Repository navigation
Replies: 1 comment
|
The inconsistent junctions come from computing the blend against something that is not the surface (a fitted plane) and from having no rule for the place where the surface stops. Both go away if the blend is a rolling ball whose contacts are found by projection onto the two real meshes: the ball centre is the point at signed distance 1. A true tangent blend between the surface of revolution and an arbitrary mesh
Two things that bite in practice:
2. Where the receiving surface is locally missing The ball tells you. A tangent point is valid only if (a) every face around it faces the centre (angle between face normal and 3. Tangent-plane continuity as a constraint
4. Measuring tangency breaks without confusing tessellation with defects A dihedral angle alone is meaningless: on a smooth surface sampled with edge length
Results (pip
The seam is smoother than the blend's own tessellation - i.e. G1 up to discretisation. Sections in each feature's tilt plane (grey: union, red: blend):
Taper at the plate edge - radius along the contact curve and the two rims in top view:
If a global answer is acceptable, the morphological closing ( Script: |




The inconsistent junctions come from computing the blend against something that is not the surface (a fitted plane) and from having no rule for the place where the surface stops. Both go away if the blend is a rolling ball whose contacts are found by projection onto the two real meshes: the ball centre is the point at signed distance
rfrom both, its two tangent points say exactly where each surface is trimmed, and the arc between them is the blend section - tangent to both surfaces by construction, not by smoothing afterwards. MeshLib has no fillet function, but everything below is native and local (no voxels): oneboolean, twocutMesh, a band of quads, one stitch; 0.1 s per feature in P…