基于POD-RBF方法的管道内壁几何识别

针对天然气、石油等管道内部被腐蚀问题,基于本征正交分解-径向基函数(POD-RBF)提出了一种管道内壁几何识别方法.考虑静磁场并建立管道的简化有限元模型,构建变几何样本库,实现了POD-RBF对任意形状的响应预测.该方法在降阶分析的同时避免了迭代过程中因几何的改变需反复求解刚度矩阵,在很大程度上提高了计算效率.采用灰狼优化(GWO)算法对目标函数实施优化,避免了在变几何过程中灵敏度的求解.算例结果显示,该文方法可高效准确地反演管道内壁的几何形状,即使在引入噪声后GWO算法仍具有较好的稳定性.

超声波防垢管道参数实验研究

采用循环水模拟装置,在相同流量、相同温度条件下,研究了管道形状、管道口径对超声防垢率的影响.结果表明,管道形状、口径对超声防垢率均有一定影响.在相同频率、相同功率超声作用下,直管处局部防垢率比弯管、变截面管处局部防垢率高,但对循环系统,总的防垢率相差不多;在相同频率、相同声强超声作用下,直管口径越大,直管处局部防垢率越高.

基于PCA-BP神经网络的管道内壁几何形状识别

针对石油和天然气等管道内部因腐蚀等因素造成的结构性缺陷问题,基于主成分分析(PCA)和反向传播(BP)神经网络提出了一种有效的管道内壁几何识别方法:首先,借助CAD二次开发实现随机管壁几何模型的自动化建模,并结合二维恒定磁场有限元分析来获取测点磁场响应,建立了自适应样本库生成系统;然后,采用主成分分析法对测点磁场响应数据进行降维处理,去除了数据中的冗余信息;最后,利用Levenberg-Marquardt算法改进的BP神经网络基于测点磁场响应与管壁几何参数之间的映射关系对管壁几何形状进行反演识别.算例结果表明,基于PCA-BP神经网络的管道内壁几何反演分析模型可快速准确地预测管道内壁的几何形状.即使对于带有不同程度随机误差的复杂几何,该方法仍具备强大的识别性能.

Characteristics of turbulent flow distribution in branch piping system

Flow distribution in branch piping system is affected by flow characteristics and different geometric variations. Most of the flow distribution studies are performed with one-dimensional analysis to yield overall information only. However, detailed analysis is required to find effects of design parameters on the flow distribution. For this aspect, three-dimensional turbulent flow analysis was performed to assess turbulence model performance and effects of upstream pressure and branch pipe geometry. Three different turbulence models of standard k-e model, realizable k-e model and standard k-co yield similar results, indicating small effects of turbulence models on flow characteristics analysis. Geometric variations include area ratio of main and branch pipes, branch pipe diameter, and connection shape of main and branch pipes. Among these parameters, area ratio and branch diameter and shape show strong effect on flow distribution due to high friction and minor loss. Uniform flow distribution is one of common requirements in the branch piping system and this can be achieved with rather high total loss design.