A Combined Method for Analog Circuit Fault Diagnosis Based on Dependence Matrices and Intelligent Classifiers

This paper presents a novel combined method for analog circuit fault diagnosis based on dependence matrices and intelligent classifiers (DM-IC). The main purpose of this method is to obtain preliminary ambiguity groups by constructing a dependence matrix, to achieve a more accurate diagnostic result by using several simple intelligent classifiers. Considering the presence of imperfect test points, a new 0-1-X dependence matrix is defined and an expansion-reduction algorithm is proposed to obtain preliminary ambiguity groups. For each of these preliminary ambiguity groups, the algorithm of separability discrimination analysis based on Fisher discriminant analysis is proposed to determine whether an ambiguity group can be further classified and how many classes should be classified. Furthermore, a new test-point optimization algorithm is proposed to simplify the input nodes, and the classifier is constructed for each separable ambiguity group. This combined method can break down a difficult classification problem into several simpler classification problems. In this way, only a small number of classifiers need to be constructed, and at most one classifier is required to participate in the diagnosis process. Our experimental results clearly demonstrate the superiority of the DM-IC method, which has a 100% correct classification of fault classes, fewer ambiguity groups, and smaller ambiguity groups in the sample circuit.