Self-organizing neural network for fault location in electrical circuits

A novel approach to fault location in analog dynamic circuits based on the application of self-organizing neural network has been presented. Important features are a very good generalization property and fast speed. Once the network has been trained, the recognition of the fault is done immediately, irrespective of the size of the circuit. The network is able to detect faults in the nonideal circuit, in which the tolerance of elements is taken into account. Two cases of analog circuits have been simulated and checked: the RLC circuit at multiple measurement points and the measurement done at external nodes of the circuit for multiple frequencies. The results of numerical experiments are given and discussed.