Systematic analysis of the impact of slurry coating on manufacture of Li-ion battery electrodes via explainable machine learning

• Main Authors: Faraji Niri, M. (Author), Kendrick, E. (Author), Marco, J. (Author), Reynolds, C. (Author), Román Ramírez, L.A (Author)
• Format: Article
• Language: English
• Published: Elsevier B.V. 2022
• Subjects: Coatings; Electrochemical electrodes; Electrode coating; Electrode manufacturing; Explainable machine learning; Ions; Learning algorithms; Li-ion battery; Li-ion battery electrodes; Lithium-ion batteries; Machine learning; Machine-learning; Manufacture; Manufacturing process; Optimised manufacturing; Optimized manufacturing; Slurry coating; Slurry mix; Systematic analysis
• Online Access: View Fulltext in Publisher

 The manufacturing process strongly affects the electrochemical properties and performance of lithium-ion batteries. In particular, the flow of electrode slurry during the coating process is key to the final electrode properties and hence the characteristics of lithium-ion cells, however it is given little consideration. In this paper the effect of slurry structure is studied through the physical and rheological properties and their impact on the final electrode characteristics, for a graphite anode. As quantifying the impact of the large number of interconnected control variables on the electrode is a challenging task via traditional trial-and-error approaches, an explainable machine learning methodology as well as a systematic statistical analysis method is proposed for comprehensive assessments. The analysis is based upon an experimental dataset in lab-scale involving 9 main factors and 6 interest variables which cover practical range of variables through various combinations. While the predictability of response variables is evaluated via linear and nonlinear models, complementary techniques are utilised for variables importance, contribution, and first and second order effects to increase the model transparency. While coating gap is identified as the most influential factor for all considered responses, other subtle relationships are also extracted, highlighting that dimensionless numbers can serve as strong predictors for models. The impact of slurry viscosity and surface tension on electrode thickness, coat weight and porosity are also extracted, demonstrating their importance for electrode quality. These variables have been rarely considered in previous works, as the relationships are difficult to extract by trial and error due to interdependencies. Here we demonstrate how model-based analysis can overcome these difficulties and pave the way towards an optimised electrode manufacturing process of next generation Lithium-ion batteries. © 2022