How to get a clean RegionDifference product

Question

I want to get the difference between the following two regions, an annulus and a crown:

Here's the code:

Rin = 0.75;
width = 0.25;
Rout = Rin + width;
nTriangles = 24;
basepoints = Rin*{Cos[#], Sin[#]} & /@ Range[0, 2 Pi, 2 Pi/nTriangles];
basesegments = {basepoints[[# + 1]], 
     basepoints[[Mod[# + 1, nTriangles] + 1]]} & /@ 
   Range[1, nTriangles];
triad[segment_, height_] := {segment[[1]], 
  segment[[2]], # + 
     height Normalize[#] &[(segment[[1]] + segment[[2]])/2 ]}
tiltangle = ArcCos[1];
flatsegmentsE = {{Cos[tiltangle] #[[1, 1]], 
      Cos[tiltangle] #[[1, 2]]}, {Cos[tiltangle] #[[2, 1]], 
      Cos[tiltangle] #[[2, 2]]}} & /@ 
   basesegments[[1 ;; nTriangles ;; 2]];
flatsegmentsO = {{Cos[tiltangle] #[[1, 1]], 
      Cos[tiltangle] #[[1, 2]]}, {Cos[tiltangle] #[[2, 1]], 
      Cos[tiltangle] #[[2, 2]]}} & /@ 
   basesegments[[2 ;; nTriangles ;; 2]];
triads = triad[#, width] & /@ Join[flatsegmentsE, flatsegmentsO];
flattriangles = Triangle[#] & /@ %;
(Create a region using DiscretizerGraphics on Graphics)
flatpicture = Graphics[flattriangles]
regiontriangles = DiscretizeGraphics[flatpicture]
regionAnnulus = 
  ImplicitRegion[x^2 + y^2 <= Rout^2 && x^2 + y^2 >= Rin^2, {x, y}];
 RegionDifference[regionAnnulus, regiontriangles];

RegionDifference works, but its output is not 'clean', which is to say there are some small hanging slivers of 'region' left at the inner tips:

I subsequently want to export this region as an STL file with meshing, so I need to get a clean RegionDifference product. Can someone please help?

Answer 1

It looks like you have introduced non-manifold geometry into your model. I explain non-manifold in more detail in my answer here. Even full-featured CAD packages will fail to create non-manifold geometry, as shown here:

We can try an approach using FEMAddOns as shown below. The FEM mesher will tend to create watertight and isotropic meshes.

(*Uncommented the following function if FEMAddOns not installed*)
(*ResourceFunction["FEMAddOnsInstall"][]*)
Needs["FEMAddOns`"];
bmeshann = 
  ToBoundaryMesh[regionAnnulus, MaxCellMeasure -> {"Length" -> .01}];
bmeshtri = 
  ToBoundaryMesh[RegionUnion @@ flattriangles, 
   MaxCellMeasure -> {"Length" -> .001}];
bmesh = BoundaryElementMeshDifference[bmeshann, bmeshtri];
mesh = ToElementMesh[bmesh, "MeshOrder" -> 1];
mesh["Wireframe"]
FindMeshDefects[MeshRegion[mesh]]

As you can see, FindMeshDefects finds many tiny faces on the inner ring.

We can mitigate the problem by adding a small margin to the inner radius of the annulus.

regionAnnulus = 
  ImplicitRegion[
   x^2 + y^2 <= Rout^2 && x^2 + y^2 >= (1.005 Rin)^2, {x, y}];
bmeshann = 
  ToBoundaryMesh[regionAnnulus, MaxCellMeasure -> {"Length" -> .1}];
bmeshtri = 
  ToBoundaryMesh[RegionUnion @@ flattriangles, 
   MaxCellMeasure -> {"Length" -> .01}];
bmesh = BoundaryElementMeshDifference[bmeshann, bmeshtri];
mesh = ToElementMesh[bmesh, "MeshOrder" -> 1];
mesh["Wireframe"]
FindMeshDefects[MeshRegion[mesh]]

By adding the margin to the inner radius, we have eliminated errors detected by FindMeshDefects.

Conversion to a 3D object using RegionProduct

If there is a desire to extrude the 2D mesh into a 3D object, one could use RegionProduct to accomplish this task as shown in the following workflow:

(*Tensor product mesh from:https://wolfram.com/xid/0rs5ccudm-eqv31q*)
pointsToMesh[data_] := 
  MeshRegion[Transpose[{data}], 
   Line@Table[{i, i + 1}, {i, Length[data] - 1}]];
rv = pointsToMesh[Subdivide[0, 1, 1]];
SetDirectory[NotebookDirectory[]];
Export["test.stl", RegionProduct[MeshRegion[mesh], rv]];
stl = Import["test.stl"];
bm1 = ToBoundaryMesh[stl];
(*Color surfaces by feature angle*)
groups = bm1["BoundaryElementMarkerUnion"];
temp = Most[Range[0, 1, 1/(Length[groups])]];
colors = ColorData["BrightBands"][#] & /@ temp;
bm1["Wireframe"["MeshElementStyle" -> FaceForm /@ colors, 
  PlotRange -> {{-1.5, 1.5}, {0, 1.5}, {-1.2, 1.2}}]]
FindMeshDefects[MeshRegion[bm1]]

Answer 2

This is nothing compared to @Tim Laska's detailed answer above, but I found that a quickfire way of removing the small slivers from the geometry was to use the option MaxCellMeasure -> Infinity while turning the region into a mesh, as such:

mr = DiscretizeRegion[regionCones, MaxCellMeasure -> Infinity]

giving me

This works probably because the slivers are small, and setting a large value for MaxCellMeasure forces the meshing algorithm to ignore them. I don't expect this to work for generic non-manifold geometry. For that, refer to @Tim Laska's answer above.