2025 - Sustainable Industrial Processing Summit
SIPS2025 Volume 14. Intl. Symp on Multiscale, Modelling, Nanotechnology and Modelling Materials
| Editors: | F. Kongoli, D. Bammann, R. Das, J.B. Jordon, R. Prabhu, A. Rajendran, P. Trovalusci, M. de Campos |
| Publisher: | Flogen Star OUTREACH |
| Publication Year: | 2025 |
| Pages: | 214 pages |
| ISBN: | 978-1-998384-64-8 (CD) |
| ISSN: | 2291-1227 (Metals and Materials Processing in a Clean Environment Series) |
CD shopping page
INTERPRETABLE MACHINE LEARNING APPROACH FOR EXPLORING PROCESS-STRUCTURE-PROPERTY RELATIONSHIPS IN METAL ADDITIVE MANUFACTURING
Xiaopeng Li1;1UNSW SYDNEY, Sydney, Australia;
Type of Paper: Keynote
Id Paper: 15
Topic: 1
Abstract:
Establishing process-structure-property (PSP) relationships is essential for optimizing manufacturing techniques, yet it often requires extensive, costly experimentation. This is particularly true for additive manufacturing (AM), where numerous process parameters complicate the task. Our research introduces an interpretable machine learning strategy to predict and refine the process window for laser powder bed fusion (LPBF), while also delineating PSP relationships. We utilized Gaussian process regression (GPR) to model various inputs, such as process parameters and microstructural features, to predict key mechanical properties. The adaptability of the GPR model, through hyperparameter tuning for each input, facilitates feature selection and enhances model transparency. This methodology not only identifies pivotal factors influencing mechanical performance but also clarifies PSP relationships in additive manufacturing alloys, offering insights for customizing final material properties. Our approach is versatile, applicable across different additive manufacturing techniques and materials, and opens the door to achieving new mechanical properties and deeper PSP understanding.