高速铁路轨道几何检测数据自动预处理方法研究  

Research on Automatic Preprocessing Methods for Track Geometry Inspection Data of High Speed Railway

作　　者：许贵阳[1] 史天运[2] 刘金朝[1] 曲建军[1] 

机构地区：[1]中国铁道科学研究院基础设施检测研究所,北京100081 [2]中国铁道科学研究院电子计算技术研究所,北京100081

出　　处：《中国铁道科学》2013年第6期8-12,共5页China Railway Science

基　　金：国家自然科学基金资助项目(51178464);国家"八六三"计划项目(2011AA11A102);国家科技支撑计划项目(2011BAG05B02-01)

摘　　要：针对人工预处理海量的高速铁路轨道几何检测数据存在效率低和结果多样性等问题,研究检测数据自动预处理的方法。基于相关系数最大化原理,提出对检测数据里程自动修正的方法;利用傅里叶变换和逆傅里叶变换构造1个高通滤波器,实现对单项检测数据趋势项的自动滤除;基于绝对平均值对单一异常值数据进行自动识别和和滤除;对于非检测速度过低造成检测数据中可能出现成片的异常值,设计基于移动窗的识别方法。利用给出的检测数据自动预处理方法对实测的高速铁路轨道几何检测数据进行预处理,结果表明:该方法不但能有效识别和剔除分布在检测数据中的无效数据,而且能完整地保留有用信息,验证了该方法的可行性。Automatic preprocessing methods are investigated for solving the problems of poor efficiency and result diversity derived from manually editing the massive invalid values of track geometry inspection data of high speed railway. A method based on the maximum principle of correlation coefficient is applied to automatically correct the mileage of inspection data. A high-pass filter is constructed by combining Fourier transform and inverse Fourier transform for automatically filtering the trend term of single inspection data. Absolute mean value method is applied for diagnosing and removing the single abnormal data. The patches of abnormal value of track geometry in the case of non-low velocity are identified and deleted by a new method based on moving window method. The proposed automatic preprocessing methods are applied for analyzing the measured track geometry inspection data of high speed railway. The results show that the methods not only can effectively distinguish and eliminate the invalid data scattered in the inspection data, but also retain the integrity of useful information, which has validated the feasibility of the methods.

关 键 词：轨道几何不平顺 检测数据 自动预处理 高速铁路 

分 类 号：U216.3[交通运输工程—道路与铁道工程] TP274.2[自动化与计算机技术—检测技术与自动化装置]