摘要
滚动轴承是机械设备的重要零部件之一,对其进行及时有效地监测和诊断对机械设备的安全运行有着重大意义。针对多源信息融合导致的高维性、信息冗余等问题,提出了一种基于局部联合稀疏边缘嵌入(locally joint sparse marginal embedding,LJSME)的轴承故障诊断方法。LJSME利用L 2,1范数来重构类间矩阵和类内矩阵并引入局部图保留高维特征间的邻域关系,且将L 2,1范数作为目标函数的正则项以保证特征提取的联合稀疏性,从而提高特征的敏感性和鲁棒性。首先从轴承振动信号中提取由时域和频域信息组成的高维特征数据集;随后利用LJSME提取高维特征空间数据集中的敏感低维特征;最后利用K-近邻分类器实现滚动轴承的故障模式识别。通过两组滚动轴承故障数据集对所提出的方法进行验证,与其他三种降维算法相比,所提算法能够有效地提取滚动轴承振动信号的敏感性特征。
In order to address the high dimensionality and information redundancy caused by information fusion in existing bearing fault diagnosis methods,a novel bearing defection diagnosis method based on locally joint sparse marginal embedding(LJSME)was proposed.In the method of LJSME,L 2,1-norm was used to reconstruct within-class and between-class matrices and local graphs were introduced to preserve the neighborhood relations among high-dimensional features,and L 2,1-norm was used as the regular term of the objective function to obtain the joint sparsity of feature extraction,thus ensuring the effectiveness of feature extraction.In the method first a high-dimensional feature dataset consisting of time-domain and frequency-domain information from the bearing vibration signal was extracted;then,sensitive low-dimensional features in the high-dimensional feature space dataset were extracted by using LJSME;and finally,the fault pattern recognition of rolling bearings was achieved by using a K-nearest neighbor classifier.The proposed method was validated by two sets of rolling bearing fault datasets,and the proposed algorithm can more effectively extract sensitive features of rolling bearing vibration signals compared with other three dimensional reduction algorithms.
作者
周宏娣
张航
钟飞
ZHOU Hongdi;ZHANG Hang;ZHONG Fei(School of Mechanical Engineering,Hubei University of Technology,Wuhan 430068,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2023年第14期124-130,共7页
Journal of Vibration and Shock
基金
国家自然科学基金(52005168)。
