一种基于改进FP-Growth算法的动车组故障预测研究

摘要

动车组的故障预测和健康管理是目前的研究热点,其中,故障预测的关键是寻找动车组故障信息和状态信息之间的关联关系.频繁模式增长(FP-Growth)算法是关联规则挖掘中的经典算法之一,用来挖掘频繁项集.针对动车组故障数据提出了一种改进的FP-Growth(IFP-Growth,Improved FP-Growth)算法,采用先序遍历FP-tree的方法产生条件模式基.实验结果表明,IFP-Growth算法能够有效提高动车组故障数据挖掘的效率,并且能够有效地挖掘动车组故障信息和状态信息之间的关联关系.%Prognostics and Health Management(PHM) of EMU is the hotspot of current research. The key of fault prediction is to find the relation between fault information and status information of EMU. The FP-Growth algorithm is one of the classical algorithms in association rule mining. It is used to excavate frequent item sets. This paper proposed an improved FP-Growth (IFP-Growth) algorithm for EMU fault data. It adopted pre-traversing FP-tree to generate conditional pattern bases. The experimental results showed that the IFP-Growth algorithm could effectively improve the efficiency of data mining of EMU faults and find the relation between fault information and status information of EMU.