[Question] - Wall boundary condition

[bakhtiyar-k]

Hello. I have a question about the walls during ground-truth simulations. Do you perform simulations with plane walls or walls represented as particles?

[benjaminum]

Everything is represented with particles.

[bakhtiyar-k]

So you convert the wall surface to static particles, introduce fluid particles, then run simulations with splishsplash to gather data. Compress the data and finally supply it as input to the network. Is it correct?

[benjaminum]

yes

[bakhtiyar-k]

Thank you

[bakhtiyar-k]

Still a bit confused. In the code, there is the run_simulator function that uses a scene.json file and an output directory. The json file contains information about fluids in .bgeo (particle data) and a rigid body in .obj (geometry data) format, so in simulations, fluid particles hit geometry planes, not static particles?

[benjaminum]

The geometry in the .obj is converted to particles here

DeepLagrangianFluids/scripts/run_network.py

Line 45 in d651c6f

points, normals = obj_surface_to_particles(x['path'])

[bakhtiyar-k]

yes, but it is for a trained network, right?

I mean in the data generation stage,

DeepLagrangianFluids/datasets/create_physics_scenes.py

Line 430 in d651c6f

run_simulator(os.path.abspath(scene_output_path), sim_directory)

scenes are run with fluid particles interacting with each other and with boundary planes that are not converted to particles yet?

[bakhtiyar-k]

After a scene was created, we have data only of fluid particles? The .bgeo geometry file is later converted to particles

DeepLagrangianFluids/datasets/create_physics_records.py

Line 19 in d651c6f

box, box_normals = numpy_from_bgeo(os.path.join(scene_dir, 'box.bgeo'))

and added to the feat_dict ?

DeepLagrangianFluids/datasets/create_physics_records.py

Line 56 in d651c6f

feat_dict['box'] = box.astype(np.float32)