基于多域空间状态矩阵奇异值与局部保持投影的滚动轴承故障特征提取方法

发布时间：2018-03-21 09:28

本文选题：总体经验模态分解　切入点：奇异值分解　出处：《机床与液压》2017年09期 　论文类型：期刊论文

【摘要】：针对滚动轴承振动信号复杂且难以从中提取有效故障特征的问题,提出了一种总体经验模态分解(EEMD)、奇异值分解(SVD)和局部保持投影(LPP)相结合的故障特征提取方法。首先,对振动信号进行EEMD分解,利用EEMD分解后的固有模态分量(IMF)分别构造时域、频域和时频域空间状态矩阵;其次,利用SVD提炼时域、频域和时频域空间状态矩阵中的故障信息,筛选其中累加百分比大于90%的奇异值组成多域有效奇异值数组,构造多域奇异值特征矩阵;然后,利用LPP约简多域奇异值特征矩阵,提取低维、高区分度的故障特征;最后,利用支持向量机(SVM)对提出的故障特征提取方法进行评估。实验结果证明了该方法提取的故障特征可有效反映滚动轴承的故障状态。
[Abstract]:In view of the complex vibration signals of rolling bearings and the difficulty of extracting effective fault features from them, a fault feature extraction method is proposed, which combines the total empirical mode decomposition (EEMD), singular value decomposition (SVD) and local preserving projection (LPP). The vibration signal is decomposed by EEMD, and the time-domain, frequency-domain and time-frequency-domain state matrices are constructed by using the inherent modal components after EEMD decomposition. Secondly, the fault information in time-domain, frequency-domain and time-frequency spatial state matrix is extracted by SVD. The singular values with a cumulative percentage of more than 90% are selected to form a multi-domain effective singular value array, and the multi-domain singular value characteristic matrix is constructed. Then, the low-dimensional and high-resolution fault features are extracted by using the LPP reduction multi-domain singular value feature matrix. The proposed fault feature extraction method is evaluated by support vector machine (SVM), and the experimental results show that the proposed method can effectively reflect the fault state of rolling bearing.
【作者单位】：青岛理工大学机械工程学院;
【基金】：国家自然科学基金资助项目(51075220) 高等学校博士学科点专项科研基金项目(20123721110001) 青岛市科技计划基础研究资助项目(12-1-4-4-(3)-JCH)
【分类号】：TH133.33

【相似文献】