# Butterfly - test mesh points inside geometry

Hi all,

I have a problem where the ‘generate test points’ component creates points that are inside the refined Snappyhexmesh of the object. This creates some results that have super high pressures which need to be filtered out in order to get the coloured vector display in the example file. the problem seems to come up in two ways

1 . the meshing of the test geometry is too coarse, and triangles are formed where the centre point is actually inside the object - see third image. I think this can be resolved by just increasing the density of the test mesh so it’s about equal to the snappyhexmesh at the surface. Is this right?

1. For the test object plane that is cutting obliquely through the object, the points actually move off the plane toward the mesh. This seems to happen regardless of whether ‘move test mesh’ is true or false. When I move the plane to where the points are, they move again next time the solution is run. As you can see in my example, if these points move toward the centre of the sphere, then they are inside the object and skew the result.

What am I doing wrong here? Guessing this question has been answered somewhere before but I couldn’t find anything specifically answering this question so far.

thanks
Nick

200531_Sphere.gh (441.1 KB)

Hi,

I always have the same question while simulating with complicated volumes.
For me, I do nothing with the redundant points, I just adjust the range of the color gradient to make sure the color looks reasonable.
I hope to see some other new solution.

Jiewei

Thanks for commenting @ljwssdsw I was looking into a workaround like this also but yours is clean.

Sometimes just phrasing the questions helps to focus on the solution… I noticed - pretty dumb - that I had a small offset (_distBaseSrf) in the honeybee_generate test points component of 0.01m which was causing some points to be inside the sphere. I’m not sure what the function of this offset is.

I made that zero. Also just increased the mesh density of the test surface. It takes longer to run the solution but maybe it’s possible to somehow create some refinement of the test mesh.

Hi Nick -

It sounds like you should check the normals of that object and most likely flip them.
-wim