Exploring the impact of coating factors on lithium-ion battery quality

The Nextrode project has developed data driven models for a comprehensive explanation of control factor impact in the manufacture of lithium-ion battery electrodes and cells. This is an important step towards transparent manufacturing and a powder to power digital twin. The team has combined experimental data, domain knowledge and artificial intelligence to understand the impact of slurry factors such as density, viscosity, surface tension, solid weight percent as well as coating process control factors such as coating speed and coating gap on characteristics of electrodes and cells.

In particular, novel machine learning models with explanation capabilities have been developed by researchers at the Universities of Warwick and Birmingham. A customised design of experiments has been performed to generate a rich dataset that is optimal for a machine learning model. Ensemble models with cross validation and explanations empowered by techniques such as Shapley Additive Explanations and Accumulated Local Effects have been trained for prediction, correlation analysis and extrapolations.

The researchers have concluded that given the slurry and coating process characteristics, the electrode and cell features can be predicted with an accuracy up to 89%. Coating gap is the strongest contributor (70%) to the definition of the thickness, porosity, and energy capacity and this is followed by viscosity (20%). The approach has facilitated a systematic way of quantifying the impact of the factors on the responses. Based on the results battery manufacturers will now be able to predict more accurately the performance of cells by monitoring a small number of manufacturing variables. The research was published in Energy Storage Materials.

Image: Pilot-scale cathode coater of WMG, University of Warwick. Photo courtesy of WMG.

Case study published December 2023.