Microstructure optimization with constrained design objectives using machine learning-based feedback-aware data-generation

Flow diagram of the proposed methodology. There are two phases in the sampling process. In the first phase, two data generation algorithms are explored for dataset creation, namely partition and allocation. In the first step, we execute our data-generation algorithms (depicted by orange arrows) to generate many valid solutions for each set of constraint. In the second phase, a machine learning-guided feedback-aware sampling approach (depicted by green allows) extracts a combination of non- zero ODF dimensions. ODFs generated with these subset of combinations lead to optimal or near-optimal solutions.