A finite-element-inspired bipartite graph learned simulator for manufacturability assessment in large-deformation sheet forming
The research developed CAtt-BiGNN, a bipartite graph neural network simulator inspired by explicit finite element methods. It learns coupled nodal displacement and elemental thinning fields during sequential simulation steps (rollout) for large-deformation sheet forming. This method treats nodes and elements as distinct entities connected by directed edges, allowing for accurate prediction of how the geometry changes and where material thinning occurs.