基于人工神经网络的湍流大涡模拟方法

大涡模拟方法(LES)是研究复杂湍流问题的重要工具,在航空航天、湍流燃烧、气动声学、大气边界层等众多工程领域中具有广泛的应用前景.大涡模拟方法采用粗网格计算大尺度上的湍流结构,并用亚格子(SGS)模型近似表达滤波尺度以下的流动结构对大尺度流场的作用.传统的亚格子模型由于只利用了单点流场信息和简单的函数关系,在先验验证中相对误差较大, 在后验验证中耗散过强. 近几年来,机器学习方法在湍流建模问题中得到了越来越多的应用.本文介绍了基于人工神经网络(ANN)的湍流亚格子模型的最新进展.详细地讨论了人工神经网络混合模型、空间人工神经网络模型和反卷积人工神经网络模型的构造方法.借助于人工神经网络强大的数据插值能力,新的亚格子模型的先验精度和后验精度均有显著提升. 在先验验证中,新模型所预测的亚格子应力的相关系数超过了0.99,在预测精度上远高于传统的大涡模拟模型. 在后验验证中,新模型对各类湍流统计量和瞬态流动结构的预测都优于隐式大涡模拟方法、动态Smagorinsky模型、动态混合模型等传统模型.因此, 人工神经网络方法在发展复杂湍流的先进大涡模拟模型中具有很大的潜力.      

基于组合神经网络的雷诺平均湍流模型多次修正方法

求解雷诺平均(Reynolds-averaged Navier-Stokes, RANS)方程依然是工程应用中有效且实用的方法, 但对雷诺应力建模的不确定性会导致该方法的预测精度具有很大差异. 随着人工智能的发展, 湍流闭合模型结合机器学习元素的数据驱动方法被认为是提高RANS模型预测性能的有效手段, 然而这种数据驱动方法的稳定性和预测精度仍有待进一步提高. 本文通过构建一个全连接神经网络对RANS方程中的涡黏系数进行预测以实现雷诺应力的隐式求解,该神经网络记作涡黏系数神经网络(eddy viscosity neural network, EVNN). 此外, 也使用张量基神经网络(tensor basis neural network, TBNN)预测未封闭量与解析量之间的高阶涡黏关系, 并利用基张量保证伽利略不变性. 最后, 采用多次修正的策略实现修正模型对流场预测的精度闭环. 上述方法使用大涡模拟(large eddy simulation, LES)方法产生的高保真数据, 以及RANS模拟获得的基线数据对由EVNN和TBNN组合的神经网络进行训练, 然后用训练好的模型预测新的RANS模拟的流场. 通过与高保真LES结果进行对比, 结果表明, 相比于原始RANS模型, 修正模型对后验速度场、下壁面平均压力系数和摩擦力系数的预测精度均有较大提升. 可以发现对雷诺应力线性部分的隐式处理可以增强数值求解的稳定性, 对雷诺应力非线性部分的修正可以提升模型对流场各向异性特征预测的性能, 并且多次修正后的模型表现出更高的预测精度. 因此, 该算法在数据驱动湍流建模和工程应用中具有很大的应用潜力.      

k-ω SST湍流模式三维激波分离流动修正

“激波?边界层分离”是航空气动领域的典型湍流非平衡流动问题, 准确模拟激波分离对于跨声速飞行器气动性能评估和优化设计具有重要意义. 然而传统涡黏性湍流模式中涡黏性系数的定义方式并不适用于非平衡流动, k-ω SST湍流模式为此引入的Bradshaw假设在应用于三维强逆压梯度和较大分离流动时反而限制了雷诺应力的生成, 导致包括k-ω SST在内的常用涡黏性湍流模式均无法对此类流动进行准确模拟. 同时, 现有的非线性雷诺应力本构关系也并不能有效提高模拟精度. 为此, 针对k-ω SST模式分别提出了基于Bradshaw假设和基于长度尺度的两种激波分离流动修正方法. 前者通过提高Bradshaw常数的方式放宽了对雷诺应力生成的限制, 后者则从湍流长度尺度概念出发, 利用混合长度理论、湍动能生成/耗散之比和一种新定义的长度尺度之比构造了ω方程耗散项修正函数, 提高了模式在三维激波分离流动中的建模长度尺度. 两种方法对ONERA M6机翼跨声速大攻角流动均能得到较雷诺应力模式更好的模拟结果. 进一步的雷诺应力分析表明, 三维激波分离流动中“主雷诺应力分量”的概念不再成立, 各雷诺应力分量大小接近. 网格收敛性分析、对其他攻角状态的验证以及湍流平板边界层壁面律验证进一步确认了所提出的两种修正方法的合理性、有效性和通用性.      

高阶矩湍流模型研究进展及挑战

高阶矩模型是湍流模式理论研究中的难点和前沿.自周培源先生首次建立一般湍流的雷诺应力输运方程起,为了更精确的预测复杂流动,人们从未间断过对高阶矩模型的研究.尤其进入新世纪以来,随着计算机硬件水平的飞跃和高精度数值算法的突破,湍流模拟方法正由RANS向LES转变.而无论对于RANS框架、LES框架还是两者混合,高阶矩模式都...     

含有周期涡结构的湍流雷诺应力与平均速度梯度之间的迟滞效应

针对存在时空周期涡旋结构的湍流场，对湍流雷诺应力与平均速度梯度之间的空间迟滞效应进行了实验研究，发现二者流向相位差呈周期变化．在低速回流式水槽中，利用二维高时间分辨率粒子图像测速(TimeResolved Particle Image Velocimetry, TRPIV)技术，对Re=324的圆柱尾流流场进行测量．经过49个周期166个相位的8 215个PIV瞬时流场，经过周期相位平均，得到一个周期内不同相位典型的湍流雷诺应力和平均速度梯度的空间分布．利用湍流雷诺应力相位平均图像和平均速度梯度图像在不同时间相位下的空间互相关函数最大值对应的流向空间距离，得到湍流雷诺应力与平均速度梯度之间沿流向的空间相位差，并绘制流向相位差随周期相位的演变过程．本文验证了湍流复涡黏模型的合理性，为建立符合周期涡旋结构物理机理的湍流模型研究提供了实验依据．     

基于车身绕流的低雷诺数湍流模型改进研究

本文将汽车绕流模块化为各典型局部流动,通过常用湍流模型对各典型局部流动进行数值模拟,结果验证了湍流模型对转捩的捕捉能力是准确模拟汽车绕流的关键. 在分析汽车绕流分离及转捩机理的基础上,优化了稳态和瞬态求解方法,改进了湍流模型对转捩的预测能力,进而提高了湍流模型在汽车流场模拟上的精度. 针对汽车绕流的稳态问题,将流线曲率因子及 响应阈值引入 LRN $k$-$\varepsilon $ (low Reynolds number $k$-$\varepsilon $) 模型,获得了一种能够更准确预 测转捩的改进低雷诺数湍流模型 (modified LRN $k$-$\varepsilon $),改善了原模型对湍流耗散率的过强依赖性及全应力发展预测不足等问题;针对汽车绕流瞬态求解,通过分析 RANS/LES 混合湍流模型的构造思想及特点,引入约束大涡模拟方法,结合本文提出的改进的 LRN $k$-$\varepsilon $ 湍流模型,提出了一种能准确捕捉转捩现象 的转捩 LRN CLES 模型. 分别将改进的模型用于某实车外流场和风振噪声仿真中,通过 Ansys Fluent 求解器计算,并将计算结果与常用湍流模型的仿真结果、HD-2 风洞试验结果和实车道路实验结果进行对比,表明改进后的湍流模型能够更准确模拟复杂实车的稳态和瞬态特性,为汽车气动特性的研究提供了可靠理论依据及有效数值解决方法.      

EARSM和DRSM在亚音速旋涡流动中的应用

基于非结构网格,采用带曲率修正的显式代数应力非缌出漠型（EARSM）和微分雷诺应力模型（DRSM）,建立了用于模拟大攻角旋涡流动的计算方法。分别以尖前缘和钝前缘的65°三角翼为例,验证了EARSM模型和DRSM模型在两种典型亚音速计算状态下对复杂涡系干扰的模拟能力和旋涡的产生、发展、破裂过程。分别利用SA、BSL两种线性湍流模型对相关问题进行了计算;通过对多种计算的流场与气动力详细结果的比较分析,就几种湍流模型对大攻角复杂旋涡流动的预测能力和敏感性等进行了评估。结果表明：EARSM和DRSM能较好地预测旋涡流动的发展破裂,强于经典线性湍流模型;对边界层发展状况预测较差,导致旋涡启动位置较早。基于算例计算,对两种先进湍流模型的改进提出了适当的建议。所得结论将可为进一步开展大攻角旋涡流动模拟方法的研究提供参考。     

基于雷诺应力模型的平衡大气边界层模拟

基于雷诺应力湍流模型（简称RSM模型）,研究了平衡大气边界层风场数值模拟问题．假设流体不可压,且不计雷诺应力输运方程中的对流项、浮力产生项、系统旋转产生项和扩散项,在准各向同性的条件下,推导出RSM模型湍动能k的表达式是标准k-ε模型k常数表达式的0.893倍．考虑k沿高度变化的修正,根据在标准k-ε模型中满足水平均匀性的湍流来流边界条件,提出在RSM模型中产生平衡大气边界层的湍流来流边界条件．基于空风洞的数值模拟结果表明,与工程上常用的湍流来流边界条件相比,基于本文提出的湍流来流边界条件得到的风场水平均匀性更优,且在整个流域内,得到的雷诺应力剖面更合适．从而验证了该湍流来流边界条件的适用性．     

旋转湍流中雷诺应力模型压力应变关联项的修正研究

利用张量的不变量理论,推导得出传统雷诺应力模型中压力应变关联项模型应用于旋转湍流模拟中的一些基本问题,即在纯旋转条件下,传统模型所描述的初始各向异性的湍流中雷诺应力张量演化规律是一个无衰减振荡过程,而快速畸变理论推导结果显示,其演化应是一个阻尼振荡衰减的过程。以衰减雷诺应力为目的,构造出包含旋转率张量高阶量的关联项。然后,结合变形率张量的高阶项,将修正模型扩展至椭圆形流线类型流动。最后,将修正模型应用于轴向旋转圆管内湍流流场的模拟,并将结果与实测结果进行了对比。     

倾斜吹吸控制下湍流边界层减阻的直接数值模拟

雷诺切应力是壁湍流高摩擦阻力的重要来源, 有理论认为可以通过壁面生成负雷诺应力(数值上为正)的方式来削弱湍流流场中雷诺应力的分布, 以此获得流动减阻. 而通过对雷诺平均运动方程的法向二次积分, 可以发现壁面生成正雷诺应力(数值上为负)对壁面摩擦阻力系数才有负贡献. 文中在湍流边界层流动的控制区域下边界设置一系列倾斜狭缝, 利用该装置通过周期性吹吸的方法产生壁面生成正(负)雷诺应力, 并采用直接数值模拟方法考察和验证上文提到的减阻理论. 文中采用的湍流边界层流动模型, 其流动雷诺数(基于外流速度及动量损失厚度)从300 发展到860. 文中通过多组数值模拟算例, 考察了射流强度和频率对壁面摩擦阻力系数的影响, 并对比了壁面生成正或负雷诺应力对流动的影响. 研究表明, 壁面生成正雷诺应力控制的减阻率能达到3.26, 而壁面生成负雷诺应力控制的减阻效果较壁面生成正雷诺应力控制的要差; 壁面生成的正雷诺应力对壁面摩擦阻力有负贡献, 而壁面生成的负雷诺应力对壁面摩擦阻力有正贡献; 通过考察控制的收支比, 发现控制方案不能获得能量净收益.      

Joint RANS/LES Approach to Premixed Flame Modelling in the Context of the TFC Combustion Model

We present an original timesaving joint RANS/LES approach to simulate turbulent premixed combustion. It is intended mainly for industrial applications where LES may not be practical. It is based on successive RANS/LES numerical modelling, where turbulent characteristics determined from RANS simulations are used in LES equations for estimation of the subgrid chemical source and viscosity. This approach has been developed using our TFC premixed combustion model, which is based on a generalization of the Kolmogorov’s ideas. We assume existence of small-scale statistically equilibrium structures not only of turbulence but also of the reaction zones. At the same time, non-equilibrium large-scale structures of reaction sheets and turbulent eddies are described statistically by model combustion and turbulence equations in RANS simulations or follow directly without modelling in LES. Assumption of small-scale equilibrium gives an opportunity to express the mean combustion rate (controlled by small-scale coupling of turbulence and chemistry) in the RANS and LES sub-problems in terms of integral or subgrid parameters of turbulence and the chemical time, i.e. the definition of the reaction rate is similar to that of the mean dissipation rate in turbulence models where it is expressed in terms of integral or subgrid turbulent parameters. Our approach therefore renders compatible the combustion and turbulent parts of the RANS and LES sub-problems and yields reasonable agreement between the RANS and averaged LES results. Combining RANS simulations of averaged fields with LES method (and especially coupled and acoustic codes) for simulation of corresponding nonstationary process (and unsteady combustion regimes) is a promising strategy for industrial applications. In this work we present results of simulations carried out employing the joint RANS/LES approach for three examples: High velocity premixed combustion in a channel, combustion in the shear flow behind an obstacle and the impinging flame (a premixed flame attached to an obstacle).     

Assessment of two‐equation turbulence modelling for high Reynolds number hydrofoil flows

This paper presents an evaluation of the capability of turbulence models available in the commercial CFD code FLUENT 6.0 for their application to hydrofoil turbulent boundary layer separation flow at high Reynolds numbers. Four widely applied two‐equation RANS turbulence models were assessed through comparison with experimental data at Reynolds numbers of 8.284×106 and 1.657×107. They were the standard k–εmodel, the realizable k–εmodel, the standard k–ωmodel and the shear‐stress‐transport (SST) k–ωmodel. It has found that the realizable k–εturbulence model used with enhanced wall functions and near‐wall modelling techniques, consistently provides superior performance in predicting the flow characteristics around the hydrofoil. Copyright © 2004 John Wiley & Sons, Ltd.     

Development and Analysis of Wall Models for Internal Combustion Engine Simulations Using High-speed Micro-PIV Measurements

The performance, efficiency and emissions of internal combustion (IC) engines are affected by the thermo-viscous boundary layer region and heat transfer. Computational models for the prediction of engine performance typically rely on equilibrium wall-function models to overcome the need for resolving the viscous boundary layer structure. The wall shear stress and heat flux are obtained as boundary conditions for the outer flow calculation. However, these equilibrium wall-function models are typically derived by considering canonical flow configurations, introducing substantial modeling assumptions that are not necessarily justified for in-cylinder flows. The objective of this work is to assess the validity of several model approximations that are commonly introduced in the development of wall-function models for IC-engine applications. This examination is performed by considering crank-angle resolved high-resolution micro-particle image velocimetry (µ-PIV) measurements in a spark-ignition direct-injection single cylinder engine. Using these measurements, the performance of an algebraic equilibrium wall-function model commonly used in RANS and LES IC-engine simulations is evaluated. By identifying shortcomings of this model, a non-equilibrium differential wall model is developed and two different closures are considered for the determination of the turbulent viscosity. It is shown that both wall models provide adequate predictions if applied inside the viscous sublayer. However, the equilibrium wall-function model consistently underpredicts the shear stress if applied in the log-layer. In contrast, the non-equilibrium wall model provides improved predictions of the near-wall region and shear stress irrespective of the wall distance and the piston location. By utilizing the experimental data, significant adverse pressure gradients due to the large vortical motion inside the cylinder (induced by tumble, swirl and turbulence) are observed and included in the non-equilibrium wall model to further improve the model performance. These investigations are complemented by developing a consistent wall heat transfer model, and simulation results are compared against the equilibrium wall-function model and Woschni’s empirical correlation.     

Applications of URANS on predicting unsteady turbulent separated flows

Accurate prediction of unsteady separated turbulent flows remains one of the toughest tasks and a practi cal challenge for turbulence modeling. In this paper, a 2D flow past a circular cylinder at Reynolds number 3,900 is numerically investigated by using the technique of unsteady RANS （URANS）. Some typical linear and nonlinear eddy viscosity turbulence models （LEVM and NLEVM） and a quadratic explicit algebraic stress model （EASM） are evaluated. Numerical results have shown that a high-performance cubic NLEVM, such as CLS, are superior to the others in simulating turbulent separated flows with unsteady vortex shedding.     

Comparisons and applications of numerical simulation methods for predicting aerodynamic heating around complex configurations

Numerical simulation methods of aerodynamic heating were compared by considering the inuence of numerical schemes and turbulence models,and attempting to investigate the applicability of numerical simulation methods on predicting heat flux in engineering applications. For some typical cases provided with detailed experimental data,four spatial schemes and four turbulence models were adopted to calculate surface heat flux. By analyzing and comparing,some inuencing regularities of numerical schemes and turbulence models on calculating heat flux had been acquired. It is clear that AUSM+-up scheme with rapid compressibilitymodified high Reynolds number k-ω model should be appropriate for calculating heat flux. The numerical methods selected as preference above were applied to calculate the heat flux of a 3-D complex geometry in high speed turbulent flows. The results indicated that numerical simulation can capture the complex flow phenomena and reveal the mechanism of aerodynamic heating. Especially,the numerical result of the heat flux at the stagnation point of the wedge was well in agreement with the prediction of Kemp-Riddel formula,and the surface heat flux distribution was consistent with experiment results,which implied that numerical simulation can be introduced to predict heat flux in engineering applications.     

Simulation and Modelling of Turbulent Trailing-Edge Flow

Computations of turbulent trailing-edge flow have been carried out at a Reynolds number of 1000 (based on the free-stream quantities and the trailing-edge thickness) using an unsteady 3D Reynolds-Averaged Navier–Stokes (URANS) code, in which two-equation (k–ε) turbulence models with various low-Re near wall treatments were implemented. Results from a direct numerical simulation (DNS) of the same flow are available for comparison and assessment of the turbulence models used in the URANS code. Two-dimensional URANS calculations are carried out with turbulence mean properties from the DNS used at the inlet; the inflow boundary-layer thickness is 6.42 times the trailing-edge thickness, close to typical turbine blade flow applications. Many of the key flow features observed in DNS are also predicted by the modelling; the flow oscillates in a similar way to that found in bluff-body flow with a von Kármán vortex street produced downstream. The recirculation bubble predicted by unsteady RANS has a similar shape to DNS, but with a length only half that of the DNS. It is found that the unsteadiness plays an important role in the near wake, comparable to the modelled turbulence, but that far downstream the modelled turbulence dominates. A spectral analysis applied to the force coefficient in the wall normal direction shows that a Strouhal number based on the trailing-edge thickness is 0.23, approximately twice that observed in DNS. To assess the modelling approximations, an a priori analysis has been applied using DNS data for the key individual terms in the turbulence model equations. A possible refinement to account for pressure transport is discussed. This revised version was published online in July 2006 with corrections to the Cover Date.     

The development of algebraic stress models using a novel evolutionary algorithm

This work presents developments to a novel evolutionary framework that symbolically regresses algebraic forms of the Reynolds stress anisotropy tensor. This work contributes to the growing trend in machine-learning for modelling physical phenomena. Our framework is shown to be computational inexpensive and produce accurate and robust models that are tangible mathematical expressions. This transparency in the result allows us to diagnose issues with the regressed formulae and appropriately make amendments, as we further understand the regression tools. Such models are created using hybrid RANS/LES flow field data and a passive solving of the RANS transport equations to obtain the modelled time scale. This process shows that models can be regressed from a qualitatively correct flow field and fully resolved DNS is not necessarily required. Models are trained and tested using rectangular ducts, an example flow genus that linear RANS models even qualitatively fail to predict correctly. A priori and a posteriori testing of the new models show that the framework is a viable methodology for RANS closure development. This a posteriori agenda includes testing on an asymmetric diffuser, for which the new models vastly outperform the baseline linear model. Therefore this study presents one of the most rigorous and complete CFD validation of machine learnt turbulent stress models to date.