The manufacturing process of lithium ion battery (LIB) electrodes impacts their architecture and the practical properties of the cells, such as their energy and power densities, their durability and safety. Therefore, it is crucial to optimize this manufacturing process in a proper manner. Such an optimization is highly complex because of the very significant amount of parameters and numerous interdependencies between the process steps, such as the slurry mixing, the casting, the drying, the calendering, the assembly, the electrolyte filling and the solid electrolyte interphase formation. As a consequence, traditional trial and error approaches can lead to high scrap rates in LIB cell manufacturing.ARTISTIC is a digital twin for inverse design of LIB manufacturing process that we keep developing since several years.[1] ARTISTIC is supported on a combination of physics-based models and machine learning. The physics-based models simulate the dynamics, with 3D resolution, of the slurry mixing, the slurry coating and its drying, the resulting electrode calendering, the cell infiltration by the liquid electrolyte, the formation step and the resulting electrochemical performance of the virtually produced electrodes and cells. These physics-based models are linked with each other in a sequential manner, and they are supported on a combination of methods, such as Coarse Grained Particle Dynamics, Discrete Element Method, Lattice Boltzmann Method and Finite Element Method. Deep learning is used to derive time-dependent 3D surrogate models mimicking the behavior of the physics-based models with much less computational cost. Furthermore, a wide diversity of machine learning techniques are used to accelerate the calibration of the physics-based models with experimental observables (e.g. slurry viscosity and density, electrode porosity, tortuosity factor and conductivity) and to unravel correlations between manufacturing parameters and electrode and cell properties. All these pieces are integrated in a single computational infrastructure performing the inverse design. For that purpose, the deep learning surrogates and the machine learning models are coupled to a multi-objective Bayesian Optimizer. The latter allows predicting which manufacturing parameters to adopt (e.g. slurry formulation, drying rate, calendering degree) in order to reach desired properties for the electrodes (e.g. loading, porosity, conductivity) and for the cells (energy and power densities). We have demonstrated this pionneering digital twin for multiple manufacturing steps and electrode chemistries (e.g. NMC111, NMC622, LFP, graphite, silicon-graphite blends), with the strong support of experimental databases acquired in our LIB manufacturing pilot line.[2] We have also demonstrated the transferability of our approach to Sodium Ion and Solid State Batteries, and we start to adapt it for dry processing methods. In my talk, I will present the latest updates of this research and development effort. Furthermore, I will also present the latest updates of the ARTISTIC Online Calculator, a free service to perform battery manufacturing simulations from an internet browser, as well as of our Virtual and Mixed Reality tools to train students, researchers and operators in the battery manufacturing field. These updates will include the presentation of the first "battery manufacturing metaverse", a multi-user battery manufacturing pilot line in Virtual Reality designed to perform practices with the MSc. students of the i-MESC Erasmus+ programme (Interdisciplinarity in Materials for Energy Storage and Coversion).[3]References[1] https://www.erc-artistic.eu/[2] https://www.modeling-electrochemistry.com/publications[3] https://i-mesc.eu/
Read full abstract