About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
AI-Enabled Materials Processing: Integrating Accelerated Experimental Workflows and Processing-Aware Machine Learning
|
| Presentation Title |
Machine learning in sintering and ceramic processing: from gradient descent optimizations to hybrid methods with finite element simulations. |
| Author(s) |
Charles Maniere |
| On-Site Speaker (Planned) |
Charles Maniere |
| Abstract Scope |
Evaluating sintering models experimentally presents challenges due to parameter interdependence and the separation between the thermally activated mechanisms of sintering and the powder-specific characteristics embedded in the sintering moduli. To circumvent the need for specialized tests, like instrumented sinter-forging, inverse learning approaches using gradient-descent optimization enable the extraction of model parameters from dilatometry data. The resulting moduli accurately reflect the powder’s intrinsic behavior, thereby ensuring reliable sintering models.
Another challenging aspect in finite element simulations of sintering is the inherent divergence of the model when applied to highly complex structures. To address this issue, we have developed a hybrid FEM-deep learning approach to efficiently predict sintering distortion in complex geometries. This model employs finite element simulations to generate synthetic datasets, which are then used to train a neural network model. The trained model is subsequently integrated with advanced methods such as topology optimization to design shapes that resist sintering distortion. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Machine Learning, Process Technology |