摘要
轴承早期的故障信号容易被噪声所淹没,导致其故障特征难以被提取,为此,提出了一种基于交叉小波变换(XWT)与改进变分模态分解(IVMD)联合去噪的信号处理方法。首先,对双通道的原始信号进行了XWT处理,得到了小波相干谱,通过包络谱曲线确定了最佳模态数K;将传统VMD优化为IVMD,利用IVMD将两个通道中峭度值较大的信号分解成为多个固有模态分量(IMFs),再对每个IMF与峭度值较大的信号进行XWT处理;然后,将得到的小波相干谱图与双通道原始信号的小波相干谱图进行了比较,从原始信号中去除了识别出的噪声分量,实现了降噪和故障特征增强的目的;最后,利用K邻近(KNN)算法进行了滚动轴承故障分类,其故障识别率达到了97.51%,与IVMD、VMD-XWT方法相比,该方法故障识别率分别提高了10.83%、4.62%。研究结果表明:该方法可以明显降低噪声干扰,能更好地提取轴承早期的故障信息。
Aiming at the problem that the early fault signal of bearing was easily submerged in noise,and the fault feature was difficult to extract,a joint denoising signal processing method based on cross wavelet transform(XWT)and improved variational mode decomposition(IVMD)was proposed.Firstly,XWT was performed on the original signal of two channels to obtain the wavelet coherence spectrum.The optimal mode number K was determined by the envelope spectrum curve,and the traditional VMD was optimized to IVMD.The signal with large kurtosis value in the two channels was decomposed into multiple intrinsic mode components(IMFs)by IVMD,the XWT was performed for each IMF and signals with large kurtosis value.Then,the obtained wavelet coherence spectrum was compared with the wavelet coherence spectrum of the dual channel original signal,and the identified noise components were removed from the original signal to achieve the purpose of noise reduction and fault feature enhancement.Finally,K-nearest neighbor(KNN)algorithm was used to classify rolling bearing faults,and the fault recognition rate reached 97.51%,which was 10.83%and 4.62%higher than IVMD and VMD-XWT respectively.The results show that this method can reduce noise interference and extract early fault information of rolling bearing better.
作者
王鹏博
刘自然
刘玉明
吕振礼
WANG Peng-bo;LIU Zi-ran;LIU Yu-ming;LV Zhen-li(School of Mechanical and Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China)
出处
《机电工程》
CAS
北大核心
2023年第2期292-298,共7页
Journal of Mechanical & Electrical Engineering
基金
河南省自然科学基金资助项目(182300410234)。
关键词
滚动轴承故障诊断
故障特征提取
降噪
故障特征增强
交叉小波变换
改进变分模态分解
K邻近算法
固有模态分量
rolling bearing fault diagnosis
fault feature extract
noise reduction
fault feature enhancement
cross wavelet transform(XWT)
improved variational modal decomposition(IVMD)
K-nearest neighbor(KNN)algorithm
intrinsic mode components(IMFs)
