基于D-CA和R-EEMD的液压系统故障识别  

Fault Feature Extraction of Hydraulic Systems Based on D-CA and R-EEMD Methods

作　　者：柴凯[1] 张梅军[1] 黄杰[1] 冯霞[1] 

机构地区：[1]解放军理工大学野战工程学院,南京210007

出　　处：《噪声与振动控制》2015年第1期204-208,224,共6页Noise and Vibration Control

基　　金：2011年国家自然科学基金项目(51175511)

摘　　要：剖析液压系统故障特征,采用了一种双相关分析(D-CA)和改进的集合经验模态分解(R-EEMD)相结合的液压系统故障提取新方法。该方法首先对原信号进行自相关分析,突出信号中的周期成分和去噪,利用支持向量回归机(SVR)延拓来改进的EEMD对原信号的自相关函数进行分解;得到理论意义上的固有模态函数(IMF)。再通过求取IMF分量与自相关处理的信号的频域而非传统时域上的互相关系数,去除虚假IMF分量。最后对去除虚假分量重构信号进行Hilbert谱分析提取信号的故障特征。该方法提高了信噪比,减少了IMF的数量,抑制了端点效应,成功地提取了液压系统故障特征频率。A new method for hydraulic-system fault-feature extraction was proposed based on double-correlation analysis（D-CA） and refined ensemble empirical-mode-decomposition （R-EEMD）. Firstly, the original signal was processed byadaptive correlation analysis to extrude the periodic components and eliminate the noise. Secondly, the R-EEMD based onsupport vector regression （SVR） was used to analyze the adaptive correlation processing signal to obtain the theoretical intrinsicmode function （IMF）. Thirdly, the false IMF components were removed by extracting the correlation coefficient ofIMFs and the adaptive correlation processing signal in the frequency domain instead of traditional time domain. Finally, thereconstructed signal was analyzed by the Hilbert spectrum to extract the fault features. Simulation and experimental resultsshow that this method can increase signal-to-noise ratio, reduce the number of IMFs, depress the end effect, and effectivelyextract the faults feature frequencies of the hydraulic system.

关 键 词：振动与波 故障诊断 双相关分析 集合经验模态分解 

分 类 号：TB53[理学—物理] TN911.7[理学—声学]