The recycling of lithium-ion batteries involves mechanical processing at different stages. After the feedstock has been shredded and dried, the crushed particles are sorted into fractions based on particle size and density. Our goal is to develop an optimized particle separation system supported by a digital inline sensor unit. By digitally monitoring the particle properties during the mechanical sorting, the air classifier can be regulated in real-time to achieve an improved separation of valuable copper and aluminum electrode fractions. The challenges are the real-time data processing and the fast 3D movement of the particles in the airstream.
At a lab-scale test setup, a Matlab-based program for image segmentation and classification of the individual particles at the classifier was developed. The workflow succeeded to separate background from particles as well as to differentiate particles based on color thresholds. Additional infrared reflectance and photoluminescence imaging further enhanced the detectability and distinguishability. Previously confused aluminum and plastic particles revealed characteristic features in both their reflectance properties as well as after excitation by UV light which can be detected spectroscopically. Our evaluation showed that the detection of separator foils and other plastic particles was about 80 % more accurate when UV light excitation was used.
After successfully completed test measurements, the lab setup was transferred to a zig-zag air classifier including an industry camera system with homogeneous illumination. The algorithms were tested under industrial conditions and used to control and optimize the mechanical sorting process by adjusting the operation parameters of the air classifier.
This work is financially supported by Bundesministerium für Bildung und Forschung within the project DIGISORT as part of the GreenBatt cluster under grant numbers 03XP0337A, 03XP0337B, 03XP0337C.