Mixed Integer Programming Method for Fault Diagnosis of Linear Analogue Circuits

Introduction: There are broadly two approaches for analogue fault diagnosis [1, 4, 5]. They are the fault analysis approach or simulation-after-test, and the fault dictionary approach or simulation-before-test. In this Letter, we are concerned with the first approach. In one kind of fault analysis approach called fault verification, equations of the nominal network are obtained. If k or less elements or nodes are faulty (a node is said to be faulty if at least one of the elements connected to it is faulty), then all submatrices of order k and above of the equations of the network are to be tested for singularity. From this information, faults can be identified. In general, this is time consuming as all possible submatrices have to be tested. Moreover, this is a brute force method. We formulate it as a 0-1 mixed integer programming problem which can be solved efficiently in a systematic way.