基于MRSVD-SVD与VPMCD的交叉滚子轴承故障诊断研究  

Fault diagnosis of cross-roller bearings based on MRSVD-SVD and VPMCD

作　　者：何冬康 甘霖 类志杰 邓其贵[1] 和杰 HE Dong-kang;GAN Lin;LEI Zhi-jie;DENG Qi-gui;HE Jie(School of Mechanical and Electrical Engineering,Liuzhou Vocational&Technical College,Liuzhou 545006,China;Special Equipment Safety Inspection Institute in Yunnan Province,Kunming 650228,China;Yunnan Huimin Labor Service Co.,Ltd.,Kunming 650228,China)

机构地区：[1]柳州职业技术学院机电工程学院,广西柳州545006 [2]云南省特种设备安全检测研究院,云南昆明650228 [3]云南惠民劳务服务有限公司,云南昆明650228

出　　处：《机电工程》2023年第1期47-54,共8页Journal of Mechanical & Electrical Engineering

基　　金：广西高校中青年教师基础能力提升项目(2022KY1036,2018KY0986);柳州职业技术学院校级科研课题(2022KB04)。

摘　　要：针对奇异值分解(SVD)提取工业机器人交叉滚子轴承振动信号微弱故障特征分量时,出现奇异值分辨率不足的问题,提出了一种基于最大分辨率奇异值分解(MRSVD)-奇异值分解(SVD)与变量预测模型模式识别(VPMCD)的工业机器人交叉滚子轴承的故障诊断方法。首先,以最大奇异值分辨率原则将一维振动信号构造成了Hankel矩阵,采用奇异值分解方法对Hankel矩阵进行了分解,得到了其奇异值序列,根据奇异值曲率谱理论选择有效奇异值,并进行了重构,得到了经降噪后的高信噪比信号,以重构信号构建了相空间矩阵,进行了二次奇异值分解,得到了其故障特征分量;然后,计算了故障特征分量的特征参数,构建了其特征向量;最后,采用了VPMCD分析了特征向量,完成了对交叉滚子轴承故障类型的识别,并与其它方法进行了识别准确率对比。研究结果表明:采用该方法对工业机器人交叉滚子轴承进行故障诊断,得到的故障类型识别准确率为98.66%,比SVD与共振解调相结合方法提高了9%;该方法通过构建最大奇异值分辨率矩阵提高了奇异值分辨率,可完整提取出工业机器人交叉滚子轴承振动信号的微弱故障特征分量,获得了更高的故障类型识别准确率。Aiming at the problem of insufficient singular value resolution of singular value decomposition(SVD)extracting the weak fault information from the cross-roller bearings for industrial robots,a new method based on maximum resolution singular value decomposition(MRSVD)-SVD and variable predictive model class discriminate(VPMCD)fault diagnosis method for industrial robot cross-roller bearings was proposed.Firstly,one-dimensional vibration signals were constructed into Hankel matrix based on the principle of maximum singular value resolution,the Hankel matrix was decomposed by SVD to obtain singular value sequence.According to the singular value curvature spectrum theory,the effective singular value was selected to reconstruct the denoised signal with high signal-noise ratio(SNR).The phase space matrix was constructed from reconstructed signals.The fault feature components were obtained by quadratic SVD processing of the reconstructed signals.Then,the characteristic parameters of fault feature components were calculated to construct feature vectors.Finally,VPMCD was used to analyze feature vectors and identify fault types.The experimental results show that the method is applied to fault diagnosis of industrial robot cross-roller bearings.The fault type identification accuracy is 98.66%,which is 9%higher than the method combining SVD and resonance demodulation.It is shown that the proposed method improves the singular value resolution by constructing the maximum singular value resolution matrix,and completely extracts the weak fault feature components of the vibration signals of the industrial robot cross-roller bearing,thus achieving higher fault type identification accuracy.

关 键 词：滚动轴承 圆柱滚子轴承 最大分辨率奇异值分解 奇异值分解 变量预测模型模式识别 HANKEL矩阵 

分 类 号：TH133.33[机械工程—机械制造及自动化]