SNShentong Ni

Shentong Ni

*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


Shentong Ni has a total of 24 co-authors. Here are the top co-authors by number of shared publications:

  • Jie Wang

    School of Mechanical Engineering, Sichuan University, Chengdu, Sichuan, China; Yibin Industrial Technology Research Institute of Sichuan University, Yibin, China

    3 shared publications
  • Boqiang Zhang

    School of Mechanical Engineering, Sichuan University, Chengdu, China

    2 shared publications
  • Dongxiao Pang

    Drilling & Production Technology Research Institute, CNPC Chuanqing Drilling Engineering Co., Ltd, Guanghan, China

    2 shared publications
  • Jie Wang

    School of Mechanical Engineering, Sichuan University, Chengdu, China

    2 shared publications
  • Lichun Jia

    Drilling & Production Technology Research Institute, CNPC Chuanqing Drilling Engineering Co., Ltd, Guanghan, China

    2 shared publications
  • Shunjie Xu

    School of Mechanical Engineering, Sichuan University, Chengdu, Sichuan, China; Yibin Industrial Technology Research Institute of Sichuan University, Yibin, China (Equal contributor)

    2 shared publications

Latest Publications

  • A Lightweight YOLOv13 Variant for Real-Time Cavings Detection in Drilling Environments

    by Shentong Ni, Shunjie Xu, Lichun Jia, Dongxiao Pang, Zhilin Li, Jie Wang

    Summary: Real-time monitoring of cavings is critical for preventing borehole instability and ensuring drilling safety. To achieve these functions while meeting the requirements for edge deployment, we propose YOLOv13-CavDet, a novel lightweight deep learning architecture optimized for real-time performance. The network builds upon the state-of-the-art YOLOv13 framework and incorporates a series of targeted, efficiency-oriented enhancements. Specifically, a feature illumination normalization (FIN) module is integrated to decouple illumination from reflectance, restoring weak textures obscured by adverse lighting. Concurrently, a specular edge fusion (SEF) module is designed to amplify structural boundaries and microreflections while suppressing background noise from the shaker mesh. To preserve the fine-grained features of small to medium targets during multiscale fusion, an edge-weighted fusion (EWF) method is introduced. Besides, to achieve a strictly minimal computational budget, the redundant P5 detection head was removed. Experimental evaluations on authentic data sets demonstrate that YOLOv13-CavDet achieves superior detection performance with the lowest computational footprint among benchmarked models, requiring only 8.323 M parameters and 21.6 giga floating-point operations (GFLOPs). This represents an approximate 68% reduction in floating-point operations compared with lightweight baselines like ShuffleNet. In addition, runtime benchmarking shows that the model achieves 150 FPS on an NVIDIA RTX 4060 GPU, confirming its real-time capability and indicating strong potential for future deployment on resource-constrained platforms.

  • Robust drilling pump fault diagnosis with feature domain anomaly repair and reliability aware pooling under overrange signals

    by Bin Chen, Shentong Ni, Yupeng Zhang, Dongyi Wang, Jie Wang

    As a key component in oilfield production, the health of drilling pumps directly affects the safety and efficiency of drilling operations. However, harsh field conditions often cause over-range impacts, sensor saturation, and flat-line segments in vibration measurements, leading to incomplete or unreliable monitoring signals. These degraded measurement segments can severely limit the performance of deep-learning-based diagnostic models. To address this issue, this paper proposes an anomaly signal detection and repair network (ASDR-Net), a feature-level diagnostic framework for drilling-pump signals under over-range and incomplete-signal conditions. Instead of explicitly reconstructing raw vibration waveforms, ASDR-Net handles invalid information in the intermediate feature domain. Specifically, the feature-level anomaly localization module identifies invalid feature intervals caused by over-range corruption, the adaptive non-dependent bidirectional repair module restores local feature continuity using neighboring contextual information, and the dynamic weighted pooling strategy reduces the influence of potentially unreliable repaired features during global aggregation. Experiments on real drilling-pump data show that ASDR-Net maintains stronger diagnostic robustness than baseline and recent incomplete-data-oriented methods under increasing anomaly rates; notably, it achieves 64.8% accuracy even at an anomaly rate of 70%. Additional validation on the public Case Western Reserve University bearing dataset further supports the generalizability of the proposed feature-domain anomaly handling strategy. These results indicate that ASDR-Net provides an effective solution for fault diagnosis under degraded measurement quality and high anomaly-rate incomplete signals.

  • A Fault Diagnosis Method for Incomplete Data via Feature Compensation and Weighted Pooling

    by Shentong Ni, Minghong Xiao, Boqiang Zhang, Yupeng Zhang, Jie Wang, Xia Fang

    Severe signal loss in industrial remote monitoring significantly degrades the performance of fault diagnosis models, particularly when data loss rates exceed 50%. To address this challenge, a Feature Compensation Weighted Pooling Network (FCWPN) is proposed, shifting the diagnostic paradigm from signal reconstruction to robust fault feature recovery. This method utilizes Feature Loss Localization (FLL) to identify missing data segments and a Bidirectional Restoration Fusion (BRF_FC) module to compensate for gaps using contextual information from both directions. Additionally, a Global Weighted Average Pooling (GWAP) mechanism dynamically prioritizes reliable original data over compensated estimates. Experimental results on the Case Western Reserve University (CWRU) dataset demonstrate superior robustness: even with an extreme Signal Loss Rate (SLR) of 80%, FCWPN maintains a diagnostic accuracy of 88.8%, significantly outperforming traditional WDCNN and LSTM models. Furthermore, generalization tests on the Paderborn University dataset confirm the model’s versatility. These findings indicate that bidirectional compensation and dynamic weighting offer a reliable solution for machinery fault diagnosis in information-sparse environments.

  • A Robust Deep Learning Framework for Real-Time Detection of Cavings in Water-Based Mud

    by Shunjie Xu, Shentong Ni, Lichun Jia, Dongxiao Pang, Zhilin Li, Jie Wang

    Summary: Real-time monitoring of cavings is critical for preventing borehole instability and ensuring drilling safety. Unlike previous studies that only focus on detecting cleaned cavings, this study aims to directly detect cavings at the discharge end of the shale shaker. However, this presents numerous challenges for object detection researchers, including harsh field conditions, heavy mud coating on cavings, low target-to-background contrast, high-speed motion blur, and variable illumination. To address these challenges, we propose YOLOv12-CavDet, a novel deep learning architecture for real-time cavings detection. The network builds upon the YOLOv12 framework and introduces a series of synergistic enhancements tailored for complex drilling environments. Specifically, an adaptive multiscale fusion (AMFusion) module is incorporated to alleviate feature confusion caused by low contrast and motion blur. In parallel, a contextual self-calibrated attention (CSCAttn) mechanism is designed to robustly extract salient features from multiscale, irregularly shaped cavings. Furthermore, a corner-aware localization loss (CALLoss) is developed to improve the bounding box accuracy of targets with indistinct boundaries. Experimental evaluations on a custom data set collected from an active drilling site demonstrate that YOLOv12-CavDet outperforms several state-of-the-art object detection models, achieving a precision exceeding 91.5%. Notably, YOLOv12-CavDet demonstrates remarkable robustness under extreme environmental conditions, such as dense fog and low-light scenarios, consistently achieving high-confidence detection results. This research provides a practical solution for the real-time, automated monitoring of wellbore stability, thereby enhancing the safety and efficiency of drilling operations.


  • Open Access

    Risk identification and quantitative assessment method of offshore platform equipment

    by Shentong Ni, Yang Tang, Guorong Wang, Liu Yang, Bo Lei, Zhidong Zhang