wind tunnel blockage corrections
TRANSCRIPT
Wind tunnel blockage corrections: Review and applicationto Savonius vertical-axis wind turbines
Ian Ross n, Aaron Altman
Mechanical and Aerospace Engineering, University of Dayton, 300 College Park, Dayton, OH 45469-0238, USA
a r t i c l e i n f o
Article history:
Received 28 July 2010
Received in revised form
31 December 2010
Accepted 26 February 2011Available online 1 April 2011
Keywords:
Low speed wind tunnel
Wind tunnel blockage corrections
Vertical-axis wind turbine
Aerodynamics
Bluff-body aerodynamics
Savonius
a b s t r a c t
An investigation into wake and solid blockage effects of vertical axis wind turbines (VAWTs) in closed
test-section wind tunnel testing is described. Static wall pressures have been used to derive velocity
increments along wind tunnel test section which in turn are applied to provide evidence of wake
interference characteristics of rotating bodies interacting within this spatially restricted domain.
Vertical-axis wind turbines present a unique aerodynamic obstruction in wind tunnel testing, whose
blockage effects have not yet extensively investigated. The flowfield surrounding these wind turbines is
asymmetric, periodic, unsteady, separated and highly turbulent. Static pressure measurements are
taken along a test-section sidewall to provide a pressure signature of the test models under varying
rotor tip-speed ratios (freestream conditions and model RPMs). Wake characteristics and VAWT
performance produced by the same vertical-axis wind turbine concept tested at different physical
scales and in two different wind tunnels are investigated in an attempt to provide some guidance on
the scaling of the combined effects on blockage. This investigation provides evidence of the effects of
large wall interactions and wake propagation caused by these models at well below generally accepted
standard blockage figures.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Aerodynamics is an active and influential science, contributing
to major aspects of wind turbine design. For an aerodynamicist
the art of manipulating and adapting a moving fluid to optimize
energy extraction can be challenging to achieve. Wind turbines
have been researched since the earliest known ancient humans
attempted to harness wind energy through diversified means.
One of the manners to achieve this goal was through vertical-axis
wind turbines (VAWT). The present research details the evolu-
tionary steps in improving the efficiency of wind tunnel testing
vertical-axis wind turbines. Fig. 1 (CAD models based on the
designs of TFC energy), displays two such VAWT models similar in
concept to designs devised by the Finnish engineer Sigurd J. Savonius
(Savonius, 1931).
Recently there has been a resurgence of interest regarding
sources of renewable energy, with numerous universities, com-
panies and research institutions carrying out extensive research
activities. These activities have led to a plethora of designs of
wind turbines based mostly on computational aerodynamic
models. Still largely restricted to an experimental subject, verti-
cal-axis wind turbines are appearing more frequently in the
civilian and military market as research into their cost-effective-
ness and simplicity progresses.
At present, there are two primary categories of modern wind
turbines, namely horizontal-axis (HAWT’s) and vertical-axis
(VAWT’s) wind turbines. The main advantages of the VAWT is
its single moving part (rotor), where no yaw mechanisms are
required, its low-wind speed operation and the elimination of the
need for extensive supporting tower structures, thus significantly
simplifying the design and installation. Blades of straight-bladed
VAWTs can be of uniform airfoil section and untwisted, making
them relatively easy to fabricate or extrude, unlike the blades of
HAWTs, which are commonly twisted and tapered airfoils for
optimum performance.
The motivation for the current research stems from an inves-
tigation into the flow blockage influence on performance of
relatively inefficient VAWTs. In order to improve the conceptual
approach, previous knowledge of bluff body aerodynamics has
been applied to a rotational frame-of-reference for VAWT con-
cepts. Savonius stated in his 1931 paper published by the Journal
of Mechanical Engineering, ‘‘The S-Rotor and its application’’, that the
maximum efficiency possible was only 31%. Following Savonius,
numerous others investigated the effect of geometric parameters
such as blade numbers, blade gap-size and overlap ratio upon flow
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/jweia
Journal of Wind Engineeringand Industrial Aerodynamics
0167-6105/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2011.02.002
Abbreviations: HAWT, Horizontal-axis wind turbine; TSR, Tip speed ratio; LSWT,
Low-speed wind tunnel; TFCE, Twenty first century energy; RPM, Revolutions per
minute; VAWT, Vertical-axis wind turbinen Corresponding author. Tel.: þ1 734 478 1734.
E-mail addresses: [email protected] (I. Ross),
[email protected] (A. Altman).
J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538
behavior. Due to the complex nature of flowfield around Savonius
turbine produced by its geometrical shape, Fig. 2 (Fujisawa and
Gotoh, 1992), theoretical work in modeling the aerodynamics
around these wind turbines is indeed quite scarce. There are
several developed theories to analyze the Darrieus and propeller
type turbines, where lift is the dominant force. The blade element
momentum theory (BEM) predicts the performance of a Darrieus
turbine well, following the ‘‘Wind Energy Handbook’’ (Burton et al.,
2001). In order to perform an aerodynamic analysis of flowfields
around VAWTs and their interaction with closed test-section
wind tunnels, a sample batch of wind turbines/bluff body geome-
trical shapes have been constructed in the present study for wind
tunnel testing, involving:
(1) a qualitative comparison of tip-speed ratio as a function of
Reynolds number using flow visualization of the wake and
flow regions around VAWTs and
(2) an investigation into the relationships that exist between tip-
speed ratio, wind tunnel velocity, the coefficient of power, the
coefficient of torque and static pressures inside a wind tunnel
test-section with the influence of blockage ratio.
A solid blockage effect is commonly observed in wind tunnel
testing that in turn produces an increase in the local wind velocity
in the working section. This increase is ideally accounted for by a
theoretical wind tunnel blockage factor or ratio of which several
developed techniques will be discussed later. Numerous accounts
of questionable accuracy have been debated throughout the
literature concerning low-speed wind tunnel testing of rotating
bluff bodies, especially of VAWT types. The possibility of pre-
viously undocumented variable deleterious effects of wind tunnel
blockage in VAWT testing is observed in tests performed for this
paper and are subsequently presented and discussed. Severe
effects will be documented when models occupy a percentage
of the tunnel cross-sectional area significantly less than the
presently accepted heuristics.
The present research details evolutionary steps in improving
the practicality in testing sub-scale VAWTs as well as an inves-
tigation into the methodology behind correcting for the flow
constraining effect. Ideology has been studied from the ‘Wall
pressure signature method’ originally proposed by Hackett et al.
(1979). The goal is to advance existing solid/wake blockage
correction methodologies in order to appropriately or knowledge-
ably apply them to rotating test models and VAWT concepts,
which exhibit unique viscous and unsteady turbulent flow con-
ditions. The question remains: ‘‘Do the wind tunnel wall surfaces
interact with the model flow to the extent of impacting the
efficiency of the rotor, therefore calling into question accurate
comparison to real-scale prototypes?’’
Solid blockage is created by a reduction in the test-section area
for flow to pass compared to an undisturbed freestream. By
continuity, Bernoulli’s equation and all of the associated assump-
tions, velocity increases in the vicinity of a model, Fig. 3 (upper)
(flow sketches based on the observations of Fujisawa and Gotoh
(1992) and descriptions detailed in ESDU 80024, 1998).
Nomenclature
a blade overlap
AR aspect ratio
Aswept swept area of a turbine
C rotor blade chord
Cp pressure coefficient
Dr turbine (rotor) diameter
D0 end plate diameter
H height of turbine
r turbine radius
Re reynolds number
T torque
UN
freestream velocity
qN
freestream dynamic pressure
l tip speed ratio
Fig. 1. Savonius vertical-axis wind turbine concepts TFC energy: (left) 3-bladed
and (right) 2-bladed conventional Savonius—(CAD models based on the designs of
TFC energy).
Fig. 2. Savonius rotor configuration and geometrical parameters (Fujisawa and
Gotoh, 1992).
Fig. 3. VAWT influence upon streamlines (top) solid blockage, (bottom) wake
blockage (flow sketches based on the observations of Fujisawa and Gotoh, 1992
and descriptions detailed in ESDU 80024, 1980).
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538524
Wake blockage is difficult to model blockage for a stationary
body, but when the test concerns a dynamic rotating bluff body
producing large wake disturbances the modeling becomes increas-
ingly more difficult to predict the degradable effect on the flow.
The wake generated has a lower mean velocity than the free-
stream. By continuity, velocity outside this wake has a higher
speed than the flow inside the wake region for a constant mass
flow and higher velocity in the freestream yields lower pressure
assuming conditions that satisfy Bernoulli’s equation, Fig. 3 (lower)
The following is a summation of formulae adopted by numer-
ous authors in their wind tunnel experimental campaigns of
VAWTs, which appears to be the de facto standard in wind
turbine data reduction. The formulae given below were adapted
from those originally given by Blackwell et al. (1997) and
modified for unit conversions from RPM to radians per second:
freestream dynamic pressure q1 ¼1
2rairU
21 ð1Þ
tip-speed ratio l¼ROðrad=sÞ
U1
¼O
60pDR =U1
ÿ � ð2Þ
power extracted Pextracted ¼OT ¼O
60ð2pÞrad=s
� �
T ð3Þ
power available Pavailable ¼ q1U1Aswept ð4Þ
power coefficient ðmeasure of efficiencyÞ CPðefficiencyÞ ¼PextractedPavailable
¼O=60 ð2pÞrad=sÿ �
T
q1U1Aswept¼ lCT ð5Þ
(Blackwell et al., 1997) Modification of a blockage corrected
freestream velocity and dynamic pressure
V1 ¼ V1 uncorrectedð1þeÞ
q1 ¼ q1 uncorrectedð1þeÞ2 ð7Þ
where Aswept ¼DRHR and epsilon (blockage correction factor) shall
be discussed in a later section.
2. Review of literature
Previous means have been proposed to analyze performance
optimization of HAWTs. Progress has also been curved towards
VAWT applications concerning aerodynamic efficiency and per-
formance regarding flow separation and alleviating adverse
effects on energy production. There remains no extensive readily
available literature concerning specific Savonius aerodynamic
model applications to wind tunnel blockage corrections, but
rather there is literature concerning the generalities of the
Savonius rotor concept. A representative selection relevant to
the present research will be first reviewed.
Fujisawa and Gotoh (1992) experimented with flow visualiza-
tion for static and rotating Savonius two bladed rotors. The
rotation effect is discussed in comparison with the measured
pressure distribution on the blade surfaces. It was suggested that
the flow separation region observed on the blade surface was
reduced due to rotation and flow through the overlap. Fujisawa
and Gotoh explain how flow separation regions contribute to
torque production of the rotating rotor and weakened flow
through the overlap acts as a resistance, which together with
the stagnation effect on the front side contributes to the rotor’s
power production capability.
Finally it was shown that the attached flow region on the
convex side of the rotor grows with TSR, contributing to improved
torque performance at low TSR. In addition, deterioration of the
torque performance at large tip-speed ratios is caused by the
decrease in stagnation torque and in the pressure recovery effect
by flow through the overlap. The main flow was visualized by
smoke-wire and the wake flow by injecting smoke just upstream of
the rotor. The static and rotating observations were concentrated on
Fig. 4. Savonius flow patterns: (a) free stream flow, (b) internal flow, (c) flow model and (d) Cp distribution (coefficient of pressure—flow visualization used to compare
wake as a function of rotor angle and wind speed; (left) static rotor and (right) rotating) (Fujisawa and Gotoh, 1992).
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 525
smoke patterns inside the rotor; measured pressure coefficients on
blade surfaces and points of separation and stagnation can be seen
in Fig. 4 (Fujisawa and Gotoh, 1992), and are used as a baseline for
comparison to the flow visualization efforts in the present study.
Undertaking a United States Energy Research and Develop-
ment Contract, Blackwell et al. (199)7 carried out an in depth
investigation of low-speed wind tunnel testing of Savonius type
rotors of two/three stages and two/three blades at different
Reynolds numbers whilst measuring variables: torque, RPM and
tunnel conditions, whereby for the first time a wind tunnel
blockage correction factor has been employed in VAWT testing.
Blackwell presents the data in the form of power and torque
coefficients and as a function of speed ratio (or angular position
for static starting torques); it was concluded that increasing
Reynolds number and/or aspect ratio improves the performance.
Rotary positioning of the turbine to obtain the static or non-
rotating torque as a function of blade position relative to the free-
stream flow was also performed. At the time there was no
universally accepted length scale to Blackwell’s knowledge with
which to calculate a Re for a Savonius rotor and suggests that the
angular position at which stall occurs is a function of Reynolds
number. It was proposed that increasing the test Re number
generally improved aerodynamic performance across the range
of Re being tested.
The work of Biswas et al. (2007) proposed an in-depth review
of wind tunnel testing on three bladed Savonius designs, with
experiments conducted on blade overlap conditions in the range
16–35%. Power coefficients have been calculated with and with-
out a wind tunnel blockage correction factor for tunnel inter-
ference, which is cited from Blackwell’s earlier efforts. It has been
stated that tunnel blockage effect is an important parameter for
wind tunnel performance analysis of VAWTs, whose effect is
much more severe in low-speed wind tunnel applications. This
study provided results on the performance of rotor evaluated
from variation of Cp with TSR at various overlaps. Allowing for
the blockage correction, maximum coefficient of power was
reduced on average by 5% from its initial value, which is a
significant reduction when dealing with initial Cp values no larger
than 30%.
3. Review of existing wind tunnel blockage methodologies
It is defined that the total blockage correction factor is the sum
of the velocity increment (blockage factor) caused by wake
blockage and solid blockage; however these are incredibly diffi-
cult factors to assess for unusual geometries such as the Savonius
rotor and the associated flowfields around them. It has long been
a standard for low-speed wind tunnel testing to operate within
an area-ratio of (tunnel cross-section to swept area of a model)
1–10%, proposed by Pope and Harper, (1966) in their text
‘‘Low-Speed Wind Tunnel Testing’’ and earlier by Pankhurst and
Holder (1952) in their text ‘‘Wind-Tunnel Technique: An Account of
Experimental Methods in Low- and High-Speed Wind Tunnels’’, both
provided various solid/wake blockage correction techniques.
Two types of test-section commonly used when testing in
wind tunnels, namely the closed test-section and the open test-
section (or blockage tolerant test-section) provide large variations
when referring to blockage allowances. The open test-section or
open jet type of wind tunnel has the capability to allow the
conditions inside the test section to be largely unaffected by
larger blockage percentage static models because of the ability to
leak flow and expand the flow around objects within the test-
section as opposed to a flow constriction problem occurring with
the closed test-section type as shown in this study. Because of the
ability to allow the flow to expand, models can generally be
allowed to exhibit higher blockage percentage in open type
testing.
Sandia laboratories initiated the resurgence of vertical-axis
wind turbines in the United States. They also set the standard for
blockage corrections for VAWTs, this being a blockage correction
factor stated by Pope and Harper as a generic correction for the
testing of any unusual shape. The following section discusses this
original blockage method and focuses on the wall pressure
method (Hackett and Wilsden, 1975) modified for this study with
the aim of providing a more detailed assessment of partitioning
solid and wake blockage when the flow behavior increasingly
becomes more three-dimensional, highly separated, unsteady and
turbulent.
A review of recent developments in the calculation of low-
speed solid-wall wind tunnel interference conducted by Hackett
(2003) detailed an extensive interpretation of wall pressures by
Ashill and Weeks (1982), where it is shown by assuming x be
the distance in the streamwise direction and y the distance along
the wall in the direction normal to x (For vertical surfaces, z
replaces y and w replaces v). Here, Hackett assumes the pressure
p is at the wall using Prandtl’s classical assumption for boundary
layers and u and v are velocities in the x and y directions,
respectively.
3.1. Pope and Harper blockage correction factor
Correcting velocity readings Pope and Harper (1966) and
subsequent data modifications to allow for these changes are
shown:
velocity correction V ¼ Vu ð1þetÞ ð8Þ
dynamic pressure correction q¼ qu ð1þ2etÞ ð9Þ
reynolds number correction R¼ Ru ð1þetÞ ð10Þ
Drag coefficient correction: (From the dynamic pressure effect
plus the wake gradient term):
CD0 ¼ CD0u 1ÿ3esbÿ2ewbð Þ ð11Þ
et ¼ solid blockageþwake blockage¼ esb þ ewb ð12Þ
Pope explains: ‘‘for finding the blockage corrections for wind
tunnel models of unusual shapes the following is suggested:’’
et ¼1
4
model frontal area
test section areað13Þ
3.2. Maskell correction
Maskell (1965) was the first to address the problems with non-
streamline flow bodies, such as bluff-body testing in closed wind
tunnel sections and that of partially stalled shapes such as wings.
When the high-lift characteristics of particular delta wing aircraft
models of small aspect ratio were tested in different wind tunnels
at the Royal Aircraft Establishment (RAE), marked differences
were observed at the onset of stall beginning at the wing tips and
spreading inboard with increasing incidence. The different results
could be reconciled only through a wall interference factor, which
is equivalent to the increase in velocity of an undisturbed stream
much larger than previous standard estimations. Maskell’s
research goal was to establish a more convincing existence of
this interference factor and the need for corrections, by relating
effective increase in the dynamic pressure q of the stream due to a
solid blockage constraint. Maskell’s theory holds true for nearly
all two-dimensional bluff-body flows and for situations of close
axis symmetric wake downstream for three-dimensional flows
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538526
with the equation for corrected wind velocity given below.
Alexander (1978) provided an adaption to Maskell’s method by
comparing the drag of flat plates normal to the freestream and
that of the drag of Savonius rotors normal to the freestream,
applying the term m, an extrapolated value from Fig. 5. For small
values of blockage ratio, ðS=Cr0:045Þ Maskell gives m¼3.15
(constant value). Alexander suggests that due to restriction on
the wake by the tunnel walls at high S/C values the value of m
falls, reaching a value close to 2.0 for S/C¼0.3 (30% blockage).
V2c
V2¼
1
1ÿm S=Cÿ � ð14Þ
where: Vc is the corrected wind velocity, V the undisturbed wind
velocity, S the flat plate or wind tunnel maximum frontal area, C
the wind tunnel working section cross sectional area, V the
undisturbed wind velocity, m¼(B/S) the extrapolated value from
Fig. 5 and B the wake area normal to wind.
3.3. Hackett, Lilley and Wilsden method
Lockheed scientists Hackett, Lilley and Wilsden produced an
updated blockage correction methodology (Hensel, 1951), by adopt-
ing sources and sinks to represent an equivalent body surface in a
stream, and static pressures measured at the sidewalls are used to
construct a relatively simple singularity set to represent the test
article and then calculate the wall effects based on that singularity set
(Hackett et al., 1979). They showed that tunnel wall static pressures
may be used to infer wake geometry and hence wake blockage using
a row of pressures along the center of the tunnel sidewall, giving the
axial distributions of both solid and wake blockages with a velocity
peak just aft the model. Through a wind tunnel testing campaign
involving models of varying size and blockages up to 10%, wall
pressure signatures were used to determine source, sink and
strengths with wind tunnel span and locations.
Essentially the concept resolves pressure signatures into
their solid and wake counterparts signifying the symmetric and
Fig. 6. Effects at a wind tunnel wall of solid/bubble and viscous wake blockage (Hackett and Wilsden, 1975).
Fig. 5. Flat plates and rotors relationship of m vs. S/C (Alexander, 1978).
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 527
anti-symmetric regions with the parameters formulated from
these parts a velocity increment expression is obtained. Hackett
and Wilsden, (1975) provide a theoretical method in determining
wind tunnel solid/bubble and viscous blockage from wall and roof
pressure measurements (Fig. 6) involving lifting and non-lifting,
powered and non-powered models. In order to calculate corrected
pressure coefficients
CpcðxÞ ¼CpuðxÞÿ1
1þðDuðxÞ=U1Þ2
" #
þ1 rearranged for velocity increment :
Du
U
¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1ÿðCPemptyÿCPmodel
Þq
ÿ1 ð15Þ
4. Experimental set-up
A high precision VAWT test bed facility has been installed at
the University of Dayton low-speed wind tunnel laboratory,
housing an Eiffel-type tunnel with a contraction ratio of 16:1
and a working section 76 cm (30 in.)�76 cm (30 in.)�244 cm
(96 in.) length. The inlet freestream turbulence intensity is less
than 0.1% and tunnel maximum velocity is 40 m/s. Four VAWT
models have been considered in this study, with the aim to obtain
a data base of pressure signatures at varying fixed and dynamic
RPM operating conditions. Fig. 1 represents a rapid-prototype
model created for TFCE’s prototype being tested for area-ratio
effects on wind tunnel experiments, whereby three identical
models of varying size, 1/20th, 1/30th and 1/40th to a full-scale
prototype; blade height 3 m (100)� rotor diameter 6 m (200)
providing swept areas; 15.2 cm (6 in.)�30.5 cm (12 in.), 10.2 cm
(4 in.)�20.3 cm (8 in.) and 7.6 cm (3 in.)�15.2 cm (6 in.), respec-
tively, This produces solid blockage values in the University of
Dayton wind tunnel ranging from 2%, 3.5% and 8%. The 1/40th scale
model, 7.6 cm (3 in.)�15.2 cm (6 in.) has been tested in two wind
tunnels, the University of Dayton facility (producing a blockage of
3.5%) and in a smaller tunnel at TFCE laboratories an open circuit,
closed-test section wind tunnel having a working section 45.7 cm
(18 in.) height�45.7 cm (18 in.) width and tunnel maximum velo-
city of 45 m/s, producing a blockage of 5.5%. The larger 2-bladed
Savonius model has been compared for extreme blockage testing
conditions, occupying 10% of the wind tunnel cross-sectional area.
The test system can be seen in Fig. 7, situated below the wind
tunnel test-section; a spindle driven by the turbine passes through
an air bearing producing a theoretically non-friction system, con-
tinues into an Interface T11 bearingless rotary torque transducer
with a 2 Nm torque capacity and a magtrol hysteresis braking
system with a 3.15 Nm (450 oz-in) loading capability. Load is
electronically applied upon the hysteresis brake by the use of a
function generator, applying negative torque on the turbine accurate
to 0.01 V increments. Rotors are tested at constant RPM conditions
with varying freestream velocities and are tested under dynamic
Fig. 8. Static pressure wall tap locations.
Fig. 7. Turbine torque and RPM wind tunnel testing facility.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538528
loading and unloading conditions using a dataq acquisition system
interfaced with a PC for real-time analysis of the system.
Eighteen static pressure taps run along the center-line of the
test-section sidewall, with increased spatial frequency in close
proximity to the model, Fig. 8. Differential pressure readings are
digitally displayed accurate to 0.1 N/m2 using an AEROLAB pres-
sure transducer array system with 300 kN/m2 (3 bar) rated
transducers. These have been compared to an installed pitot-
static tube output display on a U-tube manometer reading of the
undisturbed freestream conditions forward of the model. The
traditional definitions of differential pressure and pressure coeffi-
cient were used following Anderson (2007)for the wall pressure
measurement analysis.
5. Results, analysis and discussion
The formulae presented in this paper have been applied to the
measured torque and RPM data from preliminary testing of four
concept models, in order to assess VAWT efficiencies and power
production capability. Power curves are plotted for comparison.
Testing of a 1/40th scale model in both TFCE and the UD LSWT has
provided marked differences in efficiency characteristics and
torque readings. This is the first instance of a possible influence
of blockage factor on the efficiency of a VAWT model. The same
model is also used to determine the influence of blockage ratio on
the power curves.
5.1. Power and torque coefficient
Fig. 9 displays normalized coefficient of power and velocity
ratio (TSR). As expected with normalizing data the curves coalesce
for the 2% blockage test, giving a peak performance value as 4.5%
efficient at extracting energy from the freestream and curves for
freestreams from 22 to 40 m/s (50–90 mph) lie closely together
with fixed RPM testing. At this low test Reynolds number, such
numbers are typical. However, running tests on the same model
in a smaller wind tunnel with a reduction of tunnel cross-
sectional area by almost half and operating at 5.5% blockage
produces marked differences both in trend and absolute values. It
can be argued that this is beyond the critical blockage size,
exhibiting a shift in efficiency peak as wind speed increases and
displays curves collapsing only at the lower tip-speed ratio
region. There is a clear jump in turbine efficiency when the
blockage ratio is increased. Power coefficient increases if the true
velocity experienced by the model surface is above what is
expected or programmed by the wind tunnel operator.
Fig. 10 displays raw torque loading data as a function of turbine
RPM. As expected, reducing the rotational speeds of the turbines
through increased loading translates to a torque loading capability
Fig. 9. Comparison of two wind tunnel results for 1/40th scale model—details power coefficients increasing as function of blockage ratio increase 2–5.5%.
Fig. 10. Variation of torque and free-spin with turbine RPM—details torque trends at varying wind speeds and resonance regions. (For interpretation of the references to
color in this figure legend, the reader is referred to the web version of this article.)
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 529
detailing the peak positive torque producing region increasing as a
function of increasing freestream speeds. Resonance frequencies
exist between the wind tunnel fan and the turbine models (high-
lighted by the red shaded region in the figure).
Power coefficient varied dramatically between the 2 and
3 bladed designs in this study. The 2 bladed classical ‘two-bucket’
or ‘scoop’ design commonly observed as the popular Savonius
design in the literature produces power coefficients in the
range 0–0.17, operating at tip-speed ratios between 0.2 and 1.1.
The 3 bladed rotor in this study produced power coefficients in
the range 0.02–0.08, operating at tip-speed ratios between
0.1 and 0.6. The variation in power coefficient between 3 bladed
rotor (0.02–0.08) used extensively in this research compared to
the 2 bladed rotor (0.15–0.2) is primarily an influence of the
design of the Savonius rotor itself. It is common in the literature
to find that testing has been completed using the ‘classical’
2 bladed rotors.
A logical reason for the 2 bladed concept achieving higher tip-
speed ratio and higher power coefficients is due to its more
successful design, in this study the blades have been fabricated as
pure semi-circular shapes, producing much higher torque values
and subsequently much higher power coefficients. The 3 bladed
rotors in this study have a blade design with little of the classical
‘scoop’ shape as seen in the literature. The disadvantage of this
design is most likely a major factor in the poor torque production
and subsequent poor power coefficient achieved during testing.
5.2. Blockage area-ratio
Assessing solid and/or wake effect on induced velocity distribu-
tions Fig. 11, provides evidence of a rightward shift in the efficiency
peak when a body surface-area normal to the free stream is placed in
the tunnel is increased. That is, the freestream velocity increases due
to higher levels of flow constriction because of a larger body in the
flow. This is a positive step in comparing the influence of blockage on
artificially increasing efficiency of VAWTs due to increased velocity
and pressure differences as a function of an equivalent body within
the tunnel represented by a pressure signature. Again, with relation to
power coefficient and tip-speed ratio, with an actual flow speed
higher than what is used in calculation produces higher Cp values.
5.3. Wall pressure signature
The shift of VAWT efficiencies in wind tunnel testing indicated
by changes in pressure signature is also investigated. The next
stage in assessing a realistically accurate blockage correction
Fig. 11. Comparison of power coefficient vs. tip-speed ratio—increasing power curves at 80 mph freestream with varying blockage ratios.
Fig. 12. Comparison of wall static pressures for a 10% 2-bladed Savonius model at 30 mph freestream.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538530
factor is to record wall pressures along the center-line of a wind
tunnel closed test-section wall. The goal is to represent the
coefficient of differential pressures and relate these to a velocity
increment, closely resembling the work of Hackett and Wilsden
(1975). Because of the gross asymmetry of the flow created from
this unique type of model, pressure readings from both wind
tunnel sidewalls were incorporated into the study.
Comparing static pressure readings upstream of the model
reveals that the values are lower than those of an empty test
section. This provides evidence for the possibility of wake
propagation far upstream of the model reaching into the wind
tunnel contraction. Static pressure readings reveal a large pres-
sure decrease just aft of the models. This relates to an increased
local freestream velocity, which is a product of both flow
constriction due to solid body interaction and the propagating
wake from a rapidly spinning model influencing the freestream.
5.4. Nondimensionalized pressure coefficient as
a function of location
Fig. 15 shows a sample analysis of one model, the 10% [largest
solid blockage] geometry model. The plots describe flow behavior as
a function of wind tunnel velocity (70 mph freestream velocity),
model RPMs (2000–2530 rpm) and wind tunnel test-section side-
wall location (X/B). Comparing static pressure readings upstream of
the model reveals that the values are lower than those of an empty
test section, which provides possible evidence of upstream wake
propagation far upstream of the model reaching the tunnel contrac-
tion. At slow speeds, 13–22 m/s (30–50 mph) the normalized curves
of pressure coefficient do not coalesce neatly, Figs. 12–14, which
could be a factor of instrument range; however with increasing
freestream velocity there is a functional relationship with a larger
pressure decrease and increased RPM.
Fig. 14. Comparison of wall static pressures for a 10% 2-bladed Savonius model at 50 mph freestream.
Fig. 13. Comparison of wall static pressures for a 10% 2-bladed Savonius model at 40 mph freestream.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 531
Negative pressure coefficient is likely due to higher freestream
velocities with a model present. This aerodynamic characteristic
has been observed with all the models; the close trend between
8% and 10% blockage is shown in Fig. 15. A pressure coefficient of
zero would indicate that the pressures along the tunnel sidewall
are equivalent to those of the empty test section, providing the
conclusion that the model would have no aerodynamic influence
at all on the freestream velocity. These plots reveal large pressure
decreases just aft the models. Following incompressible flow
assumptions, this would relate to increased local freestream
velocity, which is a product of both constriction of the flow due
to solid body interaction and the wake propagating from a rapidly
spinning model.
5.5. Comparison of Cp as a function of blockage area-ratio
Fig. 16 compares pressure differential when a model is spinning
at 1000 rpm in the test section with a 27 m/s (60 mph) freestream.
Analysis of this pressure distribution reveals that the 10% and 8%
solid blockage models have a profound influence on the freestream
pressure, and that the smaller models 3.5% and 2% have a lesser
influence, as logically expected. Using this evidence and normalizing
the values with a dynamic pressure lead to a formulation of a
pressure coefficient signature inside the tunnel for eachmodel at the
27 m/s freestream condition. Fig. 16 details high negative pressure
coefficients displaying increased freestream velocity in the tunnel.
The static tap position just aft of the model center-line was
eliminated with some degree of confidence from the analysis being
a spurious data point due to the transducers’ limited transient
capability demanded by the high RPM turbine. The results recorded
for the 3.5% and 2%model show that the pressure readings approach
sensor sensitivity. Thus absolute values are likely questionable;
however the overall trend should still be identifiable as considerably
less than those for the larger model.
5.6. Pressure coefficient vs. TSR as a function of longitudinal location
Plotting Cp with TSR shows the influence of rate of rotation
upon the flow conditions within the wind tunnel test-section.
This influence has been shown as a function of increased TSR and
compared along the longitudinal static pressure port positions
upstream and downstream of the model center-line.
Fig. 17 shows the relationship for a 10% area blockage model at
a freestream of 22 m/s (50 mph). It can be shown that as TSR
increases over the range 0.3–1.1 the pressure coefficient reduces
in absolute value, therefore providing the conclusion that faster
spinning models have a reduced influence upon the freestream (at
least on the side of the pressure taps). This is the evidence that
the wake propagation from the turbine is better contained at
higher RPMs. Results show lower Cp values obtained at higher
TSR, which shows that an artificially higher freestream velocity is
present at higher TSR.
Results from the remaining wind tunnel models support this
theory, displaying similar trends. A wide range of freestream wind
Fig. 15. Comparison of wall static pressures—shows possible contributions to reduced pressures from solid/bubble and by wake blockage for a 10% 2-bladed Savonius
model at 70 mph freestream.
Fig. 16. Comparison of wall pressure coefficients at 60 mph freestream velocity and 1000 rpm—four models.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538532
speeds has been studied and reveals strong homogeneity in the
flow irrespective of the freestream conditions. Thus the influence
of model rotation is consistent over a wide range of wind speeds,
however; it is also important to observe that the rate of change of
pressure coefficient is much faster down to a TSR of 0.55 than it is
from 0.55 upwards.
5.7. Linear regression model
As a step towards quantifying the effects of blockage ratio
upon the efficiency of VAWT models, model RPM has been
studied for its effects on the freestream pressure distributions.
Using the results from the previous phase of the data analysis,
correlations were created from a linear regression as a function of
longitudinal location. Slopes were created from the previous plots
of Cp vs. TSR. Fig. 18 shows the upper, lower and mid-range data
using the slope equations as a function of TSR for the 8% area
blockage model from 27 to 36 m/s (60–80 mph). The data pro-
vides similarity between the slope equations obtained by the
linear regression.
5.8. Cp–TSR slope as a function of wind speed
Using the individual slope equations obtained at each long-
itudinal location and across wind speeds, the absolute value of the
Cp–TSR slope has been analyzed as a function of wind speed for 8%
area blockage model. From the extrapolation of the functional
relationship between pressure coefficient and tip-speed-ratio
from 8% blockage model a relationship independent of wind
tunnel wind speed conditions indicated by horizontal lines is
clear. Similar results were obtained for the 3.5% and 10% blockage
models (Fig. 19).
5.9. Application of velocity corrections
To assess the effectiveness of the correction methods selected
in this paper, the formulae for velocity corrections for each
method have been applied to the wind tunnel test data. In order
to fully integrate the velocity increments into the data reduction
process, the tip-speed ratio, torque coefficient and power coeffi-
cient have been modified to accept updated wind tunnel free-
stream conditions based on each method. Figs. 20 and 21 display
the overall results covering all correction methods and details
their effectiveness at coalescing the maximum power coefficient
regions across percentage blockage values.
Fig. 20 provides a comparison of the decrease in peak power
coefficient from applying a correction method for the 10 and 8%
rotor model, upper and lower plots, respectively. Fig. 21 displays
similar plots for the 3.5 and 2% rotor models (upper and lower).
Clear trends can be observed with the data in this form, the absolute
wall pressure method reduces the power coefficient with more
severity at lower wind speeds and this trend is also observed with
Fig. 17. 10% blockage model, TSR vs. Cp—shown is a selection of wind tunnel locations that display the overall trend well, comparing at x-location along wind tunnel at
freestream 50 mph.
Fig. 18. Correlated TSR vs. Cp relations compared at x-location along wind tunnel at freestream 60, 70 and 80 mph—compares upper, lower and mid-range data for the
3 bladed 8% model across x-location along wind tunnel, showing one relationship that with increasing TSR, Cp decreases and as wind speed increases, slope angle
decreases.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 533
the delta wall pressure method, although the reductions are much
smaller in the range 0–10% with no reduction needed for the 3.5%
and 2% rotors. The Pope method produced no correlating trend
with wind speed and presumably provided inadequately small
corrections. The Maskell method similarly shows no variation with
wind speed but provided larger reductions in power coefficient.
Fig. 22 displays the effectiveness of two methods, Pope and
Maskell, firstly by showing uncorrected power curves on the left
Fig. 19. Plots of Cp vs. TSR slopes—comparison of Cp vs. TSR slopes across freestream at 10%, 8%, 3.5% and 2%.
Fig. 20. Comparison of percentage decrease in coefficient of power on applying
correction methods—(upper) comparing the methods for correcting the 10% rotor
model across wind speeds 30–70 mph and (lower) a similar comparison for the 8%
rotor model.
Fig. 21. Comparison of percentage decrease in coefficient of power from applying
correction methods—(upper) comparing the methods for correcting the 3.5% rotor
model across wind speeds 60–90 mph and (lower) a similar comparison for the 2%
rotor model.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538534
alongside corrected power curves, middle and right. The power
curves are calculated for the three-bladed rotor results for a wind
tunnel velocity fixed at 60 mph and loading the rotors to achieve
the range of tip-speed ratios plotted. Table 1 displays the Maskell
results covering all the rotor cases, it details the effectiveness and
influence of blockage percentage as the major influencing factor
for Maskell corrections.
The coalescing trend shown in Fig. 22 (right) is produced by
applying Alexander’s adaption of Maskell’s method, in which
they have shown a close analogy between corrections for a flat
plate normal to the freestream can be applied to correcting
Savonius rotors that have an equivalent frontal area as the flat-
plate. For the current study, Fig. 22 (right) displays correction
results using m values extrapolated from Fig. 5. This provides a
very effective end result, reducing the performance of the 8%
blockage model successfully into the region of a much smaller
blockage-ratio model.
5.10. Flow visualization
A laser sheet was produced using a New Wave Solo-PIV
Nd:YAG laser with an energy output 15–200 mJ and a single
cylindrical concave optic. The laser sheet was pulsed at 10 Hz
using a DG535 four channel digital delay/pulse generator coupled
with a 1600 PCO charge-coupled device (CCD) camera fitted with a
25 mm wide angle lens at a distance of approximately 142 cm
vertically from the test-section floor. Smoke was seeded at 34 kN/m2
(5 psi) using an oil-based fluid vi-count smoke generator charged
with a nitrogen supply. Figs. 23–25 provide a selection of images.
Images were captured to compare model RPM at fixed freestream
conditions and observations made about the flow region between
the rotor and the wind tunnel sidewalls. The 2-bladed model
(Fig. 23) exhibited high degrees of streamline bending around the
reverse of the blades.
In most instances this flow is turned fully into the opposing
freestream direction. At high model RPM this flow phenomenon
produces an adverse pressure gradient that could explain a
smaller wake influence when compared to the low RPM condi-
tions. In Fig. 23, when RPMs are decreased the rotor acts
increasingly like a static bluff body in the flow, producing a Von
Karman type bluff body alternating vortex street downstream.
This scales well with the smaller rotor models, Fig. 24.
Interestingly, the flow visualizations reveal similarities to the
results of , there is a common occurrence of strong asymmetry of
the wake; however Fujisawa’s published images have a restricted
FOV, so it is unclear if the results supported a sidewall interaction and
subsequently does not provide an analysis of rotation as an influence
on wake propagation. The results in the figure confirm that the initial
choice of wind tunnel sidewall used for pressure tapping was perhaps
in error (Fig. 25). The images show a much wider wake on the
opposing (right) side, indicating the right sidewall pressures need to
be obtained before reaching any conclusions.
6. Conclusions and recommendations
A good foundation to base further testing and implementation of
modified and improved existing blockage methodologies for static
Fig. 22. Power curves for 3-blade Savonius rotors at 60 mph freestream—(left) uncorrected data, (middle) Pope correction and (right) Maskell method correction.
Table 1
Results of correcting performance of Savonius rotors operating in a restricted flow
closed-test-section wind tunnel using the Maskell method.
Blockage
(S/B (%))
Wind speed
(ft/s (mph))
TSR Initial peak
power
coefficient
Maskell
D power
(%)
Maskell updated
peak power
coefficient
10 103.15 (70) 0.74 0.1316 ÿ59.37 0.0826
10 89.33 (60) 0.56 0.1356 ÿ59.37 0.0851
10 74.58 (50) 0.63 0.1521 ÿ59.37 0.0903
10 59.26 (40) 0.67 0.1588 ÿ59.37 0.0943
10 44.21 (30) 0.67 0.1657 ÿ59.37 0.0983
8 117.30 (80) 0.263 0.0674 ÿ41.88 0.0475
8 103.15 (70) 0.311 0.0726 ÿ41.88 0.0512
8 89.33 (60) 0.251 0.0606 ÿ41.88 0.0427
8 73.30 (50) 0.298 0.0624 ÿ41.88 0.044
3.5 132.00 (90) 0.313 0.0533 ÿ19.50 0.0446
3.5 117.30 (80) 0.295 0.0524 ÿ19.50 0.0439
3.5 103.15 (70) 0.269 0.0507 ÿ19.50 0.0424
3.5 89.33 (60) 0.318 0.0532 ÿ19.50 0.0445
2 132.00 (90) 0.319 0.0448 ÿ10.25 0.0406
2 117.30 (80) 0.273 0.0437 ÿ10.25 0.0396
2 103.15 (70) 0.305 0.0449 ÿ10.25 0.0407
2 89.33 (60) 0.301 0.044 ÿ10.25 0.0399
2 73.30 (50) 0.287 0.0431 ÿ10.25 0.0391
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 535
testing has been provided for application to dynamic wind tunnel
models with a possible further application to dynamic flapping wing
and rotating bluff-bodies being tested in restricted flow domains in
closed test sections. The ever-present research goal remains in
quantifying a blockage correction to apply to rotating bluff-body
models in closed test-section, low-speed wind tunnel testing.
Fig. 23. CCD camera images across laser sheet, 10% model at 20 mph with yellow dotted line denoting boundaries of the wake—Influence of RPM, left: free-spinning model
at 800 rpm, middle: 500 rpm and right: 100 rpm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 24. CCD camera images across laser sheet (a)–(c) compares free spinning rotors at high RPMs to relatively static/very low RPM loaded rotor: (a) 8% rotor at 50 mph:
(left) free spin 880 RPM, (right) 100 RPM; (b) 3.5% rotor at 60 mph: (left) free spin 1400 RPM, (right) 100 RPM; (c) 29% rotor at 80 mph: (left) free spin 2150 RPM,
(right) 100 RPM.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538536
Firstly, the performance characteristics of the sample VAWT
concepts were obtained through a campaign of dynamic and
static loadings of the rotors under varying wind tunnel freestream
conditions. The next phase involved static wall pressure measure-
ments taken along the test-section sidewalls to provide a com-
prehensive pressure signature database of test models under
varying freestream conditions and rotor RPMs. Wake character-
istics produced by the same vertical-axis wind turbine concept
have been investigated at different physical scales in an attempt
to provide some guidance on the scaling of the combined effects
on blockage with supporting flow visualizations.
The results suggest that the precise critical point at which
blockage causes a departure from the expected results has not
been absolutely identified. For operating models of 2% and 3.5%
solid blockage there are no evident issues due strictly to blockage.
Results from this investigation give evidence that at 8% and 10%
the blockage area-ratio would cause some difference in results
due to large pressure drops and increases in freestream velocity
which is not observed at smaller area-ratio testing, and below a
certain wind speed the curves would no longer coalesce. Plots
present a visual verification providing evidence of an adequate
upstream test section but inadequate length downstream for the
asymptotic condition when testing larger models. This has not
been validated with flow visualization due to restricted FOV
downstream.
Correlations of pressure coefficient as a function of tip-speed
ratio have been provided and their susceptibility to wind speed
and longitudinal location along the wind tunnel has been
observed forward and aft the rotor models.
It is the ultimate aim of this study to quantify the shift in
efficiency curves and to define a trend behind shifting efficiencies
based upon a functional dependency of solid-body flow interac-
tion, wind tunnel speed and wake constriction due to wind tunnel
wall interference. It was found that wake constriction for a bluff-
body has a stronger influence from model rotation than from
freestream conditions. In reference Table1, the following conclu-
sions can be made:
� Initial assessment of the Pope and Harper (1966) correction
method led to the conclusion that the derived formula for
velocity increments does not effectively account for wake
blockage influences; however the method reduces peak power
coefficients somewhat effectively.
� Assessment of a wall pressure signature method (WPM)
adapted from theory provided by Hackett and Wilsden
(1975), provides a logical trend in severity of corrections. It
is shown that corrections based on the correlated pressure
coefficient techniques detailed in this paper show that correc-
tion severity decreases with increasing wind speeds and
increasing RPM. This supports the earlier pressure signature
results.
� Corrections have been assessed based on an adapted Maskell
(1966) method for correcting large bluff-body shapes. Special
attention has been focused on the analogy supplied by
Alexander (1978)of comparing the correction of a flat-plate
normal to the freestream to that of a Savonius rotor occupying
an equivalent frontal area. The derivation of a corrected
velocity based on this method produces data revealing strong
coalescing trends, a result that begins to show characteristics
of plotted normalized coefficients.
For future testing of the VAWT concept it would be logical,
following results of this study, that for closed test-section wind
tunnels one should be aware of the deleterious effect caused by
wake interaction and model rotation effects, and in order to
precisely recommend a maximum area-ratio to adopt with closed
test-section experiments, further work is required to assess if
corrections can be achieved successfully and accurately with
existing blockage techniques proposed in this study.
Acknowledgments
The authors gratefully acknowledge the developmental fund-
ing and equipment support from Twenty First Century Energy
(TFCE) and Innovative Scientific Solutions, Incorporated (ISSI), the
continued research efforts complementary to this study carried
out by the University of Dayton Research Institute (UDRI) and
support from the Department of Mechanical and Aerospace
Engineering at the University of Dayton.
References
Ashill, P.R., Weeks, D.J., 1982. A method of determining wall interference correc-tions in solid-wall tunnels from measurements of static pressure at the walls.AGARD-CP-335.
Alexander, A.J., 1978. Wind tunnel corrections for Savonius rotors. In: Proceedingsof the Second International Symposium on Wind Energy Systems, Paper E6,October 3rd–6th, 1978, pp. E6-69–E6-80.
Anderson, J., 2007. Fundamentals of Aerodynamics Fourth Edition McGraw-Hill,New York.
Burton, T., Sharpe, D., Jenkins, N., Bossanyi, E., 2001. Wind Energy Handbook.Wiley Blackwell.
Blackwell, B.F., Sheldahl, R.E., Feltz, L.V., 1997. Wind tunnel performance data fortwo- and three-bucket Savonius rotors. United States Energy Research andDevelopment Administration under Contract AT (29-1)—789, July 1977.
Biswas, A., Gupta, R., Sharma, K.K., 2007. Experimental investigation of overlap andblockage effects on three-bucket Savonius rotors. Journal of Wind Engineering31 (5), 363–368.
ESDU Data Memorandum 80024, 1998. Blockage corrections for bluff bodies inconfined flows. Endorsed by the Royal Aeronautical Society, Issued November(1980) with Amendments A, March 1998.
Fig. 25. 10% rotor 50 mph: (left) static and (right) left wall interaction at 1000 rpm (TSR 0.53), compares sidewall interaction, white strips marking 1 and 2 in. from
sidewall surface.
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538 537
Fujisawa, N., Gotoh, F., 1992. Visualization study of the flow in and around aSavonius rotor. Journal of Experiments in Fluids 12, 407–412.
Hackett, J.E., Lilley, D.E., Wilsden, D.J., 1979. Estimation of tunnel blockage from wallpressure signatures: a review and data correlation. NASA CR-15, 224, March 1979.
Hackett, J.E., Wilsden, D.J., 1975. Determination of low speed wake blockage correctionsvia tunnel wall static pressure measurements. AGARD Fluid Dynamic PanelSymposium on Wind tunnel Design and Testing Techniques, London, England.
Hackett, J.E., 2003. Recent developments in the calculation of low-speed solid-walled wind tunnel wall interference in tests on large models part I:evaluation of three interference assessment methods. Progress in AerospaceSciences 39, 537–583.
Hensel, R.W., 1951. Rectangular-wind-tunnel blocking corrections using thevelocity ratio method. NACA TN 2372, June 1951.
Maskell, E.C., 1965. A theory of the blockage effects on bluff bodies and stalledwings in a closed wind tunnel. ARC R and M 3400.
Pope, A., Harper, J.J., 1966. Low Speed Wind Tunnel Testing. John Wiley and Sons,New York.
Pankhurst, R.C., Holder, D.W., 1952. Wind-Tunnel Technique: an Accountof Experimental Methods in Low-and High-Speed Wind Tunnels. Pitman,London.
Savonius, S.J., 1931. The S-rotor and its applications. Journal of MechanicalEngineering 53 (5).
I. Ross, A. Altman / J. Wind Eng. Ind. Aerodyn. 99 (2011) 523–538538