You indicate here that each of the 1000 cases represent different geometries and inlet velocities.
Do you have the underlying mesh datasets with patches for visualization? (.obj, .stl, .vtk etc). As it looks like the input coordinates (Heatsink_Output_XYZ.npy) are just a single point cloud (the same 19517 points for all 1000 cases), so no geometry variation across the cases?
Regarding the 3D heatsink dataset.
You indicate here that each of the 1000 cases represent different geometries and inlet velocities.
Do you have the underlying mesh datasets with patches for visualization? (.obj, .stl, .vtk etc). As it looks like the input coordinates (Heatsink_Output_XYZ.npy) are just a single point cloud (the same 19517 points for all 1000 cases), so no geometry variation across the cases?