Prediction of liquid level distribution in horizontal gas-liquid stratified flows with interfacial level gradient

A one-dimensional momentum equation has been derived based on a two-fluid model and used to predict the axial distribution of liquid level or void fraction in steady, cocurrent, gas-liquid stratified flows in horizontal circular pipes and rectangular channels. The equation is carefully formulated to incorporate the effect of interfacial level gradient. Two different critical liquid levels are found from the momentum equation and are adopted as a boundary condition to calculate the liquid level or void fraction distribution upstream of the channel exit. The predicted void fraction distributions are compared with the experimental data obtained in a rectangular channel in this work and other data reported for large-diameter pipes. Good agreement is shown for air-kerosene, air-water and stream-water stratified flows with a smooth gas-liquid interface.     

关于二相流、多相流、多流体模型和非牛顿流等概念的探讨

本文分析了单相流、二相流和多相流等概念上的差异,也分析了单流体模型、双流体模型和多流体模型等概念上的差异,指出前面三种概念是按流动介质的客观物理构成划分的,而后者是按主观采用的研究方法划分的．目前这些概念在使用中存在一些混乱,如二相流与多相流,多相流与多流体模型等．本文还研究了扩散模型、非牛顿流模型和颗粒流模型等,指出前两种模型在分类上属于单流体模型,分析了非牛顿流模型、扩散模型和双（多）流体模型的特点和应用范围,最后,以泥石流为例讨论了以上概念的应用．     

融合相似性原理的涡轮叶型流场预测方法研究

计算流体力学(CFD)方法是涡轮叶片等设计阶段性能评估的重要手段. 然而, 基于CFD的数值仿真方法通常比较耗时, 难以满足涡轮叶型设计阶段快速迭代的需求. 为实现快速性能评估并克服纯数据驱动预测模型泛化能力不足的问题, 受到物理增强的机器学习思路的启发, 将相似性原理与深度学习模型相结合, 提出了一种泛化能力强的涡轮叶型流场预测新方法. 以涡轮叶片表面等熵马赫数分布预测为例, 提出采用相似性原理对叶型几何变量和气动参数进行归一化, 进而在归一化参数空间构建训练样本集与深度学习预测模型, 由此建立统一的流场预测模型, 对几何尺寸、边界条件差异较大的叶型气动性能进行评估. 在完成模型训练后, 对归一化条件下不同工况/不同形状叶型的流场、真实环境下不同工况/不同尺寸叶型的流场以及GE-E³低压涡轮不同截面叶型的流场进行预测, 结果表明预测结果的分布曲线与CFD评估结果吻合良好, 平均相对误差在1.0%左右, 由此验证了所提出的融合相似性原理的流场预测模型的精度与泛化能力.     

The interaction of background ocean air bubbles with a surface ship

A two-fluid model suitable for the calculation of the two-phase flow field around a naval surface ship is presented. This model couples the Reynolds-averaged Navier–Stokes (RANS) equations with equations for the evolution of the gas-phase momentum, volume fraction and bubble number density, thereby allowing the multidimensional calculation of the two-phase flow for monodisperse variable size bubbles. The bubble field modifies the liquid solution through changes in the liquid mass and momentum conservation equations. The model is applied to the case of the scavenging of wind-induced sea-background bubbles by an unpropelled US Navy frigate under non-zero Froude number boundary conditions at the free surface. This is an important test case, because it can be simulated experimentally with a model-scale ship in a towing tank. A significant modification of the background bubble field is predicted in the wake of the ship, where bubble depletion occurs along with a reduction in the bubble size due to dissolution. This effect is due to lateral phase distribution phenomena and the generation of an upwelling plume in the near wake that brings smaller bubbles up to the surface. © 1998 John Wiley & Sons, Ltd.     

Investigation on effect of drag models on flow behavior of power-law fluid–solid two-phase flow in fluidized bed

In this study, a Eulerian-Eulerian two-fluid model combined with the kinetic theory of granular flow is adopted to simulate power-law fluid–solid two-phase flow in the fluidized bed. Two new power-law liquid–solid drag models are proposed based on the rheological equation of power-law fluid and pressure drop. One called model A is a modified drag model considering tortuosity of flow channel and ratio of the throat to pore, and the other called model B is a blending drag model combining drag coefficients of high and low particle concentrations. Predictions are compared with experimental data measured by Lali et al., where the computed porosities from model B are closer to the measured data than other models. Furthermore, the predicted pressure drop rises as liquid velocity increases, while it decreases with the increase of particle size. Simulation results indicate that the increases of consistency coefficient and flow behavior index lead to the decrease of drag coefficient, and particle concentration, granular temperature, granular pressure, and granular viscosity go down accordingly.     

一种考虑虚拟质量力的两相压力波速经验模型

前人建立的两相压力波速经验模型未考虑虚拟质量力,本文考虑虚拟质量力、管壁弹性、管壁粗糙度等因素,通过求解双流体模型的小扰动,提出了一种新的气液两相压力波速经验模型．以一个具体的工程实例为背景,运用数值方法对其求解,得到的计算结果与前人实测的实验数据一致．结果表明,当空隙率较小时(0＜Φ＜15%)时,虚拟质量力对压力波速的影响不大,当空隙率较大时(Φ≥15%),考虑虚拟质量力计算的压力波速远大于不考虑虚拟质量力计算的压力波速．经验公式也可达到准确求解压力波速的目的．     

EULERIAN–LAGRANGIAN COMPUTATIONS ON PHASE DISTRIBUTION OF TWO-PHASE BUBBLY FLOWS

A comprehensively theoretical model is developed and numerically solved to investigate the phase distribution phenomena in a two-dimensional, axisymmetric, developing, two-phase bubbly flow. The Eulerian approach treats the fluid phase as a continuum and solved Eulerian conservation equations for the liquid phase. The Lagrangian bubbles are tracked by solving the equation of motion for the gas phase. The interphase momentum changes are included in the equations. The numerical model successfully predicts detailed flow velocity profiles for both liquid and gas phases. The development of the wall-peaking phenomenon of the void fraction and velocity profiles is also characterized for the developing flow. For 42 experiments in which the mean void fraction is less than 20 per cent, numerical calculations demonstrate that the predictions agree well with Liu's experimental data. © 1997 by John Wiley & Sons, Ltd.     

EFFECTS OF TENSOR CLOSURE MODELS AND 3-D ORIENTATION ON THE STABILITY OF FIBER SUSPENSIONS IN A CHANNEL FLOW

IntroductionFlowoffibresuspensionshasbeenveryfamiliarinmanyindustrialfields.Fibreadditivesplayanimportantroleindragreductioninmanytypesofflow[1- 3].Inthesuspensions,somebehavioroftheflowmaybealteredbythefibres.Oneoftheimportantexamplesisthehydrodynamicsta…     

颗粒离散元法研究进展

本研究旨在探索一种基于图神经网络(GNN)加速离散元法(DEM)计算的新模型, 以提高颗粒流模拟的计算效率和精度. 传统DEM方法尽管精确, 但计算耗时长. GNN具有模拟DEM的天然优势, 在GNN中颗粒表示为节点, 颗粒的相互作用表示为边. 提出的加速模型包含了两个GNN, 分别是颗粒-颗粒图神经网络(P-P GNN)和颗粒-边界图神经网络(P-W GNN), 能够分别学习颗粒-颗粒和颗粒-边界接触信息. 通过水平滚筒、倾斜滚筒和漏斗堆积3种颗粒流场景的模拟, 验证了该模型的有效性和优越性. 结果表明, 利用GNN模型能够有效地捕捉颗粒流中的复杂接触关系, 极大地提高了计算速度, 相较于传统DEM实现了约30倍的加速. 其次, 模型在不同颗粒流场景下单步预测均表现出高精度, 并且在宏观特征预测上表现优异, 能够准确预测颗粒流的休止角、温度以及重心变化等. 另外, 文章还研究了超参数对预测结果的影响, 如临界距离和GNN的层数, 合适的临界距离可以限制颗粒穿过边界, GNN的层数在3 ~ 10层对预测结果没有显著影响, 这为进一步优化GNN模型应用于颗粒流模拟提供了研究基础.

     

基于离散单元法模拟颗粒流的图神经网络计算加速方法

本研究旨在探索一种基于图神经网络(GNN)加速离散元法(DEM)计算的新模型, 以提高颗粒流模拟的计算效率和精度. 传统DEM方法尽管精确, 但计算耗时长. GNN具有模拟DEM的天然优势, 在GNN中颗粒表示为节点, 颗粒的相互作用表示为边. 提出的加速模型包含了两个GNN, 分别是颗粒-颗粒图神经网络(P-P GNN)和颗粒-边界图神经网络(P-W GNN), 能够分别学习颗粒-颗粒和颗粒-边界接触信息. 通过水平滚筒、倾斜滚筒和漏斗堆积3种颗粒流场景的模拟, 验证了该模型的有效性和优越性. 结果表明, 利用GNN模型能够有效地捕捉颗粒流中的复杂接触关系, 极大地提高了计算速度, 相较于传统DEM实现了约30倍的加速. 其次, 模型在不同颗粒流场景下单步预测均表现出高精度, 并且在宏观特征预测上表现优异, 能够准确预测颗粒流的休止角、温度以及重心变化等. 另外, 文章还研究了超参数对预测结果的影响, 如临界距离和GNN的层数, 合适的临界距离可以限制颗粒穿过边界, GNN的层数在3 ~ 10层对预测结果没有显著影响, 这为进一步优化GNN模型应用于颗粒流模拟提供了研究基础.     

Numerical modelling of cavities on axisymmetric bodies at zero and non-zero angle of attack

The flow about submerged, fully cavitating axisymmetric bodies at both zero and non-zero angle of attack is considered in this paper. A cavity closure model that relates the point of detachment, the angle that the separating streamline makes with the body and the cavity length is described. The direct boundary element method is used to solve the potential flow problem and to determine the cavity shape. A momentum integral boundary layer solver is included in the formulation so that shear stresses can be incorporated into the drag calculations. The numerical predictions based on the proposed closure model are compared with water tunnel measurements and photographs.     

Multiphase CFD-simulation of bubbly pipe flow: A code comparison

CFD simulations of dispersed bubbly flow on the scale of technical equipment become feasible within the Eulerian two-fluid framework of interpenetrating continua. For practical applications suitable closure relations are required which describe the interfacial exchange processes. Implementations of such closures have been provided in major commercial codes for years, but more recently there is a growing interest also in open source packages among which in particular OpenFOAM has become widely known.In the present work a set of closure relations suitable for adiabatic bubbly flow has been implemented in OpenFOAM. Selection of closure models has been based on previous experience with ANSYS-CFX. Great effort has been made to match all details of the models so that only residual differences due to different numerical procedures would be expected in the results. Unfortunately this was not the case and the value of one empirical model parameter had to be changed in order to obtain similar results. Under this provision the new open source implementation is validated and shown to be comparable to commercial codes.     

基于人工神经网络的非均匀固相应力模型

最小多尺度理论EMMS已经被引入多相质点网格法MP-PIC中, 建立了非均匀EMMS固相应力模型. 但现有的非均匀固相应力模型计算中, 中间步骤繁琐且花费时间长. 采用人工拟合的方式能获得非均匀固相应力表达式, 但需要人为确定拟合变量和拟合函数, 且针对于非均匀固相应力这种高度非线性函数所得到的拟合精度不高、用于MP-PIC模拟的结果相比原EMMS固相应力模型结果存在偏差. 针对上述问题, 本文提出通过机器学习的方法, 规避对固相体积分数的局部分布情况的表征, 并提出和建立能考虑颗粒浓度详细分布的人工神经网络ANN固相应力模型. 首先, 基于局部颗粒浓度和颗粒非均匀分布指数建立了双变量的ANN固相应力模型; 进一步将当前网格及其周边网格颗粒浓度组成的序列来详细表征颗粒浓度分布, 并建立相应的ANN固相应力模型. 然后, 将两种模型与EMMS固相应力模型进行了对比并测试了网格分辨率和粗化率对模型的影响. 研究表明: 基于ANN固相应力模型的模拟结果对EMMS固相应力模型结果有较高的还原度, 同时具有一定的网格分辨率无关性和粗化率无关性.