基于集合经验模式分解和K-奇异值分解字典学习的滚动轴承故障诊断

针对经典K-奇异值分解算法构造的字典中原子形态受噪声、谐波干扰影响,进而降低冲击故障特征提取精度的问题,提出了基于集合经验模式分解和K-奇异值分解字典学习的冲击特征提取方法。该方法首先利用集合经验模式分解与Hurst指数对振动信号进行预处理,剔除谐波干扰;其次,利用经典K-奇异值分解算法和预处理信号构造超完备字典;然后,利用K-均值聚类算法对字典中的原子进行筛选;最后,利用正交匹配追踪算法实现冲击故障特征的稀疏表示。实验分析和工程应用验证了所提方法的有效性和实用性。

Rolling Bearing Fault Diagnosis Based on Ensemble Empirical Mode Decomposition and K-Singular Value Decomposition Dictionary Learning

The morphology of atom in dictionary constructed by the K-singular value decomposition algorithm is affected by noise and harmonic interference, which reduces the extraction precision of the fault feature. To solve this problem, a method of impulse feature extraction based on ensemble empirical mode decomposition and K-singular value decomposition dictionary learning is proposed to realize the fault diagnosis of rolling bearing. Firstly, ensemble empirical mode decomposition and Hurst exponent are used to preprocess the vibration signal to remove the harmonic interference; Then, the preprocessed signal K-singular value decomposition algorithm is learned by K-singular value decomposition algorithm to construct the over-complete dictionary; Next, the dictionary is analyzed K-means clustering algorithm to remove the atoms with smaller kurtosis values; Finally, the orthogonal matching pursuit algorithm is used to realize the sparse representation of impulse fault features. Experiments analysis and engineer application verify the effectiveness and practicability of the proposed method.