Using Physics‐Informed Neural Networks to Predict the Damage in Single Particle Models for Li‐Ion Batteries Active Materials

In order to ensure the longevity of active materials, microscale processes that occur during lithiation and delithiation must be better understood. These processes include chemical diffusion, the mechanical equilibrium of the host material, and fatigue phenomena in the material, which result in damage. Since experimental investigation of these individual but strongly intercoupled effects is hardly possible given the current state of research, this work presents the mathematical, theoretical equations for a 1D single-particle model with their mutual feedbacks and a phase field approach for the damage calculation. To solve the resulting system of equations, a Physics-Informed Neural Network (PINN) is used, which combines the advantages of Artificial Neural Networks (ANN) of fast execution and grid freedom with the advantages of conventional numerical methods such as interpretability and a flexible handling of boundary conditions. It is shown that a PINN is capable of learning the complex coupled governing differential equations. NMC622 is investigated as the active material on the cathode side. It was found that during lithiation with a stress-free surface, notably high currents lead to damage in the particle. ... mehrThis damage then occurs predominantly in the center of the particle and is mainly favored by a high level of lithiation.