Stability of Supersonic Boundary Layers Over Blunt Wedges P.Balakumar* NASA Langley Research Center, Hampton, VA 23681 Receptivity and stability of supersonic boundary layers over blunt flat plates and wedges are numerically investigated at a free stream Mach number of 3.5 and at a high Reynolds number of 106/inch. Both the steady and unsteady solutions are obtained by solving the full Navier-Stokes equations using the 5th-order accurate weighted essentially non-oscillatory (WENO) scheme for space discretization and using third-order total-variation-diminishing (TVD) Runge-Kutta scheme for time integration. Computations are performed for a flat plate with leading edge thicknesses of 0.0001, 0.001, 0.005 and 0.01 inches that give Reynolds numbers based on the leading edge thickness ranging from 1000 to 10000. Calculations are also performed for a wedge of 10 degrees half angle with different leading edge radii 0.001 and 0.01 inches. The linear stability results showed that the bluntness has a strong stabilizing effect on the stability of two-dimensional boundary layers. The transition Reynolds number for a flat plate with a leading edge thickness of 0.01 inches is about 3.5 times larger than it is for the Blasius boundary layer. It was also revealed that boundary layers on blunt wedges are far more stable than on blunt flat plates. Introduction Transition from laminar to turbulent state in shear flows occurs due to evolution and interaction of different disturbances inside the shear layer. Though there are several mechanisms and routes to go from a laminar to a turbulent state, most of them generally follow these fundamental processes: • Receptivity • Linear instability • Nonlinear instability and saturation • Secondary instability and breakdown to turbulence In the receptivity process, unsteady disturbances in the environment such as acoustic waves and turbulence interact with the inhomogeneities in the geometry such as roughness and generate instability waves inside the shear layer. In quiet environments, the initial amplitudes of these instability waves are small compared to any characteristic velocity and length scales in the flow. In the linear instability stage, the amplitudes of these instability waves grow exponentially downstream and this process is governed by the linearized Navier-Stokes equations. Further downstream, the amplitudes of the disturbances become large and the nonlinear effects inhibit the exponential growth and the amplitude of the waves eventually saturate. In the next stage, these finite amplitude saturated disturbances become unstable to two- and/or three-dimensional disturbances. This is called the secondary instability and beyond this stage the spectrum broadens, due to complex interactions and further instabilities, and the flow becomes turbulent in a short distance downstream. In previous studies1,2, the interactions of two and three-dimensional * Aerospace Engineer, Flow Physics and Control Branch, NASA Langley Research Center, MS 170, AIAA Member. 1 of 20 American Institute of Aeronautics and Astronautics acoustic disturbances with and without isolated two-dimensional roughness elements in a supersonic boundary layer were investigated. The simulations showed that the linear instability waves are generated very close to the leading edge. The wavelength of the disturbances inside the boundary layer first increases gradually and becomes longer than the wavelength for the instability waves within a short distance from the leading edge. The wavelength then decreases gradually and merges with the wavelength for the Tollmien_Schlichting wave. The initial amplitudes of the instability waves near the neutral points, the receptivity coefficients, are about 1.20 and 0.07 times the amplitude of the free-stream disturbances for the slow and the fast waves respectively. It was also revealed that a small isolated roughness element does not enhance the receptivity process for the given nose bluntness. In this paper, the stability and generation of instability waves by acoustic disturbances in supersonic boundary layers over blunt flat plates and wedges with a free stream Mach number of 3.5 are investigated. The transition process mainly depends on the boundary layer characteristics and on the frequency, wave number distributions, and the amplitudes of the disturbances that enter the boundary layer. The boundary layer profiles depend on the flow parameters such as Mach number, Reynolds number, wall temperature, and model geometry. In supersonic and hypersonic boundary layers one important geometrical parameter is the leading edge bluntness. The effects of bluntness on transition have been studied experimentally and numerically by many researchers3-8. It was found that the bluntness generally stabilizes the boundary layer. The critical Reynolds numbers for blunt cones are much higher compared to that for sharp cones. However, the transition Reynolds number increased only by a factor of two compared to the sharp cones. It was identified that the entropy layer that is formed near the bow shock region persists for a long distance downstream and makes the boundary layer more stable compared to the sharp cone case. After the entropy layer and the boundary layer that is developing along the surface merges together, the boundary layer becomes unstable. It was also found that in addition to first and the second modes instability waves, other inviscid type disturbances grow inside the entropy layer. It is also observed that with increasing bluntness the stabilizing trend is reversed in axi-symmetric boundary layers. Another influence of the bluntness is in the generation of instability waves near the leading edge region. The objectives of this work are to estimate the stabilizing effect of the bluntness on the supersonic boundary layers over blunt flat plates and wedges and to estimate the receptivity coefficient of the instability waves generated near the leading edge. To investigate the effect of the Reynolds number based on the nose bluntness, simulations are performed at different leading edge thickness b = 0.0001, 0.001, 0.005 and 0.01 inches and at a unit Reynolds number of 1.0*106/inch for a flat plate and at nose radii of r 0 = 0.001 and 0.01 for a 10 degrees wedge. One simulation is also performed at a higher unit Reynolds number of 4.0*106/inch with 0.01 inches bluntness. This causes the Reynolds number based on the nose radius to vary from 102 to 4*104. The results consist of: (1) mean flow profiles, linear stability and transition onset Reynolds numbers for flat plates and wedges at different bluntness, and (2) receptivity coefficients for different bluntness. A schematic diagram of the computational set up is depicted in Fig. 1. Governing Equations. The equations solved are the three-dimensional unsteady compressible Navier-Stokes equations in conservation form ∂ ∂ ( ) Q + F −F = 0. (1) ∂t i ∂x ji vji j 2 of 20 American Institute of Aeronautics and Astronautics ⎧ ρ⎫ ⎧ ρu ⎫ ⎧ 0 ⎫ j ⎪ ⎪ ⎪ρ E⎪ ⎪ (ρE+ p)u ⎪ ⎪ uτ + vτ + wτ − q ⎪ ⎪ ⎪ ⎪ j ⎪ ⎪ 1j 2j 3j j⎪ Here Q = ⎨ ρu⎬ [F ]= ⎨ ρuu +δ p⎬ [F ]=⎨ τ ⎬ . (2) i ji j 1j vij ⎪ 1j ⎪ ⎪ ρv⎪ ⎪ ρvu +δ p⎪ ⎪ τ ⎪ j 2j 2j ⎩⎪ρ w⎭⎪ ⎩⎪ρ wuj +δ3jp⎭⎪ ⎩⎪ τ3j ⎭⎪ Here (x, y, z) are the Cartesian coordinates, (u, v ,w) are the velocity components, ρ is the density, and p is the pressure. E is the total energy given by u2 + v2 +w2 E= e+ , 2 e= c T, p = ρRT. (3) v Here e is the internal energy and T is the temperature. The shear stress and the heat flux are given by ⎧ ∂u ∂u 2 ∂u ⎫ ∂T τ =μ⎨ i + j − δ k⎬, q =−k . (4) ij ∂x ∂x 3 ij∂x j ∂x ⎩ ⎭ j i k j The viscosity (μ) is computed using Sutherland’s law and the coefficient of conductivity (k) is given in terms of the Prandtl number Pr. The variables ρ, p, T and velocity are non-dimensionalised by their corresponding reference variables ρ, p , T and RT respectively. The reference value for length is ∝ ∝ ∝ ∞ computed by νx /U , where x is a reference location. For the computation, the equations are 0 ∞ 0 transformed from physical coordinate system (x, y, z) to the computational curvilinear coordinate system (ξ,η,ζ) in a conservative manner and the governing equations become ∂ ∂ ( ) Q + F −F = 0. (5) ∂t i ∂x ji vji j The components of the flux in the computational domain are related to the flux in the Cartesian domain by Q J [ ] [ ] Q = i , F = F , (6) i J ji J ji ⎡∂ (ξ,η,ζ)⎤ where J = . ⎣⎢ ∂(x,y,z)⎦⎥ Solution Algorithm The governing equations are solved using a 5th order accurate WENO scheme for space discretization and using a third order, total variation diminishing (TVD) Runge-Kutta scheme for time integration. These methods are suitable in flows with discontinuities or high gradient regions. These schemes solve the governing equations discretely in a uniform structured computational domain in which flow properties are known point wise at the grid nodes. They approximate the spatial derivatives in a given direction to a higher order at the nodes, using the neighboring nodal values in that direction, and they integrate the resulting equations in time to get the point values as a function of time. Since the spatial derivatives are independent of the coordinate directions, the method can easily add multidimensions. It is well known that approximating a discontinuous function by a higher order (two or more) polynomial generally 3 of 20 American Institute of Aeronautics and Astronautics introduces oscillatory behavior near the discontinuity, and this oscillation increases with the order of the approximation. The essentially nonoscillatory (ENO) and the improvement of these WENO methods are developed to keep the higher order approximations in the smooth regions and to eliminate or suppress the oscillatory behavior near the discontinuities. They are achieved by systematically adopting or selecting the stencils based on the smoothness of the function, which is being approximated. Ref.9 explains the WENO and the TVD methods and the formulas and Ref.10 gives the application of the ENO method to the N-S equations. Ref.11 describes in detail the solution method implemented in this computation. At the outflow boundary, characteristic boundary conditions are used. At the wall, the simulation uses viscous conditions for the velocities and a constant temperature condition, and it computes density from the continuity equation. This wall temperature is approximately equivalent to adiabatic wall condition. In the spanwise direction, symmetric and periodic conditions are used at the boundaries. In the mean flow computations, the simulation prescribes the free-stream values at the upper boundary, which lies outside the bow shock. In the unsteady computations, it superimposes the acoustic perturbations to the uniform mean flow at the upper boundary. The procedure is to first compute the steady mean flow by performing unsteady computations using a variable time step until the maximum residual reaches a small value ~10- 11. These computations use a CFL number of 0.4. The next step is to introduce unsteady disturbances at the upper boundary of the computational domain and to perform time accurate computations to investigate the interaction and evolution of these disturbances downstream. The grid is generated using analytical formulae. The grid stretches in the η direction close to the wall and is uniform outside the boundary layer. In the ξ direction, the grid is symmetric about the leading edge and very fine near the nose and is uniform in the flat region. The grid is uniform in the spanwise direction. The outer boundary that lays outside of the shock follows a circle near the nose region with its vertex located a short distance upstream of the nose and follows a parabola downstream of the nose to capture the boundary layer accurately. The computational domain extends from x = -0.015 to 72.0 inches in the axial direction depending on the transition Reynolds numbers. Calculations were performed using different grid sizes varying from (2001*251*11) to (4001*301*11) depending on the size of the domain. Due to the very fine grid requirement near the nose, the allowable time step is very small and the computations become very expensive to simulate the unsteady computations in the entire domain at once. To overcome this, calculations are performed in two steps. First, the computations are done near the nose region with a very small time step. Second, the flow properties in the middle of this domain are fed as inflow conditions for the second larger domain and the computations are carried out with a larger time step. The symmetric acoustic field that impinges on the outer boundary is taken to be in the following form. p′ =Real{p˜ eiαacx±iβz+iεacy−iωt} ac ac (7) +Real{p˜ eiαacx±iβz−iεacy−iωt}. ac Here α , β , ε are the acoustic wavenumber, and ω is the frequency of the acoustic disturbance. The ac ac ac wavenumber in the y-direction ε determines the incident angle of the acoustic waves and in this paper ac computations are performed for zero incident angle, ε = 0. ac Results Computations are performed for supersonic flows over semi-infinite flat plates and wedges with blunt leading edges. Table 1 gives the flow parameters and Fig. 1 shows the schematic diagram of the computational set up. The leading edge of the flat plate is modeled as a super ellipse of the form 4 of 20 American Institute of Aeronautics and Astronautics (x−a)4 y2 + =1. (8) a4 b2 Here b is the half thickness of the plate and computations are performed for several values of thickness b = 0.0001, 0.001, 0.005, 0.01 inches. It should be noted that sharp flat-plate experiments, in general, employ beveled leading edges with very small leading edge radii. In the computations, the beveled edges are modeled as thin flat plates with small bluntness. The aspect ratio a/b is taken as 10 hence the blunt leading edge is joined with the straight portion of the plate at x = 10b. For a free stream unit Reynolds number of 1.0*106/inch, these leading edge thicknesses give the Reynolds numbers based on the leading edge thickness Re equal to 102, 103, 5*103 and 104. b Simulations are also performed for M = 3.5 over a wedge with 10 degrees half angle. The conditions behind the oblique shocks for different wedge angles are given in Table 2. The nose region of the wedge is modeled as a circle of the form (x−r )2+ y2 =r2. (8) 0 0 Here r is the radius of the leading edge bluntness. Simulations are performed for nose radii r = 0.001 0 0 and 0.01 inches at a unit Reynolds numbers 1.0*106/inch. This provides nose Reynolds numbers of 103, 104. To achieve a higher nose Reynolds number of 4*104, one computation is performed at a unit Reynolds number of 4.0*106/inch with a nose radius of r = 0.01 inches. Different cases are summarized 0 in Table 3. Table 1 Flow parameters used in the computation. Free stream Mach number: M =3.5 ∝ Free stream Reynolds number: Re =1.0*106/in. ∝ Free stream density: ρ=2.249*10-2 lbm/ft3 ∝ Free stream pressure: p =187.74 lbf/ft2 ∝ Free stream velocity: U =2145.89 ft/s ∝ Free stream temperature: T =156.42 °R ∝ Free stream kinematic viscosity: ν =1.7882*10-4 ft2/s ∞ Wall temperature: T =476 °R w Prandtl number: Pr= 0.72 Ratio of specific heats: γ=1.4 ν x Length scale ∞ 0 = 5.892*10−5 ft. (x = 0.5 in.) U 0 ∞ The boundary layer thickness at x=1 in.: δ= .01275 in. 0 Non–dimensional frequency F=1*10-5 is equivalent to 41.0 kHz 2πν f The non-dimensional frequency F is defined as F= ∞ , U2 ∞ 5 of 20 American Institute of Aeronautics and Astronautics where f is the frequency in Hertz. Table 2 Conditions downstream of the shock. Wedge Shock angle, Mach Pressure Density Temperature Unit angle, deg. number ratio ratio ratio Reynolds deg. number ratio 10 24.384 2.904 2.269 1.767 1.284 1.28 20 34.602 2.298 4.442 2.648 1.677 1.34 30 47.755 1.655 7.665 3.438 2.229 1.12 Table 3 Parameters in the computations. Nose radius Unit Reynolds number r in. Reynolds based on nose 0 number radius /in. 0.0001 1.0*106 102 0.001 1.0*106 103 0.005 1.0*106 5*103 0.01 1.0*106 104 0.01 4.0*106 4*104 Linear instability The linear stability results for the similarity boundary layer over the wedge at conditions downstream of the shock are presented in Fig. 2. The stability diagram for a flat plate is given in Ref. 1. The figure depicts the neutral stability diagram in the (Re, F) plane for different wave angles 0, 45 and 60 degrees. The figure also shows the N-Factor curves and the growth rates for the most amplified disturbances. Here, the variables are non-dimensionalized by the variables downstream of the shock. To obtain the variables non-dimensionalized by the free stream values as given in Table 1, the variables in this section should be multiplied by the appropriate factors. The frequency variable F has to be multiplied by 1.2 to obtain in terms of free stream values. The critical Reynolds number is about 209 and this occurs for an oblique wave of angle 60 degrees. The most amplified frequencies are in the range F = 0.075-1.25*10-5 and the spanwise wave number of the most amplified wave is about β = 0.025. This corresponds to 0.178 inches in dimensional units and is equivalent to about 14 boundary layer thicknesses. It is also observed that at higher Reynolds numbers Re > 1000, only the low frequency disturbances F < 3.0*10-5 are unstable. This implies that an acoustic disturbance with frequencies less than 120 kHz may be the relevant frequency 6 of 20 American Institute of Aeronautics and Astronautics range for generating instability waves inside the boundary layer. The frequency of the most amplified wave is about 40-50 kHz and the maximum N-factor at x = 12 in. (Re = 3464) is about 8.6. Mean flow and linear stability for the flat plate Figure 3 shows the mean flow density contours extracted from the Navier-Stokes computations. Figure 3(a) shows the entire domain for the leading edge bluntness of b = 0.01 in. and Figs. 3 (b), (c), (d) show the flow field near the leading edge for the leading edge bluntness b = 0.01, 0.001, 0.0001 inches. The leading edge shocks are located approximately at 7.7*10-4, 6.0*10-6, and 4.0*10-7 in. upstream of the leading edge. Figure 4 shows the Mach number distributions along the boundary layer edge for the three different bluntness cases. The Mach number distributions reach constant values within a short distance from the leading edge. The density profiles at different axial locations are plotted in Figs. 5(a)-(c) for the different bluntness cases b =0.0001, 0.001 and 0.01 inches in the similarity coordinates. The compressible Blasius similarity profile is also included for comparison and Fig. 5(d) shows the profiles for b = 0.01 inches in the physical coordinate. It is seen that very close to the leading edge, there exists a strong shock and that the associated compression is followed by an expansion over the leading edge and the shock becomes weaker away from the leading-edge region. The boundary layer profiles with b = 0.0001 inches slowly approach the Blasius similarity profile close to x = 3 inches. With increasing bluntness, the profiles did not approach the similarity profiles within the computational domain, which is closer to the transition onset point. The difference between the profiles with the bluntness and the similarity profiles increases with the bluntness. Figure 6 shows the growth rate and the N-Factors for the most amplified disturbances computed using the profiles obtained from the numerical simulation for different bluntness b = 0.0001, 0.001, 0.005, 0.01 inches. For comparison, the results for the Blasius similarity profiles, which model a sharp leading edge, are also shown in the figure. It is seen that the stability results and the N-Factor results obtained using the Navier-Stokes mean flow with the smallest bluntness b = 0.0001 agree reasonably well with those obtained using the Blasius similarity profiles. The frequency and the spanwise wave number for the most amplified wave are about F = 1.25*10-5 and β = 0.025 in both cases. However, for larger bluntness, there is a significant difference both for the mean flow and the stability results as the growth rate becomes smaller and the N-Factor curve moves downstream. For the smaller bluntness b = 0.001, the N-Factor curve remains closer to the similarity curve. For the larger bluntness cases b = 0.005 and b = 0.01 the growth rates become smaller and the N-Factor curves move further to the right. The most amplified frequency and the spanwise wave number are (0.80*10-5, 0.02), (0.325*10-5, 0.01) (0.325*10-5, 0.01) for b = 0.001, 0.005 and 0.01 inches respectively. This shows that the frequencies and the spanwise wave numbers of the most amplified disturbances become smaller with increasing bluntness. The growth rate curves are similar to the Blasius profile for smaller bluntness b = 0.0001 and 0.001 and at higher bluntness b = 0.005 and 0.01 they take a different shape. The growth rates first increase and plateau for a long distance and cause the disturbances to grow. For comparison, the growth rate and the N-Factor curve for F = 0.30*10-5 and β = 0.01 for the similarity profiles are also shown in the figure. This has a larger growth rate than that for the bluntness case, however they have the same growth rate plateau at larger axial distances. It may be that at larger distances the boundary layer profiles have the similar stability characteristics as for the similarity profiles. Boundary-layer transition data on a flat plate and on a cone, and free stream noise levels and the power spectral distribution of the free stream noise are presented in Ref. 12. The data shows the transition Reynolds number for a flat plate in quiet conditions is about 11*106. If this is used in the correlation to obtain the N-Factor at the transition, the Fig. 6 gives an N-factor of 8.0 for a constant spanwise wave number. The transition Reynolds number obtained using this N-factor for different bluntness cases are summarized in Table 4 and is plotted in Fig. 7. The ratio between the transition Reynolds number with bluntness and the transition Reynolds number for the similarity profile, (Re ) /(Re ) is about 1.05, T b T Similarity, 7 of 20 American Institute of Aeronautics and Astronautics 1.30, 2.48, 3.49 respectively for Re = 102, 103, 5*103, 104. In the experiment12, when the thickness of the b beveled leading edge was increased from 0.0001 inches to 0.001 inches, the transition Reynolds number increased from 11*106 to 14.0*106. This is an increase by a factor of 127. In the computations, when the half thickness of the flat plate was increased from 0.0001 inches to 0.001 inches the transition Reynolds number increased by a factor of 1.20. Even though the beveled edges are modeled as a thin flat plate in the computations, the computational results agree very well with the experimental results. Table 4 Transition Reynolds number for the blunt flat plates. Leading edge Reynolds Transition Transition Ratio thickness b in. number based location X (in.) Reynolds (Re ) /(Re ) T b T Similarity on b number *106 0.0 0 11.00 11.00 1.00 0.0001 102 11.77 11.77 1.05 0.001 103 14.12 14.12 1.30 0.005 5*103 27.30 27.30 2.48 0.01 104 38.37 38.37 3.49 Mean flow and Stability for the Wedge Figure 8 shows the mean flow density contours computed using the WENO code. The figures 8(a-c) show the results for the wedge angle of 10 degrees at different nose radii r = 0.001 and 0.01 inches. 0 Smaller nose radii cases r = 0.001 and 0.01 are performed at a unit Reynolds number of 1.0*106/inch 0 which yield Reynolds numbers based on the nose radius of 103 and 104. Figure 8(d) show the results obtained at a higher unit Reynolds number of 4.0*106/inch with r = 0.01 which yields the Reynolds 0 number based on the nose radius of 4.0*104. Figure 8(a) shows the density contours in larger domain and other figures show the flow field near the nose region. One interesting observation is that the inviscid density contours and the shock locations are same between Figs. 8(b) and (d) which are obtained with the same bluntness b = 0.01 but at different unit Reynolds numbers 1.0 and 4.0*106. The leading edge shocks are located approximately at 0.0008 and 0.005 in. upstream of the leading edge. Figure 9 shows the Mach number distributions along the boundary layer edge for the different bluntness cases. The inviscid Mach number obtained from the shock conditions is also shown in the figure. The Mach number distributions reach constant values within a short distance from the leading edge. This suggests that the edge conditions are not the cause for the stabilization of the boundary layers in the cases with small bluntness. The density profiles at different axial locations are plotted in Fig. 10(a)-(c) for the different bluntness cases r 0 =0.0001, 0.01, 0.01(Re/in. =4*106) inches in the similarity coordinates. The compressible Blasius similarity profile is also included for comparison and Fig. 10(d) shows the profiles for r = 0.01 inches in 0 the physical coordinate. With increasing bluntness, the profiles did not approach the similarity profiles within the computational domain, which is closer to the transition onset point. The difference between the profiles with the bluntness and the similarity profiles increases with the bluntness. Figure 11 shows the growth rate and the N-Factors for the most amplified disturbances computed using the profiles obtained from the numerical simulation for different bluntness r = 0.001, 0.01, 0.01 0 (Re/in. = 4*106) inches. For comparison, the results for the Blasius similarity profiles, which model a sharp leading edge, are also shown in the figure. The frequency and the spanwise wave number for the 8 of 20 American Institute of Aeronautics and Astronautics most amplified wave are about F = 0.90*10-5 and β = 0.025 for the similarity profiles. However, for larger bluntness, there is significant difference both for the mean flow and the stability results when the growth rate becomes smaller and the N-Factor curve moves downstream. Foe the smaller bluntness r = 0.001, 0 the N-Factor curve remains closer to the similarity curve. For the larger bluntness cases r = 0.01 and r = 0 0 0.01 (4 mil) the growth rates become smaller and the N-Factor curves move further to the right. The most amplified frequency and the spanwise wave number are (0.60*10-5, 0.02), (0.082*10-5, 0.005) (0.175*10- 5, 0.0125) for r = 0.001, 0.01 and 0.01 (4 mil) inches respectively. This shows that the frequencies of the 0 most amplified disturbances become smaller with increasing bluntness. The growth rate curves are similar to the Blasius profile for smaller bluntness r = 0.001 at higher bluntness r = 0.01 and 0.01 ( Re/in. = 0 0 Re/inch = 4*1064 mil) they take a different shape as observed in the flat plate case. The growth rates first increase and plateau for a long distance and causes the disturbances to grow. The transition Reynolds numbers obtained using the N-factor of 8.0 for different bluntness cases are summarized in Table 5 and plotted in Fig. 7. The ratio between the transition Reynolds number with bluntness and the transition Reynolds number for the similarity profile, (Re ) /(Re ) is about 1.42, 7.01, 11.96 respectively for T b T Similarity, Re = 103, 104, 4*104. Previous experiments3 and the stability calculations5 showed that the transition b Reynolds number for a blunt cone at a Mach number of 8 with nose Reynolds numbers of 30,000 increased by a factor of 1.7~2.0 compared to a sharp cone. This implies that the bluntness effects are much stronger in flows over wedges than in flows over cones. Table 5 Transition Reynolds number for the blunt wedges. Nose radius Reynolds Transition Transition Ratio r in. number based location X (in.) Reynolds (Re ) /(Re ) 0 T b T Similarity on nose radius number *106 0.0 0 10.40 10.40 1.00 0.001 103 14.38 14.38 1.42 0.01 104 73.00 73.00 7.02 0.01 (4 mil) 4*104 28.50 114.0 10.96 Interaction of three-dimensional acoustic waves with the flat plate and wedge boundary layers. After the mean flow is obtained, three-dimensional slow and fast acoustic disturbances are separately introduced at the outer boundaries and time accurate simulations are performed. The non-dimensional frequency and the spanwise wave number are F=1.25*10-5 and β = 0.025. These parameters give the largest N-factor close to the experimental transition location. To remain in the linear regime, very small initial amplitude of p˜ /p =1.0*10−7 is prescribed for the free-stream acoustic waves. Even with these ac ∞ small initial amplitudes, nonlinearity starts to develop near the end of the computational domain x ~ 10.0 inches. Details about the acoustic disturbances and the analysis to compute the wave numbers of the instability waves generated inside the boundary layer are described in Ref. 1. Figure 12 shows the results for the evolution of the unsteady fluctuations obtained from the simulation for the slow wave at a fixed time for the case b = 0.0001 inches. Figure 12(a) shows the contours of the density fluctuations in the entire domain and Fig. 12(b) depicts the results inside the boundary layer. The perturbation field can be divided into four regions. One region is the area outside the shock where the acoustic waves propagate uniformly. The second region is the shock layer across which the acoustic waves are transmitted. The third region is the area between the shock and the boundary layer. 9 of 20 American Institute of Aeronautics and Astronautics This region consists of transmitted external acoustic field and the disturbances that are radiated from the boundary layer. The fourth region is the boundary layer where the boundary layer disturbances evolve. Figure 13 shows the amplitude of the pressure fluctuations along the wall in a log scale. Figure 13 also includes the results from the parabolized stability equations (PSE) computations obtained for the same mean boundary layer profiles. The figures clearly show the initial generation and the eventual exponential growth of the instability waves inside the boundary layer. The slow wave whose wavelength is closer to the wavelength of the instability wave transforms into instability waves smoothly. The fast waves whose wavelengths are much larger are initially modulated by short wavelength disturbances. These short waves transform into instability waves and grow exponentially downstream. The growth of the disturbances agrees very well with the PSE results about one acoustic wavelength downstream of the neutral point. Following the PSE results up to the neutral point, the initial amplitude of the instability waves at the neutral point can be estimated. From these values the receptivity coefficients defined by the initial amplitude of the pressure fluctuations at the wall at the neutral point non-dimensionalised by the free-stream acoustic pressure can be evaluated. (p ) C = wall n (15) recpt,pwall p ac The computed receptivity coefficients for the slow and the fast waves are C =1.20 recpt,p ,S wall C =0.07 (16) recpt,p ,F wall Similarly, the receptivity coefficients based on the maximum density fluctuations inside the boundary layer normalized by ρ are calculated. ac C =8.40 recpt,ρ ,S max C =0.47 (17) recpt,ρ ,F max The ratio of the receptivity coefficient between the slow and the fast modes are about 17.0. As expected, the slow modes, whose phase speeds are close to the neutral stability waves of the boundary layer, excite the instability waves more efficiently than the fast waves. Figure 14 shows the evolution of wall pressure fluctuations induced by the slow acoustic wave for larger bluntness cases b = 0.001 and b = 0.01 inches and the non-dimensional frequencies for these cases are F=1.25*10-5 and 0.75*10-5 respectively. The computed receptivity coefficients for the bluntness b = 0.001 and 0.01 are C =1.18 recpt,p ,S wall C =0.33 recpt,p ,S wall This shows that for smaller bluntness the receptivity coefficients are almost the same and for the larger bluntness case the receptivity coefficient is about three times smaller than for the smaller bluntness cases for this frequency. The most amplified frequency for the bluntness b = 0.01 is about F=0.30*10-5 and the receptivity coefficient will be even smaller than this due to the attenuation of the disturbances in the leading edge region. Figures 15 and 16 show the results for the blunt wedge cases with a small bluntness r = 0.001. 0 The non-dimensional frequency and the spanwise wave number are F=1.25*10-5 and β = 0.025. Figure 15 shows the results for the evolution of the unsteady fluctuations obtained from the simulation for the slow wave at a fixed time. Figure 15(a) shows the contours of the density fluctuations in the entire domain and Fig. 15(b) depicts the results inside the boundary layer. It is seen that the region between the boundary 10 of 20 American Institute of Aeronautics and Astronautics