Time-series Data Clustering Method Of Tunnel Boring Machine And Its Application

The tunnel boring machine(TBM)is a systematic large-scale tunnel construction machine with complex technology and high added value.Its intelligent operation relies on data mining technology.Analyzing the operation data of TBM with the help of data mining technology is of great significance for improving the operation,analysis and maintenance level of TBM.The clustering research in data mining is very important for the grouping and classification of the operating data of the TBM and the analysis of the operating state,and it is an important prerequisite step for data analysis.However,the time series data of TBM has the problems of high dimensionality and strong correlation between parameters.Traditional data mining methods are difficult to effectively apply,and there is a lack of relevant algorithm research for TBM operating data.This paper is funded by the National Key Research and Development Program of the Ministry of Science and Technology(No.2018YFB1702502).Aiming at the problems of a large number of abnormal data of TBM operation data and difficulty in temporal and spatial classification,an outlier detection and clustering algorithm suitable for TBM operation data is studied.The operation data of the tunnel between Houting and Songgang of Shenzhen Metro Line 11 is used to verify the algorithm.The main work of this paper is as follows:(1)Aiming at the problem of many outliers in operation data,an outlier detection method based on sliding window is proposed.Firstly,the sliding window method is utilized to complete the segmentation processing of time series data.Secondly,the segmented data features are extracted for abnormal sub-sequence judgment,and finally the Gath-Geva clustering algorithm is used to realize the outlier recognition,Experiments show that this algorithm has a higher abnormal recognition rate.(2)Aiming at the chaotic operation data,this paper takes the advancing speed and cutter head speed as the research object,a time series clustering method based on similarity measurement is proposed to realizes the clustering of the operating data,and seek similar tunneling modes of TBM.First,the dynamic time warping algorithm is used to evaluate the similarity between time series,and then the Gath-Geva clustering algorithm is used to complete the time series clustering grouping.Experimental analysis shows that this clustering algorithm in this paper has higher clustering accuracy and can achieve data partitioning more accurately.(3)This paper takes the multi-parameter data of the TBM as the research goal,constructs a fragment clustering method of multiple time series data,realizes the segmentation and clustering of the data,and realizes the analysis of the operational status of the TBM.Based on the dynamic factor model,the common factor sequence between multiple operating parameters is solved to realize data dimensionality reduction.The FCM clustering algorithm is used to complete the fragment cluster analysis of the common factor sequence.The results show that this method can accurately identify the status change of the TBM,which is of great significance for dividing the operating data.(4)This article takes the result of segment clustering as the research object,and establishes a propulsion speed prediction model based on least squares support vector machine for each segment of data.Compared with the results of unsegmented clustering,this prediction model in this paper has higher prediction accuracy.It provides a new prediction method for parameter prediction of TBM.