Ideas for AI models in CFD
 

Hello,

I’m exploring AI for physics using pytorch, I’m playing around with a combustion use case in which I contributed using GNN to infer the surface flame density. I’m running out of simple idea to implement on my free time 🙂*. If anyone has a simple idea to explore modeling with AI for physics I will be glad to discuss. Here is the work I’ve done around GNNs for the combustion use-case [ https://github.com/rmslm/ai-physics ] and here is the dataset I used [ https://www.coe-raise.eu/open-data ].

I implemented it for structured, unstructured meshes and I implemented basic mesh splitting for the mesh to be fitted inside a single GPU memory and batched for training.

I can provide a dockerfile if some of you want to explore this configuration. Please let me know :)

Any idea ? perhaps using open source solver like openfoam or one with a python interface ?

thanks
Regards.

If you could explain in more details that would be interesting. As a casual warehouse operator, I feel kind of insecure about folowing links.