Research On Fault Diagnosis For Jointless Track Circuit Based On Improved PSO-SVM

Track circuit is one of the most important and critical infrastructure for the railway signaling system. It occupies an important position in the train control system, but also is one of the most prone to failure devices in railway signaling equipment. It is a key factor that affects the rail transport efficiency and safety performance. However, the current of fault diagnosis technology is not perfect, still can’t meet the needs of the development of China’s railway transportation. Therefore, a scientific, efficient and accurate diagnosis has important significance for the track circuit fault.The track circuit compensation capacitor disconnection and insulation damaged section were mainly studied in this article. Depending on the current detection method is not perfect, extreme lack of advanced and effective data processing methods, etc, the main work of this article are as follows:(1) This article describes the structure and working principle of ZPW-2000A system, on this basis, established an equivalent model of two-port network theory for track circuit. Compared the current and voltage values obtained by simulating model and the test data from experimental platform, verifying the correctness of the built model, and then it analyzed the common track circuit faults, gave a key talk for insulation damage and compensation capacitor section break, simulated breakage compensation capacitor, insulating section breakage and the track circuit normal state of induced voltage curve by MATLAB. Compared the simulation results and analyzed the influence of these two fault on the locomotive signal induced voltage amplitude.(2) This article summarizes three fault feature extraction method, selects wavelet transform theory as the appropriate method of track circuit fault feature extraction, then decomposition the locomotive signal induced voltage amplitude signal curve based on wavelet transform theory, and the extraction of more sensitive characteristic parameters for fault diagnosis from the decomposition of the detail signal is used for input variables of fault diagnosis model.(3) The classification principle of support vector machine is explained, and the fault diagnosis model of the non-insulated track circuit based on S VM is established. And then the process of fault diagnosis is expounded from the aspects of the selection of parameters, the pretreatment of input data and the determination of output fault sets. Finally, it designs the random selection and cross validation two determine the parameter model method, and use MATLAB to simulate the two models, analysis and comparison of two kinds of simulation results.(4) This paper summarizes the basic theory of PSO algorithm, based on these theoretical knowledge, aiming at the problem that the parameters of support vector machine is difficult to be selected, simulated annealing (SA) algorithm and PSO algorithm integration, is proposed to optimize the parameters of SVM. By using the global search ability and local search ability of SAPSO, the optimization of the parameters c and g in SVM can be realized quickly, and the optimal solution of the parameters is obtained. The fault diagnosis model of the SAPSO-SVM non insulated track circuit is designed, and the accuracy of the SAPSO-SVM model to track circuit fault diagnosis is obtained by MATLAB simulation, and the simulation results are compared with the simulation results of SVM fault diagnosis model.