As EV cells get larger and more energy dense, attaining uniform electrolyte saturation still remains a challenge both during cell development as well as during series production. In collaboration with battery manufacturing partners, Liminal has demonstrated how our customers can use ultrasound inspection and machine learning analytics to detect poorly saturated cells during the electrolyte soaking process in seconds per cell, instead of waiting until after formation or cycle testing. Additionally, with data collected post-formation, we used physics-assisted machine learning (ML) methods to develop models that predict cycle life performance with over 85% accuracy, a 25% improvement over what was achieved through similar ML methods using only electro-chemical metrics. Assessing saturation and cell quality, and detecting outliers early can save cell manufacturers ~$4.50/kWh, based on our TEA models. Ultrasound measurements are non-invasive and can be implemented with >10 ppm (parts per minute) throughput, enabling cell manufacturers to inspect 100% of cells during production, quickly ensuring their quality and reliability.
Moreover, as Liminal's ultrasound inspection can provide real-time cell quality assessment, battery scientists can iterate through multivariate design-of-experiments (DOE) in days as opposed to waiting months for long-term cycling results. Rapid assessment of the impact of material or process changes on electrolyte soaking time, saturation quality, and formation protocol development can reduce the time and cost to scale new materials, designs, and processes from R&D to production