Multiphysics-Informed Machine Learning for Capacity Degradation of Silicon Anode

Anodic materials such as silicon (Si) are promising to further increase the storage capacity and performance of existing Lithium Ion Batteries (LIBs). However, the alloying lithiation/delithiation mechanism of Si can lead to substantial increase in the volume of the silicon during the charging cycles, which induce failures and capacity degradation of the LIBs. The volumetric change can result in large stresses being developed within anode materials, which can cause the formation of cracks in the anode and delamination of the anode from the metal substrate. Coupling with mechanical failure of anode, Solid-Electrolyte Interface (SEI) grows and causes further capacity fade in the battery. In this study, we will develop physics-informed machine learning model for investigating the coupled failure modes of Si anode on the capacity degradation of silicon anode under various operation conditions. Multi-physics-based 3D FE models are built to explore the coupling effects of mechanical damage and SEI layer growth in Si anode on the Ni substrate. The effect of SEI layer will be investigated through coupling electrochemical simulation for SEI growth with the solid mechanics FE simulation for internal crack and delamination. Based on the FE simulation results, multiple Gaussian Process Regression (GPR) models will be developed, which combine to estimate the State of Health (SoH) under the interactive effects of the three failure modes in the Si anode.