Effect of Surface Contamination on the Performance of a Section of a Wind Turbine Blade

Full text

Scientia Iranica B (2011) 18 (3), 349–357
Sharif University of Technology
Scientia Iranica
Transactions B: Mechanical Engineering
www.sciencedirect.com
Effect of surface contamination on the performance of a section of a
wind turbine blade
M.R. Soltani a,A.H. Birjandi b,∗,M. Seddighi Moorani c
aDepartment of Aerospace Engineering, Sharif University of Technology, Tehran, P.O. Box 11155-8639, Iran
bDepartment of Mechanical and Manufacturing Engineering, University of Manitoba, Winnipeg, R3T 5V6, MB, Canada
cDepartment of Mechanical Engineering, University of Sheffield, Sheffield, S1 3JD, UK
Received 29 June 2010; revised 12 December 2010; accepted 6 March 2011
KEYWORDS
Wind turbine;
Contamination model;
Performance drop;
Experimental test;
Pressure distribution;
Roughness effects.
Abstract A series of low speed wind tunnel tests were conducted on a section of a 660 kW wind turbine
blade to examine the effects of distributed surface contamination on its performance characteristics. The
selected airfoil was tested with a clean surface, two types of zigzag roughness, strip tape roughness and
distributed contamination roughness. The straight and zigzag leading edge roughness models simplify the
contamination results in an early turbulence transition. In this study, surface contamination was simulated
by applying 0.5 mm height roughness over the entire upper surface of the airfoil. The distribution density
varied from the leading edge to the trailing edge of the model. Our data show that this particular airfoil
was very sensitive to surface contamination and its maximum lift coefficient decreased up to 35%, while
the stall angle of attack increased slightly. The surface contamination, however, caused very smooth stall
characteristics and less lift drop in the post stall region. In contrast to the clean model, the maximum lift
coefficient of the roughened airfoil increased with Reynolds number. The effects of zigzag roughness and
strip tape roughness were less than that of the distributed contamination roughness.
©2011 Sharif University of Technology. Production and hosting by Elsevier B.V. All rights reserved.
1. Introduction
Around 1985, the first observations were made that wind
turbines apparently could have more than one power level for
the same wind speed. This phenomenon typically happens in
high winds. The first publication on this phenomenon is due
to Madsen [1]. At several wind turbine parks in California, he
noticed different power levels, the lowest of which was about
half the designed level. Several initiatives were undertaken to
understand and solve the problem; for example, the study of
Dyrmose and Hansen [2], the Joule project on Multiple Stall [3]
∗Corresponding author.
E-mail address: umbirjaa@cc.umanitoba.ca (A.H. Birjandi).
and the analysis published by Riso [4]. Since then, the cause,
however, has remained uncertain.
Dirt and contamination accumulate on the wind turbine
blade when it operates in the wind field. The main sources
of contamination are insect impact, ageing, sand impact, and
rain contaminants. This contamination has great influence on
rotor performance. When insects, smog and dirt accumulate
along the leading edge of the blade, the power output may
drop up to 40% of its clean value [5]. Although there are
some computational methods for surface roughness simulation
to predict corresponding losses, exact blade contamination
computational simulation is impractical so far. It is still
believed that wind tunnel testing gives more realistic results
than computational schemes, since contamination distribution
modeling is very difficult. Therefore, in general experimental
methods, for contamination effect studies, different types of
leading edge roughness, i.e. strip tape, zigzag tape, etc., are
used on the model surface [6]. In this way, the transition point
moves toward the leading edge and causes early trailing edge
separation; a phenomenon that may not occur when the blade
is operating in the wind field.
In this investigation, the blade section of an under construc-
tion Horizontal Axis Wind Turbine (HAWT) was selected and
various experiments were conducted to examine the effect of
different roughness models on the performance characteristics
1026-3098 ©2011 Sharif University of Technology. Production and hosting by
Elsevier B.V. All rights reserved. Peer review under responsibility of Sharif
University of Technology.
doi:10.1016/j.scient.2011.05.024

350 M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357
Nomenclature
AAxial force
cChord length
DDrag force
hDistance between the model and the wind tunnel
floor
LLift force
NNormal force
V∞Free stream velocity
εVelocity blockage increment
ρDensity
σπ2
48 c
h2
xLongitudinal distance from leading edge
Clcorr Corrected lift coefficient
ClLift coefficient
Clmax Maximum lift coefficient
ClαLift coefficient slope
CpPressure coefficient
Re Reynolds number based on chord length
of the turbine. Experimental study of the real contamination
model in the wind tunnel and its differences with common lead-
ing edge roughness models are the main concerns of this paper.
To the authors’ knowledge, there is no information available
about surface pressure distribution and aerodynamic character-
istics of this blade, neither experimental nor numerical.
2. Contamination sources
There are many sources of contamination for wind turbine
blades, but the main source is insect impact. Insect impact to
the blade is very probable, as wind turbines are usually installed
in agricultural farms or open areas. There is a hypothesis called
‘‘The Insect Hypothesis’’, which explains insect contamination
patterns on wind turbine blades [7]. However, the phenomenon
is attributed to the weather-dependent flying behavior of
insects. The study assumes that the contamination of wind
turbine blades increases only when insects are flying during
turbine operation. Insects mostly fly when there is no rain, little
wind and when the weather is not too cold, at temperatures
above 10 °C. If the turbine operates under these conditions,
insects will increasingly contaminate the blade, mainly near its
stagnation line. Since insects have a mass density much larger
than air, they follow a straight path when they crash on the
frontal area of the airfoil, near to x/c=0%. At low wind
speeds (low angles of attack), the stagnation point is also near
to x/c=0%. Near the stagnation point, the flow is insensitive
to contamination and, hence, the power is not significantly
affected. This is because the flow near the stagnation point is
very stable and, hence, its characteristics will not be affected by
the contamination phenomenon. Above a certain wind speed,
when insects rarely fly, the contamination remains constant. At
high wind speeds, the angle of attack along the blade is large
and the suction peak shifts to the contaminated area. The flow
speed in the suction peak is high, so that the contamination
causes high frictional drag in the boundary layer. Moreover, the
positive pressure gradient beyond the suction peak destabilizes
the flow; this means that the flow will stagnate at smaller
angles of attack and will separate from the blade surface sooner.
For this situation, the flow disturbance depends a great deal
on the level of contamination, which results in a reduction of
the maximum lift coefficient. The smaller the maximum lift
coefficient, the lower the power level. This can also explain the
two or more power levels stated earlier. The design level will
be achieved again when the blades are cleaned or when it rains
during turbine operation.
Icing is another source of contamination for turbine blades.
This condition takes place in cold weather and may persist up
to 25 days in some places [8]. This type of contamination is also
simulated by leading edge roughness. Jasinski et al. [9] applied
aluminum oxide grit over the model leading edge for modeling
first stage icing phenomena. Their results showed that icing
reduces maximum lift coefficient slightly; however, it increases
the airfoil drag coefficient significantly. The experiment was
conducted with low particle density and, hence, did not show
any change in the stall angle of attack. The most recent
investigations on ice accumulation patterns on wind turbine
blades can be found in [10,11].
Aging and sand impact inside the wind are other sources of
contamination. However, these phenomena have not been con-
sidered seriously. Some cracks appear on the wind turbine blade
surface after several years of operation, which deteriorates the
performance characteristics of the blade. Furthermore, in dry
areas, sand storms are typical phenomena, and when sand im-
pacts the wind turbine blade, the surface will roughen.
In this paper, icing conditions have been ignored, but other
sources’ influences on the pressure distribution of the blade
section have been assessed. As pressure distribution around an
airfoil only contributes from drag, drag results are not presented
in the current paper. Indeed, we measured drag coefficient
using rakes, and the data can be found in [12]. The effects of
contamination on pressure distribution and lift diagrams are
the main focus of this paper.
3. Contamination modeling
There are several ways to model the wind turbine blade
surface contamination, but most of these models do not
simulate the actual one. There is little information about real
surface contamination distribution. Nearly all experimental
tests that deal with contamination effects use the leading edge
roughness on the model. Contamination at the leading edge
of the blade will in general lead to a premature transition
of the laminar boundary layer, and result in an early flow
separation that affects the maximum lift capability. To simulate
this in the wind tunnel, transition is usually fixed by putting
a roughness strip or zigzag tape on one or both sides of the
model. The Reynolds number, based on distributed roughness,
called the critical roughness height Reynolds, should be at least
600 according to the Braslow method [13]. The applied grit
roughness causes early transition and will hardly increase the
momentum thickness of the starting turbulent boundary layer.
Therefore, a small increase in drag and a slight reduction in
maximum lift will be achieved [6].
It is extremely difficult to simulate real blade surface
contamination by distributing various roughness devices, such
as roughness strip or zigzag tape on the leading edge of the
model. Pollution is distributed continually over the airfoil from
the leading edge to the trailing edge, but the point is that this
distribution is not homogeneous. For the airfoil with surface
contamination tests, a standard roughness pattern developed
by the National Renewable Energy Laboratory [14] is usually
used. This pattern has been generated using a molded insect
pattern taken from a wind turbine in the field, and it is seen
that the contamination density varies along the chord. Leading

M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357 351
Figure 1: Contamination distribution model on the airfoil.
edge contamination density is about 4 times denser than that
of the trailing edge. For example, on the airfoil model with
250 mm chord at the leading edge, there are about 18 roughness
particles in each square centimeter, while at the trailing
edge, there are only 4.5 roughness particles in each square
centimeter. Actually, the roughness particles are distributed
randomly with no pattern. The roughness distribution pattern
used on the present model surface is shown in Figure 1. Based
on the average particle size from the field specimen, standard
roughness no. 36 lapidary grit was chosen for the roughness
element, giving a particle height to chord ratio of about 0.0019
for a 250 mm chord model. Therefore, for the lowest Reynolds
number tested in this study, Re =0.43×106, the corresponding
Reynolds number of the roughness used is about 800 which is
still larger than the one proposed by the method of Braslow [13].
To apply the roughness pattern on the model, a flat plate
was used, and a roughness pattern was put on its surface;
small holes were drilled for each roughness. After drilling the
roughness pattern on the plate surface, double sided adhesive
tape was put on one of the plate sides, and the grit was fully
distributed on the other side, such that at least one grit particle
was placed in each hole (Figure 1). The banderole was then
taken off the plate surface and put on the airfoil surface.
4. Experimental apparatus
All experiments were conducted in a low speed wind tunnel
in Iran. It is a closed circuit tunnel with a rectangular section test
chamber of 80 ×80 ×200 cm3. The test section speed varies,
continuously, from 10 to 100 m/s, corresponding to a maximum
Reynolds number of up to 5.26 ×106per meter. The inlet of
the tunnel has a 7:1 contraction ratio, with four large, anti-
turbulence screens and a honeycomb in its settling chamber, to
reduce the tunnel turbulence level to less than 0.1% in the test
section. Figure 2 shows the tunnel used for this study.
The model considered in the present study has 25 cm of
chord, 80 cm span and corresponds to a section of a 660 kW
wind turbine blade under construction in Iran. The actual blade
has a length of 23.5 m and the selected airfoil profile was
extracted from 70% of the span. The airfoil profile is not known,
and we were not able to get its profile from the designer.
However, based on our measurements, it is very similar to
that of NACA 6-series airfoils. The shape shown in Figure 3,
along with its coordinates provided in Table 1, comes from
our own measurements of the blade and various drawings. The
measured coordinates of the model were compared with the
true coordinates, and the differences were less than 0.1 mm.
Figure 2: Schematic view of the wind tunnel.
Figure 3: Pressure port locations on the airfoil section. (a) Side view; and (b)
top view.
The upper and lower surfaces of the model are equipped with
64 static pressure orifices, with 0.8 mm inner diameter. The
pressure ports are located along the chord at an angle of 20°,
with respect to the model span, to minimize the disturbance
from the upstream pressure taps (Figure 3). Data are obtained
using 143PC01D Honeywell pressure transducers. Each piece
of transducer data was collected via a terminal board and
transferred to the computer through a 64 channel, 12-bit
Analog-to-Digital (A/D) board; capable of an acquisition rate
up to 500 kHz. The model angle of attack varied between
−5°and 25°, with steps of 1°. At each angle of attack,
at least 400 data points, at a frequency of 100 Hz for
each transducer, were collected and ensemble averaged. An
extensive series of experiments were conducted to ensure that
the pressure-measuring frequency and data acquisition time
were sufficient [15]. Various tests were carried out to ensure
data repeatability under different conditions [16]. Finally, all
data were corrected for solid tunnel walls and wake blockage
effects, using the method explained in [17]. Furthermore, using
the method explained in [18], both the single sample precision
and the bias uncertainty in each measured variable were
estimated and then propagated, that is being taken into account,
into the pressure coefficient, CP, variations. The maximum
overall uncertainty calculated in this way for the CPdata was
less than ±3% of the total CPvalues. The airfoil surface static
pressure distribution was measured at free stream velocities of
30–80 m/s corresponding to the Reynolds no. of 0.43 −1.15 ×
106. Reynolds numbers are defined based on chord length and
free stream velocity.
5. Data reduction
Pressure transducers measure the pressure distribution
on the upper and lower surface of the airfoil. The pressure

352 M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357
Table 1: Airfoil coordinates.
Upper surface Lower surface
X(mm) Y(mm) X(mm) Y(mm)
0 0 0 0
0.2373 2.0590 0.6340 −1.6438
0.7202 3.0431 1.1788 −2.3382
1.2032 4.0272 1.7237 −3.0325
1.7678 4.7800 2.1588 −3.4428
2.3324 5.5328 2.5939 −3.8531
2.9489 6.1704 3.0482 −4.1982
3.5655 6.8080 3.5024 −4.5434
4.3057 7.4523 4.2252 −4.9856
5.0459 8.0965 4.9479 −5.4279
5.8139 8.6782 6.5040 −6.2023
6.5819 9.2598 8.0373 −6.8079
8.0692 10.2447 9.5706 −7.4136
9.5564 11.2295 12.8494 −8.4286
12.8166 13.0869 19.7995 −10.1510
19.6816 16.3225 27.2350 −11.5782
27.1602 19.1714 37.9679 −13.0681
37.8442 22.3823 48.8495 −14.0660
48.7973 24.8576 61.9521 −14.7957
61.8951 26.9892 74.9676 −15.0581
74.9813 28.3457 87.6924 −14.8545
87.6897 28.9632 100.2241 −14.2051
100.2340 28.9001 112.7318 −13.1432
125.2256 26.9699 125.2153 −11.7538
137.7432 25.2686 137.7158 −10.1241
150.2118 23.1738 150.2002 −8.3382
162.7302 20.7246 162.6921 −6.4705
175.2017 18.0065 175.1817 −4.5981
187.7105 15.0692 187.6548 −2.8079
200.1854 11.9850 200.1443 −1.2169
212.6640 8.8136 212.6346 0.0400
225.1556 0.8230 250 0
237.6642 2.7033
250 0
on the upper surface is denoted by Puand similarly Pl
is the corresponding quantity on the lower surface. The
dimensionless pressure coefficient, Cp, is:
Cp=P−P∞
0.5ρ∞V2
∞
,(1)
where ρ∞,V∞and P∞are free stream density, velocity and
pressure, respectively. The normal component of the pressure
contributes to the total normal force, N, while the axial
component of the pressure contributes to the total axial force,
A. Normal and axial forces are defined as:
N=∫(Pl−Pu)cos(θ)dx,(2)
A=∫(Pl−Pu)sin(θ)dx,(3)
where θis the angle between the normal to the chord line and
the direction of P, which is the normal to the surface element.
The sign convention for θis positive, when measured clockwise
from the normal line to the direction of P. Eqs. (2) and (3) hold
for pressure forces on an airfoil per unit span. Lift force, L, is
the perpendicular component of resultant aerodynamic force to
V∞. The parallel component of the resultant aerodynamic force
is called drag force, D. The angle between Land N, and between
Dand A, is called angle of attack, α. The relation between these
two sets of components is:
L=Ncos(α) −Asin(α), (4)
D=Nsin(α) +Acos(α). (5)
Pressure data does not include the skin friction of the airfoil.
Therefore, in lift and drag calculations, we neglect the shear
Figure 4: Effect of Reynolds number on lift coefficient.
stress contribution. However, comparison between numerical
simulations and experimental results shows an acceptable
compatibility, less than 3% difference. This compatibility
indicates that the number of pressure holes is sufficient.
The dimensionless lift coefficient, Cl, is defined as follows:
Cl=L
0.5ρ∞V2
∞c,(6)
where cis the chord length of the airfoil. In wind tunnel tests,
air speed increases in the vicinity of the model due to blockage
effects. The lift coefficient was corrected using the method
explained in [17].
Clcorr =(1−2ε−σ )Cl,(7)
where εis the velocity blockage increment and σis:
σ=π2
48 c
h2
,(8)
where his the airfoil distance from the wind tunnel floor.
6. Results and discussion
6.1. Clean airfoil
This section highlights the experimental data for various
Reynolds numbers. Figure 4 shows the effects of Reynolds
number on the lift coefficient variations for a clean airfoil.
At low angles of attack, Reynolds number does not affect
the lift coefficient significantly. However, at high angles of
attack and especially near static stall, its effect is pronounced
(Figure 4). For post stall conditions, however, Reynolds number
does not have a significant effect on the lift data, and the
variations remain almost the same, except for the lowest
Reynolds number tested, Re =0.43 ×106. The maximum lift
coefficient for the clean airfoil is about 1.26, at the Reynolds
number of 0.43 ×106, and it occurs at an angle of attack
of about 11°. However, it decreases to 1.18 at the Reynolds
number of 1.15 ×106. The stall angle of attack for the
aforementioned Reynolds number is about 10.5°(Figure 4).
This slight reduction in maximum lift coefficient (6.7%), due to
Reynolds number, is probably caused by the transition bubble
formation over the clean airfoil surface and its motion along
the chord, in correspondence with the angle of attack change.
Timmer and Schaffarczyk [19] have mentioned this slight
reduction in the maximum lift coefficient, due to the increase

M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357 353
Figure 5: Pressure distribution on airfoil at low angles of attack.
in Reynolds number, for a DU 97-W-300Mod airfoil. They
predicted transition locations on the upper and lower surfaces
of the airfoil at various Reynolds numbers. Immediately prior
to stall, there is a significant difference in the position of
the upper surface transition bubble, in correspondence with
the Reynolds number change, while on the lower surface, the
Reynolds number does not affect transition bubble location
significantly. Assuming that the magnitude of the maximum lift
coefficient is well predicted, apparently, this rather large shift in
upper surface transition location, with the associated increase
in boundary layer thickness, counterbalances the favorable
effect of higher Reynolds number on boundary layer thickness
and stability.
For the present model, when tested at low angles of attack
and at low Reynolds number, less than 106, a laminar transition
bubble is formed over both its upper and lower surfaces, as
seen in Figure 5. When laminar flow changes to a turbulent
one, surface static pressure increases (Figure 5). As the angle
of attack increases, the upper surface laminar transition bubble
moves toward the leading edge, while the one on the lower
surface moves toward the trailing edge. Increasing the angle
of attack decreases the strength of these transition bubbles in
the upper and lower surfaces; the lower surface having more
reduction. The transition bubble strength has been defined by
the pressure difference between the two sides of the bubble.
For a Reynolds number of 0.85 ×106and zero angle of attack,
the laminar transition bubble on the upper surface is located
between 0.6<x/c<0.65 from the leading edge, while for
the lower surface, it is located between 0.5<x/c<0.6
from the leading edge (Figure 5). When the angle of attack is
increased to 5°, the upper surface laminar transition bubble
moves toward the leading edge and is located at a distance
between 0.5<x/c<0.55 from the leading edge, while the
lower surface transition bubble has moved rearward, located
between 0.6<x/c<0.7 from the leading edge. Figure 6
shows the pressure distribution over the clean airfoil at 10.5°
and 14°angles of attack. From this figure, it is clearly seen that
for the 10.5°angle of attack, the flow over the entire upper
surface of the airfoil is turbulent; however, no sign of separation
is still evident from this figure. At a higher angle of attack,
α=14°,Figure 6 shows a separated flow over almost 50% of
the upper surface of the model. The pressure patterns over the
lower surface of this model for the aforementioned two angles
of attack, 10.5°and 14°, are almost identical, and no sign of
Figure 6: Pressure distribution on airfoil at high angles of attack.
Figure 7: Reynolds effect on laminar transition bubble location.
transition is observed from the present surface pressure data
(Figure 6).
When the Reynolds number increases, the laminar transition
bubble for the upper surface shifts toward the leading edge,
while that of the lower surface moves toward the trailing edge
of the model (Figure 7). The pressure difference ahead and
aft of the transition bubble is higher on the upper surface, in
comparison with that of the lower surface. When the angle of
attack is near the stall angle, that is 10°for the highest Reynolds
number tested, the laminar transition bubble is formed closer to
the model’s leading edge. This early laminar transition bubble
on the upper surface increases static pressure over the airfoil
and as a result the maximum lift coefficient decreases slightly
(Figure 4).
6.2. Zigzag roughness, 60°and 90°
Before using the contamination distribution model, two
types of zigzag roughness were used. They both had the
same geometric characteristics, except their zigzag angle was
different, 60°and 90°, respectively. The zigzag tape width was
3 mm, and their entire width was 12 mm, as shown in Figure 8.
These tapes were located on the suction side of the airfoil at
x/c=5% from the leading edge. The height of these zigzag

354 M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357
Figure 8: Shape of 60°zigzag tape.
Figure 9: Pressure distribution over clean and the roughened airfoils.
tapes was about 0.07 mm. After these tapes were glued on
the airfoil their surfaces were covered by 0.45 mm average
diameter grits to simulate the leading edge roughness.
The zigzag roughness at the leading edge of the airfoil
forces the flow transition from laminar to turbulent and adds
thickness to the boundary layer. Flow velocity decreases as
soon as it passes over the roughness, thus static pressure
increases slightly. Figure 9 shows pressure distribution over
the clean airfoil and the one with the 90°zigzag roughness
installed on its upper surface. The pressure distribution over
the lower surface is the same for both conditions, but the
upper surface pressure has the same value just before the
zigzag roughness (Figure 9). Immediately after the roughness,
the pressure increases, because the flow regime has changed
to turbulent. The effect of zigzag roughness on the pressure
distribution over the airfoil is clearly seen from Figure 9. A
laminar transition bubble exists on the upper surface of the
clean airfoil, at about x/c∼
=0.55 from the leading edge, but this
bubble does not appear on the airfoil with roughness installed
at its leading edge. Furthermore, it is seen that the pressure
distributions for both cases, clean and with zigzag roughness
airfoils, are almost identical for x/c>0.6, where flow over the
clean airfoil grows turbulent.
Figures 10 and 11 show variations of lift coefficient, with the
angle of attack for two different roughness levels, 90°and 60°
zigzag tapes, and at various Reynolds numbers. The data are
also compared with the clean model data. From these figures,
it is clearly seen that in contrast to the clean case data shown
in Figure 4, for the roughened model, the Reynolds number
also has some effects at low angles of attack. As the Reynolds
number increases, the slope of the lift coefficient, Clα, for both
types of zigzag roughness, increases slightly (Figures 10 and 11).
6.2.1. Pressure distribution near the stall condition
Adding roughness to the leading edge of the airfoil changes
aerodynamic characteristics of the airfoil, especially near the
Figure 10: Effect of Reynolds number on lift coefficient for 60°roughness.
Figure 11: Effect of Reynolds number on lift coefficient for 90°roughness.
stall condition. For the clean condition, the stall is relatively
sharp and the stall angle of attack does not vary significantly
with Reynolds number (Figure 4). However, for the roughened
airfoil, the stall angle for all Reynolds numbers varies smoothly,
as can be seen in Figures 10 and 11. Roughness forces the
laminar flow to change to turbulent flow, which has more
energy. Furthermore, roughness increases the momentum
thickness of the starting turbulent boundary layer. So for
small angles of attack, roughness reduces lift, but at high
angles of attack, this early transition helps the flow to remain
attached. The absolute value of static pressure in the turbulent
boundary layer is higher than static pressure in the laminar
boundary layer. Thus, the lift coefficient of the model with
zigzag roughness decreases for almost all angles of attack; stall
however occurs at higher angles of attack. In contrast to the
clean airfoil, the maximum lift coefficient for the rough airfoil
increases as the Reynolds number is increased (Figures 10 and
11). When the Reynolds number is increased, the effect of zigzag
roughness on the maximum lift coefficient decreases. At the
Reynolds number of 0.43×106,90°zigzag roughness decreases
the maximum lift coefficient from 1.26 to 0.9 for the clean
condition, corresponding to a reduction of about 28.6%. But for
the Reynolds number of 1.15×106, the maximum lift coefficient
of the roughened model is 1.05, while for the clean case, it

M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357 355
Figure 12: Effect of Reynolds number on pressure distribution.
Figure 13: Comparison between 60°and 90°roughness condition.
is 1.18, corresponding to a reduction of about 11%. Thus it is
seen that zigzag roughness has more effect on maximum lift
coefficient when the Reynolds number is low. Furthermore, for
cases with the highest Reynolds number, the turbulent flow
separates at higher angles of attack. This phenomenon is shown
for two different Reynolds numbers, while the airfoil has been
equipped with a 90°zigzag roughness in Figure 12.
6.2.2. Lift coefficient after stall
The effects of 90°and 60°zigzag roughness on the lift
coefficient are very similar to each other, but they are not
exactly the same. They have the same effect prior to stall, but
the stall characteristics of two roughness levels are not the
same (Figure 13). The stall phenomenon for 90°roughness is
smoother and takes place at a higher angle of attack, when
compared with that of 60°roughness. In addition, in the post
stall region, the lift coefficient of 60°roughness decreases faster
than that of 90°(Figure 13). As a result, for the 90°zigzag
roughness model, the lift coefficient values in the post stall
region are higher. In comparison to the clean model, the lift
coefficient of 90°roughness is slightly higher for the post stall
angles of attack, that is α=14°–19°(Figure 13). In addition,
Figure 14: Effect of Reynolds number on lift coefficient variation.
Figure 15: Comparison between 90°and strip tape roughness conditions.
the stall characteristics of the model with different roughness
levels differ significantly from those of the clean one.
6.3. Strip tape roughness
Strip tape roughness was located in the same position as
the 90°and 60°zigzag ones. The strip tape was 12 mm in
width, 0.1 mm in height, and was covered with 0.5 mm of grit.
Figure 14 shows the variation of the lift coefficient with angle of
attack for four different Reynolds numbers, for the case where
strip tape roughness was glued onto the model. The results
are further compared with the clean airfoil for the Reynolds
number of 1.15 ×106. As the Reynolds number increases, the
maximum lift coefficient and the corresponding stall angle of
attack increases slightly (Figure 14). However, their value, Clmax,
for the roughened model is considerably lower than that of the
clean one, while the stall angle of attack has been increased. The
lift curve slopes, Clα, for the strip tape roughness, at all Reynolds
numbers, are also less than those of the clean one (Figure 14).
In Figure 15, the lift coefficient data for the strip tape and
90°zigzag roughness are compared with each other. It is seen
by inspection that the performance degradation caused by the
strip tape is much more than that of the zigzag one. In addition,
the stall characteristic of the zigzag roughness is smoother than

356 M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357
Figure 16: Lift coefficient variation for airfoil with surface contamination.
the strip roughness. The amount of pressure rise after the strip
tape is more than that of the zigzag one; hence, it leads to an
earlier and sharper stall phenomenon when compared to 90°
zigzag roughness (Figure 15). Furthermore, from Figure 15, it
is clearly evident that flow separation at the trailing edge, for
the model with strip tape roughness, occurs at smaller angles
of attack than for the one with zigzag roughness.
6.4. Surface contamination distribution model
In a real contamination case, pollution density is more at
the leading edge and less at the trailing edge. However, for
the present model, a random distribution of 0.5 mm diameter
roughness was used over the entire upper surface of the airfoil.
All tests were repeated under the same conditions adopted in
previous cases. However, in this paper only a few figures are
presented.
By adding the surface contamination model of the roughness
on the airfoil, the maximum lift coefficient decreased by about
35% (Figure 16). Unlike the other roughness patterns, the
maximum lift coefficient does not change significantly as a
function of the Reynolds number for this specific airfoil. As
seen from Figure 16, this type of roughness distribution reduces
Clmax, as well as Clα, significantly, when compared to the clean
data. Furthermore, the stall pattern is very smooth and the Cl
drop is not drastic. For the post stall zone, the airfoil with surface
contamination shows an increase in Cl, while that of the clean
model continues to decrease, as the angle of attack increases
beyond the static stall. This is because of pressure reduction in
the wake area and pressure growth over the lower surface of
the airfoil at high angles of attack.
For the airfoil with surface contamination, the flow begins
to separate at very low angles of attack because the surface
roughness consumes the free stream energy and increases
its pressure over the suction side of the model (Figure 17).
Pressure growth over the airfoil surface brings the separation
point toward the leading edge. Lift coefficient does not change
significantly with Reynolds number. The stall angle of attack is
almost independent of Reynolds number and occurs at about
α=13°.Figure 18 shows pressure distribution over the
airfoil with surface contamination in the post stall region.
When the angle of attack is increased from 17°to 25°, the
pressure over the suction side of the airfoil increases and the
minimum pressure peak is diminished. At 25°angle of attack,
Figure 17: Pressure distribution over airfoil with surface contamination.
Figure 18: Pressure distribution over airfoil with surface contamination in the
post stall region.
flow separation takes place earlier than for 17°angle of attack.
Therefore, pressure in the separated region of the airfoil at 25°
angle of attack is lower than for 17°angle of attack, while
in the post stall region, wake pressure reduction is dominant,
and the lift coefficient increases again after a slight reduction
(Figures 16–18).
Figure 19 shows the effect of various roughness models
on the stall angle of attack; they do not follow a regular
pattern. Stall angle of attack decreases by increasing the
Reynolds number for the clean airfoil, the airfoil with surface
contamination and the strip roughness models. However, this
pattern changes for the 60°and 90°zigzag roughness models.
The highest stall angle of attack for 90°zigzag roughness occurs
at a Reynolds number of Re =1.15 ×106, while the lowest
stall angle of attack occurs when the model is clean and tested
at the same Reynolds number, Re =1.15 ×106. Maximum
lift coefficient vs. Reynolds number for different types of
roughness is plotted in Figure 20. Reynolds number does
not have a significant effect on the maximum lift coefficient.
The maximum lift coefficient of the clean airfoil and the
airfoil with surface contamination decreases by increasing the
Reynolds number, but for other types of surface roughness, as
the Reynolds number increases, the maximum lift coefficient
decreases slightly (Figure 20).

M.R. Soltani et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 349–357 357
Figure 19: Roughness effect on stall angle of attack.
Figure 20: Roughness effect on maximum lift coefficient.
7. Conclusions
Various tests were conducted on a wind turbine blade
section to investigate contamination effects on its surface
pressure distribution, and corresponding lift variations. There
are many conventional ways, such as zigzag roughness or
strip insertion, for simulating the effects of contamination
accumulation on the wind turbine blade. In these models
usually roughness is installed near the leading edge, where it
forces the laminar flow to become turbulent. These models
cannot represent all contamination effects on performance
characteristics properly. In reality, contamination on the
wind turbine blade covers the whole blade with a higher
contamination density near the leading edge of the blade. The
exact contamination was modeled based on the field data, for
contamination distribution on a wind turbine blade.
Our results showed that the present airfoil is very sensitive
to roughness. When roughness is installed on the model, its
performance degrades considerably. Furthermore, roughness
postpones the stall phenomenon, and the Clvariation with
angle of attack in this region, in contrast to the clean one, is
very smooth. These results are seen for all roughness models,
but the amounts of roughness effect are not the same. 90°
zigzag roughness has the least effect on the maximum lift
coefficient of the model. On the other hand, the effect of
the so-called exact contamination model is significant, and
reduces maximum lift coefficient up to 35%. Results show that
conventional contamination modeling does not simulate the
exact phenomenon, and that contamination distribution should
be modeled based on field measurements.
References
[1] Madsen, H.A. ‘‘Aerodynamics of a horizontal axis wind turbine in natural
conditions’’, Riso M 2903, Riso National Laboratory, Roskilde, Denmark,
(1991).
[2] Dyrmose, S.Z. and Hansen, P. ‘‘The double stall phenomenon and how to
avoid it’’, IEA, Joint Action Aerodynamics of Wind Turbines, Lyngby, Denmark
(1998).
[3] Snel, H. and Corten, G.P., et al. ‘‘Progress in the Joule project: multiple stall
levels’’, European Wind Energy Conference, EWEC, Nice, France, pp. 141–145
(1999).
[4] Bak, C. and Madsen, H.A., et al. Double stall, Riso-R-1043, EN (1998).
[5] Hansen, C. and Butterfield, C.P. ‘‘Aerodynamics of horizontal-axis wind
turbine’’, Annual Review of Fluid Mechanics, 25, pp. 115–149 (1993).
[6] Van Rooij, R.P.J.O.M. and Timmer, W.A. ‘‘Roughness sensitivity considera-
tions for thick rotor blade airfoils’’, Journal of Solar Energy Engineering, 125,
pp. 468–478 (2003).
[7] Corten, G.P. and Veldkamp, H.F. ‘‘Insects cause double stall’’, European
Wind Energy Conference, EWEC, Copenhagen, Denmark, pp. 470–474
(2001).
[8] Tammelin, B. and Santti, K. ‘‘Rime accretions on the fells, BOREAS’’,
An International Expert’s Meeting on Wind Power under Icy Conditions,
Enontekio, Finland (1992).
[9] Jasinski, W.J. and Noe, S.C., et al. ‘‘Wind turbine performancee under icy
conditions’’, 35th AIAA, Aerospace Sciences Meeting & Exhibit, Reno, Nevada,
USA (1997).
[10] Karaj, A.G. and Bibeau, E.L. ‘‘Phases of icing on wind turbine blades
characterized by ice accumulation’’, Renewable Energy, 35, pp. 966–972
(2010).
[11] Wang, K. ‘‘Convective heat transfer and experimental icing aerodynamics
of wind turbine blades’’, Ph.D. Thesis, University of Manitoba, Canada
(2008).
[12] Soltani, M.R. and Mahmoudi, M. ‘‘Measurements of velocity field in the
wake of an oscillating wind turbine blade’’, Aeronautical Journal, 114,
pp. 493–504 (2010).
[13] Braslow, A.L. and Knox, E.C. ‘‘Simplified method for determination of
critical height of distributed roughness particles for boundary-layer
transition at mach numbers 0–5’’, NACA TN 4363, USA (1958).
[14] Janiszewska, J.M. and Ramsey, R., et al. ‘‘Effects of grit roughness and
pitch oscillations on the S814 airfoil’’, NREL Report TP-442-8261, Golden,
Colorado, USA (1996).
[15] Soltani, M.R. and Rasi, F., et al. ‘‘An experimental investigation of time
lag in pressure-measuring systems’’, 2nd Ankara International Aerospace
Conference, Ankara, Turkey (2005).
[16] Askari, F. ‘‘Experimental study of a wind turbine blade section, Master of
Science Thesis’’, Aerospace Department, Sharif University of Technology,
Tehran, Iran (2004).
[17] Barlow, J.B. and Rae, W.H., et al., Low-Speed Wind Tunnel Testing, 3rd ed.,
John Wiley & Sons Publ. (1999).
[18] Thomas, G.B. and Roy, D.M., et al., Mechanical Measurements, 5th ed.,
Addison-Wesley Publishing Company (1993).
[19] Timmer, W.A. and Schaffarczyk, A.P. ‘‘The effect of roughness at high
Reynolds numbers on the performancee of aerofoil DU 97-W-300Mod’’,
Wind Energy, 7, pp. 295–307 (2004).
Mohammad Reza Soltani has a Ph.D. in Aerodynamics from the University
of Illinois at Urbana-Champaign, USA. His research interests include Applied
Aerodynamics, Unsteady Aerodynamics Wind Tunnel Testing, Wind Tunnel
Design, and Data Processing. He is now Professor and Head of the Aerospace
Engineering Department at Sharif University of Technology, Tehran, Iran.
Amir Hossein Birjandi is a Ph.D. candidate at the University of Manitoba in
Canada. He obtained M.S. and B.S. degrees in Aerospace Engineering from Sharif
University of Technology in Tehran. He is interested in Renewable Energy
Resources, and worked on Wind Turbines as his M.S. project. He is now
working on Tidal Vertical Turbines. His research interests include Experimental
Aerodynamics and Unsteady Aerodynamics.
Mehdi Seddighi Moorani is a Ph.D. student at Aberdeen University in the
UK. He graduated with a B.S. degree in Aerospace Engineering from Malek-
e-Ashtar University, Iran, and received his M.S. degree in Aerodynamics from
Sharif University of Technology, in Tehran, where he served one year as Senior
Research Assistant after graduation. His research interests include Unsteady
Turbulent Flow, Direct Numerical Simulation, and Aerodynamics of Wind
Turbines.