Research On Contactor Fault Recognition Based On Electromagnetic Signals

As a kind of important low-voltage electrical equipment, especially used widely in the power system, the contactor lays the irreplaceable position. The contactor is always in a poor working environment when used, which affects the normal operation and life of the contactor. And that also affects the social production and human life. In particular, the contact fault occurs frequently. Therefore, this paper identifies the fault of contactor contact parts. The traditional contactor fault identification mainly depends on artificial periodic inspection, which has low efficiency and poor real time. In this paper, analyze the contactor electromagnetic signal in order to recognize different types of faults. According to the characteristics of electromagnetic signal, apply wavelet transform and Hilbert-Huang for feature extraction, and combine the simulation signals and real signals to verify the validity. The recognition process includes signal denoising, contactor electromagnetic signal feature vector extraction, and finally the pattern recognition work. The main contents of this paper include:(1) Analysis the contactor fault classification and the electromagnetic signal, including the introduction of the contactor classification and structure, summarying common faults and processing method, and studying the production of the contactor arc and the contactor electromagnetic switch signal.(2) The wavelet packet transform is applied to the feature extraction of the contactor electromagnetic signal. Firstly, wavelet packet threshold denoising method is used to remove the noise from the signal, and study the effect of threshold and threshold function on the denoising, and improve the threshold function. Extract wavelet packet energy distance as the feature vector of the signal; calculate the energy time-frequency matrixes based on the wavelet, and extract singular value of the matrix as the feature vector, which is as input to identify the fault.(3) The Hilbert-Huang is applied to extract the feature vector of the contactor electromagnetic signal. The effective method is adopted to suppress the mode mixing and the end effect in the decomposition process to ensure the validity of the signal decomposition. Extract the multi-domain feature of the signal. And extract the IMF singular spectrum entropy, the IMF energy spectrum entropy and the Hilbert spatial entropy by combining the HHT and information entropy.(4) Apply the support vector machine to identify the fault. And the improved LSSVM introduces the least square linear combination to speed up the recognition speed. Compare the common parameter optimization algorithm, and finally, the improved grid search method is used for the identification of this paper. The feature vector extracted from the previous LSSVM were used for final identification. The results show that the classification method based on HHT can get better than that based on wavelet.