Research On Hoist Fault Diagnosis Method Based On Time Series Symbolization

Mine hoist is one of the important equipments in the process of mine production.The safety and reliability of mine hoist can directly influence the economic benefit ofcoal mines and the safety of workers. Therefore, it is of theoretical and practicalsignificance to study the work of mine hoist fault diagnosis method.A large number of time series data, which represent the running state, areproduced in the operation process of mine hoist. The data contains the objective lawand knowledge that can be used for mine hoist fault diagnosis. Therefore, this thesisconducts the related research work from the perspective of time series data mining,and the main work includes:a. Time series segmentation method based on sliding window needs to calculatethe error values of all data points repeatedly，resulting in a lower efficiency of thetime series segmentation. This thesis proposed a time series sliding windowsegmentation method based on maximum vertical distance. Experimental results showthat the method can divide the time series efficiently, and has better computingefficiency.b. According to time series segmentation, a time series symbolization methodbased on local feature clustering is presented. This method does isometric processingfor each subsequence and uses multiple slope value to express each subsequence.K-means clustering algorithm is used to cluster the segmentation subsequence andgive symbols to corresponding category. Finally realize the symbolic representation oftime series. Experimental results show that this method can effectively solve theproblem of symbolic time series, and has obvious advantages.c. For the classification problem of symbolic time series, a symbolic time seriesclassification method based on support vector machine is proposed. Experimentalresults show that compared with the classification method based on distancemeasurement, the proposed classification method has better classification accuracyand computing efficiency.d. Finally, this thesis applies the research results of time series data mining tomine hoist fault diagnosis. Experimental results show that the proposed faultdiagnosis method can accurately discriminate the different components correspondingfault of mine hoist.