Session T16: CFD: LES, DNS, Hybrid RANS/LES II

Towards Sustainable Aviation: LES of an Open Fan Blade at Flight Reynolds Number using Exascale Supercomputing

As the aviation industry aims for net zero CO₂ by 2050, a new aircraft engine architecture using an open fan, rather than a ducted fan typically used today, promises a significant reduction in the energy required for flight and hence an associated reduction in fuel consumption and emissions. We present wall-resolved LES of a full 3D open fan blade using the high-order, unstructured solver GENESIS. In contrast to prior work at moderate Re representative of wind tunnel conditions, the present LES is at full-scale flight Re and was performed at the Oak Ridge Leadership Computing Facility’s (OLCF) exascale supercomputer Frontier. The LES provides unique insights into the detailed flow physics, turbulence model shortcomings, as well as the aerodynamic and acoustic performance of the open fan blade at flight Re, years in advance of when full-scale flight testing typically occurs in an engine development program.   

Dynamic slip-based wall-modeled Large-Eddy Simulations of turbulent flows with separation and reattachment in a Discontinuous Galerkin framework

Slip-based wall modeling is a promising wall-modeled large-eddy simulation (WMLES) approach to predict complex turbulent flows (Bose and Moin PoF 2014). Slip-wall model predictions are, however, sensitive to the choices of model coefficient, and the proposed dynamic models are not robust even for attached boundary layers (Bae et al. JFM 2019). In this work, we apply a new dynamic slip-wall model to predict smooth body separation in benchmark test flows subject to favorable and adverse pressure gradients over curved geometries. The new dynamic model is formulated in a consistent Discontinuous Galerkin framework and leverages the universal scaling relationship of Pradhan and Duraisamy JFM 2023. The model dynamically estimates the universal scaling parameter and the slip-wall coefficient is based on an optimal finite-element projection framework. The wall model coupled with the dynamic Smagorinsky model gives predictions for the mean velocities and Reynolds normal and shear stresses that match well with the available experiments and DNS in the reverse flow and post-reattachment regions. Proper length scales and a slip Reynolds number in the model formulation allow for robust model performance on significantly under-resolved LES meshes across a range of high Reynolds numbers.   

Large Eddy Simulation of the flow over a maneuvering 6:1 prolate spheroid

We study the flow around a 6:1 prolate spheroid under two maneuvers - pitch-up and streamwise acceleration. A kinetic energy conserving Large Eddy Simulation (LES) method is used to ensure precise simulation of small-scale turbulence and unsteady modes. Overset grid capabilities are employed to facilitate six degrees of freedom motion to simulate maneuvers. The goal is to develop an understanding of the vortex dynamics and interactions resulting from the 3D cross-flow separation, surface pressure and loading behavior of the spheroid. The work will also explore Reynolds number dependency and effect of maneuver specific parameters such as pitch-up rate and acceleration.   

Towards Immersed Boundary Modeled-LES: a priori Analysis

We present preliminary work on the development of a consistent approach for Immersed Boundary-Modeled LES method designed to incorporate IBs, turbulence closures, and wall-models consistently. To this end, we perform a priori analysis of volume-filtered DNS data of turbulent channel flow at friction Reynolds number of 5200. Volume-filtering differs from regular LES filtering in the treatment of the solid-fluid interface and represents the basis of the Volume-Filtering Immersed Boundary Method. Volume-filtered quantities are found by convolving with a filter kernel pre-multiplied by a phase-indicator function. Applying this procedure to the Navier-Stokes equations leads to volume-filtered equations that depend on the fluid volume fraction and exhibit additional terms representing a residual viscous stress tensor and a bodyforce concentrated on the solid-fluid interface akin to an Immersed Boundary forcing. Further, the resulting volume-filtered velocity has a slip on the solid-fluid interface which also controls the IB forcing term. To inform future modelling attempts, we characterize these unclosed terms using a priori volume-filtered DNS data with filter sizes from 50 to 300 wall units. In particular, we investigate the function dependence of the slip velocity on filter size, as it is needed to model the IB forcing term.   

Strategies for Tractable Turbulent Inflows in High-Fidelity Simulations

Imposing realistic turbulence is a challenging and expensive aspect of high-fidelity simulations. Modern turbulent-inflow methods still require substantial streamwise distances for profiles to fully develop. This computational burden leads to cost-versus-accuracy compromises. The inner-layer relaxes faster than the outer-layer in turbulent boundary layers, leading to many applications foregoing complete outer-layer relaxation. Additionally, commonly used methods require a priori statistics that are seldom available at targeted conditions. These challenges are addressed by the proposed work. Several cost-saving strategies are explored: sponge zones for faster development, wall-modeled inflow regions to alleviate grid requirements, and inflow pressure treatments to damp artificial acoustic contamination. Next, strategies for generating inflow turbulent statistics are explored. The proposed work explores the differences in relaxation distance when using lower-fidelity methods to generate inflow stresses compared to scaling high-fidelity databases to different conditions. This will aid future practitioners in determining the best way to generate inflows when databases are not available at desired Mach or Reynolds numbers.   

Optimal slip wall model for LES in the presence of pressure gradient effects

The slip wall model for large-eddy simulation of turbulent flow is based on rigorous derivation from LES principles, in contrast to traditional equilibrium wall models which are based on RANS principles. It has been demonstrated that slip wall models outperform equilibrium wall models in prediction of turbulent separation, while maintaining superior grid convergence properties.

In practice, modeling the slip coefficient has proven difficult in coarsely resolved high Reynolds number flows; therefore, understanding the scaling behavior of an ideal slip length model is important to inform model development. Prior work observed in equilibrium boundary layers that the ideal slip length followed a consistent scaling behavior across a wide range of friction Reynolds numbers and grid resolutions. The approach is extended to study the behavior of an ideal slip wall model in the presence of pressure gradients and non-equilibrium effects. Additionally, a consistent slip boundary condition for the subgrid-scale stress is derived and is shown to improve skin-friction predictions in favorable pressure gradients. Applications to the Boeing speed bump and the NASA High Lift Common Research Model are presented.   

Assessment of Wall Modeled Large-Eddy Simulation of Large Experimental Facilities

High-fidelity large-eddy simulations of high Reynolds number flows in large wind and water tunnels become infeasible without wall modeling. In these facilities, turbulent boundary layers grow and may be subject to wall curvature. We consider two well-documented test cases of large facilities in this study. First, is the empty test section of the William B. Morgan Large Cavitation Channel. The 72 m long facility includes a unique, high-curvature inlet nozzle after the first turn and a long outlet diffuser separated by a 100 ft long test section with a cross-sectional area of approximately 100 sq ft. The BeVERLI Hill case features a wall-mounted hill configuration in the center of a 12 m long wind tunnel. We employ the curvilinear immersed boundary method with large-eddy simulation to model the turbulence. Due to high curvature in the boundary conditions, the simplifying equilibrium turbulent boundary layer assumption is invalid. We assess and validate several wall models including an equilibrium wall stress model, a volume-averaged wall model, and a non-equilibrium boundary layer wall model. The velocity profiles and energy spectra are compared and errors in the wall models are quantified.   

Logistic Map and Bifurcation Parameters of a New 3-D Kinetic-based Discrete Dynamical System for Sub-grid-scale Modeling

Discrete dynamical systems (DDS) are of significant interest for their ability to model complex, turbulent-like behaviors while maintaining mathematical simplicity. Hydrodynamic-based DDSs have been employed for turbulence modeling. This form of turbulence model has no closure problem and its bifurcation parameters appearing on the sub-grid scales (SGS) have physical interpretations, and their values can be deduced by high-pass filtering of resolved-scale results. In this work, we develop a 3-D kinetic-based DDS using the lattice Boltzmann method, leveraging its capacities to handle complex flow domains and its suitability for GPU parallel computation. Such capacities are critical for solving intricate flow systems such as porous-media flows and cardiovascular blood flows in which pulsation presents. We derive a logistic-like map of the DDS from the lattice Boltzmann equation using the Galerkin procedure. The DDS features two bifurcation parameters: a splitting factor β, which separates the large and small scales, and a rescaling factor θ. We optimize the combination of β and θ using Lagrangian DNS data from five pipe flows, utilizing L¹ and L² norms, skewness, and flatness, with Taylor Microscale Reynolds Numbers ranging from 450 to 750. The parameter set is then verified using an independent DNS case. These results from this study suggest that the DDS has potential applicability in predicting SGS physics relevant to complex flows such as pulsatile flows and turbulence.   

Conjugate Heat Transfer Simulations over ice characterized rough surfaces

Accurate modeling of ice accretion is important for the safe and efficient design of aircraft and propulsion systems. Heat transfer predictions obtained from the fluid flow solvers are used as input in ice accretion codes. In glaze ice conditions, the freezing rates and resulting ice shapes are highly sensitive to the input values of the heat-transfer coefficient. An accurate prediction of heat transfer on iced airfoils is crucial for correctly predicting the ice accretion process. In this work, we present wall-modeled large-eddy simulations of conjugate heat transfer (CHT) over a developing boundary layer on a surface characterized by ice roughness. Results from these simulations show that WMLES with CHT can represent temperature profiles and heat fluxes on surfaces characterized by ice roughness. The smooth to rough transition on the iced surface leads to a high increase in heat-transfer. Distributions of friction coefficient and Stanton number show that the Reynolds Analogy is less accurate with increasing roughness height. For surfaces with low conductivity, overall heat transfer is reduced. Additionally, the temperature distribution becomes more heterogeneous. These results can be used to further improve heat-transfer models used for ice accretion predictions.