Analog fault feature extraction and classification based on LMD and LVQ neural network

This research paper proposes a new fault diagnostic system for analog circuits. Local Mean Decomposition (LMD) is used to decompose every faulty response into a set of product function (PF)signals, fault feature vector of the fault is built by the calculation of energy and kurtosis of each PF. Adopting the same method, all the fault feature vectors are obtained. Then, these vectors are trained in a learning vector quantization (LVQ) neural network. Via validating the performance of diagnosis with a benchmark circuit and comparing with other fault diagnostic methods, consequences prove that this proposed system has a higher testing precision for analog circuits.