| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Defects and Properties of Cast Metals
|
| Presentation Title |
A Machine Learning Approach for Prediction of the Size and Locations of Porosity in High Pressure Die Casting |
| Author(s) |
Utkarsh Godwal, Shishira Bhagavath, Supriyo Roy, Bita Ghaffari, Larry Godlewski, Mei Li, Peter D Lee, Shyamprasad Karagadde |
| On-Site Speaker (Planned) |
Utkarsh Godwal |
| Abstract Scope |
High Pressure Die Casting (HPDC) is a rapid manufacturing process producing thin and intricate components for automotive and aerospace applications. Due to the complex shapes and high cooling rates, the metal solidifies rapidly, leading to the formation of features such as porosity and shear bands. Most approaches for porosity prediction are limited to component-level predictions, this work combines local-level micro-scale models with a macro-scale model using a Machine Learning (ML) framework. The ML model is trained using a deformable-grid one-dimensional finite volume model of combined gas and shrinkage pore growth as a function of pressure, temperature, cooling rate, etc., predicted from the transient component simulation. The model is further trained using X-ray microtomography (XMT) data from a macro-level component. The model successfully predicts pore sizes and locations on a component level spanning the entire range of XMT observed porosity. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Modeling and Simulation, Solidification |