Wenhui Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Wenhui Li has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Dazhi Wang
Northeastern University
3 shared publications - An-An Liu
School of Electrical and Information Engineering, Tianjin University, Tianjin, China
2 shared publications - Likui Qiao
North University of China
2 shared publications - Ruidong Chen
School of Electrical and Information Engineering, Tianjin University, Tianjin, China
2 shared publications Sihan Wang
College of Information Science and Engineering, Northeastern University, Shenyang, China
2 shared publications- Ce Liang
Northeastern University, College of Information Science and Engineering,Shenyang,China
1 shared publication
Latest Publications
- Open Access
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1342-1356
What Concepts Lie Within? Detecting and Suppressing Risky Content in Diffusion Transformers
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1342-1356
Open AccessWhat Concepts Lie Within? Detecting and Suppressing Risky Content in Diffusion Transformers
- 30 September 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (39) , 396107
A Meteorological-state dual-branch power forecasting method for wind turbines oriented toward operating state perception
Wind turbine power output in real wind farms is jointly influenced by meteorological conditions, operating states, and control processes, making it difficult for conventional unified modeling methods to reliably characterize complex power evolution patterns. To address this issue, this paper proposes a meteorological-state dual-branch wind power forecasting method oriented toward operating state perception. Prediction samples are first constructed from supervisory control and data acquisition (SCADA) data, and the input variables are divided into meteorological features and state features according to their physical attributes, so as to characterize external wind resource variations and internal turbine response processes, respectively. A meteorological branch and a state branch are then constructed to perform differentiated temporal representation learning for the two types of features. On this basis, a gated fusion mechanism is introduced to adaptively coordinate the contributions of meteorological information and operating-state information to power forecasting. To improve the characterization of rapid power variations, a trend-constrained term based on first-order differences is incorporated into the mean squared error loss, enabling the model to account for both prediction amplitude errors and power variation trends. Experiments are conducted under short-span and long-span data scenarios using measured SCADA data from two direct-drive wind turbines in the same wind farm, and the proposed method is compared with representative deep learning and Transformer-based forecasting models. The experimental results show that the proposed method achieves consistently favorable forecasting performance across the two evaluated turbines and different temporal data spans. In the long-span experiments for Turbines 2 and 10, the R 2 values reach 0.9437 and 0.9341, respectively, demonstrating favorable error control capability, trend fitting ability, and performance consistency across the two turbines.
- 30 September 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (39) , 396107
A Meteorological-state dual-branch power forecasting method for wind turbines oriented toward operating state perception
Wind turbine power output in real wind farms is jointly influenced by meteorological conditions, operating states, and control processes, making it difficult for conventional unified modeling methods to reliably characterize complex power evolution patterns. To address this issue, this paper proposes a meteorological-state dual-branch wind power forecasting method oriented toward operating state perception. Prediction samples are first constructed from supervisory control and data acquisition (SCADA) data, and the input variables are divided into meteorological features and state features according to their physical attributes, so as to characterize external wind resource variations and internal turbine response processes, respectively. A meteorological branch and a state branch are then constructed to perform differentiated temporal representation learning for the two types of features. On this basis, a gated fusion mechanism is introduced to adaptively coordinate the contributions of meteorological information and operating-state information to power forecasting. To improve the characterization of rapid power variations, a trend-constrained term based on first-order differences is incorporated into the mean squared error loss, enabling the model to account for both prediction amplitude errors and power variation trends. Experiments are conducted under short-span and long-span data scenarios using measured SCADA data from two direct-drive wind turbines in the same wind farm, and the proposed method is compared with representative deep learning and Transformer-based forecasting models. The experimental results show that the proposed method achieves consistently favorable forecasting performance across the two evaluated turbines and different temporal data spans. In the long-span experiments for Turbines 2 and 10, the R 2 values reach 0.9437 and 0.9341, respectively, demonstrating favorable error control capability, trend fitting ability, and performance consistency across the two turbines.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 5102-5112
Cognitive Distillation for Information Forensics: Towards Improved Hateful Meme Detection
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 5102-5112
Open AccessCognitive Distillation for Information Forensics: Towards Improved Hateful Meme Detection
- 26 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (26) , 266109
Sample-weighted prototypical network with EMD distance: a few-shot fault diagnosis approach for wind turbine generators
To address the challenges of limited fault data and the difficulty of few-shot learning in real-world wind turbine operations, this study proposes a convolutional standard Transformer encoding framework based on a prototypical network, termed convolution-enhanced Transformer encoder (CNTE)-PN. Adopting a meta-learning paradigm, the method constructs multiple N -way K -shot tasks from limited samples to enhance the model’s generalization capability. To efficiently extract discriminative features from multivariate time series signals of wind turbines, a feature extraction module named CNTE is designed. In this module, one-dimensional convolution is employed to capture local temporal patterns, while the Transformer encoder is used to model global dependencies, enabling a unified representation that integrates both local and global features. Additionally, a sample weighting strategy based on intra-class variation is introduced to dynamically generate class prototypes. The Earth mover’s distance, derived from optimal transport theory, is applied to quantify the distributional discrepancy between query samples and prototypes, thereby improving diagnostic accuracy under few-shot conditions. To verify the effectiveness of the proposed method, this study first constructs a multivariate dynamic response dataset under parameter perturbations of a 5 MW wind turbine based on the OpenFAST platform, and then conducts comparative experiments and ablation analyses. On this basis, real wind farm SCADA data are further introduced for experimental validation. The results show that CNTE-PN achieves superior performance in terms of diagnostic accuracy, robustness, and feature representation capability, demonstrating its application potential for multivariate operating condition identification under few-shot scenarios.
- 26 June 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (26) , 266109
Sample-weighted prototypical network with EMD distance: a few-shot fault diagnosis approach for wind turbine generators
To address the challenges of limited fault data and the difficulty of few-shot learning in real-world wind turbine operations, this study proposes a convolutional standard Transformer encoding framework based on a prototypical network, termed convolution-enhanced Transformer encoder (CNTE)-PN. Adopting a meta-learning paradigm, the method constructs multiple N -way K -shot tasks from limited samples to enhance the model’s generalization capability. To efficiently extract discriminative features from multivariate time series signals of wind turbines, a feature extraction module named CNTE is designed. In this module, one-dimensional convolution is employed to capture local temporal patterns, while the Transformer encoder is used to model global dependencies, enabling a unified representation that integrates both local and global features. Additionally, a sample weighting strategy based on intra-class variation is introduced to dynamically generate class prototypes. The Earth mover’s distance, derived from optimal transport theory, is applied to quantify the distributional discrepancy between query samples and prototypes, thereby improving diagnostic accuracy under few-shot conditions. To verify the effectiveness of the proposed method, this study first constructs a multivariate dynamic response dataset under parameter perturbations of a 5 MW wind turbine based on the OpenFAST platform, and then conducts comparative experiments and ablation analyses. On this basis, real wind farm SCADA data are further introduced for experimental validation. The results show that CNTE-PN achieves superior performance in terms of diagnostic accuracy, robustness, and feature representation capability, demonstrating its application potential for multivariate operating condition identification under few-shot scenarios.
- 16 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Magnetics
- Vol. 61 (3) , 1-10
Modeling of Permanent Magnet Eddy-Current Coupler Based on Unsupervised Physics-Informed Radial-Based Function Neural Networks
Physics-informed neural networks (PINNs) have significant potential for modeling and parameter design in engineering field. While most existing PINNs research focuses on fluid mechanics and thermodynamics, few studies explore its application in electromagnetic field modeling of electromagnetic devices. Modeling the permanent magnet eddy current coupler (PMECC) to predict its performance characteristics based on geometric parameters and material properties is crucial for its design and optimization. An unsupervised modeling method for PMECC based on physics-informed radial basis neural networks (PIRBFNNs) is presented in this work. The modeling and solving of static magnetic field for devices excited by permanent magnets (PMs) is realized, which solves the problem of the traditional PINN fully-connected structure with many parameters and difficult training. We use the magnetic vector potential as the solution objective without providing the magnetic field boundary parameters and without labeling data, which is an unsupervised learning paradigm. The magnetic field distribution and performance of the PMECC can be computed using only the structural parameters. The experimental results show that the proposed PIRBFNN method is basically consistent with the results of the finite element numerical method and the analytical method. Additionally, a transfer learning experimental study was conducted to validate the effectiveness of the network components and training methods proposed in this paper. Furthermore, the proposed method can be applied to the modeling and analysis of various devices utilizing PM excitations.
- 16 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Magnetics
- Vol. 61 (3) , 1-10
Modeling of Permanent Magnet Eddy-Current Coupler Based on Unsupervised Physics-Informed Radial-Based Function Neural Networks
Physics-informed neural networks (PINNs) have significant potential for modeling and parameter design in engineering field. While most existing PINNs research focuses on fluid mechanics and thermodynamics, few studies explore its application in electromagnetic field modeling of electromagnetic devices. Modeling the permanent magnet eddy current coupler (PMECC) to predict its performance characteristics based on geometric parameters and material properties is crucial for its design and optimization. An unsupervised modeling method for PMECC based on physics-informed radial basis neural networks (PIRBFNNs) is presented in this work. The modeling and solving of static magnetic field for devices excited by permanent magnets (PMs) is realized, which solves the problem of the traditional PINN fully-connected structure with many parameters and difficult training. We use the magnetic vector potential as the solution objective without providing the magnetic field boundary parameters and without labeling data, which is an unsupervised learning paradigm. The magnetic field distribution and performance of the PMECC can be computed using only the structural parameters. The experimental results show that the proposed PIRBFNN method is basically consistent with the results of the finite element numerical method and the analytical method. Additionally, a transfer learning experimental study was conducted to validate the effectiveness of the network components and training methods proposed in this paper. Furthermore, the proposed method can be applied to the modeling and analysis of various devices utilizing PM excitations.
- Open Access
- 27 November 2023
- Published by MDPI AG in Mathematics
- Vol. 11 (23) , 4793
A Parameterized Modeling Method for Magnetic Circuits of Adjustable Permanent Magnet Couplers
The contactless transmission between the conductor rotor and the permanent magnet (PM) rotor of an adjustable permanent magnet coupler (APMC) provides the device with significant tolerance for alignment errors, making the performance estimation complicated and inaccurate. The first proposal of an edge coefficient in this paper helps to describe the edge effect with better accuracy. Accurate equivalent magnetic circuit (EMC) models of the APMC are established for each region. Models of magnetic flux, magnetic resistance, and eddy current density are established by defining the equivalent dimensional parameters of the eddy current circuit. Furthermore, the concept of magnetic inductance is proposed for the first time, parameterizing eddy currents that are difficult to describe with physical models and achieving the modeling of the dynamic eddy current circuit. The magnetic resistance is subdivided into two parts corresponding to the output and slip according to the power relationship. Furthermore, eddy current loss and dynamic torque models are further derived. The method proposed in this paper enables the APMC to be modeled and calculated in a completely new way. The correctness and accuracy of the model have been fully demonstrated using finite element simulation and an experimental prototype. In addition, the limitations of the proposed method and the reasons are fully discussed and investigated.
- 27 November 2023
- Published by MDPI AG in Mathematics
- Vol. 11 (23) , 4793
Open AccessA Parameterized Modeling Method for Magnetic Circuits of Adjustable Permanent Magnet Couplers
The contactless transmission between the conductor rotor and the permanent magnet (PM) rotor of an adjustable permanent magnet coupler (APMC) provides the device with significant tolerance for alignment errors, making the performance estimation complicated and inaccurate. The first proposal of an edge coefficient in this paper helps to describe the edge effect with better accuracy. Accurate equivalent magnetic circuit (EMC) models of the APMC are established for each region. Models of magnetic flux, magnetic resistance, and eddy current density are established by defining the equivalent dimensional parameters of the eddy current circuit. Furthermore, the concept of magnetic inductance is proposed for the first time, parameterizing eddy currents that are difficult to describe with physical models and achieving the modeling of the dynamic eddy current circuit. The magnetic resistance is subdivided into two parts corresponding to the output and slip according to the power relationship. Furthermore, eddy current loss and dynamic torque models are further derived. The method proposed in this paper enables the APMC to be modeled and calculated in a completely new way. The correctness and accuracy of the model have been fully demonstrated using finite element simulation and an experimental prototype. In addition, the limitations of the proposed method and the reasons are fully discussed and investigated.
- 13 January 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Magnetics Letters
- Vol. 13 (1949307X) , 1-5
Multifrequency AC Magnetic Flux Leakage Testing for the Detection of Surface and Backside Defects in Thick Steel Plates
The detection of surface and backside defects in steel plates is essential for safety maintenance. Current ac magnetic flux leakage (ACMFL) testing systems generally operate at a high frequency but cannot detect backside defects owing to the skin effect. Low-frequency ACMFL is limited by low sensitivity and low efficiency in detecting deep defects. This letter presents a multifrequency ACMFL testing method combined with the merits of low- and high-frequency excitation to detect surface and backside defects. In early work, the optimal frequency for detecting the surface defect with a depth of 4 mm or less was 400 Hz. Through simulation analysis, the optimal low frequency for detecting backside defects is determined. Additionally, the influence of the excitation intensity on the response signal is investigated. Relatively small excitation power should be used to obtain a reliable detection waveform. Finally, the experimental results of surface and backside defects detection validated the practicality of the method.
- 13 January 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Magnetics Letters
- Vol. 13 (1949307X) , 1-5
Multifrequency AC Magnetic Flux Leakage Testing for the Detection of Surface and Backside Defects in Thick Steel Plates
The detection of surface and backside defects in steel plates is essential for safety maintenance. Current ac magnetic flux leakage (ACMFL) testing systems generally operate at a high frequency but cannot detect backside defects owing to the skin effect. Low-frequency ACMFL is limited by low sensitivity and low efficiency in detecting deep defects. This letter presents a multifrequency ACMFL testing method combined with the merits of low- and high-frequency excitation to detect surface and backside defects. In early work, the optimal frequency for detecting the surface defect with a depth of 4 mm or less was 400 Hz. Through simulation analysis, the optimal low frequency for detecting backside defects is determined. Additionally, the influence of the excitation intensity on the response signal is investigated. Relatively small excitation power should be used to obtain a reliable detection waveform. Finally, the experimental results of surface and backside defects detection validated the practicality of the method.
- Top Cited PapersOpen Access
- 3 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 23717-23725
Fault Diagnosis of Bearings Based on Multi-Sensor Information Fusion and 2D Convolutional Neural Network
Intelligent operation and maintenance is an important part of Industry 4.0. In order to realize the intelligent of plant equipment, it will make full use of artificial intelligence methods to evaluate the operating status of the equipment. Fault diagnosis of industrial equipment represented by bearings is critical in smart manufacturing. Early, online and accurate diagnostics can save the plant a lot of time and expense. With the development of sensor technology and deep learning technology, multi-sensor information fusion and convolutional neural network (CNN) provide a solution to the above problems. In this paper, based on the characteristics of mechanical vibration signal propagation in space, a new multi-sensor information fusion method is proposed to implement fault classification. This method constructs the time-domain vibration signals of multiple sensors from different position into a rectangular two-dimensional matrix, and then uses an improved 2D CNN to realize signal classification. The method is validated on the open dataset Case Western Reserve University, the University of Cincinnati IMS bearing database and the dataset form designed bearing fault test rig, has achieved prediction of 99.92%, 99.68%, and 99.25% respectively. Compared with the traditional 1D, 2D CNN and other fault classification methods, the model can utilize less data and computational complexity, achieve higher fault prediction accuracy.
- 3 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 23717-23725
Top Cited PapersOpen AccessFault Diagnosis of Bearings Based on Multi-Sensor Information Fusion and 2D Convolutional Neural Network
Intelligent operation and maintenance is an important part of Industry 4.0. In order to realize the intelligent of plant equipment, it will make full use of artificial intelligence methods to evaluate the operating status of the equipment. Fault diagnosis of industrial equipment represented by bearings is critical in smart manufacturing. Early, online and accurate diagnostics can save the plant a lot of time and expense. With the development of sensor technology and deep learning technology, multi-sensor information fusion and convolutional neural network (CNN) provide a solution to the above problems. In this paper, based on the characteristics of mechanical vibration signal propagation in space, a new multi-sensor information fusion method is proposed to implement fault classification. This method constructs the time-domain vibration signals of multiple sensors from different position into a rectangular two-dimensional matrix, and then uses an improved 2D CNN to realize signal classification. The method is validated on the open dataset Case Western Reserve University, the University of Cincinnati IMS bearing database and the dataset form designed bearing fault test rig, has achieved prediction of 99.92%, 99.68%, and 99.25% respectively. Compared with the traditional 1D, 2D CNN and other fault classification methods, the model can utilize less data and computational complexity, achieve higher fault prediction accuracy.
- Open Access
- 24 April 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 78367-78377
Modeling and Performance Analysis of an Axial-Radial Combined Permanent Magnet Eddy Current Coupler
The structure and magnetic circuits of axial-radial combined permanent magnet eddy current couplers are more complicated than those of axial or radial permanent magnet eddy current couplers, the degree of coupling is deeper, which will affect the device performance. In this paper, a novel analytical model for an axial-radial combined permanent magnet eddy current coupler is proposed based on the equivalent magnetic circuit method. The distribution of the magnetic field, the relationship between the slip and torque, the influence of the structural parameters on the transmission performance, the 3-D correction and thermal compensation are studied. Combined with the concept of the micro-element method, the air gap magnetic density, eddy current density and torque are accurately calculated, which solves the problem that the existing model cannot analyse the axial-radial combined structure. To show the effectiveness of the proposed analytical model, the transmission performance of an axial-radial combined permanent magnet eddy current coupler prototype is investigated, and the analytical calculation results are compared with the results of the finite element method simulations and experimental tests.
- 24 April 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 78367-78377
Open AccessModeling and Performance Analysis of an Axial-Radial Combined Permanent Magnet Eddy Current Coupler
The structure and magnetic circuits of axial-radial combined permanent magnet eddy current couplers are more complicated than those of axial or radial permanent magnet eddy current couplers, the degree of coupling is deeper, which will affect the device performance. In this paper, a novel analytical model for an axial-radial combined permanent magnet eddy current coupler is proposed based on the equivalent magnetic circuit method. The distribution of the magnetic field, the relationship between the slip and torque, the influence of the structural parameters on the transmission performance, the 3-D correction and thermal compensation are studied. Combined with the concept of the micro-element method, the air gap magnetic density, eddy current density and torque are accurately calculated, which solves the problem that the existing model cannot analyse the axial-radial combined structure. To show the effectiveness of the proposed analytical model, the transmission performance of an axial-radial combined permanent magnet eddy current coupler prototype is investigated, and the analytical calculation results are compared with the results of the finite element method simulations and experimental tests.
- Open Access
A Composite Vectors Modulation Strategy for PMSM DTC Systems
The operation performance of permanent magnet synchronous motors (PMSMs) driven by direct torque control (DTC) are affected by torque error compensation and flux error compensation in each control period. The error compensational effects provided by different vectors under different control strategies are analyzed in this paper. The precondition of accurate error compensation is applied to the proposed novel composite vectors modulation strategy (CVM) for a PMSM DTC system. In CVM-DTC, the operating conditions of PMSM are divided into three cases according to the relationships between the errors and the actual error compensations, including steady-state case, dynamic-state case, and transient-state case. In order to establish the novel CVM-DTC strategy smoothly, the effect factors are introduced and used to represent the error compensational effects, which are obtained through the proposed effect factors’ controller. The analysis of error compensational effects provided by single active vector and synthetic voltage vector are described in detail while the PMSM is operated in different operation conditions. Finally, the effectiveness of the novel CVM-DTC strategy is verified through the experimental results in a 100-W PMSM drive system.
Open AccessA Composite Vectors Modulation Strategy for PMSM DTC Systems
The operation performance of permanent magnet synchronous motors (PMSMs) driven by direct torque control (DTC) are affected by torque error compensation and flux error compensation in each control period. The error compensational effects provided by different vectors under different control strategies are analyzed in this paper. The precondition of accurate error compensation is applied to the proposed novel composite vectors modulation strategy (CVM) for a PMSM DTC system. In CVM-DTC, the operating conditions of PMSM are divided into three cases according to the relationships between the errors and the actual error compensations, including steady-state case, dynamic-state case, and transient-state case. In order to establish the novel CVM-DTC strategy smoothly, the effect factors are introduced and used to represent the error compensational effects, which are obtained through the proposed effect factors’ controller. The analysis of error compensational effects provided by single active vector and synthetic voltage vector are described in detail while the PMSM is operated in different operation conditions. Finally, the effectiveness of the novel CVM-DTC strategy is verified through the experimental results in a 100-W PMSM drive system.