See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290091412

Numerical simulation of normal shock wave/turbulent boundary-layer

interaction over a bump surface

Article in Indian Journal of Engineering and Materials Sciences · October 2003

CITATIONS READS

3 4

1 author:

R. C. Mehta
Noorul Islam University
108 PUBLICATIONS 706 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reentry aerodynamics, Venting, Inverse heat conduction problem View project

Drag Reduction for Spike Attached to Blunt-Nosed Body at Mach 6, Experimental investigation on spiked body in hypersonic flow View project

All content following this page was uploaded by R. C. Mehta on 18 December 2020.

The user has requested enhancement of the downloaded file.

Indian Journal of Engineering & Materials Sciences
Vol. 10, October 2003, pp. 353-358

Numerical simulation of normal shock wave/turbulent boundary-layer interaction

over a bump surface
R C Mehta
Aerodynamics Division, Vikram Sarabhai Space Centre, Trivandrum 695022, India
Received 2 December 2002; accepted 16 August 2003
Numerical computation of the normal shock wave/turbulent boundary layer interaction over a bump is discussed here.
The time-dependent, two-d imensional Reynolds averaged compressible Navier-Stokes equations have been solved
employing three-stage Runge-Kutta time-stepping scheme in-conjunction with a finite volume discretization of spatial
coordinates. The closure of these equations is obtained using the Baldwin-Lomax turbulence model. Artificial dissipation
terms are added to the numerical scheme to maintain numerical stability and they contain a blend of second and fourth
differences of the state vector with an appropriate pressure switch that detects the presence of strong pressure gradients. The
formation of the A.-shock is simulated by setting downstream pressure condition using an adjustable aerofoil. The numerical
solution captures the A.-shock associated with the separated boundary layer inside the channel, which is very sensitive to the
exit condition of the flow. Compari sons have been made with availab le experimental data such as interferogram and wall
pressure distributions. They are found to be in good agreement.

The fluid dynamics of normal shock wave/turbulent The transonic channel flow was experimentally
boundary layer interaction problem has become more investigated by Delery 5 and Liu and Squire6 . An arc
and more important in many practical situations. was placed on the channel floor to produce
Examples include the flow around transonic aerofoils, convergent-divergent geometry. A bow-shaped shock
in diffusers of centrifugal compressors or transonic occurred near the end of the constriction; this
flow over the bulbous heat shield of a satellite launch interacted with boundary layer to form a A shape
l
vehicle . Viscous effects play a major role in shock and caused local separation. An investigation of
transonic flow s where they can strongly affect the flow topology based on the visualizations together
whole flowfield . This problem is especially important with the measurements of static wall pressure, and
in supercritical flows, where the strong viscous schlieren pictures provides an explanation of the
interaction occurring near the shock roo t entails a observed interaction phenomena. Numerical analysis
rapid thickening of the boundary layer and can lead to is required in order to supplement the experimental
its separation if the shock is strong enough. This can results . Numerical solution of the full Navier-Stokes
produce dramatic changes in the aerodynamic equations is the most appropriate procedure to predict
characteristics of an aircraft or other aerodynamic the complex flow field structure formed by shock
devices, which , if not handled properly can cause separated flow. Considerable advances have been
catastrophic problems in flight safety . Since the made in the development of efficient and accurate
phenomenon is so important, a large number of numerical codes which solve the Navier-Stokes
experimental investigations have been carried out equations. However, the numerous and systematic
since the late 1940s as summarised by Green 2 . applications made in transonic flows involving shock
However, owing to the complexity of the fluid wave/turbulent boundary layer interaction have
dynamics problem, our understanding is still far from frequently led to poor capture of A-shock with shock
complete and further research is required. The associated separated flow region. Rodi and Srinivas7
interaction of shock wave on a curved surface have solved numerically normal shock wave/turbulent
produces a turbulent boundary layer at transonic boundary layer interaction over a curved surface using
Mach number. An experimental and topological Baldwin-Lomax model and two equation turbulence
investigation of normal shock wave/turbulent model. Comparing the numerical results with the
boundary-layer on a convex plate in a narrow wind experimental data they have found that both the
tunnel has been preformed by Doeffer, and Dallmann3 turbulence models give identical surface pressure
and Bohning and Zierep 4 . distribution over the bump.
354 INDIAN 1. ENG. MATER. SCI., OCTOBER 2003

A complex flow field is 'Observed during are the state vector Wand inviscid flux vectors F and
experimental studies over the bump configuration due G. The viscous flux vectors Rand S are:
to f'Onnation of a transonic l-shock, its interaction
with the turbulent boundary layer, and flow separation
in the circular arc region. The above literature survey
o o
reveals iIhat experimental and numerical efforts have
been made ItO analyse the complex fluid mechanics R= , s= (j yy
'Over l!he bump. Most of the experimental
investigations were aimed to find oul: the back
pressure for the fOimatioll1 of the l-shock and shock-
induced flow separation.
The present: method employs a computational fluid where O"xx, O"yy and Lxy are components of the stress
dynamics approach to analyse such a complex fluid vector, qx and qy are components of the heat flux
dynamics problem of nonnaI shock wave separated vector, u and v are axial and normal velocity
brrbulenl: boundary layer. The time~dependent, two- components in x and y directions, respectively, e is
dimensi'Onal, compressible, tuJbulent Reynolds- total energy per unit volume. Thus, the viscous terms
averaged Navier-Stokes equations are solved using a in the above equations become:
finite volume discretization in conjunction with a
three-stage Runge-Kutta time-stepping scheme. The
closure 'Of ilhese equation is achieved using the
(J
2
==--(11+11) - -2-
3 r- r-, ay ax
[av au]
Baldwin-Lomax turbulence model. Artificial
dissipation terms are addedro the numerical scheme
ItO maint:aiin numerical stability and they contain a
blend 'Of second and fourth differences of the state
(J
yy
2
=_-(11+11)
3 r- r-,
- [auax -2-
av]
ay
vector with an appropriate pressure switch that detects
the present 'Of strong pressure gradients. The main
purpose of the present numerical simulation is to au av]
study the fonnation of the l-shock. The back pressure T xy=T yx=(I1+ 11,) [ay + ax
condition is fumd from the experimental results. A
,comparis'On is made with the experimental results
'Obtained from the wind ltiwmel data of Liu and
Sqoire
16 qx=-CP -.11 + - 1l,)%;T
• ( Pr Pr, ax
Tlteore1ical
Gov~tequalilmS
q v=-CP - + -11 11, )aT
The time dependent:, two-dimensional,
compressible, ru.rbul.ent, Reynolds-averaged Navier-
- ( Pr Pr, ay
Stokes equations can be written :as:
where Cp is specific heat at constant pressure, Pr and
dU dF a:G Pr( are the molecular and turbulent Prandtl numbers
... (1)
at i ax 1--=0
ay are assumed to take constant value of 0.72 and 0.9
respectively. The coefficient of molecular viscosity Jl
is calculated according to Sutherland's law as:
where,
1.4xlO1:T 3/2]
p pu {Jv 11 (T +110)
2

I' Pu +p
(JU pUll
u= . G= In the present preliminary investigation, a two-layer
pv F= puv Pv 2 +p
algebraic model8 has been used as a turbulent closure.
{Je I
(pe+p)u (pe+p)v This model, which utilizes vorticity distribution to
 MEHTA: NUMERICAL SIMULATION OF NORMAL SHOCK WAVB 355

determine the scale lengths, has been used previously Numerical algorithm
and is reported to offer acceptable engineering The flowfield code employs a finite volume
solutions9 ,1O. descretization technique. The spatial and temporal
In the inner region, the eddy viscosity is given by: terms are decoupled using the method of lines. Using
a finite volume approach. the governing equations are,
discretized in space starting from an integral
formulation and without any intermediate mapping. A
cell centre scheme is used to store the flow variables.
where (0 is the vorticity function, L is the distance On each cell face. the convective and diffusive fluxes
normal to the surface, and D is the Van Driest's are calculated after computing the necessary flow
damping factor is given by: quantities at the face centre. These quantities are
obtained by a simple averngjng of adjacent cell-centre
values of dependent variables II . The: numerical
procedure reduces to central differencing on a
rectangular and smooth grid The discretized solution
to the governing equations results in a set of volnme-
The subscript w indicates values at the wall. In the averaged state variables of mass. momentum. and
outer region, the eddy viscosity is determined by: energy. which are in balance with their area-averaged
fluxes (inviscid and viscous) across the cell faces. The
entire spatial discretization scheme is second order
accurate. In viscous calculations. dissipating
properties are present 'due to diffusive terms. Away
The coefficient F w is calculated as the minimum of from the shear layer regions. the physical diffusion is
the following two values. generally not sufficient to prevent the odd-even point
decoupJing of centred schemes. Thus, to maintain
numerical stability and to prevent oscillations near
shocks OJ' stagnation points. artificial dissipation
terms 12 are included as a blend of Laplacian and
The scale length 4nax is the maximum value of L biharmonic operntOJ' in a manner analogous to the
I I)
when the function F (= L.D(O attains as maximum second and fourth differences.
F max' The Klebanoff intermittency correction factor is Time-marching scheme-The spatial discretization
given by: described above reduces the governing flow equations
to semidiscrete ordinary differential equations. The
integration is performed using an efficient multistage
schemel2 • The following three-stage Runge-Kutta
time-stepping method is employed for the present
work:
The effective viscosity is then given by:

u(l)=u· -O.6At(Res€O)-D(O»

U(2)=U"-O.6At(Res{l)-D(O})
The temperature is related to pressure and density
by the perfect gas equation of state as: u(3)=U· -l.OAt(Resf2)-D«O)

1 P( u 2 +v 2 )]
p=(y-1) [ e-'2 ... (2)

where n is the cmrent time le.veI. n+1 is the new time

The ratio of the specific heats r was assumed level. and residual Res is the sum of the inviscid and
constant and equal to 1.4. viscous fluxes. In order to minimize the computation
356 INDIAN J. ENG. MATER. SCI., OCTOBER 2003

time, the expensive evaluation of the artificial

dissipation terms D is can-ied out only at the first
intermediate stage (0) and then frozen for the
subsequent stages. A conservative choice of the
Courant-Friedritchs-Lewy number (1.4) was made to
achieve a stable numerical solution. A local time-step ... (4)
was used to obtain steady state solution.

Initial and boundary conditions-The above

numerical procedure is employed to simulate flow i=1,2, .. ....nx
field inside a two-dimensional transonic channe1 6 . A j=1 ,2, .... ..mY
schematic sketch of the test section of the blow down
wind-tunnel with an adjustable aerofoil is depicted in
F i~ . 1: The fOlmation of the A.-shock is obtained by
where Yi,w and Yi,lII are wall and middle surface points,
adJustmg the angle of incidence of the aerofoil which respectively, fJ is stretching factor. nx and my are total
gives con-esponding back pressure in the downstream number of grid points in x and Y directions,
of the channel. The top wall is flat. On the bottom respectively. The m is referred to as the mid point of
wall, a bump of circular arc of 73 mm radius with a the channel. The above expression is used again to
chord length of 80 mm is mounted. The settling generate grid from top surface to the middle of the
channel and then they are joined together. These
chamber stagnation pressure is 1.481 x 10 5 Pa and
stretched grids are generated in an orderly manner.
stagnation temperature, T= is 290 K. The peak Mach
The present numerical analysis is carried out on
number is calculated from the surface static pressure
132x52 grid points. The grid stretching factor is
and . upstream stagnation pressure using isentropic
selected as 5. This spatial resolution is adequate for
relatIOn. These test conditions are used to set the
fine resolution of the steady flow fie ld features. The
boundary conditions and computational region in the
minimum spacing for the fine mesh is dependent upon
numerical algorithm. The stagnation pressure and
the boundary layer thickness. The finer mesh near the
temperature are used as the inlet boundary condition
wall helps to resolve the viscous effects. The coarse-
at the upstream of the channel. A stagnation pressure
mesh helps reducing the computer time. Fig. 2 shows
ratio of 0.78 is used as the outflow condition at the
the grid arrangement used in the numerical
downstream of the channel. At the surface, a pseudo-
computation. The convergence criterion is based on
cell approach is employed to specify no-slip, zero
the difference in density values, p, at any grid point
pressure gradient, and adiabatic condition.
between two successive iterations, that is,
'+'
Ip -p11<10-5
_ h
, were l IS
' th' . mdex.
e IteratIon .
Computational grid
One of the controlling factors for the numerical
Results and Discussion
simulation is the proper grid arrangement. In order to Flow characteristics
initiate the numerical simulation of flow alon ba the Figs 3 and 4a show the enlarged view of the
body, the physical space is discretized into computed velocity vector and Mach number contours.
nonuniform spaced grid points . These body oriented An extended separation bubble formed due to the wall
grids are generated using a finite element method in
conjunction with homotopy scheme '3 , ' 4. The typical
co mputational space is defi ned by a number of grid
points in the Cartesian coordinate system. Using these
~L--~
To=290K
TOP _ , \_ _
-
..:2._ __ __ _

surface points as the reference nodes, the normal q 625

Fb =1-48 xl0sPa
coordinate is then described by exponentially .----'--~~~~---.;
stretched field points, (Xi.j, Yi) extending outwards up
to an outer computational boundary. The stretching of 80
grid points from bottom of the channel to the middle AGdimenssions oW in mm
of the channel is obtained usin ba the followin ba
expression: Fig. 1 - Schematic sketch of two-dimensional bump
 MEHTA: NUMERICAL SIMULATION OF NORIvIAL SHOC K WAVE 357

- .093
.093 - .184
.134 - .275
.275 - .336
.45i - .548
.548 - .639
.639 - .730
.' I .730 - .821
.821 - .912
.9'12· 1.003
1.003 - 1.094
1094-1 .185
1.185 - 1.276
1.276-

Fig. 2 - Computational grid

Fig. 4a - Mach contours

Fig. 4b - Interferogram of flow field inside the channel

--. -- - -
.,f
: ; ~ :

'qH ~ i
Fig. 3 - Velncily vector plots

curvature effect of the bump can be observed in Eg.

4a. The normal shock is strong enough to separate the
boundary layer under the shock foot. This is referred
to as shock induced flow separation . The shock is
bifurcated a fi_-shock formed. The reattachment
process turns the supersonic flow parallel to the flat
floor and compresses the flow to subsonic as seen in
the Mach contour and vector plots . There is no tlow
Fig. 5 - Pressure variation o',Ier the bottom wall
separation found on the top surface of the channel.
Calculated Mach contour plot is compared with
interferogram Fig. 4b of Liu and Squire6 and shows ongm of the x-coordinate is set at the mid-chord
that the A shock is well captured. Figs 3 and 4a position of the circular arc model and normalized with
delineate all the flow characteristics of the shock the half of the cord length. The normal shock is strong
induced separated flow is obtained with present enough to separate boundary layer that is evidenced
numerical simulation. by the plateau of the wall pressure distribution in Fig.
5. Tne plateau cOlTesponds to the main part of the
Surl'ace pressure distribution separation bubble. The plateau is half of the cord
The surface static pressure distribution on the top length. Figs 5 and 6 compare the calculations for the
and on the bump surface normalized by the settling Liu and Squire experimental measurements 6 . -- The
chamber total pressure is shown in Figs 5 and 6. The calculated results are not in good agreement with the
358 INDIAN 1. ENG. MATER . SCI., OCTOBER 2003

090 interaction can be used as a guide to improve the

- Numerical
modelling of two-dimensional tlows. On the other
o.es L o ExpE"r!ment aI16l

hand, they can be helpful to test three-dimensional

I
J
I Navier-Stokes codes.
--
0.60
°0
o
0... 0.75 0 0 00
o
"6:- 00 Nomenclature
o
070 o
o c cord length
o
065 i e specific energy

0.60
-
-2
--,-----,---,--"----,-,
., 1
J F
G
M
vector of x-direction fluxes
vector of y-direction fluxes
Mach number
x/0.5e
Po stagnation pressure
Pr Prandtl number
Fig. 6 - Pressure variation over the top wall
p static pressure
q heat flux
experiments in the region near the A-shock. However, T temperature
the overall pressure distribution compares well with t time
the experimental results 6 . The pressure distribution on U conservative variables in vector form
11, V vel ocity components
the bottom wall of the channel and the density x, Y Cartesian coordinates
contours con films position of the A-shock and P density
recirculating zone. ,The existing turbulence models fl molecu lar viscosity
can be refined further to predict more accurately subscripts
inner
sepGration zone of the complex shock wave/turbulent o outer
boundary layer interaction processes. The Baldwin- t turbulent
Lomax model predicts higher value of pressu re on the w wall
top surface . The calculated and measured values of
pressure distributions are in good agreement. The References
press ure profiles indicate that the Baldwin-Lomax 1 Mehta R C, j Spncecraft Rockets, 34 (1997) 561 -564.
2 Green 1 E. Prog Aerospuce, 11 (1970) 235.
model under predicts the separation region.
3 Doerffer P & Dallmann U, MAA J, 27 (9) (1989) 1206- 1212 .
Therefore, further improvement is required to include 4 Bohning R & Zierep J, Normal shock. lllrbll!ent boundary-
the non-equilibrium effects in the algebraic turbulence layer interaCTion at a curved wall, C()/ilpuTation ot' viscous-
model. inviscid IIlteracrion. AGARD CP·29! (1980) pp. 17.1-17.8.
5 Delery], Investigation of strOilg/ turbulent bOLli/dory-layer
interaction ill 2D transonic flows with emphasis on turin. len!
Conclusions phenumena. AIAA Paper 81-1245 (1981).
A numerical experi ment of two-dimensional G Lu i X & Squire L C, J Fluid ivlech, 187 ( 1988) 467-486.
turbulent viscous flow inside a channel is performed 7 Rorli \V & Srinivas K, Z. FllIgwiss Weltralltllforsch, 13
(1989) 101-119.
by employing a three-stage Runge-Kutta time- S Baldwin B S & Lomax H, Thin layer approximation and
stepping scheme. Turbulence closure is achieved algebraic model fo r separated TUrbulent pow, AIAA P3.per
using the Baldwin-Lomux turbulence model. 78-257 (l9'J1l).
Artificial dissipation terms are added to the numerical 9 Mehta R C, COIl1{'lIf Fluids. 30 (2001) 689-709.
10 Puroh it S C, j Spacecr(lft RockeTS, 23 ( 19)36) 590-596.
scheme to maintain numerical stability and they are
11 Peyret R & Viviand H, Computational merhodsfor .fluidflow
contained as a blend of second and fourth differences (Springer-Verlag, New York) , 1993.
of the state vector with an appropriate pressure switch 12 Jameson A, Schmidt W & Turkel E, Numerical solution of
thdt detects the present of strong pressure gradients . Euler equatio/ls by finit e volume methods using RlIIlge- Kla ta
The flow fieid visuaiization of the :c-shock and lime stepping schemes, AIAA P3per g 1- 1259 (198 1).
13 Mehta R C, J Spacecrqfi Rockets, 35 (1998) 102- 105.
separated tlow region helps in a systematic 14 Mehta R C, A quasi-three-dimensional automatic grid
understanding of the flow structure. This informaLion generation method, Proc 25'h Nat Inl Conf Fiu id M ech, IlT
on the normal shock wave/turbulent boundary layer Delhi (1998) 89-93 .

View publication stats