Bake it till you make it - Graph neural networks for composite cure modeling
Ashish Hegde
PhD Student
Motivated by the need to minimize the development time and production costs of aerospace composite parts without compromising their quality, this project focuses on leveraging the power of AI-based surrogates for modelling material behavior.
The aim of this research is to use graph neural networks in a MeshGraphNet-based architecture to develop data-driven and physics-informed surrogate models that will simulate the curing of composites. Fast predictions of a trained, robust AI model will facilitate high-throughput process design through cure-cycle optimization, mould design optimization, and minimizing process-induced defects. This would accelerate the design and manufacturing cycle significantly compared to the current reliance on computationally expensive Finite Element Modeling (FEM).
