LES of shock wave / turbulent boundary layer interaction

9
under free-stream conditions M ∞ = 2.95 and Re δ0 = 63560 (Zheltovodov
& Yakovlev, 1986). A constant-current hot-wire measurement
was employed to obtain total temperature and mass-flux fluctuations,
based on the technique of Kovasznay (1953). The mass-flux fluctuation
was decomposed into density and velocity fluctuations using the strong
Reynolds analogy. The mean skin-friction measurements were performed
by the Global Interferometry Skin Friction technique (GISF) (Borisov
et al., 1993).
Heat transfer measurements for the small model were performed by
Zheltovodov et al. (1987) at free-stream conditions M ∞ = 2.9 , Re δ0 =
86520 and M ∞ = 4, Re δ0 = 100000. The data include adiabatic and
heated-wall temperature distributions and the heat transfer coefficient.
An experimentally obtained flow field structure and the measurement
stations are sketched in figure 1.4 for a 25 ◦ -degree compression
ramp. The positions of the stations in terms of the downstream coordinate
measured along the wall from the compression corner position,
are summarized in table 1.1. Obviously, the big model allowed more
detailed measurements of the mean flow. The Reynolds numbers based
on the undisturbed boundary layer thickness are Re δ0 = 144000 and
Re δ0 = 63560 for big and small models respectively. Despite of significant
Reynolds number differences, the mean flow structure is rather
similar if scaled in δ 0 as shown by Zheltovodov & Yakovlev (1986), so
both experiments may be used together for a detailed analysis.
1.3 Prediction capabilities
As pointed out by Dolling (1998, 2001) and confirmed by a more recent
comprehensive analysis (Knight et al., 2003) the numerical prediction of
SWTBLI by statistical turbulence modeling is yet unsatisfactory. For
situations with shock-induced flow separation computational results employing
Reynolds-averaged (RANS) turbulence modeling exhibit a large
scatter of predicted separation lengths for various geometrical configurations.
Although numerous computations based on the Reynoldsaveraged
Navier-Stokes equations (RANS) have been performed, currently
only weak and moderate interactions, characterized by low supersonic
Mach number or small flow deflection angles, can be predicted by
RANS computations without specific a posteriori adjustment of turbulence
models. For strong interactions, the results of RANS computations
show generally a significant disagreement with experimental data