A dimension-augmented physics-informed neural network (DaPINN) with high level accuracy and efficiency

Abstract: Physics-Informed Neural Networks (PINNs) have shown great potential in solving Partial Differential Equations (PDEs). However, they often suffer from slow convergence and low accuracy when dealing with high-dimensional multi-physics problems. In this paper, we propose DaPINN (Dimension-augmented PINN), which incorporates a dimension-augmented strategy to construct the loss function. We apply this method to simulate the coupled electrical-thermal-stress fields in VLSI interconnects. Experimental results demonstrate that DaPINN achieves significantly higher accuracy and faster convergence speed compared to standard PINN and other variants.

Download Paper

Recommended Citation:

```bibtex @article{guan2023dimension, title={A dimension-augmented physics-informed neural network (DaPINN) with high level accuracy and efficiency}, author={Guan, Weilong and Yang, Kaihan and Chen, Yinsheng and Liao, Shaolin and Guan, Zhong}, journal={Journal of Computational Physics}, volume={491}, pages={112360}, year={2023}, publisher={Elsevier}, doi={10.1016/j.jcp.2023.112360} }

Recommended citation: W. Guan, K. Yang, Y. Chen, S. Liao, Z. Guan. (2023). "A dimension-augmented physics-informed neural network (DaPINN) with high level accuracy and efficiency." Journal of Computational Physics. 491: 112360.
Download Paper