Mesh deep Q network: A deep reinforcement learning framework for improving meshes in computational fluid dynamics

Meshing is a critical, but user-intensive process necessary for stable and accurate simulations in computational fluid dynamics (CFD). Mesh generation often bottleneck CFD pipelines. Adaptive meshing techniques allow the mesh to be updated automatically produce an solution problem at hand. Existing classical adaptive require either additional functionality out of solvers, many training simulations, or both. Current machine learning substantial cost data generation, are restricted scope flow regime. Deep Q Network (MeshDQN) developed as general purpose deep reinforcement framework iteratively coarsen meshes while preserving target property calculation. A graph neural network based used select vertices removal interpolation bypass expensive each step improvement process. MeshDQN requires single simulation prior coarsening, making no assumptions about regime, type, solver, only requiring ability modify directly pipeline. successfully improves two 2D airfoils.