基于VMD消噪处理的滚动轴承早期故障识别

摘要

A novel method for early fault recognition of rolling bearing is proposed.The method integrates the variational mode decomposition (VMD) denoising method and the kernel fuzzy C-means (KFCM) clustering method.Firstly,a new method for determining the present scale K of VMD is proposed by the leaked energy and cross-correlation coefficient function synthetically.Shortage of the traditional empirical selection method for K in VMD is overcome.Secondly,the faulty vibration signal is decomposed by VMD to obtain the K band-limited intrinsic mode function (BIMF) components.And in order to find BIMF including noise,the principle is utilized that the normalized energy concentration ratio of auto-correlation coefficient function is greater than 0.9.Then,the BIMF including noise are eliminated and the residual BIMF components are reconstructed to realize denoising.Finally,the root mean square value and the normalized energy of every sample are calculated.The two-dimensional characteristic vector sample sets are obtained,and they are put into KFCM clustering algorithm for fault diagnosis.Analysis of measured fault data of bearings show that the proposed method can realize the early fault diagnosis of rolling bearing effectively when compared with the empirical mode decomposition (EMD) method.%提出了一种基于变分模态分解(VMD)消噪和核模糊C均值(KFCM)聚类相结合的滚动轴承早期故障识别方法.首先提出一种通过综合运用泄漏能量和互相关系数函数确定VMD预设尺度数K的新方法,弥补了VMD方法通常按经验选取预设尺度数方法的不足;然后对振动信号进行VMD分解得到K个限带的内禀模态函数(BIMF)分量,利用归一化的自相关系数函数能量集中比大于0.9的原则确定含有噪声的BIMF分量,并剔除这些含噪BIMF分量,再将剩余的BIMF分量叠加进行信号重构,实现了信号的消噪;最后计算各样本重构信号的均方根值和归一化能量值得到二维特征向量样本集,并输入到KFCM聚类器进行故障诊断.利用实测轴承故障数据进行验证,结果表明与经验模态分解(EMD)方法相比,可以有效地实现滚动轴承早期故障诊断.