Application of artificial intelligence techniques to probeless fault diagnosis of printed circuit boards

This thesis describes investigations which led to the development of a failure diagnosis expert system for printed circuit boards which exploits functional test data. The boards considered are highly complex mixed signal (analogue and digital) systems. The data is output from automatic test equipment which is used to test every board subsequent to manufacture.The use of a conventional machine learning technique produced only limited success due to the very large search space of failure reports. This also ruled out the use of some conventional knowledge-based approaches. In addition, there was a requirement to track changes m printed circuit board design and manufacture which also ruled out some techniques.Our investigations lead to the development of a system which tracks changes by learning in a more restricted search space derived from the original space of reports. The system performs a diagnosis by matching a failure report with information about previously seen reports. Both exact and inexact matching were investigated. The matching rules used are heuristic. The system also uses basic circuit connectivity information in conjunction with the matching procedure to improve diagnostic performance especially in cases where matching fails to identify a unique component