Fault diagnosis for jointless track circuit based on intrinsic mode function energy moment and optimized LS-SVM

The track circuit electrical insulated joint faults will lead to the false acceptance of control information for trains and it will affect the safe operation of trains. In this paper the locomotive signal induced voltage model is established based on uniform transmission line theory and the amplitude envelopes of the induced voltage when the electrical insulated joint has malfunctioned are simulated. The fault features extraction is achieved through empirical mode decomposition (EMD) method because of its adaptive advantage. The induced voltage amplitude envelope signals are decomposed into several intrinsic mode functions (IMFs) and the IMF energy moments are used as fault characteristics. The least squares support vector machines (LS-SVMs) are built to realize the multi-class classification. Moreover, the optimal parameters of LS-SVM model are obtained by using the improved PSO algorithm. The experiment shows that the fault diagnosis method for track circuit proposed in this paper is effective and the accuracy is higher than the conventional track circuit fault diagnosis approaches.