Physic Informed Deep Learning
Physics-informed deep learning (PIDL) combines the power of deep neural networks with the constraints of physical laws to solve complex scientific and engineering problems. Current research focuses on improving the accuracy and efficiency of PIDL models, exploring various architectures like convolutional and recurrent neural networks, and developing novel training strategies such as adaptive weighting and physics-guided initialization. This approach is proving valuable across diverse fields, enhancing the accuracy and efficiency of simulations in areas ranging from fluid dynamics and traffic flow estimation to material science and medical imaging. The integration of physical knowledge into deep learning models leads to more robust, reliable, and interpretable results compared to purely data-driven approaches.
Papers
Kolmogorov Arnold Informed neural network: A physics-informed deep learning framework for solving forward and inverse problems based on Kolmogorov Arnold Networks
Yizheng Wang, Jia Sun, Jinshuai Bai, Cosmin Anitescu, Mohammad Sadegh Eshaghi, Xiaoying Zhuang, Timon Rabczuk, Yinghua Liu
Physics-Informed Deep Learning and Partial Transfer Learning for Bearing Fault Diagnosis in the Presence of Highly Missing Data
Mohammadreza Kavianpour, Parisa Kavianpour, Amin Ramezani