New concept to analog fault diagnosis by creating two fuzzy-neutral dictionaries test

This paper presents research on analog fault detection and location. The problem of distinguishing between healthy and faulty analog circuits has always been very complicated. Moreover, if a test locates a faulty element, a great problem occurs. The most common approach based on pattern recognition, especially on mean square error measure, cannot distinguish all faulty circuits from healthy ones. Normally, the dictionary has to include thousands of patterns and even then, the level of fault detection/location is not satisfactory. A neutral network classifier has been proposed to solve the problem. Its generalization ability allows reducing the dictionary size significantly. This paper shows how to create a neutral dictionary test for detection and location. Moreover, at the first stage of classification, the fuzzy logic is utilized to transform a measurement vector into a zero - one range (a continuous value). Output of a neutral network is coded. It reduces the number of output neutrons.