Neural Network-Based Modeling of Diffusion-Induced Stress in a Hollow Cylindrical Nano-Electrode of Lithium-Ion Battery

Abstract Understanding the interaction between mechanical deformation and mass transport, such as diffusion-induced stress, is crucial in the development of advanced battery materials and electrochemical devices. Mathematical modeling and solving the related coupling problems have played important roles in advancing the understanding of the interaction between mechanical deformation and mass transport. As the complexity of mathematical modeling continues to increase, numerical methods used to solve the related coupling problems are likely to encounter significant challenges. This work explores the feasibility of designing a neural network specifically for solving diffusion-induced stress in the electrode of lithium-ion battery via deep learning techniques. A loss function is constructed from the spatiotemporal coordinates of sampling points within the solution domain, the overall structure of the system of partial differential equations, boundary conditions, and initial conditions. The distributions of stress and lithium concentration in a hollow-cylindrical nanoelectrode are obtained. The high degree of conformity between the numerical results and those from the finite element method is demonstrated.