Fault Diagnosis of a Logical Circuit by Use of M-Sequence Correlation with Neural Network

Abstract

The author has studied the fault diagnosis of a logical circuit (board) by use of M-sequence correlation.
In this paper, the author proposes a new method for the diagnosis and repair of a faulty logic board by use of M-sequence correlation with Neural Network (NN). Let us denote the position in the circuit fault may occur as the fault candidates, and assume that the faults occur only there. In this method, the fault position is estimated as the fault candidates which has more than 50% fault occurrence ratio. This method by use of NN has the advantage that the teaching data can be generated from only a part of faulty logic circuit, which means teaching data can be generated without question even if the number of fault candidates increases. Simulations were carried out for this fault repairing method and shown as follows: 1. It is possible to repair arbitrary faulty logic boards with high probability by use of M-sequence correlation and NN even if the teaching data of NN was calculated from only a part of faulty logic boards. 2. This method has great ability in repair, even if the estimation error of fault probabilities is not small due to a few learning iterations of NN.
Therefore this fault diagnosis method can be very useful for actual faulty logic boards.