ZWZhaoqian Wu

Zhaoqian Wu

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

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


Zhaoqian Wu has a total of 47 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • GDTS-YOLOv8: Gather-and-Distribute Mechanism With Two-Stage Self-Fine-Tuning Strategy for Multiscale Capacitor Defect Detection Under Imbalanced Data

    by Gen Chen, Fei Jiang, Shaohui Zhang, Zhi Zhong, Zhaoqian Wu, Weihua Li

    Capacitors are essential in industrial applications due to their ability to store and release electrical energy rapidly. However, capacitors, with their complex manufacturing processes and heterogeneous characteristics, pose two critical challenges to conventional detection algorithms: 1) the difficulty in detecting small-scale defects amid multiscale defect variations and 2) the inherently imbalanced distribution of defects. To address these issues, the GDTS-YOLOv8 detection framework is introduced, which leverages a gather-and-distribute (GD) mechanism and a two-stage self-fine-tuning (TS) strategy with a frozen backbone. First, GD is constructed to enhance the efficiency of multiscale feature extraction and fusion in the YOLOv8 feature fusion layer. Centered on optimizing cross-level information exchange and reducing information loss, this mechanism can effectively address the issue of poor detection performance for small targets in multiscale scenarios. Second, TS is proposed to address the inherent imbalance problem, comprising: 1) a rare-defect upweighting pretraining stage with label reduction of frequent defects and 2) a full-data fine-tuning stage integrating label reintroduction of frequent defects with backbone freezing for optimal performance. Experimental results confirm the proposed method’s effectiveness, enhancing capacitor detection precision by over 5% (75.3% peak accuracy) and demonstrating significant improvements in small-defect detection and imbalanced sample handling.

  • Predicting surface pressure fields and lift coefficients of subsonic airfoils by machine-learning-enhanced compressive sensing framework

    by Yue Zhang, Shaohui Zhang, Fei Jiang, Zhaoqian Wu, Weihua Li

    Aerodynamic performance assessment is pivotal in aircraft design optimization, yet traditional airfoil optimization methods remain computationally expensive and inefficient. To address this challenge, we introduce a machine-learning-enhanced compressive sensing (ML-CS) framework that synergizes data-driven learning with compressive sensing to accurately predict surface pressure fields and lift coefficients of subsonic airfoils. A geometry-informed sparse sampling strategy is developed using airfoil camber-thickness curve parameterization, enabling efficient feature extraction from limited training data. The development of the ML-CS framework employs a three-stage process: first, the ML-CS reconstruction efficiency is optimized, achieving more than 92% accuracy in pressure coefficient prediction using only ten sparse sensors; second, a Kriging-surrogate-assisted sensor value prediction method that robustly preserves prediction accuracy despite missing sensor data is proposed; third, a lift coefficient prediction module that reduces errors by 35.5% compared to conventional surrogate models while simultaneously reconstructing surface pressure distributions is proposed. The framework demonstrates superior efficiency and generalizability, offering a paradigm shift in aerodynamic surrogate modeling.


  • MTSNN: A Few-Shot Fine-Grained Diagnosis Framework for Cross-Machine Fault Diagnosis With Heterogeneous Fault Categories

    by Jipu Li, Ke Yue, Zhaoqian Wu, Fei Jiang, Zhi Zhong, Shaohui Zhang, Weihua Li

    Deep learning (DL)-based intelligent fault diagnosis (IFD) methods are extensively used in mechanical systems. In practical industrial contexts, DL models are constrained by three primary limitations. (1) The performance of DL models heavily relies on a significant volume of high-quality labeled data during the training process. (2) The diagnostic capabilities of DL models are insufficient to handle the complex and variable conditions encountered in mechanical systems. (3) Current DL-based IFD approaches primarily focus on diagnosing coarse-grained faults, while less attention is given to fine-grained fault detection. In response to these limitations, a new Meta-Transfer Spiking Neural Network (MTSNN) is introduced for cross-machine fine-grained fault diagnosis in mechanical equipment. Firstly, a convolutional spiking neural network (CSNN) is constructed to extract the representative features from raw signals by leveraging the low energy consumption, native ability to process time-series data, and high robustness to noise of spiking neural network. Secondly, a novel comprehensive loss function called meta-prototype contrastive loss is proposed, which integrates prototypical learning loss, meta contrastive learning loss, and entropy information maximization loss, allowing the model to effectively transfer diagnostic knowledge across different machines in limited samples scenario. The experimental evaluation conducted on three datasets reveals that the proposed MTSNN significantly outperforms other methods in diagnosing faults, particularly when applied to different machines with heterogeneous fault categories.


  • A tool wear monitoring method driven by workpiece surface quality inversion and EfficientNet

    by Fei Jiang, Zejian Xiao, Xiaohui Yan, Zhaoqian Wu, Shaohui Zhang, Rizwan Ullah

    Tool wear status monitoring plays an important role in the safety of machine and the efficiency of production. Due to the limitation of sensor installation and complex operation conditions, traditional machine vision methods are difficult to directly act on the tool itself to obtain corresponding wear status information. Meanwhile, the surface quality of the machined workpiece can provide indirect insights into the wear status of the tool to a significant degree. Therefore, an indirect tool wear status monitoring method is proposed in this paper by workpiece surface quality inversion and EfficientNet. Firstly, the surface texture data of workpiece are obtained in the form of image by an industrial camera, and the representative features that indirectly reflect tool wear status are further highlighted by using the weighted average image grayscale and the Laplacian operator. Secondly, to monitor tool wear effectively, we have optimized three key settings, width, model depth, and input image resolution, of the EfficientNet backbone network. This optimization boosts its ability to understand and generalize, ensuring it can handle various surface textures and extract crucial information accurately. This, in turn, enhances the accuracy of our tool wear monitoring system. Finally, the improved EfficientNet network is further utilized as a pre-trained network to achieve tool wear state monitoring. Utilizing the powerful representation ability of the pre-trained model and the surface texture features of workpiece, the model can accurately identify and judge the wear degree of the tool. The tool experiment results show that the proposed tool wear monitoring method has achieved remarkable results with recognition accuracy of 97.77%. While other popular methods such as VGG19, ResNet50, and ResNet152, whose training accuracies are 95.66%, 96.37%, and 91.27%, respectively. The comparative results highlight the excellent performance of the proposed method in recognition rate, training time and generalization performance.


  • Hybrid contrastive representations and SBO-based samples generation for rotating machinery anomaly detection based on driven-end current signal

    by Weiqi Lin, Shaohui Zhang, Tao Li, Fei Jiang, Zhaoqian Wu, Zhi Zhong

    While deep learning methods based on cross-entropy function have made great advancements in maintaining equipment reliability, their learning paradigm appears less suitable for anomaly detection tasks, which focus on extracting specific target features and ignore the discriminability between classes. Besides, incomplete dataset still hinders the robustness of intelligent detection models, especially when the training dataset only contains normal samples. Therefore, a hybrid contrastive representation and soft Brownian offset (SBO)-based samples generation (HCRS) method based on driven-end current signals is proposed to distinguish normal and abnormal samples in incomplete data scenarios where only normal signals are available. In the proposed HCRS detection framework, an autoencoder is initially trained using only normal current samples. It is then combined with the SBO method to generate abnormal samples, relieving the issues of incomplete dataset. Subsequently, a supervised contrastive learning-based deep feature extractor is trained using both the generated abnormal samples and the collected normal samples. This process aims to learn high-level fine-grained representations with discriminability. Finally, these learned representations are utilized to train a data-driven classifier, enabling effective anomaly detection in rotating machinery. In addition, experiments on two datasets suggest that the proposed HCRS can effectively achieve higher accuracy anomaly detection with only normal current signals and outperform existing detection approaches.


  • Source-Free Progressive Domain Adaptation Network for Universal Cross-Domain Fault Diagnosis of Industrial Equipment

    by Jipu Li, Ke Yue, Zhaoqian Wu, Fei Jiang, Zhi Zhong, Weihua Li, Shaohui Zhang

    Recently, transfer learning-based intelligent fault diagnosis methods have been extensively adopted in the realm of industrial equipment. A fundamental assumption that the source and target domains have matching fault types is effectively resolved. Unfortunately, existing methods fail to account for two limitations in real-world applications: (1) The existing methods are limited to specific domain adaptation scenarios, which makes it difficult to achieve satisfactory results. (2) The existing methods do not consider data privacy protection because they require both source and target samples during the training stage. To address these challenges, a novel source-free progressive domain adaptation network (SPDAN) is proposed to simultaneously handle multiple domain adaptation scenarios without accessing source samples. Firstly, a neighbor searching-based trustworthy pairs construction is utilized to provide the high-confident nearest fault samples. Secondly, an instance alignment-based domain shift reduction is used to eliminate the data distribution discrepancy of different domains. Finally, an information entropy-based novel fault detection is employed to identify unknown fault samples. Experiments on two bearing datasets validate the proposed SPDAN. The experiments confirm that the proposed SPDAN can successfully operate in multiple domain adaptation scenarios without relying on source samples, making it a highly promising approach for diagnosing faults in industrial equipment.


  • CNCR-YOLO: A Comprehensive Optimization Strategy for Small-Target Defect Detection in Injection-Molded Parts

    by Fei Jiang, Wenjing Ye, Peicong Lu, Shaohui Zhang, Zhaoqian Wu, Haifei Zhu

    Defect detection in injection molded parts is crucial for quality control in industrial manufacturing. However, it poses challenges due to process complexity, particularly for detecting small defects. To address this challenge, an efficient detection algorithm named as CNCR-YOLO is proposed. Firstly, to both improve the quality and quantity of injection molded parts defect samples, StyleGANv3 data generation and traditional augmentation method are utilized as pre-processing. Moreover, inspired by YOLOv5, CNCR-YOLO combines the strength of Normalized Wasserstein Distance and Complete Intersection over Union metrics together to raise the CI-NWD loss function for better optimization; Finally, CBAM attention mechanism and Res2net module are embedded into YOLOv5 to enhance the ability of target feature learning and multi-scale feature extraction. Experimental results demonstrate that the proposed method outperforms the other algorithms, achieving an improvement of 1.5% in Mean Average Precision (MAP), notably 2.3% for small target defects. Additionally, its generality and robustness are further validated on an industrial PCB defect detection dataset.


  • Speed harmonic suppression of PMSM and its application in ball screw vibration suppression

    by Zhaoqian Wu, Wang Guo, Zhenhua Liu, Canyang Jiang, Fei Jiang, Shaohui Zhang

    The speed ripple of surface-mounted permanent magnet synchronous motor (SPMSM) causes vibration of the ball screw, which degrades the accuracy and stability of the ball screw system. The harmonic injection method can suppress the speed ripple by decreasing the speed harmonic of SPMSM. When the speed ripple is suppressed, the vibration of the ball screw is also reduced. However, large amount of calculation, complex parameter adjustment, and long suppression time limit its performance in suppressing vibration of ball screw. Therefore, in this paper, we propose a speed harmonic suppression method based on the online measurement of the transfer function from inject voltage harmonic to speed harmonic. In the proposed method, the transfer function is measured after the motor reaches a stable operating condition. And then, the designed speed harmonic controller parameters are set according to the measurement results. Compared with the existing method, the method based on online measurement has less calculation, simple parameter adjustment, fast suppression speed, and good performance in suppressing the vibration of ball screw. Experimental verifications are carried out in the forward and backward condition of the ball screw.


  • An Improved Displacement Sensor Accuracy Calibration Method for Ball Screw Positioning Error Compensation

    by Zhaoqian Wu, Canyang Jiang, Fei Jiang, Shaohui Zhang, Jinglun Liang, Rizwan Ullah

    In a ball screw precision positioning system, a high-precision external displacement sensor can effectively identify the positioning error of the ball screw. However, high accuracy comes with a high cost. Ultrasonic sensors offer an inexpensive, in-line technique that can be integrated with neural networks, to improve the precision of ultrasonic sensor. In this paper, an ultrasonic sensor accuracy calibration mothed is proposed, namely, frequency feature extraction deep residual shrinkage network (FFE-DRSN), which used probability statistics and test data to enhance the resolution of ultrasonic sensors. The efficacy of the developed methods is validated through experiments of the ball screw positioning error compensation with different work conditions. The experimental results verify that the FFE-DRSN precision correction model improves positioning accuracy by at least 78% compared to the uncompensated case.