田惠东, 邹敏, 陈哲涵, 马飞, 刘博深. 基于A-RAFT模型的垂直管道输送测速方法[J]. 工程科学学报. DOI: 10.13374/j.issn2095-9389.2024.03.27.001
引用本文: 田惠东, 邹敏, 陈哲涵, 马飞, 刘博深. 基于A-RAFT模型的垂直管道输送测速方法[J]. 工程科学学报. DOI: 10.13374/j.issn2095-9389.2024.03.27.001
Vertical pipeline velocity measurement method based on A-RAFT model[J]. Chinese Journal of Engineering. DOI: 10.13374/j.issn2095-9389.2024.03.27.001
Citation: Vertical pipeline velocity measurement method based on A-RAFT model[J]. Chinese Journal of Engineering. DOI: 10.13374/j.issn2095-9389.2024.03.27.001

基于A-RAFT模型的垂直管道输送测速方法

Vertical pipeline velocity measurement method based on A-RAFT model

  • 摘要: 流速作为流动特性的重要参数之一,在垂直管道提升效率研究中有着重要地位。将以管道固液两相流为研究对象,研究管道测速的方法。结合深度学习技术,提出了基于注意力机制的A-RAFT神经网络模型,提高了网络对速度场突变区域的估计能力;构建了一个虚实结合的数据集,用于训练神经网络模型。对新提出的模型与数据集进行评估,结果显示:该模型在合成的图像上实现了高精度的速度场计算,与现有其它模型相比,精度提高了15.6%;开展了垂直管道固体颗粒输送的模拟实验,本模型在实验中所采集的真实流场数据上同样展现了准确的估计性能,流速相对误差分别为3.38%、-2.36%与-2.90%。证明该方法在估计精度和模型的泛化能力上均获得了验证。这一研究能够为能源开采、隧道掘进、污水处理以及长距离管道运输等领域中的固液两相流特性分析提供新思路。

     

    Abstract: As one of the important parameters of flow characteristics, flow velocity plays a significant role in the study of vertical pipeline efficiency improvement. This paper focuses on the measurement method of pipeline flow velocity by taking solid-liquid two-phase flow in pipelines as the research object. Combined with deep learning technology, an A-RAFT neural network model based on attention mechanism is proposed, which improves the network's estimation ability for the mutation area of velocity field. A dataset combining virtual and real data is constructed to train the neural network model. The evaluation results of the proposed model and dataset show that the model achieves high-precision velocity field calculation on synthetic images, and the estimation error accuracy is improved by 15.6% compared with other existing models. Simulation experiments for solid particle transport in vertical pipelines are conducted, and the model also demonstrates accurate estimation performance on the collected real flow field data in the experiments, with relative errors of flow velocity being 3.38%, -2.36%, and -2.90%, respectively. This proves that the method has been verified in terms of estimation accuracy and model generalization ability. This research can provide new ideas for the analysis of solid-liquid two-phase flow characteristics in fields such as energy exploitation, tunnel boring, sewage treatment, and long-distance pipeline transportation.

     

/

返回文章
返回