An intelligent detection approach for end-of-life power battery shell bolts

Abstract

With the rapid growth of the new energy vehicle industry, the number of end-of-life power batteries, which serve as the technological core, is also increasing significantly. Unfortunately, this rise in retired power batteries has led to severe environmental pollution and resource wastage. The detection of shell bolts in power batteries has thus become a crucial step in the recycling and disassembly process. To address this issue, this research proposes a detection method for end-of-life power battery shell bolts. Based on market analysis, the target bolt for the retired power battery shell was identified. The bolt images were collected and preprocessed to create a custom dataset on the experimental platform. Four popular object detection algorithms were compared, and the YOLOv8 model is selected to improve with EMA module. The improved YOLOv8 model achieves 98.9% for mAP_0.5, which increases more than 2 percentage points. Based on the repeatability of bolt recognition, this detection method can be used for the identification of bolts in other battery shells, providing a theoretical foundation for promoting the robotic disassembly of battery shells.