ENGINEERING AND CFD

Abstract

Flow noise during takeoff and landing of commercial aircraft can be substantially reduced by the use of porous surface layers in suitable sections of the airfoil. On the other hand (passive) porosity and roughness of surfaces tend to have an adverse effect on the boundary layer and thus on the lift of wings. This motivates the need to be able to predict the aerodynamic effects of porous segments of the wing surface by numerical methods for the aerodynamic design of wings, considering porosity and roughness as a function of Reynolds number. The application of a RANS approach for this task requires additional modeling terms such as the permeability and the Kozeny-Carman parameter as well as the turbulent fluctuations of the velocity field to be determined by DNS/LES simulations which are conducted in this project by the Lattice-Boltzmann method (LBM). For the simulation of the flow at the pore scale, we use advanced LB models developed in our group which show substantially improved performance with respect to stability as well as numerical dispersion properties especially in the case of high Reynolds numbers. Due to the inherent requirements of resolving both the turbulence on the scale of an airfoil and the flow inside the pore-scale resolved porous medium, the resulting simulations require more than a billion grid nodes on a locally refined three-dimensional mesh leading to massively parallel simulations.

Introduction

The motivation to replace the solid trailing edge with a porous material comes from the field of acoustics. Herr showed for a NACA0012-type airfoil [1], that the noise can be reduced to a considerable degree when the trailing edge is made porous. This research was continued, among others, in the Collaborative Research Center SFB880 [2] for an airfoil with asymmetric profile (the so-called DLR-F16). Several acoustic results for this airfoil have been published in [3]. The identical wind tunnel model of the DLR-F16 airfoil was also investigated for its aerodynamic performance in [4].

A key objective of the project is the identification of aeroacoustics effects of the porous trailing edge using resolved pore-scale simulations. The simulations reconstruct the configuration of the wind channel and resolve pores of porous inlay. This is possible with a “model-free” numerical approach in the sense of a global LES with local DNS in the region of the trailing edge.

The DNS/LES aeroacoustics simulations conducted in this project are intended to serve both as a basis for a fundamental understanding of aeroacoustics and aerodynamic effects of anisotropic and/or graded porous material used for the trailing edge as well as to generate calibration data for Large scale (airplane) (V)RANS modeling that in turn should contribute to a substantial reduction of noise during takeoff and landing.

Simulation of DLR-F16 airfoil with porous trailing edge

For the simulation of DLR-F16 airfoil with the porous trailing edge we used a weakly compressible cumulant lattice Boltzmann model [5] that provides high numerical stability and very low numerical diffusion and dispersion and therefore is particularly suitable for the simulation of turbulent flows. The model is implemented in the massive parallel framework VirtualFluids [6]. The software shows about 80% parallel efficiency using up to 60,000 CPU cores.

The simulation was carried out for a Reynolds number Re=1x10⁶ and Mach number Ma=0.15. These parameters were used in a wind tunnel experiment by Mößner et al. [4] and are relevant for commercial aircraft. Our simulation reproduced wind tunnel conditions with a length of 2.1 m and a height of 1.3m. In the spanwise direction, we used a periodic boundary condition. The simulation used a cartesian grid with a local refinement around the airfoil and in a porous trailing edge. The finest level of the grid had a resolution of 23 μm to resolve pores of the trailing edge. The grid consisted of 2.7 billion grid points (corresponds to approx. 73 billion degrees of freedom). The simulation needed about 1.4 Terabyte RAM and used 16,384 CPU cores of supercomputer SuperMUC at LRZ.

The simulation provides detailed information about flow above and inside a porous trailing edge.