Fuzzy logic-based filtering for defect detection in automated X-ray inspection during the production of lithium-ion battery pouch cells

This paper presents a fuzzy logic-based filtering method for improving defect detection in two-dimensional (2D) automated X-ray inspection (AXI), a nondestructive imaging technique widely used in the inline quality control of lithium-ion batteries (LIBs) during manufacturing. Enhancing the accuracy of defect detection in 2D AXI is crucial for ensuring battery safety; hence, the proposed method combines the outputs of three conventional detection algorithms with a rule-based fuzzy filtering strategy to generate robust defect maps. Experiments were conducted on various types of defects in LIB pouch cells to demonstrate the feasibility of the algorithm. The proposed method significantly improved the detection accuracy, with processing times meeting the strict requirements for inline inspection, as confirmed by qualitative and quantitative evaluations. Specifically, the method achieved a Dice similarity coefficient of 0.98 and an intersection over union of 0.94 for globular foreign material detection, representing improvements of ∼4.3 and 6.8%, respectively, over conventional algorithms. Furthermore, it achieved an average inference speed of 0.84 frames per second, which is 5.3 times faster than a recent comparative approach. Thus, the proposed method offers a high detection accuracy and practical efficiency, rendering it suitable for industrial deployment in automated inspection systems.