You are on page 1of 10

J. Wind Eng. Ind. Aerodyn.

145 (2015) 209–218

Contents lists available at ScienceDirect

Journal of Wind Engineering


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

Blockage-tolerant wind tunnel measurements for a NACA 0012 at high


angles of attack
J.M. Rainbird n, J. Peiró, J.M.R. Graham
Department of Aeronautical Engineering, Imperial College, London SW7 2AZ, UK

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

Article history: Good quality post-stall aerofoil force data at low Reynolds numbers is needed for the analysis of vertical-
Received 24 June 2014 axis wind turbines (VAWTs). Discrepancies between existing studies are shown to affect modelled per-
Received in revised form formance of VAWTs, with wind tunnel blockage identified as a possible cause.
1 June 2015
New data is presented for two different-sized NACA-0012 aerofoils, taken in blockage-tolerant and
Accepted 10 June 2015
Available online 21 July 2015
conventional solid-walled wind tunnels. The tolerant tunnel has transversely slotted walls and can be
configured with either single or double slatted walls, with adjustable wall porosity. Conventional tunnel
Keywords: results require processing with blockage corrections that are less than ideal for application to stalled
Post-stall aerofoils. The tolerant tunnel does not need any corrections. Results are given for the best single and
NACA-0012
double slatted wall tunnels, chosen based on which tunnel wall porosity gives the closest force mea-
Low Reynolds numbers
surements for the two aerofoils.
Vertical axis wind turbines
Blockage tolerant wind tunnels The single slatted wall data is similar to that from the corrected conventional tunnel. The double
Blockage corrections slatted tolerant tunnel has the best performance overall based on similarity of results for the aerofoils but
post-stall force peaks are significantly lower than for the conventional tunnel.
A relationship between post-stall lift and drag peak magnitude and blockage is hypothesised for
conventional tunnel data that persists even after the application of corrections.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction designated as type A is per the reference, while type B is a lift


characteristic modified to exhibit 10% lower post-stall peaks in Cl
For most applications, aerofoils are not expected to encounter at incidences of 45° and 135°. Fig. 1(b) shows the resulting turbine
deep stall, which typically starts at an incidence, α, of approximately torque coefficient CQ against λ from the BEM code using Cl types A
20°, so testing rarely extends beyond the range −20° ≤ α ≤ + 20° and B (the same Cd was used for both runs). Fig. 1(c) shows
(see, for example, Althaus, 1980). Post-stall operation is expected of development of λ with time from a standing start. The 10%
wind turbine blades during start-up, with the inboard regions of reduction in post-stall lift peaks is sufficient to impact on the
horizontal-axis wind turbine (HAWT) blades and the full length of prediction of successful self-starting, the Type B turbine failing to
accelerate along its entire power curve due to production of
vertical-axis (VAWT) blades experiencing angles deep into stall at
negative torque at 0.5 < λ < 1.7.
low tip-speed ratios, λ (Dominy et al., 2007). VAWT performance,
Note that self-start is defined in this paper after Bianchini et al.
particularly during start-up, is sensitive to small changes in aerofoil
(2011) as when a turbine accelerates through its entire power curve
characteristics at certain post-stall angles (Hill et al., 2009), making
to its fastest equilibrium state unaided. Throughout this
accurate post-stall data desirable for the study of these machines.
paper, the term “post-stall” refers to aerodynamic characteristics in
To demonstrate the sensitivity of VAWTs to post-stall aerofoil
the range of incidences between stall angles, approximately
performance, the start-up of a turbine has been modelled using a ±10° ≤ α ≤ ± 170°. The term “deep stall” refers to the range of
blade-element momentum model (Strickland's, 1975 multiple- incidences ±20° ≤ α ≤ ± 160° and “the immediate vicinity of stall”
streamtube method) with aerofoil coefficients based on Sheldahl to 10° < α < 20°.
and Klimas's (1981) post-stall study. Fig. 1(a) shows lift and drag Little aerofoil data exists covering a full range of incidences
coefficients, Cl and Cd , against α for the NACA 0018 aerofoil. The Cl from 0° to 180°, there is a particular lack of data taken at the low
Reynolds numbers at which typical VAWT blades operate. For
n
Corresponding author. Room 363, Roderic Hill Building, Imperial College Lon-
example, the commercially available Urban Green Energy Visionair
don, South Kensington Campus, London SW7 2AZ, UK. 3 turbine (http://www.urbangreenenergy.com/products/vision-
E-mail address: j.rainbird11@imperial.ac.uk (J.M. Rainbird). air3) has a 1.8 m diameter and 0.38 m blade chord. In accelerating

http://dx.doi.org/10.1016/j.jweia.2015.06.006
0167-6105/& 2015 Elsevier Ltd. All rights reserved.
210 J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218

Fig. 1. Effects of small changes in post-stall lift on VAWT performance. (a) Aerofoil coefficients (Sheldahl and Klimas, 1981). (b) Turbine Cp envelope. (c) Turbine start-up
behaviour.

from standing to λ = 2 in a wind speed of 6 m/s, it will experience information about experimental configuration could be found (see
a maximum blade Reynolds number of only 4.5  105. Timmer, 2010 for a review of NACA 0012 data and Lindenburg,
Fig. 2 reproduces Cl and Cd from the studies of Bergeles et al. 2000 for a more comprehensive review including cambered
(1983), Critzos et al. (1955), Massini et al. (1988) and Sheldahl and aerofoils). The studies are summarised in Table 1 along with that
Klimas (1981) for the NACA 0012 and Pope's (1947) study of the of Althaus (1980) which does not extend into deep stall and is
NACA 0015, results for which should be similar to the NACA 0012, referenced later in this paper.
particularly post-stall. Output from NASA's AERODAS model The table contains information on factors that can affect aero-
(Spera, 2008) for an infinite aspect ratio, 12% thick aerofoil, and foil force measurements taken in wind tunnels, namely the aspect
thin aerofoil lift theory are also included. ratio of the wing under test (a relevant consideration even when
There is excellent agreement with thin aerofoil theory for all trying to reproduce two-dimensional flows) and the chord-to-
studies pre-stall but post-stall the spread of data is far wider. At height ratio, c /H , of the aerofoil and tunnel. The length-to-height
the second lift peak at α ≈ 45°, values of Cl vary between 0.98 and ratio, L /H , of the tunnel test section and the tunnel's turbulence
1.18, a difference of 20%. At peak drag at α ≈ 90°, Cd is between 1.81 intensity can also affect readings but information on these were
and 2.08, a 15% variation. The variation in the lift peak is more not commonly given in the references.
than sufficient to impact on accurate modelling of VAWT start-up, Sensible experiment design and data processing can limit the
suggesting more accurate post-stall data is required to simulate impact of these factors. Good wind tunnels will have minimal
start-up with confidence. These uncertainties in aerofoil char- turbulence intensity and the test section length will allow suffi-
acteristics are perhaps the reason that, in spite of experimental cient distance between the tunnel contraction and models under
studies showing that VAWTs with straight, symmetrical blades are test. The impact of aspect ratios can be minimised by considering
capable of self-starting (Hill et al., 2009; Rainbird, 2007; Dominy how stall cells are likely to form on the aerofoil when designing an
et al., 2007), accurate modelling of successful start-up can be experiment (by applying the findings of Weihs and Katz, 1983, for
impossible without modification of the input aerofoil coefficient example).
data: Bianchini et al. (2011) give their blades “virtual camber” The impact of blockage cannot always be reduced as easily;
while Rossetti and Pavesi (2013) apply the post-stall model of increasing tunnel size is impractical, while reducing model size
Viterna and Corrigan (1982) to their blades. leads to increased inaccuracies in profile reproduction and
The selected studies shown in Fig. 2 represent all investigations reductions in Reynolds numbers achievable in a given tunnel.
of symmetrical aerofoils post-stall that the authors could find, Instead, results from closed wall tunnels are processed with
other than Zhou et al.'s (2011) work at very low Reynolds numbers blockage corrections or, less commonly, blockage tolerant test
and Poisson-Quinton and de Sievers's (1967) study, for which no sections are used. The ratio c /H is an indicator of blockage; the
J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218 211

Fig. 2. Cl and Cd vs. α for the c /H = 0.32 aerofoil from this study, alongside the studies of Bergeles et al. (1983), Critzos et al. (1955), Massini et al. (1988), Sheldahl and Klimas
(1981) and Pope (1947).

data presented in Fig 2 and Table 1 suggests that higher blockage closed and open jets are of opposite signs (ESDU, 1978), one would
induces higher post-stall peaks, even in blockage-corrected data. expect that free air conditions can be approximated using semi-
permeable walls.
Parkinson's (1990) tolerant section design has slots perpendicular
to the flow in certain sections of the tunnel walls that allow flow to
2. The tolerant tunnel
exit and enter the main channel, with the walls made up of a regular
array of evenly spaced aerofoils. The shape of the array components
Tolerant test sections offer a means of minimising blockage
avoids flow separation around them. Plenums of stagnant air outside
effects without the need for corrections, which are not intended
the slotted walls, enclosed from the external environment, ensure
for use with extensively separated flows (ESDU, 1978). The sim-
mass conservation in the flow along the tunnel. The design is ver-
plest tolerant designs use passive techniques. Since corrections for satile, significantly reducing blockage for a wide range of model sizes
in a given tunnel with an appropriate open area ratio (Hameury,
Table 1
1987) (OAR, a measure of open to slatted wall areas, defined as g /s
Summary of studies referenced in this paper. All NACA 0012 except Pope (NACA
0015). using the notation in Fig. 3 where g is the gap between slat aerofoils
and s is the spacing between them). Appropriate OARs are settled on
Source Post-stall peak Re (  106) c /H AR through experiments on models of different sizes but with like
Cl Cd

Althaus (1980) – – 0.15 0.33 1.85


Bergeles et al. (1983) 1.08 1.90 0.76 0.14 1.70
Critzos et al. (1955) LTPT 1.13 2.08 1.80 0.07 6.00
Critzos et al. (1955) LTPT 1.15 1.96 0.50 0.07 6.00
Critzos et al. (1955) Langley 7  10 1.14 2.07 1.36 0.10 7.00
Massini et al. (1988) 1.18 1.99 0.96 0.13 2.00
Sheldahl and Klimas (1981) 1.10 1.83 0.36 0.07 6.00

c /H = 0.27 (this study) 1.01 1.96 0.15 0.27 3.51


c /H = 0.32 (this study) 1.09 2.04 0.15 0.32 2.96

Pope (1947) 0.98 1.81 1.23 0.17 1.67


Fig. 3. Parkinson's slatted ceiling tunnel configuration for aerofoil testing.
212 J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218

shapes, and may be specific to model shape rather than tunnel. e


perspex windows
This makes experiments more costly and time consuming than for in tunnel walls
the equivalent solid-walled tunnel experiments, as the tunnel must 457mm
be configured before main experiments can be run. The two-
dimensionality of the design prevents development of significant
435mm
transverse flows, which can impede the quality of readings
taken from a experiment attempting to reproduce two-dim-
end plates
ensional flows.
Parkinson's design was originally developed (as shown in
Fig. 3) with a slatted ceiling and solid floor for the testing of
aerofoils pre-stall (by his student Williams, 1975). The floor was aerofoil
left solid as otherwise it was judged that the aerofoil under test holes in windows
would draw the shear layer from a bottom plenum chamber into 2mm
the main test section upstream of the aerofoil, degrading the flow extended
around it (Williams, 1975). An OAR of 0.6 was judged best for struts
unstalled aerofoils. Preliminary work on stalled aerofoils at α = 20°
235mm
used an OAR of 0.7 (Parkinson, 1984).
A slatted floor and ceiling configuration with OARs of 0.59 was hole in window
later used with aerofoils (Malek, 1983), following potential flow
modelling that suggested the quality of blockage reduction was R=70mm
better than for the original design. Limited experimental results 105mm R=45mm
were presented to support this prediction, with no comparisons
made to Williams' single slatted wall work.
A slatted floor and ceiling design was also utilised for bluff body aerofoil
testing (Hameury, 1987) of circular cylinders, circular cylinders with
splitter plates and flat plates normal to the flow, with promising 76mm
results for a variety of sizes of all three models taken using OARs of end plate
0.5–0.7.
extended strut

Fig. 4. Aerofoil mounting in the wind tunnel, front-on and side-on views.
3. Experimental apparatus

A Parkinson test section has been used to collect data for the
new results presented in this paper. The test section has a square
cross section with side length of 457 mm. It has a slatted floor and
ceiling configuration with a L /H ratio of 2.33, a plenum depth to
tunnel height ratio, p/H , of 0.39 and NACA 0015 slat aerofoils with
80 mm chords. It can be modified to a solely slatted ceiling con-
figuration or a conventional closed section by replacing the slatted
floor, or both floor and ceiling with solid panels. The wind tunnel
has an open circuit design and the turbulence intensity in the
working section has been measured at 0.15%, while the flow uni-
formity is within 0.5% of the centreline flow across the central half
of the tunnel's cross section.

3.1. Aerofoils

The NACA 0012 profile was chosen for testing as there is more Fig. 5. Experimental set-up in and around the wind tunnel.
existing data at extreme incidence angles for it than for any other
section, in spite of the sharp leading edge stall it experiences at the
Reynolds number used in this study (Re = 1.5 × 105). Leading edge itself bolted to a Nutem Model 528 three-component force bal-
stall can be unpredictable, potentially occurring at different angles ance, see Fig. 5 for a diagram of the experimental set-up. Using the
of attack on different runs of the same experiment, though no such
original struts and cross-beam, the balance is rated to an accuracy
repeatability issues were encountered in this study. The larger of
of 70.15 N for lift and 70.05 N for drag. New extended struts and
the two aerofoils used has a chord of 147 mm, giving a c /H ratio of
beam were manufactured to provide sufficient clearance around
0.32 and has a hollow resin construction. The smaller has a
the bottom plenum chamber. These allow larger vibrations to
124 mm chord (c /H = 0.27) and is made of wood. Both aerofoils
have a span of 435 mm and are fitted with small end plates, pri- develop in unsteady flow than did with the original apparatus.
marily intended to inhibit leakage from the tunnel at the aerofoil Care has been taken with data acquisition to limit the effects of
mounting points. See Fig. 4 for a diagram of the aerofoil mounting. these vibrations. Readings were taken at 100 Hz (a much higher
rate than the highest frequencies of the main vibrations induced)
3.2. Model mounting and data acquisition for 10 s after a period of settling. This was sufficient to capture
approximately four periods of the lowest frequency of fluctuating
The aerofoils were mounted at the half chord to a hub assembly force induced in any reading (these occurred in the α = 90° Cd
that allowed for rapid adjustment of the angle of incidence. The reading). The data was digitised using a National Instruments NI
hub in turn connected to a pair of struts mounted on a cross-beam, USB-6229 data acquisition board.
J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218 213

The incidence of the aerofoil was adjusted manually at the hub corrections for flows with “some degree of separation,” and that
assembly. Tunnel velocity was also adjusted manually before each they have been used with c /H ratios of up to 0.35 (the largest in
reading and was measured across the contraction of the tunnel this study is 0.32). Though it is clearly inappropriate to apply
using a micromanometer. corrections to fully separated flows, use of them is made in post-
The force balance was calibrated before experiments using stall studies in the absence of any better alternatives. The wake
weights applied in the directions of expected lift and drag forces. blockage correction used here is based in part on Maskell's (1963)
Due to significant drift in the load cells, unloaded readings for the work on bluff bodies and stalled wings, so this element of the
balance were taken before and after each run of the tunnel. Unloa- corrections at least is applicable when flow is separated.
ded values for each incidence were calculated by time apportioning
the difference between the start and end unloaded readings. This 4.2. Correction-free data
drift is the largest source of potential error in readings taken from
the apparatus. Total potential error in the system is estimated at a Given the limitations of correcting for separated flows, the
maximum of approximately 3%. Other than in the immediate post- possibility of conducting closed jet experiments that could provide
stall region between stall and α = 15°, where stall cells can cause accurate readings without the need for corrections has been
inconsistencies in force measurements and differences between investigated. A data set was judged to be sufficiently accurate if the
runs were as high as 8%, differences between repeated runs for the maximum correction to any coefficient reading was smaller than
same aerofoil and tunnel configuration were within this maximum. 5%. The adjustment to drag at α = 90° is typically the largest, both
in magnitude and by percentage, so by limiting this to 5%, all other
adjustments can be judged to be at least similarly limited. Working
4. Blockage corrections from the corrections of Rae et al. (1984), a quartic inequality in c /H ,
given by
When the tunnel is configured as a conventional closed test
section, results have been corrected for lift interference, solid ⎛ c ⎞4 ⎛ c ⎞3 2 ⎛ c ⎞2
0.95ζ 2⎜ ⎟ + 1.9K wcCduζ ⎜ ⎟ + (2.9ζ + 0.95K wc
2
Cdu )⎜ ⎟
blockage and wake blockage using formulas given in Rae et al. ⎝H ⎠ ⎝H ⎠ ⎝H ⎠
(1984). Total blockage (ϵsc + ϵ wc ) is assumed by Rae et al. to be
small enough to neglect terms of its square. These have been ⎛c ⎞
restored for the current study as the assumption does not hold + 1.9K wcCdu⎜⎜ ⎟⎟ < 0.05
post-stall. ⎝H ⎠ (1)
Compressibility factors have been omitted throughout due to
was established, where
the low Mach numbers of the experiments (maximum 0.09). Data
reproduced from other sources has been subjected to equivalent π2
ζ=Λ
corrections by the original authors, other than Pope who used a 48 (2)
tunnel equipped with a “breather” to atmosphere that minimised
solid and wake blockage, so corrected for streamline curvature Λ is a constant for a given profile, equal to 0.237 for the NACA 0012
only. (Allen and Vincenti, 1944), K wc is the wake blockage ratio factor, a
constant specified as 0.25 by Rae et al. (1984), and Cdu is the
The force balance requires the attachment of a pitch strut to the
uncorrected drag coefficient. See Appendix A for a full derivation.
rear of the aerofoil under test to measure moment. An extended
The inequality can be solved for c /H for given values of Λ and
strut stretching through the lower plenum chamber of the tunnel
Cdu. For a NACA 0012 aerofoil with Cdu = 2 at α = 90° the result is
would have fouled on slat aerofoils at certain OARs. Though the
c /H < 5.2% . This is prohibitively low, the wind tunnel used in the
correction formula for incidence is reliant on moment, omitting it
experiments detailed in this paper has a 457 mm square cross
from the formula was found to be acceptable, causing a maximum
section, so would require an aerofoil chord of less than 23.8 mm to
difference in corrected incidence of 0.1°. No other corrections are
achieve a ratio within the acceptable range. Small aerofoils are
moment-dependent. Because of this, and given that the moment is
harder to profile accurately and are a limiting factor on the max-
of limited importance to VAWT modelling, no attempt was made
imum Reynolds number achievable with a given wind tunnel.
to measure it.

4.1. Limitations of corrections


5. Results
The corrections used are based on potential flow theory, and so
rely on the following assumptions: Measurements of lift and drag were taken for the c /H = 0.27
and c /H = 0.32 aerofoils for 0° < α < 180°, the test section config-
 Flow is inviscid, a reasonable assumption for high Reynolds ured with a slatted ceiling of OAR 0.43, 0.51, 0.59, 0.67 and 0.84
number flows away from boundary layers. (tunnel configurations denoted C-0.43, C-0.51, C-0.59, C-0.67 and
 Flow is incompressible, acceptable for Mach numbers less than C-0.84 respectively); with a slatted floor and ceiling of OAR 0.43,
0.3. 0.51, 0.59, and 0.67 (denoted FC-0.43, FC-0.51, FC-0.59, and FC-
 Flow is fully attached, appropriate only up to the incidence at 0.67) and with conventional solid walls (where blockage correc-
which separation occurs. tions are applied to data).
 The wind tunnel test section is sufficiently long for flows far
upstream and downstream of the model to be parallel to the 5.1. Selection of data for inclusion in this report
tunnel walls. Suitability is obviously apparatus dependent.
 The aerofoil is modelled as a single vortex, source or doublet, so Results are presented for the solid walled tunnel (raw data and
it must be small in comparison to the tunnel. This is inap- with corrections applied) and for C-0.59, FC-0.43 and FC-0.59
propriate for larger c /H ratios. configurations. Solid walled data is included to allow direct com-
parisons to existing studies. The C-0.59 and FC-0.43 configurations
ESDU (1978) recommends caution when using corrections for were chosen for inclusion as the best of the C and FC configura-
separated flows, stating that practical use has been made of tions, based on experimental results (see Section 6 for
214 J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218

Fig. 8. Cl vs. α for selected post-stall studies.


Fig. 6. Effect of aerofoil size on ratio of lift coefficients (theory)/ratio of lift curve
slopes (experiments).
so results from the FC-0.59 arrangement are also presented as this
is the nearest OAR achievable with the current apparatus.
methodology used in this judgement) and through use of a
numerical simulation of the tolerant tunnel. Calculations were
completed using a source/vortex panel method based on Hess and 5.2. Comparisons to existing data
Smith (1967). Fig. 6 shows the code output for the three config-
urations in the form of a plot of normalised lift curve slope against The apparatus has been validated by comparing results from the
c /H . The normalised lift curve slope is the gradient of the pre-stall solid walled test section configuration to existing data taken in
plot of Cl against α α from a given tunnel type, divided by the same similar conventional tunnels. Fig. 7 shows results for the c /H = 0.32
gradient for unconstrained flow: aerofoil taken in the solid walled tunnel, corrected for blockage
using the methods of Section 4. Also included are the data of Shel-
a0 Cl/α
= dahl and Klimas (1981) (as before, taken at Re = 3.6 × 105) and
a 0FA ClFA/αFA (3) Althaus's (1980) study taken at Re = 1.5 × 105. As well as the same
Unconstrained flow variables are given the subscript “FA” for “free Reynolds number, it also shares similar c /H and aspect ratios (see
air”. Table 1) to the larger aerofoil of the current study.
C-0.59 and FC-0.43 give values of a0/a0FA near to unity for Lift of the Re = 1.5 × 105 studies is in close agreement pre-stall,
c /H < 0.35 for the slatted ceiling and slatted floor and ceiling with a gradient steeper than 2π up to α ≈ 2°, then a shallower
arrangements of the tunnel respectively. This implies that they gradient between there and stall, occurring at α = 11° with a ClMAX
replicate the free stream well for aerofoil/tunnel size pairings of of approximately 0.9. Stall ClMAX is consistent with the trends
the order used in the experiments detailed in this paper. established between it and Reynolds number in McCroskey's
Since the code uses a potential flow method, it is unsuitable for review of wind tunnel results for the NACA 0012 (McCroskey,
judging performance in the post-stall range of incidences. As 1987). The steep gradient at α < 2° is caused by a laminar
mentioned, Parkinson's bluff body tunnel with slated floor and separation bubble on the suction surface of the aerofoil which
ceiling was found to work best with OARs of 0.5–0.7. An OAR of moves towards the leading edge with increasing incidence. This is
0.563 was judged to be the best overall for flat plates with c /H of a feature of the NACA 0012's performance at low Reynolds num-
up to 0.33 (Hameury, 1987). A similar tunnel should be capable bers (none of the steeper gradient is exhibited by Sheldahl and
of handling similar sized aerofoils at extreme angles of incidence, Klimas's, 1981 data taken at Re = 7.0 × 105).

Fig. 7. Cl and Cd vs. α for selected low Reynolds number studies.


J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218 215

Fig. 9. Cd vs. α for selected post-stall studies.


Fig. 11. Cl and Cd vs. α for c /H = 0.27 and c /H = 0.32 aerofoils taken in the C-0.59
tolerant tunnel.
The agreement in the region 10° < α < 15° is reasonable; the
leading edge stall behaviour of the NACA 0012 makes small dif- stall lift and drag peak magnitude, but Sheldahl and Klimas (1981)
ferences here unsurprising. The agreement between Sheldahl and completed experiments at three Reynolds numbers (5 × 105 and
Klimas (1981) and the others is surprising given the much larger
7 × 105 alongside the 3.6 × 105 reproduced here) and found post-
Reynolds number used and is likely due to the compensatory
stall performance to be independent of the variable.
effect of the difference in blockage, with a c /H of 0.07 compared to
Results for drag are similar to lift: pre-stall all studies are in
0.32–0.33 for the current study and Althaus (1980).
reasonable agreement, while post-stall results for the c /H = 0.32
Figs. 8 and 9 show plots of lift and drag respectively covering 0–
aerofoil are similar to those of Critzos et al. but higher than
180° of incidence. Included are results for the c /H = 0.32 aerofoil,
Sheldahl and Klimas's.
alongside Sheldahl and Klimas (1981) (as in Fig. 7) and Critzos et
al.'s (1955) Re = 1.8 × 106 study. The second lift peak of the current
5.3. Solid walled tunnel
study at α ≈ 45° is similar in magnitude to that of Critzos et al. but
higher than that of Sheldahl and Klimas. The two earlier studies
Fig. 10(a) shows results from the solid walled tunnel for the
share a common c /H of 0.07 so the differences between them are
c /H = 0.27 and c /H = 0.32 aerofoils, both raw and corrected,
surprising. Reynolds number effects could be responsible, Critzos
labelled “r” and “c” respectively. There is a divergence in the
et al. (1955) completed experiments at Re = 5 × 105 as well as
characteristics of the two aerofoils at incidences below 15.5°,
1.8 × 106 in the same tunnel(the Langley LTPT, see Table 1), finding
particularly in the region 3.5° < α < 15.5°. Fig. 10(b) shows details
a slight positive correlation between Reynolds number and post-
from the lower incidence end of the plots up to α = 30°. The
smaller aerofoil has a rougher surface finish, the differences in the
performance at low incidence are consistent with those between
smooth and rough aerofoils (Hoerner, 1985). The unusual beha-
viour of the c /H = 0.27 aerofoil persisted in all tunnel configura-
tions; pre-stall results for it are therefore not analysed further.
Post-stall data for the two aerofoils is comparable since surface
roughness has little impact on separated flow.
Blockage corrections collapse the data from the two aerofoils
well in the post-stall region. The raw drag peak is 18% larger for
the larger aerofoil, this difference reduces to 4% for the corrected
peaks. As the studies presented in Fig. 2 and Table 1 suggest, post-
stall lift and drag peaks are higher for the higher blockage
experiment even after corrections have been applied.

Fig. 10. Cl and Cd vs. α for c /H = 0.27 and c /H = 0.32 aerofoils taken in the solid
walled test section, raw data (r) and corrected data (c). (a) Full range of incidence Fig. 12. Cl and Cd vs. α for c /H = 0.27 and c /H = 0.32 aerofoils taken in the FC-0.43
tested, 0 ≤ α ≤ 180° . (b) Detail in the range 0 ≤ α ≤ 30°. tolerant tunnel.
216 J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218

Table 2
Absolute differences between c /H = 0.27 and c /H = 0.32 data for selected tunnels.

Tunnel configuration Sum of differences, 20° < α < 160°

Cl Cd Total

Solid walled 3.77 6.13 9.89


C-0.59 0.37 0.80 1.17
FC-0.43 0.19 0.37 0.56
FC-0.59 0.43 0.52 0.95

from a reference were assessed to see which tunnel provided the


minimum deviation.
Here, the focus is on force measurements of lift and drag and the
tunnels must be judged on their performance over the full range of
Fig. 13. Cl and Cd vs. α for c /H = 0.27 and c /H = 0.32 aerofoils taken in the FC-0.59
post-stall incidences, not just at α = 90°. Since only two aerofoils
tolerant tunnel.
have been tested, there is no need for a standard deviation analysis,
differences between the two are a sufficient measure of data spread.
5.4. Tolerant tunnel As such, Table 2 has been compiled, containing sums of the absolute
differences between readings of lift and drag for the two aerofoils at
Fig. 11 contains results from the C-059 configured tolerant tun- each data point in the range 20° < α < 160°. FC-0.43 is shown to
nel, alongside the corrected results from the solid walled tunnel for give the best results over the range of incidences.
the c /H = 0.32 aerofoil to allow comparison with earlier figures. The Pre-stall, an experimental equivalent of the a0/a0FA output of
plots are similar to the corrected solid walled tunnel; the spread of the tunnel panel code has been calculated with m obtained from a
data for the two aerofoils is far less than for the raw solid walled
linear regression analysis of results in the range of incidences
data, with higher post-stall peaks (4% higher at the drag peak) for
−2° ≤ α ≤ 6° and a0FA approximated using thin aerofoil theory.
the larger aerofoil. Pre-stall results for the large aerofoil are very
Results for solid walled, C-0.59 and FC-0.43 tunnels can be seen in
similar to the corrected data, the gradient of the curve being slightly
Fig. 6. Due to the surface roughness issue inducing shallow pre-
lower, with stall occurring earlier (at 10.3° opposed to 10.8°) and at a
stall gradients for the c /H = 0.27 aerofoil, its results are lower than
lower peak (0.85 vs. 0.89). Both aerofoils exhibit lower post-stall lift
theory for all tunnel configurations. The results for the c /H = 0.32
peaks than the corrected solid-walled data. There is strong agree-
aerofoil are near-identical for the solid walled and FC-0.43 tunnels,
ment between the two tunnel configurations for drag of the
but the C-0.59 does not correct the lift curve gradient as much as
c /H = 0.32 aerofoil, with identical drag coefficient peaks of 2.04 at
predicted.
α = 90°.
The short L /H ratio of the test section used for these experi-
Figs. 12 and 13 show results for the FC-0.43 and FC-059 double
ments (2.14, compared to 2.67 for Parkinson's original) is the likely
slotted wall configurations. There is a striking data collapse for the
cause of this less-than-expected reduction of blockage. When
FC-0.43 tunnel, with results nearly identical for the two aerofoils
building tunnels to Parkinson's design, others have found an
for post-stall lift and for drag, other than for 60° < α < 110° where
insufficiently long slatted section to cause similar problems (Par-
the larger aerofoil exhibits a 5% higher drag peak. Hameury (1987)
kinson et al., 1992).
found that a 33% blockage normal flat plate diverted flow to such
The results from the tolerant tunnel suggest that its use is valid
an extent that it caused some of the slat aerofoils of his tunnel to
for the testing of post-stall aerofoils, the C-0.59 configuration
stall, suggesting that this is beyond the limit of tolerance of the
tunnel. At these extreme incidences, the largest aerofoil could be providing results very close to those achieved using current best
similarly affecting the slats of the tunnel used in these experi- practice, without reliance on correction methodologies that are
ments. Though agreement between the aerofoils is excellent, post- not strictly applicable in the post-stall region. Current best practice
stall peaks are significantly lower than that for the data from is well established for unstalled aerofoils where flow is attached
conventional tunnels, the c /H = 0.32 lift peak is 21% higher for the and corrections can be used with confidence. The tolerant tunnel
corrected solid-walled data while drag peak is 23% higher. Pre-stall provides results comparable to the corrected conventional tunnel
results for the larger aerofoil are close to those from the corrected data in this region of incidences, with no need for corrections.
solid-walled tunnel and the C-0.59 configuration, with a ClMAX of Judging the tunnels on their ability to collapse data from dif-
0.85 at 10.5°. ferent model sizes, the FC-0.43 tunnel is best, but post-stall peaks
The collapse of data for the FC-0.59 tunnel is less impressive, in lift and drag are significantly lower than those taken in solid-
with a greater spread between post-stall lift and drag, and a lower walled tunnels.
stall point (ClMAX = 0.82 at α ¼10.20) than for the other
configurations.
7. Conclusions

6. Discussion Results have been presented in a study to establish a reliable


set of post-stall aerofoil data, using conventional, solid walled
Hameury selected the FC-0.563 tunnel configuration by testing tunnels and a passive, tolerant design with slatted walls. The data
three pressure tapped flat plates with c /H of 0.083, 0.194 and 0.333 taken using the conventional tunnel agrees closely to prior studies
normal to the flow. Various parameters, including base pressure in the pre-stall range of incidences and is within a spread of
coefficients, drag coefficients, Strouhal number and required published data in the post-stall range. Results are promising for
blockage corrections were plotted against OAR and the tunnel the tolerant tunnel, with data close to the best data achieved with
providing the most similar results for the range of model sizes was solid walled tunnels, but produced in a manner suited to both
judged to be the best. Standard deviation of pressure distributions attached and separated flow.
J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218 217

The C-0.59 and FC-0.43 tunnel configurations both give pre- Using Eq. (A.3) and expanding (1 + ϵB)2 gives
stall lift curve slopes and stall angles nearly equal to corrected,
(1 + ϵB)2 = 1 + 2ϵsc + 2ϵ wc + ϵ2sc + 2ϵsc ϵ wc + ϵ2wc (A.5)
solid walled data, with no need for corrections. The C-0.59 con-
figuration again gives post-stall results close to corrected, solid Substituting (A.5) in (A.4) and rearranging:
walled data with no need for corrections. Judging the tunnel
configurations by Parkinson's own criteria, the FC-0.43 performs 2.9ϵsc + 1.9ϵ wc + 0.95ϵ2sc + 1.9ϵsc ϵ wc + 0.95ϵ2wc < 0.05 (A.6)
best, producing the most similar data values for two different sized
Wake blockage is defined in terms of c /H (Rae et al., 1984):
models. Post-stall the FC-0.43 gives peaks in lift and drag sig-
nificantly lower than corrected, solid walled data. c
ϵ wc = K wc Cdu
Parkinson's tolerant tunnel design has been applied to aerofoils H (A.7)
in the deep stall region of incidences for the first time. The quality of Through manipulation, ϵsc can be given similarly
the data obtained suggests that as a test of concept, the study has
2
been successful and the tunnel design is suitable for deep stall ⎛c ⎞
ϵsc = ζ ⎜ ⎟
experiments. Given that no reliance is placed in corrections that are ⎝H ⎠ (A.8)
acknowledged to be unsuitable for separated flows, it presents a
with ζ defined in Eq. (2).
better way of conducting post-stall aerofoil research in the future.
Substituting Eqs. (A.7) and (A.8) into (A.6) and rearranging
With further refinement the design could be used to provide post-
gives a quartic inequality in c /H :
stall data of sufficient quality to accurately model the start-up of
VAWTs. ⎛ c ⎞4 ⎛ c ⎞3 2 ⎛ c ⎞2
0.95ζ 2⎜ ⎟ + 1.9K wcCduζ ⎜ ⎟ + (2.9ζ + 0.95K wc
2
Cdu )⎜ ⎟
Using the new data from this study and a thorough review of ⎝H ⎠ ⎝H ⎠ ⎝H ⎠
existing NACA 0012 studies, a positive correlation between post-
stall lift and drag peak magnitudes and blockage is hypothesised ⎛c ⎞
which persists even after the application of corrections. + 1.9K wcCdu⎜⎜ ⎟⎟ < 0.05
⎝H ⎠ (A.9)

Data download
Appendix B. Supplementary data
Data from Figs. 10(a)– 13 is available for download in .xls for-
mat from https://www.researchgate.net/publication/277555480_ Supplementary data associated with this article can be found in
Data_from_Blockage-tolerant_wind_tunnel_measurements_for_a_ the online version at http://dx.doi.org/10.1016/j.jweia.2015.06.006.
NACA_0012_at_high_angles_of_attack.

References
Acknowledgements
Allen, H., Vincenti, W., 1944. Wall Interference in a Two-Dimensional-Flow Wind
This work was supported by the Environmental Services Tunnel, With Consideration of the Effect of Compressibility. Technical Report TR
Association Education Trust. 782, NACA.
Althaus, D., 1980. Profilpolaren für den Modellflug, vol. 1. Neckar-Verlag, Villingen-
Schwenningen.
Bergeles, G., Athanassiadis, N., Michos, A., 1983. Aerodynamic characteristics of
Appendix A. Derivation of a formula for calculating the NACA 0012 aircraft in relation to wind generators. Wind Eng. 7 (4), 247–262.
Bianchini, A., Ferrari, L., Magnani, S., 2011. Start-up behavior of a three-bladed
required c /H ratio for blockage corrections to be less than 5% H-Darrieus VAWT: experimental and numerical analysis. In: Proceedings of the
ASME Turbo Expo, pp. 6–10.
First, an inequality is set up as Critzos, C., Heyson, H., Boswinkle, R., 1955. Aerodynamic Characteristics of NACA
0012 Airfoil Section at Angles of Attack from 0 to 180 Degrees. Technical Report
Cdu − Cd TN 3361, NACA.
< 0.05 Dominy, R., Lunt, P., Bickerdyke, A., Dominy, J., 2007. Self-starting capability of a
Cdu (A.1) Darrieus turbine. Proc. Inst. Mech. Eng. Part A: J. Power Energy 221 (1), 111–120.
ESDU, 1978. Lift-Interference and Blockage Corrections for Two-Dimensional Sub-
For drag data taken using a force balance in a solid walled tunnel sonic Flow in Ventilated and Closed Wind-Tunnels. Technical Report 76028,
(when comparing to source, note that zero terms for solid walled Engineering Sciences Data Unit.
tunnels have been omitted here and terms of ϵ2B have been inclu- Hameury, M., 1987. Development of the Tolerant Wind Tunnel for Bluff Body
Testing (Ph.D. thesis). University of British Columbia.
ded here but are omitted in the source) (Rae et al., 1984): Hess, J., Smith, A., 1967. Calculation of potential flow about arbitrary bodies. Progr.
Cdu(1 − ϵsc ) Aerospace Sci. 8, 1–138.
Cd = Hill, N., Dominy, R., Ingram, G., Dominy, J., 2009. Darrieus turbines: the physics of
(1 + ϵB)2 (A.2) self-starting. Proc. Inst. Mech. Eng. Part A: J. Power Energy 223 (A1), 21–29.
Hoerner, S., 1985. Fluid-Dynamic Lift. Hoerner Fluid Dynamics, Vancouver.
where ϵB is the total blockage factor (Rae et al., 1984): 〈http://www.urbangreenenergy.com/products/visionair3〉.
Lindenburg, C., 2000. Stall coefficients: aerodynamic airfoil coefficients at large
ϵB = ϵsc + ϵ wc (A.3) angles of attack. In: IEA Symposium on the Aerodynamics of Wind Turbines.
National Renewable Energy Laboratory, Colorado, USA.
and ϵsc and ϵwc are the solid blockage and wake blockage ratios Malek, A., 1983. An Investigation of the Theoretical and Experimental Aerodynamic
Characteristics of a Low-Correction Wind Tunnel Wall Configuration for Airfoil
respectively. Substituting Eq. (A.2) into Eq. (A.1):
Testing (Ph.D. Thesis). University of British Columbia.
Cdu(1 − ϵsc ) Maskell, E., 1963. A Theory of the Blockage Effects on Bluff Bodies and Stalled Wings
Cdu − in a Closed Wind Tunnel. Technical Report R. & M. No. 3400, British ARC.
(1 + ϵB)2 Massini, G., Rossi, E., D'Angelo, S., 1988. Wind tunnel measurements of aerodynamic
< 0.05 coefficients of asymmetrical airfoil sections for wind turbine blades extended
Cdu
to high angles of attack. In: European Community Wind Energy Conference,
Herning, Denmark, pp. 241–245.
McCroskey, W., 1987. A Critical Assessment of Wind Tunnel Results for the NACA
⟹0.95(1 + ϵB)2 < 1 − ϵsc (A.4) 0012 Airfoil. Technical Report, NASA Technical Memorandum 100019.
218 J.M. Rainbird et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 209–218

Parkinson, G., 1984. A tolerant wind tunnel for industrial aerodynamics. J. Wind Spera, D., 2008. Models of Lift and Drag Coefficients of Stalled and Unstalled Airfoils
Eng. Ind. Aerodyn. 16 (2–3), 293–300. in Wind Turbines and Wind Tunnels. Technical Report NASA/CR-2008-215434,
Parkinson, G., 1990. The tolerant tunnel: concept and performance. Can. Aeronaut. NASA.
Space J. 36 (3), 130–134. Strickland, J., 1975. Darrieus Turbine: A Performance Prediction Model Using Mul-
Parkinson, G., Kong, L., Cook, N., 1992. Configuration criteria for a blockage-tolerant tiple Streamtubes. Technical Report SAND75-0431, Sandia National
wind tunnel. J. Wind Eng. Ind. Aerodyn. 40 (2), 215–219. Laboratories.
Poisson-Quinton, P., de Sievers, A., 1967. Étude aerodynamique d'un élément de Timmer, W., 2010. Aerodynamic characteristics of wind turbine blade airfoils at
pale d'hélicoptère. In: AGARD Conference Proceedings No. 22, Göttingen, Ger- high angles-of-attack. In: 3rd EWEA Conference-Torque 2010: The Science of
many, pp. 4.1–4.35. Making Torque from Wind. European Wind Energy Association, Heraklion,
Pope, A., 1947. The Forces and Pressures over a NACA 0015 Airfoil Through 180 Crete, Greece.
Degrees Angle of Attack. Technical Report E-102, Georgia Institute of Viterna, L., Corrigan, R., 1982. Fixed pitch rotor performance of large horizontal axis
Technology. wind turbines. In: Workshop on Large Horizontal-Axis Wind Turbines, vol. 1,
Rae, W., Pope, A., Barlow, J., 1984. Low Speed Wind Tunnel Testing, 2nd ed. Wiley, pp. 69–85.
New York, pp. 349–361. Weihs, D., Katz, J., 1983. Cellular patterns in post-stall flow over unswept wings.
Rainbird, J., 2007. The Aerodynamic Development of a Vertical Axis Wind Turbine. AIAA J. 21 (12), 1757–1759.
MEng Project Report, University of Durham, UK. Williams, C., 1975. A New Slotted-Wall Method for Producing Low Boundary Cor-
Rossetti, A., Pavesi, G., 2013. Comparison of different numerical approaches to the rections in Two-Dimensional Airfoil Testing (Ph.D. thesis). University of British
study of the H-Darrieus turbines start-up. Renew. Energy 50, 7–19. Columbia.
Sheldahl, R., Klimas, P., 1981. Aerodynamic Characteristics of Seven Symmetrical Zhou, Y., Alam, M., Yang, H., Guo, H., Wood, D., 2011. Fluid forces on a very low
Airfoil Sections Through 180-Degree Angle of Attack for Use in Aerodynamic Reynolds number airfoil and their prediction. Int. J. Heat Fluid Flow 32 (1),
Analysis of Vertical Axis Wind Turbines. Technical Report SAND80-2114, Sandia 329–339.
National Laboratories.

You might also like