AI and Machine Vision Integrated System for MLCC Defect Inspection

This paper presents an inspection system developed for the final stage of Multi-Layer Ceramic Capacitor (MLCC) production. It performs electrical capacitance measurement and visual inspection of bulk-supplied MLCC chips, separating defective units before packaging. To meet the required inspection rate of over 3,800 units per minute, a high-speed vision system (100 fps) and advanced image processing algorithms were implemented. Visual inspection uses strobe lighting and a hybrid method combining traditional feature extraction with AI to detect defects. The system identifies electrode length issues, chipping, and cracks with 98.98% accuracy. Tests on 3,245 non-defective and 543 defective samples showed a false positive rate of 0.65% and a false negative rate of 3.2%, with a processing speed of 3,875 units per minute. A multi-threaded architecture enables real-time inspection with high accuracy, making the system suitable for high-volume MLCC manufacturing.