| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
Hierarchical Machine Learning Framework for Optimizing Material Properties. |
| Author(s) |
Zahra Zanjani Foumani, Mahsa Amiri, Ramin Bostanabad, Lorenzo Valdevit |
| On-Site Speaker (Planned) |
Zahra Zanjani Foumani |
| Abstract Scope |
Building predictive and interpretable process-property links is crucial in additive manufacturing for achieving desired part properties. Traditional methods are often time-consuming, expensive, and limited by data scarcity and lack of generalizability. To address these challenges, we develop a hierarchical and probabilistic machine learning (ML) framework that learns from multiple datasets.
Our goal is to leverage the proposed framework to quantify the effects of process parameters on critical tensile properties like yield strength and ductility, and optimizing them despite their competing nature. To reduce the reliance on expensive tensile data, we leverage hidden strength mechanics. This involves (1) gathering surface data on hardness and porosity from small cuboids in a high-throughput setting, and (2) integrating this data into our hierarchical framework to improve learning the complex relationships between process parameters and tensile properties. We demonstrate the efficacy of our approach by experimentally validating the optimized parameters suggested by the model.
|
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Modeling and Simulation, Computational Materials Science & Engineering |