Research On Data Compression And Trend Association Analysis Algorithm For High-speed Train

In order to ensure the normal operation of high-speed train,a lot of sensors are installed in the train to constantly monitor train running status.If the possible existed problems of other components can be discovered according to the state associations among components from train monitoring data,then timely maintenance can be arranged,which can provide a basis for assessing and monitoring the status of high-speed train operation,and also reduce train running fault.Train monitoring data has a wealth of attributes and the number of data is large.Therefore,in order to find out the trend association between attributes from the train status,the following two aspects of train monitoring data are studied in this thesis:1)Solving the storage issues related with a large number of high speed train monitoring data;2)Mining the trend association between monitoring data attributes to provide a basis for assessing and monitoring the status of high-speed train operation.Specific works of this thesis can be summarized as follows:Firstly,in order to solve the large capacity issues in the storage of monitoring data,and given that the characteristics that the monitoring data of train is mainly consisted of double data,the double think mantissa as integer(DTMI)compression algorithm is proposed based on think mantissa as integer(TMI)algorithm.Experiment results verify that DTMI algorithm has higher compression efficiency compared with common compression algorithms.In order to further improve the compression speed of a large number of monitoring data,the parallel double think mantissa as integer(P-DTMI)algorithm is put forward on the basis of DTMI.Experimental results show that the P-DTMI algorithm can improve the processing speed compared with DTMI.Secondly,in order to explore the trend association relationship among the properties of train monitoring data,the FP-Growth algorithm is used to mining the trend relationship between data attributes.First of all,considering that monitoring data may exist missing items at some time,the neighbor interpolation method is adopted to fill in the missing items;the rule of i and-i is used to express the trend.Then,the confidence and lifting are adopted to remove invalid strong association rules,and reduce the association relationship misrecognized.At last,as for the operation performance of FP-growth algorithm reduces gradually and even cannot run in the case of a large number of input data,the parallel research of FP-growth algorithm is carried out in the Hadoop cluster environment.Experimental results show that the trend association relationship of monitoring data attributes obtained by experiment is coincide with the actual experience,which indicates the availability of algorithm;In addition,the parallel FP-growth can guarantee normal operation with a faster operation speed in the case of a large number of input data,which avoids the limitation caused by association analysis of a small amount of data and can obtain valuable information from the train monitoring data.Finally,in order to apply the data compression algorithm and trend association mining algorithm proposed above,the implementation scheme and prototype software system of high speed train data storage and trend association analysis are designed in this thesis.Meanwhile,the availability of algorithm is verified and design ideas for the development of similar systems are provided.