Two-center Radial Basis Function Network For Classification of Soft Faults in Electronic Analog Circuits

In this paper, a new neural network architecture with two-center radial basis functions (TCRB functions, TCRBF) in the hidden layer was presented. The special shape of TCRB function was introduced to enhance the efficiency of soft faults classification in electronic analog circuits. The application of TCRB functions in neural network classifier gives possibility to reduce the number of neurons in its hidden layer in comparison to radial basis function network with Gaussian basis functions. Additionally there is an improvement in testability of circuit under test (CUT) through decreasing the classification error. This article shows results obtained for lowpass 2^th order filter.