Analog Circuit Fault Diagnosis Based On HHT And SVDD

To address the probrem of ineffectiveness and low recognition rate in analog circuit fault diagnosis, this paper proposes a new fault diagnosis model. Combined the feature extraction with SVDD(support vector data description) classifier to classify circuit faults,then locate the faults of components. This paper did the following research:(1) To solve the problem of nonlinearity, non-stationary and poor component tolerances in analog circuit fault detection, HHT(Hilbert Huang Transform) is proposed to process this type of signal. Aiming at the problems of boundary effect 、illusive component and mode mixing caused by HHT, endpoint continuation method, correlation coefficient method, Ensemble Empirical Mode Decomposition algorithm are proposed to improve this appearance.Then get the energy of hilbert spectrum and marginal spectrum through improved HHT to indicate different circuit faults.(2) Different kernels function, kernel parameters and punishment coefficient contribution to different performance of SVDD. Find the optimization of that and combined kernel functions for SVDD classifier.This paper use the improved HHT to extract feature vectors, some of them input into SVDD for training, some for detection. By the way of simulation, the method has a higher correct rate of diagnosis than normal.(3) A method based on OEMD and SVDD classifier optimized by Kernel Fuzzy C means(KFPCM) is proposed. It will get the intrinsic mode functions with strict orthogonality in feature extraction.KFPCM algorithm is used for preclustering, traversing the feature vectors to data of high similarity, and reducing the impact of noise, then input the sample into optimized SVDD classifier for training and diagnosis.