Application of LES Methods to Military Aircraft Flow Problems

Abstract

An overview of the status of application of large eddy simulation methods in the military aircraft industry is presented. The use of large eddy simulation (LES) methods is expanding in the industry. A major challenge for LES application is adequate mesh resolution of attached wall boundary layers. The industrial environment demands large numbers of simulations over complex configurations for design, and for configurations requiring accurate attached boundary layer simulation, LES is rarely practical today. Hybrid Reynolds averaged Navier-Stokes (RANS)/LES methods help to address the constraints on near wall resolution, but bring new constraints and challenges to the fore. LES is currently used primarily for applications where unsteady flow data is required and attached boundary layer simulation is either not required or is not a major driver to solution accuracy. Applications where LES is used successfully in industry include the generation of unsteady loads, aero-optics, active flow control and acoustics. Examples of application of LES to these flows are provided. I. Introduction Computational fluid dynamics (CFD) methods are used to simulate a wide variety of flow phenomena in the military aircraft industry. A large portion of the applications are for analysis, design and optimization of aircraft at or near cruise conditions. For these applications, flow over the majority of the vehicle is attached, and regions of separation are limited in extent. Reynolds averaged Navier-Stokes (RANS) turbulence models are efficient and provide sufficient accuracy to meet most program needs for these applications. However, CFD is also increasingly being called upon to simulate complex flow phenomena with large scale separations and high levels of unsteadiness where RANS based turbulence models are not particularly accurate. In addition, for many of these flows of interest, unsteady flowfield information is required. Unsteady RANS is frequently of limited usefulness in these regimes because the turbulent viscosity damps a wide range of the smaller turbulent scales, and the remaining scales of unsteadiness tend to be periodic and very large scale. Frequency spectra show that the energy is concentrated in a few narrow bands, and RMS levels of the fluctuations are typically poorly predicted. Large eddy simulation (LES) methods hold significant promise for meeting the need for improved accuracy of massively separated flows and for generating accurate time resolved flowfield information. Because the larger scales of turbulence are directly simulated, the sub grid scale turbulence is more isotropic, and therefore simpler to model accurately. Of course the drawback to LES is the significant computational resources required to generate solutions. While steady state RANS flow solvers employ a variety of techniques to accelerate convergence, LES flow solvers must accurately resolve the transient flow field over closely spaced discrete time increments defined by the physical scales and convection speeds of the smallest resolved eddies in the flow field. In addition, the duration of the simulation must be much longer than a RANS simulation in order to capture a sufficient duration of the flow variables to give meaningful mean flowfield and turbulent statistics. Finally, the grid resolution requirements for LES are much more demanding than for RANS, particularly near the wall for attached boundary layers. Near the wall in a turbulent boundary layer there is not a broad range of turbulent scales, and accurate simulation requires mesh resolution in the normal, span-wise and stream-wise directions approaching the resolution required for direct numerical simulation. Since LES requires much finer computational meshes and many more computational iterations than does RANS simulations, the computer time required for a large eddy simulation is many orders of magnitude greater than the time required for a RANS simulation.