1
$\begingroup$

I'm trying to model a 3D die with ImplicitRegion[] in order to find out its center of mass. However, when I apply RegionCentroid[] to this region it takes too long to give an output. So I thought DiscretizeRegion[] would might render the computation of the centroid easier without losing precision.

θ=(63/100)Pi;
    depth=9/10;
    r = 22/10;
    offset = 1/2;
    offset61 = 3/10;
    offset62 = 6/10;
    ℛ= ImplicitRegion[
    (*Cube*)
        -1<=x<=1 &&
        -1<=y<=1 &&
        -1<=z<=1 &&
    (*Sphere*)
        x^2+y^2+z^2<=r&&
    (* face 1*)
        Not[(x^2+y^2)(Cos[θ]^2)-((Sin[θ]^2)((z-depth)^2))<=0&&0<(z-depth)<=1] &&
    (* face 2*)
        Not[((x-offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y+depth)^2))<=0&&-1<(y+depth)<=0]&&
        Not[((x+offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y+depth)^2))<=0&&-1<(y+depth)<=0]&&
    (*face3*)
        Not[((y-offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x+depth)^2))<=0&& -1<(x+depth)<=0]&&
        Not[((y)^2+(z)^2)(Cos[θ]^2)-((Sin[θ]^2)((x+depth)^2))<=0&& -1<(x+depth)<=0]&&
        Not[((y+offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x+depth)^2))<=0&& -1<(x+depth)<=0]&&
    (*face4*)
        Not[((y-offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x-depth)^2))<=0&& 0<(x-depth)<=1]&&
        Not[((y+offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x-depth)^2))<=0&& 0<(x-depth)<=1]&&
        Not[((y+offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x-depth)^2))<=0&& 0<(x-depth)<=1]&&
        Not[((y-offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((x-depth)^2))<=0&& 0<(x-depth)<=1]&&
    (* face 5*)
        Not[((x-offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y-depth)^2))<=0&&0<(y-depth)<=1]&&
        Not[((x)^2+(z)^2)(Cos[θ]^2)-((Sin[θ]^2)((y-depth)^2))<=0&&0<(y-depth)<=1]&&
        Not[((x-offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y-depth)^2))<=0&&0<(y-depth)<=1]&&
        Not[((x+offset)^2+(z+offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y-depth)^2))<=0&&0<(y-depth)<=1]&&
        Not[((x+offset)^2+(z-offset)^2)(Cos[θ]^2)-((Sin[θ]^2)((y-depth)^2))<=0&&0<(y-depth)<=1]&&
    (*face 6*)
        Not[((x+offset62)^2+(y+offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]&&
        Not[((x-offset62)^2+(y+offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]&&
        Not[((x+offset62)^2+(y-offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]&&
        Not[((x-offset62)^2+(y-offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]&&
        Not[((x)^2+(y+offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]&&
        Not[((x)^2+(y-offset61)^2)(Cos[θ]^2)-((Sin[θ]^2)((z+depth)^2))<=0&&-1<(z+depth)<=0]
    , {x,y,z}];
(*b=RegionPlot3D[ℛ,Axes->True, AxesLabel->{x,y,z},PlotStyle->Directive[Opacity[0.9]], PlotPoints->100]*)

When I try to discretize this region I get the following error:

DiscretizeRegion::drf: DiscretizeRegion was unable to discretize the region ImplicitRegion[<<2>>].

I'd like to know:

  1. How to circumvent this error?
  2. Is there a better method to obtain the centroid of this solid?
$\endgroup$
2
$\begingroup$

You can use the Finite Element mesh generator.

Needs["NDSolve`FEM`"]
mesh = ToElementMesh[\[ScriptCapitalR]]

enter image description here

That already gives a hint to what the problem is because ToElementMesh does not Quiet everything away. Then using the following returns something that does not look terrible:

mesh = ToElementMesh[\[ScriptCapitalR], {{-1.1, 1.1}, {-1.1, 
     1.1}, {-1.1, 1.1}}];
mesh["Wireframe"]

enter image description here

We can convert that to a MeshRegion:

enter image description here

Computing the center

rc = RegionCentroid[MeshRegion[mesh]]
{-0.0007281574436766132`, -0.00209881562008262`, \
0.0035201682587072554`}

Looks like it is slightly off center. I'd play a bit with the boundary resolution to see how much that affects the result.

A visual:

Show[
 mesh["Edgeframe"],
 Graphics3D[{Red, PointSize[0.02], Point[rc]}]
 ]

enter image description here

I did not get very far with a symbolic integration; try numeric integration. You may want to experiment with giving bounds to the ImplicitRegion.

For the statistics buffs: What are my odds for what numbers?

$\endgroup$
  • $\begingroup$ The only thing is that ToElementMesh gave me a "unable to discretize ImplicitRegion" error and a very coarse mesh, unlike yours... $\endgroup$ – Fred Nov 28 '18 at 2:28
  • $\begingroup$ @Fred what version are you on? $\endgroup$ – user21 Nov 28 '18 at 7:01
  • $\begingroup$ Version: 11.3.0.0 (Student Edition); Platform: Mac OS X x86 (32-bit, 64-bit Kernel) $\endgroup$ – Fred Nov 28 '18 at 14:06
  • $\begingroup$ @Fred, just double checking this is with the bounds specified to ToElementMesh[R,bounds], right. What happens if you use SetSystemOptions[ "FiniteElementOptions" -> {"SymbolicProcessing" -> 10.}]. If that does not help. I'd send this in to support. $\endgroup$ – user21 Nov 28 '18 at 14:15
  • $\begingroup$ I got "FiniteElementOptions" -> {"CacheInterpolationElements" -> True, "DefaultBounds" -> {-1, 1}, "DefaultElementMeshOrder" -> 2, "DefaultExtrapolationHandler" -> {Indeterminate &, "WarningMessage" -> True}, "DefaultNumberOfElements" -> 20, "InterpolationToleranceFactor" -> 0.0001, "MinimumElementMeshQuality" -> 0., "SymbolicProcessing" -> 10.} and reevaluated everything... Still the same. I've included the bounds {{-1.1, 1.1}, {-1.1, 1.1}, {-1.1, 1.1}} $\endgroup$ – Fred Nov 28 '18 at 14:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.