Research on Fault Diagnosis Approach of Analog Circuit Based on Neural Network

According to circuit decomposition technology and crossover tearing technology, a fast approach of module level fault diagnosis for analog circuit based on BPNN (back propagation neural network) is presented. Because the neural networks can create the fault dictionary, memorize and verify it simultaneously, computation time is drastically reduced. The new approach is based on parallel diagnosis, so fault modules can be located quickly and effectively. Two examples are given to illustrate the approach for both small and large-scale circuits.