Research On Dimensionality Reduction And Classification Method Of Rolling Bearing Fault Datasets

With the rapid development of technologies such as big data and industrial Internet,as the core and key equipment of the process industry,the degree of intelligence of rotating machinery is rapidly improving.This trend makes the amount of condition monitoring data that needs to be collected during its operation grow with each passing day,prompting the research on intelligent fault diagnosis technology of rotating machinery must step into the era of industrial big data as soon as possible.However,this leads to a new problem of how to effectively extract valuable fault state information and decision knowledge from mass data resources with low value density.The rolling bearing is the most commonly used key component in the rotating machinery,and its slight defect may lead to the failure of the whole mechanical system.therefore,research on dimensionality reduction and classification problems of reducing the size of fault data sets for rolling bearings,has very important scientific research significance and engineering application value for promoting the development of intelligent fault decision-making technology in the sustainable direction of data science.Based on the above reasons,this study takes the high-dimensional data set of rolling bearing fault vibration signals as the research object,and carries out research work on the dimensionality reduction and classification methods of the fault data set.Th e general situation of the whole research work is as follows:(1)Aimed at the problem that traditional dimensionality reduction methods cannot effectively retain the local and global geometric structure characteristics of the dataset,geodesic distance(GD)is selected as a metric in this study,and propose an improved t-SNE fault datasets dimensionality reduction method D-t-SNE.The characteristic of this method is to embed the GD index into the t-SNE algorithm,by using this index,the algorithm has the performance of keeping the local and global geometric structure characteristics of the two datasets basically unchanged before and after dimension reduction,thereby effectively reducing the difference between the probability distribution of the high-dimensional and low-dimensional datasets and reducing the classification errors caused by them.The effectiveness of the method is verified by the UCI data set and the bearing fault data set.(2)Aiming at the difficulty of fault classification caused by the lo w utilization rate of effective information in high-dimensional fault datasets,Linear Principal Component Discriminant Analysis is proposed to reduce the dimension of the fault datasets.The feature of this method is to integrate the idea of discriminabil ity between classes and principal component calculation into the Linear Discrim Inant Analysis algorithm,Through these two ideas,the algorithm has the ability to eliminate relevant information and redundant features,Thus the valuable fault state information and the main components of the characteristics that can reflect the running state of the machine can be retained better.Two different data sets are using to verify the effectiveness of the proposed method.(3)Aiming at the difficulty of applying adva nced fault diagnosis technology and the low utilization rate of effective information in monitoring data,which leads to the failure of mechanical equipment can not be found in time.Develop a rolling bearing fault diagnosis system based on My SQL database,and embed D-t-SNE and LPCDA algorithms into the system.The system is composed of a database and functional modules corresponding to each step in the fault diagnosis process.Using this system can quickly and accurately diagnose the fault of the mechanical equipment,and it can also check the database in the database panel according to the current abnormal state of the mechanical equipment.The feasibility of the system is verified by the fault diagnosis experiment of rolling bearing.