【作者】 赵会军；

【导师】 张国忠；

【作者基本信息】 中国石油大学 ， 油气储运工程， 2008， 博士

【摘要】 为了确保成品油顺序输送管道高效、安全运行,深入全面准确地了解成品油管道运行特性以及工况变化规律具有重要意义。论文对成品油管道顺序输送运行特性进行了理论分析、数值计算与动态模拟研究。本文提出对顺序输送一维和二维混油浓度以及水力瞬变基本微分方程组数值计算的自适应网格算法,考虑由于混油界面位置发生变化,水击波速也随之变化等因素,推导出计算公式。该方法能够自动根据研究问题中解的特征,生成疏密程度不均的网格点。对于物理量空间变化剧烈或尺度变化较小的区域,网格点通过自动加密以提高网格的分辨率,从而使网格的分布与解的特征相吻合。自适应网格法能够更容易捕捉到压力场、速度场以及浓度场中梯度变化大区域的前沿位置,提高了数值精度,同时也减少计算机运行时间。联立用自适应网格法解的水力瞬变方程、混油对流扩散方程和考虑摩擦生热的热力瞬变方程,给出顺序输送管流耦合的热力-水力-混油耦合模型,建立了相应的算法,该模型结合了特征线和自适应网格法的优点,具有较高地精度。利用人工神经网络的非线性映射能力,建立了人工神经网络混油粘度预测模型,该模型能够预测各种因素非线性影响和不同混油浓度下的混油粘度,与传统经验公式相比,该方法具有误差小,并能够同时考虑温度变化等因素的影响。以Navier-Stokes方程、质量输运方程为基础,从动量质量耦合问题着手,采用κ-ε紊流模式理论,并利用PHOENICS软件进行管内顺序输送特殊工况下的混油数值模拟,计算结果清晰地反映出管内两种油品交界面处的对流扩散传质现象。并重点分析了竖直管道内油品的输送次序对混油的影响,停输、直角弯管、盲支管对混油浓度的影响,为进一步研究混油段经过中间泵站的变化奠定了基础。论文利用人工神经网络技术建立油品掺混浓度-质量控制指标预测模型。首先,通过实验,取得大量实验数据,从而确定掺混量对油品使用性能质量指标的影响,建立了质量指标预测的神经网络模型。其次,在质量预测模型的基础上,通过改变质量预测模型的输入输出量建立混油掺混比例的预测模型即混油浓度预测模型。根据实际的实验数据,文中共建立了三个浓度预测模型,即汽油掺入柴油时,保证柴油闪点合格的浓度预测模型;柴油掺入汽油时,保证汽油终馏点合格的浓度预测模型;当低标号柴油掺入高标号柴油时,保证高标号柴油凝点合格的浓度预测模型。最后,将模型预测值与实验实测值以及经验公式计算值进行了比较,结果表明神经网络模型的预测值与实测值相对误差较小,与其他方法相比,操作简单方便,可以用来控制混油回掺比例,确定混油切割点。在上述研究的基础上,开发了成品油顺序输送管道相关计算软件。更多还原

【Abstract】 To guarantee the pipeline operation optimally, safely and efficiently, it is meaningful to understand the characteristics and rules of batch transportation of multi-product pipeline completely, deeply and exactly. The fundamental theory, numerical computation and dynamic simulation of batch transportation of multi-product pipeline have been studied.The adaptive grid method was adopted in the hydraulic transients, the one-dimensional contamination and two-dimensional contamination.The formula have been presented in the paper, taking into account the varying of propagation velocity of water hammer with the moving of the contamination interface and the other factors. It is much easier for the adaptive grid method to control the size and the density of the mesh and the generation of the grid can be finished quickly and automatically according to the characteristics of solutions. More grid points are redistributed in the large gratitude solution regions in response to numerical solutions and thus the larger the weight function on the grids,the smaller the corresponding grid distance is. The movement of the grids can be written into the governing equations so that the new positions can be worked out at every step allowing the grids to vary continuously and smoothly with the field changing, and thus the forwards position of gratitude solution region of the pressure field, velocity field and concentration field can be captured easily. The result shows that the technique has excellent qualities in improving accuracy of numerical solutions and reducing CPU time. The thermal transient equation, in which the temperature increment due to friction is taken into consideration, is derived on the bases of the continuity equation, momentum conservation equation and energy conservation equation. By combining the thermal transient equation, the hydraulic transient equation and the contamination equation, the coupling thermal-hydraulic-contamination model is offered, in which he corresponding adaptive grid method is also developed and the advantages of characteristics method and adaptive grid technique are combined.The prediction model of contamination viscosity has been established by using the formidable non-linearity mapping ability of artificial neural network. This model can predict contamination viscosity in each given point of the non-linearity influence factor. Its error is smaller than other way and it can completely meet the actual project need.The new contamination mathematical model of batch transportation is established based on the Navier-Stokes equation, the mass transportation equation and the k ?εturbulent flow theory, and the numerical simulation of contamination in pipeline under special operating mode has been investigated with the help of PHOENICS. In the interface of two different oils, its result can reflect clearly the phenomenon of convection and mass diffusion. The influences of transportation order in the erect pipeline and shutdown and blind pipeline and right angle pipe to contamination are analyzed prominently. The results provide a foundation for further study of the influence of tubular specimens and valves to the contamination as the interface flow through the intermediate stations.Acceptable concentration level of mixtures prediction models have been established in the paper originally by using BP neural networks. These models can be used to control doping percentage as well as determine cutting point. Firstly, the performance indicators of pure oil quality are determined according to the experimental data, and quality index forecast models are established with the help of BP neural networks. Secondly, Acceptable concentration level of mixtures prediction models are established on the basis of quality prediction models by changing the input and output of the quality forecast neural network models. Based on the experimental data, three prediction models have been established in this paper, which are flash point forecasting model respect to the incorporation of diesel and gasoline, dry point forecasting model respect to the incorporation of diesel and gasoline, solidifying point prediction model respect to the corporation of low-grade diesel oil and high-grade diesel. Finally, neural model prediction values have been compared with measured values and values predicted by the other method. The result shows that the neural network prediction model is more accurate than other method.The models are simple and convenient to be used to determine cutting time and dissolving ratio.In accordance with the above-mentioned studies, the calculation programs have been developed in the paper.更多还原