Development And Application Of Wheelset Bearing Fault Diagnosis System

For most rotating machinery and equipment,bearings are indispensable core parts,and the running condition of bearings can often reflect the overall health condition of mechanical equipment.In the actual working process of mechanical equipment,if a certain direction is pressed for a long time,it is easy to cause misalignment of bearings.Large friction between parts often causes wear of bearing inner ring,outer ring and rolling body;The equipment is subject to environmental interference for a long time,such as impact,water erosion,etc.,and the surface of inner ring,outer ring and rolling body is easy to peel off.If problems can be found in the early stage of equipment failure and damaged parts can be replaced in time,accidents can be avoided to a certain extent.In this paper,based on signal decomposition,eigenvalue extraction and intelligent fault diagnosis,a set of wheelset bearing fault diagnosis system is developed based on wheelset bearing signals,and the application test is carried out in combination with actual working scenes.The main research contents of this paper are as follows:(1)First,the empirical mode decomposition and singular spectrum decomposition theory are described.By establishing simulation signals for analysis,it is verified that singular spectrum decomposition can effectively extract each frequency component of the signal.At the same time,the process of signal reconstruction and feature extraction based on the improved kurtosis criterion is described.(2)The theoretical analysis process of multi-layer support vector machine was described in detail,and the advantages and disadvantages of multi-layer support vector machine in fault bearing classification were analyzed.A two-layer support vector machine model algorithm was proposed based on the comprehensive consideration of recognition accuracy and time cost.By using the improved kurtosis criterion,the component signals obtained from the empirical mode decomposition and singular spectrum decomposition are reconstructed,and the highly sensitive characteristic parameters are selected to establish the feature vectors and form the sample data.The classification model algorithm was trained on the Western Reserve public data set and the laboratory bearing data set respectively to output the recognition accuracy.The identification accuracy of test samples was compared and analyzed,which laid a foundation for the engineering application of wheelset bearing fault diagnosis.(3)An experimental platform for wheelset bearing fault diagnosis was designed and developed.According to the actual engineering application requirements,the overall structure design of the system is completed,including the selection of sensors,data acquisition board cards and industrial computer,database analysis and design,and the construction of wheel bearing fault diagnosis experimental platform.By collecting the vibration and sound data of truck wheelset bearing for data analysis,the real-time online monitoring of multi-characteristic parameters was realized.Combined with the database to complete the development of data storage,parameter configuration,real-time monitoring,vibration analysis,sound analysis,data management,model training,report generation modules.The validity of the system in wheelset bearing fault diagnosis is preliminarily verified by the experimental test on the wheelset bearing experimental platform.