07/07/2026
By May 2026, partners in charge of the sorting and discharging tests received diverse electric vehicle (EV) and domestic battery samples (covering around 3 cell formats and 8 chemistries). In parallel, shipments are on their way to feed dismantling tests and black mass production. Knowing what type of battery is entering the processing line is essential for safety, efficiency and material recovery.
LEITAT has been developing an AI-supported approach to identify and classify batteries according to relevant characteristics such as format and chemistry. The sorting model is trained on an image dataset, which is expected to grow progressively to more than 16,000 images.
Early results show ~70% accuracy on unseen test data, the model proving capacity to distinguish major battery chemistries. Simultaneously, partners are building a detailed materials library to improve the model’s capacity to differentiate structurally similar Li-ion chemistries. In the coming months, the range of chemistries and battery brands will be expanded, including further work on battery casing identification, particularly for domestic cells.
In parallel, GPA Innova/DLYTE have been working on safe discharging methods for domestic batteries. They have presented a baseline experimental set-up, which allows voltage and temperature monitoring during discharge, as well as the integration of a cooling system to improve safety. Initial tests using a Li-ion battery (~70% charged) confirmed the feasibility of their discharging approach, safely reducing its charge to ~0.5 V in about 1.5 hours.
© photo: Adobe Stock Photos
