Abstract
This Ph.D. thesis presents the results we obtained during the last three years in the context of Data Science, and especially in the field of Artificial Intelligence, supporting the optimization of the lithium-ion battery manufacturing process and cell performance. The global work was funded by ALISTORE-ERI, a network that federates institutions to perform cross-cutting high-level research in the fields of batteries. The thesis was co-shared between the Laboratoire de Réactivité et Chimie des Solides (LRCS) (Amiens, France) and CIDETEC Energy Storage (San Sebastian, Spain). Our work focused on the development of computational approaches to accelerate the identification of manufacturing parameters' effect on battery performances, electrode properties forecasting, and the inverse design for manufacturing parameters optimization. All generated data used in the studies came mainly from two sides: the European Research Council-funded ARTISTIC project led by Prof. Alejandro A. Franco at LRCS, which offered synthetic data generated by physics-based models and also experimental datasets, and CIDETEC Energy Storage, a research organization specialized in advanced battery technologies, which provided also experimental data. All together, these data allowed to explore some parts of the manufacturing parameters space for the study on how parameters affect the electrode properties, giving insights for the development of predictive modeling supported by Artificial Intelligence techniques. Those models were found relevant for setting up optimization processes to obtain results based on the specific various battery applications. Lastly, the so-generated data enabled to develop a database storage architecture for simulations to optimize the data management of the ARTISTIC Project. Moreover, such data correspond to various chemistry materials studied such as Nickel Manganese Cobalt Oxide (LiNi1−x−xMnxCoxO2 (NMC111, NMC622, NMC811)) and Graphite, and Lithium-Iron-Phosphate (LiFePO 4 ). In the end, the overall work presented intended to reduce the gap between experiments and simulations where results obtained by modeling were analyzed through experimental characterizations. Because this thesis targets various studies based on a specific part of the manufacturing parameters space, it harvested proofs of concept that can be transferred to other complex studies regarding battery applications