The Fourth Industrial Revolution (or Industry 4.0) is the ongoing automation of traditional manufacturing processes in industry thanks to machines that can analyze and diagnose production quality performance without the need of human intervention. This smart automation is needed for efficient and autonomous management of massive production systems such as the ones found in lithium ion battery giga-factories.

This PhD project aims at developing and demonstrating a novel computational approach for quasi-real-time physics-informed simulation of the battery electrode manufacturing process and of the influence of its parameters on the electrode architecture, paving the way to a concrete Battery Industry 4.0. This approach will be supported on surrogate models powered by machine learning ML methods, able to mimic the results of physics-based manufacturing process models. The surrogate models will be trained on data from the existing LRCS physics-based manufacturing models. This approach takes inspiration from a recent LRCS work reporting a 3D-resolved ML model able to correctly predict in few seconds the complex 3D trajectory of electrolyte upon its impregnation in electrodes (to be compared with the several days needed when a physics-based Lattice Boltzmann Method is used). The computational performance and prediction accuracy of the surrogate models will be assessed through their real-time usage along battery manufacturing processes in the pilot line of CIDETEC. The quasi-real time calculation capabilities of these models will also permit their integration in Bayesian Optimization (BO) algorithms allowing to automatize in quasi-real time the optimization of the manufacturing parameters as function of the desired electrode properties. The performance of these BO algorithms will be evaluated thanks to experimentations carried out in the above-mentioned pilot line and in the one available at LRCS. The primary location of the PhD student will be LRCS (registration in the Université de Picardie Jules Verne) for the development of the computational aspects. The PhD student will expend 3 periods of 3 months in CIDETEC to test and improve the surrogate models in real-usage conditions at the pilot line level.

Supervisor(s) contact: FRANCO, Alejandro A. alejandro.franco@u-picardie.fr & ASCH, Mark mark.asch@u-picardie.fr