| About this Abstract |
| Meeting |
2023 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2023)
|
| Symposium
|
2023 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2023)
|
| Presentation Title |
Inexpensive Mechanistic-Knowledge-Agnostic Machine Learning in Additive Manufacturing |
| Author(s) |
Jeremy Cleeman, Rajiv Malhotra |
| On-Site Speaker (Planned) |
Jeremy Cleeman |
| Abstract Scope |
Machine Learning (ML) enables deployable modeling of parametric effects in additive manufacturing. Multi-Fidelity Learning (MFL) reduces the cost of training such a ML model by initially training it on a large amount of low-fidelity (LF) modeling data and then fine-tuning it based on a small high-fidelity (HF) experimental dataset. But current approaches for generating the LF data incur high model development cost (≈ decades) due to the need for deep mechanistic understanding of the process. We address this issue by pushing the boundary of the functional similarity necessary between the LF and HF data. This capability is demonstrated for two problems in Fused Filament Fabrication, i.e., predicting the printed road’s size and modeling the compression induced by an emerging in-situ rolling technique. We envision that our approach will enable accelerated and inexpensive understanding and adoption of new aspects or performance metrics in additive manufacturing despite limited mechanistic knowledge. |
| Proceedings Inclusion? |
Definite: Post-meeting proceedings |