HZHongyu Zhong

Hongyu Zhong

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hongyu Zhong has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • GAFT: a small-sample bearing fault diagnostic method based on LoRA fine-tuning and GAN-TrAdaBoost

    by Hongyu Zhong, Samson Yu, Michael Shi, Rui Yuan, Mingcheng Ling, Min Zhao

    Intelligent fault diagnosis methods are indispensable for ensuring the reliable operation of industrial equipment. Currently, the growing complexity of intelligent fault diagnostic methods has substantially increased the demand for training data. However, collecting and labeling data from industrial equipment is both time-consuming and costly, leading to challenges associated with small-sample conditions. This limitation poses a significant hurdle to the development of data-driven intelligent fault diagnosis methods. Inspired by the pre-training and fine-tuning paradigms used in large language models, this study proposes a novel generative adversarial fine-tuning (GAFT) framework for bearing fault diagnosis under small-sample conditions. The framework combines generative adversarial network (GAN)-TrAdaBoost and low-rank adaptation (LoRA) fine-tuning to improve data efficiency. GAN-TrAdaBoost employs synthetic data generated through GAN-based data augmentation to provide robust synthetic data support for the pre-trained model. These synthetic data are then fed into the diagnostic model, where TrAdaBoost dynamically adjusts the weights of synthetic data with varying fidelity. Following this, the diagnostic model undergoes fine-tuning using LoRA, which parameterizes weight updates through low-rank decomposition to reduce the ratio of trainable parameters. Comprehensive experiments on two rotational machinery datasets demonstrate that the proposed GAFT consistently outperforms state-of-the-art methods in both diagnostic accuracy and data efficiency. Through the provision of high-fidelity training data and the refinement of model architecture and fine-tuning techniques, this method delivers a scalable and effective framework for intelligent fault diagnosis in bearing systems.


  • Dual-Level Polynomial Resampling Extracting Transform and Its Application to Bearing Fault Diagnosis at Variable Speeds

    by Yubo Ma, Samson S. Yu, Rui Yuan, Hongyu Zhong, Huawei Wu, Hongan Wu

    Bearing fault diagnosis in construction machinery presents significant challenges due to the variable rotating speeds of mechanical equipment and the influence of industrial noise. The key aspect of this diagnosis lies in accurately extracting and identifying the instantaneous fault characteristic frequency (IFCF) from the background noise of the vibration signal. To address this issue, we propose a novel time-frequency (TF) analysis (TFA) method called dual-level polynomial resampling extracting transform (D-PRET). The D-PRET method offers a high-quality time-frequency representation (TFR), combining energy concentration, precise IFCF estimation, and effective elimination of noise interference. This approach ensures reliable extraction and identification of bearing IFCFs, leading to a more dependable fault diagnosis result. Numerical simulations and two experimental cases demonstrate the effectiveness of D-PRET in bearing fault diagnosis applications.


  • Recursive demodulated synchro spline-kernelled chirplet extracting transform: a useful tool for non-linear behavior estimation of non-stationary signal and application to wind turbine fault detection

    by Yubo Ma, Huawei Wu, Rui Yuan, Hongyu Zhong, Hongan Wu

    Non-linear behavior is widespread in many kinds of signals from nature and engineering fields. Although the high energy-concentration level of various advanced time-frequency (TF) analysis (TFA) techniques currently developed ensure a fine representation of non-linear behavior of time-varying component (TVC) of the signal, it is far from sufficient to solely consider the single aspect of energy-concentration level, because the actual signal composition is always more complicated, especially for some thorny difficulties such as strong noise interference and the early weak TVC, etc., these negative factors bring significant challenges to reveal the non-linear behavior of TVC of practical signals. A new TFA method aimed at this issue, called recursive demodulated synchro spline-kernelled chirplet extracting transform (RDSSCET), is proposed in this paper. The proposed RDSSCET is developed on the frame of synchro spline-kernelled chirplet extracting transform (SSCET) and a newly designed external-internal nested double iteration mechanism, which effectively addresses the limitation of SSCET in handling multicomponent signals while also exhibiting superior high energy concentration and noise robustness. As such, the proposed RDSSCET can yield a more favorable outcome when revealing the non-linear behavior of TVC, particularly for weak TVC with strong noise interference. Comparison analysis results in numerical simulations verified the performance of RDSSCET. Its effectiveness in real applications is fully tested via two real-world sound signals and a practical case of wind turbine fault detection.


  • A time-saving fault diagnosis using simplified fast GAN and triple-type data transfer learning

    by Hongyu Zhong, Samson Yu, Hieu Trinh, Rui Yuan, Yong Lv, Yanan Wang

    Existing intelligent fault diagnosis approaches demand substantial data for training diagnostic models. However, factors such as the inherent characteristics of bearings, operating conditions, and privacy security make collecting comprehensive fault-bearing data very difficult. Although generating synthetic data through generative adversarial networks (GANs) is feasible, the data generation of GANs is a time-consuming process. To address these challenges, a fault diagnosis framework based on GAN and deep transfer learning (DTL) is proposed, termed the simplified fast GAN triple-type data transfer learning (SFGAN-TDTL) method. Initially, an SFGAN is proposed as a replacement for traditional GANs. The time-frequency image data generated by SFGAN serve to augment the training dataset, offering faster and higher-quality data generation compared to traditional GANs. To further reduce time consumption for GAN-based methods, the TDTL method is proposed. Differing from DTL, which utilizes synthetic data to construct a pre-trained model and conducts targeted fine-tuning with real data, TDTL employs open-source data, synthetic data, and real data to fill the weights of the task-insensitive layer, task-sensitive layer, and fully connected layer, respectively. Numerical results demonstrate that SFGAN-TDTL maintains higher diagnostic accuracy while significantly reducing time consumption.


  • Self-Iterated Extracting Wavelet Transform and Its Application to Fault Diagnosis of Rotating Machinery

    by Bowen Li, Rui Yuan, Yong Lv, Hongan Wu, Hongyu Zhong, Weihang Zhu

    Time–frequency (TF) analysis (TFA) is a valuable tool for capturing nonstationary features in time-varying signals. However, for strongly amplitude-modulated and frequency-modulated (AM–FM) signals, how to achieve accurate characterization of time-varying features and how to obtain highly concentrated TF representations (TFRs) are the primary focuses. To address these issues, a novel method called the self-iterated extracting wavelet transform (SIEWT) is proposed in this article. In the proposed method, an iterative precise and high-resolution method for calculating the instantaneous frequency (IF) of a signal is presented. The SIEWT is based on the synchrosqueezing wavelet transform (SWT) and constructs an iterative IF estimator. On this basis, a self-iterated procedure is designed by the Rényi entropy to obtain a more accurate frequency estimation operator, meanwhile reducing the computational cost. After that, the acquired IF estimate is inserted into the extraction operator to further minimize energy dispersion and generate TFR with high TF resolution. The SIEWT also retains the signal reconstruction ability, its accurate estimation of IF plays a critical role in achieving precise signal reconstruction. The effectiveness of the proposed method is thoroughly confirmed by simulation and analysis of real signals, with comparisons made against classical TF analysis methods.


  • A lightweight GAN-based fault diagnosis method based on knowledge distillation and deep transfer learning

    by Hongyu Zhong, Samson Yu, Hieu Trinh, Rui Yuan, Yong Lv, Yanan Wang

    Generative adversarial networks (GANs) have shown promise in the field of small sample fault diagnosis. However, it is worth noting that generating synthetic data using GANs is time-consuming, and synthetic data cannot fully replace real data. To expedite the GAN-based fault diagnostics process, this paper proposes a hybrid lightweight method for compressing GAN parameters. First, three modules are constructed: a teacher generator, a teacher discriminator, and a student generator, based on the knowledge distillation GAN (KD-GAN) approach. The distillation operation is applied to both teacher generator and student generator, while adversarial training is conducted for the teacher generator and the teacher discriminator. Furthermore, a joint loss function is proposed to update the parameters of the student generator by combining distillation loss and adversarial loss. Additionally, the proposed KD-GAN method is combined with deep transfer learning (DTL) and leverages real data to enhance the diagnostic model’s performance. Two numerical experiments are performed to demonstrate that the proposed KD-GAN-DTL method outperforms other GAN-based fault diagnosis methods in terms of computational time and diagnostic accuracy.