Selection of input stimulus for fault diagnosis of analog circuits using ARMA model

The paper addresses the problem of fault diagnosis of analog circuits based on dictionary approach. The proposed approach first identifies an adequate set of test frequencies to optimize the process of detection and isolation of simulated fault scenarios. The circuit under test (CUT) is then excited by an input stimulus composed of a set of sinusoidal waveforms with the selected test frequencies. The circuit response, at different fault scenarios, is preprocessed by an autoregressive moving average (ARMA) model to yield a set of features formulating the fault dictionary. Collected features are utilized to train and test a back-propagation (BP) neural network (NN) based classifier. Demonstrative results from soft fault simulation of two active circuit examples prove the excellent effectiveness of the proposed algorithm.