Detached-Eddy Simulation of a Coflow Jet Airfoil at High Angle of Attack

Abstract

TO IMPROVE aircraft performance, innovative technology should be pursued to drastically reduce the weight and fuel consumption of the aircraft and to significantly increase aircraft mission payload and stall margin. Both military and commercial aircraft will benefit from the same technology. Flow control is the most promising route to bring significant performance improvement to aircraft [1–7]. Recently, Zha et al. [8–10] andWang et al. [11] have developed a promising airfoil flow control technique using a coflow jet, which significantly increases lift, stall margin, and drag reduction. The coflow jet airfoil is designedwith an injection slot near leading edge and a suction slot near trailing edge on the airfoil suction surface. The slots are opened by translating a large portion of the suction surface downward. A high-energy jet is injected tangentially near the leading edge and the same amount of mass flow is drawn in near the trailing edge. The turbulent shear layer between the main flow and the jet causes strong turbulence diffusion and mixing under the severe adverse pressure gradient, which enhances lateral transport of energy from the jet to the main flow and allows the main flow to overcome the severe adverse pressure gradient and remain attached at high angles of attack (AOA). The high-energy jet induces high circulation and hence generates high lift. The energized main flow fills the wake and therefore reduces drag. The coflow jet (CFJ) airfoil achieves net-zero mass-flux flow control and can significantly reduce the penalty to the propulsion system bydumping zero jetmass flow. In [11], the CFJ airfoils with three different slot sizes are computed using Spalart–Allmaras (S-A) model. It is observed that at lowAOA, the computed results agree well with the experiment. But at high AOA, the deviation of the lift and drag is large because a Reynoldsaveraged Navier–Stokes (RANS) model is not able to simulate the flow separation accurately. In particular, the predicted stall AOA by RANS is significantly smaller than the measured one. Recently, in [12], the DES predicts the stalled airfoil flow at high AOA significantly better than the RANS model. To overcome the disadvantages of RANSmodels and to avoid the intensive CPU requirement for LES, Spalart et al. developed the detached-eddy simulation (DES) strategy [13], which is a hybrid RANS and LES method. Near the solid surface within the wall boundary layer, the unsteady RANSmodel is realized in DES. Away from thewall surface, themodel automatically converts to large-eddy simulation (LES). By using the RANSmodel near thewall, the mesh size as well as the CPU time can be tremendously reduced. By using the LESmodel away from thewall, the regions of massive separation and other free shear flows are properly simulated. Its application for turbulence simulation has already achieved encouraging success, as shown in thework of Tarvin et al. [14], Spalart [15,16], Forsythe et al. [17], Viswanathan et al. [18], Squires et al. [19,20], Hansen and Forsythe [21], Subbareddy and Candler [22],Wang and Zha [23,24], and Im and Zha [12]. These flows using DES include those for airfoils, cylinders, forebodies, baseflows, etc. The results are qualitatively and quantitatively better than the solutions using RANS. The objective of this paper is to simulate the flowfield of a CFJ airfoil at high AOA using DES in order to better capture the jet mixing and flow separation. A low-diffusion E-CUSP scheme [25,26] with fifth-order weighted essentially nonoscillatory (WENO) scheme [27,28] is employed for the inviscid fluxes. The computed lifts atAOA 30, 37, and 39 agree better with the experiment than those using S-A one-equation RANS. The computed drag usingDES is also improved greatly, even though the deviation remains very large. The computed stall AOA using DES is 37 , which agrees very well with the experiment and is much better than the one using S-A one-equation RANS. The computed results using DES show significantly improvements at high AOA compared with those using S-A one-equation RANS.