Research on the estimation of thermoelastic physical parameters based on a GNN-PINN hybrid model
Abstract
This paper proposes a GNN-PINN hybrid model for estimating thermoelastic parameters in BGA and QFN packaging structures, enabling quantitative analysis of solder fatigue in electronic devices. For BGA packaging, the PCBA is simplified as a non-homogeneous elastic plate, and the hybrid model captures micro-mechanical behavior at material interfaces under multi-scale, multi-field conditions. For QFN packaging, a layered graph network with direction-sensitive message passing and principal stress constraints is developed to extract directional equivalent parameters. Simulations reveal that QFN structures experience notable deformation under thermal cycling, particularly at high temperatures.
Fields of Study & Subjects
Electronic Packaging and Soldering Technologies
Model Reduction and Neural Networks
Mechanical and Thermal Properties Analysis
Publication details
- Journal / source
- International Conference on Advanced Manufacturing Technology and Electronic Information (AMTEI)
- Pages
- 93–93
- Publication date
- 14 January 2026