Intelligent fault diagnosis of rotating components in the absence of fault data: A transfer-based approach. (March 2021)

• Record Type: Journal Article
• Authors: Deng, Minqiang; Deng, Aidong; Zhu, Jing; Shi, Yaowei; Liu, Yang
• Is Part Of: Measurement. Volume 173(2021)
• Publication Date: 2021-03
• Subjects: Fault diagnosis -- Machine learning -- Transfer learning -- Order analysis -- Rolling bearing -- Gearbox; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2020.108601
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: An OSTFD method is proposed to build the intelligent fault diagnosis model. The proposed transfer learning algorithm is independent of fault samples. The transfer strategy in the OST algorithm is determined by defect orders. The OSTFD achieves promising performance in diagnosis scenarios lacking fault data. Abstract: This paper focuses on the intelligent fault diagnosis (IFD) of rotating components in the absence of fault data. Specifically, an Order Spectrum Transfer based Fault Diagnosis (OSTFD) method is proposed to establish IFD models for the target component by exploiting the monitoring data of other related machines. Considering the variable operating conditions, Bandwidth Fourier Decomposition method and Hilbert Order Transform algorithm are introduced in OSTFD to extract the envelope order spectrum (EOS) that is insensitive to unsteady speed and load for pattern recognition. Then, based on the fault mechanism, a novel Order Spectrum Transfer algorithm is proposed to transform the fault characteristics (EOS) of the target data to the source domain, in which the classifier based on one-dimensional convolutional neural network is trained. Experimental results based on four benchmark datasets demonstrate the effectiveness and superiority of the proposed OSTFD in actual applications lacking complete samples.