Wei Fan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Wei Fan has a total of 131 co-authors. Here are the top co-authors by number of shared publications:
- Chao Chen
School of Mechanical Engineering, Jiangsu University, Zhenjiang, China
14 shared publications - Zhike Peng
Shanghai Jiao Tong University
10 shared publications - Yuyong Xiong
Shanghai Jiao Tong University
8 shared publications - Hui Wang
Shanghai Aerospace Electronic Technology Institute, Shanghai, China
6 shared publications - Saisai Chen
School of Mechanical Engineering, Jiangsu University, Zhenjiang, China
5 shared publications - Changbo He
Anhui University
4 shared publications
Latest Publications
- Open Access
- 29 September 2026
- Published by Elsevier in SSRN Electronic Journal
Millimeter-Wave Impact Displacement Sensor: Doppler-based Phase Fusion Method for Ultra-high-speed Large Displacement Measurement
Accurate measurement of large displacements at ultra-high-speed is essential in explosion-induced impact scenarios. The emerging interferometric phase tracking-based millimeter-wave vibrometry can accurately quantify tiny displacement, however, ultra-high-speed large displacement can produce rapid phase variations spanning numerous phase periods, posing challenges for reliable phase tracking in this scenario. Although Doppler frequency reflects the rate of phase change and can therefore provide a prior for phase tracking, displacement reconstruction based solely on Doppler integration is susceptible to cumulative errors. To this end, a Doppler-based phase fusion method(DPF) for ultra-high-speed large displacement measurement is proposed, creating a unique millimeter-wave impact displacement sensor. Firstly, DPF constructs a phase-increment prior from the Doppler-frequency ridge of corrected I/Q signals, which provides predictive information to guide the phase evolution. To reduce the influence of unreliable local phase estimates and limit the accumulation error, a phase reconstruction method is proposed to adaptively fuse phase observations with the Doppler prediction, allowing Doppler information to support phase continuity while phase observations constrain cumulative integration errors. Simulation results demonstrate consistent reconstruction accuracy, with a maximum relative RMSE of 0.46%. Experimental results further confirmed its effectiveness and robustness, yielding an average absolute displacement error of 1.016 ± 0.757 mm with a total displacement of 795mm, offering a promising approach for ultra-high-speed impact displacement measurement.
- 29 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMillimeter-Wave Impact Displacement Sensor: Doppler-based Phase Fusion Method for Ultra-high-speed Large Displacement Measurement
Accurate measurement of large displacements at ultra-high-speed is essential in explosion-induced impact scenarios. The emerging interferometric phase tracking-based millimeter-wave vibrometry can accurately quantify tiny displacement, however, ultra-high-speed large displacement can produce rapid phase variations spanning numerous phase periods, posing challenges for reliable phase tracking in this scenario. Although Doppler frequency reflects the rate of phase change and can therefore provide a prior for phase tracking, displacement reconstruction based solely on Doppler integration is susceptible to cumulative errors. To this end, a Doppler-based phase fusion method(DPF) for ultra-high-speed large displacement measurement is proposed, creating a unique millimeter-wave impact displacement sensor. Firstly, DPF constructs a phase-increment prior from the Doppler-frequency ridge of corrected I/Q signals, which provides predictive information to guide the phase evolution. To reduce the influence of unreliable local phase estimates and limit the accumulation error, a phase reconstruction method is proposed to adaptively fuse phase observations with the Doppler prediction, allowing Doppler information to support phase continuity while phase observations constrain cumulative integration errors. Simulation results demonstrate consistent reconstruction accuracy, with a maximum relative RMSE of 0.46%. Experimental results further confirmed its effectiveness and robustness, yielding an average absolute displacement error of 1.016 ± 0.757 mm with a total displacement of 795mm, offering a promising approach for ultra-high-speed impact displacement measurement.
- 16 September 2026
- Published by SAGE Publications in Structural Health Monitoring
A cross-domain bearing fault diagnosis method under sample imbalance based on multi-view information cross-integration ensemble learning
Reliable cross-domain bearing fault diagnosis is crucial for rotating machinery health monitoring in complex industrial systems. However, practical applications often face severe sample imbalance and domain distribution shifts, the combination of which frequently causes the model to be biased toward majority classes. To overcome these challenges, we propose multi-view information cross-integration ensemble learning (MICE). Firstly, MICE employs a multi-view cross-integration strategy to generate mixed signals, thereby extracting complementary feature representations while reducing redundancy. Secondly, to handle sample imbalance, a consistency voting entropy optimization strategy analyzes prediction scores to set dynamic thresholds and generate high-confidence pseudo-labels. This process enables the model to effectively capture critical minority-class features. Finally, a dual-path collaborative learning mechanism integrates inter-view feature transfer with inter-domain alignment, and employs a confusion-class focusing strategy to explicitly distinguish confusable categories. Extensive experiments on two public datasets validate the superiority and strong cross-domain generalization capability of the proposed MICE method.
- 16 September 2026
- Published by SAGE Publications in Structural Health Monitoring
A cross-domain bearing fault diagnosis method under sample imbalance based on multi-view information cross-integration ensemble learning
Reliable cross-domain bearing fault diagnosis is crucial for rotating machinery health monitoring in complex industrial systems. However, practical applications often face severe sample imbalance and domain distribution shifts, the combination of which frequently causes the model to be biased toward majority classes. To overcome these challenges, we propose multi-view information cross-integration ensemble learning (MICE). Firstly, MICE employs a multi-view cross-integration strategy to generate mixed signals, thereby extracting complementary feature representations while reducing redundancy. Secondly, to handle sample imbalance, a consistency voting entropy optimization strategy analyzes prediction scores to set dynamic thresholds and generate high-confidence pseudo-labels. This process enables the model to effectively capture critical minority-class features. Finally, a dual-path collaborative learning mechanism integrates inter-view feature transfer with inter-domain alignment, and employs a confusion-class focusing strategy to explicitly distinguish confusable categories. Extensive experiments on two public datasets validate the superiority and strong cross-domain generalization capability of the proposed MICE method.
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 3515311
Integrated Bearing Condition Monitoring and Fault Diagnosis via Distribution-Adaptive Composite Correlation Under Mixed Gaussian–Laplace Noise
Bearing condition monitoring is essential in ensuring the reliable operation of machinery. Yet most vibration-analysis techniques implicitly assume Gaussian noise—a premise often violated in practice, where signals contain Laplace components or Gaussian–Laplace mixtures. To overcome this limitation, we propose a Distribution-Adaptive Composite Correlation (DACC) framework that simultaneously performs condition monitoring and early fault diagnosis under mixed Gaussian–Laplace noise. First, the classical noise-resistant correlation (NRC) is generalized to remain robust against heavy-tailed Laplace disturbances and broadband Gaussian interference. This generalized NRC is fused with the frequency-domain energy ratio to form a composite correlation that accentuates fault-related impulses while suppressing stochastic fluctuations. Integrating the composite correlation over a sliding window yields a scalar health index that evolves monotonically with bearing degradation. An exponentially weighted moving-average (EWMA) chart is then applied to track this index to indicate incipient anomalies. Experiments on simulated signals with pure Laplace, and Gaussian-Laplace mixed noise, as well as on public bearing datasets, show that DACC detects degradation earlier and with lower false-alarm rates than state-of-the-art methods, demonstrating its effectiveness for integrated bearing-condition monitoring and fault diagnosis in practical applications.
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 3515311
Integrated Bearing Condition Monitoring and Fault Diagnosis via Distribution-Adaptive Composite Correlation Under Mixed Gaussian–Laplace Noise
Bearing condition monitoring is essential in ensuring the reliable operation of machinery. Yet most vibration-analysis techniques implicitly assume Gaussian noise—a premise often violated in practice, where signals contain Laplace components or Gaussian–Laplace mixtures. To overcome this limitation, we propose a Distribution-Adaptive Composite Correlation (DACC) framework that simultaneously performs condition monitoring and early fault diagnosis under mixed Gaussian–Laplace noise. First, the classical noise-resistant correlation (NRC) is generalized to remain robust against heavy-tailed Laplace disturbances and broadband Gaussian interference. This generalized NRC is fused with the frequency-domain energy ratio to form a composite correlation that accentuates fault-related impulses while suppressing stochastic fluctuations. Integrating the composite correlation over a sliding window yields a scalar health index that evolves monotonically with bearing degradation. An exponentially weighted moving-average (EWMA) chart is then applied to track this index to indicate incipient anomalies. Experiments on simulated signals with pure Laplace, and Gaussian-Laplace mixed noise, as well as on public bearing datasets, show that DACC detects degradation earlier and with lower false-alarm rates than state-of-the-art methods, demonstrating its effectiveness for integrated bearing-condition monitoring and fault diagnosis in practical applications.
- Open Access
- 23 April 2026
- Published by MDPI AG in Agriculture
- Vol. 16 (9) , 935
The Application of AI Technology Across the Entire Technical Chain of Combine Harvesters: A Systematic Review
As complex agricultural machinery, traditional combine harvesters face numerous challenges during operation due to their reliance on manual observation. To meet the demands of modern agriculture, intelligent combine harvesters have emerged. Intelligent sensing uses multi-sensor fusion and deep learning to monitor crop lodging, feed rate, loss rate, and impurity content. Under suboptimal conditions, multi-source fusion strategies improve perception reliability. Information processing and decision-making enable dynamic optimization of operational parameters and reduce harvest losses. Multi-machine coordination transforms single-machine operations into fleet control, while remote monitoring leverages a cloud edge collaboration architecture to enable status visualization, remote control, and predictive maintenance for faults. Unmanned operations utilize high-precision positioning and intelligent path planning to improve fleet efficiency and field coverage. However, the field still faces common challenges, including insufficient real-time processing capabilities for multi-source heterogeneous data, poor adaptability to complex agronomic scenarios, and limited economic feasibility. In this review, we examine the complete technology chain, which includes intelligent perception, intelligent decision-making and coordination, remote monitoring, and unmanned operations. We conduct a comparative analysis of the current state of these systems and the challenges they face, providing a systematic reference for future research and industrial applications.
- 23 April 2026
- Published by MDPI AG in Agriculture
- Vol. 16 (9) , 935
Open AccessThe Application of AI Technology Across the Entire Technical Chain of Combine Harvesters: A Systematic Review
As complex agricultural machinery, traditional combine harvesters face numerous challenges during operation due to their reliance on manual observation. To meet the demands of modern agriculture, intelligent combine harvesters have emerged. Intelligent sensing uses multi-sensor fusion and deep learning to monitor crop lodging, feed rate, loss rate, and impurity content. Under suboptimal conditions, multi-source fusion strategies improve perception reliability. Information processing and decision-making enable dynamic optimization of operational parameters and reduce harvest losses. Multi-machine coordination transforms single-machine operations into fleet control, while remote monitoring leverages a cloud edge collaboration architecture to enable status visualization, remote control, and predictive maintenance for faults. Unmanned operations utilize high-precision positioning and intelligent path planning to improve fleet efficiency and field coverage. However, the field still faces common challenges, including insufficient real-time processing capabilities for multi-source heterogeneous data, poor adaptability to complex agronomic scenarios, and limited economic feasibility. In this review, we examine the complete technology chain, which includes intelligent perception, intelligent decision-making and coordination, remote monitoring, and unmanned operations. We conduct a comparative analysis of the current state of these systems and the challenges they face, providing a systematic reference for future research and industrial applications.
- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-11
Curvature-Driven Dynamic Blade Tip Clearance Measurement Using Microwave Sensor
As a core parameter for performance evaluation of rotating blades and turbines, high-precision noncontact measurement of blade tip clearance (BTC) faces severe challenges. Microwave sensing is promising for BTC measurement. However, static BTC measurements suffer from DC offsets, nonlinear amplitude modulation, and harmonic distortion, and dynamic clearance extraction is degraded by critical phase mutations. To address these issues, this article proposes a curvature-feature-driven correction algorithm based on analytic geometry, namely curvature-adaptive dynamic correction (CADC). CADC provides a parameter-free correction framework that combines a curvature-driven bidirectional compensation mechanism with a curvature-weighted covariance correction matrix, thereby mitigating the above interferences and phase mutations. Simulations verify robust calibration under nonlinear amplitude modulation and high-order harmonic interference, outperforming several state-of-the-art microwave BTC methods. Experiments show that a microwave sensor system with CADC achieves an average static displacement error of$4.150~\mu $ m over a 0–10 mm clearance range. For dynamic BTC measurement, the average absolute error per blade is$2.12~\mu $ m with an average repeatability error of$1.23~\mu $ m, and stable performance is maintained up to 3000 r/min.- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-11
Curvature-Driven Dynamic Blade Tip Clearance Measurement Using Microwave Sensor
As a core parameter for performance evaluation of rotating blades and turbines, high-precision noncontact measurement of blade tip clearance (BTC) faces severe challenges. Microwave sensing is promising for BTC measurement. However, static BTC measurements suffer from DC offsets, nonlinear amplitude modulation, and harmonic distortion, and dynamic clearance extraction is degraded by critical phase mutations. To address these issues, this article proposes a curvature-feature-driven correction algorithm based on analytic geometry, namely curvature-adaptive dynamic correction (CADC). CADC provides a parameter-free correction framework that combines a curvature-driven bidirectional compensation mechanism with a curvature-weighted covariance correction matrix, thereby mitigating the above interferences and phase mutations. Simulations verify robust calibration under nonlinear amplitude modulation and high-order harmonic interference, outperforming several state-of-the-art microwave BTC methods. Experiments show that a microwave sensor system with CADC achieves an average static displacement error of$4.150~\mu $ m over a 0–10 mm clearance range. For dynamic BTC measurement, the average absolute error per blade is$2.12~\mu $ m with an average repeatability error of$1.23~\mu $ m, and stable performance is maintained up to 3000 r/min.- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 74 (4) , 3698-3709
A Unified Ratio Differentiation Framework for Robust I / Q Imbalance Correction Under Unknown Amplitude Attenuation in CW Doppler Radar
In continuous wave (CW) Doppler radar measurement systems, correcting amplitude and phase imbalances is essential for accurate signal demodulation. In close-range scenarios, however, static in-phase/quadrature ( $I$ / $Q$ ) signals often suffer from unknown amplitude attenuation modulation, which degrades the effectiveness of conventional parameter estimation methods. To address this challenge, we present a unified ratio differentiation-based imbalance correction framework (RD-ICF) for robust $I$ / $Q$ imbalance correction. The framework exploits the ratio differentiation (RD) process to equalize attenuation across the $I$ / $Q$ signals, ensuring that imbalance parameters can be reliably estimated regardless of unknown attenuation. Within this framework, two complementary implementations are developed: the RD-equation (RD-EQ) method and the RD-elliptic correction (RD-EC) method. RD-EQ provides a direct equation-solving approach, while RD-EC incorporates elliptic correction (EC) and decision coefficients to achieve higher estimation accuracy and reduced sensitivity to sampling frequency. Simulation and experimental results demonstrate that the unified framework effectively overcomes the limitations of existing orthogonal imbalance correction techniques in the presence of unknown amplitude attenuation. In close-range experiments, it achieves a demodulation RMSE below $1.59~\mu $ m, an average nonlinearity error under 0.05% across a 9.36mm measurement range, and reduces RMSE to one-third compared to state-of-the-art methods.
- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 74 (4) , 3698-3709
A Unified Ratio Differentiation Framework for Robust I / Q Imbalance Correction Under Unknown Amplitude Attenuation in CW Doppler Radar
In continuous wave (CW) Doppler radar measurement systems, correcting amplitude and phase imbalances is essential for accurate signal demodulation. In close-range scenarios, however, static in-phase/quadrature ( $I$ / $Q$ ) signals often suffer from unknown amplitude attenuation modulation, which degrades the effectiveness of conventional parameter estimation methods. To address this challenge, we present a unified ratio differentiation-based imbalance correction framework (RD-ICF) for robust $I$ / $Q$ imbalance correction. The framework exploits the ratio differentiation (RD) process to equalize attenuation across the $I$ / $Q$ signals, ensuring that imbalance parameters can be reliably estimated regardless of unknown attenuation. Within this framework, two complementary implementations are developed: the RD-equation (RD-EQ) method and the RD-elliptic correction (RD-EC) method. RD-EQ provides a direct equation-solving approach, while RD-EC incorporates elliptic correction (EC) and decision coefficients to achieve higher estimation accuracy and reduced sensitivity to sampling frequency. Simulation and experimental results demonstrate that the unified framework effectively overcomes the limitations of existing orthogonal imbalance correction techniques in the presence of unknown amplitude attenuation. In close-range experiments, it achieves a demodulation RMSE below $1.59~\mu $ m, an average nonlinearity error under 0.05% across a 9.36mm measurement range, and reduces RMSE to one-third compared to state-of-the-art methods.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-13
A Novel Multiscale Residual Active Deep Q -Learning Network for Intelligent Bearing Fault Diagnosis With Small Training Data
In bearing fault diagnosis (BFD), deep reinforcement learning (DRL) offers a promising solution by applying rewards to optimize fault decision-making strategies. However, the lack of sufficient training data as well as the ability to explore more comprehensive fault representation of key samples will draw limitations on DRL-based diagnostic model. To tackle these aforesaid issues, this study proposes a multi-scale residual active deep Q-learning network (MRADQN), which can not only select high-value samples from unlabeled candidates for expanding the number of training samples, but also extract local and global features from these newly-reconstructed samples. Specifically, MRADQN consists of an active deep Q-learning (ADQN) module and a multi-scale residual convolutional neural network (MRCNN) module. To pick out the most beneficial samples, ADQN designs a new state and action strategy based on entropy and cosine distance to calculate the uncertainty and similarity of samples, respectively, with the aim of quickly and accurately selecting the most significant data for labeling. For improving the ability regarding feature learning, a multi-scale residual block is designed in MRCNN to obtain rich feature representations. The experimental analyses show that MRADQN has strong fault detection capabilities, exceeding 99% diagnostic accuracy. And compared with many methods, MRADQN attains superior results when facing limited training data.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-13
A Novel Multiscale Residual Active Deep Q -Learning Network for Intelligent Bearing Fault Diagnosis With Small Training Data
In bearing fault diagnosis (BFD), deep reinforcement learning (DRL) offers a promising solution by applying rewards to optimize fault decision-making strategies. However, the lack of sufficient training data as well as the ability to explore more comprehensive fault representation of key samples will draw limitations on DRL-based diagnostic model. To tackle these aforesaid issues, this study proposes a multi-scale residual active deep Q-learning network (MRADQN), which can not only select high-value samples from unlabeled candidates for expanding the number of training samples, but also extract local and global features from these newly-reconstructed samples. Specifically, MRADQN consists of an active deep Q-learning (ADQN) module and a multi-scale residual convolutional neural network (MRCNN) module. To pick out the most beneficial samples, ADQN designs a new state and action strategy based on entropy and cosine distance to calculate the uncertainty and similarity of samples, respectively, with the aim of quickly and accurately selecting the most significant data for labeling. For improving the ability regarding feature learning, a multi-scale residual block is designed in MRCNN to obtain rich feature representations. The experimental analyses show that MRADQN has strong fault detection capabilities, exceeding 99% diagnostic accuracy. And compared with many methods, MRADQN attains superior results when facing limited training data.