Recently, a research team led by Professor YANG Hangzhou from the School of Physics at Northwest University, in collaboration with a team led by Professor WANG Guoqing from the University of Electronic Science and Technology of China and a team led by Researcher LI Wei from the Beijing Institute of Space Mechanics & Electricity (BISME), has made significant progress in the field of using Physics-Informed Neural Networks (PINNs) to solve fluid equations with complex boundaries.
This method provides a new paradigm of high accuracy and high efficiency for solving fluid dynamics problems under complex boundary conditions. It not only promotes the development of PINNs themselves but also lays a solid theoretical foundation for their widespread application in engineering design and scientific computing. In the future, the team plans to further extend this method to solving a broader range of partial differential equations, such as those in heat conduction and solid mechanics.
The related findings, entitled “Hybrid Boundary Physics-Informed Neural Networks for Solving Navier-Stokes Equations with Complex Boundary,” have been accepted by NeurIPS 2025, a top-tier international conference on Artificial Intelligence and Machine Learning. The School of Physics of NWU is the primary completion unit for the paper. The first author is NWU master's student ZHOU Chuyu. Professor YANG Hangzhou is the corresponding author responsible for the article. The co-corresponding authors are XIN Guoguo, WANG Guoqing, and LI Wei. LI Tianyu from the University of Electronic Science and Technology of China is a co-first author. The co-authors include LAN Chenxi (University of Electronic Science and Technology of China), NAN Pengyu and DU Rongyu (NWU), and LIU Xun (Beijing Institute of Space Mechanics & Electricity).
