07-01-2022, 12:02 AM
(13-12-2021, 08:44 AM)josemendez Wrote: Hi!
Yep, 25000 particles is *way* too much. This will not only cause your simulation to run extremely slow, but it will also be pretty much impossible to get any accurate collision detection going on because particles will be really small.
You're using voxelization as your surface sampling method, with a resolution of 64 (!!). So 64x64x64 voxels will be generated, and the ones that intersect the cube's surface will be turned into particles. Some quick math tells you that will generate 64x64x6 = 24576 particles, give or take (each face spans 64x64 voxels, and there's 6 faces). Enabling voxel render mode in the blueprint editor makes this easy to see:
Your cube mesh only has 5x5 vertices at the top face, any additional particles will be wasted so there's absolutely no reason to go above a resolution of 5 in this case, when using voxel sampling. Even the default value of 16 would be too high. Alternatively, switch to vertex sampling.
Reducing the resolution had a *dramatic* effect on the frame rate. It now runs easily > 60 FPS. However, I'm still not able to cause depressions in the surface of the cube (for reference, here's what the mesh looks like).
I tried playing with the shape matching constraints without any noticeable difference. This is what the softbody looks like in the inspector tab:
This is the blueprint used for the softbody:
And this is what my softbody looks like in "Edit" mode:
Should I be creating particle groups and playing with the different properties in edit mode (e.g., assigning particles different values for deformation resistance and plastic yield)? Or do you think it is an issue with the mesh itself?