Hello,
I am working on some test case with Damien. I am simulating the case of a submerged sphere (D=3.76cm, center 5cm below the free surface in a constant stream of 0.45m/s). I created the geometry using Salome.

First problem : salome encounters issue when using dynamics (with 3 layers) to mesh inward. So I created 3 spheres by hand and meshed as if it was Lennard Jones BC. I still have some leaks with this technique. I am working on it. Particles enter the sphere probably because spacing is too large (so I reduced it).
Another solution which appears to work ok was to create the smallest sphere (inner sphere of the 3 layers) and ask salome to mesh outward. On the inner sphere I reduced the dr such that the outward sphere has the correct dr (assuming particle are just moved following the local normals). 
Second problem : with a mesh with no leaks, I was able to simulate the case but obtained forces about 3 times larger than expected. The expected force should be from experiments and Code Saturne simulations around 0.048N. I measured 0.12N… Any idea why the force is so much larger? I have no turbulence in the model for now.
How is the force evaluated in GPUSPH?
The goal afterward is to add a ‘paddle’ to this sphere and let it swim with a force model following a Hill’s like law.
Thanks in advance for help,
Rémi