摘要
研究目的:历史数据匹配是轨道几何状态数据挖掘的前提与基础,目前基于外部特征的匹配难以实现亚米级的匹配精度,而基于相似性的匹配算法复杂度过高。针对上述问题,本文以静态轨检数据为研究对象,通过动态时间规整度量数据相似性,并以规整路径估计里程偏差,从而构造了一种静检历史数据的精确匹配方法。在此基础上,讨论匹配性能评价问题,并在相关性与精确性评价的基础上引入同步性的指标。最后,采用某高铁2021年1月15日~7月7日间的日常检查数据对上述匹配方法进行验证。研究结论:(1)与基于外部特征的匹配相比,本文所述匹配方法的相关性、同步性与精确性更优;(2)通过改变波形模拟维修作业,显示该方法对动道扰动与数据丢失具有一定的稳健性;(3)本文所述方法无需借助外部特征,因此具有成本低、适应性好的特点,可用于轨道状态信息的深度挖掘。
Research purposes:Historical data matching is the prerequisite and basis for data mining of orbital track geometric state.At present,matching based on external features is difficult to achieve sub-meter matching accuracy,and matching algorithms based on similarity are highly complex.To solve the above problems,this paper took static track inspection data as the research object,measured the similarity of data through dynamic time warping,and estimated the mileage deviation with the wrapped path,thus constructing an accurate matching method of static inspection historical data.On this basis,this paper discussed the matching performance evaluation,and introduced the synchronicity indicators on the basis of correlation and accuracy of the evaluation.Finally,the daily inspection data of a high-speed railway between January 15,2021 and July 7,2021 were used to verify the preceding matching method.Research conclusions:(1)Compared with methods based on external features,the matching method described in this paper has better correlation,synchronization and accuracy.(2)Simulating maintenance operation by means of deforming the waveform,this method is robust to the maintenance operation and data loss.(3)This method is dispensed with external features,so it has the characteristics of low cost and good adaptability,and can be used for deep mining of track state information.
作者
魏晖
杨飞
吴仕凤
朱磊
姚晗星
任晓毅
WEI Hui;YANG Fei;WU Shifeng;ZHU Lei;YAO Hanxing;REN Xiaoyi(Jiangxi University of Technology,Nanchang,Jiangxi 330098,China;China Academy of Railway Sciences,Beijing 100081,China;China Railway Nanchang Group Co.Ltd,Nanchang,Jiangxi 330103,China;Nanchang University,Nanchang,Jiangxi 330031,China)
出处
《铁道工程学报》
EI
北大核心
2022年第9期19-25,共7页
Journal of Railway Engineering Society
基金
江西省自然科学基金(20202BABL204056)
国铁集团科技研究开发计划课题(K2019G010)
关键词
高速铁路
数据匹配
动态时间规整
静态检查
匹配性能
high speed railway
data matching
dynamic time warping
static inspection
matching performance
