一种轨旁设备可靠性度量方法的设计与实现  

Design and Implementation of A Reliability Measurement Method for Trackside Equipment

作　　者：马迪迪 赵静 林亚龙 王婧雯 MA Di-di;ZHAO Jing;LIN Ya-long;WANG Jing-wen(Hefei Urban Rail Transit Corporation,Hefei 230001)

机构地区：[1]合肥市轨道交通集团有限公司,合肥230001

出　　处：《环境技术》2023年第12期24-30,共7页Environmental Technology

摘　　要：为轨旁设备是指安装在铁路轨道旁边的设备,用于监测和控制铁路运行。轨旁设备的作用是确保铁路的安全运行,提供准确的信号和信息给列车驾驶员,以及监测轨道的状态和列车的位置。度量多尺度的特征使得其可靠性度量一直是一个难题。提出多尺度特征融合下的轨旁设备可靠性度量方法。采用互补完全经验模态分解方法对设备的多维源信号进行分解,获得一系列IMF。引入独立成分分析方法对这些IMF进行处理,通过计算各维信号的模糊熵,可以去除高频噪声并重构信号,消除多维源信号中的噪声和干扰。将经过去噪后的设备IMF信号输入卷积神经网络中,利用Softmax函数量化处理历史运行数据和状态特征集,并提取信号特征,完成多尺度特征融合。采用灰色关联分析方法计算融合后特征序列中参考状态之间的灰色关联度,完成可靠性度量。实验结果表明:所提方法的去噪效果好、度量精度高、度量效率高,保证轨旁设备安全、可靠运行。Trackside equipment refers to equipment installed beside Railway track for monitoring and controlling railway operation.The role of trackside equipment is to ensure the safe operation of the railway,provide accurate signals and information to the train driver,and monitor the status of the track and the position of the train.Measuring the reliability of multi-scale features has always been a challenge.Propose a reliability measurement method for trackside equipment based on multi-scale feature fusion.Using the complementary complete empirical mode decomposition method to decompose the multi-dimensional source signals of the equipment,a series of IMFs are obtained.Introducing independent component analysis method to process these IMFs,by calculating the fuzzy entropy of each dimensional signal,high-frequency noise can be removed and the signal can be reconstructed to eliminate noise and interference in multi-dimensional source signals.Input the denoised equipment IMF signal into the Convolutional neural network,use the Softmax function to quantify and process the historical operation data and state feature set,and extract the signal features to complete multiscale feature fusion.Use the grey correlation analysis method to calculate the grey correlation degree between the reference states in the fused feature sequence,and complete the reliability measurement.The experimental results show that the proposed method has good denoising effect,high measurement accuracy,and high measurement efficiency.

关 键 词：互补完全经验模态分解 独立成分分析 卷积神经网络 多尺度特征融合 可靠性度量 

分 类 号：U231.7[交通运输工程—道路与铁道工程]