Fault diagnosis of a logical circuit by use of a pseudorandom signal and a neural network

This paper describes a new method of pseudorandom testing of a digital circuit by use of a correlation method and a neural network. The authors have recently proposed a new method of fault diagnosis in a logical circuit by applying a pseudorandom M-sequence to the circuit under test, calculating the cross-correlation function between the input and the output, and comparing the cross-correlation functions with the references. This method, called the M-sequence correlation (MSEC) method, is further extended by using a neural network in order not only to detect the existence of faults, but also to find the place or location of the faults. The authors investigated the effects of using parts of the fault patterns to train the neural network to be able to detect faults. It is shown that more than 95% of faults can be detected even when only 60% of the possible training data are used.