A global parametric faults diagnosis with the use of artificial neural networks

A method of a global parametric faults diagnosis in analogue integrated circuits is presented in this paper. The method is based on basic features calculated from a circuit´s under test time domain response to a voltage step, i.e. locations of maxima and minima of circuit under test response and its first order derivative. The testing and diagnosis process is executed with the use of an artificial neural network. The neural network is supplied with extracted basic features. After evaluation and discrimination, the neural network outputs indicate the circuit state. The proposed diagnosis method has been verified with the use of exemplary integrated circuits - an operation amplifier muA741 and an integrated band-pass filter.