J. Wind Eng. Ind. Aerodyn.

118 (2013) 35–43

Contents lists available at SciVerse ScienceDirect

Journal of Wind Engineering

and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

Novel design and experimental validation of a contraction nozzle

for aerodynamic measurements in a subsonic wind tunnel
María Rodríguez Lastra n, Jesús Manuel Fernández Oro, Mónica Galdo Vega,
Eduardo Blanco Marigorta, Carlos Santolaria Morros
Universidad de Oviedo, Área de Mecánica de Fluidos, Campus de Viesques, 33271 Gijón, Asturias, Spain

art ic l e i nf o a b s t r a c t

Article history: A novel design for a contraction nozzle, based on a logarithmic profile, is presented and developed to
Received 7 November 2012 enhance aerodynamic measurements in a low-speed wind tunnel. The improvements obtained with this
Received in revised form new proposal are validated, both, numerically and experimentally. As a starting point, four different wind
14 February 2013
tunnel contraction profiles were firstly considered and tested using the Computational Fluid Dynamics
Accepted 7 April 2013
Available online 15 May 2013
(CFD) package ANSYS FLUENTs. Both polynomial, due to its classical inclusion for wind tunnels, and
logarithmic profiles, due to its expected enhancement, have been studied in terms of avoidance of
Keywords: separation of the boundary layer, procurement of a maximum level of exit-flow uniformity, and
Nozzle minimum turbulence levels at the outlet. Numerical comparison between obtained results shows the
Contraction
benefits of the new logarithmic profile developed by the authors, which was finally employed to
Wind tunnel
construct the nozzle. To characterize its real performance, intensive experimental measurements have
CFD
Experimental measurements been conducted using pressure transducers and both single and dual hot wire anemometry. The pressure
coefficient along the nozzle sidewall reveals an optimal evolution, matching perfectly with the
theoretical design. In addition, low levels of turbulence and high flow uniformity is confirmed at the
nozzle discharge. Turbulence intensities below 0.7% are obtained for the whole range of velocities
available in the wind tunnel, and even practically constant uniform flow is obtained for all the traverses
tested for validation. These indicators conclude that the contraction designed with the new profile is a
good enhancing of the wind tunnel benefits.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction a crucial milestone to assure high-quality flow characteristics

and, therefore, it deserves special attention. Typically, converging
Wind tunnels represent an essential tool for fluid flow research nozzles have been constructed using a pair of cubic polynomials
and model testing. They allow providing significant data about flow where, for a fixed length and contraction ratio, the location of the
characteristics around immersed bodies and aeroacoustic noise, joining point has been found to be determinant in the optimiza-
when the test section is also conditioned to perform acoustic tion of the designs (Bell and Mehta, 1989; Morel, 1975, 1977;
measurements. In the context of modern designs of wind tunnels, Ramaeshan and Ramaswamy, 2002). Years ago, the analytical
the Fluid Dynamics group of the University of Oviedo has developed resolution of inviscid flow equations to study the flow field
its own facility in 2008 integrating both aerodynamic and aero- required a lot of time. For that reason, design charts for nozzles
acoustic experiments. This facility is provided with a characteristic were developed and recurrent calculations were avoided (Bell and
open section of 1 m2 and anechoic paneling to increase its experi- Mehta, 1989; Fang, 1997; Fang et al., 2001; Morel, 1975, 1977;
mental capabilities. A number of constructive modifications have Ramaeshan and Ramaswamy, 2002). However, due to the great
been performed since then to improve current measurements, computational advances and the development of more efficient
including the design and building of an optimized contraction that codes, CFD has currently become an essential tool to optimize
could increase the quality of the air flow in the test section. wind tunnel contractions. It is an accurate way to analyze the flow
The design of the optimal contraction section in a wind tunnel behavior within the nozzle and, moreover, allows geometric
is critical to obtain valuable data from the facility. A good choice is parameters to be optimized without relatively increasing costs of
computation. Because of the smooth transition of the side walls
and the high Reynolds number reached within the nozzle, bound-
ary layer growth is controlled and viscous computations are
n
Corresponding author. Tel.: +34 985182661. expected to be reasonably simple (no swirl flow, without separa-
E-mail address: rodriguezlmaria@uniovi.es (M. Rodríguez Lastra). tion and unsteadiness, first order turbulence modeling…).

0167-6105/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jweia.2013.04.008
36 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43

Concerning the definition of the geometrical profile for the 2.1. Test chamber
nozzle, it must be advised that due to some transversal restrictions
in the wind tunnel layout, it was not possible to use a pair of The test chamber is 4.2 m long and has a cross-sectional area of
conventional cubic polynomials. Alternatively, other mathematical 4.45
2.80 m2. The dimensions of the test section allow working
fittings were explored, keeping in mind the necessity to assure the inside with different equipments and without interference in the
most significant characteristics of a well-designed nozzle: reduc- air free stream discharged from the nozzle.
tion of mean and fluctuating velocity variations, and increase of
the flow mean velocity (Mehta and Bradshaw, 1979).
2.2. Settling chamber
In this paper, four different contraction profiles (three 4th-order
polynomials—FOP-and one logarithmic derivative profile—LDP-)
The settling chamber is the largest chamber of the wind tunnel,
were studied using the Computational Fluid Dynamics (CFD) package
with a characteristic cross-sectional area of 27.78 m2. Minimum
ANSYS FLUENTs. The design criteria considered for the analysis of
velocities ensure no perturbations of the mean flow and provide
the results were the avoidance of separation of the boundary layer,
accurate flow conditions at the nozzle entrance. The chamber is
the procurement of a maximum level of exit-flow uniformity, and
composed of the following parts: the honey comb, the screens and
obtaining a minimum turbulence level at the outlet (Mehta and
the relaxation duct. The combined effect of these components is to
Bradshaw, 1979). The numerical results obtained show the benefits of
reduce the turbulence coming from the fan, by breaking down
the new profile developed by the authors, and justify its final
larger eddies into smaller ones. The screen holes (see Fig. 2)
selection for construction in the wind tunnel.
present a characteristic size of 30 mm.
Further validation has been attempted to certificate the good-
ness of the LDP design. Experimental measurements using pres-
sure taps and hot wire anemometry have revealed the good 2.3. Mufflers
performance of the nozzle in terms of flow uniformity and low
turbulence levels (below 0.7%); thus concluding that the contrac- Two mufflers, one between the settling chamber and the fan,
tion section developed here was the best option possible for the and another in the longest duct, reduce the noise coming from the
enhancement of the existing wind tunnel.

2. Description of the wind tunnel

The wind tunnel of the Fluid Dynamics Group at the University

of Oviedo is a closed loop circuit, arranged in a vertical layout, as
shown in Fig. 1. The total length of the tunnel is 24.6 m, with 8.3 m
high and maximum operative velocities in the range of 22 m/s for
the test section; i.e., a maximum Reynolds number of 1.7
106,
based on the outlet hydraulic diameter of the nozzle (i.e., the
outlet width). A full scheme of the facility is also depicted in Fig. 1. Fig. 2. Screen holes.

Fig. 1. Wind tunnel facility.

 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43 37

fan to the test section. This acoustic isolation makes it possible to y ¼ b1 þ b2 z½lnðb2 zÞ−b3  ð2Þ
perform aeroacoustic measurements within the wind tunnel.
The polynomial in Eq. (1) includes five coefficients (a0 to a4)
that must be fixed according to the following geometrical con-
2.4. Fan straints: inlet (i) and outlet height (ii), zero acceleration of the air
flow at the outlet (iii), zero variation of acceleration of the air flow
The fan used is of axial anti-stall type, with the impeller at the outlet (iv) and a particular inlet curvature (v). Mathemati-
mounted directly on the motor shaft. It provides a variable air cally, these conditions can be expressed as follows:
flow modifying the pitch of the blades. The 30 kW driven motor
establishes maximum flow rates in the range of 26 m3/s (blades yðz ¼ 0Þ ¼ hi
are full-opened) and a total-to-static pressure increment of 850 Pa. yðz ¼ lÞ ¼ ho
aðz ¼ lÞ ¼ 0
2.5. Nozzle a’ðz ¼ lÞ ¼ 0
y’ðz ¼ 0Þ ¼ tgðαÞ ð3Þ
The new nozzle had to be placed between the settling chamber
and the test section. Its length was determined by the separation where hi and ho are half inlet and outlet height respectively, l is the
between the chambers (1.505 m). Different contraction ratios were nozzle length, z is the longitudinal coordinate in the nozzle, a is
required for enabling diverse experiments, so the nozzle was the flow acceleration and α is the inlet angle. The angles chosen for
constructed with the possibility to modify its lateral span. Thus, α were 401, 501 and 601.
the distance between the two vertical faces had to be: Eq. (2) was obtained establishing that the nozzle profile follows
2.745 70.2 m at the inlet and 1.000 70.2 m at the outlet (Fig. 3). a logarithmic derivative. This shape was chosen due to its
As a consequence, the contraction ratio (in terms of cross-sectional smoothed curvature, with only three parameters defined as
area) in the nozzle is variable, ranging from 9.5:1 to 6.25:1. b1 ¼ hi
b2 ¼ ðhi −ho Þ=l
3. Nozzle contours studied b3 ¼ lnðhi −ho Þ þ 1 ð4Þ
which are a group of coefficients that adjust the offset and the
The most extended contraction profiles used for wind tunnels, slope of the logarithmic function.
which have been widely tested and recommended for this type of A representation of the four contraction profiles is shown in
application in the literature, are those based on a pair of two cubic Fig. 4. In this plot, the non-dimensional length of the nozzle (z/L)
polynomials (Bell and Mehta, 1989; Morel, 1975, 1977; Ramaeshan and the non-dimensional vertical coordinate (y/h) are represented.
and Ramaswamy, 2002). Boerger nozzles (Boerger, 1973) are also an Obviously, the z-coordinate increases in the downstream direction.
interesting alternative, widely used in the car industry, because they The coordinate system is established with the origin at the center
provide the shortest possible contraction length for a prescribed flow
quality, using a shape quite similar to that given by the cubic
polynomials. However, these families of profiles could not be adapted
in our case due to some unavoidable geometrical restrictions of the
wind tunnel. In particular, the existing structure that connects both
chambers limits the space in a way that a two cubic polynomial
nozzle cannot be placed in it. To overcome this problem, a fourth
order polynomial profile (FOP), given by Eq. (1), and a logarithmic
derivative profile (LDP), given by Eq. (2), were explored as reasonable
alternatives for our design.
y ¼ a0 þ a1 z þ a2 z2 þ a3 z3 þ a4 z4 ð1Þ

Fig. 4. Contraction profiles.

Fig. 3. Scheme of the versatile nozzle. Fig. 5. LDP model meshed.

38 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43

line of the inlet plane. The remaining x-coordinate is defined for As boundary conditions, a gauge total pressure of 296.45 Pa
the spanwise direction. at the beginning of the mesh (settling chamber inlet) with
a turbulence intensity of 2.5% and a turbulent length scale of
0.25 m were imposed, both obtained with hot wire measurements
4. Numerical methodology and validation at the inlet. An atmospheric (0 Pa) gauge pressure at the nozzle
outlet was also imposed. Both are pressure conditions instead of
4.1. Geometry and mesh velocity ones in order not to condition the simulation results.

ANSYS ICEMCFDs meshing software was used for geometry

and mesh generation (contraction outlet section ¼1
1 m2 and 5. Numerical results
contraction inlet section ¼ 2.745
2.745 m2). Due to the symmetry
presented, only one quarter of the nozzle was modeled in order to The flow uniformity of each model at the outlet was studied
reduce the computational load. Part of the settling duct was also and the results obtained are shown in Fig. 7. Due to the geome-
included in the geometry for accuracy purposes, as shown in Fig. 5. trical similarity of the contraction profiles, it was expected a
Every model was meshed using a structured mesh, avoiding similar velocity tendency at the outlet of the nozzle. The flow
highly skewed elements. A boundary layer grid dependence study, velocity of all models remains greater than 99% of Umax until
with five different density meshes, was performed using the reaching 96% of the outlet height, thus providing a boundary layer
velocity profiles along the vertical plane of the outlet as control size of 4% of the outlet height. Showing a detail of the velocity
variable. The integral turbulent scales and turbulence intensities at profiles in the boundary layer area, some differences can be
the inlet, measured previously with a hot wire probe (detailed observed. At this point the LDP profile presents slightly higher
forward), were used for defining a high-density mesh. Further velocity values than the polynomial ones, achieving a better flow
mesh refining was conducted toward the nozzle sidewalls. The uniformity.
comparison of the numerical data obtained for each mesh is The velocity contours of the LDP model at the middle plane are
illustrated in Fig. 6. depicted in Fig. 8. The flow coming from the settling chamber
The model with the coarser mesh (y+ of 631.6) is the one that arrives at the contraction inlet with a mean velocity of 2.9 m/s;
shows the biggest difference in the velocity profile. The velocity afterward it is accelerated throughout the nozzle until its way out.
starts to drop off when y/h is 0.92, while for the rest of the models At streamwise sections close to the nozzle inlet, the flow velocity
it occurs at y/h¼0.97. Results show that refining the mesh from a increases as it nears the wall. As long as the air advances towards
value of y+ ¼0.297 do not provide more accurate values of velocity the outlet, this tendency decreases being reversed near the outlet.
at the boundary layer. The velocity values are slightly lower near As the nozzle structure consists of convex surfaces, the possible
the wall for the y+ ¼ 82 model than for the three finest mesh separation region spans the entire length of the contraction.
models. For them the results remain invariant. Taking into account For that reason it is fundamental for the profile to lead the flow
these outcomes, the coarser mesh with the best results (y+ of avoiding adverse pressure gradients which are the essential
0.297) was the chosen one for continuing the profile study, due to conditions for separation. Precisely, for testing the avoidance of
the staking strategy used in the contours (see Fig. 5, detail). separation at the nozzle walls, the skin friction coefficient was
considered (Fig. 9). Simulation results show that the LDP and the
4.2. Boundary conditions and models FOP (α¼501) do not present separation, indicated by the positive
skin friction coefficient values over the wall (Sargison et al., 2004).
The Computational Fluid Dynamics (CFD) package ANSYS- However, LDP offers more balanced results without drastic
FLUENTs was used for solving the 3D steady Navier–Stokes increasing gradients of the coefficient. On FOP (α ¼601) and FOP
equations in conjunction with the k-epsilon turbulence model (α¼401) there is a region at the wall, near the outlet, that shows
and enhanced wall treatment on the structured meshes. This a discontinuity of the skin friction coefficient distribution. These
turbulence model was chosen because it obtains the production areas seem to show corner vortices, which are very common in
and dissipation of turbulence and allows pressure gradients wind tunnel contractions. In particular, the comparison of pressure
leading to boundary layer separations to be predicted. A Simple distributions on the nozzle lateral wall for the cases with α¼401
pressure–velocity coupling algorithm was selected and second and α¼ 501 reveals the impact of the corner vortex on the pressure
order discretization was applied for the equations to increase the distribution (local negative values of pressure coefficient are more
accuracy of the results. aggressive in the 401-cases leading to the appearance of such

Fig. 6. Numerical comparison of velocity profiles on vertical plane. Fig. 7. Velocity profiles at the outlet on vertical plane.
 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43 39

Fig. 8. Velocity contours of LDP on middle plane.

Fig. 10. Impact of corner vortex on pressure distribution.

Fig. 11. Dimensionless turbulent kinetic energy at the outlet on vertical plane.

dimensionless turbulent kinetic energy of both profiles maintains a

constant value of 0.22
10−3 until it is reached 96% of the contraction
length. From that point it increases its value to 7.7
10−3.
As a conclusion, numerical results obtained with the simulation
indicate that the logarithmic profile (LDP) and the polynomial
Fig. 9. Skin friction coefficient contours at the nozzle upper wall. profile (FOP (α¼501)) provide very good results in terms of flow
quality for experimental tests. However, the outcomes show that
vortices)—See Fig. 10. In addition, a visualization of the core of the the flow guided by LDP nozzle is slightly better conditioned.
vortices has been superimposed to illustrate this phenomenon. Accordingly, this was the shape chosen for building the nozzle.
Note that these structures are not present in the case with 501.
The turbulence generated in the air flow is also an important
factor to be considered. The dimensionless turbulent kinetic energy 6. Experimental setup
profiles have been analyzed at the outlet on the vertical plane
(Fig. 11). As expected, the predicted values for FOP (α¼ 601) and Following the numerical results, the LDP was the chosen profile
FOP (α¼401) become higher earlier due to the appearance of corner for constructing the wind tunnel contraction (Fig. 12).
vortices. The LDP and the FOP (α¼501) models produce less Ten pressure taps were installed along the middle section of a
turbulence at the outlet, reaching almost the same values. The sidewall, and another two were put at the inlet and the outlet,
40 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43

Fig. 12. Wind tunnel contraction constructed.

respectively. These taps, connected to a pressure transducer,

provided the wall static pressure distribution and the dynamic
pressure at the nozzle outlet.
The velocity distribution and the turbulence intensity at the
outlet of the contraction were measured with a home-made
constant temperature X-wire probe, connected to a TSI IFA 100.
The probe had a cut-off frequency of 64 kHz. The IFA 100 outputs
where connected to a National Instruments acquisition card that
gave the information to a computer. All the aspects of calibration
of the probe, capture, record and conversion of measurements
were driven by a MATLABs code, particularly developed for this
kind of applications. Some of the authors work on that particular
Fig. 13. Experimental setup for hot wire measurements.
field (use and design of these experimental techniques) is pre-
sented in references (Blanco et al., 1998; Fernández Oro et al.,
2007a, 2007b). Fig. 13 shows the experimental setup placed in the
test section, inside the wind tunnel. The X-probe was placed at
several positions on the center line, between the middle point and
the top wall (see Figs. 14 and 15).
Different sampling frequencies at 5, 10, 20 and 50 kHz were
tested. The lower frequencies (5–10 kHz) were found not to
provide enough accurate data about the turbulence of the flow,
while the higher frequencies (20–50 kHz) did not show that
problem. It was necessary to find a compromise between the
accuracy of the results and the time and space needed for
recording the data so finally, a 20 kHz sampling frequency was
chosen.

7. Experimental results Fig. 14. Assembly and detail of the X-probe.

7.1. Longitudinal pressure distribution

Good agreement was found between both distributions all over
The static pressure measured along the middle section of a the curve (see Fig. 16). The static pressure at the inlet of the nozzle
sidewall was compared with the theoretical static pressure (ideal becomes dynamic pressure at the outlet without significant losses.
flow with no losses). The dimensionless parameter Cp was used to
normalize the data 7.2. Nozzle outlet velocity and turbulence measurements
P−P ∞
Cp ¼ ð4Þ A hot wire probe was positioned at the middle line in vertical
ð1=2ÞρU 2∞
plane and was traversed from the center of the nozzle outlet to the
where P is the static pressure measured along the wall, P ∞ and U ∞ top wall. This instrument was used to perform velocity and
are is the static pressure and the mean velocity measured at the turbulence measurements, which provide information about the
outlet of the contraction, respectively. influence of the walls on the flow.
 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43 41

Fig. 15. Scheme of measurement lines.

1 Fig. 17. Time series of signals recorded at x ¼0.5 m in vertical plane (y/h ¼0, 0.9,
Experimental
and 0.99).
Theoretical

0.8

0.6
Cp

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
z/l
Fig. 16. Experimental and theoretical pressure coefficient along a sidewall.

A first study of the recorded signals showed, as expected, that Fig. 18. Power spectral density at the center of the nozzle and near the wall.
the reduction of velocity fluctuations was remarkable as the probe
approached the center of the nozzle. This fact revealed, at first
sight, the significant reduction of the turbulence level throughout
the outlet. A fragment of three signals time spectra, recorded at
three particular positions, is depicted in Fig. 17. There is no
significant variation of the flow velocity in signals at y/h¼0 and
0.9, in contrast with the data dispersion near the outlet wall
(y/h¼0.99).
Fig. 18 compares the power spectral densities of the free stream
region (y/h¼ 0) and the viscous region (y/h¼ 0.99). A moving
average filter (with an eleven-order span) was applied for redu-
cing the signal noise. The PSD level at the viscous region is lower
throughout all the frequency range, related with the flow velocity.
Both signals show a decay of the PSD with the frequency, with a
broadening of the spectra which means the evolution of the flow Fig. 19. Normalized wind speed, turbulent intensity measured at x¼ 0.5 m in
searching a fully-developed state. It was verified that both slopes vertical plane.

where greater than the −5/3 exponential decay (Davidson, 2004).

Mean velocity values measured at each position and normal- non-uniformity, based on the difference between the minimum
ized by the maximum velocity are presented in Fig. 19. It is also and the maximum nozzle outlet velocity outside the boundary layer
showed at the same figure the turbulent intensity level, estimated (Mathew et al., 2005), was found to be good (2.3%). The turbulence
by intensity level estimated is low even at positions near the wall. It is
pffiffiffiffiffiffiffi practically constant from the center to y/h¼0.8, where turbulence
u′2 intensity roughly reaches 0.6%. From y/h¼0.8 until the last measure-
TI ¼ 100 ð5Þ
u ment position, the mean TI rises up to 1%. The turbulence intensity
where u is the instantaneous velocity in the outlet at the location was also measured at the center of the nozzle inlet. At that position
of the hot wire, and u′ represents the turbulent fluctuation. the TI was found to be 2.5%, which indicates that the LPD profile
Velocity values remain almost constant from the center of chosen reduces significantly the turbulence level coming from the
the contraction outlet until y/h¼0.99. The test section flow settling chamber. These results indicate that there is no separation of
42 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43

the boundary layer, verifying the numerical ones. This is apparently

in disagreement with the expected behavior according to separation
boundaries for two-dimensional contractions of Rouse and Hassan
(1949). However, Wolf (1994), found out that for five different
contractions with shorter lengths than theoretically required, there
was no separation of the boundary layer as well.
The integral length scale of the largest eddies of the flow was
determined by
Z ∞
L¼U ACFðτÞdτ ð6Þ
0

u′ðtÞu′ðt þ τÞ
ACFðτÞ ¼ ð7Þ
u′2

where the overbar denotes the time-averaging value and τ is the time
lag that is used to construct the ACF. This formulation assumes that Fig. 20. Distribution of integral length scale (75% confidence) measured at x¼ 0.5 m
the average eddy size lies through the correlation of two velocity in vertical plane.
signals (Taylor's hypothesis). To estimate the integral length scale in
Eq. (6), it is necessary to evaluate the correlation coefficient Eq. (7)
of the velocity fluctuations (in time), which theoretically must be
defined up to infinity. The area under the correlation gives the value
of the integral scale. Because in practice, this is limited, it is necessary
to adopt a particular criterion. Following the guidelines of Tropea
et al. (2007) (Handbook of Experimental Fluid Mechanics), three
options are suggested: (a) “stop at the first zero crossing, i.e., at the
first time separation, τmax, for which the correlation coefficient
vanishes. However, in many cases the correlation coefficient starts
Fig. 21. Turbulent intensity level for different wind speeds measured at the center
to exhibit oscillations before this zero is attained” (like in our of the nozzle outlet.
database); (b) “in that case, a second, more-convenient possibility is
to define τmax as the value for which the autocorrelation coefficient
reaches its first minimum”; and (c) “another possibility is to estimate is presented in this paper. A numerical comparison of the novel
the integral scale as the value for which the correlation coefficient design developed by the authors with polynomial profiles, were
attains 1/e of its maximum (equal to 1 for zero separation), i.e., to the carried out. Results showed the benefits of the logarithmic profile
value expected if an exponential decay of the correlation coefficient is (LDP) in terms of flow uniformity, separation of boundary layer
assumed”. In our measurements, we have observed a clear corre- and turbulence levels. Moreover, two polynomial profiles (FOP
spondence between both limited criteria: the first minimum is many (α¼601) and FOP (α¼401)) tended to produce corner vortices in
times coincident with the 1/e limit. On the contrary, if we seek for the regions near the outlet.
crossing with the x-axis, the area under the curve is notably extended Experimental measurements with pressure transducers and
and the overall integral value excessively enlarged. both, single and dual, hot wire anemometry were accomplished
It is evident for the high degree of disparity and the inherent to characterize the nozzle. The pressure coefficient distribution
uncertainty to define an exact value of the integral length scale. along the sidewall revealed a perfect agreement with the theore-
Only an approximate order of magnitude and an overall trend can tical design, and the velocity and turbulence levels remained low
be given regarding this parameter. Consequently, to provide a all over the discharge. Turbulence intensity levels for the whole
reference result, we have performed a sort of averaged between all range of velocities available were found to be lower than 0.7%.
the different scales attained with sub-methods (b) and (c), which These outcomes conclude that the novel contraction developed is
have been additionally smoothed. In Fig. 20, the mean value of the a good enhancing of the wind tunnel benefits.
distributions and those upper and lower limits representing
approximately a 75% confidence level are represented. The integral
scale is approximately 50 mm in length in central positions and Acknowledgments
drops rapidly toward 10 mm close to the endwall. On the contrary,
it is enlarged as we move toward the nozzle centerline with typical The work reported in this article is funded by the Spanish
values in the range of 0.1–0.15 m, a fraction of the outlet hydraulic Ministry of Economy and Competitiveness under the subprogram
diameter for the nozzle (1 m). Predoctoral Research Grants (Fig. 22).
The variation of the turbulence with the wind speed was also
studied. For this survey, the X-probe was fixed at the center of the
contraction outlet. Measurements were made with twenty differ-
ent wind speeds. The results are plotted in Fig. 21. One can note
that the turbulence intensity levels do not differ much. They
maintain an approximately constant level of 0.69%.

8. Conclusions

The design, fabrication and characterization of a nozzle to

enhance aerodynamic measurements in a low-speed wind tunnel, Fig. 22
 M.R. Lastra et al. / J. Wind Eng. Ind. Aerodyn. 118 (2013) 35–43 43

References Mathew, J., Bahr, C., Carroll, B., Sheplak, M., Cattafesta, L., 2005. Design, fabrication,
and characterization of an anechoic wind tunnel facility. In: Proceedings of the
11th AIAA/CEAS Aeroacoustics Conference.
Bell, J.H., Mehta, R.D., 1989. Boundary-layer predictions for small low-speed
Mehta, R.D., Bradshaw, P., 1979. Design rules for small low-speed wind tunnelsAer-
contractions. AIAA Journal 27, 372–374.
onautical Journal 83, 443–449.
Blanco, E., Ballesteros, R., Santolaria, C., 1998. Angular range and uncertainty
Morel, T., 1975. Comprehensive design of axisymmetric wind tunnel contractions.
analysis of non orthogonal crossed hot wire probes. ASME Journal of Fluids
ASME Journal of Fluids Engineering 97, 225–233.
Engineering 120, 90–94.
Morel, T., 1977. Design of two-dimensional wind tunnel contractions. ASME Journal
Boerger, G.-G., 1973. Optimierung von Windkanalduesen fuer den Unterschallbereich.
Ph.D. Dissertation. University Bochum. of Fluids Engineering 99, 371–378.
Davidson, P.A., 2004. Turbulence: An Introduction for Scientists and Engineers. Ramaeshan, S., Ramaswamy, M.A., 2002. A rational method to choose optimum
Oxford University Press. design for two dimensional contractions. ASME Journal of Fluids Engineering
Fang, F., 1997. A design method for contractions with square end sections. ASME 124544–546
Journal of Fluids Engineering. 119, 454–458. Rouse, H., Hassan, M.M., 1949. Cavitation-free inlets and contractions. Mechanical
Fang, F., Chen, J.C., Hong, Y.T., 2001. Experimental and analytical evaluation of flow in a engineering 71, 213–216.
square-to-square wind tunnel contraction. Journal of Wind Engineering and Sargison, J.E., Walker, G.J., Rossi, R., 2004. Design and calibration of a wind tunnel
Industrial Aerodynamics 89, 247–262. with a two dimensional contraction. In: Proceedings of the 15th Australasian
Fernández Oro, J.M., Argüelles Díaz, K.M., Santolaria Morros, C., Blanco Marigorta, E., Fluid Mechanics Conference, vol. 13.
2007a. Unsteady flow and wake transport in a low-speed axial fan with inlet guide Tropea, C., Yarin, A.L., Foss, J.F. (Eds.), 2007. Measurement of Turbulent Flows.
vanes. Journal of Fluids Engineering 129, 1015–1029. Springer (Chapter 10).
Fernández Oro, J.M., Argüelles Díaz, K.M., Santolaria Morros, C., Blanco Marigorta, E., Wolf, T., 1994. Design of a variable contraction for a full-scale automotive wind
2007b. On the structure of turbulence in a low-speed axial fan with inlet guide tunnel. Journal of Wind Engineering and Industrial Aerodynamics 56, 1–21.
vanes. Experimental Thermal and Fluid Science 32, 316–331.