[sminder]

Hi,

I'm trying to combine machine learning with fluid mechanics. SU2 seems to be a good start as it supports python. One of my questions is how can I change the boundary conditions with my own script? For example, if I use a neural network to adjust the boundary conditions, how can I integrate a custom python script to define a dynamic boundary condition in SU2, is it feasible and how hard it is?

Many thanks!
Li

[wallym]

Hi Li,

Thanks for your post! Since the boundary conditions are defined in the configure script, you should be able to have a simple function that reads the configure file, replaces/writes the boundary conditions, and either runs/restarts SU2.

While Im sure there are other ways to do this more efficiently, this should work and be relatively easy.

[sminder]

Originally Posted by wallym

Hi Li,

Thanks for your post! Since the boundary conditions are defined in the configure script, you should be able to have a simple function that reads the configure file, replaces/writes the boundary conditions, and either runs/restarts SU2.

While Im sure there are other ways to do this more efficiently, this should work and be relatively easy.

Thank you for your reply, I wonder is there a function that I can update the boundary conditions without reading the file? As I have to change the boundary condition every time step. For example, can I just update the variable like "configure.MACH_NUMBER=XX"?

Many thanks,
Li

[CFD Lab]

Good day, I am looking for a comprehensive course in which I can learn, How to implement Machine Learning with Computational Fluid Dynamics. I am also looking for some research Ideas. Looking for a kind response thanks