Abstract: Metal additive manufacturing (AM), e.g., laser-based powder bed fusion (L-PBF), offers an enabling opportunity for making complex metal parts or customized alloys with design freedom. The unique thermal cycle of rapid heating, fast solidification, and melt-back during metal AM may cause very complex metal pool dynamics, such as steep temperature gradient and high cooling rate, intense Marangoni flow, and intrinsic cyclic heat treatment. The complex very complex kinetic process and thermal history may lead to various quality issues of the printed parts. Therefore, the understanding and prognosis of metal pool dynamics remain the central intractable problem for printing high-quality metal parts or new alloys. Computational fluid dynamics (CFD) models may help to understand the complex thermo-mechanical process physics, but require the calibration of model parameters and are computationally expensive for real-time prognosis. On the other hand, machine learning has the potential to handle high-dimensional and massive process data for efficient surrogate modeling and decision-making. However, pure data-driven machine learning models suffer from black-box or explainability, are inherently computation-intensive and storage-intensive, and need a large amount of high-quality labeled training data to achieve a good performance. A deep knowledge gap exists between machine learning modeling and computational modeling in the prediction of melt pool dynamics. To take full advantage of ML methods while leveraging the physical laws underpinning melt pool dynamics, this talk presents a physics-informed machine learning (PIML) approach to integrate deep learning with the governing equations of the melt pool for forward prediction of the temperature and velocity fields in the melt pool. The PIML approach may also inverse learning of unknown model constants (e.g., Reynolds number and Peclet number) of the governing equations. The robust PIML algorithm also shows fast convergence by enforcing physics via soft penalty constraints.
Biographical Sketch: Dr. Yuebin Guo is Henry Rutgers Professor of Advanced Manufacturing and Leads the New Jersey Advanced Manufacturing Institute at Rutgers University-New Brunswick, USA. Prior to Rutgers, he served as the Assistant Director for Research Partnerships at the U.S. Advanced Manufacturing National Program Office (AMNPO). He was also an Alexander von Humboldt Fellow at RWTH Aachen and Fraunhofer IPT, Aachen, Germany. His research focuses on manufacturing processes, digital twins, physics-informed machine learning, and materials informatics. He is the author of more than 300 peer-refereed technical publications in these areas. He is a recipient of numerous awards, including the SME Sargent Progress Award, ASME Federal Government Swanson Fellow, Tau Beta Pi Outstanding Faculty, NSF CAREER, SAE Teetor Educational Award, and SME Outstanding Young Manufacturing Engineer. He is an elected fellow of ASME, SME, and CIRP.