An enhanced selective ensemble deep learning method for rolling bearing fault diagnosis with beetle antennae search algorithm

• Published in: Mechanical systems and signal processing Vol. 142; p. 106752
• Main Authors: Li, Xingqiu, Jiang, Hongkai, Niu, Maogui, Wang, Ruixin
• Format: Journal Article
• Language: English
• Published: Berlin Elsevier Ltd 01.08.2020; Elsevier BV
• Subjects: Algorithms; Beetle antennae search algorithm; Deep learning; Enhanced selective ensemble deep learning; Enhanced weighted voting combination strategy; Fault diagnosis; Machine learning; Noise reduction; Performance enhancement; Roller bearings; Rolling bearing; Search algorithms; Teaching methods; Thresholds; Fault diagnosis; Enhanced selective ensemble deep learning; Enhanced weighted voting combination strategy; Beetle antennae search algorithm; Rolling bearing
• ISSN: 0888-3270; 1096-1216
• DOI: 10.1016/j.ymssp.2020.106752

•Deep base models are constructed to adaptively learn latent features from vibration signals.•Multiple diverse deep base models are acquired by variants of AEs and Bootstrap.•EWV with class-specific thresholds is designed to implement selective ensemble.•BAS algorithm is used to optimize the class-specific thresholds of the EWV. Rolling bearing fault diagnosis is a meaningful yet challengeable task. To improve the performance of rolling bearing fault diagnosis, this paper proposes an enhanced selective ensemble deep learning method with beetle antennae search (BAS) algorithm. Firstly, multiple deep base models are constructed to automatically capture sensitive features from raw vibration signals. Secondly, to ensure the diversity of the base models, sparse autoencoder, denoising autoencoder and linear decoder are used to construct different deep autoencoders, respectively, and also Bootstrap is used to design distinctive training data subsets for each base model. Thirdly, an enhanced weighted voting (EWV) combination strategy with class-specific thresholds is proposed to implement selective ensemble learning. Finally, BAS algorithm is used to adaptively select the optimal class-specific thresholds. Experimental bearing data are used to verify the effectiveness of the proposed method. The results suggest that the proposed method can more accurately and robustly recognize different kind of faults than both the individual base models and other existing ensemble learning methods.