LES for Supersonic Ramp Flow Control Using MVG at M_infty=3 and Re_theta=2400

qin li,chaoqun liu

2014-01-01

Abstract:In order to obtain the supersonic intake flow with the least total pressure losses caused by the shock-induced or –boundary-layer-interacted separation, flow controls are recently intensively studied by experiments [1] and numerical simulations [2] using control elements with the sub-boundary layer length scale in normal direction. These elements differ from the conventional ones, whose height are of the level of the boundary layer width. The micro vortex generator (MVG), which is one of representative methods in supersonic flow controls, has been used to control the separation zone caused by incident shock interacts with the turbulent boundary layer of the flat plate [1,2]. The supersonic ramp with turbulent boundary layer inflow is one of the prototype geometries which are usually found in supersonic inlet. The complex separation flow at the corner will cause poor total pressure recovery and finally result in performance losses of the engine. This work presents a numerical research on using the MVG with 70o declining angle of the rear edge, to control the separation of the supersonic ramp flow. The objective is to test the effects of the control by using MVG and investigate the mechanism. In order to make a detailed research on the problem, the methodology of so-called Monotone Integrated Large Eddy Simulation (MI-LES) is adopted. The main part of algorithm is to use a kind of high order, low dissipation scheme to simulate the large eddies of the flow, while the discontinuous structures like shock waves should be captured essentially free of oscillations. In this study, a fourth-order bandwidth-optimized WENO-type scheme was used. The dispersive and dissipative characteristic of the scheme is compared with the 5 th order WENO scheme in figure 1. The results show that although the WENO scheme displays some improvement on dispersive characteristics, the presented 4 th order scheme has distinct lower dissipation, which is often more favorable in LES. a) dispersive characteristics b) dissipative characteristics Figure 1 the comparison of the spectrum of the schemes The benchmark test was performed on inviscid supersonic cylinder flow at M∞=3, whose results are shown as follows: Figure 2 the comparison about the pressure distribution Figure 3 the comparison about the shock location Figure 4 the convergence history Numerical results are compared with the incremental solution by Prof. Rusunov in figure 2 and 3, which is usually thought as the “theoretical” solution of the problem. In order to generate a desirable turbulent boundary inflow, the Rescaling-Recycling Method (RRM) will be used. The basic idea from Lund is to rescale the downstream profile and reintroduce into the upstream inlet. According to the method, the boundary is divided into two layers: the inner layer and the outer layer. In each layer the profile is decomposed into the mean part and fluctuating part. Different scaling laws are satisfied in different parts. In this work, the extended version of the RRM by Xu & Martin [3] will be adopted. Using the method, the satisfactory turbulent boundary layer inflow can be obtained with the given boundary layer width and surface friction coefficient. To simulate the boundary flow, a kind of open boundary condition based on characteristics is used on out-boundary and upper-boundary, combining with the standard sponge layer technique [4] . And the adiabatic condition is used on body surface. Grid system is an important and difficult task in numerical simulation. For the ramp grid, the analytical approach by Adams [4] was used, in which smooth and clustering techniques are adopted on the ramp corner and grid concentration control is used in boundary layer, as show in figure 5. Preliminary simulations are made using grids nstreamwise×nnormal×nspanwise=420×121×64, the final grid number will be around 2048x256×128. a) the global view b) the local view Figure 5 the grid system of the ramp flow In order to describe the shape MVG accurately as possible, the body-fitted grid type is chosen to generate the grid system of the MVG with 70o declining angle of the rear edge. The careful design had been done and large amounts of effects had been spent on the surface and spatial grid generation. Finally, separately generated grids are assembled together to form the MVG-Ramp grids. The sketch of the carefully designed grid system can be seen in figure 6. a) surface grids of the MVG b) the global view c) the local grid intersection Figure 6 the grid system of the half MVG-Ramp configuration Based on the above grids, a grid-refinement procedure was performed using the developed three-dimensional (3-D) grid solver, to make the grids smoother, more uniform and orthogonal to the body surface. Such procedures are found to be very important in 3-D computation from practice. The main flow conditions for MVG-Ramp flows are M∞=3 and Reθ=2400. Using the MPI technique, the parallel computation is conducted on 128/256 computing nodes in the Texas Advanced Computing Center (TACC). For comparison, the ramp flow without the MVG was simulated first. The following investigations are made and corresponding results are obtained: 1) The separating pattern of the flow in sense of mean flow, as shown in figure 7. The multiple levels of spanwise separations are found to evolve slowly with the time, and the streamwise vortices caused by the Görtler instability are found also; Figure 6 wave structures of the mean flow Figure 7 the separation pattern of the mean flow 2) The complex wave structures. The clear images are gained about the main shock, the separation shock, over-expansion jets, the compression wave fans, and their interactions; 3) The large scale shock movement. The frequency information of the shock motion is recorded and analysed using the Fourier transformation. Preliminary analyses are carried out on the connection between the unsteady shock movement and the flow; 4) The quantitative turbulent flow analyses. The typical turbulent boundary parameters are calculated like the boundary thickness, shape factor, total pressure distribution, the friction force coefficient, the turbulent Mach number, and the Reynolds stresses, etc, and comparisons are made with results from the available references. When the MVG is placed ahead the ramp, the separating flow of the ramp are obviously influenced by the induced streamwise vortices. The following investigations are made and corresponding preliminary results are obtained: 1) Flow structures. The basic shock wave, expansive wave system can be seen in figure 8 using the half part of the field datum. Being consistent with the results in experiments by Barbinsky [1], two main counter-rotating streamwise vortices are discovered, as shown in figure 9 The wake-like flow corresponding to the low-momentum deficit is found also to be placed between and above the two main vortices. The two dominated vortices reside inside the boundary for a long distance, while the height of wake location is of that of the boundary layer and lift slowly when moving downstream. a) spatial streamlines b) cross-section streamlines Figure 8 the isobar contour on the symmetric plane and the clouds-map of the pressure on the half part of the body surface Figure 9 the vortex structures generated by the MVG 2) Due to the impact of streamwise vortices, the separating bubble in the ramp corner is suppressed and seems to be broken up into cells. The topology is largely different from the uncontrolled case. 3) Quantitative comparisons will be made between the uncontrolled ramp flows to the controlled ones by MVG. At certain downstream cross section, the various boundary layer width and shape factor are calculated. Total pressure distributions are computed and compared also along the same grid lines at different spanwise locations. The results show that some “healthy” tendency of boundary layer of the controlled flow appears. The current computational results are very preliminary with coarse grids. A more complete LES results with carefully designed fine grids of 2048x256x128 will be reported and analyzed in the final paper. Through the numerical simulation, it can be seen that MVG pretends to be a hopeful way in supersonic ramp flow control to suppress the separation region and improve the quality of boundary layer. Acknowledgements This work is supported by AFOSR grant FA9550-08-1-0201 supervised by Dr. John Schmisseur. The authors are grateful to Texas Advanced Computing Center (TACC) for providing computation hours. Reference 1 H. Barbinsky, Microramp Control of Supersonic Oblique Shock-Wave/Boundary-Layer Interactions, AIAA J., Vol. 47, No. 3, March 2009. 2 S. Ghosh, J. Choi, and J. R. Edwards, RANS and Hybrid LES/RANS Simulations of the Effeccts of Micro Vortex Gernerators Using Immersed Boundary Methods, AIAA 2008-3728. 3 S. Xu and M. P. Martin, Inflow Boundary Conditions for Compressible Turbulent Boundary Layers, AIAA 03-3963. 4 N. A. Adams, Direct Numerical Simulation of Turbulent Compressible Ramp Flow, Theoret. Comput. Fluid Dynamics (1998) 12:109-129.