Analysis of Airfoils Modifications on Horizontal Axis Wind Turbine Blades 
 
Carlos M. Guzmán Claudio 
Master of Engineering in Mechanical Engineering 
Dr. Moisés Angeles  
Department of Mechanical Engineering 
Polytechnic University of Puerto Rico 
 
One modification consists in to use dimples on 
Abstract ⎯ This project investigates the 
the airfoil surface. In a golf ball, dimples act as 
combination of upper surface dimples and trailing 
vortex generators. These vortices force the air 
edge serrations of a 3D thin airfoil and their effect 
around the ball to transition from laminar to 
on aerodynamic performance. Dimples are known to 
turbulent at an earlier phase relative to a smooth ball. 
reduce a golf ball’s drag while a serrated trailing 
The turbulence mixes higher velocity air into the 
edge is known to reduce noise. A flatback design is 
surface of the ball, delaying the boundary layer 
also studied to see if its properties apply to a thin 
separation, as visualized in Figure 1. This delay then 
airfoil. The airfoil in the present study is NREL S825, 
reduces the wake drag of the ball, thus allowing the 
to be 3D-printed with six combinations of the 
ball to go farther and higher than a smooth 
proposed modifications and tested in a wind tunnel. 
counterpart [1]. 
This study has found that the combination of 
dimples, flatback, and serrations has increased the 
maximum lift coefficient by 9.85%; however the 
maximum lift-to-drag ratio has been reduced by 
14.4%. 
Key Terms ⎯ Dimples, Flatback, NREL S825, 
Serrated Trailing Edge.  
INTRODUCTION 
The objective of this work is to study the 
behavior of a wind turbine blade when it is modified 
with upper surface dimples, flatback trailing edge, 
and trailing edge serrations. These modifications  
should help improve the performance of a blade by Figure 1 
reducing the drag wake. The dimples should act as Flow Separation on A Sphere and a Golf Ball 
vortex generators, delaying boundary layer 
There is an effect on airfoil performance when 
separation; the flatback shape may reduce the 
introducing a variety of dimples to the upper surface 
pressure gradient along the surface of the blade; and 
of the wing, such as circular and squared shaped 
the serrations could break up vortices behind the 
dimples [2]. Notably, having either inwards or 
blade by forcing the wake to dissipate energy. 
outwards dimple on the upper surface delays flow 
The combination of these modifications would separation from the aircraft wing, resulting in 
give a better understanding of how to approach increased coefficient of lift and decreased coefficient 
designing future wind turbine blades for higher of drag [3]. Reference [4] observed that a more 
efficiency.  continuous drop in pressure is produced by the 
dimpled surface, allowing the airfoil to stall at higher 
angles than the smooth counterpart. A symmetrical 
airfoil at an angle of attack of 20° with its upper One way to generate a flatback airfoil is to cut 
surface mostly covered in dimples decreased the off, truncating, a segment of the trailing edge of the 
drag coefficient by 20.5% (Figure 2) and increased airfoil. Another way to create a flatback airfoil is to 
its lift coefficient by 34.19% increase the thickness of the trailing edge while 
( maintaining the mean camber line of the airfoil 
(Figure 6). Thickness must be added along the 
camber line to prevent adverse boundary layer 
effects.  In reference [6], the author studied the 
effects of modifying a 35% thickness-to-chord (t/c) 
airfoil in the various forms. A comparison of these 
airfoils can be seen in Figure 4: 
Figure 3) [5].  • The base TR-35 sharp-trailing edge airfoil 
• TR-35.80 truncated at x/c = 80%, resulting t/c = 
44% and tTE/c = 10% 
• TR-44 a sharp-trailing edge TR-35 airfoil 
thickened to t/c = 44% 
• TR-35-10 is a blunt trailing edge airfoil with a 
 t/c = 35% and tTE/c = 10% 
Figure 2 
(a) Drag Coefficient versus Angle of Attack and (b) Drag 
Coefficient Decrement of Model 2 with Respect to Model 1 
versus Angle of Attack [5] 
 
Figure 4 
TR Series Airfoils [6] 
Figure 3 
This truncation method proved to hinder the 
(a) Lift Coefficient versus Angle of Attack and (b) Lift 
Coefficient Increment of Model 2 with Respect to Model 1 performance of the airfoil due to the loss of 
versus Angle of Attack [5] maximum camber, as seen in Figure 5. The TR-44 
airfoil is too thick for practical use. The truncated 
The next modification involves a flatback 
TR-35.80 resulted with a higher maximum lift, but 
trailing edge. The idea behind blunting the trailing 
with a significantly reduced lift slope. The TR-35-10 
edge is to improve structural integrity of the root 
possesses superior lift performance along the entire 
region of the blade by increasing the cross-sectional 
angle-of-attack range.  
area, resulting in greater buckling resistance. 
Blunting the trailing introduces an increased base 
drag at higher Reynolds Number, so it is best used 
for the root region of the blade. It also comes with 
the benefit of increased lift performance and greater 
resistance to leading edge soiling [6]. 
  
Figure 7 
Figure 5 
Base Pressure of a Slotted and an M-shaped Serrated 
Effect of Trailing Edge Modification on Lift, Re = 4.5e6 [6] 
Trailing Edge [9] 
 
Figure 6 
Blunt Trailing-edge Airfoil with T/c=35%, Tte/c=10% [6] 
The subsequent modification proposed is the 
serrating the trailing edge which comes from a need  
to reduce the aerodynamic noise profile of the blade, Figure 8 
mainly from the tip, as its periodic nature can 
Base Pressure of A 60o and 120o Sawtooth Serrated Trailing 
become an annoyance to nearby residences [7]. A Edge [9] 
variety of serrations can be generated, such as 
In reference [8] two airfoils, namely a NACA 
rectangular, sawtooth, and M-shaped. The most 
0012 and a NACA 65(12)-10, were studied in a wind 
common of these is the sawtooth shape.  They serve 
tunnel with a variety of trailing edge forms. These 
to induce a span-wise pressure gradient, forcing 
forms include a sawtooth, sinusoidal, and slotted 
vortices to collide and breakup. Figure 7 and Figure 
sawtooth serrations of various frequencies. The 
8 show a few example comparisons of different 
study shows that, for the NACA 65(12)-10, 
kinds of possible serrations and their effects on base 
serrations don’t significantly change the drag 
drag reduction. Figure 7 shows how an M-Shaped 
coefficient and the lift coefficient reduces up to 15% 
serration has lower base pressure loses over a 
between angles of -5° to 10°, however at higher 
rectangular serration. Figure 8 shows how the 
angles the drag increases with larger wavelength 
differences in angle on a sawtooth serration impact 
serrations. Figure 9 shows these results for various 
the overall base drag on a flatback blade.  
sawtooth serrations. The sinusoidal serration has 
similar performance to the sawtooth serration. The 
slotted-sawtooth resulted in up to 30% reduction in 
lift coefficient over the entire range on angles 
studied.  
 
Figure 9 
 
Lift and Drag Coefficients for A NACA 65(12)-10 Airfoil 
Figure 11 
Fitted with Different Sawtooth Serrations [8] 
Upper Surface Dimple Dimensions, Side View 
A follow up study [9] investigated the wake 
development of the previously mentioned airfoil 
serrations. It found that for the NACA 65(12)-10, the 
overall wake development for both the sawtooth and 
slotted sawtooth serrations is less turbulent due to 
the interactions of the tip-flow and the root-flow.  
 
Figure 12 
Upper Surface Dimple Dimensions, Top View 
For the purposes of this project, the dimple 
characteristics will be fixed to:  
𝜀 = 30% of the chord 
h = .1114in 
 r = .1856in 
Figure 10 d = 1.5in 
Wake Turbulent Kinetic Energy for NACA 65(12)-10 Airfoil a = .75in 
at a=10°. [8] 
 
METHODOLOGY 
Flatback Conversion 
Dimple Geometry For this project, the flatback airfoils will be 
To determine the position and dimension of the created by thickening the rear segment of the airfoil, 
upper surface dimples, the following characteristics past the maximum thickness point. This method 
are required. The term 𝜀 refers to the x-coordinate of provides a reduced pressure gradient along the airfoil 
the dimple’s center relative to the chord line of the compared to the original and maintains the camber 
airfoil. To generate the dimple the terms r and h are line shape and thickness ratio, which is not the case 
used to define the radius and the depth of on a truncated airfoil. 
indentation, respectively. As the dimples will be To transform the airfoil, the flatback profile 
applied along the entire span of the blade the term d equation will be used. This equation takes the 
will denote the center-to-center distance between original set of coordinates that define the selected 
each dimple (seen in Figure 11 and Figure 12). profile and adds thickness pass a desired point in a 
symmetrical manner [10]: 
With these conditions, it is now possible to calculate 
𝑇𝐸/100
?̅?𝑓𝑏 = ?̅?𝑜𝑟
(1) 
± 𝑎(?̅?) 
2 the constants for a(?̅?): 
where: 1𝐴 =  
(1 − 𝐵)𝑛 − 𝑛(𝜀 − 𝐵)𝑛−1
𝑓𝑏 (6) ?̅? : y-coordinate of flatback profile (non-
 
dimensional) 
?̅?𝑜𝑟: y-coordinate of original profile (non- 𝐵 = 𝜀(1 − 𝑛) (7) 
dimensional) 
𝐶 = 1 − 𝐴(1 − 𝐵)𝑛 (8) 
TE: Desired trailing edge thickness with respect to 
the chord (%) 
For the present research, a script developed in 
?̅?: x-coordinate of profile (non-dimensional) MATLAB was used to generate the conversion. The 
input variables for the Flatback conversion are as 
𝑎(?̅?): Distribution factor (non-dimensional) 
follows: 
The distribution factor 𝑎(?̅?) is a function that 
TE =4.2% t/c 
distributes the additional thickness along the chord 
length. The distribution factor is defined as: ε = 29.9% x/c 
n = 0.5 
𝑎(?̅?) = 𝐴(?̅? − 𝐵)𝑛 + 𝐶?̅? (2) 
Serration Geometry 
The constants A, B, and C are computed by 
The serration to be used in this project will be 
applying boundary conditions on the profile, which 
an M-Shaped trailing edge [12]. As the airfoil will 
are the point where the thickness addition begins and 
possess a flatback shape, the trailing edge will have 
the trailing edge. The parameter n is used to adjust 
a thickness, h, which is the basis of the equations 
the transition smoothness. It was tested to be 
used to define the serration dimensions. The term b1 
smoothest when n = 0.5 [11]. The term ε is used to 
is the width of the separation between each M-shape 
denote the starting point along the x-axis from the 
and b2 is the width of the M-shape itself. The term a 
leading edge where the thickness is added to the 
is the depth of the serration while γ is angle of the 
airfoil.  
serration. Figure 13 illustrates the parameters for this 
To set up the boundary conditions, a few geometry: 
conditions need to be met. It is desired that the 
distribution factor begins just after the point ε and 
finishes at the trailing edge. It is also desired that at 
point ε the distribution factor is parallel to the x-axis 
to ensure the thickness is added smoothly after the 
starting point. The boundary conditions are as 
follows: 
 
 ?̅? = 𝜀 → 𝑎 = 0 (3) 
Figure 13 
 ?̅? = 0 → 𝑎 = 1 (4) Dimensions of M-shaped Serrated Trailing Edge [9] 
The design is characterized by the optimizations 
 𝑑𝑎
= 𝐴𝑛(?̅? − 𝐵)𝑛−1 + 𝐶 = 0 (5) that reduce the drag generated by the flatback shape 
𝑑?̅?
by 46%. [12] The optimized parameters are related A linear behavior can be seen from -5° up to 8° with 
as follows: the maximum lift coefficient at 12°. The angle for 
zero lift occurs at -4.62°. A sudden lift coefficient 
𝑎
= 1.9 (9)  drop is seen between 17° and 18°. For drag, a linear 
ℎ
behavior is also seen from -2° up to 17°, when the 
𝑏 = 𝑏 = 3.66ℎ (10) drag coefficient increase occurs. Afterwards, the 1 2
linear behavior continues with a slightly steeper 
𝛾 = 40° (11) slope. The effect of the sudden coefficient spike on 
drag is more pronounced compared to lift. 
 
0.5
The serrations will be part of a splitter plate 
0.4
integrated into a flatback blade and will be aligned 
with camber of the blade. Effectively, it will be a 0.3
3mm plate protruding at an angle of 16.44° 0.2
downwards. 
0.1
Wind Tunnel Testing 0
The wind tunnel used is a Flotek 1440, an open Angle of Attack (deg)
circuit system with a 12” x 12” x 36” test section Control Lift Control Drag
 
with an airspeed of approximately 93 fps. The 
models are placed vertically inside the test section.  Figure 14 
A force balance is used to measure the lift and drag Lift and Drag Curve for the Control Blade, Re = 2.95E5 
on the blade over a time interval of 15 seconds, When introducing the dimples to the control 
taking the last 5 seconds for evaluation. Wind tunnel model, no significant differences can be observed. 
speed, lift and drag measurements of the 5 second The linear region is near identical for both lift and 
interval are then averaged and tabulated. Each blade drag. The angle where lift coefficient drop occurs is 
is tested three times from an angle of attack ranging increased to 19° and the maximum lift coefficient is 
from -15° to 30°. As we are assuming a standard reduced by 5.88%.  
atmospheric condition, the Reynolds Number will be 
around 296,000. All models were 3D printed with 0.5
PLA, then progressively sanded down with 80, 150, 0.4
320, and 500 grit sandpaper. An aluminum insert 
0.3
was manufactured to fit inside the blades and 
threaded to hold the blade onto the force balance 0.2
beam. 0.1
0
RESULTS 
Angle of Attack (deg)
Wind tunnel experimentation results are Dimpled Lift Control Lift Dimpled Drag Control Drag
 
presented. Observing the control model was the first 
step; then compared the variations against the Figure 15 
control. The dimpled variations with their smooth Lift and Drag Comparison Between the Dimple Control 
counterpart were compared too.  Blade and the Control Blade, Re = 2.95E5 
The S825 control blade’s aerodynamic The Flatback modification to the S825 airfoil is 
characteristics are shown in Figure 14 and Figure 15. effectively identical to the control model for both lift 
Lift Coefficient Lift Coefficient
Drag Coefficient Drag Coefficient
and drag, main difference being in the lift coefficient 
0.5
drop occurring at higher angles between 19° and 20°. 
0.4
 
0.3
0.5 0.2
0.4 0.1
0
0.3
0.2 Angle of Attack (deg)
Flatback Dimple Lift Control Lift Flatback Dimple Drag Control Drag
0.1  
0
Figure 17 
Angle of Attack (deg) Lift and Drag Comparison Between the Flatback Dimple 
Flatback Lift Control Lift Flatback Drag Control Drag Blade and the Control Blade, Re = 2.95E5 
 
 
Figure 16 
Lift and Drag Comparison Between the Flatback Blade and 0.5
the Control Blade, Re = 2.95E5 
0.4
Introducing the dimples to the flatback design 
0.3
has increased the lift coefficient on positive angles 
0.2
of attack by a noticeable, yet not significant, average 
of 3.71%. However, the drag coefficient increases by 0.1
an average of 10.28% over the control blade. 0
At near stall angles between 9° and 13°, the Angle of Attack (deg)
blade seems to be suffering from some form of Flatback Dimple Lift Flatback Lift Flatback Dimple Drag Flatback Drag
 
instability, based on the fact that the data collected is 
not as consistent compared to the rest of the domain. Figure 18 
The Flatback Dimple model also shares the Lift and Drag Comparison Between the Flatback Dimple 
characteristic with its smooth counter that the Blade and the Flatback Blade, Re = 2.95E5 
pressure drop occurs between angles of 19° and 20°. The M-shape serrations have markedly 
When comparing the two flatback designs with each introduced an improvement in lift performance for 
other, the increase in lift reduces to 1.93% and drag an average of 7.43%, with the maximum lift 
coefficient increases to 5.94%. Both the Flatback increasing by 9.94% over the control. However, the 
blade and the Flatback Dimpled blade have similar drag coefficient has increased for an average of 
aerodynamic behavior.  22.62%. Interestingly, the pressure drop occurring at 
17° does not have as pronounced an effect on the lift 
as it did on the previous blades. This drop can still 
be seen on the drag curve. The zero-lift angle 
serrated blade occurs at -5.56°, 16.14% further 
behind what was observed in the control blade. 
Lift Coefficient
Drag Coefficient
Lift Coefficient Lift Coefficient
Drag Coefficient
Drag Coefficient
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
Angle of Attack (deg) Angle of Attack (deg)
Serrated Lift Control Lift Serrated Drag Control Drag Serrated Dimple Lift Serrated Lift Serrated Dimple Drag Serrated Drag
  
Figure 19 Figure 21 
Lift and Drag Comparison Between the Serrated Blade and Lift and Drag Comparison Between the Serrated Dimple 
the Control Blade, Re = 2.95E5 Blade and the Serrated Blade, Re = 2.95E5 
Adding the dimples to the serrated blade does The following Table 1 organizes the main 
not significantly alter the aerodynamic performance aerodynamic characteristics for each blade model, 
to the blade. In comparison with the smooth serrated namely the Maximum lift coefficient and its angle, 
blade the average lift only increases by 0.77% while the zero-lift angle, and the maximum lift-to-drag 
the drag increases by 4.79%. When compared to the ratio with its angle. In general, both Serrated blades 
control blade, lift and drag are increased by 12.6% and the Flatback dimple blade have improved 
and 31.55%, respectively, for positive angles. The maximum lift coefficient. Only the Serrated dimple 
maximum lift coefficient is 10% higher compared to blade has a noticeable reduction in the angle of 
the control blade, occurring at 10.31°.  maximum lift coefficient and the zero-lift angle. Due 
to the large increases in drag for all blade variations, 
0.5
the Cl/Cdmax of the control blade is the highest of the 
0.4 group. It does not appear that the dimples assist in 
0.3 the aerodynamic performance of the blades in this 
study; they only managed to increase drag. It also 
0.2
seems that the serrations used only increases drag by 
0.1
an average of 0.20% while reducing total lift by an 
0 average of 0.072% compared to the flatback design.  
Angle of Attack (deg) Table 1 
Serrated Dimple Lift Control Lift Serrated Dimple Drag Control Drag
 Aerodynamic Characteristics of Experimented Blades 
Figure 20 Model Characteristic Clmax αClmax αCL=0 CL/CDmax αCL/CDmax 
Lift and Drag Comparison Between the Serrated Dimple 
Blade and The Control Blade, Re = 2.95E5 Control 1.32 12.00 -4.45 10.42 2.00 
 Control Dimple 1.24 13.00 -4.63 9.65 3.00 
Flatback 1.33 13.00 -4.57 9.65 5.00 
Flatback Dimple 1.39 13.00 -5.15 9.15 5.00 
Serrated 1.44 12.00 -5.17 9.35 1.00 
Lift Coefficient Lift Coefficient
Drag Coefficient
Drag Coefficient
Lift Coefficient
Drag Coefficient
Serrated Dimple 1.45 10.31 -6.15 8.92 4.16 
Comparing all these blades, side by side, we get 
the following charts for lift and drag coefficients, 
Figure 22 and Figure 23 respectively. These charts 
are polynomial trendlines based on the data acquired 
during the experiments. As previously stated, we can 
see how similar the two Flatback models and the 
Control Dimple model are to the Control Blade and 
how much more lift is generated by the two Serrated 
models. We can also see that all increase in lift is 
accompanied by a larger increase in drag for each of 
the blades compared to the control. 
 
Lift Coefficient, Re = 295,000
Control Airfoil
Flatback Airfoil
Dimpled Flatback Airfoil
Dimpled Control Airfoil
Serrated Airfoil
Dimpled Serrated Airfoil
Angle of Attack (deg)
 
Figure 22 
Trendlines for Lift Coefficient of All Experimented Blades 
Drag Coefficient, Re = 295,000
Control Airfoil
Flatback Airfoil
Dimpled Flatback Airfoil
Dimpled Control Airfoil
Serrated Airfoil
Serrated Dimple Airfoil
Angle of Attack (deg)
 
Figure 23 
Trendlines of Drag Coefficient for All Experimented Blades 
 
DISCUSSION the maximum thickness of the airfoil occurs. Flow 
separation does not occur near this area when 
The researcher speculates that the position of 
reaching the maximum lift angle; therefore, the 
the upper surface dimples is one of the major 
recirculation provided by the dimples does not 
causes of unexpected results. These dimples were 
impact the boundary layer. The thickness of the 
placed far forward on the chord line, right where 
Drag Coeffifient
Lift Coefficient
serrations might also have negatively affected the [7]  J. Mathew, A. Singh, J. Madsen and C. A. León, 
"Serration Design Methodology for Wind Turbine 
results of the experiments. Both of these features Noise Reduction," Journal of Physics: Conference 
should demand further investigation.  Series, September 2016.  
[8]  X. Liu, M. Azarpeyvand and R. Theunissen, "On The 
The original objective of this experiment was Aerodynamic Performance of Serrated Airfoils," in 
The 22nd International Congress on Sound and 
to improve the aerodynamic characteristics of a Vibration, Florence, Italy, 2015.  
wind turbine blade by introducing modifications [9]  X. Liu, H. K. Jawahar and M. Azarpeyvand, "Wake 
to the upper surface and to the trailing edge of the Development of Airfoils with Serrated Trailing 
Edges," 2016. 
blade. During the course of the study, the [10]  O. C. Seix, "Aerodynamic study on the design and 
investigator found that these modifications did not optimization of flatback airfoils for wind turbine 
applications," Barcelona, 2015. 
improve the lift performance of the blade enough 
[11]  E. S. Ferry, "Estudio aerodinámico y optimización de 
to justify the increase of drag that has been perfiles flatback aplicado a aerogeneradores.," 
Universitat Politècnica de Catalunya, 2014. 
observed. The serration geometry studied seems 
[12]  C. v. Dam, D. L. Kahn and D. E. Berg, "Trailing Edge 
to not have any drag improvements over a Modifications for Flatback Airfoils," Sandia National 
flatback airfoil. The dimpled implemented on Laboratories, Albuquerque, 2008. 
each blade did little action to reenergize the flow.   
 
The investigator believe that research into the 
 
dimples should be more focused in varying the 
geometry and size of the dimple, the height or 
depth of indentation, the angle of these shapes 
relative to the flow direction, their pattern 
geometry, and their positioning relative to the 
chord line. These future works will be attended in 
the aerodynamic laboratory of the PUPR. 
REFERENCES 
 
[1]  J. Scott, "Ask Us - Golf Ball Dimples & Drag," 13 
February 2005. [Online]. Available: 
http://www.aerospaceweb.org/question/aerodynamics/
q0215.shtml. [Accessed July 19, 2019]. 
[2]  S. S. Mahamuni, "A Review on Study on Aerodynamic 
Characteristics of Dimple Effect on Wing," 
International Journal of Aerospace and Mechanical 
Engineering, vol. 2, no. 4, July 2015.  
[3]  E. Livya, G. Anitha and P. Valli, "Aerodynamic 
Analysis of Dimple Effect on Aircraft Wing," 
International Journal of Mechanical, Aerospace, 
Industrial, Mechatronic and Manufacturing 
Engineering, vol. 9, no. 2, pp. 350-353, 2015.  
[4]  M. Mashud, D. Mondal and M. E. Haque, 
"Experimental Investigation of Airfoil with Multiple 
Dimples on the Upper Surface," in International 
Conference of Mechanical Engineering and 
Renewable Energy, Chittagong, Bangladesh, 2017.  
[5]  X. Y. Wang, S. Lee, P. Kim and J. Seok, "Aerodynamic 
Effect of 3D Pattern on Airfoil," Transactions of the 
Canadian Society for Mechanical Engineering, vol. 
39, no. 3, pp. 537-545, 2015.  
[6]  C. v. Dam, "Airfoils for Structures - Passive and Active 
Load Control for Wind Turbine Blades," in Sandia 
Blade Workshop, Albuquerque, 2004.