Improving X-ray inspection of printed circuit boards by integration of neural network classifiers

摘要

For six sigma quality in printed circuit board (PCB)-production, X-ray inspection of solder joints is a powerful method to assure a high standard in fabrication. Neural network classifiers are able to adapt inspection tasks by presentation of typical training patterns. Neural networks are integrated into a X-ray inspection system both to increase defect recognition accuracy, as well as to minimize manual adjustments of the system. The experiments carried out on different surface mount technology (SMT) device types prove the capability of neural-network-based approaches to correctly segment objects (solder joints etc.), and to detect defects (solder voids etc.).