Track Circuit Compensation Capacitor Fault Detection Based On Extreme Learning Machine

The track circuit is an important signal device,and its working state determines whether the railway system can operate normally.As the outdoor signal equipment,the compensation capacitor is mainly used to reduce the reactive power generated by the signal transmission on the rail,improve the "signal-to-interference ratio" of the track circuit,and increase the transmission distance of the high-frequency signal in the orbit.Failure of the compensation capacitor may cause the red band to malfunction,which jeopardizes driving safety.Previous studies on fault detection of the compensation capacitor,the main research point is to locate the fault capacitor by using the amplitude envelope of the locomotive signal.Based on the adjustment state of the track circuit,this paper models and analyzes the voltage change at the receiving end of the capacitance value,and proposes a method based on the capacitance estimation of the ground equipment.The specific research contents are as follows:1.Taking the experimental line of ZPW-2000 A track circuit as the research object,the four-terminal network model of each structure of the track circuit is established,and the signal transmission model of the track circuit under the adjusted state is obtained through the cascaded model.Then,the fault value of the compensation capacitor under the adjustment state is simulated and analyzed.2.By increasing the impedance simulation of the track circuit,the experimental line compensates for the capacitance drop,and uses the LABVIEW-based hardware and software devices to collect the various parameter voltage data.Verify the accuracy of the model based on the collected data3.A four-terminal network model is established for the operating line track circuit.The variation of the voltage at the receiving end and the setting principle of the capacitance of the compensation capacitor are analyzed when the capacitance of each position compensation capacitor decreases and the line is broken.4.The structure and principle of the Extreme Learning Machine(ELM)algorithm are introduced.For the problem of weak generalization ability,the particle swarm optimization(PSO)algorithm is used to optimize the ELM hidden layer parameters.On the basis of the measured data of the experimental circuit of the track circuit,the main rail output and the small rail output voltage are used as the neural network input,and the capacitance value and position are compensated as the output.The PSO-ELM has a good fitting effect through experiments.For the actual operating line,the simulation data is used to verify the feasibility of implementing the capacitance estimation method based on the ground equipment.5.The four-terminal network model of the track circuit in the shunt state is established,and the shunt current change when the compensation capacitor capacity decreases is analyzed.Based on the analysis of the application status of compensation capacitor fault location based on vehicle equipment,a fault detection method based on vehicle-ground correlation compensation capacitor is proposed.Through the experimental comparison and analysis,the optimal activation function and the number of hidden layer nodes of PSO-ELM are established,and the performance of the algorithm is compared with the performance of ELM and BP algorithms respectively.Finally,based on Matlab’s App Designer design tool,the simulation of the compensation capacitor fault detection system is completed.