A neural network approach for fault diagnosis of large-scale analogue circuits

Abstract

An approach for fault diagnosis of large-scale analogue circuits using neural networks is presented in the paper. This method is based on the fault dictionary technique, but it can deal with soft faults due to the robustness of neural networks. Because the neural networks can create the fault dictionary, memorize and verify it simultaneously, computation time is drastically reduced. Rather than dealing with the whole circuit directly, the proposed approach partitions a large-scale circuit into several small sub-circuits and then tests each sub-circuit using the neural network method. The principle and diagnosis procedure of the method are described. Two examples are given to illustrate the method for both small and large-scale circuits.