[adrian.delser]

Hi everyone,

I'd like to extract the uncorrected (non divergence free) velocity field from pisoFoam (or other solvers) at each time step. The goal is to correct it using alternate methods outside of pisoFoam using a neural network (therefore in python), and inject it back into pisoFoam for time advancement once corrected.

Any advice on efficient python-openfoam communication would also be appreciated, although communication optimisation would be for a later stage if the neural network can outperform (with some compromise between accuracy and time) the existing pressure/velocity correction in openfoam.


Thanks!