| Title |
Development of an AI Model for Quality Prediction and Root Cause Tracing of Cathode Materials Based on FeatureMAP Image Analysis |
| Authors |
송정윤(Jeong-Yun Song) ; 김수림(Soo-Rim Kim) ; 오현우(Hyun-Woo Oh) ; 이용귀(Yong-Kwi Lee) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.8.127 |
| Keywords |
Cathode materials for secondary batteries; Feature map; Recurrence plot; EfficientNetV2; Grad-CAM |
| Abstract |
The secondary battery cathode material production process faces limitations, as repetitive trial runs following post-production defect identification lead to significant losses of expensive raw materials and time. Although data from raw materials, manufacturing processes, and final quality are sequentially accumulated, their disparate occurrence times and formats make it difficult to identify integrated correlations between production conditions and quality. In this study, we propose a FeatureMAP structure that visualizes process time-series data using the Recurrence Plot technique and develop an AI model for quality grade prediction using the EfficientNetV2 algorithm. Specifically, a sliding window technique was applied to preserve the temporal continuity of the process while effectively augmenting limited manufacturing data. Furthermore, we ensured the traceability of defect causes by identifying key factors influencing quality through Grad-CAM.A total of 16,341 images were utilized for the experiment, through which the proposed model achieved a high classification accuracy of 95.07%. |