Neural Network Based Algorithm of Soft Fault Diagnosis in Analog Electronic Circuits

摘要

This paper addresses the fault diagnosis in analog electronic circuits based on neural network. Iterative algorithm of solving the system of diagnostic equations, using integral sensitivity matrix is also proposed in this paper. To obtain the system of diagnostic equations, the integral sensitivity matrix is used. The integral sensitivity matrix is constructed based on a result of the comparisons of signals measured at the nodes of circuit under test (CUT) and appropriate signals of circuit with nominal parameter values. The response of the circuit with nominal values of parameters is recorded and trained using a feed forward neural network. The proposed method has capability to detect single and multiple faulty components, also it is applicable to a wide class of electronic devices and can be integrated into modern computerized diagnostic tools.