基于VMD和形态学滤波的轨道车辆轴箱振动信号盲分离  

Fault Diagnosis of Vibration Signal of Railway Vehicle Axle Box Based on VMD and Morphological Filtering

作　　者：黄如艳 钟倩文 罗文成 郑树彬[1] 彭乐乐 HUANG Ruyan;ZHONG Qianwen;LUO Wencheng;ZHENG Shubin;PENG Lele(School of Urban Rail Transit,Shanghai University of Engineering Science,Shanghai 201620;Changzhou Luhang Rail Transit Technology Co.,Ltd.,Changzhou 213164)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620 [2]常州路航轨道交通科技有限公司,常州213164

出　　处：《计算机与数字工程》2023年第6期1416-1421,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:51907117,51975347);上海市科委重点支撑项目(编号:18030501300)资助。

摘　　要：轨道车辆轴箱振动信号包含大量振动信息,为能有效定位识别其中的故障信号,论文提出基于变分模态分解(Variational Mode Decomposition,VMD)和形态学滤波结合的盲分离算法。VMD将观测信号分解为若干个本征模态分量并用形态学对分量进行滤波,利用奇异值分解(Singular Value Decomposition,SVD)降维以估计源信号个数,将估计出的源信号与观测信号组成新的多维信号,通过快速独立成分分析(Fast Independent Component Analysis,FastICA)实现轴箱振动信号的分离。仿真信号验证算法可行性,对比VMD,EEMD和小波包算法,该算法在准确率和效率上具有优越性。构建车轮扁疤仿真数据,轴承实验平台信号和噪声混合多元复杂轴箱振动信号,较为准确地分离出各故障信号,误差最大为2.6%。The vibration signal of rail vehicle axle box contains much information,mainly is the similar characteristic frequen-cies signal such as wheel-rail excitation and bearing vibration are mixed together,and it is difficult to diagnose faults with conven-tional detection methods.The blind separation algorithm based on variational mode decomposition(VMD)and morphological filter-ing is proposed to achieve fault signal separation.VMD decomposes the observation signal into several eigenmode components and filters the components with morphology,and uses singular value decomposition(SVD)to reduce the dimensionality to estimate the number of source signals.Finally,the estimated source signal is combined with the observation to be new multi-dimensional signal,and the new signal is separated by Fast Independent Component Analysis(Fast ICA).It is verified by simulation signals that this al-gorithm has superior accuracy and efficiency than VMD,EEMD and wavelet packet algorithms;it constructs a multi-source alias-ing signal with similar characteristic frequencies of wheel flat scars and bearing experimental vibration signals and noise.It can accu-rately separate each fault signal,and the maximum error is 2.6%.

关 键 词：轴箱 VMD SVD 本征模态分量 仿真信号 FASTICA 

分 类 号：TN911.7[电子电信—通信与信息系统]