Research On Fault Diagnosis System Of EMU's Axlebox Based On Image Recognition

The development of high-speed railway not only provides people with a convenient and comfortable travel environment,but also brings a huge impetus to the economic development of cities along the route.The birth of the high-speed train "Fuxing" in June 2017 opened a new era for China's railway technology and equipment.In the process of increasing the speed of the train,the safety of the train is still an important issue that cannot be ignored.The axle box is an important device for the running section of high-speed rail vehicles.Its reliability greatly affects the safe operation of the train.With the increase in the mileage of high-speed railways and the number of EMUs,the problem of axle box failures has shown an increasing trend.Therefore,it is very important to perform fast and effective fault monitoring on the axlebox.In this paper,based on the study of the axle box wear mechanism,the use of ferromagnetic technology to separate abrasive particles from the grease and create abrasive images.Then it is processed with computer image processing technology to obtain the characteristic parameters of the abrasive particles.Finally,the fault diagnosis system of axlebox was built by using improved Bp neural network.By using this method,the lubrication condition of the axle box can be effectively judged,and the automatic identification of the abrasive particles in different wear forms can be realized.The paper first introduced the basic structure of the axlebox and the common abrasion of the axlebox.On this basis,this paper chooses ferromagnetic analysis as the method of axlebox fault detection.In order to avoid the defects of the traditional ferromagnetic analysis,the computer image processing technology and the ferrography analysis technology are further selected as the main research direction of this paper.Then the paper introduces the specific problems of axlebox fault identification,including the specific content of computer image processing technology,and the specific methods of fault classification of axlebox.The paper selects the improved Bp neural network as the fault identification method,which improves the accuracy of fault identification.Finally,based on the above theoretical basis,the paper designs a set of axlebox fault diagnosis system,including image preprocessing,fault automatic identification and result processing module.And verify the effectiveness of the system based on the actual samples collected.