boolean fail

[MackWang0102]

21.zip
I failed to perform Boolean operations using MeshLib.SDK (C#) (see attachment: body.stl cut 21.stl)

[Fedr]

Hi, both input meshes have orientation problems, and that's why the boolean fails:

1. Body.stl is not a closed surface. The side wall is one piece with two holes (top and bottom). The two caps that should close those holes are separate pieces with inverted normals: the bottom cap faces +Z and the top cap faces -Z. Because they face the wrong way, they can't be stitched to the wall without making the surface non-manifold. MeshLib keeps them as separate components, so Body has 3 components and 4 holes.

2. 21.stl is closed, but it is turned inside out: all of its normals point inward, so its signed volume is negative (-0.0134).

Both meshes load fine after these fixes, and DifferenceAB returns a closed mesh. Tested with the MeshLib 3.1.3.429 NuGet package:

var body = MR.MeshLoad.fromAnySupportedFormat("Body.stl");
var cut  = MR.MeshLoad.fromAnySupportedFormat("21.stl");

// Body: keep only the side wall and close its two planar holes again
var wall = MR.MeshComponents.getLargestComponent(new MR.MeshPart(body));
var fixedBody = new MR.Mesh();
fixedBody.addMeshPart(new MR.MeshPart(body, wall));
MR.fillHoles(fixedBody, fixedBody.topology.findHoleRepresentiveEdges(), new MR.FillHoleParams());

// 21: turn it outside out
cut.topology.flipOrientation();

var res = MR.boolean(fixedBody, cut, MR.BooleanOperation.DifferenceAB);

Result: closed mesh, 3692 triangles, volume 3.04591 (Body 3.04640 minus the 0.00048 overlap with 21).

The cleanest fix is in the exporters. Body.stl came from Eyeshot (devDept), so the cap orientation should be fixed there. 21.stl is turned inside out wherever it was created. A quick way to check any input before a boolean: mesh.topology.isClosed() must be true and mesh.volume() must be positive.

[MackWang0102]

Thank you for your prompt reply. Currently, when I sequentially subtract 4_0.stl, 4_1.stl, 6_0.stl, 6_1.stl, 7_0.stl, 7_1.stl, 8_0.stl, and 8_1.stl from body.stl, an error occurs. The error message reads: "Bad contour on 2 mesh B faces, probably mesh A has self-intersections on contours lying on these faces."

Could you please offer some suggestions on how to perform the Boolean operation successfully?
Uploading mesh.zip…

[Fedr]

The attachment didn't upload. Your comment only shows the [Uploading mesh.zip…]() placeholder. Could you attach mesh.zip again and check that it turns into a link before you post?

In the meantime, here's what that error usually means. On some step, the mesh you are cutting (the result of the previous subtractions) and the next cutter meet in a degenerate way. Usually two cutters share a face or touch along an edge, for example 4_0 and 4_1. Then the new cut runs exactly along the edge of an earlier cut. Some things to try:

1. Check every cutter the same way as 21.stl above: topology.isClosed() must be true and volume() must be positive. Flip any cutter with a negative volume.
2. After each subtraction, check the result with MR.findSelfCollidingTriangles(mesh), so you know which step first produces a bad mesh.
3. If two cutters touch or overlap, unite them first with MR.BooleanOperation.Union, then subtract the united cutter from the body in one operation.
4. If cutters share a face exactly, move one of them by a tiny amount (much smaller than your tolerance, e.g. 1e-4 of the model size) so the surfaces no longer coincide.

With the files I can tell you exactly which step fails and why.

[MackWang0102]

mesh.zip

[MackWang0102]

mesh.zip

[Fedr]

Thanks for the files. This time all meshes are fine: each is closed, oriented outward, and has no self-intersections. I reproduced the error: the subtractions of 4_0 ... 7_1 succeed, and 8_0 fails.

The reason is that the six small cutters 6_0, 6_1, 7_0, 7_1, 8_0 and 8_1 all have faces passing through the same point (0, 0, 4), the centre of the top of Body. Each subtraction leaves very thin triangles around that point, and after 7_1 two of them intersect. The next subtraction then hits that self-intersection and stops.

The fix is to unite the six small cutters first. Unions of these simple meshes don't produce degenerate triangles. Then subtract the united tool from the body:

MR.Mesh Load(string name) => MR.MeshLoad.fromAnySupportedFormat(name + ".stl");

var body = Load("Body");

// the six small cutters meet at one point on the top of the body, so unite them first
var tools = Load("6_0");
foreach (var name in new[] { "6_1", "7_0", "7_1", "8_0", "8_1" })
{
    var u = MR.boolean(tools, Load(name), MR.BooleanOperation.Union);
    if (!u.valid()) throw new Exception(name + ": " + u.errorString);
    tools = u.mesh;
}

foreach (var cutter in new[] { Load("4_0"), Load("4_1"), tools })
{
    var res = MR.boolean(body, cutter, MR.BooleanOperation.DifferenceAB);
    if (!res.valid()) throw new Exception(res.errorString);
    body = res.mesh;
}

With MeshLib 3.1.4.297 (the latest NuGet), the result is a closed mesh with 34028 triangles, volume 2.38746, and no self-intersections. Version 3.1.3.429 also succeeds, but its result has 3 small self-intersecting triangles near the body wall, so please update.

[MackWang0102]

Thank you for your reply.