Research On Fault Diagnosis Of Hump Retarder Based On Improved BP Neural Network

With the rapid development of the railway,the railway freight is growing day by day.Because of the special situation that Chinese railways are the railways with passenger and freight traffic,it is necessary to pay more attention to the transport efficiency of the freight train.The marshalling station is the heart of railway freight,therefore the disintegration and marshalling capacity of marshalling station must be improved to increase efficiency of the railway freight.The disintegration and marshalling of the marshalling station need the high efficiency of the hump,and the retarder is the critical equipment of hump,so the rates of utilization and fault are increased.The low efficiency of maintenance by on-site maintenance personnel who depend on their experience cannot satisfy the current requirements of freight humping,meanwhile,the control of hump rolling and the safe link of rolling car have been required to higher level.Therefore,it is necessary to make use of the computer to diagnose the faults of retarder.The views are summarized according to the data collected in the field which take advantage of the improves BP neural network to train and test,in order to diagnose the faults of retarder accurately,and rush to repair as soon as possible to increase the efficiency of maintenance.First of all,the faults are analyzed in machinery and electric of retarder,and the common faults are summarized.Secondly,the improved BP neural work model is established to determine that there are nine nodes in the input layer,thirteen nodes in the hidden layer,and six nodes in the output layer.The Bayesian algorithm is adopted to improve the performance function of network,and the LM algorithm is used to improve the convergence rate and generalization ability of network,the characteristic variable is selected to analyze and summarize according to the faults of retarder.Then,the improved BP neural network is trained according to the collected data,the untraining data is adopted to test and improve the generalization ability of the BP neural network,and the MATLAB simulation testing is used to improve the accuracy of BP neural network fault diagnosis.The simulation result can see that the fault diagnosis of the improved BP neural network is quick,accurate and good at application to the hump retarder.At last,the fault diagnosis system of the improved BP neural network to the hump retarder is designed,which can be applied in marshalling station to diagnose the faults and maintain quickly,the fault diagnosis system also can improve the safety and efficiency.In the thesis,the improved BP neural network which is based on the Bayesian algorithm is proposed to analyze the faults of the hump retarder for the first time.It is concluded that the characteristics of rapid convergence and high accuracy of the improved BP neural network through the MATLAB simulation verification.At the same time,the fault diagnosis system of hump retarder is designed according to the improved BP neural network.It has been tested that the designed fault diagnosis system has greatly improved the efficiency of fault diagnosis,which not only reduces the operation cost,but also improves the safety.It can be seen that the system proved a better technology to realize the Auto-control of the hump.