一种新的声发射信号消噪及故障诊断方法

摘要

在旋转机械故障诊断中,声发射信号极易受到噪声的干扰.针对经验模态分解(EMD)易产生模态混叠现象,提出了一种基于经验小波变换(Empirical Wavelet Transform,EWT)的消噪和旋转机械声发射碰摩故障诊断的方法.利用了EMD和小波变换的优点,通过对傅里叶频谱进行自适应划分,并构建小波滤波器组来提取声发射信号所包含的不同固有模态分量,可有效消除模态混叠现象,同时对分量进行Hilbert变换从而实现声发射信号的消噪和故障诊断.采用该方法对仿真信号进行加噪声和消噪处理,在同信号源下,对比基于dB4全阈值消噪、dB4默认软阈值消噪、dB4对高频系数处理消噪和EMD消噪效果.并将该方法应用到实际的声发射碰摩信号中.仿真和实验分析结果表明:EWT方法可以有效地分解出信号的固有模态,分解出的模态少,并且不存在难以解释的虚假模态,消噪效果优于其他方法,并且在声发射故障诊断中也有较大的优势.%Acoustic emission signals are highly susceptible to noise interference in rotating machinery fault diagnosis.The empirical mode decomposition (EMD) suffers from the problem of mode mixing.To solve the problem,this paper achieved a method for de-noising and fault diagnosis of rotating machinery AE signals based on the empirical wavelet transform.This method took the advantages of the EMD and the wavelet transform,classifying the Fourier spectrum by its adaptive property,constructing the wavelet filter bank to extract the different intrinsic mode components of acoustic emission signals,which can eliminate the mode mixing phenomenon.Then the Hilbert transform was carried on the component of the acoustic emission signal so as to realize the de-noising and fault diagnosis.This method was used for de-noising and compared with the result from the global threshold value de-noising,the default threshold value de-noising,the tackle high frequency coefficient de-noising based on dB4.Results show that:Intrinsic modes of the signal can be decomposed effectively through the EWT method.The number of decomposed mode is less and there is no mode that is difficult to explain.Furthermore,de-noising effect is superior to other methods and has great advantage in AE signal fault diagnosis.