Research On Bearing Fault Diagnosis Based On Support Vector Machine Under Large Scale Data

With the continuous development of manufacturing technology,many large and complex mechanical equipment have been designed and manufactured.Bearing is an essential part in all kinds of rotating mechanical equipment.Bearing has always been known as the "joint of industry",as the importance part of mechanical equipment,because its failure accounts for more than 40% of all faults,so timely fault identification of bearing is an important research content of operation and maintenance of rotating mechanical equipment.Because of the high complexity and long working time of most mechanical equipment systems,largescale original data are obtained from the monitoring of mechanical equipment bearings,which has an important impact on the efficiency of fault diagnosis.Therefore,this thesis studies the bearing fault diagnosis based on support vector machine under large-scale data.Support vector machine own well study property and generalization,and does not need accurate mathematical model,so it is very suitable for complex mechanical fault diagnosis.Because only a small amount of support vector data in the training sample data affects the construction of support vector machine hyperplane,most of the training sample data are redundant.That is to say,bearing fault diagnosis based on support vector machine can improve the accuracy and efficiency of bearing fault diagnosis based on support vector machine as long as the data is processed properly.In this thesis,the bearing vibration data is used as the original database,through the combination of K-means clustering algorithm and fast convex hull algorithm for data preprocessing,redundant data is removed,the complexity and amount of the original data are reduced,and the processed data set is suitable for support vector machine.Because the features of vibration signal data belong to high dimension,the redundant features have an impact on the diagnosis efficiency,and the parameters of support vector machine also have an impact on its diagnosis performance,so the differential evolution algorithm is used to optimize the parameters of support vector machine and the features of vibration signal synchronously,so as to improve the efficiency of support vector machine.The major job and research study are as follows:(1)Aiming at the problem of large redundancy of large-scale data,an algorithm scheme of clustering convex hull hyperplane is proposed.Firstly,the K-means clustering algorithm is used to cluster the original data,and the convex hull hyperplane is constructed for each cluster.Then the redundant data points inside the convex hull are removed,and the convex hull vertices are retained.The processed data is used for fault diagnosis combined with support vector machine algorithm.This scheme can effectively remove redundant data,shorten the training time of support vector machine,reduce the computer memory requirements,and enhance the workpiece of the arithmetic.(2)The selection of support vector machine parameters and vibration signal characteristics seriously affects the accuracy and efficiency of fault diagnosis.In order to improve the diagnosis efficiency of support vector machine,this paper uses differential evolution algorithm to optimize the parameters and signal features of support vector machine decision model synchronously,to find the optimal combination of parameters and features,to improve the diagnosis efficiency,and to provide theoretical support and application cases for the establishment of fault diagnosis system.