Model Transfer Between Material Systems for Distortion Prediction in Laser-Based Additive Manufacturing

• Activity Number: 297 - SPEED: Food, Environment, Biomedical Imaging and Physical System Visualization/Learning, Part 1
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Arman Sabbaghi* and Jack Francis and Linkan Bian
• Companies: Purdue University and Mississippi State University and Mississippi State University
• Keywords: 3D printing; Bayesian data analysis; effect equivalence; transfer learning

Abstract

The distortion of materials in laser-based additive manufacturing (LBAM) is a critical issue that affects the geometric integrity of the product and is known to be material-dependent. One key challenge in LBAM research is the stark difference of material properties, which makes learning about a new material system given past experiments on an old material system difficult. We propose a physics-based, data-driven, Bayesian transfer learning methodology to leverage past experiments to learn about new material systems. Our methodology effectively infers the material differences in terms of a lurking variable that is learned directly from the experiments. We validate our methodology by predicting the distortion of disks fabricated using Ti-6Al-4V and 316L stainless steel. Our methodology is the first framework for transferring knowledge between material systems, which is critical due to the wide range of materials used for the various applications of LBAM.