Professional Documents
Culture Documents
by
IAN E. HOSCH
TALAT SALAMA
VIRGINIA SISIOPIKU
HOUSSAM TOUTANJI
NASIM UDDIN
A DISSERTATION
BIRMINGHAM, ALABAMA
2009
Copyright by
Ian E. Hosch
2009
ii
DESIGN OF HIGHWAY OVERHEAD CANTILEVER-TYPE SIGN SUPPORT
STRUCTURES FOR FATIGUE LOADS
IAN E. HOSCH
ABSTRACT
The 2001 edition of the American Association of State Highway and Transporta-
tion Officials (AASHTO) Standard Specifications for Structural Supports for Highway
Signs, Luminaires and Traffic Signals has been revised in its entirety through a major re-
search project conducted under the auspices of the National Cooperative Highway Re-
search Program (NCHRP 17-10). A major part of the revision included updated provi-
sions and criteria for extreme wind loads and new provisions and criteria on fatigue de-
sign. These provisions differ considerably from those in previous editions of the specifi-
The impact of the fatigue criteria on the design of overhead sign structures has not
been fully evaluated. The fatigue design loads do not adequately reflect the stresses gen-
erated on these structures from wind-induced fatigue loading. In addition, the provisions
do not account for the variety of support structures in design, each with different configu-
ration, sizes, shapes, and material properties that influence vibration behavior. As a re-
sult, the vulnerability of sign support structures to wind-induced fatigue loading is not
fully realized.
The main goal of the project was to conduct theoretical and experimental pro-
support structures subjected to wind induced fatigue loads. A theoretical program was
iii
developed that took into account the variety of sign supports structures in design, as well
experimental program was developed to evaluate the accuracy of the theoretical study. A
finite element analysis program was conducted to simulate the wind-induced loading en-
vironment and the response of sign support structures to this environment. Alterations
were made to the model that could not be done experimentally due to costs and time re-
straints. The results of the finite element analysis were compared to the theoretical and
The developed information and criteria on fatigue design of sign support struc-
tures were used to develop fatigue design loads to provide an improved and more reliable
design method. Recommendations are made to update the current specifications to in-
Keywords: Cantilever-type sign support structures, fatigue design, fatigue load, natural
iv
DEDICATION
mother Sandra J. Haigh who gave me strong will and unconditional support during my
time as a student.
v
ACKNOWLEDGEMENTS
It was important to acknowledge the following people who helped with the re-
Richard Hawkins and Dr. Jason Kirby who helped with the important technical
Daniel Jones and Ricky Love of the Alabama Department of Transportation who
The Morrison Family, Brett and Dave, who helped specifically for making the ex-
Andrew Sullivan for setting up and reviewing the traffic recorders required with
vi
TABLE OF CONTENTS
Page
ABSTRACT....................................................................................................................... iii
DEDICATION .....................................................................................................................v
ACKNOWLEDGMENTS ................................................................................................. vi
CHAPTER
1 INTRODUCTION .........................................................................................................1
vii
Page
Creamer et al (1979)........................................................................................26
viii
Page
ix
Page
x
Page
xi
Page
APPENDIX
xii
LIST OF TABLES
Table Page
6.3 Estimated Microstrain Range at Fatigue Wind for Anchor Bolts --------------------- 87
6.4 Anchor Bolt Stress Range and Failure Index for Natural Wind ---------------------- 88
6.5 Anchor Bolt Microstrain Ranges from the Truck Tests -------------------------------- 91
6.6 Anchor Bolt Stress Range and Failure Index for Truck Tests------------------------- 92
6.7 Estimated Microstrain Range at Fatigue Wind for Section AA --------------------- 100
6.8 Section AA Stress Range and Failure Index for Natural Wind --------------------- 102
6.9 Section AA (Normal) Microstrain Ranges for the Truck Tests --------------------- 104
6.10 Section AA (Shear) Microstrain Ranges from the Truck Tests --------------------- 105
6.11 Section AA Stress Range and Failure Index for Truck Tests ----------------------- 107
6.12 Estimated Microstrain Range at Fatigue Wind for Chords ------------------------- 115
6.13 Chord Stress Range and Category E Failure Index for Natural Wind ------------- 117
6.14 Chord Stress Range and Category B Failure Index for Natural Wind ------------ 118
6.15 Chord 1 Microstrain Ranges from the Truck Tests ----------------------------------- 119
xiii
Table Page
6.16 Chord 2 Microstrain Ranges from the Truck Tests ----------------------------------- 120
6.17 Chord 3 Microstrain Ranges from the Truck Tests ----------------------------------- 120
6.18 Chord 4 Microstrain Ranges from the Truck Tests ----------------------------------- 121
6.19 Chord Stress Range and Category E Failure Index for Truck Tests---------------- 122
6.20 Chord Stress Range and Category B Failure Index for Truck Tests --------------- 123
6.21 Experimental Stress Ranges for Natural Wind Tests --------------------------------- 124
6.22 Experimental Stress Ranges for Truck-Induced Wind Tests ------------------------ 124
8.1 Collection Dates and Times for Natural Wind ---------------------------------------- 158
xiv
Table Page
13.1 Experimental and Theoretical Results for Natural Wind Gust --------------------- 312
13.2 Comparison of Results from Experimental and Theoretical Programs ----------- 321
13.3 Design Case Description for Natural Wind Gust ------------------------------------- 330
13.4 Design Case Results for Natural Wind Gust ------------------------------------------ 331
13.5 Design Case Description for Truck-Induced Wind Gust ---------------------------- 335
xv
Table Page
13.6 Design Case Results for Truck-Induced Wind Gust --------------------------------- 337
14.1 Experimental Stress Ranges for Natural Wind Tests -------------------------------- 341
14.2 Experimental Stress Ranges for Truck-Induced Wind Tests ----------------------- 342
xvi
LIST OF FIGURES
Figure Page
3.1 Wind velocity PSD for annual mean wind velocity 11 mph. -------------------------- 31
3.2 Force PSD for annual mean wind velocity 11 mph. ------------------------------------ 33
xvii
Figure Page
xviii
Figure Page
4.33 Typical test setup with van data acquisition system. ----------------------------------- 66
6.4 All anchor bolt microstrain range vs. wind velocity. ----------------------------------- 84
6.9 Depth chart of anchor bolt strain range in truck tests. ---------------------------------- 90
xix
Figure Page
6.19 Estimated microstrain range at fatigue wind for Section AA. ---------------------- 101
6.20 Failure indexes for natural wind of Section AA. ------------------------------------- 103
6.21 Depth chart of Section AA normal strain in truck tests. ----------------------------- 105
6.22 Depth chart of Section AA shear strain in truck tests. ------------------------------- 106
6.31 Estimated strain range at fatigue wind for SG-C2-5. -------------------------------- 115
6.33 Failure indexes of chord gauges for natural wind. ----------------------------------- 118
6.35 Comparison of results between natural wind and truck gusts. --------------------- 125
xx
Figure Page
8.4 Strain values per 0.5 ph wind velocity intervals. -------------------------------------- 167
xxi
Figure Page
8.16 Cross section of post at Section AA and Section BB. ------------------------------- 186
8.19 Wind velocity vs. wind pressure for Section AA. ------------------------------------ 191
8.20 Wind velocity vs. wind pressure for Section BB. ------------------------------------ 192
8.21 Wind velocity vs. wind pressure for Rosettes. ---------------------------------------- 192
8.22 Wind velocity vs. wind pressure for all sections. ------------------------------------ 193
9.2 Wind velocity PDS for annual mean wind velocity. --------------------------------- 202
xxii
Figure Page
9.3 Force PDS for annual mean wind velocity. ------------------------------------------- 204
9.4 Force PDS using limit-state wind velocity 37 mph. --------------------------------- 206
9.7 Best fit line approximating the average wind velocity PDS. ----------------------- 211
9.8 Theoretical plot of the experimental average wind velocity PDS. ----------------- 212
9.9 Approximation of the experimental average wind pressure PDS. ----------------- 213
9.13 RMS wind pressure VRS for 1.82% damping. --------------------------------------- 221
9.15 RMS wind pressure VRS with damping equal to 0.5%. ---------------------------- 225
9.18 Peak-to-peak VRS plots versus average wind velocity. ----------------------------- 229
9.20 Fatigue load comparison between experimental and VMS. ------------------------ 231
9.21 Fatigue wind pressure VRS for 1.82% damping. ------------------------------------ 232
9.22 Fatigue wind pressure VRS for damping range. -------------------------------------- 233
9.23 Fatigue wind VRS of Supports Specifications (2% damping). --------------------- 234
10.1 Truck and driver for truck-induced wind test. ---------------------------------------- 237
xxiii
Figure Page
10.8 Cross section of strain gauge location Section AA and Section BB. -------------- 254
11.3 Structural response time history due to the Control impulse. ---------------------- 267
12.5 Modal shapes of the first five modes from FEA. ------------------------------------- 281
12.6 Average wind pressure PDS for FEA input. ------------------------------------------ 296
12.8 VRS for the aluminum FEA model with 0.5% damping. --------------------------- 299
xxiv
Figure Page
13.2 Fatigue wind pressure VRS for damping range. -------------------------------------- 319
13.4 Direction of wind loading for natural wind gust. ------------------------------------- 330
13.9 Direction of wind loading for truck-induced wind gust. ---------------------------- 336
14.1 Comparison of results between natural wind and truck gusts. --------------------- 344
14.3 Fatigue wind pressure VRS for damping range. -------------------------------------- 350
xxv
LIST OF ABBREVIATIONS
AC accelerometer
DIA diameter
DOF degree-of-freedom
FT Fourier transform
NA not available
RMS root-mean-square
SG strain gauge
xxvi
S-N stress vs. number of cycles
xxvii
1
CHAPTER 1
INTRODUCTION
Problem Statement
specifically for sign support structures. Likewise, there is little knowledge and under-
standing of this subject. In addition, there is a lack of accountability for the variety of
sign support structures in design each with different configuration, sizes, shapes, and ma-
terial properties that influence vibration behavior. Sign support structures are highly
flexible with very low damping properties which make them highly susceptible to wind-
induced fatigue loading, and as such the vulnerability of these structures is not com-
pletely realized. As a result, currently published fatigue design equations do not ade-
quately reflect the stresses generated on these structures from wind-induced fatigue load-
ing.
The 2009 AASHTO Standard Specifications for Structural Supports for Highway
Signs, Luminaires and Traffic Signals (1-6) (hereafter referred to as Supports Specifica-
tions) include fatigue design criteria for cantilevered overhead sign structures that could
significantly impact the design of these structures. Fatigue loadings due to four wind
Galloping, and
Vortex shedding.
Fatigue loads due to natural wind and truck-induced wind gusts are applicable to
overhead sign structures and are the focus of this study. Galloping may occur in overhead
sign structures, but 4-chord trusses (which are analyzed in this study) are not susceptible
to galloping. Vortex shedding is not applicable to the overhead sign structures of this
study.
In April 2006, a cantilevered overhead sign structure located at the I-565 and I-65
Interchange, Exit 340B, (Figure 1.1) failed due to fatigue of the anchor bolts. Fatigue
fracture of the anchor bolts due to combined axial stresses and bending in the bolts was
noted. The photographs in Figure 1.2 show the cantilever overhead sign that failed. A
close-up of the bolt fracture is shown in Figure 1.3. The bolt layout was designed using
an earlier version of the Supports Specifications that did not consider fatigue design.
The main goal of this project was to evaluate the performance of cantilever-type
highway overhead sign structures subjected to wind-induced fatigue loads resulting from
natural wind and truck-induced wind gusts, and to develop more reliable fatigue loads for
design. The project involved detailed theoretical as well as experimental programs to ad-
Nashville
I-565
I-65
Huntsville
Decatur
Exit
340B
Birmingham
Objective
The main objectives of this work can be enumerated in the following three initia-
tives:
fatigue loads.
2. Use this information to develop fatigue design loads to provide an improved and
3. Compare the developed criteria with the Supports Specifications (1), and recom-
Study Initiatives
Study initiatives of this study to accomplish the main objectives are enumerated
as follows:
2. Instrument one cantilever-type highway overhead sign structure with the nec-
take field measurements under different natural wind and truck induced wind
gust conditions.
3. Evaluate the fatigue resistance of the structure to wind loading using the ex-
most susceptible to fatigue and evaluate their performance under fatigue con-
ditions, and establish sections of the structure best suitable for calculating fa-
5. Compare the experimental fatigue load criteria to the results and conclusions
6. Create finite element models and check the developed fatigue load criteria
cations of the structure. Make alterations to the FEA model and evaluate the
Work Plan
Summary
tures was developed. The theoretical program addressed the variety of sign support struc-
ture used in design each with different configurations, sizes, shapes, and material proper-
ties. Finite element models were developed to perform loading simulations of the devel-
(shown in Figure 1.4) was selected for field measurement. The structure was instru-
mented with electric strain gages to determine stresses at critical locations. Accelerome-
ters and anemometers were mounted on the structure to determine structural dynamic
properties and wind velocities behavior. Two types of applied loads were considered in
measuring of forces in base plate vicinity including the anchor bolts. Recorded data was
analyzed and compared to developed analytical methods, used to verify the procedures of
Project Tasks
The project involved an extensive analytical program to evaluate the fatigue loads
on sign support structures. An experimental program was performed to check the accu-
8
racy of the theoretical program on these structures. Fatigue loading criteria were devel-
oped and recommendations were proposed to provide a more reliable fatigue design
method. The objectives for this project were accomplished through the following tasks:
structures to wind-induced fatigue loading. The model was created in such a way as to
account for the variety of sign support structures, each with different sizes, shapes, con-
figurations and material properties. Emphasis was placed on the vibration of the structure
due to wind-induced fatigue loading. The design fatigue load was determined based on
the dynamic characteristics of the structures. The infinite-life approach to fatigue design
was performed.
The loading criteria were used as input for computer software to check critical
stresses in the main members, and for evaluating the accuracy of the loading criteria. The
SAP 2000 v. 10 finite element analysis (FEA) computer software was used (11). A three-
dimensional full-scale cantilever structure was modeled in the FEA program. The model
was based on the shop drawings provided by ALDOT, and was same structure used in the
experimental program for field measurements and testing. A modal analysis was con-
ducted as well as static and dynamic loading simulations including material alterations.
The site selection of the experimentally tested cantilever structure was located at
the same place where the previous one failed (see Figure 1.1).
9
The support structure was instrumented with strain gauges, accelerometers, and
anemometers. Electric strain gages were placed at maximum stress locations on the verti-
cal and horizontal structural members of the overhanging section, and in the vicinity of
the base plate to determine strains under the different loading conditions. Accelerometers
were placed on the post and the cantilever overhang to determine the major dynamic
properties. Wind speed readings were recorded using anemometers placed 4 ft (1.22 m)
above the top of the post (to prevent shielding effects) for ambient readings, and along
the cantilever section for truck induced gusts measurements. Each transducer provided
The overhead sign structure was tested under loading conditions: a) Natural wind
gust, and b) Truck induced wind gust. The natural wind data was taken over an extended
time period in an effort to capture the predominant natural wind gusts. Wind data from
the National Weather Service near Huntsville, AL, in the form of the annual mean wind
velocity for the area, was determined to help schedule testing days, and to compare to the
wind measurements taken to distinguish if the measured results were representative of the
wind environment. A standard semi–trailer vehicle was used to apply the truck induced
wind gusts. Varying truck speeds were used and the corresponding structural response
The data was analyzed using the critical stresses determined in the structural
the fatigue resistance was performed first to understand the distribution of stresses in the
structure, and to determine areas of the structure that were most vulnerable to fatigue
damage. A modal analysis was also performed using the experimental measurements.
Structural dynamic characteristics of the structure were determined such as the modal
frequencies, modal shapes, and damping characteristics. The analysis was used in the de-
Fatigue loading was developed from the data analysis after accomplishing an un-
derstanding of the structural behavior to wind loading and the distribution of stresses. The
infinite life approach to fatigue design was utilized and applied. The data collected of
natural wind gust were analyzed as a randomly occurring continuous load, whereas the
truck induced data were evaluated based on principles related to transient loading envi-
ronments.
Design fatigue load recommendations for natural wind and truck-induced gusts
were developed after the completion of Task 5. The comparison of the analytical results
to the experimental data was utilized and fatigue design criteria were developed that en-
compassed the results of these efforts. The recommendations addressed the “accuracy” of
The effect of the proposed provisions was assessed and explained by performing
fatigue load calculations for design. The examples compared fatigue loads using the fa-
tigue provisions of Supports Specifications and the fatigue loads according to the pro-
The fatigue provisions of Supports Specifications have not been fully used or
evaluated by most state DOTs, and as such, insufficient information is available on the
design of structural supports using the current fatigue provisions of the specifications.
The provisions were developed primarily through analytical methods, and therefore an
urements has not been performed. In addition, the provisions are only applicable to spe-
cific types of support structures, and changes in size, shape, and material properties
would result in significant different vibration behavior and subsequent fatigue load. A
study is needed to develop fatigue loads that apply to all type of support structures, and
reflect reliable loading criteria to be used in design. A methodology is needed that reveals
the vulnerability of sign supports structures to wind-induced fatigue loading, from which
Dissertation Outline
chapters with a brief description of the work involved. The dissertation is presented in the
following outline:
of the project are stated with a breakdown of the tasks needed to accomplish the
projects objectives.
the project is provided. The most relevant material is presented in a form as to de-
scribe the brief historical development of the fatigue guidelines for support struc-
tures.
duction to the fatigue provisions for natural wind and truck-induced gusts in the
Chapter 5 Structural Testing for Natural Wind and Truck-Induced Gusts: A de-
scription on the development of the testing program is provided. The testing pro-
support structure using the experimentally collected data is provided. The analysis
the proper approach to calculate the fatigue load from the experimental data.
performed using the experimentally collected data. The analysis helped to under-
stand the dynamic behavior of the structure needed primarily for the analytical
program.
Gust: The procedure used in the reduction of data collected experimentally for
natural wind gust. The methodology and calculation of the fatigue load due to
Chapter 9 Theoretical Calculation of the Fatigue Load due to Natural Wind Gust:
An analytical program was developed to calculate the fatigue load due to natural
wind guts. The method was developed to account for the variety of sign support
14
structures in design, each with different configurations, sizes, shapes, and material
properties.
Gust: The procedure used in the reduction of data collected experimentally for
truck-induced wind gust. The methodology and calculation of the fatigue load due
Gust: An analytical program was developed to calculate the fatigue load due to
truck-induced wind guts. The method was developed to account for the variety of
sign support structures in design, each with different configurations, sizes, shapes,
Chapter 12 Finite Element Analysis: The finite element analysis program is pre-
sented. The analysis was conducted to verify the structural analysis used to calcu-
late the fatigue load. The analytical programs were also verified by changing the
for natural wind and truck-induced gusts are presented. Design examples are pro-
vided to compare with the current provisions in the Supports Specifications and to
of the project. Ranking schedules are also included that helped in scheduling test-
CHAPTER 2
LITERATURE REVIEW
Chapter Overview
sign support structures. A spreadsheet was developed to categorize the reviewed studies
by breaking down the research into important areas of interests. Interests included rec-
structures. The cataloging helped to develop and systemize the research program. Also
provided in the literature review chapter are descriptions of the most relevant studies that
played a crucial role in the development and progress of the fatigue provisions of the
Supports Specifications, as well as other projects that were considered important to the
Various research papers and documentations on sign support structures were re-
viewed before developing the research program. Important aspects that had relevancy to
this project were identified during the review process. Each of the documents reviewed
were tagged and categorized regarding the predetermined interests of this study. A
spreadsheet was created that helped to label the reviewed studies with respect to the
tagged aspects. Figure 2.1 displays the spreadsheet showing the properties of interests for
17
this project and the color coded categorizing method. The numbers at the top of the
spreadsheet reference the papers (Figure 2.2) from which the color coded categories were
indentified (7-10, 12, 15-17, 19, 20, 33, 36, 37, 50, 52, 53, 55, 59, 60, 64). Only the most
important references related to supports structures are listed in the figure. When reference
to a particular subject was needed during the project, the spreadsheet was used by utiliz-
ing the legend with respect to the property of interest, and identifying the documentation
that contained information on the subject. This process helped to allow the research to
Referenced Papers
Property of Interest
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Cantilever-Type Highway Sign Support Structure
Cantilever-Type VMS Support Structure ∆
Bridge-Type Highway Sign Support Structure
Bridge-Type VMS Support Structure ∆
High Mast Structure
Traffic Signal Structure
Luminare Structure
Strain Guage Transducers
Anemometer Transducers
Accelerometer Transducers
Pressure Transducers
Galloping
Vortex Shedding
Anchor Bolt Study
Natural Wind Gusts: Fatigue Loading ∆ ∆ ∆ ∆ ∆
Natural Wind Gusts: Fatigue Resistance ∆
Truck-Induced Wind Gusts: Fatigue Loading ∆ ∆ ∆ ∆ ∆ ∆ ∆ ∆
Truck-Induced Wind Gusts: Fatigue Resistance ∆
Horizontal Loading for Truck Gust ∆
Suction Pressure for Truck Gust ∆
Variation in Truck Type
Spectral Analysis
Natural Frequencies of Support Structures ∆
Damping of Support Structures
Modal Shapes of Support Structures ∆
Steel Structure
Aluminum Structure
Design Example
Legend
Involved Experimental Program
Invovled Analytical Program
Involved Experimental & Analytical Programs
∆ Includes Equation
Referenced Papers
Cook, Ronald A., Bloomquist, D., Agosta, A.M., Taylor, K.F., Wind Load Data for Variable Message Signs . Report
1
Number 0728-9488. Florida Department of Transportation, Research Management Center, April, 1996.
Azzam, D., Fatigue Behavior of Highway Welded Aluminum Light Pole Support Structures. Dissertation, University of Adron,
2
May 2006.
Creamer, B. M., Frank, K. H., Klingner, R. E., Fatigue Loading on Cantilever Sign Support Structures from Truck Wind
3 Gusts. Research Report Number 209-1F. Texas State Department of Highways and Public Transportation, Transportation
Planning Division, April, 1979.
Dexter, R. J., Ricker, M. J., Fatigue-Resistant Design of Cantilevered Signal, Signs, and Light Supports . NCHRP Report
4
469, The Transportation Research Board, Washington D.C. 2002
Zalewski, B., Huckelbridge, A., Dynamic Load Environment of Bridge-Mounted Sign Support Structures. Report No.
5
ST/SS/05-002. Ohio Department of Transportation, Office of Research and Development, September 2005.
Foutch, D.A., Kim, T.W., LaFave, J.M., Rice, J.A. Evaluation of Aluminum Highway Sign Truss Designs and Standards
6 for Wind and Truck Gust Loadings. Research Report N0. 153. Illinois Department of Transportation, Bureau of materials
and Physical Research, December 2006.
Ginal, S. Fatigue Performance of Full-Span Support Structures Considering Truck-Induced Gust and Natural Wind
7
Pressures. Thesis, Marquette University. December 2003.
Kaczinski, M.R., Dexter,R.J., and VanDien, J.P. Fatigue Resistant Design of Cantilevered Sign, Signal and Light
8
Supports. NCHRP Report 412. Transportation Research Board. Washington D.C. 1998.
South, S.M. Fatigue Analysis of Overhead Sign and Signal Structures. Report No. 115. Illinois Department of
9
Transportation, Bureau of Materials and Physical Research. May 1994.
Fouad, F.H., Calvert, E.A., and Nunez, E. Structural Supports for Highway Signs, Luminaires, and Traffic Signals.
10
NCHRP Report 411, Transportation Research Board. Washington D.C. 1998.
Edwards, J.A., and Bingham, W.L. Deflection Criteria for Wind Induced Vibrations in Cantilever Highway Sign
11
Structures. Report No. FHWA/NC/84-001, Center for Transportation Engineering Studies, North Carolina State University,
DeSantis, P.V., and Haig, P.E., Unanticipated Loading Causes Highway Sign Failure. Proceedings of ANSYS Convention,
12
1996.
Albert, M.N., Manuel, L., Frank, K.H., and Wood, S.L., Field Testing of Cantilevered Traffic Signal Structures under
13 Truck-Induced Gust Loads . Report No. FHWA/TX-08/0-4586-2, Center for Transportation Research, University of Texas
at Austin, 2007.
Cali, P., and Covert, E.E., On the Loads on Overhead Sign Structures in Still Air by Truck Induced Gusts. Wright
14
Brothers Facility Report 8-97, Massachusetts Institute of Technology.
Ramy, A.S., Fatigue Resistant Design of Non-Cantilevered Sign Support Structures. Thesis, University of Alabama at
15
Birmingham, 2000.
Fisher, J.W., Nussbaumer, A., Keating, P.B., and Yen, B.T., Resistance of Welded Details Under Variable Amplitude Long-
16
Life Fatigue Loading. NCHRP Report 354, The Transportation Research Board, Washington D.C., 1993.
Irwin, H.P., and Peeters, M. An Investigation of the Aerodynamic Stability of Slender Sign Bridges, Calgary. LTR-LA-
17
246, national Research Council Canada-Aeronautical Establishment, 1980.
McDonald, J.R., Mehta, K.C., Oler, W., and Pulipaka, N., Wind Load Effects on Signs, Luminaires and Traffic Signal
18 Structures. Texas Department of Transportation Report No. 1303-1F, Wind Engineering Research Center-Texas Tech
University, Lubbaock, TX, 1995.
Gilani, A.S., Chavez, J.W., and Whittaker, A.S., Fatigue-Life Evaluation of Chaneable Message Sign Structures, Volume
19 1 - As Built Structures. Report No. UCB/EERC-97/10, Earthquake Engineering Research Center, University of California,
Berkeley, CA, 1997.
Kashar, L., Nester, M.R., Johns, J.W., Hariri, M., and Freizner, S., Analysis of the Catastrophic Failure of the Support
20 Structure of a Changeable Message Sign. Structural Engineering in the 21st Century, Proceedings of the 1999 Structures
Congress, New Orleans, LA, 1115-118, 1999.
Many aspects that were considered important by the researchers were not found in
the literature and were therefore excluded from the spreadsheet. Some aspects were sim-
ply dropped from the spreadsheet as the research progressed because of relevancy,
whereas other aspects were added which resulted in the breakdown shown in the Figure
2.1.
19
The research paper breakdown served as a useful tool during the development of
the research program. Relevant information on gathered literature documents was easily
and quickly identified to extract information. The breakdown also helped to identify areas
where research is needed or is lacking. Only the most prominent papers as they related to
the research performed with this project was included in the breakdown. A complete list
of reviewed papers is provided in the List of References at the end of the dissertation.
Research Review
A description of the reviewed studies that played a crucial role in the development
of the fatigue provisions in the Supports Specifications are listed as follows. Other pro-
jects included in the literature review were considered important to this research and are
described.
The fatigue provisions for truck-induced wind gust in the Supports Specifications
were initially based on this study. The research focused on a cantilever-type overhead
VMS support structure. The structure failed due to fatigue loading for which prompted
the study. After the structure was replaced, large deflections were observed due to wind
gust created from passing trucks. In the analysis, the researchers assumed that the veloc-
ity of the wind gusts onto the structure was equal in magnitude to the speed of the truck,
in addition to a gust factor equal to 1.3 to account for head winds. The wind pressure was
calculated using the fundamental fluid mechanics relationship between wind force and
the square of the wind velocity. The resulting wind pressure was doubled to account for
20
the upward deflection of the sign plus the downward deflection due to the pull of gravity.
The concluded value represented a pressure range to be used for fatigue design, and is
where
PTG = design fatigue pressure load due to truck - induced wind gust
C d = drag coefficient
The researchers performed a FEA model of the structure using the software pack-
age ANSYS. The cantilevered end of the structure was observed in the field to deflect
vertically about 1 ft (0.305 m) in length after exposed to wind gust from passing trucks.
Experimentation was not performed to validate the tip deflection other than visual obser-
vation. To help verify the observations, the researchers back-calculated the wind pressure
that would theoretically produce a 1 ft (0.305 m) deflection, and inputted the pressure
into the ANSYS program of the modeled structure. The results matched the wind pres-
sure calculated from Eq. 2.1 and was concluded as the appropriate design load based on
Davenport (1961)
Davenport’s research was not focused on overhead sign support structures spe-
cifically, but rather focused on the characteristics of wind velocity. His research has pro-
vided an accurate model for simulating wind velocity behavior. The model was used in
21
the development of the fatigue provisions for natural wind gust in the Supports Specifica-
tions. It simulated the randomness along with the gustiness and turbulence commonly
associated with wind velocity. The model was developed in the form of a power density
spectrum generally used for predicting randomly occurring events. The power density
spectrum was created using experimentally measured wind velocity time histories gath-
ered from sites located around the world. An empirical formulation was developed as a
function of the mean wind velocity, and is shown in Eq. 2.2. Terrain coefficients were
also indentified the formulation process, and are shown in Table 2.1 (13).
4κV102 x 2
Sv ( f ) = 4
[Eq. 2.2]
f (1 + x )
2 3
where
S v ( f ) = wind velocity power spectral density at any height
f = frequency
V10 = mean wind velocity at a stadard height of 10 meters above ground leve
lκ = surface drag coefficient (Table 2.1)
1200 f f
x= 2
with 2 in cycles per meter.
V10 V10
Type of Surface κ α
Open unobstructed country (e.g., prairie-type grass-
0.005 0.15
land, arctic tundra, desert)
Country broken by low clustered obstructions such as
0.015 - 0.020 0.27 – 0.31
trees and houses (below 10 m high)
Heavily built-up urban centers with tall buildings 0.050 0.43
22
Kaczinski et al (1998)
The research performed by Kaczinski reported in NCHRP Report 412 formed the
framework of the fatigue provisions for galloping, vortex shedding, natural wind gust,
and truck-induced wind gusts in the Supports Specifications. The majority of his work
related to natural wind and truck-induced wind gusts was entirely theoretical. By using
the research performed by DeSantis and Haig, the fatigue provisions for the truck-
induced gust were created. Equation 2.1 was recommended as the appropriate pressure
Davenport’s wind velocity power density spectrum was used to create the fatigue
provisions for natural wind gust. The velocity spectrum was transformed into a wind
force spectrum. The spectrum was applied as input into an FEA program to generate a
stress response spectrum for four particular overhead sign support structures. The stress
spectrum was formed at different locations in the modeled structure and compared. The
fatigue load was then developed using the infinite-life approach. Equation 2.3 was devel-
oped from the results and was recommended as the appropriate pressure load to use for
where
PNW = design fatigue pressure load due to natural wind gusts
C d = drag coefficient
I F = importance factor
23
Report 412 also focused on developing S-N curves (stress vs. number of cycles)
for anchor bolt connection details. The results were used to create the constant-amplitude
for anchor bolt design and structural analysis were provided. In addition, Report 412 in-
troduced “Importance Factors” to be used with the Supports Specifications. Fatigue de-
sign examples are provided for different types of supports structures including overhead
sign, traffic signals, and luminaires. The examples provided a thorough design procedure
using the proposed fatigue design equations. Stresses at critical details were calculated
Fouad et al (1997-2005)
The research by Fouad et al in NCHRP project 17-10 and 17-10(2) (32, 33)
evaluated fatigue design of sign support structures. The research was compiled into a
specification based on the Allowable Stress Design (ASD) methodology, the basis of
which the current Supports Specifications were formed. The impact of the new specifica-
tions was analyzed with significant conclusions. The research brought together important
fatigue concerns such as galloping, vortex shedding, natural wind, and truck-induced
gusts. Many design examples were made which provided the framework for engineers to
use for fatigue evaluations. The work performed by Fouad et al compiled all relevant in-
formation on fatigue of sign support structures into a single stand-alone document. It now
resides as a well established document for fatigue design that is specific to sign support
structures, from which helped to propel future enhancements and innovations to fatigue
design of these structures. The researchers also provided valuable information on trans-
24
forming the specification into a Load and Resistance Factor Design (LRFD) methodology
in the future. The focus was to first develop a much needed stand-alone specification us-
ing the ASD method, and then upgrade to LRFD in the future (21-35).
The research performed by Dexter and Ricker reported in NCHRP Report 469
formed the framework of the current Supports Specifications. Their research involved
experimental and analytical programs. They concluded that the fatigue provisions for
natural wind gusts in the Supports Specifications were accurate. However, the truck-
values. A reduction in the truck provisions were recommended by Eq. 2.4. The reduction
did not reflect the experimental values gathered in the report because they were signifi-
cantly less the fatigue provisions, and it was believed the committee would not accept
them. The authors were also concerned with the type of structures used in the analysis.
The fatigue load was determined from analysis on variable message signs which are sig-
nificantly heavier than conventional signs. The reduction of the fatigue design equation
where
PTG = design fatigue pressure load due to truck - induced wind gust
C d = drag coefficient
I F = importance factor
25
Report 469 also provided a reduction to the fatigue load to accommodate for
height of the structure above the ground. They recommended a linear reduction in wind
pressure starting from 19.7 ft (6.00 m) above ground level to zero at approximately 32.8
ft (10.0 m) above ground level. A recommended area on the support structure to apply the
truck load was also provided. It was proposed to apply a uniformly distributed load to the
12 ft (3.66 m) area of the structure that produces the maximum stress range. This was
recommended as opposed to applying the load to portions of the structure that is not di-
Cook et al (1996)
The research performed by Cook et al is indentified in this review due to its rele-
vancy to fatigue design due to truck-induced gusts. Their research determined the truck-
induced wind gusts through direct measurement of wind pressure. Pressure transducers
were placed on a bridge spanning a highway to measure the wind pressure from passing
trucks. Importantly, the bridge was extremely rigid so that the measured values were not
influenced by the type of structure. Twenty-three loading events were measured and re-
corded.
The results provided a clear view of the wind behavior created from passing
trucks. It revealed a biaxial loading event with a vertically applied pressure and a hori-
zontally applied pressure. A suction event was also recorded as the truck passed under-
neath the structure. The results indicated a ramped loading impulse. Variation of the
truck-induced wind gust with respect to height above ground level was also measured. A
10% pressure reduction per foot above 17 ft (5.18 m) above ground level was found (12).
26
Creamer et al (1979)
The work performed by Creamer et al was very influential for future research.
Their research focused on fatigue induced loading due truck wind gusts. Where Cook’s
determined the truck wind pressure directly using pressure transducers, Creamer’s work
determined the truck wind pressure indirectly. They measured strain variations at critical
locations of a cantilevered structure due to truck wind gusts, and back calculated for wind
pressure. A ramped impulse function was developed and inputted into a FEA model of
the tested structure. The loading simulation produced similar strains on the structure as
CHAPTER 3
Chapter Overview
gusts fatigue provisions within the AASHTO Standards Specifications is provided. This
was believed to be an important and required task so that a complete understanding of the
developed fatigue provisions was obtained before addressing their accuracy. The provi-
sions for natural wind were formed indirectly through theoretical calculations. They were
developed using the infinite-life approach to fatigue design. Spectral analysis formed the
framework of the development procedure. The truck-induced wind gust fatigue provi-
Infinite-Life Approach
The method used to develop the natural wind provisions was based on spectral
analysis in collaboration with the infinite-life approach to fatigue design. The provisions
are adequate within certain limitations, as they were formed using four particular catego-
rized structural types, each of which were chosen to represent the population of the cate-
gory selected. The structural response of two overhead signal support structures, one can-
tilever-type overhead sign support structure, and one luminaire support structure to natu-
28
ral wind excitation were analyzed, the transmitted stresses of each were averaged, and the
The AASHTO fatigue code was based on the constant amplitude fatigue limit
(CAFL). Tests to developed plots of the stress vs. the number of cycles (S-N curves) for
common connection details were used to determine this value. Fatigue stress ranges that
occurred below the CAFL were estimated to have an infinite life. Stresses ranges at criti-
cal locations such as welded or bolted connections must be under the CAFL to be in
compliance with the code. Changes in the design must be made if they are not to lower
The CAFL tests were based on constant amplitude, whereas amplitudes that occur
due to natural wind are random with variable amplitudes. To account for variable ampli-
tude loading, the infinite-life approach adopted the findings of NCHRP Report 354 where
variable amplitude fatigue tests were performed on a variety of full scale connection de-
tails. From the results, it was determined that failure would still occur if the variable am-
plitudes exceeded the CAFL of the connection detail 0.05% of the time. However, if the
CAFL was exceeded 0.01% of the time, the specimen was said to have an infinite life
(19). In view of the results, the infinite-life approach involved estimating the stress
ranges that would occur onto the structure with a 0.01% exceedence probability, includ-
ing dynamic amplification. These ranges were referred to as the fatigue limit-state load
ranges. The structure was designed such that the limit-state stress ranges did not exceed
The first step in determining the fatigue limit-state load ranges to develop the
AASHTO fatigue provisions was to estimate the structural excitation due to natural wind.
This involved predicting the natural wind environment for which the structure is to be
exposed. This was done using a spectral analysis. Simulating the natural wind force can
be difficult due to the randomness, turbulence and gustiness of natural wind velocity. A
static load that represents the force due to wind can be easily calculated, as well as a dy-
namic load in terms of a periodic wave such as a sine wave. This type of examination
does not account for the gustiness and turbulent nature of the wind itself. In most cases, a
gust factor is used to account for the random nature of wind excitation, specifically for
spectrum (PDS) curve (13) shown in Eq. 3.1 that simulated the turbulent nature of wind
velocity.
4κV102 x 2
Sv ( f ) = 4
[Eq. 3.1]
f (1 + x )
2 3
where
S v ( f ) = wind velocity power density spectrum at any height
f = frequency
V10 = mean wind velocity at a stadard height of 10 meters above ground level
κ = surface drag coefficient (Table 3.1)
1200 f f
x= 2
with 2 in cycles per meter.
V10 V10
30
Type of Surface κ α
Open unobstructed country (e.g., prairie-type grass-
0.005 0.15
land, arctic tundra, desert)
Country broken by low clustered obstructions such as
0.015 - 0.020 0.27 – 0.31
trees and houses (below 10 m high)
Heavily built-up urban centers with tall buildings 0.050 0.43
Random excitation is best analyzed over a frequency spectrum. PDS curves are
commonly used for dynamic loading analysis, but more specifically suited to random vi-
bration. In random excitations, there are no periodic systems within the excitation that
can be analyzed at specific frequencies because the excitation is applied randomly over a
spectrum of frequencies. To account for this, the natural wind fatigue analysis involved
an examination of the total energy, or power, in the random excitation over a frequency
band. A typical PDS curve is plotted with the ordinate in units of the parameter (force,
acceleration, velocity, displacement, etc.) squared divided by the bandwidth (e.g., N2/Hz),
and the abscissa in units of frequency (Hz). The peaks on a PDS curve identify the fre-
quency range(s) at which the majority of the energy lies within the particular excitation.
The area under a PDS curve is equal to the mean square value, and the square root of this
area is equal to the root-mean-square (RMS) value: a value of which plays an important
role in determining the magnitude of loads that are transmitted onto a structure from vi-
Davenport developed his wind velocity PDS curve from at least 70 experimental
wind velocity data collections from various locations around the world. His intention was
to develop a model which simulated the turbulence and gustiness of wind velocity. He
developed Eq. 3.1 from the 70 experimental data collections. The equation is a function
31
of wind velocity frequency with respect to a mean wind velocity at a specified height. His
formulation is plotted in Figure 3.1 for frequencies ranging from 0 to 10 Hz, an open ter-
rain (see Table 3.1), and an annual mean wind velocity of 11 mph (5 m/s).
FIGURE 3.1 Wind velocity PDS for annual mean wind velocity 11 mph.
Once the behavior of the wind velocity environment was estimated, the PDS was
transformed into a wind force PDS by using principles related to fluid mechanics. An in-
duced force onto a structure due to natural wind, referred to as drag, is proportional to
1
FD = ρC d AV 2 [Eq. 3.2]
2
32
where
FD = drag force
kg
ρ = density of air = 1.22
m3
C d = drag coefficient
A = area of exposed surface
V = wind velocity at any height.
By utilizing the proportionality between drag pressure and wind velocity, and accounting
for the turbulent wind velocity and its variance about the mean wind velocity, a wind
pressure PDS was developed from Davenport’s wind velocity PDS shown in Eq. 3.3. The
plotted equation is shown in Figure 3.2 for an annual mean wind velocity of 11 mph (5
m/s) and normalized for exposed area and the drag coefficient.
The PDS curve accounts for the gustiness and turbulence of wind velocity over a
above ground level. Since most support structure are at or around 32.8 ft (10.0 m) in
height, the PDS curve was well suited for these types of structures. Yet, the PDS can be
used at any particular height by using the power law profile for approximating variation
where
V = wind velocity at height z
α = surface coefficien t in Table 4
z = height above ground.
FIGURE 3.2 Force PDS for annual mean wind velocity 11 mph.
In this case, where the objective was concentrated on formulating a design code
for fatigue wind, the wind velocity variable in the pressure PDS equation was taken at the
standard height of 32.8 ft (10.0 m) above ground level, and kept uniform across the ex-
posed façade. The purpose of which was to provide a simplified design equation for
commercial use. Some conservative formulation exists as the wind velocity typically in-
Once the natural wind environment was estimated, the next step was to apply the
PDS to the infinite-life approach in determining the limit-state stress ranges. Since the
force spectrum was based primarily on the annual mean wind velocity, the wind velocity
that was exceeded 0.01% of the time was found and referred to as the limit-state wind
velocity. The force spectrum was calculated using the limit-state wind velocity and used
Wind velocity is random in nature, but it can be predicted though statistical rela-
tionships. It has been found through many experiments that the magnitude of the wind
velocity vector will form a Rayleigh distribution (14, 52, 54). By using the Rayleigh dis-
tribution, the 0.01% exceedence probability can be found through the relationship in Eq.
−πv 2
PE (v) = e 4V
2
[Eq. 3.5]
where
PE (v) = probability
v = wind velocity corresponding to the probability
V = mean wind velocity.
An analysis was conducted to determine which annual mean wind velocity to use in
Eq. 3.5 to determine the limit-state wind velocity (wind velocity with a 0.01% ex-
ceedence probability). The annual mean wind velocities of major U.S. cities were ana-
lyzed. It was found that an annual mean wind velocity of 11 mph (5 m/s) was exceeded in
only 19% of the U.S. cities analyzed and was therefore chosen. By plugging in 11 mph in
Eq. 3.5, and solving for the wind velocity corresponding to the 0.01% probability, a limit-
35
state wind velocity was found to be equal to 37 mph (17 m/s). The force spectrum was
then formed using the limit-state wind velocity (see Figure 3.3) and was used as the natu-
Structural Excitation
Once the excitation was determined, the spectral density of the response was cal-
culated. This was done through finite element analysis for the signal, sign, and luminaire
support structures. Each structure had different dynamic characteristics such as natural
frequency and critical damping percentages, all of which are based on the size, shape, and
material of the structure. The response PDS was calculated in units of stress squared di-
36
vided by Hz. The area underneath the response PDS curve is equal to the variance of the
response about the mean. The square root of the area is equal to the RMS. If the mean is
equal to zero, then the RMS is equal to the standard deviation (58, 62). Since the re-
vibration, the developed response PDS curves were narrow-banded at the natural fre-
quency of the structures. Therefore, the stress range was calculated as a constant ampli-
tude in the form of a sinusoid. For any sinusoid, the peak stress amplitude is equal to the
square root of two times the RMS (assuming a zero mean for analysis purposes). The
stress amplitude was then multiplied by two to account for a peak-to-peak stress range.
The RMS was determined from the response PDS for each structure, and the effective
where
S reff = effective stress range or limit - state stress range
σ rms = RMS of the response
Equation 3.7 was calculated for the four type of structures analyzed. The analysis
was normalized for the drag coefficient and exposed area. The result ranged from 3.6 to
6.3 psf (170 to 300 Pa) using 1% and 2% damping percentages. An average was taken
and used for the code. The final equation (Eq. 3.7) was multiplied by the drag coefficient,
importance factor, and a wind factor to account for other annual mean wind velocities.
37
2
v
PNW = 5.2C d I F [Eq. 3.7]
11
where
PNW = design fatigue load due to natural wind, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean wind velocity other than11 mph (5 m/s)
occurring over a finite length of time. The loading was created when semi-trailer trucks
passed underneath the structure causing the structure to vibrate. The vibration generated
stresses in the structure that can potentially accumulate fatigue damage over time.
The Supports Specifications design equations for truck-induced gusts was origi-
nally developed from the work performed by Desantis and Haig (16), and later revised
from the work presented in NCHRP Report 469 (14). It assumes that the wind loading
onto the structure was equivalent to the speed of the passing truck, plus a gust factor of
1.3 to account for head wind. For instance, a truck traveling at 65 mph (29.1 m/s) would
produce a 65 mph (29.1 m/s) wind onto the structure. The result was an 18.3 psf (876 Pa)
magnitude static pressure. The code only accounted for a vertical force applied upward
(ground up) onto the structure. Assuming that the upward motion of the structure result-
ing from the truck gust was equivalent to a proceeding downward motion, the 18.3 psf
(876 Pa) was doubled to account for a peak-to-peak range equaling 36.6 psf (1760 Pa).
NCHRP Report 469 investigated the fatigue load due to truck gusts performed by
researchers on VMS cantilever structures. The research indicated that the current code
38
was too conservative based on their conclusions. A reduction of the loading was recom-
2
V
PTG = 18.8C d I F [Eq. 3.8]
65
where
PTG = design fatigue load due to truck gust, psf (Pa)
C d = drag coefficient
I F = importance factor
V = truck speed, mph (m/s)
The reduction was determined based on experimental evidence from VMS struc-
tures. Strain measurements on a cantilevered VMS support structure were recorded dur-
ing random truck events. An equivalent static load was calculated that would produce the
same strain range measured experimentally. The resulting load was to be applied verti-
cally to all horizontal areas (underside of the structure) along a 12 ft length (3.7 m), or
equal to the width of a travel lane. A reduction in the load with respect to height of the
structure above ground level was also permitted. It was discovered the pressure load de-
creased linearly with height, with the maximum occurring at 19.7 ft (6 m) above ground
CHAPTER 4
Chapter Overview
A cantilever-type highway overhead sign structure was selected for field meas-
urement. The structure was instrumented with electric strain gages to determine stresses
tural dynamic behavior, and anemometers were attached to measure wind velocity vec-
tors (magnitude and direction). Each transducer provided time history streamlines that
A four chord truss cantilever-type highway overhead sign support structure was
chosen for analysis. The structure was located at northbound Exit 340B on highway I-65
near Decatur, Alabama USA. A picture of the structure evaluated for this project is
Geometric Properties
The truss overhang was made of 3.5 in (88.9 mm) diameter steel pipe supported
by a 24 in (610 mm) steel pipe shaft support. The shaft was support by a 1.25 in (31.8
mm) thick by 35 in (889 mm) diameter base plate with eight 1.5 in (38.1 mm) anchor
bolts connected to a concrete foundation. The structural configurations are shown in Fig-
14 ft 25 ft
10.5 ft
2 ft Four Chord
Cantilever
Truss
11.5 ft 4 ft
4 ft
Post
Sign 26.75 ft
4 ft
3 ft
10.75 in
37.75 ft
3 ft
Material Properties
The post and truss section was made of API-5L-X52 steel pipe. All plates were
structural steel ASTM A572 Gr. 50. The anchor bolts were AASHTO M314-90 Gr. 55
(essentially the same as ASTM F1554 Gr. 55). The W-shape and T-shape sections used
for the sign-to-truss connection were made of A572 Gr. 50 steel. The sign was aluminum
wrought alloy designated as 6061-T6. The concrete pile foundation was conventional
concrete with #9 size rebar Gr. 60. All of the major required material properties of the
4, 000 psi
Pile
Concrete Compressive 3,600,000 NA NA
Foundation
Strength
Pile ASTM A706
Rebar 29,000,000 60,000 80,000
Reinforcement Gr. 60
Instrumentation Process
There was a unique advantage associated with this project. The researchers were
able to instrument the structure before and during assembly, which allowed for a thor-
bled structure in the field were avoided as a result. All strain gauges, except on the anchor
bolts, were placed on the structure at the ALDOT laboratory in Montgomery, Alabama
USA before transportation to the assembly site (Figure 4.5 – 4.6). The anchor bolts were
instrumented on site the day after the foundation was poured and the anchor bolts set
(Figure 4.7). The post support was placed on the foundation and bolted to the anchor
bolts after the anchor bolts were gauged. A period of 14 days (28 days after pouring)
44
transpired after the post support was erected for concrete curing of the foundation. After
the curing stage was completed, the truss overhang and sign was attached to the post sup-
port (Figure 4.8 – 4.9). A lift bucket was used to place the accelerometers and anemome-
Strain Gauges
Uni-axial strain gauges and single-plane rectangular strain gauge rosettes (meas-
uring in three primary directions: 0°, 45°, and 90°) were used to measure the strain re-
sponse of the structure. They were placed on the structure at locations where strains were
most critical, and especially applicable to determining the causative load. These locations
included: truss chord members near the truss-to-support connection, bottom of the pole
support member near the base plate, and the anchor bolts. A list of all strain gauges and
their locations along with a schematic detailing these locations are provided in Appendix
A.
48
A total of 16 uni-axial strain gauges were used on the truss chord members lo-
cated near the truss-to-support connection: four gauges per chord placed along the cir-
cumference of the member. These gauges measured strain in a single direction, oriented
along the longitudinal length of the chord member, in an attempt to capture tensile and
compressive strain readings. A schematic of the gauge locations are shown in Figure
4.11, with the actual gauged structure at this location shown in Figure 4.12. A picture of
Placed on each
Chord at 90º
around Chord
Circumference
Centerline
of 1st Truss
Panel
3’- 4.5”
Post Support
Uni-axial and rosette strain gauges were used for the post support. The gauges
were attached to the structure in the laboratory before shipment to the assembly site. Two
levels, located nearest to the base plate, were gauged along the circumference of the
member in accordance to the schematic in Figure 4.14. This was done in an attempt to
capture tensile and compressive strains near the base plate. Forty-five degree rectangular
rosettes were used to measure transverse shear and torsion strains. A close up of Section
AA viewing a rosette and Section BB located 4 in (102 mm) above Section AA viewing a
uni-axial strain gauge is shown in Figure 4.15. Two levels were chosen for redundancy
issues in the case of strain gauge measurement failure since the structure was tested out-
51
side, and the testing duration lasted over multiple months. A picture of the assembled in-
Section AA Section BB
4 in Section BB
Section AA
12 in
Anchor Bolts
Strain gauges were attached to the anchor bolts in the field, after the concrete
foundation concrete was placed. A picture of the anchor bolts before placement into the
concrete foundation is shown in Figure 4.17, and after placement in the foundation before
gauging in Figure 4.18. The material specifications of the anchor bolts are provided in
Table 4.1. The concrete foundation reached 11 ft (3.35 m) into the ground, with the rein-
A total of eight anchor bolts were used for the structural foundation connection.
Uni-axial strain gauges were attached to each anchor bolt in the orientation shown in Fig-
ure 4.19, totaling eight strain gauges. The gauges were attached as close as possible to the
foundation due to assembly procedures and leveling processes. The gauges were located
tangent to the outside rim of the anchor bolt group formation, with the sensitive end ori-
ented along the longitudinal length of the bolt. The threads at the gauge location were
shaved (Figure 4.20) and cleaned for gauge attachment. This was done for each bolt. A
picture of a finished instrumented anchor bolt is shown in Figure 4.21, and of the assem-
Support
Base Plate
Strain Gauge
Anchor Bolt
Foundation
Orientation
Anemometers
Anemometers were used to measure wind velocity and direction. The WindSonic
ultra-sonic wind and direction sensor was used for this application (see Figure 4.23). The
anemometer calculated speed and direction by measuring the time needed for generated
sound pulses to travel from one transducer to the other within the air gaseous medium. A
combined total of six anemometers were required for wind measurements of natural wind
and truck induced wind gusts. Two were placed above the post support structure, and four
on the truss section of the support structure. Steel extensions were fabricated to fit the
anemometers to the structure. A list of the anemometers identification and locations are
their locations can also be seen in Figures 4.24 – 4.26. The mounting instructions pre-
Anemometers AN-1 and AN-2 were used to measure the wind created from trav-
eling trucks. They were placed along the centerline of the traffic lane above where the
trucks traveled. The distance was 30 ft (9.14 m) from the centerline of the post support.
The anemometers were attached onto the truss structure using steel extensions as to not
60
be blocked by the sign. Two anemometers were used for redundancy issues, as well as for
providing differing planes of measurement: AN-1 measured the horizontal plane whereas
AN-2 measured the vertical plane. This enabled a three dimensional wind velocity vector
Anemometers AN-3 and AN-4 were used to measure the ambient wind environ-
ment. The anemometers were attached to a steel extension that reached 4 ft (1.22 m)
above the structure to avoid any effects for wind dynamics (turbulence, vortices, etc.)
caused by the structure. Two anemometers were used for redundancy as well as for pro-
viding differing measurement planes: AN-3 measured the vertical plane whereas AN-4
measured the horizontal plane. This enabled a three dimensional wind velocity vector to
be measured and used in the data analysis. Each anemometer provided wind velocity and
direction angle measured from an identified North direction marked on the instrument
with measurement recordings in compass bearings. For all anemometers, the North com-
pass on the instrument was directed from the front façade of the sign, opposing the direc-
tion of traffic.
Accelerometers
They were of the piezoelectric type that uses a piezoelectric crystal mounted to a small
mass from which the voltage output is converted to acceleration. Each accelerometer had
a maximum capacity of 96.5 ft/sec2 (3 G). The data was used to determine major dynamic
Modal shapes
ter only gave the natural frequency of the member upon which it rests. A combined total
of six unidirectional accelerometers were required for the structure. They were placed at
AC-1
AC-2
Centerline
Post
AC-3
Centerline
Truss
Three accelerometers were placed at location AC-1 (see Figure 4.27) to measure
the vertical (perpendicular to the direction of traffic), horizontal (parallel to the direction
of traffic), and longitudinal (transverse to the direction of traffic) accelerations of the post
member. Two accelerometers were placed at location AC-2 to measure the vertical and
horizontal directions (Figure 4.28). One accelerometer was placed at location AC-3 to
measure the horizontal direction for indentifying torsion behavior of the overhang truss
(Figure 4.29). An overview of locations AC-2 and AC-3 is shown in Figure 4.30.
threaded screws. Electrical insulation between the accelerometers and the test surface was
built into the transducer device. All mounting surfaces were flat to avoid distortion that
may produce strains which could affect the accelerometer’s response. Carefulness was
taken when screwing the accelerometers to the attachment surface as to not overreach the
Mounting blocks were used at each location for accelerometer attachment (see
Figure 4.31). For instance at location AC-1 where three accelerometers were needed, all
three were attached to one single attachment block. A picture of this location with the at-
tached mounting block is shown in Figure 4.32. Once the truss section of the structure
was erected, the AC-1 accelerometers were screwed onto the mounting block with the
sensitive ends (measurement direction) oriented in the proper directions. The same type
of block was used at locations AC-2 and AC-3. Due to the round surface at these loca-
tions (steel pipe truss web member), a small flat steel plate was welded to the side of the
members for which the mounting blocks and accelerometers were attached. The size of
the plate was manufactured as small as possible, with enough space to accurately and se-
63
curely mount the accelerometers and the plate to the web member, but without the possi-
bility of creating significant additional wind drag. The initial mounting instructions pre-
All instrumentation was wired into a data acquisition system. The data was con-
verted into engineering units by the data acquisition system, filtered, and stored onto the
hard drive of a portable computer. Data was then saved onto Maxwell CD-R data storage
disks to be distributed after testing. A white van was used to hold the data acquisition
system and computer during testing. It was driven to and parked underneath the sign
structure on the side of the highway for testing. Figure 4.33 shows a typical test setup
with the van and all instrumentation wiring fed through a side hole in the van and hooked
to the acquisition system. A close up of the data acquisition system and computer is
The data acquisition was capable for collecting all data from instrumentation si-
multaneously, which was required for the fatigue tests. The maximum number of chan-
Anemometers: 8 channels
Accelerometers: 6 channels
Total Channels = 62
64
FIGURE 4.33 Typical test setup with van data acquisition system.
67
It was important to be specific about the sampling rate of the data acquisition sys-
tem. The vibration behavior of sign support structures are typically around 1 to 10 Hz,
including all modes of vibration. In view of this, data were collected at 60 samples per
second for all instrumentation, which was at least six times greater than the highest an-
ticipated modal frequency of 10 Hz. The Nyquist frequency was therefore 30 Hz, allow-
ing all frequencies recorded below 30 Hz to be accurate, avoiding aliasing and other
prominent sampling errors. A digital low-pass Butterworth filter was used and set at 20
CHAPTER 5
Chapter Overview
The sign support structure was tested under loading conditions: a) Natural wind
gust, and b) Truck induced wind gust. The wind data was taken over an extended time
period in an effort to capture the predominant natural wind gusts. Wind data from the Na-
tional Weather Service near Huntsville, Alabama USA, in the form of the annual mean
wind velocity for the area, was determined to help schedule appropriate testing days be-
tween ALDOT and the UAB research team, and to compared to the wind measurements
taken for accuracy so that a proper representation of the natural wind environment was
established. A standard semi-trailer vehicle was used to apply the truck induced wind
gusts. Varying truck speeds was used and the corresponding load data was recorded.
At least 32 hours of wind velocity data was desired for data collection. The inten-
tion was to have the collected data distributed between the following sampling intervals:
A 20 to 30 mph (8.94 – 13.4 m/s) was considered as high wind according to the
labels. The wind velocity having a 0.01% exceedence probability in accordance to the
infinite-life approach of the Supports Specifications was 38 mph (17 m/s). Capturing this
wind was considered probable, and likely during sampling of the categorized “high wind”
label; however such conditions warranted extreme weather that may be unsuitable for
field testing due to safety issues. The intention was to capture the vibration behavior of
the sign support structure in response to vibratory induced wind velocities, typically
above 10 mph (4.47 m/s), and form a relationship between wind velocity and structural
response stress ranges. Low wind data was wanted in order to establish a wind velocity
The next issue was determining appropriate testing days for natural wind. A good
distribution of measured wind velocities, and structural response, was needed. Since wind
cannot be produced manually, the days that were scheduled for testing were chosen based
on a trial and error process, which was unavoidable. ALDOT needed at least a two week
advance to schedule times. In addition, testing could not be performed during precipita-
tion.
A ranking schedule was prepared to help with the scheduling decisions. It was
1 being the worst day to test for natural wind with respect to fatigue loading, and
10 being the best day to test for natural wind with respect to fatigue loading.
The ranking system was developed from historical data collected by the National
Weather Service Forecast Office located in Huntsville, Alabama USA. Climatology re-
cords ranging from early 2008 to 1999 were used in its development. It was prepared for
the months of October 2008 to March 2009 for testing. All ranking schedules used are
provided in Appendix D.
A ranking of 10 indicated that, from the historical records, the averaged natural
wind velocity, maximum wind velocity, gusts occurrences and direction were ideal for
testing with respect to this project. A ranking of 1 indicated the opposite. Most rankings
were from 5 and up, which indicated their percentage of the “ideal” event preferred when
The proposed method for application of the raking schedule for field testing was
as follows:
1. Determine the best possible day(s) for which natural wind testing is possible in
2. Look on the UAB ranking schedule associated with the required month of natural
wind testing.
3. Determine day(s) with the best ranking (preferably rank 8 and up) from the UAB
schedule
71
4. Develop a coincidence between the day(s) chosen from the ALDOT work sched-
ule and the day(s) selected from the UAB rank schedule (preferably choosing a
5. Check local weather updates and forecasts (approaching fronts, storms, etc…).
6. Chose optimal days from the information gathered from steps 1 through 5 to test.
Also included on the schedule was a moving average taken over a three day pe-
riod. This was provided to illustrate a smoother pattern of test day rankings. It helped the
ALDOT officer to choose a string of optimal testing days where the moving average was
high (preferably 8 and up). Even though the early months of 2009 during the projected
project schedule looked favorable, it was recommended to begin testing as soon as possi-
It was important to note that the UAB ranking schedule was only a forecast that
was based on averaged historical records. It showed a historical pattern of ideal wind oc-
currences in relation to the required needs of this project. Yet, the actual wind conditions
of the chosen testing time periods did to a large degree reflect the UAB ranking. There
were however scheduling delays and cancelations during the testing program due to pre-
cipitation.
Test Procedure
The test procedure for determining the fatigue load from natural wind gusts went
as follows. An appropriate testing day was scheduled between UAB and ALDOT with
the help of the ranking schedule. A testing day was typically scheduled at least two weeks
72
in advance. Even though natural wind was highest during storm events, scheduled days
were canceled for extreme weather as precipitation created large errors in the data collec-
On the scheduled testing day, ALDOT drove the data acquisition van to the site
and began attaching wiring from the instrumentation to the van. This took typically one
to two hours. The testing began after instrumentation was operated and all transducers
were checked for accuracy. All instrumentation was measured simultaneously at 60 sam-
ples per second. Data recording intervals were set at no smaller than 45 minutes each,
AN-1 and AN-2 were used to measure the ambient wind velocity and direction during
testing. All strain gauges and accelerometers were used to measure the structural re-
sponse. Data was saved onto a CD for storage and distribution, and later used for deter-
mining the fatigue load due to natural wind back in the lab.
After data collection was finished for the scheduled event, all wiring was discon-
nected from the data acquisition system. Instrumentation connections were then stored
properly next to the structure for future testing, and ALDOT researchers left with the data
acquisition van. This procedure continued for each scheduled day until at least 32 hours
The data collection event involving truck induced gusts was a staged occurrence.
Random truck events were also considered for possible truck measurements, but due to
the location of the sign, trucks were not able to reach a speed that would create signifi-
73
cant structural response. For this reason, a truck and driver was hired to run underneath
the sign structure at various speeds to measure the fatigue load due to truck induced
gusts. A picture of the truck and driver is shown in Figure 5.1. The height of the semi-
trailer was 13.5 ft (4.11 m) width equal to 8.5 ft (2.59 m), with a 53 ft (16.2 m) long box
trailer.
Several truck types were considered for collection, based on geometric character-
Cab + trailer
The selected truck (Figure 5.1) was representative of the majority of trucks using the
Test Procedure
The test procedure for truck induced wind gust analysis was a cyclic event. The
truck drove loops underneath the sign structure a various speeds while the data acquisi-
tion system was recording structural responses from the instrumentation. A range of
speeds from 60 mph to 70 mph was used for the test. Three runs per speed were con-
ducted, totaling nine truck runs. Traffic counters were placed across the road underneath
the sign structure (see Figure 5.2) to record truck speed during each run, and to compare
to odometer readings for accurately identifying the truck speed. The traffic counters were
set to run parallel with the data acquisition system to have simultaneous time readings
during each truck run. They were JAMAR Trax Flex HS recorders with 0.365 in (9.271
mm) tubes spaced at 36 in (91.4 cm), and taped at 12 in (30.5 cm) intervals.
Due to obstacles at the site, ALDOT used traffic control for the testing to prevent
interference from cars during each run. Constant communication between the cab and the
ALDOT officer at the structure site was kept to indicate when the truck was approaching
so that the passing run could be marked on the data and later isolated in the lab for analy-
sis. Detailed instructions of the truck induced gust test procedure prepared for ALDOT
The truck tests were scheduled on a low wind day so that the truck gust could be
isolated from the data. This typically was on days with natural wind velocities averaging
less than 10 mph (4.47 m/s). Structural excitation was created for greater natural wind
velocities, and therefore created unreadable truck gust impulses from the data collection
as natural wind generally controlled structural response. A picture of one of the truck
runs is shown in Figure 5.3. Anemometers AN-3 and AN-4 measured wind gust created
from the truck run, and all strain gauges and accelerometers were used to measure the
structural response. The data was later isolated in the lab to determine the fatigue load
CHAPTER 6
FATIGUE RESISTANCE
Chapter Overview
An analysis of the resistance of the support structure was first performed with the
collected experimental data. This helped to understand basic operational principles of the
structure including:
Structural behavior,
Best areas of the structure to developed fatigue loads from the measured data.
Particular attention was concentrated on the connection details such as the anchor bolts,
The stresses at these critical locations were compared to the constant-amplitude fatigue
service loading. Sections of the structure that were most vulnerable to service loading
were identified. Areas of the structure were identified as the best case for developing fa-
was a stress limitation used in fatigue analysis. The stress occurring at particular details
due to an applied fatigue load was compared to the limitations. The detail was considered
safe to significant fatigue damage if the stress was less than the endurance limit. Design
changes must be made such as material strength or shape changes if the stress is above
were used for the analysis. They were based on stress versus number of cycles (S-N)
curves on a variety of connection details for support structures. The curves were devel-
oped from experimental testing in which a constant-amplitude load was applied to the
The specifications divided the variety of details into separate stress categories.
The constant-amplitude fatigue thresholds for the relevant stress categories for the tested
support structure in this project are listed in Table 6.1 with reference to the detail (1).
Failure Index
The failure index was calculated at the instrumented locations. This helped to
value the resistance of the structure to fatigue loading. The failure index was defined as
the ratio of the fatigue stress in the structure divided by the constant-amplitude fatigue
threshold for the particular connection detail (Eq. 6.1). The fatigue threshold was ex-
tracted from the Supports Specifications depending on the type of connection under
evaluation.
79
Fatigue Stress
Failure Index = ≤1 [Eq. 6.1]
Constant - Amplitude Fatigue Threshold
The fatigue stresses were calculated using the experimentally measured strain
values. Calculated three second peak-to-peak strain ranges were used for the natural wind
gust analysis, and the maximum peak-to-peak ranges measured during the truck tests
were used for the truck-induced gusts analysis. A detailed description on the methodol-
ogy in developing the three second peak-to-peak strain ranges and wind velocity averages
80
Wind Gust.
The support structure had eight 1.5 in (38.1 mm) diameter AASHTO M314-90
Gr. 55 anchor bolts (see Figure 6.1). All eight were instrumented with uni-axial strain
gauges oriented along the axial length of the bolt. A layout of the anchor bolts is shown
in Figure 6.2, and the labeling used for the analysis is shown in Figure 6.3.
45°
22.5°
Important to anchor bolt analysis is the clearance distance of the anchor bolt from
the bottom of the base plate to the foundation. This distance is generally between 1 in
(25.4 mm) and 3 in (76.2 mm) in practice. The clearance distance for the tested support
structure was measured in the field and is listed in Table 6.2 for each anchor bolt. General
structural analysis of anchor bolts involves purely axial stresses, but as the clearance
length of the anchor bolt increases, bending stress are distributed onto the bolt which are
not accounted for in the analysis. This in turn would create error in the structural analysis
results by the design engineer, and therefore fatigue related evaluations may be inaccu-
Direction of Traffic
Roadway
Front Face x
of Support
Structure y
AB-4 AB-5
AB-3 AB-6
AB-2 AB-7
AB-1 AB-8
The experimentally measured strain values of the anchor bolts were used for this
analysis. Each anchor bolt (totaling eight) was evaluated. The stress range was deter-
mined and plotted versus wind velocity. A regression was performed on the plot to esti-
mate the stress range at the fatigue wind of 38 mph (17 m/s). The stress range at this
value was compared to the endurance limit of the particular detail (stress category D).
Strain ranges were determined from the experimentally collected natural wind
data on the anchor bolts. The maximum peak-to-peak strain range was calculated at three
second intervals. The simultaneously occurring three second wind velocity was also cal-
culated to juxtapose the anchor bolt strain ranges. The maximum strain ranges and wind
velocities were plotted and averaged together at 1 mph (0.447 m/s) intervals. This was
done for the total collected data measurements using all eight anchor bolts. The resulting
plot of the averaged maximum peak-to-peak strain range versus the three second wind
FIGURE 6.4 All anchor bolt microstrain range vs. wind velocity.
As seen in Figure 6.4, anchor bolts AB-8 and AB-1 demonstrated the largest
strain ranges during the experimental measurements. As seen in Table 6.2, these anchor
bolts had the largest clearance from the foundation to the base plate, and was believed to
be the reason for the large strain ranges measured experimentally. The wind data was di-
rected predominately on the front face of the structure (see Chapter 8); however anchor
bolts AB-8 and AB-1 show the largest ranges, followed by AB-2.
Based on the Supports Specifications, and the detailed fatigue loading analysis in
Chapter 8, the fatigue wind was determined to be 38 mph (17 m/s). The velocity was de-
termined as the wind velocity with a 0.01% exceedence probability from the annual mean
wind velocity equal to 11 mph (5 m/s) in accordance to the infinite-life approach to fa-
tigue design philosophy. The fatigue wind velocity was not measured during the experi-
85
mental tests, but wind velocities slightly lower than the fatigue wind were measured, and
therefore a regression analysis was performed on the data obtained in Figure 6.4 to esti-
mate the strain occurring at this wind. The estimation process was performed in similar
fashion to the process conducted in Chapter 8 for computing the wind pressure at the fa-
tigue wind.
By using the fundamental fluid mechanics relationship between pressure and the
velocity squared, a regression was performed on the experimental data to form a best fit
line that follows this relationship. The independent variable (x-axis, wind velocity) was
transformed into wind velocity squared and plotted with the microstrain ranges to “lin-
earize” the data. A linear best fit line was produced from the transformed data plot and
recorded (see Figure 6.5). The independent variable was transformed back into its origi-
nal form and the trendline equation was used to estimate the microstrain range at the fa-
tigue wind. The resulting plot is shown in Figure 6.6 for AB-8, the anchor bolt with the
largest recorded microstrain ranges. At the fatigue wind of 38 mph (17 m/s), the strain
range was estimated at 242.065(10-6), or 242.065 microstrain. The estimated strain range
result for all anchor bolts is listed in Table 6.3, and is plotted in a bar graph (Figure 6.7)
TABLE 6.3 Estimated Microstrain Range at Fatigue Wind for Anchor Bolts
FIGURE 6.7 Estimated microstrain range at fatigue wind for anchor bolts.
The stress range at the fatigue wind was determined from the estimated mi-
crostrain ranges, and compared to the constant-amplitude fatigue thresholds for the an-
chor bolt stress category D. Stress category D is for mechanically fastened connections
such as “anchor bolts or other fasteners in tension” (1). The stress range was calculated
88
for each of the eight anchor bolts tested. It is important to point out that if the fatigue
stress range was above the threshold, the design did not comply with the Supports Speci-
fications fatigue provisions with respect to the limits placed on the fatigue stress ranges
for anchor bolts. This is true regardless if the support structure was designed for fatigue
using the in-place fatigue design equation for natural wind. In such cases, believed to be
due to the large clearance between the foundation and the base plate, an additional bend-
ing stress was developed in the bolt because of an increased influence of out-of-plane
The fatigue stress ranges calculated from the experimental microstrain ranges es-
timated at the fatigue wind is listed in Table 6.4. The values are listed with the constant-
amplitude fatigue thresholds of steel anchor bolts for the stress category D.
TABLE 6.4 Anchor Bolt Stress Range and Failure Index for Natural Wind
The failure index is also provided in Table 6.4, which was equal to the stress
range induced onto the anchor bolt at the fatigue wind divided by the constant-amplitude
fatigue threshold for the category detail (stress category D). A failure index greater than
89
one would indicate a fatigue stress range occurring greater than the allowable threshold.
As listed in the table, the failure index for each anchor bolt was less than one, with the
exception of AB-8 which was equal to one and was border line in the threshold design
check. A visual on the failure index variation with anchor bolt is provided in Figure 6.8.
The maximum peak-to-peak strain ranges were determined from the experimen-
tally collected data of the truck tests. Each of the eight anchor bolts was evaluated. A de-
tailing of the maximum peak-to-peak strain range procedure is provided in Chapter 10:
in the chapter, only four of the nine truck tests were usable on account of prevailing natu-
90
ral wind velocities. The truck runs utilized in this analysis was measured on low wind
days. A plot of the truck test strain results is provided in Figure 6.9, showing the relation-
ship between the truck speed and the resulting maximum peak-to-peak strain range in a
depth chart. The microstrain ranges per anchor bolt as measured from the truck tests are
listed in Table 6.5. The maximum range was recorded on AB-7 of truck test Truck 4
equaling 175.713 microstrain, when the truck was traveling at the highest tested speed of
70 mph (31.3 m/s). As seen in the figure, the majority of the strain occurred at the higher
truck speeds.
FIGURE 6.9 Depth chart of anchor bolt strain range in truck tests.
91
TABLE 6.5 Anchor Bolt Microstrain Ranges from the Truck Tests
The maximum peak-to-peak strain range measured from all truck tests was used
to calculate the stress range and determine the failure index for the truck tests. The results
are listed in Table 6.6 and illustrated in the bar chart of Figure 6.10. The results indicate
92
no significant effect from truck gusts on the anchor bolts with respect to the fatigue
TABLE 6.6 Anchor Bolt Stress Range and Failure Index for Truck Tests
post support was instrumented at 12 in (305 mm) at Section AA and 16 in (406 mm) at
Section BB. Only Section AA was evaluated in this chapter because it was closer to the
weld. A total of eight strain gauges at each section was placed, which included 4 rosettes.
A general layout of the gauges is shown in Figure 6.12, and the labeling methodology
shown in Figure 6.13. Each strain gauge on Section AA was evaluated to provide an es-
Section AA Section BB
4 in Section BB
Section AA
12 in
Strain Gauge
The evaluation was performed in the same fashion as the anchor bolts. The stress
range was determined for natural wind gusts and plotted versus wind velocity intervals
for natural wind gusts. A regression was performed on the plot to estimate the stress
range at the fatigue wind of 38 mph (17 m/s). The maximum stress ranges were deter-
mined for the truck-induced gusts. The stress ranges were compared to the endurance
limit of the particular detail (stress category E') and the failure index was calculated.
Strain ranges were determined from the experimentally collected natural wind
data on the post support. The maximum peak-to-peak strain range was calculated at three
second intervals. The simultaneously occurring three second wind velocity was also cal-
culated to juxtapose the strain ranges. The maximum strain ranges and wind velocities
were plotted and averaged together at 1 mph (0.447 m/s) intervals. This was done for all
measured data of the strain gauges on the post support. A total of 12 readings were made:
8 normal stress values, and 4 shear stress values. The shear stress values were determined
from the rosette strain gauges. The gauges were rectangular strain rosettes arranged in a
45° pattern as shown in Figure 6.14. The strain transformation equation, Eq. 6.2, was
used to calculate the experimental shear strain measurement from the rosettes.
With θa equal to -135°, θb equal to -90°, and θc equal to -45°, Eq. 6.2 becomes:
96
ε x = εa − εb + εc
ε y = εb [Eq. 6.3]
γ xy = ε a − ε c
Equation 6.3 was used to calculate the experimental peak-to-peak three second shear
strain ranges.
The resulting plots of the averaged maximum peak-to-peak microstrain range ver-
sus the three second wind velocity at 1 mph (0.447 m/s) intervals is shown in Figure 6.15
for the uni-axial normal strain values and in Figure 6.16 for the rosette shear strain val-
ues.
97
location B demonstrated the largest normal strain ranges during the experimental meas-
urements. Since the loading was directed on the front face of the structure, this outcome
seems reasonable because of the resulting moment about the x-axis. Rosette gauges R2 at
location C and R4 at location G demonstrated the largest shear strain ranges (Figure
6.16). This is because of the torsion created about the post, in addition to the shear cre-
ated by the wind force directed onto the front face of the structure.
98
The fatigue wind was determined to be 38 mph (17 m/s) in Chapter 8: Experimen-
tal Calculations of the Fatigue Load due to Natural Wind Gust. It was found as the wind
velocity with a 0.01% exceedence probability from an annual mean wind velocity equal
phy. This wind velocity was not measured during the experimental tests, and therefore a
regression analysis was performed on the data obtained in Figure 6.15 and Figure 6.16.
The estimation process was performed in similar fashion to the process conducted in the
anchor bolt section, as well as a detailed description provided in Chapter 8 for computing
By using the fundamental fluid mechanics relationship between pressure and the
velocity squared, a regression was performed on the experimental data to form a best fit
99
line that followed this relationship. The independent variable (x-axis, wind velocity) was
transformed into wind velocity squared and plotted with the microstrain ranges to “lin-
earize” the data. A linear best fit line was produced from the transformed data plot and
recorded (see Figure 6.17). The independent variable was transformed back into its origi-
nal form and the trendline equation was used to estimate the microstrain range at the fa-
tigue wind. The resulting plot is shown in Figure 6.18 for SGR-AA-11, one of the strain
gauges with the largest recorded microstrain ranges as an example. At the fatigue wind of
38 mph (17 m/s), the strain range was estimated at 46.854 (10-6), or 46.854 microstrain.
The estimation strain range result for all strain gauges is listed in Table 6.7, along with a
bar chart (Figure 6.19) to demonstrate the variability in strain values as it relates to the
Microstrain Range
Section AA
(in/in)
SGR-AA-3 41.1571
SG-AA-5 45.921
SGR-AA-7 20.9966
Uni-axial SG-AA-9 36.6934
(Normal Strain) SGR-AA-11 46.8539
SG-AA-13 31.621
SGR-AA-15 20.921
SG-AA-1 30.4554
R1 41.0425
Rosette R2 52.5839
(Shear Strain) R3 42.9721
R4 47.0807
101
FIGURE 6.19 Estimated microstrain range at fatigue wind for Section AA.
The stress range at the fatigue wind was determined from the estimated mi-
crostrain ranges in Table 6.7, and compared to the constant-amplitude fatigue thresholds
for the anchor bolt stress category E'. This was done for each of the strain gauges listed in
the table. If the fatigue stress range was above the threshold, the design did not comply
with the Supports Specifications fatigue provisions with respect to the limits placed on
the fatigue stress ranges for column-to-base-plate fillet-welded connection. This is true
regardless if the support structure was designed for fatigue using the in-place fatigue de-
The fatigue stress ranges calculated from the experimental microstrain ranges es-
timated at the fatigue wind is listed in Table 6.8. The values are listed with the constant-
amplitude fatigue thresholds of steel anchor bolts for the stress category E'.
102
TABLE 6.8 Section AA Stress Range and Failure Index for Natural Wind.
Fatigue Mi-
Fatigue Stress Constant-
crostrain
Section AA Range Amplitude Fatigue Failure Index
Range
(ksi) Threshold (ksi)
(in/in)
SGR-AA-3 41.1571 1.194 2.6 0.459
SG-AA-5 45.921 1.332 2.6 0.512
SGR-AA-7 20.9966 0.609 2.6 0.234
SG-AA-9 36.6934 1.064 2.6 0.409
SGR-AA-11 46.8539 1.359 2.6 0.523
SG-AA-13 31.621 0.917 2.6 0.353
SGR-AA-15 20.921 0.607 2.6 0.233
SG-AA-1 30.4554 0.883 2.6 0.340
R1 41.0425 1.190 2.6 0.458
R2 52.5839 1.525 2.6 0.587
R3 42.9721 1.246 2.6 0.479
R4 47.0807 1.365 2.6 0.525
The failure index is also provided in Table 6.8, and is shown graphically in Figure
6.20. A failure index greater than one would indicate a fatigue stress range occurring
greater than the allowable threshold. As listed in the table, the failure index for each
gauge was less than one and subsequently passes the threshold design check. This was
only an estimation of the strain at the base-plate connection as the strain gauges were lo-
cated 12 in (305 mm) above the connection. The failure indexes were approximately
The maximum peak-to-peak strain ranges were determined from the experimen-
tally collected data from the truck tests. Each of the strain gauges was evaluated. The mi-
crostrain ranges per gauge as measured from the truck tests are listed in Table 6.9 and
Table 6.10. The maximum normal strain range was recorded on SGR-AA-11 of truck test
Truck 4 equaling 12.087 microstrain, when the truck was traveling at 70 mph (31.3 m/s).
The maximum shear strain range was recorded on R2 of 13.337 microstrain for the same
truck test. A depth chart showing the induced strain range with respect to truck speed is
shown in Figure 6.21 for the normal strain values measured by the uni-axial strain
gauges, and Figure 6.22 for the shear strain values measured by the rosette strain gauges.
104
TABLE 6.9 Section AA (Normal) Microstrain Ranges from the Truck Tests
TABLE 6.10 Section AA (Shear) Microstrain Ranges from the Truck Tests
The maximum range measured for each gauge was used to calculate the stress
range to determine the failure index for the truck tests. The results are listed in Table
6.11. The results indicate no significant effect from truck gusts on the column-to-base-
plate fillet-welded connection with respect to the fatigue threshold of the detail (Figure
6.23). The column was large enough in size to withstand the truck-induced wind loading.
Aluminum material could be used in place of steel provided that the natural wind fatigue
stress ranges were lower than 1 ksi (7 MPa) which was not the case (see Table 6.8)
107
TABLE 6.11 Section AA Stress Range and Failure Index for Truck Tests
Fatigue Mi-
Fatigue Stress Constant-
crostrain
Section AA Range Amplitude Fatigue Failure Index
Range
(ksi) Threshold (ksi)
(in/in)
SGR-AA-3 11.371 0.330 2.6 0.127
SG-AA-5 10.636 0.308 2.6 0.119
SGR-AA-7 8.803 0.255 2.6 0.098
SG-AA-9 10.140 0.294 2.6 0.113
SGR-AA-11 12.087 0.351 2.6 0.135
SG-AA-13 7.821 0.227 2.6 0.087
SGR-AA-15 8.040 0.233 2.6 0.090
SG-AA-1 7.400 0.215 2.6 0.083
R1 10.660 0.309 2.6 0.119
R2 13.337 0.387 2.6 0.149
R3 10.545 0.306 2.6 0.118
R4 12.306 0.357 2.6 0.137
Two areas of concern were evaluated with the chord-to-column connection. The
first concern was analyzing the chord-to-plate weld (Figure 6.24). The next concern was
the net section of the bolted connection (Figure 6.25). The failure index in regards to the
constant-amplitude fatigue thresholds for the connection detail was checked for both. The
connection involved four chords. Each chord was strain gauged with uni-axial gauges
oriented along the axial length of the chord. A schematic of the instrumentation layout is
shown in Figure 6.26. Each chord was labeled one through four (Figure 6.27), and the
individual strain gauges were labeled per chord as shown in Figure 6.28 of a cross section
Chord-to-Plate
Weld
Placed on each
Chord at 90º
around Chord
Circumference
Centerline
of 1st Truss
Panel
3’- 4.5”
Chord 3
Chord 1
Chord 4
Chord 2
Traffic
SG-C1-10
SG-C1-2
SG-C1-1
SG-C1-3 Chord 1 Chord 3 SG-C1-9
SG-C1-11
SG-C1-4 SG-C1-12
z
y
SG-C1-6 SG-C1-16
SG-C1-15
SG-C1-5
SG-C1-7 SG-C1-14
The evaluation was performed in the same fashion as the anchor bolts and Section
AA. The stress range was determined for natural wind gusts and plotted versus wind ve-
locity intervals for natural wind gusts. A regression was performed on the plot to estimate
112
the stress range at the fatigue wind of 38 mph (17 m/s). The maximum stress ranges were
determined for the truck-induced gusts. The stress ranges were compared to the endur-
ance limit of the particular detail and the failure index was calculated.
Strain ranges were determined from the experimentally collected natural wind
data on the chords. The maximum peak-to-peak strain range was calculated at three sec-
ond intervals. The simultaneously occurring three second wind velocity was also calcu-
lated to juxtapose the chord strain ranges. The maximum strain ranges and wind veloci-
ties were plotted and averaged together at 1 mph (0.447 m/s) intervals. This was per-
formed for all measured data of the strain gauges on the chords. A total of 13 readings
were made: 4 gauges per chord, with 3 gauges not responding during testing (see Figure
6.28). The resulting plots of the averaged maximum peak-to-peak microstrain range ver-
sus the three second wind velocity at 1 mph (0.447 m/s) intervals are shown in Figure
6.29.
As seen in Figure 6.29, strain gauge SG-C2-5 located on the inside of Chord 2
demonstrated the largest strain ranges during the experimental measurements. Most of the
gauges showed similar results. Since the loading was directed on the front face of the
structure, this outcome appeared reasonable. The separation at the higher wind velocities
was due to a lessoned amount of collected data for the velocity intervals.
113
The fatigue wind was determined to be 38 mph (17 m/s) in Chapter 8: Experimen-
tal Calculations of the Fatigue Load due to Natural Wind Gust. It was found as the wind
velocity with a 0.01% exceedence probability from an annual mean wind velocity equal
phy. This wind velocity was not measured during the experimental tests, and therefore a
regression analysis was performed on the data obtained in Figure 6.29. The estimation
process was performed in similar fashion to the process conducted in Chapter 8 for com-
By using the fundamental fluid mechanics relationship between pressure and the
velocity squared, a regression was performed on the experimental data to form a best fit
line that followed this relationship. The independent variable (x-axis, wind velocity) was
114
transformed into wind velocity squared and plotted with the microstrain ranges to “lin-
earize” the data. A linear best fit line was produced from the transformed data plot and
recorded (see Figure 6.30). The independent variable was transformed back into its origi-
nal form and the trendline equation was used to estimate the microstrain range at the fa-
tigue wind. The resulting plot is shown in Figure 6.31 for SG-C2-5, the strain gauge with
the largest recorded microstrain ranges as an example. At the fatigue wind of 38 mph (17
m/s), the strain range was estimated at 46.4367(10-6), or 46.4367 microstrain. The estima-
tion strain range result for all strain gauges is listed in Table 6.12. A bar graph is pro-
vided in Figure 6.32 to illustrate the variability of the strain ranges measured by the
gauges.
Microstrain Range
Chord Chord Strain Gauge
(in/in)
SG-C1-1 33.3139
1 SG-C1-2 36.5662
SG-C1-3 34.407
SG-C2-5 46.4367
2 SG-C2-6 40.8636
SG-C2-8 39.6164
SG-C3-10 35.5541
3 SG-C3-11 39.496
SG-C3-12 33.5019
SG-C4-13 37.6525
SG-C4-14 39.1607
4
SG-C4-15 41.6753
SG-C4-16 39.7908
116
The stress range at the fatigue wind was determined from the estimated and the
1. The weld of the chords to the structural plate. The weld ran parallel to the normal
For Concern 1, the stress category E was used because of the welded connection with ax-
ial and bending loads applied parallel to the weld. Stress category B was used for Con-
cern 2 because of the bolted joints. Although the strain gauges were not located directly
onto these locations, it was felt that an estimate could be made of its fatigue resistance
117
criteria. The resulting failure indexes of each strain gauge for Concern 1 and Concern 2 is
listed in Table 6.13 and Table 6.14, respectively. A plot showing their relationship to
each other and the potential impact of the design to fatigue resistance concerning the two
areas is provided in Figure 6.33. The failure indexes in Table 6.13 and Table 6.14 were
approximately 25%, proving that the fatigue stress in the chord-to-column connection
TABLE 6.13 Chord Stress Range and Category E Failure Index for Natural Wind
Fatigue Mi-
Chord Fatigue Stress Constant-
crostrain
Strain Range Amplitude Fatigue Failure Index
Range
Gauge (ksi) Threshold (ksi)
(in/in)
SG-C1-1 33.3139 0.966 4.5 0.215
SG-C1-2 36.5662 1.060 4.5 0.236
SG-C1-3 34.407 0.998 4.5 0.222
SG-C2-5 46.4367 1.347 4.5 0.299
SG-C2-6 40.8636 1.185 4.5 0.263
SG-C2-8 39.6164 1.149 4.5 0.255
SG-C3-10 35.5541 1.031 4.5 0.229
SG-C3-11 39.496 1.145 4.5 0.255
SG-C3-12 33.5019 0.972 4.5 0.216
SG-C4-13 37.6525 1.092 4.5 0.243
SG-C4-14 39.1607 1.136 4.5 0.252
SG-C4-15 41.6753 1.209 4.5 0.269
SG-C4-16 39.7908 1.154 4.5 0.256
118
TABLE 6.14 Chord Stress Range and Category B Failure Index for Natural Wind
Fatigue Mi-
Chord Fatigue Stress Constant-
crostrain
Strain Range Amplitude Fatigue Failure Index
Range
Gauge (ksi) Threshold (ksi)
(in/in)
SG-C1-1 33.3139 0.966 16 0.060
SG-C1-2 36.5662 1.060 16 0.066
SG-C1-3 34.407 0.998 16 0.062
SG-C2-5 46.4367 1.347 16 0.084
SG-C2-6 40.8636 1.185 16 0.074
SG-C2-8 39.6164 1.149 16 0.072
SG-C3-10 35.5541 1.031 16 0.064
SG-C3-11 39.496 1.145 16 0.072
SG-C3-12 33.5019 0.972 16 0.061
SG-C4-13 37.6525 1.092 16 0.068
SG-C4-14 39.1607 1.136 16 0.071
SG-C4-15 41.6753 1.209 16 0.076
SG-C4-16 39.7908 1.154 16 0.072
The maximum peak-to-peak strain ranges were determined from the experimen-
tally collected data from the truck tests. Each of the strain gauges was evaluated. The mi-
crostrain ranges per gauge as measured from the truck tests are listed in Table 6.15 – 6.19
for each chord respectively. The maximum strain range was recorded on SG-C2-5 of
truck test Truck 3 equaling 12.075 microstrain, where the truck was traveling at 70 mph
(31.3 m/s). A plot depicting the distribution of strain with respect to the truck speed is
The maximum range measured for each gauge was used to calculate the stress
range to determine the failure index for the truck tests. The results are listed in Table 6.19
for Concern 1 (stress category E) and Table 6.20 for Concern 2 (stress category B). The
results indicate no significant effect from truck gusts on the chord connection (welded
TABLE 6.19 Chord Stress Range and Category E Failure Index for Truck Tests
TABLE 6.20 Chord Stress Range and Category B Failure Index for Truck Tests
Discussion of Results
An analysis of the fatigue resistance of the structure was performed. The stress
provided by the Supports Specifications. A failure index was calculated to provide a rela-
tionship between the gauged locations with a control value. The distribution of the stress
in the structure was evaluated and the most appropriate location for determining the fa-
The largest strain recorded was located on the anchor bolts followed by the post
and chords as expected. The distribution of stress followed a clear and distinct path to the
foundation. The major findings are listed in Table 6.21 and 6.22 for natural wind and
truck-induced gusts. The table lists the stress ranges determined at the average wind ve-
124
locity and the fatigue wind velocity. Maximum ranges at the specified location are listed
for each wind velocity, as well as the average stress range from all gauges at the section.
For the truck tests, the stress ranges on averaged increased with the speed of the
truck. No significant loading was generated from the truck gusts as compared to the natu-
125
ral wind gust measurements. The stress generated from natural wind was as much as four
times greater than truck-induced wind gusts. This is illustrated in the bar graph of Figure
6.35, showing the stress ranges for the fatigue wind as compared to the stress ranges for
the 70 mph (31.3 m/s) truck speed. A significant increase in stress range for both natural
wind and truck gusts was observed for the anchor bolts. This was due to the anchor bolts
with large clearances as compared to the other bolts, resulting in an unbalanced distribu-
tion of stress. The bolt with the largest stress ranges were located closest to the neutral
FIGURE 6.35 Comparison of results between natural wind and truck gusts.
Significant conclusions were made after the fatigue behavior of the structural
members was evaluated. The results helped to direct the project as to which gauges to use
126
in calculated the fatigue loads due to natural wind and truck-induced wind gusts. The
depth charts helped to distinguish the areas of the structure that were most vulnerable to
fatigue loading. The analysis detailed the stress flow in the structure from the point of
loading to the connections. Large strain ranges were measured in the chords, post, and
anchor bolts, with the largest found in the anchor bolts. Strain values measured in the
chords were slightly less than the strains in the post indicating a smooth stress transfer
The stress ranges in the column and chords were not significant enough to create
fatigue damage, however the stress ranges in the anchor bolts did show reasonable evi-
dence of possible fatigue failure. It was a fatigue failure of the anchor bolts that caused
the previous structure to fail due to a combined stress of axial and bending. The problem
areas with this project were associated to the bolts with clearances equal to 3 in (76.2
mm) or more. It was believed that if the clearance lengths were decreased, a better distri-
bution of the stress from the column to the foundation would occur, thereby reducing the
overloading of stress onto one anchor bolt. The stress distribution would better reflect the
structural analysis during the design phase of the support structure, and subsequently in-
crease the accuracy of the fatigue analysis estimation made by the engineer.
One aspect needed in calculation of the fatigue load was a lessoned variability of
the strain readings from one gauge to the next as was found with the anchor bolt meas-
urements. Although the distribution would not be the same because of the location to the
neutral axis from one gauge to the other, a preconceived pattern illustrating a balanced
state should be easily observed. Likewise, strain readings in the post were decided as the
most noteworthy from the other instrumented locations in regards to fatigue load calcula-
127
tions. The strains in the anchor bolts and chords varied significantly from the balanced
state, with extreme differences found in the anchor bolts. In view of the results, it was
concluded that the best location for calculating the fatigue loads from the measured data
CHAPTER 7
Chapter Overview
A modal analysis of the structure was conducted from the experimental data. The
accelerometers measurements were used for the study. Important dynamic properties
were gathered from the analysis to be used in the development of the fatigue loads due to
the structural system. It was obtained through an experimental technique using measured
acceleration data obtained from operational dynamic forces. The major parameters that
The investigation of these aspects helped to devise and develop the fatigue loading due to
The data obtained from the accelerometers were used for the analysis. Although
the strain gauges could be used, the only parameters extracted from the analysis would be
the modal frequencies and modal damping values. The modal shapes cannot be deter-
They were of the piezoelectric type that uses a piezoelectric crystal mounted to a small
mass from which the voltage output is converted to acceleration. Each accelerometer had
a maximum capacity of 96.5 ft/sec2 (3 G). The locations of the accelerometers were stra-
tegically chosen, as the accelerometer only gave the natural frequency of the member
upon which it rests. A combined total of six unidirectional accelerometers were required
for the structure. They were placed at particular locations to measure each possible de-
gree of freedom in vibration direction. Accelerometer locations are shown in Figure 7.1.
130
AC-1
AC-2
Centerline
Post
z
AC-3
y Centerline
Truss
x
Three accelerometers were placed at location AC-1 (see Figure 7.1) to measure
the vertical (perpendicular to the direction of traffic), horizontal (parallel to the direction
of traffic), and longitudinal (transverse to the direction of traffic) accelerations of the post
member. Two accelerometers were placed at location AC-2 to measure the vertical and
horizontal directions (Figure 7.2). One accelerometer was placed at location AC-3 to
measure the horizontal direction for indentifying torsion behavior of the overhang truss
(Figure 7.3). An overview of locations AC-2 and AC-3 is shown in Figure 7.4.
131
threaded screws. Electrical insulation between the accelerometer and the test surface was
built into the transducer device. All mounting surfaces were flat to avoid distortion that
may produce strains which could affect the accelerometer’s response. Carefulness was
taken when screwing the accelerometers to the attachment surface as to not overreach the
Mounting blocks were used at each location for accelerometer attachment. For
instance at location AC-1 where three accelerometers were needed, all three were at-
133
tached to one single attachment block. A picture of this location with the attached mount-
ing block is shown in Figure 7.5. Once the truss section of the structure was erected, the
AC-1 accelerometers were screwed onto the mounting block with the sensitive ends
(measurement direction) oriented in the proper directions. The same type of block was
used at locations AC-2 and AC-3. Due to the round surface at these locations (steel pipe
truss web member), a small flat steel plate was welded to the side of the members for
which the mounting blocks and accelerometers were attached. The size of the plate was
manufactured as small as possible, with enough space to accurately and securely mount
the accelerometers and the plate to the web member, but without the possibility of creat-
ing significant additional wind drag. The initial mounting instructions prepared for AL-
The number of degrees of freedom was chosen to represent the dynamic behavior
of the structural system. They were chosen based on the expected modal shapes of the
system. Five dynamic vibratory shapes were identified as crucial dynamic behavior that
1. Torsion of the cantilevered truss and sign about the support shaft (moment about
z-axis)
2. Vertical vibration of the truss and shaft in the plane perpendicular to the ground
3. Horizontal vibration of the truss and shaft in the plane parallel to the ground
4. Longitudinal vibration of the truss and shaft along the axial length of the truss
5. Torsion vibration of the truss about the longitudinal axis (moment about x-axis)
The placements of the accelerometers were identified based on the five assumed vibra-
tory responses and are shown in Figure 7.1. The associated degrees of freedom and their
AC-2: 2y, 2z
AC-3: 3y
135
Six degrees of freedom were indentified with the cantilever-type sign support structure of
this study.
Spectral Analysis
A spectral analysis was conducted first to obtain natural frequencies and their or-
der of appearance. These values were used in identifying and determining modal shapes.
Only the first four modes were discovered from the spectral plots. From the analysis of
the plots, the first two modes showed the primary means of vibration.
The data collected for each accelerometer was used in the analysis. The collected
time duration used was 10 minutes. In order for completeness and accurate representation
of the dynamic behavior of the structural system, the chosen event must excite all major
natural frequencies. The event was representative a random vibration occurrence, and as-
sumed as a continuous function. Hanning windows were used to ensure the signal begins
and ends at zero. The Fourier transform (shown as Eq. 7.1) was used to transform the
∫ x (t ) e
− j 2πft
X(f) = dt [Eq. 7.1]
−∞
where
X ( f ) = fourier transform
x(t ) = continous time series
j = −1
f = frequency where - ∞ < f < ∞
t = time
The spectral plot of the forced event causing the structural vibration is shown in
Figure 7.6. The data used for this analysis was taken from the anemometer placed 4 ft
(1.22 m) above the post support, AN-4. The measured data was in velocity units (speed
per second), from which the drag pressure was calculated (Eq. 7.2). According to the ori-
entation of the equation, the value calculated was in the form of the drag pressure divided
PD 1
= ρV 2 [Eq. 7.2]
CD 2
where
PD = drag pressure
C D = drag coefficient
slug kg
ρ = density of air, 0.0023657 3
1.22 3
ft m
V = wind velocity
137
The plot shows significant pressure around the lower frequencies, showing a broadband
spectrum evident of the gustiness and turbulent nature of wind behavior, and drastically
A typical spectral plot of the response of the structure to the forced event, meas-
ured in acceleration units from the accelerometers, is shown in Figure 7.7. The plots
shows a significant broadband spectrum around the lower frequencies evident of the
the natural frequencies of the structural system evident of the resonance effect. These
spikes are identified as the natural frequency of vibration of the support structure. The
resulting response of the structure is a combination of the background turbulence and the
resonant vibration.
138
The next step was compiling the modal frequencies of the support structure meas-
ured from each accelerometer during structural vibration. The frequencies were identified
by the spikes on the spectrum plots. Four major natural frequencies were found and are
listed Table 7.1 in their order of appearance along the frequency domain. The spectrum
plots of the frequencies are shown in Figure 7.8 and Figure 7.9 for accelerometers AC-2-
Y and AC-2-Z, located on the top horizontal strut of the cantilevered truss overhang.
The plots demonstrate a dominance of vibration associated with the first two mo-
dal frequencies. Figure 7.8 was from measured acceleration in the y direction (horizontal,
frequency of 1.61 Hz identified as mode 1. Figure 7.9 was from acceleration measured in
mode 1 in Figure 7.7 was greater than mode 2 in Figure 7.8, which revealed the majority
of vibration was controlled by mode 1 during the measured sample. Other frequencies of
vibration were excited, but the implication of these vibrations was small.
Modal Shapes
The modal shapes were indentified for each of the four modes of vibration de-
scribed in the spectral analysis section. The modal shapes illustrated the direction and
shape of the vibrating structure. This was done using a system known as Quadrature
Picking using frequency response functions (FRF) formulated from the force and re-
sponse spectral plots. The shapes were found by using the imaginary component of the
Fourier transform (Eq. 7.1) to identify the modal direction and placement of the point on
the structure where one accelerometer with respect to another accelerometer was located
during resonant vibration. The strategic placement of the accelerometers, and the direc-
tion they measure, before testing during the development of the research program was
The FRF is described as the ratio of the output spectrum to the input spectrum. In
the context of this project, the FRF was the spectral acceleration divided by the spectral
wind pressure. This particular FRF, shown in general terms in Eq. 7.3, is commonly re-
ferred to as accelerance:
X (ω )
H (ω ) = [Eq. 7.3]
F (ω )
where
H (ω ) = accelerance
X (ω ) = spectral acceleration response
F (ω ) = spectral force excitation
ω = frequency
The process required remedial steps before formulating Eq. 7.3 in order to reduce
noise errors in the output spectra. Through using the method of least squares, the result-
ing accelerance was estimated as the cross spectrum of the response and the force excita-
tion divided by the autospectrum of the excitation, as shown in Eq. 7.4 (39). This helped
( F )( X ) GFX (ω)
H 1 (ω) = = [Eq. 7.4]
( F )(F ) GFF (ω)
142
where
H 1 (ω ) = accelerance estimator
F = spectral force excitation input
X = spectral accleration response output
G FX (ω ) = cross spectrum of response and excitation
G FF (ω ) = autospectrum of excitation
The spectral force excitation was in the form of a Fourier transform shown in Fig-
ure 7.6. The spectral response was the Fourier transform of the accelerometer data de-
scribed in Figure 7.8 and 7.9 as an example. The accelerance was calculated for each ac-
celerometer located on the post and truss overhang. Real and imaginary components re-
sulted in the Fourier transform calculation, from which the imaginary component was
used to identify modal shapes in a process known as Quadrature Picking. The accelerance
becomes imaginary at the modal frequencies, whereby the amplitude of the imaginary
component is proportional to the modal displacement of the point on the structure where
the accelerometer was attached. A basic outline of the modal shape was achieved with
each modal frequency by first calculating the accelerance for each accelerometer, identi-
fying the modal frequencies in the accelerance, plotting the imaginary components at
each identified frequency, and then comparing the modal displacements between the ac-
celerometers (39).
Mode 1
support, and AC-2-Y and AC-3-Y on the truss overhang. The plots of the imaginary
components at the mode indicated a larger modal displacement of the truss in the y direc-
143
tion than the shaft. The plots of the imaginary components are shown together in Figure
7.10 for AC-1-Y and AC-2-Y. Accelerometer AC-3-Y was identical to AC-2-Y. The
comparison of the modal displacements between the two accelerometer locations demon-
strated a torsion vibration of the truss about the support shaft by the relatively large mo-
dal displacement of the truss to the shaft. Each accelerometer plot indicated a negative
modal displacement in the y direction. A general illustration of the outlined modal shape
obtained from experimental data is shown in Figure 7.11. The mode is best described as
torsion about the support shaft with a slight bending of the shaft in the y direction.
AC-1-Y
AC-2-Y
z
AC-3-Y
x
y
Mode 2
support, and AC-2-Z on the truss overhang. The plots of the imaginary components at the
mode indicated an upward modal displacement of the truss in the positive z direction, fol-
lowed by a backward modal displacement of the shaft in the negative x direction. This
was indicative of a vertical rocking motion of the truss and shaft support. Mode 2 was
slightly picked up by AC-1-Z on the shaft, but was less significant than the other acceler-
ometers. This mode was not measured from the accelerometers oriented in the y direc-
tion, indicating no movement in this direction of this mode. The plots are shown together
in Figure 7.12 for AC-1-X and AC-2-Z. A general illustration of the outlined modal
AC-1-X
AC-2-Z
x
y
Mode 3
support, and AC-2-Y and AC-3-Y on the truss overhang. The plot of the imaginary com-
ponent for the post showed a significant modal displacement in the positive y direction,
whereas the modal displacement of the truss was opposite in direction and less signifi-
cant. The plots are shown together in Figure 7.14 for AC-1-Y and AC-2-Y. AC-3-Y was
identical to AC-2-Y. The modal shape is best described as a horizontal twist of the truss,
but also described by a horizontal rocking of the post support with the truss displacement
dependent and rocking backward in response. A general illustration of the outlined modal
AC-1-Y
AC-2-Y
AC-3-Y
x
y
Mode 4
support, and AC-2-Z on the truss overhang. The plots of the imaginary components at the
mode indicate an upward modal displacement of the truss in the positive z direction, fol-
lowed by a forward modal displacement of the shaft in the positive x direction. This was
indicative of an outward and inward clamping motion of the truss and shaft support.
Mode 4 was slightly picked up by AC-1-Z on the shaft, but was less significant than the
other accelerometers. This mode was not measured from the accelerometers oriented in
the y direction, indicating no movement in this direction of this mode. The plots are
shown together in Figure 7.16 for AC-1-X and AC-2-Z. A general illustration of the out-
lined modal shape obtained from experimental data is shown in Figure 7.17.
148
AC-1-X
AC-2-Z
x
y
Modal Damping
A modal damping analysis was performed from the experimentally collected ac-
celerometer data in order to determine the damping ratio, ξ, of the structure for further
fatigue analysis in this project. The damping ratio was used in developing a dynamic
model of the structure to simulate the response to wind loading events, and to determine
the fatigue load based on the dynamic characteristics. This process is discussed in detail
in Chapter 9: Theoretical Calculation of the Fatigue Load to due Natural Wind Gusts.
There are many ways to calculate the modal damping of the structure. For the
purpose of this project, only the damping of the first two major vibratory modes was
needed. This was because Mode 1 and Mode 2 formed the majority of the vibration, as
seen in Figures 7.8 and 7.9. Their modal frequencies were nearly identical, but this did
not suggest that the rate at which the vibration decays (damping) were identical. It is also
important to point out that, by looking at Figures 7.8 and 7.9, the majority of vibration
was in the first modal frequencies, and therefore the vibration shape was controlled pre-
The transient events used in the damping analysis were the measured response of
the truss from experimental testing of the fatigue load due to truck gusts. A large truck
was driven periodically under the sign structure and the response of the structure was
measured using anemometers, strain gauges, and accelerometers. This was done on a
relatively low wind day as to not have external effects from natural wind gusts. The result
was a noticeable transient event as seen from the response data. A total of four truck runs
were used for the analysis, from which two damping calculations were made per mode
(mode 1 and mode 2), equaling a total of eight damping calculations per mode, and 16
150
damping calculations total. An average was taken from the eight values for each and used
for the remainder of the project as the damping value of the sign structure for the mode of
interest.
An understanding of the forced excitation was developed before the damping cal-
culations were made. An experimentally measured transient event was isolated and plot-
ted on amplitude versus time graph. An example of a typical truck transient event meas-
ured by the accelerometers, and used for this analysis, is shown in Figure 7.18.
The event depicted in Figure 7.18 was representative of a free vibration system of oscilla-
tory motion. The system was underdamped and periodic, since the majority of vibration
was controlled by a single frequency; in this case, Mode 1 equal to 1.61 Hz. Other modal
151
frequencies were present due to the ruggedness of the plot, but were insignificant to the
major modal shape of Mode 1. The duration of the event was roughly 10 seconds. It was
measured using accelerometer AC-2-Y, which measured oscillatory motion in the y direc-
damped, periodic dynamic system. The expression shown in Eq. 7.5 is a representation of
the exponential decay in amplitude with time in conjunction with the periodic motion of
where
x(t ) = amplitude as a function of time (G)
A0 = initial amplitude (G)
ξ = damping ratio
ω n = natural frequency (rad/sec)
ω d = damped frequency (rad/sec)
t = time (sec)
ϕ 0 = initial phase (rad)
The exponential expression in Eq. 7.5 (shown as Eq. 7.6) is representative of the expo-
nential decay of the amplitude peaks with increasing time. The cosine expression repre-
sents the periodic motion of the system. In relation to the measured transient event, Eq.
where
y = amplitude as a fuction of time of the exponential expression (G)
A = initial amplitude of the exponential expression (G)
ξ = damping ratio
ω = natural frequency (rad/sec)
t = time (sec)
A regression analysis was done to determine the damping ratio. The positive peak
amplitudes of each wavelength was extracted from the time history graphs and plotted
separately. The time was equalized to start from zero so that the amplitude, A0, could be
153
realized. An example of the resulting plot is shown in Figure 7.20. A trendline was fitted
to the plotted peaks and the equation of the trendline was extracted. The exponent of the
trendline equation was set equal to the absolute value exponent of Eq. 7.6. By knowing
the natural frequency of the structure, ω, which was determined in the Spectral Analysis
and Modal Shapes sections of this chapter, the damping ratio, ξ, was calculated. A se-
quence of the calculation events, using the trendline equation in Figure 7.20 as an exam-
where
A = 0.0447 G
-ξωt = −0.244 x
t=x
and with
ω = 1.61 Hz = 10.116 rad/sec (Mode 1)
therefore
− ξωt = −0.244 x
0.244 [Eq. 7.8]
ξ= = 0.02412 = 2.41%
10.116
This process was performed for the upper positive peaks and lower negative peaks
for each transient truck event measured. The above example was for AC-2-Y, measuring
vibration in the horizontal direction of Mode 1. The damping for Mode 2, using AC-2-Z,
was also performed in the same manner. The results of the modal damping analysis are
listed in Table 7.2. The larger damping value for Mode 1 was attributed to an additional
aerodynamic damping of the sign as it vibrated horizontally about the shaft support.
155
Top/Bottom
Transient Damping Ratio
Accelerometer Mode Peak Ampli-
Event (%)
tudes
Top 2.41
AC-2-Y Mode 1
Bottom 2.30
1
Top 1.38
AC-2-Z Mode 2
Bottom 1.63
Top 2.20
AC-2-Y Mode 1
Bottom 2.19
2
Top 0.65
AC-2-Z Mode 2
Bottom 0.67
Top 1.50
AC-2-Y Mode 1
Bottom 1.38
3
Top 0.543
AC-2-Z Mode 2
Bottom 0.495
Top 1.246
AC-2-Y Mode 1
Bottom 1.335
4
Top 0.398
AC-2-Z Mode 2
Bottom 0.922
Average Damping Ratio for Mode 1 (1.61 Hz): Shaft Torsion 1.820
Average Damping Ratio for Mode 2 (1.64 Hz): Vertical Rocking 0.836
156
CHAPTER 8
Chapter Overview
The calculation of the fatigue load due to natural wind from the experimental data
is presented in this chapter. Only collected data with wind velocity greater than 9 mph (4
m/s) and directed onto the front face of the structure was used. The fatigue load due was
based on an upper to lower peak–to-peak stress range. A three second average was used
for data reduction. The data was broken down into two components 1) structural excita-
tion, and 2) structural response. The anemometers made up the excitation, whereas the
The equivalent static wind load approach was used in determining the fatigue load
due to natural wind. The excitation and response of the structure was measured experi-
mentally. The behavior was dynamic in nature. An equivalent static wind load was back-
calculated from the measured response values that would produce the same dynamic re-
The experimental data collected was analyzed in the same fashion as the devel-
opment of the natural wind fatigue provisions in the Supports Specifications, except for
157
1. Wind velocity directionality unit vectors were developed to describe loading ori-
entations,
3. Equivalent static wind pressures were back-calculated using the stress ranges and
4. The pressures were categorized to their corresponding wind velocities and plotted,
termined, and
The intention for the likeness in analysis to the Supports Specifications was to enable a
direct comparison to be made between the code and the experimental data. This would
The data collected from the strain gauges form the basis of the analysis. Strain
gauges were placed on the chord members, the supporting shaft, and the anchor bolts,
totaling 44 gauges. It was determined that the most consistent measurements were found
in the gauges located on the post (see Chapter 6: Fatigue Resistance). The usage of the
post gauges also allowed for a more straightforward back-calculation of the equivalent
The step-wise procedure for determining the fatigue load generalized in the nu-
meration presented earlier is detailed in the following sections. The detailing describes
the procedure performed by the researcher starting from the experimentally collected raw
data to the back-calculation of the equivalent static wind load performed at the conclu-
Structural Excitation
A total of 36.75 hours of excitation and response data was recorded for the natural
wind fatigue load program. Since the acquisition system recorded data at a rate of 60
samples per second, the collected hours corresponded to 7,938,000 data points per in-
strument and 492,156,000 total data points of all 62 instruments. The collection dates and
times of the data samples are shown in Table 8.1. The wind data was collected in 45 min-
ute intervals except for data collected on 4/8/2009 and 4/9/2009, which was taken in 25
minute intervals. The collection data on these dates were designated for truck-induced
wind gust testing, however natural wind data was collected between truck runs. A total of
53 data collection intervals were accomplished: 44 intervals of 45 minutes each, and nine
The data collected from the anemometers were used to distinguish between usable
and discarded data to apply for wind pressure back-calculation. Data was considered us-
Wind velocity was directed onto the front face of the structure.
The data was broken down into wind velocity and direction. It was discovered from the
collected structural response data that significant structural vibration was only induced
with wind velocities greater than 9 mph (4 m/s). Only wind velocities directed onto the
front of the structure were used for evaluation. The resulting usable data that complied
with the above limitations was three hours and thirty minutes, equaling a total of
The natural wind data collected was nicely spread between the desired wind ve-
locities. Wind rose diagrams were developed for all data collected to help in determining
usable data directed onto the front face of the structure. The wind rose diagram in Figure
8.1 is of all 36.75 hours collected, whereas the wind rose diagram in Figure 8.2 is of the
data determined usable and used for the analysis. The data demonstrated in the wind rose
diagrams were measured using the ambient wind anemometer AN-4. The directions are
shown in compass bearings. The North end of the anemometer was oriented opposing
The average wind velocity of the data considered usable was 12.96 mph (5.79
m/s). The measurement was taken from anemometer AN-4 located 4 ft (1.22 m) above
the structure equaling 32.75 ft (9.98 m) above ground level. The wind velocity was ori-
ented onto the front face of the structure as shown in Figure 8.2. A maximum wind veloc-
0° (North)
Counts
180° (South)
0° (North)
Counts
270° (West) 90° (East)
180° (South)
The collected excitation data determined as usable was further reduced for analy-
sis. The direction of the wind velocity vector was transformed from the compass bearings
used for experimental measurement, to polar bearings for data analysis. A wind direc-
tionality unit vector was developed to use in the back-calculation for determining the
Averaging Time
The data streamlines were averaged every three seconds in order to be in compli-
ance with the drag coefficients and height coefficients available in the Supports Specifi-
The magnitude and direction of the wind velocity vector was broken down into
was oriented in the direction away from the front face of the structure and opposing the
direction of traffic. An average of the components for every three seconds was taken. The
wind velocity vector was then reformed for each three second window.
The reasoning for this procedure was twofold. First, it solved the cross-over prob-
lem. In compass bearings, 360° was the same as 0° on the direction scale. A simple aver-
age of values without a breakdown into components would not take care of this issue. For
example, the average of 315° and 45° equals 180°, which would indicate the wind was
blowing opposite (South) of the true direction (North). A breakdown into components
solved the cross-over problem. Secondly, low wind velocities did not excite as much vi-
bration in the structure as higher wind velocities. For that reason, the X and Y compo-
nents of the breakdown took into account the magnitude of the wind velocity and not just
direction. The restructuring of the wind velocity vector over the three second averaging
window placed more emphasis on the larger magnitudes which would create more sig-
nificant vibration, and less emphasis on the smaller velocities which had less significant
results.
163
Once the three second window averaging was accomplished, the reformed three
second wind velocity vector was transformed from its compass bearings to polar bear-
The 180° addition was to account for incoming wind as the anemometer measured the
wind direction along compass bearings from the direction it was blowing.
Equation 8.1 was applied to anemometer AN-4 measuring the horizontal plane.
Anemometer AN-3 measured the vertical plane, and AN-3 was not working for the diag-
nosed usable data. When it was working, the wind velocity direction was primarily be-
tween 350° and 10° (cross-over through zero), and averaged close to zero. In view of this,
After the excitation was transformed from compass bearings to polar bearings, a
wind directionality unit vector was formed. The unit vector was developed into Cartesian
coordinates (i, j, k). This was done in order describe mathematically where the wind ve-
locity, and subsequent wind pressure load, was directed for back-calculation of the
equivalent static wind load. Using the polar coordinates, the unit vector was calculated by
λ = (λ x i + λ y j + λ z k )
V V y V
= x i + j + z k [Eq. 8.2]
V V V
= [(cos β cos α )i + (cos β sin α ) j + (sin β )k ]
where
λ = wind directionality unit vector
α = wind orientation along the horizontal plane (xy - plane) measured from AN - 4
β = wind orientation along the vertical plane (zy - plane) measured from AN - 3
A schematic of the coordinate system is shown in Figure 8.3 for better understanding of
and the development of the unit vector are shown in Table 8.2.
A wind velocity directionality unit vector was formed for each three second averaging
window for all usable data collected. The unit vector represented the direction of the
structural excitation with respect to the coordinate system shown in Figure 8.3. The mag-
nitude of the wind velocity was proportional to the magnitude of the wind pressure to be
Structural Response
The structural response analysis involved the strain gauge data corresponding to
the usable wind data (data was measured simultaneously in the field). The same three
second averaging window of the structural excitation was used for the analysis. The ma-
jor outcome was to determine the experimentally measured stress ranges from the meas-
ured structural excitation to back-calculate the equivalent static wind load and subsequent
Data Offsetting
The first step involved an offsetting procedure to generate the true strain time his-
tory from the collected raw data. Wind velocity occurs randomly in nature and cannot be
controlled manually. For this reason, the zeroing of the gauges before testing in the field
was not representative of a true no-strain condition. The structure was continually vibrat-
ing due to the wind velocity presence, and the zeroing of the scales occurred during
backward and sometimes forward vibratory movement which was unavoidable. The solu-
tion was to perform an offsetting procedure on the collected data to offset the strain value
Important for this process was the simultaneous collection of strain and wind ve-
locity data. The strain data collected during each run was filtered with respect to its corre-
sponding collected wind velocity. This was done in +/- 0.5 mph (0.224 m/s) wind veloc-
ity magnitude intervals. The strain values that occurred during each interval were deter-
mined and averaged together. The values were plotted on a wind velocity versus strain
diagram (see Figure 8.4). For example, all strain values occurring during a 3.5 to 4.5 mph
(1.56 to 2.01 m/s) interval were averaged and plotted as a 4 mph (1.79 m/s) data point.
The data formed a parabola, which adhered to the fluid mechanics relationship between
FIGURE 8.4 Strain values per 0.5 mph wind velocity intervals.
A regression analysis was performed to determine a best fit line of the plotted
data. This was done through a transformation regressor linearization process. The inde-
pendent variable (wind velocity on the abscissa axis) was squared. It was plotted versus
its corresponding averaged strain (see Figure 8.5). The linearization of the data proved
the purely parabolic nature of the data. A best fit line was then constructed as a linear
predictor. The intercept of the trendline on the ordinate axis indicated the strain value to
offset. For example, the offset for the data run presented in Figure 8.5 was 1.1501 mi-
crostrain. This means that when the strain gauges were zeroed during the testing proce-
dure, the structure was vibrating due to the continuous wind excitation, and at the mo-
ment of zeroing an approximate magnitude of 1.1501 microstrain was induced onto the
structure. This is evident from Figure 8.5 at the zero squared wind velocity point.
168
Returning to Figure 8.4, a line was plotted to fit the data based on the information
gathered from the transformation (Figure 8.5). As shown in Figure 8.5, the slope of the
trendline was the slope of the parabola, and the intercept of the trendline was the intercept
of the parabola. Plotting a line on Figure 8.4 using the transformation regressor values is
y = 0.0351x 2 + 1.1501
Once the trendline was produced for the parabolic data, the offset (1.1501 for this
data run example) was subtracted from the trendline to produce an accurate zeroed time
history. The offset plot is shown in Figure 8.7. As a side note, the plot in Figure 8.7 can
be used to project strain values for higher wind velocities than that used for fatigue analy-
sis. For instance, Figure 8.8 shows the plot projected to a 90 mph (40.2 m/s) wind, which
y = 0.0351x 2
1. Filter the collected strain data in +/- 0.5 mph (0.224 m/s) intervals corresponding
3. Averaged strain values versus the wind velocity interval were plotted to test cor-
4. The data was transformed into a linear plot by squaring the independent variable
(wind velocity),
5. The transformed wind data was plotted with its corresponding averaged strain to
6. A regression was performed on the transformed data to generate a linear best fit
trendline,
7. The trendline was projected to determine its intercept with the ordinate axis, and
8. The intercept value was subtracted from the raw strain data time history stream-
The offsetting procedure described was done for each strain gauge time history of
the usable data collection. It produced the true strain values to account for zeroing the
instrumentation during wind velocity excitation. To lessen the magnitude of the offset,
ALDOT was encouraged to zero the instrumentation and begin data collection when the
wind velocity seemed to be low. Since zeroing at exactly a zero wind velocity was practi-
Strain Ranges
The next step was analyzing the offset data to determine the strain range. The off-
set data was averaged every three seconds. The maximum and minimum values within
each three second window were determined and subtracted from each other. The result
was the maximum peak-to-peak range within each three second window. This was done
for all usable offset data and strain gauges. Other parameters such as the standard devia-
tion and peak-to-standard deviation ratio were determined for each three second window
The rosette gauges located at Section AA were used to measure shear stresses in
the shaft. Before the peak-to-peak range was determined, a strain transformation was
needed to calculate the shear strain. The gauges were rectangular strain rosettes arranged
in a 45° pattern as shown in Figure 8.10. The strain transformation equation, Eq. 8.3, was
used to calculate the experimental shear strain measurement from the rosettes.
With θa equal to -135°, θb equal to -90°, and θc equal to -45°, Eq. 8.3 becomes:
ε x = εa − εb + εc
ε y = εb [Eq. 8.4]
γ xy = ε a − ε c
Equation 8.4 was used to calculate the experimental shear stress in the shaft for which the
stress ranges were determined and the wind pressure magnitude, P, was found.
174
The equivalent static wind load that would produce the same peak-to-peak stress
range determined experimentally was back-calculated. The three second ranges and wind
directionality unit vectors were used for this calculation. Two measures formed the basis
of the process. First, the stress range was determined from the experimentally measured
strain ranges as described in the Structural Response section of this chapter. Second, a
theoretical structural analysis was performed on the structure to determine the stresses at
the gauged locations for comparison with the experimental values. The calculation in-
normal and shear stress values for each three second interval was developed at the loca-
tion of the gauge to be used in the comparison. Importantly, the stress values were devel-
175
oped with the wind pressure magnitude, P, kept as a variable. The two measures, experi-
mental stress and theoretical stress, were set equal to each other and the wind pressure
magnitude, P, was solved (see Eq. 8.5). The instrumented locations along the shaft sup-
port (Section AA and Section BB) were used exclusively for this evaluation because of
natural wind velocity and using them to form force vectors along with the wind direction-
ality unit vectors. The wind pressure magnitude, P, was kept as a variable. This produced
point loads with direction at each exposed area. Drag and height coefficients provided by
the Supports Specifications, which were developed using three second wind averages in
previous studies, were determined for each member exposed to wind and used in the
analysis. The idea was to simulate as accurately as possible the wind loading conditions.
Equations of equilibrium were developed and the acting moment vector (resultant) was
solved. The moment vectors were used to develop stress elements in a combined loading
analysis and unsymmetrical bending at each strain gauged location, and Eq. 8.5 was used
The front and the sides of the structure were exposed to wind pressure from the
wind data determined as usable. Each face was divided into individual segments for
structural analysis. The segmented division depended on the type of member exposed.
The front face of the structure contained three segments, front sign, front truss, and front
shaft. The side of the structure was segmented into two areas: East side face and the West
side face. An illustration of the exposed areas to wind pressure used for the structural
analysis is shown in Figures 8.11 through 8.13. Each exposed area, area centroid, and its
distance to the strain gauges at Section AA and Section BB were calculated. The results
are shown in Table 8.2. The shaft segments were further broken down into three seg-
ments to account in variation of wind pressure with height (detailed in the next section).
Effective areas of the areas listed in Table 8.3 were calculated. The effective area
represented the exposed area (Table 8.3) multiplied by the drag coefficient of the mem-
bers that make up the area, and the height coefficient of the area centroid above ground
level. The back-calculation was based on three second ranges and wind directionality unit
vectors, and therefore the drag coefficients and height coefficients in the Supports Speci-
Drag coefficients. All drag coefficients were calculated using Table 3-6 of the
Supports Specifications. Fifty year design life was used in the calculation with a velocity
conversion factor equal to 1.00 (no conversion needed). The results are listed in Table
179
8.4. A value of 1.123 was used for the sign, 1.2 was used for the chord members in the
overhanging truss, and 1.1 was used for all web members and shaft support.
Height coefficients. All height coefficients were calculated based on the Table 3-5
of the Supports Specifications. Exposure condition C (open terrain with scattered obstruc-
tions) was used for the calculation. The height coefficient represents a change in turbu-
lence of wind pressure, becoming less turbulent and more stable as the height above
ground level increases. The coefficient used in the Supports Specifications was defined
using the power law of the wind velocity profile with respect to height above ground
level shown in Eq. 8.6. The curved profile with height defined by Eq. 8.6 was simplified
into a stepped profile by the Supports Specifications and is shown in Figure 8.14 demon-
strating the values used for this structure. The exposure definition only applied to the
180
shaft support as the centroid of the truss and sign was located between 24.6 ft (7.50 m)
and 32.8 ft (10 m) where the coefficient was equal to 1.0. All height coefficients used in
the structural analysis for each area segment is listed in Table 8.5.
α
h
vw = v10 [Eq. 8.6]
h10
where
v w = wind velocity
v10 = wind velocity at a reference height of 32.8 ft (10 m); normalized at 1.0
h = height above ground level
h10 = reference height of 32.8 ft (10 m)
α = terrain constant equal to 0.16 for neutral air above flat open coast.
Effective area calculation. The effective area was defined as the exposed area of
the segment multiplied by the drag coefficient of the members that make up the area
segment, and the height coefficient of the centroid location of the area above ground
Drag Effective
Area Height Co-
Segment Location Label Coefficient, Area, Ae,i
(in2) efficient, Kz
Cd (in2)
Front
Sign Ae,1 26,108 1.123 1.0 29,320
Sign
Front Chord Ae,2-1 2,016 1.2 1.0 2,419
Truss Web Ae,2-2 1,118 1.1 1.0 1,230
Bottom Ae,3-1 4,435 1.1 0.87 4,244
Front
Middle Ae,3-2 2,362 1.1 0.94 2,442
Shaft
Top Ae,3-3 1,195 1.1 1.0 1,315
Bottom Ae,4-1 4,435 1.1 0.87 4,244
East
Middle Ae,4-2 2,362 1.1 0.94 2,442
Shaft
Top Ae,4-3 1,195 1.1 1.0 1,315
Bottom Ae,5-1 4,435 1.1 0.87 4,244
West
Middle Ae,5-2 2,362 1.1 0.94 2,442
Shaft
Top Ae,5-3 1,195 1.1 1.0 1,315
The effective areas and the wind directionality unit vectors were used to develop
acting force vectors for each three second interval. Equations of equilibrium were
formed. Using the position vectors listed in Table 8.3, representing the distance of the
centroid of the effective area to the strain gauge locations, the active moment at the strain
gauge location was calculated. This process is described by the following system of equa-
tions.
Acting force vector. The acting pressure vector was defined as the wind direction-
ality unit vector times the pressure magnitude, P, as shown in Eq. 8.7.
Knowing the pressure is equal to the force divided over an area, the force vector was de-
fined in Eq. 8.8 in terms of the effective area calculated in Table 8.6, and keeping the
F = λ (ΣAe ,i P)
[Eq. 8.8]
= [λ x (ΣAe ,i P )i + λ y (ΣAe ,i P) j + λ z (ΣAe ,i P)k ] lb
Fx = λ x (ΣAe ,i P ) lb
F y = λ y (ΣAe ,i P ) lb
Fz = λ z (ΣAe ,i P ) lb
equilibrium were created from which the acting moment equations were solved. A typical
free body diagram of the wind loading on the structure can be generalized as shown in
Figure 8.15 with all force components shown in their positive sense. The acting moment
vector (resultant in Figure 8.15) was calculated from the equations of equilibrium formed
using the free body diagram. The pressure magnitude, P, was kept variable during the
calculation process. This was done for each three second interval for all usable data
streamlines.
184
the stress and strain values at the strain gauge locations. Section AA and Section BB were
primarily used for this calculation due to the placement of the strain gauges with respect
to in-plane and out-of-plane displacement. The stress element development involved the
Torsion stress,
Of the four loading conditions bulleted, the analysis involved almost exclusively
bending moment and torsion. The bending moment was analyzed as an unsymmetrical
bending. It was broken down into two components, a drag moment (moment about the x-
axis) and lift moment (moment about the y-axis). Since no vertical variability in wind di-
rection was measured using anemometer AN-3, the normal stress component of the com-
bination was not present. Transverse shear stress was present and used in the wind pres-
sure magnitude, P, evaluation with the rosette strain gauges at Section AA.
Material Properties
The material properties needed for development of the stress element are listed in
Table 8.7. A cross section of the analysis location is shown in Figure 8.16.
The stress and strain equations relevant to the combined loading analysis are
listed in Table 8.8. The stress element derived from the structural analysis was developed
The stress element was formed for each strain gauged location. For Section AA
and Section BB, the strain gauges were placed circumferentially on the outer surface of
the shaft. The gauges were spaced at 45° from each other for each section. The strain
gauge locations and labeling used for the analysis are illustrated in Figure 8.17. The uni-
axial gauges were oriented along the longitudinal axis (z-axis) of the post. The middle
gauges of the rosettes were oriented along the same direction. The stress at each lo-
cation using the relevant stress equations listed in Table 8.8 was calculated. The values
were placed on a stress element and added together to determine the normal and shear
stresses at the gauged location (see Figure 8.18). The values were calculated with the
Strain Gauge
The wind pressure magnitude, P, was solved using Eq. 8.5. Strain gauges at Sec-
tion AA involved uni-axial and rosette gauges. Both types of gauges were used for nor-
mal stress values, whereas the rosette gauges were used for shear stress values. The ro-
settes were placed at locations A, C, E, and G. Section BB had only uni-axial gauges and
were used for the normal stress values. An example calculation for the rosette gauge at
Wind Directionality
Experimental
Unit Vector [Theoretical
Wind Pressure
Max. Max. Stress
Section Magnitude, P
Strain Stress Value]*P
i j k (psf)
Range Range (psi)
(µin/in) (psi)
AA 12.351 358.179 73,613.22 0.701
BB -0.682 -0.731 0 13.717 397.793 72,638.24 0.789
Rosette 14.420 158.620 32,695.94 0.699
The example is a snap-shot of a typical three second interval of the diagnosed us-
able data. A summary of the back-calculation procedure to determine the values in Table
1. The wind directionality unit vector was formed using the measured anemometer
compass bearings. The compass bearings were transformed into polar bearings
and the directionality unit vector was developed through trigonometric composi-
tions.
190
ing was performed to determine the stress element at the strain gauged location.
The theoretical value was calculated using the segmented areas and equations of
equilibrium with the wind pressure magnitude, P, kept as a variable. The wind di-
rectionality unit vector (determined experimentally) was used to establish the di-
rection of the wind pressure magnitude to correlate with the measured strain val-
ues.
3. The resulting theoretical normal stress and shear stress of the element from a
AA and Section BB, and torsion and transverse shear for the rosettes, was
summed.
4. The experimentally measured maximum strain range within the three second in-
terval was found and multiplied by the modulus of elasticity for Section AA and
Section BB (normal stress), and the modulus of rigidity for the rosette (shear
stress).
5. The wind pressure magnitude, P, was solved by dividing the experimental stress
The five step process to determine the equivalent static wind pressure was performed for
all three second intervals of the three hours and thirty minutes of usable data. A wind di-
rectionality unit vector was formed for each interval. This resulted in a total of 4,200
The maximum wind pressure magnitude, P, occurring within each three second
interval was calculated for every stain gauge at Section AA and Section BB, including
the rosettes. The calculation was performed in the manner described by the five step
process listed. The wind pressure magnitudes resulting from the five steps were plotted
versus the corresponding three second wind velocity (average wind for the three second
duration) and are shown in Figure 8.19 through Figure 8.21 for Section AA, Section BB,
FIGURE 8.19 Wind velocity vs. wind pressure for Section AA.
192
FIGURE 8.20 Wind velocity vs. wind pressure for Section BB.
FIGURE 8.22 Wind velocity vs. wind pressure for all sections.
A profound curve can be seen from the figures, characteristic to a parabola which
pressure and the velocity squared. The spread observed at higher velocities was due to the
A regression analysis was performed to determine a best fit line of the plotted data
to simulate the parabolic curve. This was done through a transformation regressor lineari-
zation process similar to the process performed for the offsetting procedure. An average
was taken of the data points for all sections and gauges. The independent variable (wind
velocity on the abscissa axis) was squared and was plotted versus its corresponding wind
The linearization of the transformation proved the parabolic nature of the data. A
best fit line was then constructed as a linear predictor to acquire a linear equation of the
transformed data. Reversing the transformation, and using the best fit line equation, a
parabolic trendline of the data was produced and is shown in Figure 8.24. The slope of
the developed trendline was equal to 0.0025, with a y-intercept equal to 0.402.
195
Infinite-Life Approach
The infinite-life approach was used in the same fashion as the Supports Specifica-
tions. The wind velocity that was exceeded only 0.01% of the time was calculated using a
−πv 2
2
PE (v) = e 4v [Eq. 8.9]
where
PE (v) = probability
v = wind velocity
v = mean wind velocity.
196
The wind velocity that exceeds the mean wind velocity 0.01% of time was calculated us-
ing Eq. 8.9, and referred to as the fatigue wind velocity (referred to as the limit-state wind
velocity in the Supports Specifications. Using the same annual mean wind velocity equal
to 11 mph (5 m/s) as the Supports Specifications, the fatigue wind velocity calculated
with Eq. 8.9 was found to be 38.0 mph (17 m/s). As shown in Figure 8.25, the equivalent
static wind pressure magnitude corresponding to a 38 mph (17 m/s) wind velocity was
It was concluded from the results of this project that the fatigue load due to natural wind
gust determined from experimental analysis was equal to 4.01 psf (192 Pa).
197
CHAPTER 9
Chapter Overview
A description of the theoretical program for evaluation of the fatigue load due to
natural wind is presented in this chapter. The fatigue load due to natural wind was calcu-
Calculation of the Fatigue Load due to Natural Wind Gust. The theoretical program was
developed as a hybrid of experimental and theoretical data. The process was developed
similar to the Supports Specifications natural wind fatigue provisions with adaptations to
account for the variety of sign support structures in design, each with different configura-
The objective was to provide a unified design method for fatigue loading on
highway overhead sign support structures due to natural wind gust. Overhead sign sup-
port structures are highly flexible with low damping properties, which makes them sus-
ceptible to vibratory induced fatigue loading. The magnitude of this load is dependent on
the dynamic behavior and characteristics of the structure itself. The intention was to de-
velop a relationship between fatigue loading and the dynamic response in terms of ran-
cantilever- and bridge-type overhead sign support structures, as well as variable message
sign (VMS) structures (Figure 9.1). For analysis purposes, these structures were ap-
proximated as a SDOF system because the modes of vibration are significantly separated
such that the vibration in response to randomly applied wind loading is controlled pre-
dominately by a single modal shape. For that reason, the modal shapes were estimated as
vibrating independently from each other in single global directions (10, 20).
Cantilever-Type
Overhead
Sign Support
The fatigue provisions for natural wind in the Supports Specifications are ade-
quate within certain limitations. They were developed based on four particular catego-
rized structural types. The structural response of one overhead signal support structure,
one cantilever-type overhead sign support structure, and two luminaire support structures
to natural wind excitation were analyzed. The transmitted stresses of each structure were
averaged, and the fatigue provisions were developed from the averaged results (52). The
Supports Specifications are therefore only applicable to structures of the type mentioned
with the same dynamic properties, most importantly the natural frequency and critical
damping percentage (a.k.a. damping ratio). Differences in these properties, such as the
case with bridge-type sign support structures and VMS support structures, are not ac-
This study provides a detailed approach to handle cases that have different dy-
namic properties than those used to develop the Supports Specifications. Cantilever-type
sign support structures can have different configurations and made with various materials
and cross sectional shapes. These parameters will dramatically affect the magnitude of
the fatigue load. Thus, a method that incorporates the specific dynamic properties of the
structure is needed for estimating the fatigue load. Bridge-type sign support structures
and VMS support structures, which are not covered by the Supports Specifications, can
also be addressed with the proposed design method. The primary differences in these
and mass, which are directly related to the natural frequency and damping of the struc-
ture.
200
Methodology
the Davenport natural wind velocity power density spectrum (PDS) curve for simulating
natural wind excitation in conjunction with the infinite-life approach for fatigue loading
(13, 52). The method presented employed the same PDS excitation and infinite-life ap-
proach, as well as a PDS excitation developed from the experimental wind velocity data
collected. However, the response of the structure due to this excitation was evaluated dif-
ferently in this research than used in the Supports Specifications. The analysis of the re-
sponse was based on principles of random vibration in utilization of the vibration re-
sponse spectrum (VRS) (39, 40-49, 56). This was done in order to account for the
uniqueness and individuality of sign structures regarding their dynamic properties, which
has significant affect on stresses generated from natural wind fatigue loading. The ap-
proach is equivalent to determining an equivalent static wind load, which produces the
same response on the structure as a randomly applied dynamic wind load. The fatigue
design equivalent static wind load is chosen from the VRS in terms of the natural fre-
quency and damping ratio of the structure. The proposed method can be used as a tool to
determine the appropriate design fatigue load for the particular structure in question dur-
Structural Excitation
A comparison was made between the structural excitation developed for the Sup-
ports Specifications and the data collected with this project. Descriptions on the devel-
Supports Specifications
The method used to development the natural wind provisions was based primarily
on spectral analysis in collaboration with the infinite-life approach to fatigue design. The
estimation of the structural excitation due to natural wind involved predicting the natural
wind environment for which the structure was to be exposed. This was done using a spec-
tral analysis based on A.G. Davenport’s wind velocity power density spectrum shown in
4κV102 x 2
Sv ( f ) = 4
[Eq. 9.1]
f (1 + x )
2 3
where
S v ( f ) = wind velocity power spectral density at any height
f = frequency
V10 = mean wind velocity at a stadard height of 10 meters above ground level
κ = surface drag coefficient (Table 9.1)
1200 f f
x= 2
with 2 in cycles per meter.
V10 V10
Type of Surface κ α
Open unobstructed country (e.g., prairie-type grass-
0.005 0.15
land, arctic tundra, desert)
Country broken by low clustered obstructions such as
0.015 - 0.020 0.27 – 0.31
trees and houses (below 10 m high)
Heavily built-up urban centers with tall buildings 0.050 0.43
202
Davenport developed the wind velocity PDS curve from 70 experimental wind
velocity data collections from various locations around the world. His intention was to
develop a model which simulated the turbulence and gustiness of wind velocity. He de-
veloped Eq. 9.1 from the 70 experimental data collections. The equation is a function of
wind velocity frequency with respect to a mean wind velocity at a specified height. His
formulation is shown in Figure 9.2 for frequencies ranging from 0 to 10 Hz, an open ter-
rain (see Table 9.1), and an annual mean wind velocity of 11 mph (5 m/s):
FIGURE 9.2 Wind velocity PDS for annual mean wind velocity.
Once the behavior of the wind velocity environment was estimated, the PDS was
transformed into a wind force PDS by using principles related to fluid mechanics. The
drag force induced onto a structure due to natural wind is proportional to wind velocity
1
FD = ρCd AV 2 [Eq. 9.2]
2
where
FD = drag force
kg
ρ = density of air = 1.22
m3
C d = drag coefficient
A = area of exposed surface
V = wind velocity at any height.
By utilizing the proportionality between drag pressure and wind velocity, a wind pressure
PDS was developed from Davenport’s wind velocity PDS shown in Eq. 9.3. The plotted
equation is shown in Figure 9.3 for an annual mean wind velocity of 11 mph (5 m/s) and
S F ( f ) = ρ 2 C d2 A 2V 2 S v ( f ) [Eq. 9.3]
204
The PDS curve accounts for the gustiness and turbulence of wind velocity over a
spectrum of frequencies, and was based on an averaged wind velocity taken at a specified
height above ground level. Since most support structure are at or around 32 ft (10 m) in
height, the PDS curve is well suited for these types of structures. Yet, the PDS can be
used at any particular height by using the power law profile shown in Eq. 9.4 for ap-
where
V = wind velocity at height z
α = surface coefficien t in Table 8.1
z = height above ground.
205
In this case, where the objective was concentrated on formulating a design code
for fatigue wind, the wind velocity variable in the pressure PDS equation was taken at the
standard height of 32 ft (10 m) above ground level, and kept uniform across the wind ex-
posed façade of the structure. The purpose of which was to provide a simplified design
equation for commercial use. Some conservative formulation exists as the wind velocity
Once the natural wind environment was estimated, the next step was to apply the
PDS to the infinite-life approach. Since the force spectrum was based primarily on the
annual mean wind velocity, the wind velocity that was exceeded 0.01% of the time was
found and referred to as the limit-state wind velocity, or fatigue wind. The force spectrum
Wind velocity is random in nature, but it can be predicted though statistical rela-
tionships. It has been found through many experiments that the magnitude of the wind
velocity vector will follow a Rayleigh distribution (54). By using the Rayleigh distribu-
tion, the wind velocity that has a probability of exceedence equal to 0.01% was found
through the relationship in Eq. 9.5 as a function of the annual mean wind velocity.
−πv 2
PE (v) = e 4V
2
[Eq. 9.5]
where
PE (v) = probability
v = wind velocity corresponding to the probability
V = mean wind velocity.
206
An analysis was conducted to determine which annual mean wind velocity to use in
Eq. 9.5 to determine the limit-state wind velocity (wind velocity with a 0.01% ex-
ceedence probability). The annual mean wind velocities of 59 major U.S. cities were ana-
lyzed. It was found that an annual mean wind velocity of 11 mph (5 m/s) was exceeded in
only 19% of the U.S. cities analyzed and was therefore chosen. By plugging in 11 mph in
Eq. 9.5, and solving for the wind velocity corresponding to the 0.01% probability, a limit-
state wind velocity was found to be equal to 38 mph (17 m/s). The force spectrum was
then formed using the limit-state wind velocity (see Figure 9.4) and was used as the natu-
ral wind velocity prediction model for structural excitation for the natural wind fatigue
A PDS was developed using the wind velocity data collected with this research.
Only the data considered usable (wind data directed on the front face of the structure)
was used for the analysis, and was the same excitation and response data used to deter-
mine the fatigue load from the experimentally collected data presented in Chapter 8. A
velocity PDS of each data collection event was developed and a best fit trendline was
created of the curves. The best fit line was then transformed into a wind pressure PDS
using the same procedure used for the Supports Specifications excitation. The trendline
was compared to the Supports Specifications and both were used for the structural re-
Each wind velocity time history of the usable collected data was developed into a
PDS. The time domain was transformed into the frequency domain through the Fourier
transform (Eq. 9.6). The PDS was calculated by taking the Fourier transform and multi-
plying it by its conjugate, dividing by its period, and then taking the limit as the period
∫ x(t )e
− j 2πft
X(f ) = dt [Eq. 9.6]
−∞
208
where
X ( f ) = Foureir transform
x(t ) = time history
f = frequency
t = time
j = imaginary
lim1
S( f ) = X ( f )X *( f ) [Eq. 9.7]
T →∞T
where
S ( f ) = Power density spectrum
X ( f ) = Fourier transform
f = frequency
T = Period
X * ( f ) = Complex conjugate of the Fourier transform
Power spectral density curves are particularly useful for this application. They are ideal
for random vibration analysis due to the inherent statistical properties of the time history
that can be extracted in relation to the vibratory nature of the structure. The area under
the PDS curve is equal to the mean square value. The square root of the mean square
value is equal to the root-mean-square (RMS). For cases where the mean is equal to zero,
the RMS is equal to the standard deviation (18, 39, 42). The developed PDS curves using
Eq. 9.6 and Eq. 9.7 are shown in Figure 9.5, along with the average PDS.
209
A best fit line was developed that approximated the PDS average curve in Figure
9.5. A mathematical expression was needed that followed the curvature of the average
PDS curve. On the log-to-log plot in Figure 9.5, the average PDS curve was viewed as bi-
linear. The magnitude of the ordinate (y-axis) and abscissa (x-axis) data points were
transformed into a log-to-log format so that the ordinate and abscissa axes would be a
linear relationship on a standard linear plot without altering the curvature of the plot. This
was done by taking the logarithm with a base 10 of the ordinate and abscissa values. The
resulting plot is shown in Figure 9.6. The transformed plot (Figure 9.6) was subdivided
into two sections that were observed to be linear. A linear trendline was fit to each sec-
tion. The equation of the trendline was extracted and used as an approximation of the bi-
210
linear curvature of the PDS plot. The logarithmic ordinate and abscissa axes were then
transformed back to its original values using the logarithmic identity in Eq. 9.8.
The next step was plotting the trendline equations onto logarithmic axes. Equiva-
lent power equations, that represented the linear equations on a logarithmic axis, were
developed of the two sectioned plot. The resulting plot is shown in Figure 9.7. The power
equations in Figure 9.7 as a result describe the true curvature of the plot on both a loga-
rithmic and linear axis, and can be used as a mathematical expression approximating the
average wind velocity PDS. The approximation was capped at the lower frequencies by a
211
constant distribution (see Figure 9.8) and used for the remainder of the analysis. As
shown in the figure, the theoretical simulation of the experimental wind velocity PDS av-
erage closely followed the curvature and was viewed as a close approximation of the ex-
perimental curve.
FIGURE 9.7 Best fit line approximating the average wind velocity PDS.
212
FIGURE 9.8 Theoretical plot of the experimental average wind velocity PDS.
The theoretical wind velocity PDS approximation was converted into a wind pres-
sure PDS in the same fashion as the Supports Specifications force excitation PDS. Equa-
tion 9.3 was used for the wind velocity-to-pressure conversion, resulting in the plot
It is important to point out that the plot in Figure 9.9 represented the average wind
pressure PDS and not the 0.01% exceedence probability wind pressure PDS used in the
infinite-life approach to fatigue design. It simulated the wind gustiness and turbulence of
an average wind velocity equal to 12.96 mph (5.79 m/s). Based on the experimental na-
ture of the plot, and the process in its development, the 0.01% exceedence probability
was not found exclusively using the average excitation plot in Figure 9.9, but rather
213
found by creating wind pressure PDS curves for each natural wind event measured ex-
perimentally and approximated much like the process used to find the experimental fa-
tigue load. The distribution of each wind pressure PDS was found and plotted according
to the corresponding wind velocity. The distribution was simulated and the value at the
0.01% exceedence probability was found. A more detailed description of this process is
A comparison was made between the wind pressure excitation used in the Sup-
ports Specifications and the one developed in this project. The average wind velocity was
used for the comparison. The code used an average wind velocity equal to 11 mph (5
214
m/s) whereas the experimental excitation was based on an average wind velocity equal to
12.96 mph (5.79 m/s). A plot of the comparison is shown in Figure 9.10.
The comparison plot shows the experimental PDS was greater than the excitation
used in the Supports Specifications. This should be the case since the average wind veloc-
ity of the experimental PDS was greater. However, the fatigue load was based on the
wind velocity with a 0.01% exceedence probability and not the average. Based on the re-
sults obtained later in the analysis of this chapter, the PDS equation for the Supports
Specifications excitation generated a PDS for the fatigue wind that was greater than the
experimental PDS, which explained the slightly larger fatigue load due to natural wind.
The experimental PDS was based on an approximation, found through a fitted curve of
215
experimental results that estimated what the PDS would be for a 0.01% exceedence prob-
Structural Response
Stresses are induced onto a structure when it vibrates, and because of the high
flexibility and low damping properties of sign support structures, the vibratory stresses
are enhanced as a direct result of the structure’s dynamic characteristics. In the structural
response analysis, the excitation on the structure from the randomly applied load, and the
subsequent random vibration response of the structure, were approximated through basic
principles of structural dynamics in utilization of the VRS. The VRS in this context was a
tool for determining the load transmitted onto the structure from the vibratory response
created by the natural wind excitation described in the Structural Excitation section of
this chapter.
The first step in the formulation of the VRS was determining the response of the
structure from the wind pressure PDS excitation. The vibration behavior of sign support
structures when excited were approximated as a SDOF system in each of their major
global directions. This was because the majority of vibration was in the first two modes
shapes of the first two modes were in distinct individual directions, which behaved inde-
pendently from each other (mode 1 horizontal, and mode 2 vertical). In addition, the am-
plitude of the first modal frequency, the horizontal vibratory motion of the truss which
created torsion about the shaft support, was much larger than the amplitude of the second
216
modal frequency, and therefore controlled much of the vibration in response to the natu-
ral wind loading condition. This is not to imply that the second modal shape did not play
a part, or had no influence in the response behavior, but the significance of this mode as
well as the other higher modes of vibration were small in comparison to the first. In view
The implication of developing the VRS was to account for the different types of
sign support structures, each with differing configurations, sizes and shapes, and materi-
als that have a direct influence on the dynamic characteristics of the structure. In the VRS
development, the dynamic characteristics of the structure such as the natural frequency
and damping ratio were kept variable in the calculation. In line with the directive needed
in the VRS development, the response of a SDOF system due to the wind pressure PDS
excitation developed in the Structural Excitation section of this chapter was calculated
f
2
1 + 2ξ
fn
Uˆ PPSD ( f ) = YˆPPSD ( f )
2 2 2
f f [Eq. 9.9]
1 − + 2ξ
f n fn
217
where
Uˆ PPSD ( f ) = response pressure PSD as a function of frequency, Pa 2 /Hz (psf 2 /Hz)
Yˆ ( f ) = excitation pressure PSD as a function of frequency, Pa 2 /Hz (psf 2 /Hz)
PPSD
f = frequency, Hz
ξ = damping ratio
f n = natural frequency of the structure, Hz
Equation 9.9 was used to calculate the response pressure PDS for individual natu-
ral frequencies and damping ratios of the particular structure of interest. It was formed
using fundamental Structural Dynamics related to a SDOF system. The divisor term in
Eq. 9.9 is referred to as the transfer function. It was multiplied by the excitation PDS
equaling the response PDS. The units were in pressure. It represented the pressure load in
the form of a PDS curve that was transmitted onto the structure resulting from the vibra-
tory response created by the wind pressure PDS excitation. Dynamic amplification of the
structure due to the applied loading was therefore inherent in Eq. 9.9, accounting for the
dynamic properties that govern the response behavior. A plot of the excitation and re-
sponse calculated using Eq. 9.9 is shown in Figure 9.11, using a the natural frequency of
the first mode equal to 1.61 Hz and the modal damping of the first mode equal to 1.82%
The spike shown in the response curve of Figure 9.11 was located at the natural
frequency used in the Eq. 9.9. Other frequency and damping values can be used in the
equation to account for other support structures with differing dynamic properties. The
results would have different spikes than the one depicted in Figure 9.11 of a 1.61 Hz
natural frequency. In addition, the damping value influences the width and height of the
spike. The spike would subsequently shrink for higher damping values, and increase for
The area under the response curve calculated using Eq. 9.9 is equal to the mean
square value. The square root of the area curve is referred to as the root-mean-square
(RMS). The RMS was a value used in this study for determining peak amplitudes within
amplitudes about a mean value in response to the turbulence and gustiness of the natural
wind excitation. The RMS of the response pressure PDS for individual natural frequen-
cies and damping ratios was determined by integrating the response pressure PDS over
the frequency domain, and then taking the square root, as shown by Eq. 9.10 (39, 412).
f
2
1 + 2ξ
∞
fn
U PRMS ( f n , ξ ) = ∫ YˆPPSD ( f ) df [Eq. 9.10]
0
2 2
f
2
f
1 − + 2ξ
f n fn
where
U PRMS ( f n , ξ ) = overall response pressure RMS as a function of the natural frequency
and damping ratio, Pa (psf)
For the response curve in Figure 9.11, using a natural frequency equal to 1.61 Hz and
critical damping percentage equal to 1.82%, the square root of the area under the curve
(RMS) calculated using Eq. 9.10, was equal to 0.284 psf (13.6 Pa). The calculated RMS
value is commonly referred to as the overall level of the structural response PDS. The
overall level was calculated for SDOF dynamic systems with individual natural frequen-
cies and damping percentages and plotted to form the VRS that was used to calculate the
fatigue load.
220
The VRS is a plot of the RMS of the response (Eq. 10) versus the range of natural
frequencies used in the equation. The resulting VRS provides the transmitted pressure
RMS onto the structure due to the dynamic amplification of the structure in relation to its
specific dynamic characteristics. Take for example Figure 9.12. The same base input was
applied to n number of structures. However, each structure had different mass and stiff-
ness properties, which gave it different natural frequencies of vibration, fn, and therefore
each individual structural response to the base excitation was different. Meaning, there
were different response PDS curves for each natural frequency (calculated from Eq.
9.10). For each response curve, the RMS was determined, and then plotted along with its
corresponding natural frequency to form the pressure VRS shown in Figure 9.13.
ü1 ü2 ün
m1 m2 mn
k1 c1 k2 c2 kn cn ÿ
The VRS in Figure 9.13 was from the excitation developed from the experimen-
tally collected measurements, and used a critical damping percentage of 1.82% represent-
ing the tested structure. The plot shows a decrease in RMS wind pressure as the natural
frequency increase in large part due to the lessoned proximity of the resonant frequency
of the structure to the peak amplitude of the excitation PDS. The proper use of the VRS is
to select the natural frequency of the structure of the first modal shape with vibratory mo-
tion in the direction of the structure. For the tested structure, the correct frequency to use
was the first modal frequency equal to 1.61 Hz. The corresponding ordinate value was
Natural Frequency
Sign support structures have a variety of modal shapes, but because of the large
separation between modes with vibration in the direction of the loading, they vibrate pre-
dominately independent from each other in distinct single directions. The appropriate
natural frequency to use in the VRS must correspond to the modal shape that has motion
in the direction of the natural wind loading. The most critical loading scenario for natural
wind is directed normal to the plain of the sign (in the direction of traffic), most com-
monly referred to as the horizontal modal shape (see Figure 9.14). For sign support struc-
tures, the horizontal modal shape is generally around 1 to 3 Hz, which typically corre-
sponds to the first modal shape for cantilever-type structures, and the second modal shape
Direction of
Wind Loading Base Plate
Finite element software (i.e., SAP2000) can be used to estimate the appropriate
modal shapes and their associated natural frequencies to use with the VRS curves. If FEA
software is not available, fundamental structural dynamics of a SDOF system (Eq. 9.11)
can be used for estimating these values (10, 17, 33, 56, 57, 61, 63).
1 K
fn = [Eq. 9.11]
2π M
where
f n = natural frequency, Hz
K = generalize d stiffness, N/m (lb/ft)
M = generalize d mass, kg (slug)
dal shapes for overhead sign support structures using Eq. 9.11 can be found in the work
performed by Creamer et al. (10), which also contains useful calculation examples. It is
understood that Eq. 9.11 would be predominately used by engineers in commercial appli-
For example purposes, the first two modal frequencies of the structure were calcu-
lated using Eq. 9.11. The values were compared to the natural frequency values deter-
mined from the experimental modal analysis in Chapter 7: Experimental Modal Analysis.
The values are similar, proving that Eq. 9.11 can be used as an estimation of these
values if FEA software is not available. In most cases, the modal natural frequencies with
horizontal and vertical vibratory directions will typically be around 2.0 to 3.0 Hz for can-
tilever- and bridge-type sign support structures, with the frequency of the horizontal
Damping also had an effect on the response of the structure as it vibrates. Take for
example the VRS plot in Figure 9.15, with the same excitation but a damping value de-
creased to 0.5%. With natural frequency equal to 1.61 Hz, the RMS wind pressure be-
came equal to 0.412 psf (19.7 Pa), a significant increase. This was true for any structural
an increased value), the spike in the response PDS increased in height and width, creating
additional area under the curve as seen in Figure 9.16 as compared to Figure 9.11, and
FIGURE 9.15 RMS wind pressure VRS with damping equal to 0.5%.
and the structure will slowly stop vibrating after a period of time. The length of time de-
pends on the damping value, and the vibration frequency. During this time, stress is in-
duced onto the structure. For structures with low damping ratios, longer periods of vibra-
tion at high amplitudes will result, during which time potentially damaging stress is
transmitted onto the structure. If the stress is higher than the endurance limit of the con-
nection detail, the fatigue life of the structure is decreased with each cycle of vibration.
With higher natural frequencies of vibration, the transmitted load is decreased; but more
importantly, damping becomes less important. Damping becomes more of a factor as the
tively high flexibility and subsequent low natural frequencies (1 to 3 Hz). What’s more,
their damping ratios are mostly below 2.0%. A low damping will allow the structure to
vibrate longer at high amplitudes, and thus produce more stress that could potentially
cause fatigue damage. Damping ratios can vary depending on the structural material, and
therefore actual values can be obtained from experimental data of comparable structures
if available.
The design equation for fatigue due to natural wind loading must be representa-
tive of the amplitude peak-to-peak stress range that is induced on the structure during
common everyday vibration. The majority of the structural vibration in response to natu-
ral wind excitation is controlled predominantly by a single modal frequency, the basis for
227
which validates approximating the response as a SDOF system. As a result, the response
wind pressure PDS resembled a narrow-banded spectrum concentrated about the natural
frequency. The RMS value of the spectrum embodies the variance of the vibration ampli-
tudes from a mean, which symbolizes the response of the structure due to the turbulence
and gustiness of natural wind. During a single transient event from a common everyday
natural wind gust, the peak-to-peak ranges of the response will initiate at its largest value
and then decay at rate indicative of the damping ratio. The design fatigue equation must
exemplify the largest range created on the outset of this event. In view of this, given that
the response is predominately controlled by a single modal frequency, and the gustiness
and turbulence of natural wind is typified by the RMS value, the averaged initial peak-to-
the initial peak amplitude was approximated as the square root of two times the RMS
value, and then doubled to form the initial peak-to-peak range, as shown by Eq. 9.12.
where
Pressure Range = peak - to - peak pressure range, Pa (psf)
U PRMS ( f n , ξ ) = overall response pressure RMS as a function of the natural frequency
and damping ratio calculated using Equation 3, Pa (psf)
The plot in Figure 9.13 represents the wind pressure in terms of the RMS values
only, and does not represent peak-to-peak ranges. To predict the peak-to-peak pressure
range for design considerations, the VRS chart of RMS values (Figure 9.13) was factored
228
by 2.8 (Eq. 9.12). The resulting plot is shown in Figure 9.17. The tested structure with a
natural frequency equal to 1.61 Hz, and a critical damping percentage equal to 1.82%
would require a wind pressure representing the peak-to-peak amplitude equal to 0.800 psf
(38.3 Pa).
Infinite-Life Approach
The infinite-life approach to fatigue design requires the wind pressure from an
average wind velocity that has a probability of exceedence equal to 0.01%. This value
was found by using an approximation through a best fit line of individual VRS of differ-
ing average wind velocities. The VRS in Figure 9.17 was of an average wind velocity
equal to 12.96 mph (5.79 m/s). This average represents all usable collected data. To find
229
the 0.01% VRS, a range of VRS curves were developed of differing average wind veloci-
ties. This was done by dividing the usable data into 10 minute sections, each with ranging
average wind velocities, and forming the peak-to-peak amplitude VRS (as in Figure 9.17)
for each 10 minute segment. A plot was made of the wind pressure extracted from the
VRS for the 1.61 Hz natural frequency and 1.82% damping tested structure versus the
corresponding average wind velocity of the segment used to develop the VRS. The result
of the plot is shown in Figure 9.18 along with the trendline developed from experimen-
tally collected strain data detailed in Chapter 8: Experimental Calculation of the Fatigue
The peak-to-peak VRS points in Figure 9.18 follow the experimental trendline
developed from the strain gauges in Chapter 8. They mostly lie underneath experimental
trendline which was to be expected because the VRS points are based on a SDOF,
whereas the experimental trendline was based on actual structural response measurements
of a true multiple degree-of-freedom system. The proximity of the VRS points show that
approximating the structure as a SDOF is justified. The VRS point in black represents the
average VRS described in detail and shown in Figure 9.17. The variation in ordinate val-
ues of the VRS points was due to the accuracy of the wind velocity PDS regression
analysis.
The scatter closely followed the experimental trendline for the experimentally
made response measurements. The wind velocity with a 0.01% exceedence probability
was found to equal to 38 mph (17 m/s) in Chapter 8, which was equivalent to a 4.01 psf
(192 Pa) fatigue wind pressure load. In applying the same trendline procedure in Chapter
8, by transforming the domain by squaring the wind velocity to “linearize” the data,
forming a linear trendline to the transformed data, and extracting the trendline equation
(Figure 9.19), the resulting wind pressure at a 38 mph (17 m/s) wind velocity was found
The theoretical VMS calculated value of 3.80 psf (182 Pa) as compared to the ex-
perimental calculated value of 4.01 psf (192 Pa) was very close with a 5.24 percent dif-
ference (calculated from the experimental to the theoretical). The proximity reveals the
justification of approximating the response of the structure as a SDOF. The 3.80 psf (182
Pa) VMS fatigue load using the 0.01% exceedence probability is a factor of 4.75 times
the peak-to-peak RMS value from Figure 9.17. To account for other support structures
with differing dynamic properties, the peak-to-peak RMS amplitude VMS in Figure 9.17
was factored by 4.75. The result is the fatigue load VMS (Figure 9.21) that is in compli-
ance with the 0.01% exceedence probability and the infinite life approach.
The fatigue load is extracted from the VRS plot in Figure 9.21 and used for de-
sign. The VRS plot in Figure 9.21 represents fatigue loads from an annual fatigue wind
velocity equal to 38 mph (17 m/s). By knowing the natural frequency of the structure, the
fatigue load can be determined that is based on the dynamic characteristics of the struc-
ture. The plot is of a 1.82% critical damping percentage. Vibration response spectrums
for other damping values ranging from 0.1% to 2% were created and are shown in Figure
9.22. The trend shows damping to become more influential as the natural frequency de-
creases. The fatigue values vary greatly and were therefore important to obtain a true
section was based on the wind excitation developed from this study. The same process
was done for the current Support Specifications wind excitation as well. The resulting
fatigue load VRS is shown in Figure 9.23. The plot was made for a 2.0% critical damping
percentage as was used in the development of the fatigue load. The plot also shows varia-
The code does not specify fatigue loading alteration with the dynamic characteristics of
the structure. This process was specific and original to the researchers of this project.
10000
2% Critical Damping
1000
Pressure Range (Pa)
100
1 m/s
10 2 m/s
3 m/s
4 m/s
5 m/s
6 m/s
1 7 m/s
8 m/s
9 m/s
10 m/s
0.1
0.001 0.01 0.1 1 10
Natural Frequency (Hz)
CHAPTER 10
Chapter Overview
A detailed description on the processes and results of determining the fatigue load
due to truck induced gusts is presented in this chapter. The measurements made experi-
mentally from the truck gusts tests were used in the analysis. The method for data reduc-
tion is presented, along with an analysis of the results. The data was broken down into
two components: 1) structural excitation, and 2) structural response. The fatigue load was
The fatigue load was determined from the experimentally measured data collected
during the truck gust tests. Only data collected during low wind days, generally less than
8 mph (3.58 m/s), was used. The strain gauge instrumentation on the shaft support Sec-
tion AA and Section BB was used for the fatigue analysis. This was because of the
placement of the gauges and their orientation along the geometry of the structure in rela-
tion to the truck induced load. The measured strain ranges were used to back-calculate
the required wind pressure to produce the measured range. The maximum load calculated
A total of nine truck-induced wind gust tests were performed within two days
time. Each test run was at a specific truck speed. The allocation of test runs with respect
The individual truck runs specific to truck speed, and the order at which they were per-
formed is provided in Table 10.1. The table lists the truck speed as read on the speedome-
ter in the truck cab, and by JAMAR Trax Flex HS recorders using 0.365 in (9.27 mm)
diameter tubes spaced at 36 in (914 mm) that were placed across the road at the sign loca-
tion.
It was discovered from the truck data that the natural wind velocity on Day 1 of
testing was too strong, and transient truck events could not be isolated from the data.
Ironically, this day was one of the highest natural wind speed days, and therefore the col-
lected data was used in the fatigue load due to natural wind analysis. Day 2 proved to be
successful due to the low wind velocity which made transient truck events easy to isolate.
In view of this, only data collected on Day 2 was used for the fatigue load due to truck
gusts analysis. This amounted to four truck runs deemed usable with one run at a truck
speed of 60 mph (26.8 m/s), one run at a truck speed of 65 mph (29.1 m/s), and two runs
Truck Specimen
A semi-trailer was used for the experiment. A picture of the truck and driver is
shown in Figure 10.1. The height of the semi-trailer was 13.5 ft (4.11 m) width equal to
Structural Excitation
The major projects analyzing fatigue loading due to truck-induced gusts have
been gathered and reported in NCHRP Report No. 469 (14) which comprised the basis of
the current Supports Specifications fatigue provisions for truck-induced gusts. Impor-
tantly, the conclusions on truck gust fatigue in Report 469 were based on experimental
evaluation of cantilever-type VMS support structures, and not sign support structures.
Many specific observations related to truck gusts on support structures but were inde-
pendent to the type of structure were made during in Report 469. These observations
Report 469 concluded that the full strength of the truck gusts pressure of 18.8 psf
(900 Pa) was produced on the underneath exposed area of the structure at 19.7 ft (6 m) or
less above the roadway. The pressure decreased linearly to zero at a height of 32.8 ft (10
m) above the roadway. The pressure load applied in the horizontal direction was found to
be insignificant compared to the fatigue load due to natural wind, and was excluded. The
underneath exposed area to truck gusts was determined to cover 12 ft (3.7 m) in length or
the length of the traffic lane, whichever was greater. The provisions indicated that the
truck gust pressure should be applied over the exposed length located where the force
would create the worst-case scenario (largest moment arm). It was specified for most
The objective of the truck gust tests performed in this project was to evaluate the
accuracy of the truck gust provisions in the Supports Specifications as applied to cantile-
ver-type sign support structures. Many of the observations specific to the truck gusts
239
structural excitation reported in Report 469 were used in this project. These observations
Fatigue pressure was applied onto the underneath horizontal exposed area of the
structure,
The exposed area was directly above the traffic lane used by the truck, and
The gust was assumed maximum at the bottom of the sign and decreases linearly
The above excitation specifics were used in the evaluation of the truck gusts on cantile-
ver-type sign supports structures of this project. These observations had no influenced, or
were influenced in their discovery, by the structural response of the structure tested when
the observations were made, and were therefore independent on the type of support struc-
ture. In view of this, these observations were assumed as factual for this project, and were
used in the evaluation of the cantilever sign structure to truck-induced wind gusts.
Anemometers AN-1 and AN-2 were placed on the structure above the traveling
lane to evaluate the speed and direction of the truck gusts. The purpose of the measure-
ments was to reinforce the accuracy of the gust pressure back-calculated from the strain
gauge measurements. AN-1 was oriented to measure the wind speed and direction on the
horizontal plane (xy plane), whereas AN-2 was oriented to measure the wind speed and
direction on the vertical plane (zy plane). Their locations are shown in Figure 10.2.
Transient truck gust events were significant and obvious in measurements made
240
from the strain gauges and accelerometers. The fatigue truck gust pressure was deter-
mined from these measurements. Analysis of the truss anemometers AN-1 and AN-2 in-
dicated no significant transient event. This may be due to the short pulse of the truck
transient and the sampling frequency of the anemometers. The instruments were set at 60
samples per second in line with the strain gauges and accelerometers. The anemometer
had a maximum sampling frequency inherent in the instrument of 4 Hz, and therefore the
remaining samples were repeated until the sampling period was renewed. The limitation
was not discovered by the researchers until after the anemometers were bought and
placed on the structure. This is viewed as the main contributor to the failure of the in-
Structural Response
The measurements made from the strain gauges comprised the majority of the
structural response. Strain gauges located on the shaft support were used primarily be-
cause of their location in line with the symmetry of the structure, and their orientation to
the truck excitation. The evaluation involved exclusively the response to a vertically ap-
plied pressure onto the exposed horizontal area on the underneath section of the structure.
The exposed are was equal to 12 ft (3.7 m) located directly above the traffic lane used by
the truck specimen. The analysis involved a back-calculation from the strain gauge meas-
urements to determine a lift moment from which the truck gust pressure was determined.
The strain gauge measurements encompassed a strain range created form the truck tran-
AN-4
AN-3
y
AN-2
x AN-1
Traffic
Direction
Stress Range
range determined from the strain gauge measurements. Only the working truck tests col-
lected on a low wind day were used for the analysis. This amounted to four tests of one at
60 mph and 65 mph (29.1 m/s), and two at 70 mph (31.3 m/s). The strain gauges at loca-
tion C and G (see Figure 10.3) for Section AA and Section BB were the sole gauges used
for the analysis. This was based on their orientation to the loading event, and their place-
Strain gauges at the other locations, particularly locations A and E were not used
because of the unknown direction of truck gusts (because of the anemometer failure) and
the unknown gradient of the pressure with height. These two factors dramatically influ-
enced the drag moment (moment about the x axis, see Figure 10.2) that was needed in
calculating the unsymmetrical bending needed for the other gauges not oriented perpen-
dicularly along the x-axis. The back-calculation with strain gauges located at C and G did
not require the determination of the drag moment along the x axis and allowed for a more
straightforward calculation.
Strain Gauge
Transient Events
The evaluation of the transient events used for the truck gust analysis involved
determining the maximum peak-to-peak strain range. A typical transient is shown in Fig-
243
ure 10.4, showing in-plane vibration (mode 2, rocking) to the vertical upward force cre-
ated by the truck gusts as measured from strain gauge SGR-AA-7 located on Section AA
at location C. The length of the transient event was inconsequential for determining the
fatigue load. Due to the nature of the load, its repetition, and randomness, the fatigue load
must represent the maximum peak-to-peak stress range of the structural response. As
long as the endurance limit of the detail is greater than the maximum measured peak-to-
peak range, then significant fatigue deterioration due to truck gust is avoided during the
lifespan of the structure. The peak-to-peak strain range for each truck event is listed in
Table 10.2.
The fatigue load due truck gusts was back-calculated from the measured response ranges
The equivalent static wind load that would produce the same maximum peak-to-
peak stress range determined experimentally was back-calculated. Two measures formed
the basis of the process. First, the stress range was determined from the experimentally
measured strain ranges as described in the Structural Response section of this chapter.
Second, a theoretical structural analysis was performed on the structure to determine the
stresses at the gauged locations for comparison with the experimental values. A stress
element of normal values was developed at the location of the gauge to be used in the
245
comparison. Importantly, the stress values were developed with the wind pressure magni-
tude, P, kept as a variable. The two measures, experimental stress and theoretical stress,
were set equal to each other and the wind pressure magnitude, P, was solved (see Eq.
10.1). The instrumented sections along the shaft support (Section AA and Section BB) at
locations C and G (Figure 10.3) were used exclusively for this evaluation because of their
orientation to in-plane displacements with respect to the vertically applied truck gust.
The structural analysis involved determined exposed areas and using them to form
the moment vectors needed in determining the load. The wind pressure magnitude, P,
was kept as a variable throughout the process, and solved for using Eq. 10.1 at the end of
the procedure. Point loads with direction at each exposed area was developed with drag
coefficients provided by the Supports Specifications for each member exposed to wind
and used in the analysis. The idea was to simulate as accurately as possible the wind
loading conditions. Equations of equilibrium were developed and the acting moment vec-
tor (resultant) was solved. The moment vectors were used to develop stress elements at
each strain gauged location. Equation 10.1 was used to solve for the wind pressure mag-
nitude.
246
The wind pressure was applied onto a horizontal area located on the underneath
portion of the truss overhang, in compliance with the research collected in NCHRP Re-
port 469 (14) and used in the Supports Specifications. The exposed underneath area
equaled 12 ft (3.7 m) in length. It was located directly above the traffic lane used by the
truck, as shown in Figure 10.5. The truck ran directly underneath the truss anemometers
AN-1 and AN-2, which were placed along the centerline of the second panel. The hori-
zontal area therefore included all exposed horizontal surfaces of the truss 6 ft (1.83 m) on
each side of the anemometer location. This included the two chords, one horizontal strut,
and three horizontal diagonal struts. The sign was not included in the calculation. A
breakdown of the exposed area used in the analysis is shown in Figure 10.6, with values
Exposed
Area
Anemometers
AN-1 and AN-2
12 ft
12 ft (3 Panels @ 4 ft)
Chord
3 ft
I I
Horizontal Diagonal Exposed
Strut Strut Area
Chord A1 1,008
Bottom Horizontal
A2 47.34
Truss Strut
381 0 309 381 0 305
Horizontal
Diagonal A3 298.8
Strut
The effective areas of the exposed areas listed in Table 10.3 were calculated. The
effective area represented the exposed area multiplied by the drag coefficient of the
Drag coefficients. All drag coefficients were calculated using Table 3-6 of the
Supports Specifications. Fifty year design life was used in the calculation with a velocity
conversion factor equal to 1.00 (no conversion needed). The worst case scenario was ap-
plied by maximizing the wind speed and member diameter. A maximum wind speed of
70 mph (31.293 m/s) was used, assuming the wind gust equals the speed of the truck, and
the largest diameter member equal to 3.5 in (88.9 mm). The results of the drag coefficient
Chord A1 1.10
Effective area calculation. The effective area was defined as the exposed area of
the segment multiplied by the drag coefficient of the members that make up the area
Drag Effective
Area
Segment Location Label Coefficient, Area, Ae,i
(in2)
Cd (in2)
Chord A1 1,008 1.10 1,108.8
Horizontal
A2 47.34 1.10 52.074
Bottom Strut
Truss Horizontal
Diagonal A3 298.8 1.10 328.68
Strut
Total Effective Area 1489.554
The effective areas were used to develop acting moment vectors representing a vertically
applied truck gust, and equations of equilibrium were formed. Using the position vectors
listed in Table 10.3 of the distance of the centroid of the effective area to the strain gauge
locations, the active moment at the strain gauge location was calculated. This process is
Acting force vector. The acting pressure vector was defined as a wind directional-
ity unit vector times the pressure magnitude, P, as shown in Eq. 7.7.
Knowing the pressure is equal to the force divided over an area, the force vector was de-
fined in Eq. 10.3 in terms of the effective area calculated in Table 10.5, and keeping the
F = λ (ΣAe,i P)
[Eq. 10.3]
= [λx (ΣAe,i P)i + λy (ΣAe,i P) j + λz (ΣAe,i P)k ] lb
The components of the force vector in Eq. 10.3 were as follows, with the only z-
component (vertically applied force) applicable to this analysis as the x and y components
Fx = λ x (ΣAe ,i P ) lb
F y = λ y (ΣAe ,i P ) lb
Fz = λ z (Σ Ae ,i P ) lb
equilibrium were created from which the acting moment equations were solved. A typical
free body diagram of the wind loading on the structure can be generalized as shown in
Figure 10.7 with all force components shown in their positive sense. The acting moment
vector (resultant in Figure 10.7) was calculated from the equations of equilibrium formed
using the free body diagram. The pressure magnitude, P, was kept variable during the
calculation process.
252
Reaction
MR
FR
Fz
x
y
the stress and strain values at the strain gauge locations. Locations C and G at Section
AA and Section BB was used for this calculation due to the placement of the strain
gauges with respect to in-plane displacement. The stress element development involved
the combination of the bending stress caused by the upward force and its moment arm,
down into two components, a drag moment (moment about the x-axis) and lift moment
(moment about the y-axis). Only the lift moment was needed because strain gauges used
in the calculation and their location and orientation on the structure. Other gauges were
not used because of the uncertainty of the horizontal pressure (applied on the vertical face
of the sign) distribution. The vertical pressure however was calculated with a high level
of accuracy.
Material Properties
The material properties needed for development of the stress element are listed in
Table 10.6. A cross section of the analysis location is shown in Figure 10.8.
FIGURE 10.8 Cross section of strain gauge location Section AA and Section BB.
The stress and strain equations relevant to the combined loading analysis are
listed in Table 10.7. The stress element derived from the theoretical structural analysis
Stress Element
A stress element was developed using the equations in Table 10.7. The stress
elements were formed with the wind pressure, P, as a variable. The values were placed on
a stress element and added together to determine the normal stresses at the gauged loca-
FIGURE
FIGURE 10.9 Typical
10.9 Typical stress element.
stress element.
The stress elements, formed with the wind pressure, P, kept as a variable, were
compared to the experimentally measured normal stress ranges depicted in Table 10.2.
The wind pressure, P, was calculated using Eq. 10.1. The results of the calculation are
Strain Gauge
Strain Gauge Truck Test Truck Speed Wind Pressure
Location on
Section Run (mph) (psf)
Section
C 7.917
Truck 6 60
G 6.979
C 6.652
Truck 7 65
G 5.783
Section AA
C 5.442
Truck 8 70
G 4.892
C 3.871
Truck 9 70
G 3.519
C 8.348
Truck 6 60
G 7.286
C 7.130
Truck 7 65
G 6.127
Section BB
C 5.819
Truck 8 70
G 4.921
C 4.119
Truck 9 70
G 3.494
The results in Table 10.8 represent the equivalent static wind load to create an
equivalent stress range on the structure as a result of dynamic amplification and vibration
of the structure in response to the truck-induced gust impulse. The maximum pressure
was found to be equal to 8.348 psf (400 Pa) resulting from a truck speed of 60 mph (26.8
m/s). The maximum pressure was considered as the fatigue load, as it was representative
of the maximum stress range measured experimentally from trucks traveling underneath
the structure.
The pressure values calculated from the applicable strain gauges indicated that
trucks traveling at 70 mph (31.3 m/s) created slightly less response than trucks traveling
at 60 mph (26.8 m/s). This occurrence was only measured with the post gauges, as the 70
mph (31.3 m/s) truck speed controlled on the anchor bolts and chords, and therefore more
257
tests may have been needed. Alternatively, this may be due to the natural frequency of
the structure as compare to the impulse frequency of the truck-induced gusts. Trucks
traveling slower may produce a truck gust impulse with a frequency closer to the resonant
frequency of the structure than trucks traveling at faster speeds. In view of this, the re-
sults would indicate a relationship between the back-calculated wind pressure and the dy-
Gust.
258
CHAPTER 11
Chapter Overview
design fatigue load due to truck-induced wind gusts on highway overhead sign support
structures. Truck-induced wind gust is a loading generated on the structure from wind
gust created when trucks travel underneath them at highway speeds. Semi-trailer com-
mercial trucks were the focus of the research. It was considered a fatigue load due to the
large number of trucks using the highway system, and the significant wind load generated
on the overhead sign structure with each passing truck. The theoretical study primarily
involved analysis of sign support structures in relation to mechanical vibration and utili-
zation of the shock response spectrum (SRS). The fatigue load due to truck gust was ex-
tracted from the SRS as an equivalent static wind load and used for design.
Research Significance
The fatigue load is determined primarily from the structural dynamic characteris-
tics of the particular structure in question. Operational overhead sign support structures
are highly flexible with low damping properties, which make them susceptible to vibra-
tory induced fatigue loading caused by wind gust impulses generated from passing
trucks. The magnitude of this load is dependent on the dynamic behavior and characteris-
259
tics of the structure. The analysis becomes especially crucial when the frequency of the
impulse matches the natural frequency of vibration of the structure. Frequencies of vibra-
tion and the frequencies of the truck gust impulses are similar, typically around 1 to 3 Hz,
and therefore resonance issues are a major cause of fatigue damage. A method that incor-
porates the response of the structure based on its dynamic properties is needed for esti-
mating the fatigue load. The intention of the study was to develop a relationship between
fatigue loading and the structural dynamic response in terms of mechanical vibration as a
Cantilever-
Type
Overhead
The fatigue provisions for truck gusts within the Supports Specifications were se-
lected from the National Cooperative Highway Research Program (NCHRP) Report 412
(1, 16, 52), and later modified due to the research detailed in the NCHRP Report 469 (14,
16). It was stated in Report 412 that the design fatigue load in regards to truck gusts was
message sign (VMS). The Supports Specifications are only applicable to highway signs
with the same structural excitation and dynamic response characteristics reported. Differ-
ences in these characteristics, such as the case with other cantilever- and bridge-type sign
support structures, cannot be accounted for nor are addressed in the current Supports
Specifications.
This chapter details a universal approach to fatigue design to handle cases that
have different structural dynamic properties than those used to develop the current Sup-
ports Specifications. Cantilever-type sign support structures can have different configura-
tions and are made with various materials and cross sectional shapes. These parameters
will dramatically affect the behavior of the structure and subsequent magnitude of the
fatigue load. A method for estimating the fatigue load that incorporates the specific dy-
namic properties of the structure can as a result account for the variety of sign structures
in design. For that reason, the proposed method was considered universal. It is applicable
to any type of sign support structure of varying configuration, material, strength, mass,
and stiffness whereby its response approximated as a SDOF system. Bridge-type sign
support structures, which are not covered by the Supports Specifications, can also be ad-
dressed with the proposed design method. The primary differences in these structures in
Since all of these dynamic features are directly related to the natural frequency of the
structure, the method described in this paper can be used to determine the design fatigue
load due to truck gusts in specific reference to the dynamic characteristics of the particu-
Methodology
The framework of the developed method can be broken down into two major
components: 1) the truck gust excitation of the structure, and 2) the dynamic response of
the structure. Research data collected from the literature was used to simulate the truck
gust excitation for this chapter. A typical truck gust pressure loading was developed in
the form of a time history streamline. Different design truck speeds were employed in the
time history simulation, resulting in truck gust impulses occurring at different frequen-
cies, which in turn would excite different vibratory responses in the structure.
The response of the structure and consequent fatigue load due to the truck gust
excitation was evaluated differently than the methods described in the literature. The pro-
posed method was entirely new and unique with respect to overhead sign support struc-
tures. The structural response was theoretically calculated based on principles of me-
chanical vibration, and the fatigue load was determined through utilization of the SRS
(39, 40-49, 56). The load is extracted from the SRS in terms of the natural frequency of
the structure to account for the variety of sign support structures in design, and the esti-
mated speed of the truck to account for different loading impulses. The extracted value is
in the form of an equivalent static wind load which produces the same response onto the
structure as the dynamic impulse. Dynamic amplification was included within the fatigue
262
load chosen from the SRS, since the SRS was developed based on the frequency of vibra-
tion of the structure in response to the loading event. The proposed method is used as a
tool to easily determine the appropriate design fatigue load due to truck gust for the par-
Structural Excitation
An accurate depiction of the truck gust wind excitation without influence from the
type of structure and its dynamic behavior was needed in the development of the theoreti-
cal procedure. Experimental measurements made with this project were not applicable as
the strategically placed anemometers (AN-1 and AN-2) did not pick truck wind gusts.
This was believed to be due to the sampling frequency of the anemometers. The research
performed by Cook et al, in their collaboration with the Florida Department of Transpor-
tation (12) was considered by the researchers as the best candidate for replacement.
Cook’s research involved experimentally determining the truck gust pressure load
through direct measurement of wind pressure from passing trucks. The same process was
conducted by the researchers, but to measure wind velocity. With Cook’ research, the
truck wind pressure was evaluated at varying heights above ground level using pressure
transducers placed on a rigid road bridge spanning over the highway. Time history wind
pressure impulses from 23 truck gust loading events were recorded. The findings in
Cook’s research were closely similar to other experimental research efforts on the subject
performed by different investigators in other locations (10, 14, 17), especially with
Cook’s research indicated a biaxial fatigue loading event. The results indicated a
horizontal and vertical component of the wind gust pressure impulses. The horizontal
component was projected onto the facade within the vertical plane that was oriented per-
pendicular to the ground. The vertical component was projected onto the facade within
the horizontal plane that was oriented parallel to the ground (12). Only the vertical pres-
sure impulses was used with this study, as the natural wind fatigue loading criteria con-
trols wind loading in the horizontal direction. All strain values measured experimentally
were significantly higher (nearly twice as much) with natural wind gusts than measured
Some small liberties were taken with this study in order to simplify the pressure
impulses recorded by Cook so that they could be functional for numerical applications.
The time history impulses were transformed into a linear composition by enveloping the
samples by the triangular single cycle loading shown in Figure 11.2. This was done simi-
lar to the research performed by Ginal (37), except impulses from trucks traveling at
other speeds were also employed. The loading simulation in Figure 11.2 represented a
linear increase in wind pressure as the truck approaches, followed by an equivalent (in
magnitude) suction as the truck passes and separates from the structure. This is indicative
a wind pressure directed upward onto the underside facade of the structure as the truck
from the structure (12, 37). The impulse in Figure 11.2 represents the measured values
The wind pressure impulse resulting from a truck traveling at 70 mph (31.3 m/s)
was used as an impulse Control (dashed line in Figure 11.2). The Control reflects an en-
velopment procedure of the recorded time history samples of trucks that traveled at this
from truck gusts defined at a 90% confidence level that the actual maximum absolute
value of pressure and impulse duration will be equal to or below the Control at least 95%
of the time (12). The envelopment resulted in a linear increase and subsequent decrease
to the maximum impulse pressure equal to 1.50 psf (71.8 Pa) as shown in Figure 11.2.
These values were used for both the approach pressure and the separation pressure of the
Control for symmetrical simplification of the simulated impulse. The duration of the
Trucks traveling at other speeds generated wind gust impulses with different fre-
quencies and durations than the Control (see Figure 11.2), and would as a consequence
excite different responses on the structure. Impulses generated at other truck speeds were
identified and employed in this method to account for a variation in excitation frequen-
cies. The impulses were normalized based on the Control impulse at 70 mph (31.3 m/s)
using the mechanics of fluids relationships between wind pressure and the square of wind
constant acceleration.
Structural Response
The structural response from truck gust loads described in the previous section
(see Figure 11.2) was approximated through basic principles of structural dynamics. For
analysis purposes, the dynamic behavior of highway overhead sign support structures was
approximated by a SDOF system. This was because the modes of vibration measured in
Chapter 7: Experimental Modal Analysis were significantly separated such that the vibra-
shape in independent directions. For that reason, the modal shapes were estimated as vi-
brating independently from each other in single global directions (10 ,20, 39). In this
study, the structural response of a SDOF system due to the truck gust load was calculated
Pi = 2e −ξωn ∆t cos(ω d ∆t ) Pi −1
− (e − 2ξωn ∆t ) Pi − 2
+ 2ξω n ∆tpi [Eq. 11.1]
ω
+ ω n ∆te −ξωn ∆t n (1 − 2ξ 2 ) sin(ω d ∆t ) − 2ξ cos(ω d ∆t ) pi −1
ω d
where
Pi = response pressure as a function of natural frequency and critical damping percentage, psf (Pa)
pi = truck induced wind gust pressure impulse excitation, psf (Pa)
ωn = undamped natural frequency of the structure, rad/sec
ωd = ωn 1 − ξ 2 = damped natural frequency of the structure, rad/sec
ξ = percent critical damping
t = time, sec
Eq. 11.1 calculates the response pressure time history for individual natural frequencies
and critical damping percentages of the particular structure of interest. It can be seen as
an equivalent pressure load that is transmitted onto the structure from the vibration re-
sponse created by the truck-induced excitation. The equation embodied full dynamic am-
An example of the vibratory response time history of a typical sign support struc-
ture calculated using Eq. 11.1 is shown in Figure 11.3. The 70 mph (31.3 m/s) Control
impulse was used as the excitation for this example. The natural frequency of the struc-
ture was chosen with a modal shape in the direction of the vertical loading, and was equal
to 1.64 Hz found in Chapter 7: Experimental Modal Analysis. This modal shape is best
described as a rocking vibratory motion in the direction perpendicular to the ground (par-
allel to the vertical component loading). The plot shows response time histories of the
structure for an undamped case, and for a critical damping percentage equal to 1.0%.
267
FIGURE 11.3 Structural response time history due to the Control impulse.
Sign support structures typically have critical damping percentages around 1.0%
or less. Even for 1.0% damping however, the difference in magnitude of the maximum
pressure range observed between the two response cases (Figure 11.3) was very small,
and significantly lessens as the damping percentage decreases. This revealed that the
maximum peak response was not sensitive to structural damping, and therefore the un-
damped case was used in the SRS development, which adheres to other common SRS
development case studies (39, 47). For such instances, the damped natural frequency pa-
rameter, ωd, in Eq. 11.1 became equal to the undamped natural frequency, ωn.
The SRS in this context was a tool for determining the maximum load transmitted
onto the structure from the vibration response created by truck gusts. The SRS was a plot
268
of maximum peak values determined from the response time histories calculated using
Eq. 11.1 for a range of SDOF systems (see Figure 11.4) with varying undamped natural
frequencies under a common loading (39, 47). The SRS plot for the vertically directed
truck gust impulse is shown in Figure 11.5. The SRS shown in Figure 11.5 is a plot of the
maximum response pressure ranges versus natural frequencies. The natural frequency
domain extends from 0 to 10 Hz, and individual SRS curves for truck speeds ranging
from 50 mph to 100 mph (22.4 m/s to 44.7 m/s) are plotted based on the impulses pre-
ü1 ü2 ün
m1 m2 mn
k1 c1 k2 c2 kn cn ÿ
The pressure range, and subsequent stress range, was of interest when dealing
with fatigue analysis. It is important to clarify the curves in Figure 11.5 represent the ini-
tial range of the positive to negative peaks (see Figure 11.3) for each impulse. The stress
generated from the vibration caused by this pressure range will be induced onto the struc-
ture with each passing truck, with little influence from the damping properties of the
structure. The damping will allow the structure to slowly stop vibrating over a period of
time by gradually decreasing the magnitude of this range, as shown in Figure 11.3 for the
1% damping case. However, if the initial pressure range occurring at the onslaught of the
gust impulse generates a stress that is greater than the endurance limit of the material and
connection detail, then the fatigue life of the structure is greatly decreased with each
270
passing truck. And therefore, the design fatigue load must be representative of the initial
range of values plotted in Figure 11.5 in terms of the natural frequency of vibration of the
structure which, as indicated in the figure, has a significant effect on the magnitude of the
initial range.
The fatigue load is extracted from the SRS as the ordinate value corresponding to
the natural frequency of vibration with a modal shape in the direction of the excitation.
Sign support structures have a variety of modal shapes, but because of the large separa-
tion between modes with vibration in the direction of the loading, they vibrate predomi-
nately independent from each other in distinct single directions. The appropriate natural
frequency to use in the SRS must correspond to the earliest modal shape that has motion
in the direction of the truck-induced wind loading. In this case, the truck-induced load
was directed vertically onto the underneath facade of the structure. The appropriate mo-
dal shape would be a vertical vibratory motion (perpendicular to the direction of traffic)
generally associated with the second modal shape of typical cantilever-type sign struc-
tures (see Figure 11.6), and the third modal shape of typical bridge-type sign support
structures.
271
Base Plate
Direction of
Wind Loading
Finite element software (i.e., SAP2000) can be used to estimate the appropriate
modal shapes and their associated natural frequencies to use with the VRS curves. If FEA
software is not available, fundamental structural dynamics of a SDOF system (Eq. 11.2)
can be used for estimating these values (10, 17, 33, 56, 57, 61, 63).
1 K
fn = [Eq. 11.2]
2π M
where
f n = natural frequency, Hz
K = generalize d stiffness, N/m (lb/ft)
M = generalize d mass, kg (slug)
272
dal shapes for overhead sign support structures using Eq. 11.2 can be found in the work
performed by Creamer et al. (10), which also contains useful calculation examples.
For example purposes, the first two modal frequencies of the structure were calcu-
lated using Eq. 11.2. The values were compared to the natural frequency values deter-
mined from the experimental modal analysis in Chapter 7 Experimental Modal Analysis.
The results were similar, proving that Eq. 11.2 can be used as an estimation of
these values if FEA software is not available. In most cases, the modal natural frequen-
cies with horizontal and vertical vibratory directions will typically be around 2.0 to 3.0
Hz for cantilever- and bridge-type sign support structures, with the frequency of the hori-
The SRS plots in Figure 11.5 have varying peaks which are dependent on the du-
ration and frequency of the impulse excitation. It would be advisable to design structures
that have natural frequencies outside of the frequencies plotted at these peaks. For the 70
mph (31.3 m/s) Control, the maximum possible structural response from this impulse
would occur onto a structure that has a natural frequency with a vertical vibratory modal
273
shape around 1.60 Hz which is extremely close the structure resonant frequency of the
tested structure at 1.64 Hz. The transmitted pressure range from the response of the im-
pulse for this particular example structure was equal to 8.33 psf (399 Pa), taken as the
in the Supports Specifications (1, 14). Accommodation for height above ground level is
also provided in the 2002 revisions. The height provisions conservatively match the find-
ings in Cook’s report (1, 12, 14). Whereas Cook’s findings indicated a 10% reduction in
wind pressure per foot increase starting from 17 ft (5.2 m) above ground level (12), the
Supports Specifications allow for a linear reduction in wind pressure starting from 19.7 ft
(6 m) above ground level to zero at 32.8 ft (10 m) (1, 14). For that reason, the 2002 revi-
sions on pressure reduction due to height above ground level were considered applicable
to the method presented in this chapter. The design traveling speed of the truck to use in
the SRS is generally taken as the speed limit. A Control speed of 70 mph (31.3 m/s) was
11.5 for design choice, but also to illustrate the loading patterns in the SRS.
274
CHAPTER 12
Chapter Overview
sign support structure. The FEA software package SAP2000 v. 10 was used for the study
(11). The analysis was performed to address three primary areas of interests:
2. Check the accuracy of the natural wind and truck-induced fatigue loading back-
calculations, and
3. Compare the developed equivalent static wind load to its dynamic load counter-
The first process was performed to test the validity of the FEA model. The last two proc-
esses were performed for both the natural wind loading and truck-induced wind loading.
The FEA served as a model for determining stresses in locations that were not experi-
mentally instrumented. Validation of the FEA models concentrated on those areas that
were instrumented and tested experimentally. Changes in the models, such as materials,
sizes and shapes, sign area, lumped mass, damping, etc. were also evaluated that would
otherwise not be performed experimentally due to the time and cost of such projects.
275
Model Development
sign structure. The model was created from the shop drawings of the structure provided
Geometry
The basic geometry of the model is shown in Figure 12.1. Body constraints were
used to connect the sign to the W-shape sections, and the W-shape sections to the truss.
Body constraints cause the constrained joints to move together as one three-dimensional
object (11). A close up of the connection is shown in Figure 12.2. Stiffeners were used
for the sign to reduce its flexibility needed in the modal analysis so that the global modal
behavior of the structure as a whole could be evaluated. The truss-to-post connection was
an extremely rigid bolted plate connection. It was modeled using steel plates connected to
the chord ends at three points spaced symmetrically along the weld. Body constraints
were used to connect the plates to the post support to simulate the rigid action so that all
joints cannot displace relative to each other (11). A close up of this connection is shown
in Figure 12.3.
276
The base connection of the post support to the foundation used eight anchor bolts
to attach the post base plate to the 70 in (1,778 mm) diameter by 6 ft (1.83 m) long con-
crete pile foundation. In the FEA model development of this connection, the support post
was first connected directly to the pile. The pile was restrained in all translation and rota-
tional degrees-of-freedom at the end as the only base support for the structure. However,
this connection was too rigid when comparing the FEA modal analysis to the experimen-
tal modal analysis detailed in Chapter 7: Experimental Modal Analysis. The connection
was then modified by modeling the anchor bolts and base plate. Body constraint connec-
tions were used to connect the post support to the base plate. The anchor bolts were con-
nected to the base plate at nodal locations. The anchor bolts were modeled with a 2 in
(50.8 mm) clearance from the bottom of the base plate to the top of the concrete surface.
Rigid links were used to connect the anchor bolts to the pile foundation. This modeled
connection gave it the flexibility that better simulated the existing structural behavior as
Element Type
Beam and shell elements were used in the model. Beam elements were used for
the truss and support post, as well as for the anchor bolts, pile foundation, sign W-shape
connections, and sign stiffeners. Shell elements were used for the aluminum sign, W-
Material Definition
The primary material of the structure was steel. The sign was made of aluminum
and the foundation was made of concrete. The post and truss section was made of API-
5L-X52 steel pipe. The plates were made of structural steel ASTM A572 Gr. 50. The an-
chor bolts were made of AASHTO M314-90 Gr. 55 (essentially the same as ASTM
F1554 Gr. 55). The W-shape and T-shape sections used for the sign-to-truss connection
were made of A572 Gr. 50 steel. The sign was made of aluminum wrought alloy desig-
nated as 6061-T6. The concrete pile was made of conventional concrete with #9 size re-
bar Gr. 60. All of the major required material properties of the designations given are
the model, and to determine the natural frequencies of the structure and their correspond-
ing modal shapes. The natural frequencies were compared to the values determined ex-
perimentally for accuracy. The first five modal shapes for the model are shown in Figure
12.5, and the natural frequencies, natural periods, and a description of the modal shapes
x
y
Mode 4 Mode 5
FIGURE 12.5 Modal shapes of the first five modes from FEA.
Period Frequency
Mode Mode Shape
(sec) (Hz)
1 Shaft Torsion 0.63577 1.57
2 Vertical Rocking 0.61645 1.62
3 Horizontal 0.25484 3.92
4 Longitudinal 0.19125 5.23
5 Truss Torsion 0.13408 7.46
There was a distinct difference in modal shapes and frequencies for each mode,
which demonstrates an accurate dynamic behavioral simulation. The x-axis was defined
as oriented along the longitudinal length of the structure, the y-axis oriented opposite the
direction of traffic, and the z-axis oriented vertically along the axial length of the post
support. The first modal shape was indicative of a horizontal action of the sign and truss
282
creating a torsion behavior about the z-axis. The second modal shape was indicative of a
vertical rocking of the truss in the plane perpendicular to the ground about the y-axis. The
third modal shape was indicative of a horizontal rocking of the sign, truss, and shaft in
the direction of traffic about x-axis. The forth modal shape was indicative of a longitudi-
nal rocking of the sign about the y-axis. And the fifth modal shape was indicative of a
Only four of the five of the indentified modes of vibration were induced in the
field. The fifth mode, torsion of the truss overhang, was not activated when analyzing the
experimental data from wind induced excitations. The primary vibration was in the first
two modal shapes, with the first mode predominately controlling the vibration with larger
amplitudes. The modal shapes in each direction were significantly separated from each
other such that the structure vibrated primarily in single directions, independently from
each other.
A comparison between the modal shapes determined from the experimental and
FEA analyses is provided in Table 12.3. A close comparison between theoretical and ex-
perimental results is shown. The modal shapes were identical in shape, however the am-
plitudes of vibration between the FEA and experimental results was not compared. The
first two modal shapes are the most important with respect to fatigue loading because the
majority of the vibration of the structure was within these modal shapes with the largest
amplitudes of vibration, and therefore inducing the largest fatigue stress on the structure.
283
determining the equivalent static wind load for natural wind and truck-induced gusts. The
process was performed by inputting into the program the equivalent static pressure load
determined for natural wind and truck-induced gusts, and then extracting from the output
the strain at the locations of the strain gauges used to calculate the load. The output strain
from the FEA was compared to the experimentally obtained values, and the accuracy of
the back-calculation was evaluated. The analysis was seen as a structural analysis “dou-
ble check” to test the accuracy of the back-calculation performed, as well as the accuracy
of the FEA model. A successful comparison would indicate that the equivalent static
wind load determined for natural wind and truck-induced gusts was correct, and the FEA
model was accurate in depicting structural behavior, in addition to the modal analysis
comparison.
284
The equivalent static wind load determined from Chapter 8: Experimental Calcu-
lation of the Fatigue Load due to Natural Wind Gust was inputted into the SAP 2000
program. The FEA output was extracted at locations that were used to calculate the
equivalent static wind load from the experimental data. The output was compared to the
strain determined experimentally at these locations, and the accuracy of the back-
as follows:
1. Collect the equivalent static wind load developed from Chapter 8: Experimental
2. Collect the strain range value that corresponds to the equivalent static wind load
gathered in Step 1 from the experimental data at the particular strain gauge loca-
3. Input the equivalent static wind load into the FEA model as the static structural
excitation.
4. Run a linear-static solution of the FEA model with the equivalent static wind load
excitation.
5. Extract the strain value from the FEA solution at the strain gauge location deter-
mined in Step 2.
6. Compare the FEA strain value determined in Step 5 with the experimental value
gathered in Step 2.
285
The equivalent static wind load determined from Chapter 8: Experimental Calcu-
lation of the Fatigue Load due to Natural Wind Gust was found to be equal to 4.01 psf
(192 Pa). The load was determined through a regression analysis to predict the fatigue
wind load for a 38 mph (17 m/s) wind velocity in compliance with the infinite-life ap-
proach to fatigue design for support structures. This wind velocity however, did not occur
during the experimental testing, and no experimental stress was available to complete the
comparison. A three second wind velocity and direction equal to approximately 14 mph
(6.26 m/s) was arbitrarily chosen in its place, which is slightly larger the average three
second wind velocity of the total collection. By using Eq. 12.1, developed from the re-
gression in Chapter 8 for determining the equivalent static wind pressure at the fatigue
wind, the equivalent static wind pressure for a 14 mph (6.26 m/s) was calculated to be
where
y = dependent variable equal to the equivalent static wind pressure, psf (Pa)
x = independent variable equal to the wind velocity, mph (m/s)
The average maximum three second strain range corresponding to the three sec-
ond wind velocity equal to 14 mph (6.26 m/s) was determined to be equal to 7.90 (10-6).
Strain gauge SGR-AA-3 at Section AA located 12 in (305 mm) above the base plate weld
286
was used because of its location along the symmetry of the structure, and its orientation
An equivalent static wind pressure load equal to 0.892 psf (42.7 Pa) was used as
input into the SAP2000 FEA program. The load was applied to the sign, post support, and
truss including the exposed chords, vertical struts and diagonal struts of the web. The
pressure load for each member was multiplied by the drag coefficient of the particular
member it was applied. The same drag coefficients used to develop the load and Eq. 12.1
The resulting equivalent static wind load was applied horizontally as a uniformly
distributed load to each member. It was applied as a uniformly distributed pressure to the
sign, and was applied as a uniformly distributed linear load (force per length) to the
chords, vertical struts, diagonal struts, and post support. Only the members exposed on
the vertical plane as seen from the elevation view of the structure applied.
model of the support structure. The location on the model that corresponded to the strain
gauge used for the comparison (SGR-AA-3) was on the front face (exposed face) of the
support post at 12 in (305 mm) above the base plate. The post was modeled as a beam
element in the FEA model, and therefore the moment at this location was extracted from
the solution. The stress value was calculated using Eq. 12.2, which accounted for the
287
normal stress at the section when the dead weight of the structure is excluded, and the
force was applied purely horizontally. The normal strain was calculated using Eq.12.3
assuming the stress remains within the elastic limit of the steel material. The results of the
Mc
σ= [Eq. 12.2]
I
where
σ = bending moment stress, psi (Pa)
M = bending moment along the x - axis extracted from the FEA, lb - ft (N - m)
c = distnce from the neutral axis to the outer most fiber located at SGR - AA - 3, in (mm)
I = moment of inertia about the x - axis at Section AA, in 4 (mm 4 )
σ
ε= [Eq. 12.3]
E
where
ε = normal strain, in/in, (mm/mm)
σ = normal stress, psi (Pa)
E = modulus of elasticity of steel, psi (Pa)
288
Extracted from
Calculated Values from FEA Bending Moment
FEA
σ ε
Bending Moment c I E
Eq. 12.2 Eq. 12.3
(lb-ft) (in) (in2) (psi)
(psi) (in/in)
A comparison was made with the FEA strain value calculated in Table 12.4 with
the experimentally measured strain range at the specific strain gauge location used for the
analysis. The results of the comparison are provided in Table 12.5. The comparison re-
sulted in a 0.759% difference between the two values. The closeness of the two values
indicate the back-calculation to determine an equivalent static wind load that would pro-
duced the strain range measured experimentally was accurate. The results also further
prove the accuracy of the FEA model with the measured structural behavior of the sup-
port structure.
The same comparison procedure for natural wind gust was used for truck-induced
wind gust. The equivalent static wind load determined from Chapter 10: Experimental
Calculation of the Fatigue Load due to Truck-induced Wind Gust was inputted into the
SAP 2000 program. The FEA output extracted at locations that were used to calculate the
equivalent static wind load from the experimental data. The output was compared to the
strain determined experimentally at these locations, and the accuracy of the back-
as follows:
1. Collect the equivalent static wind load developed from Chapter 10: Experimental
2. Collect the strain range value that corresponds to the equivalent static wind load
gathered in Step 1 from the experimental data at the particular strain gauge loca-
3. Input the equivalent static wind load into the FEA model as the static structural
excitation.
4. Run a linear-static solution of the FEA model with the equivalent static wind load
excitation.
5. Extract the strain value from the FEA solution at the strain gauge location deter-
mined in Step 2.
6. Compare the FEA strain value determined in Step 5 with the experimental value
gathered in Step 2.
290
The equivalent static wind load determined from Chapter 10: Experimental Cal-
culation of the Fatigue Load due to Truck-induced Wind Gust was found to be equal to
8.348 psf (400 Pa). The load was applied to the end 12 ft (3.66 m) of the overhang truss.
The pressure was applied to the horizontal area on the underneath portion of the truss.
The experimental strain range used to back-calculate the 8.348 psf (400 Pa) was
equal to 9.245(10-6). The range was measured on the post support at Section AA located
12 in (305 mm) above the base plate. Strain gauge SGR-AA-7 was used because of its
location along the symmetry of the structure, and its orientation to the in-plane displace-
An equivalent static wind pressure load equal to 8.348 psf (400 Pa) was used as
input into the SAP2000 FEA program. The load was applied to the exposed horizontal
area on the underneath façade of the truss overhang along the end 12 ft (3.66 m) section.
The load was applied vertically upward (away from ground) as a uniformly distributed
linear load (force per length). Only the truss section including the chords, horizontal strut,
and horizontal diagonal struts along the end 12 ft (3.66 m) section of the truss were
loaded. The pressure load for each member was multiplied by the drag coefficient of the
particular member it was applied. The drag coefficients used in Chapter 10 were applied.
291
model of the support structure. The location on the model that corresponded to the strain
gauge used for the comparison (SGR-AA-7) was on the side face of the support post at 12
in (305 mm) above the base plate, measuring the in-plane strain of the post due to the ver-
tically applied truck gust. The post was modeled as a beam element in the FEA model,
and therefore only the moment at this location was extracted from the solution. The stress
value was calculated using Eq. 12.4. The equation accounts for the normal stress at the
section with the dead weight of the structure excluded, and the force applied purely verti-
cally. The normal strain was calculated using Eq.12.5 assuming the stress remains within
the elastic limit of the steel material. The results of the calculations are shown in Table
12.6.
Mc P
σ= + [Eq. 12.4]
I A
where
σ = bending moment stress, psi (Pa)
M = bending moment along the x - axis extracted from the FEA, lb - ft (N - m)
c = distnce from the neutral axis to the outer most fiber located at SGR - AA - 3, in (mm)
I = moment of inertia about the x - axis at Section AA, in 4 (mm 4 )
P = axial force, lb (N)
A = cross sectional area, in 2 (mm 2 )
σ
ε= [Eq. 12.5]
E
292
where
ε = normal strain, in/in, (mm/mm)
σ = normal stress, psi (Pa)
E = modulus of elasticity of steel, psi (Pa)
A comparison was made with the FEA strain value calculated in Table 12.6 with
the experimentally measured strain range at the specific strain gauge location used for the
analysis. The results of the comparison are provided in Table 12.7. The comparison re-
sulted in a 2.53% difference between the two values. The closeness of the two values in-
dicate the back-calculation to determine an equivalent static wind load that would pro-
duced the strain range measured experimentally was accurate. The results also further
prove the accuracy of the FEA model with the measured structural behavior of the sup-
port structure.
293
An evaluation of the theoretical methods developed for this project was per-
formed using FEA. The VRS for natural wind gusts and SRS for truck-induced wind
gusts were developed using principles related to Structural Dynamics. A SDOF was used
in their formulation. The spectrums (VRS and SRS) provide an equivalent static wind
load based on the dynamic characteristic of the structure that would produce an equiva-
To test the accuracy of the spectrums, dynamic loading was inputted into the FEA
SAP2000 program and critical response values were extracted from the solution. Using
the same loading conditions and dynamic characteristics of the structure, the equivalent
static wind load was extracted from the VRS for natural wind gust, and SRS for truck-
induced wind gust, and was applied as input into the FEA program. A linear-static solu-
tion was solved, and the same response values extracted from the previously run dynamic
solution was performed for the linear-static solution and compared. The comparison
helped to determine the accuracy of the VRS and SRS methods developed with this pro-
ject. In addition, structural parameters such as material properties were altered and the
comparison processes were performed again. The comparison procedure therefore in-
2. Cantilever-type support structure (Figure 12.1) made with the same material
By changing the material properties, the dynamic characteristics of the structure were al-
tered, and a true test of the VRS and SRS was performed and evaluated without addition
A dynamic loading FEA solution was conducted to test the hypothesis that the use
of the VRS for natural wind gusts is an accurate approach in determining an equivalent
static wind load for fatigue. Details on the development of the VRS are provided in Chap-
ter 9: Theoretical Calculation of the Fatigue Wind Load due to Natural Wind Gust. The
develop the VRS. A dynamic load was inputted into the program as a power spectral den-
sity function, based on the wind pressure power spectral density excitation developed
from the experimentally collected data, and used to develop the VRS in Chapter 9. The
response of the FEA model for deflection at the tip of the post support and the moment at
the base plate was analyzed and recorded. This response was then compared to the linear-
static response of the equivalent static wind load taken from the VRS. A generalized de-
1. Input the dynamic wind pressure power spectral density function developed in
Chapter 9: Theoretical Calculation of the Fatigue Wind Load due to Natural Wind
2. Run a spectral solution in the FEA program and extract the displacement of the tip
of the post support and the bending moment at the base plate.
Modal Analysis including the modal frequencies and critical damping percent-
ages.
4. Use the values gathered in Step 3 to extract the equivalent static wind load from
5. Input the equivalent static wind load determined in Step 4 as a static pressure load
6. Run a linear-static solution in the FEA program and extract the displacement of
the tip of the post support and the bending moment of the base plate.
7. Compare static solution values determined in Step 6 with the dynamic solution
8. Change the model from steel to aluminum and perform steps 1 through 7 again.
With exception to Step 3, perform a FEA modal analysis to determined the modal
frequencies of the new structure, and assume a critical damping percentage for the
The loading (dynamic and static) was applied horizontally to the front façade of
the structure. The wind pressure power density spectrum developed in Chapter 9: Theo-
retical Calculation of the Fatigue Wind Load due to Natural Wind Gust is shown in Fig-
ure 12.6, and was used as input in the FEA program for the dynamic loading. The spec-
trum in the figure was developed from the average wind velocity equal to 12.96 mph
(5.79 m/s). Likewise, the average peak-to-peak range VRS was used for the comparison
and not the VRS developed from the regression analysis to simulate the wind pressure
The spectrum was inputted into the SAP program as a dynamic power density
spectrum and applied to the exposed structural members on the front façade of the struc-
ture. This included the sign, post support, truss chords, and truss vertical and diagonal
struts.
Steel model. The equivalent static wind load was extracted from the VRS (Figure
12.7) and inputted into the FEA program. The load was extracted from the VRS based on
the earliest modal frequency with a modal shape in the direction of the wind loading. This
corresponded to mode 1 with a modal frequency of 1.57 Hz. The modal frequencies
solved by the FEA program were used instead of the experimentally obtained values to be
consistent with the dynamic power density spectrum solution. The dynamic solution will
use the modal frequencies solved by the program to determine resonance areas in the
spectrum. A damping ratio value of 0.015 (1.5%), which was slightly less than the ex-
perimentally determined value of 1.86%, was used in the VRS as well as in the FEA.
The extracted VRS equivalent static wind load for the modal frequencies and
damping criteria was equal to 0.83 psf (39.7 Pa). The VRS load was multiplied by the
drag coefficient of the member it was applied, and loaded horizontally as a uniformly dis-
tributed load on the front face of the exposed members to the wind load. This included
the sign, truss members outside of the sign, and the post support.
298
Aluminum model. The equivalent static wind load was extracted from the VRS
(Figure 12.8) and inputted into the FEA program. A FEA modal analysis was performed
on the model to determine the modal frequencies. The modal frequencies did signifi-
cantly change (see Table 12.8), which was good for the comparison, but the modal shapes
FIGURE 12.8 VRS for the aluminum FEA model with 0.5% damping.
The load was extracted from the VRS based on the earliest modal frequency with
a modal shape in the direction of the wind loading. This corresponded to mode 1 with a
modal frequency of 1.127 Hz for the aluminum model. The damping ratio was changed
300
from 0.0015 (1.5%) used in the steel model to a value of 0.005 (0.5%). This value was
obtained from the literature on damping analysis of aluminum support structures (20).
The extracted VRS equivalent static wind load for the modal frequencies and
damping criteria was equal to 1.332 psf (63.8 Pa). The VRS load was multiplied by the
drag coefficient of the members it was applied, and was loaded horizontally as a uni-
formly distributed load on the front face of the exposed members to the wind load. This
included the sign, truss members outside of the sign, and the post support.
Comparison of Results
The results between the dynamic power spectral density analysis and the static
VRS analysis were compared. The comparison involved the displacement at the tip of the
post support in the y-axis direction (see Figure 12.5 for axis orientation), and the bending
moment at the base plate about the x-axis. The results are listed in Table 12.9. As shown
in the table, the results from the PDS dynamic solution and the VRS linear-static solution
show very close agreement, which substantiate the hypothesis that the equivalent static
wind load extracted from the VRS based on the dynamic characteristics of the structure
will produced the same response on the structure as a dynamic random excitation due to
wind pressure.
301
Critical
Material Natural Tip Base
Damping Analysis Case
Model Frequency Deflection Moment
Percentage
PDS Dynamic
0.09352 in 76.3 kip-in
Solution
1.57 Hz 1.5%
Steel VRS Static
0.0869 in 72.0 kip-in
Solution
Percent Difference from Dynamic FEA 7.08 % 5.64 %
PDS Dynamic
0.3388 in 114.6 kip-in
Solution
1.127 Hz 0.5%
Aluminum VRS Static
0.3344 in 115.6 kip-in
Solution
Percent Difference from Dynamic FEA 1.30 % -0.873 %
The small difference in results between the static VRS and the dynamic PDS was
due to the development of the VRS, and the method of the FEA PDS response calcula-
tion. The VRS was based on a simplification of the structure as a SDOF system, and as a
result had only one resonance spike along the frequency domain within the calculated
response PDS curve (see Figure 12.9). The FEA SAP2000 program calculated the re-
sponse PDS curve using a multiple degree-of-freedom system, which generated multiple
modal spikes.
For example, A SAP2000 output response PDS spectrum of the response bending
moment of the post support member is shown in Figure 12.10. From this view, the
smaller frequency spikes cannot be seen because they are significantly smaller than the
major response of the structure in the form of the first modal frequency. When zooming
into at a smaller window of the spectrum, as shown in Figure 12.11, the smaller spikes at
other modal frequencies are seen but are very small in amplitude as compared to the ma-
jor mode 1 spike in Figure 12.10. Subsequently, with more than one spike, the dynamic
302
FEA had a larger area under the spectrum curve to calculate the root-mean-square value,
which is proportional to the magnitude of the response values. The dynamic FEA indi-
cated a larger variance of vibration amplitudes because of the increased number of ex-
cited modes. As a result, the resulting response values (tip deflection and base plate mo-
ment) of the dynamic FEA were larger the VRS static values developed using a SDOF
system.
The small difference in results further justified the simplification of the problem
into a SDOF system for these structures. The resonance spikes in the dynamic FEA re-
sponse PDS were significantly separated from each other such that the vibration modes of
the structure in the y-direction were nearly independent. The largest resonance spike was
associated to the modal shape in the direction of the loading, and therefore was the pre-
dominate mode of vibration. Even though there were other activated modal shapes within
the vibration, they subsequently had very little effect on the overall response. In view of
this, approximating the response as a SDOF system using the VRS was validated. Proper
use of the VRS in the analysis is to select the earliest natural frequency that has a modal
shape in the direction of the excitation, and read off the corresponding equivalent static
A dynamic loading FEA solution was conducted to test the hypothesis that the use
of the SRS for truck-induced wind gust is an accurate approach in determining an equiva-
lent static wind load for fatigue. Details on the development of the SRS are provided in
Chapter 11: Theoretical Calculation of the Fatigue Wind Load due to Truck-Induced
Wind Gust. The comparison served as a mathematical check on the dynamic response
calculations used to develop the SRS. A dynamic load was inputted into the program as
transient time history function, the same function used to develop the SRS. The response
of the FEA model for deflection at the tip of the post support and the moment at the base
plate was analyzed and recorded. This response was then compared to the linear-static
305
response of the equivalent static wind load taken from the SRS. A generalized description
1. Input the dynamic transient time history function developed in Chapter 11: Theo-
retical Calculation of the Fatigue Wind Load due to Truck-Induced Wind Gust as
2. Run a dynamic solution in the FEA program and extract the displacement range
of the tip of the post support and the bending moment range at the base plate.
Modal Analysis including the modal frequencies and critical damping percent-
ages.
4. Use the values gathered in Step 3 to extract the equivalent static wind load from
the SRS developed in Chapter 11: Theoretical Calculation of the Fatigue Wind
5. Input the equivalent static wind load determined in Step 4 as a static pressure load
6. Run a linear-static solution in the FEA program and extract the displacement of
the tip of the post support and the bending moment of the base plate. Note: the
7. Compare static solution values determined in Step 6 with the dynamic solution
8. Change the model from steel to aluminum and perform steps 1 through 7 again.
With exception to Step 3, perform a FEA modal analysis to determine the modal
The loading (dynamic and static) was applied vertically to the horizontal area 12
ft (3.66 m) from the end of the truss structure. The wind pressure transient time history
function developed in Chapter 11: Theoretical Calculation of the Fatigue Wind Load due
to Truck-Induced Wind Gust is shown in Figure 12.12, and was used as input in the FEA
program for the dynamic loading. The 70 mph (31.3 m/s) was used for the steel model,
whereas the 60 mph (26.8 m/s) was used for the aluminum model.
The transient time history function was inputted into the SAP program as a dy-
namic time history and was loaded in 0.01 second time steps. The loading was applied to
the exposed horizontal area on the underneath face of the structure. This included the
Steel model. The equivalent static wind load was extracted from the SRS (Figure
12.13) and inputted into the FEA program. The load was extracted from the SRS based
on the earliest modal frequency with a modal shape in the direction of the wind loading.
This corresponded to mode 2 with a modal frequency of 1.62 Hz. The modal analysis
solved by the FEA program was used instead of the experimentally obtained values to be
consistent with the FEA dynamic solution. The dynamic solution will use the modal fre-
quencies solved by the program to determine resonance. The steel model used the 70 mph
(31.3 m/s) time history plot. Damping was not considered in the analysis as the initial
The extracted SRS equivalent static wind load for the modal frequencies and truck
speed criteria was equal to 8.33 psf (399 Pa). The SRS load was multiplied by the drag
coefficient of the member it was applied, and loaded vertically as a uniformly distributed
load on the underneath horizontal face of the exposed members to the wind load. This
Aluminum model. The equivalent static wind load was extracted from the SRS
(Figure 12.13) and inputted into the FEA program. A FEA modal analysis was performed
on the model to determine the modal frequencies. The modal frequencies did signifi-
cantly change (see Table 12.8), which was good for the comparison, and the modal
The load was extracted from the SRS based on the first modal frequency with a
modal shape in the direction of the wind loading. This corresponded to mode 2 with a
modal frequency of 1.212 Hz for the aluminum model. The 60 mph (26.8 m/s) time his-
tory was used for the aluminum model to test the accuracy of the SRS with changing
speeds. The extracted SRS equivalent static wind load for the modal frequency was equal
309
to 5.89 psf (282 Pa). The SRS load was multiplied by the drag coefficient of the members
it was applied, and was loaded vertically as a uniformly distributed load on the under-
neath horizontal area of the exposed members to the wind load. This included chords,
Comparison of Results
The results between the dynamic time history analysis and the static SRS analysis
were compared. The comparison involved the displacement at the tip of the post support
in the x-axis direction (see Figure 12.5 for axis orientation), and the bending moment at
the base plate about the y-axis. Evaluation of the time history solution involved extracting
the maximum peak-to-peak range of displacement and moment values to use in the com-
parison because the SRS was based on a peak-to-peak range. Damping was not consid-
As shown in the table, the results from the time history (TH) dynamic solution
and the SRS linear-static solution show close agreement, which substantiate the hypothe-
sis that the equivalent static wind load extracted from the SRS based on the dynamic
characteristics of the structure will produced the same response on the structure as a dy-
namic random excitation due to wind pressure. In view of this, approximating the re-
sponse as a SDOF system using the SRS was validated. Proper use of the SRS in the
analysis is to select the first modal frequency that has a modal shape in the direction of
the excitation, and read off the corresponding equivalent static wind load ordinate.
310
CHAPTER 13
Chapter Overview
A complete overview of the research performed with this project was completed
and is presented. A discussion of the results from the project is provided in this chapter,
followed by proposed fatigue design equations for natural wind and truck-induced wind
gusts. The review encompassed both the experimental and theoretical results, and a de-
sign methodology was developed. Design examples are provided to show the use of the
proposed fatigue provisions, and comparisons are made with the Supports Specifications.
The proposed fatigue design equations for natural wind gusts were developed
from the experimental and theoretical programs with this project. The recommendations
are presented in two forms. The first form is a more generalized design equation, whereas
the second equation is a more detailed design equation that accounts for the dynamic
Discussion of Results
The evaluation of the results involved primarily the comparison between the
equivalent static wind loads developed from the experimental and theoretical programs.
312
The equivalent static wind load was determined for a fatigue wind equal to 38 mph (17
m/s). The fatigue wind was determined using the infinite-life approach to fatigue design
of support structures as the wind velocity with a 0.01% exceedence probability from an
annual mean wind velocity of 11 mph (5 m/s). The wind pressure at the fatigue wind was
determined in both programs. The results of the analysis are listed in Table 13.1.
TABLE 13.1 Experimental and Theoretical Results for Natural Wind Gust
The results from the two programs were similar with a percent difference from the
experimental equal to 5.24%. The experimental was larger than the theoretical because
the VRS was developed using a SDOF to simulate the structural response. The experi-
mental was based on the real structure, which was a multiple degree of freedom dynamic
system. The closeness of the two values indicated that approximating the structure as a
on the structure that were vulnerable to fatigue. A complete distribution of stress within
the structure from the wind loading was established. Large variances in measured strain
ranges were realized in the chord members and anchor bolts taking into account the bal-
anced state with respect to the neutral axis created from the loading. This was especially
true for the anchor bolts, and was believed to be due to the clearance differences between
313
anchor bolts and the base plate. The strain measurements made on the post support at
Sections AA and BB were more consistent and a more straightforward calculation was
accomplished.
The fatigue load due to natural wind gust was determined from the measurements
made on the post support. The location of these gauges, and the orientations to in-plane
and out-of-plane vibrations made the calculation straightforward. The results of the back-
calculation for the wind pressures were consistent with each gauge located along the cir-
cumference of the post. A wind pressure value of 4 psf (191 Pa) was determined from the
analysis. This represented an equivalent static wind load that produced the same stress
range on the structure as measured experimentally from dynamic vibration due to ran-
domly applied natural wind loading. The value was determined at the fatigue wind of 38
mph (17 m/s) in compliance with the infinite-life approach to fatigue design for support
structures.
The same methodology in calculating the 4 psf (191 Pa) wind pressure was per-
formed with this project as with the development of the fatigue load due to natural wind
adopted by the Supports Specifications. The difference was attributed to the attainment of
the load through experimental versus theoretical means. The fatigue load of the Supports
sis was used, that was based on a multiple degree-of-freedom dynamic system. The same
method of computation was performed with this project, but by using experimentally
The formulation of the VRS in calculating the fatigue load due to natural wind
load resulted in a successful comparison with the experimental results. The methodology
314
was a hybrid using both experimental and theoretical applications. The wind velocity data
collected experimentally was used to develop wind pressure power density spectrum ex-
citation by the same procedure performed by the Supports Specifications. The difference
was in the structural response analysis. The VRS was comprehensive and universal in
that the method can be applied to any support structure of different configuration, size,
shape, and material properties. The fatigue load is extracted from the VRS based on the
dynamic characteristics of the structure, which was directly related to the stiffness and
The results of the VRS approach produced an equivalent static wind load that was
slightly less than the experimentally obtained value. This was due to the formulation of
the VRS as a SDOF. The response of the structure was approximated as a SDOF,
whereas the real structure is a multiple degree-of-freedom system. As proved by the FEA
dynamic loading comparisons and the close similarity between the VRS value and the
The design fatigue load provisions were developed with this project in response to
the overview and discussion of the results. The established fatigue design equation was
The general fatigue design equation was based primarily on the experimental results. A
more detailed approach was developed using the VRS, which was considered accurate
315
due to the close comparison between the results of the experimental and theoretical re-
search program of this project. This approach can be used if the dynamic characteristics
of the structure, such as the modal frequencies and shapes, as well as critical damping
The general fatigue design equation for natural wind gusts is shown as Eq. 13.1. It
view. The load is applied as a uniformly distributed load to be easily applied by the engi-
neer. It is recommended for cantilever-type sign support structures; specifically for sup-
port structures with a modal frequency of 1.61 Hz, and a critical damping percentage of
1.82%.
2
v
PNW = 4C d I F [Eq. 13.1]
11
where
PNW = design fatigue load due to natural wind gust, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean three second wind velocity, mph (m/s)
The equation takes the form similar to the Support Specifications for easy impli-
cation. The drag coefficients and importance factors are not changed, and are provided in
316
the Supports Specifications. Accommodation for other annual mean wind velocities other
than 11 mph (5 m/s) is also provided in the equation by a simplified ratio calculation.
A more detailed design equation was developed to account for the variety of sign
support structures in design, each with different configurations, cross sectional shapes,
and material properties. Since these factors have significant influence on the dynamic
characteristics of the structure, and because sign supports structures are highly flexible
with low damping properties, a method was needed in determining the fatigue load based
on the vibration behavior of these structures. The approach presented in this section ac-
counted for the dynamic behavior of the structure in terms of the modal frequency and
critical damping percentages, and the fatigue load was determined based on these proper-
of natural wind gusts in that it accounts for the variety of support structures in design.
Vibration response spectrums were developed that presented the fatigue load
based on the dynamic properties of the structure. The load is extracted from the VRS in
terms of the earliest modal frequency of the structure with a modal shape in the direction
of the loading. The most critical loading scenario for natural wind is directed normal to
the plain of the sign (in the direction of traffic), exciting the modal shape most commonly
referred to as the horizontal modal shape. For sign support structures, the horizontal mo-
dal shape is generally around 1 to 3 Hz, which typically corresponds to the first modal
shape for cantilever-type structures, and the second modal shape for bridge-type struc-
tures. Finite element software (i.e., SAP2000) can be used to estimate the appropriate
317
modal shapes and their associated natural frequencies to use with the VRS curves. If FEA
software is not available, fundamental structural dynamics of a SDOF system (Eq. 13.2)
can be used for estimating these values (10, 17, 33, 56, 57, 61, 63).
1 K
fn = [Eq. 13.2]
2π M
where
f n = natural frequency, Hz
K = generalized stiffness, N/m (lb/ft)
M = generalized mass, kg (slug)
VRS, as the equivalent static wind load increases as the modal frequency decreases be-
cause of the proximity to the broadband amplitude of the wind pressure PSD excitation
because of their relatively high flexibility and subsequent low natural frequencies (1 to 3
Hz). What’s more, their damping ratios are mostly below 2.0%. A low damping will al-
low the structure to vibrate longer at high amplitudes, and thus produce more stress that
could potentially cause fatigue damage. It is recommended for design purposes to assume
a damping ratio equal to 0.015 (1.5%). However, damping ratios can vary depending on
the structural material, and therefore actual values can be obtained from experimental
A systematic procedure for determining the fatigue design load due to natural
Step 1: annual mean wind velocity. The first step involves determining the annual
mean wind velocity of the site. The National Weather Service Offices near the site can be
used to determine this value. It is recommended to use a value of 11 mph (5 m/s), how-
Step 2: modal analysis. The second step involves a modal analysis of the design-
ing support structure. The earliest modal frequency of the modal shape in the direction of
the natural wind loading is needed, as well as the critical damping percentage. Finite ele-
ment analysis software can be helpful in this step. If FEA is not available, the modal fre-
quency and modal shape can be estimated using Eq. 13.2. If experimental data of compa-
rable structures are not available, it is recommended to use a damping ratio estimated at
0.015 (1.5%).
Step 3: vibration response spectrum. The third step involves the VRS shown in
Figure 13.1 and Figure 13.2. A VRS constant, PVRS, is extracted from the spectrum as the
ordinate value corresponding to the natural frequency and damping ratio determined in
Step 2. If using a damping ratio of 1.5% (recommended), use the VRS in Figure 13.1.
The VRS in Figure 13.2 is provided for other damping ratios if needed.
319
Step 4: fatigue design equation. The final step is plugging all information gath-
ered from Steps 1-3 into the fatigue design equation for natural wind gusts based on the
detailed approach. Input the VRS constant, PVRS, determined from Step 3 into Eq. 13.3
for each member along the facade of the structure exposed to natural wind.
2
v
PNW = PVRS C d I F [Eq. 13.3]
11
where
PNW = design fatigue load due to natural wind gust, psf (Pa)
PVRS = transmitted pressure constant extracted from the VRS, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean three second wind velocity, mph (m/s)
The proposed fatigue design equations for truck-induced wind gust were devel-
oped from the experimental and theoretical programs with this project. The recommenda-
tions are presented in two forms. The first form is a generalized design equation, whereas
the second equation is a more detailed design equation that accounts for the dynamic
Discussion of Results
The evaluation of the results involved primarily the comparison of the results
from the experimental and theoretical programs. In the experimental program, the truck-
induced gust fatigue load was developed from measured strain ranges collected during
the truck tests. An equivalent static wind load was developed that would produce an
equivalent strain range as measured experimentally. The analytical program involved de-
velopment of the SRS, where the structure was approximated to behave dynamically as a
SDOF system. The equivalent static wind load was extracted from the SRS based on the
13.2.
The percent difference of the theoretical value to the value developed from the
experimental results was 0.216%, indicated good similarity between the two programs.
The theoretical program and SRS was seen as a good approximation for determining the
fatigue load. There was very little variation in the experimental value with respect to
truck speed. The maximum range at the location used for the back-calculation was ob-
served at 60 mph (26.8 m/s), yet values occurring from the higher speeds tested were ex-
tremely close. The majority of analysis of data at the other instrumented locations on the
322
structure including the anchor bolts and chord members showed higher range values oc-
curring at the highest speed of 70 mph (31.1 m/s). In view of this, the design equation
was developed using the maximum range measured on the post member, and since the
measured values were similar with truck speed, the 70 mph (31.3 m/s) was used as the
reference speed for this load. As a result, accounting for higher speeds would not unnec-
The fatigue resistance analysis indicated that the most suitable location to calcu-
late the fatigue load was from strain measurements on the post member. This was due to
the consistency of the data at this location, the orientation to in-plane vibration, and also
A vertical load equal to 8.348 psf (400 Pa) resulting from a truck speed of 60 mph
(26.8 m/s) was calculated. There was evidence in the data of a horizontal load, but the
distribution of this load with height above ground level was not experimentally measured.
The maximum height of the pressure load significantly influences the magnitude of the
calculated load. The recorded strain ranges due to a horizontal load were much larger
during the natural wind tests than for the truck-induced wind tests. As a consequence, no
experimental value was obtained in the analysis for a horizontal component, assuming
The general fatigue design equation for truck-induced wind gusts is shown as Eq.
13.4. It is to be applied vertically to the horizontal area on the underneath façade of the
323
traffic lane, whichever is greater) located directly over the traffic lane. The load is applied
height of 19.7 ft (6.00 m) above the roadway, and decreases linearly to zero at a height of
32.8 ft (10 m). Equation 13.4 is recommended for cantilever-type sign support structures;
specifically for support structures with a modal frequency of 1.64 Hz for vertical rocking
2
v
PTG = 8.4C d I F [Eq. 13.4]
70
where
PTG = design fatigue load due to truck - induced wind gust, psf (Pa)
C d = drag coefficient
I F = importance factor
v = traveling speed of the truck, mph (m/s)
The equation takes the form similar to the Support Specifications for easy impli-
cation. The drag coefficients and importance factors are not changed, and are provided in
the Supports Specifications. Accommodation for other traveling speeds other than 70
mph (31.3 m/s) is provided in the equation if needed, however it is not recommended use
A more detailed design equation was developed to account for the variety of sign
support structures in design, each with different configurations, cross sectional shapes,
and material properties. Since these factors have significant influence on the dynamic
characteristics of the structure, and because sign supports structures are highly flexible
with low damping properties, a method was needed in determining the fatigue load based
on the vibration behavior of these structures. The approach presented in this section ac-
counts for the dynamic behavior of the structure in terms of the modal frequency. The
fatigue load is determined based on these properties. The approach was considered a uni-
fied design approach to fatigue design for support structures in that it accounts for the
Shock response spectrums were developed that presents the fatigue load based on
the dynamic properties of the structure. The load is extracted from the SRS in terms of
the earliest modal frequency of the structure with a modal shape in the direction of the
loading. This shape is commonly referred to as the vertical rocking modal shape and has
vibratory motion in-plane to the vertically applied truck load. For sign support structures,
the vertical modal shape is close to the horizontal modal shape, and is generally around 1
to 3 Hz, which typically corresponds to the second modal shape for cantilever-type struc-
tures, and the third modal shape for bridge-type structures. Finite element software (i.e.,
SAP2000) can be used to estimate the appropriate modal shapes and their associated
natural frequencies to use with the SRS curves. If FEA software is not available, funda-
mental structural dynamics of a SDOF system (Eq. 13.5) can be used for estimating these
values (10, 17, 33, 56, 57, 61, 63). A conservative approach would be to use the peak
325
amplitude from the SRS that corresponds to the truck speed regardless of the natural fre-
1 K
fn = [Eq. 13.5]
2π M
where
f n = natural frequency, Hz
K = generalized stiffness, N/m (lb/ft)
M = generalized mass, kg (slug)
A systematic procedure for determining the fatigue design load due to truck-
induced wind gusts based on the proposed detailed approach may be given as follows.
Step 1: design speed of the truck. The first step involves determining the design
speed of the truck. A speed of 70 mph (31.3 m/s) is recommended however other speeds
can be used if needed. It is not recommended to use speeds less than 70 mph (31.3 m/s).
Step 2: modal analysis. The second step involves a modal analysis of the support
structure. The earliest modal frequency with a modal shape in the direction of the truck
gust loading is needed. The damping ratio is not required for truck-induced gusts. Finite
element analysis software can be helpful in this step. If FEA is not available, the modal
Step 3: shock response spectrum. The third step involves the SRS shown in Figure
13.3. A SRS constant, PSRS, is extracted from the spectrum as the ordinate value corre-
Step 4: fatigue design equation. The next step is plugging all information gathered
from Steps 1-3 into the fatigue design equation for truck-induced gusts based on the de-
tailed approach. Simply input the SRS constant, PSRS, determined from Step 3 into Eq.
13.6 for each member along the facade of the structure exposed to truck-induced wind
gusts. It is to be applied in the same fashion as the general design equation for truck-
where
PTG = design fatigue load due to truck - induced wind gust, psf (Pa)
PSRS = transmitted pressure constant extracted from the SRS, psf (Pa)
C d = drag coefficient
I F = importance factor
values determined from Eq. 13.6 for structures higher than 19.7 ft (6.00 m). The truck
gust pressure is assumed maximum at a height of 19.7 ft (6.00 m) above the roadway, and
Design Examples
Examples for determining the fatigue load are presented for natural wind and
truck-induced wind gusts. The general and detail approaches were used and compared to
each other as well as to the design fatigue equation in the Supports Specifications. The
comparison tested the accuracy of the Supports Specifications with respect to the experi-
mentally determined values, as well as the theoretical approaches developed with this
project. A variety of case studies of different types of structures were used in the com-
parisons, each with different structural dynamic properties that match real structures in
operation.
328
The fatigue design equations for natural wind and truck-induced gusts are pro-
vided in Eq. 13.7 and Eq. 13.8 (1). The fatigue load was calculated using these equations
for the case studies presented, and the results were compared to the fatigue design equa-
2
v
PNW = 5.2C d I F [Eq. 13.7]
11
where
PNW = design fatigue load due to natural wind, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean wind velocity other than11 mph (5 m/s)
2
V
PTG = 18.8C d I F [Eq. 13.8]
65
where
PTG = design fatigue load due to truck gust, psf (Pa)
C d = drag coefficient
I F = importance factor
V = truck speed, mph (m/s)
329
The above equations were used for the comparison, and were applied in the manner
Six different design cases were performed to illustrate the developed design ap-
proaches for calculating fatigue loads due to natural wind, and more importantly to com-
pare the results to the Supports Specifications. A variety of overhead sign support struc-
tures with different dynamic properties and exposure conditions are provided by the six
design cases listed in Table 13.3. The dynamic properties reflect sign support structures
that were evaluated in previous studies (10, 20, 38, 52). The design cases were chosen to
the Supports Specifications. This was done primarily by altering the natural frequencies
and damping ratios, which represents a practical scenario as these structures are not exact
The natural wind loading for the six cases was directed onto the exposed front fa-
cade of the structure (normal to the front plain of the sign and support members, in the
direction of traffic), as shown in Figure 13.4. A fatigue pressure due to natural wind was
applied to the post support only for this example. A drag coefficient equal to 1.10 was
used for the 2.0 ft (0.61 m) diameter member, and an importance factor equal 1.0. The
results of the analysis using the general and detailed approach (Eq. 13.1 and Eq. 13.3)
and the Supports Specifications (Eq. 13.7) are given in Table 13.4. The VRS used for
Design
Case Description
Case
Annual mean wind velocity equal to 11 mph
Case 1 Structural natural frequency of 2.00 Hz in the direction of loading
0.02 (2%) damping ratio
Annual mean wind velocity equal to 13 mph
Case 2 Structural natural frequency of 2.00 Hz in the direction of loading
0.02 (2%) damping ratio
Annual mean wind velocity equal to 11 mph
Case 3 Structural natural frequency of 0.802 Hz in the direction of loading
0.02 (2%) damping ratio
Annual mean wind velocity equal to 11 mph
Case 4 Structural natural frequency of 2.00 Hz in the direction of loading
0.005 (0.5%) damping ratio
Annual mean wind velocity equal to 11 mph
Case 5 Structural natural frequency of 0.802 Hz in the direction of loading
0.005 (0.5%) damping ratio
Annual mean wind velocity equal to 13 mph
Case 6 Structural natural frequency of 0.802 Hz in the direction of loading
0.005 (0.5%) damping ratio
Direction of
Wind Loading
The frequencies listed in Table 13.3 represent the earliest modal frequency with a
modal shape in the direction of the wind loading (see Figure 13.4). The equivalent static
wind loads extracted from the VRS figures were based on these frequencies.
Comparisons between the general and detail fatigue design equation was per-
formed, as well as a comparison with the Supports Specifications. Case 1 was considered
a control case because it closely reflected the conditions in the development of the Sup-
ports Specifications. The development of the natural wind fatigue provisions in the Sup-
ports Specifications used structures with a natural frequency of 2.00 Hz, and a critical
damping percentage equal to 2%. The structural dynamic characteristics specified for
Case 1 were considered representative of the support structures in operation. The results
334
in Table 13.4 indicated the Supports Specifications overestimated the fatigue load as
Large differences resulted between the general and detailed equations developed
with this research, and with the Supports Specifications. The comparison was especially
through Case 6, resulted in very large differences. The general equation was based on the
tested structure with a damping percentage equal to 1.82%, which was close to 2.00%.
The case studies with a damping percentage equal to 0.5% resulted in an increased value
ferences between the approaches. Structures with a lower natural frequency encroached
closer to the broadband peak frequency of the wind, and therefore resulted in higher
resonant behavior. This along with a reduced damping set up an environment where the
As seen in Figure 13.1, the design fatigue load calculated using the Supports
Specifications was conservative for structures with a natural frequency greater than 0.65
Hz. This value will increase for critical damping percentages less than 1.5%. With struc-
tures with damping percentages equal to 2%, the Supports Specifications was conserva-
tive for natural frequencies greater than 0.5 Hz. Likewise, for increasing damping per-
centages, the Supports Specifications becomes more conservative at lower structural fre-
quencies. In general, by increasing the natural frequency of the structure and the damping
The frequency of the structure becomes further away from the peak amplitude of the
wind pressure frequency, and resonance becomes less influential to the response.
Three different design case scenarios were performed to illustrate the proposed
design approach for calculating fatigue loads due to truck gusts, and to compare the re-
sults to the Supports Specifications. A description of the design cases are listed in Table
13.5. The pressure loading calculated using Eq. 13.4 and Eq. 13.6 along with Figure 13.3
of the detailed approach was used for the comparison to the pressure load calculated us-
Design
Case Description
Case
Truck speed equal to 65 mph
Case 1
Structural natural frequency of 2.00 Hz in the direction of loading
Annual mean wind velocity equal to 65 mph
Case 2
Structural natural frequency of 0.802 Hz in the direction of loading
Annual mean wind velocity equal to 70 mph
Case 3
Structural natural frequency of 2.00 Hz in the direction of loading
Annual mean wind velocity equal to 70 mph
Case 4
Structural natural frequency of 0.802 Hz in the direction of loading
The same sign support structures depicted in the natural wind gust design example
section (Table 13.3) was used for the comparison. A four-chord truss overhang was as-
sumed for the case scenarios. The chords were 3.5 in (88.9 mm) in diameter and were a
distance of 20 ft (6.1 m) above ground level. This height was at the reduction threshold
336
according to the Supports Specifications (1), and therefore no accommodation for height
was calculated.
The calculations reflect the design fatigue load to be applied vertically to the hori-
zontal area on the underneath portion of the overhead truss above the traffic lane (see
Figure 13.9). A drag coefficient equal to 1.10 (calculated using the Supports Specifica-
tions provisions for drag coefficients) and an importance factor equal to 1.0 was used for
the load calculation. A summary of the results of the comparison are listed in Table 13.6.
Direction of
Wind Loading
The results from both the general and detailed design equations were significantly
less than the Supports Specifications, as the equations of this project were developed
from experimental evidence on cantilever sign support structures. When comparing the
general and detailed design equations to each other, very interesting conclusions are
gathered. The general equation is recommended for structures with the earliest modal
frequency with a modal shape in the direction of loading equal to 1.64 Hz. As seen in the
SRS of Figure 13.3, the fatigue load due to truck-induced gusts will be significantly re-
duced for structures with natural frequencies as furthest away, greater or smaller, from
approximately 1.6 Hz for trucks traveling at 70 mph (31.3 m/s). Structures with frequen-
cies at 1.6 Hz will see the largest vibration response due to resonance with the truck in-
duced wind pulse. In view of the results gathered with natural wind gusts, it is advisable
to design structures with modal frequencies of the first two modes of vibration higher
than 1.60 Hz to comply with both natural wind and truck-induced wind gust loading.
338
CHAPTER 14
Chapter Overview
into sections corresponding to the major tasks performed in this project including:
Fatigue resistance,
Modal analysis,
1. The theoretical models using the VRS for natural wind gust and SRS for truck-
induced wind gust were accurate in determining the fatigue load for the experi-
2. General and detailed fatigue design equations were developed from the theoretical
and experimental programs for natural wind. The general equation is recom-
mended for structures with a modal frequency of the earliest modal shape in the
339
with natural frequencies less than 1.61 Hz and critical damping percentages less
than 1.82%. The detailed equation can also be used for natural frequencies and
damping percentages greater than 1.82% as the general equation is slightly con-
3. General and detailed fatigue design equations were developed from the theoretical
and experimental programs for truck-induced gust. The general equation is rec-
ommended for structures with a modal frequency of the earliest modal shape in
the direction of loading equal to 1.64 Hz or greater. The detailed equation is rec-
ommended for structures with natural frequencies significant greater or less than
1.64 Hz.
4. The Supports Specifications when compared to the developed equations for natu-
ral wind can be conservative for structures with natural frequencies greater than
0.50 Hz. The limiting frequency begins to increase for structures with damping
5. The Supports Specifications when compared to the developed equations for truck-
6. The fatigue stress generated on the tested cantilever structure from natural wind
gust was as much as four times greater than the fatigue stress generated from
7. Support structures with natural frequencies greater than 1.60 Hz are less vulner-
able to fatigue loading. The higher the natural frequency of the structure, the less
8. Structural vibration became momentous for wind speeds greater than 9 mph (4
9. The anchor bolts with clearances 3 in (76.2 mm) or larger had stress ranges over
was due to an increased significance of the bending moment stress in the anchor
bolt, and was especially critical during natural wind loading where high torsion
Fatigue Resistance
An analysis of the fatigue resistance of the structure was performed. The stress
provided by the Supports Specifications. A failure index was calculated to provide a rela-
tionship between the gauged locations with a control value. The distribution of the stress
in the structure was evaluated and the most appropriate location for determining the fa-
The largest strain recorded was located on the anchor bolts followed by the post
and chords, as expected. The major findings are listed in Table 14.1 and 14.2 for natural
wind and truck-induced gusts. The table lists the stress ranges determined at the average
wind velocity and the fatigue wind velocity. Maximum ranges at the specified location
are listed for each wind velocity, as well as the average stress range from all gauges at the
341
section. The stress range was defined as the peak-to-peak range from the measured re-
stress distribution throughout the structure. Strains measured in the chords near the
chord-to-column connection matched well with the measured strains on the post. The
strains on the post were slightly larger in proportion to the additional exposed wind area
at these locations. The chords recorded loads created from the exposed area of the sign
and truss, whereas the post recorded loads from the sign, truss, and post members.
The structure as a whole was within the threshold limitations, with the exception
of the anchor bolts. The stress distribution between anchor bolts from the base plate to the
concrete foundation was not consistent. The anchor bolts with large clearances of 3 in
(76.2 mm) or more were heavily loaded, much more than the other bolts. The majority of
stress was concentrated in these bolts and controlled the overall stability of the structure.
342
This was believed to be due to the larger influence of out-of-plane forces induced onto
the bolt from the base plate, which created large bending moments in the bolt.
The most consistent instrumented location on the structure was the post member.
The best strain results were produced at these locations, showing the balanced state of
strain with respect to the proximity of each gauge to the neutral axis created from the
magnitude and direction of the load. A high degree of variability in stress ranges were
found in the anchor bolts due to the inconsistent clearances between the bolts. A degree
of variability was also observed in the chords. In view of the results, the fatigue load
back-calculation methodology was performed using the strain gauge readings on the post
member. The calculation was more straightforward, and provided consistent results.
For the truck tests, the stress ranges on averaged increased with the speed of the
truck. As compared to the natural wind gust measurements, no significant loading was
343
generated from the truck gusts. This is illustrated in the bar graph of Figure 14.1, showing
the stress ranges for the fatigue wind as compared to the stress ranges for the 70 mph
(31.3 m/s) truck speed. A significant increase in stress range for both natural wind and
truck gusts was observed for the anchor bolts. This was due to the anchor bolts with large
stress. The values shown in the figure represent the maximum stress ranges recorded at
each section.
Modal Analysis
Experimental and analytical modal analysis was conducted on the tested structure.
Modal frequencies, modal damping, and modal shapes were indentified from the analysis.
The results of the two programs (experimental and analytical) matched well with each
The experimental program involved forming FRF from the structural excitation
and response measurements. The accelerometer data proved to be the best instrumenta-
tion to use for the response measurements. Modal frequencies and damping values can be
determined from the strain gauge data, of which matched the accelerometer data, how-
The analytical program for modal analysis involved FEA using the software
package SAP2000. The modal shapes and modal frequencies were close to the measured
values, indicating an accurate modal analysis with both approaches. This helped to vali-
date the modal was performing accurately in depicting the structural behavior under load.
344
FIGURE 14.1 Comparison of results between natural wind and truck gusts.
The fatigue load due to natural wind gusts was determined. A general design
equation and a more detailed design equation using the VRS were developed. Compari-
sons were made between the two approaches and with the Supports Specifications fatigue
design equation. The results indicated Supports Specifications were conservative for cer-
tain situations.
It was concluded from the analysis that the design fatigue load calculated using
the Supports Specifications was conservative for structures with a natural frequency
greater than 0.50 Hz. This value increased for critical damping percentages less than
1.82%. Likewise, for increasing damping percentages, the Supports Specifications be-
came more conservative at lower structural frequencies. By increasing the natural fre-
345
quency of the structure and the damping percentage, the vulnerability of the structure to
fatigue damage was decreased. Stress is induced onto the structure when it vibrates;
thereby increasing the frequency and damping values will result in smaller vibration am-
plitudes. More importantly, the modal frequency of the structure becomes further away
from the resonant frequency of excitation and response, and therefore resonance has a
lessoned influence.
The general fatigue design equation for natural wind gusts is shown as Eq. 14.1. It
view. The load is applied as a uniformly distributed load to be easily applied by the engi-
neer. It is recommended for cantilever-type sign support structures; specifically for sup-
port structures with a modal frequency of 1.61 Hz, and a critical damping percentage of
1.82%.
2
v
PNW = 4C d I F [Eq. 14.1]
11
where
PNW = design fatigue load due to natural wind gust, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean three second wind velocity wind velocity, mph (m/s)
346
The equation takes the form similar to the Support Specifications for easy impli-
cation. The drag coefficients and importance factors are not changed, and are provided in
the Supports Specifications. Accommodation for other annual mean wind velocities other
A more detailed design equation was developed to account for the variety of sign
support structures in design, each with different configurations, cross sectional shapes,
and material properties. Since these factors have significant influence on the dynamic
characteristics of the structure, and because sign supports structures are highly flexible
with low damping properties, a method was needed in determining the fatigue load based
on the vibration behavior of these structures. The approach presented in this section ac-
counted for the dynamic behavior of the structure in terms of the modal frequency and
critical damping percentages, and the fatigue load was determined based on these proper-
ties. The approach was considered a unified design approach to fatigue of natural wind
Vibration response spectrums were developed that presented the fatigue load
based on the dynamic properties of the structure. The load is extracted from the VRS in
terms of the first modal frequency of the structure with a modal shape in the direction of
the loading. The most critical loading scenario for natural wind is directed normal to the
plain of the sign (in the direction of traffic), which excites the modal shape most com-
monly referred to as the horizontal modal shape. For sign support structures, the horizon-
tal modal shape is generally around 1 to 3 Hz, which typically corresponds to the first
347
modal shape for cantilever-type structures, and the second modal shape for bridge-type
structures. Finite element software (i.e., SAP2000) can be used to estimate the appropri-
ate modal shapes and their associated natural frequencies to use with the VRS curves. If
FEA software is not available, fundamental structural dynamics of a SDOF system (Eq.
14.2) can be used for estimating these values (10, 17, 33, 56, 57, 61, 63).
1 K
fn = [Eq. 14.2]
2π M
where
f n = natural frequency, Hz
K = generalized stiffness, N/m (lb/ft)
M = generalized mass, kg (slug)
VRS, as the equivalent static wind load increases as the modal frequency decreases be-
cause of the proximity to the broadband amplitude of the wind pressure PSD excitation
because of their relatively high flexibility and subsequent low natural frequencies (1 to 3
Hz). What’s more, their damping ratios are mostly below 2.0%. A low damping will al-
low the structure to vibrate longer at high amplitudes, and thus produce more stress that
could potentially cause fatigue damage. It is recommended for design purposes to assume
a damping ratio equal to 0.015 (1.5%). However, damping ratios can vary depending on
348
the structural material, and therefore actual values can be obtained from experimental
A systematic procedure for determining the fatigue design load due to natural
Step 1: annual mean wind velocity. The first step involves determining the annual
mean wind velocity of the site. The National Weather Service Offices near the site can be
used to determine this value. It is recommended to use a value of 11 mph (5 m/s), how-
Step 2: modal analysis. The second step involves a modal analysis of the design-
ing support structure. The earliest modal frequency of the modal shape in the direction of
the natural wind loading is needed, as well as the critical damping percentage. Finite ele-
ment analysis software can be helpful in this step. If FEA is not available, the modal fre-
quency and modal shape can be estimated using Eq. 14.2. If experimental data of compa-
rable structures are not available, it is recommended to use a damping ratio estimated at
0.015 (1.5%).
Step 3: vibration response spectrum. The third step involves the VRS shown in
Figure 14.2 and Figure 14.3. A VRS constant, PVRS, is extracted from the spectrum as the
ordinate value corresponding to the natural frequency and damping ratio determined in
349
Step 2. If using a damping ratio of 1.5% (recommended), use the VRS in Figure 14.2.
The VRS in Figure 14.3 is provided for other damping ratios if needed.
Step 4: fatigue design equation. The final step is plugging all information gath-
ered from Steps 1-1 into the fatigue design equation for natural wind gusts based on the
detailed approach. Simply input the VRS constant, PVRS, determined from Step 3 into Eq.
14.3 for each member along the facade of the structure exposed to natural wind.
2
v
PNW = PVRS C d I F [Eq. 14.3]
11
where
PNW = design fatigue load due to natural wind gust, psf (Pa)
PVRS = transmitted pressure constant extracted from the VRS, psf (Pa)
C d = drag coefficient
I F = importance factor
v = annual mean three second wind velocity, mph (m/s)
350
The fatigue load due to truck-induced wind gusts was determined. A general de-
sign equation and a more detailed design equation using the SRS were developed. Com-
parisons were made between the two approaches and with the Supports Specifications
fatigue design equation. The results indicated Supports Specifications were highly con-
servative.
It was concluded from the analysis that the fatigue load due to truck-induced gusts
was significantly reduced for structures with natural frequencies as furthest away, greater
or smaller, from approximately 1.6 Hz for trucks traveling at 70 mph (31.3 m/s). Struc-
tures with frequencies at 1.6 Hz experienced the largest vibration response due to reso-
nance with the truck induced wind pulse. In view of the result with natural wind gusts, it
is advisable to design structures with modal frequencies of the first two modes of vibra-
tion higher than 1.60 Hz to comply with both natural wind and truck-induced wind gust
loading.
The general fatigue design equation for truck-induced wind gusts is shown as Eq.
14.4. It is to be applied vertically to the horizontal area on the underneath façade of the
traffic lane, whichever is greater) located directly over the traffic lane. The load is applied
height of 19.7 ft (6.00 m) above the roadway, and decreases linearly to zero at a height of
32.8 ft (10 m). Equation 14.4 is recommended for cantilever-type sign support structures;
352
specifically for support structures with a modal frequency of 1.64 Hz for vertical rocking
2
v
PTG = 8 . 4C d I F [Eq. 14.4]
70
where
PTG = design fatigue load due to truck - induced wind gust, psf (Pa)
C d = drag coefficient
I F = importance factor
v = traveling speed of the truck, mph (m/s)
The equation takes the form similar to the Support Specifications for easy impli-
cation. The drag coefficients and importance factors were not changed, and are provided
in the Supports Specifications. Accommodation for other traveling speeds other than 70
mph (31.3 m/s) is provided in the equation if needed, however it is not recommended use
A more detailed design equation was developed to account for the variety of sign
support structures in design, each with different configurations, cross sectional shapes,
and material properties. Since these factors have significant influence on the dynamic
characteristics of the structure, and because sign supports structures are highly flexible
with low damping properties, a method is needed in determining the fatigue load based
on the vibration behavior of these structures. The approach presented in this section ac-
353
counts for the dynamic behavior of the structure in terms of the modal frequency. The
fatigue load is determined based on these properties. The approach is considered a unified
design approach to fatigue design for support structures in that it accounts for the variety
of structures in design.
Shock response spectrums were developed that presents the fatigue load based on
the dynamic properties of the structure. The load is extracted from the SRS in terms of
the earliest modal frequency of the structure with a modal shape in the direction of the
loading. This shape is commonly referred to as the vertical rocking modal shape and has
vibratory motion in-plane to the vertically applied truck load. For sign support structures,
the vertical modal shape is close to the horizontal modal shape, and is generally around 1
to 3 Hz, which typically corresponds to the second modal shape for cantilever-type struc-
tures, and the third modal shape for bridge-type structures. Finite element software (i.e.,
SAP2000) can be used to estimate the appropriate modal shapes and their associated
natural frequencies to use with the SRS curves. If FEA software is not available, funda-
mental structural dynamics of a SDOF system (Eq. 14.5) can be used for estimating these
1 K
fn = [Eq. 14.5]
2π M
where
f n = natural frequency, Hz
K = generalized stiffness, N/m (lb/ft)
M = generalized mass, kg (slug)
354
A conservative approach would be to use the peak amplitude from the SRS that corre-
sponds to the truck speed regardless of the natural frequency of the structure.
A systematic procedure for determining the fatigue design load due to truck-
induced wind gusts based on the proposed detailed approach may be given as follows.
Step 1: design speed of the truck. The first step involves determining the design
speed of the truck. A speed of 70 mph (31.3 m/s) is recommended however other speeds
can be used if needed. It is not recommended to use speeds less than 70 mph (31.3 m/s).
Step 2: modal analysis. The second step involves a modal analysis of the design-
ing support structure. The earliest modal frequency with a modal shape in the direction of
the truck gust loading is needed. The damping ratio is not required for truck-induced
gusts. Finite element analysis software can be helpful in this step. If FEA is not available,
the modal frequency and modal shape can be estimated using Eq. 14.5.
Step 3: shock response spectrum. The third step involves the SRS shown in Figure
14.4. A SRS constant, PSRS, is extracted from the spectrum as the ordinate value corre-
Step 4: fatigue design equation. The next step is plugging all information gathered
from Steps 1-3 into the fatigue design equation for truck-induced gusts based on the de-
tailed approach. Simply input the SRS constant, PSRS, determined from Step 3 into Eq.
14.6 for each member along the facade of the structure exposed to truck-induced wind
gusts. It is to be applied in the same fashion as the general design equation for truck-
where
PTG = design fatigue load due to truck - induced wind gust, psf (Pa)
PSRS = transmitted pressure constant extracted from the SRS, psf (Pa)
C d = drag coefficient
I F = importance factor
values determined from Eq. 14.6 for structures higher than 19.7 ft (6.00 m). The truck
gust pressure is assumed maximum at a height of 19.7 ft (6.00 m) above the roadway, and
The FEA model of the support structure was used to double check the back-
calculations for natural wind and truck-induced wind gusts. It was also used to compare
dynamic loading to the equivalent static wind load gathered from the VRS and SRS.
Verification of the model was established by comparing the experimental modal analysis
with the FEA. Alterations as needed primarily concerned the connection of the structure
It was found that the back-calculation from the measured strain values to the wind
pressure load was accurate. The wind pressure developed from the experimental data was
inputted in the FEA as a uniformly distributed load applied to the exposed portion of the
structure to wind gusts. The strain at the instrumented location was extracted from the
linear-static solution and compared to the strain values measured experimentally. The re-
sults are listed in Table 14.3. The loading produced from the average wind velocity was
used in the natural wind comparison because experimental strain at the fatigue wind was
357
not measured. The results showed close agreement, indicating the structural analysis used
The FEA was also used to verify the use of the VRS and SRS as an accurate ap-
proach in determining the fatigue load based on the dynamic characteristics of the struc-
ture. The methods were based on approximating the dynamic behavior of the support
structure as a SDOF. Material properties of the model were changed and the analysis was
performed again. The results indicated the VRS and SRS were accurate in determining
The results from this study were seen as a positive step towards more reliable
loads used in the design of sign support structures. The theoretical program was devel-
oped with the intention for application with the variety of sign supports structures used in
design. An experimental program was produced to check the accuracy of this program.
The results indicate that the developed methodology can be used to design sign supports
358
structures using on more realistic design loads. The impact of the new criteria would
greatly enhance the design community to ensure the safety as well as economy in design.
also a major consideration. The VRS and SRS for natural wind and truck-induced wind
gusts respectively can be developed into a table format that can be easily and readily used
by the design engineer to determined design fatigue loads based on the particular struc-
ture in design. The table format would account with variability in sign supports structures
Addressing the accuracy of the theoretical program including the VRS and SRS
was the major recommendation for future research. The idea was to developed a universal
methodology that could be used for the variety of sign supports structures in design in-
cluding bridge-type sign supports structures as well as VMS support structures. As this
research proved, the methodology was accurate for application with cantilever-type sign
were to apply this method to the full variety of sign supports structures. The method was
believed to be applicable to any structure that can be approximated as a SDOF and is ex-
posed to natural wind and truck-induced wind loading. Research is needed to accurately
and scientifically prove that bridge structures and VMS structures actually fall into this
category.
359
LIST OF REFERENCES
7 Albert, M.N., Manuel, L., Frank, K.H., and Wood, S.L., Field Testing of Cantile-
vered Traffic Signal Structures under Truck-Induced Gust Loads. Report No.
FHWA/TX-08/0-4586-2, Center for Transportation Research, University of Texas
at Austin, 2007.
8 Azzam, D.; Fatigue Behavior of Highway Welded Aluminum Light Pole Support
Structures. Dissertation, University of Akron, May 2006.
360
9 Cali, P., and Covert, E.E., On the Loads on Overhead Sign Structures in Still Air
by Truck Induced Gusts. Wright Brothers Facility Report 8-97, Massachusetts In-
stitute of Technology.
11 CSI Analysis Reference Manual. Computers and Structures, Inc., Berkeley, Cali-
fornia, January, 2007.
12 Cook, Ronald A., Bloomquist, D., Agosta, A.M., Taylor, K.F., Wind Load Data
for Variable Message Signs. Report Number 0728-9488. Florida Department of
Transportation, Research Management Center, April, 1996.
15 Dexter, R. J., Johns, K. W. Development of Fatigue Design Load Ranges for Can-
tilevered Sign and Signal Support Structures, Journal of Wind Engineering and
Industrial Aerodynamics, 77&78, 1998, pp. 315-326.
16 DeSantis, P.V., and Haig, P.E., Unanticipated Loading Causes Highway Sign
Failure. Proceedings of ANSYS Convention, 1996.
17 Edwards, J.A., and Bingham, W.L. Deflection Criteria for Wind Induced Vibra-
tions in Cantilever Highway Sign Structures. Report No. FHWA/NC/84-001, Cen-
ter for Transportation Engineering Studies, North Carolina State University, 1984.
18 Fackler, W.C., Pusey, H.C., Volin, R.H., Schell, E.H. Equivalence Techniques for
Vibration Testing. The Shock and Vibration Information Center, Naval Research
laboratory, Code 6020, Washington D.C., 1972.
21 Fouad, Fouad H.; and Calvert, Elizabeth A., New Wind Design Criteria for Traffic
Signal Support Structures, UTCA Report Number 04219, University Transporta-
tion Center for Alabama, Tuscaloosa, AL (August 31, 2005) 49pp.,
http://utca.eng.ua.edu/projects/final_reports/04219fnl.pdf.
22 Fouad, Fouad H., and Calvert, Elizabeth. “Design of Cantilevered Overhead Sign
Supports,” Transportation Research Record, Issue No. 1928, Transportation Re-
search Board, Washington, D.C., 2005.
23 Fouad, Fouad H.; and Calvert, Elizabeth A., AASHTO 2001 Design of Overhead
Cantilevered Sign Supports, UTCA Report Number 02216, University Transporta-
tion Center for Alabama, Tuscaloosa, AL, March 30, 2004,
http://utca.eng.ua.edu/projects/final_reports/02216fnl.pdf.
24 Fouad, Fouad H.; Calvert, Elizabeth; Davidson, James S.; and Delatte, Norbert.
“Proposed Revisions to AASTHO 2001 Supports Specifications,” Compendium of
Papers CD-ROM, Transportation Research Board, 83rd Annual Meeting, Wash-
ington, DC, Jan. 11-15, 2004.
25 Fouad, Fouad H., and Calvert, Elizabeth. “Wind Load Provisions in the 2001
AASHTO Supports Specifications,” Transportation Research Record, Issue No.
1845, Transportation Research Board, Washington, D.C., 2003.
26 Fouad, Fouad H.; Davidson, James S.; Delatte, Norbert; Calvert, Elizabeth A.;
Chen, Shen-En; Nunez, Edgar; and Abdalla, Ramy, “Structural Supports for
Highway Signs, Luminaires, and Traffic Signals.” NCHRP Report 494, Transpor-
tation Research Board, Washington, D.C., 2003.
27 Fouad, Fouad H.; Davidson, James S.; Delatte, Norbert; Calvert, Elizabeth A.;
Chen, Shen-En; Nunez, Edgar; and Abdalla, Ramy, Structural Supports for High-
way Signs, Luminaires, and Traffic Signals: Draft Final Report, NCHRP Project
17-10(2), Prepared for National Cooperative Highway Research Program, Trans-
portation Research Board, National Research Council, The University of Alabama
at Birmingham, Birmingham, AL, May 2002.
28 Fouad, Fouad H. and Elizabeth Calvert. “Impact Of The New Wind Load Provi-
sions On The Design Of Structural Supports,” Compendium of Papers CD-ROM,
Transportation Research Board, 82nd Annual Meeting, Washington, DC, Jan. 12-
16, 2003.
29 Fouad, Fouad H. and Elizabeth Calvert. “Wind Load Provisions in the 2001 Sup-
ports Specifications,” Compendium of Papers CD-ROM, Transportation Research
Board, 82nd Annual Meeting, Jan. 12-16, 2003.
362
30 Fouad, Fouad H.; and Calvert, Elizabeth A., Evaluating the Design Safety of
Highway Structural Supports, UTCA Report Number 00218, University Transpor-
tation Center for Alabama, Tuscaloosa, AL, August 2001,
http://utca.eng.ua.edu/projects/final_reports/00218report.htm.
31 Fouad, Fouad H.; Nunez, Edgar; and Calvert, Elizabeth A. “Proposed Revisions
to AASHTO Standard Specifications for Structural Supports for Highway Signs,
Luminaires, and Traffic Signals,” Transportation Research Record, Issue No.
1656, Transportation Research Board, Washington, D.C., 1999.
32 Fouad, Fouad H.; Mehta, Kishor C.; and Calvert, Elizabeth A., Wind Loads Re-
port: Final Draft. NCHRP Project 17-10(2), Prepared for National Cooperative
Highway Research Program, Transportation Research Board, National Research
Council, The University of Alabama at Birmingham, Birmingham, AL, Sept.
1999.
33 Fouad, Fouad H.; Calvert, Elizabeth A.; and Nunez, Edgar, “Structural Supports
for Highway Signs, Luminaires and Traffic Signals.” NCHRP Report 411, Trans-
portation Research Board, Washington, D.C., 1998.
34 Fouad, Fouad H.; Calvert, Elizabeth A.; and Nunez, Edgar, Structural Supports for
Highway Signs, Luminaires, and Traffic Signals: Draft Final Report. NCHRP
Project No. 17-10, Department of Civil and Environmental Engineering, The Uni-
versity of Alabama at Birmingham, Birmingham, AL, Sept. 1997.
35 Fouad, Fouad H.; Calvert, Elizabeth A.; and Nunez, Edgar, Structural Supports for
Highway Signs, Luminaires, and Traffic Signals: Draft Final Specification.
NCHRP Project No. 17-10, Department of Civil and Environmental Engineering,
The University of Alabama at Birmingham, Birmingham, AL, Sept. 1997.
36 Gilani, A., and Whittaker, A., “Fatigue-Life Evaluation of Steel Post Structures I:
Background and Analysis.” Journal of Structural Engineering, ASCE, New York,
NY, 2000.
46 Irvine, T. Enveloping Data Via the Vibration Response Spectrum. Vibration Data
Publications, October 11, 1999. www.vibrationdata.com.
48 Irvine, T. Equivalent Static Loads for Random Vibration. Vibration Data Publica-
tions, March 13, 2008. www.vibrationdata.com.
51 James, M.L., Smith, G.M., Wolford, J.C., Whaley, P.W., Vibration of Mechanical
and Structural Systems: With Microcomputer Applications. Harper & Row, Pub-
lishers, Inc., 1989.
52 Kaczinski, M. R., Dexter, R. J., and Van Dien, J. P. Fatigue Resistant Design of
Cantilevered Sign, Signal and Light Supports, NCHRP Report 412, Transportation
Research Board, Washington, D.C., 1998.
364
53 Kashar, L., Nester, M.R., Johns, J.W., Hariri, M., and Freizner, S., Analysis of the
Catastrophic Failure of the Support Structure of a Changeable Message Sign.
Structural Engineering in the 21st Century, Proceedings of the 1999 Structures
Congress, New Orleans, LA, 1115-118, 1999.
55 McDonald, J.R., Mehta, K.C., Oler, W., and Pulipaka, N., Wind Load Effects on
Signs, Luminaires and Traffic Signal Structures. Texas Department of Transpor-
tation Report No. 1303-1F, Wind Engineering Research Center-Texas Tech Uni-
versity, Lubbaock, TX, 1995.
56 Paz, M. Structural Dynamics: Theory and Computation, 4th Ed. Chapman & Hall,
1997.
62 Walpole, R.E., Myers, R.H., Myers, S.L., Ye, K., Probability & Statistics for En-
gineers& Scientists, 7th Ed. Prentice Hall, Inc., 2002.
63 Yang, C. Y. Random Vibration of Structures. John Wiley & Sons, Inc., 1986.
APPENDIX A
A detail of the instrumentation layout and identification are provided. They were used to
Strain Gauges
367
368
369
370
371
Accelerometers
372
Anemometers
373
374
APPENDIX B
The support tube requires three 3 equally spaced holes, tapped M5, 7.5 mm from
Pass the cable (fitted with the 9 way Clipper plug) through the tube.
Note: the customer must fit appropriate strain relief to the cable.
Connect the plug by twisting it whilst pushing it gently into the socket on the
WindSonic.
When it locates, twist the outer sleeve clockwise to connect and lock the plug.
Fix the WindSonic to the tube using the 3 stainless steel screws provided (Maxi-
a position clear of any structure, which may obstruct the airflow or induce turbu-
lence.
transmitters. A site survey may be required if there is any doubt about the strength
APPENDIX C
Stud Mounting
The accelerometer is fixed to the test surface by means of a threaded metal screw
In preparing the stud mounting, the test surface must be drilled and tapped
Distortion of the accelerometer as mounted may produce strains that affect the ac-
To ensure that the test surface is very flat (which can be done by grinding or lap-
ping)
To prevent the mounting stud from bottoming in the transducer case—this can
lead to strain
To screw the accelerometer onto the test surface using the torque recommended
by the manufacturer
post support), where the structural surface is rounded, a solid mounting block can
be fabricated which is rounded to this same contour on one side and flat on the
The size of the mounting plate (made of steel preferably) should be as small as
possible, with enough space to accurately and securely mount to the web member,
Location A
Location B
Location C
Location D
379
Location A
Line y
Line x
1 2
Accelerometer 1 1
Accelerometer 2 2
Accelerometer placed as close as possible to the center of the truss mounting plate
(next to accelerometer 1)
Location B and C
Line z
Line y
5
Line x
Accelerometers mounted to flat surface of the mounting plate (with one side fab-
Accelerometer 3 3
Accelerometer 3 placed at the mid-span of web member (in line with Accelerome-
ters 1 and 2)
Accelerometer 4 4
Accelerometer 4 placed at the mid-span of web member (in line with accelerome-
ters 1 and 2)
Accelerometer 5 5
Accelerometer 5 placed at the mid-span of the web member (in line with acceler-
ometers 1 and 2)
Location D
Line z
2 ft
Accelerometer 6 6
Accelerometer 6 mounted 2 ft from top of post along the centerline of the post
Accelerometer mounted to flat surface of the mounting plate (with one side
APPENDIX D
The ranking schedule used between UAB and ALDOT for determining appropri-
APPENDIX E
Loop Instructions
1. Continue on I-65 North (over Tennessee River) and take Exit 340: 6 miles
2. Get in right lane and Travel Underneath Sign at Required Speed (see fig-
ure below)
3. After passing underneath sign, continue straight (do not take ramp) and
5. Take left after exiting and travel over bridge crossing I-65
7. Continue on I-65 South (past sign) over Tennessee River and take Exit
334: 17 miles
Nashville
Sign
Decatur
Huntsville
North
6. Call ALDOT Daniel when arriving at Exit 334 (call once for OK, twice to answer
because of problem)
8. Call ALDOT Daniel when arriving at the beginning of Tennessee bridge (call
13. Call ALDOT Daniel when leaving Exit 334 (call once for OK, twice to answer
because of problem)
14. Call ALDOT Daniel when arriving at the beginning of Tennessee bridge (call