Aspects of PCB automated quality control using potential function based clustering algorithms

Abstract

Video monitoring and inspection of printed circuit boards (PCBs) in manufacturing process is a complex task, being affected by different types of disturbances. In quality control, an automated system based on video inspection should be able to identify different states of good or faulty PCBs. Copper tracks have complex shapes, making analyzing process more difficult. In addition, clustering techniques must deal with unwanted copper material between conductive tracks, which affects dynamic behavior of the circuit. In this paper, clustering algorithms based on potential functions are used in automated video inspection of PCBs copper tracks. Simulation results are presented for two oblong clusters of copper tracks, which are not well separated and difficult to classify. Two situations are discussed, regarding a good PCB and a faulty one, with some copper material between the tracks, which is close enough but contactless with the tracks. Potential function based algorithms (PFBA) are able to identify the complex shapes of copper tracks in good PCBs. Also, they work well in heavy classifying process of the unwanted copper clusters placed very close to conductive tracks.