一种旋转机械振动信号的有效消噪方法  

An Improved Signal Denoising Method in Mechanical Fault Diagnosis

作　　者：张鹏瑞 杨智刚 

机构地区：[1]甘肃建筑职业技术学院基础课部,甘肃兰州730050 [2]甘肃省特种设备检验研究院,甘肃兰州730020

出　　处：《测控技术》2015年第8期45-48,共4页Measurement & Control Technology

摘　　要：提出了一种基于奇异值分解(SVD)、Mallat算法和经验模态分解的信号降噪方法。首先,采用香农熵判据寻求最佳小波分解,对带噪部分小波系数进行经验模态分解,提取出信号趋势分量;其次对小波系数剩余部分采用奇异值分解方法降噪,并根据奇异值差分谱自适应的选择奇异值进行重构,将重构后的信号和趋势项叠加作为新的小波系数;最后进行小波重构得到最终的消噪信号。运用模拟信号和齿轮箱断齿故障信号进行仿真,结果表明该方法能够准确地选择用于重构的奇异值个数,并能有效去除信号噪声,保留特征信号的细节信息,尤其对含有趋势项的故障特征有很大实用性。A denoising approach based on singular value decomposition（SVD）,Mallat algorithm and empirical mode decomposition is presented.Firstly,the original signal is decomposed into the best wavelets by Shannon entropy criterion.And the wavelet coefficients of the noisy part are decomposed by EMD,and the trend components are extracted from some noisy wavelet coefficients.Then,SVD is used to denoise the detrended wavelet coefficients,and the difference spectrum of singular value is used to select the singular values adaptively.After de-noising,the de-trended parts are added to the trend parts to get denoised wavelet coefficients.Finally,all wavelet coefficients are reconstructed to get finally de-noising signal.Results of simulation with this method show that this method can accurately reserve the signal detail information,and can effectively remove signal noise.Especially it is very important for mechanical faults with trend characters to denoise.

关 键 词：奇异值分解 MALLAT 经验模态分解 熵判据 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TP391.9[自动化与计算机技术—控制科学与工程]