A machine learning-based digital twin of the manufacturing process: metal powder-bed fusion case
| Autor: | Omar Fergani, Katharina Eissing |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
engrXiv|Engineering
bepress|Engineering bepress|Engineering|Computational Engineering engrXiv|Engineering|Computer Engineering bepress|Engineering|Mechanical Engineering|Manufacturing engrXiv|Engineering|Computational Engineering bepress|Engineering|Computer Engineering engrXiv|Engineering|Manufacturing Engineering |
| Popis: | Additive manufacturing processes are enabling the manufacturing of geometries that were impossible to manufacture before. However, the technology is also facing challenges due to process inconsistencies, such as local overheating which creates large scrap rates. An intelligent process relies on a digital twin [1, 2] to achieve a robust process planning avoiding substantial costs related to trial and error to overcome those obstacles. Such a digital twin of the process hence helps to accomplish a print first-time-right. We build a digital twin of the AM process - including the full print path - using a machine learning algorithm trained on synthetic, physics-based and experimentally validated data. This allows for the prediction of overheated regions for a complete workpiece within a reasonable time and the application of improved scan strategies. We outline the general method on an exemplary aerospace part made of In718, successfully predicting overheated regions and correcting it through an alternative exposure strategy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|