Fault Diagnosis Of Track Circuit Based On Fuzzy Neural Network Optimized By Harris Hawks Optimization Algorithm

As the important ground equipment of the Chinese Train Control System(CTCS),the normal working state of track circuits is of great significance for ensuring the safe operation of trains and improving driving efficiency.Due to the complex components of the track circuit and the equipment mostly located outside,the track circuit is easily affected by environmental factors,leading to frequent failures.At present,on-site identification of track circuit faults still relies heavily on the experience of maintenance personnel and there are problems with low diagnostic efficiency and accuracy.In order to improve maintenance efficiency and reduce the labor intensity of staff,it is necessary to introduce intelligent algorithms to quickly and accurately locate track circuit faults.Therefore,this thesis takes the high-voltage pulse track circuit as the research object and proposes a fault diagnosis model of track circuit based on the Fuzzy Neural Network(FNN)optimized by improved Harris Hawks Optimization Algorithm(IHHO).The main research contents of this thesis are as follows:(1)The track circuit four-terminal network model is established based on the structure and principle of the high-voltage pulse track circuit,and the validity of the simulation model is verified.This thesis expounds on the composition and working principle of the high-voltage pulse track circuit,introduces three working states of the track circuit,and analyzes the common faults and causes of the high-voltage pulse track circuit.Based on the four-terminal network theory,a simulation model of high-voltage pulse track circuit is established,and the voltage signals at the power transmitting and receiving ends are derived.Comparing and verifying the simulation data with the measured data,the results show that the error of the simulation model is less than 5%,which is in good agreement with the on-site measured data.(2)The standard Harris Hawks Optimization Algorithm(HHO)is improved and the performance of the improved HHO algorithm is verified.The optimization performance of the Genetic Algorithm(GA),Particle Swarm Optimization(PSO),and the standard HHO algorithm are tested through benchmark tests of single-peak and multi-peak test functions.The test results prove that the HHO algorithm has better optimization accuracy.However,due to the weakened searchability of the standard HHO algorithm in the later stages of iteration,it is easy to fall into local optima,resulting in slowed convergence speed.Therefore,in this thesis,the standard HHO algorithm is improved by designing Logistic chaotic mapping to initialize the population,nonlinear escape energy updating strategy,and Cauchy-Gaussian mutation mechanism.The effectiveness of the improved HHO algorithm is verified through optimization experiments using benchmark test functions.(3)The high-voltage pulse track circuit fault diagnosis model based on the IHHO-FNN(Improved Harris Hawks Optimization Algorithm-Fuzzy Neural Network)is constructed,and comparative experiments are carried out to verify that the diagnostic performance of the IHHO-FNN model is superior.The accuracy and stability of fault diagnosis can be further improved by applying the IHHO algorithm to optimize the weights and other parameters of the Fuzzy Neural Network.Compared with the HHO-FNN(Harris Hawks Optimization Algorithm-Fuzzy Neural Network)model,the IHHO-FNN model has a smaller mean squared error and faster convergence speed.The comparison experiment of fault diagnosis between the IHHO-FNN model and HHO-FNN,PSO-FNN(Particle Swarm Optimization-Fuzzy Neural Network),and GA-FNN(Genetic Algorithm-Fuzzy Neural Network)models is set up.The results show that the overall diagnostic accuracy of the track circuit of the IHHO-FNN model is 97.33%,and the diagnostic time is less while maintaining a higher diagnostic accuracy.To further validate the diagnostic performance of IHHO-FNN,the IHHO-FNN model is compared with the fault diagnosis results of Deep Belief Network(DBN),Extreme Learning Machine(ELM),and Support Vector Machine(SVM).Experimental results show that the IHHO-FNN model proposed in this thesis performs better in terms of precision,recall,F-1 score,and missed detection rate,which verifies the feasibility and effectiveness of the IHHO-FNN model proposed in this thesis in the track circuit fault diagnosis.