针对滚动轴承复合故障诊断中的故障特征分离和提取难题,课题组提出一种基于多尺度形态滤波(multiscale morphological filtering,MMF)和K-SVD字典学习的复合故障特征分离与提取方法.首先,利用多尺度形态学滤波的尺度差异对信号进行分解,实现复合故障特征分离;其次,通过特征能量因子(feature energy factor,FEF)筛选出最佳尺度分量,并利用K-SVD分别构建学习字典库;然后,通过正交匹配追踪算法(orthogonal matching pursuit,OMP)从字典库中重构出信号;最后,结合迭代求差思想,对复合故障进行分离和特征强化.仿真和实验分析表明该方法能够自适应地分离并准确提取滚动轴承复合故障特征.与经典变分模态分解(variational mode decomposition,VMD)方法对比,该方法具有更好的鲁棒性.
Feature Extraction Method for Bearing Composite Faults Based on Multi-Scale Morphological Filtering and K-SVD
In view of the problem of feature separation and extraction in composite fault diagnosis of rolling bearings,a composite fault feature separation and extraction method based on multi-scale morphological filtering(MMF)and K-SVD dictionary learning was proposed.Firstly,the scale difference of multi-scale morphological filtering was used to decompose the signal at multiple scales,and the composite fault feature separation was achieved;Secondly,the optimal scale component was selected by feature energy factor(FEF),and the corresponding learning dictionary library was constructed through K-SVD;Then,the orthogonal Matching pursuit(OMP)algorithm was used to complete the reconstruction of the signal from the dictionary library;Finally,combined with the idea of iterative difference,the composite faults were separated and feature were enhanced.Simulation and experimental analysis show that the proposed method can self-adaptively separate and accurately extract the composite fault characteristics of rolling bearings.Compared with the classical VMD method,the proposed method has better robustness.
bearingcomposite faultMMF(Multi-scale Morphological Filtering)FEF(Feature Energy Factor)dictionary learningOMP(Orthogonal Matching Pursuit)
万方数据
维普
