引入稀疏原子特征融合的滑动轴承摩擦故障状态监测

发布时间：2018-06-25 11:45

本文选题：滑动轴承 + 状态监测　；参考：《航空动力学报》2017年10期

【摘要】：从信息融合理论出发,将特征的稀疏表达作为特征融合参数,提出一种结合K奇异值分解(KSVD)和最大相关最小冗余准则(mRMR)的轴承摩擦故障特征融合算法。该算法采用KSVD对信号稀疏化,将稀疏系数对应的字典原子作为特征融合的参数,用以表达非线性故障信息;针对字典原子集的优化选择问题,基于互信息的mRMR提出一种确定最优原子集的原子数目的准则;最后,通过最大化原则融合稀疏系数,提取故障状态监测的有效信息。轴承摩擦故障模拟实验的结果表明,所提方法能够更好地融合不同特征的故障信息,相比于单特征和其他融合特征方法,提高了约12%的故障识别率。
[Abstract]:Based on the information fusion theory, the sparse representation of features is taken as feature fusion parameters, and a feature fusion algorithm based on K singular value decomposition (KSVD) and maximum correlation minimum redundancy criterion (mRMR) is proposed. The algorithm uses KSVD to sparse signals and takes dictionary atoms corresponding to sparse coefficients as feature fusion parameters to express nonlinear fault information. Based on mutual information, mRMR proposes a criterion to determine the number of atoms in the optimal atomic set. Finally, the effective information of fault state monitoring is extracted by the principle of maximization fusion of sparse coefficients. The simulation results of bearing friction fault show that the proposed method can better fuse the fault information of different features and improve the fault identification rate by about 12% compared with the single feature and other fusion feature methods.
【作者单位】：军械工程学院车辆与电气工程系;西南交通大学机械工程学院;
【基金】：国家自然科学基金(51205405,51305454)
【分类号】：TH133.31

【相似文献】