Research Of Inspection Method On Solder Joints Quality Based On BP Network

Electronic device is developing toward lighter, thinner and littler volume, which demands highly on the density and functionality of package I/Os. So it is important factor to assure the quality of the solder joint. Recently it becomes a hot point to find the defects of the solder joint in time and effectively. Inspection methods on solder joints quality include two types: destructive testing and non-destructive testing. Because the destructive testing damages the electronic device to some extent, so it is not fit for carrying out in inspecting vast products. Therefore it is an exigent requirement for electronic industry to design some advanced non-destructive testing methods and technologies nowadays.Inspection method on solder joints quality based on the structured illumination is adverted extensively as a kind of non-destructive testing recently. In the paper we research the inspection method on solder joints quality based on the structured illumination. There are two hard problems in solder joints inspection based on structured illumination. One is how to extract features of solder joints effectively, and the other is how to classify the features of solder joints effectively. The following methods are proposed to solve the two problems.Firstly, we improve the algorithm for fast convergence of back propagation by enlarging error in the research of BP network. The result of emulational experiment shows that the algorithms are effective. The algorithm solves the problems of too much training and the low precision of BP network applied in solder joints inspection.Secondly, features extracting algorithm based on histogram, features extracting algorithm based on shape and features extracting algorithm based on texture are imported to solve the drawback of traditional features extracting algorithms based on structured illumination with a layer circle that make use of image information deficiently. We extract abundant features information. The result of experiment shows that the algorithms are effective, and abundant features information is extracted.Finally, BP network is imported in the inspection of solder joints quality owing to the good performance of pattern recognizing, generalization and self-adapter in contrast with traditional inspection algorithms. Combining improved error enlarged...