| About this Abstract |
| Meeting |
2024 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2024)
|
| Symposium
|
2024 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2024)
|
| Presentation Title |
A Fast Data-driven Residual Stress Prediction for Laser Powder Bed Fusion Additive Manufacturing Based on the Modified Inherent Strain Method |
| Author(s) |
Praveen S. Vulimiri, Shane Riley, Florian X. Dugast, Albert C. To |
| On-Site Speaker (Planned) |
Praveen S. Vulimiri |
| Abstract Scope |
Metal additive manufacturing processes, such as laser powder bed fusion or directed energy deposition, melt and fuse material to an existing structure to build a part sequentially. The repeated heating and cooling cycles introduce thermal stress, which can cause the part to distort and crack. While simulation can help predict the stress, the computational time required could potentially take longer than manufacturing the part. In this work, a data-driven, geometry agnostic mean-variance estimation model was developed to predict the residual stress in a few seconds. The model was trained on over 800 geometries from the Princeton University ModelNet database, simulated using the layerwise inherent strain method. For unseen parts, the normalized error of the predicted stress is around 10% on average, and 95% of errors are within two standard deviations of the predicted variance at each element. |
| Proceedings Inclusion? |
Definite: Post-meeting proceedings |