Available online at www.sciencedirect.

com

Procedia Engineering 31 (2012) 762 768

International Conference on Advances in Computational Modeling and Simulation

Studies of Compression Corner Flowfields Using THREE

Turbulent Models
Changxian Zhanga*, Jinglei Xua, Lin Gaoa, Ge Gaoa
a

National Key Lab. of Science and Technology on Aero-Engine Aero-thermodynamics, School of Jet Propulsion,
BUAA,Beijing,100191,China

Abstract
This paper presents the numerical investigations on shock-wave/turbulent boundary-layer interaction flow over a
compression corner using SA, SST, and WJ EASM k- turbulence models, and compared with experiment data. The
applicability of the three different turbulence models has been estimated comprehensively for this kind of the complex
flow. The results indicate that the wall pressure distribution, the friction coefficients and the positions of separation and
attachment with the three models agree well with the experimental data.

2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Kunming
University of Science and Technology Open access under CC BY-NC-ND license.
Keywords: supersonic flow; turbulence models; shock; turbulent boundary layers; separation fluid ;

* Corresponding author. Tel.: +86-10-82339943

E-mail address: [email protected].

1877-7058 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2012.01.1099

763

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

1. Introduction
Shockwave turbulent boundary layer interaction induced by two-dimensional compression corner in the
supersonic flows is classical benchmark for evaluating turbulence model performances. Configurations
same as the compression corner, such as aircrafts inlet, transonic compressor and control surface between
the airfoils and fuselages produce a large number shockwave structures. Separation flows resulted in
shockwave turbulent boundary layer interaction give rise to turbulence complexity and affect aircraft
configuration design, aerodynamic analysis, propulsive efficiency and thermal load computations.
Experimental and numerical studies on the compression corner in supersonic or hypersonic flow have been
implemented since the middle of the last century. Settles et. al[1-3] carried out a series of experiments in
supersonic wind tunnel in Princeton University. Attachment flow, incipient separation flow and large
separation flow induced by interaction between shockwave and turbulent boundary layer were studied at
high Reynolds numbers. Elfstrom, G.M. et. Al [4] implemented experiment of compression corner flow at
the Mach number of 9.22 in the No.2 supersonic wind tunnel in Imperial College. The effects of wall
temperature and attack angle were investigated in their research works. Fernholz et. Al [5] provided
numerous experimental data which can supplement theoretical study on the shockwave turbulent boundary
layer interaction flow. Experiments cost enormous time and money, and can not solve extremely complex
turbulent flows in reality. With the developments of the computer science and Computational Fluid
Dynamics, numerical study of the shockwave turbulent boundary layer interaction becomes easier to
implement. Furthermore, numerical simulation has lower cost compared with the experiment study. Dolling
et. Al [6] summarized conditions of development in compression corner flow based on the Reynolds
averaged equations, and drew a conclusion that most of turbulence models can not give rise to accurate
results comprehensively. Yan et. Al [7] did not predict separation region precisely in the numerical
simulation of two-dimensional compression corner flow using several eddy viscosity models. This work
calculated the two-dimensional compression corner flow using WJ EASM k- [8] model, Spalart-Allmaras
[9] model and Shear Stress Transport [10] model.)The performances of these three turbulent models in the
calculation are compared and analyzed in the present research work.
2. Governing Equations and Turbulence Models
2.1. Governing Equations
The governing equations are the unsteady compressible Faver-filtered Navier-Stocks equations, written
in nondimensional variables and convention forms as follow:
wU
w

 U ui 
wt wxi

(1)

w
w
 Uui    Uui u j  pG ij 
wt
wx j
w( U E ) w

(( U E  p)u j )
wt
wx j

w
(W ij  W ijT )
wx j
w
((W ij  W ijT )ui  q j  qTj )
wx j

2
3

P (ui , j  u j ,i  uk , k G ij )

W ijT

Pt ui , j  u j ,i  uk , k G ij  U kG ij

2
3

(3)
(4)

W ij

(2)

2
3

(5)

764

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

Where, W7ij is the Reynolds stress tensor and Pt is determined by different modeling method in eddy
viscosity models.
2.2. Turbulent Models
The RANS (Reynolds-averaged Navier-Stokes equations) is the main method for numerical simulating
turbulent flow at present because of lower request for computer hardware, easier achievement and better
robustness. Eddy viscosity model and Reynolds stress model are created on the basis of different treatments
with the Reynolds stress in the Navier-Stocks equations. Linear viscosity model based on the Boussinesq
theory is easy to be carried out in procedure and possesses good robustness. This model is suitable for shear
flow in which shear stress is dominant, but is inappropriate in the separation flow, large reverse pressure
gradient flow and secondary flow. The SA and SST models used in the present work belong to linear eddy
viscosity model. On the other hand, Reynolds stress model built transport equation for Reynolds stress,
which describe complex turbulent flow more precisely and more actually, however, it is more expensive
and need rigid numerical scheme in order to avoiding divergence. Therefore, there are fewer applications in
the numerical study of engineering flows.
Actually researchers focus on merits containing easy-to-use in linear eddy viscosity model and better
simulating capability in Reynolds stress model. Explicit algebraic stress model in which lesser differential
terms in the Reynolds stress transport equations is ignored and local equilibrium hypothesis is applied made
a success to a certain extent. This model reserves pressure-strain ratio correlation term same as secondary
moment transport equations, and constructs coefficients based on averaged strain rate tensor and rotation
velocity invariant. Furthermore, second-order term describes anisotropic turbulence features. WJ EASM k model used in present research work is one of this type models which is reasonable to employ in
turbulent flow involving large reverse pressure gradient and anisotropy. This paper presents the
performances of WJ EASM k- Podel, SA model and SST model in the two-dimensional compression
corner flow.
2.3. Numerical Method
We use finite-volume approach to discretize the time-dependent governing equations which are written
as a strong conservation form. The computational algorithm is the explicit, upwind-difference scheme with
the convection terms discretized by Roes flux-difference-splitting technique [13], and with diffusion terms
evaluated by the second-order central-difference scheme. 086&/0RQRWRQLF8SVWUHDP6FKHPHIRU
Conversation Law) extrapolation technique is employed to maintain accuracy, monotonicity and robustness
in cases of shockwave discontinuities. We determine the steady state solution of the differential equations
for each computational cell by using five-stage Runge-Kutta marching algorithm, in which the Runge-Kutta
coefficients are set equal to 1/4, 1/6, 3/8 , 1/2, and 1/1, respectively.
3. Computational Results and Analysis
3.1. Computational model

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

Present numerical simulation is based on an experiment of two-dimensional 24 compression corner

carried out by Settles et. al in the supersonic tunnel at Princeton University. This experiment contains
abundant and complex flow behaviors in which characteristic parameters are summarized as follows:
incoming flow Mach number M=2.84, total pressure P0=6.89h105Pa, total temperature T0=262k, wall
temperature Tw=276k, characteristic boundary layer thickness G=23mm, and unit Reynolds number Re=6.3
h107. A 180h110 nonuniform Cartesian grid generated algebraically is employed in the calculations. The
first grid interval perpendicular to the wall surface is 2.0h10-7m which guarantees y+ less than 1.0.
Compression waves caused by supersonic incoming air currents impact to the compression ramp and
assemble to shockwaves. Boundary layer develops much thicker, and means velocity is reduced behind the
shockwave. Furthermore, adverse pressure gradient becomes much greater resulted in shockwave effect.
Distinct separation flow behind the shockwave is appearing due to the large attack angle of 24. Flow
features can not spread to upperstream of supersonic flow field. On the contrary, large adverse pressure
transfers flow information to upperstream in subsonic region of boundary layer. In this way, pressure
gradient and boundary layer thickness become much greater, and shockwave located a new place to
maintain equilibrium. Air flow backwards and form new compression region after reattachment. Fig.1 is a
flow sketch map of two-dimensional compression corner flow under the supersonic incoming parameters.
In addition, experimental values come from reference [4].

Fig.1 The flow field of shock wave/turbulence boundary layer interaction

3.2. Computational Results and Analysis

Some flow features such as separation, reattachment, shockwave location and wall surface friction are
mostly concerned in the research work of shockwave turbulent boundary layer interaction. Thus,
comparison between the computational results of three turbulence models and experimental values are
shown as follows. By this means, the performances of turbulence models simulating shockwave turbulent
boundary layer interaction will be illustrated.
3.2.1. Flow Separation Region
Three models can obtain separation region in the present numerical simulation which illustrated these
models have capability to predict shockwave turbulent boundary layer interaction flow in a certain extent.
The locations of separation and reattachment are shown in the table 1.

765

766

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

Table 1 Separation locations and length
Separation location

Reattachment location

Separation length

Exp.

-1.47

0.48

1.95

SA

-1.92

1.97

3.89

SST

-2.71

1.77

4.48

WJ

-2.10

0.98

3.08

Fig.2 shows separation region around the corner. Three turbulence models represent common features of
predicting separation point too early and predicting reattachment fall late. Comparatively speaking, SST is
the worst to calculate separation point in three models and SA gives rise to computational result according
with experimental value. For reattachment point, WJ is better than other two model. This three turbulence
models are widely used in present fluid engineering field, but their performances for describing separation
flow induced by shockwave turbulent boundary layer interaction are not satisfactory. The main reason is
that the anisotropic turbulence viscosity and Reynolds stresses are not taken account. In a word, WJ is the
best one of three turbulence models at predicting the locations of separation, reattachment and length of
separation region.
3.2.2. Surface Pressure Distribution
The disagreement of surface pressure distribution calculated by SST and experimental values is most
visible compared with two other models. Pressure rising location illustrates this appearance in Fig.3. WJ
predicts pressure influence location in according with experiment exactly. Simulation results of WJ and SA
behind apex of corner are very similar and better than data coming from SST. Frankly speaking, pressure
rising values of three turbulence model at ramp surface are less than experimental data given in reference
[4]. Comprehensively analyzed, WJ and SA obtain the relative accurate shockwave location compared with
SST. The reason is that advanced turbulence model such as SST (in comparison with k-, k-) always
underestimates turbulent viscosity and causes pressure recovery later.

-2

-1

(a)SA

0
-2

-1

(b)SST
Fig.2 Streamlines around separated area

-2

-1

(c)WJ

767

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

SST
SA
WJ EASM K-w
EXP

0.003

Cf

Pwall/Pf

0.002

0.001
2

SST
SA
WJ EASM k-Z
exp

1
-6

-4

-2

Fig. 3 Pressure distributions on the ramp

-8

-6

-4

-2

Fig.4 Skin friction distributions on ramp

3.2.3. Wall Surface Friction Coefficients Distribution

Fig.4 shows the comparison of wall surface friction resulted in three turbulence models and
experimental values. SA presents better performance in simulating separation region before the apex of
corner than the others. WJ predicts variation trends of the wall friction coefficients exactly although the
values are different from experiment. The performance of SST in calculating wall friction coefficients is the
worst among present three models. Researchers did not explain wall friction coefficients around the
backflow region in their experiment, so the comparison between the numerical simulation and experiment
can not be carried out in this region. It is certain that calculated results of wall friction based on these three
models are very similar in the separation region before the apex of corner. WJ gives rise to minimum value
of friction coefficient and makes friction variation trend go upward rapidly, which is according with
experimental data proximately behind the reattachment point. The computational results of wall friction
coefficients based on SA and SST are not satisfactory especially behind the reattachment point.
4. Conclusions
We studied shockwave turbulent boundary layer interaction flow induced by two-dimensional
compression corner, which contains complex turbulent separation flow, large adverse pressure gradient and
strong anisotropic turbulent viscosity in present paper. Three turbulence models exhibit different
performances in predicting this type flow. Conclusions are summarized as follows:
(1) Separation point obtained by SA is according with one of WJ. But, SST predicts separation location
too early. Three models did not calculate reattachment point exactly. Comparatively speaking, WJ gives
rise to the best result among these turbulence models, SST the second, and SA the third.
(2) Shockwave locations predicted by SA and WJ satisfy experimental value perfectly. SST captures
shockwave location shifting to upstream excessively.
(3) Before the separation point, wall friction coefficients given by SA are slightly greater than
experimental data. Calculated results by using WJ and SST are according with experiment. Behind the
reattachment point, we got the closest wall friction coefficients to experimental values using WJ, but SA
shows the worst performance.

768

Changxian Zhang et al. / Procedia Engineering 31 (2012) 762 768

Three turbulence models used in present paper have capability to simulate complex separation flow
induced by shockwave turbulent boundary layer interaction to a certain extent. SA is more suitable for
separation region before apex of corner, but, displays very bad performance behind the reattachment point.
WJ describes flow features well either separation or reattachment flow. SST did not obtain satisfactory
results neither separation nor reattachment flow. We can draw a conclusion based on present research that
WJ is prefer to SA and SST in calculating separation flow resulted in shockwave turbulent boundary layer
interaction. For attachment flow, we can get good numerical simulation results using SA model.
References
[1] Settles G S, Fitzpatrick T J, Bogdonoff S M. Detailed study of attached and separated compression corner flowfields in high
Reynolds number Supersonic Flow[J]. AIAA Journal, 1979,17(6):579-585.
[2] Settles G. S., An Experimental Study of Compressible Turbulent Boundary Layer Separation at High Reynolds Number, Ph. D.
Dissertation, Princeton University, Princeton, N. J., Sept. 1975.
[3] Settles G S, Dodson L J. Hypersonic shock/boundary-layer interaction database.[R]. NASA-CR-177577,1991.
[4] Elfstrom G. M. Heat Transfer from Hypersonic Turbulent Flow at a Wedge Compression Corner[J]. Journal of Fluid
Mechanics, 1972 53(1), 113-127.
[5] Fernholz, H. H.; Finley, P. J.; Dussauge, J. P.; Smits, A. J.; Reshotko, E., A survey of measurements and measuring techniques
in rapidly distorted compressible turbulent boundary layers, AGARDograph 315,1989.
[6] Dolling D. High-speed turbulent separated flows: consistency of mathematical models and flow physics. AIAA[J]
1998;36:72532.
[7] Yan H, Knight D, Zheltovodov A. Large eddy simulation of supersonic compression corner using ENO scheme. Third AFOSR
International Conference on DNS and LES, Arlington, TX, August 2001.
[8] Stefan Wallin and Arne V. Johansson, An explicit algebraic Reynolds stress model for incompressible and compressible
turbulent flows. Journal of Fluid Mechanics, 403, pp 89-132 .
[9] Spalart P R, Allmaras S R. A one-equation turbulence model for aerodynamic flows[R]. AIAA-92-0439, 1992.
[10] Wilcox D C. Reassessment of the scale determining equation for advanced turbulence models. AIAA Journal,1988,26(11):
1299-1310.