Wavelet analysis for diagnostic problems

In this paper the problem of off-line diagnosis of analog circuits affected by catastrophic multiple faults is considered. Information on the circuit behavior is given by the voltage measurements in a set of available test points. Because of the huge amount of data involved in each voltage acquisition, it is necessary to apply a compression technique to reduce the volume of the data to be given to the diagnostic system. Our diagnostic system consists of a neural network, trained to recognise catastrophic single faults, and used to diagnose multiple faults. The network has as many input nodes as the features extracted by each circuit image and the output nodes are as many as the number of circuit components. In this paper we investigate the suitability of the wavelet analysis to the problem of data compression of measurements for detecting faults in analog circuits. We also present some results obtained with the proposed diagnostic system in diagnosing multiple faults in the test circuit examined and compare these results with those previously obtained.