Deep learning aided mesh generation
Unstructured nonuniform mesh generation for FE (Finite Element) and FV (Finite volume) under guidance of deep learning. Once a neural network is trained, this method generates non-uniform mesh without time spent on error estimation . It is also able to produce higher quality mesh than traditional mesh refinement when use proper training dataset and deep learning models. We explore problems from 2D Possion's equation to 2D and 3D linear elasiticy. We also see the possibilities in more complex problems such as unsteady fluids.