Article
Computer Science, Artificial Intelligence
Jinyang Jiao, Jing Lin, Ming Zhao, Kaixuan Liang, Chuancang Ding
Summary: Driven by industrial big data and intelligent manufacturing, deep learning approaches have made significant achievements in machine fault diagnosis. However, current diagnosis models trained on specific datasets may not perform well on others due to domain discrepancy. The CAAN method based on adversarial domain adaptation shows effectiveness in addressing this issue and demonstrates superiority in experiments on different mechanical systems.
Review
Chemistry, Analytical
Yu Guo, Jundong Zhang, Bin Sun, Yongkang Wang
Summary: Deep Transfer Learning (DTL) is a novel paradigm in machine learning that combines deep learning's superior feature representation with transfer learning's knowledge transference. Adversarial Deep Transfer Learning (ADTL) emerged as a response to the challenges faced by early DTL paradigms. This review categorizes ADTL into non-generative and generative models, and examines its recent advancements in the Intelligent Fault Diagnosis (IFD) field.
Article
Computer Science, Artificial Intelligence
Zhenhua Fan, Qifa Xu, Cuixia Jiang, Steven X. Ding
Summary: In this study, a new fault diagnosis network called DWQDAN is proposed, which addresses the challenges of limited labeled data and domain shifts using a weighted quantile discrepancy metric and adversarial learning. Experimental results demonstrate that DWQDAN outperforms other methods on public datasets and can be applied effectively for practical fault diagnosis by adjusting parameters.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Chemistry, Multidisciplinary
Heng Chen, Lei Shi, Shikun Zhou, Yingying Yue, Ninggang An
Summary: This paper introduces a multi-source consistency domain adaptation neural network MCDANN, which uses sub-domain division alignment and multi-source prediction consistency to improve the transfer accuracy of the fault diagnosis model.
APPLIED SCIENCES-BASEL
(2022)
Article
Mathematics, Applied
Ziwei Deng, Zhuoyue Wang, Zhaohui Tang, Keke Huang, Hongqiu Zhu
Summary: The paper presents a cross-domain fault diagnosis method based on transferred stacked autoencoder, which extracts features from the source domain data to establish a model and then performs domain adaptation with target domain data, demonstrating its effectiveness and superiority through experiments.
APPLIED MATHEMATICS AND COMPUTATION
(2021)
Article
Engineering, Mechanical
Weihua Li, Ruyi Huang, Jipu Li, Yixiao Liao, Zhuyun Chen, Guolin He, Ruqiang Yan, Konstantinos Gryllias
Summary: Deep Transfer Learning combines the advantages of Deep Learning in feature representation and Transfer Learning in knowledge transfer, making DL-based fault diagnosis methods more reliable and robust. However, further research is needed to explore the potential of DTL-based approaches in Intelligent Fault Diagnosis.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Engineering, Industrial
Biliang Lu, Yingjie Zhang, Zhaohua Liu, Hualiang Wei, Qingshuai Sun
Summary: This paper proposes a universal unsupervised domain adaptation (UUDA) method to address the issues of domain and label space shift in fault diagnosis. By utilizing outlier threshold learning, domain-invariant sampler, and adversarial classifier training, it achieves exceptional performance in handling label space inconsistencies across different domains.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Computer Science, Artificial Intelligence
Zhenyu Wu, Hongkui Zhang, Juchuan Guo, Yang Ji, Michael Pecht
Summary: A deep adversarial transfer learning model (Deep Imba-DA) is proposed in this paper to tackle the issues of class imbalances and distributional discrepancies in bearing fault diagnosis.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Yu Wang, Xiaojie Sun, Jie Li, Ying Yang
Summary: This study introduces a deep adversarial domain adaptation network (DADAN) to transfer fault diagnosis knowledge using domain-adversarial training and a supervised instance method for better feature alignment.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Electrical & Electronic
Wentao Mao, Yamin Liu, Ling Ding, Ali Safian, Xihui Liang
Summary: The article introduces a new deep transfer learning method, SDANN, which integrates DANN and structured relatedness information among multiple failure modes to improve transfer learning. By utilizing a new loss function, relatedness matrix, and symmetry constraint regularizer, the method achieves stability on insufficient data and good diagnostic performance.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Mechanical
Wanxiang Li, Zhiwu Shang, Maosheng Gao, Fei Liu, Hu Liu
Summary: The main challenge of fault diagnosis models based on partial domain adaptation (PDA) is to promote positive transfer in the shared label space and avoid negative transfer caused by the mismatch between the outlier and the target label spaces. To address this challenge, this paper proposes a partial deep transfer diagnosis model (MICDDA) based on multi-representation structure intraclass compact and double-aligned domain adaptation. Experimental results in two case studies show that the MICDDA can effectively weaken the transfer of source outlier knowledge to the target domain, capture rich and homogeneous compact fault features in PDA diagnosis scenarios, and effectively improve diagnostic performance.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Review
Engineering, Aerospace
Siyu Zhang, Lei Su, Jiefei Gu, Ke LI, Lang Zhou, Michael Pecht
Summary: In practical mechanical fault detection and diagnosis, transfer learning combined with deep learning can improve the performance of the target task while reducing the demand for large-scale supervised data and high computation power. However, direct transfer may lead to a significant reduction in detection performance due to domain differences. Domain adaptation strategies can address this issue by transferring distribution information from the source domain to the target domain. This survey reviews various current domain adaptation strategies combined with deep learning and analyzes their principles, advantages, and disadvantages, as well as their application in fault diagnosis.
CHINESE JOURNAL OF AERONAUTICS
(2023)
Article
Engineering, Mechanical
Dongdong Wei, Te Han, Fulei Chu, Ming Jian Zuo
Summary: This study focuses on intelligent fault diagnosis of machines in the context of changing working conditions. A multiple source domain adaptation method is proposed to learn fault-discriminative but working condition-invariant features to address the data distribution shift issue.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Chemistry, Analytical
Wangpeng He, Jing Chen, Yue Zhou, Xuan Liu, Binqiang Chen, Baolong Guo
Summary: In this paper, a novel fault diagnosis method combining a generative adversarial network and transfer learning is proposed to address the problems of insufficient data and data distribution variations under different working conditions. The method generates dummy samples and extracts transfer fault characteristics to improve the accuracy and applicability of the diagnostic model.
Article
Computer Science, Interdisciplinary Applications
Yafei Deng, Delin Huang, Shichang Du, Guilong Li, Chen Zhao, Jun Lv
Summary: Recently, deep transfer learning approaches have been widely developed for mechanical fault diagnosis issue. DA-GAN model shows great superiority in dealing with mechanical partial transfer problem in both TIM and TDM.
COMPUTERS IN INDUSTRY
(2021)
Article
Mathematics
Abhijeet Ainapure, Shahin Siahpour, Xiang Li, Faray Majid, Jay Lee
Summary: This paper proposes a new cross-domain fault diagnosis method with enhanced robustness. By introducing noisy labels to significantly increase the generalization ability of the data-driven model, promising diagnosis performance can be obtained even with strong noise interference and low-quality data. Experimental validation on rotating machinery datasets demonstrates the applicability and performance of the proposed method in real industrial environments.
Article
Engineering, Electrical & Electronic
Wei Zhang, Jiaxuan Zhang, Xiang Li
Summary: This study proposes a novel deep-learning-based method for intelligently identifying the oil film coefficient of SFDs. Experimental results demonstrate that the method can accurately handle high-dimensional data and achieve high identification accuracy.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Mechanical
Xiaofei Liu, Yaguo Lei, Naipeng Li, Xiaosheng Si, Xiang Li
Summary: A novel convolutional vector fusion network (C-VFN) is proposed in this paper to improve the prediction accuracy of remaining useful life (RUL) of machinery by dynamically evaluating the degradation sensitivity of each feature.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Industrial
Wei Zhang, Ziwei Wang, Xiang Li
Summary: In this paper, a blockchain-based decentralized federated transfer learning method is proposed for collaborative machinery fault diagnosis. The experimental results indicate that the proposed method is effective in data privacy-preserving collaborative fault diagnosis of multiple users, offering a promising tool for applications in the real industrial scenarios.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Engineering, Industrial
Xu Li, Chi Zhang, Xiang Li, Wei Zhang
Summary: In the past decades, significant developments and applications have been made in data-driven machinery fault diagnosis methods. However, due to data privacy concerns and the disparity in data conditions among different users, collaborative training of powerful models has become challenging. To address these issues, a federated transfer learning method is proposed, which involves generating fake data, communicating only the models, and introducing prediction alignment and consensus schemes. Experiments on real-world machinery datasets validate the effectiveness of the proposed method, making it promising for real industry applications.
JOURNAL OF MANUFACTURING SYSTEMS
(2023)
Article
Engineering, Mechanical
Naipeng Li, Yaguo Lei, Xiang Li, Xiaofei Liu, Bin Yang
Summary: Nonparametric degradation modeling automatically formulates the main degradation pattern of condition monitoring signals and offers higher flexibility in describing complex degradation processes. Functional principal component analysis (FPCA) is a commonly used nonparametric modeling technique, but is limited by the requirement that different units must share the same scale. This paper proposes a new degradation modeling method based on FPCA by axis rotation, which addresses the issues of time function existence and flexibility in noisy signals, and utilizes a particle filtering algorithm for non-Gaussian state estimation.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Editorial Material
Mathematics
Xiang Li, Shuo Zhang, Wei Zhang
Article
Automation & Control Systems
Shuhui Wang, Yaguo Lei, Bin Yang, Xiang Li, Yue Shu, Na Lu
Summary: This paper proposes a graph neural network-based data-cleaning method to address the difficulty of acquiring high-quality training data for deep learning in mechanical fault diagnosis. The method includes two stages: group anomaly detection and graph clustering, which effectively detect anomalous data and rectify incorrect labeling, improving the performance of mechanical fault diagnosis.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Yafei Zhu, Xiang Li, Yewei Zhang, Wei Zhang
Summary: In this article, a model based on convolutional neural network is proposed to predict the health index of lithium-ion batteries. By training with source domain and target domain data, the model gradually reduces the domain gap and achieves accurate predictions.
IEEE SENSORS JOURNAL
(2023)
Article
Automation & Control Systems
Bin Yang, Yaguo Lei, Xiang Li, Clive Roberts
Summary: This article proposes a deep targeted transfer learning (DTTL) method for fault diagnosis, which addresses the issue of data distribution shift and facilitates diagnosis knowledge transfer across related machines. The method relaxes the strict assumption that all target domain data are unlabeled by introducing labeled one-shot target domain samples called anchors. DTTL includes a domain-shared residual network, a target-domain clustering module, and a targeted adaptation module to correct the joint distribution shift. Experimental results on transfer diagnosis tasks across different bearings demonstrate that DTTL achieves higher diagnosis accuracy compared to other methods.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Engineering, Mechanical
Xiwei Li, Yaguo Lei, Mingzhong Xu, Naipeng Li, Dengke Qiang, Qubing Ren, Xiang Li
Summary: This paper proposes a fault diagnosis method for automotive transmissions considering gear shifting, and effectively addresses the challenges introduced by gear shifting through a spectral variation sparsity indicator (SVSI) and a weighted health indicator (WHI). The experimental results demonstrate that the proposed method is capable of detecting faults prior to inspection and accurately identifying faulty gears, making significant contributions to the field of condition monitoring and fault locating for automotive transmissions.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Automation & Control Systems
Xiang Li, Wei Zhang, Xu Li, Hongshen Hao
Summary: Intelligent machinery prognostics and health management methods have gained attention in recent years. This article proposes a partial domain adaptation method for remaining useful life prediction with incomplete target-domain data.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)