Article
Computer Science, Artificial Intelligence
Zhiqiang Zhang, Qingyu Yang
Summary: The article introduces an intelligent fault diagnosis method based on reconstruction sparse filtering (RSF), which extracts diverse features by constraining the basis vectors, enabling precise description of the health conditions of rotating machinery and achieving significant performance improvement.
APPLIED SOFT COMPUTING
(2022)
Article
Engineering, Multidisciplinary
Chun Cheng, Wenyi Liu, Weiping Wang, Michael Pecht
Summary: A novel fault diagnosis method using deep variant sparse filtering network (DVSFN) is introduced in this paper, which can overcome the deficiencies of traditional methods and is verified with rolling bearing and planetary gearbox datasets. Experimental results show that this method can achieve higher accuracy and robustness.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2021)
Article
Automation & Control Systems
Taewan Kim, Seungchul Lee
Summary: This article proposes a new unsupervised clustering and domain adaptation framework to address data deficiency and domain issues in deep-learning-based fault diagnosis methods. The framework consists of two steps: unsupervised clustering and domain adaptation. In unsupervised clustering, an expectation-maximization adversarial autoencoder is used for feature extraction and subspace mapping, followed by clustering using a Gaussian mixture model. In domain adaptation, a domain synchronization technique based on symmetric Kullback-Leibler divergence is employed to infer the relationship between source and target domain clusters. Experimental results on two rolling-element-bearing datasets validate the effectiveness of the proposed method, particularly its ability to perform domain adaptation without retraining, which holds promise for real industrial applications.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Hardware & Architecture
Chun Cheng, Yan Hu, Jinrui Wang, Haining Liu, Michael Pecht
Summary: This study developed a generalized sparse filtering (GSF) method and applied it to rotating machinery fault diagnosis, demonstrating its excellent performance through experiments. Additionally, the influences of normalization parameters on diagnostic performance were investigated to determine the best parameter combinations.
JOURNAL OF SUPERCOMPUTING
(2021)
Article
Chemistry, Multidisciplinary
Jiantao Lu, Weiwei Qian, Shunming Li, Rongqing Cui
Summary: A new intelligent fault diagnosis method of rotating machinery is proposed based on enhanced KNN, which combines parameter-based and case-based methods effectively. The method includes a dimension-reduction stage using sparse filtering for feature extraction, and a case-based reconstruction algorithm to adaptively determine the nearest neighbors for different testing samples. Experimental results on vibration signal datasets of bearings validate the effectiveness of the proposed method.
APPLIED SCIENCES-BASEL
(2021)
Article
Engineering, Mechanical
Zhongwei Zhang, Mingyu Shao, Chicheng Ma, Zhe Lv, Jilei Zhou
Summary: In this study, a novel domain adaptation approach is proposed for fault diagnosis of rotating machinery. The approach utilizes a deep sparse filtering model to extract fault features and a domain classifier to perform domain shift, while Z-score standardization and CORAL are employed as preprocessing tools. The effectiveness of the approach is verified through experimental vibration data from a bearing and a gear dataset.
NONLINEAR DYNAMICS
(2022)
Article
Automation & Control Systems
Guowei Zhang, Xianguang Kong, Jingli Du, Jinrui Wang, Shengkang Yang, Hongbo Ma
Summary: This study proposes an intelligent fault diagnosis method based on sparse feature learning, called adaptive multispace adjustable sparse filtering (AMSASF). The method automatically captures rich and complementary features under multiple spaces using multispace sparse filtering and improves the robustness of the algorithm by adaptively assigning different importance to different sparse spaces using attention mechanism. The sparsity is adjusted to increase the inter-class distance and obtain more discriminative features.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Automation & Control Systems
Li Wang, Sai Ma, Qinkai Han
Summary: The article introduces an enhanced sparse low-rank representation approach for weak fault feature extraction in rotary machinery. By utilizing a weighted dual approximation regularization to suppress noise and irrelevant harmonics, it effectively extracts weak fault features.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Engineering, Mechanical
Zhiqiang Zhang, Shuiqing Xu, Hongtian Chen
Summary: Representation learning has powerful potential in intelligent fault diagnosis of rotating machinery. However, current sparse filtering-based methods have limitations in handling complex signals and selecting relevant features for fault classification. To address these issues, label-induced sparse filtering (LISF) is proposed, which utilizes discriminant information from labels and measures feature importance with a projection matrix. Experimental results demonstrate the effectiveness of LISF in learning discriminative features and achieving excellent diagnosis results. Moreover, LISF can automatically select key features for fault classification, improving diagnosis efficiency.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Electrical & Electronic
Hongru Cao, Haidong Shao, Bin Liu, Baoping Cai, Junsheng Cheng
Summary: This paper proposes a clustering-guided novel unsupervised domain adversarial network to address the problem of unsupervised partial transfer fault diagnosis. The network, constructed using domain-specific batch normalization, eliminates domain-specific information and enhances alignment between source and target domains. Additionally, an embedded clustering strategy is designed to learn tightly clustered target-domain features and suppress negative transfer.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Multidisciplinary
Baokun Han, Zongling Liu, Zongzhen Zhang, Jinrui Wang, Huaiqian Bao, Zujie Yang, Shuo Xing, Xingwang Jiang, Bo Li
Summary: This research proposes a parallel network model based on intrinsic component filtering (PICF) to address the long upgrade time and retraining required in fault diagnosis systems. Experimental results demonstrate that this method achieves higher fault classification accuracy under small sample training and when increasing fault types.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2023)
Article
Engineering, Mechanical
Zongzhen Zhang, Jinrui Wang, Shunming Li, Baokun Han, Xingxing Jiang
Summary: Sparse optimization based early fault diagnosis method has gained attention recently. By using objective function and nonlinear blind deconvolution algorithm, the fault representation and convergence stability can be improved. The proposed method utilizes an improved sigmoid function, Gaussian fitting window function, and L1/2 penalty to enhance the weight vector's distribution and performance. Experimental results confirm the effectiveness of the method, which significantly improves noise adaptability, computation effectiveness, and the robustness of fault diagnosis.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Mechanical
Chun Cheng, Wei Zou, Weiping Wang, Michael Pecht
Summary: In this study, a deep sparse filtering network (DSFN) is proposed for intelligent fault diagnosis of rotating machinery. Pre-training with sparse filtering and fine-tuning using back-propagation algorithm can help DSFN adaptively learn discriminative features from datasets and achieve higher diagnostic accuracy with fewer training samples compared to classical methods.
PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART D-JOURNAL OF AUTOMOBILE ENGINEERING
(2022)
Article
Engineering, Mechanical
Yi Qin, Jiahong Yang, Jianghong Zhou, Huayan Pu, Xiangfeng Zhang, Yongfang Mao
Summary: In actual industrial scenarios, the centralized learning paradigm for remaining useful life (RUL) prediction of rotating machineries faces challenges due to difficulty in collecting sufficient degradation data and isolated data island among different users. To address this, a dynamic weighted federated RUL prediction framework is proposed, which uses a cloud server and multiple edge clients to train prediction models without accessing raw data. Experimental results on gear and bearing datasets demonstrate the effectiveness and superiority of the proposed method.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Automation & Control Systems
Jiahong Chen, Jing Wang, Jianxin Zhu, Tong Heng Lee, Clarence W. de Silva
Summary: This article tackles the issue of cross-domain fault diagnosis of rotating machinery by developing an unsupervised domain adaptation method to mitigate domain shifts between data domains. It aims to improve the transferability of labeled data in the source domain by maximizing mutual information between the target feature space and the entire feature space, while minimizing feature-level discrepancies between the two domains to enhance diagnosis accuracy. Experiment results demonstrate the feasibility and superior performance of the proposed method using public datasets and real-world adaptation scenarios.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2021)
Article
Automation & Control Systems
Xiaoli Zhao, Minping Jia, Peng Ding, Chen Yang, Daoming She, Zheng Liu
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2020)
Article
Automation & Control Systems
Xiaoli Zhao, Minping Jia, Zheng Liu
Summary: This article proposes an intelligent fault diagnosis method for electromechanical systems based on a new semisupervised graph convolution deep belief network algorithm, which can achieve high diagnostic accuracy with a small amount of labeled data.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Xiaoli Zhao, Jianyong Yao, Wenxiang Deng, Peng Ding, Yifei Ding, Minping Jia, Zheng Liu
Summary: This paper proposes an intelligent fault diagnosis method for gearboxes under variable working conditions based on adaptive intraclass and interclass convolutional neural network (AIICNN). By applying intraclass and interclass constraints to improve sample distribution differences, and using an adaptive activation function to enlarge the heterogeneous distance and narrow the homogeneous distance of samples, the feasibility of the proposed method is verified through experimental data.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Yifei Ding, Peng Ding, Xiaoli Zhao, Yudong Cao, Minping Jia
Summary: This article proposes a new framework for predicting the remaining useful life of bearings based on a multisource domain adaptation network (MDAN). By learning domain-invariant features and supervision from multiple sources, MDAN achieves better generalization performance. Case studies and comparisons with other methods validate the effectiveness of the proposed approach.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Automation & Control Systems
Xiaoli Zhao, Jianyong Yao, Wenxiang Deng, Peng Ding, Jichao Zhuang, Zheng Liu
Summary: This article proposes a new algorithm called multiscale deep graph convolutional networks (MS-DGCNs) to alleviate the problem of significant disordered fluctuations in the measured signals of the rotor-bearing system. An intelligent fault diagnosis method based on MS-DGCNs is designed to improve feature representations and accuracy. Experimental results demonstrate the higher accuracy and generalization of the method.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Automation & Control Systems
Yudong Cao, Minping Jia, Peng Ding, Xiaoli Zhao, Yifei Ding
Summary: Deep neural networks have been effective in fault classification and remaining useful life (RUL) prediction for mechanical equipment. However, traditional deep learning models are limited by network depth and require retraining for updating parameters. To address these issues, an incremental learning method based on temporal cascade broad learning system (TCBLS) is proposed. This method achieves high prediction accuracy, saves training time consumption, and handles newly acquired data without retraining.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Engineering, Mechanical
Chaoyang Weng, Baochun Lu, Qian Gu, Xiaoli Zhao
Summary: This paper proposes a novel hierarchical transferable network (HTNet) to effectively identify the fault pattern and severity of rolling bearings, providing reasonable repair plans. The HTNet utilizes a two-layer hierarchical structure and an adaptive subnet selection module to establish the correlation between fault patterns and severity levels. It also introduces a hierarchical domain adaptation method to extract domain-invariant features from different classification tasks, resulting in better performance and transferability under variable working conditions.
NONLINEAR DYNAMICS
(2023)
Article
Automation & Control Systems
Peng Ding, Minping Jia, Yifei Ding, Yudong Cao, Jichao Zhuang, Xiaoli Zhao
Summary: Few-shot learning based machinery prognostics are feasible for intelligent operation and maintenance with scarce monitoring data. To improve the reliability of predictive maintenance, a novel Bayesian approximation enhanced probabilistic meta-learning (BA-PML) algorithm is proposed to estimate uncertainty in few-shot prognostics. This algorithm consists of a designed base probabilistic predictor and an episodic training strategy.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)
Article
Engineering, Electrical & Electronic
Yudong Cao, Jichao Zhuang, Minping Jia, Xiaoli Zhao, Xiaoan Yan, Zheng Liu
Summary: This article proposes a complex graph neural network (CGNN-PIP) based on the picture-in-picture strategy for the remaining useful life (RUL) prediction of rotating machinery under multichannel signals. The classical graph convolution operation is upgraded to extract deep degenerate feature representations, and the picture-in-picture strategy is designed to guide graph construction. The effectiveness and superiority of the proposed method are verified through two case studies on different run-to-failure datasets. Results show that CGNN-PIP can reasonably construct the topology map of complex domain data and extract temporal and structural information reflecting equipment degradation. Comparisons with state-of-the-art methods demonstrate advantages in prediction accuracy and training consumption.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Electrical & Electronic
Chaoyang Weng, Baochun Lu, Qian Gu, Xiaoli Zhao
Summary: In this article, a novel multisensor fusion transformer (MsFT) with self-attention mechanism is proposed for rotating machinery fault diagnosis. Experimental results show that the MsFT exhibits excellent robustness against noise interference, and outperforms other methods in terms of accuracy and reliability.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Automation & Control Systems
Xingjun Zhu, Xiaoli Zhao, Jianyong Yao, Wenxiang Deng, Haidong Shao, Zheng Liu
Summary: This method proposes an intelligent fault diagnosis method for servo motor-cylindrical roller bearings under variable working conditions based on adaptive multiscale convolution manifold embedding networks (AMCMENet). It applies designed multiscale convolutional neural networks (MSCNN) to extract features and uses intraclass and interclass constraints for reprocessing, improving the distribution differences of samples. The method also optimizes the parameters of the constructed locality sensitive discriminant analysis algorithm module using the particle swarm optimization algorithm. Experimental results show that the proposed method performs better under cross working conditions.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)
Article
Engineering, Electrical & Electronic
Xiaoli Zhao, Minping Jia, Zheng Liu
Summary: The article introduces an intelligent fault diagnosis method for rotating machinery based on semisupervised deep sparse auto-encoder (SSDSAE) with local and nonlocal information. Vibration spectrum signals are fed into the SSDSAE algorithm for fault feature extraction, and the extracted sparse discriminant features are used for fault diagnosis with a back-propagation (BP) classifier. The method utilizes weighted cross-entropy (WCE) techniques to improve the generalization performance of the fault diagnosis model and is validated with experimental data.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Information Systems
Xiaoli Zhao, Minping Jia, Zheng Liu
Article
Automation & Control Systems
Yifei Ding, Peng Ding, Xiaoli Zhao, Yudong Cao, Minping Jia
Summary: This article introduces a multisource domain adaptation network (MDAN) for improving prognostics and health management of rotating machinery. MDAN effectively utilizes historical data from multiple sources, learns domain-invariant features, and achieves better generalization in the target domain.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)