Research On Fault Detection Of Electrical Contacts Based On Infrared Thermal Imaging Technology

All of joints on electric equipments and electric wires are difined as electrical contacts. It has the characteristics of large numbers, various types and high incidence of fault. Therefore, the condition monitoring and fault diagnosis of electrical contact is crucial to safe and stable operation of the power system. A new method based on infrared image fetures ofelectrical contacts and probabilistic neural network(PNN) for fault detection is first time put forward in this paper. This paper study and analysis three categories of electrical contact fault, such as High-voltage isolation switches, Hydraulic Clamp and High-voltage bushing cap. Taking air pollution index(API) and ambient temperature into consideration, which have an great effect on the accuracy of testing results, the method is illustrated detailedly from theoretics and experiments. The outstanding virtue of this method is that can recognize fault grades of electrical contacts accurately and frequently.A highly sensitive infrared camera is used to get infrared images of electrical contact. Then, infrared image denoising method based on genetic wavelet of TLS(Total Least Squares) is presented for electrical contact infrared image denoising. This method took the image that denoised by TLS as male and denoised by wiener filter as female to genetic operation. The dominant gene which extract from TLS wavelet denoising and wiener filtering be called as optimal offspring and decode it into image. Experimental results show that the method effectively removes the noise, has higher SNR(signal-to-noise rate) and smaller MSE(minimizes the mean squared error), compared to conventional methods.In order to extract object region from infrared image which contains complicated information taken at the scene, a improved adaptive genetic algorithm(IAGA) to segment electrical contact image be proposed. To ensure the diversity of population a new select method based on both similarity and fitness be used. The evolution factor be introduced into traditional adaptive genetic algorithm, which enables the improved adaptive genetic algorithm to adjust the possibilities of crossover and mutation adaptively according the individual fitness and evolution generations. The optimal threshold segmentation was transformed into an optimization problem and the improved algorithm which was highly efficient in optimization was used into find the optimal threshold value and accomplish the segmentation of images.Faulty insulator image segmentation experiments have proved that the new algorithm can greatly shorten the time for optimization, overcome shortcomings of prematurity and slow convergence speed and improve the efficiency of image segmentation.With the air pollution index, environment temperature and temperature data from infrared image be taken into accounted, PNN(probabilistic neural network) classifier are designed to check the failuregrades of the electrical connector Contacts. For the three different types of electrical contact, the mapping relationship between the different air pollution levels, different ambient temperatures and failure grades be established by the great nonlinear mapping ability of PNN. The excellent processing capability of PNN on fault tolerance and structural adaption make it achieve the highly accurate detecting of the failure multi-grades.