Colloquiums and Conferences
Combining Physics-informed Neural Networks with the Fixed Stress Splitting for Solving Biot's Model
Time: 2025-03-22 18:20:00
Title: Combining Physics-informed Neural Networks with the Fixed Stress Splitting for Solving Biot's Model
Speaker: Prof. Mingchao Cai (Invited by Prof. Qi Ye)
Affiliation: Morgan State University, USA
Time: March 21, 10:00–11:00
Venue: Lecture Hall, Room 111, West Building, School of Mathematical Sciences
Speaker Biography:
Mingchao Cai is a professor and doctoral supervisor at Morgan State University, USA. His research focuses on numerical analysis and the computation of fluid, elastic and biomechanical systems. Since 2017 he has received more than USD 1.2 million in research funding from the NSF, NIH and the Army Research Office. He has published more than 30 papers in influential SCI journals such as SIAM Journal on Numerical Analysis, SIAM Journal on Scientific Computing, Mathematics of Computation and Journal of Biomechanics, as well as two book chapters, and has given over 80 plenary or invited talks. He serves as a review panellist for the NSF computational mathematics programme and organised the CBMS conference Deep Learning and Numerical PDEs (2023).
Abstract:
Biot's model serves as the cornerstone of poroelasticity. Our objective is to address this model using physics-informed neural networks (PINNs). By leveraging the fixed-stress splitting iterative method, we design distinct loss functions for the displacement and pressure variables. This approach enables the training of two separate, compact neural networks. The fixed-stress iterative algorithm is employed to couple these variables, enhancing the accuracy of the neural network approximations. Our numerical experiments validate the efficacy of this method, and we provide error analysis to underscore the capability of fixed-stress splitting-based PINNs in accurately approximating solutions to Biot's model. Additionally, we briefly explore the application of PINNs to inverse problems associated with Biot's model.