The objective of this thesis is to inspect solder joints on SMT(surface mounting technology) board rapidly and correctly. As the classification of defects, there are overflow, poor wedge, cold joint, floating, bridging, near short etc. Conventional method has some disadvantage that the inspection time is long and the types of inspection are limited. In this thesis for improving these problems, the inspection is done by injecting the laser beam to solder joint and analyzing the pattern of the reflected laser beam on sensor box. In order to classify the reflected pattern, the self-organizing neural network is used. All procedures are simulated on a computer and the success ratio of correct classification is above 80 percents.
