Research On Fault Data Of EMU Based On High Utility Pattern Mining

High-speed railways are developing rapidly in China.Because of the diverse landforms and environments,the railway vehicles and lines are large-scaled and refined.To deepen the research on the intelligent operation and maintenance of the Electrical Multiple Unit(EMU),the sensor devices on the running vehicles cooperate with the ground equipment to monitor and collect a large number of operating parameters and fault information.How to use these data to analyze and dig out effective conclusions is an important research direction in Prognostics and Health Management technology of EMU.In the field of data mining,frequent pattern mining algorithms can be used to find relevant items,which are suitable for analyzing EMU-related data.But traditional algorithms cannot cope with the rapidly growing data volume.In contrast,the high utility frequent pattern mining algorithm uses the profit information of the items to mine more important frequent pattern from the data.To further explore the quantitative relationship between items in the dataset,the high utility quantitative frequent pattern mining algorithm becomes a popular research topic.But this kind of algorithm still has low performance in time and memory dimensions and is not easy to expand.In the background of such demand,this paper conducts in-depth research on the high utility quantitative frequent pattern mining algorithm and finds effective conclusions with reference to actual business on the real dataset of EMU.The specific works are carried out from the following aspects:(1)In the static database,a high utility quantitative frequent pattern mining algorithm RHUQI-Miner(Related High Utility Quantitative Itemset Mining)based on the Related Degree is proposed.The performance and practicability of the algorithm are improved.To find the frequent pattern with a higher related degree and optimize the search space,RHUQI-Miner proposes the concept of Related Degree,constructs the Item Related Degree Structure,and gives a pruning optimization strategy.Then,the fixed pattern length strategy is used,so that the algorithm can control the length of the output frequent pattern according to the actual situation;(2)In the incremental database,the incremental high utility quantitative frequent pattern mining algorithm IRHUQI-Miner(Incremental Related High Utility Quantitative Itemset Mining)is proposed to realize the incremental application of the RHUQI-Miner algorithm.It improves the scalability of the algorithm and provides a solution for dealing with the dynamically growing database;(3)Combined with external environmental information such as temperature and location,the algorithm is used on the EMU fault data.The relationship between the fault information and operating parameters is analyzed,which provides data support for differentiated and precise maintenance strategies.Moreover,the vehicle operation safety is improved and the maintenance cost is reduced.