Fault diagnosis of circuits using statistical parameters and implementation using classifiers — A survey

Abstract

In this paper a survey on fault diagnosing techniques of electronic circuits are presented which are related mainly to industrial applications. Diagnozing the faults in circuit boards is very essential for achieving better reliability and easy maintainance of electronic systems. The circuit fault finding diagnosis is treated as the pattern recognition case and uses machine learning methodology. Increasing integration densities in advanced chips and high frequency of operation lead to subtle manifestation of faults in circuits of board level. Fault diagnosis by functional test is therefore necessary in place of structural test for board-level testing. Recent machine learning techniques use new algorithms to improve the success rate of functional fault diagnosis, reduces cost of product through successful servicing of circuits. Earlier methods of diagnosing fault using Artificial Neural Networks (ANNs) and Fuzzy logic were resulting in ambiguous or incorrect diagnostic results which take long time to debug and sometimes wrong repair actions, which significantly increase cost of servicing. Some key capabilities of these methods like capabilities of detecting, locating and identification of faults, the capabilities of detecting single or multiple faults, soft or hard faults, the capabilities of diagnosing linear or nonlinear circuits, etc are taken into account. Different classifications of fault diagnosis are reviewed. Various techniques for circuit fault diagnosis are also addressed in this paper. Survey on techniques of fault diagnosis provides a better insight for further research and application areas.