Article
Automation & Control Systems
Shijin Li, Jianbo Yu
Summary: Transfer learning-based process fault diagnosis has been widely studied, but there is still a challenge in handling multisource domain adaptation under various working conditions. This article proposes a novel transfer learning model, FC-MSDA, for process fault diagnosis. It introduces a common feature extractor, a feature selection module, domain specific feature generators, and a class-level distribution alignment loss to address the challenges. The experimental results demonstrate the effectiveness of FC-MSDA in process fault diagnosis.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Automation & Control Systems
Ziling Huang, Zihao Lei, Guangrui Wen, Xin Huang, Haoxuan Zhou, Ruqiang Yan, Xuefeng Chen
Summary: In this study, a novel fault diagnosis method is proposed that utilizes multisource information fusion and classification label information. By extracting features and fusing them, along with a joint loss function, the method effectively addresses fault diagnosis under polytropic working conditions. Experimental results demonstrate the method's great potential.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Engineering, Electrical & Electronic
Changqing Shen, Yu Xia, Xingxing Jiang, Zaigang Chen, Lin Kong, Zhongkui Zhu
Summary: This study proposes a new multi-source transfer learning method, named OMSSN, to address bearing fault diagnosis under variable working conditions.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Buyao Yang, Tiantian Wang, Jingsong Xie, Jinsong Yang
Summary: This article proposes a novel domain adaptation model DAHAN for high-speed train bogie fault diagnosis. The method combines local and global domain adaptation to achieve hybrid domain adaptation and improve transfer learning performance at low signal-to-noise ratio. An optimized training procedure is constructed to enhance the stability and real-time performance of the proposed model. The effectiveness and advantage of DAHAN are verified using the Case Western Reserve University (CWRU) bearing dataset, and the proposed method shows significant performance improvement and applicability on train bogie fault diagnosis.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Automation & Control Systems
Zheng Chai, Chunhui Zhao
Summary: This article introduces a fault-prototypical adapted network for cross-domain industrial intelligent fault diagnosis using deep transfer learning. Experimental results show that the proposed approach learns transferable feature representations that reduce domain discrepancy and improve diagnosis performance on target data.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Rui Wang, Weiguo Huang, Jun Wang, Changqing Shen, Zhongkui Zhu
Summary: This study proposes a novel fault diagnosis method based on multisource domain adaptation, which achieves the diagnosis of bearing faults under time-varying working conditions by learning transferable features and reducing domain shift. Experimental results demonstrate the effectiveness, robustness, and superiority of the proposed method on multiple real datasets.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Engineering, Mechanical
Yu Xia, Changqing Shen, Dong Wang, Yongjun Shen, Weiguo Huang, Zhongkui Zhu
Summary: This study introduces a deep learning-based fault diagnosis method that utilizes multi-source transfer learning to address the issues of insufficient labels and different distributions in data.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Automation & Control Systems
Yaoxiang Yu, Liang Guo, Yongwen Tan, Hongli Gao, Jiangquan Zhang
Summary: In real industrial applications, obtaining massive labeled data for fault diagnosis of machineries is difficult. Therefore, transfer learning is introduced to apply knowledge learned from labeled datasets to unlabeled data. However, there are challenges such as unknown label space, limited fault types in labeled datasets, and difficulty in applying ideal datasets to real data. To solve these problems, a new transfer learning model called multisource partial transfer network is proposed, which consists of a common module and three domain-specific modules for feature extraction, fault diagnosis, and domain adaptation.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Engineering, Electrical & Electronic
Danya Xu, Yibin Li, Yan Song, Lei Jia, Yanjun Liu
Summary: Intelligent Fault Diagnosis System utilizes multi-source unsupervised domain adaptive network to address fault diagnosis issues under different source domain conditions. The method considers source domain variances, uses source domain data and a small amount of target domain data to extract feature information.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Automation & Control Systems
Yidan Hu, Ruonan Liu, Xianling Li, Dongyue Chen, Qinghua Hu
Summary: A task-sequencing meta-learning method is proposed in this article to address the few-shot fault diagnosis problem. By training a meta-learning model over a series of learning tasks, the method is able to adapt and generalize knowledge with only a few examples. The effectiveness of the method is validated through experiments.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Computer Science, Information Systems
Yang Xiao, Qingfeng Wang, Shuai Wang, Wenwu Chen
Summary: This paper proposes a multi-source domain enhanced method for fault diagnosis of rotor systems. By using improved adaptive variational mode decomposition and weighted semi-supervised transfer component analysis, the method enhances fault diagnosis accuracy and domain generalization performance. Experimental results demonstrate that the method has good application and promotion value in solving cross equipment, cross working condition, and cross domain diagnostic tasks.
Article
Engineering, Electrical & Electronic
Lu Zhang, Hua Li, Jie Cui, Wei Li, Xiaodong Wang
Summary: An innovative class subdomain adaptation network (CSAN) is proposed to solve the problem of inconsistent distribution of bearing data under variable working conditions. The CSAN model consists of a lightweight channel convolution neural network (CCNN) for feature extraction and classification, and an innovative domain adaptation algorithm as a loss function embedded in the network. Experimental results demonstrate the superiority of the proposed CSAN model in achieving optimal performance and desirable generalization performance under changeable working conditions.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Automation & Control Systems
Chao Zhao, Weiming Shen
Summary: This article proposes an adversarial mutual information-guided single domain generalization network for machinery fault diagnosis, which learns domain-invariant representations to address domain shift problems. A domain generation module is designed to generate fake target domains with significant distribution discrepancies, and an iterative min-max game of mutual information is implemented to learn generalized features for resisting unknown domain shift. Extensive diagnosis experiments on two mechanical rigs validated the effectiveness of the proposed method.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Yong Feng, Jinglong Chen, Shuilong He, Tongyang Pan, Zitong Zhou
Summary: This paper proposes a globally localized multisource domain adaptation method with category shift for cross-domain fault diagnosis. By constructing a GlocalNet, which consists of a feature generator and three classifiers, multisource information is comprehensively fused. The Wasserstein discrepancy of classifiers is optimized locally and accumulative higher order multisource moment is used globally to achieve multisource domain adaptation at domain and class levels, thus reducing the shift on domain and category. A distilling strategy is presented to refine the classifier at sample level, and an adaptive weighting policy is employed for reliable result.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Shijin Li, Jianbo Yu
Summary: In this article, a novel deep transfer network (DTN) with adaptive joint distribution adaptation (AJDA) is proposed to solve the problem of process fault diagnosis under varying working conditions. The proposed method effectively reduces domain shifts by considering both marginal and conditional distributions adaptatively, resulting in more accurate diagnostic results.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Automation & Control Systems
Shunyi Zhao, Ke Li, Choon Ki Ahn, Biao Huang, Fei Liu
Summary: This article develops Bayesian estimation algorithms for estimating unforeseen signals in sensor outputs without tuning. A novel iterative algorithm using inverse Wishart distribution and variational inference technique is proposed to adaptively replace the effects of tuning parameters.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Automation & Control Systems
Wanke Yu, Chunhui Zhao, Biao Huang, Min Wu
Summary: In this study, a robust dissimilarity distribution analytics (RDDA) method is proposed for incipient industrial fault detection. The probabilistic model of the RDDA method is formulated with Laplace distribution, which provides better robustness compared to Gaussian distribution based models. By using variational inference, maximum likelihood estimations of latent variables and model parameters can be derived. A monitoring strategy is established based on static and dynamic statistics, utilizing dissimilarity between distributions of datasets. The proposed RDDA method is more suitable for practical industrial applications due to its consideration of missing data problems. Experimental results demonstrate the method's ability to accurately detect incipient faults using historical data with missing values.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Automation & Control Systems
Chao Jiang, Yusheng Lu, Weimin Zhong, Biao Huang, Dayu Tan, Wenjiang Song, Feng Qian
Summary: Inferential modeling plays a significant role in estimating quality-related process variables in modern manufacturing. This article proposes a new nonlinear extension of probabilistic slow feature analysis (PSFA) under the deep learning framework to enhance dynamic feature extraction and improve prediction accuracy by incorporating variational inference and Monte Carlo inference. The proposed model considers the relevance of inputs with outputs to enhance prediction performance. The model is validated through an industrial hydrocracking process and achieves a significant reduction in root mean squared error compared to PSFA.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Automation & Control Systems
Ronghu Chi, Huaying Li, Na Lin, Biao Huang
Summary: In this work, a data-driven indirect iterative learning control (DD-iILC) is proposed for a repetitive nonlinear system using a proportional-integral-derivative (PID) feedback control in the inner loop. A linear parametric iterative tuning algorithm for the set-point is developed from an ideal nonlinear learning function using an iterative dynamic linearization (IDL) technique. An adaptive iterative updating strategy of the parameter in the linear parametric set-point iterative tuning law is presented by optimizing an objective function for the controlled system. The convergence is proven using contraction mapping and mathematical induction, and the theoretical results are verified through simulations.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Interdisciplinary Applications
Junyao Xie, Oguzhan Dogru, Biao Huang, Chris Godwaldt, Brett Willms
Summary: Data-driven soft sensors have been widely used in the process industry for quality variable estimation. However, building reliable soft sensors for complex industrial processes under limited data conditions is challenging. To address this issue, we propose a reinforcement learning framework that leverages samples from source domains to solve the cross-domain soft sensor problem. The proposed framework incorporates a method for sample selection and soft sensor design, taking into account correlation and estimation error metrics.
COMPUTERS & CHEMICAL ENGINEERING
(2023)
Article
Automation & Control Systems
Ronghu Chi, Huaying Li, Dong Shen, Zhongsheng Hou, Biao Huang
Summary: In this article, the authors propose an indirect adaptive iterative learning control scheme to enhance the performance of P-type controllers by learning from set points. Adaptive mechanism is included to regulate the learning gain using real-time measurements. The proposed methods are used for both linear and nonlinear systems, with theoretical analysis and simulation studies provided to validate the results.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
Article
Engineering, Chemical
Seshu K. Damarla, Xi Sun, Fangwei Xu, Ashish Shah, Biao Huang
Summary: Control valve, affected by stiction, causes oscillations in closed-loop signals, leading to reduced product quality, plant throughput, and increased environmental impact. Therefore, it is crucial to detect and quantify stiction in control valves. In this study, four noninvasive and practical methods are developed using statistical tests such as F-test, t-test, modified Hotelling T2-test, and reverse arrangement test. These methods are applied to benchmark control loops from various industries and compared with existing methods. The results show that the proposed methods perform equally well or better than existing methods, with the t-test-based method and the modified Hotelling T2-test-based method being particularly effective. The proposed methods not only detect stiction but also quantify its severity, providing timely notifications to operators and assisting maintenance engineers in scheduling plant shutdowns. These methods are applicable to all control loops except for level loops.
INDUSTRIAL & ENGINEERING CHEMISTRY RESEARCH
(2023)
Article
Energy & Fuels
Stefan Jespersen, Zhenyu Yang, Dennis Severin Hansen, Mahsa Kashani, Biao Huang
Summary: To reduce the environmental impact of offshore oil and gas, stricter regulations on hydrocarbon discharge are being implemented. One approach to reducing oil discharge is by improving control systems through the introduction of new oil-in-water sensing technologies and advanced control methods. However, obtaining valid control-oriented models for de-oiling hydrocyclones has proven to be challenging, as existing models are often based on droplet trajectory analysis and do not account for the dynamics or require the measurement of droplet size distribution.
Article
Automation & Control Systems
Hongtian Chen, Biao Huang
Summary: This study develops three novel data-driven approaches for the development of fault-tolerant soft sensors in automation systems. The approaches, namely MSaS, SSaS, and IMSaS, aim to address the issue of unpredictable faults and their impact on soft sensor performance. MSaS constructs an optimal estimator of faults, SSaS removes influences from unknown sensor faults using a constructed subspace, and IMSaS is an improved version of MSaS that eliminates the effects of past prediction errors. These fault-tolerant soft sensors rely solely on system measurements and are evaluated through performance analysis and case studies.
IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY
(2023)
Article
Automation & Control Systems
Vamsi Krishna Puli, Biao Huang
Summary: Extraction of underlying patterns from measured variables is important for data-driven control applications. The proposed model can separate oscillating patterns and nonstationary variations. The methodology is applied to solve a fouling monitoring problem for an industrial oil production process.
IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY
(2023)
Article
Automation & Control Systems
Yousef Salehi, Kaiyu Zhou, Biao Huang, Xuehua Zhang
Summary: The study proposes a computer vision model to estimate flotation froth concentration, addressing practical challenges such as contaminated images, camera noise, and outliers. By restoring contaminated images, extracting froth image features, and building a regression model, the algorithm can accurately estimate froth concentration. This algorithm provides a time-saving alternative to laboratory analysis, making it essential for advanced process control applications.
JOURNAL OF PROCESS CONTROL
(2023)
Article
Automation & Control Systems
Qingyang Dai, Chunhui Zhao, Biao Huang
Summary: Due to frequent changes in operating conditions, industrial processes often exhibit time-varying behaviors, resulting in shifting data distributions. Conventional adaptive methods struggle to distinguish normal shifts from real faults when the distribution shifts widely. This study proposes an incremental variational Bayesian Gaussian mixture model (IncVBGMM) for adaptive monitoring to accommodate the changing data distribution caused by different degrees of time-varying behaviors. The proposed method effectively differentiates various types of faults from normal shifts and adapts to the time-varying dynamics.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Na Lin, Huaying Li, Ronghu Chi, Zhongsheng Hou, Biao Huang
Summary: In this work, a data-driven virtual reference setting learning (DDVRSL) method is proposed to enhance the PD feedback controller of the repetitive nonlinear system. The method utilizes an ideal nonlinear learning law and an iterative adaptation law to estimate parameters and improve robustness against uncertainties. The proposed method does not require exact mechanistic model knowledge and its convergence is proven through mathematical analysis and simulations.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Automation & Control Systems
Ronghu Chi, Xiaolin Guo, Na Lin, Biao Huang
Summary: This article addresses the challenges of data-driven control design in the presence of strong uncertainties, hard nonlinearities, and model dependency. It proposes a dynamic linearization (DL) method and an extended state observer (ESO) to handle an unknown nonlinear nonaffine system. The article presents a modified linear data model (mLDM) that accurately captures the input-output dynamics, including both linear parameter increments and unmodeled uncertainties and disturbances. The theoretical results are mathematically proven and validated through simulations.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Hongtian Chen, Hao Luo, Biao Huang, Bin Jiang, Okyay Kaynak
Summary: transfer learning has attracted attention as a new learning paradigm, and is used to develop fault diagnosis approaches for improving the safety and reliability of automation systems. This survey article provides a comprehensive review of transfer learning-motivated fault diagnosis methods and highlights open problems and potential research directions in this field. It also presents principles and a classification strategy for utilizing previous knowledge specifically for fault diagnosis tasks, aiming to contribute timely to transfer learning-motivated techniques in this area.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)