Accuracy Of Large Eddy Simulation Research Articles

Flowpattern in hydrocyclones, numerical simulations with experimental verification

This paper reports a detailed study of the flow in cyclone separators, with the use of most up to date computational fluid dynamics simulations, which are validated with positron emission particle tracking (PEPT) experiments tracing the movement of particles through the cyclone. The parameters varied were the viscosity of the carrier liquid, the flowrate and, in the numerical simulations, the inlet configurations of the cyclone, namely one and two inlets and, with the two inlets, a) both at right angles to the cyclone axis and b) angled downwards. The study reveals features of the flow, which have not been seen till now, but are necessary for the understanding and modelling of the separation and purification efficiency of cyclones. The results of the simulations and the close agreement with experiment are a testament to the reliability and accuracy of large eddy simulation (LES), even for flow features as difficult to simulate as the confined strongly swirling flows in cyclone separators. The results show that a contiguous, smooth surface of zero axial velocity does exist and has approximately the shape that has been assumed by modellers. The significant effects of fluid viscosity, underflow and modifications to the inlet are also shown.

A priori test of Large-Eddy Simulation models for the Sub-Grid Scale turbulent stress tensor in perfect and transcritical compressible real gas Homogeneous Isotropic Turbulence

In this study, the focus lies on the modeling of the subgrid-scale stress tensor in the context of real gas flows. By conducting tests on perfect and dense gas configurations using six different models, the research findings highlight the superiority of gradient models over eddy-viscosity models in capturing the structural behavior of the stress tensor. However, the models’ performance experiences a significant decline when considering the intricate thermodynamics of real gases. To address this challenge, dynamic models that take into account the characteristics of the flow are also tested, resulting in some improvements in the outcomes. Despite these advancements, the achieved results still fall short of being entirely satisfactory. In particular, the models fail to accurately predict the occurrence of large values of the subgrid-scale stress tensor in the real gas flow. The complex thermodynamic nature of real gases therefore significantly influences the subgrid-scale stress tensor and the outcomes of this study underscore the urgent need for continued research to advance our understanding and modeling capabilities of subgrid-scale terms in real gas flows, paving the way for statistically accurate Large-Eddy Simulations of industrial real gas flows using moderately refined grids.

On the robustness and accuracy of large-eddy simulation in predicting complex internal flow of a gas-turbine combustor

The objective of this study is to evaluate the effects of numerical and model setups on the large-eddy simulation (LES) predictive capability for the internal flow of a propulsion-relevant configuration. The specific focus is placed on assessing the LES technique with lower mesh resolutions, which is of technological relevance to practical industrial design. A set of Riemann flux formulations and commonly used subgrid-scale models are considered in this work to produce a hierarchy of LES setups with different dissipation effects (both numerically and physically). The LES results obtained from different setups are compared qualitatively in terms of the key flow characteristics and evaluated quantitatively against the experimental measurements. The error landscape is generated to reveal the predictive qualities of different LES setups. The study shows that the choice of numerical flux formulation plays a prominent role in governing the general flow patterns, while the effect of subgrid-scale model is mainly manifested in transient flow characteristics, such as vortex breakdown and swirl-induced vortical structures. Based on the error analysis, it is found that lower dissipative LES setup is not always beneficial to the LES accuracy. This is in contrast to the commonly accepted understanding in literature for the LES, which was established solely with canonical flow configurations.

Deep learning closure models for large-eddy simulation of flows around bluff bodies

Near-wall flow simulation remains a central challenge in aerodynamics modelling: Reynolds-averaged Navier–Stokes predictions of separated flows are often inaccurate, and large-eddy simulation (LES) can require prohibitively small near-wall mesh sizes. A deep learning (DL) closure model for LES is developed by introducing untrained neural networks into the governing equations and training in situ for incompressible flows around rectangular prisms at moderate Reynolds numbers. The DL-LES models are trained using adjoint partial differential equation (PDE) optimization methods to match, as closely as possible, direct numerical simulation (DNS) data. They are then evaluated out-of-sample – for aspect ratios, Reynolds numbers and bluff-body geometries not included in the training data – and compared with standard LES models. The DL-LES models outperform these models and are able to achieve accurate LES predictions on a relatively coarse mesh (downsampled from the DNS mesh by factors of four or eight in each Cartesian direction). We study the accuracy of the DL-LES model for predicting the drag coefficient, near-wall and far-field mean flow, and resolved Reynolds stress. A crucial challenge is that the LES quantities of interest are the steady-state flow statistics; for example, a time-averaged velocity component $\langle {u}_i\rangle (x) = \lim _{t \rightarrow \infty } ({1}/{t}) \int _0^t u_i(s,x)\, {\rm d}s$ . Calculating the steady-state flow statistics therefore requires simulating the DL-LES equations over a large number of flow times through the domain. It is a non-trivial question whether an unsteady PDE model with a functional form defined by a deep neural network can remain stable and accurate on $t \in [0, \infty )$ , especially when trained over comparatively short time intervals. Our results demonstrate that the DL-LES models are accurate and stable over long time horizons, which enables the estimation of the steady-state mean velocity, fluctuations and drag coefficient of turbulent flows around bluff bodies relevant to aerodynamics applications.

Neural networks for large eddy simulations of wall-bounded turbulence: numerical experiments and challenges.

We examine the application of neural network-based methods to improve the accuracy of large eddy simulations of incompressible turbulent flows. The networks are trained to learn a mapping between flow features and the subgrid scales, and applied locally and instantaneously-in the same way as traditional physics-based subgrid closures. Models that use only the local resolved strain rate are poorly correlated with the actual subgrid forces obtained from filtering direct numerical simulation data. We see that highly accurate models in a priori testing are inaccurate in forward calculations, owing to the preponderance of numerical errors in implicitly filtered large eddy simulations. A network that accounts for the discretization errors is trained and found to be unstable in a posteriori testing. We identify a number of challenges that the approach faces, including a distribution shift that affects networks that fail to account for numerical errors.

Toward the use of LES for industrial complex geometries. Part I: automatic mesh definition

With the constant increase of computational power for the past years, Computational Fluid Dynamics (CFD) has become an essential part of the design in complex industrial processes. In this context, among the scale resolving numerical methods, Large-Eddy Simulation (LES) has become a valuable tool for the simulation of complex unsteady flows. To generalise the industrial use of LES, two main limitations are identified. First, the generation of a proper mesh can be a difficult task, which often relies on user-experience. Secondly, the ‘time-to-solution’ associated with the LES approach can be prohibitive in an industrial context. In this work, these two challenges are addressed in two parts. In this Part I, an automatic procedure for mesh definition is proposed, whereas the Part II is devoted to numerical technique to reduce the LES ‘time-to-solution’. The main goal of these works is then to develop an accurate LES strategy at an optimised computational cost. Concerning the mesh definition, because LES is based on separation between resolved and modelled subgrid-scales, the quality of the computed solution is then directly linked to the quality of the mesh. However, the definition of an adequate mesh is still an issue when LES is used to predict the flow in an industrial complex geometry without a priori knowledge of the flow dynamics. This first part presents a user-independent approach for both the generation of an initial mesh and the convergence of the mesh in the LES framework. An automatic mesh convergence strategy is proposed to ensure LES accuracy. This strategy is built to guarantee a mesh-independent mean field kinetic energy budget. The mean field kinetic energy is indeed expected to be mesh independent since only turbulent scales should be unresolved in LES. The approach is validated on canonical cases, a turbulent round jet and a turbulent pipe flow. Finally, the PRECCINSTA swirl burner is considered as a representative case of complex geometry. First, an algorithm for the generation of an unstructured mesh from a STL file is proposed to generate a coarse initial mesh, before applying the mesh convergence procedure. The overall strategy including automatic first mesh generation and its automatic adaptation paves the way to use LES approach as a decision support tool for various applications, provided that the ‘time-to-solution’ is compatible with the applications constraint. A second paper, referred as Part II, is devoted to the reduction of this time.

Focus on software, data acquisition, and instrumentation

Focus on software, data acquisition, and instrumentation

The effect of filter anisotropy on the large eddy simulation of turbulence

We study the effect of filter anisotropy and sub-filter scale (SFS) dynamics on the accuracy of large eddy simulation (LES) of turbulence, by using several types of SFS models including the dynamic Smagorinsky model (DSM), dynamic mixed model (DMM), and the direct deconvolution model (DDM) with the anisotropic filter. The aspect ratios (AR) of the filters for LES range from 1 to 16. We show that the DDM is capable of predicting SFS stresses accurately at highly anisotropic filter. In the a priori study, the correlation coefficients of SFS stress reconstructed by the DDM are over 90%, which are much larger than those of the DSM and DMM models. The correlation coefficients decrease as the AR increases. In the a posteriori studies, the DDM outperforms DSM and DMM models in the prediction of various turbulence statistics, including the velocity spectra, and probability density functions of the vorticity, SFS energy flux, velocity increments, strain-rate tensors and SFS stress. As the anisotropy increases, the results of DSM and DMM become worse, but DDM can give satisfactory results for all the filter-anisotropy cases. These results indicate that the DDM framework is a promising tool in developing advanced SFS models in the LES of turbulence in the presence of anisotropic filter.

Deep learning method for the super-resolution reconstruction of small-scale motions in large-eddy simulation

A super-resolution reconstruction model for the subgrid scale (SGS) turbulent flow field in large-eddy simulation (LES) is proposed, and it is called the meta-learning deep convolutional neural network (MLDCNN). Direct numerical simulation (DNS) data of isotropic turbulence are used as the dataset of the model. The MLDCNN is an unsupervised learning model, which only includes high-resolution DNS data without manually inputting preprocessed low-resolution data. In this model, the training process adopts the meta-learning method. First, in the a priori test, the SGS turbulent flow motions in the filtered DNS (FDNS) flow field are reconstructed, and the energy spectrum and probability density function of the velocity gradient of the DNS flow field are reconstructed with high accuracy. Then, in the a posteriori test, the super-resolution reconstruction of the LES flow field is carried out. The difficulty of LES flow field reconstruction is that it contains filtering loss and subgrid model errors relative to the DNS flow field. The super-resolution reconstruction of the LES flow field achieves good results through this unsupervised learning model. The proposed model makes a good prediction of small-scale motions in the LES flow field. This work improves the prediction accuracy of LES, which is crucial for the phenomena dominated by small-scale motions, such as relative motions of particles suspended in turbulent flows.

CFD evaluation of mean and turbulent wind characteristics around a high-rise building affected by its surroundings

This study comprehensively investigated the accuracy of computational fluid dynamics (CFD) simulations for predicting the mean and turbulent wind characteristics around a high-rise building surrounded by low-rise street canyons. The flow field was characterized based on the complicated interaction between a high-rise building and its surrounding buildings, which has rarely been examined in previous CFD studies. We obtained the CFD results using steady Reynolds averaged Navier-Stokes equation (RANS) models and large eddy simulations (LES), which were compared with the measurement results obtained via wind tunnel experiments. Furthermore, we quantitatively investigated the impact of the presence of a high-rise building on the prediction accuracy of pedestrian wind environments, including the gust wind velocity. The results showed that the performance of both LES and RANS decreased in the presence of a high-rise building, indicating that predicting the flow around a high-rise building is more challenging than predicting the flow of a simple street canyon. Additionally, the prediction accuracies of RANS and LES for the mean wind velocity at the pedestrian level were similar in regions with high wind velocity. However, the prediction accuracy of LES in the medium- and low-wind-velocity regions was significantly better than that of RANS, which is closely related to the superior prediction accuracy of turbulent kinetic energy in LES. Although LES can predict the gust wind velocity directly with higher accuracy than that of RANS by using a gust factor, RANS can be used to predict the gust wind velocity with a certain accuracy using a peak factor.

A novel hybrid control strategy of wind turbine wakes in tandem configuration to improve power production

The wind turbines that operate in the wake region of the upstream turbines produce less power and may suffer serious structural issues due to highly unsteady flows, which can reduce the life expectancy of the turbines. In this study, a novel hybrid wake control strategy for the wind farm power generation enhancement is proposed, which is based on highly accurate large eddy simulations coupled with the actuator line method. The combined effects of yaw angle and tilt angle control methods on the performance improvement of downstream wind turbines in wind farm layouts have not yet been investigated. This work would be the first attempt to evaluate the hybrid wake-control strategy of tandem wind turbines. It is found that the vortex generation is stronger at lower tilt angles because more parts of the rotor are affected by the wakes of the upstream wind turbine. An optimisation analysis is also provided to find the optimum wake deflection angles of the upstream turbine to maximize the electrical power generation. The results show that an accumulative power production increment of 17.1% is achieved by controlling both yaw angle (θ=30o), and tilt angle (φ=24o). The power obtained in the present study is approximately 6.1% higher than previous wake control techniques. By using the hybrid control strategy, an annual energy production enhancement of 3.7% is achieved, which is higher than the previous wake-controlled wind farm layouts.

Computational Analysis of Actuation Techniques Impact on the Flow Control around the Ahmed Body

Active flow control with jet devices is a promising approach for vehicle aerodynamics control. In this work an extended computational study is performed comparing three different actuation strategies for active flow control around the square back Ahmed body at Reynolds number 500,000 (based on the vehicle height). Numerical simulations are run using a Large Eddy Simulation (LES) approach, well adapted to calculate the unsteady high Reynolds number flow control using periodic jet devices. computations are validated comparing to in-house experiments for uncontrolled and some controlled cases. The novelty of this investigation is mainly related to the in-depth study of the base flow and actuation approaches by an accurate LES method and their comparison to experiments. Here, several simulations are performed to estimate the effect of active controls on the flow topology and the drag reduction. Beside the continuous blowing jet, three periodic actuation techniques including periodic blowing and suction as well as the zero flux synthetic jet devices are explored. The slots are implemented discontinuously in order to achieve a better control efficiency linked to vortex generation. In this framework, spectral analyses on global aerodynamical quantities, rear pressure/drag coefficient behavior examination as well as wake structure investigations are performed in order to compare these jet actuations. As a result, shear layer variations are observed during the blowing phase, but the main flow topology change occurs with suction and synthetic jets. Rear back pressure is therefore substantially increased.

A dynamic spatial gradient model for the subgrid closure in large-eddy simulation of turbulence

A dynamic spatial gradient model (DSGM) is proposed for the subgrid-scale (SGS) closure of large-eddy simulation (LES). The velocity gradients at neighboring LES grids are incorporated to improve the accuracy of the SGS stress. Compared to the previous machine-learning-based multi-point gradient models, the current model is free from the need of a priori knowledge. The model coefficients are dynamically determined by the least-square method using the Leonard stress. The a priori tests show that the correlation coefficients of the SGS stress for the DSGM framework are much larger than the traditional velocity gradient model over different tested filter widths from viscous to inertial scales. The analysis of the model coefficients in the a priori test suggests that the number of the model coefficients can be significantly reduced, leading to a simpler version of the model. A small-scale eddy viscosity (SSEV) model is introduced as an artificial viscosity to mimic the flux of kinetic energy to smaller scales which cannot be resolved at an LES grid. The velocity spectrum predicted by SSEV-based implicit LES is very close to that of direct numerical simulation (DNS) data. In the a posteriori tests, both the flow statistics and the instantaneous field are accurately recovered with the SSEV-enhanced DSGM model. Compared with the SSEV-based implicit LES, the dynamic Smagorinsky model, and the dynamic mixed model, the results predicted by the current model have overall closer agreements with the filtered DNS result, suggesting that the DSGM framework is well-suited for highly accurate LES of turbulence.

Adjoint shape optimization coupled with LES-adapted RANS of a U-bend duct for pressure loss reduction

Nowadays, as industrial designs are close to their optimal configurations, the challenge lies in the extraction of the last percentages of improvement. This necessitates accurate evaluations of the performance and represents a significant higher computational cost. The present work aims at integrating Large Eddy Simulations in the optimization framework for an accurate evaluation of the flow field. The number of expensive evaluations is kept to a minimum by using the adjoint method for the evaluation of the gradient of the objective function. Divergence of the gradients due to the chaotic flow motion is avoided by an additional step which decouples the Large Eddy Simulations from the gradient calculations. An adaptation process based on a Reynolds Averaged Navier-Stokes simulation is therefore sought to mimic the more accurate Large Eddy Simulation results. The obtained field is then used in combination with an adjoint shape optimization routine. The method is tested on the design of a U-bend for internal cooling channels by minimizing its pressure loss. Starting from an optimized geometry obtained through a classical approach based on RANS evaluations, further improvements of the design are achieved with the application of the proposed strategy when performances are evaluated by means of LES.

Industrial scale Large Eddy Simulations with adaptive octree meshes using immersogeometric analysis

We present a variant of the immersed boundary method integrated with octree meshes for highly efficient and accurate Large Eddy Simulations (LES) of flows around complex geometries. We demonstrate the scalability of the proposed method up to O(32K) processors. This is achieved by (a) rapid in-out tests; (b) adaptive quadrature for an accurate evaluation of forces; (c) tensorized evaluation during matrix assembly. We showcase this method on two non-trivial applications: accurately computing the drag coefficient of a sphere across Reynolds numbers 1−106 encompassing the drag crisis regime; simulating flow features across a semi-truck for investigating the effect of platooning on efficiency.

A wall-resolved large-eddy simulation of deep cavity flow in acoustic resonance

Abstract

A Statistical Evaluation of WRF-LES Trace Gas Dispersion Using Project Prairie Grass Measurements

AbstractIn recent years, new measurement systems have been deployed to monitor and quantify methane emissions from the natural gas sector. Large-eddy simulation (LES) has complemented measurement campaigns by serving as a controlled environment in which to study plume dynamics and sampling strategies. However, with few comparisons to controlled-release experiments, the accuracy of LES for modeling natural gas emissions is poorly characterized. In this paper, we evaluate LES from the Weather Research and Forecasting (WRF) model against Project Prairie Grass campaign measurements and surface layer similarity theory. Using WRF-LES, we simulate continuous emissions from 30 near-surface trace gas sources in two stability regimes: strong and weak convection. We examine the impact of grid resolutions ranging from 6.25 m to 52 m in the horizontal dimension on model results. We evaluate performance in a statistical framework, calculating fractional bias and conducting Welch’s t-tests. WRF-LES accurately simulates observed surface concentrations at 100 m and beyond under strong convection; simulated concentrations pass t-tests in this region irrespective of grid resolution. However, in weakly convective conditions with strong winds, WRF-LES substantially overpredicts concentrations – the magnitude of fractional bias often exceeds 30%, and all but one C-test fails. The good performance of WRF-LES under strong convection correlates with agreement with local free convection theory and a minimal amount of parameterized turbulent kinetic energy. The poor performance under weak convection corresponds to misalignment with Monin-Obukhov similarity theory and a significant amount of parameterized turbulent kinetic energy.

Large-eddy simulation of pollutant dispersion in generic urban street canyons: Guidelines for domain size

Pollutant dispersion in urban street canyons has been widely investigated by large-eddy simulation (LES). Many LES studies focused on generic street canyons under a wind direction perpendicular to the street axis. Accurate LES simulations require a sufficiently large domain size to minimize the effects of the artificial boundary conditions at the domain faces on the results. As opposed to RANS simulations, there is a lack of guidelines for an appropriate domain size for LES simulations of wind flow and pollutant dispersion in street canyons. The present study systematically investigates the effect of the domain width, domain height and upstream and downstream domain lengths on the wind flow and pollutant dispersion within a generic 2.5D street canyon with spanwise periodic boundary conditions. Following a validation study, 16 LES simulations are performed for different domain sizes. The results show that the minimum requirement for the domain width is 2.5H, where H is the roof height of the street canyon. For the domain height, upstream domain length and downstream domain length, 7.5H, 5H and 10H are recommended, respectively. These guidelines should help to reduce the computational costs of this type of simulation without significantly compromising the accuracy.

Role of Inflow Turbulence and Surrounding Buildings on Large Eddy Simulations of Urban Wind Energy

Predicting flow patterns that develop on the roof of high-rise buildings is critical for the development of urban wind energy. In particular, the performance and reliability of devices largely depends on the positioning strategy, a major unresolved challenge. This work aims at investigating the effect of variations in the turbulent inflow and the geometric model on the flow patterns that develop on the roof of tall buildings in the realistic configuration of the University of Birmingham’s campus in the United Kingdom (UK). Results confirm that the accuracy of Large Eddy Simulation (LES) predictions is only marginally affected by differences in the inflow mean wind speed and turbulence intensity, provided that turbulence is not absent. The effect of the presence of surrounding buildings is also investigated and found to be marginal to the results if the inflow is turbulent. The integral length scale is the parameter most affected by the turbulence characteristics of the inflow, while gustiness is only marginally influenced. This work will contribute to LES applications on the urban wind resource and their computational setup simplification.

Some notes on numerical simulation of the turbulent cavitating flow with a dynamic cubic nonlinear sub-grid scale model in OpenFOAM

The accuracy of large eddy simulation (LES) is highly dependent on the performance of sub-grid scale (SGS) model. In the present paper, a dynamic cubic nonlinear sub-grid scale model (DCNM) proposed by Huang et al. is implemented for the simulation of unsteady cavitating flow around a 3-D Clark-Y hydrofoil in OpenFOAM. Its performance in predicting the evolution of cloud cavitation is discussed in detail. The simulation with a linear model, the dynamic Smagorinsky model (DSM), is also conducted as a comparison. The results with DCNM show a better agreement with the available experimental observation. The comparison between DCNM and DSM further suggests that the DCNM is able to predict the backscatter more precisely, which is an important feature in LES. The characteristics of DCNM is analyzed to account for its advantages in the prediction of unsteady cloud cavitation as well. The results reveal that it is the nonlinear terms of DCNM that makes DCNM capture sub-grid scale vortices better and more suitable for studying the transient behaviors of cloud cavitation than DSM.