Fault Diagnosis of Analog Circuit Based on Multi-layer Neural Networks

The theory of the fault diagnosis of analog circuit is a new offshoot of the theory of circuit networks. In this paper, the neural network is used in the fault diagnosis of analog circuit to improve the adaptive capacity. This makes the way of the directory be of use in fault, and enhances the validity of the fault diagnosis. Simulation results have shown that this claim is valid. Results indicated that the structure of BP network influents not only training process, but also assorting effecting. Overfull or fewer more neurons in hidden-layer would also reduce order of accuracy. On the condition of same training sample, two-hidden-layer is more quickly than single hidden-layer, but sometimes there would be agitation in network with two-hidden-layer. When being trained, network must have proper performance target.