Sudeepta Mondal¹, Daniel Gwynn¹, Nandana Menon¹, Asok Ray¹, Amrita Basak¹; ¹Pennsylvania State University

During material consolidation in additive manufacturing (AM), the melt pool dimensions play a critical role in determining the final grain structure and thus the resulting mechanical properties. However, maintaining desired melt-pool properties are challenging due to the inherent layer-by-layer fabrication process resulting in cyclic heating and cooling as well as thermal gain of the component being built. As a possible solution to this problem, a physics-informed machine learning (ML) assisted modeling and optimization framework was explored in this work. An analytical heat transfer model is employed for predicting the thermal distribution in a directed energy deposition process for faster computation. Thereafter, a surrogate-assisted statistical learning and optimization architecture involving Gaussian Process-based modeling and Bayesian Optimization is employed for finding the optimal set of process parameters as the scan progresses, subject to the constraint of maintaining a desired percentage of columnar growth during the build.