Understanding and controlling the process–structure–property relationship is central to advancing manufacturing. This talk presents AI/ML methods that study manufacturing processes and material properties through physically meaningful intermediate representations. We begin with droplet evolution modeling in inkjet printing, where diffusion-corrected neural operator learning characterizes process-level dynamics that govern material deposition and initial geometric fidelity. Then, we will focus on metal additive manufacturing, treating melt pool geometry and dynamics as a process-induced structural descriptor that encodes thermal history and solidification conditions. We introduce a deep learning framework with spatiotemporal attention to automatically segment melt pools from in-situ X-ray image sequences, enabling quantitative, reproducible extraction of structure-relevant features. In the end, we conclude with fatigue life prediction of aluminum alloys, where a transformer-based deep operator model leverages physics-informed features to estimate fatigue properties under cyclic loading. Together, these studies demonstrate how physics-guided AI can provide a data-driven pathway for predictive modeling and intelligent control in advanced manufacturing systems.
Bo Shen is an assistant professor in the Department of Mechanical & Industrial Engineering and the Department of Data Science at NJIT. He got his Ph.D. from Virginia Tech in 2022. He obtained his B.S. degree in Statistics from the University of Science and Technology of China. He is currently interested in developing and applying generative AI for science and engineering, specifically space weather and advanced manufacturing. His research is supported by NASA, DoD, NSF, and NVIDIA.
