高级检索

长距离注浆管道DAS监测下的数字图像堵塞识别模型研究

Research on digital image blockage recognition model under DAS monitoring of long-distance grouting pipelines

  • 摘要: 高效可靠的智能化监测对保障长距离注浆管道的正常作业具有重要意义。针对长距离注浆管道输送过程中易产生堵塞的问题,提出了一种结合分布式声波传感技术(DAS)及分频段式点云配准模型(FR-LoFTR-2D)的监测手段。采用分频段式去噪模块(FR)对管道堵塞数据进行预处理,进一步将其转换为频带能量(FBE),并基于此搭建点云配准模型进行堵塞点位图像识别,最后结合相机投影模块(2D)实现空间点坐标的降维转换以提高定位直观性,进而实现长距离注浆管道的堵塞点定位问题。为评估提出的技术方案,搭建了129.71 m的室外环管试验模型和513 m的工业性环管试验模型,分别进行配准模型的搭建及验证。测试结果表明:FR模块可以在保留特征信号的基础上,初步剔除由注浆泵震动引发的管道震动噪声。针对DAS相位转换得到的FBE三维图像,FR-LoFTR-2D模型能够有效实现特征配准及堵塞点的定位功能。在室外环管试验和工业性环管试验中,该模型的堵塞点定位置信度分别为92.4%和89.3%。在管路增长约297%的工况下,模型识别置信度下降约为3.1%。该研究应用DAS全覆盖式监测长距离注浆管道,并结合其监测数据特点搭建了FR-LoFTR-2D堵塞识别模型,为长距离注浆管道监测及堵塞点的定位问题提供了直观有效的解决方案。

     

    Abstract: The efficient and reliable intelligent monitoring of long-distance grouting pipelines is crucial for ensuring their normal operation. To address the clogging issues often encountered during the transportation process, a monitoring approach combining Distributed Acoustic Sensing (DAS) technology and a frequency-band-based point cloud registration model (FR-LoFTR-2D) is proposed. A frequency-band denoising module (FR) is used for preprocessing the clogging data, which is then converted into Frequency Band Energy (FBE). Based on this, a point cloud registration model is developed for the recognition of clogging points. Additionally, a camera projection module (2D) is employed to perform dimensionality reduction and spatial point coordinate transformation, improving the visualization of the positioning results. This approach aims to resolve the issue of clogging point localization in long-distance grouting pipelines. To evaluate the proposed technique, experimental models of a 129.71 m outdoor loop pipeline and a 513m industrial loop pipeline were constructed and used for model development and validation. The test results indicate that the FR module can effectively filter out pipeline vibration noise induced by the grouting pump, while preserving the characteristic signals. For the FBE three-dimensional images obtained from the DAS phase conversion, the FR-LoFTR-2D model can effectively achieve feature registration and clogging point localization. In the outdoor loop test and industrial loop test, the model’s clogging point localization accuracy was 92.4% and 89.3%, respectively. When the pipeline length increased by approximately 297%, the model's recognition accuracy decreased by approximately 3.1%. This study applies DAS for full-coverage monitoring of long-distance grouting pipelines and, based on the characteristics of the monitoring data, develops the FR-LoFTR-2D clogging recognition model, providing an intuitive and effective solution for monitoring and clogging point localization in long-distance grouting pipelines.

     

/

返回文章
返回