Surrogate Modelling of Lithium-Ion Battery Electrode Manufacturing by Combining Physics-based Simulation and Deep Learning

Optimization of the manufacturing process of Lithium-Ion Batteries (LIB) is crucial for advancing energy storage systems, and finding efficient approaches for its acceleration is a key area of research. Such approaches should be able to replace time-consuming and material scrap-expensive trials-and-error optimization methods. This work presents a comprehensive LIB electrode manufacturing framework that combines physics-based simulations with Deep Learning for simulating computationally and in an efficient manner the manufacturing process of LIB electrodes as a function of their formulation. This framework takes the form of a surrogate manufacturing model able to predict the impact of manufacturing parameters on the electrode architectures. The model is based on a regressor inspired Variational Autoencoder (VAE) method. The analysis of the data and predicted electrode functional metrics demonstrate the consistency of our approach with an electrode manufacturing model developed on the basis of physics. The reported framework holds significant promise in paving near real time optimization of LIB electrode manufacturing, supporting the optimization of battery cell design in pilot lines.