Beruflich Dokumente
Kultur Dokumente
Abstract: The present investigations are centered on understanding the discrepancies in shock wave boundary-layer interaction (SWBLI)
for perfect and real gas laminar flows. In view of this, the in-house-developed computational fluid dynamics (CFD) solvers are integrated
with a gradient-based optimization algorithm to predict the critical radii of SWBLI in the case of perfect and real gas flows. The
developed high-fidelity approach has been observed to be useful in the precise estimation of critical radii of bluntness. Further, studies
for SWBLI revealed that real gas effects reduce the extent of separation in comparison with the perfect gas flow and also necessitate lower
magnitudes of critical radii. It has been noted that a reduced requirement of a high entropy layer thickness and upstream overpressure
region demonstrate a need for a lower value of inversion and equivalent radii for real gas flow conditions. Therefore, a larger estimate of
the equivalent radius of SWBLI, obtained for perfect gas flow conditions, or any radius larger than that would definitely provide the
necessary separation control for real gas flows. DOI: 10.1061/(ASCE)AS.1943-5525.0001085. © 2019 American Society of Civil
Engineers.
Author keywords: Shock wave boundary-layer interaction (SWBLI); Reacting flows; Critical radii; Separation control; High-fidelity
framework.
6 7
comes equal to the equivalent radius. With this information, it 6 τ xx 7
6 7
is still difficult to predict the critical radii for a given hypersonic 6 7
6 τ 7
flow and plate-ramp configuration, since this would require in- 6 xy
PN 7
6 7
formation on the entire flow field. Further, these observations are 6 uτ xx þ vτ xy − qx − i¼1 hi Ci ūi 7
6
Ev ¼ 6 7
made specifically in connection with a perfect gas flow only and
−C1 ū1 7
one needs to consider high-enthalpy flow condition. Thus, one of 6 7
6 7
the contributions of the present investigations is to cross-check 6 7
6 · 7
the validity of indication about the flow field around critical radii 6 7
6 · 7
for a high-enthalpy flow. This process also puts forth the obser- 4 5
vation of a real gas effect on SWBLI. The effect of enthalpy has −C4 ū4
been studied in the literature (John and Kulkarni 2017), but real 2 3 2 3
0 0
gas effects have not. Therefore, a methodology that is suitable 6 7 6 7
for low- and high-enthalpy flows is not only developed to obtain 6 τ xy 7 607
6 7 6 7
estimates of critical radii of SWBLI but also to verify the resem- 6 7 6 7
6 τ 7 607
blance in flow features around critical radii in either case. In 6 yy
PN 7 6 7
6 7 6 7
view of these objectives, a high-temperature real gas flow solver, 6 uτ xy þ vτ yy − qy − i¼1 hi Ci v̄i 7 607
developed in house and integrated with an optimization pro- Fv ¼ 6
6
7;
7 S ¼ −6
6S 7
7
6 −C1 v̄1 7 6 17
cedure (CFD-O), is used to formalize the philosophy of critical 6 7 6 7
6 7 6 · 7
radius prediction for R-SWBLI. Details of the optimization 6 · 7 6 7
6 7 6 7
procedure, solver, and results are presented in the following 6 · 7 6 · 7
4 5 4 5
sections.
−C4 v̄4 S4
Computational Strategy for CFD-O The mixture pressure is evaluated using the sum of partial
pressures of individual species:
Finite-Volume Formulation XN
Ci
p ¼ Ru T ð2Þ
Estimation of the critical radii of bluntness for SWBLI for per- i¼1
MW i
fect and real gas flow situations is a major motivation of the cur-
rent investigations. To demonstrate the high-enthalpy effects in where p = pressure; Ru = universal gas constant; T = temperature;
SWBLI, two different versions of a laminar flow solver [Unstruc- N = number of species; and Ci and MW i = mass concentration and
tured Solver for Hypersonic Aerothermodynamic Simulations molecular weight of ith species, respectively. The temperature in
(USHAS)] developed in house are employed herein. The first the equation is determined by applying the Newton–Raphson
version of the solver is based on the perfect gas assumption, method iteratively to the internal energy:
which means the specific heat and specific heat ratio of gas remain
XN Z
constant with temperature and solves only Navier–Stokes equa- Ci T
tions (John and Kulkarni 2013, 2014a; John et al. 2016; John and ρe ¼ Cpi dT þ h0fi − p ð3Þ
i¼1
MW i TR
Kulkarni 2017). Another version considers fully coupled Navier–
Stokes and species continuity equations to deal with real gas where ρ = density; e = energy; T R = reference temperature; and Cpi
effects (Desai et al. 2016, 2017; Agarwal et al. 2017). High- and h0fi = specific heat at constant pressure and heat of formation of
enthalpy simulations using a similar methodology are reported species i. Here, species continuity equations provide the production
in the literature (Lee 1999; Ess and Allen 2005). An axisymmetric rate of various species. The required reaction kinetics data are in-
form of the governing equations, as shown in Eq. (1), is used to corporated into the solver, as mentioned by Dunn and Kang (1973).
develop the present solver: This solver also accounts for the temperature dependence of vari-
ous properties like specific heat, thermal conductivity, and viscosity
for each species, as elaborated by Gordon and Mcbride (1994). In
∂U ∂EI ∂FI ∂Ev ∂Fv this numerical study with a perfect gas flow solver or real gas
þ þ þS¼ þ ð1Þ
∂t ∂x ∂y ∂x ∂y flow solver, convective fluxes are obtained using an advection up-
stream splitting method (AUSM) (Liou and Steffen 1993) scheme,
whereas viscous fluxes are calculated using face values of flow
where quantities and their first derivatives, as detailed by Blazek
8 5
P/ (P )
6 4
4 3
2 2
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
0 1
−2 0
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12
(a) x (m) (b) x (m)
Fig. 2. Comparison of (a) skin friction coefficient Cf ; and (b) surface pressure distribution P=P∞ with numerical results of John and Kulkarni (2014a).
Table 1. Details of result for validation of present real gas solver into 50,400 quadrilateral cells. The flow is simulated using a com-
Separation Reattachment Plateau
putational approach similar to that described earlier. In this case, the
postshock temperature rise is enough to cause specific heat varia-
Case Point (mm) Point (mm) Pressure (Pa) tion; however, it is insufficient to activate dissociation reactions.
0.3 mm The obtained plots of surface pressure and heat flux distribution
Perfect gas 38.10 67.18 505.27 are plotted in Fig. 3. This figure clearly shows that the results ob-
Real gas 37.60 67.50 504.07 tained from the present simulation agree well with results reported
0.5 mm in the literature. Thus, the solver is validated for the high-enthalpy
Perfect gas 38.28 68.84 527.41 condition as well. Further, a detailed validation of the solver for
Real gas 38.00 69.20 526.60
reacting flow problems, including an unsteady shock tube test case
and steady flow past a sphere, are given in Desai et al. (2016, 2017).
1.4 1.4
Real gas flow solver Real gas flow solver
1.2 Experimental−Wieting, 1987 1.2 Experimental−Wieting, 1987
1 1
0.8 0.8
0
P/P0
Q/Q
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 20 40 60 80 0 20 40 60 80
(a) θ (b) θ
Fig. 3. Comparison of (a) surface heat flux Q=Q0 ; and (b) surface pressure distribution P=P0 with experimental results of Wieting (1987).
NO
Gradient evaluation
Determine λk k=k+1
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
Gradient method
k+1 k κ k
xk+1 = xk ± λkgrad. f(xk) R b =R b ± λ grad. f(Rb )
are indeed grid independent, especially for derived quantities like are illustrated for various blunt-leading-edge cases in Fig. 9. One of
wall distribution of the skin-friction coefficient. With that motive, a the major objectives of this study is to highlight and quantify this
thorough grid-independence study is conducted using four different effect for R-SWBLI. Therefore, the skin-friction coefficient was
grids with varying control volumes and first cell thickness normal obtained as an outcome of CFD simulations for each optimization
to the wall boundary. Table 3 shows the grid specifications used for cycle and is shown in Fig. 10 for two sample cases, along with the
this study. Figs. 6 and 7 show the surface distribution of coefficient reference sharp-leading-edge case. From Fig. 10, the separation
pressure (Cp ) and skin-friction coefficient (Cf ) for perfect and real bubble size is estimated for different radii. It is evident from the
gas flow solvers, respectively. It can be observed that both Cp and figure that the prediction of the separation bubble size is higher
Cf distributions are invariant with grid spacing. More importantly, with perfect gas flow simulations. As mentioned earlier, the lower
0.25 0.04
Grid 1
Grid 2
Grid 1 0.035 Grid 3
Grid 2 Grid 4
0.2 Grid 3
Skin friction coefficient (Cf )
Grid 4
0.03
Pressure coefficient (Cp )
−3
x 10
1
0.025
0.15
0.02 0
0.1 0.015 −1
0.01
−2
0.04 0.06 0.08
0.05 0.005
Zoomed view
0
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12
(a) x (m) (b) x (m)
Fig. 6. Grid-independent results of perfect gas solver (Rb ¼ 0.0 mm): (a) pressure distribution; and (b) skin friction distribution.
0.25 0.04
Grid 1
Grid 2
Grid 1 0.035 Grid 3
Grid 2 Grid 4
0.2 −3
Skin friction coefficient (Cf )
Grid 3 0.03 x 10
1
Pressure coefficient (Cp )
Grid 4
0.025
0
0.15
0.02
−1
0.1 0.015
−2
0.04 0.06 0.08
0.01
0.05 0.005
Zoomed view
0
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12
(a) x (m) (b) x (m)
Fig. 7. Grid-independent results of real gas solver (Rb ¼ 0.0 mm): (a) pressure distribution; and (b) skin friction distribution.
12
Perfect gas flow (R b−0.0 mm)
Real gas flow (R b−0.0 mm)
10
Perfect gas flow (R b−0.726 mm)
Real gas flow (R −0.726 mm)
2
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
−2
0 0 0
0 0.5 1 1.5 2 0 2 4 6 8 2 4 6 8
2
/ T/T Viscosity (m /s) x 10−5
Fig. 9. Variation of different properties in boundary layer 42 mm downstream of leading edge for various leading-edge radii cases.
b
Perfect gas flow (R −1.7 mm) thickness is lower than the HEL at the location upstream of the
0.2 b
Real gas flow (R −1.7 mm) upstream influence location. However, for a leading-edge bluntness
b
radius less than the inversion radius (Rb ¼ 0.626 mm), the boun-
0.15
dary layer remains thicker than the HEL, as shown in Figs. 15 and
16. Further, the thickness of the boundary layer approaches the
thickness of the HEL around the upstream influence location for
0.1 the inversion radius. The matching of the boundary-layer and
HEL thickness at the reference upstream influence station has been
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
Viscous perfect gas flow (R b−0.84 mm) High entropy layer edge
Inviscid perfect gas flow (R b−0.84 mm)
Boundary layer profile 0.015 0.015
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U U/U U/U
Ramp
Blunt leading edge
x=0.012 m x=0.022 m x=0.042 m
Separation bubble
Fig. 13. Variation of boundary-layer and entropy-layer thickness at three different locations for perfect gas flow with inversion radius (upstream
influence location 0.022 m).
Y (m)
0.005 0.005 0.005
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U∞ U/U∞ U/U∞ Ramp
Blunt leading edge
x=0.012 m x=0.022 m x=0.042 m
Separation bubble
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
Fig. 14. Variation of boundary-layer and entropy-layer thickness at three different locations for real gas flow with inversion radius (upstream
influence location 0.022 m).
0.01 0.01
M∞ 0.01
Y (m)
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U U/U U/U Ramp
∞ ∞ ∞
Blunt leading edge
x= 0.012 m x=0.022 m x=0.042 m
Separation bubble
Fig. 15. Variation of boundary-layer and entropy-layer thickness at three different locations for perfect gas flow with Rb ¼ 0.626 mm leading-edge
radius (upstream influence location 0.022 m).
0.01 0.01
M∞ 0.01
Y (m)
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U∞ U/U U/U∞ Ramp
∞
Blunt leading edge
x=0.012 m x= 0.022 m x= 0.042 m
Separation bubble
Fig. 16. Variation of boundary-layer and entropy-layer thickness at three different locations for real gas flow with Rb ¼ 0.626 mm leading-edge
radius (upstream influence location 0.022 m).
difference in the corresponding equivalent radius. But the need for a critical radii through an optimization process. Further, an optimi-
favorable pressure gradient at the reference upstream influence zation strategy was developed to integrate a gradient-based optimi-
location remains a valid basis for choosing leading-edge bluntness zation technique and CFD solvers developed in-house for perfect
as the equivalent radius for perfect and real gas flows. and real gas flows. As an outcome of this integration, inversion and
equivalent radii are found to be, respectively, 0.84 and 1.95 mm for
perfect gas flow and 0.781 and 1.95 mm for real gas flow. Real gas
Conclusions effects are shown to be responsible for this decrement in the sep-
aration size and thus lower leading-edge bluntness for separation
This article investigated shock wave boundary layer interaction control. Further, it is proposed that the inversion radius should have
control using leading-edge bluntness for the prediction of two a thicker high entropy layer around the upstream influence location
Y (m)
0.005 0.005 0.005
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U∞ U/U∞ U/U Ramp
Blunt leading edge ∞
x=0.012 m x=0.022 m x=0.042 m
Downloaded from ascelibrary.org by Nottingham Trent University on 08/09/19. Copyright ASCE. For personal use only; all rights reserved.
Separation bubble
Fig. 17. Variation of boundary-layer and entropy-layer thickness at three different locations for perfect gas flow with equivalent radius (upstream
influence location 0.022 m).
Viscous real gas flow (R b−1.85 mm) High entropy layer edge
Inviscid real gas flow (R b−1.85 mm)
Boundary layer profile
0.015 0.015
M 0.01 0.01
∞ 0.01
Y (m)
0 0 0
−0.5 0 0.5 1 −0.5 0 0.5 1 −0.5 0 0.5 1
U/U U/U U/U Ramp
Blunt leading edge ∞ ∞ ∞
x=0.012 m x=0.022 m x=0. 042 m
Separation bubble
Fig. 18. Variation of boundary-layer and entropy-layer thickness at three different locations for real gas flow with equivalent radius (upstream
influence location 0.022 m).
Blast wave theory (R −0.84 mm) Blast wave theory (R −0.781 mm)
b b
0.35 Blast wave theory (R −1.7 mm)
b 0.35 Blast wave theory (R −1.6 mm)
b
Blast wave theory (Rb−1.85 mm) Blast wave theory (Rb −1.7 mm)
Pressure coefficient (Cp )
0.3 0.3
Pressure coefficient (Cp )
Blast wave theory (R −1.89 mm) Blast wave theory (R −1.85 mm)
b b
0.25 Blast wave theory (R −1.95 mm) 0.25 Real gas flow (R −0.0 mm)
b b
Perfect gas flow (R −0.0 mm)
0.2 b 0.2
0.15 0.15
Upstream influence location Upstream influence location
0.1 0.1
0.05 0.05
0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
(a) x (m) (b) x (m)
Fig. 19. Comparison of surface pressure distribution from blast wave theory with (a) perfect gas flow; and (b) real gas flow.
than the local boundary layer for perfect and real gas flow. An- region lead to a need for a lower value of critical radii of the blunt-
other valuable remark concers the equivalent radius; it is worth ness of real gas flow conditions. The present investigations also
noting that a streamwise favorable pressure gradient in conjunc- revealed that the special gas dynamic features at the critical radii
tion with a thick HEL is highly desirable to reduce the separation are an independent type of gas flow. Further, the critical radii
length absolutely for separation control. It was noted that the real predicted for perfect gas flow would have the same functionality
gas effects reduce the extent of separation owing to SWBLI in in a real gas flow situation as well. Thus, only perfect gas sim-
comparison with perfect gas flow. Thus, the reduced requirements ulations are sufficient to predict the critical radii to ensure sepa-
of a high entropy-layer thickness and upstream overpressure ration control.
Liou, M., and S. C. J. Steffen. 1993. “A new flux splitting scheme.” J. Com- tion angle on laminar boundary layer separation in hypersonic flow.
put. Phys. 107 (1): 23–39. https://doi.org/10.1006/jcph.1993.1122. NASA TN D-3290. Springfield, VA: National Aeronautics And Space
Marini, M. 2001. “Analysis of hypersonic compression ramp laminar flows Administration.
under sharp leading edge conditions.” Aerosp. Sci. Technol. 5 (4): Venkatakrishnan, V. 1993. “On the accuracy of limiters and convergence to
257–271. https://doi.org/10.1016/S1270-9638(01)01109-9. steady state solutions.” In Proc., 31st Aerospace Sciences Meeting,
Needham, D. A., and J. L. Stollery. 1966. “Boundary-layer separation in Aerospace Sciences Meetings. Reno, NV: American Institute of
hypersonic flow.” In Proc., 3rd and 4th AIAA Aerospace Sciences Meeting, Aeronautics and Astronautics.
455. Reston, VA: American Institute of Aeronautics and Astronautics. Wieting, A. R. 1987. Experimental study of shock wave interference
Neuenhahn, T., and H. Olivier. 2009. “Numerical study of wall tempera- heating on a cylindrical leading edge. NASA TM A00484. Honolulu:
ture and entropy layer effects on transitional double wedge shock American Institute of Aeronautics and Astronautics.