Jansen Daniel PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 116

Master of Science Thesis

Passive Flow Separation Control


on an Airfoil-Flap Model
The Effect of Cylinders and Vortex Generators

D.P. Jansen

August 2012

Faculty of Aerospace Engineering · Delft University of Technology


Passive Flow Separation Control
on an Airfoil-Flap Model
The Effect of Cylinders and Vortex Generators

Master of Science Thesis

For obtaining the degree of Master of Science in Aerospace Engineering


at Delft University of Technology

D.P. Jansen

August 2012

Faculty of Aerospace Engineering · Delft University of Technology


Delft University of Technology

Copyright ¥ c Aerospace Engineering, Delft University of Technology


All rights reserved.
Delft University Of Technology
Department of
Aerodynamics

The undersigned hereby certify that they have read and recommend to the Faculty of Aerospace En-
gineering for acceptance a thesis entitled “Passive Flow Separation Control on an Airfoil-Flap
Model” by D.P. Jansen in partial fulfillment of the requirements for the degree of Master of Science.

Dated: August 2012

Head of department:
Prof.dr.ir.drs. H. Bijl

Supervisor:
Dr.ir. L.L.M. Veldhuis

Reader:
Dr.ir. B.W. van Oudheusden

Reader:
Ir. W.A. Timmer
Acknowledgements

This report is written as a thesis for the degree of Master of Science at the Faculty of Aerospace Engi-
neering, Department of Aerodynamics, at Delft University of Technology.
I would like to take the opportunity to thank all the people who have helped me in accomplishing this
graduation research. I would like to thank Dr. ir. L.L.M. Veldhuis, my supervisor, for his full support,
enthusiasm and sharing of his in depth knowledge on experimental research and the field of flow separation
control. I could not have wished for a better advisor for this project.
I also would like to thank Dipl. ing. S. Bernardy and L. Molenwijk for their willingness in answering my
questions and their assistance in performing the wind tunnel experiments.
Then I would like to thank my parents for their motivation and advice in times most needed. Special
thanks goes to R. Peacock for his sincere curiosity and help on the project. Finally I also want to
mention Kim-Lan, my girlfriend, because without her love and unlimited support I definitely would not
have achieved this result.

Delft, The Netherlands D.P. Jansen


August 2012

v
D.P. Jansen vi MSc. Thesis
Summary

In the on-going search for improvements in high lift device performance, the field of flow separation
control has seen explosive growth in recent years. This is due to a continuously improved understanding
of fluid mechanics, the development in experimental and computational techniques and the variety of
flow control mechanisms existing today. These multiple mechanisms can be differentiated into active and
passive flow control techniques.
In this research two passive flow separation control methods are tested on their ability to postpone flow
separation on an airfoil with trailing edge flap model of the Extra EA-400. The two methods are the
application of vortex generators on the airfoil and the positioning of vortex producing cylinders in the
slot of the airfoil flap system.
The 2D airfoil flap model was tested by experimental and numerical investigation. The flap deflection was
set at δf = 45◦ , δf = 50◦ and δf = 55◦ to test the flap at moderate to highly separated flow conditions.
Measurements in the experiment were done by pressure holes and probes and oil flow visualization.
Numerical calculations were performed in FLUENT as part of the ANSYS 13.0 package, which was also
used to create the grid. The model is a RANS analysis with the k − ω SST turbulence model and low
Re near wall treatment.
It was found in the experiments that both the vortex generators and the cylinders were able to delay flow
separation and thereby increase the lift of the airfoil flap system. Both devices are able to create vortices
that entrain flow into the boundary layer, which gets re-energized. The cylinders proved to be superior
compared to the vortex generators in the ability to postpone the point of flow separation. With respect
to size and positioning the cylinders also showed a larger bandwidth than the vortex generators at which
the devices were effective. The results from the experiments indicated that the reduced frequency for
the cylinder vortex shedding resulting in the largest delay in flow separation, matched the theoretical
predictions found in literature. Especially at high flap deflections significant lift gains were obtained up to
6% for the vortex generators and 18% for the cylinders at a deflection of δf = 55◦ . At the flap deflection
of δf = 45◦ the flow separation delay capabilities of the cylinders are less effective. The frequency tests,
oil flow visualization and boundary layer measurements were used to understand the performance and
working principles of the cylinders.
The numerical simulations showed difficulties in reproducing the results for the cylinders obtained from
the experiment. While for the attached airfoil flow the numerical simulations showed good agreement
with the experiment, the flow separation on the flap at high deflections was not accurately modelled for
both the baseline and cylinder configuration of the airfoil flap. The vortices created in the simulation
showed a different frequency than found in the experiment and only minor effects of the cylinder were
observed. Separation on the flap was very slightly delayed, although no significant mixing effects of the
vortices were visible in the boundary layer.

vii
D.P. Jansen viii MSc. Thesis
Contents

Acknowledgements v

Summary vii

List of Figures xii

List of Tables xvii

Nomenclature xxi

1 Introduction 1
1.1 The need for high lift systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Limits in high lift performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Flow separation in high lift performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Plain airfoil characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Flapped airfoil characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Flow separation control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Active flow separation control techniques . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Passive flow separation control techniques . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Opportunity of research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Layout of the report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 Experimental model setup 15


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 The wind tunnel and model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Passive flow separation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Vortex generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

ix
Contents

2.3.2 Cylinders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Measurement techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 Pressure measurements and calculation of Cl and Cd . . . . . . . . . . . . . . . . . 23
2.4.2 Frequency measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4.3 Oil flow visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Experimental results 27
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Baseline model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Vortex generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Cylinders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4.1 Flap deflection δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4.2 Flap deflection δf = 50◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.3 Flap deflection δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.4 Comparison of δf effects on cylinder performance and further investigation . . . . 47
3.5 Conclusion experimental investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4 Numerical investigation 55
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 General setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 The numerical domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.4 Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4.1 Low flap deflection δf = 15 , no flow separation . . . . . . . . . . . . . . . . . . . .

61
4.4.2 High flap deflection δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.4.3 Very high flap deflection δf = 55 ◦
. . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4.4 Conclusion on baseline model validation . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4.5 Influence of tunnel wall at large flap deflection . . . . . . . . . . . . . . . . . . . . 63
4.4.6 Modelling laminar zones in CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.5 Results for baseline and cylinder configurations . . . . . . . . . . . . . . . . . . . . . . . . 67
4.5.1 Baseline and cylinder velocity field . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5.2 Vortex development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5.3 Vortex influence on the flap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.6 Conclusion on numerical investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

5 Conclusion and Recommendations 79


5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

D.P. Jansen x MSc. Thesis


Contents

Bibliography 81

A Test matrix for the wind tunnel experiments 83

B Additional pictures on the wind tunnel and model layout 85

C Output file example of pressure measurements from Profmeasure 89

MSc. Thesis xi D.P. Jansen


Contents

D.P. Jansen xii MSc. Thesis


List of Figures

1.1 General high lift system with single slotted trailing edge and leading edge flap. . . . . . . 2
1.2 Lift polar showing the effects of a trailing edge and leading flap deflection on a plain airfoil. 2
1.3 Circular-arc mean lines A, B, C obtained from the Joukowski transformation. Figure from
Smith [1975]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Flow around a cylinder with a very high circulation, such that the two stagnation points
coincide. Figure from Smith [1975]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Visualization of flow separation on a flap. Figure from Little et al. [2009]. . . . . . . . . . 4
1.6 Visualization of a laminar separation bubble. . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.7 Visualization of flow separation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.8 Lift polar showing the general effect of increasing trailing edge flap deflection. . . . . . . . 7
1.9 Triple-slot flap system of the B727-200. Figure from Rudolph [1996]. . . . . . . . . . . . . 8
1.10 PIV visualization of flow separation postponement on a wind turbine blade by applying
synthetic jets at α = 16◦ and Re = 1.6 · 105 . Figure from Maldonado et al. [2010]. . . . . 9
1.11 Components of an electrodynamic synthetic jet. Figure from Cattafesta [2011]. . . . . . . 9
1.12 Working principle of boundary layer suction. Figure from Boermans [2008]. . . . . . . . . 10
1.13 Schematization of a piezoelectric flap. Figure from Veldhuis and van der Jagt [2010]. . . . 10
1.14 Schematization of a DBD plasma actuator. Figure from LeBeau [2007]. . . . . . . . . . . 11
1.15 Vortex generators displayed in different setups. Figures from von Stillfried et al. [2010]. . 11
1.16 Effect of counter-rotating micro VG’s on RMS anemometer output from hot-film sensors
on flap upper surface with α = 8◦ and M∞ = 0.2. Figure from Lin [1999]. . . . . . . . . . 12
1.17 Lift-enhancing tabs as used by Lee et al. [2005]. . . . . . . . . . . . . . . . . . . . . . . . . 12

2.1 The Low Speed Low Turbulence Wind Tunnel at Delft University of Technology. . . . . . 16
2.2 Cross section of the model and definition of flap parameters. . . . . . . . . . . . . . . . . . 16
2.3 The model positioned in the wind tunnel with the cylinder test setup. . . . . . . . . . . . 17
2.4 Boundary layer measurements on model lower side at x/c = 0.67 . . . . . . . . . . . . . . 18

xiii
List of Figures

2.5 The model with VG’s applied (left) and cylinders (right). . . . . . . . . . . . . . . . . . . 19
2.6 Vortex generator geometrical definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.7 Vortex generator geometry and construction method. . . . . . . . . . . . . . . . . . . . . . 20
2.8 Cylinder position and definition of coordinates. Hinge point is referenced to the airfoil
nose with [x, y] = [0.72c, 0.06c]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.9 Cylinder geometrical definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.10 Reduced frequency dependency on cylinder diameter. . . . . . . . . . . . . . . . . . . . . . 23
2.11 Test setup of the wake rake. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1 Lift polars for the baseline configurations at multiple flap deflections at Re = 2.0 · 106 . . . 29
3.2 Pressure distributions for baseline configurations at multiple a.o.a for δf = 55 , δf = 50
◦ ◦

and δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Effect of height, spacing and edge distance on the lift for multiple VG configurations at
δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4 Flap and airfoil lift contributions for multiple VG configurations with h∗ = 0.023, d∗ = 0
and δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.5 Pressure distribution for optimum VG configuration at multiple a.o.a. with λ/h = 4.0,
h∗ = 0.023, d∗ = 0 and δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.6 Pressure distributions for optimum VG and baseline configuration at low and high a.o.a.
with λ/h = 4.0, h∗ = 0.023, d∗ = 0 and δf = 55. . . . . . . . . . . . . . . . . . . . . . . . . 34
3.7 Effect of size and positioning on the lift for multiple cylinder configurations at δf = 55◦ . . 35
3.8 Effect of size and positioning on the lift for multiple cylinder configurations at δf = 55 . . ◦
36
3.9 Lift gains ∆Cl dependency against reduced frequency F + at high and low angle of attack
with r∗ = 0.090, θ = 37◦ and δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.10 Flap and airfoil lift contributions for multiple cylinder configurations with D = 0.0167

and δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.11 Pressure distributions for optimum cylinder and baseline configuration at low and high
a.o.a. with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 55◦ . . . . . . . . . . . . . . . . . . 38
3.12 Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.13 Effect of positioning on the lift for multiple cylinder configurations with D = 0.0167 and

δf = 50◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.14 Flap and airfoil lift contributions for multiple cylinder configurations with D = 0.0167

and δf = 50◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.15 Pressure distributions for optimum cylinder and baseline configuration at low and high
a.o.a. D∗ = 0.0167, r∗ = 0.090, θ = 30◦ , δf = 50◦ . . . . . . . . . . . . . . . . . . . . . . . 42
3.16 Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 50◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.17 Effect of positioning on the lift for multiple cylinder configurations with D = 0.0167 and

δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.18 Flap and airfoil lift contributions for multiple cylinder configurations with D∗ = 0.0167
and δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

D.P. Jansen xiv MSc. Thesis


List of Figures

3.19 Pressure distributions for optimum cylinder and baseline configuration at low and high
a.o.a. with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 45◦ . . . . . . . . . . . . . . . . . . 47
3.20 Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.21 Frequency measurements at multiple positions behind the cylinder. Pressure fluctuations
|∆P (f ) | are plotted against frequency with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 45◦ . 50
3.22 Frequency measurement locations at the flap at δf = 45◦ . . . . . . . . . . . . . . . . . . . 51
3.23 Total pressure measurements of the boundary layer at various locations on airfoil and flap
for the baseline and optimum cylinder configuration with D∗ = 0.0167, r∗ = 0.090, θ = 37◦
and δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.1 Visualization of the numerical domain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57


4.2 Mesh in the vicinity of the cylinder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
+
4.3 ywall for the baseline model at δf = 45 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

58
4.4 Convergence of the steady state solution. . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.5 Locations of the monitor points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6 Convergence of the velocity magnitude in the monitor points. . . . . . . . . . . . . . . . . 61
4.7 Pressure distributions for CFD and experiment at low flap deflection at δf = 15◦ . . . . . 62
4.8 Pressure distributions for CFD and experiment at high flap deflection at δf = 45 . . . . . ◦
63
4.9 Pressure distributions for CFD and experiment at very high flap deflection at δf = 55◦ . . 64
4.10 Total domain with modelled tunnel walls on upper and lower side. . . . . . . . . . . . . . 64
4.11 Pressure distributions for the CFD simulation modelled with and without tunnel wall at
δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.12 Designation of laminar and turbulent zones around the airfoil and flap at δf = 45 . ◦
. . . 66
4.13 Numerically calculated pressure distributions with/without laminar zones compared to
experiment at δf = 45◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.14 Total pressure measurements in the boundary layer visualized for numerical simulations
modelled with laminar zones and for fully turbulent simulations at δf = 45◦ . Measurements
are performed at x/c = 0.787. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.15 Mean velocity field and streamlines for the baseline and cylinder configuration at δf = 55 . 69 ◦

4.16 Mean velocity field and streamlines in the slot for the baseline and cylinder configuration
at δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.17 Mean pressure distribution for the baseline and cylinder configuration at δf = 55 . ◦
. . . 71
4.18 Mean pressure vectorization for the baseline and cylinder configuration at δf = 55◦ . . . . 71
4.19 Mean skin friction coefficient for the baseline and cylinder configuration at the flap upper
surface at δf = 55◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.20 Vorticity magnitude at different time instances at very high flap angle. δf = 55 ◦
. . . . . 73
4.21 Lift coefficient development in time for the baseline and cylinder configuration at δf = 55◦ . 74
4.22 Velocity magnitude development in time at the monitor points at δf = 55◦ . . . . . . . . . 75
4.23 Visualizations of the line rakes at x = 0.8965c and at x = 0.9365c. . . . . . . . . . . . . . 75

MSc. Thesis xv D.P. Jansen


List of Figures

4.24 Horizontal velocity component at x/c = 0.8965 for very high flap angle at δf = 55◦ . . . . 76
4.25 Horizontal velocity component at x/c = 0.9365 for very high flap angle at δf = 55 . ◦
. . . 77

B.1 Exterior view of the wind tunnel test section. . . . . . . . . . . . . . . . . . . . . . . . . . 86


B.2 Interior view, looking down stream, of the wind tunnel test section. Clearly visible are the
pitot tube and the wake rake that is positioned behind the model. . . . . . . . . . . . . . 86
B.3 Detailed view on the positioning of the cylinders. The mechanism supporting the cylinder
is clearly visible as well as the zigzag tape that was used to remove the lower surface
laminar separation bubble. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
B.4 Detailed view of the suction orifices at the wing/wall junction shown here around the airfoil
upper surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
B.5 View of the zigzag positioning of the pressure holes on the airfoil and flap model surface. 88

C.1 Raw data of pressure measurements for baseline run at δf = 55◦ . . . . . . . . . . . . . . . 92

D.P. Jansen xvi MSc. Thesis


List of Tables

2.1 Selected cylinder sizes with corresponding F + . . . . . . . . . . . . . . . . . . . . . . . . . 22

A.1 Test matrix for the vortex generators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83


A.2 Test matrix for the cylinders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

xvii
List of Tables

D.P. Jansen xviii MSc. Thesis


Nomenclature

Abbreviations
2D 2-Dimensional
3D 3-Dimensional
AC Alternating Current
CCD Charge-Couple Device
CFD Computational Fluid Dynamics
CPU Central Processing Unit
DBD Dielectric Barrier Discharge
DES Detached Eddy Simulation
EXP Experimental
FCD Flow Control Device
FFT Fast Fourier Transform
LE Leading Edge
LES Large Eddy Simulation
NACA National Advisory Committee for Aeronautics
PISO Pressure-Implicit with Splitting of Operators
PIV Particle Image Velocimetry
RAM Random-Access Memory
RANS Reynolds Average Navier-Stokes
SST Shear-Stress Transport
TE Trailing Edge
UV Ultraviolet

xix
List of Tables

VG Vortex generator

Coefficients
Cd Drag coefficient
Cl Lift coefficient
Cl0 Lift coefficient at zero angle of attack
Clmax Maximum lift coefficient
Cn Normal coefficient
cp Pressure coefficient

Greek Symbols
α Angle of attack
αs Stall angle
λ Spacing between two VG pairs
µ Dynamic viscosity
σmon Convergence parameter
τw Wall shear stress
β Angle between VG’s and flow direction
β Curvature
ω Specific dissipation rate of turbulent kinetic energy
δ Boundary layer thickness
δf Flap deflection angle
∆ Change in variable

Latin Symbols
c Airfoil chord
D Cylinder diameter
d Distance between FCD and separation point
D∗ Cylinder diameter in units of c
d∗ Distance between FCD and separation point in units of c
f Frequency
F+ Reduced frequency
h Height of a VG

D.P. Jansen xx MSc. Thesis


List of Tables

h∗ Height of a vortex generator in units of c

L Sound pressure level


Lf Length of object over which flow separation is delayed

M∞ Free stream Mach number

P Pressure fluctuation in frequency domain

p∞ Free stream pressure


ps Pressure fluctuation

pref Reference sound pressure

pt Total pressure
q∞ Free stream dynamic pressure

r Cylinder distance from reference point

r∗ Cylinder distance from reference point in units of c


Re Reynolds number

Rey Turbulent Reynolds number

S Distance between cylinder and wall

s Spacing between two VG’s in a pair

St Strouhal number

t Time
t0 Reference time

U Total velocity

U∞ Free stream velocity


x Chordwise coordinate

y Normal coordinate

y+ Inner distance variable for a boundary layer

MSc. Thesis xxi D.P. Jansen


List of Tables

D.P. Jansen xxii MSc. Thesis


CHAPTER 1

Introduction

High lift aerodynamics has always been an important aspect of aircraft design. This is due to its major
influence on aircraft performance in the essential flight phases of take-off and landing. Flap systems
seen on aircraft today are primarily based on technology from the early 1940s, but studies in high lift
design have gained a new interest. Although many of the currently used high lift devices consist of
outdated complex multi-element flap systems, there is a growing trend in recent years for simpler, more
cost effective designs. These mechanically simpler systems are still required to perform well in terms
of aerodynamic performance. Increasing computational resources has given numerical approaches like
CFD (Computational Fluid Dynamics) more possibilities to study the flow around high lift devices. This
in combination with still developing experimental techniques in wind tunnel testing, like LDV (Laser
Doppler Velocimetry) and PIV (Particle Image Velocimetry), has increased knowledge and performance
of high lift aerodynamics to comply to the current demand of effective high lift systems.

1.1 The need for high lift systems

For several decades airplanes have been applied with high lift devices. The necessity of high lift systems
lies in the fact that under certain circumstances the lift generated by the wings is not enough for the
required aircraft performance. Although the wings produce enough lift for the aircraft to fly at cruise
conditions, they do not suffice at take-off and landing. A lift increase is necessary in order to climb to the
cruise altitude at take-off and to be able to fly at the necessary low manoeuvre speeds in case of landing.
While a rise in lift for the wings is easily obtained by increasing the angle of attack, which is always
done during take-off and landing, the necessary lift improvement to obtain maximum performance of the
aircraft is unrealizable without the use of high lift systems. According to Meredith [1992] increasing the
performance of high lift devices has the following possible benefits:

• A shorter take-off and landing field length is required.

• A higher take-off weight is possible.

• An increase in range can be accomplished.

1
Chapter 1. Introduction

Figure 1.1 – General high lift system with single slotted trailing edge and leading edge flap.

Cl

Airfoil
Airfoil + trailing edge flap
Airfoil + leading edge flap

0 α

Figure 1.2 – Lift polar showing the effects of a trailing edge and leading flap deflection on a plain airfoil.

• Steeper climb or approach angles are possible, enhancing safety during landing. This steeper climb
or approach also reduces noise emission at ground level.
• A reduction in stall speed can be achieved, leading to improved slow dive performance for emergency
landings.

The most widely used high lift system is the configuration of an airfoil with a leading edge flap (slat)
and at least one, often more, trailing edge flaps as is seen in figure 1.1. There are different types of both
devices that have specific effects on the wing performance. In essence their goal is to increase the camber
of the wing and to enlarge the wing surface area. These characteristics are realized by a flap and slat
extraction and deflection, which both contribute to the gain in total lift of the wing. The lift gaining
effects can best be visualized in a lift polar, which shows the dependency of the lift coefficient on the
angle of attack. Figure 1.2 shows the lift polar for a plain airfoil and for an airfoil with trailing/leading
edge flap.
The plain airfoil shows the well-known lift development with a linear relation between Cl and α at lower
angle of attack, followed by the stagnation of lift at the stall angle until it drops at higher angle of attack.
It is seen that the trailing edge flap deployment delivers a direct increase in lift when compared to the
plain airfoil over the whole range of angle of attack. Similar to the plain airfoil lift development the lift
increases proportional with higher angle of attack until a maximum is reached. It is seen that the stall
angle at this lift maximum for the trailing edge flap deflection is lower than for the plain airfoil. Leading
edge flaps do not have an immediate effect on the lift at low angles of attack, but due to an increase of
the stall angle, realize higher maximum lift values.

D.P. Jansen 2 MSc. Thesis


1.2. Limits in high lift performance

Figure 1.3 – Circular-arc mean lines A, B, C obtained from the Joukowski transformation. Figure from
Smith [1975].

Thus leading and trailing edge flaps indeed are beneficial in obtaining a higher lift, clarifying their use
on current aircraft at take-off and landing. It is also clear that certain limits exist in lift performance,
as is shown in figure 1.2 at high angle of attack. The working principles of high lift devices need further
explanation, but first it is necessary to seek clarification concerning the observed limits that apply in lift
performance.

1.2 Limits in high lift performance

To better understand what may be attainable in high lift performance it is necessary to have knowledge
of theoretical maximum possible lift limits. Smith [1975] considered the flow around circular-arc mean
lines in potential flow conditions. The circular-arc mean lines are shown in figure 1.3. Stagnation points
are present at both ends of the arcs. By applying Joukowski airfoil theory to these arcs Smith [1975]
found the following definitions for obtainable lift:

Cl = 2π (sin (α + β)) /cosβ (0◦ ≤ β ≤ 45◦ ) (1.1)

Cl = 4πsinβsin (α + β) (45◦ ≤ β ≤ 90◦ ) (1.2)

In the case of β = 90◦ the two stagnation points are so close together that a single stagnation point is
formed. Figure 1.4 visualizes the flow under these conditions. It is this flow situation where the highest
possible lift in theory is attained for which equation 1.2 gives Clmax = 4π. In the case of an angle of
attack of α = 45◦ and β = 45◦ , which resembles quite closely the curvature of modern day flapped
airfoils, the theoretical maximum is determined by equation 1.1 as Clmax = 8.89. This still seems rather
high, since even aerodynamically advanced triple slotted flap systems have lift coefficients in the range
of Clmax = 3.5 − 4.0 at maximum, as identified by Torenbeek [1976]. So what is the reason that the
theoretical values for Clmax are far from being achieved in practice? The answer lies in the assumption
of potential flow in determining the theoretical limits. In potential flow viscous effects are not taken
into account and boundary layers around objects are not modelled. In reality these boundary layers and
viscous effects play a very important role in the behaviour of the flow around high lift systems. In the real
world the appearance of boundary layers incorporate flow phenomena like transition and flow separation.
It turns out that especially the aspect of flow separation has a dominant impact on the performance
of high lift systems. Figure 1.5 graphically illustrates this phenomenon of flow separation at a highly
deflected flap. This picture is obtained from PIV measurements on an airfoil and flap at zero incidence
on a δf = 30◦ flap deflection performed by Little et al. [2009].
It is seen that flow is separated over the complete length of the flap and a drastic loss in lift efficiency is
encountered. The general explanation for flow separation to occur is that the momentum of the flow in the

MSc. Thesis 3 D.P. Jansen


Chapter 1. Introduction

Figure 1.4 – Flow around a cylinder with a very high circulation, such that the two stagnation points
coincide. Figure from Smith [1975].

Figure 1.5 – Visualization of flow separation on a flap. Figure from Little et al. [2009].

vicinity of the wall is too low for the boundary layer to overcome the occurring adverse pressure gradient.
The flap deflection of 30◦ tested by Little et al. [2009] is representative for deflections at which flow
separation generally occurs for single slotted flaps. While equations 1.1 and 1.2 suggest that curvature
β and angle of attack α can easily be increased to values of 45◦ for maximum lift, practice shows that
this is not possible without the onset of flow separation and thus much lower lift limits apply. The flow
separation has its unfavourable impact on the system shown in figure 1.5 on a high flap deflection, but it
is also the reason for the occurrence of stall at high incidence angles as was observed earlier in figure 1.2.
Also at stall of (flapped) airfoils high adverse pressure gradients appear with the onset of flow separation
as a result.
While the theoretical limits from potential flow are not representative for real world high lift applications,
they do indicate that there is room for improvement in performance of current high lift devices in case
phenomena like flow separation are tackled. In this thesis an investigation will be performed in how flow
separation can be eliminated or postponed on an airfoil flap system with high flap deflection. For this a
further understanding of the working principles of high lift devices is required together with the way flow
separation affects the performance.

D.P. Jansen 4 MSc. Thesis


1.3. Flow separation in high lift performance

Figure 1.6 – Visualization of a laminar separation bubble.

1.3 Flow separation in high lift performance


The concept of flow separation in high lift performance has been studied from the early pioneering years
in aviation up until today. Wright et al. [1924] obtained a patent on the design of a split flap that included
a good qualitative understanding of the flow processes including flow separation. Although improvements
in mathematical analysis, experimental methods and numerical simulations have been made ever since,
modern aerodynamics still has difficulty accurately predicting flow separation. Flow separation does have
a simple definition in mathematical terms. The occurrence of flow separation is often expressed through
the definition of the wall shear stress, which has the form:

∂U
τw = µ (1.3)
∂y y=0

Flow separation occurs if the velocity gradient at the wall is zero, thus when the wall shear stress is
zero. The problem in localizing flow separation in numerical simulations is to accurately calculate this
wall shear stress. The following section will further describe in more detail the known process of flow
separation in the case for plain and flapped airfoils.

1.3.1 Plain airfoil characteristics

As explained earlier, increasing the angle of attack for an airfoil towards the stall angle for high lift
performance will eventually cause flow separation to occur on the airfoil. For a plain airfoil Torenbeek
[1976] states that there are multiple flow separation types encountered for stall: thin airfoil stall, leading
edge stall, trailing edge stall or a combination of leading and trailing edge stall. The type of stall is mainly
determined by the Reynolds number and the curvature of the nose, which affects another phenomenon
called laminar to turbulent flow transition. The way this flow transition takes place at higher angle of
attack determines the type of stall that will occur.
Leading edge stall and thin airfoil stall are quite similar. At lower incidence angles, the high adverse
pressure results in a flow transition where the laminar boundary layer separates and quickly reattaches
as a turbulent boundary layer. This is called a short separation bubble, see figure 1.6 for a graphical
illustration. For thin airfoils operating at higher incidence angles, the reattachment point gradually moves
aft producing a long separation bubble. Ultimately, the boundary layers fail to reattach on the airfoil
and flow separation has occurred as visualized in figure 1.7. For leading edge stall the short separation
bubble also transforms into a long separation bubble, but in this case the change is more abrupt.
Trailing edge stall occurs at higher Reynolds number and for thicker airfoils where the laminar to turbulent
flow transition is more focused in a point, instead of a short or long separation bubble. In this case

MSc. Thesis 5 D.P. Jansen


Chapter 1. Introduction

Figure 1.7 – Visualization of flow separation.

separation is determined by the boundary layers’ ability to cope with the general pressure rise at the
trailing edge, which leads to trailing edge separation at a higher angle of attack. Combined stall occurs
at intermediate Reynolds number and leading edge thickness where both leading edge and trailing edge
stall phenomena occur.

1.3.2 Flapped airfoil characteristics

For flapped airfoil systems high lift performance and the phenomenon of flow separation at stall is more
complicated than for single airfoil configurations. This is because the multiple elements, which have their
own flow characteristics, also influence each other. This affects the possible onset of flow separation,
which can occur on both the flap and the airfoil. Over all, more parameters are involved in determining
the high lift performance and flow separation characteristics.

Multi-element flow effects

To better understand how the parameters determine the high lift and flow separation characteristics of the
high lift system, it is necessary to more accurately identify the working effects of the multiple elements.
Like said, the elements influence each other. These effects are primarily pressure influences, which are
well described by Smith [1975]:

• Slat effect: The circulation on a forward element reduces velocities on the downstream element,
which reduce negative pressure peaks on its nose.

• Circulation effect: In turn, the circulation on the downstream element induces velocities on the
upstream element, particular near the trailing edge, which increases its circulation.

• Dumping effect: The increased velocity at the trailing edge of the forward element relieves the upper
surface pressure recovery impressed on the boundary layer, so alleviating separation problems.

• Off-the-surface pressure recovery effect: The boundary layer from forward elements is dumped at
velocities appreciably higher than free stream. The final deceleration to free stream velocity is done
in an efficient way; without this effect the boundary layer is unable to overcome the entire pressure
rise. The deceleration of the wakes occur out of contact with a wall; this is usually more effective
than the best possible deceleration with a wall.

• Fresh boundary layer effect: Each new element starts out with a fresh boundary layer at its leading
edge. Thin boundary layers can withstand stronger adverse pressure gradients than thick ones.
Hence, breaking up a boundary layer into several thinner boundary layers is favourable to the delay
of separation.

D.P. Jansen 6 MSc. Thesis


1.3. Flow separation in high lift performance

Cl

δf = 0◦
δf = 20◦
δf = 30◦
δf = 40◦

0 α

Figure 1.8 – Lift polar showing the general effect of increasing trailing edge flap deflection.

The earlier mentioned camber effect of the slat and flap deployment is a more general explanation of
the detailed described effects stated above. These effects also explain the working principles of the slat
and flap on the lift performance of a plain airfoil earlier shown in figure 1.2. Torenbeek [1976] elucidates
that in general the deflection of a trailing edge flap results in an increased circulation, which induces an
upwash at the nose. The adverse pressures increase and airfoils liable to leading edge stall will show flow
separation at an angle of attack, which is below that of the plain airfoil. The working effect of a slat
counteracts this effect by reducing the negative pressure at the nose, thereby increasing the stall angle
and maximum obtainable lift.

Flow separation on a flap

Similar to the initiation of stall at high angle of attack for a plain airfoil, the deflection for a flap shows
a behaviour where at too large angles the flow can separate. The trailing edge flap hereby acts kind of
as a plain airfoil where dependent on the flap shape, multiple types of stall can occur.
Figure 1.8 shows the general effect of multiple flap deflections for a trailing edge flap in a lift polar. It
shows that initially the lift polar shifts up for higher flap deflections at the cost of a lower stall angle.
However there is a deflection at which a further increase of the flap deflection results in flow separation
and the lift will stagnate and even drop at even higher deflections. At this point flow separation is
occurring at the flap even on low angles of attack.
Besides the flap deflection there are several other design parameters that determine the effects described
by Smith [1975] and the possible onset of flow separation. Flap chord lengths, gap sizes, slot shapes
all affect flap performance, although for trailing edge flaps the most direct and important parameter
remains the flap deflection. Numerous studies have been performed by Recant [1940] and Cahill [1949],
to accurately identify the performance of the multiple design parameters. It was also discovered that
multiple slots resulting in several flap elements were favourable in postponing flow separation compared
to single element counterparts. This led to the development of multi-slotted flap systems designed in
the 1960’s - 1970’s, where performance increase was found through all the effects described by Smith
[1975], although these concepts were not even fully understood at that time. The multi-slotted flap
systems realized higher flap deflections and were able to tackle flow separation. Typical values for triple

MSc. Thesis 7 D.P. Jansen


Chapter 1. Introduction

Figure 1.9 – Triple-slot flap system of the B727-200. Figure from Rudolph [1996].

slotted flaps were around δf = 50◦ deg to generally δf = 30◦ for single slotted flaps. This led to very
sophisticated triple slotted high lift device systems being applied by Boeing on the B727, see figure 1.9,
up to the B737 and eventually also the B747, at which the triple slotted inboard and outboard trailing
edge flaps found their complexity peak. While lift gains, identified by Torenbeek [1976], of 40% compared
to the single slotted counterparts were no exception, the development of multi-slotted flap systems also
showed downsides. The deployment of multi-slotted flaps requires a mechanically complex system. This
affects the development time and costs and puts high pressure on maintenance management. In addition,
an aerodynamic downside is that the external hinges and actuators necessary for these systems generate
parasitic drag when stowed in the cruise configuration. Of course there is also the direct weight penalty
on the aircraft for such bulky systems. This led to a change of view that nowadays simpler systems are
preferable as identified by Ashby [1996] and Rudolph [1996]. More knowledge of high lift performance
and flow separation is obtained over the past decades with the growing trend of more accurate numerical
simulations and experimental methods. This helped the development of new techniques in controlling
flow separation and some of them already find their real life applications today. Although these methods
are new, in general the goal remains the same as with the complex multi-slotted systems. That is to
find ways to delay flow separation and thereby increase the possible flap deflection and improve high lift
performance.

1.4 Flow separation control

For a couple of decades other approaches, besides the multi-slot flap systems, have been studied in order to
control flow separation. The idea is to return to simpler single or double slotted flaps and incorporate these
new flow control techniques. These approaches can be divided into two types of flow separation control
techniques: active flow separation control and passive flow separation control. Active flow separation
control techniques are based on putting energy into the flow, while passive control techniques do not
induce energy in the system. In practice, active flow separation control can lead to higher lift performance
improvements compared to passive techniques, often at the cost of increased complexity of the system.
Both flow control techniques are based on either directly increasing the momentum in the boundary layer
or by creating vortices transporting higher momentum free stream flow to within the boundary layer.
Increasing the momentum of a boundary layer will generally increase the ability to overcome the adverse
pressure gradient.

1.4.1 Active flow separation control techniques

Cattafesta [2011] classifies the active flow control techniques into the following solutions: fluidic, moving
object/surface, plasma and others. This section will briefly describe each of these control techniques.

D.P. Jansen 8 MSc. Thesis


1.4. Flow separation control

(a) Baseline situation. (b) Situation with sinusoidal actuation of syn-


thetic jets.

Figure 1.10 – PIV visualization of flow separation postponement on a wind turbine blade by applying
synthetic jets at α = 16◦ and Re = 1.6 · 105 . Figure from Maldonado et al. [2010].

Figure 1.11 – Components of an electrodynamic synthetic jet. Figure from Cattafesta [2011].

Fluidic actuators

Fluidic actuators use fluid injection or suction to obtain a certain amount of control on the flow. Although
many subclasses exist, the two most commonly used fluidic actuators are synthetic jets and boundary layer
suction/blowing.
Figure 1.11 shows a schematic of an electrodynamic synthetic jet configuration. Synthetic jets are based
on alternately ingesting and expelling fluid into the flow to create vortices and a higher momentum
boundary layer. For the actuator shown in figure 1.11 this is done by a diaphragm, which will oscillate
under influence of electrodynamic transduction. The magnet generates a magnetic field with a magnetic
flux density, which under the influence of an alternating current (AC-current) in the wound coils, results
in an alternating force induced on the coils. This causes the diaphragm to oscillate, which results in
fluid flowing in and out the cavity through an orifice or slot leading to vortices in the boundary layer.
Maldonado et al. [2010] performed PIV measurements on wind turbine blades equipped with synthetic
jets at high angle of attack. Significant improvements in CLmax were found in the order of 12% at a
delayed stall angle of 2◦ compared to the baseline configuration. The delay in flow separation on the
turbine blade is illustrated in figure 1.10.
Boundary layer suction/blowing is another way to increase the momentum in the boundary layer. With
boundary layer suction this is done by removing the low momentum flow in the vicinity of the wall,
where usually the fluid is expelled at another location. Boundary layer blowing directly increases the
momentum of the boundary layer and can even be done oscillatory to add vortices to the flow as well.
Figure 1.12 graphically explains the working principle of boundary layer suction.

MSc. Thesis 9 D.P. Jansen


Chapter 1. Introduction

Figure 1.12 – Working principle of boundary layer suction. Figure from Boermans [2008].

Figure 1.13 – Schematization of a piezoelectric flap. Figure from Veldhuis and van der Jagt [2010].

Moving object/surface actuators

There are several types of moving object/surface actuators, but the most common is the piezoelectric
flap. An AC-voltage across the piezoelectric device causes the flipping motion of the flap, which then
interacts with the flow. In this way vortices can be created in different sizes and direction depending on
the geometry and orientation with respect to the local free stream flow. A typical example of a cavity-type
piezoelectric flap is shown in figure 1.13.

Plasma actuators

Plasma actuators come in different forms, having slightly different techniques to obtain the plasma. The
most popular variant is the Dielectric Barrier Discharge (DBD) plasma actuator, which consists of two
electrodes that are separated by a dielectric material. The air passing the electrodes becomes ionized by
the voltage that is applied. The ionized air, now called plasma, produces forces in the air due to the
attained electric field gradients by the electrodes. These forces can induce velocity components to the
flow and thus can be used for effective flow control including delay of flow separation. Figure 1.14 shows
the general setup of a DBD plasma actuator. The general advantages of plasma actuators with other
active control techniques is that they are easily applicable and very compact, which gives a lot of design
freedom. The main evident disadvantage is the amount of energy necessary to supply a high voltage to
the electrodes.

1.4.2 Passive flow separation control techniques

There are various forms of passive flow separation control techniques. Unlike active flow control no
direct energy is brought into the system, meaning that passive flow control is solely based on mixing

D.P. Jansen 10 MSc. Thesis


1.4. Flow separation control

Figure 1.14 – Schematization of a DBD plasma actuator. Figure from LeBeau [2007].

(a) Counter-rotating (b) Counter-rotating (c) Co-rotating. (d) Multiple-row.


“common flow “common flow up”.
down”.

Figure 1.15 – Vortex generators displayed in different setups. Figures from von Stillfried et al. [2010].

high momentum fluid to areas of low momentum, hence to boundary layers which are on the verge of
separation.

Vortex generators

Vortex generators (VG’s) are the most commonly known passive control devices and are already used
in different industries. Although VG’s come in various shapes and sizes, in general a vortex generator
is build up as a small vertical plate positioned at an angle with respect to the local free stream flow.
With appropriate dimensioning and positioning stream wise vortices can be created which can be used to
control the flow. Vortex generators can be classified in different ways. First there exist co-rotating and
counter-rotating types, depicted in figure 1.15, where in general better results are obtained with the latter.
Then there also exists a segmentation in terms of scale, hence conventional boundary layer VG’s and more
recently the development of sub-boundary layer VG’s, also called micro vortex generators. Research in
sub-boundary layer VG’s as performed by Lin [1999] show significant improvements in reducing flow
separation comparable to their larger conventional counterparts without the effects of increased drag.
The counter-rotating micro VG’s led to lift increases up to 10% on a leading and trailing edge flapped
airfoil at accompanying drag reductions of 50%. This was achieved by a delay in the separation location
from approximately 45% flap chord to at least 85% flap chord. Figure 1.16 shows the results in flow
separation delay by means of measured anemometer RMS values tested for two Reyonolds numbers.

In general the interest in micro VG’s and creation of small-scale perturbations is quite similar to the
trend found in the field of active flow control techniques. Micro VG’s as other small-scale flow control
solutions have a small bandwidth and the location of separation needs be somewhat fixed in order for
the micro VG’s to be effective.

MSc. Thesis 11 D.P. Jansen


Chapter 1. Introduction

Figure 1.16 – Effect of counter-rotating micro VG’s on RMS anemometer output from hot-film sensors on
flap upper surface with α = 8◦ and M∞ = 0.2. Figure from Lin [1999].

Figure 1.17 – Lift-enhancing tabs as used by Lee et al. [2005].

Lift-enhancing tabs

Lift-enhancing tabs or Gurney flaps are small ’plates’, which are located generally at trailing edges of lift
generating devices. A typical Gurney flap is several boundary layers in height and is usually positioned
perpendicular to the flow.
Figure 1.17 shows the working principle of lift-enhancing tabs. Two counter rotating vortices are created
aft of the Gurney flap, entraining the flow from the airfoil upper surface around the top recirculation
region. Hence the flow stays attached over the airfoil flap surface and separation can be delayed. Ashby
[1996] found promising results, where various tabs on both the main airfoil and trailing edge flap were
tested. Lift increases up to 11% were found at some angle of attack, while the maximum lift coefficient
gained 3% with respect to the baseline configuration.
Even more active and passive flow separation control techniques exist and in general it can be said that
very interesting results are obtained with various designs. In general active flow control techniques share
the main similar disadvantage. Whether it is applying an AC-current or voltage as for the synthetic jets
and plasma actuators, or fluid being blown into the system by a pump with boundary layer blowing,
significant amount of energy needs to be put into the system. Often problems are encountered with the
practical implementation of these mechanisms. Part of these problems are alleviated by a migration of
the active flow control research field in terms of approach, as discussed by Cattafesta [2011]. Increased
flow physics understanding changed the concept from inducing large vortices to the system to focus on
creating smaller instabilities to the flow. This makes it feasible to reduce power, size and mass. However
for high-speed applications these ’small-scale’ devices currently still lack bandwidth and often have control
related issues. With passive flow control, designs are often easier applicable and design aspects like size,

D.P. Jansen 12 MSc. Thesis


1.5. Opportunity of research

mass and maintenance are concept wise less an issue. Also facets like costs and safety are considered
to be less problematic for passive flow control solutions. This is the reason why passive techniques like
vortex generators can already be found on various recent aircraft designs. This makes passive control
very if not more interesting for the near future.

1.5 Opportunity of research

In section 1.4 a variety of methods in flow separation control has been discussed including obtained results
by other research. It is identified that for the near future passive flow control can give many solutions.
One of the solutions not discussed up to this point is a passive method, researched by van der Steen
[2009], where cylinders were used on a simplified airfoil flap system to prevent separation occurring over
the flap at high flap deflections. Research shows that under specific conditions, the flow past cylinders
can generate perpendicular vortices to the free stream, more generally known as the Von-Kármán vortex
street. van der Steen [2009] has shown that these vortices can delay flow separation and can be more
effective then conventional vortex generators. This thesis will continue this research by utilizing these
passive flow separation techniques on an existing high lift device system being the airfoil flap sytem of
the Extra EA-400. As was pointed out in section 1.3 the interest lies in applying the separation control
on high flap deflections where flow separation is eminent. More specifically the goal of this thesis is: To
investigate the performance improvements of cylinders and vortex generators on the Extra EA-400 airfoil
flap system at high flap deflections by experiment and numerical simulations.
The Extra EA-400 is a six-seat single-engine monoplane produced by Extra Flugzeugbau GmbH in
collaboration with Delft University of Technology, which developed the aerodynamic design of the wing.
For the airfoil flap system of this aircraft, Delft University of Technology has a 2D prismatic wind tunnel
model. Hence, the study is a 2D investigation that will simplify the analysis compared to a 3D wing
especially for the numerical simulation. For a full 3D wing analysis, lateral effects like wing tip vortices
and wing fuselage interactions will significantly increase the complexity of the flow structures. These 3D
effects are not studied in this thesis.

1.6 Layout of the report

The layout of the report will be as follows. In chapter 2 the setup of the experiment will be discussed
which includes an overview of the wind tunnel, apparatus and the model. Further the passive flow control
devices working principles and measurement techniques will also be described. Chapter 3 will give the
results and discussion from the experiment. Then in chapter 4 the setup and results for the numerical
simulation will be discussed. Finally the conclusion and recommendations for this research will be given
in chapter 5.

MSc. Thesis 13 D.P. Jansen


Chapter 1. Introduction

D.P. Jansen 14 MSc. Thesis


CHAPTER 2

Experimental model setup

2.1 Introduction

This chapter will describe the experimental investigation performed on the airfoil flap model of the Extra
EA-400. This model is based on the aerodynamic profile NLF-MOD22B, designed by the Delft University
of Technology. In section 2.2 a brief description is given about the wind tunnel and the model. This
is followed by an explanation and some preliminary calculations for the applied passive lift devices in
section 2.3. Finally section 2.4 gives an overview of the applied experimental techniques and measurement
systems including an explanation on the required data.

2.2 The wind tunnel and model

The wind tunnel

The wind tunnel used is the Low Speed Low Turbulence Wind Tunnel of Delft University of Technology.
It is a closed vertical arranged circuit wind tunnel, operating at tunnel speeds of 0 − 120 m/s with
turbulence levels below 0.05%. The test section has dimensions of 1.80 × 1.25 m and a length of 2.4 m.
An electric motor drives the flow through the diffuser towards the settling chamber, after which the air is
accelerated towards the test section. Rotatable side plates are installed in the test section to give freedom
in adjusting the angle of attack. Figure 2.1 shows the complete layout of the wind tunnel.

The model

The NLF-MOD22B model used for the experimental and numerical research is an airfoil with a single
slotted trailing edge flap system. The NLF-MOD22B airfoil flap system is schematically depicted in figure
2.2 for its retracted and deployed state respectively, the latter also showing the geometric details of the
flap setting.

15
Chapter 2. Experimental model setup

Figure 2.1 – The Low Speed Low Turbulence Wind Tunnel at Delft University of Technology.

Figure 2.3 shows the composite wind tunnel model, which was installed in a vertical position between the
two rotatable circular endplates. The model has a chord of 0.6 m and span of 1.25 m. The 30% chord flap
can be deployed through a translation and rotation of the turntables positioned at the flap endplates.

From earlier research on the NLF-MOD22B flap system performed by Boermans and Rutten [1995] it
was found that artificial tripping of the boundary layer on the airfoil lower surface was necessary in order
to eliminate a detrimental laminar separation bubble on the flap lower surface. This was also observed
in preliminary wind tunnel tests of this research. For this reason zigzag tape with a thickness of 0.5 mm
is placed at 58% chord length. The thickness of the zigzag tape was determined by the local height of
the boundary layer. The boundary layer thickness was measured through a vertical pitot tube traverse.
Figure 2.4 shows the performed measurements including the boundary layer state before and after the
application of the zigzag tape.

Figure 2.2 – Cross section of the model and definition of flap parameters.

D.P. Jansen 16 MSc. Thesis


2.2. The wind tunnel and model

Figure 2.3 – The model positioned in the wind tunnel with the cylinder test setup.

MSc. Thesis 17 D.P. Jansen


Chapter 2. Experimental model setup

4
y [mm]

1 α = 0◦
α = 12◦
α = 0◦ zigzag at x/c = 0.58
α = 12◦ zigzag at x/c = 0.58
0
0 300 600 900 1200 1500
p tot

Figure 2.4 – Boundary layer measurements on model lower side at x/c = 0.67 .

It should be noted that the presence of the wind tunnel walls clearly affects the 2D state of the flow
through the possibility of locally early separation. This was avoided by applying boundary layer suction
at the rotatable endplates through 5 mm diameter holes, spaced 10 mm apart and from the main wing
respectively, and through 4 mm diameter holes, spaced 8 mm apart and from the flap. These suction
orifices are displayed in the picture of figure B.4 in appendix B. Tufts were applied to the model in
preliminary experiments to test the necessary suction level to prevent the occurrence of flow separation
at the wing/wall junction. In these preliminary tests suction level was increased up to the point where
tufts near the wing/wall junction showed steady behaviour with an orientation in line with the general
flow direction.
The flap is positioned at a fixed overlap and gap size of respectively 0.0c and 0.035c, which produces
the highest lift according to measurements from Boermans and Rutten [1995]. In order to ensure flow
separation on the flap to test the vortex generating devices on their performance, the flap needs to
be positioned at a large deflection. This means that it is necessary to investigate at which deflection
angle flow separation starts to occur. Boermans and Rutten [1995] performed studies on the pressure
distributions for several flap deflections and concluded that flow separation starts to occur at the flap for
a flap deflection of 40◦ . From these results the flap deflections at which the model will be tested are set
at 45◦ , 50◦ and 55◦ .

2.3 Passive flow separation methods

The airfoil flap configuration model as discussed in section 2.2 will serve as the benchmark setup for the
wind tunnel investigation. This section will describe the passive separation control devices tested in this
study.

D.P. Jansen 18 MSc. Thesis


2.3. Passive flow separation methods

Figure 2.5 – The model with VG’s applied (left) and cylinders (right).

2.3.1 Vortex generators

The vortex generators were tested at the lower side of the airfoil near the local trailing edge, as depicted
in figure 2.5. While VG’s are more often found on upper surfaces of wings, downstream of the leading
edge, the idea here is to position the devices upstream of the flap leading edge. As discussed later in this
section, vortices need a specific distance to develop and it takes time for the flow to mix with the entrained
high-energy flow. There is a certain distance from the separation point at which vortex generators become
effective. If positioned on the upper side of the airfoil it becomes difficult to reach the flap boundary
layer with vortices. Positioning on the flap itself is difficult because of the short distance with respect
to the point of separation. From the lower side of the airfoil this can be done more effectively since the
vortices have enough time develop. This makes the positioning of VG’s at the end of the lower side of
the airfoil an interesting study case.
Figure 2.5 depicts the conceptual idea that streamlines that come from the lower side of the airfoil and
continue close over the flap upper surface are entrained with a vortex shedding to delay separation over
the flap. Correct dimensioning and positioning of the VG’s should give an optimal configuration, at which
the necessary streamlines are entrained with vortex shedding.
As explained in section 2.3.1 different types of vortex generators exist. For this research only the counter
rotating VG’s were tested at several sizes. Figure 2.6 shows the geometrical definitions of VG’s, which
include:

• λ for the mutual spacing of the vortex generators. Here λ is defined as the distance between the
VG pairs.

• h to define the height of the vortex generator.

• d for the distance between the VG’s and the bottom trailing edge of the NLF-MOD22B model.

The vortex generators were all tested on different values for these parameters. For the remaining param-
eters in figure 2.6, fixed values were chosen of β = 18◦ and s = h respectively, for which van der Steen
[2009] found the best results.
The height of the vortex generators was difficult to determine. In van der Steen [2009] VG heights ranging
from h = 13 δ (micro VG’s) to h = δ proved to be successful, however these devices were applied for quite
a different flow situation, an inclined flat plate. Furthermore micro VG’s should be positioned close to
the point of separation to be effective, which is unrealizable in this case when the VG’s are positioned
at the lower end of the airfoil. It was expected that slightly larger VG’s were necessary in order for the
vortices to remain their strength through the slot and be effective over the flap. Figure 2.4 shows that
the boundary layer height at location of the VG’s is 6 mm after the transition tape is applied. From this

MSc. Thesis 19 D.P. Jansen


Chapter 2. Experimental model setup

Figure 2.6 – Vortex generator geometrical definitions.

Figure 2.7 – Vortex generator geometry and construction method.

result VG heights were chosen of 6 mm, 14 mm and 28 mm, which were tested at multiple spacing’s λ.
The complete test matrix for the VG’s is given in table A.1 in appendix A. Note that in the test matrix
and in shown results in following sections the non-dimensionalized parameters h∗ and d∗ are used. Here
h∗ and d∗ are defined as h∗ = h/c and d∗ = d/c.
The VG’s were produced from a 0.1 mm aluminium sheet. A construction method was applied similar to
van der Steen [2009]. Small rectangular plates were bend over their diagonal until the two created planes
are perpendicular, creating the delta shaped vortex generator shown in figure 2.7. The devices were
connected to the model through double sided tape, which made it easy to remove and test for different
configurations.

2.3.2 Cylinders
The cylinders were tested in the slot just in front of the trailing edge flap. Similar to the vortex generators,
the concept is to entrain the streamlines close over the flap surface with a vortex shedding developing
in the slot. This is depicted in figure 2.5. As well as for the vortex generators positioning and size
are of key importance for the cylinder performance. Figures 2.8 and 2.9 show the parameter definition
for the positioning and size of the cylinders. The positioning is expressed in the cylindrical coordinates
r and θ. The diameter is specified as D. The origin is chosen to be at the hinge of the supports, at
[x, y] = [0.72c, 0.06c] measured from the airfoil nose; The connection system will be explained later in this
section. The area of interest is also depicted in figure 2.8, which resulted in multiple selections of r, θ and
D to test the cylinders. The complete test matrix for the cylinders is given in table A.2 in appendix A.
Note that in the test matrix and in shown results in following sections the non-dimensionalized parameters
r∗ and D∗ are used. Here r∗ and D∗ are defined as r∗ = r/c and D∗ = D/c.
It must be noted that the circulation of the Von Kármán vortex street is lateral to the flow direction
instead of lengthwise, which is the case for the VG devices. Unlike for the vortex generators some
theoretical basic equations exist in order to estimate the correct cylinder size beforehand. The shedding
frequency of a vortex street behind a circular object can be determined with the following equation:

D.P. Jansen 20 MSc. Thesis


2.3. Passive flow separation methods

Figure 2.8 – Cylinder position and definition of coordinates. Hinge point is referenced to the airfoil nose
with [x, y] = [0.72c, 0.06c].

Figure 2.9 – Cylinder geometrical definitions.

MSc. Thesis 21 D.P. Jansen


Chapter 2. Experimental model setup

D [mm] F+
Cylinder I 10 3.6
Cylinder II 15 2.4
Cylinder III 20 1.8

Table 2.1 – Selected cylinder sizes with corresponding F + .

St · U∞
f= (2.1)
D

Here St is the Strouhal number, a dimensionless number that describes oscillating flow mechanisms, and
U∞ is the free stream velocity of the fluid. The Strouhal number for cylinders is 0.2 for Reynolds numbers
from 100 to 105 as given by White [2006].
It should be noted that equation 2.1 applies for free stream flow. In the vicinity of a wall, research shows
that the shedding of vortices by the cylinder is affected by the distance from the wall S, which is clarified
in figure 2.9. This was investigated by Wang and Tan [2007], which tested a cylinder wake close to a fully
developed turbulent boundary layer with thickness δ = 0.4D. The research showed that vortex shedding
takes place as in free stream when S/D > 0.8. For S/D < 0.8 the vortex shedding is asymmetrical
with stretched and underdeveloped vortices from the near wall side. At ratios of S/D < 0.3 the vortex
shedding can even be completely suppressed. The experiment further showed the Strouhal number
remains constant at about 0.2, but that the state of the turbulent boundary layer could significantly
affect flow characteristics at different S/D.
Besides taking into account the influence of the wall, or in this specific investigation the influence of the
slot boundaries and flap, the question remains at which shedding frequency the momentum exchange
in the boundary layer is optimal. Nishri and Wygnanski [1998] and Greenblatt and Wygnanski [2000]
performed studies on flow separation control through periodic excitation and identified that the vortex
frequency in a different form is more useful. They introduced the reduced frequency F + :

f Lf
F+ = (2.2)
U∞

This expression is a dimensionless form of the vortex frequency f . In equation 2.2 Lf is by definition the
length of the object over which flow separation must be delayed. Combining equation 2.2 with 2.1 leads
to the interesting result:

Lf
F + = St (2.3)
D

This means that the reduced frequency is independent of velocity, considering St is independent of
Reynolds number between 100 and 105 . Greenblatt and Wygnanski [2000] found an optimum reduced
frequency for forcing a separated flow to reattach at F + = 1. For maintaining attached flow the forcing
frequency should be between 3 < F + < 4. These results were used as a basis for estimating the cylinder
sizes that have been tested. When for Lf the length of the flap is chosen, the reduced frequency can
be plotted as a function of the cylinder diameter, see figure 2.10. From a theoretical point of view it
is preferable to test a cylinder size corresponding to F + = 1, but this would result in cylinder sizes of
40 mm and is considered impractical given the limited size of the slot. Cylinder diameters of 10 mm,
15 mm and 20 mm were selected, which have reduced frequencies in the range of 1.8 < F + < 3.6. The
diameters and corresponding reduced frequencies are summarized in table 2.1.
The cylinders were mounted at the required position through four fixed supports with adjustable can-
tilevers. The cylinders crossed the total span of the model so that the concept of 2D flow is maintained.

D.P. Jansen 22 MSc. Thesis


2.4. Measurement techniques

3
F+

0
0 10 20 30 40 50
D

Figure 2.10 – Reduced frequency dependency on cylinder diameter.

Positioning of the construction was done with care. The brackets need to be oriented far enough from
the pressure orifices to avoid flow interference effects, keeping in mind that a too large support spacing
may cause insufficient bending stiffness for the construction and possibly undesirable vibrations of the
cylinders. The cantilevers were able to rotate and, through the multiple positioning holes, have freedom of
movement in horizontal and vertical direction. In appendix B pictures are provided to see the mechanism
in more detail.

2.4 Measurement techniques

There are several experimental methods to determine airfoil flap configuration performance and to analyse
the effect of vortex generating devices, of which the latter is particularly important for this study. For
this research pressure measurements, oil flow visualization and frequency measurements were used. The
next section will shortly explain and discuss these techniques.

2.4.1 Pressure measurements and calculation of Cl and Cd

Multiple pressure measurement methods were applied which can give direct insight in local flow properties,
like flow transition and flow separation. The pressure measurements can also be used to calculate certain
performance parameters. These parameters are the dimensionless coefficients Cl and Cd , which represent
the lift and drag. All the obtained pressure quantities were measured by tubing pressure orifices to a
digital multi-manometer from Pressure Systems, which is able to directly transform the pressure offsets
to discrete values.
On the upper and bottom side of both the airfoil and flap a total of 85 pressure orifices were located,
where for each pressure hole measurements are performed at a frequency of 330 Hz in an interval of
4.0 s. The pressure holes are located on the model in a zigzag position that will prevent the pressure
holes to locally disturb each other, see figure B.5 in appendix B. The pressure measurements can be used
to accurately calculate the lift in the following way. First the measured static pressures at the pressure
orifices are translated to dimensionless pressure coefficients through:

MSc. Thesis 23 D.P. Jansen


Chapter 2. Experimental model setup

p − p∞
Cp = (2.4)
q∞

Here p is the local static pressure at the orifice and p∞ and q∞ are respectively the far field total and
dynamic pressure. The latter two are obtained by measurements with a pitot static tube. This tube is
positioned in the plane of the mid-span section on the tunnel sidewall opposite to the model lower surface
and about two chord lengths in front of the leading edge. This is visualized in the picture of figure B.2.
The pressure distribution is integrated through the trapezium rule to obtain a normal coefficient for both
the airfoil and flap with:

T Eˆ
airf oil
! " x
Cnairf oil = Cplower − Cpupper d (2.5)
c
LEairf oil

TE
ˆf lap
! " x
Cnf lap = Cplower − Cpupper d (2.6)
c
LEf lap

This gives a total normal coefficient of:

Cn = Cnairf oil + Cnf lap (2.7)

This can be used to calculate the lift using:

Cl = Cn /cos (α) − Cd · tan (α) (2.8)

To calculate the lift coefficient from equation 2.8, a drag coefficient is necessary. For the drag calculation
another pressure measurement approach is used, the so-called wake-traverse method. A total pressure
and static pressure wake rake is positioned behind the model. The dimensions of the wake rake are
shown in figure 2.11. The positioning of the wake rake in the tunnel is seen in the picture of figure B.2
in appendix B. The pitot tubes are positioned horizontally, although ideally they should be aligned with
the local flow behind the airfoil flap configuration that is deflected downwards by the downwash of the
wing. However it was very unpractical to adjust the orientation of the wake rake for each observation.
Still the total pressure tubes are allowed to have an angle misalignment of 40◦ which was assumed to be
sufficient for the measurements. The wake rake pressure measurements are used to determine the profile
drag of the model using the method proposed by Jones [1936]. The equation used in this method to
determine the drag is given by:

ˆ ò 3 ò 4 1 2
pt − p pt − p∞ y
Cd = 2 1− d (2.9)
q∞ q∞ c

This gives the necessary information to calculate the lift coefficient.


An important issue remains in the computation of the previously mentioned performance parameters.
Since the wind tunnel experiment needs to provide 2D results, some corrections need to be applied for
the presence of the tunnel wall. Apart from the local influence on the boundary layer flow separation as

D.P. Jansen 24 MSc. Thesis


2.4. Measurement techniques

(a) Front view.

(b) Top view.

Figure 2.11 – Test setup of the wake rake.

discussed in section 2.2, the wall also has two other effects, namely solid blockage and wake blockage.
The solid blockage effect is caused by an increase in the free stream velocity attributed to the volume
distribution of the body. The wake blockage effect is a serious constraint on the growth of the wake. For
streamlined bodies at small angles these effects are small, but especially for the wake blockage become of
significant performance for bluff bodies. This includes airfoil flap configurations at high flap deflections
and high angle of attack. Allen and Vincenti [1944] proposed a correction model for this situation, which
is based on a wall representation of image doublets and sources. It is applied to the pressure distribution
computation and also for the lift and drag coefficient calculation in order to get the required 2D results.

All measured (un)corrected pressures and tunnel conditions were collected by Labview from which the
aerodynamic coefficients are calculated in Profmeasure. Appendix C shows the output file from Profmea-
sure for such a single measurement, which is then read into MATLAB for further postprocessing.

2.4.2 Frequency measurements

A mapping of the frequency was created of the produced vortices aft of the cylinder. A listening tube was
connected to an external storage system and gave the possibility to analyse results on screen in Labview.
This gave a recorded sampled signal of the pressure fluctuation ∆ps against time in the vicinity of the
cylinder. This however is a noisy signal, which is difficult to analyse for measured frequencies. Therefore
the signal was transformed through a Fast Fourier Transform (FFT) in MATLAB to the frequency
domain. The FFT transformation applied here is given by the following equation:

N
Ø j−1
∆P (k) = ∆ps (j) e2πi(k−1) N (2.10)
j=1

The measurements were performed at multiple locations to gain insight on the development of a possible
vortex street. These locations are visualized in figure 3.22.

MSc. Thesis 25 D.P. Jansen


Chapter 2. Experimental model setup

2.4.3 Oil flow visualization

Oil visualization gives direct qualitative insight in the characteristics of the flow. It can confirm observa-
tions made from analyzing data of quantitative measurements. When applied properly it can be utilized
for detecting both transition as well as flow separation.
For the visualization fluid a mixture of oil with a fluorescent dye was prepared. For the respectively
low tunnel speed, a thinner was added to make sure that the fluid viscosity was not too high and that
the oil would settle quickly for each measurement. A UV light was used for enhanced visibility with
photography. The images were recorded with a Nikon D80 CCD camera at high aperture to ensure
sufficient light exposure.

D.P. Jansen 26 MSc. Thesis


CHAPTER 3

Experimental results

3.1 Introduction

This chapter will give an overview of the results obtained from the experimental investigation. Initially
the NLF-MOD22B airfoil flap system was tested for its baseline configuration. From the experiments
performed by Boermans and Rutten [1995] it was identified that flow separation occurs at the flap at a
deflection of 40◦ . Based on this result, the choice was made to test the flow separation delay capabilities
of the passive lift devices on the model for flap angles of respectively 45◦ , 50◦ and 55◦ . All tests in
the experimental investigation are performed on a tunnel speed of 50 m/s corresponding to a Reynolds
number of Re = 2.0 · 106 . An overview of the results from the baseline run tests is presented in section
3.2. Then section 3.3 and 3.4 will respectively show the effects of the performance of the passive high
lift devices with respect to the baseline run results at the selected flap deflections. Finally in section 3.5
conclusions will be drawn based on the experimental investigation.

3.2 Baseline model

The baseline model is defined as being solely the airfoil and flap with one of the three deflections 45◦ , 50◦
and 55◦ . As discussed in section 2.4 the wind tunnel experiments consisted of pressure measurements, oil
flow visualization and frequency measurements. In this section the pressure measurements and resulting
lift polars of the baseline models will be discussed to understand the effects of the angle of attack and the
three flap deflections on the model performance. Discussion of the oil flow visualization and frequency
measurements for the baseline models is done in section 3.4 where the baseline is directly compared to
the similar tests done for the cylinder configuration.

Baseline lift polars

In figure 3.1 the lift polars are shown for the three tested flap deflections. In general the lift coefficient
shows similar behaviour for all three cases. For all cases the angle of attack is increased from −6◦ to

27
Chapter 3. Experimental results

13◦ . Figure 3.1 shows that the lift increases with the angle of attack until the stall angle is reached and
abruptly drops. It is clear that the lift curve for δf = 45◦ shows the largest obtainable lift at all angles
of attack with a maximum of Cl = 3.23 at α = 12◦ . Although at lower angles of attack the flap with
δf = 50◦ can show comparable lift performance, it is noticeable that the lift starts to stagnate in this
case at α = 8◦ . This is an unexpected result for which a supplementary run was performed. However, no
differences in the lift development between the subsequent runs were measured. At a flap deflection of
δf = 55◦ it is clear that the lift is significantly lower over the complete range of angle of attack. Further
results will show that at flap angles larger than δf = 45◦ the lift becomes highly affected by the onset of
flow separation.
A remark should be made on the tests performed for the flap deflection of 55◦ . Early wind tunnel tests
revealed that the a.o.a. had to be lowered first to negative angles in order to have a semi-attached flow
on the flap. This meant that if the model was directly set at α = 0◦ at the start of the run a completely
separated flow on the flap was observed. It is a rather peculiar phenomenon and is also reported by
Baragona [2004] and Biber [1995] on other single slotted airfoil flap researches. This insinuates the
presence of a stall hysteresis caused by laminar bubble bursting on the flap. The observed stall hysteresis
was kept in mind, but no further research was done on the hysteresis by means of stall recovery tests. The
laminar bubble bursting may also be the cause the stagnation in lift found at high angle of attack for the
flap deflection of 50◦ . In fact Baragona [2004] discovered in the same setup at which hysteresis for α = 0◦
was observed, changes in the slot geometry could cause laminar bubble bursting and accompanying lift
stagnations at high angle of attack. The tests performed by Baragona [2004] and also Biber [1995] seem
to resemble the early observations in the lift polars of this wind tunnel research for the NLF-MOD22B
airfoil flap.
The next section will further elaborate on the observations made from the lift polars by discussing the
accompanying pressure distributions.

Baseline pressure distributions

Figure 3.2 shows the pressure distributions for the baseline configuration at flap deflections 45◦ , 50◦ and
55◦ for multiple angles of attack. Overall similar behaviour can be observed in the pressure distribution
development. For all deflections a large negative pressure peak develops on the leading edge of the main
airfoil, while the pressure distributions on the flap show more variety in behaviour. Focusing on the flap
pressures it is clear that flow separation on the flap upper surface is visible at all deflections. Especially
at 55◦ the sudden increase in pressure about halfway the flap chord and the high steady pressure toward
the trailing edge, clearly indicates the onset of flow separation. The flow separation is caused by the very
high adverse pressure gradient the flow needs to overcome, due to the very low peak pressures at the flap
leading edge. Similar behaviour is found at 50◦ and 45◦ though the adverse pressure gradients are less
severe and the point of separation shifts more towards the trailing edge.
For the dependency on the angle of attack the following observations can be made. With increasing angle
of attack, the stagnation point on the airfoil moves rearward along the lower surface and the suction peak
at the nose of the airfoils upper surface develops significantly resulting in a lower pressure at the upper
surface and a higher lift Clairf oil for the main element. From figure 3.1 it is visible that stall is reached
at an angle of αs = 13◦ for all the tested flap deflections. Figures 3.2a - 3.2c show that at stall, flow
separation is now found on the airfoil due to the severe adverse pressure gradient at this high angle of
attack. The stall on the airfoil is of a leading edge type, due to its sudden appearance and the onset of
separation close to the leading edge.
In general the pressure distributions on the flap show less severe changes when the angle of attack is
increased. Still, a tendency can be observed that for higher incidence angles the pressure on the upper
flap surface in fact increases and smaller lift coefficients Clf lap are created. This effect is more often
found for airfoil flap configurations for example by Boermans [1998] on an experimental research for a
NACA 63-415 airfoil with slotted flap. At higher angle of attack the pressure distribution on the flap

D.P. Jansen 28 MSc. Thesis


3.2. Baseline model

3.5
Baseline, δf = 55◦
Baseline, δf = 50◦
Baseline, δf = 45◦
3

2.5
Cl

1.5

1
−10 −5 0 5 10 15
α [◦ ]

(a) The baseline lift polars.

0.1

0
y/c

−0.1

−0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x/c

(b) The tested flap deflections.

Figure 3.1 – Lift polars for the baseline configurations at multiple flap deflections at Re = 2.0 · 106 .

MSc. Thesis 29 D.P. Jansen


Chapter 3. Experimental results

−8
Baseline, α = 0◦
Baseline, α = 6◦
−7
Baseline, α = 12◦
Baseline, α = 13◦
−6

−5

−4

Cp
−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

(a) δf = 55◦ .

−8 −8
Baseline, α = 0◦ Baseline, α = 0◦
Baseline, α = 6◦ Baseline, α = 6◦
−7 −7
Baseline, α = 12◦ Baseline, α = 12◦
Baseline, α = 13◦ Baseline, α = 13◦
−6 −6

−5 −5

−4 −4
Cp

Cp

−3 −3

−2 −2

−1 −1

0 0

1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c x/c

(b) δf = 50◦ (c) δf = 45◦ .

Figure 3.2 – Pressure distributions for baseline configurations at multiple a.o.a for δf = 55◦ , δf = 50◦ and
δf = 45◦ .

is relieved which also favours the delay of flow separation. This is due to the fact that the upper flap
pressure distribution is suppressed by the displacement effect of the main airfoil large wake above the
flap. This means that it may be possible that if the airfoil flow separation could be delayed and a higher
stall angle could be reached, for example by applying a leading edge flap, the separation on the flap at
these high incidence angles can be removed. In fact figures 3.2a - 3.2c show that at stall, the flow on the
flap is completely attached, indicating the aforementioned potential.
Focusing on figure 3.2b partly explains the lift drop at high a.o.a. observed in figure 3.1. It reveals that
at α = 12◦ the flow over the flap is largely separated resulting in a low Clf lap . This also has its effect
on the airfoil pressure distribution especially at its trailing edge, which is probably caused by a smaller
circulation effect of the flap on the main element. This at least explains that the root cause of the lift
stagnation at high angle of attack is found in the flap performance. As hypothesized in the previous
section it could be that a laminar bubble burst on the flap has caused this. Section 3.4.2 will provide
additional insight through the performed oil flow visualizations.

D.P. Jansen 30 MSc. Thesis


3.3. Vortex generators

3.3 Vortex generators

This section will describe the wind tunnel results for the airfoil flap model tested with vortex generators.
The vortex generators were only tested for a flap deflection of 55◦ . The performance will be addressed
by observing the lift polars and comparing them to the baseline situation. This is followed by a closer
inspection on the pressure distributions. The complete test matrix for the vortex generators is presented
in appendix A.

VG lift polars

Figure 3.3 shows the lift performance of two sets of multiple VG configurations presented in several lift
polars. Equivalent to the baseline tests the angle of attack is increased from −6◦ to 13◦ at which the lift
was measured. This time the model did not necessarily needed to be lowered first to negative angles to
have semi-attached flow on the flap at α = 0◦ , as was the case for the baseline model at δf = 55◦ .
As is covered in section 2.3.1 several VG configurations have been selected and tested. It can be seen
from the figures that the VG configurations all show an increase in lift at lower angles of attack but
generally have a decrease in lift near the stall angle. It should be noted that the main area of interest for
a lift increase is more around α = 0◦ and not at Clmax . This is because the model is tested with a trailing
edge flap only, hence not in combination with a leading edge slat which gives more performance at higher
angle of attack. From figure 3.3 it can be seen that VG’s with smaller and larger heights (h∗ = 0.010
and h∗ = 0.070) show only minor effect in lift improvements. The smaller height, h∗ = 0.010, results
vortices that are too small and therefore insufficient in mixing enough high-energy flow to the boundary
layer at the flap. The larger VG’s, h∗ = 0.070, are probably causing adverse vortex interaction since they
lead to very fluctuating lift values also visible in the lift polar. The spacing also shows an effect where
values should not either be too large or too small for an optimal vortex performance. VG’s at a spacing
of λ/h = 6.5 and λ/h = 4.5 do not result in a significant lift improvement. This is probably due to a
spacing that is too large and where an insufficient amount of healthy air is pumped into the boundary
layer. Only the VG configuration with spacing λ/h = 4.0 and height h∗ = 0.023 shows a significant
performance gain over the whole range of α, this setup is found as the optimum VG configuration. In
this case a gain is found at α = 0◦ with lift increases up to ∆Clo = 0.11 which is an improvement of 6.0%
compared to the baseline. At high angles improvements are obtained of ∆Clmax = 0.064 an improvement
of 2.2%. It is clear for all configurations that stall is not delayed, although this is not surprising since
the stall for this airfoil flap model is mainly determined by the airfoil stall characteristics and not the
performance of the flap.
The total lift can be subdivided into the contributions of both the flap and airfoil, Clf lap and Clairf oil .
Figure 3.4a shows that the lift gain at low incidence angles is initially both on the flap and airfoil. However
as the angle of attack increases the lift gain for the airfoil decreases while the contribution from the flap
remains fairly constant. Another visualization can be made of the airfoil and flap lift contribution by
plotting the ratio Clairf oil /Clf lap versus the a.o.a. Figure 3.4b indicates that compared to the baseline
configuration, the lift at higher angles of attack has a larger contribution from the flap lift component.

VG pressure distributions

Figure 3.6 shows the pressure distributions for the found optimum VG configuration at multiple angles of
attack. A very comparable pressure distribution development to figure 3.2 is seen for the different angles
of attack. Comparison of the flap pressure distributions of the baseline and VG configurations shows that
the tendency of a higher flap pressure at larger angle of attack for the baseline model is not found for the
airfoil flap model with VG’s. In figure 3.6 it is visible that the flap pressure remains fairly constant for
increasing angle of attack. This also explains the result found from figure 3.4b where at higher angle of
attack a larger contribution to the total lift is found from the flap lift component.

MSc. Thesis 31 D.P. Jansen


Chapter 3. Experimental results

3.5 3.5
Baseline Baseline
VG, λ/h = 6.5 VG, λ/h = 6.5, h∗ = 0.010, d∗ = 0
VG, λ/h = 4.5 VG, λ/h = 4.5, h∗ = 0.010, d∗ = 0
VG, λ/h = 4.0 VG, λ/h = 4.0, h∗ = 0.070, d∗ = 0.13
3 3

2.5 2.5
Cl

Cl

2 2

1.5 1.5

1 1
−10 −5 0 5 10 15 −10 −5 0 5 10 15
α[ ] ◦
α [◦ ]

(a) VG’s set 1 with h∗ = 0.023 and d∗ = 0. (b) VG’s set 2.

0.1 0.1

0 0
y/c

y/c

−0.1 −0.1

−0.2 −0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/c x/c

(c) VG positions set 1. (d) VG positions set 2.

Figure 3.3 – Effect of height, spacing and edge distance on the lift for multiple VG configurations at δf =
55◦ .

D.P. Jansen 32 MSc. Thesis


3.3. Vortex generators

3.5
Baseline
VG, λ/h = 6.5
VG, λ/h = 4.5
VG, λ/h = 4.0
3

2.5
15
Baseline
VG, λ/h = 6.5
C lf lap , C lairf o il

2
VG, λ/h = 4.5
VG, λ/h = 4.0
12

1.5

C lairf oil/C lf lap


9

1
6

0.5
3

0
−10 −5 0 5 10 15 0
α[◦ ] −10 −5 0 5 10 15
α[◦ ]

(a) Clf lap and Clairf oil versus α. (b) Clairf oil and Clf lap versus α.

Figure 3.4 – Flap and airfoil lift contributions for multiple VG configurations with h∗ = 0.023, d∗ = 0 and
δf = 55◦ .

−8
VG, α = 0◦
−7
VG, α = 6◦
VG, α = 12◦
VG, α = 13◦
−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 3.5 – Pressure distribution for optimum VG configuration at multiple a.o.a. with λ/h = 4.0, h∗ =
0.023, d∗ = 0 and δf = 55◦ .

MSc. Thesis 33 D.P. Jansen


Chapter 3. Experimental results

−8 −8
Baseline Baseline
VG, λ/h = 4.0, h∗ = 0.023, d∗ = 0 VG, λ/h = 4.0, h∗ = 0.023, d∗ = 0
−7 −7

−6 −6

−5 −5

−4 −4
Cp

Cp
−3 −3

−2 −2

−1 −1

0 0

1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c x/c

(a) α = 0◦ . (b) α = 12◦ .

Figure 3.6 – Pressure distributions for optimum VG and baseline configuration at low and high a.o.a. with
λ/h = 4.0, h∗ = 0.023, d∗ = 0 and δf = 55.

Figure 3.6 shows the pressure distribution of the optimum VG configuration compared with the baseline
model at α = 0◦ and α = 12◦ . It can be seen for both situations that the optimum VG configuration has
a slightly different pressure distribution on the flap compared to the baseline model. The pressure at the
nose is slightly higher, although the pressure over the rest of the flap chord is lower. This is because the
sudden pressure rise is shifted more towards the trailing edge, which indicates that flow separation seems
to be delayed. At zero angle of attack the upper surface pressure distribution of the airfoil is lowered,
possibly due to an increased circulation effect. This effect is more visible at lower angle of attack than
at α = 12◦ .

3.4 Cylinders
This section will describe the wind tunnel results for the airfoil flap model tested with vortex generating
cylinders. The cylinders were tested for flap deflections of 55◦ , 50◦ and 45◦ . The performance of the
cylinders will be addressed by observing the lift polars and comparing them to the baseline situation.
This is followed by a closer inspection on the pressure distributions and performed oil flow visualizations.
To better understand the working principles of the cylinder vortices, microphone frequency measurements
and transverse total pressure measurements were executed at a flap deflection of 45◦ . These results will
also be presented. The complete test matrix for the vortex generators is given in appendix A.

3.4.1 Flap deflection δf = 55◦

Cylinder lift polars for δf = 55◦

Figure 3.7 and 3.8 show the lift performance of several cylinder configurations in a lift polar at a flap
deflection of δf = 55◦ . Similar to the VG tests the model did not need to be lowered first to negative angles
to have semi-attached flow on the flap at α = 0◦ . It can be seen that multiple cylinder configurations
show significant improvement over the whole range of α in lift. Cylinders with varying diameters all show
a performance gain when positioned properly, hence in the range of r∗ = 0.06 ↔ 0.09 and θ = 30◦ ↔ 37◦ .
The optimum configuration is found at a D∗ = 0.0167, r∗ = 0.090 and θ = 37◦ . An interesting graph is
illustrated in figure 3.9 where the lift gains ∆Cl are visualized against corresponding reduced frequencies

D.P. Jansen 34 MSc. Thesis


3.4. Cylinders

3.5 3.5
Baseline Baseline
Cylinder, r ∗ = 0.060, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.060, θ = 55◦
Cylinder, r ∗ = 0.090, θ = 30◦ Cylinder, D ∗ = 0.0167, r∗ = 0.060, θ = 18◦
Cylinder, r ∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0333, r∗ = 0.090, θ = 37◦
3 3

2.5 2.5
Cl

Cl
2 2

1.5 1.5

1 1
−10 −5 0 5 10 15 −10 −5 0 5 10 15
α[ ] ◦
α [◦ ]

(a) Cylinders set 1 with D∗ = 0.0167. (b) Cylinders set 2.

0.1 0.1

0 0
y/c

y/c

−0.1 −0.1

−0.2 −0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/c x/c

(c) Cylinder positions set 1. (d) Cylinder positions set 2.

Figure 3.7 – Effect of size and positioning on the lift for multiple cylinder configurations at δf = 55◦ .

F + . The reduced frequency with the highest lift gain is determined in the experiment as F + = 3.6, which
is some what higher than the optimum value of F + = 1 found by Greenblatt and Wygnanski [2000]. The
frequency corresponds to an optimum cylinder diameter of D∗ = 0.0167. For this configuration the largest
lift gain is found near zero angle of attack, the main area of interest, with lift increases up to ∆Cl0 = 0.32
which is an improvement of 18% compared to the baseline. At high angles improvements are obtained of
∆Clmax = 0.24. A trend is visible in figure 3.9 that as frequency is reduced (larger cylinders) lift gains
become smaller.
The total lift can be subdivided into the contributions of both the flap and airfoil, Clf lap and Clairf oil . In
figure 3.10a the lift contributions of three configurations including the optimum, D∗ = 0.0167, r∗ = 0.090
and θ = 37◦ , are plotted. It shows comparable trends to the earlier found vortex generator airfoil and
flap lift contributions. For increasing angle of attack the cylinder lift gain for the airfoil decreases. A
similar effect is observed for the flap lift, which at this point differs slightly with the optimum VG
configuration that showed a reasonably independent flap lift to the angle of attack. As for the optimum
VG configuration, it can be seen from figure 3.10b that the contribution of the flap to the total lift
performance is increased at higher angle of attack when compared to the baseline configuration. Similar
to the case of the vortex generators the cylinders do not delay the stall of the airfoil flap model.

Cylinder pressure distributions for δf = 55◦

Figure 3.11 shows the pressure distribution for the optimum cylinder configuration compared to the
baseline model at α = 0◦ and α = 12◦ . It shows at both angles of attack that the pressure peak at

MSc. Thesis 35 D.P. Jansen


Chapter 3. Experimental results

3.5
Baseline
Cylinder, D ∗ = 0.025, r∗ = 0.090, θ = 30◦
Cylinder, D ∗ = 0.025, r∗ = 0.105, θ = 30◦
Cylinder, D ∗ = 0.025, r∗ = 0.045, θ = 42◦
3

2.5
Cl

1.5

1
−10 −5 0 5 10 15
α [◦ ]

(a) Cylinders set 3 with D∗ = 0.025.

0.1

0
y/c

−0.1

−0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x/c

(b) Cylinder positions set 3.

Figure 3.8 – Effect of size and positioning on the lift for multiple cylinder configurations at δf = 55◦ .

0.35

0.3

0.25

0.2
∆ Cl

0.15

0.1

0.05
α = 0◦
α = 12◦
0
0 1 2 3 4 5
F+

Figure 3.9 – Lift gains ∆Cl dependency against reduced frequency F + at high and low angle of attack
with r∗ = 0.090, θ = 37◦ and δf = 55◦ .

D.P. Jansen 36 MSc. Thesis


3.4. Cylinders

3.5
Baseline
Cylinder, r ∗ = 0.060, θ = 37◦
Cylinder, r ∗ = 0.090, θ = 30◦
Cylinder, r ∗ = 0.090, θ = 37◦
3

2.5
15
Baseline
Cylinder, r ∗ = 0.060, θ = 37◦
C lf lap , C lairf o il

Cylinder, r ∗ = 0.090, θ = 30◦


2
Cylinder, r ∗ = 0.090, θ = 37◦
12

1.5
C lairf oil/C lf lap

1
6

0.5
3

0
−10 −5 0 5 10 15 0
α[◦ ] −10 −5 0 5 10 15
α[◦ ]

(a) Clairf oil and Clf lap versus α. (b) Clairf oil /Clf lap versus α.

0.1 0.1

0 0
y/c

y/c

−0.1 −0.1

−0.2 −0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/c x/c

(c) Cylinder positions. (d) Cylinder positions.

Figure 3.10 – Flap and airfoil lift contributions for multiple cylinder configurations with D∗ = 0.0167 and
δf = 55◦ .

MSc. Thesis 37 D.P. Jansen


Chapter 3. Experimental results

−8 −8
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦
−7 −7

−6 −6

−5 −5

−4 −4
Cp

Cp
−3 −3

−2 −2

−1 −1

0 0

1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c x/c

(a) α = 0◦ . (b) α = 12◦ .

Figure 3.11 – Pressure distributions for optimum cylinder and baseline configuration at low and high a.o.a.
with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 55◦ .

the leading edge of the upper flap surface is significantly lower. The sudden pressure rise is also found
further towards the flap trailing edge, which suggests that flow separation is postponed. It further seems
that the pressure recovery is smoother for the cylinder configuration. The enhanced flap lift also reduces
pressures at the upper airfoil, which is probably achieved through an increased circulation effect of the
flap on the main element.

Cylinder oil flow visualization δf = 55◦

Figure 3.12 shows oil visualization pictures of the flap for the baseline and cylinder configuration at
α = 0◦ and α = 10◦ respectively. As explained in section 2.4.3 oil flow visualization gives direct visual
insight in the flow characteristics. The visible flow characteristics will now be discussed. Observation of
picture 3.12a shows that the fluorescent oil is streaked downstream from left to right up till the dividing
vertical line. This is the point where laminar to turbulent transition takes place. At this point, inside
the bubble, there is very little flow and oil is clumped up resulting in the clear dividing line. After the
bubble the flow reattaches and becomes turbulent. In this area all the oil streaks along with the uniform
flow leaving a clean surface. This is because the turbulent flow has a high shear stress that scours away
the oil. Further downstream the turbulent flow is unable to cope with the adverse pressure gradient and
separates. The separation point is clearly seen by a second dividing line where the flow has come to rest
and is clumped up again. After this point the flow is non-uniform and circulates and oil is streaked in
different directions. These oil flow patterns are seen in all the pictures, although clear differences are also
visible.
Most noticeable is the clear shift downstream of the second vertical line for the cylinder configuration
model with respect to the baseline configuration. This confirms the observations from the pressure
distributions in section 3.4.1 that flow separation has indeed been postponed. This is visible at both
α = 0◦ and α = 10◦ . This is a major indication that the cylinder vortices indeed have a positive effect
on delaying flow separation.
Comparing the oil patterns between the baseline and cylinder configurations also shows that the vortex-
entrained flap has a less evident transition line than the baseline run. This indicates two things. First,
the boundary layer growing from the flap leading edge does have a laminar zone despite the experienced
vortices entrained from the cylinder. Second, the cylinder configuration has a softer laminar to turbulent
transition, probably having a shorter bubble.

D.P. Jansen 38 MSc. Thesis


3.4. Cylinders

(a) Baseline configuration at α = 0◦ . (b) Baseline configuration at α = 10◦ .

(c) Optimum cylinder configuration at α = 0◦ . (d) Optimum cylinder configuration at α =


10◦ .

Figure 3.12 – Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 55◦ .

MSc. Thesis 39 D.P. Jansen


Chapter 3. Experimental results

3.5
Baseline
Cylinder, r ∗ = 0.090, θ = 18◦
Cylinder, r ∗ = 0.105, θ = 30◦
Cylinder, r ∗ = 0.090, θ = 30◦
3

2.5
Cl

1.5

1
−10 −5 0 5 10 15
α [◦ ]

(a) Cylinders set 1.

0.1

0
y/c

−0.1

−0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x/c

(b) Cylinder positions set 1.

Figure 3.13 – Effect of positioning on the lift for multiple cylinder configurations with D∗ = 0.0167 and
δf = 50◦ .

D.P. Jansen 40 MSc. Thesis


3.4. Cylinders

3.5
Baseline
Cylinder, r ∗ = 0.090, θ = 18◦
Cylinder, r ∗ = 0.105, θ = 30◦
Cylinder, r ∗ = 0.090, θ = 30◦
3

2.5
15
Baseline
Cylinder, r ∗ = 0.090, θ = 18◦
C lf lap , C lairf o il

Cylinder, r ∗ = 0.105, θ = 30◦


2
Cylinder, r ∗ = 0.090, θ = 30◦
12

1.5
C lairf oil/C lf lap

1
6

0.5
3

0
−10 −5 0 5 10 15 0
α[◦ ] −10 −5 0 5 10 15
α[◦ ]

(a) Clairf oil and Clf lap versus α. (b) Clairf oil /Clf lap versus α.

0.1 0.1

0 0
y/c

y/c

−0.1 −0.1

−0.2 −0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/c x/c

(c) Cylinder positions. (d) Cylinder positions.

Figure 3.14 – Flap and airfoil lift contributions for multiple cylinder configurations with D∗ = 0.0167 and
δf = 50◦ .

MSc. Thesis 41 D.P. Jansen


Chapter 3. Experimental results

−8 −8
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 30◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 30◦
−7 −7

−6 −6

−5 −5

−4 −4
Cp

Cp
−3 −3

−2 −2

−1 −1

0 0

1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c x/c

(a) α = 0◦ . (b) α = 12◦ .

Figure 3.15 – Pressure distributions for optimum cylinder and baseline configuration at low and high a.o.a.
D∗ = 0.0167, r∗ = 0.090, θ = 30◦ , δf = 50◦ .

3.4.2 Flap deflection δf = 50◦

Cylinder lift polars for δf = 50◦

Figure 3.13 shows the lift performance of several cylinder configurations in a lift polar at a flap deflection
of δf = 50◦ . It can be seen that again a lift increase with a cylinder configuration is possible over the
whole range of α. This improvement however is less evident when compared to the results of the δf = 55◦
case. The optimum configuration, found at a D∗ = 0.0167, r∗ = 0.090 and θ = 30◦ , shows a lift increase
of in ∆Cl0 = 0.14 and 0.32 in Clmax . At first glance this is opposite to the δf = 55◦ case, where the lift
increase is mainly found at low angles of attack. However as explained earlier and can be seen in figure
3.13, the baseline lift polar for δf = 50◦ at high angle of attack shows a lift stagnation. At least it is
clear from figure 3.13 the cylinder configuration near Clmax does not suffer from this lift stagnation and
therefore results in a large ∆Clmax . It is easy to see from figure 3.13 that if the stagnation would not
occur for the baseline configuration, a comparable trend in lift increase to the δf = 55◦ case would be
apparent. With regard to the optimum location and size of the cylinders, very comparable results are
found to the δf = 55◦ case. Again reduced frequencies of the cylinder vortices with F + = 3.6 are proven
to be most efficient.
The contributions of both the flap and airfoil lift performance, Clf lap and Clairf oil are visualized in figure
3.14a. It shows that in this case the performance improvement mainly lies in an increase of the airfoil lift.
In fact the flap lift does not increase at all compared to the baseline configuration until a higher angle
of attack is reached. It can be seen that as the flap lift is increased from α = 10◦ and onwards compared
to the no-cylinder situation, the lift on the airfoil shows a larger ∆Cl than at lower angles of attack.
This indicates an increased circulation effect of the flap on the airfoil at high incidence angles. This also
suggests that the lift increase found at lower α is not due to an increased circulation effect. At zero angle
of attack the baseline flap lift is even slightly higher with Clf lap = 0.546 compared to Clf lap = 0.544
for the cylinder variant. Finally figure 3.14b shows that for the optimum cylinder configuration the
contribution of the flap to the total lift performance is increased at higher angle of attack when compared
to the baseline configuration, a characteristic also observed at δf = 55◦ for the VG and cylinder tests.

D.P. Jansen 42 MSc. Thesis


3.4. Cylinders

Cylinder pressure distributions for δf = 50◦

Figure 3.15 indicates that at α = 0◦ there are just small differences in flap pressure distributions between
the original configuration and the best cylinder configuration. At the flap the negative pressure peak is
slightly reduced but longer maintained, however a general gain in negative pressure at the flap upper
surface is minimal. At α = 12◦ the pressure distributions show large differences. It is clear that these
differences mainly lie in the fact that for the baseline model separation the separation point on the flap has
evidently shifted towards the leading edge. This results in a significant increased and a longer maintained
negative pressure peak for the cylinder configuration at this angle of attack. Again the flap upper surface
pressure has a smoother recovery at both α = 0◦ and α = 12◦ . Overall comparison with the δf = 55◦
case concludes that the lift increase for the flap deflection of δf = 50◦ is mainly obtained through an
improved airfoil pressure distribution.

Cylinder oil flow visualization for δf = 50◦

The oil flow pictures in figure 3.16 show comparable phenomena as seen earlier for the δf = 55◦ visual-
izations. Clear transition and separation lines are visible accompanied between the laminar, turbulent
and flow separated zones respectively. Again the cylinder configurations at both α = 0◦ and α = 12◦ see
a shift in the separation line with respect to the original configurations. Where for the α = 0◦ case the
effect is only minor, for the α = 12◦ the situation changes from a large separation zone for the baseline
to an almost completely attached flow situation for the cylinder configuration. This indicates that the
cylinders are capable of entraining the flow to the extend of even completely removing the flow separa-
tion. It is at this angle where earlier the lift polars show a lift stagnation for the baseline model possibly
indicating the presence of a laminar separation bubble burst. Figure 3.16b shows some interesting details
of the encountered phenomenon. It is seen that the transition bubble is almost at the nose of the flap.
The flow then shortly separates after the bubble. This indicates that the flow separation initiated close
to the leading edge on the flap is indeed caused by a laminar bubble burst at high angle of attack. It
remains unclear why the bubble burst only occurs at this specific condition, hence short range of high
angle of attack at this flap deflection. The possible effect of bubble bursting is not found for the cylinder
configuration. It is clear that the cylinder vortices besides having direct positive lift effects on the system
are also capable of removing laminar bubble bursting, which results in the major improvements in delay
of flow separation seen in figure 3.16b and 3.16d.

3.4.3 Flap deflection δf = 45◦

Cylinder lift polars for δf = 45◦

Figure 3.17 shows the lift performance of several cylinder configurations in a lift polar at a flap deflection
of δf = 45◦ . It can be seen that only minor lift improvements are obtained over the whole range of
angle of attack when cylinders are used to entrain the flow with vortices. The optimum configuration,
found at a D∗ = 0.0167, r∗ = 0.090 and θ = 37◦ , showed a lift increase of 0.042 in Cl0 and 0.063 in
Clmax . The optimum position is found to be similar to the other two cases with flap deflections δf = 55◦
and δf = 50◦ . Again reduced frequencies of the cylinder vortices with F + = 3.6 are proven to be most
efficient.
The contributions of both the flap and airfoil lift performance, Clf lap and Clairf oil are visualized in
figure 3.18. It shows that the performance lies in an increase of the airfoil lift for the whole range of
α, accompanied by a flap lift increase at higher angle of attack. This behaviour is also seen in the
δf = 50◦ case, although now the performance improvements are much less evident.

MSc. Thesis 43 D.P. Jansen


Chapter 3. Experimental results

(a) Baseline configuration at α = 0◦ . (b) Baseline configuration at α = 12◦ .

(c) Optimum cylinder configuration at α = 0◦ . (d) Optimum cylinder configuration at α =


12◦ .

Figure 3.16 – Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 50◦ .

D.P. Jansen 44 MSc. Thesis


3.4. Cylinders

3.5
Baseline
Cylinder, r ∗ = 0.090, θ = 30◦
Cylinder, r ∗ = 0.090, θ = 37◦
3

2.5

Cl 2

1.5

1
−10 −5 0 5 10 15
α [◦ ]

(a) Cylinder set 1.

0.1

0
y/c

−0.1

−0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x/c

(b) Cylinder positions

Figure 3.17 – Effect of positioning on the lift for multiple cylinder configurations with D∗ = 0.0167 and
δf = 45◦ .

Figure 3.18b shows that for the optimum cylinder configuration the contribution of the flap to the total
lift performance is increased at higher angle of attack when compared to the baseline configuration.

Cylinder pressure distributions for δf = 45◦

Figure 3.19a reveals that at α = 0◦ there are small differences in flap pressure distributions between
the original configuration and the best cylinder configuration. Though the negative pressure is longer
maintained at the flap upper surface, the negative pressure peak is reduced. This explains that overall
the flap lift is not increased, although flow separation may be slightly delayed. At α = 12◦ the differences
in the pressure distributions of the baseline and best cylinder configurations are more evident. Especially
the pressure over the flap upper surface pressure shows a smoother recovery.

Cylinder oil flow visualization for δf = 45◦

The oil flow pictures in figure 3.20 again show clear transition and separation lines between the laminar,
turbulent and flow separated zones respectively. As well as for the other flap deflections the cylinder
configurations at both α = 0◦ and α = 12◦ see a shift in the separation line with respect to the original
configurations. However these shifts are less distinct which support the observations made in the pressure
distributions where also small differences were noticed. Noticeable at α = 12◦ is the baseline configuration
suffering from a small area of separated flow just before the trailing edge, while the cylinder configuration
shows completely attached flow over the whole flap.

MSc. Thesis 45 D.P. Jansen


Chapter 3. Experimental results

3.5
Baseline
Cylinder, r ∗ = 0.090, θ = 30◦
Cylinder, r ∗ = 0.090, θ = 37◦
3

2.5

15
Baseline
C lf lap , C lairf o il

2 Cylinder, r ∗ = 0.090, θ = 30◦


Cylinder, r ∗ = 0.090, θ = 37◦
12

1.5
C lairf oil/C lf lap

0.5
3

0
−10 −5 0 5 10 15 0
α[◦ ] −10 −5 0 5 10 15
α[◦ ]

(a) Clairf oil and Clf lap versus α . (b) Clairf oil /Clf lap versus α.

0.1
0.1

0
0
y/c

y/c

−0.1 −0.1

−0.2 −0.2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/c x/c

(c) Cylinder positions (d) Cylinder positions

Figure 3.18 – Flap and airfoil lift contributions for multiple cylinder configurations with D∗ = 0.0167 and
δf = 45◦ .

D.P. Jansen 46 MSc. Thesis


3.4. Cylinders

−8 −8
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦
−7 −7

−6 −6

−5 −5

−4 −4
Cp

Cp
−3 −3

−2 −2

−1 −1

0 0

1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c x/c

(a) α = 0◦ . (b) α = 12◦ .

Figure 3.19 – Pressure distributions for optimum cylinder and baseline configuration at low and high a.o.a.
with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 45◦ .

Very clear are the transition lines in the baseline oil visualizations, which are hardly visible for the
cylinder configurations. This is similar to the behaviour found at the other flap angles. Since this was
encountered at all flap angles it was decided to further examine this observation. The softer laminar
to turbulent transitions gave rise to the idea that influencing the transition itself could delay the flow
separation. It is commonly known for example that turbulent boundaries are more resistant to adverse
pressure gradients and can indeed have a positive effect on delaying flow separation. It was decided to
perform several tests with applying zigzag tape on the flap to investigate the influence of transition on
the flow separation. This was only done for an angle of δf = 45◦ . It was found however that zigzag tapes
on different locations resulting in various turbulent flow situations on the flap did not result in significant
influences on the delay of flow separation. From this could be concluded that a different transition is not
sufficient to positively affect the location of flow separation.

3.4.4 Comparison of δf effects on cylinder performance and further investi-


gation

In section 3.4.1 to 3.4.3 the pressure measurements, lift polars and oil flow visualizations have been
shown and discussed for the cylinder configuration studies. The results concerning the working effect of
the cylinders are not straightforward. It is seen that at high flap deflection of δf = 55◦ the cylinders
indeed have a positive effect on flow separation delay and total lift with an increase up to 18% in Cl0 with
∆Cl0 = 0.32 compared to the baseline. However, as the flap angle decreases to δf = 50◦ and δf = 45◦ ,
lift increases are significantly less with ∆Cl0 = 0.14 and ∆Cl0 = 0.042 respectively.
At higher angle of attack a similar but slightly different trend is observed. In general lift gains are less
at high angle of attack, ∆Clmax = 0.24 for δf = 55◦ and ∆Clmax = 0.063 for δf = 45◦ , excluding the
exceptional case of the measurements performed at δf = 50◦ where a laminar separation bubble very
negatively affected the baseline case. The difference is that at a high flap deflection of δf = 55◦ the lift
gain decreased with increasing angle of attack, while at δf = 45◦ the cylinders perform slightly better
around ∆Clmax .

MSc. Thesis 47 D.P. Jansen


Chapter 3. Experimental results

(a) Baseline configuration at α = 0◦ . (b) Baseline configuration at α = 12◦ .

(c) Optimum cylinder configuration at α = 0◦ . (d) Optimum cylinder configuration at α =


12◦ .

Figure 3.20 – Oil flow visualization on the flap for the baseline and optimum cylinder configuration at
δf = 45◦ .

D.P. Jansen 48 MSc. Thesis


3.4. Cylinders

What is certain is that there is an optimum position found for the cylinders at r∗ = 0.090 and θ = 37◦
for all flap deflections. It seems higher θ result in cylinders being too close to the airfoil wall in the slot,
reducing the capability of the cylinder to develop an effective vortex street as seen by Wang and Tan
[2007]. Too low values of θ possibly result in vortices not passing the flap upper but lower surface. The
parameter r∗ mostly affects the position of the vortex creation with respect to the point of separation.
From the experiments it is likely that for values of r∗ < 0.090 the vortex strength at the point of separation
reduces being less effective. For values of r∗ > 0.090 the vortex source moves closer to the separation
point which may be good from a view of increased vortex strength, but is probably negatively affected
by the increasingly narrower slot. These effects are similar for all the tested flap deflections.
Also the optimum size seems to be the same for all flap cases, that is D∗ = 0.0167. It should be noted
that beforehand it was expected that larger cylinders would perform better due to the fact that vortex
strength increases with larger cylinder sizes and the reduced frequency is closer to F + = 1, which is the
theoretical optimum for reattaching separated flows. However as mentioned earlier, the cylinder vortices
are affected very much by wall effects, which especially with the larger cylinders can be significantly
influenced. It may be that the larger cylinders require a larger gap size or different overlap, because they
may be too close to the airfoil or flap surface to produce an effective vortex street. The larger cylinders
can also have a direct blocking effect through the slot negatively influencing the flow. Larger gaps and
overlaps are not tested in this experimental investigation.
Up to this point some phenomena are reasonably understood and some conclusions can be drawn. Still it
remains unclear why there is a loss in effectivity for the cylinders at lower flap deflections. For this reason
additional tests, that is frequency measurements and boundary layer measurements, are performed at a
flap deflection of δf = 45◦ to gain more insight in this behaviour.

Frequency measurements

At the flap angle of δf = 45◦ frequency measurements are performed at various locations near the flap
for an angle of attack of α = 0◦ . The experimental test setup is explained in section 2.4.2. The FFT
transformed signal measured by the listening tube results in the graphs depicted in figure 3.21, where
the single sided amplitude of the pressure fluctuation, |∆P (f ) |, is plotted in the frequency domain. The
measurements are performed for both the cylinder with D = 10 mm and the baseline configuration. The
locations at which the measurements are performed are shown in figure 3.22. The locations are chosen
such that the measurements will visualize a development of the vortices along the flap.
Focusing on the signal measured at location 1, directly in the wake of the cylinder, there is clear peak
visible at a frequency of 770 Hz for the cylinder measurements. This peak indicates the vortices shed
behind the cylinder, which have a shedding frequency of 770 Hz. This is very close to the theoretical
estimation of f = 0.2·40
0.010 =
St·U∞
D = 800 Hz, based on a local slot flow velocity of V = 40 m/s1 . This
is a clear proof that vortices are indeed shed from the cylinder at the slot at the expected frequency.
More important is that from this it can be concluded that the vortices are shed at a reduced frequency
of F + = 3.6 at which the cylinders capability of reattaching separated flow is close to optimal by theory,
according to Greenblatt and Wygnanski [2000].
More information is given by the frequency measurements. Also in figures 3.21b and 3.21f, thus at location
2 and 3, it is seen that peaks are present at 770 Hz for the cylinder frequency measurements. These
are smaller when compared to the clear peak in figure 3.21a. In free stream conditions at location 6 a
clear top is observed at 770 Hz as well, although small. The peaks at the cylinder vortex frequency are
not seen at location 4 and 5. These observations indicate the following. First the reduction in pressure
fluctuations at 770Hz from location 1 to 3 reveals that the vortex strength reduces substantially over this
distance. At position 3 the vortex strength from the cylinders is at a similar level of intensity as smaller
circulation structures that start to develop due to the onset of flow separation. It seems that at location
1 The local slot flow velocity V = 40 m/s is based on data from initial numerical simulations.

MSc. Thesis 49 D.P. Jansen


Chapter 3. Experimental results

0.05 0.05
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦

0.04 0.04
|∆P(f )| [dB]

|∆P(f )| [dB]
0.03 0.03

0.02 0.02

0.01 0.01

0 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz] Frequency [Hz]

(a) Location 1. (b) Location 2.

0.05 0.05
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦

0.04 0.04
|∆P(f )| [dB]

|∆P(f )| [dB]

0.03 0.03

0.02 0.02

0.01 0.01

0 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz] Frequency [Hz]

(c) Location 3. (d) Location 4.

0.05 0.05
Baseline Baseline
Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦ Cylinder, D ∗ = 0.0167, r∗ = 0.090, θ = 37◦

0.04 0.04
|∆P(f )| [dB]

|∆P(f )| [dB]

0.03 0.03

0.02 0.02

0.01 0.01

0 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz] Frequency [Hz]

(e) Location 5. (f) Location 6.

Figure 3.21 – Frequency measurements at multiple positions behind the cylinder. Pressure fluctuations
|∆P (f ) | are plotted against frequency with D∗ = 0.0167, r∗ = 0.090, θ = 37◦ and δf = 45◦ .

D.P. Jansen 50 MSc. Thesis


3.4. Cylinders

0.4
(6)

0.3

0.2

y/c
0.1

(1) (2) (3)


0 (4)

(5)
−0.1

−0.2
0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 3.22 – Frequency measurement locations at the flap at δf = 45◦ .

3 the vortices are already too weakened to have a positive effect on flow mixing in the boundary layer. It
follows that the vortices can only be effective before this point, between 0.845 < x/c < 0.95. This could
be the clarifying answer to why the cylinders are so effective at δf = 55◦ and not at δf = 45◦ . Observing
the pressure distributions in figures 3.11, 3.15 and 3.19 shows that for δf = 55◦ the separation occurs
before x/c = 0.95, while for δf = 50◦ and δf = 45◦ the separation point is slightly behind x/c = 0.95. At
the large flap deflection of δf = 55◦ the strength of the vortices at the point of separation is sufficient to
energize it, while at δf = 45◦ this is not the case.

It may be that the quick decay of vortices is not the only effect for a loss in mixing efficiency. The
measurements at location 6 at which lower pressure fluctuations are measured than directly behind the
cylinder clarifies the following. It shows the expected behaviour of these pressure fluctuations to decrease
with increasing distance from their point of origin. The spectrum measured at location 6 is a propagated
signal originated from the vortex street passing from location 1 to 2. This means that if not directly
measured in a vortex the measured fluctuations can be lower than the actual value for the closely passing
vortex. It could be that the measurements of |∆P (f ) | along the flap could also indicate that the vortex
gradually deviate from the streamline close to the wall, at which the measurements are performed, leading
to a loss in effectivity of re-energizing the boundary layer. This effect would be more critical for the lower
flap deflections for which the vortices need to pass a larger distance along the flap surface. The exact
effect of this is not known, since the path of the vortices is not determined.

In section 3.4.4 the differences in separation delay at high a.o.a. is addressed. A point can now also be
made on the effect of the angle of attack, which at δf = 55◦ had a larger negative influence on cylinder
performance, with regard to the vortex sustainability just discussed. Observing the pressure distribution
at figure 3.11 for the baseline shows that for the high flap deflection δf = 55◦ the pressure distribution on
the flap remains quite the same compared to the case at zero angle of attack. For smaller flap deflections,
i.e. δf = 45◦ , there is a clear trend that at higher angle of attack the pressure increases on the flap, at
the leading edge as well closely after. This indicates that the encountered adverse pressure gradient by
the flow is lower. This could have a positive effect on the conservation of the vortices, which as explained
in this section has a large influence. This would then explain why at lower flap deflections the cylinder
effect is more significant at high incidence angles with respect to low angles.

MSc. Thesis 51 D.P. Jansen


Chapter 3. Experimental results

Boundary layer measurements

Additional to the frequency analysis tests, transverse total pressure measurements are performed for the
flap angle of δf = 45◦ at the trailing edge of the airfoil and several locations on the flap. Figure 3.23 shows
the results for these measurements at low and high angle of attack, α = 0◦ and α = 12◦ respectively.
Normally a boundary layer shape is expressed in velocity U/Ue . However with the reasonable assumption
that within the boundary layer the static pressure is independent in the direction normal to the surface,
the total pressure can indeed give an indication of the boundary layer development. The total pressure
ptot here is referenced with the ambient pressure pamb at the wind tunnel.
Observing figure 3.23a at position 1, the trailing edge of the main element, the total pressure measure-
ments reveal that at α = 0◦ a very standard boundary layer profile is seen and that there is no difference
in boundary layer development between the baseline and cylinder configuration. At position 2 there are
some differences between both configurations. First it is seen that there is a deviation in the total pressure
very close to the wall for the baseline configuration. This is likely due to the transition point, which is
close at this measurement point. As explained from the oil flow visualizations the laminar to turbulent
transition is bubble is quite significant for the baseline case and that shows here in the total pressure
deviation. Further from the wall the boundary layer developments are quite similar with an increase in
total pressure (increasing velocity) and comparable boundary layer heights. Outside the boundary layer
the total pressure for the baseline configuration is much higher. This is because the wake of the cylinder
directly has a negative effect on the total pressure. Going further upwards a loss in total pressure is
observed for both configurations as is caused by the wake of the airfoil. Finally at the free stream flow
similar total pressure levels are found. Overall the total pressure developments do not show specific
indications that the boundary layer for the cylinder case is entrained by vortices and gets energized. This
may agree with the observed performance at α = 0◦ for this flap angle. Here only a minor total lift gain
was measured, and as explained in the previous section, is not expected to have much vortex interaction
at this position of x/c = 0.937, because of the low intensity of the vortices at this flap deflection of
δf = 45◦ . Still some improvements would be expected. This suggests that the total pressure itself is not
sufficient enough in this case to give a good indication of the boundary layer profile. At position 3 for the
same angle of attack no improvements are seen in the boundary layer development expressed by the total
pressure. Only from figure 3.23b which are the same measurements performed at α = 12◦ at location
3 there is a small sign of velocity entrainment in the boundary layer. The increase in total pressure at
boundary layer height could indicate an improved velocity profile at this position, which according to the
oil flow visualizations should be completely attached. However this again does not fully agree with the
total pressure measurements. To conclude the performed total pressure measurements show logical and
clear phenomena outside the boundary layer, but are insufficient to accurately describe the behaviour of
the flow within the boundary layer. To improve local static pressures need to be known as well to obtain
true boundary layer velocity profiles.

3.5 Conclusion experimental investigation

Vortex generators and cylinders have been tested in different configurations to test the ability of these
devices to add momentum to the boundary layer. The hypothesis is that the shed vortices mix the
boundary layer with high momentum outside flow and should result in a delay of flow separation. The
experiment shows that flow separation indeed can be delayed by the effects of the passive lift devices and
that lift of the total system is improved.
First the baseline configuration was tested for all flap deflections. It is found that the clean model at
the tested gap, overlap and flap deflections is suspected of possible bubble bursting and stall hysteresis.
The VG and cylinder configurations do not show indications for this, perhaps improving the situation
with their effects on the flow. Baseline oil flow visualization shows signs of laminar bubble bursting at

D.P. Jansen 52 MSc. Thesis


3.5. Conclusion experimental investigation

0.25
Ptot Baseline
Ptot Optimum Cylinder

(1)

(2)
0.1

(3)
y/c

−0.05

−0.2
0.7 0.8 0.9 1 1.1 1.2
x/c
(a) α = 0◦ .

0.1
Ptot Baseline
(1) Ptot Optimum Cylinder

−0.05 (2)

(3)
y/c

−0.2

−0.35
0.7 0.8 0.9 1 1.1 1.2
x/c
(b) α = 12◦ .

Figure 3.23 – Total pressure measurements of the boundary layer at various locations on airfoil and flap
for the baseline and optimum cylinder configuration with D∗ = 0.0167, r∗ = 0.090, θ = 37◦
and δf = 45◦ .

MSc. Thesis 53 D.P. Jansen


Chapter 3. Experimental results

high angle of attack. The stall of the multi-element model is of a leading type found on the airfoil for all
configurations. The VG’s and cylinders do not increase the stall angle.
The baseline configurations at all tested flap deflections show large areas of flow separation at the flap.
For the cylinder configurations the separation is significantly delayed in most cases when compared to
the baseline cases. This is seen in the various surface pressure measurements and oil flow visualizations.
Accompanied to the effects of delays in flow separation are measurements of significant lift increases. It is
found that cylinders performance is best at high flap deflections, hence at δf = 55◦ . The VG’s are only
tested at high flap deflection of δf = 55◦ and show a significant performance increase in only some of the
tests. It can be concluded that the cylinders capabilities to delay flow separation and increase the lift
are superior to the vortex generators. This fact lies, besides in smaller noticed effects in flow separation
delay, also in the narrower bandwidth at which VG’s are proven to be effective. While for the cylinders
multiple positions and sizes are efficient in creating flow separation delaying vortices, the VG’s show a
performance gain at only a single configuration. This is at a height of 0.023c and spacing λ/h = 4, where
lift gains of ∆Clo = 0.11 are measured, an improvement of 6% . Larger spacing’s result in an insufficient
amount of healthy air to be pumped into the boundary layer, while a configuration with smaller VG
spacing’s possibly suffer from vortex interfering effects. The cylinder configurations show performance
gains of 18% with ∆Cl0 = 0.32 at α = 0◦ and δf = 55◦ while effects were less at δf = 45◦ with an
increase of only ∆Cl0 = 0.042. The optimum position is found at r∗ = 0.090 and θ = 37◦ at a size of
D∗ = 0.0167 and these settings proved to be best for all flap settings. Larger cylinder sizes are tested
only at the flap deflection of δf = 55◦ and they prove to be less effective, although reduced frequency
values are closer to the preferable F + = 1. In these cases the cylinders could block and negatively affect
the flow through the gap. Higher values of θ result in vortices that come too close to the airfoil surface,
while low values of θ lead to vortices passing the flap on the lower side instead of the upper side. r∗ can
have a large influence as well where values of r∗ > 0.090 result in the cylinders getting too close to the
airfoil and flap surface in the slot. It is found through the frequency measurements that the vortex decay
along the flap upper surface is significant and result in a vortex strength that is too small toward the flap
trailing edge to be efficient at flap deflections of δf = 45◦ . This explains the much smaller lift gains at
the lower flap deflections. In these cases a larger cylinder diameter having stronger vortices could have
been more effective, although this has not been tested. In case larger cylinders are tested or when the
cylinders are positioned closer to the point of separation on the flap (thus increasing r∗ ) should probably
be done in combination with a larger gap size and overlap setting to deny negative wall effects for the
vortex production. Changing the gap and overlap also directly affects the performance of the slot and
thereby the lift characteristics of the complete model.
The effect of the angle of attack is not completely understood. At a flap deflection of δf = 55◦ the lift
gains are higher at low incidence angles while at δf = 45◦ increases are larger at high incidence angles.
This could be due to a less severe adverse pressure gradient developing on the flap at lower flap angles
that could have a positive effect on the vorticity preservation.

D.P. Jansen 54 MSc. Thesis


CHAPTER 4

Numerical investigation

4.1 Introduction

In section 3.4 it has been proved that the cylinders can have a positive effect on the delay of flow separation
and hence increase the lift. This has been quantified and visualized by the pressure measurements,
frequency measurements and oil flow visualizations. However, more insight is necessary in order to
determine the exact working principle of the cylinders. In addition, a direct influence of the cylinder on
the local flow is investigated. To increase our understanding on these aspects a CFD (Computational
Fluid Dynamics) investigation is performed. The goal in this numerical investigation is to develop a
model which can give comparable results to the experiment and provide additional information for the
cylinder working principles.
Section 4.2 will explain the general build-up of the CFD model. Followed by section 4.3 which will discuss
the numerical domain of the model. In section 4.4 several validations on the CFD model are presented.
The results of the numerical investigation are discussed in section 4.5 followed by the conclusion.

4.2 General setup

The numerical analysis was performed by the commercially available FLUENT package, which is part of
the ANSYS 13.0 simulation software. FLUENT is able to solve the equations of mass, momentum and
energy conservation. For this model the energy equation is not solved, because the flow is assumed to be
incompressible. The mass and momentum equation were solved as unsteady to accurately capture the
vortex generation. However some initial steady computations were also performed as will be discussed in
section 4.3.
To simulate the turbulence FLUENT can apply three types of methods: Reynolds Averaging Navier
Stokes (RANS), Large Eddy Simulation (LES) or Detached Eddy Simulation (DES). For this simulation
RANS is selected as LES and DES simulations require much more computational effort. RANS models
the turbulence through the definition of a mean and fluctuating component of all the flow variables that
generates additional equations. These equations are closed by a selected turbulence model which in this

55
Chapter 4. Numerical investigation

case is done with the k − ω SST model. The model is known for its good behaviour for separated flows.
This is indicated by several numerical investigations like Catalano and Amato [2003] who performed a
turbulence model study for multi-element high lift configurations and praised the capabilities of the k − ω
SST model in such cases. The model also gives the opportunity to serve as a low-Reynolds turbulence
model where the flow is directly modelled all the way through the viscous sub-layer instead of applying the
commonly used semi-empirical wall functions to capture the boundary layer. Although this low-Reynolds
behaviour requires a higher density mesh and thus computational effort, it is known to be more accurate
in capturing flow separation.

4.3 The numerical domain

For the grid generation a two dimensional C-type structured mesh is chosen, which is illustrated in figure
4.1. The grid generation is done using the Ansys Meshing tool. The most forward point and both the
upper and lower side of the total domain are at a distance of 10c from the airfoil flap configuration, while
the rear boundary is at a distance of 15c. Multiple zones are modelled in order to create the structured
grid and to allow more control on specifying higher mesh densities for crucial areas. A Near-field zone
is defined as a high-density region close to the airfoil flap model. Also a Near-wake zone is modelled as
a high-density region. Both regions are indicated in figure 4.1b. The structured mesh leads to a grid of
approximately 600000 elements.
The choice of a multi-zone structured grid is based on the following. Although it takes more effort to
create such a grid for complex shaped geometries than an unstructured grid, it can have the following
benefits. First a structured grid gives more control on specifying the mesh and can prevent the mesh to
have cells with bad shape, hence a high skewness. Second the numerical diffusion is kept to a minimum
when the flow is aligned with the direction of the grid, which is not realizable with an unstructured
grid. This increases the accuracy and can be used to decrease the number of cells leading to shorter
computation times. Of course in the areas of severe flow separation and back flow, in the region behind
the flap in this case, the structured grid suffers from similar numerical diffusion errors as a structured grid
and more cells are necessary. Murayama et al. [2006] investigated the influence of an unstructured and
structured mesh for the numerical and experimental research of a 3-dimensional wing flap model. For the
low flap deflection both meshes resulted in numerical comparable results to experimental investigations.
Still slightly better results were found for the structured mesh in computing the lift coefficient, with
deviations in general being ≤ 1% against ≤ 1 − 3% for the unstructured mesh.
As explained in the previous section the boundary layer is treated by a low-Reynolds turbulence model.
For the high-density boundary layer mesh the ANSYS [2009] manual suggests a minimum of 10 cells
within the viscosity affected near-wall region (Rey < 200), which in this model is set at 15 cells or more
at all locations. Another constraint with the low-Reynolds turbulence model is a very small height of
the first cell often expressed through the definition of y + . The ANSYS [2009] manual explains that these
y + values should be in the order of 1 to a maximum of 4 − 5. Figure 4.3 shows the y + values found for
the baseline configuration at a flap deflection of 45◦ . It shows that the desired y + values are met for most
of the domain.

Boundary conditions and initial conditions

The far-field boundary conditions are specified with a velocity inlet of 50 m/s and a pressure outlet at
ambient pressure conditions of 101325 pa. The inlet and outlet turbulent properties are set at a turbulent
intensity of 0.1%, which corresponds to the turbulence characteristics of the Low Speed Low Turbulence
Wind Tunnel used in the experiment. Further a turbulence viscosity ratio of 1 is applied, which is common
for exterior flows as specified in the ANSYS [2009] manual. The contours of the airfoil, flap and cylinder
are given wall boundary conditions with zero slip.

D.P. Jansen 56 MSc. Thesis


4.3. The numerical domain

(a) Mesh of the total domain. The red contours in the center of the grid visualize the geometry of
the airfoil flap configuration.

(b) Mesh of the airfoil flap domain. Also visible is the Near-field zone around the airfoil and flap
and the Near-wake region behind the flap.

Figure 4.1 – Visualization of the numerical domain.

MSc. Thesis 57 D.P. Jansen


Chapter 4. Numerical investigation

Figure 4.2 – Mesh in the vicinity of the cylinder.

10
Airfoil Lower
Airfoil Upper
9
Flap Lower
Flap Upper
8

6
ywall

5
+

0
0 0.2 0.4 0.6 0.8 1 1.2
x/c

+
Figure 4.3 – ywall for the baseline model at δf = 45◦ .

D.P. Jansen 58 MSc. Thesis


4.3. The numerical domain

2.5

2
Cl

1.5

0.5

0
0 2000 4000 6000 8000 10000 12000 14000 16000
iteration

Figure 4.4 – Convergence of the steady state solution.

Steady solution

The unsteady simulations were preceded by steady simulations. The steady state solutions provide
accurate initial conditions for the complete flow field and speed up the computations performed with the
unsteady simulations. Figure 4.4 shows the convergence of the lift coefficient for a steady state solution.
It can be observed that convergence is obtained after 16000 iterations.

Unsteady solution

The unsteady simulations are calculated by the PISO solver, which is known to be efficient for unsteady
simulations as stated by ANSYS [2009]. Pressure, momentum and turbulent properties are calculated
with 2nd order spatial discretization accuracy. The transient formulation is through a 1st order implicit
method.
For the unsteady simulations setting the correct time step is of primary importance. The time step was
set to accurately capture the vortex shedding of the cylinder. From the experiment and theory a vortex
shedding frequency is found of 770 Hz and 800 Hz respectively for a cylinder with D = 10 mm at a
Strouhal number of 0.2. For this numerical simulation the experimentally found result is considered to
be a better prediction and a frequency is estimated of 770 Hz. This corresponds with a time period of
1.130 · 10−3 s for each vortex shedding. It is important to capture the shedding with a sufficient amount
of time steps. Lam and Wei [2006] suggest a minimum of 20 time steps is necessary to capture a cylinder
vortex street. Using this and including a small margin results in an applied time step of 5.0 · 10−5 s for
the numerical simulations.
After each iteration in time the solution needs to converge to steady state. For the unsteady simulations
of the configurations with cylinder, convergence is analysed by monitoring the velocity magnitude in
multiple points displayed in figure 4.5. A parameter is defined to quantify the convergence in the monitor
points:
Ui+1 − Ui
σmon = (4.1)
Ui
Figure 4.6 shows σmon between each time step. In general it is found that after 25 iterations the velocities
in monitor points 1 and 2 are converged to σmon < 1 · 10−5 . However, after this amount of iterations the

MSc. Thesis 59 D.P. Jansen


Chapter 4. Numerical investigation

Figure 4.5 – Locations of the monitor points.

D.P. Jansen 60 MSc. Thesis


4.4. Model validation

−3
x 10
1.5
monitor point 1
monitor point 2
monitor point 3
1

0.5
σ mon

−0.5

−1

−1.5
4000 4050 4100 4150 4200
iteration

Figure 4.6 – Convergence of the velocity magnitude in the monitor points.

velocity at monitor 3 mostly shows values of σmon > 1 · 10−4 and does not seem to be fully converged yet.
For this reason, it was decided to fix the amount of iterations between each time step at 50 iterations. It
is possible that less iterations may prove to be sufficient as well.

4.4 Model validation


Before the model was used to simulate airfoil flap configurations with cylinders, it was tested for some
easier case studies. This involved baseline cases where three different flap deflections of δf = 15◦ ,
δf = 45◦ and δf = 55◦ were tested. The first case study is the baseline airfoil flap configuration at a low
flap deflection of 15◦ at which flow separation does not occur.

4.4.1 Low flap deflection δf = 15◦ , no flow separation


First a low flap deflection is chosen at which flow separation is not expected on the flap. Pressure data for
this flap setting is obtained from Boermans and Rutten [1995], since this flap deflection was not tested in
the experiment of this research. The flap for this case is positioned with a gap size and overlap of 0.03c
and 0.08c respectively. Figure 4.7 shows in one plot the pressure distributions of the airfoil and flap for
the reference case and the numerical simulation.
It can be seen that at low flap angle there are some differences in the pressure distributions, but in
general they seem to agree well with each other. Firstly, it must be noted that a small error exists
for the experimental computation at the main airfoil trailing edge. This seems to be a problem in the
post-processing Profmeasure Fortran code, where an incorrect method is applied to connect the upper
and lower pressure distribution at the last pressure hole. A similar though less evident effect can be seen
at the flap trailing edge pressure distribution. It can be stated that the numerically obtained pressure
distributions are better representations at these locations. Figure 4.7 further shows that flow separation
does not occur at the flap upper surface in both the experimentally and numerically obtained pressure
distribution. A small area of separated flow is observed at the bottom of the airfoil at x/c = 0.7, which
is captured by both methods. However, the numerically calculated pressure seems to be lower when
compared to experiments.

MSc. Thesis 61 D.P. Jansen


Chapter 4. Numerical investigation

−8
Exp Baseline
CFD Baseline
−7

−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.7 – Pressure distributions for CFD and experiment at low flap deflection at δf = 15◦ .

Another observation shows that the peak pressure at the flap nose is lower for the numerically calculated
distribution, which also results in a higher Clf lap compared to the experiment. As explained earlier in
section 3.1, Smith [1975] says that the main element can be very much affected by the circulation effect
of its downstream element. The higher Clf lap could explain the slightly lower pressure distribution over
the airfoil due to an increased circulation effect. The reason for the lower peak pressure at the flap could
be a result of the earlier mentioned flow separation occurring in the slot at the bottom of the airfoil,
which the CFD model is not able to capture very accurately.

4.4.2 High flap deflection δf = 45◦

Now a flap deflection of δf = 45◦ is chosen at which flow separation definitely will occur on the flap.
Figure 4.8 shows the pressure distribution of the airfoil and flap for the experiment and the numerical
simulation. Again a small error is observed in the experimentally obtained pressure distribution at the
airfoil trailing edge. Further it is clear that the numerical simulation does not capture the flow separation
on the flap correctly. The point of separation is in front of the experimentally obtained location and the
accompanying pressure values are significantly different. Similar to the low flap angle test case the peak
pressure at the flap is lower, but for this case also the pressure near the flap trailing edge. Although the
exact separation point is not correct, the CFD simulation does show the capability to accurately model
the pressure recovery. Evaluating the pressure gradient of the flap at the recovery show similar values for
CFD and experiment. Further calculating the surface integral from both flap pressure distributions also
show that very similar lift coefficients Clf lap are obtained.
The pressure distribution on the main airfoil is very comparable to the experimental result. This can be
expected, since no flow separation occurs on the main airfoil, but it may also be a result of the comparable
Clf lap flap lift contributions. Due to the comparable flap lift contributions the net circulation effect of
the numerically and experimentally obtained flap pressure distributions can be considered similar and
hence do not result in differences on the main airfoil pressure distributions.

D.P. Jansen 62 MSc. Thesis


4.4. Model validation

−8
Exp Baseline
CFD Baseline
−7

−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.8 – Pressure distributions for CFD and experiment at high flap deflection at δf = 45◦ .

4.4.3 Very high flap deflection δf = 55◦


Finally a very high flap angle of δf = 55◦ is simulated. Also for this flap deflection the numerical
simulations do not agree perfectly with the experimental results. Again the flow separation on the flap is
not modelled correctly and also here lower pressures are found at the flap leading and trailing edge than
in the experiment. A significantly higher Clf lap from the CFD simulation can be observed and explains
the over prediction of the pressure on the main airfoil again due to its circulation effect.

4.4.4 Conclusion on baseline model validation


From the previous model validations at low and high flap angle it can be concluded that the model
especially has trouble with accurately capturing the pressure distribution on the flap. At low angles this
is already visible in an over predicted negative pressure peak at the flap nose, while at higher angles the
pressure distribution over the complete flap chord shows discrepancies with the experiment. In general
the pressure distribution on the airfoil is well captured.
This means that at this point the numerical model is quantitatively questionable. This must be noted
knowing that numerical studies incorporating cylinders will involve even more complex flow phenomena
around the flow separation. However, it must be said that there are trends observed for the baseline
model for increasing flap deflection, which imply that the model can be appropriate for indicating trends
with investigating cylinder configurations. At this point it was decided to further investigate possibilities
to improve the performance of the numerical model before cylinder configurations are tested. The over
predicted pressure at the flap nose for low and high flap deflections presumed that it was necessary to test
the influence of laminar zones in the CFD model. For the large differences in modelling flow separation at
high flap angles, it was decided to examine the possible influence of the tunnel wall and blockage effects.

4.4.5 Influence of tunnel wall at large flap deflection


The experimental results were corrected for the influence of the wind tunnel wall by the method of Allen
and Vincenti [1944]. These corrections made are due to blockage effects of the tunnel walls on the model.

MSc. Thesis 63 D.P. Jansen


Chapter 4. Numerical investigation

−8
Exp Baseline
CFD Baseline
−7

−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.9 – Pressure distributions for CFD and experiment at very high flap deflection at δf = 55◦ .

Figure 4.10 – Total domain with modelled tunnel walls on upper and lower side.

These however are only corrections on the angle of attack and the performance parameters through a
source and sink method. These corrections do not take into account possible interference effects on the
separation point, the region of transition, and the structure of the near-wake. This makes it uncertain
if the Allen & Vincenti corrections are sufficient to model the complete tunnel wall interaction with the
airfoil flap model. It may be that the actual tunnel wall, especially at high flap deflections, influences
the location of flow separation. Therefore a test is performed for the numerical simulation where instead
of far field boundaries a nearby tunnel wall is modelled. The adjusted domain is shown in figure 4.10.
It should be noted that the tunnel walls are modelled as symmetry planes to avoid the development of
boundary layers. This is done to avoid the growth of very thick boundary layers, which would otherwise
grow along the semi-infinitely long tunnel walls and result in a significantly smaller effective tunnel throat.
It must be noted that the influence of the tunnel walls was tested on an initial mesh resulting in dissimilar
computed pressure distributions when compared to the numerical baseline distribution shown in figure
4.8. This initial mesh still is suitable to test the effect of the tunnel walls.

Figure 4.11 shows that the wind tunnel walls result in a lower pressure distribution on the upper airfoil
surface. The velocity over the airfoil is increased since the effective area the flow can move through is

D.P. Jansen 64 MSc. Thesis


4.4. Model validation

−8
Exp Baseline
CFD Baseline
−7 CFD Baseline with tunnel wall

−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.11 – Pressure distributions for the CFD simulation modelled with and without tunnel wall at
δf = 45◦ .

decreased. This blockage effect decreases the pressure over the airfoil, although this effect is only small.
Further effects like a shift in the location of the separation point on the flap are not visible. This suggests
that the blockage effect does not have a significant influence on the position of separation on the flap.
For this reason, it was decided not to incorporate the wind tunnel walls in further numerical simulations.

4.4.6 Modelling laminar zones in CFD

Effects on the pressure distribution

Up till this point the flow in the numerical simulation was modelled as fully turbulent. This is very
common for numerical simulations, although it is not an exact model of reality. As discussed in section
3.1 the laminar to turbulent transition can have a large influence on the development of flow separation.
In general, laminar flow is less resistant to an adverse pressure gradient than its turbulent counterpart
and will therefore separate earlier. However the effect of laminar zones for this complex flow situation
is not exactly straightforward. A test is performed with the influence of laminar zones in CFD, where
laminar to turbulent transition locations are known from the experimental data. To test the improvement
of the model a plot is made of the pressure distribution and compared to the fully turbulent zone model
and the pressure distribution obtained from the experiment. This is shown in figure 4.13. It must be
noted that the computational pressure distributions are from calculations performed on an initial mesh.
This mesh does not have a proper meshed Near-wake behind the flap and there are too few cells. The
computation can still be used to show a trend for the incorporation of laminar zones. The laminar and
turbulent zone designation is depicted in figure 4.12.
Figure 4.13 shows that modelling laminar zones has a small but noticeable effect on the flow situation
compared to the fully turbulent numerical model. The pressure distribution on the airfoil for the laminar
zones clearly shows the transition point occurring at 0.35c. The location of the separation point at the
flap is slightly shifted backwards and the pressure is lower just behind the flap leading edge. Also visible
is the higher pressure at the flap trailing edge for the laminar zone simulation, which is matching better
with the distribution from the experiment.

MSc. Thesis 65 D.P. Jansen


Chapter 4. Numerical investigation

Figure 4.12 – Designation of laminar and turbulent zones around the airfoil and flap at δf = 45◦ .

−8
EXP Baseline, δf = 45◦
−7 CFD Baseline, δf = 45◦
CFD Baseline, δf = 45◦ , with laminar zones
−6

−5

−4
Cp

−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.13 – Numerically calculated pressure distributions with/without laminar zones compared to ex-
periment at δf = 45◦ .

D.P. Jansen 66 MSc. Thesis


4.5. Results for baseline and cylinder configurations

Effects on the boundary layer

Defining laminar zones also affects the development of the boundary layer over the airfoil and flap. To
determine this influence the total pressures of the fully turbulent and laminar zone computations are
plotted along a vertical line from the airfoil surface at the trailing edge. These can be compared directly
with the baseline total pressure measurements performed in the experiment.

0.2 0.2
EXP Ptot Baseline EXP Ptot Baseline
CFD Ptot Baseline CFD Ptot Baseline, with laminar zones
y/c

y/c

0.8 0.8
x/c x/c

(a) Ptot measured for the numerical simulation mod- (b) Ptot measured for the numerical simulation with
elled fully turbulent. modelled laminar zones.

Figure 4.14 – Total pressure measurements in the boundary layer visualized for numerical simulations mod-
elled with laminar zones and for fully turbulent simulations at δf = 45◦ . Measurements are
performed at x/c = 0.787.

Figure 4.14 shows that the value of the total pressure for both numerical cases just outside the boundary
layer coincides very well with the experimentally measured value. The main difference lies in the height
of the boundary layer. For the fully turbulent flow model the boundary layer is thicker than measured
in the experiment while for laminar zone numerical simulation it is thinner. It is easy to understand
that the latter has a smaller height result since fully developed turbulent boundary layers along the same
geometry will always be thicker than ones where also a laminar and transition region is apparent, as
explained by White [2006]. The fact that the boundary layer is thinner than observed in the experiment
suggests that the transition location could be shifted more towards the airfoil leading edge. It should be
said that the slope ∂U/∂y is modelled accurately by the laminar zone model.
The effects seen in the pressure distributions and within the boundary layer indicate that the situation is
improved compared to the fully turbulent modelling, hence the distributions seen in previous sections 4.4.2
and 4.4.3 could be adjusted to incorporate laminar zones and see improvements. Still the effects are only
small. The exact incorporation of transition itself into a numerical simulation needs more justifications
and research as is done here. For these reasons, the incorporation of laminar zones is not included in
further simulations.

4.5 Results for baseline and cylinder configurations


The numerical model has been assessed by several validation tests. It is clear that determining the
correct position of flow separation with the RANS model remains a troublesome task and does not
exactly correspond to the experimental results, especially at very high flap angle. The incorporation of
tunnel walls and laminar to turbulent transition did not significantly increase performance of the model
to capture the flow separation more accurately. Quantitatively the CFD model is too inaccurate to

MSc. Thesis 67 D.P. Jansen


Chapter 4. Numerical investigation

directly compare with experimental result. However, trends and certain flow phenomena could definitely
be captured. This section will address the results found from the cylinder configuration in CFD and
the baseline numerical simulations. At some points comparisons are made between results found from
the experiments. It must be noted that only one cylinder configuration is tested. This is the optimum
cylinder configuration found in the experiment with a diameter of 10 mm or D∗ = 0.0167 at the position
of r∗ = 0.090 and θ = 37◦ . The numerical model is developed for δf = 45◦ and δf = 55◦ and is tested
only for a zero angle of attack.

4.5.1 Baseline and cylinder velocity field

The computed mean pressure distributions for the baseline configurations at δf = 45◦ and δf = 55◦
are already shown in section 4.4. Figure 4.15a and 4.16a show the accompanying mean velocity fields
of the baseline model at a flap deflection of δf = 55◦ , also supplied with the mean stream traces. The
velocity fields indeed show the large area of separation on the flap as well as the flow-separated area
near the airfoil lower side trailing edge. Figure 4.15b and 4.16b visualize velocity fields for the cylinder
configuration, which in general shows a very similar result to the baseline configuration with comparable
zones of highly separated flow. From this picture no clear delay of flow separation can be detected. Also
from figure 4.17 showing the pressure distributions for the baseline and cylinder configuration a shift in
the separation point is hard to detect. Although some minor effects are visible, neither a clear shift in
separation nor an increase in lift is shown. This does not correspond to the experimental investigation at
which a significant shift in the separation point was detected accompanied by an increase in lift of 18%.
This gives the suggestion that the numerical model is not able to capture the flow mixing capabilities
of the vortices as captured in the experiment. Further investigation of the numerical results will clarify
this.
To better observe the minor pressure effects on the airfoil and flap, the pressures are translated to local
pressure vectors perpendicular to the surfaces in figure 4.18. Focusing on the flap reveals that the mean
negative pressure is lower on the flap nose for the cylinder model. This is caused by a decreased flow
velocity, also visible in figure 4.16b. This decreased flow velocity may be caused by the cylinder vortex
generation utilizing total energy from the flow, leading to a decreased mean velocity over the flap nose.
The net effect of this decreased velocity and higher pressure effect is that the flap upper surface pressure
recovery is relieved and in fact flow separation is very slightly delayed. This observed phenomenon is not
due to a mixing effect but may be a direct result of the presence of the cylinder in the flow.
Figure 4.19 shows the mean skin friction coefficient on the flap upper surface. As indicated in section
3.1, the mean friction coefficient is an efficient parameter to indicate the location of flow separation.
Also from this plot only a slight improvement can be seen for the cylinder configuration in the delay of
flow separation. In the next section further investigations are performed on the cylinder vortex shedding
development and an explanation is sought why the flow mixing does not seem to be captured by CFD.

4.5.2 Vortex development

An important aspect of actual flow mixing in the boundary layer is a correct simulation of the vortex
development behind the cylinder. Therefore, the vortex generation and development is closer studied in
this section. For the experiment it is measured that vortices are created behind the cylinder and that the
vortex frequency agrees very well with the theoretically determined value based on a Strouhal number
of St = 0.2. It is interesting to see if the numerical simulation also shows a clear Von-Kármánn vortex
street at similar vortex frequencies.
Figure 4.20 visualizes the vorticity magnitude in the vicinity of the cylinder at different time instances.
It is clear that vortices are shed from the top and bottom of the cylinder. A close inspection of the shed
vortices, see figure 4.20d - 4.20f, reveals that from the lower surface a very clear vortex is created which

D.P. Jansen 68 MSc. Thesis


4.5. Results for baseline and cylinder configurations

(a) Baseline model.

(b) Cylinder model.

Figure 4.15 – Mean velocity field and streamlines for the baseline and cylinder configuration at δf = 55◦ .

MSc. Thesis 69 D.P. Jansen


Chapter 4. Numerical investigation

(a) Baseline model.

(b) Cylinder model.

Figure 4.16 – Mean velocity field and streamlines in the slot for the baseline and cylinder configuration at
δf = 55◦ .

D.P. Jansen 70 MSc. Thesis


4.5. Results for baseline and cylinder configurations

−8
CFD Baseline
CFD Cylinder
−7

−6

−5

−4

Cp
−3

−2

−1

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
x/c

Figure 4.17 – Mean pressure distribution for the baseline and cylinder configuration at δf = 55◦ .

Baseline Baseline
Cylinder Cylinder

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
x/c x/c

(a) Airfoil pressure vectors. (b) Flap pressure vectors.

Baseline Baseline
Cylinder Cylinder

0.7 0.8 0.9 0.7 0.8 0.9


x/c x/c

(c) Airfoil pressure vectors in the slot. (d) Flap pressure vectors in the slot.

Figure 4.18 – Mean pressure vectorization for the baseline and cylinder configuration at δf = 55◦ .

MSc. Thesis 71 D.P. Jansen


Chapter 4. Numerical investigation

0.04
τwm ean Baseline
τwm ean Cylinder

0.03

τw
0.02

0.01

0
0.85 0.9 0.95 1 1.05 1.1
x/c

Figure 4.19 – Mean skin friction coefficient for the baseline and cylinder configuration at the flap upper
surface at δf = 55◦ .

develops in time. The vortex created from the cylinder upper surface however develops in a stretched
way and quickly dissolves as it passes the flap leading edge, which is an unexpected result. A closer look
reveals that the boundary layer generated and shed from the airfoil lower side could explain the upper
vortex behaviour. At all displayed time instances the shed airfoil boundary layer generates significant
circulation that passes the cylinder upper side and clearly distorts the cylinder upper surface vortex. It
looks like this boundary layer vorticity directs the flow and cylinder vortices vertically towards the lower
airfoil surface, which forces the upper cylinder vortex to develop close to the wall. As discussed earlier in
section 3.4, research of Wang and Tan [2007] showed that cylinders which are positioned to close the wall,
S/D∗ < 0.8, can show asymmetric vortex generation with stretched close wall vortices. It seems that this
is also the case here for the numerical simulation. It may be, that the numerical modelling of the off-
surface boundary layer is not done accurately enough and thus causes the aforementioned problem. The
development of a separated boundary layer can be significantly affected by the local meshing, however no
time was available to solve or further investigate this problem. The end result of the asymmetric vortex
generation is that the upper surface vortex does not show a circulation effect over the flap. This will be
further explained in section 4.5.3.
The stretched vortex from the cylinder upper surface also seems to cause a second problem. The frequency
at which the vortices are simulated does not match with the frequency found in the experiment. This
frequency was calculated by observing the velocity fluctuations in monitor point 1 and 2, which are
plotted in figure 4.22. It shows that the velocity change due to the upper surface cylinder vortex is hardly
captured at both the monitor points but its effect is noticeable. From this figure a vortex shedding cycle
of 0.00315 s can be determined, hence the time for the development of two vortices. This corresponds
to a vortex frequency of 635 Hz, where experiment showed a frequency of 770 Hz, a difference of 17%.
Simulations with finer meshes around the cylinder, smaller time steps or increased number of iterations
between each time step have been tested but did not show significant improvements on the found vortex
frequency. It may be that the asymmetrical vortex generation is an influence on the observed difference
in the experimentally and numerically found cylinder vortex frequency. Remarks can also be made on
the comparison of a numerical 2-D model with a 3-D occurred experimental flow phenomenon. In the
3-dimensional space of the experiment the larger and smaller flow structures are allowed to have energy
cascades in all directions, while for the numerical simulation these degradations are only allowed in two
directions. Besides 3D effects the chosen turbulence model also affects the vortex development. Extensive
studies on simple 2D free stream cylinder configurations as performed by Rahman et al. [2007] show that
vortex frequencies can already deviate for 7 − 20% depending on the chosen turbulence model. This
indicates that the way of how small-scale eddies cascade is resolved can already have a large influence on
the vortex frequencies of larger flow structures.

D.P. Jansen 72 MSc. Thesis


4.5. Results for baseline and cylinder configurations

(a) t = t0 (b) t = t0 + 0.0005 s

(c) t = t0 + 0.00095 s (d) t = t0 + 0.00175 s

(e) t = t0 + 0.00255 s (f) t = t0 + 0.00315 s

Figure 4.20 – Vorticity magnitude at different time instances at very high flap angle. δf = 55◦

MSc. Thesis 73 D.P. Jansen


Chapter 4. Numerical investigation

2.5

2
Cl

1.5

0.5

Baseline
Cylinder
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
t[s]

Figure 4.21 – Lift coefficient development in time for the baseline and cylinder configuration at δf = 55◦ .

It is time to evaluate current obtained results. It is seen in section 4.5.1 that flow separation is not
significantly delayed as was observed at δf = 55◦ for the experimental investigation. It is suggested that,
although a direct influence of the cylinder has a minor positive effect, the numerical model is not able
to capture the flow mixing effect of energizing the boundary layer. This section shows that the vortices
development necessary for a flow mixing effect is not as expected. First there is an asymmetrical shedding
with stretched vortices at the cylinder upper side. Second, and this could well be caused by the first
observation, the vortex shedding is at a much lower frequency. These two observations could be the
direct cause of the inability to have a mixing effect. In the next section the vortex influence on the flap
is researched and discussed.

4.5.3 Vortex influence on the flap

The influence of the shed vortices on the flow over the flap can be displayed in several ways. Figure 4.24
and 4.25 show line rakes of the horizontal velocity component at two positions above the flap for the
baseline and cylinder configuration at different time instances. Each figure also illustrates a zoomed in
situation for the boundary layer. The locations of these line rakes are indicated in figure 4.23. It must
be noted that in the complete time interval a single (lower cylinder side) vortex has passed over the flap.
Both figures show that the line rakes for the cylinder configuration have more variations in the horizontal
velocity than for the baseline case mostly outside the boundary layer, due to the passing of a vortex. It
is also clear from figures 4.24a, 4.24b, 4.25a and 4.25b that overall a velocity loss exists in the wake of
the cylinder. This is due to a total pressure loss, which is also observed in the experiment and shown in
figure 3.23, section 3.4.4. Figure 4.25c reveals that in the boundary layer the average negative velocity
component is slightly smaller at the point of separation indicating a minor separation delay, which could
be the result from the earlier explained decreased pressure difference with the leading edge. Figure 4.24c
shows that at the time instance of t = t0 + 0.0170 the boundary layer is given a velocity increase at the
near wall regime of the boundary layer. This could be a very small sign of a flow mixing effect though it
is observed at a very small time interval. It does show that the numerically computed boundary layer is
capable to react on a vortex entrainment from outside the boundary layer. Still the effect is minimal and
is not sufficient to have a positive effect on delay of flow separation. It should be noted that it is here at
the boundary layer where the RANS model most likely has the biggest difficulty accurately capturing the

D.P. Jansen 74 MSc. Thesis


4.5. Results for baseline and cylinder configurations

120
monitor point 1
monitor point 2
monitor point 3

90
V [m/s]

60

30

0
0 0.005 0.01 0.015
t[s]

Figure 4.22 – Velocity magnitude development in time at the monitor points at δf = 55◦ .

flow. As was seen in the previous section, where the model has difficulty determining the exact location
of the separation point, it is the near wall treatment where these models start to show deviations with
the experiments. It may be that the boundary layer does not correctly cope with the entrained vortex.
For this different near wall treatments could be tested in combination with various turbulence models. It
may be that averaging nature of the RANS models itself is not capable of capturing the desired effects.

Figure 4.23 – Visualizations of the line rakes at x = 0.8965c and at x = 0.9365c.

MSc. Thesis 75 D.P. Jansen


Chapter 4. Numerical investigation

0.1 0.1
t = t0 t = t0
t = t0 + 0.0050 t = t0 + 0.0070
t = t0 + 0.0105 t = t0 + 0.0140
t = t0 + 0.0170 t = t0 + 0.0210
0.08 0.08

0.06 0.06
ywall/c

ywall/c
0.04 0.04

0.02 0.02

0 0
0 10 20 30 40 50 60 70 80 90 100 110 0 10 20 30 40 50 60 70 80 90 100 110
V x [m/s] V x [m/s]

(a) Cylinder configuration. (b) Baseline configuration.

0.02 0.02
t = t0 t = t0
t = t0 + 0.0050 t = t0 + 0.0070
t = t0 + 0.0105 t = t0 + 0.0140
t = t0 + 0.0170 t = t0 + 0.0210
0.016 0.016

0.012 0.012
ywall/c

ywall/c

0.008 0.008

0.004 0.004

0 0
0 10 20 30 40 50 60 70 80 90 100 110 0 10 20 30 40 50 60 70 80 90 100 110
V x [m/s] V x [m/s]

(c) Cylinder configuration, zoomed in on boundary (d) Baseline configuration, zoomed in on boundary
layer. layer.

Figure 4.24 – Horizontal velocity component at x/c = 0.8965 for very high flap angle at δf = 55◦ .

4.6 Conclusion on numerical investigation

The goal of the numerical investigation is to develop a model that gives comparable results to the
experiment and provide additional information on the working principles of the cylinders. The simulations
indicate that both items remain a troublesome task for the numerical model to perform.
A 2D model is created of the NLF-Mod22B airfoil at a flap setting of δf = 45◦ and δf = 55◦ . While in
general the flow around the airfoil is calculated correctly, an extensive model validation shows that it is
not possible to accurately capture the flow separation occurring on the flap. While at low flap angles the
RANS model performs quite well it loses accuracy at higher flap angles δf > 55◦ . Further investigations
on tunnel wall effects and incorporation of laminar zones, do not show significant improvements. It seems
that the application of a complex and resource demanding near wall treatment in combination with the
k − ω SST turbulence model known for its favourable behaviour in high separation flows, is not sufficient
to correctly model the boundary layer in the strong adverse pressure gradient that arises on the flap.
This means that the baseline configuration already shows deviations with the experimentally obtained
results.
Still simulations were performed on the cylinder configurations to observe possible trends measured by

D.P. Jansen 76 MSc. Thesis


4.6. Conclusion on numerical investigation

0.1 0.1
t = t0 t = t0
t = t0 + 0.0050 t = t0 + 0.0070
t = t0 + 0.0105 t = t0 + 0.0140
t = t0 + 0.0170 t = t0 + 0.0210
0.08 0.08

0.06 0.06
ywall/c

ywall/c
0.04 0.04

0.02 0.02

0 0
−20 −10 0 10 20 30 40 50 60 70 80 90 100 110 −20 −10 0 10 20 30 40 50 60 70 80 90 100 110
V x [m/s] V x [m/s]

(a) Cylinder configuration. (b) Baseline configuration.

0.02 0.02
t = t0 t = t0
t = t0 + 0.0050 t = t0 + 0.0070
t = t0 + 0.0105 t = t0 + 0.0140
t = t0 + 0.0170 t = t0 + 0.0210
0.016 0.016

0.012 0.012
ywall/c

ywall/c

0.008 0.008

0.004 0.004

0 0
−20 −10 0 10 20 30 40 50 60 70 80 90 100 110 −20 −10 0 10 20 30 40 50 60 70 80 90 100 110
V x [m/s] V x [m/s]

(c) Cylinder configuration, zoomed in on boundary (d) Baseline configuration, zoomed in on boundary
layer. layer.

Figure 4.25 – Horizontal velocity component at x/c = 0.9365 for very high flap angle at δf = 55◦ .

MSc. Thesis 77 D.P. Jansen


Chapter 4. Numerical investigation

the computational model. The cylinder tested has the settings of the optimum configuration found in
the experiment with a diameter of 10 mm or D∗ = 0.0167 at the position of r∗ = 0.090 and θ = 37◦ .
The simulations with cylinders show difficulties concerning the vortex development and an interaction
with the modelled boundary layer. The vortices created behind the cylinder are asymmetrical and the
frequency does not correspond to measurements performed in the wind tunnel. Vortices calculated in
the simulation are at a shedding frequency of 635 Hz where the experiment shows a frequency of 770 Hz.
An undesired interference effect with the airfoil lower surface boundary layer and 3D effects are possible
explanations for these observations. It is seen that the boundary layer on the flap hardly responds to
this vortex development and a flow mixing effect is not detected. Results may be better if the vortex
generation is simulated better, however questions remain on the capability of the modelled boundary
layer to cope with the entrained vortices from the outside flow.

D.P. Jansen 78 MSc. Thesis


CHAPTER 5

Conclusion and Recommendations

5.1 Conclusion
The goal of this thesis was to investigate the performance improvements of cylinder vortex streets on
the Extra EA-400 airfoil flap system in a 2D situation. In the experiments cylinders with diameters in
the range of 10 − 20 mm were tested at flap deflections of δf = 45◦ − 55◦ and compared to the baseline
configurations. At a flap deflection of 55◦ a comparison is also made with the behaviour of regular VG’s
with heights h∗ = 0.023. The model and applicable devices were tested at an inlet velocity of 50 m/s in
a closed loop wind tunnel. A fixed gap and overlap was used for the slot parameters at 0.035c and 0.0c
respectively.
The static pressure measurements and oil flow visualizations indicate that both the cylinders and VG’s
are capable of delaying flow separation on the flap and increasing the lift. It is found that the performance
of cylinders are superior to the regular VG’s in feasible separation delay as well as in a larger bandwidth
at which cylinders are effective. At the flap deflection of 55◦ an increase in lift is seen of ∆Cl0 = 0.32
for the cylinders and 0.11 for the VG’s. The VG’s positioning and sizing should be done with precision.
Large spacing’s result in an insufficient amount of healthy air to be pumped into the boundary layer,
while a configuration with small VG spacing’s suffer from vortex interfering effects. Low VG heights
prove to be less effective as the created vortices are small and are decayed before reaching the area of
flow separation. VG’s with a height of 0.023c and spacing λ/h = 4 are found to be optimal.
The tested cylinders become less effective as the flap deflection lowers to δf = 45◦ . Frequency tests
showed that the vortex strength reduces rapidly along the flap surface, losing the ability to energize the
boundary layer. At lower flap angles the vortices need be effective further toward the trailing edge, since
the location of flow separation shifts aft with decreasing flap deflection. At δf = 45◦ a lift increase of
∆Cl0 = 0.042 is measured with the application of cylinders. The optimum configuration is found to be
the same for all flap deflections at r∗ = 0.090 and θ = 37◦ with a cylinder size of D∗ = 0.0167 (10 mm).
The reduced frequency F + at the optimum configurations is determined to be in the range of 3 − 4, which
is close to the most efficient frequency according to literature. The cylinders require attentive placement
within the airfoil flap slot. Positioning the cylinders too close to the airfoil or flap surface results in
a distorted vortex street, while cylinders at a large distance from the separation location rapidly lose
efficiency.

79
Chapter 5. Conclusion and Recommendations

A numerical investigation has been performed with ANSYS Fluent to provide additional information with
respect to the experimental research. A RANS model is chosen to reduce computation resource demands
and modelling complexity. To cope with the large areas of flow separation a k − ω SST turbulence model
with low Re near wall treatment is selected. Validations of the 2D model at several flap deflections
give varying results. At a flap deflection of δf = 15◦ the numerical model performs well and pressure
distributions are comparable to the experiment. At a flap setting of δf = 45◦ and δf = 55◦ the numerical
model proves to be poor in calculating an accurate flow situation around the flap. This results in an
incorrect prediction of the position of flow separation.
Simulations for the cylinder configurations show difficulties in calculating a realistic vortex development.
A shedding frequency is measured of 635 Hz where the experiment shows a frequency of 770 Hz. In the
numerical simulation the vortices are also affected too much by the airfoil surface in the slot, which leads
to a stretched upper vortex and as a result an asymmetrical vortex street. There is little interaction
between the vortices and the boundary layer, and no significant energizing effect of the boundary layer
is observed. It is concluded that the numerical model is not able to capture this interaction. A small
positive effect is noticed of the direct influence of the cylinder on the flow in the slot and separation is
very slightly delayed. The results of the CFD investigation indicate the following. Although not captured
in the numerical simulation, the separation delays and positive lift increments found in the experiment
are primarily due to velocity entraining effects of the vortices inside the boundary layer and not only due
to a direct pressure effect of the cylinder.

5.2 Recommendations
The investigation on passive high lift control applied on a practical airfoil flap model has shown interesting
capabilities of vortex generating devices such as VG’s and cylinders. While the research in this thesis
led to some important answers to certain questions, it also gave rise to new questions and research
possibilities. Based on the findings from this investigation the following recommendations can be made
for future studies both by experiment and numerical simulations:

• The cylinders should be tested on an airfoil flap system with a leading edge slat. The cylinders in
this research showed capabilities of flow separation at high angle of attack. It would be interesting
to see if the cylinders are capable of further increasing Clmax in combination with the stall delaying
effects of a leading edge slat. The model can be further improved to practical situations when a
full wing is applied with cylinders. In this case lateral effects on the model can be further studied.
• Larger cylinders could be applied together with a larger gap size/overlap to create larger vortex
structures in order to improve the effectiveness of the cylinders at lower flap angles.
• Flow visualization techniques like PIV (Particle Image Velocimetry) could give additional infor-
mation on the behaviour of the flow in the slot and at the flap. PIV will visualize the vortex
development for the different cylinder positions in the slot and give further insight in the vortex
decay observed by the frequency measurements. Velocity profiles within the boundary layer should
be measured for the δf = 55◦ configuration to better understand the flow mixing effects of the
cylinders close to the wall.
• It is recommended to perform a Detached Eddy Simulation (DES) or Large Eddy Simulation (LES)
for further numerical investigations. DES and LES would model small-scale eddies and resolve large
eddies directly resulting in a more accurate representation of the large vortex structures behind the
cylinder and in the wake of the flap. The DES could still have problems in the boundary layer
where it models the flow with unsteady RANS models. This could remain troublesome for the
boundary layer vortex interaction and LES may prove to be more accurate. The up scaling from
RANS to DES would require much higher demands on computation resource, with LES demanding
even more from CPU and RAM-memory capacity.

D.P. Jansen 80 MSc. Thesis


Bibliography

H.J. Allen and W.G. Vincenti. Wall interference in a two-dimensional flow wind tunnel with consideration
of the effect of compressibility. Technical report, NACA, 1944.
ANSYS. ANSYS FLUENT 12.0 User’s Guide. Inc., 2009.
D. L. Ashby. Experimental and computational investigation of lift-enhancing tabs on a multi-element
airfoil. Technical report, Stanford University Department of Aeronautics and Astronautics, 1996.
M. Baragona. Unsteady Characteristics of Laminar Separation Bubbles an Experimental and Numerical
Investigation. PhD thesis, TU Delft, 2004.
K. Biber. Physical aspects of stall hysteresis of an airfoil with slotted flap. In AIAA Paper 1995-0440,
1995.
L.M.M. Boermans. Aerodynamic design of aircraft and advanced transportation systems. 1998.
L.M.M. Boermans. Practical implementations of boundary layer suction for drag reduction and lift
enhancement at low speed. In Presentation at KATnet II Workshop, Ascot UK, 2008.
L.M.M. Boermans and P.B. Rutten. Two-dimensional aerodynamic characteristics of airfoil NLF-MOD22
with fowler flap. Technical report, TU Delft, 1995.
J. F. Cahill. Summary of section data on trailing-edge high-lift devices. Technical report, NACA, 1949.
P. Catalano and M. Amato. An evaluation of RANS turbulence modelling for aerodynamic applications.
Aerospace Science and Technology, 7:493–509, 2003.
L. N. Cattafesta. Actuators for active flow control. Annual Reviews, 43:247–272, 2011.
D. Greenblatt and I. Wygnanski. The control of flow separation by periodic excitation. Progress in
Aerospace Sciences, 36:487–545, 2000.
B.M. Jones. Measurements of profile drag by the pitot-traverse method. In British Aero. Res. Council,
1936.
K.M. Lam and C.T. Wei. Characteristics of vortices shed from a circular cylinder and an inclined flat
plate. In Computational Wind Engineering, 2006.
P. LeBeau. Flow control using plasma actuators, 2007. URL http://www.pa.uky.edu/.
B. Lee, K. Yee, W. Joo, and D. Lee. Passive control of dynamic stall via nose droop with gurney flap.
In AIAA Journal, 2005.

81
Bibliography

J. C. Lin. Control of turbulent boundary-layer separation using micro-vortex generators. AIAA Journal,
3404:99, 1999.

J. Little, M. Nishihara, I. Adamovich, and M. Samimiy. Separation control from the flap of a high-lift
airfoil using DBD plasma actuators. In AIAA Paper 2009-145, 2009.
V. Maldonado, J. Farnsworth, W. Gressick, and M. Amitay. Active control of flow separation and
structural vibrations of wind turbine blades. Wind Energy, 13:221–237, 2010.
P.T. Meredith. Viscous phenomena affecting high-lift systems and suggestions for future CFD develop-
ment. In AGARD CP-515, 1992.

M. Murayama, T. Imamura, K. Yamamoto, and K. Kobayashi. Comparison of RANS simulations of


multi-element high-lift configurations. In AIAA Paper 2006-1396, 2006.

B. Nishri and I. Wygnanski. Effects of periodic excitation on turbulent flow separation from a flap. AIAA
Journal, 36(4):547–556, 1998.
Mahbubar Rahman, Mashud Karim, and Abdul Alim. Numerical investigation of unsteady flow past a cir-
cular cylinder using 2-D finite volume method. Journal of Naval Architecture and Marine Engineering,
4:27–42, 2007.

I. G. Recant. Wind-tunnel investigation of a NACA 23030 airfoil with various arrangements of slotted
flaps. Technical report, NACA, 1940.

K. C. Rudolph. High-lift systems on commercial subsonic airliners. Technical report, NASA, 1996.

A.M.O. Smith. High-lift aerodynamics. Journal of Aircraft, 12:518–523, 1975.

E. Torenbeek. Synthesis of subsonic airplane design. Delft University Press, 1976.

M.V. van der Steen. Passive off-surface flow separation control methods on a simplified flapped configu-
ration. Master’s thesis, TU Delft, 2009.

L.L.M. Veldhuis and M. van der Jagt. Separation postponement by means of periodic surface excitation.
In ICAS, 2010.

F. von Stillfried, S. Wallin, and A. V. Johansson. An improved passive vortex generator model for flow
separation control. In AIAA Journal, 2010.

X. Wang and S. Tan. Near-wake flow characteristics of a circular cylinder close to a wall. In Journal of
Fluids and Structures, 2007.
F.M. White. Viscous Fluid Flow. McGraw-Hill, 2006.

O. Wright, J.M.H. Jacobs, and F.D. Hardesty. Airplane, 1924.

D.P. Jansen 82 MSc. Thesis


APPENDIX A

Test matrix for the wind tunnel


experiments

In this appendix the test matrix for the wind tunnel experiments are presented. The test matrix is
two-fold, one for the tested vortex generators and one for the tested cylinders.

Table A.1 – Test matrix for the vortex generators.

Run δf λ/h h∗ d∗
Baseline 55 - - -
VG 1 55 4.5 0.010 0
VG 2 55 6.5 0.010 0
VG 3 55 4.5 0.023 0
VG 4 55 4.5∗ 0.023 0
VG 5 55 9 0.023 0
VG 6 55 13.5 0.023 0
VG 7 55 6.5 0.023 0
VG 8 55 4 0.023 0
VG 9 55 4∗ 0.023 0
VG 10 55 4 0.07 0.13
VG 11 55 8 0.07 0.13

= shifted λ/2h w.r.t. previous run.

83
Appendix A. Test matrix for the wind tunnel experiments

Table A.2 – Test matrix for the cylinders.

Run δf D∗ r∗ θ
Baseline 55 - - -
Cyl 1 55 0.017 0.060 37◦
Cyl 2 55 0.017 0.060 18◦
Cyl 3 55 0.017 0.060 55◦
Cyl 4 55 0.017 0.060 0◦
Cyl 5 55 0.017 0.060 30◦
Cyl 6 55 0.017 0.090 30◦
Cyl 7 55 0.017 0.090 37◦
Cyl 8 55 0.033 0.090 37◦
Cyl 9 55 0.033 0.090 10◦
Cyl 10 55 0.033 0.105 -165◦
Cyl 11 55 0.025 0.090 30◦
Cyl 12 55 0.025 0.105 30◦
Cyl 13 55 0.033 0.105 37◦
Cyl 14 55 0.25 0.045 30◦
Cyl 15 55 0.25 0.045 42◦
Baseline 50 - - -
Cyl 16 50 0.017 0.090 30◦
Cyl 17 50 0.017 0.090 18◦
Cyl 18 50 0.017 0.090 54◦
Cyl 19 50 0.017 0.105 30◦
Cyl 20 50 0.017 0.075 30◦
Baseline 45 - - -
Cyl 22 45 0.017 0.090 30◦
Cyl 23 45 0.017 0.090 37◦

D.P. Jansen 84 MSc. Thesis


APPENDIX B

Additional pictures on the wind


tunnel and model layout

This appendix will show some additional pictures of the Low Speed Low Turbulence Wind Tunnel of
Delft University of Technology. Also some detailed pictures of the model are presented.

85
Appendix B. Additional pictures on the wind tunnel and model layout

Figure B.1 – Exterior view of the wind tunnel test section.

Figure B.2 – Interior view, looking down stream, of the wind tunnel test section. Clearly visible are the
pitot tube and the wake rake that is positioned behind the model.

D.P. Jansen 86 MSc. Thesis


Figure B.3 – Detailed view on the positioning of the cylinders. The mechanism supporting the cylinder is
clearly visible as well as the zigzag tape that was used to remove the lower surface laminar
separation bubble.

Figure B.4 – Detailed view of the suction orifices at the wing/wall junction shown here around the airfoil
upper surface.

MSc. Thesis 87 D.P. Jansen


Appendix B. Additional pictures on the wind tunnel and model layout

Figure B.5 – View of the zigzag positioning of the pressure holes on the airfoil and flap model surface.

D.P. Jansen 88 MSc. Thesis


APPENDIX C

Output file example of pressure


measurements from Profmeasure

89
Appendix C. Output file example of pressure measurements from Profmeasure

D.P. Jansen 90 MSc. Thesis


MSc. Thesis 91 D.P. Jansen
Appendix C. Output file example of pressure measurements from Profmeasure

Figure C.1 – Raw data of pressure measurements for baseline run at δf = 55◦ .

In this appendix an example of an output file for a single pressure measurement is given. This output
is produced by Profmeasure, which presents both the uncorrected pressure data and the pressure data
corrected by the method of Allen & Vincenti. The input for Profmeasure is the measured pressures and
tunnel conditions gathered by Labview. Also input is given in Profmeasure for the Allen & Vincenti
correction (in this case the cx400_d30 input file is used for all measurements). The output data was
further post-processed in MATLAB to visualize the results, which are shown in this thesis.

D.P. Jansen 92 MSc. Thesis

You might also like