; the second node is deleted and its elements are reconnected. The seam is now conformal — the two flanges share one node row.
Verification is not optional, because equivalence is destructive and silent:
- Free-edge check on the seam. Run find → edges → free edges on the seam component. Before the merge the seam shows two coincident free-edge loops; after a correct merge it shows none. Any remaining free edge on the seam means a pair was missed.
- Gap check. Re-run find → nodes → by coincident at the gap location with the same 0.1 mm tolerance. It must still return the two separate node rows. If it returns nothing, the gap has been collapsed and you must undo.
- Element quality spot check. Merging moves nodes by up to the tolerance. Check the Jacobian and warpage of the seam elements; a node pulled 0.1 mm on a 2 mm element can visibly distort a quad.
When the merge goes wrong
| Symptom | Likely cause | Action |
|---|---|---|
| Preview highlights nodes across the adhesive gap | Tolerance larger than the gap | Reduce tolerance to ≤ gap/3 and preview again |
| Free edges remain on the seam after merge | Seam drift exceeds tolerance, or a pair was outside the selected component | Re-run find by coincident at a slightly higher tolerance, inspect the missed pair individually |
| Seam elements show high warpage or low Jacobian | Node moved by the full tolerance on a small element | Undo, reduce tolerance, or use equivalence with use actual on to keep the retained node at its true position |
| Gap nodes disappeared | Equivalence reached beyond the seam selection | Undo immediately; restrict the selection to the seam component only |
The whole operation reduces to one disciplined number: a merge tolerance that sits above the seam drift and comfortably below the smallest intended gap. Preview at that tolerance, confirm the highlighted nodes are only on the seam, execute, then prove the result with a free-edge check on the seam and a coincident-node check at the gap. If both pass, the weld line is conformal and the design gaps are untouched.