Bayesian Optimization with Active Constraint Learning for Advanced Manufacturing Process Design

This study addresses the complex challenge of identifying process parameters for optimal manufacturing outcomes in advanced manufacturing, where nonlinear and costly process-to-quality relationships prevail. We introduce a novel experimental design framework that energizes the optimization of process parameters and feasibility constraint learning with a significantly reduced number of trials as compared to traditional Design of Experiments methods. Our approach is grounded in two primary methodologies: (1) active multi-criteria sample for constraint estimation and (2) Bayesian optimization-based sample for optimal parameter identification. This integration facilitates the efficient discovery of globally optimal parameter settings and outperforms multiple benchmark models in constraint estimation accuracy. The framework’s efficacy is demonstrated through application on both synthetic datasets and a real-world case study involving the synthesis of 2D materials, demonstrating its potential to enhance manufacturing efficiency and quality in complex manufacturing processes significantly.