Journal of Wind Engineering

and Industrial Aerodynamics 91 (2003) 113–126

An experimental study of the aerodynamic

influence of a pair of winglets on a
flat plate model
M.G. Savage*, G.L. Larose
Aerodynamics Laboratory, National Research Council of Canada, Ottawa, Ont., Canada K1A 0R6

Abstract

The influence of a pair of thin stationary winglets on the aerodynamics of a bluff body is
studied in this paper. The bluff body in question is a rectangular prism with a width-to-depth
ratio of nine, and is referred to as a flat plate. The aerodynamic and aeroelastic characteristics
studied experimentally were the static aerodynamic force and moment coefficients, their rate
of change with angle of wind incidence and the motional aerodynamic derivatives.
It was observed that a pair of winglets placed directly above the leading and trailing edges of
the rectangular prism improved its aeroelastic characteristics. The winglets provided
important aerodynamic damping and reduced the sensitivity of the rectangular prism to
torsional aerodynamic instability. The experimental approach and a summary of the main
findings are presented.
r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Long-span-bridges; Aerodynamic winglets; Aerodynamic derivatives; 9:1 Rectangular prism

1. Introduction

A wind tunnel investigation was carried out on a dynamic flat-plate section model
in the 5 m vertical wind tunnel of the Aerodynamics Laboratory, Institute for
Aerospace Research, National Research Council Canada. The objective of the
investigation was to evaluate the effects of aerodynamic winglets fitted to a bluff
body as a means of improvement of the aerodynamic performance. The positive
effects of such a concept have been reported elsewhere for specific cases [1–3], but no
generic experimental verification exist. The idea for the work followed from recent

*Corresponding author.

0167-6105/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 0 2 ) 0 0 3 3 9 - 2
114 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

experiments carried out on a truss-girder bridge which had a lower-than-acceptable

stability limit. The solution was to install aerodynamic winglets which also served as
cover for the sidewalks. The winglets provided enough aerodynamic torsional
damping to transform the aerodynamic instability from a single-degree-of-freedom
(sdof) torsional instability to two-degree-of-freedom (2dof) flutter. The aerodynamic
stability limit was also increased by 30%.
The work described in this paper shows the benefits of this solution applied to a
basic flat-plate model (a 9:1 rectangular prism with sharp edges), and also provides
experimental data for comparison against numerical predictions. The aerodynamic
stability of the deck with and without winglets was determined as were the motion
aerodynamic derivatives for several angles of wind incidence. The influence of the
winglets on the static aerodynamic force coefficients was also studied.

2. Model detail and instrumentation

The two-dimensional rigid section model had a width B ¼ 0:51 m; depth D ¼

0:056 m; width/depth ratio B=D ¼ 9:1 and span length L ¼ 2:134 m: The model
consisted of an aluminum frame covered with plate aluminum skin and end plates
installed at both ends. The aerodynamic winglets were designed as thin metal sheets,
0:12B in width and supported by posts at a height of 1:5D above the deck and
extended the entire length of the bridge deck. Fig. 1 shows a photograph of the
model with winglets. The model was mounted in a dynamic rig, consisting of
relatively soft springs compared to the stiffness of the model, allowing the simulation
of the vertical and torsional oscillations of a section of the deck. Since the
aerodynamic derivatives are only a function of the geometry of the deck, no attempt
was made to tune the section model to properties of a particular bridge.
The motion of the model was measured with a pair of laser displacement
transducers mounted at each end of the model. The outputs from the four
transducers were summed and differenced to give heave and pitch model responses.
All tests were conducted in a wind speed range from 2 to 12 m=s in smooth flow.
Based on the width of the plate, this corresponded to a Reynolds number range of
0.7–4:0  105 : All edges were sharp.

3. Experimental procedure

The aerodynamic derivatives represent the motion-induced forces acting on an

oscillating structure immersed in a fluid in motion. They can be best described as a
representation of the aerodynamic damping and the aerodynamic stiffness provided
by the wind for a given deck geometry.
The method employed here to extract these unsteady aerodynamic properties is
known as the initial displacement technique and is described in [4]. It involves the
measurement in smooth flow of the decay in amplitude with time of an initial
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 115

Fig. 1. View of the section model in the 5 m vertical wind tunnel and nomenclature.

displacement of the deck in heave and pitch, for increasing windspeeds. A coupled
initial displacement of approximately 0:05B in heave and 21–31 in pitch was imposed
to the deck. Time histories of motion decay were recorded over several windspeeds.
Each test series at a given wind velocity was repeated at least 5 times.

4. Analysis

The motion-induced forces on a section model of unit length under coupled

vertical (heave:z) and torsional (pitch:a) oscillations can be expressed by the
following set of equations in Scanlan’s notation:
 
1 2 n Z’ n B’a 2 n 2 n z
L ¼ rU B KH1 ðKÞ þ KH2 ðKÞ þ K H3 ðKÞa þ K H4 ðKÞ ;
2 U U B

 
1 Z’ B’a z
M ¼ rU 2 B2 KAn1 ðKÞ þ KAn2 ðKÞ þ K 2 An3 ðKÞa þ K 2 An4 ðKÞ ;
2 U U B
116 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

where L and M are, respectively, the lift force and the pitching moment per unit
length, B the deck width, r is the air density, U is the mean wind speed, K the
n
reduced frequency ð2pfB=UÞ and Hi¼1;y;4  Ani¼1;y;4 are the normalized motion
aerodynamic derivatives for heave and pitch, respectively.
A similar set of equations can be expressed in dimensional form as per [4]. The
present analysis consisted in fitting the displacement signals in a least-squares
manner with this set of equations. Eight floating parameters, four for each
degree of freedom, were employed, and these were associated with damping and
frequency. The analysis was made with both the heave and pitch traces
simultaneously.
Subsequently, the system parameters were converted into normalized aerodynamic
derivatives by subtracting values determined from the still-air traces.

5. Results

Aerodynamic derivative tests and static force coefficient tests were conducted on
the bridge deck section with and without winglets.

5.1. Static coefficients

The mean lateral, vertical and torsional aerodynamic forces acting on the section
model for a given wind speed were determined experimentally and normalized to
produce static force coefficients. For the vertical forces and overturning moment, the
tests were conducted in a dynamic rig. For three windspeeds, the static heave
deflection and rotation were measured directly with optical sensors (laser) as for the
aerodynamic derivative tests.
Following a careful calibration to determine the static stiffness of the rig, the mean
displacement and rotation were converted to wind loads and subsequently
normalized to provide the lift coefficient in the wind axis frame of reference, CL ;
and the moment coefficient CM : The procedure was repeated for angles of wind
incidence from 101 to þ101 in increments of 21: For each case, the resulting angle of
wind incidence was calculated algebraically combining the nominal angle of
incidence at which the section model was set and the static rotation due to the
wind load. For the lateral forces, the test rig was modified so that the drag forces
could be measured directly with two strain-gauge force transducers placed
downstream of the model, one at each extremity of the model. The measured loads
were normalized to provide CD in the wind axis frame of reference. The static
coefficients were also converted to body-force coefficients CX (lateral) and Cz
(vertical).
All force coefficients were normalized by the deck width, B; and by B2 for the
moment coefficients.
The static force coefficients for the bridge deck with and without winglets are
shown in Fig. 2. The winglets have little effect on the drag and lift characteristics of
the prism when compared to the effect on the pitching moment. The winglets clearly
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 117

Cx = 2 X/ ρ U 2 DL C z = 2Z/ ρ U 2 BL CM = 2M/ ρ U 2 B 2 L
3 1 0.5

Pitching Moment Coefficient, CM

Without winglets
With winglets 0.4
2.5
Drag Coefficient, Cx

0.5
0.3

Lift Coefficient, Cz
0.2
2 0
0.1

1.5 0
-0.5
-0.1
1
-1 -0.2
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 -12-10 -8 -6 -4 -2 0 2 4 6 8 10 12 -12-10 -8 -6 -4 -2 0 2 4 6 8 10 12
α α α
Cd = 2D/ ρ U 2 DL Cl = 2L / ρ U 2 BL
3 1

2.5
Drag Coefficient, Cd

0.5
Lift Coefficient, Cl

2 0

1.5
-0.5

1
-1
-12-10 -8 -6 -4 -2 0 2 4 6 8 10 12 -12-10 -8 -6 -4 -2 0 2 4 6 8 10 12
α α

Fig. 2. Static force coefficients for the flat plate section model, with and without winglets, smooth flow.
The angle of wind incidence, a; is positive ‘nose up’.

increased the rate of change of the pitching moment coefficient with angle of
incidence near 01; from 1.9 per radian without the winglets to 3.3 with the winglets,
likely providing important aerodynamic damping. The lift coefficient slope showed
only a small change from 9.1 per radian without the winglets to 8.3 with the winglets
and the coefficients appeared to have a small offset.

5.2. Aerodynamic derivatives

The procedure described in Sections 3 and 4 was used to determine the motion
derivatives for the bridge deck with and without winglets. For all experiments the
aerodynamic derivatives were estimated for initial pitch amplitudes in the 21–31
range.
The influence of the winglets on the aerodynamic derivatives was studied for a
complete range of angle of attack of the model, i.e., 51 to þ51 in increments of 11
and the measured aerodynamic derivatives are presented in Fig. 3. Fig. 4 illustrates
this influence on selected derivatives for angles of wind incidence of 11; 01 and þ11:
The most significant effect of the winglets was to increase the aerodynamic
damping and torsional stability as indicated by An2 for all the angles of wind
incidence. The slope of the H2n curve, a coupling term, was also significantly modified
while H3n and An3 were not affected.
With winglets, no sign reversal of the An2 derivatives was observed for positive
angles except for the extreme angle of þ51: This indicated a marked improvement of
118 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

the torsional aerodynamic stability of the section model. Sign reversal of the An2
derivatives was observed for all tests without winglets.

5.3. Analytical prediction

It is of interest to verify if the observed changes of the moment coefficient slopes

are enough to explain the improvement of the torsional aerodynamic stability of the
section model with winglets.
Assuming quasi-steady aerodynamics, a prediction of the aerodynamic derivatives
can be made from the relationship between the rate of change of the static force and

0 10

-5 5
H1*

H2*

-10 0

-15 -5
0 5 10 15 0 2 4 6 8 10
0 0
+5°
-2 -10 +4°
H4*

H3*

+3°
-4 -20 +2°
0°
-6 -30
0 5 10 15 0 2 4 6 8 10
1.5 0.5

1 0
A1*

A2*

0.5 -0.5

0 -1
0 5 10 15 0 2 4 6 8 10
0.5 4

2
A4*

A3*

0
0

-0.5 -2
0 5 10 15 0 2 4 6 8 10
Reduced velocity (U/fhB) (a) Reduced velocity (U/fpB)

Fig. 3. (a,b) Aerodynamic derivatives for a 9:1 rectangular prism with winglets in smooth flow.
(c) Aerodynamic derivatives for a 9:1 rectangular prism without winglets in smooth flow.
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 119

0 10
H1*

H2*
-10 0

-20 -10
0 5 10 15 0 2 4 6 8 10
10 0
-5°
5 -4°
H4*

H3*
-20 -3°
0 -2°
-1°
-5 -40
0 5 10 15 0 2 4 6 8 10
1 1

0.5
A1*

A2*

0.5
0

0 -0.5
0 5 10 15 0 2 4 6 8 10
0.5 4

0 2
A3*
A4*

-0.5 0

-1 -2
0 5 10 15 0 2 4 6 8 10
Reduced velocity (U/fhB) (b) Reduced velocity (U/fpB)

Fig. 3 (continued).

moment coefficients with angle of wind incidence and the aerodynamic derivatives as
per [5] and Eqs. (1) and (2):
Cz0 Cm0 Cz0 Cm0
H1n E ; An1 E ; H3n E ; An3 E ð1Þ
2K 2K 2K 2 2K 2

Cz0 Cm0 An1 An3

H2n E ny ; An2 E ny ; nz ðKÞE ; ny ðKÞE ð2Þ
2K 2K H1n H3n

where Cz0 ; Cm0 are, respectively, the variations of the lift and moment coefficient with
angle of wind incidence in radians and nz;y are the points of application of the
aerodynamic forces expressed as a fraction of B:
120 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

0 10

5
H1*

H2*
-10
0

-20 -5
0 5 10 15 0 2 4 6 8 10
5 0
+5°
0 -10 +4°
H4*

H3*
+3°
-5 -20 +2°
0°
-10 -30
0 5 10 15 0 2 4 6 8 10
1 1

0.5
A1*

A2*

0.5
0

0 -0.5
0 5 10 15 0 2 4 6 8 10
0.5 4

2
A4*

A3*

0
0

-0.5 -2
0 5 10 15 0 2 4 6 8 10
Reduced velocity (U/fhB) (c) Reduced velocity (U/fpB)

Fig. 3 (continued).

Fig. 5 shows some of the measured aerodynamic derivatives compared to

approximations using Eqs. (1) and (2) for the model with winglets. Although there
is some agreement for H1n ; H3n and An3 ; it is clear that it would be erroneous to use
quasi-steady aerodynamics to predict H2n and An2 : This could be partly attributed to
the large variations of the point of application, i.e. center of pressure, of the
aerodynamic forces with reduced velocity and model configuration. The point of
application of the forces at high reduced velocity has a maximum value, ny ¼ 0:39
for the case without winglets and 0.62 with winglets. An attempt was made to predict
the variations of ny with reduced velocity from the torsional aerodynamic
measurements as described in [6]. The resulting ny differed importantly from the
ny of Eq. (2) indicating that coupling plays a role in the generation of the motion-
induced forces.
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 121

In Fig. 6, the approach described by Raggett [3] was used, where the aerodynamic
derivatives of the winglet are determined theoretically and then superimposed upon
those obtained experimentally for the bridge deck section. This is based on the direct
superposable principle of the motion-induced forces on the winglets to the motion
induced forces on the rectangular prism. Cobo del Arco et al. reported in [2] the use
of a similar technique to predict the effects of aerodynamic appendages on bridge
decks. However, their analysis did not include an experimental verification. Using
the relationships proposed by Raggett [3] to calculate the main damping
contribution of the winglets,

H1n ðKÞ ¼ b=BðU=fBÞ;

An2 ðKÞ ¼ br2 =B3 ðU=fBÞ

1.5 0

1 -5
H1*
A1*

0.5 -10

0 -15
0 2 4 6 8 10 0 2 4 6 8 10

0.5 10
Without winglets
With winglets
5
H2*
A2*

0
0

-0.5 -5
0 2 4 6 8 10 0 2 4 6 8 10

3 0

2 -10
H3*
A3*

1 -20

0 -30
0 2 4 6 8 10 0 2 4 6 8 10
Reduced velocity (U/fpB) (a) Reduced velocity (U/fpB)

Fig. 4. (a) Comparisons of selected aerodynamic derivatives for a 9:1 rectangular prism with and without
winglets in smooth flow for an angle of wind incidence of þ11: (b) Comparisons of selected aerodynamic
derivatives for a 9:1 rectangular prism with and without winglets in smooth flow for an angle of wind
incidence of 01: (c) Comparisons of selected aerodynamic derivatives for a 9:1 rectangular prism with and
without winglets in smooth flow for an angle of wind incidence of 11:
122 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

1.5 0

1 -5

H1*
A1*

0.5 -10

0 -15
0 2 4 6 8 10 0 2 4 6 8 10

0.5 10
Without winglets
With winglets
5

H2*
A2*

0
0

-0.5 -5
0 2 4 6 8 10 0 2 4 6 8 10

3 0

2 -10
H3*
A3*

1 -20

0 -30
0 2 4 6 8 10 0 2 4 6 8 10
Reduced velocity (U/fpB) (b) Reduced velocity (U/fpB)

Fig. 4 (continued).

and adding this to the derivatives reported in Fig. 3 for the prism without
winglets, a reasonable agreement was observed for the combined H1n and An2 at 01
angle of wind incidence. Raggett’s expressions appeared to overestimate the
contribution of the vertical aerodynamic damping. This is unlikely to affect the
aerodynamic stability. At low reduced velocity the predictions of the An2 is
satisfactory but at reduced velocity near instability, An2 obtained from Raggett’s
expressions showed a sign reversal while the measured An2 did not. The sign reversal
would point towards torsional instability while experimentally, flutter (2dof) was
observed. Predictions of the effects of the winglets based only on the main damping
terms, (H1n and An2 ), is thus not sufficient and the coupling effects need to be
considered.
An aeroelastic approach with two-dimensional theoretical analysis should be used
to describe the aerodynamics of the model with winglets and calculate all the
aerodynamic forces. In [7] a description of a bridge deck controlled by a pair of
winglets is given, see Fig. 7.
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 123

1.5 0

1 -5

H1*
A1*

0.5 -10

0 -15
0 2 4 6 8 10 0 2 4 6 8 10

0.5 10
Without winglets
With winglets
5

H2*
A2*

0
0

-0.5 -5
0 2 4 6 8 10 0 2 4 6 8 10

3 0

2 -10
H3*
A3*

1 -20

0 -30
0 2 4 6 8 10 0 2 4 6 8 10
Reduced velocity (U/fpB) (c) Reduced velocity (U/fpB)

Fig. 4 (continued).

The aerodynamic forces acting on the bridge deck and winglets can be described
by
h.0 þ m0 h’0 þ b0 h0 ¼ L0 þ L1 þ L2 ; ð3Þ

e1 L 1 e2 L 2
a. 0 þ s0 a’ 0 þ g0 a0 ¼ M0 þ M1 þ M2 þ ; ð4Þ
I I
where L0 ; L1 ; L2 denote the aerodynamic lift acting directly on the bridge deck, wing
nos. 1 and 2, respectively. M0 ; M1 and M2 denote the moments, and I is the mass
moment per unit length. These quantities are given by
L0 ¼ H1 h’0 þ H2 a’ 0 þ H3 a0 þ H4 h0 ; ð5Þ

M0 ¼ A1 h’0 þ A2 a’ 0 þ A3 a0 þ A4 h0 ; ð6Þ
 
rU 2 Bi dCL h’i nBi ai
Li ¼  ai   ; i ¼ 1; 2; ð7Þ
2m da U U
124 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

5 4

0 3
-5

A1*
H1*

2
-10
1
-15

-20 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
8 0.5
6
0.0
4
H2*

A2*
-0.5
0
-1.0
-2
-4 -1.5
0 2 4 6 8 10 0 2 4 6 8 10

0 4.5
-5 3.5
-10
2.5
H3*

A3*

-15
1.5
-20
-25 0.5

-30 -0.5
0 2 4 6 8 10 0 2 4 6 8 10
R educed velocity R educed velocity

n
Fig. 5. Variations of H1;2;3 and An1;2;3 with reduced velocity for a 9:1 rectangular prism with winglets in
smooth flow for an angle of wind incidence of 01: The dashed lines are predictions derived from quasi-
steady aerodynamics using the static force coefficients of Fig. 2 and n ¼ 0:39:

 
rU 2 B2i dCM h’i nBi ai
Mi ¼  ai   ; i ¼ 1; 2; ð8Þ
2I da U U
a1 ¼ am1 þ a0 ; a2 ¼ am2 þ a0 ;
h1 ¼ h0  a0 e 1 ; h0 þ a0 e 2 ;
where the aerodynamic forces L0 and M0 acting on the bridge deck are represented
by Eqs. (5) and (6). The aerodynamic forces L1;2 and M1;2 acting on the winglets are
given by Eqs. (7) and (8). The time-averaged point of application of the aerodynamic
forces, n, is treated here as a constant generally taken as the upstream quarter chord.
It has been shown in [3] that this set of Eqs. (3)–(8) can model adequately the
motion-induced forces on a bridge deck with stationary or actively controlled
winglets. This includes the effects of the winglets on the damping and coupling
motion-induced forces.

6. Conclusions

The aerodynamic influence of a pair of small winglets mounted above the top
surface of a 9:1 rectangular prism was studied experimentally.
 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126 125

-4

H1* -8

-12

-16

-20
0 2 4 6 8 10 12 14 16
Reduced velocity

0.4

0.2
A2*

0.0

-0.2

-0.4
0 2 4 6 8 10
Reduced velocity

with winglets R aggett (3)

Fig. 6. Variations of H1n and An2 with reduced velocity for a 9:1 rectangular prism with winglets in
smooth flow for an angle of wind incidence of 01: The dashed lines are predictions derived from
Raggett [3].

Static force coefficients and motion aerodynamic derivatives were measured for
several angles of wind incidence.
It was observed that the winglets were an important source of aerodynamic
damping without affecting the drag and lift characteristics of the prism.
Analytical predictions of the observed additional aerodynamic damping
based on quasi-steady assumptions did not correlate with the experimental
results.
Analytical predictions based on the superposition of the motion-induced forces of
the winglets to the motion-induced forces of the rectangular prism can yield
satisfactory results. However, the results suggest that all main damping terms
(An2 ; H1n ) have to be included as well as the coupling terms (An1 ; H2n ) in the
superposition of the effects.
126 M.G. Savage, G.L. Larose / J. Wind Eng. Ind. Aerodyn. 91 (2003) 113–126

Fig. 7. Bridge deck controlled by two wings.

References

[1] G. Diana, S. Bruni, A. Colina, A. Zasso, Aerodynamic challenges in super long span bridges design,
International Symposium on Advances in Bridge Aerodynamics, Copenhagen, Denmark, 1998.
[2] D. Cobo del Arco A.C. Bengoechea Aparicio, Some proposals to improve the wind stability
performance of long span bridges, Second European and African Conference on Wind Engineering,
Genova, Italy, 1997.
[3] J.D. Raggett, Stabilizing winglet pair for slender bridge decks, ASCE Sixth Annual Structures
Congress, Orlando, FL, 1987.
[4] N.K. Poulsen, A. Damsgaard, T.A. Reinhold, Determination of flutter derivatives for the great belt
bridge, J. Wind Eng. Ind. Aerodyn. 41–44 (1992) 153–164.
[5] E. Simiu, R.H. Scanlan, Wind Effects on Structures, Wiley, New York, 1986.
[6] H.P.A.H. Irwin, Wind tunnel and analytical investigations of the response of lions’ Gate Bridge to a
turbulent wind, National Research Council of Canada Report No. LTR-LA-94, 1977.
[7] H. Kobayashi, H. Nagaoka, Active control of flutter of a suspension bridge, J. Wind Eng. Ind.
Aerodyn. 41–44 (1992) 143–151.