The impact of turbulence modeling on the numerical simulation of the crossing-shock-wave/boundary-layer interactions occuring in a Mach 4 ow on 7 £ £ 7 deg, 7 £ £ 11 deg, and 15 £ £ 15 deg double-sharp n plates is analyzed.The full Reynolds-averaged Navier-Stokes equations are solved with linear and weakly nonlinear formulations of the k-! turbulence model, on grids up to 4 £ £ 10 6 cells. The overpredicted heat transfer on the bottom plate is shown to be closely related to the main three-dimensional features developing in these ows. The stronger the interaction, the more the numerical solutions violate the realizability principles. By introducing a dependence of cµ on the velocity gradients, realizable solutions are obtained and analyzed. The heat-transfer coef cients are effectively lowered but not suf ciently to meet the experimental data. The expected impact of other corrections, based on a limitation of the turbulent length scale or some compressibility effects, is evaluated and shown to be insuf cient to ll the gap between computed and measured heat-transfer coef cients. The streamlines arriving near the wall in the regions of overpredicted heat transfer are shown to originate from the very narrow regions close to the n leading edges, in the upper part of the incoming boundary layer, and to be associated with an increase of turbulent kinetic energy when approaching the bottom wall, rather than when crossing the shock waves.
Nomenclature
C h= heat-transfer coef cient or Stanton number; Eq. (1) c l = coef cient in the constitutive equation for l t , 0.09 D t (q } ) = substantial derivative, @q } / @t @q U õ } / @X i k = turbulent kinetic energy (TKE), 1 2 ] u õ u õ P k = TKE production rate, q ] u õ u Ò S õ Ò S = strain invariant, (2S õ Ò S õ Ò ) S õ Ò = strain rate, 1 2 (@U õ / @x Ò @U Ò / @x õ ) S õ Ò = traceless strain rate, S õ Ò 1 3 S kk d õ Ò s = dimensionless strain invariant, S/ x U õ = mass-averaged velocity components U , V , W u õ = uctuating velocity components u, v, w (mass averaging) ] u õ u Ò = mass-averaged velocity correlations x Ò = Cartesian coordinates x, y, z a m = realizability factor; Eq. (8) d õ Ò = Kronecker symbol d 0 = incoming boundary-layerthickness, 3.5 mm ² = dissipation of TKE l , m = dynamic and kinematic molecular viscosities, l q m l t , m t = Dynamic and kinematic turbulent viscosities, l t q m t $ = dimensionless vorticity invariant, X / x q = mass density X = vorticity invariant, (2X õ Ò X õ Ò )