Research On Arcing Detection Method Of Pantograph And Catenary Based On Multi-sensor Information Fusion

With the rapid development of high-speed trains in China,people are always concerned about the safety of passengers in train operation.However,the detection and maintenance of pantograph and catenary which provide power to locomotive has always been a difficult problem for high-speed railway.As the main fault that damages the catenary,the arc of pantograph seriously threatens the safety of train operation and the life safety of passengers.As an indispensable part of railway detection system,arc detection and identification is very important.Aiming at the problem of high false alarm rate and high labor cost in arc detection of vehicle-mounted contact network(3C)of emu,this thesis proposes a method for arc detection of bow network based on multi-sensor information fusion,which can detect various characteristics of arc and identify arc by comprehensive analysis of DSmT theory.The main research contents of this paper are as follows:Firstly,according to the experimental requirements,the arcing simulation generator is built,and the infrared camera,HD camera,current transformer and other system hardware equipment that meet the research needs of this thesis are selected.At the same time,the detection process of arcing is introduced.Secondly,the exceeding temperature of the contact point between pantograph and catenary was collected and analyzed,and the exceeding temperature time was recorded to judge the temperature characteristics of the arc.The binary image is used to distinguish the arc light from the background color,and an improved Canny operator is proposed to detect the edge of the arc light.In the gradient calculation,the gradient amplitude in the direction of 45° and 135° is added to avoid the missing pixels effectively.At the same time,the total number of pixels in the edge and the relative position coordinates of each pixel are recorded.Calculate the abscissa of the center of gravity of the arc and judge the visible light image of the circular arc according to whether the abscissa value is located in the contact line within the imaging range of the camera.According to the characteristics of the circuit current distortion when the arc occurs,the improved F-score method is used to select the current eigenvector.The improved F-score algorithm overcomes the limitation that the traditional F-score algorithm is only applicable to the feature selection between two types of samples,making the selected eigenvector more concise and representative.Finally,the selected eigenvector is input into support vector machine(SVM)to realize the recognition of arc fault current.Finally,the target recognition framework is built based on the results of pantograph and catenary operation state,and the basic probability of pantograph and catenary operation state is assigned according to the judgment results of variouscharacteristics of the arc and the duration of the arc.The integrity constraints are introduced into the free DSmT model,and the hybrid DSmT model is constructed.In the hybrid DSmT model,the hybrid DSmT fusion algorithm is used to fuse the calculation and analysis,and the calculated results are transformed Turn it into a reliability function,judge the operation state of pantograph and catenary according to the reliability function value,and detect the arc.Case analysis and simulation results show that the arc detection system proposed in this thesis can correctly identify the arc fault,effectively improve the accuracy of arc detection,can get more practical diagnosis results,and provide reliable basis for the inspection workers.