Tamkang Journal of Science and Engineering, Vol. 12, No. 3, pp.

223-229 (2009) 223

CFD Analysis and Wind Tunnel Experiment on a

Typical Launch Vehicle Model
Selvi Rajan. S1*, Santhoshkumar. M2, Lakshmanan. N3,
Nadaraja Pillai. S4 and Paramasivam. M5
1
Scientist, Structural Engineering Research Centre, CSIR Campus,
Taramani, Chennai – 600 113, Tamil Nadu, India
2
Structural Engineer, Vibromech Engineers & Services ltd.,
Chennai, Tamil Nadu, India
3
Director, Structural Engineering Research Centre, CSIR Campus,
Taramani, Chennai – 600 113, Tamil Nadu, India
4
PhD Research Scholar, Wind Engineering Research Centre, Tokyo Polytechnic University,
Japan 243-0297
5
PDF Research Scholar, Wind Engineering Research Centre, Tokyo Polytechnic University,
Japan 243-0297

Abstract
In order to understand the physical phenomena of the wind flow over the typical launch vehicle,
the flow was simulated using both Wind tunnel and Computational Fluid Dynamics (CFD). In the
present study, tests were conducted on a 1:50 scaled model of a launch vehicle. The model was
subjected to two wind conditions, wind flow normal to the shorter plan dimension q = 0°, where the
three main cylinders of the model were one behind the other and wind flow normal to the longer plan
dimension, q = 90°, where all the three main cylinders of the vehicle are subjected to direct wind
pressure in the windward direction. Based on the CFD studies, the flow pattern and the force
coefficients were derived. To validate these results, wind tunnel tests were carried out on a 1:50 scaled
rigid and light-weight models respectively, for obtaining path lines and force coefficients. Results on
streamlines obtained based on CFD simulation and wind tunnel experiments compared very well. The
force coefficients in both directions were evaluated from CFD results showed good agreement with the
corresponding measured values based on wind tunnel experiment.

Key Words: CFD Simulation, Boundary Layer Wind Tunnel, Launch Vehicle, Flow Visualization

1. Introduction corresponding large changes in drag. In current competi-

tive environment there is a need to design the satellite
There are many simulation methods and different launch vehicles which will be capable of launching even
models are available to access the application of CFD under some unforeseen changes in the atmospheric con-
[1,2] in the field of aerodynamics and wind Engineering. ditions. One such situation was aroused when PSLV-C5
Murakami et al. [1,3] show such a simulation over the launched at the time of heavy rain from Sriharikota, In-
surface mounted cubic model with the k-e and LES mo- dia in Oct. 2003 [4]. There is always a possibility of
dels. Many studies show that the critical shapes can have storms, cyclones and tornados near the launching station
significant unsteady effects. Different shapes can cause during the designed period of launch. Before launching,
sudden changes in the size and structure of the wake and the Mobile Service Tower (MST) is taken away from the
launch pad. The Umbilical tower that is fully exposed to
*Corresponding author. E-mail: [email protected] atmospheric wind supports the launch vehicle. It be-
224 Selvi Rajan. S et al.

comes necessary to test the launch vehicles to atmo- powders and tufts. Force data may be taken simulta-
spheric wind loads under simulated Atmospheric Boun- neously with each of these flow visualisation techniques
dary Layer (ABL) flow. Although sufficient studies on [5]. Weinstein [6] used flow visualisation technique to
the flow over some critical bodies are available, there are examine rocket sled flow fields and to obtain the aerody-
not many studies on the physical phenomena of flow namic flow field around aircraft in flight. Kompenhans
around a launch vehicle for the problem of wind in hori- et al. [7] have used Particle image velocimetry (PIV) to
zontal direction. Airflow around the launch vehicle mo- record the complete flow velocity field in a plane of the
del is complicated since the Boosters, Strapons are clo- flow to obtain information about unsteady flow fields.
sely packed with main stage. The cross-section looks like Based on the literature [5], dry ice with hot water based
compound. mixture was tried initially to visualise the flow pattern. It
A typical launch vehicle model of scale 1:50 as shown was observed that when dry ice came in contact with hot
in Figure 1, was considered for the present study. Prior to water, the generated smoke was dense and it settled to
wind tunnel testing a detailed understanding on the flow the floor of the wind tunnel instead of getting transported
behaviour around the launch vehicle was thought to be by the flow. Subsequently, a mixture of glycerin and dis-
essential for proper instrumentation and collection of tilled water had been tried as the source material. The
data. There was no sufficient literature available for flow main advantage of the CFD predictions and its validation
over compound cylinder like structures. In order to have using wind tunnel experiments are discussed in this paper.
some preliminary ideas about the flow behaviour, a CFD
analysis was done as a pre-experimental study using test 2. Numerical Simulations
section inlet conditions. Through flow visualisation, flow
aspects relating to points of separation, re-attachment The computational domain is shown in Figure 2.
and wake flow can be studied besides knowing how the There are possibilities for the wind to approach from all
air moves around an object and/or what forces it exerts the four directions. Since the model is symmetric, the
on the object. Flow visualisation techniques on models computations were done only for two cases. Case_1:
in the low speed wind tunnel are mostly based on smoke, Wind approaching from west direction, i.e., wind flow
normal to the shorter plan dimension q = 0°, where the
three main cylinders are one behind the other. Case_2:
Wind approaching from south direction, i.e., wind flow
normal to the longer plan dimension, q = 90°, where all
the three main cylinders are subjected to direct wind pres-
sure in the windward direction. The computational do-
main was meshed with special meshing feature called
“embedded grids” to increase the cell density near the
model cross-section. Figure 3 shows embedded mesh re-

Figure 1. Typical launch vehicle model. Figure 2. Computational domain.

 CFD Analysis and Wind Tunnel Experiment on a Typical Launch Vehicle Model 225

gion near the model. There were about 32,000 hexahe- cients in two orthogonal directions. The launch vehicle
dral cells in the each computation domain. The numeri- model was fabricated to a geometric scale of 1:50 with
cal analysis used mass and momentum conservation equ- great care to accurately model the cross-section. The
ations in two dimensions for steady state, incompressible light-weight (hollow) model made of wood as shown in
flow. A third order QUICK scheme used for modeling Figure 4 was positioned suitably on the load cell within
the convective terms of the momentum equations. For the test section of the tunnel. Load cell was then cali-
the present study a k-e turbulence model with logarith- brated by applying static loads in both x and y directions
mic turbulent wall functions available in commercial and the corresponding calibration factors are shown in
CFD package STAR-CD was employed [8]. It contains a Figure 5. The model was subjected to various wind speeds
full description of the equations in the code, its numeri- of 2.6, 6.9, 7.4, and 10.3 m/s, with the directions of flow
cal methodology and capabilities. The k-e turbulence
model available in STAR-CD was validated by many au-
thors [8]. Mean velocity and pressure distributions were
further obtained based on the CFD computation.

3. Wind Tunnel Test

To validate CFD results, a single component load

cell was designed and fabricated to obtain force coeffi-

Figure 4. Force model under test in boundary layer wind

Figure 3. Embedded (Discontinuous) Mesh near the model. tunnel.

Figure 5. Calibration charts corresponding to x and y direction, respectively.

226 Selvi Rajan. S et al.

kept normal to the two orthogonal axes of the launch ve- were derived based on force measurement and were
hicle. Corresponding to each of the directions, mean va- computed as 0.25 and 1.2 respectively.
lues of force coefficients were derived. Reference area, The aim of the present study was to predict the quali-
A = pD2/4 (m2) tative as well as quantitative information on velocity,
where D = effective diameter pressure, turbulence parameters like turbulence kinetic
Reference dynamic pressure = ½r U2 (Pascals) energy, turbulence energy dissipation and wake. Figure
where r = density of air, 1.2 kg/m3 6 shows the mean velocity distributions for case 1 and
and U = mean wind velocity (m/s) case 2 respectively. Figures 7 and 8 show the instanta-
Mean wind force per meter, in each direction = Cf × Vf neous streamline pattern for both the cases based on CFD
where CN is the calibration factor corresponding to the simulation in comparison with flow visualization con-
direction (from Figure 5) ducted experimentally on the rigid model of the launch
and VN is the output of the strain gauge vehicle using wind tunnel. The results imply that the vor-
tex wake developing behind a group of cylinders is, to
Coefficient of force = (Cf × Vf) / [(½r U2) × A] (1)
some extent, similar to that of a single bluff body. The
Similarly, the force coefficient for the other direction same has been observed by different author [9]. Hence,
was determined. In addition, a rigid model made of the value of force coefficient for case 2, is almost same as
solid stainless steel was fabricated and positioned in the the value of drag coefficient for a single cylinder, namely
downstream side of the test section of the wind tunnel to 1.2. The value of force coefficient in the other direction
investigate the flow behaviour around the model. The was compared with the available literature on similar
model was subjected to a low wind speed of 1.2 m/s, to
enable to capture the flow images. The flow lines were
visualized using glycerin mixture and were captured us-
ing a CCD camera.

4. Results and Discussions

Pressure data and the force data are expressed in

terms of dimensionless coefficients for the purpose of
comparison. As per Bernoulli’s equation the surface pres-
sure on the body is usually expressed in the form of non-
dimensional pressure coefficient as:

(2)

where, P is the pressure at required point and P0 is the ref-

erence pressure, r is the density of air and U is the free
stream velocity. The pressure coefficients are resolved
and algebraically added to the corresponding direction to
obtain force coefficient. Results on pressure distribution
as obtained from CFD study on the model were further
analysed to obtain force coefficients for both the test
cases of wind flow normal to the shorter plan dimension
0° (case 1) and wind flow normal to the longer plan di-
mension, 90° (case 2) and these values were found to be
0.23 and 1.18. The force coefficients using wind tunnel Figure 6. Mean velocity distribution for case 1 & case 2.
 CFD Analysis and Wind Tunnel Experiment on a Typical Launch Vehicle Model 227

Figure 7. Instantaneous streamlines for case 1.

Figure 8. Instantaneous streamlines for case 2.

type of study to further validate the results [10,11]. The tion provides the critical regions of pressure and helps to
values were respectively reported as 0.38 and 0.30 as fix up the locations for pressure measurement.
base drag coefficients under subsonic wind, which can
be compared with the value as obtained from the wind 5. Conclusion
tunnel test as 0.25.
The figures shown below give the clear idea about CFD simulation was conducted on a 1:50 scaled
free shear and recirculation regions. Also one can realise model of a rocket launching vehicle for two cases of q =
the free shear in the corner regions during experimenta- 0° and q = 90°. Based on the CFD studies, the flow pat-
tion because of the some inadequacies, hence it is quite tern and the force coefficients were derived. To validate
tedious to measure the data at more points. However it these results, wind tunnel tests were carried out using
shows the sufficient information for the free shear visu- 1:50 scaled rigid and light-weight models respectively,
ally. Flow visualization studies conducted using wind for obtaining flow lines and force coefficients. Follow-
tunnel, were also shown for comparison in Figures 7 and ing are the conclusions derived:
8. The wake survey was done for a distance of 10 D be- l The force coefficients for both the test cases of
hind the model at 12 stations for both the cases. The ve- wind flow normal to the longer plan dimension q =
locity was non-dimensionalised and is shown in Figure 9 0° and wind flow normal to the shorter plan dimen-
for case 1 and case 2, respectively. The static pressures sion, q = 90° were found to be 0.23 and 1.18. Cor-
were computed over 280 pressure locations over the mo- responding to each of the directions, the force co-
del and presented in terms of non-dimensionalised pres- efficients were derived based on force measure-
sure coefficients, Cp in Figure 10. This pressure distribu- ment using wind tunnel and were found to be 0.25
228 Selvi Rajan. S et al.

Figure 9. Mean velocity defect profiles on wake centerline corresponding to case1 and case2.

Figure 10. Pressure distribution over the model.

and 1.2 respectively. These values were further measurement in all the directions besides provid-
compared well with the values available in the li- ing better understanding on the flow pattern.
terature. l Such pre-experimental CFD study helps in reduc-
l The predicted flow field around the model was ing the number of repetitive experiments in the
examined using CFD as well as based on wind collection of data besides reducing the cost and
tunnel experiment. There was a very good com- time considerably when compared to the tradi-
parison, as can be seen from Figures 7 and 8. tional approach of wind tunnel experiments, for
l The CFD predictions provide good aid for fixing certain type of typical selected studies. In the case,
up the measurement points for velocity pressure where the levels of turbulence intensities are of
 CFD Analysis and Wind Tunnel Experiment on a Typical Launch Vehicle Model 229

high importance, it is imperative to resort to wind [6] Weinstein, L. M., “Large-Field Schlieren Visualisation
tunnel experiments. from Wind Tunnel to Flight,” Journal of Visualisation,
Vol. 2, pp. 321-330 (2000).
References [7] Kompenhans, J., Raffel, M., Dieterle, L., Dewhirst, T.,
Vollmers, H., Ehrenfried, K., Willert, C., Pengel, K.,
[1] Murakami, S., Mochida, A. and Hayashi, Y., “Examin- Kahler, C., Schroder, A. and Ronneberger, O., “Parti-
ing the k-e Model by Means of a Wind Tunnel Test and cle Image Velocimetry in Aerodynamics: Technology
Large Eddy Simulation of the Turbulence Structure and Applications in Wind Tunnels,” Journal of Visual-
around a Cube,” J. Wind Eng. Ind. Aerodyn., Vol. 35, ization, Vol. 2, pp. 229-244 (2000).
pp. 87-100 (1990). [8] STAR-CD version 3.15A manuals, Computational
[2] Tamura, T., “Numerical Study of Aerodynamic Be- Dynamics Limited (2002).
havior of a Square Cylinder,” J. Wind Eng. Ind. Ae- [9] David Sumner, “Closely Spaced Circular Cylinders in
rodyn., Vol. 33, pp. 161-170 (1990). Cross-Flow and a Universal Wake Number,” Journal
[3] Murakami, S., “Numerical Simulation of Turbulent of Fluids Engineering, Vol. 126, pp. 245-249 (2004).
Flow Filed around Cubic Model Current Statues and [10] Ioannis Kitsios and John Lygeros, Computation for a
Applications of k-e Model and LES,” J. Wind Eng. Personnel Launch Vehicle Using Reachability, Ame-
Ind. Aerodyn., Vol. 33, pp. 139-152 (1990). rican Institute of Aeronautics and Astronautics Launch-
[4] Mukesh Kumar, Geospatial Evaluation of Biodiversity pad Abort Flight Envelope, HYCON-publications.
Pattern in Mining Landscape of Northern Chattisgarh, [11] Experiment on Reducing Drag on an Aerospace Launch
M. Tech Thesis, Remote Sensing and Geographic Vehicle, Dryden Flight Research Center, Edwards,
Information System Andhra University, Feb. India California.
(2005).
[5] Merzkirch, W., Flow Visualisation, Ed., 2nd ed., Aca- Manuscript Received: Jul. 16, 2008
demic Press, New York (1987). Accepted: Jun. 10, 2009