Investigation of the sensitivity of turbulent closures and coupling of hybrid RANS‐LES models for predicting flow fields with separation and reattachment

Hybrid models have found widespread applications for simulation of wall‐bounded flows at high Reynolds numbers. Typically, these models employ Reynolds‐averaged Navier–Stokes (RANS) and large eddy simulation (LES) in the near‐body and off‐body regions, respectively. A number of coupling strategies between the RANS and LES regions have been proposed, tested, and applied in the literature with varying degree of success. Linear eddy‐viscosity models (LEVM) are often used for the closure of turbulent stress tensor in RANS and LES regions. LEVM incorrectly predicts the anisotropy of Reynolds normal stress at the RANS‐LES interface region. To overcome this issue, use of non‐linear eddy‐viscosity models (NLEVM) have started receiving attention. In this study, a generic non‐linear blended modeling framework for performing hybrid simulations is proposed. Flow over the periodic hills is used as the test case for model evaluation. This case is chosen due to complex flow physics with simplified geometry. Analysis of the simulations suggests that the non‐linear hybrid models show a better performance than linear hybrid models. It is also observed that the non‐linear closures are less sensitive to the RANS‐LES coupling and grid resolution. Copyright © 2016 John Wiley & Sons, Ltd.     

Modelling turbulence around and inside porous media based on the second moment closure

To predict turbulence in porous media, a new approach is discussed. By double (both volume and Reynolds) averaging Navier–Stokes equations, there appear three unknown covariant terms in the momentum equation. They are namely the dispersive covariance, the macro-scale and the micro-scale Reynolds stresses, in the present study. For the macro-scale Reynolds stress, the TCL (two-component-limit) second moment closure is applied whereas the eddy viscosity models are applied to the other covariant terms: the Smagorinsky model and the one-equation eddy viscosity model, respectively for the dispersive covariance and the micro-scale Reynolds stress. The presently proposed model is evaluated in square rib array flows and porous wall channel flows with reasonable accuracy though further development is required.     

Coherent structures in trailing-edge cooling and the challenge for turbulent heat transfer modelling

The present paper tests the capability of a standard Reynolds-Averaged Navier–Stokes (RANS) turbulence model for predicting the turbulent heat transfer in a generic trailing-edge situation with a cutback on the pressure side of the blade. The model investigated uses a gradient-diffusion assumption with a scalar turbulent-diffusivity and constant turbulent Prandtl number. High-fidelity Large-Eddy Simulations (LES) were performed for three blowing ratios to provide reliable target data and the mean velocity and eddy viscosity as input for the heat transfer model testing. Reasonably good agreement between the LES and recent experiments was achieved for mean flow and turbulence statistics. The LES yielded coherent structures which were analysed, in particular with respect to their effect on the turbulent heat transfer. For increasing blowing ratio, the LES replicated an also experimentally observed counter-intuitive decrease of the cooling effectiveness caused by the coherent structures becoming stronger. In contrast, the RANS turbulent heat transfer model failed in predicting this behaviour and yielded significantly too high cooling effectiveness. It is shown that the model cannot predict the strong upstream and wall-directed turbulent heat fluxes caused by large coherent structures, which were found to be responsible for the counter-intuitive decrease of the cooling effectiveness.     

Unified turbulence models for LES and RANS,FDF and PDF simulations

A review of existing basic turbulence modeling approaches reveals the need for the development of unified turbulence models which can be used continuously as filter density function (FDF) or probability density function (PDF) methods, large eddy simulation (LES) or Reynolds-averaged Navier–Stokes (RANS) methods. It is then shown that such unified stochastic and deterministic turbulence models can be constructed by explaining the dependence of the characteristic time scale of velocity fluctuations on the scale considered. The unified stochastic model obtained generalizes usually applied FDF and PDF models. The unified deterministic turbulence model that is implied by the stochastic model recovers and extends well-known linear and nonlinear LES and RANS models for the subgrid-scale and Reynolds stress tensor.      

A dynamic framework for functional parameterisations of the eddy viscosity coefficient in two-dimensional turbulence

This paper puts forth a dynamic framework for investigating the subgrid scale physics of decaying two-dimensional turbulence utilising a modular approach with eddy viscosities in various functional forms. The derivation of the low-pass spatially filtered implementation of the Navier–Stokes equations is given by using the vorticity-streamfunction formulation. Two different implementations of the viscosity kernels based on the representation of the eddy viscosity terms are proposed and tested by solving a canonical two-dimensional decaying turbulence problem. It is seen that the implementation of the eddy viscosity formulation plays a distinct role in the dissipative behaviour of the different viscosity kernels. Among eddy viscosity kernels tested, we found that the Leith eddy viscosity formulation yields superior results with higher correlation coefficients.     

On the simulation of the flow around a circular cylinder at Re=140,000

The flow around a circular cylinder at Reynolds number of 1.4 × 10⁵ is examined with Reynolds-Averaged Navier–Stokes equations (RANS) and Scale-Resolving Simulation (SRS) methods. Such problem is in the upper limit of the flow regime where turbulent transition occurs in the free shear-layers and so the flow dynamics is dominated by the spatial development of vortex-shedding structure, and in particular by the Kelvin–Helmholtz rollers and turbulence onset. The objectives of this investigation are threefold: (i) determine the aptitude of distinct RANS and SRS models to simulate the correct flow regime; (ii) compare the predictions of selected methods with available experimental measurements; and (iii) examine key modelling and flow features that contribute to the observed results. The evaluated models range from RANS supplemented with linear, transition, and non-linear turbulent viscosity closures, to hybrid and bridging SRS methods. Bridging computations are conducted at various constant degrees of physical resolution (range of resolved scales). The results illustrate the complexity of predicting the present flow problem. It is shown that RANS and SRS formulations modelling turbulence in boundary-layers with the selected linear turbulent viscosity closures lead to a premature onset of turbulence which alters the flow regime of the simulations. Although the transition and non-linear RANS closures can predict the correct flow regime, the outcome of this study indicates that solely the bridging model at constant physical resolution is able to achieve an accurate and physics-based prediction of the flow dynamics. Nonetheless, the necessary degree of physical resolution makes the numerical requisites of such computations demanding.     

Unsteady CFD simulations of a pump in part load conditions using scale-adaptive simulation

The scope of this work is to demonstrate the applicability of an eddy resolving turbulence model in a turbomachinery configuration. The model combines the Large Eddy Simulation (LES) and the Reynolds Averaged Navier Stokes (RANS) approach. The point of interest of the present investigation is the unsteady rotating stall phenomenon occurring at low part load conditions. Since RANS turbulence models often fail to predict separation correctly, a LES like model is expected to give superior results. In this investigation the scale-adaptive simulation (SAS) model is used. This model avoids the grid dependence appearing in the Detached Eddy Simulation (DES) modelling strategy. The simulations are validated with transient measurement data. The present results demonstrate, that both models are able to predict the major stall frequency at part load. Results are similar for URANS and SAS, with advantages in predicting minor stall frequencies for the turbulence resolving model.     

Hybrid RANS/LES computations of plane impinging jets with DES and PANS models

The qualities of a DES (Detached Eddy Simulation) and a PANS (Partially-Averaged Navier–Stokes) hybrid RANS/LES model, both based on the k–ω RANS turbulence model of Wilcox (2008, “Formulation of the k–ω turbulence model revisited” AIAA J., 46: 2823–2838), are analysed for simulation of plane impinging jets at a high nozzle-plate distance (H/B = 10, Re = 13,500; H is nozzle-plate distance, B is slot width; Reynolds number based on slot width and maximum velocity at nozzle exit) and a low nozzle-plate distance (H/B = 4, Re = 20,000). The mean velocity field, fluctuating velocity components, Reynolds stresses and skin friction at the impingement plate are compared with experimental data and LES (Large Eddy Simulation) results. The k–ω DES model is a double substitution type, following Davidson and Peng (2003, “Hybrid LES–RANS modelling: a one-equation SGS model combined with a k–ω model for predicting recirculating flows” Int. J. Numer. Meth. Fluids, 43: 1003–1018). This means that the turbulent length scale is replaced by the grid size in the destruction term of the k-equation and in the eddy viscosity formula. The k–ω PANS model is derived following Girimaji (2006, “Partially-Averaged Navier–Stokes model for turbulence: a Reynolds-Averaged Navier–Stokes to Direct Numerical Simulation bridging method” J. Appl. Mech., 73: 413–421). The turbulent length scale in the PANS model is constructed from the total turbulent kinetic energy and the sub-filter dissipation rate. Both hybrid models change between RANS (Reynolds-Averaged Navier–Stokes) and LES based on the cube root of the cell volume. The hybrid techniques, in contrast to RANS, are able to reproduce the turbulent flow dynamics in the shear layers of the impacting jet. The change from RANS to LES is much slower however for the PANS model than for the DES model on fine enough grids. This delays the break-up process of the vortices generated in the shear layers with as a consequence that the DES model produces better results than the PANS model.     

Exact transport equation for eddy diffusivity in turbulent shear flow

Two-equation models that treat the transport equations for two variables are typical models for the Reynolds-averaged Navier–Stokes equation. Compared to the equation for the turbulent kinetic energy, the equation for the second variable such as the dissipation rate does not have a theoretical analogue. In this work, the exact transport equation for the eddy diffusivity was derived and examined for better understanding turbulence and improving two-equation models. A new length scale was first introduced, which involves the response function for the scalar fluctuation. It was shown that the eddy diffusivity can be expressed as the correlation between the velocity fluctuation and the new length scale. The transport equations for the eddy diffusivity and the length-scale variance were derived theoretically. Statistics such as terms in the transport equations were evaluated using the direct numerical simulation of turbulent channel flow. It was shown that the streamwise component of the eddy diffusivity is greater than the other two components in the whole region. In the transport equation for the eddy diffusivity, the production term due to the Reynolds stress is a main positive term, whereas the pressure–length-gradient correlation term plays a role of destruction. It is expected that the analysis of the transport equations is helpful in developing better turbulence models.     

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.      

Implicit unfactored implementation of two-equation turbulence models in compressible Navier–Stokes methods

An implicit unfactored method for the coupled solution of the compressible Navier–Stokes equations with two-equation turbulence models is presented. Both fluid-flow and turbulence transport equations are discretized by a characteristics-based scheme. The implicit unfactored method combines Newton subiterations and point-by-point Gauss–Seidel subrelaxation. Implicit-coupled and -decoupled strategies are compared for their efficiency in the solution of the Navier–Stokes equations in conjunction with low-Re two-equation turbulence models. Computations have been carried out for the flow over an axisymmetric bump using the k–ϵ and k–ω models. Comparisons have been obtained with experimental data and other numerical solutions. The present study reveals that the implicit unfactored implementation of the two-equation turbulence models reduces the computing time and improves the robustness of the CFD code in turbulent compressible flows. © 1998 John Wiley & Sons, Ltd.     

A Hybrid RANS/LES Simulation of Turbulent Channel Flow

Hybrid models combining large eddy simulation (LES) with Reynolds-averaged Navier–Stokes (RANS) simulation are expected to be useful for wall modeling in the LES of high Reynolds number flows. Some hybrid simulations of turbulent channel flow have a common defect; the mean velocity profile has a mismatch between the RANS and LES regions due to a steep velocity gradient at the interface. This mismatch is reproduced and examined using a simple hybrid model; the Smagorinsky model is switched to a RANS model increasing the filter width. It is suggested that a rapid spatial variation in the eddy viscosity is responsible for an underestimate of the grid-scale shear stress and for the steep velocity gradient. To reduce the mean velocity mismatch a new scheme is proposed; additional filtering is introduced to define two kinds of velocity components at the interface between the two regions. The two components are used to remove inconsistency in the velocity equations due to a rapid variation in the filter width. Using the new scheme, simulations of channel flow are carried out with the simple hybrid model. It is shown that the grid-scale shear stress becomes large enough and most of the mean velocity mismatch is removed. Simulations for higher Reynolds numbers are carried out with the k–ε model and the one-equation subgrid-scale model. Although it is necessary to improve the turbulence models and the treatment of the buffer region, the new scheme is shown to be effective for reducing the mismatch and to be useful for developing better hybrid simulations. Received 5 April 2002 and accepted 8 January 2003 Published online 25 March 2003 Communicated by M.Y. Hussaini     

Shape optimization of staggered ribs in a rotating equilateral triangular cooling channel

A rotating equilateral triangular cooling channel with staggered square ribs inside the leading edge of a turbine blade has been optimized in this work based on surrogate modeling. The fluid flow and heat transfer in the channel have been analyzed using three-dimensional Reynolds-averaged Navier–Stokes (RANS) equations under uniform heat flux condition. Shear stress transport turbulence model has been used as a turbulence closure. Computational results for area-averaged Nusselt number have been validated compared to the experimental data. The objectives related to the heat transfer rate and pressure drop has been linearly combined with a weighting factor to define the objective function. The angle of the rib, the rib pitch-to-hydraulic diameter ratio, and the rib width-to-hydraulic diameter ratio have been selected as the design variables. Twenty-two design points have been generated by Latin Hypercube sampling, and the values of the objective function have been calculated by the RANS analysis at these points. The surrogate model for the objective function has been constructed using the radial basis neural network method. Through the optimization, the objective function value has been improved by 21.5 % compared to that of the reference geometry.     

Turbulence modelling and role of compressibility on oil spilling from a damaged double hull tank

The viscosity plays an important role, and a multiphase solver is necessary to numerically simulate the oil spilling from a damaged double hull tank (DHT). However, it is uncertain whether turbulence modelling is necessary, which turbulence model is suitable; and what the role of compressibility of the fluids is. This paper presents experimental and numerical investigations to address these issues for various cases representing different scenarios of the oil spilling, including grounding and collision. In the numerical investigations, various approaches to model the turbulence, including the large eddy simulation (LES), direct numerical simulation and the Reynolds average Navier–Stokes equation (RANS) with different turbulence models, are employed. Based on the investigations, it is suggested that the effective Reynolds numbers corresponding to both oil outflow and water inflow shall be considered when classifying the significance of the turbulence and selecting the appropriate turbulence models. This is confirmed by new lab tests considering the axial offset between the internal and the external holes on two hulls of the DHT. The investigations conclude for numerically simulating oil spilling from a damaged DHT that when the effective Re is smaller the RANS approaches should not be used and LES modelling should be employed; while when the effective Reynolds numbers is large, the RANS models may be used as they can give similar results to LES in terms of the height of the mixture in the ballast tank and discharge but costing much less CPU time. The investigation on the role of the compressibility of the fluid reveals that the compressibility of the fluid may be considerable in a small temporal‐spatial scale but plays an insignificant role on macroscopic process of the oil spilling. Copyright © 2016 John Wiley & Sons, Ltd.     

Turbulent transport in an inclined jet in crossflow

The present study experimentally investigates a turbulent jet in crossflow relevant to film cooling applications. The jet is inclined at 30°, and its mean velocity is the same as the crossflow. Magnetic resonance imaging is used to obtain the full three-dimensional velocity and concentration fields, whereas Reynolds stresses are obtained along selected planes by Particle Image Velocimetry. The critical role of the counter-rotating vortex pair in the mixing process is apparent from both velocity and concentration fields. The jet entrainment is not significantly higher than in an axisymmetric jet without crossflow, because the proximity of the wall inhibits the turbulent transport. Reynolds shear stresses correlate with velocity and concentration gradients, consistent with the fundamental assumptions of simple turbulence models. However the eddy viscosity is strongly anisotropic and non-homogeneous, being especially low along the leeward side of the jet close to injection. Turbulent diffusion acts to decouple mean velocity and concentration fields, as demonstrated by the drop in concentration flux within the streamtube issued from the hole. Volume-averaged turbulent diffusivity is calculated using a mass–flux balance across the streamtube emanating from the jet hole, and it is found to vary slowly in the streamwise direction. The data are compared with Reynolds-Averaged Navier–Stokes simulations with standard k − ε closure and an optimal turbulent Schmidt number. The computations underestimate the strength of the counter-rotating vortex pair, due to an overestimated eddy viscosity. On the other hand the entrainment is increasingly underpredicted downstream of injection. To capture the correct macroscopic trends, eddy viscosity and eddy diffusivity should vary spatially in different ways. Therefore a constant turbulent Schmidt number formulation is inadequate for this flow.     

Improvement of double-buffer problem in LES–RANS interface region by introducing an anisotropy-resolving subgrid-scale model

The “double-buffer problem” has been regarded as a crucial concern for the strategy behind the hybrid large eddy simulation (LES)/Reynolds-averaged Navier–Stokes (RANS) model (or HLR model, for short). Such models are likely to show unphysical mean-velocity distributions in the LES–RANS interface region, where “super-streak structures” also appear that look like low-speed streaks generated in the near-wall region of wall turbulence. To overcome this difficulty, the stochastic backscatter model, in which the vortex structures in the interface region are divided into smaller scales, holds promise due to the effect of random source term imposed in the momentum equation. Although this method is effective, several parameters must be prescribed and their specification process is arbitrary and ambiguous. An alternative advanced HLR model has been proposed, in which an anisotropy-resolving subgrid-scale (SGS) model was adopted in the LES region as well as a one-equation nonlinear eddy viscosity model in the RANS region. Previous investigations indicated that this HLR model did not exhibit or, at least, largely reduced the “double-buffer problem” in the mean-velocity distribution, with no special treatment being applied. The main purpose of the present study is to reveal why this HLR model improves the predictive performance in the LES–RANS interface region. Specifically, we focus on the role of the extra anisotropic term introduced in the SGS model, finding that it plays an important role in enhancing vortex structures in the interface region, leading to a considerable improvement in model performance.     

基于基因表达式编程的数据驱动湍流建模

计算流体动力学是湍流研究的重要手段, 其中雷诺平均模拟在航空航天等实际工程中得到了广泛应用. 雷诺平均模拟的结果很大程度上依赖于湍流模型的预测精度, 而实际工程应用中常用的模型往往精度有限. 近年来, 数据驱动的湍流建模方法得到越来越多的关注. 本文介绍了基于基因表达式编程 (gene-expression programming, GEP) 方法的湍流建模相关进展. 本文首先讨论基因表达式编程应用于湍流建模的具体方法, 包括基本算法、显式代数应力模型和湍流传热两种建模框架、模型测试方法以及损失函数设置等. 在此基础上, 基因表达式编程方法被应用于涡轮叶栅尾流混合、竖直平板间自然对流、三维横向流中的射流等问题. 结果表明, GEP可以有效提升常用模型对于尾流混合损失、壁面热通量等关键参数的预测精度. 基因表达式编程方法可以显式给出模型方程, 因此模型具有可解释性强等特点. 基于双向耦合方法得到的模型还被证明具有较好的后验测试精度和鲁棒性. 基因表达式编程方法还被初步应用于大涡模拟亚格子应力和边界层转捩等问题的建模, 在不同湍流建模领域表现出很大的潜力.      

Turbulence modeling for the axially rotating pipe from the viewpoint of analytical closures

A single-point model eddy viscosity model of rotation effects on the turbulent flow in an axially rotating pipe is developed based on two-point closure theories. Rotation is known to impede energy transfer in turbulence; this fact is reflected in the present model through a reduced eddy viscosity, leading to laminarization of the mean velocity profile and return to a laminar friction law in the rapid rotation limit. This model is compared with other proposals including linear redistribution effects through the rapid pressure-strain correlation, Richardson number modification of the eddy viscosity in a model of non-rotating turbulence, and the reduction of turbulence through the suppression of near-wall production mechanisms. PACS 47.27.Eq, 47.32.-y