WIND-INDUCED VIBRATIONS OF CANTILEVERED

TRAFFIC SIGNAL STRUCTURES

by
NARENDRA PULIPAKA, B.E., M.S.C.E.
A DISSERTATION
IN
CIVIL ENGINEERING
Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
Approved

December, 1995

1^ I
^

ACKNOWLEDGMENTS

I would like to express my sincere thanks and gratitude to my doctoral

committee chairman, Dr. James R. McDonald for his patience, and
encouragement. I sincerely appreciate his guidance, which not only helped
me in successfully completing this research, but also helped me in learning
American culture and professionalism.
I extend my sincere thanks to my advisory committee members Dr.
Kishor 0. Mehta, Dr. W. Pennington Vann, Dr. Partha P. Sarkar, and Dr. Walt.
J. Oler without whose guidance and input this research would not have been
possible. I express my deep sense of respect and gratitude towards Dr.
William P. Vann, whose courses I always enjoyed. His comments after my
seminars and on my dissertation manuscript were very valuable.
I would like to thank all the visiting researchers, especially Professor
Scanlan, whose comments and suggestions have helped me take the right
direction in my research.
I acknowledge the financial support from the Texas Department of
Transportation (TxDOT) for conducting this research. My special thanks are
due to many people who helped me in conducting experiments, especiallyByron Yeatts, Scott Funderburk, Robby Manalli, Allan Wolf, Kyle Eleuterius,
Frank Wyatt, Drex Little, Mario Torres, and Eric Hunzeker. I also thank all my
friends and colleagues, who helped me in many occasions and were very
supportive throughout my stay at Texas Tech.
I thank my parents and family members who have been always there,
when I needed encouragement and support. My special thanks are due to
my wife Uma, for her support and help-which she provided in spite of herself
being very busy preparing for her USMLE exams.

Lastly, I would like to thank all my professors and teachers for their
excellent instruction and patience in answering my questions. Without them, I
would not have achieved anything. I would like to dedicate this dissertation to
all my professors and teachers.

Ill

TABLE OF CONTENTS

ACKNOWLEDGMENTS

ii

ABSTRACT

viii

LIST OF TABLES

ix

LIST OF FIGURES -

LIST OF TERMS AND ABBREVIATIONS

xii

CHAPTER
1. INTRODUCTION
1.1 General

1
.

1.2 Description of the structure

1.3 Problem description1.4 Objectives

.
-

1.5 Organization of the dissertation

2. BACKGROUND LITERATURE 2.1 General

2.2 Vortex shedding

2.2.1 Mechanism -

2.2.2 Strouhal's number -

2.2.3 Effect of surface roughness

2.2.4 Unconstrained circular cylinder

2.2.5 Other research

2.2.6 Vibration mitigation techniques

2.3 Galloping

10

2.4 Summary

16

3. FINITE ELEMENT ANALYSES -

17

3.1 General

17

3.2 Objectives

17

IV

^v^r-

3.3 Finite element models

18

3.4 Fundamental frequencies -

19

3.5 Load versus strain -

20

3.6 Arm tip displacement versus maximum stress -

21

3.7 Risk of fatigue failure

21

3.7.1 Background

21

3.7.2 Endurance limit

22

3.7.3 Fatigue of signal structures

22

3.8 Conclusions from the finite element analyses

24

LABORATORY EXPERIMENTS

25

4.1 General

25

4.2 Tow-tank experiments

4.2.1 Objectives

4.2.2 Experimental research plan

25
27

28

4.2.3 Flow visualization tests

29

4.2.4 Vortex shedding results

33

4.2.5 Force measurements

33

4.2.6 Galloping results

35

4.3 Wind-tunnel experiments -

35

4.3.1 Objectives

37

4.3.2 Aerodynamic damping

37

4.3.3 Wind tunnel facility -

39

4.3.4 Effect of turbulence -

39

4.3.5 Model

40

4.3.6 Dynamic experiments

40

4.3.7 Wind-tunnel results -

41

4.3.8 Conclusions FIELD EXPERIMENTS

41
-

43

^^J!*^.

5.1 Objectives

5.2 Field research plan -

43

44

5.3 Facilities and instrumentation

44

5.3.1 Texas Tech field site

45

5.3.2 Test signal structures

45

5.3.3 Foundation

5.3.4 Instrumentation

46

48

5.3.4.1 Wind instruments -

48

5.3.4.2 Transducers

49

5.3.4.3 Displacement transducer-

53

5.3.4.4 Tilt meter

54

5.4 Field studies -

54

5.4.1 Structural characteristics

55

5.4.1.1 Stiffness

55

5.4.1.2 Natural frequencies

55

5.4.1.3 Damping

58

5.4.2 Galloping of the structure -

58

5.4.2.1 40-ft Signal structure

58

5.4.2.2 48-ft Signal structure

63

65

65

5.5.2 Size and location of wing -

66

5.5.3. Proposed wing

67

71

5.5 Mitigation measures

5.5.1 Damping plate (Wing)

6. ANALYSIS OF RESULTS AND CONCLUSIONS

6.1 Analysis of results

6.1.1 Equivalent SDOF model

6.1.2 Calculation of structural parameters

71
-

71

73

6.1.3 Aerodynamic force and eff. damping coeff.

VI

74

6.1.4 Calculation of effective damping -

11

6.2 Conclusions -

83

REFERENCES

85

VII

ABSTRACT

Under certain conditions of wind speed and wind direction, some

cantilevered traffic signal structures undergo large amplitude vibrations in
wind speeds of 10 to 20 mph. These vibrations are steady 'up and down'
motions perpendicular to the wind direction. The vibration amplitudes
sometimes reach as high as 2 ft from peak to peak. These large amplitude
vibrations may lead to fatigue failures, in addition to being a distraction to
passing motorists.
The three primary objectives of this research are: (1) to identify the
conditions required for large amplitude vibrations in cantilevered traffic signal
structures. (2) to understand the mechanism that produces the large
amplitude vibrations. (3) to develop strategies to mitigate vibrations in
cantilevered traffic signal structures.
Preliminary structural analysis was done using finite element models of
the typical traffic signal structures to obtain the fundamental frequency and
the stress levels reached due to static loading. A better understanding of the
across-wind vibration problem, which takes place in steady winds was
obtained by conducting tow tank experiments. The vibrations are attributed to
a galloping phenomenon, which primarily takes place when the wind is
blowing normal to the cantilever arm from the back side of a traffic signal
attached with a back plate. This configuration was found to have negative
aerodynamic damping from the wind tunnel experiments conducted using a
one quarter scale dynamic model.
The most effective mitigation measure was found to be a horizontal
wing attached above the signal light. A horizontal wing attached above the
signal light increased the aerodynamic damping to make the structure
aerodynamically stable.

VIII

LIST OF TABLES

3.1

Calculated frequencies of TxDOT signal

structures with different arm lengthsCalculated load versus strain

3.3

Calculated arm tip deflections,

moments, and stresses for various load cases- Measured strains, tilt, and displacements of
40-ft and 48-ft signal structures -

5.2

Measured frequency, and damping

IX

3.2

5.1

20

20

21
56

58

LIST OF FIGURES

1.1

Typical cantilevered traffic signal structure

1.2

3-light signal head without backplate

1.3

5-light signal head with backplate

2.1

Strouhal number versus Reynolds number

2.2

Effective angle of attack on an oscillating bluff body

2.3

Force coefficient in the Y-direction for a square shape

3.1

FEM model of 40-ft signal structure

3.2

FEM model of 48-ft signal structure

4.1

General view of the tow tank

4.2

Traffic signal configurations used in tow tank experiments

30

4.3

Tow carriage mounting configurations

31

4.4

Velocity, angle of attack, and force component definitions

32

4.5

Cpy versus angle of attack for traffic signal configuration-5

36

4.6

Wind tunnel setup -

4.7

Wing shapes and dimensions

4.8

Reduced velocity versus H*i

5.1

Details of the 40-ft signal structure

46

5.2

Details of the 48-ft signal structure

47

5.3

The possible orientations of the signal structure arm -

47

5.4

Rotatable foundation

5.5

A three-cup anemometer and wind direction

vane mounted on a 19-ft pole near the test structure

49

5.6

Locations where the transducers were mounted

50

5.7

Wheatstone half-bridge circuit

50

5.8

Wheatstone full-bridge circuit

51

5.9

Pre-amplification unit used to amplify the strain signal-

5.10

Displacement transducer -

11

14

18

19

26

37

40

42

43

52
53

5.11

Tilt meter

54

5.12

Static load versus strain

5.13

Free vibration of 40-ft structure without lights. Time Versus Strain

59

5.14

Free vibration of 48-ft structure without lights, Time Versus Strain

60

5.15

Spectrum of free vibration of 40-ft structure without lights

61

5.16

Spectrum of free vibration of 48-ft structure without lights

62

5.17

Proposed wing

68

5.18

Wind direction plot -

68

5.19

Wind speed plot

5.20

RMS of strain plot

6.1

Mode shape corresponding to the lowest freq. for 48 ft structure

72

6.2

Equivalent SDOF model

6.3

Aerodynamic force -

6.4

Three locations of the wing considered -

79

6.5

Effective damping coefficient versus wind speed

80

6.6

Effective damping ratio versus wind speed

81

57

69
70

XI

72

75

LIST OF TERMS AND ABBREVIATIONS

FEM

Finite Element Method

SDOF

Single Degree of Freedom

SMA-80

Single Mast Arm signal structure designed for 80 mph wind

load.

WING

A rectangular flat-plate fixed on the traffic signal structure to

reduce wind-induced vibrations.

XII

CHAPTER 1
INTRODUCTION

1.1 General
With the development of stronger materials, more accurate computer
analysis software, and new design methods, modern structures are becoming
more and more slender. These slender structures are prone to wind-induced
vibrations. In the design of these slender structures the dynamic effects of wind
loads sometimes are significantly large and cannot be neglected. The research
work presented here addresses wind-induced vibrations of cantilevered traffic
signal structures similar to the one shown in Figure 1.1

Figure 1.1 Typical cantilevered traffic signal structure.

1.2 Description of the structure
Cantilevered traffic signal light structures (see Figure 1.1) consist of a
vertical pole (17 ft to 20 ft high) and a horizontal arm (20 ft to 48 ft long), with
signal light heads attached. The pole and the arm are made of hollow
galvanized steel sections, which are either circular or octagonal in shape, and
have tapering diameters.

Figure 1.2 Three-light signal head without backplate.

Signal light heads are usually either a 3-light head with red, green, and
yellow signals (see Figure 1.2) or a 5-light head with red, green, yellow, green
arrow, and yellow arrow signals (see Figure 1.3). Some signal light heads have

fj- .>:,}*vs^-.

Figure 1.3 Five-light signal head with backplate.

a black colored plate (backplate) attached for better visibility of the signal lights.
The signal light head shown in Figure 1.3 has a backplate and five signal lights.

1.3 Problem description

Under certain conditions of wind speed and wind direction, some
cantilevered traffic signal structures undergo large amplitude vibrations in wind
speeds of 10 to 20 mph. These vibrations are steady 'up and down' motions
perpendicular to the wind direction. The vibration amplitudes sometimes reach
as high as 20-24 inches from peak to peak. These large amplitude vibrations
may lead to fatigue failures, in addition to being a distraction to passing
motorists.
One structure failed during the course of this study. A 48-ft cantilever
traffic signal structure failed in Dalhart, Texas in November 1991. The cantilever
arm fell to the ground when a crack developed at the connection between the
horizontal arm and the vertical pole. Fortunately, no one was injured by the
collapse.
The dimensions used in the analytical studies and the full-scale test
structures were selected to meet Texas Department of Transportation (TxDOT)
specifications. However, video recordings of signal structure vibrations in other
states in the United States prove that this problem is not limited to Texas.

1.4 Obiectives
The three primary objectives of this research are:
(1)

to identify the conditions required for large amplitude vibrations in

cantilevered traffic signal structures.

(2)

to understand the mechanism that produces the large amplitude

vibrations.

(3)

to develop strategies to mitigate vibrations in cantilevered traffic

signal structures.

The first objective involved identifying the signal light configurations that
contribute to excessive vibration, based on video recordings and field
observation. The second objective was accomplished by seeking a better
understanding of structural characteristics using finite element analyses, and
wind-structure interaction through a sequence of laboratory and field tests.
Once the phenomenon was understood, the third objective was to develop
strategies for vibration mitigation and to test them in the field.

1.5 Oroanization of the dissertation

This dissertation is divided into six chapters including this introduction.
The second chapter reviews literature on wind-induced vibrations of slender
structures. The third chapter discusses various analytical studies conducted
using finite element models of the traffic signal structures. In the fourth chapter
results from the tow tank and wind tunnel experiments are reported. Chapter 5
discusses the series of full-scale field experiments. The final chapter analyzes
results and presents conclusions.

CHAPTER 2
BACKGROUND LITERATURE

2.1 General
Long cylindrical structures of any bluff cross-section (circular, square, or
other) may exhibit wind-induced oscillations in a plane normal to the wind.
Neglecting the effects of wind gusts and the wakes of other bodies, this behavior
can be attributed to one of two aerodynamic mechanisms (Scruton, 1963): vortex
shedding or galloping.
There are many research publications in the area of vortex shedding and
galloping vibrations of structures. However, none of them deal with the
vibrations of cantilevered signal structures. Most of the available publications
deal with common shapes like circular, square, or some combination of these.
Dynamic flow past these simple shapes obviously is different from flow past the
more complex shape of a signal light.
Despite the differences in the dynamic flow, one can understand the basic
characteristics of fluid-induced vibrations by reviewing the available literature on
circular, square and other sections. Literature on vortex shedding and galloping
is reviewed in the following paragraphs.

2.2 Vortex sheddina

2.2.1 Mechanism
Vortex shedding in the wake of a cylindrical object is the periodic
formation of vorticity with alternating rotational directions caused by shear layers
on opposite sides of the object. A wake that exhibits the generation and
downstream convection of alternating vortices is referred to as a "von Karman
vortex street." The frequency of vortex shedding is dependent primarily upon
the cross-stream dimension of the body and the free stream velocity. There is
also a secondary dependence on the viscosity of the fluid that can be

characterized by Reynolds number. The degree of coherence, periodicity, and

two-dimensionality of the wake vortices depends on the aspect ratio of the body,
the mean flow Reynolds number, and the background turbulence in the free
stream flow. As a bluff body sheds vortices, the pressure on each side of the
object is alternately reduced and increased. The result is a periodic forcing
function normal to the free stream that can excite motion of the body, particularly
if the structure has one of its frequencies close to the vortex shedding frequency,
resulting in a condition of resonance. The periodic forces associated with the
vortex street are small relative to the mean drag forces.

2.2.2 Strouhal number

The vortex shedding frequency is characterized by the non-dimensional
Strouhal number
S, = (fD)A/

(2.1)

where f is the shedding frequency of wake vortices, D is the cross stream

dimension of the object, and V is the free stream velocity. Figure 2.1 depicts
Walshe and Wooten's (1970) summary of the variation of Strouhal number with
Reynolds number for a stationary cylinder of circular cross-section. Over a large
range of relatively low Reynolds numbers, two-dimensional wakes exhibit regular
vortex shedding at an almost constant Strouhal number approximately equal to
0.2. At intermediate Reynolds numbers near 1x10^, the vortex shedding
process becomes random and irregular and occurs at a higher Strouhal number.
At higher Reynolds numbers, the vortex shedding becomes regular again but the
characteristic Strouhal number is slightly higher (approximately 0.27) than in the
low Reynolds number regime. These low, intermediate, and high Reynolds
number regimes are referred to as the subcritical, critical or transcritical, and
supercritical regimes, respectively.

2.2.3 Effect of surface roughness

Achenbach and Heinecke (1981) evaluated the influence of surface
roughness on the vortex-shedding frequency of circular cylinders in the
Reynolds number regime of: 6x10^ < R < 5x10^- Cylinder roughness was
observed to delay the onset of the critical flow where vortex shedding becomes
irregular. In addition, surface roughness reduced the Strouhal number in the
critical flow regime. In the subcritical region the Strouhal number was nearly
constant at St= 0.205, which is approximately equivalent to the smooth cylinder
results. Therefore, variations in surface roughness cause a measurable, but not
significant, variation from the general cylinder depicted in Figure 2.1.
.6

-"r
Random shedding

^ .4
E

Regular shedding

r)
5 -3

Regular shedding

CO

2.
Crilical

.1

Supofcrilical

Subcntical
J

Transcritical
L_i.

8 c

6 J6

.6 .8

10

10
Reynolds Number - R

Figure 2.1

Variation of Strouhal number with Reynolds number

(Walshe and Wooten, 1970)

2.2.4 Unconstrained circular cvlinder

Jones et al. (1969) studied the effect of forcing a circular cylinder to
oscillate normal to the flow over a range of frequencies and amplitudes:
0.06 < (fhD)A/ < 0.5
0.014 < (ho)/D < 0.083

10

In their nomenclature, fh is the forcing frequency, ho is the amplitude of the

forced oscillation, D is the previously defined characteristic diameter, and V is
the flow velocity. The results are summarized by them as follows:
Oscillation of the cylinder in the lift direction has no
significant effect on the mean drag coefficient. An
unsteady lift due to cylinder motion, which increases
with amplitude of motion, exists only when the
cylinder is oscillated at or relatively near the
aerodynamic Strouhal frequency for the stationary
cylinder. This lift is a negative (destabilizing)
aerodynamic damping force at cylinder frequencies
below the stationary-cylinder Strouhal frequency. As
the cylinder frequency is increased through and
above the Strouhal frequency, there is an abrupt
change to a positive (stabilizing) aerodynamic
damping force. (Jones et al., 1969)
In other words, cylinder motion tends to amplify the forces associated with
vortex shedding and to facilitate a convergence of the cylinder's frequency of
oscillation with the vortex shedding frequency for a given wind velocity. The
resonance effects are strongest in a small bandwidth around the stationary
Strouhal frequency with the greatest amplification of lift occurring when the ratio
of forced frequency to Strouhal frequency is 0.99. However, if the Strouhal
frequency is much different from the natural frequency of the structure, the
degree of constraint will not have much effect on the Strouhal number.

2.2.5 Other research

Szepessy and Bearman (1992) investigated the effect of using end plates
on circular cylinders. By varying the ratio of open length (cylinder span length
between end plates) to cylinder diameter, three-dimensional effects were
controlled in an effort to observe their effect on fluctuating lift, drag and shedding
frequency. In addition to experiments on circular cross-sections, many
researchers have published results for square or rectangular sections. Okajima
(1982) obtained the Strouhal number for rectangular cylinders with side-to8

height ratios (b/h) of 1.0 to 4.0 over a range of subcritical Reynolds numbers.
For the square cylinder (b/h = 1.0), the Strouhal number was approximately
constant at 0.13 for Reynolds numbers of 1x10^ to 2x10\ For rectangular
cylinders (b/h of 2.0 and 3.0), Okajima observed a transitional Reynolds number
where the flow separated at the leading edges with periodic reattachment on the
lower or upper surfaces synchronized with the vortex shedding. At lower
Reynolds numbers, the flow separated at the leading edges and did not reattach;
at higher Reynolds numbers, the flow continuously reattached. The
corresponding Strouhal numbers varied from 0.13 for the fully separated flow to
0.17 with the fully attached flow.

2.2.6 Vibration mitiaation techniques

Scruton (1963) dealt with practical methods of reducing wind effects on
structures and tall stacks. The effects of vortex excitation can be suppressed by
reducing the periodicity of vortex formation or by reducing the span wise
cohesiveness of the flow, that is, by forcing the flow to be three-dimensional.
Examples of devices meant to accomplish this goal are a perforated cylinder
mounted outside a plain cylinder (Price 1956), triangular spoilers on a
suspended pipe-line (Baird 1955), and helical strakes on vertical stacks,
suggested by Scruton and Walshe (1957) for which Woodgate and Meybrey
(1959) determined the optimal configuration.
Roshko (1955) studied flow past a circular cylinder at high Reynolds
number. For Reynolds numbers from 3x10^ to Ixio"^ regular vortex shedding
from a simple cylinder occurred at a Strouhal number (St) of 0.27. Figure 2.1
depicts this result as the supercritical region where regular shedding has
resumed. Roshko also noted that a splitter plate attached to the trailing edge of
a circular cylinder tended to suppress the vortex shedding. It is possible that the
splitter plate allows the flow to remain more nearly attached to the cylinder much

in the same way that the trailing portions of an airfoil aid it in maintaining
attached flow.
In addition to the above discussed techniques, many damping techniques
(both active and passive) have been successfully used to mitigate the vortex
shedding vibrations.

2.3 GalloDinc
When a structure vibrates normal to the free stream flow, the flow in turn
oscillates relative to the moving structure. This oscillating flow induces an
oscillating aerodynamic force on the structure. If that aerodynamic force tends
to diminish the vibrations of the structure, then the structure is said to be
aerodynamically stable. If that aerodynamic force tends to increase the
vibrations, the structure is said to aerodynamically unstable.
Figure 2.2 shows a bluff body in steady flow. The free stream velocity V
of the flow is horizontal (from left to right). Assume the bluff body is moving
down with a velocity y, then the air surrounding it will have a relative velocity of
y upwards. The vector sum of free stream velocity and upward air velocity
relative to the bluff body (because of downward movement of the bluff body), is
Vrei. This relative wind velocity Vrei hits the bluff body with an angle of attack a
with respect to the free stream velocity.
The lift and drag forces caused by the relative wind velocity Vrei are given
by:
L(a)=jCL(a)pv2|A

(2.2)

D(a)=^CD(a)pv2|A

(2.3)

where p is the density of the air, CL(a), and CD(a) are lift and drag coefficients for
an angle of attack a of the incident flow, and A is the characteristic area of the

10

Lift force (L)

caused by the
relative wind

Aerodynamic force (Fy) either helping the

vibration or mitigating the vibrations
Drag force (D) caused by
the relative wind

Bluff body moving

V down with a velocity
of y

Figure 2.2 Effective angle of attack on an oscillating bluff body

bluff body. The force component in line with the direction of oscillation of the
body (Fy) can be obtained as :
Fy=L(a)Cos(a)-i-D(a)Sin(a)=-CFypV^A

(2.4)

where a is the angle of attack, which in turn is given by:

a = tan"^(-^)

(2.5)

A bluff body moving downward causes positive angles of attack as shown

in the Figure 2.2. By substituting free stream velocity V = V^ei Cos(a), and the
equations 2.2 and 2.3 into Equation 2.4, the force coefficient Cpy can be written
as:
Cpy = [CL(CX) + C D(a)tan(a)] sec(a)

(2.6)

If the force Fy is negative it acts downwards, i.e., in the same direction as

the movement of the bluff body, which, in turn, means the aerodynamic force is
trying to help the structure vibrate. Under these conditions the structure is
aerodynamically unstable.
If the body has mass m per unit length, is elastically sprung, and has
linear mechanical damping, its equation of motion can be written as:
11

m[y+2Ccoy-hco^y] = -Fy

(2.7)

where ^ is the damping ratio, co is the natural circular frequency, y is the

displacement of the bluff body downwards, y is the velocity of the bluff body
(positive downwards), y is the acceleration of the bluff body (positive
downwards), and Fy is the aerodynamic force (positive when acting upwards on
the body). If the mean aerodynamic lift and drag coefficients are assumed to be
the same for a fixed body and an oscillating body, Fy is given by Equation 2.4,
and Cpy is given by Equation 2.6.
Considering small amplitude incipient motions, that is, when the velocity
of the bluff body (y) is close to zero, wherein
a = (yA/)=0

(2.8)

For this condition

Fy =

raPy^
V^OCy

(2.9)

a=0

which leads to the examination of the force coefficient (dCpy/da), which can be
found by differentiating Equation 2.6.
For small amplitudes of vibration the equation of motion can be written as
m[y-f- 2Ccoy+ co^ y] = - - p V ^ A ^ ' ^ J ' ^

(2.10)

Since the aerodynamic force on the right hand side of Equation 2.10 is a
function of the velocity of the bluff body, similar to the mechanical damping term
on the left hand side, this term is known as aerodynamic damping. The overall
damping coefficient for the structure vibrating in a steady flow can be written as
2mCco-h-pVA - ^
V da /
2

=d

(2.11)

where the first term is the mechanical damping of the structure and second term
the aerodynamic damping. The system is aerodynamically stable if d > 0 and

12

aerodynamically unstable if d < 0. Since the first term (mechanical damping) is a

positive term, instability will occur only if
'dCpy^
^ da

<0

(2.12)

This is the as well-known Den Hartog's criterion (1956), a necessary

condition (the sufficient condition is d < 0) for galloping instability.
Den Hartog (1932) recognized that certain cross-sectional shapes exhibit
galloping oscillations at near constant frequency over a wide range of wind
speeds with oscillation amplitudes that increase as a function of wind speed.
Simple vortex shedding could not account for this observed behavior. His study
dealt with transmission lines with non-circular cross-sections due to ice
accumulation. Circular cross-sections do not develop steady lateral forces
perpendicular to the free stream because of symmetry at all angles of attack.
However, non-circular shapes produce aerodynamic forces that are a function of
angle of attack. Unlike the oscillations of circular cylinders which occur only at
resonance with the periodic shedding of wake vortices, sections meeting Den
Hartog's (1956) criterion exhibit a pure plunging motion perpendicular to the free
stream at an amplitude that increases with wind speed.
As illustrated in Figure 2.3, a square cross-section exhibits this behavior
for a < 14 degrees. Consider an initial disturbance of the cylinder that results in
a downward motion. As a result of the cylinder motion, the relative velocity
between the fluid and cylinder includes a cross-stream component giving a
positive angle of attack. From Figure 2.3, it is noted that the aerodynamic force
on the cylinder at a positive angle of attack is negative and most importantly is
acting in the same direction as the cylinder motion. As the structure decelerates
during downward motion, the angle of attack and aerodynamic force decrease.
Once the downward motion ceases and the structure starts moving upwards,
angle of attack also reverses. The deflected structure moves past the neutral
position with maximum velocity where the angle of attack also is maximum.
13

Because the aerodynamic force is aligned with the motion, the structure will
overshoot the neutral position and the motion will continue in a self-perpetuating
manner. The angle of attack increases as vibration amplitude and translational

y
,

"

'

-.4

-.6

10

20

15

25

Angle of Attack a - Degrees

Figure 2.3 Force coefficient in the Y-direction for a square shape

velocity increase. The corresponding aerodynamic forces increase as well until

the angle of attack reaches approximately 15 degrees. At higher angles of
attack the aerodynamic forces become stabilizing. Therefore, the amplitude of
the galloping motion is limited to the negatively damped angles of attack only.
However, the amplitude increases as the free-stream velocity increases.
The Den Hartog criterion is not satisfied and galloping behavior does not
occur for all cross-sectional shapes. For example, a circular cylinder
experiences a pure drag with respect to the relative velocity vector.
Consequently, the transverse force is positive for positive angles of attack and
always acts in the opposite direction of the cylinder motion, thereby providing
positive aerodynamic damping. Similarly, a symmetrical airfoil produces a lift
force with respect to the relative velocity vector that is also positive for positive

14

angles of attack. Any deviation from the statically neutral position always results
in an aerodynamic force opposite to the disturbed motion.
Parkinson and Brooks (1961) presented a nonlinear analytical model
using aerodynamic coefficients to explain the galloping motion of cylinders.
They refer to their model as "quasi-steady" because of the assumption that
steady-state aerodynamic data can be applied to a dynamic equation of motion.
In Parkinson and Brook's experiment, models of spring-mounted cylinders
began plunging when resonance existed between the system's natural frequency
and the stationary vortex-shedding frequency of the cylinder. Then, as predicted
by their theory of galloping, sections that satisfied the Den Hartog criterion would
oscillate at velocity dependent amplitudes. Parkinson and Smith (1962)
improved the accuracy of the analytical model by using higher order polynomials
to approximate the aerodynamic coefficients of a square section more
accurately. Parkinson and Wawjonek (1981) noted that the quasi-steady theory
of galloping did not adequately describe experimental results for cylinders at
velocities near the vortex shedding lock-in velocity. A nonlinear theoretical
explanation of the interaction between vortex shedding and incipient galloping
was presented by Obasaju (1983).
Whereas Parkinson and Wawjonek investigated the free vibration of
cylinders in which only amplitudes were measured, Obasaju used a forcedvibration study of a square cylinder to determine the dependence of massdamping parameter on amplitude ratio and reduced velocity,
Furthermore, Obasaju linearized the forcing function used in the quasisteady theory of galloping by making an assumption of small amplitude
oscillations, and compared the resulting behavior to that predicted by vortex
shedding alone. He concluded that near the resonant reduced velocity, a
spring-mounted cylinder can perform either a high amplitude oscillation at the
vortex shedding frequency or small-amplitude galloping. Furthermore, for
certain mass-damping parameters, this response can continue somewhat above
15

the resonant velocity. Finally, although galloping is never observed below the
resonant velocity for vortex shedding, the quasi-steady theory of galloping by
itself fails to make this prediction. In Parkinson and Brooks (1961) non-linear
quasi-steady theory, the minimum speed for onset of galloping is a function of
structural damping and negative aerodynamic damping but does not take into
account the critical or lock-in velocity due to vortex shedding. Therefore, an
advantage of incorporating Obasaju's forced-vibration results is that it correctly
predicts that galloping cannot occur below vortex resonance. Also, the
seemingly chaotic behavior of cylinders having an incipient galloping velocity
greater than but very near to the vortex lock-in velocity is explained.
Novak (1969, 1972) provided the most mathematically detailed and
comprehensive development of the quasi-steady theory of galloping. Using
early theoretical and experimental results as a foundation, he successfully
provided a rigorous and complete description of the quasi-steady theory of
galloping. With the exception of Obasaju's more recent linearization refinement,
Novak's theoretical development parallels and is inclusive of all published work
and at the same time is more detailed. Even more refined are his closed-form
equations generalized for any aerodynamic cross-section and many degrees of
freedom.

2.4 Summary
The literature on vortex shedding and galloping phenomena were
reviewed. Vortex shedding vibrations occur under a narrow lock-in wind speed
range. It is the galloping phenomenon which best explains the vibrations of the
cantilevered signal structures in a wide range of winds. The nonlinear nature of
the aerodynamic forces on most shapes results in a limit to the amplitude of
galloping oscillation at a given wind speed. However, as velocity increases, so
does the limit-amplitude until, in the instance of the square at a = 15 degrees,
the aerodynamic forces resulting from displacement become stabilizing.

16

CHAPTER 3
FINITE ELEMENT ANALYSES

3.1 General
The first set of finite element analyses of signal structures was based on
current TxDOT drawings and specifications (TxDOT, 1990). Single Mast Arm
structures designed for 80 mph wind (SMA-80) with cantilever arm lengths of 20,
24, 28, 32, 36, 40, 44, and 48 ft were analyzed.
Later it was found that the dimensions of the test structures used in the
full-scale field testing did not exactly match the TxDOT specifications. A second
set of finite element analyses was conducted using the actual dimensions and
weights of the test structures as measured in the field. These measured
dimensions were used so that comparison could be made between the field
measurements and the finite element method results.

3.2 Obiectives
The primary objectives of the analyses were:
(1)

To estimate fundamental frequencies of vibration of typical signal

structures, including the ones tested in the field. The objective here was
to define the range of fundamental frequencies of signal structures with
different cantilever arms.

(2)

To obtain a relationship between a static load applied near the cantilever

arm tip and strain at strain-gage locations on the signal structure used in
the field experiments. The aim here was to compare the calculated
strains (using FEM) with the strains measured in the field experiments.
The FEM models in this analysis were based on the arm and pole
dimensions of the test structures as measured in the field.

(3)

To obtain a theoretical relationship between arm tip displacement and the

stresses in the signal structure for the purpose of evaluating potential
17

fatigue effects. The FEM model was based on a 48 ft long arm and
TxDOT specifications.

3.3 Finite element models

The cantilevered traffic signal structure, which is a statically determinate
structure, can be modeled with beam elements. Because both the vertical pole
and the cantilevered arm are tapered, a relatively fine mesh is needed to
represent the mass and stiffness variations. A commercially available FEM
software package, Stardyne for Windows, Version 4.0 (Titan Corp., 1993) was
used to perform the static and dynamic analyses. FEM models of 40-ft and 48-ft
signal structures are shown in Figures 3.1 and 3.2, respectively. The
dimensions of the test structures and the locations of the signal light heads on
the arm are shown in Figures 5.1 and 5.2. Two types of signal light heads were
used: (1) a 5-light signal head which weighs about 90 lb. at the tip of the arm,
and (2) a 3-light signal head which weighs about 55 lb. at a point 12 ft. from the
tip of the arm.

Signal Light Masses

Calculated
fundamental
frequency = 0.87 Hz

Node Numbers

Fixed Boundary Condition

:

Figure 3.1 FEM model of 40-ft signal structure

18

3.4 Fundamental frequencies

The two signal structures tested in the field were first analyzed by the
FEM using the Stardyne software. The purpose was to determine the natural
Signal Light Masses
51

Node Numbers

Fixed Boundary Condition

Calculated
fundamental
frequency = 0.81 Hz

Figure 3.2 FEM model of 48-ft signal structure

frequencies of vibration and mode shapes. Models were also developed and
analyzed for the SMA-80 standard TxDOT signal structures with different
cantilever arm lengths. These models were based on published geometry and
weights.
Table 3.1 summarizes the results of the FEM analyses. The table gives
the calculated fundamental frequencies (first mode) for standard TxDOT signal
structures with different cantilever arm lengths. The calculated fundamental
frequencies for the two structures tested in the field (see Figures 3.1 and 3.2)
were compared with the frequencies measured in the field (see Table 5.2).
Using the measured dimensions and actual weights of the structures, the field
and calculated values agree very well. Natural frequencies estimated using the
standard dimensions may not always give correct results.

19

Table 3.1 Calculated frequencies of TxDOT signal

structures with different arm lengths
Arm Length (Feet)
20
24
28
32
36
40
44
48

Freqijency (Hertz)
1.70
1.56
1.35
1.23
1.10
0.96
0.88
0.81
3.5 Load versus strain

The purpose of this study was to validate the FEM model with field
measurements. Static concentrated load increments were applied at a location
near the free end of the cantilever arm. The FEM analysis was then used to
determined the strains at gage locations on the structure where the strain gages
were mounted for the field tests. If the FEM model gives reasonable agreement
between load and strain, then the model can be used with confidence in
calculating other parameters, such as stress and displacement. Table 3.2
presents the results of the load versus strain calculations.
Table 3.2 Calculated load versus strain
Load, Ib^
0
50
60
70
80
90

Strain on vertical Dole in u strain

40-ft structure^ 48-ft structure^
0
0
49.6
31.5
59.3
37.8
69.4
44.1
79.4
50.4
56.7
89.3

Load applied 3 feet from tip of arm.

2

Strains located 13 in. above top of base plate.

20

3.6 Arm tip displacement versus maximum stress

The relationships between arm tip displacement and maximum stresses in
the vertical pole and the cantilever arm were needed to study fatigue effects
from cyclic loading. The endurance limit is a function of mean stress, stress
range and number of loading cycles. If we have a relationship between tip
displacement and maximum stress in the signal structure, conclusions regarding
potential fatigue failure can be drawn by knowing the vibration amplitude.
Table 3.3 presents the relationships between tip displacement of the arm
versus the maximum stresses in the arm and the pole.
Table 3.3

Calculated arm tip deflections, moments and stresses for various

load cases^
Arm
tip
Defl.
(in.)

Moment at pole
base and
arm connection
(in. X lb. X 10'^)

Stress at
base of
Arm
Pole
(ksi)
(ksi)

15.4

2.142

7.28

11.09

Case 1 -f- two signal^

lights (80 lb. and 50 lb.) 21.7

2.811

9.52

14.55

Case 2 with forced arm

tip deflection of

25

3.131

10.60

16.21

Case 3

28

3.42

11.58

17.73

Case 3

31

3.72

12.56

19.26

Case
1
2
3

Load
Dead load of pole and
arm only

Values are calculated for 48-ft SMA-80 signal structure

^In Case 2 stresses and deflection are due to weight of the structure and signal
lights.
3.7 Risk of faticue failure
3.7.1 Backcround
Materials under repeated loading and unloading, or under reversal of
stress, fail at stresses smaller than the ultimate strength of the material under
static loads. As the cycles of stress reversal increases, the magnitude of stress

21

at which the failure occurs decreases. This phenomenon of decreased strength

of materials under cyclic loading is called fatigue.
If cTmax and amin are the maximum and minimum values of the repeated
stress cycle, then the algebraic difference
R = CTmax - CJmin

(3.1)

is called the range of stress. The cycle is completely defined if the range and
maximum stress are given. The average stress is
Gm = 0 . 5 (Gmax + amin)

(3.2)

3.7.2 Endurance limit

The endurance limit is the maximum stress at which the material does not
fail even after infinite number of complete stress reversals (i.e., am= 0). The
endurance limit for mild steel is about 27 ksi. The endurance limit for the
galvanized steel used for the traffic signal structures should be at least 20 ksi.

3.7.3 Fatigue of sional structures

When a traffic signal structure is not vibrating, the structure is stressed
due to its self weight and the weights of the signal lights. As the signal structure
starts vibrating, it undergoes cyclic stress variations about the dead load stress.
The mean stress caused by dead load am is about 15 ksi in the 48-ft
structure arm with lights (see Table 3.3). The maximum stress amax reached
when the arm tip moves down 9 inches (in addition to the dead load deflection)
is about 20 ksi. For this amplitude of vibration, from Equations 3.1, and 3.2
amin

^ a m " amax

2 x 1 5 - 20
10 ksi
R

amax " amin

20- 10
10 ksi
22

This range of stress 10 ksi variation, which is much less than the
endurance limit of 20 ksi, should not cause any fatigue failures. Back calculating
the amplitude of vibration needed for the stress variation range (R) to be equal
to the endurance limit
R

amax-amin

am

0 . 5 (amax + amin)

20 ksl

(3.3)

we know

15 ksi
So,

amax + amin

30 ksi

(3.4)

from 3.3, and 3.4

25 ksi

amax
amin

5 ksi

Table 3.3 shows the stresses caused by arm tip deflection up to 9 inches
(in addition to the dead load deflection). But, by extrapolating the values in
Table 3.3, the amplitude of vibration needed to produce a maximum stress (amax)
of 25 ksi can be calculated as about 20 inches.
The sample calculations suggest that the fatigue failure of traffic signal
structures is not an issue unless the amplitudes of vibration are very high ( 20
inches for 48-ft arm structure, or arm length to amplitude ratio of about 30). It
should be noted that the above calculations did not take stress concentration at
the junction of the arm to pole into account. Due to this stress concentration, the
fatigue failure of signal structures can be a concern even at lower amplitudes of
vibration. Besides stress concentration, hair line cracks present in the structure
can also lead to fatigue failures under the vibration amplitudes much smaller
than the above calculated amplitudes. The cracks can be caused in many ways,
some of which are: (1) manufacturing defects, (2) improper handling during
erection, and (3) an automobile hitting the signal structure.

23

3.8 Conclusions from the finite element analyses

The fundamental frequencies of traffic signal structures with different
cantilever arm lengths were calculated using the finite element method, which
gave an idea of the range of fundamental frequencies of these signal structures
(see Table 3.1). Also the fundamental frequencies of the test structures used in
the field study were calculated (see Figures 3.1 and 3.2).
Bending strains at strain-gage locations on the signal structure used in
the field experiments due to a static load applied near the cantilever arm tip and
were calculated (see Table 3.2).
Stresses in the signal structure due to different loading conditions were
calculated (see Table 3.3) and the risk of fatigue failures was briefly addressed.
Sample calculations suggest that fatigue failures should not occur unless the
amplitudes of vibration reach very high values ( 20 inches for 48-ft arm
structure). However stress concentration at the connections, type and quality of
the welding, and other manufacturing defects can lead to fatigue failures at
smaller amplitudes of vibration. The main objective of this research is focused
towards identifying the mechanism that produces vibrations and proposing
methods to mitigate them-fatigue behavior of these structures was not studied in
depth.

24

CHAPTER 4
LABORATORY EXPERIMENTS

4.1 General
The experiments conducted for this research used three testing
environments: (1) Tow tank, (2) Wind tunnel, and (3) Field testing. The first two
are discussed in this chapter, and the full-scale field tests are discussed in
Chapter 5.
The tow-tank experiments were conducted by Dr. Walter Oler and Mr.
Steve Cook of the Mechanical Engineering Department of Texas Tech
University. Their experiments are briefly discussed here. The results from the
tow-tank experiments were used to design the field tests conducted on full-scale
signal structures. For a more detailed discussion of the tow-tank studies, see
McDonald et al. (1995)

4.2 Tow-tank experiments

The tow tank permitted the testing of a full-scale signal arm and traffic
light using water instead of air, which has the advantage of working with flow
speeds less than 10 feet per second. An actual traffic signal light along with a
portion of the arm were towed through the tow tank to simulate wind speeds in
the range of 10 to 30 mph. The resulting flow patterns were observed by
injecting dye into the water and recording on video tape. Visualization
experiments were designed to detect vortex shedding. A load transducer
capable of measuring forces in three coordinate directions determined
equivalent wind forces on the signal light arm assembly as it was towed through
the tank. These measured forces were used to check if a signal light
configuration is susceptible to galloping vibrations.
Shown in Figure 4.1, the Texas Tech Tow Tank consists of a belowground water tank with overall dimensions of 80 ft long by 15 ft wide and 10 ft
25

^^^^"

ijJU.iHj<tiWWgH_Jli^

1 ^W7^'f^^'^^'""*

' .tffp.i .-ran.

7 ^-:I^^

t*i
/

Figure 4.1 General view of the tow tank

26

deep. Steel rails mounted above the water on either side of the tank support a
motorized towing carriage which is used to propel a variety of models through
the tank. The carriage also supports the computerized motion control and data
acquisition systems and has adequate space for the operator and several
observers. The carriage and supported models can be tested at preprogrammed
speeds of 0-5 ft/s with the capacity for accelerations up to 2 ft/s^. Data
acquisition systems include a high speed pressure measurement system, load
cells for force and moment measurements, and several options for video taped
flow visualization both above and below the water surface.
In general, the tow tank can be used for any incompressible aerodynamic
study which might more typically be conducted in a wind tunnel. The 16:1 ratio
of kinematic viscosity of air to water allows tests of a specific model to be
conducted at 1/16 the speed of an equivalent wind tunnel test at the same
Reynolds number. This speed differential is particularly advantageous in flow
visualization experiments involving time dependent model motion or transient
phenomena. Flow visualization in water based experiments (tow tank, water
table, etc.) is better than that of wind tunnel experiments because color dyes do
not diffuse in water as rapidly as smoke disperses in the air. The tow tank is
particularly suited to tests involving transient model velocities which are virtually
impossible to duplicate in a wind tunnel. The Texas Tech Tow Tank has been
successfully applied to research programs concerned with parachute,
automotive, and wind turbine aerodynamics as well as the aerodynamic loading
on solar receivers and traffic signal lights.

4.2.1 Obiectives
The primary objective of the tow tank experiments was to determine the
source of the aerodynamic force causing vibrations. The two most likely
candidates were vortex shedding and galloping. Separate experimental
programs were developed to evaluate the significance of each of these
27

phenomena. A flow visualization experiment was utilized to detect the

occurrence of vortex shedding and to identify the shedding frequency. The
potential for galloping of the traffic signals was evaluated by making steady state
aerodynamic lift and drag measurements and applying the Den Hartog (1956)
criterion for negative aerodynamic damping. Both the flow visualization and
force measurements were performed for variety of signal light configurations as
shown in Figure 4.2.

4.2.2 Experimental research plan

A full-size, three-light traffic signal head was mounted on a 6 5/8 in.
diameter steel pipe in several configurations. The steel pipe simulated the
shape and size of the horizontal cantilever arm which supports the traffic signals
over road intersections. Figure 4.2 depicts the cross-sections of the test
configurations. Configurations 1 through 4 are traffic signals mounted without a
back plate; Configurations 5 through 8 are traffic signal options mounted with a
51 in. x 23 in. rectangular back plate. The back plate, which is black, provides a
contrasting background for the signal lights, when viewed toward the sun.
The signals were tested with flow from both the front and back. Hence,
the configurations were matched pairs (1&2, 3&4, 5&6, 7&8) where the only
difference was in the direction of the relative flow used in the experiment. In
Configurations 1, 2, 5 and 6 the signal heads were hung below the horizontal
arm. In Configurations 3, 4, 7 and 8, the signal heads are mounted in front of
the horizontal arm. Because the blockage ratio of the traffic signal and arm
relative to the tow tank cross section was low, the free stream flow field was
assumed to be essentially two-dimensional. However, the flow near each end of
the traffic signal was three-dimensional. This characteristic was investigated
through the flow visualization experiment. None of the experiments considered
flow parallel to the axis of the supporting arm.

28

Figure 4.3a shows how the arm-mounted traffic signal was attached to the
towing carriage. Note that the pipe arm is mounted to the bridge and extends
down into the water. A second mounting configuration for flow visualization,
Figure 4.3b, used an extension to place the signal underneath the bridge where
clear video tapes could be made through a Plexiglas viewing plane.

4.2.3 Flow visualization tests

To determine the significance of vortex shedding from traffic signal heads
and the corresponding Strouhal numbers, dye was injected at up to twenty
locations around the mounting pipe and signal head. The dye ports were turned
on in various combinations and video tapes were made of the resulting flow field.
Small-scale turbulence features were most apparent when dye was injected
close to the pipe and traffic signal. Conversely, the primary vortex frequency
was more easily observed when dye was injected approximately two
characteristic diameters downstream of the model (characteristic diameter is the
largest cross-stream dimension of the model). Furthermore, by placing dye ports
in the middle and on the ends of the traffic signal, three-dimensional features of
the flow could be investigated. Repeated experimental runs and dye port
adjustments were made for Configurations 1 through 8, to obtain the best
visualization of the vortex features. Over three hours of video tape recordings
were made during the flow visualization portion of the experiment. When a dye
port location successfully captured periodic flow reversals which could be
studied to obtain the vortex shedding frequency, the video tape of that
experimental run was analyzed frame-by-frame on a video editor. The number
of video frames between flow reversals is proportional to the period of that flow
feature and can be used to compute the Strouhal number

29

(a)

^ X

Configuration 1

(b)

Configuralion 2

o
(j.)

(e)

QxiTiguration 3

(d)

Configuration 4

Coofiguratioa 5

(0

Connguratioa 6

o
(g)

..I

Configuralion 7

(h) Coofiguratioa 8

Figure 4.2 Traffic signal configurations used in

tow tank experiments

30

t
d

O)

(/)
Q)

.-

i_
n
o '
o

.nnn

(0
N
^
CD CD
i_
"1
_J (0
O) '*>
**c
ii

o
^-^

ID

o o

N'

5=

c
g
CO

CD

"c
D
O

0
O)
CO

s
o

p^^

co

r1

c
o E

rt

i)

CD
t_

ij.

c o
o

'co'

p|l'

31

CO

c
o
CD
D

"c
0

c
o
Q.
E
o
o
0
o
D

(0

B
CO

o
_
D)
C
(0

o
0
>
"^
0
U-

32

4.2.4 Vortex sheddina results

The Strouhal number of S, =0.19 for the circular cylinder closely agrees
with published results by other researchers. While the periodic vortices shed
from the simple cylinder were easy to see, large amounts of video had to be
reviewed to find useful data for most of the signal head configurations. In
general, even though the flow velocity was well within the sub-critical Reynolds
number range, the flow was very disorganized. The variety of edges and
corners on the traffic signals produced a complex turbulent flow field in which
identification of a dominant shedding frequency was difficult. The Strouhal
numbers for the configurations 1, 2, and 4 ranged from the 0.13 to 0.28. The
Strouhal numbers for other configurations could not be identified and are not
reported.
Randomness and high frequency of vortex formation are characteristics
that make vortex shedding an unlikely candidate to produce significant winddriven oscillations of traffic signal structures. First, random and disorganized
vortex shedding is less energetic and thus less able to sustain significant
oscillations. Second, a relatively high shedding frequency implies that the
required lock-in wind velocity for signal heads would be around 5 mph. Since
the lock-in velocity is relatively low, the driving forces would also be low.
Moreover, test signal structures in the full-scale field tests were observed
to be vibrating at a wide range of wind speeds. For example, the signal structure
with a 48-ft arm was observed to be vibrating at wind speeds ranging from 9 mph
to 16 mph. Vibrations in a wide range of wind speeds can not be explained by
the vortex shedding phenomenon.

4.2.5 Force measurements

Force measurements were made by attaching the mounting pipe to a
force balance which simultaneously measured the lift, drag, and side forces. For
the orientation of the model shown in Figure 4.3, the drag is the force acting on
33

the model in the free stream direction, the lift is an axial force aligned with the
mounting pipe, and the side force acts perpendicular to the plane defined by the
mounting pipe and the free stream velocity vector. This side force is the driving
force of interest for the signal structure vibration study. In a field experiment or
during normal operation, this direction would correspond to the vertical direction
or the direction commonly associated with lift.
The traffic signal mounting arm was connected to the balance by a
coupler which could vary the angle of attack of free stream flow relative to the
traffic signal. Static angles of attack, as depicted in Figure 4.4, simulate the
conditions experienced by an object subjected to galloping. In Figure 4.4 Vx is
analogous to the wind velocity which is aligned with the ground and the x-axis of
the traffic signal. The x-y axis is always attached to the model and indicates that
in simple galloping, all motion is in the y-direction (pure plunging). Therefore,
the angle between the wind and the model is always zero. However, the motion
of the model in the y-direction (Vy) results in an induced angle of attack. The
relative velocity, Vrei, is the vector sum of Vx and Vy. It is also the velocity
produced by the tow tank (Vrei = V,ank). The angle of attack, a, is measured
between the x-axis and Vrei. In other words, when the simulated motion of the
model is in the negative y-direction, a positive induced velocity of Vy results in a
positive angle of attack.
The force of interest in galloping applications is in the y-direction because
this is the force which either dampens or drives the motion of the model.
Therefore, the coefficient, Cpy, must be computed. The necessary equations
are as follows (refer to Figure 4.4):

Vx = Vre|COSa

(4.1)

Fy = Lcosa + (D-Dtare)sina

(4.2)

CFy=Fy/(0.5pVx'A)

(4.3)

34

where A is the signal area, and D,are is the drag on the pipe with no traffic signal
mounted on it.
The tow tank was operated at 1 fps for 50 seconds and 2000 data points
were collected for each angle of attack tested. The first 4 seconds of data were
discarded to avoid acceleration dynamics and the average force was computed
for each axis. Finally, the coefficient of force in the y-direction was computed.
This procedure was completed for each of the 8 traffic signal configurations.

4.2.6 Gallopino results

Figure 4.5 depicts the results for Cpy versus angle of attack for
configuration 5. Notice in Figure 4.5 that dCpy /da<0 for all angles of attack.
This is a classic result for an object that experiences aerodynamic forces aligned
with its direction of motion when initially disturbed in a direction perpendicular to
the free stream velocity, i.e., for an object exhibiting negative aerodynamic
damping. The negative slope for Cpy indicates that the aerodynamic forces are
destabilizing and the model is susceptible to galloping oscillations. None of the
other configurations tested were found to be susceptible to galloping.

4.3 Wind-tunnel experiments

A static test setup like the one used in the tow tank experiments can only
be used to check if a given traffic signal light configuration meets Den Hortog's
criterion-which is only the necessary condition for galloping vibrations. For a
complete understanding of the galloping phenomenon, one must use a dynamic
test setup. A dynamic test setup facilitates finding the aerodynamic damping at
different wind speeds, which can be used to find the wind velocity required for
the onset of galloping vibrations and the amount of the damping to be added to
mitigate the vibrations. Also, using a dynamic test setup one can test the
effectiveness of wing (damping plate) attached to the signal lights to increase
the aerodynamic damping, and in turn mitigate the vibrations.
35

The signal was Installed witti

15 degrees of tilt because of
Interference between the back
plate and the mounting pole.

Q.IS
D

0.5

XI

Fy

0.25

Dn

On d
^.25

tr

^.5
-30

-20

10

10

An^eof Atta

Figure 4.5

Cfy versus angle of attack for

traffic signal configuration-5

36

20

30

The wind tunnel experiments were conducted with the knowledge that
only configuration 5 in Figure 4.2 is susceptible to galloping. In the tow-tank
experiments a static (fixed) model was used to measure the drag and lift forces,
and these forces were used to determine if a configuration satisfied Den
Hartog's criterion. In the wind-tunnel experiments a dynamic model was used,
i.e., the model was suspended in the wind tunnel using springs (see Figure 4.6)
which in turn are connected to load cells to measure the dynamic lift and drag
forces.
Wind Tunnel
Wing

Wing.

N^

-k|^rag

Ul
Springs

k-B-H

Signal Light Modal Backplate

Springs
(b) Side View

(a) Front View

Figure 4.6 Wind tunnel setup.
4.3.1 Obiectives

The objectives of the wind-tunnel experiments were: (1) to evaluate the

aerodynamic damping, and (2) to test the effectiveness of the 'wing' (or damping
plate) as a vibration mitigation technique, which was successfully tested in the
field (see Figure 5.17).

4.3.2 Aerodynamic dampina

If the model is given an initial displacement and suddenly released, when
no wind is blowing, we obtain the mechanical damping of the suspension
system. This damping is analogous to the structural damping of the signal
structure measured in zero wind speed conditions. The damping measured in
37

the wind tunnel when the wind is blowing at a certain speed would be different
from the mechanical damping of the system. The overall damping system when
the wind is blowing can be either higher or lower than the mechanical damping.
This can be written as :

Overall Damping = Mechanical Damping -i- Aerodynamic Damping

If the aerodynamic damping is negative, the overall damping measured

when the wind is blowing will be smaller than the mechanical damping.
Otherwise, it will be larger.
The aerodynamic damping is usually plotted as a non-dimensional
number H / , which is given by Sarkar (1992):

4m(Co-C(R))
Hi(R) =
^^ ^^^ ^^
pB

,^ ^,
(4.4)

where
HJ(R)=

a measure of the aerodynamic damping at reduced velocity of R

R=-ynB
U = wind speed.
^(R) = the overall damping of the system at reduced velocity R
p = density of the air
m = mass of the model per unit length
B = along wind dimension of the model (see Figure 4.6)
^0= damping of the system at zero wind speed.
n = frequency of the system (model suspended from the springs) at zero
wind speed.

38

From Equation 4.4, one can observe that H^' will be a negative number if
the aerodynamic damping is positive, i.e., the system is aerodynamically stable.
Similarly, if Hi is a positive number the aerodynamic damping is negative.

4.3.3 Wind tunnel facility

A three feet high and four feet wide closed section was used for these
experiments. For the setup used, wind speeds up to 90 ft/sec could be attained.
However, higher wind speeds could have been obtained by changing the gear
ratio between the electric motor to the turbine fan. All tests were conducted in
smooth, non-turbulent wind.

4.3.4 Effect of turbulence

Turbulence in the incident wind can have a profound effect on the
galloping vibrations. If the bulk of the turbulence energy is contained in
frequencies much different from the natural frequency of the structure, the effect
of turbulence can be determined by simply using a wind tunnel with the desired
level of turbulence (Blevins, 1977).
Though it would be very interesting as well as useful to study the effect of
turbulence on the galloping vibrations (or aerodynamic damping), the difficulties
in conducting the tests with a turbulent wind in the Texas Tech wind tunnel were:
(1) generating the required turbulence and more importantly (2) measuring the
turbulence.
Due to a lack of the required facilities, the wind tunnel tests were confined
to tests in the non-turbulent wind. However, Novak and Tanska (1974) reported
that the wind speeds required for the onset of galloping were lower for a
turbulent wind than for a non-turbulent wind.

39

4.3.5 Model
A one-quarter scale model similar to that of the signal light configuration 5
(see Figure 4.2) was made from wood and aluminum. Two different wings were
tested for their effectiveness in mitigating vibrations: (1) a flat plate wing, and (2)
a flat plate with rounded edges. The shapes and dimensions of the wings are
shown in Figure 4.7.

4.3.6 Dvnamic experiments

The model was suspended by the springs as shown in Figure 4.6. An
initial displacement was given to the model in the direction of the lift and it was
suddenly released to allow the model to vibrate freely. This procedure was
repeated with different wind speeds. The free vibration response (in the
direction of lift) was recorded for different wind speeds and for the cases of: (1)
no wing (2) flat plate wing, and (3) flat plate with rounded edges.

4 in.
0.25 inch tubes
p a d d e d to the
plate to obtain
rounded edges

^f-^^^^- "

0=-
^

Flat plate wing

Flat plate with rounded edges

Figure 4.7 Wing shapes and dimensions

40

4.3.7 Wind-tunnel result.^

The non-dimensional number H / which gives a measure of the
aerodynamic damping is plotted against the reduced velocity (R = U/nB) in
Figure 4.8.
For the model without a wing attached to it, H / suddenly starts increasing
beyond a reduced velocity of 20, which corresponds to a wind speed of 14 mph
in field conditions. At a reduced velocity of 25 (18 mph in the field conditions),
*

Hi became a positive number, which suggests that the signal light configuration
has negative aerodynamic damping, and is susceptible to galloping vibrations.
With the flat-plate wing attached H / decreases (becomes a bigger negative
number) in the same region of the reduced velocity, i.e., the wing has stabilized
the signal light configuration. In the case of the wing with rounded edges, H /
decreases even more, i.e., this configuration is even more stable. H / values for
an air foil were also plotted in Figure 4.8 for comparison purpose. These H /
values for an air foil were calculated using the theoretical approximation
suggested by Theodorsen (1940).
4.3.8 Conclusions from the wind tunnel experiments
As stated in Section 4.3.1, the primary objectives of the wind tunnel
testing were (1) to measure the aerodynamic damping and (2) to test the
effectiveness of the wing in increasing the aerodynamic damping and mitigating
the vibrations.
Aerodynamic damping was measured by conducting free vibration tests in
different wind speed conditions. The results were presented in a nondimensional form.
The effect of attaching a wing on the aerodynamic damping was tested by
repeating the free vibration tests after attaching a wing. Two different wings
were tested and the results were compared.
41

winn-Mtsxri^sjj:.;

x:

W
0)

U
LU
0) "D
(0 <D
CL D
C
-^
TO O
LL CC

.2
CL

LL

CD

^111
o

in
CO

CO

CO

>
CM

00

c
o

o
>
D
CD
O
ID
D

CD

in

CC
CD
"^
(D
^
13

in

lO

in

CM

CM

o
CO

.^H

42

m
CO

CHAPTER 5
FIELD EXPERIMENTS

Full-scale field experiments provide an opportunity to deal with the

behavior of the actual structure without the problems of modeling and scaling
errors. Fundamental frequency and damping can be measured directly. The
wind conditions are real, rather than simulated; flow conditions are threedimensional, and turbulent, representing the actual situation.
Field experiments are not without disadvantages. Some field conditions
are difficult to control. The wind does not always blow and rarely at the speed
and direction required by the experiment. Experiments cannot always be
repeated under the exact same conditions. Individual parameters cannot always
be isolated by holding all other parameters constant. Instrumentation is more
difficult to control under the field conditions than in the laboratory. Despite these
disadvantages field experiments provide the best information on the behavior of
structures subjected to wind loads.

5.1 Obiectives
The objectives of the field experiments were as follows:
(1)

Determine the structural characteristics, fundamental frequency and

damping of the two traffic signal structures tested.

(2)

Study the large amplitude vibrations observed under conditions of a

steady wind.

(3)

Test the effectiveness of measures designed to mitigate large

amplitude vibrations.

The structural characteristics of the two signal structures tested were

needed to evaluate the galloping phenomena and to verify results of the finite
element method (FEM) analyses. In order to evaluate mitigation strategies for
the large amplitude vibrations, it was first necessary to produce them in the field.
43

Once the vibrations were initiated, the circumstances under which vibrations are
sustained could be identified.
A number of mitigation measures were identified and tested. The fullscale field tests either verified their effectiveness or demonstrated that the
concepts did not work.

5.2 Field research plan

Results of the tow tank experiments were used to develop the field
research plan. From observations of operational signal structures and the tow
tank experiments, it was clear that the galloping phenomenon takes place under
a rather narrow set of conditions. The wind direction appeared to be the most
critical. Thus, the need to control the relative wind direction of attack, was one
of the first field conditions addressed. The ability to rotate the signal arm to a
predetermined relative wind direction of attack was essential. Otherwise, a wait
would ensue for the wind to blow from the required direction. On the other hand,
the rigidity of the foundation under service conditions had to be reproduced.
The design of the turn-table foundation had to meet these two critical criteria.
Instrumentation was needed to measure the characteristics of the wind
and the response of the signal structure to wind effects. The research plan
included an evaluation of various transducers such as displacement
transducers, electrical resistance strain gages, and a tilt meter. The existing
meteorological tower at the field test site provided the needed data on wind
characteristics.

5.3 Facilities and instrumentation

The field experiments were conducted on the two traffic signal structures.
The tests were conducted at the Texas Tech University Wind Engineering
Research Field Laboratory (WERFL). Wind characteristics were determined
from instrumentation on the 160-ft meteorological tower. Structural response
44

was measured with transducers mounted on the signal structure and the WERFL
data acquisition system.

5.3.1 Texas Tech field site

The WERFL is located on the Texas Tech University campus in a large
open field with very few obstructions within a one-mile radius of the site.
Facilities at the WERFL include a 30 x 45 ft pre-engineered metal building
mounted on a circular track so the building can be rotated. Inside the rotatable
building is a concrete block building that does not rotate. This building contains
the data acquisition system for monitoring wind instrumentation and the various
transducers.
The 160-ft meteorological tower supports anemometers at the 13, 33, 70
and 160 ft levels. The instruments provide data for determining wind speed
profile and turbulence intensity. Other weather instruments include temperature,
barometric pressure and relative humidity sensors that are mounted at the 33 ft
level.

5.3.2 Test signal structures

Two signal structures were selected for testing. The first one, hereafter
referred to as the 40-ft signal structure, was an old structure that had been taken
out of service. It was available early in the project and was used primarily as a
shake down test for the instrumentation. The dimensions of this structure do not
match current TxDOT standard specifications. All weights and dimensions had
to be measured in the field. The structure was selected for testing because it
was readily available when testing began. Details of the 40-ft signal structure are
shown in Figure 5.1.
The second signal structure tested had a 48-ft cantilever arm. Hereafter it
is referred to as the 48-ft signal structure. It was purchased directly from a
manufacturer and meets current TxDOT specifications for an SMA-80 structure.
45

Figure 5.2 shows details of the 48-ft signal structure. A series of individual tests
was designed for each of the two signal light structures in order to achieve the
stated objectives of the project. These tests are described in detail in
subsequent sections of this report.

h
8.74 in. dia.

^__l
'

9.55 in. dia.

22.4 ft.

]-^^ '" ^'^

3.94 in. dia.

(1)

All the diameters shown are outer diameters

of a hollow circular cross section.

(2)

The thickness of the shell Is 0.17 inches.

11.1 in. dia.

Figure 5.1 Details of the 40-ft structure

Both signal structures have base plates to accommodate four anchor
bolts that are normally set in a concrete pier foundation. The signal structure
foundation used in the tests had provision for four anchor bolts, except they
were attached to a rotatable steel plate that could be rigidly clamped at one of
the 24 possible locations spaced 15 degrees apart (see Figure 5.3).

5.3.3 Foundation
Because of the need to rotate the signal arm to any relative wind direction
of attack (with 15 increments) the rotatable foundation had to meet the following
criteria:
46

(1)

Capability to easily rotate the signal structure to any desired

orientation in 15 increments.

(2)

Same foundation rigidity, when clamped, as a structure under

service conditions.

10.5 in. dia.

K-

12ft
4.1 in. dia.

nm
19ft

^ 1
(XXXX)

Notes:

13 in. dia

(1)

All the diameters shown are outer diameters

of a hollow circular cross section.

(2)

The thickness of the shell is 0.239 inches.

Figure 5.2 Details of the 48-ft structure

Except for the rotatable turret bearing, the foundation was constructed to the
same specifications as one being used for actual service conditions. The

t
North

JL

*
t *

270 II r~i
'

'

i_^(,

90

I*

t i l

Q
180

\
I

15 Degree
increments

Figure 5.3 Possible orientations of the signal structure arm (plan view)
47

rotatable foundation is shown in Figure 5.4

The following procedure was followed in rotating the signal structure to
some desired orientation relative to the wind direction.
(1)

The wind direction was monitored for at least five minutes to obtain
an average value.

(2)

The four clamping bolts were removed, which allowed the structure
to rotate freely (the rotatable foundation supports the structure even
when the clamping bolts are removed).

(3)

The structure was rotated to the desired location in 15 increments.

The structure was then secured by installing the clamping bolts in
the threaded holes provided in the steel base plate and tightened.

Signal Structure
Base plate

Signal Structure Pole

^.
J5,

Signal Strucuture
Anchor Bolts

Clamping
Bolts

Top Plate
(Rotates)

Turret
Bearing
(Rotates)

Bottom Plate
(Fixed)
Concrete Pier Foundation

Figure 5.4 Rotatable foundation

Figure 5.3 shows the possible orientations of the signal structure arm. To
achieve vibrations due to galloping the cantilever arm needed to be very nearly
perpendicular to the mean wind direction.

5.3.4 Instrumentation
Instrumentation was needed to measure wind characteristics and
structural response.
5.3.4.1 Wind instruments. Anemometers mounted on the 160-ft
meteorological tower and on a 19-ft pole were used to measure characteristics
48

of the natural wind. A three-cup anemometer and wind direction vane were
mounted on a 19-ft pole that was placed near the signal structure installation
(see Figure 5.5). These instruments measured the wind speed and direction
needed to set the proper orientation of the signal structure relative to the wind
direction. Data runs requiring a continuous measurement of wind speed and
direction were taken continuously for 15-minute periods.
5.3.4.2 Transducers. Three transducers were tested in the field for their
effectiveness in measuring the response of the signal structure to wind effects.
They were (1) electrical resistance strain gages, (2) a linear variable differential
transducer and (3) a tilt meter. Figure 5.6 shows locations where the
transducers were mounted.

Figure 5.5

A three-cup anemometer and wind direction vane

mounted on a 19-ft pole near the test structure.

One set of strain gages was installed on the 40-ft signal structure. Two
gages were mounted on opposite sides of the vertical pole, in line with the
cantilevered arm of the structure 13 in. above the top of the pole base plate. As
the pole bends, the gage on the outside measures tension strains while the one
on the inside measures compression. The gages were connected into a
Wheatstone half-bridge circuit as shown in Figure 5.7. This arrangement
measured the bending strain and compensated for axial and temperature strains.

49

in Figure 5.7, Ri and R2 are bridge completion resistors and R3 and R4 are the
strain gages on the pole.

Tilt Meter

Displacement
Transducer

Figure 5.6 Instrument locations on the test structure

Two sets of strain gages were installed on the 48-ft signal structure. One
set was located 13 inches above the base of the vertical pole and the other set
was located on the cantilever arm at a distance of 11 inches from the connection

R4 Rg

Figure 5.7 Wheatstone half-bridge circuit

plate of the arm. In order to obtain a higher voltage output, a full Wheatstone
bridge was configured with four active gages. Two gages were mounted on the
tension side of the structure, as well as two on the compression side. The fullbridge Wheatstone circuit is shown in Figure 5.8.

50

R2 = Rc

Figure 5.8 Wheatstone full-bridge circuit

Temperature changes during a data run will give erroneous strain

readings if provisions have not been made for temperature compensation.
Under harsh field conditions, temperature compensation is especially important.
Temperature fluctuations affect the bridge circuit in two ways:
(1)

The bridge will output an erroneous strain reading if there are no

provisions for temperature compensation.

(2)

Temperature fluctuation may also affect the resistance of lead wires

in the circuit.

Temperature compensation can be achieved in a Wheatstone bridge if

the gages in opposite legs of the bridge experience the same temperature
changes. Lead wires should be kept as short as possible. For this reason, the
amplifier circuit boards were installed very near the locations of the gages,
rather than in the instrumentation room near the data acquisition equipment.
In the field, voltages from nearby electrical or magnetic fields can create
noise that affects the voltage output of the bridge. The bridge output due to
strain is small and can easily be overshadowed by noise voltage. The noise
effects can be eliminated by one of two well-known techniques:
(1)

By amplifying the signal close to the gage. This technique is known

as pre-amplification because the signal is amplified before running it
51

through a long signal cable. By amplifying the signal before it is

affected by noise, the signal voltage is much higher than the noise
voltage.
(2)

By using a current signal instead of a voltage signal to measure

strain. The inductive nature of the noise voltage does not
significantly affect the current carried by the electrical conduit.

The pre-amplification technique was used in this study. A model TM-1

amplifier conditioner module (Imperial Instruments) was used to amplify the
strain signal close to the test structure (see Figure 5.9)

Figure 5.9 Pre-amplification unit used to amplify the strain signal.

In order to relate strain to the loading on a structure, the strain gage

bridge must be balanced so there is zero output voltage prior to applying the
load. The balance is achieved by adjusting the resistance of one of the gages in
the Wheatstone bridge. The balance is accomplished by connecting a very high
variable resistance in parallel with one of the strain gages. The variable
resistance is adjusted until the voltage output from the bridge is zero.
52

Calibrating the strain gage bridge means establishing a relationship

between voltage output and strain in the specimen. Bridge calibration is done by
simulating a known strain, i.e., imposing a known change in resistance of one of
the gages and measuring the voltage output. The method is known as shunt
calibration.
5.3.4.3 Displacement transducer. A method was needed to measure the
large amplitude deflection at the tip of the cantilever arm. An instrument that
seemed to hold promise was a displacement transducer (UniMeasure) with a
thin wire that unrolled from a shaft as the displacement took place (see Figure
5.10), similar to a fishing reel.

Figure 5.10 Displacement transducer

A voltage proportional to the amount of wire rolled off the shaft is the
output signal. The instrument has a linear response over a relatively large
displacement of the order of 24 inches. However, since the arm tip was moving
up and down, a tension was needed in the wire to keep it taut. Even though the
tension was less than 2 lbs, it was enough to dampen the vibration due to
galloping. The sensor worked well for the free vibration tests, but could not be
used for galloping tests.

53

5.3.4.4 Tilt meter. The tilt meter is a small sensor (Applied

Geomechanics) that provides a voltage output signal proportional to the angle of
tilt of the surface on which the instrument is mounted (see Figure 5.11). The
angle of tilt of the cantilever arm can be measured by mounting the tilt meter at
the arm tip as shown in Figure 5.6.

mwsmmmm^m::iisms^it''*Mr7:v.~-;is^^':^^^^^^^

Figure 5.11 Tilt meter

5.4 Field studies

The field studies involved three distinct sets of experiments:
(1)

Determine structural characteristics of the two full-size signal

structures.

(2)

Examine the specific conditions under which galloping takes place.

(3)

Test the various mitigation concepts.

Each of these experiments is discussed below.

54

5.4.1 Structural characteristics

The three structural characteristics of interest were stiffness (expressed
as a load versus deflection relationship), natural frequencies, and damping
coefficient.
5.4.1.1 Stiffness. The two signal structures were tested for stiffness
characteristics on calm wind days. The structures were instrumented with strain
gages, a tilt meter, and a displacement transducer, as described in Section
5.3.4.
Static concentrated loads were applied at a point near the free end of the
cantilever arm of the signal structure by hanging known weights using a steel
wire. Zero load readings were taken on the three transducers. The instruments
were read as each load increment was added to obtain relationships between
load and strain, tilt angle and displacement, respectively. Results of these tests
are summarized in Table 5.1 for both the 40-ft and the 48-ft structures.
Because of small fluctuations in the transducer outputs, even on a calm day, the
data were recorded as five minute means. Calculated strains from Chapter 3
(see Table 3.2) were also shown again in Table 5.1 for the comparison of
calculated strains with the measured strains. Agreement is reasonable.
Figure 5.12 shows load versus strain relationships for the 40-ft and 48-ft
signal structures. Note that the curves are essentially linear. From these load
versus strain curves, it is apparent that the 48-ft signal structure is slightly stiffer
than the 40-ft signal structure.
5.4.1.2 Natural frequencies. Tests were conducted to obtain the natural
frequencies of vibration of the two signal structures. A concentrated load was
suspended by a wire connected three feet from the free end of the cantilever
arm. The wire was cut to quickly release the weight and allow the arm to
undergo free vibration. The three sensors were monitored to obtain time
histories of strain, tilt and displacement. The strain records were selected for
the frequency analysis because the results appeared to be the most reliable of
55

c
C/D

CD

-^ CD r^ o in
-^ r^ CO o h~
CO CO Tf in iri

O) c\j in T- o
"^ 00 "^ o in
CO CO '^' in in

E
CD
CO C30
O
CD
CL n
CO o
13
O

Q.

CO

<

o
CD

o
3

to

..;3

D)
CO

.9- CD

CO O C\J Tt CD

CO

"cO

00

- 0 0
C5) O

-4'

CO 0 0

E g^

<2
C

o
13
v_

CO

O CO CO CM O
CD Q^ ^
iD CO

CO
I

o
c

<D

i : (0
CO iz

E "

O
ZJ

CO

*-

T5

CO CX)

55 i2 (D

CO

i n 00 T- Tt r>.
,-^

^ |s> Tt O CD
CO CO ^ t i n in

D .E ' "
"^

<Si

O -n

CO
CO
CD

1-

CD ' ^

CO

-^ in i n CD N-

O :;= " -

g
'co

CO

<D

CO

I
13
CO

Qd
O
CL

13

(0
0)

CM

C\J

c
"co

' ^ in CO N- 00

CO

"(0
1

*C0

CVJ

CO

Q.
to
TD
D

CO

00

o
JO

C
CO

CO CO O

CO

CD CO Tf

- ^ CO

o j a> ci> CI) cj)

Tt i n CD r^ 00

o g^

CM
f

CO
CO
CD

CD CD h -

CJ) 00

i n CD r^ 00 cj)

in
m iw: CO
"'
2 T3 "^

3 S

in
(0

D ) h- (0
(Tt
CO ^
o o o o
o
i5 O : 9 *-' i n CD h - 0 0 cj)
-

CO - J

- -

CO JO
0) 0) CO

CO ^
- w o

56

the three instruments. The analog signals from the transducers were converted
to digital form. A spectral analysis was then performed to identify the
fundamental frequencies of vibration.
^ 100

40-ft Structure
48-ft Structure
CO

20

40
60
Load in Lb

80

100

Figure 5.12 Load Versus Strain in the Vertical Pole

The experiment was first performed on the bare pole without signal lights
attached. Figures 5.13 and 5.14 show time histories of strain for the two bare
signal structures. The two spectral analyses are shown in Figures 5.15 and 5.16
for the 40-ft and 48-ft signal arms, respectively.
The fundamental frequencies change when traffic signal lights are
mounted on the signal structures. The change depends on the mass of the
signal lights and their locations relative to the free end of the arm. Signal lights
were mounted on the two structures as shown in Figures 5.1 and 5.2. Table 5.2
compares fundamental frequencies of the bare structures and configurations of
two signal lights on each structure. The presence of the signal lights on the
structures reduced the fundamental frequencies by about 25%. Table 5.2 also
shows the calculated frequencies (using FEM) from Chapter 3 for comparison
with the measured frequencies. FEM results for 48-ft structure are very close to
that of measured frequencies, where as for 40 ft structure the results were off by
about 20%. This discrepancy could be due to the lack of exact dimensions and
material properties of 40 ft structure.

57

5.4.1.3 Damping. For conducting a dynamic analysis of a structure, the

structural damping is typically assumed based on the guidelines provided in the
design codes. The true damping of a structure can be obtained only by
conducting a free vibration experiment. The strain versus time histories of free
vibration given in Figures 5.13 and 5.14 were used to obtain damping values.
The logarithmic decrement approach (Clough and Penzien, 1980) was used to
obtain the damping ratios for the two signal structures. Damping values with
signal lights and without signal lights are tabulated in Table 5.2. Clearly the
presence of the signal lights affects the damping as well as the fundamental
frequencies.

Table 5.2 Measured frequency, Measured damping, and Calculated Frequency

% Critical Damping
Fundamental Frequency
Structure type
Without lights
With lights
Without lights
With lights
Meas^ Calc^ Meas^ Calc^
1.02
1.21
0.78
0.87
0.52
0.78
40-ft struct
0.38
0.62
0.98
1.02
0.74
0.81
48-ft struct
^Measured in the field
^Calculated using finite element models
5.4.2 Gallopino of the structure
This second series of tests in the field was designed to reproduce
galloping in the two test structures and to carefully identify the parameters that
contribute to galloping.
5.4.2.1 40-ft Signal structure. The structure with the 40-ft cantilever arm
was installed in the field first. The primary purpose was to shake down the
instrumentation and obtain qualitative information about the galloping
phenomenon in the field. In particular, it was desired to verify results observed
in the tow tank.
From the tow tank experiments signal light Configurations 1, 2, 3 and 4
(see Figure 4.2) were found to be very stable, i.e., galloping was not expected to
take place, because of the positive slope of the Cpy versus angle of the attack
58

150

<o
c
(0

"co
o
k_

o
Q.

c
o

*co

CO

-100

10

20

30

Time, Seconds
Figure 5.13 Free vibration response of 40-ft structure without signal lights
59

60

<0

c
"co

to
o
o

o
CL

c
o
c
'co

CO

10

20

30

40

50

70

Time, Seconds
Figure 5

14 Free vibration response of 48-ft structure without signal lights

60

80

<0

c:
Q
"co
CD

a
CO

1.5

2.5

i_J

3.5

4,5

Frequency, Hertz
Figure 5.15 Free vibration spectrum of 40-ft structure without signal lights
61

xlO

(O

(D

Q
CD
CL

CO

10

Frequency, Hertz
Figure 5.16 Free vibration spectrum of 48-ft structure without signal lights
62

curve over a wide range of attack angles. Because these configurations are not
susceptible to galloping, none of these configurations was tested in the field.
Configurations 6, 7, and 8 from Figure 4.2 also were not expected to
exhibit galloping. These same three configurations were set up on the structure
in the field. The cantilever arm was rotated to within 7.5 of the five-minute
mean wind direction under several wind speeds. The test was set up several
different times, but galloping was not observed in any case. The tow tank and
field tests were in complete agreement for these three configurations.
According to the tow tank studies. Configuration 5 in Figure 4.2 should
exhibit galloping, because of the negative slope of the Cpy versus angle of attack
curve. With Configuration 5 mounted on the 40-ft signal structure, galloping was
observed on several occasions with free end displacement amplitudes of 12-16
inches. From these tests, the data indicate that the only light configuration likely
to gallop is Configuration 5.
5.4.2.2 48-ft Signal structure. According to tow tank test results, the
signal light configuration that produced galloping with the largest amplitude
displacements of the free end of the cantilever arm was Configuration 5. Hence,
this configuration was used in all subsequent tests on the 48-ft signal structure.
The three primary factors required for the structure to exhibit galloping were:
(1)

Relative wind direction of attack (i.e., wind blowing from back side of
the lights or from the front side)

(2)

Wind speed

(3)

Signal light Configuration 5.

All field tests that involved galloping essentially followed the same procedure.
The wind vane on the 19-ft pole near the structure was monitored to obtain oneminute mean wind direction. The signal structure arm was rotated so the wind
direction was normal to the cantilever arm within 7.5 from the back side of the
signal light.

63

In Configuration 5, the signal light is suspended below the signal arm and
has a back plate. Wind flow is from the back side of the signal light. After
adjusting the signal structure to the most favorable wind direction, the structure
would achieve a state of steady galloping through the following sequence of
events: The free end of the cantilever arm was held steady by means of a thin
wire. The wire was slowly released to allow the signal structure to vibrate at will.
Initially, small displacements normal to the wind flow took place. The signal
structure was allowed to continue vibrating. Gradually the amplitudes of
vibration increased. If the wind speed and direction held steady, vibrations
would continue to increase in amplitude until some limiting value was achieved.
The vibration took place at a frequency very near the fundamental frequency of
the sign structure.
A change in wind speed or wind direction would alter the vibration
characteristics. With significant change in wind direction, the vibration
amplitudes would decrease and galloping would cease. A change in wind speed
resulted in a change in displacement amplitude, if wind direction held steady.
Ideally, displacements of the free end of the cantilever arm would have
been measured with the displacement transducer. However, because the
driving force associated with galloping was so small, the tension in the wire
caused significant damping of the vibrations.
To demonstrate the problem with tension from the displacement
transducer, the following experiment was conducted. The signal structure was
set up to gallop as described above. After a period of time, the displacement
amplitude reached a steady state and clear evidence of galloping. The wire to
the displacement transducer was carefully hooked to the free end of the arm
without stopping the vibration. The wire had no more than a two-pounds
tension. Suddenly the displacement amplitude was reduced and did not regain
the original displacement amplitude. The wire was then unhooked; after a few
minutes the galloping resumed and regained its original steady state amplitude.
64

From these experiments, it was concluded that the displacement

transducer could not be used for measuring displacements during galloping. All
subsequent tests relied on the strain gage readings either in the arm or the
vertical pole to derive the displacement. The strain gage readings proved to be
reliable and accurate.

5.5 Mitigation measures

Galloping has been identified as the mechanism for large amplitude
displacements in traffic signal structures under certain conditions of wind
direction, speed and structure characteristics, as discussed above. Several
mitigation measures were tested in the field to reduce the amplitude of the
vibration. These included a damping plate (wing) attached to the cantilever arm,
tuned mass dampers, and water slosh damper. The damping plate was by far
the most effective mitigation measure, and it is described below.

5.5.1 Damping plate (Wing)

Signal lights fixed as Configuration-5 (see Figure 4.2)have negative
aerodynamic damping (given by positive Hi in Figure 4.8). This negative
aerodynamic damping decreases the effective damping of the structure and
leads to large amplitude galloping vibrations. These vibrations can be mitigated
by installing a flat plate, which has positive aerodynamic damping (Shiraishi,
1971). The positive aerodynamic damping of a flat plate increases the effective
damping of the signal structure. A flat plate used for this purpose of vibration
mitigation is called a "Damping Plate" or a "Wing."
A wing can be used in two different ways to mitigate the vibration: (1) A
wing can be installed above the signal light in Configuration-5, such that the new
configuration (i.e., Configuration-5 with a wing over it) has a positive
*

aerodynamic damping (given by negative Hi in Figure 4.8). The negative

aerodynamic damping of Configuration-5, is eliminated by addition of the plate
65

over the lights. This method of vibration mitigation is called the "Elimination
method". (2) In the second method, a wing can be used elsewhere on the
cantilever arm (away from the signal lights). This method is effective only if the
negative aerodynamic force caused by the signal lights is compensated by the
positive aerodynamic damping force caused by the wing. One must calculate
the negative aerodynamic force caused by the signal lights and then determine
the size of the wing needed to compensate this force. This method of vibration
mitigation is called the "Compensation method".
Various damping plate configurations have been used by TxDOT
maintenance personnel to mitigate wind-induced vibrations with marginal
success. The typical installation consists of a plate 9 in. x 36 in. in size
mounted on a section of the bare arm away from the signal light (compensation
method). This configuration is not very effective in mitigating vibrations,
because some signal structures continue to vibrate after a wing is attached. The
reason is that the a much larger plate is required to obtain the positive
aerodynamic damping force. A larger wing 16 in. x 66 in., hereafter called
Proposed wing, was used in the field testing. The proposed wing was mounted
above the 5-signal light (elimination method).

5.5.2 Size and location of wing

The important parameters in the magnitude of the damping force that a
wing causes in order to mitigate the vibrations are its size and its location on the
arm. Equation 4.4 may be rewritten in the following form:
4m(Co-aR)) = Hi(R)pB2

(5.1)

The left-hand side of the Equation 5.1 is the aerodynamic damping force, which
is shown from the right-hand side to be proportional to the square of the wing
dimension in the direction of the wind, B. Moreover, Equation 5.1 gives the
aerodynamic damping force per unit length in the direction parallel to the arm.

66

To relate the aerodynamic damping to the dimensions of the wing, a simplified

relation can be written as follows:

FAD

(B^ L)

(5.2)

where.
FAD = the aerodynamic damping force due to the wing,
B

= width of the wing = 9 in. (TxDOT wing) = 16 in. (Proposed wing),

= length of the wing = 36 in. (TxDOT wing) = 66 in. (Proposed wing).

The above equation suggests that the aerodynamic damping force due to
the wing is proportional to the square of the width and directly proportional to
the length of the wing. By substituting the sizes of two different wings, the
Proposed wing (at the same location as TxDOT wing) generates 5 to 6 times
more damping force than the TxDOT wing.
The above discussion addresses the reason for using a larger wing, but
not the 16 in. X 66 in. size. The reason for selecting that particular size is that it
is a standard blank road sign stock used by the Texas Department of
Transportation and is readily available for the implementation.
In addition to increasing the size, by locating the wing over the signal light
near the tip of the arm, the effectiveness of the wing can be increased. Higher
effectiveness is due to two factors: (1) the signal light head which originally had
negative aerodynamic damping will have positive aerodynamic damping with the
plate installed above it, and (2) the wing which has a positive aerodynamic
damping by itself generates higher damping force due to higher vibration
velocities near the tip of the arm.

5.5.3. Proposed wing

An experiment was designed to demonstrate the effectiveness of a large
damping plate. A 16 in. wide and 66 in. long plate was mounted above the
traffic lights in Configuration 5 (see Figure 4.2), as shown in Figure 5.17.
67

Wing

|^B = 16in

3 1 3 in.

Wind
Direction

Figure 5.17 Proposed wing

On a day when the wind speed was between 10-15 mph, the signal arm
was rotated to be normal to the 5-minute mean wind direction. The wind was
essentially from the south (180) so the signal arm was rotated 270 and pointed
toward the west. Initially, the damping plate was mounted on the structure. A
continuous set of records was obtained consisting of 217 5-minute records.
After validation of the data, plots of 5-minute mean wind direction, 5-minute
mean wind speed and, 5-minute RMS of strain on the vertical pole of the 48-ft
signal structure were made as shown in Figures 5.18, 5.19 and 5.20,
respectively. These are essentially time-histories of the three parameters.

20

40

60

80

100

120

140

160

180

200

220

240

Run Number
Figure 5.18 Wind direction plot
The RMS of strain, which is the root mean square of the fluctuating strain
component for each five-minute record, is a measure of strain fluctuation about a
68

zero mean. A large RMS value implies large fluctuations (displacements) of the
signal structure.
The signal structure would not be expected to gallop unless the wind
direction was within 7.5 of due south (180). Thus, from Figure 5.18
galloping was expected from record 27 to 91 when the wind direction was
favorable for galloping. Galloping is not expected below record 27 and above
record 91 because the wind direction was not favorable. Even when the wind
direction was favorable for galloping, little or no galloping was indicated for
records 27 to 75, as shown by the relatively small values of RMS during that
time period (see Figure 5.20). These were the records of data collected when
the wing was attached.
Expect
Galbping
G

&

20

J
|

Without Wing

With Wing

100

120

140

200

220

240

Run Number
Figure 5.19 Wind speed plot

At about record 73, the plate was quickly removed from the signal
structure. RMS values in Figure 5.20 indicate a very strong galloping from
record 75 to 91. By record 91 the wind direction had shifted more than 7.5, so it
was no longer normal to the back side of the signal structure. Thereafter,
galloping diminished as observed in Figure 5.20.
The experiment clearly demonstrated the effectiveness of the large
damping plate. To be most effective, the plate should be mounted above the
signal light with at least a 3-in. separation between the damping plate and top of
the backing plate. The large wing is essential for effective mitigation of the
vibration.
69

I
r~

on
C

-<>

80

:l

60 -

Expect
Galloping

I
"H

With Wing
Without Wing

40 20
0

20

40

60

80

100

120

140

160

180

200

220

240

Run Number
Figure 5.20 RMS of strain plot
The large damping plate was clearly the most effective method for
mitigating vibration due to galloping. When galloping is observed in an existing
signal structure, a large damping plate (wing) can be installed over the signal
light to effectively reduce the vibration.
The advantages of this mitigation strategy are:
(1)

It is a relatively easy fix, requiring no knowledge of the

dynamic characteristics of the existing structure.

(2)

Materials are readily available and easy to install by

highway department maintenance personnel.

(3)

The wing is not a significant distraction to motorists.

The wing can be used elsewhere on the arm other than over the signal
lights. Except that by doing so, one is trying to compensate for the negative
aerodynamic damping of the signal lights instead of eliminating it. Obviously, for
this mitigation method to be effective one needs to calculate the negative
aerodynamic damping force caused by the signal lights and find out what size of
the plate at a given location on the arm can produce the required positive
aerodynamic damping force.

70

CHAPTER 6
ANALYSIS OF RESULTS AND CONCLUSIONS

The large amplitude vibrations of cantilever traffic signal structures have

been thoroughly studied through a literature search and a series of
experiments. The experiments were conducted using the Texas Tech tow tank,
wind tunnel, and field site to address the objectives of this research, namely :
(1)

to identify the conditions required for large amplitude vibrations in

cantilevered traffic signal light structures;

(2)

to understand the mechanism that produces the large amplitude

vibrations; and

(3)

to develop strategies to mitigate vibrations in cantilevered traffic

signal structures.

6.1 Analvsis of results

Results from finite element analyses (Chapter 3), laboratory experiments
(Chapter 4), and full-scale field experiments (Chapter 5) are analyzed in this
chapter. The discussions and calculations in this chapter are limited to the 48-ft
structure.

6.1.1 Eouivalent Single Degree of Freedom Model

The response of a signal structure (see Figure 6.1), which is continuous
(infinite degrees of freedom), is a function of both space and time. However this
problem can be simplified to an equivalent Single Degree of Freedom (SDOF)
model as shown in Figure 6.2. The equivalent single degree of freedom system
concept is based on defining the dynamic response as follows:

y(x,t) = y*(t)x|/(x)

(6.1)

where

71

y(x,t) = dynamic response at point x on the structure at time t.

y*(t) = displacement along the generalized coordinate (see Figure 6.1)
\|/(x) = normalized mode shape.

y*(t)

Figure 6.1 Mode shape corresponding to the lowest frequency for 48 ft structure

As seen in Figure 6.1, the first mode response has displacement in both X
and Y directions. The response in the X direction is very small, and is neglected
in the calculations presented in the following paragraphs. Mode shape \\f(x) is
obtained using FEM analysis and is normalized with respect to the arm tip
displacement.

Figure 6.2 Equivalent SDOF model

72

The dynamic equilibrium equation for the equivalent single degree of

freedom model shown in Figure 6.2 can be written as follows:

M * [y* (t) + 2CmC0y* (t) -f- co^ y* (t)] = P * (t)

(6.2)

where
M*

Generalized mass

CO

Circular frequency

y* (t) =

Acceleration in the direction of generalized coordinate

y* (t) =

Velocity in the direction of generalized coordinate

y*(t)

Displacement in the direction of generalized coordinate

Cm

Mechanical damping ratio (measured at zero wind speed)

P*(t)

Generalized dynamic force.

6.1.2 Calculation of structural parameters

Considering only the arm movement and using the normalized mode
shape from the FEM analysis and the known mass distribution along the arm,
the generalized mass of the 48 ft structure was calculated to be 8.42 slugs
(Clough and Penzien, 1980).
The generalized stiffness K* of the 48-ft structure was determined from
field measurements using the average load at the tip of the arm to produce a
deflection of one inch (see Table 5.1).
K* =

[(50/3.14)+(60/3.76)-f-(70/4.37)-H(80/5)+(90/5.75 )]/5
15.91 lb/in
191 lb/ft.

73

Knowing the fundamental frequency f of the 48 ft structure with signal lights

installed (see Table 5.2) and the generalized stiffness K*, the generalized mass
M* of the structure can be calculated as follows:
co=27rf =

(6.3)
1

27C VM*

So,

0.74 Hz (field measured frequency, see Table 5.2)

K*
9-^
=
8.83 slugs

M^

which closely matches with the analytical result of 8.42 slugs. The difference in
the masses can be explained by the fact that discretized finite element models
are stiffer than actual structures, which also explains the difference in the
measured and calculated frequencies in Table 5.2.
The generalized mechanical damping Cm for the 48 ft structure with the
signal lights is calculated using the field measured damping ratio ^^n = 0.62%.
(see Table 5.2) as follows:
Cm

=2M*Cm(27rf)

(6.4)

= 2 X 8.83 X 0.0062 x 2 x 71 x 0.74

= 0.51 slugs/sec
= 16.39 1b. sec/ft.

6.1.3 Aerodynamic force and effective damping coefficient

The generalized force P* is the aerodynamic force which causes the
vibrations:

P*(t)= JP(x,t)\|/(x)dx

(6.5)

74

where P(x,t) is the aerodynamic force distributed along the cantilever arm length
of L and \|/(x)is the normalized mode shape. The 48-ft traffic signal structure
arm that was tested had two traffic signal lights: (1) a 5-light signal at the tip of
the arm and (2) a 3-light signal at a point 12 ft from the tip of the arm (see Figure
5.2). Neglecting the contribution from the arm alone, the distributed
aerodynamic force on the structure can be assumed to act only at the location of
the signal lights, and at the location of the wing (see Figure 6.3).
P(x,t) .

11
c m

mWintg

n
KXXXX

Figure 6.3 Aerodynamic force

Simiu and Scanlan (1986) discussed a mathematical model to compute

this force P(x,t) for bridge decks with two degrees of freedom, i.e., vertical
displacement and torsional rotation of the bridge deck. The same mathematical
model can be used for traffic signal structure vibrations by considering only the
contribution from vertical displacement. After simplifications (neglecting terms
for torsional rotation), the final equation can be written as follows:

P(x,t) = - ^ ^ . ^ H ; Mx)y* (t) + LB(x,t)

where
P

Air density

Along wind dimension of wing

75

(6.6)

CO

Circular frequency of the structure

\|/(x)

Normalized mode shape

Flutter derivative obtained from wind tunnel

experiments (see Figure 4.8)

H|

y * (t) =

Velocity of equivalent SDOF system (Arm tip

velocity).

LB(x,t) =

Buffeting force due to turbulence.

Substituting Equation 6.6 in Equation 6.5:

P*(t) =

P^'^H;W'(x)dx
2

y*(t) + jLB(x,t)v(x)dx.

(6.7)

In the above equation, the term within the brackets is an aerodynamic damping
coefficient Ca:

Ca=--^^H;jxi/2(x)dx.
2
0

(6.8)

If the second term in Equation 6.7 is written as:

LB(t)=jLB(x,t)\|/(x)dx
0

(6.9)

then Equation 6.7 becomes

P*(t)=-Cay*(t)+LB(t).

(6.10)

Substituting Equations 6.10, 6.9, 6.8, 6.4, and 6.3 in Equation 6.2
M*y*(t) + C,y*(t) + K*y*(t)]=-C,y*(t)+L;(t)

(6.11)

which can also be written as:

M * r ( t ) + (C,.,+Cjy*(t) + K*y*(t)l = LB(t).

(6.12)

In the above equation the damping coefficient is the sum of the

mechanical damping coefficient of the structure Cm and the aerodynamic
damping coefficient Ca. If the net damping coefficient is called effective damping
coefficient Ce, Equation 6.12 can be written as:
76

M * y* (t) + Ce y* (t) + K * y* (t) = LB(t)

(6.13)

Ce = Cm + Ca.

(6.1 4)

where

The effective damping ratio ^^ in terms of percent critical damping is

given by:

Ce =

;^100

(6.15)

2 M CO

When a structure has negative aerodynamic damping (i.e., Ca < 0), the
effective damping decreases (i.e., Ce < Cm). With this reduced damping, a small
disturbance will lead to prolonged vibrations of the structure. If this reduction in
the effective damping reaches a point where the effective damping becomes
negative (i.e., Ce < 0), the structure starts vibrating by itself, which is "self

excitation."
6.1.4 Calculation of effective damping for different cases
In the above paragraphs, all the structural parameters (mass, mechanical
damping, and stiffness) were calculated from results of the field experiments and
FEM analyses. The aerodynamic damping coefficient Ca is calculated using the
flutter derivatives Hj (see Equation 6.8), which are obtained from the windtunnel experiments. Knowing the mechanical and aerodynamic damping
coefficients, the effective damping coefficient is calculated using Equation 6.14.
The effective damping ratio in terms of percent critical damping is calculated
using Equation 6.15.
From the full-scale field experiments, it was found that Configuration-5
(see Figure 4.2) exhibited galloping vibrations when there was no wing attached.
The vibrations were significantly mitigated when a wing was attached (see
Figure 5.20) above the 5-light signal at the tip of the arm. The attached wing
increases the effective damping of the structure. It is of interest to calculate the

77

increase in the effective damping after the wing is attached. In the field
experiments, the proposed wing (16 in. wide and 66 in. long) was attached
above the 5-light signal to change this shape into an aerodynamically stable
shape. A wing can also be effective at other locations, provided it generates
enough positive aerodynamic damping to compensate for the negative
aerodynamic force caused by the signal lights. It is also desirable to know the
effectiveness of the proposed wing attached at other locations away from the
signal lights. To address these questions the effective damping coefficients at
different wind speeds and different wing locations were calculated for the
following five cases:
A. A 5-signal light at the arm tip, and a 3-signal light at a point 12
ft from the tip of the arm with no wing.
B. Lights same as in Case A and the proposed wing attached
above the 5-signal light.
C. Lights same as in Case A and the proposed wing with rounded
edges located above the 5-signal light.
D. Lights same as in Case A and the proposed wing attached
halfway between the lights.
E. Lights same as in Case A and the proposed wing attached to
the tip of the arm.
Effective damping coefficients Ce were calculated and plotted with
respect to the wind speeds for the above five cases in Figure 6.5. Effective
damping ratios ^e were plotted against wind speeds in Figure 6.6. For Cases A,
B, and C, H] values were obtained from the wind-tunnel experiments. For cases
D and E, theoretical values of Hj for an airfoil of the same size as the proposed
wing was used. Airfoil data (Theodersen, 1940) was readily available and
experiments conducted by Shiraishi (1971) suggested that H[ for a flat plate is
close to that of an airfoil. In Case D (wing attached to the arm between the
78

lights), the interference of the arm was neglected, because the gap between the
wing and the arm is large enough for the wing to act independently.
Case D

Cases B & C

Case E

^X X X A >

>^AA/
12 ft.

Figure 6.4 Three locations of the wing considered

From Figure 6.5, the calculated effective damping coefficient Ce becomes
negative for Case A (structure vibrates) for wind speeds greater than 20 mph.
However, in the field experiments, the same configuration vibrates at wind
speeds above 10 mph (see Figure 5.19). This difference in the calculated and
observed wind speeds for the onset of galloping can be attributed to several
factors: (1) the flutter derivatives (Hj) were obtained in the wind tunnel in
smooth flow, whereas the field experiments were conducted in the turbulent
wind, (2) wind-tunnel experiments were conducted in two-dimensional flow
conditions, whereas in the field conditions the wind flow is three dimensional,
and (3) there could be small differences between the one-quarter model used in
the wind tunnel experiments and the complex shape of the signal lights.
Except for the difference in the onset wind speed for galloping, the wind
tunnel results explain the observations made in the field experiments i.e.,
effective damping for the signal structure with no wing attached (Case A) is

79

m
(O
CC

CD

(n
CC

I M

Lli

0)
(0
CC

CO
CC

(O
CC

f I
CN
CO

00
CO

seB

cy

Case

<D
CO

O^

Case

O
CO

cu
O

o
CD
0)
CL

CO
C\J

CC-;

0/

CO

c
C\J
13
C/)

CO
CVJ

0)
>

CO

SI

CL

eff

CX3

C
d")
0

ed.

tvj

CD
Q.
C/)
D

c
OJ

o:

CL

E
CO
D
CD
.>
O
CD
+

LU
CO

CO

CVJ

Ti 1

o in o ir> o in o
o in o in
h- CO CO in in rr 1^ in
CO
CO O J C\J 1 -

o in o

in

t -

(UPeS 'Ql) 0 'luepji^eoQ BuidoiBQ eAipe^j^

80

IT)
CO

0)
13

a>

m
0)

CO

CO

CO

CO

LJJ

CD
CO

CO

CD
CO

CO

CC

CO

I ^I I i
CVJ

5f

CO

O
\

<
aY
0

30

CD

CD

Wind Speed, mph

10

12

14

16

18

20

22

24

26

Q)

28

3se B

ai

Cas

Cas(

ri

cq';j-cvjcvjoocO"^cvj
cvi cvi cvj
T-" -r-^ T- I-'

\
cn CO Tf CVJ o
CD CD C> C>

\i1

4 6

\>

CVJ Tf
o p

(BuidoiBQ iBoiiuQ %) 5 'oiJBu BuidiuBQ eAiJoejB

81

Figure 6.6 Effecltive d amp ng ratio v ersus wind spe

negative (aerodynamically unstable) and the effective damping for signal

structure with a wing attached over the 5-light signal at the tip of the arm (Case
B) remains positive (aerodynamically stable) for all wind speeds (see Figure
6.5). Hence, the wind tunnel results explain the aerodynamic behavior of the
various cases discussed above.
Except for Case A (at wind speeds higher than 20 mph) all the other
cases have positive effective damping, which suggests that Cases B, C, D, and
E are aerodynamically stable and do not exhibit galloping vibrations. The
question arises which of these four cases (B, C, D, or E) is the best vibration
mitigation strategy? In Figure 6.5, at wind speeds higher than 20 mph, Case-E
has the highest effective damping followed by Cases C, D, and B, in order.
Although Case E exhibits highest effective damping, Case B or Case C
are recommended as a vibration mitigation strategy. The reasons are:
1. Case B and C eliminate the force causing the vibrations, whereas
Case D and E compensate the force causing the vibrations with
positive aerodynamic damping.
2. In Case D and E, signal lights without wing attached above them will
contribute to negative aerodynamic damping force, which in turn
depends on the mode shape of vibration, number of signal lights and
their location.
A wing was used only above the 5-light signal light at the tip of the arm in
the field experiments. Wings can be attached to other lights on the arm, to
further increase the effective damping. Also, by using a wing with rounded
edges one can obtain even higher damping.
The wing used in the experiments was an aluminum plate, but the results
are valid even if it is made of some other equally strong material as long as it is
of same size and shape. It may be more practical, if the manufacturers of signal
lights make wings that attach directly to the signal lights (i.e., instead of
attaching them to the arm). In this way, wing and signal lights can be made
82

compatible in terms of connections, for ease of installation and removal. Also,

lighter and cheaper materials like plastics can be used instead of aluminum
plates. The shape of the wing can be further optimized to give higher (positive)
aerodynamic damping through additional wind tunnel testing.
6.2 Conclusions
The signal structures were observed to be vibrating under a narrow set of
conditions. These conditions were first identified from the tow tank experiments.
Large amplitude vibrations occur when the wind blows from the backside of the
signal lights with a backplate attached (see Configuration 5 in Figure 4.2).
These large amplitude vibrations are due to the galloping phenomenon, which is
caused by aerodynamic instability.
The mechanism that produces the large amplitude vibrations can be
explained as follows: At low wind speeds vortex shedding or gustiness in the
wind initiates small amplitude vibrations. The signal structures exhibit the
galloping phenomenon when the Den Hartog criterion is satisfied based on
quasi-steady theory or when negative-damping criterion is satisfied based upon
aeroelastic theory. After the onset of galloping, the overall structural damping
(effective damping) reduces until it becomes negative, leading to self-excited
motion with large amplitudes. This phenomenon was explained with aeroelastic
tests done on a wind-tunnel model of the signal light. The vibration amplitudes
increase with increase in wind speed until the amplitudes reach a limiting value
due to possible non-linearity in the system. The vibration amplitudes diminish if
the wind direction changes or the wind speed reduces.
A strategy of using a wing (damping plate) mounted above the signal
lights was tested in the field experiments and found to be an effective means of
vibration mitigation. A flat-plate wing with rounded edges was found to be a
better alternative to a flat-plate wing from the wind-tunnel experiments. The use
of a wing eliminates the building up of the high amplitudes in the first place, by

83

increasing the aerodynamic damping as demonstrated by the wind tunnel

experiments.

84

_:;i.V

.tlB

REFERENCES

Achenbach, E. and Heinecke, E., 1981: "On Vortex Shedding from Smooth and
Rough Cylinders in the Range of Reynolds Numbers 6 x 10^ to 5 x 10^,"
Journal of Fluid Mechanics, Vol. 109, pp. 239-251.
Baird, R. C , 1955: "Wind-Induced Vibrations of a Pipe-line Suspension Bridge
and Its Cure," Transactions, American Society of Mechanical Engineers, Vol.
77, No. 6.
Bearman, P. W., and Obasaju, E. D., 1982: "An Experimental Study of Pressure
Fluctuations on Fixed and Oscillating Square-Section Cylinders," Journal of
Fluid Mechanics, Vol. 119, pp. 297-321.
Blevins, R. D., 1977: Flow Induced Vibration, Van Nostrand Reinhold Company,
New York, New York.
Clough, R. W. and Penzien, J., 1980: Dynamics of Structures, McGraw-Hill Book
Company, New York, NY.
Den Hartog, J. P., 1932: "Transmission Line Vibration Due to Sleet,"
Transactions, AIEE, Vol. 51, pp. 1074-1076.
Den Hartog, J. P., 1956: Mechanical Vibrations, 4th Ed., McGraw-Hill, New
York, New York.
Jones, G. W., Cincotta, J. J. and Walker, R. W., 1969: Aerodynamic Forces on a
Stationary and Oscillating Circular Cylinder at High Reynolds Numbers,
NASA Technical Report 300, National Aeronautics and Space Administration,
Houston, TX.
McDonald, J. R., Mehta, K. C , Oler, W. J., and Pulipaka, N., 1995: Wind Load
Effects on Signs, Luminaires, and Traffic Signal Structures, a report
submitted to Texas Department of Transportation (TxDOT), Austin, Texas.
Novak, M., 1969: "Aeroelastic Galloping of Prismatic Bodies," Journal of the
Engineering Mechanical Division, ASCE, Vol. 95, No. EM1, pp. 115-142.
Novak, M., 1972: "Galloping Oscillations of Prismatic Structures," Journal of the
Engineering Mechanics Division, ASCE, Vol. 98, No. EMI, pp. 27-46.
Novak, M., and H. Tanska, 1974: "Effect of Turbulence on Galloping Instability,"
Journal of Engineering Mechanics Division, ASCE, Vol. 100, pp. 27-47.
85

Obasaju, E. D., 1983: "Forced-Vibration Study of the Aeroelastic Instability of a

Square-Section Cylinder Near Vortex Resonance," Journal of Wind
Engineering and Industrial Aerodynamics, Vol. 12, pp. 313-327.
Okajima, A., 1982: "Strouhal Numbers of Rectangular Cylinders," Journal of
Fluid Mechanics, Vol. 123, pp. 379-398.
Parkinson, G. V. and Brooks, N. P. H., 1961: "On the Aeroelastic Instability of
Bluff Cylinders," Transactions ASME, Journal of Applied Mechanics, Vol. 83 ,
pp. 252-258.
Parkinson, G. V. and Smith, J. D., 1962: "An Aeroelastic Oscillator with Two
Stable Limit Cycles," Transactions ASME, Journal of Applied Mechanics, Vol.
84, pp. 444-445.
Parkinson, G. V. and Wawzonek, M. A., 1981: "Some Considerations of
Combined Effects of Galloping and Vortex Resonance," Journal of Wind
Engineering, Vol. 8, pp. 135-143.
Price, P., 1956: "Suppression of the Fluid-Induced Vibration of Circular
Cylinders," Journal of the Engineering Mechanical Division, Vol. 82, pp.1030.
Roshko, A., 1955: "Experiments on the Flow Past a Circular Cylinder at Very
High Reynolds Number," Journal of Fluid Mechanics, Vol. 10, pp. 345-356.
Sarkar, P. P., 1992: New Identification Methods Applied to the Response of
Flexible Bridges to Wind, Ph.D. dissertation submitted to The Johns Hopkins
University, Baltimore, Maryland.
Scruton, C , 1963: On the Wind-Excited Oscillations of Stacks, Towers and
Masts, Paper 16, Britain National Physical Laboratory.
Scruton, C. and Walshe, D. E. J., 1957: A Means for Avoiding Wind-Excited
Oscillations of Structures of Circular or Nearly Circular Cross-Section,
unpublished report NPL/Aero/335. (Also: British Patent No. 907,851.)
Shiraishi, N., 1971: "An investigation on Aerodynamic Response of Plate-Like
Structures in Fluctuating Gusts," Proceedings of The Third International
Conference on Wind Effects on Buildings and Structures, Tokyo, Japan.
Simiu, E. and Scanlan, R. H., 1986: Wind Effects on Structures, Second Ed.,
John Wiley and Sons, New York, NY.

86

/ . H i i V -#..

Szepessy, S. and Bearman, P. W., 1992: "Aspect Ratio and End Plate Effects
on Vortex Shedding from a Circular Cylinder," Journal of Fluid Mechanics,
Vol. 234, pp. 191-217.
Titan Corporation, 1993: Stardyne, Version 4.0, Charsworth, California 91311.
Theodorsen, T., 1940: NACA Report 685.
TxDOT, 1990: Traffic Control Standard Sheets, State Department of Highways
and Public Transportation, Austin, Texas.
Walshe, D. E. and Wootton, L R., 1970: "Preventing Wind-Induced Oscillations
of Structures," Proceedings, Institute of Civil Engineers, Vol. 47, pp. 1-24.
Woodgate, L. and Maybrey, J. F. M., 1959: Further Experiments on the Use of
Helical Strakes for Avoiding Wind-Excited Oscillations of Structures of
Circular or Nearly Circular Section, Unpublished Report NPLyAero/381.

87