Research On Switch Fault Diagnosis System Based On PSO-BP

The switch is the key equipment to realize the transfer line in the rail transit transportation infrastructure.The switch machine is the most important equipment in the daily maintenance,which has many faults in the signal discipline.For a long time,the operation state analysis and fault diagnosis of the switch machine in the signal specialty mainly rely on the information collection of the microcomputer monitoring system and the analysis and judgment of the monitoring and analysis personnel.The demand for the staff's sense of responsibility and skill level is high when manually looking at the analysis data.At the same time,there are some problems such as poor timeliness,long analysis period and weak guidance.A large number of monitoring data lack of accurate and timely analysis,resulting in some hidden dangers missing reports or fault misjudgment,resulting in increased fault delay,expanded scope of influence,and greater impact on the signaling profession.Based on the full analysis of the maintenance experience of domestic and foreign turnouts,this paper uses the mature neural network technology to carry out the technical slicing and multi-dimensional comparison of the centralized monitoring data of signals,find out the abnormal information mark points,analyze the historical false alarm data pertinently,and gradually make the function of the intelligent analysis system reach the analysis ability of the signal experts.A kind of intelligent algorithm using PSO-BP neural network is proposed The method of intelligent diagnosis of turnout fault can solve the problems of difficult judgment and long time of current turnout fault.The main research contents are as follows:Firstly,the basic structure and action mechanism of S700 K switch and its turnout are analyzed,and analyzed.Combined with the big data on site,six common failure modes and their corresponding failure curves are summarized.Then take the normal working turnout curve as a reference,divide it into seven parts,and discuss in sections to reduce the difficulty of analysis.Secondly,in order to simplify the calculation process and ensure the real-time performance of the entire model analysis,a fault feature extraction method is proposed.This method establishes a fault feature set,which reduces the calculation time and improves the calculation efficiency.A fault diagnosis framework was established,and a fault diagnosis process based on BPNN was given.Aiming at the existing problems,a particle swarm optimization algorithm was proposed to optimize it,and a fault diagnosis method based on PSO-BP model was constructed.The 220 sets of switch curves at Helens Lunan Station were selected for simulation.This system is based on PSO-BP neural network,uses Access to build a database,and uses JAVA programming to complete the fault diagnosis process,including system log-in and log-out,diagnosis,new fault entry,visualization module,statistics module,human-machineinterface,etc.The simulation results show that the PSO-BP model proposed in this paper has greatly improved the diagnosis time and accuracy compared with the traditional BP model,and the judgment accuracy rate has increased by about 10%.It is verified that the research in this paper has engineering application value.