面向大规模监测数据的高铁故障诊断技术研究  

Research on Fault Diagnosis Technology of High-speed Rail for Large Scale Monitoring Data

作　　者：李贵兵[1] 金炜东[1] 蒋鹏[1] 付小利[1] 熊定鸿[1] 谷鹏举 

机构地区：[1]西南交通大学电气工程学院,成都610031

出　　处：《系统仿真学报》2014年第10期2458-2464,共7页Journal of System Simulation

基　　金：国家自然科学基金重点项目(61134002);西南民族大学青年教师基金项目(13NZYQN17)

摘　　要：大数据及云计算等技术带来的科研方式的转变正在影响着对高铁故障诊断技术领域的研究工作,利用云计算平台对大规模高铁监测数据进行了故障检测分析的研究工作,利用hadoop平台对仿真平台产生的高铁监测数据进行了数据预处理,并在该平台下实现了并行化的EEMD与信息熵相结合的故障特征提取,然后在map-reduce计算框架下实现了对特征提取结果的KNN故障分类识别,对分类效果和运行性能指标进行了分析。实验结果表明,该方法能准确有效的对高铁故障进行识别分类,在运算速度、并行化加速度性能上都较传统方法有明显的改善。The change of research way brought by big data and cloud computing is having an impact on the research of high-speed rail fault diagnosis analysis. The research of fault detection and analysis for large scale high-speed rail monitoring data is conducted utilizing cloud computing platforms. Data preprocessing to monitoring data generated by simulation platform is realized based on hadoop, and so is Parallel fault feature extraction by the way of EEMD combining with information entropy. KNN fault classification algorithm towards fault feature under the computing framework of map-reduce is successfully achieved. The classified result and performance index is analyzed. Experimental results show that the proposed method is accurate and effective on high-speed rail fault classification, and it has significantly improved the speed of operation and parallel speed-up compare to previous method.

关 键 词：故障诊断分析 MAP-REDUCE 聚合经验模态分解(EEMD) KNN 并行化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]