Developing of an Geometry-Aware AI Model for Predicting Stress Distribution in Aluminum Wheels under Impact
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.)：￥1,100 Members (tax incl.)：￥880
- Publication code
- 20265395
- Paper/Info type
- Proceedings (Spring)
No.91-26
- Pages
- 1-6(Total 6 p)
- Date of publication
- May 2026
- Publisher
- JSAE
- Language
- English
- Event
- 2026 JSAE Annual Congress (Spring)
Detailed Information
| Author(J) | 1) Changgon Kim, 2) Zeqing Jin, 3) Bowen Zheng |
|---|---|
| Author(E) | 1) Changgon Kim, 2) Zeqing Jin, 3) Bowen Zheng |
| Affiliation(J) | 1) Hyundai Motor, 2) University of California, 3) University of California |
| Affiliation(E) | 1) Hyundai Motor, 2) University of California, 3) University of California |
| Abstract(E) | This study proposes a geometry-aware Graph Neural Network (GNN) model for predicting stress distributions under dynamic impact on aluminum wheels. The model incorporates a weighted loss function and an edge reconstruction technique to improve accuracy in stress concentration regions. Using high-fidelity 3D finite element data from 85 wheel designs, we applied edge augmentation to enhance generalization with limited samples. The GNN model achieved a MAPE of 7.0% in critical regions, significantly reducing computation time compared to conventional FE analysis. This results demonstrate the feasibility of using GNNs for early-stage structural performance prediction in automotive development. |
