Fault Diagnosis Research Of Locomotive Rolling Bearing Based On Neural Network

With the continuous development of industrial technology,the related facilities of railway transportation in our country have also been improved day by day.After several large locomotive speed into the era of high-speed rail.However,locomotive accidents have occurred from time to time,particularly major railway accidents of Wen Zhou line are vivid.Rolling bearing locomotive is a vital part and its status is good or bad determines the normal operation of the locomotive.When a rolling bearing malfunctions,it can bring about another major railway accident.Based on this,this dissertation will focus on the rolling bearing fault diagnosis of locomotive running department.Aiming at the problem of low accuracy and slow speed of locomotive rolling bearing fault diagnosis,this paper studies the application of wavelet packet technology and rough set theory technology to locomotive rolling bearing fault diagnosis.Firstly,wavelet packet decomposition is used to construct the fault feature set,and then the rough set is used to reduce the feature set to eliminate the redundant information.Then,the reduced minimum attribute set is used as the input of BP neural network improved by Levenberg-Marquardt algorithm,Establish the corresponding neural network model to achieve fault diagnosis.The test results show that compared with the ordinary BP network model,the method achieves the goal of improving the speed of fault diagnosis and the accuracy of fault diagnosis.This paper mainly completed the following aspects of research work:(1)The research background and significance of this topic are given,and the research progress and status quo of the topic at home and abroad are described.The rolling bearing and its basic structure and parameters were analyzed.The origin of the vibration of the rolling bearing,the classification of the failure and the mechanism of the failure were discussed.The fault diagnosis procedures,diagnosis performance indicators and diagnosis methods of the rolling bearing were designed.(2)The characteristics of artificial neural network and its BP algorithm are studied.By comparing the common BP network and its two improved algorithms,it is determined that the improved Levenberg-Marquardt method has higher training speed and higher precision.(3)Based on the original data published online by the Bearing Data Center of Case Western Reserve University in the United States,wavelet noise reduction was used to perform preprocessing and fault signature parameters of rolling bearings were obtained based on wavelet packet technology.By analyzing and comparing the training effects of different wavelet packet decomposition layers,it is determined that this paper uses 3-layer wavelet packet decomposition and reconstruction.Finally,using BP neural network to identify the faults of locomotive rolling bearings,the fault types are obtained.The simulation results showthat the algorithm is effective.(4)The rough set theory is introduced into the rolling bearing fault diagnosis.The rough set is used to reduce the fault feature set to eliminate the redundant information.Then the reduced attribute set is used as the input of the BP neural network to establish the corresponding neural network model.Simulation results show the effectiveness of the algorithm.(5)When using the rough set to reduce the conditional attributes,this paper designs a method to build the decision table first and then reduce it.The feasibility and time complexity of the method are analyzed to prove the effectiveness of the method.