A physics-induced surrogate modeling framework for lithium-ion battery electrode manufacturing

The manufacturing of lithium-ion battery electrodes is a complex, multi-step process. The selection of manufacturing parameters at each stage directly affects the final microstructure of the electrode and, consequently, the electrochemical performance of the cell. Here, we demonstrate the application of our Physics-Induced Electrode (PIE) framework, comprising a set of Machine Learning (ML) models that represent key manufacturing steps: slurry, drying, and calendering. For each step, the ML model uses the coordinates of the particles (representing the active material and carbon-binder aggregates at the mesoscale), along with geometrical descriptors with physical properties, to predict the resulting particle displacement. The model predictions were evaluated using statistical metrics and were found to be significantly accurate. The predicted electrode microstructures were further assessed in terms of their properties and exhibited good agreement with those obtained from purely physics-based simulations. To enhance interpretability, SHapley Additive exPlanations (SHAP) were used to analyze the contribution of the input features to the model predictions. The SHAP analysis showed that the ML predictions capture trends consistent with physical intuition for the manufacturing step analyzed. The PIE framework offers a scalable and interpretable approach that can be integrated into digital solutions to support battery electrode manufacturing pilot lines.