A novel intelligent fault diagnosis method based on fast intrinsic component filtering and pseudo-normalization. (November 2020)

• Record Type: Journal Article
• Authors: Zhang, Zongzhen; Li, Shunming; Lu, Jiantao; Wang, Jinrui; Jiang, Xingxing
• Is Part Of: Mechanical systems and signal processing. Volume 145(2020)
• Publication Date: 2020-11
• Subjects: Unsupervised learning -- Fast intrinsic component filtering -- Pseudo-normalization -- Intelligent fault diagnosis -- Lifetime consistency; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2020.106923
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Fast intrinsic component filtering (FICF) is proposed for intelligent fault diagnosis. FICF can learn the intrinsic component and show strong filtering performance for non-intrinsic component. Convolution activation and pseudo-normalization are introduced to improve the computation efficiency, diagnostic accuracy and robustness. Abstract: Unsupervised learning method can obtain the desired feature distribution by changing the objective function. Sparse optimization is an important principle. Given the importance of consistency of features of the same fault condition in the accuracy and robustness of the diagnostic results, a novel unsupervised learning algorithm named fast intrinsic component filtering (FICF) is proposed in this paper for the fault diagnosis of rotating machinery. First, fast convolution activation is introduced to improve the training efficiency. Second, the minimized l 1 / 2 -norms of the feature matrix columns and maximized l 1 / 2 -norms of the feature matrix rows are used to achieve population sparsity and lifetime consistency, respectively. Third, pseudo-normalization (PN) of the test features is proposed to guarantee that all features have similar contributions by dividing the recorded l 2 -norm of the training features. FICF can learn the intrinsic information of the samples and guarantee the sparsity of features within each sample. Meanwhile, FICF can also extract discriminative features from multiple-fault samples. A gear-box dataset and a bearing dataset are used to verify the performance of the proposed method. The verification results confirm that FICF shows high efficiency and strong robustness.