2MNet: Multi-sensor and multi-scale model toward accurate fault diagnosis of rolling bearing. (December 2021)

• Record Type: Journal Article
• Authors: Guan, Yang; Meng, Zong; Sun, Dengyun; Liu, Jingbo; Fan, Fengjie
• Is Part Of: Reliability engineering & system safety. Volume 216(2021)
• Publication Date: 2021-12
• Subjects: Fault diagnosis -- Rolling bearing -- Multi-sensor information fusion -- Deep multi-scale feature extraction and fusion; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2021.108017
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: A correlation kurtosis weighted rule is constructed to fuse multi-sensor signals. Multi-scale features can be extracted through multi-branch dilated convolutions. Multi-scale features can be reused and fused by combining the pyramid principle. Abstract: Rolling bearing is an indispensable element of rotating machinery, timely and accurate fault diagnosis of rolling bearing plays an important role in the safe and reliable operation of modern industrial systems. Considering the bottleneck that the information collected by a single sensor and single scale features extracted by conventional networks are not comprehensive, a multi-sensor and multi-scale model (2MNet) is proposed to bring a new perspective to accurate fault diagnosis. Most notably, multi-sensor vibration signals in three directions can be fused by defining a novel correlation kurtosis weighted fusion rule. Furthermore, the implication of multi-scale is twofold: one is the multi-scale feature extraction by optimizing the conventional deep residual network and adding dilated convolution, and the other is to achieve multi-scale feature fusion by combining the pyramid principle which can connect deep and shallow features. The superiority and applicability of the model are confirmed by numerical simulation and rolling bearing data.