Yongsheng Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yongsheng Zhu has a total of 109 co-authors. Here are the top co-authors by number of shared publications:
- Ke Yan
Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System, Xi’an Jiaotong University, Xi’an 710049, China
21 shared publications - Zhijun Ren
Frontier Institute of Science and Technology, Xian Jiaotong University, Xian, China
19 shared publications - Jun Hong
Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System, Xi’an Jiaotong University, Xi’an 710049, China
17 shared publications Jun Hong
Xi’an Jiaotong University
16 shared publicationsKe Yan
Xi’an Jiaotong University
15 shared publications- Dawei Gao
School of Marine Science and Technology, Northwestern Polytechnical University, Xian, China
12 shared publications
Latest Publications
- 2 June 2026
- Published by Taylor & Francis in Nondestructive Testing and Evaluation
- p. 1-43
TF-GIDI: time-frequency granular divergence for interpretable unsupervised grease-lubrication degradation monitoring and relubrication verification in rolling bearings
1. Grease-lubricated rolling bearings are critical components in rotating machinery, and their reliability depends strongly on lubrication integrity [1,2]. Loss of lubrication integrity is a major ...
- 2 June 2026
- Published by Taylor & Francis in Nondestructive Testing and Evaluation
- p. 1-43
TF-GIDI: time-frequency granular divergence for interpretable unsupervised grease-lubrication degradation monitoring and relubrication verification in rolling bearings
1. Grease-lubricated rolling bearings are critical components in rotating machinery, and their reliability depends strongly on lubrication integrity [1,2]. Loss of lubrication integrity is a major ...
- 12 May 2026
- Published by Elsevier in Advanced Engineering Informatics
- Vol. 74
Reliability-aware dynamic graph fusion and LLM-based diagnostic assistant for bearing faults
- 12 May 2026
- Published by Elsevier in Advanced Engineering Informatics
- Vol. 74
Reliability-aware dynamic graph fusion and LLM-based diagnostic assistant for bearing faults
- 16 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
An Interpretable Rolling-Bearing Lubrication-State Diagnosis Method Based on Multiscale Residual and Semantic Consistency CycleGAN
Lubrication anomaly in rolling bearings often precede mechanical damage and accelerate failure progression. Because lubrication conditions are amenable to corrective actions (e.g., grease replenishment or replacement) before severe damage occurs, reliable measurement-based monitoring of lubrication states is particularly valuable. Unlike classical bearing damage characterized by stable fault characteristic frequencies, lubrication-related changes exhibit non-stationary, high-frequency broadband measurement signatures driven by evolving micro-contact and frictional interactions. Meanwhile, labeled lubrication anomaly data remain extremely scarce in industrial settings, which together limits conventional feature learning and supervised deployment. To address these challenges, this paper proposes a Multi-Scale Residual and Semantic-Consistency CycleGAN (MSRS-CycleGAN) and a lab-to-field diagnosis framework that uses labeled laboratory multi-state signals as the source domain and healthy-only field vibration as the target domain. MSRS-CycleGAN incorporates multi-scale residual convolutional blocks to enhance sensitivity to lubrication-related high-frequency patterns and introduces a semantic-consistency loss to preserve state-discriminative structures during style transfer, thereby generating realistic lubrication anomaly signals that align with real target-domain measurements. A hierarchical strategy then performs anomaly detection via an autoencoder trained solely on healthy target-domain data, followed by lubrication-state classification using a convolutional neural network. Experiments demonstrate improved agreement between generated and real target-domain signals in spectral characteristics and superior lubrication-state diagnostic accuracy under zero-fault-data target-domain conditions.
- 16 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
An Interpretable Rolling-Bearing Lubrication-State Diagnosis Method Based on Multiscale Residual and Semantic Consistency CycleGAN
Lubrication anomaly in rolling bearings often precede mechanical damage and accelerate failure progression. Because lubrication conditions are amenable to corrective actions (e.g., grease replenishment or replacement) before severe damage occurs, reliable measurement-based monitoring of lubrication states is particularly valuable. Unlike classical bearing damage characterized by stable fault characteristic frequencies, lubrication-related changes exhibit non-stationary, high-frequency broadband measurement signatures driven by evolving micro-contact and frictional interactions. Meanwhile, labeled lubrication anomaly data remain extremely scarce in industrial settings, which together limits conventional feature learning and supervised deployment. To address these challenges, this paper proposes a Multi-Scale Residual and Semantic-Consistency CycleGAN (MSRS-CycleGAN) and a lab-to-field diagnosis framework that uses labeled laboratory multi-state signals as the source domain and healthy-only field vibration as the target domain. MSRS-CycleGAN incorporates multi-scale residual convolutional blocks to enhance sensitivity to lubrication-related high-frequency patterns and introduces a semantic-consistency loss to preserve state-discriminative structures during style transfer, thereby generating realistic lubrication anomaly signals that align with real target-domain measurements. A hierarchical strategy then performs anomaly detection via an autoencoder trained solely on healthy target-domain data, followed by lubrication-state classification using a convolutional neural network. Experiments demonstrate improved agreement between generated and real target-domain signals in spectral characteristics and superior lubrication-state diagnostic accuracy under zero-fault-data target-domain conditions.
- 1 April 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (4)
Seepage–acoustic coupling numerical simulation on heterogeneous thick sandstone aquifer formation
Accurate characterization of aquifer heterogeneity is critical for preventing water inrush hazards, yet traditional methods often lack the required resolution. To address this problem, a coupled seepage–acoustic numerical simulation framework was developed to investigate the physical mechanisms of flow-induced noise within boreholes. The influences of aquifer lithology, seepage rate, seepage location, and borehole diameter on the flow and acoustic fields were systematically analyzed. The results demonstrate that hydrodynamic noise can be decomposed into dynamic pressure (pseudo-sound) and sound pressure (radiated noise), and these two components exhibit complementary spatial characteristics. Dynamic pressure is governed by local flow velocity and reaches its maximum at the seepage location, while it decays monotonically with increasing frequency. In contrast, sound pressure reflects global pressure fluctuations and shows its minimum at the seepage location, with distinct resonant peaks appearing in the spectrum (typically 0–400 Hz). A regime transition is observed in that dynamic pressure predominates under low seepage conditions (1.45 m3/h), although the fundamental frequency characteristics remain unchanged. Lithological variations and seepage position primarily affect the signal amplitude rather than the frequency consistent with Strouhal number scaling, where finer grains together with proximity to the seepage point generate higher amplitudes. Furthermore, borehole diameter exerts opposite effects on the two components. A smaller diameter increases the radiated sound pressure because of higher average velocity, while a larger diameter enhances the local dynamic pressure due to the stabilization of vortices.
- 1 April 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (4)
Seepage–acoustic coupling numerical simulation on heterogeneous thick sandstone aquifer formation
Accurate characterization of aquifer heterogeneity is critical for preventing water inrush hazards, yet traditional methods often lack the required resolution. To address this problem, a coupled seepage–acoustic numerical simulation framework was developed to investigate the physical mechanisms of flow-induced noise within boreholes. The influences of aquifer lithology, seepage rate, seepage location, and borehole diameter on the flow and acoustic fields were systematically analyzed. The results demonstrate that hydrodynamic noise can be decomposed into dynamic pressure (pseudo-sound) and sound pressure (radiated noise), and these two components exhibit complementary spatial characteristics. Dynamic pressure is governed by local flow velocity and reaches its maximum at the seepage location, while it decays monotonically with increasing frequency. In contrast, sound pressure reflects global pressure fluctuations and shows its minimum at the seepage location, with distinct resonant peaks appearing in the spectrum (typically 0–400 Hz). A regime transition is observed in that dynamic pressure predominates under low seepage conditions (1.45 m3/h), although the fundamental frequency characteristics remain unchanged. Lithological variations and seepage position primarily affect the signal amplitude rather than the frequency consistent with Strouhal number scaling, where finer grains together with proximity to the seepage point generate higher amplitudes. Furthermore, borehole diameter exerts opposite effects on the two components. A smaller diameter increases the radiated sound pressure because of higher average velocity, while a larger diameter enhances the local dynamic pressure due to the stabilization of vortices.
- 2 January 2026
- Published by Springer Nature
- p. 1070-1082
A Lightweight Fault Diagnosis Method of Motor Based on Efficient Additive Self-attention and Separable Dilated Convolution
- 2 January 2026
- Published by Springer Nature
- p. 1070-1082
A Lightweight Fault Diagnosis Method of Motor Based on Efficient Additive Self-attention and Separable Dilated Convolution
- 18 December 2025
- Published by Elsevier in Tribology International
- Vol. 216
Mechanism study on the influence of load and contaminant on the surface damage of GCr15 bearing raceway
- 18 December 2025
- Published by Elsevier in Tribology International
- Vol. 216
Mechanism study on the influence of load and contaminant on the surface damage of GCr15 bearing raceway
- 27 October 2025
- Published by Elsevier in Advanced Engineering Informatics
- Vol. 69
A MoE-LLM-based multisensor flexible fusion fault diagnosis method for rotating machinery
- 27 October 2025
- Published by Elsevier in Advanced Engineering Informatics
- Vol. 69
A MoE-LLM-based multisensor flexible fusion fault diagnosis method for rotating machinery
- Open Access
- 15 May 2025
- Published by JVE International Ltd. in Vibroengineering PROCEDIA
- Vol. 59, 82-87
Method of detecting bubbles in a pipeline by vibration excitation
The transportation of liquids through pipelines has extensive applications in the industrial sector. However, due to factors such as localized negative pressure or discontinuities in fluid flow, the formation of bubbles within the pipeline is a common occurrence. Developing an effective method for detecting these internal bubbles is of significant practical importance. Pipelines containing bubbles can be conceptualized as multiphase flow systems, where the frequency domain characteristics of the system vary with changes in bubble volume. Based on this principle, this paper proposes a method for identifying bubbles within pipelines by analyzing the frequency domain features of mechanical vibration waves. To validate the efficacy of this method, a testing apparatus was constructed to use pressure sensors to detect the pressure signals at the inlet and outlet of a pipeline. By analyzing the amplitude ratio and phase difference of the two signals, the presence of bubbles within the pipeline can be determined. The results demonstrate that this method is effective in detecting bubbles within the pipeline.
- 15 May 2025
- Published by JVE International Ltd. in Vibroengineering PROCEDIA
- Vol. 59, 82-87
Open AccessMethod of detecting bubbles in a pipeline by vibration excitation
The transportation of liquids through pipelines has extensive applications in the industrial sector. However, due to factors such as localized negative pressure or discontinuities in fluid flow, the formation of bubbles within the pipeline is a common occurrence. Developing an effective method for detecting these internal bubbles is of significant practical importance. Pipelines containing bubbles can be conceptualized as multiphase flow systems, where the frequency domain characteristics of the system vary with changes in bubble volume. Based on this principle, this paper proposes a method for identifying bubbles within pipelines by analyzing the frequency domain features of mechanical vibration waves. To validate the efficacy of this method, a testing apparatus was constructed to use pressure sensors to detect the pressure signals at the inlet and outlet of a pipeline. By analyzing the amplitude ratio and phase difference of the two signals, the presence of bubbles within the pipeline can be determined. The results demonstrate that this method is effective in detecting bubbles within the pipeline.
- 26 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-14
Contribution Imbalance and the Improvement Method in Multisensor Information Fusion-Based Intelligent Fault Diagnosis of Rotating Machinery
The contribution of different signals to rotating machinery fault diagnosis can vary significantly, leading to suboptimal performance in multi-sensor information fusion-based intelligent fault diagnosis (MIF-IFD). This paper examines the issue of imbalanced contributions in MIF-IFD models, explores its causes, and proposes an improvement method. We introduce a contribution discrepancy module to evaluate the contribution of various sensor signals to fault identification. By controlling the training pace of high-contribution branch networks, low-contribution parts are trained sufficiently to keep up. Additionally, a distillation module is added to guide each branch network’s learning direction by using outputs from pre-trained single-sensor networks as supervisory signals. This approach helps mitigate the degradation in feature extraction ability due to imbalanced training. Experimental results show that the proposed method performs well across two datasets and is valuable for practical deployment in MIF-IFD systems.
- 26 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-14
Contribution Imbalance and the Improvement Method in Multisensor Information Fusion-Based Intelligent Fault Diagnosis of Rotating Machinery
The contribution of different signals to rotating machinery fault diagnosis can vary significantly, leading to suboptimal performance in multi-sensor information fusion-based intelligent fault diagnosis (MIF-IFD). This paper examines the issue of imbalanced contributions in MIF-IFD models, explores its causes, and proposes an improvement method. We introduce a contribution discrepancy module to evaluate the contribution of various sensor signals to fault identification. By controlling the training pace of high-contribution branch networks, low-contribution parts are trained sufficiently to keep up. Additionally, a distillation module is added to guide each branch network’s learning direction by using outputs from pre-trained single-sensor networks as supervisory signals. This approach helps mitigate the degradation in feature extraction ability due to imbalanced training. Experimental results show that the proposed method performs well across two datasets and is valuable for practical deployment in MIF-IFD systems.