Article
Engineering, Mechanical
Lang Xu, Steven Chatterton, Paolo Pennacchi
Summary: A new method combining singular value decomposition and squared envelope spectrum for bearing fault diagnosis is proposed in this paper, and its effectiveness is demonstrated through the evaluation of actual vibration signals.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Engineering, Mechanical
Bingyan Chen, Weihua Zhang, James Xi Gu, Dongli Song, Yao Cheng, Zewen Zhou, Fengshou Gu, Andrew Ball
Summary: In this paper, new detection methods of cyclostationarity are developed for rolling bearing fault diagnosis by constructing generalized envelope signals and using product envelope spectrum (PES), which improve the accuracy and robustness of fault diagnosis.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Multidisciplinary
Yongjian Sun, Shaohui Li
Summary: This article proposes a novel fault diagnosis method based on convolutional neural networks (CNN), which does not require human intervention and achieves high accuracy. By transforming vibration signals into symmetrical images and inputting them into the CNN, the fault type can be automatically diagnosed.
Article
Automation & Control Systems
Feiyu Lu, Qingbin Tong, Ziwei Feng, Qingzhu Wan
Summary: Bearing fault diagnosis is crucial for safe and reliable operation of mechanical systems, but current methods fail to effectively handle category imbalance and variable speeds in practical scenarios. A novel fault diagnosis method is proposed based on spectrum alignment (SA) and deep transfer convolution neural network (DTCNN), which uses SA features and a data augmentation module to handle unbalanced bearing data. A DTCNN model with joint distribution adaptation is built to learn domain-invariant features and achieve superior diagnostic effect and generalization ability. Evaluation results show high accuracy, F1-score, and area under curve for 14 transfer tasks, surpassing state-of-the-art methods.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Engineering, Electrical & Electronic
Junbin Chen, Ruyi Huang, Kun Zhao, Wei Wang, Longcan Liu, Weihua Li
Summary: The MSCNN-FA method proposes a feature alignment module to enhance the shift-invariance of convolutional neural networks for bearing fault diagnosis. It utilizes a multiscale convolution strategy to extract robust features and constructs a classifier with fully connected layers, outperforming other CNN-based methods in terms of diagnosis accuracy and feature robustness in bearing experiments.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Mechanical
Boyao Zhang, Yonghao Miao, Jing Lin, Yinggang Yi
Summary: The adaptive CYCBD method proposed in this article uses EHPS to accurately estimate the target cyclic frequency or period, showing robustness in fault period identification. Compared to the original CYCBD, ACYCBD can extract weak impulses without prior knowledge of the period.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Engineering, Multidisciplinary
Chunran Huo, Quansheng Jiang, Yehu Shen, Chenhui Qian, Qingkui Zhang
Summary: A modified ADC-CNN model and improved LATL training method were proposed for fault diagnosis of rolling bearings. Experimental results showed that the proposed method significantly outperformed the traditional 1D-CNN model on the PU dataset.
Article
Multidisciplinary Sciences
Meng Xu, Yaowei Shi, Minqiang Deng, Yang Liu, Xue Ding, Aidong Deng
Summary: This paper proposes an improved multiscale branching convolutional neural network for rolling bearing fault diagnosis. The method extracts and processes complex vibration signals using multiscale feature learning and lightweight dynamic separable convolution to improve the accuracy and computational efficiency of fault diagnosis.
Article
Engineering, Electrical & Electronic
Xiaolin Liu, Jiani Lu, Zhuo Li
Summary: In this paper, a fault diagnosis method based on multiscale fusion attention CNN (MSFACNN) is proposed to address the nonstationary characteristics of aircraft engine rolling bearings. By converting vibration signals into images and using multiscale convolution and attention mechanisms to extract and weigh fault features, higher recognition accuracy is achieved.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Mechanical
Fengping An, Jianrong Wang
Summary: This paper proposes a rolling bearing fault diagnosis algorithm based on the overlapping group sparse model-deep complex convolutional neural network. This algorithm solves the difficulties in extracting the composite fault signal features of rolling bearings and the problem of multi-scale information. The experimental results show that the proposed method can effectively identify rolling bearing faults under constant and changing operating conditions, and it has a higher classification accuracy compared to traditional machine learning methods.
NONLINEAR DYNAMICS
(2022)
Article
Automation & Control Systems
Chunran Huo, Quansheng Jiang, Yehu Shen, Qixin Zhu, Qingkui Zhang
Summary: Deep transfer learning is used to solve the problem of unsupervised intelligent fault diagnosis of rolling bearings. However, when the data distribution between two domains is different, the existing deep transfer learning models are not enough to complete the target domain data learning. An enhanced transfer learning method based on the linear superposition network is proposed for rolling bearing fault diagnosis. Experimental results show improved bearing fault diagnostic precision compared to traditional feature-based transfer learning methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Industrial
Yang Guan, Zong Meng, Dengyun Sun, Jingbo Liu, Fengjie Fan
Summary: With the development of technologies such as Internet of Things and big data, the fusion and cross analysis of multi-sensor signals provide the possibility for comprehensive condition monitoring and intelligent fault diagnosis of rolling bearings. In this study, an intelligent fault diagnosis method based on information fusion and parallel lightweight convolutional neural network is proposed. Experimental results show the effectiveness and superiority of this method.
JOURNAL OF MANUFACTURING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Zhuyun Chen, Wu Qin, Guolin He, Jipu Li, Ruyi Huang, Gang Jin, Weihua Li
Summary: This study proposes an explainable deep ensemble model for fault diagnosis of bearings, which combines feature learning ability and model interpretability. It uses a residual network integrated with bidirectional long short-term memory for feature extraction and fusion, and a visualization technique to provide better understanding of learned features and decision-making. Experimental results demonstrate the superior explainable capability and diagnosis performance of the proposed approach compared to other methods.
IEEE SENSORS JOURNAL
(2023)
Article
Acoustics
Yanwei Xu, Weiwei Cai, Liuyang Wang, Tancheng Xie
Summary: This study proposes an intelligent diagnosis method for rolling bearing faults based on an improved convolutional neural network and a lightweight gradient boosting machine, which improves the model's generalization ability and fault diagnosis efficiency through feature extraction and classification.
SHOCK AND VIBRATION
(2021)
Article
Engineering, Electrical & Electronic
Xiaoping Liu, Lijian Xia, Jian Shi, Lijie Zhang, Linying Bai, Shaoping Wang
Summary: This article proposes an antinoise bearing fault diagnosis method based on improved RP and a convolutional neural network (CNN). It constructs RP by obtaining different scales of approximation coefficients and detail coefficients through wavelet packet decomposition (WPD) on the vibrational signal. Redundant parts of each RP are removed based on symmetry characteristics, and the remaining parts are spliced into multiscale asymmetric RP (MARP). A fault diagnosis model for rolling bearing is established using MARP as input for the pretrained ResNet-34. The proposed method achieves an accuracy of 90% under Gaussian white noise with a signal-to-noise ratio (SNR) of above -6 dB, as validated on the Paderborn bearing dataset.
IEEE SENSORS JOURNAL
(2023)