Digitalization of Battery Manufacturing through Artificial Intelligence and Multiscale Modeling
Prof. Dr. Alejandro A. Franco1,2,3,4
1 Laboratoire de Réactivité et Chimie des Solides (LRCS), CNRS UMR 7314, Université de Picardie Jules Verne, Hub de l’Energie, 15 Rue Baudelocque, 80039 Amiens, France
2 Réseau sur le Stockage Electrochimique de l'Energie (RS2E), Fédération de Recherche CNRS 3459, Hub de l’Energie, 15 Rue Baudelocque, 80039 Amiens, France
3 ALISTORE-European Research Institute, Fédération de Recherche CNRS 3104, Hub de l’Energie, 15 Rue Baudelocque, 80039 Amiens, France
4 Institut Universitaire de France, 103 Boulevard Saint-Michel, 75005 Paris, France
The needed massive deployment of lithium ion batteries (LIBs), in particular to satisfy the demand from the Electric Vehicle sector, encourage battery manufacturers to multiply the number of giga-factories to reduce the cost of production. Such a production consists of a complex process involving multiple steps, such as the slurry preparation, its coating, drying, calendering, electrolyte filling and formation. The choice of the manufacturing parameters along the process strongly impact the overall LIB cell performance. However, the optimization of the manufacturing parameters to obtain the desired characteristics of LIB cells is currently based on a forward "trial and error" approach. This approach is inefficient in terms of time and cost due to the infinite number of possibilities for adjusting the manufacturing parameters. The integration of Industry 4.0 concepts into giga-factories such as Artificial Intelligence (AI) and multiscale modeling is essential to accelerate the manufacturing process optimization and significantly raise the work efficiency of scientists, engineers and production line operators. Here I present our efforts in developing a digital twin of battery manufacturing , supported on a combination of physical and AI/machine learning models trained with data arising from in house high throughput characterizations. This digital twin allows simulating the different steps along the manufacturing process, predicting the resulting electrode properties and evaluating the associated electrochemical performance upon battery cell cycling. The predictive capabilities of this digital twin are illustrated with results for different electrode formulations, paving the way towards manufacturing digital optimization. Finally, data standardization challenges to ease the wide use of these tools in the battery field will be briefly discussed.