Measurement Based Performance Evaluation of

a Segmental Concrete Bridge

by

Rodney Tod Davis, B. S., M. S.

Dissertation
Presented to the Faculty of the Graduate School of

The University of Texas at Austin

in Partial Fulfillment

of the Requirements

for the Degree of

Doctor of Philosophy

The University of Texas at Austin

 May, 1999

Measurement Based Performance Evaluation of

a Segmental Concrete Bridge

Approved by
Dissertation Committee:

Supervisor: John E. Breen

Michael E. Kreger

Ned H. Burns

James O. Jirsa
 Eric B. Becker

Dedication

This dissertation is dedicated to my family in Virginia.

 Acknowledgements

I would like to thank Professor Breen for the opportunity to conduct the research on the
U. S. 183 Segmental. His unequaled insight into the various topics studied, and firm grasp on the
scope of this project and my dissertation was invaluable when I occasionally became bogged down
in the numbers. I would also like to thank the sponsors of the research: The Texas Department of
Transportation and the Federal Highway Administration. I enjoyed working with all the TxDOT
engineers and field personnel, but special thanks goes to Tom Rummel and Dean Van Landuyt.
They are two very talented TxDOT engineers that provided assistance to the researchers
throughout the project. My thanks goes to everybody at Eby Construction. They obviously are a
first class company with high quality personnel, and I greatly enjoyed watching them work. Ron
Fletcher, Mike Pettit, Donnie Lamb, Joe Lee, Randy Svilar and David Patrick all made sure our
schedules were identical. The daily help from all the Eby construction workers was invaluable,
and greatly contributed to the success of the research.
The information in this dissertation could not have been reduced to a presentable format
within the time permitted for the research project without the aid of the entire U. S. 183 research
staff. My thanks goes to Valerie Andres for her work on the mainlane Y-pier, to Wade Bonzon for
his work on the segmental pier, and to Bryan Wood for his work on the mainlane thermal study. I
would also like to thank Bryan for attending every emergency field meeting held at Granger Lake.
Special thanks goes to Keith Thompson for his work on the balanced cantilever superstructure.
Keith was the last and only Master’s student working at the end of the research project. The
amount of work he successfully completed was most impressive. These four Master’s students
have published theses on each of their respective topics in greater depth than I could present in this
dissertation. More than a few sleepless nights, seven day weeks, and scorching hot days had to be
endured by these folks to install and monitor the twelve-hundred gauges on this project.

iv
 Occasionally live load tests required a major expansion of our research team with
“volunteers.” Thank you Jeff West, Andrea Schokker, Sarah Billington, Brad Wood, Brad
Koester, Chris Gilcrist, Reagan Herman, Mike Smart, Michele Parsley, Pete Schonwetter, Paul
Ziehl, Chad Corsten, and David Jauregui for the free labor. I also would to thank my
undergraduate and graduate assistants on the project who soon realized that my expected level of
workmanship was not a function of the amount of time we had to do the work. I hope Mohammed
Najah, Scott Listavich, Steve Smith, and Chris Breen took a few fond memories of the project
with them.
I would like to thank the entire staff at the Ferguson Laboratory. Laurie Golding, April
Jenkins, Denise Nealy, Ruth Goodson, Regina Forward and Sharon Cunningham provided much
needed assistance from the office. Blake Stasney, Wayne Fontenot, Ray Madonna, Wayne Little,
Pat Ball, Mike Bell, and Ryan Green were a great help on the lab floor and computer room.
I would also like to thank Campbell Scientific for the steadfast and flawless performance
of their dataloggers and other equipment. The people from Campbell Scientific were also a great
help, answering any and all questions put forth. Lastly, I would like to thank the laboratory’s
1974 Ford Pickup for its faithful service on the project. Only once did we simultaneously run out
of gas, get stuck in the mud, and have battery failure, not necessarily in that order. Other than that
incident, I can’t complain about the truck’s performance.

v
 Measurement Based Performance Evaluation of
a Segmental Concrete Bridge

Publication No._____________

Rodney Tod Davis, Ph.D.

The University of Texas at Austin, 1999

Supervisor: John E. Breen

Construction of the U. S. 183 segmental viaduct in Austin, Texas provided the

opportunity to study many topics in need of research, mostly for segmental box girder design and
construction. Field instrumentation and testing of various elements of the structure, including two
piers and four superstructure multi-span units, provided a large amount of data for evaluation.
Most of these measured results were presented and evaluated in this dissertation. Topics under
study included post-tensioning force losses, thermal gradients and their effects, live load response,
behavior of local zones and other discontinuity zones, the diffusion of post-tensioning forces, and
an evaluation of semi-continuous multi-span bridge units. Recommendations were made for
improved design and construction practices.

vi
 Table of Contents
CHAPTER 1 INTRODUCTION 1

1.1 THE U.S. 183 STUDY ........................................................................................................... 1

1.2 PURPOSE FOR THE STUDY............................................................................................... 7

1.2.1 Structural performance and safety ............................................................................... 7
1.2.2 Constructability problems ............................................................................................ 8
1.2.3 Initial cost of future bridges......................................................................................... 8
1.2.4 Maintenance problems ................................................................................................. 8

1.3 PROJECT OBJECTIVES....................................................................................................... 9

1.3.1 Identify and study major design uncertainties ............................................................. 9
1.3.2 Instrument and test structural elements ....................................................................... 9
1.3.3 Interpret measurements................................................................................................ 9
1.3.4 Make recommendations to TxDOT and FHWA ........................................................ 10

1.4 STRUCTURAL ELEMENTS UNDER STUDY ................................................................. 10

1.4.1 Mainlane pier ............................................................................................................. 15
1.4.2 Mainlane girder.......................................................................................................... 19
1.4.3 Ramp P girder ............................................................................................................ 28
1.4.4 Large ramp pier ......................................................................................................... 33
1.4.5 Transition and modified spans ................................................................................... 45

1.5 TOPICS STUDIED ON THE U.S. 183 SEGMENTAL PROJECT ..................................... 47

1.5.1 Friction losses in post-tensioning tendons ................................................................. 47
1.5.2 Thermal gradients and their effects ........................................................................... 48
1.5.3 Strut-and-tie design regions ....................................................................................... 49
1.5.4 Stress distributions in box girders.............................................................................. 49
1.5.5 Behavior of semi-continuous multiple span units ...................................................... 50

vii
1.6 SCOPE AND CHAPTER ORGANIZATION ..................................................................... 51

CHAPTER 2 INSTRUMENTATION PROGRAM 53

2.1 GENERAL APPROACH TO MEASUREMENTS ............................................................. 53

2.2 INSTRUMENTATION ........................................................................................................ 55

2.2.1 Strain gauges ............................................................................................................. 55
2.2.2 Thermocouples .......................................................................................................... 63
2.2.3 Linear potentiometers ................................................................................................ 64
2.2.4 Pressure transducers .................................................................................................. 65
2.2.5 Load cells................................................................................................................... 66
2.2.6 Data Acquisition system ............................................................................................ 66
2.2.7 Demec extensometer.................................................................................................. 73
2.2.8 Deflection measurements .......................................................................................... 75
2.2.9 Tiltmeter .................................................................................................................... 77
2.2.10 Epoxy sleeve system .................................................................................................. 78

2.3 INSTRUMENT LOCATIONS............................................................................................. 80

2.3.1 Mainlane pier ............................................................................................................. 80
2.3.2 Three-span mainlane box girder unit ......................................................................... 84
2.3.3 Segmental large ramp pier ......................................................................................... 91
2.3.4 Five-span ramp box girder unit.................................................................................. 98
2.3.5 Multiple-celled box girder units .............................................................................. 107

2.4 MATERIAL PROPERTY TESTS ..................................................................................... 109

CHAPTER 3 PRESTRESSING FORCE LOSSES IN POST-TENSIONING TENDONS 115

3.1 PROBLEM DESCRIPTION AND TECHNICAL BACKGROUND ................................ 115

3.2 LITERATURE REVIEW AND FIELD EXPERIENCE .................................................... 120

3.2.1 Friction losses .......................................................................................................... 120
3.2.2 Elastic shortening losses .......................................................................................... 126
3.2.3 Creep and shrinkage ................................................................................................ 127

viii
3.3 INSTRUMENTATION PROGRAM ................................................................................. 130
3.3.1 Bench test ................................................................................................................ 130
3.3.2 Mainlane girder tests ............................................................................................... 133
3.3.3 Ramp girder tests ..................................................................................................... 140
3.3.4 Large ramp pier tests ............................................................................................... 145

3.4 DISCUSSION OF RESULTS ............................................................................................ 149

3.4.1 Friction losses .......................................................................................................... 149
3.4.2 Elastic shortening .................................................................................................... 155
3.4.3 Long term losses ...................................................................................................... 157

3.5 CONCLUSIONS ................................................................................................................ 158

CHAPTER 4 THERMAL GRADIENTS AND THEIR EFFECTS 161

4.1 PROBLEM DESCRIPTION AND TECHNICAL BACKGROUND ................................ 161

4.1.1 Superstructure behavior and analysis techniques .................................................... 162
4.1.2 Substructure behavior and analysis techniques........................................................ 173

4.2 PREVIOUSLY MEASURED GRADIENTS AND DESIGN GRADIENTS .................... 177

4.2.1 Hoffman, McClure and West [33] ........................................................................... 177
4.2.2 Hirst and Dilger [34]................................................................................................ 178
4.2.3 Shui [35] .................................................................................................................. 178
4.2.4 Priestley [36] ........................................................................................................... 179
4.2.5 Roberts, Breen, Kreger [8]....................................................................................... 179
4.2.6 Arockiasamy and Reddy [37] .................................................................................. 180
4.2.7 Pentas, Avent, Gopu, and Rebello [38] ................................................................... 180
4.2.8 Potgieter and Gamble [39] ....................................................................................... 181
4.2.9 NCHRP 276 [40] ..................................................................................................... 182
4.2.10 AASHTO LRFD [10] .............................................................................................. 184

4.3 DATA COLLECTION AND ANALYSIS......................................................................... 185

4.3.1 Ramp P superstructure ............................................................................................. 185
4.3.2 Mainlane superstructure .......................................................................................... 207
4.3.3 Large ramp pier P16 ................................................................................................ 228

ix
 4.3.4 Mainlane pier D6 ..................................................................................................... 251

4.4 DISCUSSION OF RESULTS ............................................................................................ 257

4.4.1 Superstructure thermal gradients for longitudinal analysis...................................... 259
4.4.2 Superstructure thermal gradients for transverse design ........................................... 266
4.4.3 Pier gradients ........................................................................................................... 270

4.5 CONCLUSIONS ................................................................................................................ 274

4.5.1 Superstructure gradients for longitudinal design ..................................................... 274
4.5.2 Thermal gradients for transverse design .................................................................. 276
4.5.3 Thermal gradients for the design of piers ................................................................ 277

CHAPTER 5 LOAD RESPONSE OF BOX GIRDERS 279

5.1 INTRODUCTION .............................................................................................................. 279

5.2 PROBLEM AND TECHNICAL BACKGROUND ........................................................... 279

5.2.1 Structural response of box girders to dead load and live load ................................. 280
5.2.2 Diffusion of post-tensioning forces ......................................................................... 282
5.2.3 Behavior of multiple-cell box girders ...................................................................... 283

5.3 TYPICAL SOLUTIONS, DESIGN CODE METHODS, AND PREVIOUS

MEASUREMENTS ........................................................................................................... 285
5.3.1 Analysis of box girder structures ............................................................................. 286
5.3.2 Design code methods of analysis ............................................................................. 290
5.3.3 Previous measurements ........................................................................................... 295

5.4 DATA COLLECTION, ANALYSIS AND RESULTS ..................................................... 296

5.4.1 Mainlane girder........................................................................................................ 296
5.4.2 Ramp P girder .......................................................................................................... 323
5.4.3 Two-cell girder - Unit C15, L2 ................................................................................ 342
5.4.4 Three-cell girder - Unit C13 .................................................................................... 351
5.4.5 Large ramp pier ....................................................................................................... 361

5.5 DISCUSSION OF RESULTS ............................................................................................ 366

5.5.1 Diffusion of post-tensioning .................................................................................... 366
5.5.2 Response to dead loads and live loads ..................................................................... 368

x
 5.5.3 Behavior of multiple-cell box girders ...................................................................... 369

5.6 CONCLUSIONS ................................................................................................................ 371

5.6.1 Temporary post-tensioning for epoxying segments ................................................. 371
5.6.2 Diffusion of post-tensioning forces ......................................................................... 372
5.6.3 Girder response to dead loads and live loads ........................................................... 372
5.6.4 Performance of multiple-cell girders under live load .............................................. 373
5.6.5 Performance of the segmental pier under bending................................................... 373

CHAPTER 6 BEHAVIOR OF DISCONTINUITY (OR D-) ZONES 375

6.1 PROBLEM AND TECHNICAL BACKGROUND ............................................................... 375

6.2 PREVIOUS STUDIES ........................................................................................................... 381

6.3 DATA COLLECTION, ANALYSIS, AND DICUSSION OF RESULTS ............................ 386

6.3.1 Mainlane pier............................................................................................................ 387
6.3.2 Mainlane deviator segment....................................................................................... 394
6.3.3 Mainlane anchor segment ......................................................................................... 403
6.3.4 Large ramp pier capital............................................................................................. 417
6.3.5 Ramp interior anchor segment.................................................................................. 423
6.3.6 Ramp deviator segment ............................................................................................ 431
6.3.7 Ramp anchorage blister segment .............................................................................. 440

6.4 RECOMMENDATIONS AND CONCLUSIONS ................................................................. 449

6.4.1 Pier capitals .............................................................................................................. 449
6.4.2 Deviators .................................................................................................................. 450
6.4.3 Pier segments............................................................................................................ 451
6.4.4 Anchorage blister ..................................................................................................... 452

xi
CHAPTER 7 BEHAVIOR OF A SEMI-CONTINUOUS UNIT 453

7.1 INTRODUCTION ................................................................................................................. 453

7.2 PROBLEM AND TECHNICAL BACKGROUND ............................................................... 457

7.3 PREVIOUS STUDIES ........................................................................................................... 459

7.4 DATA COLLECTION AND ANALYSIS ............................................................................ 462

7.4.1 Live load tests........................................................................................................... 462
7.4.2 Temperature gradient effects .................................................................................... 478

7.5 CONCLUSIONS AND RECOMMENDATIONS ................................................................. 483

7.5.1 Construction procedures ........................................................................................... 483
7.5.2 Cast-in-place joint behavior under thermal loads ..................................................... 484

CHAPTER 8 SUMMARY, RECOMMENDATIONS AND CONCLUSIONS 485

8.1 PROJECT SUMMARY ......................................................................................................... 485

8.2 GENERAL CONCLUSIONS ................................................................................................ 486

8.2.1 Post-tensioning losses............................................................................................... 486
8.2.2 Thermal gradients ..................................................................................................... 487
8.2.3 Load response of box girders ................................................................................... 487
8.2.4 Behavior of D-zones ................................................................................................. 488
8.2.5 Cast-in-place joint .................................................................................................... 488

8.3 SPECIFIC CONCLUSIONS AND RECOMMENDATIONS ............................................... 488

8.3.1 Post-tensioning losses............................................................................................... 488
8.3.2 Thermal gradients ..................................................................................................... 491
8.3.2.1 Superstructure gradients for longitudinal design ........................................ 491
8.3.2.2 Thermal gradients for transverse design ..................................................... 493
8.3.2.3 Thermal gradients for the design of piers ................................................... 494
8.3.3 Load response of box girders ................................................................................... 495
8.3.3.1 Temporary post-tensioning for epoxying segments .................................... 495
8.3.3.2 Diffusion of post-tensioning forces ............................................................ 496
8.3.3.3 Girder response to dead loads and live loads .............................................. 497

xii
 8.3.3.4 Performance of multiple-cell girders under live load ................................. 497
8.3.3.5 Performance of the segmental pier under bending...................................... 497
8.3.4 Behavior of D-zones ................................................................................................. 498
8.3.4.1 Mainlane Y-pier .......................................................................................... 498
8.3.4.2 Segmental pier P16 capital ......................................................................... 498
8.3.4.3 Deviators..................................................................................................... 499
8.3.4.4 Mainlane anchor segment D5-16 ................................................................ 499
8.3.4.5 Ramp P anchor segment P16-1 ................................................................... 500
8.3.4.6 Anchorage blister ........................................................................................ 500
8.3.5 Cast-in-place joint behavior ..................................................................................... 501
8.3.5.1 Construction procedures ............................................................................. 501
8.3.5.2 Cast-in-place joint behavior under live loads and thermal loads ................ 502

8.4 OBSERVATIONS ON CONSTRUCTABILITY AND DESIGN ......................................... 502

8.4.1 Mainlane Y-pier ....................................................................................................... 502
8.4.2 Mainlane girder ........................................................................................................ 503
8.4.3 Segmental Pier P16 .................................................................................................. 505
8.4.4 Ramp P constructed in balanced cantilever .............................................................. 506
8.4.5 Transition spans........................................................................................................ 508
8.4.6 Modified spans ......................................................................................................... 509

8.5 RECOMMENDATIONS FOR FUTURE STUDIES............................................................. 509

8.5.1 Instrumentation......................................................................................................... 509
8.5.2 Future research topics ............................................................................................... 511

APPENDIX 513

REFERENCES 515

VITA 519

xiii
 List of Tables
Table 1.1 Project bid summary ........................................................................................................ 3

Table 1.2 Unit construction costs ..................................................................................................... 4

Table 2.1 Coefficients of thermal expansion................................................................................ 112

Table 2.2 Modulus of rupture test results ..................................................................................... 114

Table 3.1 Friction and wobble coefficients for post-tensioned tendons from
AASHTO [24] .................................................................................................................... 123

Table 3.2 Friction and wobble coefficients for post-tensioned tendons from AASHTO
LRFD [10] .......................................................................................................................... 123

Table 3.3 Friction and wobble coefficients for post-tensioned tendons from CEB [25] .............. 124

Table 3.4 Friction and wobble coefficients for post-tensioned tendons from ACI 318-95 [26] .. 125

Table 3.5 Friction and wobble coefficients for post-tensioned internal tendons from PTI [27] .. 125

Table 3.6 Friction and wobble coefficients for post-tensioned internal tendons from
ACI-ASCE [28] .................................................................................................................. 126

Table 3.7 Bench test results ......................................................................................................... 131

Table 3.8 In-place friction test results on span D2 ....................................................................... 135

Table 3.9 Measured losses and calculated wobble coefficients from in-
place friction test on span D2 ............................................................................................. 136

Table 3.10 Elongations from friction tests on span D2 ................................................................ 137

Table 3.11 Span D5 friction test results from strain gauges......................................................... 138

Table 3.12 Measured elongations from friction tests on span D5 ................................................ 139

Table 3.13 Elastic shortening losses in tendons in span D5 ......................................................... 140

Table 3.14 Ramp P friction test results from strain gauges .......................................................... 142

Table 3.15 Measured elongations from friction tests on Ramp P ................................................ 143

Table 3.16 Lift-off test results for tendon T3 ............................................................................... 143

Table 3.17 Elastic shortening losses–Ramp P external tendons ................................................... 144

xiv
Table 3.18 Elongation measurement based wobble coefficients for all
deviated external tendons, μ=0.25 ...................................................................................... 152

Table 3.19 Elongation measurement based wobble angles for deviated

external tendons, μ=0.35 .................................................................................................... 153

Table 3.20 Calculated and measured elastic shortening losses in tendons in span D5................. 156

Table 3.21 Calculated and measured elastic shortening losses–Ramp P external tendons .......... 157

Table 4.1 Mainlane girder D5 transverse thermal strains from the

May 20, 1995 positive gradient .......................................................................................... 227

Table 4.2 Summary of maximum top fiber gradient magnitudes ................................................. 260

Table 5.1 Axle weights and spacing for live load test trucks on Unit D2 .................................... 312

Table 5.2 Axle weights and spacing for live load test trucks on Ramp P .................................... 336

Table 5.3 Axle weights and spacing for live load test trucks on Units C15 and L2 ..................... 343

Table 5.4 Axle weights and spacing for live load test trucks on Unit C13 .................................. 352

Table 5.5 Summary of live load test results for multiple-cell girders .......................................... 370

Table 6.1 Summary of tendon forces at various limit states for deviators
tested by Beaupre et al [55] ................................................................................................ 386

Table 6.2 Measured strains in anchor segment D5-16 reinforcing bars from
stressing post-tensioning tendons in microstrain ................................................................ 408

Table 6.3 Measured strains in anchor segment D5-16 reinforcing bars from
stressing post-tensioning tendons (continued) in microstrain............................................. 410

Table 6.4 Measured strain changes in segment PC16-8 gauges from post-tensioning forces ...... 421

Table 6.5 Measured strain changes in segment PC16-8 gauges from placement of
superstructure segment P16-17........................................................................................... 422

Table 6.6 Measured strain changes in segment P16-1 gauges from placement of
superstructure segment P16-17........................................................................................... 426

Table 6.7 Measured strain changes in segment P16-1 gauges from superstructure
post-tensioning forces in microstrain ................................................................................. 427

Table 6.8 Measured strain changes in segment P16-10 gauges from

post-tensioning forces in microstrain ................................................................................. 434

xv
Table 6.9 Measured strain changes in segment P16-10 gauges from live load case 2 ................. 440

Table 6.10 Measured strain changes in segment P16-4 anchorage blister

gauges from post-tensioning of tendon T22 ....................................................................... 442

Table 6.11 Element forces from P16-4 blister STM .................................................................... 447

Table 7.1 Axle weights and spacing for live load test trucks on Unit D2 .................................... 463

Table 7.2 Axle weights and spacing for live load test trucks on Unit C13 .................................. 471

Table 7.3 Axle weights and spacing for live load test trucks on Units C15 and L2 ..................... 476

xvi
 List of Figures
Figure 1.1 Location map .................................................................................................................. 1

Figure 1.2 Span by span construction .............................................................................................. 2

Figure 1.3 Mainlane structure .......................................................................................................... 4

Figure 1.4 The San Antonio Y ......................................................................................................... 5

Figure 1.5 U.S. 183 North Ramp P constructed in balanced cantilever ........................................... 6

Figure 1.6 The AASHTO Segmental Guide Specification [9] ....................................................... 10

Figure 1.7 Mainlane superstructure section ................................................................................... 11

Figure 1.8 Ramp superstructure section ......................................................................................... 12

Figure 1.9 Cast-in-place transition spans ....................................................................................... 13

Figure 1.10 Pier types .................................................................................................................... 13

Figure 1.11 Project precasting plant............................................................................................... 14

Figure 1.12 Casting machines ........................................................................................................ 15

Figure 1.13 Mainlane pier .............................................................................................................. 16

Figure 1.14 Mainlane pier elevation .............................................................................................. 17

Figure 1.15 Mainlane pier details................................................................................................... 18

Figure 1.16 Construction of mainlane pier capital ......................................................................... 19

Figure 1.17 Mainlane girder........................................................................................................... 20

Figure 1.18 Construction of a mainlane girder............................................................................... 21

Figure 1.19 Construction of a cast-in-place deck joint................................................................... 21

Figure 1.20 Cross section and elevation–Span D5 ......................................................................... 22

Figure 1.21 Casting machine.......................................................................................................... 23

Figure 1.22 Casting machine details .............................................................................................. 24

xvii
Figure 1.23 Geometry control of the casting machine ................................................................... 24

Figure 1.24 Transporting segments ................................................................................................ 25

Figure 1.25 Locating segments on the erection trusses .................................................................. 26

Figure 1.26 Epoxying segments ..................................................................................................... 27

Figure 1.27 Stressing a pair of mainlane girder external tendons .................................................. 28

Figure 1.28 Plan of U.S. 183 at IH35 ............................................................................................. 29

Figure 1.29 Construction sequence of Ramp P .............................................................................. 30

Figure 1.30 End view of balanced cantilever ramp under construction ......................................... 31

Figure 1.31 Ramp P anchor segment dimensions .......................................................................... 32

Figure 1.32 Ramp P typical segment dimensions .......................................................................... 32

Figure 1.33 Typical large ramp pier: elevation view ..................................................................... 34

Figure 1.34 Typical sections: large ramp pier ................................................................................ 35

Figure 1.35 Pier segment casting machine ..................................................................................... 37

Figure 1.36 Typical segment-per-day casting sequence................................................................. 38

Figure 1.37 Placement of outer pier segment form ........................................................................ 40

Figure 1.38 Casting bed with steel support frame for formwork ................................................... 40

Figure 1.39 Hydraulic jacks used to align the bulkhead pier segment ........................................... 41

Figure 1.40 U-shaped tendon ducts located in pier foundation ...................................................... 42

Figure 1.41 Steel support frame for first precast segment.............................................................. 43

Figure 1.42 Cast-in-place base forming a rigid moment connection between

base segment PC16-1 and foundation .................................................................................. 44

Figure 1.43 Hydraulic ram stressing temporary post-tensioning bars ............................................ 45

Figure 1.44 Plan and section of transition Unit C13 ...................................................................... 46

Figure 1.45 Plan and section of modified Unit C15-L2 ................................................................. 47

Figure 1.46 External tendon layouts for the U.S. 183 box girders ................................................. 48

xviii
Figure 1.47 Structural response to thermal gradients ..................................................................... 48

Figure 1.48 Strut-and-tie model for diffusion of a point load ........................................................ 49

Figure 1.49 Diffusion of post-tensioning forces ............................................................................ 50

Figure 1.50 Construction of a cast-in-place continuity joint .......................................................... 51

Figure 2.1 Gauge installation ......................................................................................................... 54

Figure 2.2 Vibrating concrete near instrumentation ....................................................................... 55

Figure 2.3 Electronic strain gauge ................................................................................................. 56

Figure 2.4 Wheatstone Bridge circuit ............................................................................................ 56

Figure 2.5 “S” series gauges installed on rebar .............................................................................. 57

Figure 2.6 “S” series gauges installed on prestressing tendon ....................................................... 58

Figure 2.7 “S” series gauges installed on structural pipe ............................................................... 58

Figure 2.8 “C” series gauge construction ....................................................................................... 59

Figure 2.9 C-gauge tested against the Demec extensometer .......................................................... 61

Figure 2.10 C-gauge calibration ..................................................................................................... 62

Figure 2.11 Installation of C-gauges .............................................................................................. 63

Figure 2.12 Thermocouple circuit .................................................................................................. 64

Figure 2.13 Installation of a thermocouple .................................................................................... 64

Figure 2.14 Linear potentiometer schematic .................................................................................. 65

Figure 2.15 Pressure transducer schematic .................................................................................... 65

Figure 2.16 Load cell schematic .................................................................................................... 66

Figure 2.17 Data acquisition systems used on the project.............................................................. 67

Figure 2.18 Wiring the AM416 Multiplexer to the 21X Datalogger.............................................. 68

Figure 2.19 Wiring diagram for 350Ω strain gauges including the Wheatstone circuit ................. 69

Figure 2.20 Concrete specimen and 21X datalogger temperatures recorded during
temperature effects test ......................................................................................................... 70

xix
Figure 2.21 C-gauge output recorded during temperature effects test ........................................... 70

Figure 2.22 Wiring diagram for thermocouples ............................................................................. 71

Figure 2.23 Wiring diagram for linear potentiometers ................................................................... 72

Figure 2.24 Wiring diagram for pressure transducers and load cells ............................................. 73

Figure 2.25 The Demec extensometer............................................................................................ 74

Figure 2.26 Demec extensometer schematic .................................................................................. 74

Figure 2.27 Piano wire deflection measuring system schematic .................................................... 76

Figure 2.28 Taking deflection measurements ................................................................................ 76

Figure 2.29 GPS deflection measurement equipment .................................................................... 77

Figure 2.30 Tiltmeter gauge mounting plate .................................................................................. 78

Figure 2.31 Epoxy sleeve strain measurement system ................................................................... 79

Figure 2.32 Epoxy sleeves tested during the bench test ................................................................. 80

Figure 2.33 Elevation of Y-pier with gauge locations ................................................................... 82

Figure 2.34 Y-pier sections showing gauge locations .................................................................... 83

Figure 2.35 Plan view of Y-pier showing gauge locations ............................................................ 84

Figure 2.36 Elevation of mainlane girder D5 with deflection measurement locations .................. 85

Figure 2.37 Tendon gauge locations for mainlane girder D5 ......................................................... 85

Figure 2.38 Strain gauge locations for segment D5-9 .................................................................... 86

Figure 2.39 Thermocouple locations for segment D5-9 ................................................................. 86

Figure 2.40 Longitudinal strain gauge locations for segment D5-16 ............................................. 87

Figure 2.41 Strain gauge locations for deviator segment D5-12 .................................................... 88

Figure 2.42 Strain gauge locations for anchor segment D5-16 ...................................................... 89

Figure 2.43 Linear potentiometer locations at pier D6................................................................... 90

Figure 2.44 Strain gauge locations for the cast-in-place joint over pier D6 ................................... 90

xx
Figure 2.45 Elevation of large ramp pier P16 with segment identification .................................... 92

Figure 2.46 Segment PC16-1 sections with gauge locations ......................................................... 93

Figure 2.47 Segment PC16-5 sections with gauge locations ......................................................... 93

Figure 2.48 Segment PC16-7 sections with gauge locations ......................................................... 94

Figure 2.49 Segment PC16-8 top C-gauge locations ..................................................................... 95

Figure 2.50 Segment PC16-8 bottom C-gauge locations ............................................................... 96

Figure 2.51 Segment PC16-8 S-gauge locations on rebar .............................................................. 97

Figure 2.52 Segment PC16-8 S-gauge locations on tie-down bars ................................................ 98

Figure 2.53 Elevation of Ramp P, span P16, with segment identification ..................................... 99

Figure 2.54 Ramp P tendon gauge locations .................................................................................. 99

Figure 2.55 Anchor segment P16-1A section with gauge locations............................................. 100

Figure 2.56 Segment P16-2 sections with gauge locations .......................................................... 101

Figure 2.57 Deviator segment P16-10 sections with gauge locations .......................................... 102

Figure 2.58 Segment P16-17 sections with strain gauge locations .............................................. 103

Figure 2.59 Segment P16-17 section with thermocouple locations ............................................. 104

Figure 2.60 Deviator segment P16 section with S-gauge locations on rebar ............................... 105

Figure 2.61 Segment P16-4 anchorage blister strain gauge locations .......................................... 106

Figure 2.62 Plan and section of transition spans showing instrumentation locations .................. 107

Figure 2.63 Plan and section of modified spans showing instrumentation locations ................... 108

Figure 2.64 Concrete modulus of elasticity test specimen ........................................................... 109

Figure 2.65 Modulus of elasticity test results for Mainlane pier D6 ............................................ 110

Figure 2.66 Modulus of elasticity test results for Mainlane superstructure Unit D2.................... 110

Figure 2.67 Modulus of elasticity test results for Large Ramp pier P16 ...................................... 111

Figure 2.68 Modulus of elasticity test results for Ramp P superstructure .................................... 111

xxi
Figure 2.69 Concrete compressive strengths for mainlane D....................................................... 113

Figure.3.1 Friction loss model ..................................................................................................... 116

Figure 3.2 Prestress force loss from creep and shrinkage ............................................................ 119

Figure.3.3 Mainlane girder tendons ............................................................................................. 120

Figure 3.4 Ramp P girder tendons................................................................................................ 120

Figure 3.5 Results from previous friction tests ............................................................................ 122

Figure 3.6 Bench test ................................................................................................................... 131

Figure 3.7 Instrumentation calibration from bench test ............................................................... 132

Figure 3.8 Mainlane girder D5 tendon profiles–elevation ........................................................... 133

Figure 3.9 Mainlane girder D5 tendon profiles–end view ........................................................... 133

Figure 3.10 Ramp P tendon geometry–elevation ......................................................................... 141

Figure 3.11 Ramp P tendon geometry–end view ......................................................................... 141

Figure 3.12 Strain change in ramp tendons over time .................................................................. 145

Figure 3.13 Pier P16 elevation and tendon profiles ..................................................................... 146

Figure 3.14 Sequence of final post-tensioning of pier P16 tendons ............................................. 147

Figure 3.15 Selected axial strains in segment PC16-1 during pier post-tensioning,
north-south axis locations................................................................................................... 148

Figure 4.1 Climatic influences on superstructure thermal gradients ............................................ 163

Figure 4.2 The effects of cross section shape on thermal gradient shape..................................... 164

Figure 4.3 Components of a nonlinear thermal gradient .............................................................. 165

Figure 4.4 Effect of linear thermal gradient components on a statically determinate span ......... 165

Figure 4.5 Effect of linear thermal gradient components on a

statically indeterminate bridge structure............................................................................. 166

Figure 4.6 Calculation of self-equilibrating stresses from a nonlinear

thermal gradient when plane sections remain plane ........................................................... 167

Figure 4.7 Thermal gradient that varies across the width and depth of a cross section ............... 169

xxii
Figure 4.8 Transverse analysis for thermal gradients................................................................... 170

Figure 4.9 Girder response to an applied positive thermal gradient when warping occurs .......... 172

Figure 4.10 Girder response to an applied positive thermal gradient when warping occurs ........ 173

Figure 4.11 Response of a single column to a thermal gradient................................................... 174

Figure 4.12 Response of a single column bridge pier to a thermal gradient ................................ 174

Figure 4.13 Response of a multiple column bridge pier to a thermal gradient ............................ 175

Figure 4.14 Response of solid column to thermal gradients ........................................................ 175

Figure 4.15 Thermal gradient orientations at different times of the day ...................................... 176

Figure 4.16 Thermocouple locations and maximum recorded thermal

gradients for the Hoffman, McClure and West study [33] ................................................. 178

Figure 4.17 Thermocouple layout for the San Antonio Y study by Roberts [8] .......................... 180

Figure 4.18 Temperature distributions through the depth of the section for the US 190
Atchafalaya River Bridge study by Pentas et al. [38] ......................................................... 181

Figure 4.19 Proposed maximum solar radiation zones by Imbsen, et al. [40] ............................. 183

Figure 4.20 Recommended positive vertical temperature gradient by Imbsen, et al. [40] ........... 183

Figure 4.21 Recommended negative vertical temperature gradient by Imbsen, et al [40] ........... 184

Figure 4.22 Positive vertical temperature gradient in concrete and steel

superstructures (Figure 3.12.3-2 and Table 3.12.3-1 from the AASHTO
LRFD Bridge Design Specifications [10]) ......................................................................... 185

Figure 4.23 Thermocouples used to calculate thermal gradient magnitudes

on the Ramp P superstructure ............................................................................................. 186

Figure 4.24 Measured thermal gradient magnitudes on Ramp P for the month of March 1997 .. 187

Figure 4.25 The maximum measured positive gradient on Ramp P (from March 20, 1997) ....... 188

Figure 4.26 The maximum measured negative gradient on Ramp P (from March 6, 1997) ........ 189

Figure 4.27 Statistical occurrence of daily maximum T1,meas values on Ramp P before
application of the asphalt blacktop ..................................................................................... 190

Figure 4.28 Statistical occurrence of daily maximum T1,meas values on Ramp P after
application of the asphalt blacktop ..................................................................................... 190

xxiii
Figure 4.29 Statistical occurrence of daily minimum T1,meas values on Ramp P before
application of the asphalt blacktop ..................................................................................... 191

Figure 4.30 Statistical occurrence of daily minimum T1,meas values on Ramp P after
application of the asphalt blacktop ..................................................................................... 191

Figure 4.31 Division of the Ramp P cross-section into tributary areas for each
thermocouple gauge ........................................................................................................... 192

Figure 4.32 Comparison of measured and calculated positive thermal gradient

stresses for segment P16-2 (near diaphragm) .................................................................... 193

Figure 4.33 Comparison of measured and calculated positive thermal gradient

stresses for segment P16-10 (quarter point) ....................................................................... 195

Figure 4.34 Comparison of measured and calculated positive thermal gradient

stresses for segment P16-17 (midspan) .............................................................................. 196

Figure 4.35 Comparison of measured and calculated negative thermal gradient

stresses for segment P16-2 (near diaphragm) ..................................................................... 198

Figure 4.36 Comparison of measured and calculated negative thermal gradient

stresses for segment P16-10 (quarter point) ....................................................................... 199

Figure 4.37 Comparison of measured and calculated negative thermal gradient

stresses for segment P16-17 (midspan) .............................................................................. 200

Figure 4.38 Minimum top flange stress load combinations for P16-2, P16-10 and P16-17......... 202

Figure 4.39 Maximum top flange stress load combinations for P16-2, P16-10 and P16-17 ........ 204

Figure 4.40 Comparison of measured and calculated transverse flexural stresses

from positive and negative thermal gradients for P16-17 .................................................. 205

Figure 4.41 Measured stress changes in the Ramp P external tendons from
the maximum positive gradient .......................................................................................... 206

Figure 4.42 Measured stress changes in the Ramp P external tendons from
the maximum negative gradient ......................................................................................... 207

Figure 4.43 Thermocouples used to calculate thermal gradient magnitudes

on the mainlane girder D5 .................................................................................................. 208

Figure 4.44 Measured thermal gradients on mainlane girder D5 for the month of June 1996 ..... 209

Figure 4.45 The maximum measured positive gradient on mainlane

girder D5 (from June 17, 1996) .......................................................................................... 210

xxiv
Figure 4.46 Measured negative gradient on mainlane girder D5 (from November 11, 1995) ...... 211

Figure 4.47 Statistical occurrence of daily maximum T1,meas values on mainlane girder D5
before application of the asphalt blacktop .......................................................................... 212

Figure 4.48 Statistical occurrence of daily maximum T1,meas values on mainlane girder D5
after application of the asphalt blacktop ............................................................................. 213

Figure 4.49 Statistical occurrence of daily minimum T1,meas values on mainlane girder D5
before application of the asphalt blacktop .......................................................................... 213

Figure 4.50 Statistical occurrence of daily minimum T1,meas values on mainlane girder D5
after application of the asphalt blacktop ............................................................................. 214

Figure 4.51 Division of the mainlane girder D5 cross-section into tributary areas for each
thermocouple gauge ........................................................................................................... 214

Figure 4.52 Comparison of measured and calculated positive thermal gradient stresses for
segment D5-9 (April 20, 1995) .......................................................................................... 216

Figure 4.53 Comparison of measured and calculated positive thermal gradient stresses for
segment D5-16 (April 20, 1995) ........................................................................................ 217

Figure 4.54 Comparison of measured and calculated positive thermal gradient stresses for
segment D5-9 (June 17, 1996)............................................................................................ 220

Figure 4.55 Comparison of measured and calculated positive thermal gradient stresses for
segment D5-16 (June 17, 1996).......................................................................................... 221

Figure 4.56 Comparison of measured and calculated negative thermal gradient stresses for
segment D5-9 (November 11, 1995) .................................................................................. 223

Figure 4.57 Comparison of measured and calculated negative thermal gradient stresses for
segment D5-16 (November 11, 1995) ................................................................................ 224

Figure 4.58 Measured peak positive and negative thermal gradients in degrees Celsius for
segment D5-9 flanges and webs ......................................................................................... 226

Figure 4.59 Typical daily temperature cycle of selected thermocouples in segment PC16-5 ...... 228

Figure 4.60 Temperatures recorded through the thickness of the west wall of segment PC16-5. 229

Figure 4.61 Temperatures measured along the height of pier P16 ............................................... 230

Figure 4.62 One day cycle of thermal gradients along the north-south axis
of segment PC16-5 ............................................................................................................. 231

xxv
Figure 4.63 One day cycle of thermal gradients along the northeast-southwest axis of
segment PC16-5 ................................................................................................................. 232

Figure 4.64 One day cycle of thermal gradients along the east-west axis of segment PC16-5 .... 233

Figure 4.65 One day cycle of thermal gradients along the northwest-southeast axis of
segment PC16-5 ................................................................................................................. 234

Figure 4.66 Temperature and strain changes on the north-south axis of segment PC16-5 .......... 235

Figure 4.67 Temperature and strain changes on the east-west axis of segment PC16-5 .............. 236

Figure 4.68 Temperature and strain changes through the thickness of the north wall of
segment PC16-5 ................................................................................................................. 236

Figure 4.69 Temperature and strain changes through the thickness of the west wall of
segment PC16-5 ................................................................................................................. 237

Figure 4.70 Concentric octagonal model for temperature representation .................................... 238

Figure 4.71 Projection of temperature gradient shape onto a single axis ..................................... 239

Figure 4.72 Measured and design code maximum positive thermal gradients for pier P16 ......... 240

Figure 4.73 Measured and design code maximum negative thermal gradients for pier P16 ......... 241

Figure 4.74 Temperature changes along the east-west axis recorded by thermocouples in
segment PC16-5, maximum positive gradient .................................................................... 242

Figure 4.75 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, east-west axis, maximum positive gradient............................................ 243

Figure 4.76 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-5, east-west axis, maximum positive gradient............................................ 243

Figure 4.77 Temperature changes along the north-south axis recorded by thermocouples in
segment PC16-5, maximum positive gradient .................................................................... 244

Figure 4.78 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, north-south axis, maximum positive gradient ........................................ 244

Figure 4.79 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-5, north-south axis, maximum positive gradient ........................................ 245

Figure 4.80 Temperature changes along the east-west axis recorded by thermocouples in
segment PC16-5, maximum negative gradient ................................................................... 245

xxvi
Figure 4.81 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, east-west axis, maximum negative gradient ........................................... 246

Figure 4.82 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-5, east-west axis, maximum negative gradient ........................................... 247

Figure 4.83 Temperature changes along the north-south axis recorded by thermocouples in
segment PC16-5, maximum negative gradient ................................................................... 247

Figure 4.84 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, north-south axis, maximum negative gradient ....................................... 248

Figure 4.85 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-5, north-south axis, maximum negative gradient ....................................... 248

Figure 4.86 Mechanism for cracking of large monolithic members during curing ...................... 250

Figure 4.87 Comparison of internal curing temperature to ambient air temperature, capital
segment PC16-8 ................................................................................................................. 250

Figure 4.88 Maximum measured gradient shape during curing of capital segment PC16-8 ........ 251

Figure 4.89 Map of cracks found during curing of capital segment PC16-8 ............................... 251

Figure 4.90 One day cycle of pier D6 capital concrete temperatures........................................... 252

Figure 4.91 One day cycle of pier D6 capital pipe temperatures ................................................. 253

Figure 4.92 Pier D6 capital pipe temperature versus concrete core temperature, typical sunny
day ...................................................................................................................................... 253

Figure 4.93 Pier D6 pipe measured strains, typical sunny day..................................................... 254

Figure 4.94 Pier D6 gauge locations ............................................................................................ 255

Figure 4.95 Measured concrete strains in the pier D6 capital over a sunny day .......................... 256

Figure 4.96 Measured concrete strains at the top of pier D6 column over a sunny day ............... 257

Figure 4.97 Positive design gradients and measured gradients, no blacktop ............................... 261

Figure 4.98 Positive design gradients and measured gradients, 50mm of blacktop ..................... 261

Figure 4.99 Negative design gradients and measured gradients, no blacktop .............................. 264

Figure 4.100 Negative design gradients and measured gradients, 50mm of blacktop ................. 265

Figure 4.101 Recommended positive top flange design gradient for transverse design .............. 267

xxvii
Figure 4.102 Recommended top flange negative design gradient for transverse design.............. 268

Figure 4.103 Recommended web and bottom flange negative design gradient
for transverse design........................................................................................................... 269

Figure 4.104 Recommended positive design gradient for piers ................................................... 271

Figure 4.105 Recommended negative design gradient for piers .................................................. 272

Figure 5.1 Shear lag deformation in a simple span box girder ..................................................... 281

Figure 5.2 Top flange stresses in a continuous box girder ........................................................... 282

Figure 5.3 Response of a multiple-cell box girder to live load .................................................... 284

Figure 5.4 Response of twin single-cell box girders to live load ................................................. 285

Figure 5.5 Beam model of a box girder bridge ............................................................................ 287

Figure 5.6 Analysis of twin single-cell box girders using superposition ..................................... 288

Figure 5.7 Finite element model of a box girder bridge ............................................................... 289

Figure 5.8 Pattern of effective flange width coefficients, bf and bs (Figure 4.6.2.6.2-1 from
the AASHTO LRFD Bridge Design Specifications [10]) .................................................. 291

Figure 5.9 Cross sections and corresponding effective flange widths, bm, for flexure and
shear (Figure 4.6.2.6.2-3 from the AASHTO LRFD Bridge Design Specifications [10]) . 292

Figure 5.10 Values of the effective flange with coefficient bm/b, for the given values of b/li
(Figure 4.6.2.6.2-2 from the AASHTO LRFD Bridge Design Specifications [10])........... 293

Figure 5.11 Effective flange widths, bn, for normal forces (Figure 4.6.2.6.2-4 from the
AASHTO LRFD Bridge Design Specifications [10]) ........................................................ 294

Figure 5.12 Example of the effects of the diffusion of post-tensioning forces

(after Kristek [45]) ............................................................................................................. 295

Figure 5.13 Mainlane girder D5 tendon locations ........................................................................ 297

Figure 5.14 Longitudinal stresses in segment D5-16 from the tensioning of

external tendons 1 through 6 .............................................................................................. 298

Figure 5.15 Longitudinal stresses in segment D5-9 from the tensioning of

external tendons 1 through 6 .............................................................................................. 300

Figure 5.16 Longitudinal stresses in segment D5-16 from the tensioning of

internal bottom flange tendons 7 through 14 ...................................................................... 302

xxviii
Figure 5.17 Longitudinal stresses in segment D5-9 from the tensioning of internal bottom
flange tendons 7 through 14 ............................................................................................... 304

Figure 5.18 Longitudinal stresses in segment D5-16 from the tensioning of bottom flange,
wing and external tendons 1 through 18 ............................................................................. 306

Figure 5.19 Longitudinal stresses in segment D5-9 from the tensioning of bottom flange,
wing and external tendons 1 through 18 ............................................................................. 307

Figure 5.20 Measured and calculated longitudinal stresses in segment D5-16 from post-
tensioning and dead load .................................................................................................... 309

Figure 5.21 Measured and calculated longitudinal stresses in segment D5-9 from post-
tensioning and dead load .................................................................................................... 310

Figure 5.22 Deflections from post-tensioning and dead load....................................................... 311

Figure 5.23 Live load cases–Mainlane girders D4, D5 and D6 ................................................... 313

Figure 5.24 Live load cases–Mainlane girders D4, D5 and D6 (continued) ................................ 314

Figure 5.25 Longitudinal stresses from load case 2 ..................................................................... 316

Figure 5.26 Longitudinal stresses from load case 8 ..................................................................... 317

Figure 5.27 Deflections from load case 2 .................................................................................... 318

Figure 5.28 Deflections from load case 3 .................................................................................... 319

Figure 5.29 Deflections from load case 5 .................................................................................... 320

Figure 5.30 Deflections from load case 6 .................................................................................... 321

Figure 5.31 Deflections from load case 8 .................................................................................... 322

Figure 5.32 Deflections from load case 9 .................................................................................... 323

Figure 5.33 Diffusion of force from temporary post-tensioning in segment P16-10 ................... 324

Figure 5.34 Layout of segments and cantilever tendons in the P16 upstation cantilever ............. 326

Figure 5.35 Diffusion of post-tensioning force from the cantilever tendons in

segment P16-10 .................................................................................................................. 327

Figure 5.36 Longitudinal stresses in segment P16-2 after completion of

the P16 upstation cantilever ............................................................................................... 328

xxix
Figure 5.37 Longitudinal stresses from selected strain gauges in segment P16-2 over the
course of the cantilever construction sequence .................................................................. 329

Figure 5.38 Layout of internal and external continuity tendons in span P16 ............................... 330

Figure 5.39 Longitudinal stresses in segment P16-2 after stressing of the continuity
tendons for span P16 .......................................................................................................... 331

Figure 5.40 Longitudinal stresses in segment P16-10 after stressing of the continuity
tendons for span P16 .......................................................................................................... 332

Figure 5.41 Longitudinal stresses in segment P16-17 after stressing of the continuity
tendons for span P16 .......................................................................................................... 333

Figure 5.42 Stresses from selected strain gauges from segment P16-2 over the course
of the continuity stressing sequence ................................................................................... 335

Figure 5.43 Stresses from selected strain gauges from segment P16-10 over the course
of the continuity stressing sequence ................................................................................... 335

Figure 5.44 Stresses from selected strain gauges from segment P16-17 over the course
of the continuity stressing sequence ................................................................................... 336

Figure 5.45 Live load test cases for the ramp P girder ................................................................. 337

Figure 5.46 Measured live load longitudinal stresses in segment P16-2 ...................................... 339

Figure 5.47 Measured and calculated live load longitudinal stresses in segment P16-2 .............. 340

Figure 5.48 Measured and calculated live load longitudinal stresses in segment P16-10 ............ 341

Figure 5.49 Measured and calculated live load longitudinal stresses in segment P16-17 ............ 342

Figure 5.50 Cross section of Unit C15, L2 .................................................................................. 343

Figure 5.51 (a) Live load cases for Unit C15, L2......................................................................... 344

Figure 5.51 (b) Live load cases for Unit C15, L2 ........................................................................ 345

Figure 5.52 Deflections from load case 9 .................................................................................... 347

Figure 5.53 Deflections from load case 10................................................................................... 348

Figure 5.54 Deflections from load case 11................................................................................... 349

Figure 5.55 Deflections from load case 12................................................................................... 350

Figure 5.56 Deflections from load case 13................................................................................... 351

xxx
Figure 5.57 Cross section of Unit C13 ......................................................................................... 352

Figure 5.58 (a) Live load cases for Unit C13 ............................................................................... 353

Figure 5.58 (b) Live load cases for Unit C13 ............................................................................... 354

Figure 5.59 Deflections from load case 2 .................................................................................... 356

Figure 5.60 Deflections from load case 3 .................................................................................... 357

Figure 5.61 Deflections from load case 4 ..................................................................................... 358

Figure 5.62 Deflections from load case 5 ..................................................................................... 359

Figure 5.63 Deflections from load case 6 .................................................................................... 361

Figure 5.64 Stresses in pier segment PC16-1 from placement of

superstructure segment P16-17........................................................................................... 363

Figure 5.65 Stresses in pier segment PC16-5 from placement of superstructure

segment P16-17 .................................................................................................................. 364

Figure 5.66 Stresses in pier segment PC16-7 from placement of superstructure

segment P16-17 .................................................................................................................. 364

Figure 5.67 Stresses in pier segment PC16-8 tie-down bars from placement of
superstructure segment P16-17........................................................................................... 365

Figure 6.1 Typical strut-and-tie models [52]................................................................................ 378

Figure 6.2 Typical strut-and-tie models for segmental box girder design [52] ............................ 379

Figure 6.3 Strut types [52] ........................................................................................................... 379

Figure 6.4 Anchorage details at nodes [52].................................................................................. 380

Figure 6.5 Strut-and-tie model of a diaphragm from the San Antonio Y [7] ............................... 383

Figure 6.6 Strut-and-tie model of a deviator from the San Antonio Y [7] ................................... 384

Figure 6.7 Types of anchorage blisters [54] ................................................................................. 385

Figure 6.8 Mainlane pier D6 details ............................................................................................. 388

Figure 6.9 Simple strut-and-tie model for mainlane pier D6 ....................................................... 389

Figure 6.10 Frame model for mainlane pier D6 ........................................................................... 390

xxxi
Figure 6.11 Measured strains in pier D6 capital from the superstructure dead load .................... 391

Figure 6.12 Measured strains in pier D6 capital from live load case 5 ........................................ 393

Figure 6.13 Measured strains in the deviator reinforcing bars from superstructure post-
tensioning forces and dead load ......................................................................................... 395

Figure 6.14 Crack patterns in the segment D5-12 deviator .......................................................... 397

Figure 6.15 Measured strains in the deviator reinforcing bars from live load case 2 ................... 398

Figure 6.16 Strut-and-tie model for the segment D5-12 deviator ................................................ 400

Figure 6.17 Mainlane anchor segment D5-16 details................................................................... 404

Figure 6.18 Crack patterns in segment D5-16 diaphragm ............................................................ 405

Figure 6.19 Crack patterns in the bottom flange of segments D5-16, D5-15 and D5-14 ............. 406

Figure 6.20 Post-tensioning and dead load compressive force paths during
construction of span D5 ...................................................................................................... 412

Figure 6.21 Dead load support conditions during construction of span D5 ................................. 413

Figure 6.22 Span D5 compressive force paths at ultimate load ................................................... 414

Figure 6.23 Segment D5-16 diaphragm tie locations at ultimate load ......................................... 415

Figure 6.24 Measured bottom flange longitudinal stresses from post-tensioning forces ............. 417

Figure 6.25 Large ramp pier capital segment PC16-8 strut gauges.............................................. 418

Figure 6.26 Large ramp pier capital segment PC16-8 tie gauges................................................. 419

Figure 6.27 Ramp P girder anchor segment P16-1 details ........................................................... 424

Figure 6.28 Crack patterns in segment P16-1 .............................................................................. 425

Figure 6.29 Strut-and-tie model at pier P16 - Side view .............................................................. 429

Figure 6.30 Strut-and-tie model at pier P16 - Looking down station ........................................... 430

Figure 6.31 Ramp P girder segment P16-10 deviator details ....................................................... 432

Figure 6.32 Vertical and horizontal forces from tendons T1, T2 and T3 ..................................... 433

Figure 6.33 Crack patterns in segment P16-10 ............................................................................ 436

xxxii
Figure 6.34 Strut-and-tie model for the segment P16-10 deviator ............................................... 439

Figure 6.35 Ramp P girder segment P16-4 anchorage blister details ........................................... 441

Figure 6.36 Strut-and-Tie model for segment P16-4 anchorage blister for
service load forces from Tendon T22 ................................................................................. 445

Figure 6.37 Struts and ties near anchorage blister at ultimate ...................................................... 448

Figure 6.38 Ramp P girder segment P16-10 deviator to web detail ............................................. 451

Figure 7.1 Post-tensioning tendons on the mainlane girder ......................................................... 454

Figure 7.2 Original joint details from contract plans for U. S. 183 .............................................. 455

Figure 7.3 Actual cast-in-place joint construction ....................................................................... 456

Figure 7.4 Longitudinal sections - cast-in-place joint .................................................................. 458

Figure 7.5 Joint details on the San Antonio Y ............................................................................. 460

Figure 7.6 Deflections during live load test, from Roberts, et al. [7] ........................................... 461

Figure 7.7 Strain gauge locations in cast-in-place joint - Unit D2 ............................................... 462

Figure 7.8 Live load cases for Unit D2 ........................................................................................ 464

Figure 7.9 Cast-in-place joint stresses–load case 2 ...................................................................... 465

Figure 7.10 Cast-in-place joint stresses–load case 3 .................................................................... 466

Figure 7.11 Cast-in-place joint stresses–load case 5 .................................................................... 466

Figure 7.12 Cast-in-place joint stresses–load case 6 .................................................................... 467

Figure 7.13 Deflections from live load test on Unit D2–load case 2 ........................................... 469

Figure 7.14 Deflections from live load test on Unit D2–load case 3 ........................................... 469

Figure 7.15 Deflections from live load test on Unit D2–load case 5 ........................................... 470

Figure 7.16 Deflections from live load test on Unit D2–load case 6 ........................................... 470

Figure 7.17 Live load cases–Unit C13 ......................................................................................... 472

Figure 7.18 Deflections from live load test on Unit C13–load case 2.......................................... 473

Figure 7.19 Deflections from live load test on Unit C13–load case 4.......................................... 474

xxxiii
Figure 7.20 Live load cases–Units C15 and L2 ........................................................................... 475

Figure 7.21 Deflections from live load test on Units C15, L2–load case 9 .................................. 477

Figure 7.22 Deflections from live load test on Units C15, L2–load case 11 ................................ 477

Figure 7.23 Measured temperatures for thermal gradient load case............................................. 479

Figure 7.24 Measured strains in segment D5-9 from thermal gradient load case ........................ 480

Figure 7.25 Measured strains in segment D5-16 and cast-in-place joint

from thermal gradient load case ......................................................................................... 482

Figure A.1 US 183 Mainlane box girder daily maximum thermal gradients ............................... 513

Figure A.2 US 183 Ramp P box girder daily maximum thermal gradients ................................. 514

Figure A.3 US 183 segmental box pier P16 daily peak thermal gradients ................................... 514

xxxiv
 CHAPTER 1
INTRODUCTION

1.1 THE U.S. 183 STUDY

The data for this dissertation came from a field study of the construction and service
behavior of the U.S. 183 segmental viaduct in north central Austin. The structure is an urban
viaduct designed to separate local traffic from through traffic. U.S. 183 is the major route for
commuter traffic between IH35 and the northwest Austin suburbs (see Figure 1.1). U.S. 183,
within the limits of the project, has extensive commercial development and a narrow right-of-way.
U.S. 183 elevated was designed to provide limited access to local roads beneath the structure, with
major interchanges at IH35, Lamar Boulevard, and Loop 1. Local businesses could be accessed
from frontage roads and turnarounds beneath the elevated structure.

University of Texas at Austin

J.J. Pickle Research Campus
and Braker Ln.
Phil M. Ferguson Laboratory

US
18
3
t Rd.

Instrumented
1)

Mainlane and
d.
oo p

Burne

Blv

Pier D6
ay (L

r
ma

IH-35
La
essw

Segmental N
Expr

Viaduct Instrumented
Transition
ac
MoP

Spans

Instrumented Ramp P
Segmental
AUSTIN Viaduct
and Pier P16

US
183

Figure 1.1 Location map

1
 Twin precast segmental box girders constructed by the span-by-span method (Figure 1.2)
were selected as the structure type most suited to much of the site for several reasons. The
structure could be constructed from above the completed frontage roads while traffic could be
maintained beneath. Single column piers were used beneath the single-cell box girders allowing
the frontage roads to be built beneath the structure overhangs, maximizing the number of lanes
(12) in the narrow right-of-way. With over 10 kilometers of structure, precast concrete box girder
construction, including fabrication of erection trusses and construction of a precasting facility, was
also an economical solution. The box girder design option was the low bid at $71.3 million over
an alternate multiple precast delta and I-girder design with cast-in-place closures. Martin K. Eby
Construction was the contractor in a joint venture with Flatiron Structures Company. A summary
of all bids is shown in Table 1.1.

Erection
trusses

Figure 1.2 Span by span construction

2
 Table 1.1 Project bid summary
Contractor Design Option Bid Price

Martin K. Eby Construction Co. Precast Box Girder $71,328,098

Austin Bridge & Road Precast Box Girder $71,792,000
H. B. Zachary Company Precast Box Girder $73,456,000
Gilbert Texas Construction Co. Precast Box Girder $77,495,848
McCarthy Brothers/PCL Civil Constructors Precast Box Girder $82,893,263

An important consideration in selecting a box girder for the site was aesthetics. Austin
residents are very sensitive to potential influences on their high quality of life in the state capital.
A structure as potentially intrusive as the U.S. 183 elevated would have to be designed as a
signature structure for Austin, while minimizing visual impact. The segmental box girder option
gave the designers a girder type known for its simple form and lightweight appearance. The
designers, the TxDOT Bridge Division, used a spine beam style box girder, with narrow soffit and
long wings, to further enhance the light appearance of the structure. Architectural details
reflecting the masonry construction of the State Capitol and The University of Texas Main
Building were successfully blended into the design, especially of the single column piers, by
TxDOT engineers as seen in Figure 1.3. Table 1.2 shows the cost of the U.S. 183 box girders and
Y-shaped piers compared to the average Texas girder and pier costs [1]. The span lengths of the
segmental box girders on U.S. 183 were somewhat longer than the average span length for a
pretensioned I-girder. Also, the standard multiple circular column with bent cap piers could not
have been used at this site because of conflicts with the right-of-way requirement for the frontage
roads.

3
 Figure 1.3 Mainlane structure

Table 1.2 Unit construction costs

Structure Type Unit Cost

Pretensioned I-girder $323/m2

U.S. 183 segmental box girder $463/m2

Circular column with rectangular cap pier $420/m3

Circular column with inverted T cap pier $485/m3
U.S. 183 mainlane Y-pier $556/m3

The TxDOT Bridge Division designed the U.S. 183 elevated segmental bridge using the
knowledge they gained from the design, construction, and instrumentation of the San Antonio Y
several years before (Figure 1.4). The segmental box girder superstructure design used at San
Antonio was very similar in dimension and construction procedure to the superstructure used on
most of U.S. 183, but was built in fully continuous multispan units. Most of the U.S. 183
superstructure was constructed as simple spans. The contract for the San Antonio Y
instrumentation study was with The University of Texas at Austin, as was the study on U.S. 183.
The University of Texas at Austin, particularly the Phil M. Ferguson Structural Engineering
Laboratory, has had a long term productive bridge research relationship with TxDOT and FHWA.
The study of segmental bridges for Texas at the laboratory dates back to the early 1970's, when

4
research was conducted for the design and construction of the segmental channel spans of the JFK
Memorial Causeway in Corpus Christi. The knowledge gained from these field studies, and from
many laboratory based research projects was immediately implemented by TxDOT and the
segmental bridge industry. This research provided valuable information to designers, such as
effective shear key details for segment joints [2], ultimate load behavior of externally prestressed
girders [3] and segmental box piers [4], and the actual behavior of discontinuity (or D-)zones for
various reinforcement layouts [5] [6]. The instrumentation of the U.S. 183 segmental project was
designed to continue study of several research topics originally studied at the San Antonio Y [7],
but having several new ranges of variables. More importantly this program was also able to study
many new topics and structure types, including a segmentally constructed pier, a strut-and-tie
model type pier, and a five span continuous box girder constructed in balanced cantilever. The
latter is shown in Figure 1.5.

Figure 1.4 The San Antonio Y

5
 Figure 1.5 U.S. 183 North Ramp P constructed in balanced cantilever

The close proximity of the structure to the Ferguson Laboratory (Figure 1.1) gave
researchers the opportunity to conduct a very thorough, time intensive study of the bridge while
under construction. Staff on the project consisted of four Master's degree candidates, one Ph.D.
candidate, two supervising professors, and numerous helpers and volunteers. Over 1100 gauges
and thermocouples were installed and monitored on the project. All of the data were examined.
Obviously all of the data provided by these sensors could not be presented in this dissertation, nor
could a detailed study be performed on all imaginable topics. The topics presented herein were
considered to be the most important and useful to design engineers, construction engineers, and
constructors. Recommendations and conclusions presented were intended to aid engineers and
constructors in their decisions concerning structural behavior of segmental bridges during
construction and service, as well as providing a basis for changes in the AASHTO Guide
Specification for the Design and Construction of Segmental Bridges [8].

6
1.2 PURPOSE FOR THE STUDY
The main purpose for this study was to gather information through measurements and
analysis that would lead to the reduction of construction and maintenance costs for segmental
bridges in the State of Texas, while improving their structural performance and safety. The cost of
this research project was less than 1% of the total project cost, yet historically represents a
substantial investment for research by a state agency. The standard deviation of the lowest three
bid prices shown in Table 1.1 was 1.6% of the low bid, which is actually a small percentage for a
project of this size and complexity. This demonstrates that the variation in estimated construction
costs can be larger than an expensive research program for the project. Table 1.2 shows that the
cost of the 39m span U.S. 183 box girders, the span length selected as a compromise between
economy and aesthetics, is somewhat higher in cost than the Texas average I-girder superstructure
cast at existing plants. These I-girders are generally in the 13m to 40m span range.

1.2.1 Structural performance and safety

The evaluation of structural performance by instrumentation is actually an evaluation of
the design method used to analyze the structure. Structural performance of the traditional I-girders
on short to medium span bridges is well understood from testing, and the design is generally
straightforward, requiring few untested assumptions. Box girder design may require a complex
time-dependent three-dimensional analysis, such as for Ramp P studied on this project. The
assumptions and simplifications used in design can only be verified or calibrated using
measurements on an actual structure.

Evaluation of the level of performance and safety in a structure begins with the
evaluation of all critical elements of the bridge, starting with the main structural elements. If the
actual structural response to known loads, environmental conditions, or time dependent factors is
poorly understood or has been poorly predicted, the level of performance and, often, of safety
cannot be predicted. Predictable behavior of the main structural elements is critical in determining
the safety of the entire structure. Connections and other areas with complex details or
concentrated loads are designed to a higher degree of safety than the main members. Segmental
box girders have numerous discontinuity zones (D-zones), where plane sections do not remain
plane after loading and cracking may be expected. These D-zones, including deviators, anchorage
zones, and diaphragms, can be very complex and have been the source of many problems in the

7
past. Field measurements were taken in many D-zones on U.S. 183 to evaluate their actual
structural performance.

1.2.2 Constructability problems

Problems with structural elements or their assembly have a significant influence on
project costs. Bridge types known for their constructability problems are bid higher by all
contractors. This was the case not long ago for segmental box girders. However, the combination
of wider experience and the codification of good practice in the AASHTO Guide Specification for
Segmental Bridges [8] has greatly reduced the problems, and much of the uncertainty. Yet, since
many recommendations in the Guide Specifications were based on judgment rather than studies,
and since designers and constructors continue to innovate on new projects, construction problems
still occur on these structures. Poor constructability may also lead to poor quality and increased
future maintenance costs.

Researchers on the U.S. 183 project surveyed the constructability of the project beginning
at the precasting plant and continuing through to the end of construction. Construction problems
occur for two distinct reasons. First, the problem may be inherent with the structural design. The
engineer may be unaware of constructability problems hidden in the design, and the problems may
not even be visible to the contractor at bid time. Secondly, the construction technique and
equipment used by the contractor may not be optimal, or may not be adaptable to conditions that
may arise during construction.

1.2.3 Initial cost of future bridges

The initial bid price of a segmental bridge can be reduced by two main factors. First, the
structure can be designed to be more efficient. The instrumentation program on U.S. 183 studied
the structural response of elements that traditionally have been over-designed because of the
complexity and importance of the element. Second, the structure's form and construction method
can be made more simple and straightforward, with a minimum number of construction steps or
tasks required building the bridge. This topic is highly related to constructability, and has the
greatest influence on profit margin for the contractor.

1.2.4 Maintenance problems

Another purpose for the study was to identify sources of future maintenance problems,
and recommend ways to solve these problems for future bridges. Consideration of the life cycle

8
cost of bridges is beginning to become commonplace with highway departments. Traditional
expensive maintenance items, such as painting and deck replacement, were eliminated with the
U.S. 183 segmental box girders. Potential major future maintenance items for this bridge are
repair of cracks, especially in regions near post-tensioning anchorages and deviators, and
replacement of tendons. The researchers noted the details that performed poorly at service load
levels, and also noted the details that performed well.

1.3 PROJECT OBJECTIVES

The general objectives of this research project were to:

1.3.1 Identify and study major design uncertainties

There are many design areas where engineering judgment had to be used when U.S. 183
was designed in lieu of data from previous field or laboratory measurements. The researchers
interviewed both design engineers and construction engineers to determine the major uncertainties
on this project. These uncertainties included the accuracy of prestress friction loss calculations,
temperature gradient effects on the box girders, transverse diffusion of post-tensioning forces into
the cross section, construction stress distribution in partially completed box girders, and the actual
behavior of complex D-zones including the usefulness of strut-and-tie modeling.

1.3.2 Instrument and test structural elements

Many of the structural elements designed using questionable or possibly over-
conservative design assumptions were instrumented. Also, many instruments were placed to
record the structure's response to general load cases, in locations where the behavior was thought
to be well known, to look for anomalies in both the structure's behavior and the performance of the
instrumentation systems. Monitoring of instrumented segments generally began at the casting
yard, continued throughout construction and was continued for about a year thereafter to record
the response to thermal loads, live loads, and long term deformations of the structure.

1.3.3 Interpret measurements

Measurements were made of tendon forces, concrete surface and internal strains,
reinforcing bar strains, structural steel pipe strains, deflections, and temperatures. After data were
reduced for a given load case, the measured structural behavior could be compared to that
predicted by normal design techniques used by the industry. This comparison gave an indication

9
of the efficiency of the design, whether conservative or unconservative, and provided the actual
measured response in regions that showed signs of distress, such as major cracking.

1.3.4 Make recommendations to TxDOT and FHWA

Once the measurements had been interpreted and analysis completed, recommendations
were to be made to TxDOT and FHWA engineers, the sponsors of the research. Specific
recommendations were also to be made for modifications to current design codes. The AASHTO
Guide Specification for the Design and Construction of Segmental Concrete Bridges [8] (Figure
1.6) and the AASHTO LRFD Bridge Design Specifications [9] are the primary codes available to
guide designers in the U.S.A. Much of the research for the Segmental Guide Specification was
performed at The University of Texas at Austin. Recommendations for modifications of these two
documents are given in this document.

Figure 1.6 The AASHTO Segmental Guide Specification [8]

10
1.4 STRUCTURAL ELEMENTS UNDER STUDY
The U.S. 183 segmental viaduct consists of about 10km of precast concrete
superstructure. The U.S. 183 mainlanes are provided by two parallel bridges carrying three traffic
lanes and one full shoulder each (see Figure 1.7). Access ramps are provided at Lamar Boulevard
and IH35 using a smaller box girder designed for one lane of traffic plus a shoulder (see Figure
1.8). The transitions from the mainlane cross section to the ramp cross section were made using a
three-cell cast-in-place girder, shown in Figure 1.9, originally designed to be precast and
constructed as two single-cell boxes joined by cast-in-place top and bottom slabs.

17070mm
3 Lanes at 3658mm = 10970mm

Figure 1.7 Mainlane superstructure section

11
 7925mm
3658mm Lane

Figure 1.8 Ramp superstructure section

12
 Figure 1.9 Cast-in-place transition spans

The U.S. 183 segmental viaduct project was originally designed with three types of
segmentally constructed piers (see Figure 1.10). However, the contractor opted to cast two of the
pier types in place. One cast-in-place Y-shaped mainlane pier and one segmentally constructed
large ramp pier were instrumented, and the small ramp piers were not selected for study.

(a) Mainlane Y-Pier (b) Small Ramp Pier (c) Large Ramp Pier
Figure 1.10 Pier types
13
 The contractor’s decision to cast most of the piers in place rather than use precast
segmental construction was made due to several factors. First, the vast majority of the piers were
relatively short (below 10 meters in height) and easy to reach with only small mobile cranes.
Second, access to the piers for construction vehicles were relatively simple, with most of the
mainlane piers located in the median of the existing frontage roads where ground conditions were
excellent. Third, space was limited at the precasting and storage facility constructed for the
project. The decision was made to use the large amount of land available under the bridge right-
of-way as the construction preparation area for the cast-in-place piers. This allowed pier
construction to begin as soon as the forms were constructed and shipped.

All concrete segments were cast at a precasting facility in southeast Austin that was
constructed solely for the U.S. 183 segmental project. The precasting yard, shown in Figure 1.11,
was constructed in an inactive shallow quarry. The yard had as many as twelve casting machines
in operation, serviced by four overhead cranes and one mobile crane (see Figure 1.12). Every pair
of casting machines was surveyed from survey houses located along the central axis of the casting
yard. Segments were taken to storage from the casting machines by a straddle crane. With no
provision to double stack segments, the storage yard filled to near capacity at one point in time.
As additional erection trusses came on line, the storage yard emptied and the casting machines
were progressively staged out of service.

Precasting Yard in Background

Storage Area
Shown in
Foreground

Figure 1.11 Project precasting plant

14
 Figure 1.12 Casting machines

The casting yard quickly became a model of efficiency, with a segment produced each
day from every casting machine, except from the heavy end diaphragm casting machines. Out of
the total 3200 segments cast, only about 0.5% were rejected, usually for large voids in congested
webs or deviator blocks. Repetition, excellent construction engineering and quality assurance
personnel, including the on site TxDOT inspectors, and experienced construction workers resulted
in very high quality segments. Project success at the precast yard was important. 60% of the
project construction cost occurred at the precasting yard.

1.4.1 Mainlane pier

Bridge pier design in most state departments of transportation has become highly
standardized. Design for piers that have worked well in the past are often used, and many states
minimize creativity in this area of bridge design as a means of minimizing costs. The piers are
conservatively designed, and minimum reinforcement ratios tend to govern. Structural problems
with these standard piers do not often arise. Modern research on these standard piers has focused
on modifying construction methods, such as precasting, and improving the appearance of multi-
column and single column piers. Billington predicts that both innovative and aesthetic
modifications to existing standard Texas piers could be made with only a marginal increase in cost,

15
at $525/m3, over the cost of the standard circular column piers with rectangular or inverted T bent
cap shown in Table 1.2 [1].

The standard highway piers used to support I-girder superstructures were rejected for use
at U.S. 183 for aesthetic reasons alone. The U.S. 183 mainlane bridge piers are not, by any means,
typical bridge piers. They are functional and also innovative, with great emphasis placed on
aesthetics, as shown in Figure 1.13. They provide an opportunity to investigate the behavior of a
non-standard bridge substructure.

Figure 1.13 Mainlane pier

The piers for the mainlane portion of the U.S.183 project are Y-shaped reinforced
concrete piers with structural steel tension ties across the top of the “Y,” as shown in Figure 1.14.
The column shaft of the piers is of variable height H. The capital has a constant height of
3200mm. Each end of each pipe is anchored by steel plates at two locations, as shown in Figure
1.15. Through their form, these piers provide a visual representation of their structural behavior.
Most observers will intuitively realize that as vertical load is placed on the two sets of bearings
located beneath the box girder webs, the “Y” will tend to spread apart, placing the steel pipe ties
across the “Y” into tension. Although this behavior is intuitive, detailing and dimensioning

16
involved in the pier design may be quite difficult, especially since this pier is a composite structure
of steel and reinforced concrete.

Symmetric abt.
Longitudinal CL Pier
381mm
A 1828mm 762mm A
Pipe Assembly

2254mm
6 Sp. @ 381mm

152mm
3200mm

457mm
229mm

Construction Joint
533mm

Varies

1245mm
Reveals @ 1219 mm Typ.
"H"

1524 mm
Base

Figure 1.14 Mainlane pier elevation

17
 5182mm
1829mm 762mm

Transverse CL Pier
Symmetric abt.
1219mm
2438mm

152mm 452mm

Symmetric abt. See Pipe Assembly

Longitudinal CL Detail
View A-A

PL 38mm x 279mm
CL of Pier
2616mm

x 279mm
457mm

Extra Strong Pipe PL 19mm x 279mm

25mm
x 279mm
203mm (8”) Dia.
Face of Concrete

Pipe Assembly Detail

Figure 1.15 Mainlane pier details

Mainlane pier D6 was selected for instrumentation, and its location is shown on the map
in Figure 1.1. Pier construction began at the north end of the project and moved southward, with
piers for both C and D mainlanes constructed simultaneously. Selection of pier D6 allowed
research to begin as early during construction as possible. Access to pier D6 from the laboratory
and from the ground was also optimal. The mainlane piers were cast in two pours above the
footing for an overall height of 7700mm. The reinforcing bar cage for the capital was actually tied
inside the form, as shown in Figure 1.16. The forms used maintained the appearance, including all
chamfers and reveals, originally intended for the pier as a precast post-tensioned structural
element.
18
 Figure 1.16 Construction of mainlane pier capital

1.4.2 Mainlane girder

One three-span semi-continuous bridge unit of mainlane precast segmental box girder
was selected for instrumentation. Unit D2 (spans D4, D5, and D6) was selected for study.
Instrumented pier D6 was located at the joint between spans D5 and D6. This bridge unit was the
second mainlane bridge unit constructed on the project, allowing research to begin as early as
possible without interfering with the construction of the first couple of spans.

The general shape of the U.S. 183 mainlane girder was based on a design concept by T.Y.
Lin International originally intended for the San Antonio Y segmental project. The girder was
conceived as a spine beam with narrow soffit, shallow structural depth, short spans, and very wide
wings. The girder actually used at the San Antonio Y maintained the spine beam concept, but had
slightly better structural proportions for transverse bending of the top flange. The box girder used
for the U.S. 183 mainlane, shown in Figure 1.17, was similar to the San Antonio girder in
transverse proportions, but was 300mm deeper in structural depth to accommodate longer spans
(39m for U.S. 183 versus 33m at the San Antonio Y). One major difference between the U.S. 183
mainlane girders and the San Antonio Y mainlane girders is the lack of fully continuous spans at
U. S 183.

19
 Figure 1.17 Mainlane girder

The economical method of erecting precast segmental girders up to about 46m in length
is by the span-by-span method. This method uses a pair of erection trusses (see Figure 1.18) to
construct the bridge one span at a time. All segments are loaded onto the trusses and stressed
together generally from one end of the girder. The girder, with ungrouted tendons at this point,
must support its own dead load plus the enormous live load of the construction equipment and
segments passing over to construct the next span. The flexural stiffness benefits of a fully
continuous structure are not realized until after the most critical loads the bridge may ever see
have already been placed on the structure. Furthermore, construction of a continuous structure
requires that continuity tendon post-tensioning operations, bearing placement, and closure pours
occur directly behind the advanced erection trusses, complicating and possibly slowing down
construction of the span on the trusses, as seen at the San Antonio Y. From a design standpoint,
a continuous box girder must be designed for a thermal gradient load case that creates bending
moments over interior piers of a multi-span bridge unit opposite to those induced by live load
forces. This requires additional continuity post-tensioning over the interior piers of a continuous
bridge unit. For these reasons, TxDOT engineers designed the mainlane structure, and most of
the ramp spans, as simple spans. A two or three span partially continuous bridge unit was
created, after construction of the girders had been completed, by casting a deck slab between the
20
box girder top flanges. This “Poor-boy” continuity, as it is known to TxDOT, provides an
inexpensive and smooth riding surface at locations other than the finger joints (see the “Poor-
boy” cast-in-place deck joint under construction in Figure 1.19).

Figure 1.18 Construction of a mainlane girder

Figure 1.19 Construction of a cast-in-place deck joint

21
 The simple span construction was a great success, with the contractor constructing one
span every two days from each set of erection trusses. Construction could have proceeded even
faster except for the requirement that all heavy lifting be done at night, when the frontage roads
beneath could be narrowed to a single lane.

Mainlane spans generally consist of sixteen or seventeen segments. Span D5 is 38.9m

long with two heavy end diaphragm or anchor segments, two deviator segments, and twelve
typical segments, as shown in Figure 1.20.

38.90m
1.83m 14 segments @ 2.49m 1.83m

A D5-1 Through D5-16

A
D5-5 and D5-12
Deviator segments

D5-1 Anchor segment D5-16 Anchor segment

Elevation

17070mm
5798mm 5474mm 5798mm

254mm
2134mm

279mm 613mm
406mm
4877mm

Section A-A
Figure 1.20 Cross section and elevation–Span D5

Six deviated external and 8 straight bottom slab internal 19-15mm diameter strand
tendons, and four smaller straight wing tendons were used to post-tension the span. Transverse
prestressing of the top flange was mostly completed at the casting yard. Most of the casting

22
machines had bulkheads for pretensioning the top slab. Transverse post-tensioning was used only
in the anchor segments.

All casting machines used the short line method of casting, with a fixed bulkhead rear
form and the previous day’s segment mounted on adjustable hydraulic jacks as the front form.
The exterior web and wing form could be lowered slightly and the core form articulated inward to
release a newly cast segment from the form. The soffit form traveled with the segment as it was
rolled on tracks out of the form to the match casting position, or from the match casting position to
the finishing rack. The surveying house was located at the front of the form to measure the
segment as-cast from the previous day, and then later to adjust it correctly in the match casting
position. Reinforcing bar cages were prefabricated to the rear of the form in a jig shaped like the
exterior form. Photographs of the primary features of the casting machines are shown in Figures
1.21, 1.22, and 1.23.

Crane

Reinforcing cage
Cover fabrication jig

Finishing rack

Form

Survey house

Figure 1.21 Casting machine

23
Match casting Match casting
segment Form segment

Form

Bulkhead Hydraulic geometry Tracked bottom

control jacks form
Figure 1.22 Casting machine details

Rodman Surveyor

Match casting
segment

Geometry
control pin

Completed
reinforcing cage
and P/T ducts
Figure 1.23 Geometry control of the casting machine

24
 Segments were hauled from storage to the site using lowboy trailers with hydraulically
adjustable axles to help distribute the heavy load of the segment (490kN for a typical mainlane
segment to 620kN for an anchor segment). Segments were unloaded with a straddle crane and
stored beyond the abutment of the newly constructed portion of the bridge on the approach
roadway (see Figure 1.24).

Straddle crane

Lowboy trailer

Figure 1.24 Transporting segments

Once the twin triangular erection trusses were in place, supported by brackets on the piers,
segments were lifted onto a heavy-duty flatbed truck and hauled to the end of the previously
constructed span. The “Texas lifter”, a different type of straddle crane, would lift the segments off
the flatbed truck and move them forward over the trusses while turning them 90o. The segments
were lowered onto three sets of roller jacks, adjustable in all directions, and rolled into approximate
position (Figure 1.25).

25
 Precast
segments

Roller jacks

Erection truss

Figure 1.25 Locating segments on the erection trusses

The first anchor segment and adjacent typical segment were epoxied together, with six
post-tensioning threadbars anchored in external blisters providing the temporary clamping force
required for a good fit during epoxy curing. These two segments were then surveyed and adjusted
into their final position, allowing for bearing compression. Then, epoxying proceeded for the
remaining segments. Figure 1.26 shows epoxying in progress from the deck, with the external
temporary post-tensioning blisters on the wings in view. These temporary post-tensioning
anchorage blisters on the deck (2 per segment) were made of steel, and could be reused. Four
concrete temporary post-tensioning blisters were cast into each segment, and were located near the
juncture of the girder flanges and webs in the core of the box girder. The location of the six
external post-tensioning blisters in each typical segment is shown at the bottom of Figure 1.26.

26
 Texas lifter

Temporary P/T
blisters (steel)

Span epoxied
to this point

Worker applying
epoxy

External P/T
blisters (steel)

External P/T
blisters (concrete)

Typical Section
Figure 1.26 Epoxying segments

Prefabricated prestressing tendons were pulled from the dead end to the live end of the
girder and stressed in pairs with two jacks and two hydraulic pumps (see Figure 1.27). The
elastomeric bearing top surface to bearing plinth gap was grouted, the jacks supporting the
segments on the truss were released, and the erection trusses were prepared for launching. Trusses
were either self launching, requiring a launching nose on each truss, or were launched using a
ground-based crane. Segments were also placed on the trusses using a ground-based crane when
required, since the Texas lifter could only service one set of trusses at a time.
27
 Winch for ram

Rams

Work platform

Figure 1.27 Stressing a pair of mainlane girder external tendons

1.4.3 Ramp P girder

The alignment of the flyover Ramp P that connected northbound IH35 with northbound
U.S. 183, shown in Figure 1.28, required that the structure pass over the mainlanes and the
frontage roads of both of those highways. The congestion in the area beneath the ramp reduced
the space available for supporting substructure, requiring longer spans for the ramp than were
typical for the rest of the project. The alignment of the ramp also followed a fairly tight radius of
221m. Both of these constrictions made span-by-span construction, used on the mainlane girders
and most ramp girders, impractical for the construction of Ramp P. Therefore, balanced cantilever
construction was used to build five spans of Ramp P.

28
 und IH-35

nd IH-35
P13
38.

South-Bo

North-Bou
1m P14
N
54
.9
m
US 18
Ram 3 North
p “R -Boun
” d Main
lanes

43
(Spine
US 18 “C”)

.3
3 Sou

m
th-Bo
Pier P15 und M
ainlan
es (Sp
ine “D
”)

54.
Pier P16

m9
P17

Flyover

38.1m
Ramp “P”
P18

Figure 1.28 Plan of U.S. 183 at IH35

The balanced cantilever constructed spans of Ramp P were selected for instrumentation
for several reasons. One of these was the desire to study the method of construction itself since
little data are available on balanced cantilever erection. Also, the box girder shape, true continuity
of the five span bridge unit, and horizontal curvature were entirely different than for the mainlane
girder. Ramp P provided additional topics for study beyond those found with the mainlane girder,
such as torsional response of the box girder. Measurements for some topics, such as diffusion of
post-tensioning forces and thermal gradients, could be compared with those of the mainlane using
the different box girder shapes of Ramp P and the mainlane as the variable.

Figure 1.29 shows the span arrangement and construction sequence stages for Ramp P.
Clearance of the mainlane and frontage roads beneath Ramp P required unequal span lengths and
unique construction staging and post-tensioning sequences. Precast segments were hung in place
with ground based cranes and epoxied to previously erected segments in similar fashion to the
mainlane girder. Top flange cantilevering tendons were stressed after each pair of segments had

29
been epoxied in place, one on either end of the cantilever unit to balance moments in the pier (see
Figure 1.30).

Upstation

38.1 m 54.9 m 43.4 m 54.9 m 38.1 m

Pier P18 Pier P17 Pier P16 Pier P15 Pier P14 Pier P13

Phase I: Construction of the P17 cantilever unit and the completion of the upstation endspan.

Phase II: Construction of the P16 cantilever unit.

Phase III: Construction of the P14 cantilever unit. Construction begins on the P15 unit.

Phase IV: Completion of the downstation endspan and the P15 cantilever unit.

Phase V: The central span is completed. The free cantilever wings of the P15 and P16 units are
extended by one segment.

Phase VI: Final bearing adjustaments are made on P17. The upstation 54.9 m span is completed.
The same process is repeated on the downstation half of the bridge.

Phase VII: The last external tendons are placed and stressed. The barriers are cast and a wearing
surface is applied to the deck.

Figure 1.29 Construction sequence of Ramp P

30
 Anchored cantilevering
Ducts for cantilevering
tendons tendons

Temporary P/T
blister for bottom
flange

Figure 1.30 End view of balanced cantilever ramp under construction

Closure pours were made between cantilever units to make the girder continuous. A
combination of bottom slab internal tendons, external multi-span tendons, and bearing elevation
adjustments changed the state of stress in the box girder to that required for a five-span continuous
girder subjected to traffic loads. Fixity required between the pier capital and the girder during
construction was only maintained at piers P15 and P16 in the final structure. The ramp girder had
heavy anchor segments for post-tensioning tendons over each pier. The dimensions of these
segments are given in Figure 1.31. Typical segments had varying bottom flange thickness near the
pier to increase the moment of inertia of the section for cantilevering moments. Figure 1.32 shows
the typical segment dimensions. Instrumentation was concentrated in the half of the 54.9m span
closest to pier P16.

31
 3960 4270
1910
400
250 460 380 300

300
150 100 6.5
2130

1020 100
150 40
300 360 150
Dimensions in mm
460 760

Figure 1.31 Ramp P anchor segment dimensions

3960 4270
1910 1050 480
400
250 200 300
360 6.5
1420

100
300
2130

1100
100
B
40
300 A

460 760

Dimensions in mm

Dimension “A” varies from 360mm to 280mm

Dimension “B” varies from 880 mm to 760mm
Figure 1.32 Ramp P typical segment dimensions

Precasting of the ramp segments was performed by the short line method, nearly
identically to the mainlane segments. The segments on Ramp P were somewhat more complicated
and congested than any typical segment for spans of the bridge erected using the span-by-span
method. The segments for Ramp P had top flange cantilevering tendon ducts that transitioned in
alignment from outside the web reinforcing to their anchorage point at the centerline of the web.
These segments also had bottom flange internal tendon ducts, variable depth bottom flanges, and
sometimes vertical or horizontal deviators for guiding the external tendons along their required
profile in the horizontally curved bridge.

32
1.4.4 Large ramp pier
The large ramp pier was designed as a hollow, octagonal column cross section with
406mm thick walls. The column consists of precast typical segments of 2.44m and 1.22m lengths.
The hollow section reduced foundation costs and facilitated the transportation and erection of the
pier segments. A solid capital segment is located at the top of the column providing an anchorage
zone for pier post-tensioning tendons. The capital also served as the bearing surface for the ramp
superstructure and an anchorage for tie-down bars connecting the superstructure to the capital
during balanced cantilever erection. Figure 1.33 and Figure 1.34 show the general configuration of
the large ramp pier.

33
 A A

381
Capital
2515

25
Reveals @ 1219-mm O.C.
Column Height “H” Varies

B B

Tendons
(19 ~ 15-mm dia. Strands)
See View A-A of Figure 1.34
Varies

Bottom precast segment extends

Cast-in-Place

into cast-in-place base.

Base
1905

C C
Footing
1524

1070-mm f
3660-mm long
Drilled Shafts

Note: All dimensions in mm.

230 1070 2134 1070 230
3660

Figure 1.33 Typical large ramp pier: elevation view

34
 781 1413
P.T. Tendons

1413
1500
864 1880 864
3607

VIEW A-A

P.T. Tendons
752 781 752

406

946
203 1880 203
2286

SECTION B-B
838
762
4267
2743
2590

P.T. Tendons
762
838

1067-mm dia.
Drilled Shaft
533 2590 533
762 2134 762
3660

Note: All dimensions

SECTION C-C in mm.

Figure 1.34 Typical sections: large ramp pier

35
 Pier P16 was selected for instrumentation for several reasons. First, the pier had
unobstructed exposure to the sun over much of its height until the ramp superstructure was
constructed. This allowed many months of thermal measurements to be taken without having
shade as a variable. This pier was also adjacent to the longest cantilever on the structure, about
27.4m, so it would also be subjected to the largest unbalanced cantilever moment. Finally the pier
was located beneath the instrumented superstructure span, so a common data acquisition system
location could be used.

All of the segments for pier P16 were cast at the same precasting yard as for the
superstructure segments. After precasting, the segments were placed in storage at the same site
until pier erection activities began. The U.S. 183 project plans called for 13 segmentally
constructed piers, of which 12 were located along the flyover ramp P. Large ramp piers along
ramp P ranged in height from 8.2m to 21.6m. The piers were comprised of a total of fifty-seven
2.44m segments, seven 1.22m segments, and thirteen capital segments. Variations in pier height
were accommodated with a cast-in-place closure pour between the top of the footing and the first
precast segment. This method eliminated the need to cast variable height segments.

Precast pier segments were cast in a specially designed casting bed at the precasting
facility. The bed consisted of tandem casting platforms, each with hydraulic jacks to control
segment attitude and alignment (see Figure 1.35). Two platforms were utilized to enable
constructors to cast one pier segment per day. Figure 1.36 illustrates the typical cycle for casting
pier segments.

36
 Casting bed for
current day’s cycle Form

Casting bed for

next cycle

Match casting
segment

Figure 1.35 Pier segment casting machine

37
 Phase 1

• Bulkhead segment placed,

surveyed, and leveled.
• Bond breaker applied to
top face.

Phase 2

• Outer form placed, surveyed.

Phase 3

• Reinforcement cage inserted.

• Inner form placed.
• Concrete placed, then cured overnight.

Phase 4

• Inner form disassembled, removed.

• Outer form removed in halves.

Phase 5

• New segment moved to second bed to

serve as new bulkhead.
• Bulkhead segment moved
to yard for storage.

Figure 1.36 Typical segment-per-day casting sequence

38
 The short-line method of match casting was used to cast all pier segments. A typical
casting cycle began with the stripping of the forms and placement of the previous day’s segment
into one of the sides of the tandem casting bed using the straddle crane. This segment served as
the match casting segment against which the current day’s segment was cast. The match casting
segment from the previous day was then hauled to storage with the straddle crane. This technique
ensured a perfect fit between segments, and reduced systematic alignment errors. The match
casting segment was then checked by surveyors to ensure that it was plumb and aligned correctly.

The cycle continued with the placement of the steel outer form onto the top of the match
casting segment (see Figure 1.37). The form was supported by the segment and a steel support
cage with a set of jacks that could be leveled independently of those supporting the bulkhead
segment (Figures 1.38 and 1.39). The reinforcement cage, that had been assembled in a jig during
the previous day’s casting operations, was then lowered into the outer form. Steel ductwork for
the post-tensioned tendons and temporary post-tensioning bars, PVC drainage piping, and
electrical wiring ducts were placed at this point in the cycle. Steel pipe mandrels were inserted
into all ducts to ensure good duct alignment between segments. The inner core form was placed,
and blockouts were installed in the top face for temporary post-tensioning bar anchor plates.
Lifting hooks were also installed so that the segment could be carried by the yard’s straddle crane.
Shear keys were formed into the top segment face with small, hand-held key forms inserted into
the fresh concrete. The form was made plumb with the match casting segment by the surveyors,
then concrete was cast.

39
 Figure 1.37 Placement of outer pier segment form

Form

Steel frame
Match casting
segment

Figure 1.38 Casting bed with steel support frame for formwork

40
 Match casting
segment

Hydraulic jacks

Figure 1.39 Hydraulic jacks used to align the bulkhead pier segment

The segmental piers on the U.S. 183 project were designed to be post-tensioned from the
top of the capital segment. For this reason the post-tensioning ducts were designed with a U-
shaped bend at the base of the pier. Figure 1.40 shows this duct configuration at the pier’s
foundation. After completion of the cast-in-place foundation, an adjustable steel frame was
installed on top of the footing to support the first segment, PC16-1, on pier P16 (see Figure 1.41).

41
 I
Up STA

H H

I
Plan View

101.6-mm dia.
Tendon Ducts

Base Segment
PC16-1

Cast-In-Place
Base

Cast-In-Place
Foundation

Section H-H Section I-I

Radius = 1.02 m
Figure 1.40 U-shaped tendon ducts located in pier foundation

42
 P/T ducts

Figure 1.41 Steel support frame for first precast segment

After alignment and leveling of the first segment, a concrete base was cast around it.
During construction of the cast-in-place base, the concrete was mechanically vibrated and forced
to flow up against the bottom face and into the core void of the segment, ensuring good contact
between the base and the bottom segment. Figure 1.42 shows the cast-in-place base during
erection of the pier segment PC16-2.

43
 Epoxied surface

Cast-in-place base
Scaffold

First pier segment

Figure 1.42 Cast-in-place base forming a rigid moment

connection between base segment PC16-1 and foundation

Subsequent segments, as shown in Figure 1.42, were placed using a mobile crane.
Segment joints were epoxied like the superstructure segments, but the post-tensioning threadbars
used to provide the squeezing force on the joint were left in place permanently. Figure 1.43 shows
the hydraulic ram used for post-tensioning the threadbars. A portable scaffold was slipped up the
pier by the crane as construction proceeded.

44
 Ram

Chair

Figure 1.43 Hydraulic ram stressing temporary post-tensioning bars

Post-tensioning of the pier’s two main tendons followed the placement of the capital
segment PC16-8. Tendons consisting of 19-15mm diameter strands were cut to an approximate
length, lifted by crane to the top of the pier, and inserted into the tendon ducts. The crane was
then used to pull the tendons through the full lengths of the ducts using a fish wire. The tendons
were stressed with a ram hung from the crane.

1.4.5 Transition and modified spans

Since the U.S. 183 segmental project is an urban viaduct, limited access ramps had to be
provided to the structure at various points. The transition and modified spans joined the single-
cell box girder ramps to the single-cell mainlane girders at Lamar Boulevard and at IH35. The
transition spans were designed as three-celled box girders, thus providing the four webs necessary
where the ramps split from the mainlane girder (see Figure 1.44). The transition spans were
originally designed to be constructed as two single-cell, single winged girders with cast-in-place
top and bottom flanges completing the central cell. The contractor decided not to modify the
erection trusses used on the project to handle the eccentric load of a single winged girder, and not
to purchase the two additional casting machines required for the transition segments. Instead, he
opted to cast the entire set of transition spans in place on timber falsework. Precast or not, the
transition spans provided an opportunity to study a wide three-celled girder under various live
loads. Two span unit C13 was selected for study.

45
 Pier C36

Pier C37

Pier C38
Span C36 Span C37
40.84m 40.84m
A

Diaphragm Girder webs

A
PLAN

Varies 1994mm Varies 1994mm 5930mm

SECTION A-A
Figure 1.44 Plan and section of transition Unit C13

Modified bridge unit C15-L2 was also selected for study. This modified bridge unit was
constructed using three spans of modified ramp girder and three spans of modified mainlane
girder, with truncated wings on one side and a cast-in-place closure (see Figure 1.45). This detail
is similar to many twin single-cell box girder bridges that are joined longitudinally at their interior
wing tips to share live loads. Construction of the modified spans required little modification to the
erection equipment for the mainlanes and ramps, with the gore closure formed and cast entirely
from above. Longitudinal post-tensioning tendon layouts were very similar to those in the
mainlane girders. Transverse prestressing was provided by a combination of pretensioned strands
and post-tensioning tendons for the modified spans, and by post-tensioning only for the transition
spans.

46
 Pier L7
Pier L6
Pier L5
Pier L4

Span L6
Span L4 Span L5 39.94m
34.09m 40.20m
A

Cast-in-place gore Weblines

A
Span C41 Span C42 Span C43
34.14m 40.20m 40.07m

Pier C43

Pier C44
Pier C41

Pier C42

2050mm 3820mm Varies 5474mm 5798mm

556mm

Section A-A
Figure 1.45 Plan and section of modified Unit C15-L2

1.5 TOPICS STUDIED ON THE U.S. 183 SEGMENTAL PROJECT

The major topics studied in the current research study were:

1.5.1 Friction losses in post-tensioning tendons

Although considerable research has been conducted measuring friction losses in post-
tensioning tendons, the wide variety of possible geometric configurations for a tendon makes any
additional data useful to a designer. The external tendons on U.S. 183 were of a common design
(Figure 1.46), making friction loss test results for these tendons important. Other prestress losses
include elastic shortening and shrinkage and creep losses. Tendon strains were measured using
strain gauges, so tendon forces could be tracked over time. Gauges in the concrete of the box

47
girders also allowed comparisons between tendon strain changes and the response of the entire
bridge to prestressing forces, temperature changes, and creep and shrinkage.

19 - 15mm dia. strand tendons

Figure 1.46 External tendon layouts for the U.S. 183 box girders

1.5.2 Thermal gradients and their effects

Thermal gradients can cause stresses in continuous box girders of the same magnitude as
stresses from live loads. The daily heating and cooling of a box girder top flange causes the
bridge to deflect up and down, leading to induced moments for continuous structures (Figure 1.47)
that may be opposite in direction of moments from dead load and live load. Overestimation of the
magnitudes for the design thermal gradient for a structure can lead to great amounts of extra
prestressing. Underestimation can lead to a cracked structure. Measurements were taken on U.S.
183 to determine the actual thermal gradients as a check of the recommended design thermal
gradients, to record the frequency of the magnitudes of thermal gradients, and to measure the
response of the structure to thermal loads. The traditional analysis technique used to calculate
concrete stresses from thermal loads makes broad assumptions as to the response of the girder.

Positive bending moment

No bending moment induced by boundary conditions

Thermal Gradient in Superstructure

Bending moments in
piers and superstructure

Thermal Gradient in Piers

Figure 1.47 Structural response to thermal gradients

48
1.5.3 Strut-and-tie design regions
Concrete box girders have many local regions called D-zones, or discontinuity zones that
can not be designed using normal design techniques for beams or columns. Strut-and-tie modeling
of the regions gives the designer a useful method for the design of D-zones, such as for the anchor
zone in Figure 1.48. D-zones under study at U.S. 183 included two different deviators, two
different anchor diaphragms, an anchorage blister, and two different pier capitals. Each of these
D-zones was designed for strength and safety, yet had to be detailed to give satisfactory
performance at service load levels. Instrumentation gave an indication of why some of these D-
zones behaved very well at service load levels, while some others behaved poorly with large
cracks.

compression strut
tension tie

Depth of Girder
node

Figure 1.48 Strut-and-tie model for diffusion of a point load

1.5.4 Stress distributions in box girders

Low-level viaduct type bridges such as U.S. 183 generally have short to medium span
lengths, yet may need to have a wide deck for three lanes of traffic or more. These proportions for
a single-cell concrete box girder make predictions of stress using simple beam theory inaccurate,
particularly near the post-tensioning anchorage zones, as shown in Figure 1.49. Stress distributions
can also be difficult to predict because of the construction sequence and curved alignment of a
structure, such as on Ramp P studied at U.S. 183. Measurements were taken and stress
distributions were plotted for many load cases on two different box girders and the segmental pier.

49
 Longitudinal concrete compressive
stresses at a section 2m from
the anchorages

A A

2m
Plan

6 Deviated 19 Strand Tendons

Section A-A
Figure 1.49 Diffusion of post-tensioning forces

1.5.5 Behavior of semi-continuous multiple span units

The mainlane spans and many of the ramp spans were designed to be simple spans under
dead load and live load, with “poor boy” continuity of the deck elements over the piers. This
“poor boy” continuity was obtained by using an inexpensive cast-in-place deck slab to provide a
riding surface between these simple span girders at locations other than the finger joints, while no
connection was provided between successive girder webs or bottom slabs. The joint also served to
maintain alignment of the bridge unit, since guided bearings were provided at a minimum of
locations. Construction of this joint is shown in Figure 1.50. Measurements were taken to
determine how much negative moment this detail was carrying under live load. The moment
couple would be developed between the slab as a tension member and the elastomeric bearings in
shear. The development of a negative moment was not a performance characteristic desired by the
TxDOT design engineers. Also, the response of the joint to thermal gradients in the box girder
was measured.

50
 Parapet under construction

Precast anchor segments

Cast-in-place “Poor-Boy” joint

Elastomeric bearings

Elevation at Pier D6
Figure 1.50 Construction of a cast-in-place continuity joint

1.6 SCOPE AND CHAPTER ORGANIZATION

The topics selected for study on U.S. 183 were thought to be those most important to the
segmental bridge industry at the present time. The information gained on some of the topics, such
as friction losses and diffusion of post-tensioning forces, can be applied to many types of
structures. Research on other topics, such as deviator behavior and thermal gradient effects,
applied only to concrete box girders. The primary intent of the research was to verify the behavior
of as many structural elements as possible through measurements. Therefore, this dissertation
emphasizes the measured results, using calculated results only for purposes of comparison. It was

51
not within the scope of this dissertation to refine all design procedures that proved to be deficient.
Instead suggestions were made and data presented to help aid in future designs. Furthermore,
much of the data presented was intended to quantitatively verify design code provisions, especially
the recommended design thermal gradient.

The chapters that follow contain details of the instrumentation used and background
information on the structure, the study of the five main topics, and a summary of
recommendations and conclusions. The Appendix has a summary of the thermal gradient records
for the mainlane and ramp box girders. These chapters are:

1. Introduction

2. Instrumentation Program

3. Prestressing Force Losses in Post-Tensioning Tendons

4. Thermal Gradients and Their Effects

5. Load Response of Box Girders

6. Behavior of D-zones

7. Behavior of a Semi-Continuous Unit

8. Observations, Recommendations, and Conclusions

Appendix

Chapters 3 through 7 are generally organized as follows:

1. Problem description and technical background

2. Literature review and previous field experience

3. Instrumentation program and field measurements

4. Comparison of results with calculated results

5. Recommendations

Every chapter has been written and illustrated as an individual document. Chapter 8
provides a summary of the findings, and requires a previous understanding of the topics studied.

52
 CHAPTER 2
INSTRUMENTATION PROGRAM

2.1 GENERAL APPROACH TO MEASUREMENTS

The important measurements taken on this research project can be classified into two
main categories: Temperatures and deformations. The measurement of concrete, steel, or air
temperatures was easily made with thermocouples located in or on the material in question.
Deformations, on the other hand, had to be measured with a variety of instruments from strain
gauges to extensometers, each with unique installation requirements and wiring circuitry, as
required. Other categories of measurements included the measurement of force, using a load cell,
and the measurement of pressure, using a pressure transducer. These instruments basically are
capable of measuring deformations internally and give an electrical response proportional to force
or pressure.

The instruments used on the project can also be placed into two main categories: those
that could be read automatically with a datalogger, and those that had to be read manually.
Measurements that had to be taken at frequent intervals, such as once every minute during a live
load test, were read electronically with a datalogger. Measurements that had to be taken hourly for
long periods of time, such as cross-sectional temperatures, also were taken electronically with a
datalogger. Instruments requiring manual readings often served as a redundant system for the
electronic instruments. For example, the manually read Demec extensometer was used to measure
external strains as a check on internal strain meters. Manual reading sets often required 45
minutes to complete during each interval of a test, limiting their usefulness and inviting human
error. Most of the electronic transducers used on the project were cast within concrete, and
therefore had to be read with a datalogger. These transducers, mainly strain gauges and
thermocouples, were prefabricated at the laboratory to the greatest extent possible, then tied into
the rebar cage as is shown in Figure 2.1. The construction or precasting schedule often required
that instrumentation be installed in the course of one day.

53
 Bar cage fabrication jig

Bar cage storage jig

Figure 2.1 Gauge installation

Durability of the instrumentation and of the data acquisition systems was a primary
concern. Many of the transducers were quite fragile, especially the strain gauges. The gauges
were protected from direct impact with multiple layer coatings, but could not readily be protected
from vibration, such as during concrete placement shown in Figure 2.2, or more critically from a
hammer impact. Many strain gauges were lost entirely during concrete placement, or showed
evidence of delamination from the steel to which they were bonded. Since much of the project
was precast, many of the gauges had to be installed months ahead of the time they would be
needed to take measurements. Although most of the gauges that survived the casting procedure
worked well for months or years afterward, some of the gauges demonstrated erratic output or quit
working altogether. Thermocouples had nearly a 100% survival rate for the length of the project,
easily surviving the casting process and not suffering any long-term degradation. The dataloggers
selected for the project were built for field use, and had been used with great success on previous
field studies, especially at the San Antonio Y [10]. Many of the instrumentation systems used

54
were developed in a study by Arrellaga [11]. Testing and further refinement of the systems was
performed on this research project based on recommendations by Roberts [10] and Arrellaga [11].

Figure 2.2 Vibrating concrete near instrumentation

2.2 INSTRUMENTATION

2.2.1 Strain gauges

The electronic resistance type strain gauges selected for the project were all
Micromeasurements Group 350Ω gauges. This nominal gauge resistance was selected because
many of the anticipated measured strains were quite small, and the excitation voltage was only 5
Volts. A 350Ω gauge would increase the output voltage over a standard 125Ω gauge. The gauge
dimensions were chosen such that the gauge could be easily installed on a single wire of a 15mm,
7-wire strand. The electronic resistance gauge, as shown in Figure 2.3, works by changing its
resistance linearly with changes in gauge length. The gauges used on the project were also
thermally compensated for use on a material with a particular coefficient of thermal expansion,
which was steel and concrete in this case. The resistance of the gauge changes with its own
temperature such that thermal strain in the gauged material does not produce any appreciable
output over a wide temperature range. Therefore, strain output from the gauge has a stress

55
associated with it, since uniform contraction and expansion strains have been subtracted out
automatically.

As gauge strains, wire

elongates and wire’s
lead wires resistance increases
L L dL

Internal wire
Direction of strain
Plastic gauge backing
Figure 2.3 Electronic strain gauge

All strain gauges on the project were wired as one quarter of a Wheatstone bridge circuit,
shown in Figure 2.4. The output voltage is produced when current is not equal on the left and
right sides of the bridge. A three-wire system was used with all strain gauges to eliminate any
output from temperature induced resistance changes in the gauge lead wires.

350 Ω Electronic
Strain Gauge

Wire Resistances
Typically 3.5 Ω
35

Excitation/Input 5V
0
Ω

35
0
Ω

Ω
0
35

Output (mV)

Figure 2.4 Wheatstone Bridge circuit

56
 Strain gauges were always installed on steel. The “S” series gauges were installed
directly on important steel elements, such as reinforcing bars, while “C” series gauges were
installed on small lengths of steel embedded in the concrete with the intention of measuring
concrete strains. S-gauges were installed extensively on the heavy reinforcing bars in D-zones,
such as in the anchorage blister shown in Figure 2.5. Gauges were located where cracks were
expected to form so that the measured strain could easily be converted into a force. S-gauges were
also installed on post-tensioning tendons and post-tensioning threadbars for the same reason.
Gauges on the main external post-tensioning tendons, shown in Figure 2.6, had to be installed in
the girder with the duct cut away. Gauges on the threadbars, such as the tiedown bars in the
capital and anchor segments at pier P16, could be installed at the laboratory.

The laboratory-installed gauges almost always gave better performance and had a longer
life than gauges installed at the site under uncontrolled temperature and humidity. The
temperature of the material to be gauged apparently had the greatest influence. S-gauges were
installed on the structural steel pipes for the mainlane Y-pier, shown in Figure 2.7, in direct
sunlight. This exposure caused great temperature fluctuations in the pipe itself. The glue selected
for all the strain gauging on the project, Micromeasurements Group M-Bond 200, would work
over a wider temperature range than many of the more expensive glues. Heat was used to raise the
temperature of the material to be gauged if it was below 15oC. The glue would not set up
otherwise, and often would not bond to the gauge or the steel.

Anchorage blister
reinforcement

Strain gages

Figure 2.5 “S” series gauges installed on rebar

57
 Strain gauges

External tendons, duct removed

Figure 2.6 “S” series gauges installed on prestressing tendon

Strain gauges
under installation
on upstation pipe

Structural pipes

Figure 2.7 “S” series gauges installed on structural pipe

The “C”-series gauges consisted of a strain gauge bonded to a small steel rod, which was
threaded on each end for 2 nuts and a washer. A plastic sleeve was provided so that only the end
nuts and washers bonded to the concrete. The C-gauge was embedded in the concrete at time of

58
casting. This device, shown in Figure 2.8, was originally designed by Stone [12] as a modified
Mustran Cell. The C-gauge was tested and further refined on this project.

To Data Acquisition Unit

32mm dia. Steel Washer
26 Gauge 3-Wire 4.76mm Nut

Leadwires
Electronic Strain Gauge

Protective Shrink Tubing 4.76mm dia. Steel Round

203.2mm Effective Gauge Length

Figure 2.8 “C” series gauge construction

The gauge in its final form used throughout the project consisted of a cold rolled steel rod
4.76mm in diameter approximately 235mm long, and was threaded approximately 20mm on each
end. The effective gauge length between washers was 203mm. The strain gauge was the same
350Ω electrical resistance strain gauge used for gauging the tendon wires and reinforcing bars.
This strain gauge was bonded at the center of the steel rod. An acrylic coating and a small piece
of butyl rubber were applied to the strain gauge after the electrical leads were installed and
isolated from the steel rod. Approximately 165mm of heat shrink plastic tubing was slipped over
the strain gauge, leads, and steel rod. The nuts were #10x32 USC and the washers were
approximately 25mm in diameter. The nuts and washers were placed on each end, giving a
203mm gauge length, and tightened to yield the threads of the 4.76mm rod. The heat shrink
tubing was shrunk to hold wire leads in place and protect the strain gauge. Finally a coating of
epoxy or silicone rubber was applied to the ends of the heat shrink tubing for waterproofing.

Tests were performed to determine the overall length for the C-gauge. If the gauge were
too short it would act as a stiff inclusion in the concrete, reducing the strain measured by the gauge
below acceptable limits. The 4.76mm rod was selected because it was small enough in diameter to
reduce inclusion effects with reasonable gauge lengths, yet large enough for consistent quality
gauging. C-gauges with lengths of 102mm, 152mm, and 203mm were constructed, cast within
concrete specimens, and tested in a load machine. The test results are described in detail by

59
Andres [13]. All the gauges performed satisfactorily, but the 203mm gauge was selected for use
because of its ease of handling. A longer gauge was not considered because of its tendency to be
easily bent and yield with relatively small transverse loads between the washers.

The 203mm gauge was tested further in 152mm by 152mm by 610mm concrete prisms
tested in compression. The C-gauge was cast down the center of the specimen and Demec
extensometer points, described in Section 2.2.7, were placed on the four exterior faces. Strains
were measured mechanically with the Demec gauge and electronically by the C-gauge and a
datalogger. The results of these tests are shown in Figure 2.9. The C-gauges gave very consistent
results, and compared well to the strains measured by the Demec extensometer.

60
 10.0

8.0
Test 1
Stress in MPa

6.0

C-gauge measurements
4.0 on specimens 1, 2, and 3

2.0 Demec measurements

on specimens 1, 2, and 3

0.0
0 50 100 150 200 250 300 350
Microstrain

10.0

Demec measurements
8.0 on specimens 1, 2, and 3
Stress in MPa

6.0 C-gauge measurements

on specimens 1, 2, and 3

4.0 Test 2

2.0

0.0
0 50 100 150 200 250 300 350
Microstrain

Figure 2.9 C-gauge tested against the Demec extensometer

Since the C-gauges were threaded on each end, they were easily installed in a test frame
and pulled in tension. Every C-gauge used on the project was tested in this way and checked for
linearity and slope compared to other C-gauges in the test group. The results of these tests for
gauges C30 through C49 is shown in Figure 2.10. The one bad gauge in the plot was
disassembled, rebuilt, and tested again. Slight variations in slope on the plot were caused by

61
bending stresses in gauges that were slightly bent initially, and did not effect the accuracy of the
gauge once it was cast in concrete.
Instruments C30 - C49
1.5
Pressure transducer output (mV)

1.0

0.5

0.0

-0.5

-1.0
0.2 0.4 0.6 0.8 1.0 1.2 1.4
Strain gauge output (mV)
Figure 2.10 C-gauge calibration

The C-gauges were tied directly to the reinforcement cage, as shown in Figure 2.11.
Additional bars often had to be added to the cage to hold one end of the gauge, since its length was
only 203mm. Lead wires were run along the bars to an outlet blockout. It was found to be
essential to leave some slack where wires crossed bar intersections or changed directions to allow
for adjustments of the bar cage once it was lifted into the form. In retrospect, the C-gauges
worked extremely well in the laboratory and very well in the field when they were not damaged by
flowing concrete or vibration. The gauges’ lifetime was adequate for monitoring all construction
operations and the load distribution live load tests. However, they did not have long term stability.
For monitoring very long term service load behavior, a few vibrating wire strain gauges should
have been installed. The C-gauges were too small to be quickly installed in the field, and were too
easily bent by foot traffic and flowing concrete, especially in the girder webs. The gauge could be
fabricated in a similar manner using a 500mm piece of rebar with threaded ends for testing, and
deformed surfaces for bonding to the concrete. The life of this strain gauge may still be quite
short compared to vibrating wire gauges, which function reliably for several years.

62
 C-Gauge

Figure 2.11 Installation of C-gauges

2.2.2 Thermocouples
Type T thermocouples were selected for the project because of the temperature range of
the measurements to be taken. A typical type T thermocouple circuit is shown in Figure 2.12.
The thermocouples proved to the most durable gauge type used on the project. The thermocouple
wire selected for the project was heavy gauge, wire mesh shielded with a Teflon outer case. The
wire was tough and had high tensile strength, unlike the wire leads used for the strain gauges. The
thermocouple end at the measurement temperature location was twisted and silver soldered, then
protected with heat shrink tubing. A typical thermocouple installation is shown in Figure 2.13.
The stiff constantan wire allowed the lead to be bent away from the rebar that the thermocouple
was mounted on so that concrete temperatures could be measured without the local influence of
the highly heat conductive steel rebar. Thermocouples were also installed in small holes drilled
into the structural steel pipes in the mainlane Y-pier, installed within the strand group of some
external tendons, and hung out drain holes in the girders to measure ambient outside air
temperature.

63
 +
V Data acquisition unit measures the voltage
-

constantan wire
T2 (Temperature
to be measured)
copper wire
T1 (Reference
temperature)
Two temperatures induce a current in the circuit.
Figure 2.12 Thermocouple circuit

Lead bent away

from rebar

Thermocouple wires
soldered and protected

Figure 2.13 Installation of a thermocouple

2.2.3 Linear potentiometers

Linear potentiometers were used on the project to measure movements of the
superstructure with respect to the pier. This device, shown in Figure 2.14 gives an output voltage
signal that changes directly in proportion to the distance the plunger is moved in or out of the
housing. A linear potentiometer can accurately measure length changes on the order of
millimeters, but cannot be used to measure strains in a material because of mounting irregularities,
temperature changes, and the precision of the device itself.

64
 Signal Voltage Excitation Voltage

S+ S- E- E+

Signal varies linearly

with displacement d
from 0 volts minimum
to the excitation
voltage maximum

Figure 2.14 Linear potentiometer schematic

2.2.4 Pressure transducers

Hydraulic rams were used extensively on the project by the contractor to jack the
prestressing force into the post-tensioning tendons, and by the researchers in various laboratory
tests. A pressure transducer, shown in Figure 2.15, gives an output signal linearly changing with
pressure. This device was wired into the dataloggers during post-tensioning to record tendon force
at the same time strain gauges were read.

Signal Voltage Excitation Voltage

S+ S- E- E+

Signal varies linearly

with pressure

Pressure
source

Figure 2.15 Pressure transducer schematic

65
2.2.5 Load cells
Load cells were used during post-tensioning tests to measure force at the dead end of a
tendon. A load cell, shown in Figure 2.16, operates much like a pressure transducer but with the
signal voltage changing with applied force on the sides of the load cell. A load cell actually has
strain gauges inside wired into a Wheatstone circuit. The load cells used on the project had a hole
through their center for the tendon to pass through.

Signal Voltage Excitation Voltage

Signal varies linearly S+ S- E- E+

with force

Applied
force

Figure 2.16 Load cell schematic

2.2.6 Data Acquisition system

The data acquisition systems used on the project all consisted of Campbell 21X
dataloggers, Campbell AM416 relay multiplexers, a 12V marine battery, and a laptop personal
computer. The containment for the data acquisition systems varied at different locations on the
project, as shown in Figure 2.17. The mainlane girder had two cages containing the dataloggers,
terminal boards, multiplexers, and a battery. The data acquisition system nearest pier D6 is shown
in the top photo in Figure 2.17. Instrumentation for the mainlane Y-pier was also wired to this
data acquisition system after temporarily being wired to the weatherproof box shown in the middle
photo in Figure 2.17. This weatherproof box later housed the data acquisition system for pier P16.
Each segment of Ramp P with large amounts of instrumentation had its own data acquisition
system, with the dataloggers and multiplexers stored in individual sealed electrical junction boxes.
This allowed readings to be taken as the segments were being hung in balanced cantilever. This
system proved to be far superior to the cages and terminal boards used in the mainlane girder. The
ramp data acquisition systems were installed and tested while the segments were in storage, while

66
the mainlane girder systems had to be assembled inside the partially constructed span. All
terminal connections were also eliminated in the ramp girder data acquisition systems.

Mainlane DAS

Pier P16 DAS

(weatherproof)

Ramp P DAS

Figure 2.17 Data acquisition systems used on the project

The 21X datalogger has 8 channels for differential voltage measurements, and four
terminals for switched excitations. The number of datalogger input channels are expanded with
the use of up to four AM416 relay multiplexers. The 21X is programmed to control the
multiplexers with signals from the excitation and control ports shown in Figure 2.18. The
multiplexers are also powered through the 12V supply on the face of the 21X. Each multiplexer
has 16 full bridge channels, which can be divided into 32-quarter bridge, or differential voltage

67
channels. Each of the two COM terminals on the multiplexer are wired to a different input
channel on the 21X, providing a maximum of 128 differential voltage channels to be read with one
datalogger.

L2 L1

COM
COM
SHIELD SHIELD
H2 H1
L2 L2 L2
H2 H2 H2
14 10 L1
L1 L1 6
H1 H1 H1
SHIELD SHIELD SHIELD
L2 L2 L2
H2 H2 H2
13 9 L1
L1 L1 5
H1 H1 H1
GND
RES

CLK

12V

AM416
MULTIPLEXER

1 2 3 4 5 6 7 8
H L H L H L H L H L H L H L H L

EXCITATION CAO CONTROL PULSE INPUTS

+12

1 2 3 4 1 2 1 2 3 4 5 6 1 2 3 4

21X DATALOGGER

Figure 2.18 Wiring the AM416 Multiplexer to the 21X Datalogger

Strain gauges were wired as a quarter bridge on the multiplexer, with the remaining
three-quarters of the bridge wired on the 21X datalogger, as shown in Figure 2.19. Actually two
individual Wheatstone bridge circuits are shown in Figure 2.19, with the multiplexer acting as two
16 channel multiplexers. The H1 and L1 switched terminals are internally wired to the H1 and L1
COM, and the H2 and L2 switched terminals are wired to the H2 and L2 COM. Variations in
resistance at the relays, which would show as gauge output, are small. The AM416 multiplexer
has gold clad silver alloy contacts rated at a resistance of 50 mΩ. Current is supplied to the gauge
68
through the multiplexer switches at an amperage well below the maximum current rating for the
switch. A variation of 50 mΩ resistance at the switches would correspond to only a 14 microstrain
apparent output by the gauge being switched. All leads for the 350Ω precision resistors were
identical in size and length to keep resistances equal on all sides of the bridge circuit.

350Ω strain gauge

L2 L1

COM
(typical)

COM
SHIELD SHIELD
H2 H1
L2 L2 L2
H2 H2 H2
14 10 L1
L1 L1 6
H1 H1 H1
SHIELD SHIELD SHIELD
L2 L2 L2
H2 H2 H2
13 9 L1
L1 L1 5
H1 H1 H1
GND
RES

CLK

12V

350Ω strain
gauge AM416
(typical) MULTIPLEXER
350Ω 350Ω

350Ω
1 2 3 4 5 6 7 8
H L H L H L H L H L H L H L H L

350Ω
350Ω
EXCITATION CAO CONTROL PULSE INPUTS
+12

350Ω
1 2 3 4 1 2 1 2 3 4 5 6 1 2 3 4

21X DATALOGGER

Figure 2.19 Wiring diagram for 350Ω strain gauges including the Wheatstone circuit

A test was performed to find if apparent gauge output was being recorded because of
temperature variations at the datalogger or at the gauge. Temperature changes at the datalogger
and the completion circuit for the bridge were expected to change resistances that could not be
compensated by the Wheatstone bridge circuit. Also the strain gauges were compensated for a
coefficient of thermal expansion of the gauged material of 10.8/oC while the actual coefficient of
thermal expansion of the concrete produced by the batch plant at the casting yard was about

69
9.9/oC. Figure 2.20 gives the temperature of the datalogger and the temperature at a thermocouple
located next to a C-gauge cast within a 152mm by 305mm concrete cylinder. A 25oC temperature
variation was used because that is approximately the seasonal temperature variation at the bridge
site. The concrete specimens were given 24 hours to achieve uniform temperature. Prior to this a
thermal gradient exists and the gauges would measure a strain caused by self-equilibrating internal
stresses in the concrete cylinder. Toward the end of each 24-hour period, as seen in Figure 2.21,
the gauges in cylinders 1, 3 and 4 showed nearly identical output. The gauge in cylinder 2 showed
signs of partial debonding. Using the gauge factor of 0.0019002, the maximum gauge output
variation was about 9.5 microstrain.

40
Concrete cylinders
Temperature (Celsius)

30

20

10
21X Datalogger
0
0 12 24 36 48 60 72 84 96
Time (Hours)

Figure 2.20 Concrete specimen and 21X datalogger temperatures recorded during temperature
effects test

0.035
0.030
Concrete test cylinders 1, 3, and 4
Gauge Output (mV)

0.025
0.020
0.015
0.010
Concrete test cylinder 2
0.050
0.000
12 24 36 48 60 72 84 96
-0.005
-0.010
Time (Hours)
Figure 2.21 C-gauge output recorded during temperature effects test

70
Thermocouples were wired for differential voltage measurement, as shown in Figure
2.22. The thermocouple circuit does not require any excitation or completion circuits,
and is the simplest device used on the project.

Copper
Constantan L2 L1

COM
COM
SHIELD SHIELD
H2 H1
L2 L2 L2
H2 H2 H2
14 10 L1
L1 L1 6
H1 H1 H1
SHIELD SHIELD SHIELD
L2 L2 L2
H2 H2 H2
13 9 L1
L1 L1 5
H1 H1 H1
GND
RES

CLK

12V

Cu

Thermocouple AM416
(typical)
MULTIPLEXER
Copper

Constantan

1 2 3 4 5 6 7 8
H L H L H L H L H L H L H L H L

EXCITATION CAO CONTROL PULSE INPUTS

+12

1 2 3 4 1 2 1 2 3 4 5 6 1 2 3 4

21X DATALOGGER
Figure 2.22 Wiring diagram for thermocouples

71
 The excitation and the signal for linear potentiometers have a common ground, therefore
the device can be read as a single ended measurement. With the wiring shown in Figure 2.23, one
multiplexer can switch 32 channels of linear potentiometers.

L2 L1

COM
COM
SHIELD SHIELD
H2 H1
L2 L2 L2
H2 H2 H2
14 10 L1
L1 L1 6
H1 H1 H1 E+
SHIELD SHIELD SHIELD -
E-,S S+
L2 L2 L2
H2 H2 H2
13 9 L1
L1 L1 5
H1 H1 H1 E+ S+
E-,S
-
GND
RES

CLK

12V

Linear pot
(typical)
AM416
MULTIPLEXER

1 2 3 4 5 6 7 8
H L H L H L H L H L H L H L H L

EXCITATION CAO CONTROL PULSE INPUTS

+12

1 2 3 4 1 2 1 2 3 4 5 6 1 2 3 4

21X DATALOGGER
Figure 2.23 Wiring diagram for linear potentiometers

Pressure transducers and load cells are both full bridge devices requiring a separate
excitation circuit and differential voltage signal circuit. The wiring used for these devices is

72
shown in Figure 2.24. Each device requires one entire channel on either the 21X datalogger or the
AM416 multiplexer.

E+
E-
Load cell or
S+ pressure transducer
S-

1 2 3 4 5 6 7 8
H L H L H L H L H L H L H L H L

EXCITATION CAO CONTROL PULSE INPUTS

+12
1 2 3 4 1 2 1 2 3 4 5 6 1 2 3 4

21X DATALOGGER
Figure 2.24 Wiring diagram for pressure transducers and load cells

2.2.7 Demec extensometer

Despite the reliability of the C-gauges, gauges were often damaged during casting. Often
the damage was obvious because the signal was lost entirely, or the steel rod for the C-gauge was
bent enough to cause plastic strains in the steel beneath the gauge. This often led to erratic
behavior of the gauge. Gauges often debonded from the steel after a certain amount of time. If
one of the two latter problems occurred, it would not be known until analysis of the data began. It
is difficult to make the judgment that a gauge has gone bad by visually inspecting the raw data or
even the plotted data in some cases. A simple and reliable backup to the concrete strain gauge was
to use a Demec extensometer.

A Demec extensometer shown in Figure 2.25, is a mechanical device that measures strain
on the surface of concrete or any material. Two steel points are installed into the concrete after it
has hardened. These points can be glued to the concrete surface. However, HILTI® brand Hit
anchors were used on this project to anchor the points into the concrete to a depth of about 32mm.
A small 0.8mm diameter hole was drilled into the steel nail head of the Hit anchor. The Hit
anchor was installed into the concrete at the precasting yard storage area. A 10mm hole was
73
drilled into the concrete. A small amount of epoxy was injected into the hole. The Hit anchor was
placed in the hole and hammered firm. Another Hit anchor was then placed a distance of 400mm
away from the first point, and in the direction that the strain was to be measured. The Demec
extensometer has points on each of its ends that fit into the holes in the Hit anchors, shown in
Figure 2.26. One of the points can pivot to accommodate movement between the Hit anchors, and
a dial gauge on the reader registers the amount of movement. The Demec extensometer
mechanically measures changes in the distance between the two 0.8mm holes in the Hit anchors
and gives readings in terms of strain. The Demec extensometer used in the U.S. 183 study had an
accuracy of 4x10-6 m/m. Demec measured strains gave alternative strain data to compare with the
electronic data from the C-gauges.

Figure 2.25 The Demec extensometer

Dial Gauge

Dial Gauge Plunger

Detail showing how the
Demec extensometer
point fits into a hit anchor.

Movable point can pivot

Measurement
400 mm Length in Zero Position

Figure 2.26 Demec extensometer schematic

74
 Demec strains are less accurate and slower to read than data from than the electronic
strain gauges. Because Demec readings are taken manually, they can suffer from human error.
Also the drilled Hit anchors were damaged on several occasions, and tended to accumulate dirt
and grout dust in the drilled locating holes. The Demec extensometer gave values of strains at the
surface level of the concrete, whereas the concrete gauges were embedded a certain depth and
provided strain readouts for concrete beneath the surface level. Therefore, there was not a direct
comparison between strain values from the two instruments located adjacent to one another. The
Demec points or gauge are also not temperature compensated. Thermal strains that have no stress
associated with them will appear in the data from the Demec points. Therefore, Demec points
were most useful for reading short term changes in concrete strain under controlled loading
conditions, where the temperature of the concrete did not vary significantly. If read properly by a
person who has experience with the Demec gauge, it can provide useful comparison data to that
from the electronic concrete strain gauges. A detailed study of the use and accuracy of Demec
points was performed by Arrellaga [11].

2.2.8 Deflection measurements

The taut wire baseline system was used to measure deflections in the mainlane girders,
and the transition and modified girders. This system, shown in Figure 2.27, could not be used in
Ramp P because of the horizontal curvature. The system consisted of a piano wire anchored at the
centerline of bearing at one end of a span and hung through a pulley at the centerline of bearing at
the other end of the span. A 222N weight drew the wire consistently tight from test to test.
Measurements were made using a digital extensometer, shown in Figure 2.28, mounted on a rail
that allowed the gauge roller to travel up and down. A reference plate made of magnetic stainless
steel was mounted on the bottom of the top flange of the girder. A magnet on the top end of the
gauge shaft held the instrument to the reference plate while the cross arm was lowered by the
reader down onto the wire. Measurements were taken at the span quarter points, as well as near
the span ends to check the geometry of the wire from test to test. This deflection measurement
system was very accurate and reliable, although somewhat dangerous when in use.

75
 Measurement locations

Pulley hanger Wire Fixed hanger

Weight

CL Bearing (typ.)
Elevation
Mainlane Spans D4, D5, D6
Figure 2.27 Piano wire deflection measuring system schematic

Magnetic
mount

Sliding arm

Figure 2.28 Taking deflection measurements

Deflections were also measured with Global Positioning System survey equipment during
several tests. The measurements were taken by Michael Hyzak from Applied Research
Laboratories of The University of Texas in Austin as an exercise in the use of GPS. Accuracy of
the equipment is about 5mm, which was on the same order of magnitude as the measured span
deflections under live load. The GPS receiving units were placed at various points along the spans
during testing, as seen in Figure 2.29.

76
 Live load trucks
GPS receiving unit

Ramp P deck

Figure 2.29 GPS deflection measurement equipment

2.2.9 Tiltmeter
The tiltmeter used to measure slope and twist of the pier P16 cantilevers during
construction was a Model 800P Portable Tiltmeter manufactured by Applied Geometrics. Use of
the tiltmeter also came through the cooperation of the Applied Research Laboratories. The 800P
Tiltmeter uses electrolytic resistance cells to measure angles from the baseline gravity vector (a
straight line towards the center of the Earth, or in other words, a very precise plumb line). The
precision of the 800P is smaller than 1 microradian (1 mm in 1 km). Ceramic tiltplates (also
manufactured by Applied Geometrics) were cemented to the deck of the bridge. The tiltplates,
shown in Figure 2.30, are mounts for the tiltmeter that allow precise placement and orientation of
the tiltmeter for every measurement. The 800P tiltmeter device has indexing bars attached to its
bottom surface so that it can be precisely fitted to the tiltmeter plates every time measurements are
taken. Four measurements were made with the tiltmeter at each tiltplate during tests, and these
measurements were used to calculate the magnitude and direction of tilt. The data were read using
a voltmeter. Further study of the use of tiltmeters for bridge instrumentation has been conducted
by Hyzak, and is currently unpublished.

77
 Figure 2.30 Tiltmeter gauge mounting plate

2.2.10 Epoxy sleeve system

An epoxy sleeve strain measurement system was used to measure tendon strains in the
mainlane girder. The system consists of a pair of cylindrical sleeves cast in epoxy around a group
of strands. Two sets of Demec points are mounted in each pair of epoxy sleeves. The system was
originally developed by Arrellaga [11], and used by Roberts [10] at the San Antonio Y. Because
of problems with grouting the ducts after removal of an entire section of duct at the San Antonio
Y, only half the duct was removed on U.S. 183, as seen in Figure 2.31. Although tests in the
laboratory, such as the bench test shown in Figure 2.32, successfully used the epoxy sleeve system
to measure strains, field installation allowed only a partial depth sleeve, which was very difficult
to seal. Each epoxy sleeve took about 2 hours to construct. The time would have been better
spent installing several more electronic strain gauges at each location. Data from the epoxy
sleeves used on the mainlane were quite erratic.

78
Figure 2.31 Epoxy sleeve strain measurement system

79
 Data acquisition
system
Tendon

Strain gauges
Epoxy sleeves

Computer

View A-A

A A
Anchorage zone with
plate and trumpet

Ram

Hydraulic pump

Figure 2.32 Epoxy sleeves tested during the bench test

2.3 INSTRUMENT LOCATIONS

2.3.1 Mainlane pier

The downstation right side quadrant of mainlane pier D6 was instrumented with C-
gauges, S-gauges, and thermocouples. Figure 2.33 shows that C-gauges were placed at six
sections along the height of the pier. These gauges were intended to follow the flow of
compressive forces from the bearing pedestal to the footing. Demec points were also located on

80
the outside of the pier at the same elevations as the C-gauges. When the pier was redesigned as a
cast-in-place pier, heavy reinforcement was added to the capital, particularly on the inside face of
the Y. Strain gauges S101 through S106 were installed on the hooked bars detailed in the left half
of the capital in Figure 2.33. These S-gauges and the C-gauges were located as shown in the
sections in Figure 2.34. Several thermocouples were also installed to measure the thermal gradient
between the surface concrete and the core of the pier. The structural steel pipes had an array of
strain gauges installed on their surface, both inside and outside of the concrete. The S-gauges on
the pipes were intended to measure the force in the pipes, and measure the change in force in the
pipes near the anchor plates at the end of each pipe. The location of the pipe gauges is shown in
Figure 2.35. Thermocouples were also installed in the pipes so that thermal induced stresses could
be evaluated. Even though the loading on the pier was always symmetric from the superstructure,
both sides of the pier should have been gauged for redundancy.

81
 See Figure 2.35 for
instrument designation

(T103) (T102)

654 mm 654 mm 654 mm

(T101)
S105 C133 C129 C130
(S106) (C131) (C132) (C128)
A A
S103
(T105, T104)
(S104) C122 C126 C124
B (C125)(C123) (C127) B
S101 C116 C117 C118
(S102) (C119) (C120) (C121)
C C
(T107, T106)

152 mm
(T108)
(T109) 3 Equal Spaces
(Each Level)

D C110 C111 C112 C115 D

4420 mm
E C104C105 C106 C109
E
2286 mm
152mm

F C101 C102
C103
F

LEGEND
C-gauge ( ) Indicates Instrument
Strain Gage on Transverse CL Pier
Thermocouple

Note: Sections A-A through F-F shown in Figure 2.34

Figure 2.33 Elevation of Y-pier with gauge locations

82
 1219mm 457mm
2438 mm
1219 mm
203 mm 864 mm 152 mm
2134 mm

1219mm
S106 C131 C132 C128
T105 T104

1626 mm
1219 mm
C133 C129 Transverse CL Pier C109
S105 C130 C108

813 mm
C107

457mm
SECTION A-A C104

Longitudinal C C105 C106

L Pier

203 mm
1008mm
SECTION D-D
340mm
C125

C127
C126 C123

1219mm
1898mm

S104 2438mm
C124

1219 mm
C122

203mm 864mm 152mm

S103
340mm

SECTION B-B

1626mm
1219 mm
Transverse CL Pier C115
C114

813mm
C113
852 mm 242 mm C110

Longitudinal C C111 C112

L Pier
203mm
C121
C119
C120

SECTION E-E
1219 mm
1703 mm

S102
T107 T106
S101
2642 mm
C117
C118
C116

237 mm

1321 mm
264 mm 2115 mm
SECTION C-C
1829 mm

LEGEND
1302 mm

Transverse C
L Pier C103
914 mm

C-Gauge
Strain Gauge C102

Thermocouple C101
Longitudinal CL Pier

SECTION F-F
Figure 2.34 Y-pier sections showing gauge locations

83
 762 mm 1829 mm

Longitudinal CL Pier

1219 mm
S125
S126

2438 mm
A S127 S123 S121 S119
S128 S124 S122 S120
S129
S130

S117
S118
A S113 S111 S109 S107
S114 S112 S110 S108
S115
S116
I) II) III) IV) V)
PLAN VIEW OF CAPITAL

S130
S127 S128 S122 S120
S124
S129 S126 S125 S123 S121 S119
S118 S115 S116 S110 S108
S112
S117 S114 S113 S111 S109 S107
I) II) III) IV) V)
SECTION A-A
Figure 2.35 Plan view of Y-pier showing gauge locations

2.3.2 Three-span mainlane box girder unit

The mainlane girder for span D5, shown in Figure 2.36, had two segments heavily
instrumented, D5-9 and D5-16. Deviator segment D5-12 also had considerable instrumentation.
The taut wire measurement system was used in span D5, as well in spans D4 and D6, during post-
tensioning of tendons and during live load tests. Tendon gauges and epoxy sleeve mounted
Demec points are shown in Section A-A in Figure 2.37. Gauges were located along the tendon to
measure friction losses through the deviators during post-tensioning, and to measure changes in
tendon force from live loads.

84
 3 and 7 -15mm dia strand tendons Up Station
19 - 15mm dia. strand tendons

B C D

A A

D5-9 D5-12
Deflection D5-16
measurement B C
D
locations
Figure 2.36 Elevation of mainlane girder D5 with deflection measurement locations

Tendon instrument
Gauges Gauges location (typ.)
Gauges
S56,S57,D11,D12 S68,S69,D23,D24 S80,S81
S66,S67,D21,D22 Deviator (typ.)
S54,S55,D9,D10 S78,S79
S52,S53,D7,D8 S64,S65,D19,D20 Tendons (typ.) S76,S77

DAS DAS

Gauges Gauges Gauges

S50,S51,D5,D6 S62,S63,D17,D18 S74,S75
S48,S49,D3,D4 S60,S61,D15,D16 S72,S73
S46,S47,D1,D2 S58,S59,D13,D14 S70,S71

SECTION A-A
Figure 2.37 Tendon gauge locations for mainlane girder D5

Section B-B in Figure 2.38 shows the location of the longitudinally oriented C-gauges
and Demec points in segment D5-9. The right half of the section was more heavily instrumented
than the left half. Gauges were spaced closely above the right web to measure changes in strain
transversely due to shear lag effects from bending moments. C-gauges were also located
transversely around the cross section to measure strains from transverse bending, tension and
compression in the webs and flanges. Section B-B in Figure 2.39 shows the location of
thermocouples in segment D5-9. This segment was chosen for the thermal gradient study because

85
it’s heat flow would not be influenced by large volumes of concrete such as the deviator beams
and diaphragms, and could use the same data acquisition system as the midspan C-gauges.
Thermocouples were placed densely on the right side of the section, including across web and
flange widths, and vertically up the left web for redundancy and directional effects.

17,680mm

2642mm 2642mm 3035mm 2121mm 3@610mm 2184mm 2184mm

76mm
C29 C28 C27 C15 C16 C17 C18 C19 C20 C21

D42 D41 D40 584mm

2134mm

C22
DAS

9
C- Gauge D43

D3
1168mm
D44 D37 D38
Demec C23
C26 C25 C24 229mm
1981mm 1981mm

76mm
502mm

SECTION B-B
Figure 2.38 Strain gauge locations for segment D5-9

17,680mm

5798mm 3042mm 2121mm 2427mm 3682mm 114mm

25mm
25mm

T29 T31 T34 T5 T2

T28 T30 T33 T6 T3
T27 102mm T32 T35 T7 T4
T26
25mm

T8 T1
133mm

T25 T10,T9,T11
2@308mm
25mm
206mm

T24 DAS
mm

2@308mm

T23 T13,T12,T14
978

3@326mm T22
T18
T20 T16
T19 T15
T21
1911mm 1911mm
T17
Thermocouple
SECTION B-B
Figure 2.39 Thermocouple locations for segment D5-9

86
 The longitudinally oriented C-gauges in D5-16, the upstation anchor segment, are shown
in Figure 2.40. The anchor segment has a short length of typical section cast onto it downstation
from the diaphragm for match casting purposes. This is where the C-gauges were located. The
Demec points shown in Figure 2.40 were actually located in segment D5-15. The distribution of
these gauges was similar to that in segment D5-9, but was symmetric for redundancy. The density
of the longitudinally oriented C-gauges should have been more dense, especially in the top flange,
as was done later for the Ramp P girder.

17,680mm
3@610mm 3@610mm
2184mm 2184mm 2121mm 2121mm 2184mm 2184mm

76mm
C49 C48 C47C46 C45 C44 C30 C31 C32 C33 C34 C35 C36

D54 D53 D52 584mm

2134mm

C43 C37
DAS
1
C- Gauge D55
D49 D5 1168mm
D56 D50
Demec C42 C38
C41 C40 C39 229mm
1981mm 1981mm

76mm
502mm

SECTION D-D
Figure 2.40 Longitudinal strain gauge locations for segment D5-16

The deviator segment shown in Figure 2.41 was instrumented with strain gauges near the
deviation pipes on the right hand side of the section only. Gauges were installed directly on the
rebars at locations of anticipated cracks. Section G-G in Figure 2.41 shows that the gauges were
installed on three transverse planes. Since the deviator was designed as an inverted T-beam, S-
gauges should also have been installed on the heavy top bar near the centerline of the cross-
section. This was later done on the deviator in Ramp P. However, cracks only occurred in the
anticipated locations.

87
 G

102mm to

356mm
140mm
S1 S3
S2 S4 S5 S6

279mm
Strain gauge G

SECTION C-C

914mm

Gauges S1-S6 Gauges S13-S18

Gauges S7-S12
SECTION G-G
Figure 2.41 Strain gauge locations for deviator segment D5-12

Strain gauges were also located directly on the heavy bars in the anchor segment D5-16,
as shown in Figure 2.42. The diaphragm was designed using the strut-and-tie method. The bars
placed to act as ties from the design were the ones that received the most instrumentation. Gauges
were located on these bars at locations of anticipated cracks, such as near chamfered corners.

88
 Gauges S19-S45 External tendons

#4 (typ.)

#8 (typ.)

#5 (typ.)
#11 Future tendons

Internal tendons
Figure 2.42 Strain gauge locations for anchor segment D5-16

Linear potentiometers were placed around the bearing plinths at pier D6 to measure
girder end rotations during live load tests, and measure any horizontal movement of the
superstructure. The location of these instruments is shown in Figure 2.43. The movements were
so small that the measurements were of little value during the live load tests, but did provide good
data for daily thermal fluctuations. This data were used to time the casting of the cast-in-place
deck joints to reduce tension in the joint during the initial curing period.

89
 Legend
Vertically oriented linear pot
Horizontally oriented linear pot

Up Station
Anchor segment
L10 bearing plinth (typ.)
L102 L9

L8
L7 L4

L3

L101
L6 L5 L2 L1

Plan View - Pier D6

Figure 2.43 Linear potentiometer locations at pier D6

The cast-in-place deck joint between spans D5 and D6 was instrumented with an array of
longitudinally oriented C-gauges, similar in transverse distribution to those in the top flange of
segment D5-16. The deck joint was expected to go into tension and compression from thermal
forces, and possibly from live load forces. Gauges were located top and bottom of the deck joint,
as shown in Figure 2.44, to measure bending strains as well.

17,680mm
3@610mm 3@610mm
2184mm 2184mm 2121mm 2121mm 2184mm 2184mm
229mm
(typ.)

C74 C72 C70C68 C66 C64 C62 C60 C58 C56 C54 C52 C50
C75 C73 C71C69 C67 C65 C63 C61 C59 C57 C55 C53 C51
3@51mm (typ.)
51mm clr. (typ.)

C- Gauge
DAS

Cast-in-place joint

Figure 2.44 Strain gauge locations for the cast-in-place joint over pier D6
90
2.3.3 Segmental large ramp pier
The segmental large ramp pier P16 was instrumented at three sections in the column, and
throughout the capital. Figure 2.45 shows an elevation of the pier and the location of the
instrumented segments. Figure 2.46 shows the instrumentation in segment PC16-1. The gauges
were located just above the cast-in-place base. The segment is heavily instrumented with C-
gauges, and lightly instrumented with thermocouples to compare its temperature with that of the
other instrumented segments. Figure 2.47 shows the instrument locations in segment PC16-5.
This segment was selected for the thermal gradient study. It had optimal exposure to the sun, and
was sufficiently far from the capital and the shaded sections of the bottom of the pier to prevent
vertical heat flow. Figure 2.48 shows the gauge locations in segment PC16-7. The gauge
distribution is identical to that in segment PC16-1, with the plane of gauges located 610mm
beneath the solid capital segment.

91
 2515mm
PC16-8
PC16-7
C C
PC16-6
PC16-5

B B
PC16-4
PC16-3
PC16-2

Tendons
(19 ~ 15-mm dia. strands)
2438mm
PC16-1

(Typ.)

A A

Figure 2.45 Elevation of large ramp pier P16 with segment identification
92
 Post-tensioning
Up Station tendon ducts (4) Up Station
C400 T400
Instrumentation blockout
C418 C401 C403 and duct T407 T401
C402
C419 C404

C406
C416

C415 C417 C407 C405 T406 T402

C414 C409
C412
C413 C411 C408 T405 T403

C410 T404

Section A-A Section A-A

Legend
Concrete Strain Gauge Oriented Horizontally
Concrete Strain Gauge Oriented Vertically
Thermocouple
Figure 2.46 Segment PC16-1 sections with gauge locations

Post-tensioning Up Station
Up Station tendon ducts (4) T410
T411
T431 T412 T413
C430 Instrumentation Blockout T414
T432
C431 and Duct T433 T415
C432
T417
C434
C440

C439 C441 C435 C433 T428 T430 T418 T416

T429

C438
C437 T427 T421
T426 T420
C436 T425 T424 T419
T423
Section B-B T422
Legend Section B-B
Concrete Strain Gauge Oriented Horizontally
Concrete Strain Gauge Oriented Vertically
Thermocouple
Figure 2.47 Segment PC16-5 sections with gauge locations

93
 Post-tensioning
Up Station tendon ducts (4) Up Station
T440
C450 Instrumentation Blockout
C468 C451 C453 and Duct T447 T441
C452
C469 C454

C456
C466

C465 C467 C457 C455 T446 T442

C464 C459
C462
C463 C461 C458 T445 T443

C460 T444

Section C-C
Legend Section C-C
Concrete Strain Gauge Oriented Horizontally
Concrete Strain Gauge Oriented Vertically
Thermocouple
Figure 2.48 Segment PC16-7 sections with gauge locations

The capital segment PC16-8 was heavily instrumented with C-gauges, S-gauges, and
thermocouples. Figure 2.49 shows the location of the C-gauges intended to measure compressive
forces from the girder bearings and main post-tensioning anchorages, and the anchor plates of the
tie down bars. The C-gauges in Figure 2.50 are similar in distribution to the gauges in segment
PC16-7, and were intended to measure compressive strains from the superstructure dead load axial
forces and bending moments.

94
 Up Sta
B
C539 C533
C540 C534

C541
C544
C543

C542
A A
C536
C538
C537 C535

DYWIDAG Tie-Down Bars B

for superstructure moment
connection
Centerline
Compressive Struts Tendon
C542
(C543) C541
anchorage
(C544)
914mm

C533 C543 C542

(C539) (C544) (C541)
C535
(C537)

C536 C534
(C538) Pier Tendons
(C540)

Section B-B Section A-A

(Gauges C533, C535, C537, C539 located at 75mm clear cover.)
(Gauges C534, C536, C538, C540 located at 455mm clear cover.)
Legend
C-gauge oriented along compressive struts
( ) denotes gauges on rear strut
Concrete Strain Gauge
(Top View)
Figure 2.49 Segment PC16-8 top C-gauge locations

95
 Post-Tensioned
Tendons
Up Sta Up Sta
C501 C517
C509 C525
C508
C502 C524 C518
75 mm 830mm 1600mm

C516 C532
C510 C526
2510mm

C523 C519
C507 C515 C511 C503 C531 C527

B B
C514 C512 C528
C506 C504 C530 C520
A A C513 C522 C529
C505 C521
2390mm 2510mm

Section A-A Section B-B

Figure 2.50 Segment PC16-8 bottom C-gauge locations

Since the capital is an excellent example of a strut-and-tie structure, the heavy reinforcing
bars located in the top and bottom of the pier intended to act as ties were instrumented. Figure
2.51 shows the S-gauge locations on these tie bars. The ramp girder anchor segments were post-
tensioned to the capital with the 16 threadbars shown in Figure 2.52 to form a moment connection.
These threadbars were cast into the capital, then coupled to bars extending up to the deck of the
anchor segments. Force in these bars was monitored at three elevations by the strain gauges
shown in Figure 2.52.

96
 B B

A A

406mm
2970mm

m
8m
50
2010mm
#11

Up Sta P.T. bar tie-downs Up Sta

S501 #10
S544 S541
S508 S502

S549 S545
S509
S510

S546
S552

S550
S548
S514 S515
S507 S513 S516 S503
S551
S511 S547
S512
S504 Post-Tensioning
S506 S543 S542
Tendons
S505
Section A-A #5 Section B-B
685mm

Steel Strain Gauge

950mm

Figure 2.51 Segment PC16-8 S-gauge locations on rebar

97
 B

Up STA
S538 S517
(S539) (S518)
((S540)) ((S519))

S535 S522
S537 S536 S521 S520

S534 S529 S528 S523

A S530S527 A
S531 S524
(S532) (S525)
((S533)) ((S526))
S523 S520
S530 S527 B (S528) (S521)
S531 (S529) (S528) S524 ((S529)) ((S536))
((S536)) ((S521)) (S523) S524 (((S534))) S517
(S534) (((S537)))
(S527) (S522)
((S537)) (((S535))) (((S522))) ((S520)) ((S528)) ((S535))
(((S538))) (((S517))) (((S531))) (((S538)))

S532 S525 S518

S525
(S539) (S518) (S539)
(S532)
S533 S526 S519
(S540) (S519) S526
(S540)
(S533)

Section A-A Section B-B

Notes:
Gauges denoted by ( ) located “behind” gauges shown.
Gauges denoted by (( )) located “behind” gauges denoted by ( ).
Guages denoted by ((( ))) located “behind” gauges denoted by (( )).
Figure 2.52 Segment PC16-8 S-gauge locations on tie-down bars

2.3.4 Five-span ramp box girder unit

The half span of Ramp P upstation from pier P16 was heavily instrumented. Three full
cross-sections were instrumented for axial, bending and torsional strains. Three D-zones were
also instrumented for strains, and one section was thoroughly instrumented for concrete
temperatures. Tendons on either side of one vertical deviator were instrumented with strain
gauges to measure friction losses during post-tensioning, and strain changes during live load tests.
The instrumented segments are shown in the elevation in Figure 2.53, along with the location of
98
tiltmeter plates on the deck used during balanced cantilevering. The location of tendon gauges is
shown in Figure 2.54.

1.9 m 11.5 m 11.5 m 2.0 m

460 mm 460 mm 460 mm 460 mm
P16-1A
A B C D

P16-2 P16-4 P16-6 P16-8 P16-10 P16-12 P16-14 P16-16 P16-17

A B C D

Upstation
Pier P16 Sections A, B, & C - Instrumented for Strains
Section D - Instrumented for Temperature

Tiltmeter plate location (symmetric about pier P16)

Figure 2.53 Elevation of Ramp P, span P16, with segment identification

Instrumented locations
(3 gauges on each tendon)

19 - 15mm dia. strand external tendon (typ.)

Plan Layout
Upstation
Elevation Layout

Center
P16-1A Interior Deviator Horizontal Line of
Anchorage Vertical Deviator Span
Segment Beam Beam

Figure 2.54 Ramp P tendon gauge locations

Anchor segment P16-1A was designed by the strut-and-tie method, but was post-
tensioned vertically. This greatly reduced the cracking in this segment when compared to the
mainlane anchor segments. Strain gauges were located on the diaphragm face shown in Figure
99
2.55 at locations were minor cracking did occur. These few S-gauges were wired into the data
acquisition system in segment P16-2. P16-2 was instrumented with a symmetrical array of C-
gauges and Demec points, as shown in Figure 2.56. Eight strain gauge rosettes were in the
segment to measure shear strains and torsional effects.

3960mm

Cantilever tendon ducts

#9 #7

S629 S630 S631

Transverse S628
S634
prestressing
2130mm

S627
S633
External post-tensioning #6
tendon ducts S626
S632

S625

Figure 2.55 Anchor segment P16-1A section with gauge locations

100
 Dimensions in mm
410 300 300 410
100 840 860 360 510 580 580 510 360 860 840 100
C608 (L)
C609 (D)
C601 C602 C603 C604 C605 C606 C607C610 (T) C614 C615 C616 C617 C618 C619 C620

C611 (L)
C648 C647 C612 (D) C621 C622
C646 C613 (T) C627 (L) C623
C643 (L)
C628 (D)
C640 (L) C644 (D) C635 (L) C624 (L)
C629 (T)
C641 (D) C645 (T) C636 (D) C625 (D)
C641 (T) C626 (T)
C637 (T)
C639 C630

C638 C632 C631

Section A-A

150 8 @ 952.5mm 150

D625 D626 D627 D628 D629 D630 D631 D632 D633

D617 D612
D618 D613 D611
D615
20 622 D614
D6 D D60 D60
9 8
19
D6 D623
D616D606 D60
Upsta. D605 7
21 D624
D6 D61
0
.
sta Ups
Up ta .
Upsta.
D603
Section A-A
D602
D604
D601

Figure 2.56 Segment P16-2 sections with gauge locations

Segments P16-10 and P16-17 were instrumented with an array of C-gauges and Demec
points identically to segment P16-2, as shown in Figures 2.57 and 2.58. Segment P16-17 is the
midspan segment and P16-10 is at the quarter point. The gauging density in these segments was
sufficiently increased over that of the mainlane, but more cross-sections needed to be instrumented
since the gauges in P16-2 were quite close to the heavy end diaphragm, and the gauges in P16-10
were too close to the vertical deviator.

101
 Dimensions in mm
410 300 380 340
80 830 800 410 610 560 510 560 420 840 760 100
C708 (L)
C709 (D)
C701 C702 C703 C704 C705 C706 C707 C710 (T) C714 C715 C716 C717 C718 C719 C720

C711 (L) C721 C722

C748 C747 C712 (D)
C746 C743 (L) C713 (T) C727 (L) C723
C740 (L) C744 (D) C728 (D) C724 (L)
C741 (D) C745 (T) C735 (L) C729 (T) C725 (D)
C736 (D)
C742 (T) C737 (T) C726 (T)

C739 C730

C738 C732 (L) C731

C733 (D)
C734 (T)
Section B-B
D713
D715
D714
D716
Upsta.

150mm 8 @ 952.5mm 150mm

D725 D726 D727 D728 D729 D730 D731 D732 D733

D717 D712
D718 D711
D70 D70
20 722 9 8
D7 D D70
19 7
D7 D723 D706
21 D71
D7 D724 D705 0
. Ups
s ta ta.
Up
Upsta.
D703
D702
D704 Section B-B
D701

Figure 2.57 Deviator segment P16-10 sections with gauge locations

102
 Dimensions in mm
360 330 280 430
100 810 860 360 580 560 560 560 410 910 740 80

C808 (L)
C809 (D)
C801 C802 C803 C804 C805 C806 C807C810 (T) C814 C815 C816 C817 C818 C819 C820

C811 (L)
C848 C847 C821 C822
C812 (D)
C846 C843 (L) C813 (T) C827 (L) C823
C840 (L) C844 (D) C828 (D) C824 (L)
C841 (D) C845 (T) C835 (L) C829 (T)
C825 (D)
C842 (T) C836 (D)
C837 (T) C826 (T)

C839 C830

C838 C835 (L) C831

C836 (D) D813
C837 (T) D815
Section C-C D814
D816
Upsta.
150mm 8 @ 952.5mm 150mm

D825 D826 D827 D828 D829 D830 D831 D832 D833

D817 D812
D818 D811
D80 D80
9 8
20 2 D80
D8 D82 9 7
1 D823 D806
D8 D81
0
1 D824 D805
D82 Ups
ta .
ta.
Ups
Upsta.
D803
D802 Section C-C
D804
D801

Figure 2.58 Segment P16-17 sections with strain gauge locations

Segment P16-17 was selected for the thermal gradient study. The array of thermocouples,
shown in Figure 2.59, was more dense than the mainlane and symmetrical. This instrumentation
was as thorough as has been done in any study to date, and was possible because 128 channels were
available for the gauges in segment P16-17.

103
 T802 T805 T810 T813 T816 T821 T824
T803 T806 T811 T814 T817 T822 T825
T801 T804 T807 T808 T812 T815 T818 T819 T823 T826 T827

T809 T820

T852 T828
T853 T829
T854 T830
T849 T831
T850 T832
T851 T833
T846 T834
T847 T835
T848 T843 T840 T837 T836
T844 T841 T838
T845 T842 T839
Section D-D
Figure 2.59 Segment P16-17 section with thermocouple locations

The deviator beam in segment P16-10 was similar in instrumentation to the deviator on
the mainlane, with the addition of gauges mounted on the heavy top bars, and C-gauges located in
the web, as shown in Figure 2.60. Unfortunately, gauges S719, S720 and S721 were all destroyed
during segment casting. Concrete for the deviator was poured through a hole in the center of the
top flange form, and had to be knocked off the deviator bars between lifts so that it would not dry
on the bars.

104
 Looking Up Station

S713 S716
(S714) (S717) 203mm
((S715)) ((S718)) 2@229mm

3 - #9 12 - #5 A

S701
(S702)
((S703))
S704
(S705)
Deviator pipes for
S719
((S706)) S707 S710 A (S720) external tendons
(S708) (S711)
((S721))
((S709)) ((S712))

C751 C752 C753

C750 C754
Web guages

S716 S717 S718

S719 (S713) S720 (S714) S721 (S715)
C749 C755
914mm x 356mm
Deviator beam
#9
Tendon duct #5

S724 S725
(S722) (S723)

S710 S711 S712

(S707) (S708) (S709)
((S704)) ((S705)) ((S706))
(((S701))) (((S702))) (((S703)))

Section A-A
Figure 2.60 Deviator segment P16 section with S-gauge locations on rebar
105
 The bottom slab post-tensioning tendon anchorage blister in segment P16-4 was the third
D-zone in Ramp P to be instrumented. S-gauges were located on bars that were expected to go
into significant tension. C-gauges were located in the bottom slab and webs of the segment behind
the anchorage blister to measure the tensile strain in this region. The location of these gauges is
shown in Figure 2.61.

S618
#5 (TYP.)
C655 #4 (TYP.)
S619
C654 S609
S610
S611
S620
S621
S612
S622 S613 S614
C653
C652

C650
C649
C651 #5

S624
S615
S616 S617

9 - 15mm dia. strand tendon

356mm high x 2046mm long
anchorage blister

Figure 2.61 Segment P16-4 anchorage blister strain gauge locations

106
2.3.5 Multiple-celled box girder units
The cast-in-place transition spans shown in Figure 2.62 were instrumented for deflections
only during a live load test using the taut wire deflection measuring system. Wires were hung
down the center of each of the three cells in each span. Deflections were measured at the quarter
points of the span, as well as at the girder ends.
Pier C36

Pier C37

Pier C38
Span C36 Span C37
40.84m 40.84m
A

Deflection measurement points Piano wire

A
PLAN

Varies 1994mm Varies 1994mm 5930mm

Piano wire Gauge mounting

plate

SECTION A-A
Figure 2.62 Plan and section of transition spans showing instrumentation locations

Three modified spans with cast-in-place gore closure strip were also instrumented for
deflections only, as shown in Figure 2.63. The wires were hung down the center of each single
celled box girder.

107
 Pier L7
Pier L6
Pier L5
Pier L4

Span L6
Span L4 Span L5 39.94m
34.09m 40.20m
A

Cast-in-place gore Deflection

Piano wire
closure strip measurement point

A
Span C41 Span C42 Span C43
34.14m 40.20m 40.07m

Pier C43

Pier C44
Pier C41

Pier C42

2050mm 3820mm Varies 5474mm 5798mm

556mm

Piano wire Gauge mounting plate

Section A-A

Figure 2.63 Plan and section of modified spans showing instrumentation locations

108
2.4 MATERIAL PROPERTY TESTS
Three 152mm by 152mm by 533mm concrete prism samples were made for each
segment instrumented on the project. These specimens were tested for their moduli of elasticity
and coefficients of thermal expansion using the Demec extensometer and a test machine. A
concrete prism test specimen is shown in Figure 2.64

152mm 152mm

67mm
533mm

400mm

Demec Points Demec Points

67mm

Figure 2.64 Concrete modulus of elasticity test specimen

The results of the modulus of elasticity tests are shown in Figure 2.65 through Figure
2.68. The concrete for the mainlane pier did not come from the batch plant at the precasting yard.
The mainlane pier concrete (Figure 2.65) was not as stiff as that used for the precast segments
(Figure 2.67). Interestingly enough, the modulus of elasticity of the concrete decreases as the
specimens age, and dry out. This was seen in every specimen.

109
 50000
45000
40000
Modulus of Elasticity in MPa

35000
30000 Pier D6 Capital

25000
20000 Pier D6 Column
15000
10000
5000
0
Aug-94 Nov-94 Mar-95 Jul-95 Nov-95 Mar-96 Jul-96 Nov-96
Date
Figure 2.65 Modulus of elasticity test results for Mainlane pier D6

50000
45000
Modulus of Elasticity in MPa

40000
35000
30000
D5-9
25000 D5-16
D5-5
20000 D5-12
D6-9
15000
10000
5000
0
Jun-94 Sep-94 Jan-95 Apr-95 Jul-95 Oct-95 Feb-96 May-96 Aug-96

Date
Figure 2.66 Modulus of elasticity test results for Mainlane superstructure Unit D2

110
 50000
45000
Modulus of Elasticity in MPa

40000
35000
30000
PC16-8
25000
PC16-7
20000 PC16-5
PC16-1
15000 PC16-BASE
10000
5000
0
Jul-95 Oct-95 Feb-96 May-96 Aug-96 Dec-96 Mar-97 Jun-97
Date

Figure 2.67 Modulus of elasticity test results for Large Ramp pier P16

50000
45000
Modulus of Elasticity in MPa

40000
35000
30000
25000 P16-4
P16-2
20000 P16-17
P16-10
15000 P16-1A

10000
5000
0
Jul-96 Aug-96 Oct-96 Dec-96 Jan-97 Mar-97 Apr-97 Jun-97

Date

Figure 2.68 Modulus of elasticity test results for Ramp P superstructure

111
 Coefficient of thermal expansion tests were also performed on several of the test
specimens. The results are in Table 2.1. The concrete in the mainlane pier once again proved to
be substantially different in this material property compared to the concrete in segments from the
precasting yard.

Table 2.1 Coefficients of thermal expansion

Coefficient of Thermal
Segment Expansion (oC-1)

Pier D6 Column 8.28x10-6

Pier D6 Capital 8.35x10-6
D5-5 9.59x10-6
D5-9 9.79x10-6

D5-12 9.99x10-6
D5-16 10.40x10-6
D6-9 10.40x10-6

PC16-1 9.90x10-6
PC16-5 9.90x10-6
PC16-7 9.00x10-6

PC16-8 10.01x10-6
P16-2 9.00x10-6
P16-10 9.72x10-6
P16-17 9.36x10-6

Concrete compressive strengths (f c' ) measured at the precasting yard for mainlane D are
shown in Figure 2.69. Very few of the compressive test cylinders were tested beyond the 7 day
break because the 28 day design strength of 37.9Mpa was usually achieved by this point in time.
Concrete strengths of over 62Mpa were the norm at 28 days. Strengths probably continued to
increase beyond the 28-day strength since 25% of the cementitious material was fly ash.

112
 80

70
Cylinder Strength (Mpa)

60

50

40

30
28 Day Design Strength

20

10

0
0 100 200 300 400 500 600 700
Time (hours)

Figure 2.69 Concrete compressive strengths for mainlane D

The results of modulus of rupture tests are given in Table 2.2 for selected concrete
samples. The specimens used in these tests were the same as used in the modulus of elasticity
tests and the coefficient of thermal expansion tests. Table 2.2 presents the average measurements
from three tests specimens for various segments, except for segments P16-2 and PC16-5 that used
two specimens. The test results were not very consistent among specimens from the same
segment, as indicated by the large standard deviations in the far right column of Table 2.2, or
among various segments. The average rupture tensile strain for the concrete specimens from the
precasting yard was 172με, while the average for the PC-16 cast-in-place base concrete specimens
was higher averaging 200με.

113
 Table 2.2 Modulus of rupture test results

Concrete Tensile Modulus of Tensile Strain Standard

Age Stress Elasticity at Rupture Deviation (με)
Segment
(days) (Mpa) (Mpa) (με)
P16-1a 948 5.49 37000 148 9.9
P16-2 1006 7.48 41000 183 12.4
P16-4 1001 7.48 42000 178 13.3
P16-10 995 6.24 37000 169 5.6
P16-17 989 6.40 37500 171 13.8
PC16-Base 1100 7.10 35500 200 17.0
PC16-1 1372 5.87 36500 161 26.8
PC16-5 1363 6.36 37000 172 1.6
PC16-7 1329 6.99 37500 186 24.1
PC16-8 1255 7.70 42500 181 29.2
Average 6.71 38300 172
(except base)

114
 CHAPTER 3
PRESTRESSING FORCE LOSSES IN POST-TENSIONING TENDONS

3.1 PROBLEM DESCRIPTION AND TECHNICAL BACKGROUND

Prestressing force losses in post-tensioning tendons may be divided into two categories,
losses occurring at the time of stressing, and losses occurring over time thereafter. When using
low-relaxation strands, the largest prestressing losses in a post-tensioned structure usually come
from friction losses during stressing. Friction loss of the post-tensioning force in a tendon occurs
when the tendon comes in contact with its duct or other guiding device during stressing. The
normal force between the tendon and duct can be small, such as for a straight tendon with its duct
internal to the concrete. The normal force can also be large, especially for external tendons
deviated at discrete locations in a bridge girder. The coefficient of friction between tendon and
duct can vary substantially depending on the normal force generated between the two materials.
The three basic elements that are involved in the friction of unlubricated solids are [14]:

1. the true area of contact between mating rough surfaces;

2. the type and strength of bond formed at the interface where contact occurs; and

3. the way in which material in and around the contacting regions is sheared and
ruptured during sliding.

Because these variables are difficult to quantify, design coefficients of friction for various
duct materials and normal forces have been determined empirically, and are usually presented as a
range of values. The engineer must select an appropriate coefficient of friction when designing a
tendon. The accuracy of his selection may not be known until the bridge is under construction.

Friction loss during post-tensioning of a tendon may substantially reduce the prestressing
force in the tendon. This occurs in tendons with large cumulative angular deviations over the
length of the tendon. Multispan deviated tendons may lose 40% to 50% of the prestressing force
at the jacking end of the tendon. Accurate calculation of the friction losses in this case is essential
for an economical, safe, and constructable design.

Elongations measured during post-tensioning give an indication of the cumulative friction

loss in a tendon. The AASHTO Guide Specifications for the Design and Construction of
Segmental Concrete Bridges [8] require measured elongations to be within 5% of those calculated

115
by the engineer, unless otherwise approved by the owner. This specification is insensitive to the
level of prestress force friction loss in a tendon. The tolerance is difficult to satisfy for two
distinct reasons in tendons with high friction losses. First, the engineer may not be able to select a
friction coefficient with sufficient accuracy. Second, the constructor may not install ducts to the
degree of accuracy that the engineer assumed in his calculation. These factors will not have a
significant impact on the accuracy of elongation calculations for tendons with little friction loss.
Selection of the proper modulus of elasticity for the tendon and knowledge of the friction losses
near the live end anchorage may become more important in this case. Consistently high friction
losses on a project can be overcome by increasing overall prestressing force by adding additional
strands to the tendon, where possible, or by lubricating the tendon and duct with graphite to reduce
the friction.

The commonly used friction loss model for bridge design calculations is shown in Figure
3.1. The model assumes that the tendon to duct contact surface is planar when computing the
normal force. The total normal force for a multistrand tendon in a circular duct would be
somewhat larger.

Tendon Friction Force = μN

Normal Force = N
A

P
P-dP
dα

Tendon

Duct

Section A-A
Figure 3.1 Friction loss model

116
Equilibrium in the tendon for the differential length of duct in Figure 3.1 gives:

dP = -μN (3-1)
where
dP = differential change in tendon force
μ = coefficient of friction
N = normal force
N = Pdα
P = tendon force
dα = differential angle change
Integrating both sides gives:

Px/P0 = e-μα (3-2)

This equation gives the relationship between the known tendon force at point 0 and the
tendon force at some point x as a function of the cumulative angle change α between point 0 and
point x. To account for duct placement construction tolerances, an additional angle change is
added to α in one of three ways. First, the angle change can be an absolute value β, the wobble
angle, applied at deviators or other locations.

Px/P0 = e-μ(α+β) (3-3)

Second, the angle change can be a percentage k of the design angle change.

Px/P0 = e-μ(α+kα) (3-4)

Third, the angle change can be a function of the length of the duct L, as is used for internal
tendons, with an appropriate wobble coefficient K.

Px/P0 = e-μα-KL (3-5)

The accuracy of these equations depends on proper selection of the coefficients of

friction μ, and the wobble angle β or coefficients of wobble k or K.

Deformation of a structure also occurs when prestressing is applied. Tendons are usually
stressed in some sequence. Thus, tendons stressed later will have an effect on the force in tendons
stressed earlier in the sequence. This prestress force change is known as the elastic shortening
loss, and is mainly a function of the geometry of the structure and the modulus of elasticity of the
concrete, and obviously the magnitude and sequence of the applied post-tensioning forces.

117
 Long term prestress losses occur when the length of a tendon changes from creep and
shrinkage of the girder concrete, and because of prestressing steel relaxation. Long term prestress
losses are a function of material properties as well as the geometry of the structure. The use of
low relaxation strand has reduced the relaxation loss in a tendon to below 3% of the jacking force
[8]. Therefore, a reasonable estimate of the relaxation loss by the engineer using current code
equations is quite adequate. Accuracy of the relaxation loss calculation has little effect on the
accuracy of the total tendon force loss estimate.

On the other hand, creep and shrinkage induced losses in post-tensioning tendons may be
difficult to predict at the time of the structure’s design. The magnitude of both creep and
shrinkage strains causing post-tensioning losses is directly related to the amount of water
remaining in the pore volume of the paste in the concrete at the time of stressing [15]. The
concrete elements in a post-tensioned structure may have been cast well in advance of the time
prestressing or other loads were applied, such as is usual in a precast segmental bridge. Creep and
shrinkage losses may be low in this case because much of the water in the concrete not used for
hydration has escaped into the atmosphere, and the crystalline structure of the paste is dense,
restricting movement of the remaining water. On the other hand, a cast-in-place segmental bridge
is usually prestressed very early, as soon as the concrete comes to the required strength, so that
construction may continue. The accuracy of creep and shrinkage loss calculations may become
more important than the accuracy of friction loss calculations in the cast-in-place case. The
designer must estimate the casting and post-tensioning schedule for the project at the time of
design, and estimate the concrete creep and shrinkage behavior for concrete available at the
particular project site. Figure 3.2 shows the effect of concrete creep and shrinkage on prestress
force for concrete of various ages.

118
 Increasing loss in prestressing force
with decreasing age of concrete
at time of prestressing

Tension

Prestressing tendon
Stress
Compression

Decreasing age of concrete Strain

at time of prestressing
Concrete
girder

Figure 3.2 Prestress force loss from creep and shrinkage

Simple methods for roughly estimating the prestress loss from creep and shrinkage strains
are included in concrete design codes and other special publications. The accuracy of these
methods is sufficient for preliminary designs and many final designs. More accurate time
dependent analyses will be required to predict losses and control deflections on some structures,
such as bridges constructed in balanced cantilever.

The post-tensioned box girders under study on U.S. 183 had both internal and external
tendons, as shown in Figures 3.3 and 3.4. The internal tendons were placed essentially on straight
profiles in the bottom flange of the girders and in the wings. The external tendons were deviated
at two or more points on various profiles. The main objectives for the research on these tendons
were:

1. to measure friction losses in the anchorage zones,

2. to determine the magnitude and consistency of wobble friction along the tendon
profiles,

3. to measure tendon force changes other than friction losses including elastic
shortening, creep and shrinkage, and

4. to compare the measured results to values calculated using normal design

techniques.

119
 3 and 7 -15mm dia. strand wing tendons
19 - 15mm dia. strand bottom flange internal tendons
19 - 15mm dia. strand deviated external tendons

Span D5

Dead end Up Station Tendon instrument Jacking end

location (typ.)

Figure.3.3 Mainlane girder tendons

12 - 15mm dia. strand internal cantilevering tendons

19 - 15mm dia. strand external deviated tendons
Tendon instruments (typ.)

Up Station 9 - 15mm dia. strand bottom

P16 flange internal tendons P17

Figure 3.4 Ramp P girder tendons

3.2 LITERATURE REVIEW AND FIELD EXPERIENCE

3.2.1 Friction losses

Post-tensioning tendons can be classified into two basic groups, internal and external.
Internal tendons are in ducts cast within the concrete of the bridge girder over their entire length.
Ducts may be curved on a vertical profile in response to the shape of the bending moment diagram
for the member. Ducts may be placed essentially straight, such as in the top and bottom slabs of
box girders. With some exceptions, internal tendon profiles have high radii of curvature (>30m)
resulting in low normal forces between strand and duct, and the tendon and duct are in contact
over much of the length of the profile.

External tendons are located in ducts that are cast within the concrete only at discrete
points, such as at a deviator or saddle. Required curvature changes only occur at these points,

120
resulting in curvature profiles with low radii (8m), and tendon to duct contact over little of the
length of the profile. Coefficients of friction and wobble are much different than those for internal
tendons.

Internal tendons
Internal tendons were commonplace before the use of external tendons became common
practice. A duct commonly used in the 1960's in the United States for internal tendons was
flexible steel tubing. High wobble loss was experienced by the industry, and was experimentally
verified by Bezouska [16]. Wobble friction coefficients (K) were found to be 0.0049m-1 when
used with a coefficient of friction of 0.25. Soon afterward, the State of California and the rest of
the country began using rigid or semi-rigid steel duct in bridge construction. This duct could be
tied in place more accurately than the flexible duct, and remained in place when concrete was cast.
Wobble friction coefficients (K) were then found by Bezouska [17] to drop to 0.0007m-1 when
used with an assumed curvature friction coefficient of 0.25. After a review of their construction
records, California eventually decided to use a friction coefficient of 0.20 with no wobble term for
all internal tendons.

More modern field tests performed by Dywidag [18] gave a coefficient of friction of 0.24
in galvanized semi-rigid duct when the wobble coefficient was assumed to be 0.0007m-1. Tran
[19], testing a laboratory prepared girder with galvanized semi-rigid duct, found that the wobble
coefficient was closer to 0.0013m-1, and the coefficient of friction 0.16. Assuming K=0.0013m-1
in Dywidag's test girder, their coefficient of friction becomes 0.17, similar to Tran's coefficient of
friction.

Harstead, Kummerle, Archer, and Porat [20] performed full-scale tests in galvanized
semi-rigid steel duct with strand tendons. Using an assumed K of 0.0013m-1, their test results also
give a coefficient of friction of 0.17 in a duct with a minimum radius of 20m, and 0.24 in another
duct with a minimum radius of 6m. Bezouska [17] also saw the relationship between increased
curvature and increased friction loss. Tests by Yasuno, Kondo, Tadano, Mogami, and Sotomura
[21] in specimens with minimum radii of curvature similar to those of Tran [19] (8m–21m) gave
μ=0.14 and K=0.0011m-1 as friction coefficients.

Test results for internal segmentally constructed tendon profiles by Davis, Tran, Breen,
and Frank [22] corresponded well with previously measured results, shown in Figure 3.5, with a
coefficient of friction of 0.16 and wobble coefficient of 0.0016m-1. The coefficient of friction

121
between galvanized semi-rigid duct and strand tendons apparently changes from values typical
when normal forces correspond to curvature radii between about 15m to 40m radii in draped
internal ducts (with an upper bound of μ=0.25) to much higher values under high normal forces
which accompany radii less than 10m. Friction coefficients used in Equation 3-5 for external
deviator pipes, with radii usually less than 10m, are approximately μ=0.25 with K=0.0007m-1 or
greater for the wobble coefficient, as recommended by AASHTO [23].

Internal Tendon Friction Tests

0.30

0.25 1 - Bezouska 1966

2 - Bezouska 1971
4 3 - Tran
Friction Coefficient

4 - Harstead et al
0.20
5 - Yasuno et al
4 6 - Davis
1 3 2
0.15
6
5 5 1
5 1
0.10
1
1
0.05
Assumed wobble coefficient of 0.0013/m too
conservative for nearly straight tendon profiles
1
0.00
0 10 20 30 40 50 60 70 80 90
Approximate Minimum Radius of Curvature (m)
Figure 3.5 Results from previous friction tests

External tendons
Coefficients of friction used by design engineers for external tendons are largely based on
field experience. This is reflected in the AASHTO Guide Specification [23]. AASHTO suggests
using μ=0.25 and no wobble or additional angle change term when lubrication is used at the
deviator. A realistic coefficient of friction and wobble term for unlubricated external tendons was
suggested by Roberts, Breen, and Kreger [7]. They suggested using the usual coefficient of
friction of 0.25, but based on field observations they recommended supplementing it with an
inadvertent angle change term (β from Equation 3-3) of 0.04 radians or 2.29 degrees added to the
intended angle change at each deviator.

122
 Design coefficients for internal and external tendons recommended by various authorities and
shown in Tables 3.1 through 3.6 tend to be somewhat conservative, or give a range of values. In
Table 3.3, the CEB recommended coefficient of friction is modified, depending on the degree of
duct filling. Higher tendon area to duct area ratios generate higher total normal forces because the
tendon bears on a greater part of the perimeter of the duct. Therefore, they multiply the basic
coefficient of friction for flat surfaces by a squeezing factor.

Table 3.1 Friction and wobble coefficients for post-tensioned tendons from AASHTO [23]
Wobble
Materials Friction Coefficient μ Coefficient
K (m-1)
For strand in galvanized metal sheathing 0.15–0.25* 0.0007
For deformed high strength bars in galvanized 0.15 0.0007
metal sheathing
For strand in internal polyethylene duct 0.23 0.0007
For strand in straight polyethylene duct (external 0 0
to the concrete)
Rigid steel pipe deviators 0.25** 0.0007
* A friction coefficient of 0.25 is appropriate for 12 strand tendons. The coefficient is less for larger tendon
and duct sizes.
** Lubrication will probably be required.

Table 3.2 Friction and wobble coefficients for post-tensioned tendons from AASHTO
LRFD [9]
Type of Tendons and Wobble Coefficient, K Curvature Coefficient, μ
Sheathing (m-1) (1/RAD)
Tendons in rigid and semi-rigid
galvanized ducts 0.0007 0.05–0.25
- 7-wire strands
Pre-greased tendons
- wires and 7-wire strands 0.0010–0.0066 0.05–0.15
Mastic-coated tendons
- wires and 7-wire strands 0.0033–0.0066 0.05–0.15
0.25
Rigid steel pipe deviators 0.0007 Lubrication probably
required

123
 Table 3.3 Friction and wobble coefficients for post-tensioned tendons from CEB [24]
Physical
Tendon Type Coefficient of Coefficient of Wobble
Friction* Friction** Coefficient
μ0 μ K (m-1)
Cold drawn wire 0.13 0.17 0.005–0.010
Strand 0.15 0.19 0.005–0.010
Deformed bar 0.50 0.65 0.005–0.010
Smooth and round bar 0.25 0.33 0.005–0.010
Monostrands (single or grouped) 0.05–0.07 0.006–0.010
in slabs or reservoirs
Greased multistrand or wire
tendons (e.g. in nuclear 0.13–0.15 0.004–0.008
containments)
Dry multistrand or wire tendons
(e.g. in nuclear containments with Factors as for Factors as for
dry air as subsequent corrosion ordinary p.t. tendons ordinary p.t.
protection) tendons
External multistrand tendons:
naked dry strands over steel 0.25–0.30 0
saddle***
External multistrand tendons:
greased strands over steel 0.20–0.25 0
saddle***
External multistrand tendons: dry 0.12–0.15 0
strands inside plastic pipe over
saddle***
External multistrand tendons: 0.05–0.07 0
bundle of monostrands over
saddle***
* Values can be multiplied by 0.9 if slight lubrication is present, e.g. by soluble oil.
** The coefficient of friction μ is the product of the physical coefficient of friction μ0
and the squeezing factor χ, where χ is dependent on the degree of filling of the
duct. Where more exact investigations are not available, this factor can be
assumed to be 1.3 to 1.35 for tendons filling the duct between 50% and 60%. The
values in the table assume 50% filling.
***These values correspond to a saddle radius of 2.5m to 4.0m. For lower radii further
test evidence is needed.

124
Table 3.4 Friction and wobble coefficients for post-tensioned tendons from ACI 318-95 [25]
Curvature
Tendon Duct Tendon Wobble Coefficient Coefficient
Type K (m-1) μ
Grouted Metal sheathing Wire tendons 0.0033–0.0049 0.15–0.25
Grouted Metal sheathing High strength bars 0.0003–0.0020 0.08–0.30
Grouted Metal sheathing 7 - wire strand 0.0016–0.0066 0.15–0.25
Unbonded Mastic coated Wire tendons 0.0033–0.0066 0.05–0.15
Unbonded Mastic coated 7 - wire strand 0.0033–0.0066 0.15–0.15
Unbonded Pregreased Wire tendons 0.0010–0.0066 0.05–0.15
Unbonded Pregreased 7 - wire strand 0.0010–0.0066 0.05–0.15

Table 3.5 Friction and wobble coefficients for post-tensioned internal tendons from PTI [26]
Recommended Recommended
Type of Duct Range of Range of for for
Values Values Calculations Calculations
μ K (m-1) μ K (m-1)
Flexible tubing non- 0.18–0.26 16–66 x 10-4 0.22 25 x 10-4
grouted
Flexible tubing 0.14–0.22 10–23 x 10-4 0.18 16 x 10-4
galvanized
Rigid thin wall tubing 0.20–0.30 3–16 x 10-4 0.25 10 x 10-4
non-galvanized
Rigid thin wall tubing 0.16–0.24 0–13 x 10-4 0.20 7 x 10-4
galvanized
Greased and wrapped 0.05–0.15 16–49 x 10-4 0.07 33 x 10-4

125
 Table 3.6 Friction and wobble coefficients for post-tensioned internal
tendons from ACI-ASCE [27]
Usual Range Usual Suggested Suggested
Type of Steel Type of Duct or of Observed Range of Design Design
Sheath Values Observed Values Values
K (m-1) Values K (m-1) μ
μ
Wire cables Bright metal 0.0016–0.0098 0.15–0.35 0.0066 0.30
sheathing
Wire cables Galvanized metal 0.0016–0.0098 0.15-0.35 0.0049 0.25
sheathing
Wire cables Greased or asphalt 0.0098 0.25–0.35 0.0066 0.30
coated and wrapped
High strength Bright metal 0.0003–0.0016 0.08–0.30 0.0010 0.2
bars sheathing

High strength Galvanized metal 0.0003–0.0016 0.08–0.30 0.0007 0.15

sheathing
bars
Galvanized Bright metal 0.0016–0.0066 0.15–0.30 0.0049 0.25
strand sheathing

Galvanized Galvanized metal 0.0016–0.0066 0.15–0.30 0.0033 0.20

sheathing
strand

Other sources of friction

Substantial friction losses occur within the anchorage region of the tendon. Individual
strands are deviated to enter the anchor head causing a normal force between the strands and the
walls of the trumpet and anchor plate. At the live end, the force lost in the tendon can be 2% to
3% of the jacking force, as measured by Roberts, et al. [7]. A similar loss occurs at the dead end
anchorage but is of little consequence. The engineer must include the live-end loss in order to
accurately predict friction losses and elongations in the girder.

3.2.2 Elastic shortening losses

Elastic shortening losses for a post-tensioning tendon depends on the time that tendon
was stressed relative to the other tendons in the structure. It also depends on the change in strain
in the concrete along the tendon profile in question for an internal tendon, or the translation of the
anchorage and deviation points for an external tendon. Calculation of the elastic shortening stress

126
loss is usually calculated for a group of ungrouted post-tensioning tendons using an equation such
as Equation 3-6 suggested by Zia, Preston, Scott and Workman [28].

ES = 0.5Esfcpa/Eci (3-6)

ES = Stress change in a tendon from elastic shortening of the girder

Es = Modulus of elasticity of the prestressing steel

Eci = Modulus of elasticity of the concrete at the time of stressing

fcpa = Average compressive stress in the concrete along the member length at the
center of gravity of the tendons immediately after the prestress has been
applied

This is also the equation used in the AASHTO Standard Specification for Highway
Bridges [23] for use in bridge design. The equation requires that the designer know the modulus
of elasticity of the concrete and the stress in the concrete along the tendon profile. The stress in
the concrete may be difficult to predict in girder cross-sections with significant shear lag and the
resultant poor diffusion of post-tensioning forces. The modulus of elasticity of the concrete
depends greatly on the modulus of elasticity of the coarse aggregate used at the site. The concrete
modulus of elasticity may differ from the modulus calculated from the simple equation in
AASHTO [23].

3.2.3 Creep and shrinkage

Losses due to creep and shrinkage depend on environmental factors, material properties
of the concrete and its constituents, and the proportions of the structural member. AASHTO [23]
uses Equation 3-7 for shrinkage loss in post-tensioning tendons. It is a function of relative
humidity only. The 0.80 factor is intended to reduce the loss because of the potential age of a
post-tensioned member at the time of stressing. This equation is adequate for preliminary designs.

SH = 0.80 (117000-1030RH) (3-7)

SH = Tendon loss due to shrinkage in MPa

RH = Average relative humidity in %

Equation 3-8 for the stress loss in a tendon from shrinkage was given by Zia, Preston,
Scott and Workman [29] for ACI–ASCE Committee 423. The equation estimates the amount of
water leaving the concrete pore volume as a function of time, relative humidity and shape,

127
although it does not account for differences in concrete mixes, such as water to cement ratio or
percentage of aggregate.

SH = 8.2 x 10-6 KshEs(1-0.0018V/S)(100-RH) (3-8)

SH = Tendon stress lost due to shrinkage in Mpa

Es = Modulus of elasticity of the prestressing tendon

V/S = Volume to surface area ratio of the concrete

RH = Average relative humidity in %

Ksh is a function of the time elapsed between moist curing and

prestressing

ACI Committee 209 studied the effects of concrete constituents and other factors on
creep and shrinkage [29]. Equation 3-9 was developed for estimating shrinkage strain in concrete
and includes these factors:

εsh = εshu St Sh Sth Ss Sf Se Sc (3-9)

εshu = 0.0008 for moist cured concrete

εshu = 0.00073 for steam cured concrete

St = t/(35+t), t = time in days from age 7 days
(moist cured)
St = t/(55+t), t = time in days from age 1 to 3 days
(steam cured)
Sh = 1.4 - 0.01RH, 40% < RH < 80%
Sh = 3.0 - 0.03RH, 80% < RH < 100%
Sth = 1.0, minimum thickness < 152mm
Sth = 0.84, minimum thickness < 229mm
Sth = 0.80, minimum thickness < 356mm
Sth = 0.72, minimum thickness < 508mm
Sth = 0.70, minimum thickness < 610mm
Ss = 0.89 + 0.00157*(Slump), Slump in mm
Sf = 0.33 +F/75, F < 50%

128
 Sf = 0.88 +F/430, F > 50%
F = % fines by weight of total aggregate
Se = 0.95 + A/120, A = air content in %
Sc = 0.72 + C/14544, C = cement content in N/m3

These equations shown above are for girders free to contract along their length. The
calculation of shrinkage losses becomes more complicated when additional boundary conditions
are placed on continuous girders, such as a moment connection to a pier.

The calculation of prestressing losses from creep is similar to that for shrinkage. The
ACI Committee 209 Report [29] gives the following equation for calculating creep strain in
concrete:

Ct = εcr/εelastic = creep coefficient (3-10)

εelastic = immediate strain response to a constant load

εcr = creep strain from a constant load

Ct = Cu Kt Ka Kh Kth Ks Kf Ke (3-11)
Cu = 2.35, ultimate creep coefficient
Kt = (t0.6)/(10 + t0.6), t = days of load application
Ka = 1.25 ti-0.118, (moist cured)
Ka = 1.13 ti-0.095, (steam cured)
ti = age in days when loaded
Kh = 1.27 - 0.0067H, for H > 40%
H = relative humidity
Kth = 1.0, minimum thickness < 152mm
Kth = 0.85, minimum thickness < 305mm
Kth = 0.75, minimum thickness < 457mm
Kth = 0.70, minimum thickness < 610mm
Ks = 0.82 + (slump)/381, slump in mm
Kf = 0.90 + F/500, % fines by weight
Ke = 1.0, for air content < 6%
Ke = 0.45 + A/11, for air content A > 6%

129
 Once again, this equation provides the means for calculating the creep strain in plane
stress members. The problem becomes more complicated when the actual boundary conditions
are applied.

3.3 INSTRUMENTATION PROGRAM

Measurements were taken on tendons tested in the laboratory, on tendons in two spans in
the mainlane superstructure, and on tendons in one span of Ramp P. Span D5 on the mainlane and
span P16 on Ramp P had strain gauges installed permanently on the tendons. Tendons tested in
the bench test at the laboratory and in span D2 during the in-place friction test were only
temporarily instrumented for strain readings during stressing.

3.3.1 Bench test

Tendons under study on U.S. 183 were 19-15mm strand tendons. The stressing
equipment, tendons, and instrumentation for measuring tendon strains were tested and calibrated
in a bench test shortly before construction of the bridge superstructure. The bench test was
probably unnecessary from a construction standpoint since the first girder to be constructed had
nearly straight tendons in the bottom flange. Calibration of the ram could have been verified
during the in-place friction test. From the researchers' standpoint, the bench test provided the
means for measuring friction losses in the anchorage zones and ram separately from friction in the
duct profiles. The bench test also allowed side by side calibration of the strain gauges, epoxy
sleeve system, load cell and pressure transducer used to measure tendon strains in the field tests.

The bench test, shown in Figure 3.6, was conducted at the Ferguson Structural
Engineering Laboratory at the University of Texas at Austin. The test specimen was a 19-15mm
strand tendon 22.5m long. Dead end and live end bulkheads were constructed from steel and were
seated on either end of an access passage beneath the laboratory's structural load floor. The
tendon was hung freely in a duct down the length of the access passage. Both the live end and
dead end anchor plates and trumpets were cast into steel pipe lined cylinders of concrete to more
accurately represent the anchorage zones. A load cell was placed between the anchor head and the
anchor plate at the dead end to measure tendon force. Pressure on the ram was measured with a
pressure transducer. Strain in the tendon was measured using electronic strain gauges on
individual wires of the tendon, a Demec extensometer with Demec points mounted in epoxy
sleeves around the tendon, and by elongation measurements with a ruler. As the tendon was
stressed, the individual strands of the tendon were deviated in the trumpet in order to match the
130
geometry of the anchor head. This deviation caused a normal force and friction to occur between
the tendon and the trumpet. The loss in tendon force between the live end and the dead end of the
tendon is shown in Table 3.7. Confirming measurements by Roberts [10], the loss through the
stressing equipment and anchorage hardware was over 2%. The load cell was calibrated face to
face with the ram in a previous test. The load cell gave a slightly nonlinear response in this test,
with maximum error being 0.93% of the jacking force. This error was accounted for in the tests to
follow, when this load cell was used, by using the actual nonlinear calibration curve. The live end
and dead end anchorage zone losses in Table 3.7 were calculated based on the angle changes of
the strands in the trumpet and anchor plate. The angle change at the dead end was less than that at
the live end because the load cell increased the distance of the anchor head to the anchor plate.

22.5m
Load cell Concrete reaction struts
Steel frame Freely suspended duct and 19 - 15mm strand tendon

Anchor and trumpet assembly Steel frame

Anchor head See Detail below
Ram

Anchor plate Anchor head

Trumpet
Strand deviation

Tendon

Friction between Ram nose

strands and trumpet
Detail
Figure 3.6 Bench test

131
 Table 3.7 Bench test results
Measured Measured Total Loss Calculated Live Calculated Dead
Ram Force Load Cell (%) End Anchorage End Anchorage
(kN) Force (kN) Zone Loss (kN) Zone Loss (kN)
0 0 0 0 0
445 441 0.8 2.5 1.5
890 887 0.3 1.9 1.1
1334 1335 0.0 -0.6 -0.4
1779 1761 1.1 11.4 6.6
2224 2176 2.1 30.4 17.6
2669 2607 2.3 39.3 22.7
3114 3050 2.0 40.6 23.4
3558 3486 2.0 45.6 26.4
3959 3871 2.2 55.8 32.2

The bench test provided an apparent modulus of elasticity of the tendon for use with the
electronic strain gauges. The strain gauges were placed parallel to the helical outer wires of the
prestressing strands. As would be expected, the strains measured by the strain gauges differed from
those measured from the elongation of the entire tendon. The strains measured from the epoxy
sleeve system were somewhat inconsistent in both the bench test and in tests to follow. The epoxy
sleeve system measured strains therefore were used only as a backup to the electronic gauges. The
final results for the instrumentation in the bench test are shown in Figure 3.7.
1600

1400 Strain gauges

1200
Epoxy sleeve Demecs
Tendon Stress in MPa

1000
Elongation measurement
800

600
Measured Moduli of Elasticity
400
Strain gauges 208,900MPa
200 Epoxy sleeves 201,700MPa
Elongation 190,500MPa
0
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
Tendon Strain

Figure 3.7 Instrumentation calibration from bench test

132
 The tendon modulus of elasticity determined from the elongation measurements, at
190,500MPa, was 1.3% lower than the AASHTO [23] suggested value of 193,000MPa.

3.3.2 Mainlane girder tests

The mainlane girders were constructed as simple spans. The internal and external tendon
profiles are shown in Figures 3.8 and 3.9.

Instrument location 1 in span D5

Instrument location 2 in span D5

Instrument location 3 in span D5

155mm

Jacking end
1387mm
10.17m (10.30m) 19.75m (17.55m) 10.17m (10.30m)
40.84m (38.90m)
CL pier CL
Note: Lengths for Span D2 shown without parenthesis pier
Lengths for Span D5 shown in parenthesis
Figure 3.8 Mainlane girder D5 tendon profiles–elevation

2 @457mm 1295mm
at anchorage
1295mm at

3 4 5 6
anchorage

1 2

7 8 9 10
2 @229mm 1295mm 11 12 13 14

Section Looking Up Station

Figure 3.9 Mainlane girder D5 tendon profiles–end view

133
 All tendons were 19-15mm strand tendons, with the exception of the four wing tendons.
The six deviated external tendons had no intended angle change at any point other than at the two
deviators. A summary of the deviation angles is given in Tables 3.9 and 3.11.

The eight bottom flange internal tendons followed the geometry of the bridge girder
except for an angle change at the dead end of the profile. This angle change was provided to ease
installation of the tendons. The tendons were pulled into their ducts from the dead end to the live
end. In retrospect, the angle change was unnecessary and required that the dead end anchor
segment have different anchorage geometry than the live end anchor segment.

In-place friction tests on span D2

Strand forces for four profiles of 19-15mm strand tendons were measured in span D2 on the
mainlane. Ram pressure was measured at the live end with a pressure transducer and a hydraulic
dial gauge. Dead end force was measured with a load cell. The entire test setup was identical to
that of the bench test with the exception of the tendon length and profile. Two bottom flange
internal tendons were tested (tendons 12 and 13), and two deviated external tendons were tested
(tendons 2 and 3). The results of these tests are shown in Table 3.8. The anchorage zone losses
shown in columns (4) and (5) of Table 3.8 were assumed to be proportional to those determined in
the bench test. The live end anchorage zone losses were calculated by dividing the appropriate
measured live end force in column (1) of Table 3.8, by the measured live end force from the bench
test in column (1), then multiplying this ratio by the live end anchorage loss in column (4),
measured in the bench test. The dead end anchorage zone losses in column (5) of Table 3.8 were
calculated by dividing the appropriate measured dead end force in column (2) of Table 3.8, by the
measured dead end force from the bench test in column (2), then multiplying this ratio by the dead
end anchorage loss in column (5), measured in the bench test. The losses at the deviators or dead
end curvature shown in column (6) of Table 3.8 were determined by taking the difference between
the total measured losses and these assumed anchorage losses. Based on these measurements and
assumptions the average force loss through two deviators was 7.5% of the live end force. The
average loss in tendons 12 and 13, caused by the curvature in the duct near the dead end of the
tendon, was only 0.4%, much less than the losses assumed to occur in the live and dead end
anchorage zones.

134
 Table 3.8 In-place friction test results on span D2
(1) (2) (3)=(1)-(2) (4) (5) (6)=(3)-(4)-
Measured Measured Measured Assumed Assumed (5)
Test Live End Load Cell Total Loss Live End Dead End Calculated
Tendon Force Force (kN) Anchorage Anchorage Deviators 1
(kN) (kN) Loss Loss and 2 Loss
(kN) (kN) or Dead End
Curvature
Loss,
(kN) or % of
live end
force
Bench 3959 3871 88 56 32 -
2 3904 3512 392 55 29 308 7.89%
3 3818 3461 357 54 29 274 7.18%
Ave. 2- 3861 3487 374 55 29 290 7.51%
3
12 3824 3733 91 54 31 6 0.16%
13 3849 3742 107 54 31 22 0.57%
Ave. 3837 3738 99 54 31 14 0.36%
12-13

The friction coefficients and wobble angles and coefficients, used in Equations 3-3, 3-4 and
3-5, have been calculated based on the measured and calculated values given in Table 3.8. The
friction coefficients and wobble angles and coefficients are presented in Table 3.9. The angle
change given in column (1) in Table 3.9 is the intentional angle change, and was calculated using
the contract drawings. The percentage loss in column (2) is the loss occurring over the length of
tendon under consideration. For the deviated external tendons 2 and 3 this percentage loss
includes only the losses occurring in the deviator pipes. For the internal tendons 12 and 13 only
the loss occurring within the curvature in the ducts near the dead end of the tendon was
considered. The friction coefficient of 0.25 in column (3) was assumed for the deviated external
tendons. Using the percentage loss in column (1), the assumed friction coefficient from column
(3), and Equations 3-3 and 3-4, the wobble angles β in column (4) and the wobble coefficients k in
column (5) were calculated. The average β for each deviator was 0.036 radians, which compares
well with Robert’s [7] recommended value of 0.04 radians per deviator. The friction
coefficient μ (column (3) of Table 3.9) was calculated for the internal tendons 12 and 13 assuming
no wobble losses. The percentage loss for tendons 12 and 13 in column (2) was so small as to be

135
unbelievable. The assumed dead end or live end anchorage zone losses for these tendons from
Table 3.8 must have been too large, or errors were made during the tendon force measurements.

Table 3.9 Measured losses and calculated wobble coefficients from in-place friction test on
span D2
(1) (2) (3) (4) (5) (6)
Angle Calculated μ β
Test Tendon Change % Loss Fiction (Radians) k K (m-1)
(Degrees) Coefficient
Bench 0 0 0.25*** 0 0 -
Deviated
External
2 14.162** 8.00%** 0.25*** 0.086** 0.349 -
3* 14.102** 7.28%** 0.25*** 0.056** 0.228 -
Straight
Internal
12 11.242 0.16% 0.0082 - - 0
13 11.242 0.58% 0.0296 - - 0
* Visible misalignment of anchorage and deviator noted
** Includes 2 deviators
*** Assumed friction coefficient

Table 3.10 gives the measured and the calculated elongations for tendons 2, 3, 12 and 13.
Column (1) of Table 3.10 contains the elongations measured during stressing of the tendons,
adjusted for elastic shortening of the box girder and wedge seating at the dead end. The measured
elongations in column (1) of Table 3.10 were consistently shorter than the calculated elongations
in column (3). The elongations in column (3) were calculated using the losses from Table 3.8 and
the tendon’s elastic modulus determined from the bench test shown in column (4). The predicted
elongations in column (3) assumed that each deviator performed identically. This was most likely
not the case. Also, the loss in the live end anchor segment was probably larger than that seen in
the bench test because of the additional duct length in the anchor segment beyond the anchorage
device. The measured moduli of elasticity in column (2) of Table 3.10 were calculated using the
measured elongations in column (1), and the losses from Table 3.8. The measured moduli were
inconsistent between the deviated external tendons and the internal tendons. The average
measured modulus of elasticity was larger for the external tendons, indicating the tendon force
was lower than expected over its entire length. The average modulus measured for the internal
tendons indicates that the bench test modulus may have been lower than the actual modulus of
elasticity for these tendons.
136
 The wobble angle β in column (6) of Table 3.10 was calculated using only the design
geometry of the tendons from Table 3.9, an assumed live end anchorage zone loss of 2%, the
tendon’s elastic modulus determined from the bench test, and the measured elongations. The
unrealistically large β angles in column (6) indicate that the bench test modulus was different from
that of the in-place friction test tendons. The load cell may also have given erroneous readings.
There is a question as to the accuracy of the load cell since nonlinear output was observed during
the bench test and load cell calibration. It is likely that the load cell was not functioning properly
since the 11.242o tendon deviation in tendons 12 and 13 produced only about 0.5% force loss. In a
best case scenario this friction loss should have been about 3%, assuming a coefficient of friction
of 0.15 and no wobble. The load cell calibration was not checked after the friction tests.

Table 3.10 Elongations from friction tests on span D2

(1) (2) (3) (4) (5) (6)
Test Measured Measured Calculated Bench Assumed Resultant
Tendon Elongation Modulus Elongation Test Friction Wobble
(mm) (MPa) (mm) Modulus Coefficient Angle
(MPa) μ β (Radians)
Deviated External
2 275 203,400 294 190,500 0.25 0.307*
3 277 198,200 288 190,500 0.25 0.166*
Ave. 276 200,800 291 190,500 0.25 0.237*
2-3
Straight Internal
12 296 193,500 301 190,500 0.15 0.211
13 293 196,800 303 190,500 0.15 0.563
Ave. 294.5 195,200 302 190,500 0.15 0.387
12-13
* Angle applied at each deviator (2 total)

Mainlane girder friction tests on span D5

Strand forces for six sets of 19-15mm strand deviated external tendons were measured on
span D5 on the mainlane. Ram pressure was measured at the live end with a pressure transducer,
and tendon strains were measured with strain gauges and epoxy sleeve systems on either side of
the deviators. The strain measurements taken between the live end anchorage and the first
deviator gave an indication of the effective wedge seating loss beyond the anchorage zone itself.
Very little strain change was measured in the two tendon lengths beyond the first deviator that was
caused by wedge seating. The wedge seating length shown in column (6) of Table 3.11 was
137
calculated from the measured strain change in gauges on the adjacent length of tendon. The
average wedge seating length was 5.19mm with a small standard deviation of 0.55mm. The
wobble coefficients determined in this test using the strain gauge data are shown in columns (5)
and (6) of Table 3.11. The percentage loss through the deviators measured by the strain gauges,
shown in column (2) of Table 3.11, was consistently nearly twice the loss measured with the load
cell on span D2. The location of the gauges on the top strands of each tendon may not have
provided strains representative of those in the entire tendon, even though the data were very
consistent. Arrellaga has shown that substantial force differences can exist between the various
wires and strands in the same tendon [11]. The calculated wobble angles and coefficients β and k
in columns (5) and (6) of Table 3.11 varied considerably from tendon to tendon because the
friction coefficient μ in column (3) was assumed to be constant at 0.25. Thus, the calculated
wobble friction also have included any reduction or increase in friction resulting from errors in the
selection of μ. The standard deviations for the measured wobble values were on the order of the
wobble values themselves, and the averages were unrealistically large.

Table 3.11 Span D5 friction test results from strain gauges

(1) (2) (3) (4) (5) (6)
Test Intentional Measured Assumed Calculated Calculated Calculated
Tendon Angle Loss* Friction β∗ k Wedge
Change* Coefficient Seating
μ Length
1 7.122o 7.16% 0.25 0.173rad 1.39 -
2 7.017o 7.22% 0.25 0.177rad 1.45 4.50mm
3 6.966o 4.84% 0.25 0.077rad 0.63 5.84mm
4 6.966o 6.78% 0.25 0.159rad 1.31 5.28mm
5 6.975o 5.62% 0.25 0.110rad 0.90 5.13mm
6 7.040o 6.80% 0.25 0.159rad 1.29 -
Ave. 1-6 6.40% 0.143rad 1.16 5.19mm
σ 1-6 0.040rad 0.32 0.55mm
* Measured through one deviator

Another way of determining the wobble coefficients and angles is to compute them based on
measured elongations. Table 3.12 gives the measured elongations in column (1) and the wobble
angles and coefficients β and k in columns (4) and (5) calculated using these elongations and the
elastic modulus found in the bench test. Once again, the losses in span D5 appear to be larger than
those measured in span D2. However, the measured elongations were consistent with those

138
measured in span D2 when the differences in tendon length were accounted for. The assumed loss
distribution over the length of the tendons, and the modulus of the steel strand, both have
uncertainties. Because of the calculation method any variation in actual strand area would be
perceived as a change in elastic modulus. The average values for β and k in Tables 3.11 and 3.12
actually compare quite well on average, even though the standard deviation of the individual
wobble angles and coefficients is 28% of the average based on the strain gauge data, and 22% of
the average based on the elongation data.

Table 3.12 Measured elongations from friction tests on span D5

Test (1) (2) (3) (4) (5)
Tendon Measured Bench Test Assumed Calculated Calculated
Elongation Modulus (MPa) Friction β∗ (Radians) k
(mm) Coefficient
1 262 190,500 0.25 0.190 1.53
2 267 190,500 0.25 0.116 0.95
3 265 190,500 0.25 0.146 1.20
4 264 190,500 0.25 0.164 1.35
5 267 190,500 0.25 0.116 0.95
6 267 190,500 0.25 0.115 0.94
Ave. 1-6 265 190,500 0.25 0.141 1.15
σ 1-6 2.1 0.031 0.25
* Measured through one deviator

Elastic shortening losses

The elastic shortening losses measured in the instrumented external tendons are given in
Table 3.13. Strain changes in the tendons were symmetric about the longitudinal centerline of the
bridge since the tendons were stressed in pairs using two rams and one hydraulic pump.
Instrument locations 1, 2, and 3 are shown in Figure 3.8. Tendon locations and deviations are
shown in Figure 3.9. In general, the data were very consistent. Stressing of the deviated external
tendons tended to cause more elastic shortening loss in the deviated legs of the previously stressed
tendons at gauge locations 1 and 3. Stressing of the bottom slab internal tendons produced more
elastic shortening loss in the middle leg of the instrumented deviated tendons than stressing of the
deviated tendons. This is because the bottom slab internal tendons had greater eccentricity with
respect to the girder cross-section than did the deviated external tendons.

139
 Table 3.13 Elastic shortening losses in tendons in span D5
Strain (Stress) Change from Elastic Shortening
Average for All Stressed Tendons at
Tendon Each Gauge Location in με (MPa)
Stressed 1 2 3
Deviated
2 and 5 -48.9 -51.9 -41.2
1 and 6 -38.2 -52.1 -39.9
Ave. Deviated -43.6 (-9.1) -52.0 (-10.9) -40.6 (-8.5)
Straight
10 and 11 -39.9 -58.7 -40.1
9 and 12 -40.7 -59.3 -40.7
8 and 13 -41.4 -58.7 -40.2
7 and 14 -37.7 -57.6 -37.4
Ave. Straight -39.9 (-8.3) -58.6 (-12.2) -39.6 (-8.3)

3.3.3 Ramp girder tests

The five span continuous girder on Ramp P was constructed in balanced cantilever from
the four interior piers, then made continuous with cast-in-place closures, and both internal and
external post-tensioning. The interior anchor segment anchorage zones for the external tendons
were constructed as deviation saddles. The anchorages deviated downward at the live end to
facilitate stressing inside the box girder core. Therefore, even though the tendons used were 19-
15mm strand tendons, the anchor zone configuration did not exactly match that of the bench test.

Friction losses
The test program consisted of strain measurements of six external tendons on either side
of two deviators using multiple electronic strain gauges. Measurements were taken near the
downstation deviator and the adjacent horizontal deviator in span P16 (see Figure 3.10). Tendons
T1 and T2 were continuous over spans P16 and P17, and were jacked from the downstation end
only, near the instrumentation. Tendon T3 was continuous over spans P14, P15 and P16, and was
jacked from both ends. The geometry of the tendons through the deviators under study is shown
in Figures 3.10 and 3.11, with the angle changes given in column (1) of Table 3.14. The results
from the friction tests are also given in Table 3.14.

140
 Jacking end for tendons T1 and T2
Jacking end for tendon T3 is at P14 for first pull
Tendon instrument location H and P17 for second pull

Tendon instrument location I

Tendon instrument location J

Up Station
P16 P17

14.36m 7.70m 10.74m

54.86m
Figure 3.10 Ramp P tendon geometry–elevation

1321mm
1016mm
Dimensions at CL Pier P16
965mm

T3R T4 T4 T3L
T1R T1L
T2R T2L
1460mm
1232mm
1079mm

2 @ 229mm 203mm

Section Looking Down Station

Horizontal curvature not shown
Figure 3.11 Ramp P tendon geometry–end view

141
 Table 3.14 Ramp P friction test results from strain gauges
(1) (2) (3) (4) (5)
Test Tendon Angle Change Measured % Assumed Calculated Calculated
Loss Friction β∗ k
Coefficient (Radians)
T1L 4.870o 18.2 0.25 0.719 8.45
T1R 7.449o 15.6 0.25 0.548 4.22
T2L 4.344o 6.6 0.25 0.197 2.60
T3R 9.695o ** 11.8 0.25 0.333 1.97
Ave. 4.31
* Additional angle β is applied at each deviator and each anchor or pier segment (9 total for
tendons T1L, T1R, T2L and T2R, and 14 total for tendons T3L and T3R)
**Angle change through two deviators

The strain gauges consistently measured a large amount of friction loss through the
deviators, as shown in column (2) of Table 3.14, as was the case in span D5. The measured loss
was as high as 18% through the deviator for tendon T1L. Expected loss through this deviator
would be 3% using μ=0.25 and β=0.04 in Equation 3-3. Since the gauges were all located on the
top strands of the tendon, and these were the strands in direct contact with the deviator pipe, the
measured values may be representative of these outer strands and may not be accurate for the
tendon stress on average. The average wobble coefficient k=4.3 in column (5) of Table 3.14 is
unrealistic. A wobble of this magnitude would require a construction error resulting in an angle
change of about 40o at each deviator.

The wobble angles β and wobble coefficients k , shown in columns (4) and (5) of Table
3.15, calculated based on the bench test tendon modulus and the measured elongations, also
indicated high friction in the ramp P tendons. The results were based on the measured elongation
data in column (1) of Table 3.15, and the bench test modulus in column (2). The calculated
wobble coefficients in Table 3.15 were sensitive to the value of the tendon elastic modulus. Based
on the elongations and dead end force measurements on the mainlane girder D2, the bench test
modulus was most likely smaller than the elastic modulus of the tendons in ramp P. Therefore, the
average calculated wobble coefficient k=1.1 in column (5) of Table 3.15 most likely was
unrealistically large. The friction loss in tendons T3L and T3R, measured by a lift-off test when
stressing the downstation end of the tendon, revealed much less apparent friction loss. The results
of the lift-off test are shown in Table 3.16. The calculated wobble terms in columns (4) and (5)
are small, with k=0.11 on average. Since the downstation end of the tendon was stressed many

142
hours after the upstation end, the force distribution along the tendon may have changed over time.
This phenomenon has been measured both in the laboratory [22] and in the field [10], and is more
pronounced in tendons with high friction loss. Force near the live end of these tendons may have
decreased, while force at the dead end increased. Therefore, using the lift-off test to measure
friction loss may lead to overestimation of the average stress in the tendon because of the lapse in
time.

Table 3.15 Measured elongations from friction tests on Ramp P

(1) (2) (3) (4) (5)
Test Tendon Measured Bench Test Assumed Calculated β∗ Calculated
Elongation Modulus Friction (Radians) k
(mm) (MPa) Coefficient
T1L 593 190,500 0.25 0.0404 0.393
T1R 550 190,500 0.25 0.132 1.190
T2L 558 190,500 0.25 0.128 1.287
T2R 552 190,500 0.25 0.144 1.374
T3L 798 190,500 0.25 0.124 0.976
T3R 760 190,500 0.25 0.157 1.161
Ave. 1.06
* Additional angle β is applied at each deviator and each anchor or pier segments (9 total for
tendons T1L, T1R, T2L andT2R, and 14 total for tendons T3L and T3R)

Table 3.16 Lift-off test results for tendon T3

(1) (2) (3) (4) (5)
Test Tendon Dead End Assumed Total Angle Calculated β Calculated
Force/ Live Friction Change ∗ k
End Force Coefficient (Radians) (Radians)
T3L 0.645 0.25 1.775 -0.0195 -0.011
T3R 0.558 0.25 1.893 0.441 0.233
Ave. 0.023** 0.11
* Total additional angle β for all deviators and anchor or pier segments
** Applied at each of the 14 deviators and saddles

Elastic shortening losses

The measured elastic shortening losses are given in Table 3.17. Losses in some tendons
could not be measured because gauges were lost when three of the tendons failed during stressing.
Instrument locations H, I, and J are shown in Figure 3.10. The elastic shortening strains from
stressing each of the external tendons are not directly comparable to each other. Since the bridge

143
was built on a substantial horizontal curve, no two tendon profiles were alike. The first tendon
stressed, tendon T1L, had the most elastic shortening loss at -30.4MPa. This elastic shortening
loss was small, at less than 2.5% of its stress after jacking. The total elastic shortening stress loss
at location I was only 1.4% on average for the group of 6 tendons.

Table 3.17 Elastic shortening losses–Ramp P external tendons

Strain (Stress) Change from Elastic Shortening
Tendon Gauge in Each Stressed Tendon in με (MPa)
Stressed Location T1L T1R T2L T2R T3R
T1R H -16.4
T2L H -19.6 -17.8
T2R H -17.8 -26.0 -21.5
T3R H -16.4 -19.6 -17.1 -15.2
T3L H -8.6 -7.6 -6.3 -8.9 -12.0
Total Loss H -78.8 -71.0 -44.9 -24.1 -12.0
(-16.5) (-14.8) (-9.4) (-5.0) (-2.5)

T1R I -27.2
T2L I -30.4 -30.4
T2R I -31.1 -35.5 -27.6
T3R I -33.2 -32.3 -32.3 -33.6
T3L I -23.8 -23.8 -25.6 -22.8
Total Loss I -145.7 -122.0 -85.5 -56.4
(-30.4) (-25.5) (-17.9) (-11.8)

T3L J -36.7

Long term external tendon strains

Although almost all of the span D5 tendon gauges became debonded from the tendon
after only a week beyond the time of stressing, the gauges on the Ramp P external tendons
performed well for months after stressing. Figure 3.12 shows a plot of tendon stress over time for
tendon T1 on either side of a vertical deviator. The most pronounced loss apparent in the figure is
the initial friction loss between the gauges at location H and at location I at about 200MPa,
although one of these plots is for T1L and the other is for T1R and the tendons were stressed at
different times. About 40MPa of the total time dependent loss at location H occurred in the first
month after stressing. Tendon T1 at location I only lost about 20MPa in the first month after
stressing. The tendons were not grouted during this month, so it is conceivable that some of the
144
force in the tendon at location H migrated through the deviator to the tendon length at location I
during this time period because of tendon slippage through the deviator. The additional loss seen
to occur between April 1997 and June 1997 may be due in part to the average seasonal
temperature variation and the difference in the coefficients of thermal expansion between the
tendon steel and the girder concrete. The total loss over the ten month period was about 70MPa
on average, amounting to 6% of the original tendon force following stressing. The tendon forces
plotted in Figure 3.12 appear to be stable by August 1997.

Tendon T1 Stresses
1200 Gauge Location H

1000
Tendon stress in MPa

800 Gauge Location I

600

400

200

0
Oct 96 Dec 96 Feb 97 Apr 97 Jun 97 Aug 97 Oct 97 Dec 97
Figure 3.12 Strain change in ramp tendons over time

3.3.4 Large ramp pier tests

Pier P16 was the only post-tensioned pier studied on the project. The pier had two 19-
15mm strand U-shaped tendons. The U-bend was provided by rigid pipes cast in the footing, as
shown in Figure 3.13. The anchorage zone geometry was straight, so that a direct comparison
could be made between anchorage zone losses measured in the bench test and the pier tendon
tests. Tendons were cut to an approximate length, lifted by crane to the top of the pier, and
inserted into the tendon ducts. The crane was then used to pull the tendons through the full
lengths of the ducts. After installation of the post-tensioning anchorage heads and wedges, the
tendons were stressed in the order shown in Figure 3.14. The tendon ducts were placed so as to
cross each other within the footing in plan. Thus, the anchorages at points 1 and 2 belonged to the
same tendon.

145
20879mm

19 - 15mm dia.
strand tendons

180o bend in
footing

Figure 3.13 Pier P16 elevation and tendon profiles

146
 Up Sta

C400

3 C402 2

C412
1 4

C410

Plan View - Pier P16

Figure 3.14 Sequence of final post-tensioning of pier P16 tendons

Friction losses
The force at the dead end of the tendon was measured after stressing the other end by
performing a lift-off test with the ram. The lift-off stress on the dead end of the tendon was about
49% of the live end stress on average after one pull. Subtracting out the anchorage zone losses
based on the results of the bench test, the measured coefficient of friction μ in the rigid steel pipe
with 180o angle change was 0.240 in the first tendon and 0.214 in the second tendon. This agrees
well with the usual friction coefficient μ=0.25 chosen for rigid pipe deviators in the first tendon.
The friction loss in the U-bend pipe of the pier was of little consequence since the tendons were
stressed from both ends, and no substantial friction occurred at any point other than in the U-bend
pipe.

Elastic shortening losses

Concrete strains were measured once every minute during the post-tensioning process. In
addition, a pressure transducer was attached to the hydraulic ram to measure jacking forces at the
four tendon anchor locations. Figure 3.15 shows a plot of readings taken by the pressure
transducer and strain gauges oriented along the vertical axis of the pier in segment PC16-1 over

147
the course of the post-tensioning process. The tendons were stressed in the order indicated
adjacent to the plot of the pressure transducer output.

60.0 200

Pressure Transducer Reading (MPa)

180
20.0
Strain Change (mm/mm x 10-6)

160
-20.0 140

120
-60.0
C400 100
-100.0 C402
80
C410
C412 60
-140.0 2
1
3 4
Pressure 40
-180.0 Transducer
20

-220.0 0
1:50 AM

2:00 AM

2:10 AM

2:20 AM

2:30 AM

2:40 AM

2:50 AM

3:00 AM

3:10 AM

3:20 AM

3:30 AM

3:40 AM

3:50 AM
May 1, 1996
Figure 3.15 Selected axial strains in segment PC16-1 during pier post-tensioning, north-south
axis locations

As shown in Figure 3.15, gauges located along the axis running north-south through the
pier’s cross-section showed several well-defined trends during post-tensioning. The different
strain responses after stressing at location 1 and 3 that are seen on opposite sides of the cross
section were caused by the eccentricity of effective prestress force on the pier due to friction
losses. In a frictionless tendon the forces at 1 and 2 would be equal after stressing from one end.
Thus the axial stress would be uniform. However, due to friction the tendon force was not
equalized until the tendon was jacked from location 2, and then again from location 4. Because
position 1 was primarily eccentric with respect to the east-west centroidal axis, large differential
strains occurred in the north and south faces.

After stressing at position 2 was complete, the post-tensioning loads experienced by the
pier were nearly symmetric about the east-west axis. During stressing at point 2, concrete strains
at gauges C410 and C412 were increasing because of the eccentricity of the jack at point 2 was

148
beyond the Kern point for the cross section. This strain increase produced a gain in force of the
tendon length anchored at point 1. The same behavior was seen in the tendon anchored at points 3
and 4. Concrete strain changes measured by the gauges during stressing at locations 3 and 4
produced an elastic shortening loss in the tendon anchored at points 1 and 2 of approximately
82με or 16MPa.

3.4 DISCUSSION OF RESULTS

The results from these post-tensioning force and strain measurements can be compared to
the results of other studies, field data, and typical design solutions based on methods
recommended in bridge design codes. These comparisons should give bridge designers some
insight into the accuracy of their design calculations, and most importantly show which variables
have the greatest impact on accuracy.

3.4.1 Friction losses

Friction losses were measured for internal tendons, for external deviated tendons, and in
the anchorage zones. The impact of wedge seating length is also directly related to the frictional
performance of the tendon.

Internal tendons
The internal tendons studied on the project were straight bottom slab tendons. Curvature
changes over the majority of the length of the tendon were very small. Almost all of the friction
loss in these tendons was due to losses in and near the anchorage regions, not from wobble friction
along the length of the duct. The segments for this project were cast using the short line method.
The bulkhead had permanent holes cut for locating the bottom slab tendons. The semi-rigid steel
ducts were held in place between the bulkhead and match casting segment with an inflatable
mandrel. This mandrel was capable of holding the ducts solidly in place during the casting
process. Measured wobble losses in these ducts were very small, as seen in Table 3.9. The
modulus of elasticity of the tendons calculated from the elongation data for tendons 12 and 13 in
span D2 was somewhat higher than that calculated in the bench test. If the tendon modulus of
elasticity is assumed to be 193,500MPa, as measured for tendon 12 in Table 3.10, the β angles in
Table 3.10 would decrease substantially. β for tendon 12 would become 0.0013 radians, which
relates to K=0.0000048m-1. β for tendon 13 would become 0.2260 radians, or K=0.00084m-1.
The change in elastic modulus from 190,500MPa to 193,500MPa represents a 1.6% increase in

149
stiffness. Such a change could have resulted from a 0.8% or 0.04mm increase in wire diameter,
which is within the wire ASTM Specification A 416-74 [30] limits for prestressing wire.

Most of the internal tendon research data presented in Section 3.2 primary was concerned
with determining the coefficients of friction and wobble in draped internal tendons. Draped
profile internal tendons primarily are used in non-segmental cast-in-place box girders, although
they can also be used in segmental precast and cast-in-place girders. The ducts can not be held in
place using mandrels and therefore are subject to placement error during construction. Wobble
coefficients were found in other studies to range from 0.0007m-1 to 0.0016m-1 for commonly used
semi-rigid steel duct.

An appropriate approach for the design of tendons similar to the U.S. 183 bottom slab 19
strand tendons would not include these latter high wobble loss coefficients if the use of mandrels
during construction is to be specified, and their effectiveness verified. When using such mandrels,
a wobble loss coefficient of 0.0007m-1 and a friction coefficient of about 0.16 should be chosen for
use along the majority of the length of the tendon. Friction will be generated from curvature
changes due to the vertical and horizontal alignment of the bridge, as well as in the live end
anchorage region. Friction calculation using these coefficients would provide a conservative
result compared to the measured data presented in Table 3.9.

External tendons
The external tendons used in the U.S. 183 box girders were 19-15mm strand tendons, a
common size and efficient tendon for the span-by-span construction used on the mainlane girders.
The large jacks used to stress these tendons were hung from rigging at the open end of the
mainlane girder, and stressed from one end only. On the other hand, the multispan 19 strand
external tendons in Ramp P had to be stressed with these same large rams from inside the core of
the girder. Handling and clearances became a problem, particularly because of the obstructions on
the bottom slab and the horizontal curvature of the bridge.

Since the short-line method of casting was used, the forms holding the rigid ducts for the
deviators and anchor segments allowed little adjustment for geometry changes. Small placement
errors were bound to occur, and had a measurable impact on friction. Roberts recommended that
an additional inadvertent angle change of 0.04 radians be applied at each deviator to account for
duct misalignment. This recommendation was primarily based on results measured in girders
constructed by the span-by-span method. The inconsistency of the measured moduli of the

150
tendons in span D2 shown in Table 3.10 indicates that a misalignment loss may have occurred in
the live end anchor segment, and therefore the β values in Table 3.9 may actually be smaller than
shown. If 0.04 radians of additional angle change is applied only at the two deviators, the
calculation of the dead end force as measured by the load cell in span D2 is quite accurate. The
average force in the tendon, however, is lower than the calculation predicts, and the elongations
are lower than predicted as well.

Using the modulus of elasticity measured for tendon 12 in Table 3.10, at 193,500MPa, and
the elongation data for all the deviated external tendons tested in spans D2, D5, and Ramp P, the β
values shown in Table 3.18 were calculated. The table uses the commonly applied coefficient of
friction for external tendons of μ=0.25. The galvanized steel deviator pipes used in the first
several spans of mainlane D spine were inadvertently bent with a 2m radius instead of the 7.5m
radius drawn in the plans. The large β angles in Table 3.18 for the mainlane span D2 and D5
tendons indicate that the 0.25 coefficient of friction chosen was too small for the tight radius pipe
bend. The large β angles in Table 3.18 would have revealed themselves as visible duct
misalignments in the bridge, but these were not apparent. For these tight bends, the coefficient of
friction is more likely closer to 0.35, the maximum of the range specified by ACI-ASCE [27] for
wire cables on metal sheathing. The wobble angles for all the span D2 and D5 deviated external
tendons based on the measured elongations are recalculated in Table 3.19 using 0.35 as the friction
coefficient. These β angles appear to be more realistic for spans D2 and D5, with an average of
β=0.0373 radians. The standard deviation of the wobble angle β was high, at about σ=0.04
radians, regardless of the friction coefficient used.

151
Table 3.18 Elongation measurement based wobble coefficients for all deviated external
tendons, μ=0.25
Test Measured Assumed Total Assumed Resultant
Tendon Span Elongation Elastic Angle Friction Wobble
(mm) Modulus Change Coefficient Angle β∗
(MPa) α (rad) μ
2 D2 275 193,500 0.2472 0.25 0.2217
3 D2 277 193,500 0.2461 0.25 0.1034
1 D5 262 193,500 0.2486 0.25 0.1271
2 D5 267 193,500 0.2449 0.25 0.0634
3 D5 265 193,500 0.2432 0.25 0.0843
4 D5 264 193,500 0.2432 0.25 0.0994
5 D5 267 193,500 0.2435 0.25 0.0540
6 D5 267 193,500 0.2457 0.25 0.0530
Ave. D2, D5 0.1008
σ D2, D5 0.0554
T1L Ramp P 593 193,500 0.9256 0.25 0.0277
T1R Ramp P 550 193,500 0.9982 0.25 0.1181
T2L Ramp P 558 193,500 0.8945 0.25 0.1144
T2R Ramp P 552 193,500 0.9423 0.25 0.1297
T3L Ramp P 798 193,500 1.7748 0.25 0.1139
T3R Ramp P 760 193,500 1.8926 0.25 0.1469
Ave. Ramp P 0.1085
σ Ramp P 0.0415
* β angle applied at each deviator segment (2 total for all tendons on spans D2 and D5), or
each deviator and anchor segment (9 total for tendons T1L, T1R, T2L, and T2R, and 14 total
for tendons T3L and T3R)

152
 Table 3.19 Elongation measurement based wobble angles for deviated external tendons,
μ=0.35
Assumed Total Assumed Resultant
Test Measured Elastic Angle Friction Wobble
Tendon Span Elongation Modulus Change Coefficient Angle β∗
(mm) (MPa) α (rad) μ
2 D2 275 193,500 0.2472 0.35 0.1231
3 D2 277 193,500 0.2461 0.35 0.0385
1 D5 262 193,500 0.2486 0.35 0.0553
2 D5 267 193,500 0.2449 0.35 0.0131
3 D5 265 193,500 0.2432 0.35 0.0255
4 D5 264 193,500 0.2432 0.35 0.0363
5 D5 267 193,500 0.2435 0.35 0.0038
6 D5 267 193,500 0.2457 0.35 0.0027
Ave. D2, D5 0.0373
σ D2, D5 0.0392
* β angle applied at each deviator segment (2 total for all tendons on spans D2 and D5)

The high β values for the tendons of ramp P in Table 3.18, averaging β=0.11 radians, may
be correct. The horizontal and vertical curvature of the bridge would have required that most
every deviator pipe have a different radius bend, and that all pipes be carefully measured into
place. This was not practical, nor was the provision for adjustment of the holes in the deviator
form that held the pipes in place. To simplify the construction of deviators without increasing
friction, “diabolos” have been used to replace the bent deviator pipes [31]. A diabolo is a deviator
pipe with a trumpet bell shape on each end. The radius of the bell can be designed to
accommodate many deviation angles, so a few standard diabolos can be used in all the deviator
segments on a project. These diabolos pipes were used successfully on the Second Stage
Expressway System (SES) elevated bridges in Bangkok, Thailand. High density polyethylene
ducts passed continuously through the diabolos on these bridges.

Based on the measured data, the external deviated tendon design for spans D2 and D5 could
be performed accurately by choosing 0.30 or 0.35 as the coefficient of friction, and using the
β=0.04 at each deviator as suggested by Roberts [10]. The normal coefficient of friction of 0.25
should not be used in this case because of the tight radius bend in the deviator pipes. This tight
radius bend was detrimental to friction loss, as well as to the service level performance of the

153
deviator. Also, the β angle need not be applied at the live end anchor segment since the duct at
that location is short and straight, even though the potential for misalignment does exist.

The friction performance for the external tendons tested in Ramp P would be much more
difficult to predict than that of the mainlane tendons. Both the elongation and strain gauge data
showed that friction loss was large through the deviators. On the other hand, the lift-off data for
tendons T3L and T3R revealed that the friction forces did not stay locked in at the deviators and
pier segments. The force in the tendon tended to average out over time, and a normal design
calculation using μ=0.25 and β=0.04 applied at all deviators and pier segment saddles would
predict the force in the dead end of the tendon quite well. Unfortunately, the short measured
elongations meant that the average stress in the tendon was below that calculated using the
coefficients above. These long tendons were good candidates for lubrication with graphite since
the elongations were consistently short. The lubrication may not have helped in this case since
much of the friction loss may have occurred very near the live end.

The saddle shaped duct immediately adjacent to the inclined jacking anchorage did not
allow strands that tangled upon insertion into the anchor head to untangle, as was the usual case
with the straight anchorages and ducts of the mainlane tendons. Furthermore, the large jack used
for the 19 strand tendons within the core of the girder had to be hoisted up to be level with the
anchorage plate by a cable attached to its lifting flange, then rotated to avoid hitting its hydraulic
connections on the top flange of the girder. This necessary maneuver added a twist to the tangle
that had probably already formed. Total elongation including removal of slack was over 1m for
tendons T3L and T3R. Strands could be heard rearranging themselves in the anchor segment
saddle during the entire stressing process. Entangled strands evidently were being pulled into the
anchor head. This resulted in three failed tendons, with breakages immediately beyond the anchor
head. The long elongations, deviation saddle adjacent to the anchorage, and difficult access to the
large tendons were to blame. The tangled strands in contact with the saddle shaped duct adjacent
to the stressing anchorage probably increased friction at this location, and reduced the average
stress over the entire length of the tendon.

Misalignments and duct obstructions were noted, but not consistently of the magnitude
needed to produce the β angles in Table 3.18. Most likely the coefficient of friction between
tendon and duct in Ramp P was closer to 0.30 than 0.25. A design β angle higher than 0.04

154
radians may be warranted in spans with significant horizontal curvature, unless a diabolo type
deviator pipe is used.

Anchorage zone losses

The friction losses occurring in the anchorage zone as the strands of the tendon are flared
to enter the anchor head need to be considered. This loss at the jacking end of the tendon reduces
the stress over the entire length of the tendon, and reduces elongations accordingly. The loss at
the jacking end anchorage of the 19-15mm strand tendons was measured to be 2.2% of the jacking
force. Losses for different tendon and stressing systems may vary from this, but losses at the
stressing end of strand type tendons are generally in the 2%-3% range [10]. The anchorage zone
loss measured for the 3-strand wing tendon tested in mainlane span D2 was 1.74% for live end.
The loss measured for the 7 strand wing tendon was slightly larger at 2.33% for the live end.

Wedge seating losses

The wedge seating lengths shown in Table 3.11 indicate that the average seating length
for this jack and anchorage system was about 5.2mm. Designers usually assume a wedge seating
displacement of 6mm after the wedges are power seated. All measured seating displacements
were smaller than 6mm, so the common design value would be appropriate in this case.

3.4.2 Elastic shortening

Table 3.20 gives the measured and calculated strain change in the deviated external
tendons of span D5 due to elastic shortening of the girder from the tensioning of adjacent tendons.
The measured data, given in Table 3.13, are quite consistent from tendon to tendon. The elastic
shortening stress losses are usually calculated for a group of similar tendons that are stressed in
some sequence. The total elastic shortening of the girder along the tendon path is calculated from
the tensioning of the entire tendon group, and one half of this strain is assumed to act on every
tendon as an average elastic shortening loss. The calculated elastic shortening strains for the
tendons in Table 3.20 are from the stressing of a pair of external tendons or a pair internal tendons,
all 19-15mm strand tendons. This way the calculated values could be directly compared to the
measured values. Gauge locations 1, 2 and 3 are shown in Figure 3.8.

155
 Table 3.20 Calculated and measured elastic shortening losses in tendons in span D5
Strain Change from Elastic Shortening
Average for All Stressed Tendons at
Pair of Tendons Each Gauge Location in με
Stressed 1 2 3
Deviated measured -43.5 -52.0 -40.6
External calculated -31.4 -71.8 -37.2
Bottom Slab measured -39.9 -58.6 -39.6
Internal calculated -49.3 -88.9 -49.3

The calculated strains in Table 3.20 for the central leg of the tendons are generally larger
in magnitude than the measured strains. This difference exists for three reasons. First, the
modulus of elasticity of the concrete used in the calculations was taken from test results of
concrete samples. The modulus of elasticity of the samples, though taken from the same batch of
concrete as the structure, consistently was lower than that of concrete in the structure. Differences
in curing conditions and moisture content at the time of loading resulted in significantly different
concrete stiffnesses. A design engineer usually has little information available to aid in the
selection of the proper modulus of elasticity. This can adversely effect the accuracy of the
calculations. Fortunately, elastic shortening losses are not usually substantial, and great accuracy
is not needed. The second reason that the magnitude of the calculated values would be larger than
the measured is that shear lag effects were not accounted for in the calculation. Shear lag effects
would reduce or increase the elastic shortening of the installed tendons, depending on the tendon
stressed, when compared to an ideally behaving beam.

The third and most important reason for the difference between the measured and the
calculated strains is the slippage of the ungrouted tendons in the deviator pipes. The calculation
assumed that the tendons would not move with respect to the deviator. It is evident that the
tendons did slip, but not enough to cause equal strain change in the central and inclined legs of the
tendons. A conservative approach for design would be to calculate the minimum and maximum
possible elastic shortening strains of each leg of the tendons using either infinite or zero friction at
the deviator pipes. This is easily done since the strain change in each leg must be calculated for
the infinite friction case. The worst case of loss could then be used for the design of the girder.

Table 3.21 gives the calculated and the measured elastic shortening loss strains for the
deviated tendons of Ramp P. Gauge locations H and I are shown in Figure 3.11. The same
characteristics are seen in these external tendons as were seen in the external tendons in the
156
mainlane span D5. The calculated elastic shortening losses in the inclined and horizontal part of
the tendons tended to be inaccurate because of slipping of the tendons in the deviator, and a low
assumed modulus of elasticity of the girder concrete.

Table 3.21 Calculated and measured elastic shortening losses–Ramp P external tendons
Pair of Strain Change from Elastic Shortening
Tendons Gauge in Each Stressed Tendon in με
Stressed Location T1 T2 T3
T2 H measured -40.6
H calculated -45.5
T3 H measured -26.1 -23.8
H calculated -32.1 -28.8

T2 I measured -63.8
I calculated -79.8
T3 I measured -56.6 -57.2
I calculated -74.3 -74.3

3.4.3 Long term losses

Losses due to creep and shrinkage should be analyzed after two years or more of data
collection so that seasonal effects can be separated from the permanent losses. Since the project
segments were all precast 5 months or more before erection of the structural elements under study,
most of the shrinkage strain had occurred in the concrete segments before any strain gauges were
recording data. Also, the concrete was quite aged before loading, thereby greatly reducing its
creep potential. For these reasons, long term losses may not be as important as seasonal losses and
gains in stress from average ambient temperature changes.

As seen in Figure 3.12, tendon stress losses from creep and steel relaxation are only about
6% of the total tendon force. It is doubtful that the state-of-the-art methods for creep calculation
would be useful or necessary for this case. Steel relaxation could be responsible for up to 3% of
the long term loss. To predict the seasonal stress change in the tendons, the designer must know
the coefficient of thermal expansion of the concrete, which is mostly a function of the coarse
aggregate, the coefficient of thermal expansion of the prestressing steel, the seasonal temperature
fluctuation, and the time of year and temperature of the materials on the day of stressing.
Approximations will be necessary since the bridge design would be completed well before
construction began.

157
3.5 CONCLUSIONS
The following conclusions have been made based on the measured data, other field and
laboratory data, and field observations:

1. Measurements of live end anchorage zone friction losses in the laboratory bench test and
in-place friction test indicated that an assumed design loss of 2% would be sufficient,
unless actual live end losses are known from previous measurements.

2. Measured wedge seating losses were slightly less than the design value of 6mm. The
current design value is adequate.

3. The bench test proved to be of little value to all parties except the researchers, other than
as a basic calibration trial of the various pieces of the stressing system. The modulus of
elasticity determined in the bench test did not prove to be representative of most of the
tendons used in the structure, presumably because of slight variations in strand area. The
in-place friction test was much more useful for providing information to the engineers
and constructors. Accurate elongation calculations must be based on the results of an in-
place friction test, otherwise the measured elongation tolerance may not be easily met.

4. Wobble friction in the straight internal ducts of the structure was quite small. These
ducts were effectively held in position during concrete placement by inflatable mandrels.
Friction coefficients for internal tendons in ducts constructed using mandrels can
conservatively be taken as μ=0.16 and K=0.0007m-1. For draped internal ducts, friction
coefficients are μ=0.16 and K=0.0013m-1 for monolithic girders and K=0.0016m-1 for
segmental girders, based on other studies [19] [22].

5. The friction coefficient for external tendons in smoothly bent deviator pipes with
consistent radius can be taken as μ=0.25. The friction coefficient in the sharply bent
deviator pipes used in some of the U.S. 183 girders, at about a 2m radius, generated a
coefficient of friction of about μ=0.35. The sharp radius bend also caused large cracks in
the deviator concrete, and should be avoided.

6. The extra wobble angle β=0.04 radians suggested by Roberts [7] was found to be
sufficient when applied at each deviator of the mainlane girders, if the proper coefficient
of friction (0.35) was used in the calculation. The wobble angle β=0.04 radians was
recommended based on studies of girders constructed span-by-span with straight or large

158
 radius horizontal geometry. The β angles measured in Ramp P, with a horizontal
curvature of 221m, were higher at β=0.11 radians with an assumed friction coefficient of
μ=0.25. The horizontal curvature of the girder makes accurate deviator pipe placement
more difficult, thereby warranting a higher design β angle. The β angle should be
applied at all deviators and saddles. The use of a diabolo, or double trumpet bell shaped
deviator pipe, would help reduce the β angles on curved structures. The diabolo style
deviator pipe was not necessary for the mainlane girders because the total friction loss
was small for the one span tendons, and the bridge alignment was nearly straight.

7. Anchorage of the long 150m (3 span) external tendons in ramp P adjacent to a deviation
saddle in the anchor segment proved to be unacceptable. The large elongations caused
entangled tendons to be drawn close to the back of the anchor head where they broke.
Straight anchorage geometry would have allowed the 19 strand tendons to untangle to
some extent in the long distance between the anchorage and the deviator. No strand
breakages of this type occurred in any of the 14 tendons in each of 162 spans of mainlane
girder. The mainlane girders had straight anchorage geometry. If deviation saddles are
required adjacent to a live end anchorage where a first pull must be made, the length of
elongation should be limited to that of one span. Proper support of unstressed tendons
over their deviated length prior to stressing would help reduce the total elongation
substantially by reducing the slack length.

8. Elastic shortening loss calculations for the external deviated tendons were found to be
inaccurate if slippage was not assumed to occur at the deviators. The measured values
fell between the cases calculated using a deviator with infinite friction and zero friction
between the deviator pipe and the stressed tendon. The more conservative loss from
these two cases should be used for design.

9. Long term losses were found to be small when compared to other losses for the girders
under study. The segments were well aged before they were erected and prestressed.

159
160
 CHAPTER 4
THERMAL GRADIENTS AND THEIR EFFECTS

4.1 PROBLEM DESCRIPTION AND TECHNICAL BACKGROUND

Structures of any common building material undergo strain changes when the
temperature of all or part of the structure changes. Whether or not forces are generated in the
structure depends on the distribution of temperature change in the structure, the shape of the
structure, and the boundary constraints on the structure. In general, heat is conducted well in steel
structures, tending to cause more uniform temperature changes in the whole structure when one
part or surface is heated or cooled. Concrete structures, on the other hand, conduct heat poorly.
Concrete structures can maintain temperature gradients from uneven heating or cooling that are
larger and are maintained longer than in steel structures of similar proportion. Hybrid structures,
such as steel plate girder bridges with concrete decks, have characteristics of both concrete and
steel structures.

A thermal gradient is defined as a temperature distribution throughout the depth of a

cross section relative to the minimum temperature, measured or predicted, which is taken as the
zero reference point of the thermal gradient. When structures such as the U.S. 183 box girders are
designed, both an average temperature change in the structure, to reflect seasonal temperature
changes, and nonlinear design thermal gradient through the depth of the structure must be applied.
The thermal gradient design cases are used to predict stress changes in a girder that occur over a
relatively short period of time, perhaps daily. The numerical values for these design cases are
given in the AASHTO Guide Specification [23] or the AASHTO LRFD Guide Specification [9].
The design thermal gradients are discussed in Section 4.2.9 and 4.2.10. The accuracy of the
thermal gradient design calculation, in predicting what is actually occurring in the structure over
its lifetime, relies on both realistic design values and correct solution techniques. The
measurements taken using thermocouples on U.S. 183 were intended to check the current
superstructure design code values for thermal gradient design. Two different cross sections were
instrumented, one on the mainlane girder D5 and the other on ramp P. The strain measurements
taken in the structure were also compared to results calculated using common design methods.
There has been much concern by the bridge industry as to the accuracy of the current
recommended design thermal gradients for concrete box girders. Overly conservative design

161
thermal gradients can increase the cost of a bridge, mostly through unnecessary prestressing.
Thermal gradients can cause bending moments in continuous structures that are opposite in sign
or the same in sign as the bending moments from live loads, thus enlarging the entire moment
envelope.

Traditionally the bridge industry has ignored thermal gradient effects in bridge piers.
Substructure elements should respond to thermal gradients just as would a superstructure member.
Temperature and strain measurements were taken in mainlane pier D6 that is solid in cross
section, and in ramp pier P16 that is voided in cross section. The measured thermal gradients
were compared to those measured in the superstructure for both distribution and magnitude. The
importance of the stress changes from the measured thermal gradients was then evaluated, since a
pier is a bending and compression member and not just a bending member like the superstructure.

4.1.1 Superstructure behavior and analysis techniques

Thermal gradients in a box girder or other bridge girder are caused by changes in
weather conditions combined with daily heating and cooling. Conditions needed for the
development of a positive thermal gradient are shown in Figure 4.1(a). A positive thermal
gradient is defined to exist when the temperature of the box girder deck surface (Tdeck) is hotter
than the minimum average temperature at some depth in the section (Tmin). Tmin usually occurs
somewhere near the top of the girder webs, as seen in Figure 4.1(a). The sun primarily heats the
deck surface over the course of a day. The maximum possible magnitude of the positive thermal
gradient at deck level for a bridge girder occurs when a long period of cold and cloudy weather is
followed by a day of clear and sunny weather. The deck level magnitude of the positive thermal
gradient is larger when the total amount of solar radiation reaching the structure over the day is
increased. The location of the bridge determines the peak amount of solar radiation. Altitude,
latitude, and relative humidity are all important in determining the peak amount of solar radiation.
The magnitude of the gradient is decreased when heat is removed from the deck by wind or rain.

The maximum measured deck level magnitude of the positive thermal gradient is defined
as T1,meas, and is calculated from the measured temperatures Tdeck-Tmin, as shown in Figure 4.1(a).
Deck level temperatures (Tdeck) were taken 25mm below the actual deck level concrete surface.
The measured soffit level magnitude of the positive thermal gradient is defined as T4,meas, and is
calculated from the measured temperatures Tsoffit-Tmin, as shown in Figure 4.1(a). Since much of
the cross sectional area of a box girder is in the top flange, the deck level magnitude of the

162
positive thermal gradient has much greater influence on the structure’s response than the
magnitude of the thermal gradient at other depths in the cross section.

The conditions necessary for the formation of a negative thermal gradient are shown in
Figure 4.1(b). A negative thermal gradient exists when the deck surface of the girder (Tdeck) is
colder than the maximum average temperature in the webs (Tmax). Negative gradient conditions
of maximum negative deck level magnitude (T1,meas) occur when a relatively warm bridge girder
is cooled rapidly by cold rain on the deck. Both peak positive and negative deck level thermal
gradient magnitudes occur most frequently in the spring, when solar radiation intensity is high and
weather conditions change radically.

Solar
SUN
Radiation

Reflected Light Measured Measured

Clear Skies Radiation Temperature Thermal
Wind
Distribution Gradient
Tdeck

Tdeck-Tmin
T1,meas=
Re-Radiation Tmin

Tsoffit-Tmin
Tsoffit

T4,meas=
(a) Positive Thermal Gradient Conditions

Sudden Stormy Cold Front

Following a Long Period
of Warm Weather
Measured Measured
Precipitation Temperature Thermal
Distribution Gradient
Tdeck
T1,meas=
Tdeck-Tmax

Re-Radiation Tmax

Strong
Wind
Tsoffit
Tsoffit-Tmax
T4,meas=

(b) Negative Thermal Gradient Conditions

Figure 4.1 Climatic influences on superstructure thermal gradients

163
 The deck level magnitude of the thermal gradients is also dependent on the shape of the
girder cross-section. Figure 4.2 shows that an increase in the number of webs or a decrease in
web spacing will lead to more heat conduction from the deck into the webs, reducing T1,meas for
the positive thermal gradient case. In fact, measurements showed a decrease in deck level
temperature over the girder webs under positive thermal gradient conditions when compared to
other portions of the deck surface. The massiveness of the cross section will also have an effect
on the shape and deck level magnitude of the thermal gradients. Design thermal gradients are
only modified over the structural depth, and are assumed constant across the width of a girder.

Applied Heat from Solar Radiation

Thermal
Gradient

Heat
Conduction

Applied Heat from Solar Radiation

Thermal
Gradient

Heat
Greater distance
Conduction
for heat energy to
travel.

Figure 4.2 The effects of cross section shape on thermal gradient shape

Design of a girder with the random cross section in Figure 4.3 for a thermal gradient
usually begins by breaking the thermal gradient into the sum of three parts. The thermal gradient
in Figure 4.3 is divided into a uniform temperature, a linear temperature gradient that is zero at the
neutral axis of the girder, and a nonlinear temperature distribution that results in self-equilibrating
stresses in the cross section. The division of the applied thermal gradient into these three parts
greatly simplifies the analysis of the girder, if transverse plane sections in the girder are assumed
to remain plane under the thermal loading. Figure 4.4 shows the structural response of a simple
164
span bridge to the first two of the three components of a nonlinear thermal gradient. The figure
shows that the uniform temperature change results in only an expansion of the girder with no
forces generated in the girder or at the supports. Similarly, the linear gradient causes a curvature
change in the girder. The girder deflects into a circular arc with no resultant internal stresses and
no reactions at the supports. The analysis of the simple span for thermal effects is elementary,
except in the case of the nonlinear thermal gradient.

T1 = Tave + Tlinear + TNL

y T1 Tave Tlinear TNL

z Neutral
Axis
= + +

Cross-Section Thermal Uniform Linear Non-Linear

Gradient Temperature Temperature Temperature
Gradient Distribution

Figure 4.3 Components of a nonlinear thermal gradient

Contraction (-) Expansion (+)

Tα

= +

Linear Uniform Uniform

Temperature = Elongation + Curvature
Gradient Strains Strains
Strains Deformed
Shape

No Gradient

Zero Reactions at Supports

Figure 4.4 Effect of linear thermal gradient components on a statically determinate span

165
 Figure 4.5 shows the response of a three-span continuous structure under the same linear
thermal gradient loading as shown for the girder in Figure 4.4. In this case, the uniform
elongation from the uniform temperature change is accommodated by expansion bearings at three
of the four bearing locations. On the other hand, the girder cannot assume its new equilibrium
position in a circular shape because of the dead weight of the structure or restraint at the bearings.
The new equilibrium shape of the girder causes bending in the girder and reactions at the
bearings.

Three-Span Structure

L/3 L/3 L/3

δ δ

Support Reactions
Necessary to Make δ = 0
Unrestrained Deformation with Uniform Curvature

Mrestraint

(+)

Moments Due to Restraining Force at Center Supports

Actual Deformation
Figure 4.5 Effect of linear thermal gradient components on a statically
indeterminate bridge structure

166
 A girder with various boundary conditions subjected to a nonlinear thermal gradient can be
analyzed the same way any girder is analyzed for other loads. Figure 4.6 shows how the stresses
generated by the nonlinear thermal gradient are broken down into three components. Basically,
the fixed end forces for the member are calculated. Once these fixed end forces are known, the
member can be assembled into a structural analysis model. The thermal gradient stresses that do
not contribute to the fixed end forces are the self-equilibrating stresses, and are found by default.
The fixed end forces and self-equilibrating stresses can be calculated using the following
equations.

T1α y
Area, A
y Inertia, I
T(y)α z

Center of Gravity

Applied Thermal Cross-Section

Gradient Shape (Section A-A)
Strains
y
Compression (-) Tension (+)
A
M M
P x P

σ (y)=EαT(y)
Fixed End Fixed End
Reactions Reactions
Temperature-
A
Induced
Fully Restrained Stress
Member Distribution

Compression (-) Tension (+)

y

- - =

EαT(y) - P/A - My/I = σSE(y)

Self-Equilibrating
Stresses
Figure 4.6 Calculation of self-equilibrating stresses from a nonlinear thermal
gradient when plane sections remain plane
167
The axial fixed end force P is calculated by:

P= ∫
depth
EαT(y)b(y)dy (4-1)

The fixed end moment M is calculated by:

M= ∫
depth
EαT(y)b(y)ydy (4-2)

and the self-equilibrating stresses are calculated by:

σSE(y) = EαT(y) - P/A - My/I (4-3)

where:
y = distance measured perpendicular to the longitudinal axis at
the center of gravity of the cross section
T(y) = temperature at a depth y
b(y) = net section width at a depth y
E = modulus of elasticity
α = coefficient of thermal expansion
σSE(y) = self-equilibrating stress at a depth y
A = cross sectional area

I = moment of inertia
If the thermal gradient varies over the vertical and transverse axes of the member, as
shown in Figure 4.7, these equations simply change to:

P= ∫ ∫
depth width
EαT(z,y)dzdy (4-4)

The fixed end moment about the z-axis is calculated by:

Mz = ∫ ∫
depth width
EαT(z,y)zdzdy (4-5)

168
The fixed end moment about the y-axis is calculated by:

My = ∫ ∫
depth width
EαT(z,y)ydzdy (4-6)

and the self-equilibrating stresses are calculated by:

σSE(z,y) = EαT(z,y) - P/A - Mzy/Iz - Myz/Iy (4-7)

where:
y = vertical distance measured to the center of gravity
of the cross section
z = transverse distance measured to the center of
gravity of the cross section
T(z,y) = temperature at transverse distance z and depth y
σSE(z,y) = self-equilibrating stress at transverse distance z
and depth y
Iz = moment of inertia about z axis
Iy = moment of inertia about y axis

Thermal
T(z,y)
Gradient
(view from top) y

z
Thermal
Gradient
(view from side)
Cross-Section

Figure 4.7 Thermal gradient that varies across the

width and depth of a cross section

The magnitude of structural response of a bridge girder to thermal gradients is greatly

dependent on the peak magnitude of the thermal gradient at deck level. The deck of a box girder
contributes greatly to the moment of inertia of the cross section, so temperature induced stresses
integrated over the full deck width have great influence on the longitudinal stresses in the entire
cross section. Transverse stresses from thermal gradients through the deck, webs and bottom slab

169
must also be calculated. In this case the axial and bending stiffnesses of the slabs and webs are
dependent only on their thicknesses, thus simplifying the calculations and reducing their
sensitivity to nonlinear thermal gradients from that of a flanged section. The magnitude of the
measured thermal gradients through the webs and bottom slabs of the U.S. 183 box girders were
found to be moderate to small. The deck level magnitude of the thermal gradient and average
temperature change in the top slab between webs was found to be quite large. Therefore, the
thermal gradient design case is an important part of the box girder's transverse design. Figure 4.8
shows a typical model of a box girder used for transverse design for all load cases. The fixed end
forces from the design thermal gradients in the top slab, webs and bottom slab are calculated
using the same approach used for the longitudinal analysis. Bending moments induced in the top
slab are especially important since they must be combined with truck wheel loads. The moments
produced at the top of the webs from the temperature expansion of the top slab can also be large.

Transverse
Thermal TInterior
Girder Tambient
Gradients
TInterior < TAmbient (Outside air temperature)
Girder

Fixed end
restraining forces
are computed for
each member.

Two-Dimensional
Frame Model

The fixed end forces are

then applied to the frame
model and the behavior
under the transverse
gradients is determined.

Figure 4.8 Transverse analysis for thermal gradients

170
 Transversely plane sections of a box girder are usually assumed to remain plane under a
thermal load to greatly simplify the calculations. If warping of the cross section will occur, usual
design procedures allow removal of portions of the cross section, such as the wing tips, to account
for the reduced stiffness of the girder. The designer proceeds from this point assuming once again
that plane sections will remain plane. This assumption has proven to be a valid design method for
computing box girder response to external loads, dead loads and even post-tensioning loads.
When the fixed end forces are calculated for these known loads, equilibrium must be satisfied
regardless of any warping occurring in the cross section. Error in the calculation may occur after
the elements are assembled in the structural model, since the assumed stiffnesses of the elements
may be incorrect. For the thermal gradient load case, the fixed end forces for the member are
calculated assuming plane sections are going to remain plane. If this is not the case, the fixed end
forces will not be correct, and therefore the assumed loading on the structure will not be correct.

An example of warping in a cross section is shown in Figures 4.9 and 4.10. The applied
thermal gradient in the top slab is very large when compared to the rest of the box girder. If
continuity to the webs is released, the top slab will bow upward both longitudinally and
transversely. If the transverse curvature is ignored, since it will have little effect on the
longitudinal response of the girder for this example, the top slab would have the shape shown in
Figure 4.9. The forces required to achieve continuity with the webs and bottom slab result in the
final deformed shape shown at the bottom of Figure 4.9. The forces and stresses at the top slab to
web juncture are complex. Longitudinal stresses may vary considerably across the top slab, as
shown in Figure 4.10. The measured longitudinal stresses taken from the U.S. 183 box girders
were compared to the same stresses calculated using common design procedures to evaluate the
effects of warping.

171
 Applied Positive
Gradient

Resulting
Distributed
Forces to
Bring the Undeformed
Deformed Shape
Pieces Into
a Consistent
Shape

Side View
(View A-A)
A

Span Shown with Cut Between

Top Flange and Rest of Section

B B

Final Deformed Shape

Figure 4.9 Girder response to an applied positive thermal gradient when warping occurs

172
 Local Axial
Forces and
Bending
Moments
Over Webs
Resultant
Warped
Shape

Plan View of Top Flange (Section B-B)

Resultant Normal
Stress Pattern
Figure 4.10 Girder response to an applied positive thermal gradient when warping occurs

4.1.2 Substructure behavior and analysis techniques

Bridge piers are subjected to thermal gradients and average temperature changes. For
common bridges, the thermal gradient design load for substructures is ignored altogether, even
though a pier would respond to a thermal gradient much in the same way that a superstructure
girder would behave. For example, the flagpole type single pier in Figure 4.11 would behave
similarly to the simple span girder of Figure 4.4. The analysis procedure would be the same for
calculating the internally balanced stresses in the cross section. Figure 4.12 shows a single pier
subjected to a thermal gradient that will produce bending stresses both in itself and in the
monolithically connected superstructure. Figure 4.13 shows a two column bent subjected to a
thermal gradient. The thermal gradient case on the left of the figure produces only internally
balanced stresses, while the case on the right produces both bending stresses and internally
balanced stresses. A thermal gradient case that is capable of producing large internally balanced
stresses in massive solid piers is demonstrated in Figure 4.14. Temperature changes on the
surface of the pier produce stresses analogous to those produced by the mating of two materials
with dissimilar thermal expansion properties subjected to a uniform temperature change.

173
 Solar
Radiation

Gradient
No Gradient Applied

Gradient Shape

Column
Cross-Section

Fixed Base

Figure 4.11 Response of a single column to a thermal gradient

Solar
Radiation Rigidly-Connected
Superstructure

Gradient No Gradient
Applied

Gradient Shape

Column
Cross-Section

Non-Rigid Soil/Structure
Interface

Figure 4.12 Response of a single column bridge pier to a thermal gradient

174
 Gradient
Solar Applied Solar
Radiation Radiation
No Gradient

No Gradient Gradient Shape Gradient Gradient Shape

Applied
Cross-Section Cross-Section

Sunlight Striking Face of Bent Sunlight Striking Side of Bent

Figure 4.13 Response of a multiple column bridge pier to a thermal gradient

Shrunk-on Bronze Concrete “Shell”

Uniform ΔT ΔTcore ≠ ΔTshell

αsteel ≠ αbronze αcore = αshell

Steel Bolt Concrete “Core”

Figure 4.14 Response of solid column to thermal gradients

The shape and magnitude of the thermal gradient in a vertical pier will differ from that of
a superstructure member because of its varying exposure to the sun over the course of the day.
The largest magnitude of positive thermal gradient in a superstructure girder occurs when an
extended cloudy cool period is followed by a clear, calm and hot day with high solar radiation
intensity. These conditions would also produce a large magnitude thermal gradient in a bridge
pier, but the changing orientation of sun exposure would tend to reduce the magnitude of the
gradient from that produced with a single orientation. Figure 4.15 shows the thermal gradient
shapes that might be produced at different times of the day on the octagon piers of ramp P. The
heat from the sun exposure at sunrise and mid-morning on the east face of the pier dissipates into
the cross section by afternoon, effectively reducing the potential magnitude of the gradient at this

175
time. Radiation from the sun also strikes the pier at an acute angle during mid-afternoon, the time
when peak positive gradient magnitudes are produced in the superstructure.

Sun
Sun Sun

Sun
Sun

Sunrise Mid-Morning Noon Mid-Afternoon Sunset

S
Pier Cross-Section
Thermal
Gradient E W
Shape
N
Figure 4.15 Thermal gradient orientations at different times of the day

Analysis of a bridge pier for thermal gradients would proceed identically to that for the
superstructure. One major difference is that the design thermal gradient in the superstructure will
vary only over the depth of the girder and not the width. A design thermal gradient applied to a
pier will no doubt have its orientation governed by sun exposure, and not the pier's major or minor
bending axes. The response of a pier to a thermal gradient may also differ from that of a
superstructure box girder. Piers are generally of a compact cross sectional shape, so the
assumption that plane sections remain plane might very well be valid. Also, the peak magnitude
of the thermal gradient is applied at deck level and over the large width of the deck in a
superstructure box girder, resulting in an appropriately large response of the entire girder. The
compact shape of a bridge pier results in proportionately less of the cross section being subjected
to the peak magnitude of the gradient. The temperature measurements taken from the
instrumented piers on U.S. 183 allowed a direct comparison of the magnitudes and shapes of the
superstructure thermal gradients and the pier thermal gradients. The importance of the thermal
gradient design case depends on the magnitude of the thermal gradient selected, and magnitude of
the stresses produced in the pier when compared to the stresses from other load cases.

176
4.2 PREVIOUSLY MEASURED GRADIENTS AND DESIGN GRADIENTS
Several research studies have been conducted to determine the magnitude, shape, and
frequency of positive and negative thermal gradients in concrete bridges. Some studies were
conducted on actual bridges, others on laboratory constructed girders, and some studies were
analytical. The frequency and duration of readings for these studies varied considerably, as did
the number of thermocouples. The distance of the topmost thermocouple to the deck surface no
doubt affected the accuracy of some of the measured peak thermal gradient magnitudes, based on
an inspection of the gauge location drawings. Very little has been done to verify a structure's
response to thermal gradients.

4.2.1 Hoffman, McClure and West [32]

These researchers at Penn State constructed a full-scale segmental box girder with the
cross section shown in Figure 4.16. Thermocouples were placed primarily across the thicknesses
of the top slab, webs and bottom slab. Readings were taken eighteen times daily from October
25, 1978 to October 16, 1979. The peak positive and negative thermal gradient magnitudes are
plotted in Figure 4.16. The peak negative thermal gradient deck level magnitude is only a small
fraction of the magnitude of the peak positive gradient at deck level. The peak positive gradient
magnitude at deck level measured about 29oC, much higher than the 22.8oC recommended by the
AASHTO LRFD design specification for the Penn State area. The shape of the gradient may
actually vary considerably from that plotted because of the absence of any gauges between the
tops of the webs and the middle of the webs.

177
 6525mm

No Topping

Thermocouple

Maximum recorded negative gradient

January 4, 1979
Height above soffit, inches (mm)

60”
(1524)

40”
(1016)
Maximum recorded positive gradient
July 7, 1979
20”
(508)

-20o 0o 20o 40o 60o

(-11.1) (0) (11.1) (22.2) (33.3)
Temperature Difference, Fahrenheit (oC)
Figure 4.16 Thermocouple locations and maximum recorded thermal
gradients for the Hoffman, McClure and West study [32]

4.2.2 Hirst and Dilger [33]

Their study compared the results from an analytical model to the measured thermal
gradients from a light rail bridge in Canada and a box girder in Australia. The results from the
analytical model compared well with the measured results, thus verifying their method. The
maximum measured positive gradient magnitude at deck level was only 10oC.

4.2.3 Shui [34]

Temperature measurements were taken on three bridges in various parts of the country.
These were the Kishwaukee River Bridge in Illinois, the Denny Creek Bridge in Washington, and
178
the Linn Cove Viaduct in North Carolina. The measured thermal gradients were similar for all
with a maximum positive thermal gradient magnitude at deck level of 11oC and a maximum
negative gradient magnitude at deck level of -5.5oC.

4.2.4 Priestley [35]

This was another study where analytical results were compared to measurements. The
measurements were taken on a quarter scale box girder subjected to laboratory controlled ambient
temperature and radiation intensity. Once again, good correlation was found between the
analytical results and the measured results. The peak simulated positive thermal gradient was
32oC at deck level for use in New Zealand. The addition of blacktop reduced the peak thermal
gradient magnitude at deck level at the rate of 0.2oC per millimeter of blacktop.

4.2.5 Roberts, Breen, Kreger [7]

This study measured temperatures in two box girders on the San Antonio Y segmental
viaduct in Texas. The results of this study would be expected to compare directly with the
measurements taken on U.S. 183, since Austin and San Antonio are similar in climate, elevation
and latitude. The box girders under study were also similar in proportion to those on U.S. 183,
with long wings and massive fillet area, but were 356mm less in depth at 1778mm. A fine
distribution of thermocouples, seen in Figure 4.17, was placed down the center of the webs and
readings were taken hourly for several years, including one year without any blacktop. The
maximum positive thermal gradient magnitudes measured at deck level were 12.2oC without
blacktop and 15.6oC with 50mm of blacktop. The maximum measured magnitudes of negative
thermal gradients at deck level were -8.3oC without blacktop and -4.4oC with 50mm of blacktop.
The first year of data without blacktop most likely did not include the peak positive gradient
possible for this area, judging by the variability of the data from U.S. 183 over several years.

179
 3’-0” 8”-0” 12”
1” 1” 1”
6 1/2” 3” 3” 3”
6 1/2” 8” 1”
1”
12” Thermocouple
5’-10”
12”
12”
1’-2” 1”
7” 3” 3”
1” 1”
Type I Segment Thermocouple Layout
Segments 11C-5 and 11C-10

12’-0” 15’-0” 12”

1” 1” 1”
8 1/2” 3” 4”
8” 4”
8 1/2” 1”
12”
1”
5’-10”

12” Thermocouple

12”
12” 1”
9” 4” 4”
1” 1”
Type III Segment Thermocouple Layout
Segments 44A-6 and 44A-15
Figure 4.17 Thermocouple layout for the San Antonio Y study by Roberts [7]

4.2.6 Arockiasamy and Reddy [36]

An analytical study using a program called FETAB (Finite Element Temperature
Analysis of Bridges) was used to predict temperatures measured in two bridges. Thermocouples
were retrofitted into each structure. The maximum recorded positive thermal gradient magnitude
at deck level was 25.6oC. Temperature variations were recorded transversely across the deck.

4.2.7 Pentas, Avent, Gopu, and Rebello [37]

A precast pretensioned girder bridge was instrumented in Louisiana with the cross
section and gauge locations shown in Figure 4.18. Measurements were taken approximately once
a month for two years. The maximum positive thermal gradient magnitude at deck level was
about 13oC, and the maximum negative thermal gradient magnitude at deck level was about -4oC.

180
These measurements do not compare directly to those of a box girder because the I-girders were
exposed to ambient air on all surfaces.

Expansion Joint 2 Thermocouple Locations

Expansion Joint 3

Expansion Joint 4
D C
B A

Array of Gauges E W
Thermocouple locations

120 Ambient Temp.

Girder Temp.
Slab Temp.
100
Temperature (oF)

Max.
80

60 Min.

40

20
10/22 12/16 2/21 4/15 6/10 8/25 12/1 4/19
1987 1988 1989
Date of 24 Hour Continuous Temperature Measurements

Figure 4.18 Temperature distributions through the depth of the section for the US 190
Atchafalaya River Bridge study by Pentas et al. [37]
181
4.2.8 Potgieter and Gamble [38]
Data from 26 weather stations from around the country was used as input in an analytical
model to predict thermal gradients in concrete bridges and the frequency of peak magnitude
thermal gradients. Two days of field measurements were taken on the Kishwaukee River Bridge
in Illinois to verify the results of the model. The results of this study were used to develop the
design recommendations in the NCHRP 276 report. The researchers recommended further field
studies be performed to verify the model for all areas of the country.

4.2.9 NCHRP 276 [39]

This report gave an overview of thermal effects in concrete. The recommendations for
design thermal gradients given in the NCHRP report were adopted by AASHTO to be used to
design the nation's concrete bridges. The positive design thermal gradients were taken from the
results of the Potgieter and Gamble study [39] and vary with location and thickness of blacktop.
The various thermal gradient zones are shown in Figure 4.19. The recommended shape and
magnitude for the positive design thermal gradients for box girders are shown in Figure 4.20.
Temperatures T2 and T3 should be reduced by 2.8oC for beam and slab bridges. The shape and
values of the negative thermal gradients shown in Figure 4.21 were based on the gradient
specified in the British Standard BS 5400 [40]. There was no substantial analytical or measured
basis for the specified negative gradients or their shapes. They were chosen based on engineering
judgment. It is not clear why the temperatures in the webs and bottom slab are functions of the
amount of blacktop.

182
 3 2

3
1 2
4 2

3
Figure 4.19 Proposed maximum solar radiation zones by Imbsen, et al. [39]
0.2m 0.1m

T1
Zone 1
1.0m or (d - 0.2)m

T2 mm Temperature, oCelsius
of
Blacktop T1 T2 T3
T3
0 30 11 6
d > 0.61m

D3

50 24 11 5
100 17 8 4

Zone 2
mm Temperature, oCelsius
0.2m

of
Blacktop T1 T2 T3
2.8 o C
0 26 9 5

50 20 9 5
Gradients are for box girders 100 14 8 4

Zone 4 Zone 3
mm Temperature, oCelsius mm Temperature, oCelsius
of of
Blacktop T1 T2 T3 Blacktop T1 T2 T3
0 21 8 4 0 23 9 5

50 16 8 4 50 18 9 4
100 12 9 4 100 13 9 4

Figure 4.20 Recommended positive vertical temperature gradient by Imbsen, et al. [39]

183
 T1 Zone 1

0.2m 0.1m
mm Temperature, oCelsius
of
T2 T1 T2 T3 T4
Blacktop

0.45 d T3 0 15.0 3.9 1.1 7.8

50 12.2 3.9 1.1 8.3
d >0.6m

100 8.9 2.8 0.6 6.7

Zone 2
T3
0.45 d

mm Temperature, oCelsius
0.1m 0.2m

of
Blacktop T1 T2 T3 T4
T2
0 12.8 3.3 1.1 5.6
T4
50 10.0 3.3 1.1 6.1

100 7.2 2.8 0.6 5.0

Zone 4 Zone 3
mm Temperature, oCelsius mm Temperature, oCelsius
of of
Blacktop T1 T2 T3 T4 Blacktop T1 T2 T3 T4
0 10.6 2.8 1.1 3.3 0 11.7 3.3 1.1 4.4

50 8.3 2.8 0.6 4.4 50 9.4 3.3 1.1 5.6

100 6.1 3.3 0.6 4.4 100 6.7 3.3 0.6 4.4

Figure 4.21 Recommended negative vertical temperature gradient by Imbsen, et al [39]

4.2.10 AASHTO LRFD [9]

The design thermal gradients adopted for the AASHTO LRFD design specifications
were also based on the recommendations of NCHRP 276, with some simplifications. The
AASHTO LRFD thermal gradient values are given in Figure 4.22. The upper part of the design
gradient shape is composed of two lines. The bottom portion of the thermal gradient has a
magnitude T3=0, unless determined otherwise by a site-specific study, but need not exceed 2.8oC
for the positive thermal gradient case. A full-scale research project would have to be done prior
to the design of the bridge for an accurate estimate of T3. The negative thermal gradients are
derived by multiplying the applicable positive gradient by -0.5. This approximation reduces the
amount of calculation involved for computing stresses almost in half, depending on the method
used. The dimension A varies with the depth of the superstructure. A is equal to 300mm for

184
concrete superstructures that are 400mm in depth or greater. A is equal to 100mm for concrete
superstructures that are less than 400mm in depth.

3 2

3
1 2
4 2

T1
100mm

Plain Concrete 50 mm Asphalt 100 mm Asphalt

T2 Surface (2” Asphalt) (4” Asphalt)
t
Superstructure

Zone T1 oC T2 oC T1 oC T2 oC T1 oC T2 oC
A
Depth

1 30.0 7.8 23.9 7.8 17.2 5.0

2 25.6 6.7 20.0 6.7 13.9 5.6

Steel 3 22.8 6.1 18.3 6.1 12.8 6.1

200mm

girder
structures 4 21.1 5.0 16.1 5.0 12.2 6.1
only
T3

Figure 4.22 Positive vertical temperature gradient in concrete and steel superstructures
(Figure 3.12.3-2 and Table 3.12.3-1 from the AASHTO LRFD Bridge Design
Specifications [9])
4.3 DATA COLLECTION AND ANALYSIS

4.3.1 Ramp P superstructure

Temperature measurements were taken hourly in one cross section of the ramp P
superstructure with a dense grid of thermocouples. Strain measurements were also taken at three
sections in the girder at the same time intervals. The thermocouples used to evaluate thermal
gradient magnitude are shown in Figure 4.23. The six thermocouples located at the top of the

185
webs were usually at the lowest temperature on average of any horizontal plane of thermocouples.
This was particularly true at the times peak positive gradients were occurring, and when negative
gradients of any importance were occurring. The positive gradient shape shown in Figure 4.23
was typical for positive gradients. T1,meas is the measured deck level magnitude of the thermal
gradient. T1,meas is the average temperature measured with the top thermocouples minus the
average temperature (Tmin) measured by the six baseline thermocouples. Similarly, T4,meas is the
soffit level gradient occurring at the same time as T1,meas.

T1,meas
Top Thermocouples

Typical Positive
Middle “Baseline” Gradient Shape
Thermocouples

Bottom Thermocouples T4,meas

Figure 4.23 Thermocouples used to calculate thermal gradient magnitudes on

the Ramp P superstructure

The deck level thermal gradient magnitude, soffit level thermal gradient magnitude, and
the ambient air temperature beneath the box girder are plotted in Figure 4.24 for the month of
March 1997. One deck level positive gradient magnitude of interest during this month was T1,meas
created when hot asphalt was applied to the deck. This gradient magnitude was the largest deck
level positive gradient magnitude measured for the ramp P girder, but should be considered a
construction load and not a normal service load. The maximum measured positive T1,meas on ramp
P under normal conditions occurred after the asphalt blacktop was in place on March 20, 1997. A
full year of data for this girder without blacktop could not be measured because of the
construction schedule. The maximum possible positive T1,meas for this girder without blacktop
most probably was not measured. Maximum T1,meas on March 20, 1997 followed a period of cool
weather. The maximum T1,meas for the days following this cool period were all similar in
magnitude, and persisted until the onset of the next cool period. In general, the spring of 1997 did
not have conditions favorable for producing maximum positive thermal gradients. The positive
thermal gradients measured in the mainlane superstructure for the springs of 1995 and 1996 were
substantially higher in magnitude than those measured in the spring of 1997. Persistent mild,

186
cloudy and rainy weather was to blame. The T1,meas values for the entire study period are
summarized in the Appendix.

35
30
Temperature (oC)

25
20
15
10
5
0
Ambient
Asphalt Applied Maximum T1,meas
to Deck

20
15
10
5
0
-5
Temperature (oC)

Minimum T1,meas
-10
Deck Level Thermal Gradient Magnitude, T1,meas

Soffit Level Thermal Gradient Magnitude, T4,meas

10
5
0
-5
-10
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
March, 1997
Figure 4.24 Measured thermal gradient magnitudes on Ramp P for the month of March 1997

The measured thermal gradient of March 20, 1997 is tabulated for each thermocouple
location in Figure 4.25. T1,meas values plotted over the width of the top slab show that the webs
acted to draw heat from the top of the cross section, thus a line of thermocouples located up the
centerline of a web would not have measured the peak deck level temperatures. Both the positive
design thermal gradient shape and magnitude recommended by the AASHTO LRFD design

187
specification, shown in the plot at the right in Figure 4.25, do not accurately represent the positive
thermal gradient measured in this girder. The massiveness of the top flange to web fillets may
have had some effect on this. The maximum measured negative thermal gradient from March 6,
1997 is presented in Figure 4.26. Once again, the recommended negative design thermal gradient,
which was patterned after the shape of the positive design gradient, poorly represented the
measured negative gradient in shape and magnitude. Substantial negative thermal gradients occur
when there are extreme changes in weather over very short periods of time, so the gradient shape
is not entirely predictable. Maximum positive thermal gradients occur when a relatively stable,
cool and cloudy period is followed by a day of bright sunshine and resultant warmer weather.
The positive thermal gradient develops from a cross section of nearly uniform temperature, and
therefore the shape is much more predictable.
Temperature (oC)

16
12 T1,meas
8
T1,meas = 13.8 oC
4
0
Positive Gradient Design
March 20, 1997 Positive
Thermocouple
4:30 pm Measured
Thermal
Positive
Gradient
Thermal
Gradient
0 4 8 12 16 20
Temperature (oC)
T802 T805 T808 T810 T813 T816 T819 T821 T824
13.4 15.8 13.1 15.5 14.6 15.3 10.8 10.9 14.7
T801 T803 T806 T811 T814 T817 T822 T825 T827
3.1 5.7 4.7 3.4 8.9 8.6 3.8 6.1 4.7
T804 T807 T809 T812 T815 T818 T820 T823 T826
4.8 3.9 3.2 5.9 6.0 5.2 1.9 3.4 5.1
T854 T853 T852 T828 T829 T830
1.7 -0.1 -0.9 -1.5 -0.6 1.5
T851 T850 T849 T831 T832 T833
2.3 0.6 -0.5 -1.5 -0.4 1.7
T848 T847 T846 T834 T835 T836
3.6 1.5 -0.2 -1.3 -0.1 3.1
T843 T840 T837
-0.6 -1.4 0.1
T844 T841 T838 Temperatures for thermocouples
Thermocouple 0.3 0.2 0.5
Temperature T845 T842 T839 are presented in degrees Celsius.
3.4 2.9 3.4

Figure 4.25 The maximum measured positive gradient on Ramp P (from March 20, 1997)

188
 Temperature (oC)
T1,meas
0
-2
-4
-6 T1,meas = -7.6 oC
-8
-10

Negative Gradient Design

March 6, 1997 Negative
Thermocouple Measured Thermal
7:30 am
Negative Gradient
Thermal
Gradient
-12 -8 -4 0
Temperature (oC)
T802 T805 T808 T810 T813 T816 T819 T821 T824
-8.8 -8.3 -6.2 -7.8 -8.4 -8.2 -5.7 -6.5 -8.8
T801 T803 T806 T811 T814 T817 T822 T825 T827
-6.4 -6.2 -4.1 -4.5 -5.7 -5.0 -3.4 -5.7 -6.7
T804 T807 T809 T812 T815 T818 T820 T823 T826
-6.5 -4.7 -1.3 -3.2 -4.3 -3.2 -1.2 -7.3 -9.4
T854 T853 T852 T828 T829 T830
-1.0 0.0 1.2 1.2 0.3 -1.6
T851 T850 T849 T831 T832 T833
-2.1 -1.6 -0.3 0.0 -1.1 -2.8
T848 T847 T846 T834 T835 T836
-3.6 -2.9 -1.3 -0.9 -1.9 -4.3
T843 T840 T837
-1.3 -0.6 -0.8
Thermocouple T844 T841 T838 Temperatures for thermocouples
-2.2 -1.8 -2.4
Temperature T845 T842 are presented in degrees Celsius.
T839
-4.4 -3.8 -4.5

Figure 4.26 The maximum measured negative gradient on Ramp P (from March 6, 1997)

The statistical occurrence of peak positive and negative thermal gradient deck level
magnitudes is of primary concern when selecting gradient magnitudes for design. Figure 4.27
shows the occurrence of maximum daily T1,meas values from November 1996 to March 1997, when
no blacktop was in place. This time range did not allow a complete or even realistic distribution
of gradients. Figure 4.28 presents the occurrence of maximum daily T1,meas values from March
1997 through February 1998, with 50mm of blacktop in place. This graph reveals an entirely
different distribution than that in Figure 4.27, and is more representative of an actual year. These
figures show that the peak magnitude T1,meas values occurred on small percentage of days over the
course of a year, but T1,meas values only two or three degrees smaller than the peak occurred
frequently.

189
 25 %
T1,meas (Daily maximum)
Occurrence Ratio

20 % Before asphalt overlay

Database contains 98 days:
15 %
November 1996 through
April 1997.
10 %

5%

0%
<0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Temperature Ranges (oC)
Figure 4.27 Statistical occurrence of daily maximum T1,meas values on Ramp
P before application of the asphalt blacktop

25 %
T1,meas
Occurrence Ratio

20 %
After 50 mm (2”) asphalt overlay
15 % Database contains 270 days:
March 1997 through February 1998.
10 %

5%

0%
<0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Temperature Ranges (oC)

Figure 4.28 Statistical occurrence of daily maximum T1,meas values on Ramp P

after application of the asphalt blacktop

Figures 4.29 and 4.30 give the occurrence distribution of peak daily minimum T1,meas
values. Figure 4.29 demonstrates once again that the daily peak minimum deck level thermal
gradient magnitude occurred only a small percentage of days. The distribution of daily minimum
T1,meas values, measured without the blacktop in place, differed from the distribution of daily
maximum T1,meas values in that minimum T1,meas values of half the magnitude of the peak value
occurred most of the time. Figure 4.30 shows that the magnitude of peak daily minimum deck
level gradient, with the 50mm of blacktop in place, was about half that without the blacktop.

190
 25 %
Occurrence Ratio

T1,meas
20 %
Before asphalt overlay
Database contains 98 days:
15 % November 1996 through April 1997.

10 %

5%

0%
>0 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15
Temperature Ranges (oC)
Figure 4.29 Statistical occurrence of daily minimum T1,meas values on Ramp P before
application of the asphalt blacktop

40 %

35 %

30 %
Occurrence Ratio

T1, meas
25 %
After 50 mm (2”) asphalt overlay
Database contains 270 days:
20 %
March 1997 through February 1998.

15 %

10 %

5%

0%
>0 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15
Temperature Ranges (oC)

Figure 4.30 Statistical occurrence of daily minimum T1,meas values on Ramp P

after application of the asphalt blacktop

The primary reason the design thermal gradients have been established is for the
calculation of cross sectional stresses and transverse stresses. Strain gauges were placed both
longitudinally and transversely at three cross sections in the five span continuous ramp P girder.
These cross sections were all in one half of span P16. The first cross section was in segment
191
P16-2 near the face of the anchor segment diaphragm over pier P16. The second cross section
was located at the quarter point of the span. The third cross section was located at midspan,
which was the same location as the plane of thermocouples. The measured strains converted to
stresses are plotted against calculated stresses in the figures to follow. The gauges used
automatically subtracted out the strain that occurred from unrestrained thermal expansion, leaving
only a strain caused by stresses. Stresses have been calculated using both the measured maximum
gradients and the gradients recommended by AASHTO LRFD. The measured maximum
gradients were taken as the change in temperature at each thermocouple over a certain period of
time, and were similar to the maximum positive and negative gradients presented in Figures 4.25
and 4.26. The measured temperature changes were distributed over the cross sections as shown in
Figure 4.31, each area taking the temperature change of one thermocouple. The design method
used was the same one recommended in the AASHTO LRFD and presented in section 4.1.1.
Plane sections were assumed to remain plane in this analysis. Since the structure was continuous,
the longitudinal concrete stresses were the sum of bending stresses and self-equilibrating stresses.

Neutral Axis
Thermocouple

Figure 4.31 Division of the Ramp P cross-section into tributary areas for each
thermocouple gauge

The measured and calculated stresses for segment P16-2 for the maximum positive gradient
load case are plotted in Figure 4.32. The plot of top fiber stresses shows good correlation between
the measured and the calculated values, except at the wing tips. The design thermal gradient
stresses were uniform across the cross section, since the gradient did not change over the width of
the girder and plane sections were assumed to remain plane. The stresses calculated using the
measured thermal gradient were distributed transversely in proportion to the measured
temperature change at each gauge. The measured stresses definitely revealed a reduction in
sectional stiffness at the wing tips. The measured stresses were almost constant between webs in
the top flange. Very little strain change was measured in the lower portion of the webs and the
192
bottom flange. The location of the web strain gauges near the exterior of the girder resulted in a
strain measurement higher in compression than the average across the web width. The
o
temperature differential across the thickness of the webs was about 4 C. The stress distribution
from thermal stresses was not as readily interpreted at this section for thermal gradients as it was
for other load cases because of the proximity of the web gauges to the exterior of the concrete.
(+ compression, - tension)

10
Stress (MPa)

-5
(+ compression, - tension)

10
Stress (MPa)

Positive Gradient
-5 March 20, 1997
4:30 pm

Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
Gauge
10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.32 Comparison of measured and calculated positive thermal gradient stresses
for segment P16-2 (near diaphragm)
193
 Figure 4.33 shows that the distribution of measured stresses in segment P16-10 at the
quarter point of the span was not predicted well by the standard calculation procedure, even on
average. The measured stresses were low at the wing tips, as was seen in segment P16-2, and
peaked over the webs in the top flange. The distribution of measured stresses in the lower portion
of the webs and bottom flange was also similar to that in segment P16-2. The measured and
predicted stresses for the positive thermal gradient case for segment P16-17 near midspan are
plotted in Figure 4.34. The distribution of measured stresses in this segment was very similar to
that of segment P16-10. The heavy end diaphragm located 610mm from the gauges in segment
P16-2 may have been responsible for a reduction in warping at this section and a smoother
distribution of stresses across the top flange. Also, the deviator beam in segment P16-10 may
have contributed to the peaks in stress over the webs, which were somewhat larger than those in
segment P16-17 that had no deviator beam.

194
 (+ compression, - tension)
15

10
Stress (MPa)

-5

(+ compression, - tension)

10
Stress (MPa)

-5 Positive Gradient
March 20, 1997
4:30 pm
Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
15 10 5 0 -5 Gauge
Stress (MPa)
(+ compression, - tension)

Figure 4.33 Comparison of measured and calculated positive thermal gradient stresses for
segment P16-10 (quarter point)

195
 (+ compression, - tension)
Stress (MPa) 10

-5

(+ compression, - tension)

10
Stress (MPa)

-5 Positive Gradient
March 20, 1997
4:30 pm
Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
10 5 0 -5 Gauge
Stress (MPa)
(+ compression, - tension)

Figure 4.34 Comparison of measured and calculated positive thermal gradient stresses for
segment P16-17 (midspan)

The response of the girder to the maximum negative thermal gradient case was generally
similar to the response to the positive gradient, but opposite in direction. The measured and
predicted stresses for the negative gradient case are plotted in Figures 4.35 through 4.37 for the
three-instrumented sections. The average top flange stress in segment P16-2 was predicted

196
reasonably well by the calculated method using the measured gradient, as was the bottom flange
stress. The measured stresses and the calculated stresses did not compare well in the webs
because of the negative gradient through the thickness of the web. The measured stresses over the
webs in the top flange were irregular in segments P16-10 and P16-17, possibly attributable to
plane sections not remaining plane, reduction of thermal forces by warping, inelastic behavior in
the congested web area, or forces from the deviator beam.

197
 (+ compression, - tension)
Stress (MPa) 10

-5

(+ compression, - tension)

10
Stress (MPa)

-5 Negative Gradient
March 6, 1997
7:30 am
Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
10 5 0 -5 Gauge
Stress (MPa)
(+ compression, - tension)

Figure 4.35 Comparison of measured and calculated negative thermal gradient stresses
for segment P16-2 (near diaphragm)

198
 (+ compression, - tension)
0
Stress (MPa)

-5

-10

-15

(+ compression, - tension)
Stress (MPa)

10

-5 Negative Gradient
March 6, 1997
7:30 am
Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
10 5 0 -5 Gauge
Stress (MPa)
(+ compression, - tension)

Figure 4.36 Comparison of measured and calculated negative thermal gradient stresses
for segment P16-10 (quarter point)

199
 (+ compression, - tension)
10

5
Stress (MPa)

-5

(+ compression, - tension)

10
Stress (MPa)

-5 Negative Gradient
March 6, 1997
7:30 am
Measured
Gradient Stress
Calculated
Gradient Stress
Design
Gradient Stress

Concrete Strain
10 5 0 -5 Gauge
Stress (MPa)
(+ compression, - tension)

Figure 4.37 Comparison of measured and calculated negative thermal gradient stresses
for segment P16-17 (midspan)

The magnitudes of the measured thermal gradients in ramp P were lower than the AASHTO
LRFD recommended gradient magnitudes, but the peak stresses at points in the section were
marginally higher to much higher. Figure 4.38 shows the service load stresses from combinations
of dead load, prestressing, live loads and the negative gradient case. The plots show both the
measured stresses and the stresses predicted by normal calculation methods using the AASHTO
200
LRFD design thermal gradients. The measured stresses near the diaphragm in segment P16-2 all
fell below the stresses predicted by calculation. These stresses also did not approach the 0.45f 'c
service level stress limit. The measured stress at one point in quarter point segment P16-10
exceeded the service level stress limit by a small amount, but did not approach the inelastic limit
of the concrete at 0.7f 'c. Stresses in midspan segment P16-17 all fell well below the service level
stress limit. The load case at midspan did not include the live load case because the stresses
would have opposed those caused by the negative thermal gradient.

201
 P16-2 (Near diaphragm): Dead Load + Prestressing + Live Load Case 1 + Negative Gradient
45
(+ compression, - tension) 40 f’c = 44.8 MPa
Inelastic Range
35 of Concrete
Stress (MPa)

30 0.7f’c = 31.4 MPa

25
20
0.45f’c = 20.2 MPa
15
10
5
0

P16-10 (Quarter point): Dead Load + Prestressing + Negative Thermal Gradient

45
(+ compression, - tension)

40 f’c = 44.8 MPa

Inelastic Range
35 of Concrete
Stress (MPa)

30 0.7f’c = 31.4 MPa

25
20
0.45f’c = 20.2 MPa
15
10
5
0

P16-17 (Midspan): Dead Load + Prestressing + Negative Thermal Gradient

45
(+ compression, - tension)

40 f’c = 44.8 MPa Inelastic Range

35 of Concrete
Stress (MPa)

30 0.7f’c = 31.4 MPa

25
20
0.45f’c = 20.2 MPa
15
10
5
0

Measured Stress
Calculated Stress
Concrete Strain Gauge

Figure 4.38 Minimum top flange stress load combinations for P16-2, P16-10 and P16-17
202
 The stresses are plotted for various load combinations including the positive thermal
gradient case in Figure 4.39. Since the maximum measured positive thermal gradient magnitude
at deck level was closer to the design value than was the maximum negative thermal gradient
magnitude, the measured stresses shown in Figure 4.39 exceeded the service level stress limit in
more instances. The measured stresses projected well into the inelastic range in quarter point
segment P16-10. No signs of distress in the concrete were noticed in this area prior to the
application of the blacktop, and none were noticed from the inside of the girder.

203
 P16-2 (Near diaphragm): Dead Load + Prestressing + Positive Gradient
45
(+ compression, - tension) 40 f’c = 44.8 MPa
Inelastic Range
35 of Concrete
Stress (MPa)

30 0.7f’c = 31.4 MPa

25
20
15 0.45f’c = 20.2 MPa
10
5
0

P16-10 (Quarter point): Dead Load + Prestressing + Live Load Case 1 + Positive Gradient
45
(+ compression, - tension)

40 f’c = 44.8 MPa

Inelastic Range
35 of Concrete
Stress (MPa)

30 0.7f’c = 31.4 MPa

25
20
0.45f’c = 20.2 MPa
15
10
5
0

P16-17 (Midspan): Dead Load + Prestressing + Live Load Case 1 + Positive Gradient
(+ compression, - tension)

45
40 f’c = 44.8 MPa Inelastic Range
35 of Concrete
Stress (MPa)

30 70% f’c = 31.4 MPa

25
20
0.45f’c = 20.2 MPa
15
10
5
0

Measured Stress
Calculated Stress
Concrete Strain Gauge

Figure 4.39 Maximum top flange stress load combinations for P16-2, P16-10 and P16-17
204
 The transverse strains in the cross section were measured with strain gauges near each
face of the top flange, bottom flange, and webs. The strains converted to stresses are plotted for
midspan segment P16-17 in Figure 4.40 for the peak positive and negative thermal gradient cases.
The measured stresses were compared to stresses calculated using a model of the cross section
such as the one shown in Figure 4.8. The measured stresses did not compare well with the
calculated stresses. For one, the calculated stresses assumed that no stresses would be created in
the transverse direction from the thermal gradient response in the longitudinal direction. The
primary difference between the measured and the calculated stresses was that the measured
stresses seemed to contain an additional amount of uniform compression or tension, depending on
the location. The calculated stresses were dominated by bending stresses.

Positive Gradient (March 20, 1997 4:30 pm)

3.68 7.13

-3.16 -0.42
-1.04
-0.34

1.25
-0.08 -1.41 1.34

1.62 1.18

Negative Gradient (March 6, 1997 7:30 am)

-4.78 -1.68

-0.92 1.43
Measured
-1.68
Transverse Stress (Mpa) 0.00
Calculated -0.96
0.83
Transverse Stress (Mpa)
1.12 -0.34
(values in italics)
+ Compression
-1.24 -1.59
- Tension

Figure 4.40 Comparison of measured and calculated transverse flexural stresses from positive
and negative thermal gradients for P16-17

205
 The strain changes in the external prestressing tendons were measured during the
maximum positive and negative thermal gradient cases. Figure 4.41 shows that the stress change
in any of the tendons during the maximum positive gradient day was less than 11MPa, or only
0.6% of the guaranteed ultimate tensile strength of the tendon. Figure 4.42 shows that the stress
change in the tendons from the maximum negative thermal gradient was somewhat smaller at
9.5MPa. Fatigue from daily thermal stress cycles should not be a problem. Ryals [41]
recommended a fatigue limit stress of 69MPa for external tendons based on traffic induced stress
cycles.

P16 Interior Positive Thermal Gradient

Anchorage Diaphragm Upstation (March 20, 1997 4:30 p.m.)

Vertical Horizontal
Deviator Beam Deviator Beam

Center-line of Span
12.0
(+ compression, - tension)

8.0
Temperature Change
Stress (MPa)

4.0
Within Interior of Box: 0.0 oC
0.0
-4.0
-8.0
-12.0
-16.0
T1 (Right) T2 (Right) T3 (Right)
T1 (Left) T2 (Left) T3 (Left)

Figure 4.41 Measured stress changes in the Ramp P external tendons from the maximum
positive gradient

206
 P16 Interior Negative Thermal Gradient
Upstation (March 6, 1997 7:30 am)
Anchorage Diaphragm

Vertical Horizontal
Deviator Beam Deviator Beam

Center-line of Span
12.0
(+ compression, - tension)

8.0
Stress (MPa)

4.0
0.0
-4.0 Temperature Change
Within Interior of Box: -5.0 oC
-8.0
-12.0 T1 (Right) T2 (Right) T3 (Right)
T1 (Left) T2 (Left) T3 (Left)
-16.0

Figure 4.42 Measured stress changes in the Ramp P external tendons from the maximum
negative gradient

4.3.2 Mainlane superstructure

Temperatures were measured in the mainlane girder D5 hourly with the thermocouples
distributed as shown in Figure 4.43. T1,meas and T4,meas were calculated with respect to the average
temperature of the baseline thermocouples in the web. These baseline thermocouples located at
the bottom of the webs recorded lower temperatures on average (Tmin) during periods of peak
positive gradients than the thermocouples located at the top of the webs. The baseline
thermocouples at the top of the webs recorded higher temperatures (Tmax) than those at the bottom
of the webs by about 2oC during periods of peak negative gradient. This behavior was different
than that of the ramp P cross section, and was probably due to the huge mass of concrete at the
root of the wings and the top of web fillet area. Heat absorbed through the top slab was retained
in the fillet area, unable to cool completely by the end of the night. Concrete surfaces on the top
interior of the box girder continued to increase in temperature until after midnight during periods
of high solar intensity. The typical positive gradient shape, shown in Figure 4.43, was similar to
that of ramp P. The shape differed over that portion of the girder depth between the bottom of the
top flange between webs and the bottom of the root of the wing. The shape of the negative

207
gradients of any significant magnitude was irregular, and was dependent on the gradient present in
the cross section prior to the weather event that caused the high magnitude negative gradient.

Deck Level
Thermocouples
T1,meas

Negative Gradient
Baseline for Tmax
Typical Positive
Positive Gradient Gradient Shape
Baseline for Tmin

Soffit
T4,meas
Thermocouples
Figure 4.43 Thermocouples used to calculate thermal gradient magnitudes on the
mainlane girder D5

T1,meas, T4,meas and the ambient air temperature are plotted in Figure 4.44 for the month of
June 1996. The maximum magnitude of the positive gradient at deck level occurred on June 17th
after several days of stable weather. The maximum daily T1,meas values for that entire week were
nearly of the same magnitude. The daily ambient temperature variation was moderate at 10oC,
but the solar radiation was at its strongest of the year. T1,meas values of nearly this magnitude were
recorded in 1995 with no blacktop in place, and in other months in 1996. The spring of 1997 did
not produce T1,meas values of significant magnitude during positive thermal gradient conditions.
The T1,meas values recorded over the duration of the entire study are summarized in the Appendix.

208
 Ambient Air Temperature
40
35
Temperature (oC)

30
25
20
15
10
5
Maximum T1,meas recorded
June 17, 1996 at 15.39oC

T1,meas
20
15
Temperature (oC)

10

5
0

-5
-10

T4,meas
Temperature (oC)

10
5

0
-5
-10
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
June 1996

Figure 4.44 Measured thermal gradients on mainlane girder D5 for the month of June 1996

209
 The maximum magnitude positive thermal gradient from June 17, 1996 is shown in
Figure 4.45. Both the magnitude and the shape of the measured positive gradient differed from
that of the design thermal gradient from AASHTO LRFD, shown in the plot at the right in Figure
4.45. T1,meas was lower over the webs, as was seen in the ramp P girder. T4,meas was quite small at
about half that of the ramp P girder. The difference in elevation of the bottom of the top flange
between webs and the bottom of the root of the wings produced an abrupt change in shape of the
positive thermal gradient. This profile change was not a characteristic of ramp P, which had
proportions similar to more common box girders. The mainlane girder had the proportions of a
spine girder by design for appearance.

20
16
12
8 Deck level gradient magnitudes in oC
4 T1,meas=15.39oC
0

Positive gradient Average gradient

Design
measured June 17, 1996 temperature
gradient
through section

0 10 20
(T29) (T31) (T34) (T5) (T2) oC
15.10 18.54 15.32 13.88 14.10
(T28) (T30) (T33) (T6) (T3) (T1)
9.60 12.04 10.49 6.49 4.43 7.60
(T27) (T32) (T35) (T7) (T4)
1.32 1.65 4.76 4.10

(T26) (T8)
2.10 2.37
(T25) (T10) (T9) (T11)
0.76 (Thermocouple) 1.21 1.26 3.43
(T24) Gradient temp oC (T13) (T12) (T14)
-0.13 -1.07 -0.60 2.32
(T23)
-0.68
(T22) (T20) (T16)
-0.85 -1.25 -1.46
(T19) (T18) (T15)
-1.07 -0.40 -0.96
(T21) (T17)
0.10 0.54

Figure 4.45 The maximum measured positive gradient on mainlane girder D5 (from
June 17, 1996)
210
 One of the maximum magnitude measured negative gradients, from November 11, 1995,
is shown in Figure 4.46. The shape of the negative gradient shown in the plot at the right in the
figure was regular in shape only within the depth of the top flange. The rest of the gradient shape
was the sum of the previously existing gradient shape prior to the event that caused the high
magnitude negative gradient, and the negative gradient caused by rapid cooling of the entire
exterior of the box girder. The soffit gradient magnitude (T4,meas) was nearly as large as T1,meas.

0
-2
-4 Deck level gradient temperature in oC
-6
-8 T1,meas=-6.27oC
-10

Design
Negative gradient gradient
measured Nov. 11, 1995 Average gradient
temperature
through section

-10 -5 0
(T29) (T31) (T34) (T5) (T2) oC
-5.63 -7.32 -8.75 -4.65 -5.02
(T28) (T30) (T33) (T6) (T3) (T1)
-2.21 -3.42 -0.06 -0.61 -4.96
(T27) (T32) (T35) (T7) (T4)
-0.19 1.72 1.84 1.01 -3.33

(T26) (T8)
1.65 1.77
(T25) (T10) (T9) (T11)
0.67 (Thermocouple) 1.50 0.79 -2.96
(T24) Gradient temp oC (T13) (T12) (T14)
0.27 0.93 -0.58 -4.25
(T23)
0.27
(T22) (T20) (T16)
-1.17 0.32 0.27
(T19) (T18) (T15)
-0.80 -0.55 -1.14
(T21) (T17)
-4.23 -3.26

Figure 4.46 Measured negative gradient on mainlane girder D5 (from November 11, 1995)

211
 The statistical distribution of peak daily thermal gradient deck level magnitudes for the
mainlane box girder was similar to that seen on ramp P. The daily peak T1,meas values were
somewhat larger on average on the mainlane for the positive gradient case, and the daily peak
T1,meas values were about the same in magnitude and distribution for the negative gradient case.
Figure 4.47 gives the distribution of daily peak T1,meas values for the positive gradient condition
before the blacktop was in place. Gradient magnitudes measured in redundant months were only
given half weight for the statistical distribution plots. The plot in Figure 4.47 includes two spring
seasons, which in this case tended to unrealistically raise the percentages of the higher valued
gradient magnitudes. High deck level gradient magnitudes were consistently measured in both
the spring of 1995 and 1996, while the spring of 1997 did not have such large magnitudes. Figure
4.48 gives the statistical distribution of maximum daily T1,meas values after the blacktop was in
place. The maximum measured T1,meas value with blacktop was in fact larger than the maximum
T1,meas value without the blacktop. The distribution of daily peak deck level positive gradient
magnitude substantially shifted to the lower temperatures after the 50mm of blacktop was in
place. The distribution of daily minimum T1,meas values, shown in Figure 4.49, revealed that very
few days had a substantial negative thermal gradient, with minimum daily T1,meas values in the
-7oC to -9oC range occurring a low percentage of the time. Figure 4.50 shows that any
appreciable deck level negative gradient magnitudes had all but been eliminated once the 50mm
of blacktop was placed, with over half of the days having a minimum T1,meas value greater than
-1oC.

25 %
T1,meas
Occurrence Ratio

20 %
Before asphalt overlay
Database contains 410 days:
15 % March 1995 through May 1996

10 %

5%

0%
<0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Temperature Ranges (oC)

Figure 4.47 Statistical occurrence of daily maximum T1,meas values on mainlane girder D5
before application of the asphalt blacktop

212
 25 %
T1,meas
Occurrence Ratio
20 %
After 50mm asphalt overlay
Database contains 500 days:
15 % June 1996 through December 1997

10 %

5%

0%
<0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Temperature Ranges (oC)

Figure 4.48 Statistical occurrence of daily maximum T1,meas values on

mainlane girder D5 after application of the asphalt blacktop

30 %

25 %
T1,meas
Occurrence Ratio

20 %
Before asphalt overlay
Database contains 410 days:
15 % March 1995 through May 1996

10 %

5%

0%
<0 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -16
Temperature Ranges (oC)

Figure 4.49 Statistical occurrence of daily minimum T1,meas values on mainlane

girder D5 before application of the asphalt blacktop

213
 40 %

35 %

30 %
Occurrence Ratio

25 %
T1,meas
20 %
After 50mm asphalt overlay
Database contains 500 days:
15 % June 1996 through December 1997

10 %

5%

0%
<0 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -16
Temperature Ranges (oC)

Figure 4.50 Statistical occurrence of daily minimum T1,meas values on

mainlane girder D5 after application of the asphalt blacktop

Both measured and calculated stresses are presented in the figures to follow. Stresses
were calculated using the measured temperature changes and the cross-sectional model in Figure
4.51. This calculation assumed that the measured temperatures were distributed evenly within the
elements. The necessary integrations for the calculation also used these elements, and plane
sections were assumed to remain plane. Stresses varied across the width of the cross section
depending on the temperature change within each element. Stresses were also calculated using
the appropriate AASHTO LRFD design thermal gradients. Plane sections were assumed to
remain plane in this calculation, and there was no transverse variation of temperature in the design
gradient. Therefore, no transverse variation of stress was calculated, as can be seen in the figures.

thermocouple Neutral axis

Figure 4.51 Division of the mainlane girder D5 cross-section into tributary areas
for each thermocouple gauge
214
 Stresses are plotted for the positive thermal gradient case from April 20, 1995 in Figures
4.52 and 4.53. The positive thermal gradient from this day produced an extreme gradient through
the thickness of the top flange between webs. The maximum temperature change at one
thermocouple, located at the center of the top flange, was 17.1oC for the 9 hour interval under
consideration. Figure 4.52 shows the stresses in segment D5-9 near midspan. The thermocouples
were also located in this segment. Peak measured stresses in the top flange generally fell below
those calculated using the AASHTO LRFD design gradient of 25.6oC. The measured stresses in
the top slab were higher over the webs than in the wings. The measured soffit gradient magnitude
of 3.3oC was higher than the 2.8oC recommended by AASHTO LRFD, yet the measured bottom
flange stresses did not exceed those calculated using the AASHTO LRFD gradient. The
measured stresses were actually predicted quite well by the calculation using the AASHTO LRFD
gradient, even though the gradient shape differed substantially at some points. The best
comparison can be made using the plots of stress over the depth of the webs. The location of the
strain gauges about 80mm below the surfaces of the concrete did not allow measurement of the
peak top and bottom fiber stresses predicted by the calculations. The measured stresses in the top
flange did compare well to the calculated values at the appropriate depth beneath the top fiber.

215
 (+ compression, - tension)
10
Stress (MPa)
5

-5

Segment D5-9
Positive Gradient
April 20, 1995

Measured
(+ compression, - tension)

Gradient Stress
10 Calculated
Stress (MPa)

Gradient Stress
5 Design
Gradient Stress
0
Concrete Strain
Gauge
-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.52 Comparison of measured and calculated positive thermal gradient

stresses for segment D5-9 (April 20, 1995)

216
 (+ compression, - tension)
10
Stress (MPa)
5

-5

Segment D5-16
Positive Gradient
April 20, 1995

Measured
(+ compression, - tension)

Gradient Stress
Calculated
10
Gradient Stress
Stress (MPa)

Design
5 Gradient Stress

0 Concrete Strain
Gauge
-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.53 Comparison of measured and calculated positive thermal gradient

stresses for segment D5-16 (April 20, 1995)

Although the positive thermal gradient in the heavy end diaphragm segment D5-16 at the
longitudinal strain gauge locations probably differed somewhat from the gradient measured in

217
segment D5-9, the measured stresses in these two segments were almost identical. The thermal
gradient induced stresses in segment D5-16 for April 20, 1995 are plotted in Figure 4.53. The
strain gauges in this segment were located only 90mm from the heavy end diaphragm, yet the
measured stresses differed substantially from those in segment D5-9 only near the neutral axis of
the girder. The gauges located near the neutral axis of the girder were tied to the exterior plane of
the web bar cage. Therefore, the strains measured by these gauges may not reflect the true
average strain change of all points at this depth in the section, since a positive thermal gradient
was present across the thickness of the web. The measured stresses were also somewhat lower
near the wing tip than the wing tip stresses in segment D5-9. The good comparison between the
gauges at these two locations gives credibility to the measurements, especially since the internally
balanced stresses being measured should have been similar along the length of the girder. The
shape of the plot of measured and calculated stresses down the depth of the web of the mainlane
differed from that for the ramp P girder. The simple span mainlane girder had only self-
equilibrating stresses, so a balance of tensile and compressive stresses was produced over the
depth of the web seen at the bottom of Figure 4.52. The shape of the curve for the measured and
calculated stresses was exactly like the shape of the thermal gradient. The measured and
calculated stresses in the ramp P girder seen at the bottom of Figures 4.32, 4.33 and 4.34 were
influenced by a combination of bending stresses and self-equilibrating stresses because the ramp P
girder was continuous across the piers, and had a moment connection to some of the piers.

Stress changes from the day of the maximum positive thermal gradient deck level
magnitude on June 17, 1996 are shown in Figures 4.54 and 4.55. At this point in time there was
50mm of blacktop in place, and the top slab closure pour had been cast between girders at the
joint locations. The 50mm of blacktop changed the shape of the positive gradient from that of
April 20, 1995. The most important difference was the top slab temperature variation. The 50mm
of blacktop insulated the top flange concrete from heat loss during the night, nearly eliminating
the formation of a daily negative gradient. The daily maximum T1,meas values with or without
blacktop were nearly the same, at 17.1oC without blacktop and 18.5oC with 50mm of blacktop.
The measured stresses from the June 17, 1996 gradient were similar to those measured for the
April 20, 1995 gradient, with one major exception. Most of the gauges located in the top flange
over the webs measured strains that would be well into the plastic range of the concrete. Most of
these data points were not plotted in Figures 4.54 and 4.55 because they were off the selected
scale. This behavior was also measured on the ramp P girder. It is possible that the concrete in

218
the heavily congested area over the webs had fatigued and was effectively relieving thermal
stresses because of cracked concrete. This may have been the case only in the concrete area
immediately adjacent to the heavy web stirrup bars, which was where the strain gauges giving the
large measurements were tied in place. It is also possible that the strain gauges themselves had
fatigued and were giving erroneous readings, even though they were designed for a long fatigue
life over a wide temperature range. In all cases the only gauges giving the high strain readings
were located over the webs in the top flange. The measured stresses through the depth of the
webs for the June 17, 1996 gradient were similar in segments D5-9 and D5-16. These stresses
were also similar to those measured for the April 20, 1995 positive gradient. The measured
stresses compared better to the stresses calculated using the actual measured temperature changes
than to the design gradient, as was also the case for the April 20, 1995 gradient.

219
 (+ compression, - tension)
10
Stress (MPa)
5

-5

Segment D5-9
Positive Gradient
June 17, 1996
(+ compression, - tension)

Measured
Gradient Stress
10 Calculated
Stress (MPa)

Gradient Stress
Design
5
Gradient Stress

0 Concrete Strain
Gauge
-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.54 Comparison of measured and calculated positive thermal gradient

stresses for segment D5-9 (June 17, 1996)

220
 (+ compression, - tension)
Stress (MPa) 10

-5

Segment D5-16
Positive Gradient
June 17, 1996

Measured
(+ compression, - tension)

Gradient Stress
10 Calculated
Gradient Stress
Stress (MPa)

Design
5 Gradient Stress

0 Concrete Strain
Gauge
-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.55 Comparison of measured and calculated positive thermal gradient

stresses for segment D5-16 (June 17, 1996)

Calculated and measured stresses for the negative gradient case from November 11, 1995
are plotted in Figures 4.56 and 4.57. In general, the measured stresses were lower than the
221
calculated stresses, especially near the neutral axis of the girder. The measured stresses in both
segment D5-9 (in Figure 4.56) and segment D5-16 (in Figure 4.57) were closer to the stresses
calculated using the measured gradient than that calculated using the AASHTO LRFD gradient.
Neither the shape nor the magnitude of the AASHTO LRFD gradient was very similar to the
actual gradient. Once again, the measured stresses in the top flange decreased from webline to
wingtip, indicating plane sections were not remaining plane even at midspan.

222
 (+ compression, - tension)
10
Stress (MPa)
5

-5

Segment D5-9
Negative Gradient
November 11, 1995

Measured
(+ compression, - tension)

Gradient Stress
10 Calculated
Gradient Stress
Stress (MPa)

Design
5 Gradient Stress

0 Concrete Strain
Gauge
-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)
Figure 4.56 Comparison of measured and calculated negative thermal gradient
stresses for segment D5-9 (November 11, 1995)

223
 (+ compression, - tension)
Stress (MPa) 10

-5

Segment D5-16
Negative Gradient
November 11, 1995

Measured
(+ compression, - tension)

Gradient Stress
Calculated
10 Gradient Stress
Stress (MPa)

Design
5 Gradient Stress

Concrete Strain
0
Gauge

-5

10 5 0 -5
Stress (MPa)
(+ compression, - tension)

Figure 4.57 Comparison of measured and calculated negative thermal gradient

stresses for segment D5-16 (November 11, 1995)

The top flange of the mainlane box girder experienced thermal gradients through its
thickness of magnitudes large enough to warrant consideration during the transverse design of the

224
section. Figure 4.58 shows the peak positive and negative gradients measured through the top
flange, and the corresponding gradients at other points in the section. These gradients occurred
when no blacktop was in place. Measurements indicated that these peak gradients would be
similar to those with 50mm of asphalt in place for the top flange. The positive thermal gradients
measured through the thickness of the webs and bottom flange were of insignificant magnitude.
The magnitude of the peak positive gradient in the top flange was 17.5oC, exceeding the
maximum measured deck level gradient over the depth of the entire section on that day.
Significant magnitude negative thermal gradients occurred in the webs and bottom flange, as seen
in the center section of Figure 4.58, with the peak negative gradient magnitude occurring in the
top flange. Since the actual thermocouple locations were never closer than 25mm from any
concrete surface, the peak positive and negative gradient magnitudes were most probably larger
than those presented in all cases.

225
 7.4 5.6 9.0

-10.1 -8.4
-0.1 -0.8
1.8
-0.2
-0.3 -0.4
2.3
1.2
1.1
Positive gradient from
May 20, 1995

-4.9 -5.5 -5.1

5.0 5.5
-4.6
0.9 -5.5

1.7 -6.7
1.1 1.9
-3.8
-5.0
Negative gradient from
January 18, 1996

C1 C3

C2 C4
C5
C6
C7
C8
C9
C13 C11
C10
C14 C12

Figure 4.58 Measured peak positive and negative thermal gradients in degrees
Celsius for segment D5-9 flanges and webs

Table 4.1 gives the measured strains from the May 20, 1995 positive gradient, and the strains
predicted by analysis using the measured gradient. The magnitude of the measured strains was
less than the calculated strains at nearly every gauge location. The analytical model predicted
compression on the outside face and tension on the inside face of the box girder, and the gauges
measured a similar response. This was different from the response measured on ramp P. For the

226
positive gradient case the ramp P girder had compression in the top flange and tension in the
bottom flange, as well as bending and self-equilibrating stresses. Two major differences existed
between the mainlane girder and the ramp girder. First, the mainlane girder was significantly
wider for its depth than the ramp P girder, which has an effect on the distribution of bending
stresses around the section. Secondly, the mainlane box girder had a longitudinal crack at the top
of both webs where the webs met the fillet on the inside of the girder. The reduction in stiffness
from the crack would contribute to a reduction in bending stresses and axial stresses in the top
flange and bottom flange by reduced bending moments in the webs. The crack occurred while the
segments were still in the form when the concrete was relatively weak. The crack was caused by
moments generated by the transverse pretensioning forces in the top flange. The prestressing
strands were cut when the concrete was less than 24 hours old. The ramp P box girder did not
crack from the pretensioning forces because the distance between webs was much shorter,
reducing the elastic shortening from the prestressing. The analytical model used for calculating
the strains in Table 4.1 did not account for this reduction in stiffness at the tops of the webs,
which is common practice.

Table 4.1 Mainlane girder D5 transverse thermal strains from the

May 20, 1995 positive gradient
Gauge Location Measured Strains in με Calculated Strains in με
C1 -99
C2 113
C3 -36 -63
C4 39 61
C5 8 26
C6 -32 -32
C7 5 19
C8 -16 -28
C9 12
C10 -9 -18
C11 18 11
C12 -12 -14
C13 12
C14 -10 -15

227
4.3.3 Large ramp pier P16
Measurements were taken to determine the importance of thermal gradient effects in the
voided octagon pier P16. The relationship of the sides of the pier to the sun's exposure greatly
influenced the shape and magnitude of the gradient over the course of a day. Figure 4.59 shows
the temperatures measured by the thermocouples just under the concrete surface in pier segment
PC16-5 near the midheight of the pier for June 17, 1996. A general and smooth cooling trend
occurred overnight, with the gauges all converging to nearly the same temperature. The morning
sun exposure first warmed the east face of the pier. As the sun rose, its radiation hit the pier at an
increasingly acute angle. For this reason the south face of the pier underwent the lowest
temperature change of any side of the pier that was directly exposed to the sun. The west face of
the pier underwent the greatest temperature change due to all day ambient heating and direct
exposure to the sun's radiation at the hottest part of the day. Conditions were ideal for the
development of a large positive thermal gradient on this day, as was also measured on the
mainlane girder.

June 17, 1996

44.0 S
SE T410 SW W
Thermocouple Temperature (oC)

42.0
T431 T413
40.0
E Up STA W E SW
38.0 T428 T416
SE
36.0

34.0
S

32.0
T428
30.0 T431
T410
28.0 T413
T416
26.0
12 PM
1 PM
2 PM
3 PM
4 PM
5 PM
6 PM
7 PM
8 PM
9 PM
10 PM
11 PM
12 AM
1 AM
2 AM
3 AM
4 AM
5 AM
6 AM
7 AM
8 AM
9 AM
10 AM
11 AM

12 AM

Figure 4.59 Typical daily temperature cycle of selected thermocouples in segment PC16-5

The gradient measured through the thickness of the west wall in segment PC16-5 was
also substantial, plotted in Figure 4.60. The interior thermocouples T417 and T418 did not cool to

228
the same temperature as the exterior of the concrete overnight, and changed little over the course
of the day. The vertical distribution of concrete temperatures on the west face is given in Figure
4.61. The temperature measured near midheight of the pier by thermocouple T417 was very
similar to the temperature recorded by T442 located only 610mm from the massive solid capital
segment. The superstructure was not in place so no shading of the pier's west face occurred
during the entire day. Temperatures measured 610mm from the solid base of the pier were
substantially lower than the temperatures further up the pier because of the heat loss to the footing
and ground, and from shading by the north and southbound lanes of IH35. Temperatures recorded
by thermocouple T511 in the solid capital segment did not match the peak changes at T417 and
T442 because of heat loss to the solid core of that segment.

June 17, 1996

45.0

43.0
Thermocouple Temperature (oC)

41.0
Up STA T418 T416
39.0 T417
W
37.0

35.0

33.0

31.0

29.0 T416
T417
27.0 T418

25.0
1 AM
2 AM
3 AM
4 AM
5 AM
6 AM
7 AM
8 AM
9 AM

12 PM

10 PM
11 PM
12 AM

10 AM
11 AM

1 PM
2 PM
3 PM
4 PM
5 PM
6 PM
7 PM
8 PM
9 PM

12 AM

Figure 4.60 Temperatures recorded through the thickness of the west wall of segment PC16-5

229
 June 17, 1996
37.0
Thermocouple Temperature (oC)
T402 : Base Segment
36.0 T417 : Gradient Segment
T442 : Top Segment
35.0 T511 : Capital Segment
Location of
thermocouple
34.0 in each segment

33.0

32.0
E W
31.0 Up STA

30.0

29.0
1 AM
2 AM
3 AM
4 AM
5 AM
6 AM
7 AM
8 AM
9 AM

12 PM
1 PM
2 PM
3 PM
4 PM
5 PM
6 PM
7 PM
8 PM
9 PM
10 PM
11 PM
12 AM

10 AM
11 AM

12 AM
Figure 4.61 Temperatures measured along the height of pier P16

Detailed plots of the daily temperature changes in segment PC16-5 for June 17, 1996 are
given in Figures 4.62 through 4.65. The plots are intended to show the gradient existing on a
vertical plane of the pier. The maximum gradient magnitude that occurred on the west face was
10oC, which was considerably smaller than the 15oC positive gradient magnitude (T1,meas)
measured on the mainlane on this same day.

230
 Relative Temperature (oC)
Relative Temperature (oC) 10.0 12:00 AM 10.0 6:00 AM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
N S N S
Relative Temperature (oC)

Relative Temperature (oC)

10.0 11:00 AM 10.0 3:00 PM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
N S N S
Relative Temperature (oC)
Relative Temperature (oC)

10.0 7:00 PM 10.0 11:00 PM

8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
N S N S
S T410
T411
T412

Up STA

T424
T423
N T422

Figure 4.62 One day cycle of thermal gradients along the north-south axis of segment PC16-5

231
 Relative Temperature (oC)
Relative Temperature (oC) 10.0 12:00 AM 10.0 6:00 AM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
NE SW NE SW
Relative Temperature (oC)

Relative Temperature (oC)

10.0 11:00 AM 10.0 3:00 PM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
NE SW NE SW
Relative Temperature (oC)
Relative Temperature (oC)

10.0 7:00 PM 10.0 11:00 PM

8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
NE SW NE SW

SW
T413

T415 T414

Up STA

T427

T425 T426
NE
Figure 4.63 One day cycle of thermal gradients along the northeast-southwest axis of
segment PC16-5

232
 Relative Temperature (oC)
Relative Temperature (oC) 10.0 12:00 AM 10.0 6:00 AM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
E W E W
Relative Temperature (oC)

Relative Temperature (oC)

10.0 11:00 AM 10.0 3:00 PM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
E W E W
Relative Temperature (oC)
Relative Temperature (oC)

10.0 7:00 PM 10.0 11:00 PM

8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
E W E W

E W
T429 T417
T428 T430 T418 T416
Up STA

Figure 4.64 One day cycle of thermal gradients along the east-west axis of segment PC16-5

233
 Relative Temperature (oC)
Relative Temperature (oC) 10.0 12:00 AM 10.0 6:00 AM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
SE NW SE NW
Relative Temperature (oC)

Relative Temperature (oC)

10.0 11:00 AM 10.0 3:00 PM
8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
SE NW SE NW
Relative Temperature (oC)
Relative Temperature (oC)

10.0 7:00 PM 10.0 11:00 PM

8.0 8.0

6.0 6.0

4.0 4.0

2.0 2.0

0.0 0.0
SE NW SE NW

T431
T432
SE
T433

Up STA
T421
T420
NW
T419

Figure 4.65 One day cycle of thermal gradients along the northwest-southeast axis of
segment PC16-5

Strain gauges were installed in the pier to measure the response to thermal gradients. The
pier's cross-section was regular and compact, and was expected to behave in a predictable manner.

234
Figure 4.66 shows the temperature changes on the north and south face of pier P16 for June 17,
1996. The temperature changes were nearly identical on both faces of the pier. The measured
strain changes on both faces were also nearly identical. The strain change from unrestrained
thermal expansion was not included in the presented strain. Temperature increases on the exterior
of the pier resulted in compressive stresses that would be opposed by internally balancing stresses
in the interior concrete. The same behavior was seen on the west face of the pier, and is plotted in
Figure 4.67. Strain gauge C433 responded to the temperature gradient from the entire section, as
well as the gradient in the west wall. In general the gauge went into compression as the concrete
on the west face was heated. The pattern was not as easily comprehended as the gauge measured
response of the north and south walls because of the cooling of the east wall and other
temperature changes in the section. Gauge C439 had evidently become debonded from its
mounting rod, and appeared to be accurately giving the thermal strain change for an unrestrained
length of concrete.

June 17, 1996

150 10.0

Thermocouple Change (oC)

Strain Change (microstrain)

Thermocouple Temperature Changes

125 5.0
S
100 0.0 C430 T410

75 -5.0
C430
50 C436 -10.0
T410 Up STA
25 T422 -15.0

0 -20.0 C436 T422

Concrete Strain Changes N
-25 -25.0

-50 -30.0
1:00 AM
2:00 AM
3:00 AM
4:00 AM
5:00 AM
6:00 AM
7:00 AM
8:00 AM
9:00 AM
12:00 AM

10:00 AM
11:00 AM
12:00 PM
1:00 PM
2:00 PM
3:00 PM
4:00 PM
5:00 PM
6:00 PM
7:00 PM
8:00 PM
9:00 PM
10:00 PM
11:00 PM
12:00 AM

Figure 4.66 Temperature and strain changes on the north-south axis of segment PC16-5

235
 June 17, 1996
150 10.0

Thermocouple Change (oC)

Strain Change (microstrain)
125 5.0

100 0.0
S
C439 C433
75 -5.0
C433 Thermocouple Temperature Changes
50 C439 -10.0
T416 Up STA
25 T428 -15.0
T428 T416
0 -20.0
N
-25 -25.0
Concrete Strain Changes
-50 -30.0
1:00 AM
2:00 AM
3:00 AM
4:00 AM
5:00 AM
6:00 AM
7:00 AM
8:00 AM
9:00 AM
12:00 AM

10:00 AM
11:00 AM
12:00 PM
1:00 PM
2:00 PM
3:00 PM
4:00 PM
5:00 PM
6:00 PM
7:00 PM
8:00 PM
9:00 PM
10:00 PM
11:00 PM
12:00 AM
Figure 4.67 Temperature and strain changes on the east-west axis of segment PC16-5

Figure 4.68 shows the measured temperatures and strain response of the north wall.
With little gradient along the north-south axis, the strains appeared to be associated purely with
those from self-equilibrating stresses. Figure 4.69 gives this same data for the west wall of the
pier. Even though the temperature changes through the thickness of the west wall were similar to
those shown for the north wall in Figure 4.68, the strains measured were from self-equilibrating
stresses in response to temperature changes in the entire cross section.

June 17, 1996

150 10.0
Thermocouple Change (oC)
Strain Change (microstrain)

Thermocouple Temperature Changes

125 5.0

100 0.0 S

75 -5.0
C436
50 C438 -10.0
T422 Up STA
25 T424 -15.0
C438 T424
0 -20.0 C436 T422

-25 -25.0
N
Concrete Strain Changes
-50 -30.0
1:00 AM
2:00 AM
3:00 AM
4:00 AM
5:00 AM
6:00 AM
7:00 AM
8:00 AM
9:00 AM
12:00 AM

10:00 AM
11:00 AM
12:00 PM
1:00 PM
2:00 PM
3:00 PM
4:00 PM
5:00 PM
6:00 PM
7:00 PM
8:00 PM
9:00 PM
10:00 PM
11:00 PM
12:00 AM

Figure 4.68 Temperature and strain changes through the thickness of the north wall of
segment PC16-5

236
 June 17, 1996
150 10.0

Thermocouple Change (oC)

Strain Change (microstrain)
125 5.0

100 0.0
Up STA
C433
75 -5.0
C433 Thermocouple Temperature Changes
C435
50 C435 -10.0 T418
T416
25 T418 -15.0
T416
0 -20.0 N
-25 -25.0
Concrete Strain Changes
-50 -30.0
1:00 AM
2:00 AM
3:00 AM
4:00 AM
5:00 AM
6:00 AM
7:00 AM
8:00 AM
9:00 AM
12:00 AM

10:00 AM
11:00 AM
12:00 PM
1:00 PM
2:00 PM
3:00 PM
4:00 PM
5:00 PM
6:00 PM
7:00 PM
8:00 PM
9:00 PM
10:00 PM
11:00 PM
12:00 AM
Figure 4.69 Temperature and strain changes through the thickness of the west wall of
segment PC16-5

In order to compare the measured temperature gradient, which was nonlinear along two
axes, to common singly nonlinear design gradients, an equivalent measured gradient had to be
derived. The concentric octagon method shown in Figure 4.70 was used. Interpolation of
temperatures was only employed within the same octagon shell. This interpolation procedure was
used because temperatures varied greatly through the thickness of the walls. The temperatures
from points of known temperature as well as interpolated temperatures could then be projected
into a gradient shape as demonstrated in Figure 4.71. Irregularities caused by the octagon shape
of the pier were eliminated for the sake of simplicity and applicability of the gradient to piers of
different cross section.

237
 Octagonal Shells

No interpolation
between shells

Measured Temperature
Temperatures interpolated
Locations
around shell perimeter
Figure 4.70 Concentric octagonal model for temperature representation

238
 Known Temperatures

E W

Interpolated
Temperatures

Temperatures averaged
along these lines
16.0
14.0
Temperature (oC)

12.0
10.0
8.0
6.0
Dashed
4.0
segment
hape
2.0 Gradient S used in stress
0.0 calculations
Figure 4.71 Projection of temperature gradient shape onto a single axis

Using this method, the maximum magnitude positive gradient occurred in April 1996 at
o
13.1 C, a time at which the pier's strain gauges had not been activated. The shape of the gradient
and those recommended by AASHTO LRFD and its parent NCHRP 276 report are plotted in
Figure 4.72. The magnitude of the measured gradient was only half of that recommended for
superstructure design. The shape fell between the shapes recommended by AASHTO LRFD and
NCHRP 276. The minimum magnitude negative gradient was measured in April 1996 at -6.8oC.
This value was close to the minimum magnitude negative gradients measured in the

239
superstructure girders. The shape of this gradient is plotted in Figure 4.73. This time the general
shape followed the original NCHRP 276 shape more closely, even though it was not derived from
tests or analysis. The magnitude of the minimum measured negative gradient was, once again,
only about half of that recommended for superstructure design.

26 25.6
24
22
20
Temperature Change (oC)

18
16
14 April 1996
AASHTO-LRFD 13.1
12 NCHRP 276
10
8
6
4
2
0
East Face West Face

E W
Figure 4.72 Measured and design code maximum positive thermal gradients for pier P16

240
 East Face West Face
0
-1
-2
-3
Temperature Change (oC)

-4
-5
-6
-7 -6.8

-8
-9
-10
April 1996
-11 AASHTO-LRFD
-12 NCHRP 276

-13 -12.8
-14

E W
Figure 4.73 Measured and design code maximum negative thermal gradients for pier P16

To gain insight into the actual response of the pier to thermal gradients, the measured
strain gauge data were plotted along with the strains predicted by three methods of analysis. The
first method was the finite element method using temperature changes interpolated along the
octagonal shells in Figure 4.70. The second method was the standard “classical” method
described in Section 4.1.1, and commonly used by engineers. The gradient used was the one-
dimensional gradient derived in Figure 4.71 and plotted in Figure 4.72. The third method was the
primary bending axis method, which used the actual two-dimensional measured gradient to
241
determine the primary axis about which self-equilibrating stresses and curvature strains were
symmetric. Plane sections were assumed to remain plane in this third method.

Figure 4.74 gives the temperature changes along the east-west axis of segment PC16-5
that occurred over a portion of the day on June 17, 1996. The time period began in early morning
when temperatures in the cross section were most uniform, and ended at the time of the maximum
positive gradient. The measured strain changes from segment PC16-1, plotted in Figure 4.75,
were taken over this same time period. The gradient used to calculate the three analytical results
plotted in Figure 4.75 was derived from the temperature changes that occurred over the time
period in segment PC16-5. This gradient was not the thermal gradient that existed in the segment
at the end of the time period. Figure 4.75 shows that the measured strain changes compared fairly
well with the finite element results, even though the gradient in segment PC16-1 was not the same
as the gradient from segment PC16-5 used in the analysis. Figure 4.76 gives a plot of the same
analytical results and the measured strains in the segment where the gradient under consideration
did occur. The measured results in the east and west walls of segment PC16-1 were actually
almost identical to the strains measured in segment PC16-5 shown in Figure 4.76. Once again,
the finite element results compared most favorably with the measured results, followed by the
primary bending axis method.
T416

18
Temperature Change (oC)

16
14
12
T430

T418
T416
T428

10
W
T428

8 E
T417
T429
T417
T429

6 Up STA
T430

4
T418

2
0
-2
East Face West Face
Segment PC16-5
Figure 4.74 Temperature changes along the east-west axis recorded by
thermocouples in segment PC16-5, maximum positive gradient

242
 80

C417
60
Strain Change (microstrain)

40

T
20 C415

C417
C415

C407
C407
0
-20 E W

C
-40
Up STA
-60 Measured Strain Change
-80 Finite Element Model
Classical Method
-100 Primary Bending Axis Method
-120
East Face West Face
Segment PC16-1

Figure 4.75 Calculated strain changes and strain changes recorded by strain gauges
in segment PC16-1, east-west axis, maximum positive gradient

80
C441
Strain Change (microstrain)

60
40
T
C433

20

C441

C433
C435
C435

0
-20 W
E
C

-40
Up STA
-60 Measured Strain Change
Finite Element Model
-80 Classical Method
-100 Primary Bending Axis Method
-120
East Face West Face
Segment PC16-5

Figure 4.76 Calculated strain changes and strain changes recorded by strain gauges
in segment PC16-5, east-west axis, maximum positive gradient

The temperature changes in segment PC16-5 along its north-south axis for the same time
period on June 17, 1996 are plotted in Figure 4.77. The positive gradient condition produced a
very unsymmetric gradient with respect to the east-west axis, and produced an almost
symmetrical gradient condition along the north-south axis. The strain gauge output and analytical
results shown in Figures 4.78 and 4.79 were dominated by strains that were self-equilibrating
within the width of the south and north walls. The primary bending axis method predicted strains
that compared well to the finite element method strains for this symmetrical gradient, but did not
compare well with the results calculated by the classical method along either axis. Once again, the

243
measured strains shown in Figures 4.78 and 4.79 compared well to the finite element method
results and the primary bending axis results.

S
18 T410
T411
16 T412
Temperature Change (oC)

14
12
T422

10

T410
8
Up STA
T423

T411
4

T412
T424

2
T424
0 T423
T422
-2
North Face South Face N
Segment PC16-5
Figure 4.77 Temperature changes along the north-south axis recorded by
thermocouples in segment PC16-5, maximum positive gradient

S
80
Strain Change (microstrain)

C400
C402

60 C402
40
C412

T
C410

20
0
-20
C

-40
C400

Up STA
-60 Measured Strain Change
Finite Element Model
-80 Classical Method
-100 Primary Bending Axis Method C412
-120 C410
North Face South Face
N
Segment PC16-1
Figure 4.78 Calculated strain changes and strain changes recorded by strain gauges
in segment PC16-1, north-south axis, maximum positive gradient

244
 S

C438
80
Strain Change (microstrain) C430
60 C432
40

T
C432
20
0
-20

C
C430
-40
Up STA
C436

-60 Measured Strain Change

Finite Element Model
-80 Classical Method
-100 Primary Bending Axis Method C438
-120 C436
North Face South Face
N
Segment PC16-5

Figure 4.79 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-5, north-south axis, maximum positive gradient

To evaluate the pier's structural response to negative gradients, a time period was selected
that produced a large magnitude negative gradient. The negative gradient occurred on the night of
June 3-4, 1996, when intense daytime heating left the pier quite warm, and a large nighttime
temperature drop cooled the exterior of the pier rapidly. The west face of the pier was the
warmest after the daytime heating, and cooled the greatest during the night, creating an
unsymmetrical negative gradient in the pier. The temperature change along the east-west axis in
segment PC16-5 for the time period under consideration is plotted in Figure 4.80. Figure 4.81
shows that the measured strains in segment PC16-1 compared well to the strains predicted by the
finite element method and the primary bending axis method in the east wall. The strain measured
by gauge C407 was less than predicted by these two methods of analysis. A similar response
occurred for the positive gradient case plotted in Figure 4.75. The small magnitude of the strain
measured by C407 could have been correct for the gradient actually occurring in segment PC16-1,
or the gauge could have been poorly bonded in the concrete and was giving poor results. The
measured strains in the west wall plotted in Figure 4.82 also had a magnitude smaller than that
predicted by the various methods of analysis. Reasonable correlation existed between the
measured strain from gauge C441 and calculated strains in the east wall.

245
 1
Temperature Change (oC)
0

T418
T430
-1
-2
T429

T430

T418
T416
T428
-3

T417
-4 E W

T417
T429
-5 Up STA
-6
T428

-7

T416
-8
-9
East Face West Face
Segment PC16-5
Figure 4.80 Temperature changes along the east-west axis recorded by
thermocouples in segment PC16-5, maximum negative gradient

50
Strain Change (microstrain)

40 Measured Strain Change

Finite Element Model
30 Classical Method T
20 Primary Bending Axis Method

C417
C415

C407
10
C415

0 W
E
C

-10
C407

Up STA
-20
C417

-30
-40
-50
East Face West Face
Segment PC16-1
Figure 4.81 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, east-west axis, maximum negative gradient

246
 50
Measured Strain Change
Strain Change (microstrain)

40
Finite Element Model
30 Classical Method

T
20 Primary Bending Axis Method

C441

C433
C435
C435
10

C433
0 W
E

C
-10
Up STA
-20
C441

-30
-40
-50
East Face West Face
Segment PC16-5
Figure 4.82 Calculated strain changes and strain changes recorded by strain gauges
in segment PC16-5, east-west axis, maximum negative gradient

The negative gradient temperature changes along the north-south axis from June 3-4, 1996
are plotted in Figure 4.83. The gradient condition was nearly symmetric. The measured strains in
segments PC16-1 and PC16-5, shown in Figures 4.84 and 4.85, continued to compare well with
the finite element method calculated strains, and reasonably well with the primary bending axis
method. The classical method did not allow enough refinement of the input gradient to compare
well with the measured data. Modification of the method, such as was the case for the primary
bending axis method, would have to be done to get realistic results from an analysis of a two-
dimensional gradient.

S
1 T410
T411
0 T412
Temperature Change (oC)

T430

T418

-1
T417

-2
T429

-3
-4
-5 Up STA
T416

-6
T428

-7
T424
-8 T423
T422
-9
North Face South Face N
Segment PC16-5
Figure 4.83 Temperature changes along the north-south axis recorded by
thermocouples in segment PC16-5, maximum negative gradient

247
 S
Strain Change (microstrain) 50 C400
40 Measured Strain Change
Finite Element Model C402
30 Classical Method

T
20 Primary Bending Axis Method
10
C410

C400
0
C412

C
-10

C402
Up STA
-20
-30
-40 C412
-50 C410
North Face South Face
N
Segment PC16-1
Figure 4.84 Calculated strain changes and strain changes recorded by strain gauges in
segment PC16-1, north-south axis, maximum negative gradient

S
50
C436

C430
Measured Strain Change
Strain Change (microstrain)

C430
40 Finite Element Model C432
30 Classical Method
T

20 Primary Bending Axis Method

10
C432

0
C

-10
Up STA
-20
-30
C438

-40 C438
-50 C436
North Face South Face
N
Segment PC16-5
Figure 4.85 Calculated strain changes and strain changes recorded by strain gauges
in segment PC16-5, north-south axis, maximum negative gradient

Substructure elements are often massive in section, whereas superstructure elements are
made with efficient sections to reduce dead load moments and shears. The dead weight of a pier
seldom has much effect on the foundation costs of a bridge. Also, the additional costs of forming
a voided pier section are often greater than the savings in concrete volume. For these reasons, the
cross sections of many piers, or portions thereof such as the capital segment on pier P16, are often
cast solid. All the piers on the U.S. 183 project were originally designed to be voided in section
and precast. Because of site conditions, the contractor elected to cast most of the piers in place,
with solid sections. The ramp P piers, including P16, were the only piers precast on the project.
The design used voided piers to reduce weight for transport, which required an increase in

248
concrete strength over that required for larger solid piers. The capital segment was solid in
section to handle the multiple post-tensioning anchorages cast within it and the resultant
complicated stress paths. The concrete for the ramp capital and ramp column segments had
nearly the highest design strength of any concrete on the project at 51.7MPa. Concrete strengths
generally fell well above their 28 day design strengths to make certain the one day strength of the
concrete would allow the segments to be removed from their forms after only one night of curing.
The concrete used was very reactive because of its self-generated heat, and always far exceeded
the one-day strength requirement. Vertical cracks were noticed in the voided octagon column
segments upon removal of the forms. The cracks generally followed an internal soft inclusion in
the section, especially the 200mm PVC drainpipes. The cracks were caused by thermal gradients
during curing. The concrete was taking its initial set at about uniform temperature, so no thermal
gradient existed initially. Once the concrete began to react at an increased rate because of its self-
generated heat, the average temperature of the concrete increased. The steel exterior form
radiated this heat to the atmosphere creating a negative gradient. The concrete, still very weak
shortly after its initial set, cracked vertically from tensile stresses. The transverse steel area was
low at 470mm2/m.

Based on these observations, the thermocouples in the capital segment were activated to
measure the curing temperature gradients. The solid capital segments were also found to be
cracked as they sat in the form. The cracks formed at the architectural reveals in both the vertical
and horizontal directions were probably caused by the mechanism demonstrated in Figure 4.86.
The core of the segment was found to get quite hot during curing with a maximum of about 80oC.
A plot of the core temperature over time is given in Figure 4.87. This plot shows that the core
concrete did not reach the ambient air temperature for a full week. The shape of the maximum
negative gradient in the capital segment is plotted in Figure 4.88, two days after the concrete was
cast. This gradient magnitude was about 35oC. The gradient that occurred in the form was
smaller than this, but occurred at a time when the concrete was very weak in tension, and was
sufficient to crack the concrete. The cracking pattern seen on the capital segment is shown in
Figure 4.89. The cracks occurred mostly at the midpoint of each side, both vertically and
horizontally. Once the gradient had subsided, the tension condition on the exterior face ended and
the cracks closed. Over time, the cracks opened back up slightly because of shrinkage.

249
 Cracks Form at Reveals
Due to Volume Changes

Shell Concrete Contracts Hot Interior

(Cooled by Ambient Air) Expands

Architectural
Reveals

No Internal Heating High Internal

Curing Temperature

Core Concrete
Shell Concrete
Figure 4.86 Mechanism for cracking of large monolithic members during curing

No Data
90.0 Forms
Removed
80.0

70.0

60.0 Core Temperature (T539)

Temperature (oC)

50.0 Forms in
Place
40.0

30.0

20.0

10.0 Casting Complete Ambient Air

10:00 am Sept. 18
0.0
18-Sep

19-Sep

20-Sep

21-Sep

22-Sep

23-Sep

24-Sep

25-Sep

26-Sep

Figure 4.87 Comparison of internal curing temperature to ambient air

temperature, capital segment PC16-8

250
 36.0 S
Relative Temperature (oC) T523
30.0 T531

24.0 T537
T539
18.0
Up STA T535
12.0
T527
6.0 T519
N
0.0 Section Located 915mm
12:00 PM September 20, 1995
S N from Bottom Face

Figure 4.88 Maximum measured gradient shape during curing of capital segment PC16-8

Cracks
Blockouts

1
8 2

1 2 3 4
7 3

6 4
5

Up STA 5 6 7 8

Figure 4.89 Map of cracks found during curing of capital segment PC16-8

4.3.4 Mainlane pier D6

The mainlane Y-pier provided an opportunity to study a steel and concrete composite
structure for thermal effects. The superstructure was not placed on the pier for almost a year after
it was built, so the pier experienced a variety of climatic conditions under direct sunlight. The
two most important types of measurements taken were those that demonstrated the behavior of the
pier when the structural steel pipe ties changed temperature with respect to the concrete, and when
the exterior of the solid concrete column and capital changed temperature with respect to the core
of the concrete.

251
 The difference in temperature between the exterior concrete of the capital and the core
concrete is shown in Figure 4.90 for a typical sunny day in March. Thermocouples T104 and
T106 had western exposure. The core concrete temperature changed little over the course of the
day. The exterior concrete changed about 8oC, with a maximum positive gradient of about 8oC
existing through the section. The 203mm steel pipe tie temperature, measured by T103 in Figure
4.91, changed about 11oC on the same day. The temperature of the steel pipe decreased rapidly as
its embedment length into the concrete increased, with The temperature at T101 being only 3oC
greater than that at T105. Figure 4.92 shows the change in steel pipe tie temperature versus the
temperatures measured at the core concrete of the capital. The maximum gradient is about 10oC.

30
T104
25 T105
T103 T102 T101
T106
Temperature (C)

20
T107
T105
15 T104

T107
T106
10 T108
T109

0
12 PM 6 PM 12 AM Thermocouple
12 AM 6 AM

Figure 4.90 One day cycle of pier D6 capital concrete temperatures

252
 30
T101 - Inside concrete
25 T102 - Face of concrete T103 T102 T101
T103 - Center of pipe
Temperature (C)

20 T105
T104
15
T107
T106
T108
10 T109

Thermocouple
0
12 AM 6 AM 12 PM 6 PM 12 AM

Figure 4.91 One day cycle of pier D6 capital pipe temperatures

March 11,1995
12
10 T103 T102 T101
T103
8 T105
Temperature (C)

T105
6 T107
T104
T109
4
T107
T106
2 T108
T109
0

-2
-4 Thermocouple
12 2 4 6 8 10 12 2 4 6 8 10 12
AM AM AM AM AM AM PM PM PM PM PM PM PM

Figure 4.92 Pier D6 capital pipe temperature versus concrete core temperature,
typical sunny day

Strain gauges were installed on the structural steel pipe ties, with the locations shown in
Figure 4.93, to measure the structure's response to the heating and cooling of the pipes. Figure
4.93 shows a change of 12 microstrain in compression at S129 and S130 located at the center of
the pipe. This change was a response to the heating of the pipes over the day, and the thermal
gradient in the capital. This strain related to a stress change in the pipes of only 2.4Mpa. The
peak strain change for gauges S119 and S120 was at 4:00PM, when the sun was striking the
253
anchor plate for the pipe located on the surface of the pier capital. The strain measured at S119
and S120 showed that the steel pipes were bonded to the concrete at this point in time because of
their rough-galvanized finish.

March 11, 1995

15

10

5
Microstrain

S119, S120
-5
S121, S122

-10 S123, S124

S126, S127, S128

-15 S129, S130

-20
12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

S130
S127 S128

S129 S126 S125

S125 I) II)
S126
A S127 S123 S121 S119
S128 S124 S122 S120 S122
S129 S124
S130 S123 S121
III) IV)
S120
A
S119
I) II) III) IV) V) V)

PLAN VIEW OF CAPITAL VIEW A-A

Figure 4.93 Pier D6 pipe measured strains, typical sunny day

254
 Strain gauges were also located throughout the pier capital at the locations shown in
Figure 4.94. Figure 4.95 shows that the magnitude of the strain changes over the course of this
sunny day of the core concrete gauges C119 and C120 was similar to the exterior gauges C116
and C117 on the northern face, but opposite in direction. Exterior gauges on the western face of
the capital measured considerable compression, with a change of 122 microstrain for C121 over
the day. This strain related to a stress of about 3.3MPa, or about 20% of the normal service load
level allowable stress.

S105 C133 C129 C130

(S106) (C131) (C132) (C128)
A A
S103
(S104) C122 C126 C124
B (C125)(C123) (C127) B
S101 C116 C117 C118
(S102) (C119) (C120) (C121)
C C
LEGEND
C-Gauge
Strain Gauge

C121
C119
C120
C125

C127
C126 C123

S106 C131 C132 C128 S102

S104
C124
C122

S101
C133 C129
C117
C118
C116

S105 C130 S103

SECTION A-A SECTION B-B SECTION C-C

Figure 4.94 Pier D6 gauge locations

255
 March 11, 1995
60

40
Strain change in microstrain

20

-20

-40
C116 - Exterior, northern face
-60
C117 - Exterior, northern face
-80 C118 - Exterior, western face
-100

-120
12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

60

40
Strain change in microstrain

20

-20

-40
C119 - Interior
-60 C120 - Interior
C121 - Exterior, western face
-80

-100

-120
12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

Figure 4.95 Measured concrete strains in the pier D6 capital over a sunny day

Similar behavior was seen in the pier column. Gauges were placed in the top of the
column at the locations shown in the section at the bottom of Figure 4.96. Strain gauges C114
and C113 showed very similar change, even though C113 was at the quarter point of the column's
thickness and C114 was at the center of the core. The strain changes measured by C111 and C112
256
were much larger and opposite in direction than the core gauges. The strain change measured by
C112 was similar to that measured by C118 in Figure 4.95. The inoperable gauge located on the
west face of the column would no doubt have produced strains similar to C121 in the capital. The
positive thermal gradient did not appear to extend very deeply into the concrete, as was seen on
other parts of the project. The massiveness of the core concrete kept thermal stresses low in the
core and high on the exterior.

March 11, 1995

20

10
Strain change in microstrain

-10

-20

-30

-40
C111 - Exterior, northern face
-50 C112 - Exterior, northern face
-60 C113 - Interior
C114 - Interior
-70

-80
12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

Transverse CL Pier C114

C113

Longitudinal C C111 C112

L Pier

SECTION AT TOP OF COLUMN

Figure 4.96 Measured concrete strains at the top of pier D6 column over a sunny day

4.4 DISCUSSION OF RESULTS

The position of the thermocouples and strain gauges in the segments was usually
governed by the location of the reinforcing bars and the anticipated path of heavy concrete flow
into the forms. Many of the gauges were attached to extra bars added strictly for the purpose of
257
securing the gauges and wiring at points in the concrete where no bars existed. A reasonable
amount of cover concrete, usually 50mm, was allowed for these extra bars. The gauges,
particularly the thermocouples, were often extended beyond these bars or the bar cage into the
cover region. The thermocouples could be extended about 25mm beyond the bar to which it was
attached before the flow of concrete would move it out of position. The position of the gauges
near the surface of the top slab of the superstructure segments was checked by hand just prior to
the passing of the finishing machine. The flow of concrete in front of the finishing machine
required that the gauges be placed no closer than 25mm from the surface. The thermocouples at
other locations in the concrete could not be checked for position after the concrete was placed,
and therefore could only be extended into the cover concrete a maximum of 25mm beyond their
attachment point to a reinforcing bar, and was usually seen to be slightly bent after the passing of
concrete to about 35mm clearance to the form.

The temperatures measured by these gauges were not the actual surface temperatures at
points adjacent to the gauges. The actual surface temperatures would have been somewhat higher
or lower, depending on the conditions. A linear interpolation could be made from two gauges to
predict the actual surface temperature, but would be unconservative because of the nonlinear
temperature distribution that always existed during periods of intense heating or cooling. For this
reason the peak temperatures (T1,meas) of the measured gradients presented earlier were in most
cases somewhat lower than the actual peak temperatures. The analysis using the measured
gradients also probably predicted stresses that were somewhat low compared to stresses
calculated from the gradient that actually existed. The position of the strain gauges in the
concrete was also governed by the flow of concrete into the form. The mounting rods for the
strain gauges were easily bent if subjected to the direct flow of concrete, and therefore could not
be placed in the cover region. The gauges were usually tied to the heavy bars in the reinforcing
cage. For this reason the strain gauges could not be placed directly under the surface of the
concrete, and attempts to do so resulted in damaged gauges most of the time from foot traffic and
concrete flow. The location of the gauges, usually about 50mm or more under the concrete
surface, prevented the measurement of the peak thermal stresses. A large stress gradient existed
through the concrete adjacent to the surface being heated or cooled. A gauge only 50mm away
from the surface would measure a strain change substantially lower than that at the actual concrete
surface. The following discussions consider the reduced sensitivity of the measurements because
of the proximity of the gauges to the surface of the concrete.

258
4.4.1 Superstructure thermal gradients for longitudinal analysis
Positive gradients
The maximum measured deck level positive gradient magnitude (T1,meas), both with and
without blacktop, was about 14oC for ramp P and 16oC for the mainlane girder. A linear
interpolation that took into account the concrete cover over the thermocouples usually resulted in
a 2oC to 3oC increase over these measured values. The actual increase was probably higher than
this because of the nonlinear shape of the temperature distribution, especially for the case without
blacktop. With this increase, the maximum T1,meas for the mainlane would have been close to the
20oC deck level positive gradient magnitude (T1) recommended by the AASHTO LRDF guide
specification [9] with 50mm of blacktop. The AASHTO LRFD recommended T1 of 25.6oC for
plain concrete surfaces would be about 6oC too high when compared to the maximum measured
results. The measured results without the blacktop contained only one year of data, so slightly
higher T1,meas could have been possible.

The daily temperature changes of the top flange surface were found generally to be more
extreme when no blacktop was in place during periods of daily positive gradient formation, as is
reflected in the statistical distribution plots of Figures 4.47 and 4.48. Higher T1,meas values
occurred on average without blacktop. Regardless of this fact, the maximum absolute temperature
change over the course of a day measured at the top layer of thermocouples was 17.1oC without
blacktop and 18.5oC with 50mm of blacktop.

The statistical distribution of maximum daily T1,meas values, shown in Figures 4.47 and 4.48,
revealed that large deck level gradient magnitudes, from 40% to 95% of the measured maximum,
occurred most of the time when no blacktop was present. After 50mm of blacktop was added,
daily maximum T1,meas values of 60% of the magnitude of the measured maximum or less occurred
most of the time. This indicated that the positive thermal gradient case without blacktop would
warrant a higher design gradient than with blacktop. Table 4.2 gives maximum T1,meas values
measured in the mainlane and the ramp girder. The 95% fractiles for the measured values are also
given in Table 4.2. The 95% fractile value is often chosen as the appropriate design value for
other types of loads. The maximum T1,meas values shown were most likely 2oC to 3oC lower in
magnitude than at the actual concrete surface. In general, maximum T1,meas values were similar for
both the no blacktop and 50mm of blacktop cases, but the 95% fractile was about 1oC lower for
the case with the blacktop. Based on the 95% fractile results of Table 4.2 adjusted 2oC for the
difference between the point of measurement and the actual concrete surface, the appropriate
259
Central Texas values for positive thermal gradient are T1=16oC for positive gradient with no
blacktop, and T1=15oC for positive gradient with 50mm of blacktop. These are substantially
below the AASHTO LRFD values.

Table 4.2 Summary of maximum top fiber gradient magnitudes

Positive Positive Negative Negative
Gradient Gradient Gradient Gradient
No Blacktop 50mm No Blacktop 50mm
Blacktop Blacktop

Mainlane
Peak Measured 15.9oC 15.4oC -8.0oC -4.6oC
95% Fractile 13.9oC 12.8oC -6.5oC -3.2oC
Ramp P
Peak Measured 13.4oC 13.8oC -8.9oC -5.8oC
95% Fractile 11.7oC 11.9oC -6.8oC -4.3oC
AASHTO LRFD 25.6oC 20.0oC -12.8oC -10.0oC

The shape of the design gradient is of less importance, within reasonable limits, than the
proper selection of the T1 value because of the influence of the width of the deck. The positive
gradient shape originally recommended by NCHRP 276 [39] did appear to be more accurate than
the AASHTO LRFD simplified positive gradient shape for the mainlane superstructure, as can be
seen in Figures 4.97 and 4.98. The AASHTO LRFD shape more accurately represented the
positive gradients measured on ramp P. The measured positive gradients plotted in Figures 4.97
and 4.98 were typical for commonly recurring positive gradients, both with and without blacktop.
The box girder shape of ramp P was of more typical proportions to other segmental box girders
than the mainlane box girder, so the AASHTO LRFD may be the better shape, especially if the
dimension “A” from Figure 4.22 was adjusted to match the top of the girder web and bottom of
the fillet. A value for T3 in Figure 4.22 should be used in all cases. The value of T3 should be a
function of the appropriate deck gradient temperature T1 with no blacktop, 3oC being suitable for
central Texas. Although some warping of the girders was noted based on the measured strains,
the stresses predicted by using the AASHTO LRFD design gradient did correspond fairly well
with the measured values in the mainlane girder, when the proximity of the gauges to the surface
of the concrete had been accounted for. The calculated stress results for the ramp P girder thermal
gradient cases poorly predicted the measured stresses at locations away from the anchor segment
diaphragm. A reduction in the positive gradient magnitudes recommended by NCHRP 276 or
AASHTO LRFD for central Texas would be justified based on the temperature measurements.
260
Superstructure Depth (mm) 2000

1500

NCHRP 276

1000 AASHTO LRFD

Mainlane D5 from 4/20/95
Ramp P from 2/16/97
500

0
0 5 10 15 20 25 30

Temperature (Celsius)

Figure 4.97 Positive design gradients and measured gradients, no blacktop

2000
Superstructure Depth (mm)

1500
NCHRP 276
AASHTO LRFD
1000 Mainlane D5 from 6/17/96
Ramp P from 3/20/97

500

0
0 5 10 15 20 25 30
Temperature (Celsius)
Figure 4.98 Positive design gradients and measured gradients, 50mm of blacktop

261
 The stresses occurring from the maximum positive gradient cases with and without
blacktop may not have been equal, even though the maximum T1,meas values were similar.
Calculations usually assume that a stress free state exists at the point in time of the day when no
thermal gradient exists between the top flange and the webs. In actuality the state of stress would
not be known at any time, although the concrete would tend to creep toward a minimal state of
stress at some average state of thermal strain. This state of stress would be very difficult to
calculate, even if aided by measured strains. With this in consideration, it would be logical to
assume that a girder with higher daily temperature changes in the top flange, would have greater
thermal stresses generated. This was not reflected by the measured stress changes in Figures 4.52
through 4.55 because only the changes measured during the period of time from no gradient to
maximum positive gradient at deck level were presented. The maximum positive gradient
magnitudes were similar in each case, as were the changes in stress during the time periods under
consideration on April 20, 1995 and June 17, 1996.

Strains measured by some gauges in the top flange over the webs in both the ramp P
girder and the mainlane girder were very large, well into the plastic range for the concrete. This
behavior was only seen in the top flange over the webs, and the strains transitioned smoothly back
down into the elastic range at the other strain gauge locations in the top flange. This can be seen
in Figure 4.55 for the mainlane, and in Figure 4.33 for the ramp P girder. The irregular measured
stress distributions across the top flanges from the thermal gradient loads were not predicted by
the common design technique recommended in the AASHTO LRFD. Further research into
analysis techniques for thermal loads is urgently needed. Until an improved analysis is available,
the obvious change to the current AASHTO LRFD or NCHRP 276 [39] positive design gradient
T1 values documented herein should be deferred. Structures designed using these current design
positive gradients combined with current analysis techniques show no signs of distress.

Negative gradients
The shape and magnitudes for minimum negative thermal gradients did not compare well
with the current design gradients. The shape of the negative gradient is difficult to predict
because substantial negative gradients are often produced by extreme weather events. The
negative gradient caused by an extreme weather event is the superposition of the effects of the
event and the gradient existing previously. The shape is also greatly affected by the exposure of
the bottom of the wings, sides of the webs, and bottom of the bottom flange to the cold condition.
The magnitude of the negative gradient within the thickness of the top flange is more important
262
than the exact shape of the entire gradient when applied to a superstructure. The AASHTO LRFD
negative gradient shape, however, is much different from an actual negative gradient shape, and
would predict uniform curvature strains and subsequently self-equilibrating strains that may differ
substantially from those actually occurring.

The minimum T1,meas values without blacktop were about -9oC for ramp P and -8oC for
the mainlane superstructure. Measured daily minimum T1,meas values less than -5oC occurred
about 85% of the time without blacktop in place. Measured daily minimum T1,meas values less than
-7oC with no blacktop and below –5oC with 50mm of blacktop occurred 95% of the time, as can
be seen in Table 4.2. The AASHTO LRFD T1 value shown in Table 4.2 is derived by multiplying
the positive gradient value by -0.5. A more realistic value, based on the measured magnitudes
with no blacktop, would be -0.3 times the current AASHTO LRFD or NCHRP 276 T1 value. If
the previously recommended positive gradient design T1 values of 16oC with blacktop and 15oC
with 50mm of blacktop are adopted, then the negative gradient factor should be –0.4 with no
blacktop and –0.3 with 50mm of blacktop for the deck level. The shape and magnitudes from
NCHRP 276 could be used for the remainder of the gradient. During negative gradient
conditions, the entire exterior of the box girder is rapidly cooled. The deck is cooled faster than
the bottom flange or webs with the addition of a cold rain. The interior of the box girder is cooled
very little during this period, and may actually be increasing in temperature because of previous
heating of the girder. The depth of the girder that cools the slowest on average is in the fillet area
at the tops of the webs, and at the top of the bottom flange. The bottom surface of the bottom
flange undergoes a temperature drop only slightly smaller than that of the deck surface. These
characteristics were reflected in the measured gradient shapes in Figure 4.99.

263
 -0.30 x NCHRP 276 positive
gradient peak magnitude

2000
Superstructure Depth (mm)

1500
NCHRP 276
AASHTO LRFD
1000
Mainlane D5 from 1/19/96
Ramp P from 10/26/97
500

0
-14 -12 -10 -8 -6 -4 -2 0
Temperature (Celsius)

Figure 4.99 Negative design gradients and measured gradients, no blacktop

The magnitude of the measured daily minimum T1,meas values was reduced to -3oC or less
about 85% of the time after the addition of 50mm of blacktop, and -5oC or less about 95% of the
time, as seen in Table 4.2. The current recommended design negative gradient T1 value for the
U.S. 183 girders with 50mm of blacktop is -10oC, which would be very conservative. Based on
the distribution graphs in Figures 4.30 and 4.50, and the 95% fractile from Table 4.2, the design
negative thermal gradient T1 value with 50mm of blacktop also should be –5oC, which is slightly
less than –0.3 times the current AASHTO recommended positive gradient T1 value at deck level.
The NCHRP 276 shape and values should be used for the remainder of the gradient. The negative
gradients plotted in Figure 4.100 showed that the shape of the measured gradient was nearly the
same for the case with 50mm of blacktop and with no blacktop. The temperatures in the bottom
flange were predicted well by the NCHRP 276 negative gradient.

264
 -0.30 x NCHRP 276 positive
gradient peak magnitude

2000
Superstructure Depth (mm)

1500
NCHRP 276
AASHTO LRFD
1000 Mainlane D5 from 10/27/97
Ramp P from 10/26/97

500

0
-14 -12 -10 -8 -6 -4 -2 0
Temperature (Celsius)

Figure 4.100 Negative design gradients and measured gradients, 50mm of blacktop

The measured stresses for the negative gradient case from November 11, 1995 shown in
Figures 4.56 and 4.57 compared well with the calculated stresses using the measured gradient.
The stresses calculated using the current AASHTO LRFD gradient were too conservative. The
calculated stresses for the negative gradient case in the ramp P girder did not compare well to the
measured stresses at any location. The calculated bottom flange stresses for ramp P in Figures
4.35, 4.36 and 4.37 showed that the AASHTO LRFD negative gradient predicted too much
compression in the bottom fibers from uniform curvature induced stresses. The stresses
calculated from the measured gradient gave proper emphasis to the self-equilibrating stresses
actually occurring, and predicted tension in the bottom fibers. The shape originally recommended
by NCHRP 276 was more realistic than the AASHTO LRFD shape, particularly in the bottom
flange and lower part of the webs. In general, the calculated stresses using the AASHTO LRFD
design negative gradient were conservative in the simple span mainlane girder, and
unconservative as well as unrealistic in the continuous ramp P girder. Based on this observation,
analytical study of the structural response of box girders to thermal loads is urgently required.

265
Until an improved analysis is available the obvious change in negative design gradient T1 values
recommended and documented herein should be deferred.

4.4.2 Superstructure thermal gradients for transverse design

Positive gradients
Positive thermal gradient magnitudes measured through the thickness of the top flange at
points between the webs were nearly as large as the positive gradient magnitudes through the
depth of the entire box girder. The maximum measured positive gradient magnitude in the top
flange of the mainlane was about 17oC, both with 50mm of blacktop and without. The
extrapolated temperature difference between the top fiber and bottom fiber was almost 20oC.
Based on the statistical distribution of top fiber temperature gradients through the depth of the
girder, the 95% fractile positive gradient magnitude would be about 18oC without blacktop, and
17oC with 50mm of blacktop. The positive gradients through the webs and bottom flange of the
mainlane were much less in magnitude, which would be expected since these elements did not
have direct exposure to the sun for most or all of the day. The web and bottom flange positive
gradient magnitudes occurring with the maximum positive gradient in the top flange were less
than 4oC, and could be ignored for design purposes. The maximum measured positive gradient
magnitude in the top flange of the ramp P girder was substantially less than that of the mainlane
because the spring of 1997 did not produce positive gradients as large as the two previous springs.
Also, the ramp P box girder top flange thickness was only 203mm, compared to 254mm on the
mainlane girder.

The analytical method used to calculate stresses from the measured positive gradients
tended to overestimate stresses for the mainlane, probably because of the longitudinal cracks at
the top of the webs, and underestimate stresses for the ramp P girder. Also, this two-dimensional
method of analysis did not consider the relationship of transverse stresses and longitudinal
stresses. Measured top flange stress changes from the positive gradient were about 12% of the
allowable compressive stress in the mainlane girder, and 57% of the allowable in the ramp P
girder. Calculated top flange stresses using the measured maximum positive gradients were
similar in both girders at about 30% of the maximum allowable compressive stress in the
concrete. These high percentages indicate that the positive gradient through the thickness of the
top flange should be considered as a design case.

266
 The design values for the thermal gradient load case for transverse design of box girders
are not explicitly given in the AASHTO Guide Specification [23], AASHTO LRFD Guide
Specification [9], or the AASHTO Segmental Guide Specification [8]. The NCHRP 276 positive
gradient shape for beam and slab bridges is plotted in Figure 4.101, along with the measured
positive gradients and a proposed positive top flange gradient for design in Central Texas. The
proposed positive gradient has a peak deck level magnitude of 18oC for no blacktop and 17oC
with 50mm of blacktop. This peak temperature decreases linearly at a gradient of -0.072oC/mm
for no blacktop and –0.068oC/mm for 50mm of blacktop. Thus, the bottom fiber of a 250mm
thick top flange will be 0oC. Top flanges thicker than 250mm can be assumed to have a linear
temperature gradient of 18oC or 17oC, depending on the presence of blacktop. The measured
gradient shown in Figure 4.101 is nearly parallel to the proposed gradient. The temperature at the
bottom gauge location in the measured gradient plot was arbitrarily assigned to zero. Use of the
proposed gradient results in a decreased total temperature change over the top flange depth for top
flanges thinner than 250mm, and also provides a realistic average temperature change in the top
flange, based on the measured data. The NCHRP 276 temperatures for beam and slab bridges are
plotted in Figure 4.101 for comparison purposes because the gradient magnitude through the
thickness of the top flange was measured to be more severe than the gradient through the same
depth of concrete directly over the webs. The NCHRP 276 positive gradient shape for box girders
was intended to be applied down the centerline of the webs.

250
Top Flange Depth (mm)

200 Mainlane D5 from

5/20/95
150 Proposed positive
gradient, no blacktop
100 NCHRP 276 for beam
bridges
50 NCHRP 276 for box
girders

0
0 5 10 15 20 25
Temperature (Celsius)

Figure 4.101 Recommended positive top flange design gradient for transverse design

267
Negative gradients
The magnitude of the measured negative thermal gradient through the thickness of the
top flange of the mainlane box girder was more severe than the minimum T1,meas for the entire
depth of the girder. The minimum measured negative gradient in the top flange was about -12oC
in the mainlane girder and –9oC in the ramp P girder with no blacktop. Based on the statistical
distribution of top fiber temperature gradient magnitudes through the depth of the mainlane and
ramp P girders, the 95% fractile negative gradient magnitude would be about -10oC without
blacktop, and -5oC with 50mm of blacktop for the mainlane 254mm top flange. Significant
tensile stress changes were measured in ramp P from the negative gradient, with a maximum top
fiber stress change of 6.4MPa in tension. This tensile stress would be important when designing
the transverse prestressing force and eccentricity for the top flange. The proposed top flange
negative gradient for Central Texas in Figure 4.102 has a peak deck level temperature of –10oC
for no blacktop and –5oC with 50mm of blacktop. The slope of the proposed negative top flange
gradient, at 0.040oC/mm for no blacktop and 0.020oC/mm for 50mm of blacktop, is slightly less
conservative than the slope of the measured gradient in Figure 4.102. Top flanges thicker than
250mm can be assumed to have a linear temperature gradient of -10oC or -5oC, depending on the
presence of blacktop.

250

200 Top Flange Depth (mm)

Mainlane D5 from 150

1/18/96
NCHRP 276 100

Proposed negative
gradient (no blacktop) 50

0
-15 -10 -5 0
Temperature (Celsius)

Figure 4.102 Recommended top flange negative design gradient for transverse design

268
 The measured negative gradients through the thicknesses of the flanges and webs
differed from the positive gradient case in that the negative gradients through the bottom flange
and webs were of a significant magnitude. Since the negative gradients measured in the webs and
bottom flange were quite large, a negative gradient in these elements should be considered for
design. The magnitude of the web and bottom flange gradients was about 75% of the top flange
negative gradient when no blacktop was in place. The addition of blacktop had no influence on
the gradients in the webs or bottom flange. The proposed negative gradient in Figure 4.103 is
derived from the proposed negative gradient for the top flange with no blacktop plotted in Figure
4.102. The recommended peak magnitude temperature in Figure 4.103 is 75% of the magnitude
recommended for the top flange at deck level with no blacktop. The slope of the gradient is
0.030oC/mm for webs and flanges up to 250mm thick. The overall magnitude of the temperature
gradient predicted using the proposed gradient in Figure 4.103 compared well to the measured
temperature difference across the webs and bottom flange. The measured gradients were not
linear, as in the top flange, and therefore the proposed gradient would give slightly unconservative
moments from the linearly changing component of the gradient, and slightly conservative
moments and axial forces from the constant component of the gradient.

Mainlane web gradient

from 1/18/96
Mainlane bottom flange 400
gradient from 1/18/96
Proposed gradient in 350
Web or Bottom Flange
Thickness (mm)
web
300
Proposed gradient in
bottom flange 250
NCHRP 276
200
150
100
50
0
-10 -8 -6 -4 -2 0
Temperature (Celsius)

Figure 4.103 Recommended web and bottom flange negative design gradient for
transverse design

269
4.4.3 Pier gradients
The measured thermal gradients in the piers under study were found to be less severe
than those measured in the superstructures for a given day. The positive gradient in the large
ramp pier P16 had peak magnitude on the west face of the pier at the end of the afternoon.
Substantial positive gradients were measured on any face of the pier that had sun exposure,
including the east face. Positive gradient magnitudes measured with the sun exposure on the east
face were about 80% of the magnitude of the maximum positive gradient that was to occur later in
the day. The positive gradient existing when the sun exposure was on the south face was only
about 50% of the maximum.

The magnitudes of the positive thermal gradients measured in pier P16 were only about
75% of those T1,meas values measured in the superstructure girders on the same days. The
superstructure gradients were higher because the lower portions of the girders were shaded most
of the day, while the pier received direct sunlight whenever skies were clear. The shape of the
positive gradients in pier P16 was about the average of the NCHRP 276 gradient and the
AASHTO LRFD gradient, with the peak magnitude reduced by half. The recommended shape for
a positive design gradient, based on the results from pier P16, is shown in Figure 4.104. The
recommended positive gradient shape is the same as recommended for the superstructure in
Section 4.4.2 without blacktop, but with the entire gradient shape is multiplied by 75%. The
temperature on the far side from the sun exposure should remain at 2.8oC. Until further
refinement of the common analysis technique used for finding stresses from thermal gradients has
been performed, this measurement based gradient cannot be recommended. Instead, 75% of the
NCHRP design positive gradient for no blacktop should be used, but with the far face temperature
remaining at 2.8oC. The peak magnitude of this recommended shape is about 6oC higher than the
maximum measured magnitude for three reasons. First, the time period over which data were
taken may not have produced the highest possible positive gradient, although the spring of 1996
did produce substantially higher positive gradients in the mainlane girder than the spring of 1997.
Second, the thermocouples used to measure the concrete temperatures were placed 25mm under
the actual surface, so surface temperatures were probably higher on the exterior and lower on the
interior than presented. Third, the method commonly used by design engineers to determine
stresses from the one dimensional positive gradient consistently was found to underestimate the
peak stresses to an actual gradient when compared to the measured results, and to calculated
results using more refined design methods.

270
 2000
Pier Width (mm)

1500
Pier P16 from April 1996
Proposed gradient
AASHTO LRFD
1000
NCHRP 276

500

0
0 5 10 15 20 25 30
Temperature (Celsius)

Figure 4.104 Recommended positive design gradient for piers

The negative gradients measured in pier P16 were not effected by the orientation of the
sun to any extent. Also, negative gradients of significant magnitude did not occur as often as
large positive gradients. The peak magnitude of the negative gradients was nearly as large as
those seen in the superstructure, and the shape was predicted well by the NCHRP 276 negative
gradient shape, with the exception of the peak magnitude. The same negative design gradient
(with –7oC at deck level) is recommended for piers, based on the measured results from pier P16,
as is recommended for superstructure box girders without blacktop in Section 4.2.2. Once again,
the original NCHRP 276 values should be used until the analysis technique can be improved.

271
 2000

Pier Width (mm)

1500
Pier P16 from April 1996
Proposed gradient
AASHTO LRFD 1000
NCHRP 276

500

0
-14 -12 -10 -8 -6 -4 -2 0
Temperature (Celsius)

Figure 4.105 Recommended negative design gradient for piers

Thermal gradients in the solid sections of the mainlane pier and the ramp pier P16 capital
were slightly smaller than those measured in the voided sections. Although the concrete in the
core of these solid sections did not warm or cool appreciably over the course of the day, the
exterior of the concrete was also not able to be heated or cooled to the extent of the voided piers
because of heat loss to the core concrete. The net effect was a slight reduction in thermal gradient
over the voided pier. Since the mainlane pier was not instrumented with the intent to define a
thermal gradient shape, and the ramp pier capital segment PC16-8 was not part of a continuously
solid vertical pier, the best recommendation for design thermal gradients for a solid pier would be
those already given for the voided section. The peak magnitudes were found to be similar for
both the voided and the solid sections. With the massive core of the solid section piers, the
stresses of primary concern would be the self-equilibrating stresses on the surface of the concrete.
These are determined largely by the magnitude of the gradient, and not by the shape of the
gradient. The shape of the gradient is more important for the voided section, where large areas
away from the neutral bending axis of the section increase the proportion of bending stresses to
self-equilibrating stresses.

272
 Positive thermal gradients tend to cause compressive stresses on the surface of the
concrete, while negative gradients cause tensile stresses on the surface of the concrete.
Compressive stresses on the mainlane pier D6 from a positive gradient were measured as large as
2.7MPa, or about 0.1fc' which was higher than the total axial stress from the superstructure at
1.9MPa. Compressive stresses on the ramp pier P16 were calculated to be as high as 3.9MPa
from the measured peak positive gradient, or about 0.2 of the allowable compressive stress in this
post-tensioned concrete. Since the design of the pier for balanced cantilever construction
moments from the superstructure was dominated by control of tensile stresses in the pier, the
additional compressive stress from the positive gradient was of no consequence, even during
construction. Tensile stress from the maximum measured negative gradient was calculated to be
-2.3MPa on the surface of pier P16. This tensile stress was calculated to cause tension in the pier
concrete only during the out of balance cantilever construction load case. Significant negative
gradients occur infrequently, so the addition of the negative gradient load case to the construction
loads may not be necessary, depending on the bridge and the duration of construction.

In general, the importance of the addition of compressive stress from the thermal
gradient load case would only be important for prestressed pier designs of uncommon shape or
proportion. The decision to add the thermal gradient load case to the design should be made by
the engineer. Similarly, the importance of thermal induced tensile stresses in a compression
member is of little importance, except for special load cases. The probability of a substantial
negative gradient occurring during the special load case, especially during construction, should
also be determined by the engineer.

Cracking on the piers or superstructure members from daily thermal loads was not
observed anywhere on the project. Cracking may occur in the future because of concrete fatigue,
given the daily stress range on the concrete surfaces. Since the stresses in the concrete from
temperature, as well as shrinkage, are a function of exposed area and volume of the pier, a pier
design that does not include the thermal gradient case should use a nominal amount of reinforcing
steel in each direction and on each face of the concrete that is a function of both surface area and
concrete volume. The AASHTO LRFD [9] gives Equation 5.10.8.2-1 for computing nominal
steel percentages that meets this requirement, although for members with least dimension less
than 1220mm. This equation requires a minimum steel area on each face of a member of
0.379Ag/fy, where Ag is the area of concrete in mm2 between bar spacings and fy the yield strength
of the reinforcing steel in MPa. The 265mm2/m reinforcing requirement in the AASHTO Guide
273
Specification [23] is primarily intended for walls, and may be totally insufficient for controlling
local tensile stresses in more massive elements.

The 470mm2/m of transverse reinforcement used in the large ramp piers, including P16,
was inadequate for controlling temperature gradient induced stresses during the concrete curing
period. The transverse steel area used was almost twice that recommended by AASHTO [23],
and was twice that used in the original design of the piers. The AASHTO LRFD minimum steel
requirement dictates a steel area of 372mm2/m, which would also have been inadequate in
controlling the curing stresses. The negative thermal gradient magnitude measured in segment
PC16-8 during curing was -35oC. The high strength and resultant reactivity of this concrete
caused this extreme gradient. A lesser negative gradient measured while the segment was in the
form was severe enough to cause cracking in the young concrete before the form was removed.
The presence of these cracks could be detrimental in a corrosive environment to a precast pier
made otherwise of very high quality concrete, since the cracks tend to open with time from
shrinkage. Concrete with initial reactivity reduced by the use of fly ash, lower design strength, or
other methods such as cooling may prove more beneficial to the structure than the use of a more
efficient section and high early strengths. The amount of mild reinforcing steel used for
temperature and shrinkage crack control should be increased for piers made with more reactive
concrete, since the thermal gradient during curing may be a function of the design strength.

4.5 CONCLUSIONS
The following conclusions have been made based on the measured thermal gradients in
the various structural elements under study, and the measured response of the structure to these
gradients and the current AASHTO LRFD design thermal gradients.

4.5.1 Superstructure gradients for longitudinal design

Positive thermal gradients
1. The peak measured positive gradient deck level magnitudes (T1,meas) were similar for both the
cases without and with 50mm of blacktop, although thermal data without blacktop were
limited.

2. The NCHRP 276 or AASHTO LRFD recommended design positive gradient T1 value for
girders with 50mm of blacktop accurately predicted peak gradients at 20oC, but the
recommended T1 value without blacktop at 25.6oC was somewhat high, based on the limited

274
 measurements. The distribution of daily positive gradient deck level magnitudes over time
indicates that the thermal gradient case without blacktop deserves a higher design gradient
magnitude (at T1=16oC) than the case with 50mm of blacktop (at T1=15oC).

3. The AASHTO LRFD recommended positive gradient shape more accurately represented the
measured positive gradient shape of the ramp P girder, and the NCHRP 276 recommended
positive gradient shape better represented the shape measured on the mainlane. In either
case, a temperature gradient in the bottom slab should be considered, 3oC being suitable for
the soffit gradient magnitude in central Texas.

4. Calculated stresses using the AASHTO LRFD recommended design positive thermal gradient
compared well to stresses measured on the mainlane girder, but compared poorly and
unconservatively to stresses measured on the ramp P girder, even though the design gradient
T1 value was larger than the actual gradient. Evidence of sectional distortion or warping was
measured in every thermal gradient case. Also, soon into the life of the girders, high strains
were measured in response to thermal gradients in the top slab over the webs. These high
strains were measured in both the mainlane girder and the ramp P girder, and would indicate
plastic behavior in the concrete near the top of the heavy web reinforcement anchored in the
deck.

5. Based on the measured positive gradients alone, a reduction in the magnitude of the design
positive thermal gradient T1 value would be warranted in some cases. An analytical study of
the structural response to thermal loads needs to be performed before any reduction in the
design positive thermal gradients could be recommended. The effects of cross-sectional
shape, diaphragms, continuity, and potential plasticity should be considered in this study.

Negative thermal gradients

1. The shape of the AASHTO LRFD recommended negative thermal gradient did not compare
well to measured negative thermal gradient shapes in either the mainlane girder or the ramp P
girder. The negative gradient shape recommended by NCHRP 276 better represented the
actual shape of the negative gradient, especially in the lower part of the cross section.

2. Based on the measured negative thermal gradients, the peak top fiber gradient temperature T1
recommended by AASHTO LRFD or NCHRP 276 was too extreme. The peak top slab
negative gradient temperatures were closer to -0.3 times the NCHRP 276 recommended
positive gradient temperatures for the appropriate case without or with 50mm of blacktop.
275
 All points of the negative gradient other than the top fiber temperature would be represented
fairly accurately with the NCHRP 276 negative gradient shape. T1 values should be –7oC
without blacktop and –5oC with 50mm of blacktop in Central Texas.

3. Based on the unconservative calculated stresses in the ramp P girder when compared to the
measured stresses, no change to the current recommended design negative gradient from
NCHRP 276 can be recommended, pending further study of box girder response to thermal
gradients.

4.5.2 Thermal gradients for transverse design

1. Measured stresses from both positive and negative thermal gradients through the thicknesses
of the top slab, webs and bottom slab were large enough to warrant a design thermal gradient
for transverse design.

2. Based on the measured temperatures, a positive thermal gradient should only be applied to
the top flange, and the gradient shape should be linear. A recommended shape and
magnitude for the positive gradient is given in Section 4.4.2 for Central Texas.

3. Based on the measured temperatures, significant negative thermal gradients occurred

simultaneously in the top flange, webs, and bottom flange. The negative gradient should be
assumed to be linear in the top flange, as described in Section 4.4.2. The negative gradients
occurring in the webs and bottom flange should also be assumed to be linear, and can be
obtained by multiplying the magnitude and slope of the recommended top flange negative
gradient for no blacktop by 0.75.

276
4.5.3 Thermal gradients for the design of piers

1. Significant thermal gradients and thermal induced stresses were measured in the voided
segmental pier P16 and the solid mainlane pier D5.

2. A positive thermal gradient for the design of voided piers can be derived by multiplying the
entire recommended positive gradient shape for box girders in Section 4.5.1 with no blacktop
by 0.75, with the exception of the far fiber temperature which should remain at 3oC.

3. The recommended negative thermal gradient shape for superstructures with no blacktop from
Section 4.5.1 can be used for the design of voided piers.

4. Although few data were taken to define the shape of thermal gradients in solid pier sections,
the magnitude of the thermal gradients were measured to be similar to those of the voided
pier. The recommended thermal gradients for the design of voided piers in 2). and 3). above
are recommended for the design of solid piers until further studies can be done.

5. The stresses produced by the thermal gradients in the piers were mostly inconsequential for
pier design, except for one construction load case during the construction of the balanced
cantilever superstructure of ramp P. The decision to use a thermal gradient load case for the
design of piers should be made by the engineer.

6. Daily thermal induced stress changes in the piers were measured to be on the same magnitude
as those produced by the superstructure dead load. In order to control surface stresses and
concrete fatigue cracking, a nominal amount of transverse steel should be selected for a pier
based on both the concrete volume and surface area, such as by the AASHTO LRFD
Equation 5.10.8.2-1.

7. Negative thermal gradients that occurred during curing of the pier P16 segments were large
enough to crack the concrete segments while in the form. Negative thermal gradient
o
magnitudes were measured as high as -35 C shortly after removal from the form. An area of
transverse steel calculated by the AASHTO LRFD Equation 5.10.8.2-1 would not have been
enough to prevent the cracking, since the transverse area of steel actually in the pier segments
exceeded the amount found by this equation. A designer should consider the negative
gradient produced in higher strength concrete elements during curing. Transverse steel

277
should be increased to handle the thermal stresses, or provisions should be made to reduce the
heat of hydration.

278
279
Measurement Based Performance Evaluation of
a Segmental Concrete Bridge

Volume II

by
Rodney Tod Davis, B. S., M. S.

Dissertation
Presented to the Faculty of the Graduate School of
The University of Texas at Austin
in Partial Fulfillment
of the Requirements
for the Degree of

Doctor of Philosophy

The University of Texas at Austin

May, 1999

280
 CHAPTER 5
LOAD RESPONSE OF BOX GIRDERS

5.1 INTRODUCTION
The structural response of the U.S. 183 box girders and voided pier P16 during
application of dead loads, post-tensioning forces, and live loads was measured. The measurement
of the distribution of stresses across the width of the cross sections was the primary purpose for
the instrumentation. Four different superstructure box girders were studied. The first was the
mainlane three span semi-continuous unit D2. This unit was constructed using the span-by-span
method, which dictated that self-weight and post-tensioning forces be applied over a short period
of time after the girder had been assembled. The second was the five-span continuous ramp P
girder constructed in balanced cantilever. Construction of the span P16, the span under study,
occurred over several weeks because of construction staging. Application of segment dead loads
and post-tensioning forces occurred over very short periods of time though, making strain
measurements easy to interpret. A live load test was performed on both mainlane unit D2 and
ramp P. Live load tests were also performed on the two-cell box girder unit C15/L2, and the
three-cell box girder unit C13 to study the sharing of live load moments transversely from
unsymmetrical load cases. Finally, measurements were taken on segmental pier P16 during the
dead load application from a superstructure segment on Ramp P. The out of balance moment from
the installation of this final segment on the span P16 cantilever created large instantaneous stress
changes in pier P16.

5.2 PROBLEM AND TECHNICAL BACKGROUND

Box girder bridges must be designed differently than multiple girder bridges because of
potentially large horizontal shear deformations from bending moments, the distance post-
tensioning forces take to diffuse into the cross section, and the lack of redundant load paths in
single-cell box girders. Simplified conservative methods can be used to determine the distribution
of live load moments among girders in multiple girder bridges. A single-cell box girder must be
designed to take the entire live load moment on the span, regardless of the location of the load
across the girder width. Shear lag deformations will alter the stress distribution on the cross
section from that of a simple beam for live load, dead load, and post-tensioning forces, and will

279
change the moment diagram on continuous span box girder bridges. The box girder analysis must
include these considerations in order to accurately predict peak stresses.

5.2.1 Structural response of box girders to dead load and live load
The cross sectional shape of many box girders does not allow the designer to assume that
plane sections will remain plane during a longitudinal analysis of a girder. The top flange width
of a box girder is dictated by the roadway geometry above, while span lengths are dictated by pier
height and ground obstructions. Box girders frequently are quite wide when compared to their
span length. This proportion requires that the distance between webs, and the width of the girder
wings also be large when compared to the span length. Shear deformations in the top flange
become important to the overall bending behavior of the girder.

Figure 5.1 shows a typical U.S. 183 mainlane girder subjected to a point load at midspan.
The plan view shows that the deflected shape of the top flange is influenced by shear deformations
or shear lag, and tends to relieve longitudinal stresses at points in the flanges away from their
intersection with the webs. This effect is most prevalent over the part of the span under high
moment gradient, where shear is high. The portions of the flanges under higher bending stress
tend to have their stress relieved to some extent by shear deformations toward portions of the
flanges under lower stress. For a live load case on the simple span, shear is high over the entire
span length. For the dead load case, shear is high near the bearing reactions where the moment is
low, and bending stresses are low. Measurements on the mainlane girder D5 revealed the effect of
shear lag on the bending stiffness of the girder for the live load, dead load and post-tensioning
force cases. The mainlane girders would be considered quite wide for their span lengths. Their
span length to overall width ratio was about 2.3. Single-cell box girders with a span to width ratio
of 10 or larger are generally considered to be free of significant shear lag effects for both simple
and continuous span girders.

280
 Live load

Elevation

A
Deflected shape of top flange

A
Plan

Live load

Section A-A
Figure 5.1 Shear lag deformation in a simple span box girder

Shear lag effects can be more important in continuous structures. Both bending moment and
moment gradient tend to be highest at interior piers. Figure 5.2 shows a continuous single-cell
box girder subjected to a lane load. The pier reaction causes an abrupt point of negative moment
at the pier. The average top fiber stress in the girder, as calculated using beam theory, also peaks
abruptly at the centerline of the pier. Since plane sections do not remain plane in the girder
because of shear lag, the longitudinal stresses at Line A and Line B are not equal. The stresses at
Line A peak at the centerline of the pier because shear deformations within the web are small,
especially when compared to the shear deformations at Line B. The tensile stresses along Line B
are reduced from those at Line A, which effectively reduces the girder's bending stiffness,
changing the moment diagram, and increasing stresses at points in the cross section close to the
webs.

281
 A
Live load

A
Pier reaction

Elevation

Line A
Stress (Tension + )

Average stress
from beam theory

Line B

Top Flange Stresses

Live load

Line A
Line B
Section A-A
Figure 5.2 Top flange stresses in a continuous box girder

5.2.2 Diffusion of post-tensioning forces

Post-tensioning tendons in box girders are often located very near, if not in, the girder
webs. The diffusion of the post-tensioning forces into an entire span is effected by shear lag in the
girder, as a result of the girder shape, as was the case for dead loads and live loads. Since the
282
bending moments from the post-tensioning are designed to oppose the bending moments from
dead load and live load, the peaks in stress located near the girder webs caused by shear lag will
cancel themselves out to some extent. For this reason shear lag in post-tensioned box girders is
often not considered to be of any great consequence [26].

A more complicated problem is the calculation of stresses from post-tensioning forces at

points very near the anchorage. A designer often needs to know when the post-tensioning force
can be considered to be acting on the entire cross section. The force from the anchorage may have
to diffuse from a modified portion of the cross section into the typical cross section, such as forces
from an external tendon anchorage in a diaphragm. Also the concrete in the diffusion region may
be cracked, further complicating the problem. For these reasons, strut-and-tie modeling is often
used when designing anchorage zones. This truss analogy method will not easily predict cross
sectional stresses at points relatively near the anchorage.

Post-tensioning anchorages may also be located within the thickness of the normal cross
section of the bridge. St. Venant's Principle is often applied when trying to calculate the diffusion
of the post-tensioning force into the webs and flanges of a box girder. These members are
essentially plates loaded on edge. St. Venant's Principle allows the substitution of one loading
with another statically equivalent loading more convenient for calculating member stresses at
points beginning some distance away from the location of the original load. In the simplest
example of St. Venant's Principle, a plate of some small thickness, long length, and width b is
concentrically loaded with a point load on each end to produce an axial force along the length of
the plate. At a distance b away from the point load, the axial stress distribution is nearly uniform
across the width of the plate. The minimum stress is 0.973 of the average stress, and the
maximum stress is 1.027 of the average stress [42]. For practical engineering calculations the
point load could be replaced with a uniformly distributed load at a distance b away from the load.
It also follows that for a wide plate with evenly spaced point loads at spacing b, the cross sectional
stresses would be nearly uniform at a distance b away from the point loads. The cross sectional
shape of a box girder, and the location and spacing of anchorages complicates the problem
considerably from the previous two examples.

5.2.3 Behavior of multiple-cell box girders

Box girders may have one or more cells. Multiple cell box girders with evenly spaced
webs and a bearing at each web centerline behave only slightly differently than a multiple girder

283
and slab bridge. The transition spans on U.S. 183 occurred where the ramp girders intersected
with the mainlane girders. The design solution to this merger was a three-cell box girder with a
variable width center cell, and bearings located only under the outer cells. The result was a girder
that was neither a multiple-cell box girder in the traditional sense, nor the union of two single-cell
box girders with a gore closure between wing tips. The transition girder was difficult to analyze
for live load because loads on one side of the girder would be partially shared by the other side of
the girder. Since the transition spans were originally designed to have cast-in-place top and
bottom slabs between two precast single-cell and single-winged box girders, all the longitudinal
post-tensioning was located in the outer two cells of the girder. Figure 5.3 shows by a sketch of
the deformed shape how live loads on one side of the girder would be distributed to the other side
of the girder at points away from the piers. The amount of distribution of live load depends on the
transverse bending stiffness of the central top and bottom slabs, the torsional stiffness of the girder
cells, and the stiffness of the bearings.

Live Load

Section Near Piers

Live Load

Section Near Midspan

Figure 5.3 Response of a multiple-cell box girder to live load
284
 Twin single-cell box girders are commonly joined at the wing tips and post-tensioned
transversely. This allows any roadway geometry needed without the use of a longitudinal
expansion joint or two parapets at the median. Also, significant sharing of live load moments can
be realized. Figure 5.4 shows the cross section of a modified span on U.S. 183. A gore closure
connects the truncated wings of the ramp girder and the mainlane girder. As for the modified
spans, live loads on one girder can be distributed to the other girder at points away from the piers.
The amount of sharing depends on the longitudinal bending stiffness of each girder, the transverse
bending stiffness of the central top slab, and the torsional stiffness of each girder. The stiffness of
the bearings effects the stiffness of all these other components.

Live Load

Section Near Piers

Live Load

Section Near Midspan

Figure 5.4 Response of twin single-cell box girders to live load

5.3 TYPICAL SOLUTIONS, DESIGN CODE METHODS, AND PREVIOUS

MEASUREMENTS

A tremendous amount of resource material is available to engineers describing methods

of structural analysis for box girders. The methods discussed in this chapter are the most

285
commonly used methods that are widely used by engineers for box girder designs. The results
using these design methods are evaluated by comparing them to the measured data.

5.3.1 Analysis of box girder structures

The methods used for the analysis of box girder bridges range greatly in complexity.
Some bridges may be designed using a two-dimensional frame model, with only a rough
estimation of moment redistribution from creep. This type of bridge would be constructed by the
span-by-span method, and made continuous with closure pours at the piers or left as simple spans.
Bending moments in the girders are dominated by positive moments from the construction as a
simple span. In other box girder bridges, such as the three-span continuous bridge in Figure 5.5,
the two dimensional frame model would have to include provisions for creep and vertical
displacements of the girder at the bearing locations. This bridge would be built by the balanced
cantilever method, as indicated by the post-tensioning tendon profiles in the figure. Deflection
calculations must include creep in order to construct the bridge to the correct geometry. The
bending moments in the structure are tuned by continuity post-tensioning, bearing elevation
changes, and redistribution from concrete creep. The two dimensional frame model at the bottom
of Figure 5.5 has nodes located at all the points needed for the application of dead loads and post-
tensioning forces during construction, and at enough points to accurately model the changes in
cross section. The nodes are also located at the survey points used to track structural deflections
during construction. The cross sectional properties of bridges designed using a frame model are
usually modified for shear lag effects using one of the design code recommended methods.

Box girder bridges on a horizontal curve may have to be designed using a three
dimensional frame model, or a two-dimensional frame model with added consideration for the
torsional moments. Torsion effects need not be considered when the factored torsional moment is
less than one third of the factored torsional cracking moment of the girder [8]. One problem with
three dimensional frame models of box girder bridges is that the center of gravity of the girder in
bending most likely will not be at the shear center for the girder in torsion.

286
 Structure

Superstructure Element

Rigid Element
Bearing Element
Node

Foundation Boundary
Conditions

Model
Figure 5.5 Beam model of a box girder bridge

Box girders subjected to unsymmetrical loading or torsional loading can be analyzed using
several techniques, including folded plate methods, and analysis of the structure as a continuum.
For some box girder bridges certain load cases can be solved by resolving the loads into the
superposition of many loads. These individual load cases are easily solved using common design
techniques. Figure 5.6 shows a box girder bridge consisting of two single-cell girders connected
with a gore closure at their wing tips. The live load case is symmetrical to the bridge as a whole,
but will cause transverse bending, longitudinal bending and torsion in each of the girders. The
loads and boundary conditions are modified into the superposition of six easily solved cases (cases
2, 4, 6, 8, 10 and 11).

287
 P P Live Load + Impact
Near Midspan

DL1 + DL1 +
DL2 DL2
Elastomeric
Cast-in-place
Bearings
Closure

Case 1 Bearing Reactions

P

DL1 +
DL2

Case 1 = Case 2 + Case 3

P P

Case 3 = Case 4 + Case 5

P/2 P/2 P/2 P/2

M/2 M/2 M/2 M/2

Case 5 = Case 6 + Case 7

P/2 P/2
L

T M/2 T M/2
T=M+PL/2

Case 7 = Case 8 + Case 9

P/2 P/2

M/2 T M/2 Brace

Rigid Brace Reactions

Case 9 = Case 10 + Case 11

Figure 5.6 Analysis of twin single-cell box girders using superposition
288
 Some bridges, because of their geometric complexity, can only be modeled accurately using a
continuum model. In modern times the finite element method has become the predominant
method used to solve continuum models. The method lends itself to computerization because it
inherently accepts the general case. The shape functions are always polynomials, not a family of
custom functions used by other methods designed for hand solutions, so the stiffness matrix of
each element is easily generalized. Also, the global stiffness matrix is directly assembled from the
element stiffness matrices, and is symmetric. General boundary conditions are easily applied after
the matrix is assembled, unlike the finite difference method. The continuum is divided into many
finite elements, with a grid across the girder cross-section and along the girder length. Figure 5.7
shows a bridge that may need to be designed using a finite element program. The bridge geometry
includes a horizontal curve, skewed piers and bearings, and a light rail live load on one side. The
girder is unsymmetrical with inclined webs and external post-tensioning. One benefit of using the
finite element method is that stresses in the D-zones are known, and can be used to develop STM
for use in the design of the D-zones such as anchorage diaphragms.

A C
L Pier
CL Pier

Plan - Curved Girder

Light
Roadway Rail Ped.

Section A-A
Figure 5.7 Finite element model of a box girder bridge
289
5.3.2 Design code methods of analysis
Design codes, such as the AASHTO Guide Specification for the Design and Construction
of Segmental Concrete Bridges [8], do not rigidly specify the method of analysis to be used by the
engineer. However, design aids for analysis of a girder as a beam or part of a frame are given. To
compensate for shear lag effects, an effective flange width method is given in the AASHTO LRFD
Bridge Design Specification [9]. In this method portions of the flanges are removed if the ratio of
the flange width to effective span length exceeds 0.1. Figure 5.8 shows the modified shape of a
flange after using the method for simple span, continuous and cantilever girders. The girder
widths bs and bf are found by determining the original flange width b from Figure 5.9, dividing it
by the effective span length li from Figure 5.8, then selecting values from the graph in Figure 5.10.
The distance a is equal to b, the original flange width, but not to exceed 0.25l, a quarter of the
actual span length. This method tapers the flange width from bf at the central part of the span to bs
near the piers. Section properties are therefore continually changing over a portion of the span
near the piers, which may also be the case from a variable depth bottom flange. Other methods,
such as the Ontario Highway Bridge Design Code method [43], use a step instead of a taper,
somewhat simplifying section property calculations. Based on calculations and measurements by
Roberts [10], the stepped flange width change simplifies calculations and is sufficiently accurate.
Stress distributions in the flanges are determined using part c) of Figure 5.9 when using the
AASHTO LRFD method. The extreme fiber stresses calculated using the modified section
properties are assumed to be the maximum stresses, and occur only over the width of the webs.

290
 System Pattern of bm/b
Single-span Girder
a a
li = 1.0 l

bf
bs bs

l
Continuous End Span a 0.1l 0.1l
Girder li = 0.8 l
bs bf bf
bs
Interior Span
li = 0.6 l
l l
Cantilever arm a
li = 1.5 l
bs bf

l
Figure 5.8 Pattern of effective flange width coefficients, bf and bs (Figure
4.6.2.6.2-1 from the AASHTO LRFD Bridge Design Specifications [9])

291
 b1 b2 b2 b1
bm1 bm2 bm2 bm1

d0
a)

bm3 bm3

b3 b3

b1 b2 b2 b1

bm1 bm2 bm2 bm1

b) d0

bm3 bm3

b3 b3

bm1 bm2 Constant Stress bm2 bm1

b0 b0
in bm as a result
c) of flexure

Linear distribution of stresses in the top flange

Figure 5.9 Cross sections and corresponding effective flange widths, bm, for
flexure and shear (Figure 4.6.2.6.2-3 from the AASHTO LRFD Bridge Design
Specifications [9])

292
 0.7
b
For > 0.7 : bmf = 0.173 li
0.6 li
bms = 0.104 li
0.5
bf
b
li 0.4

0.3

0.2
bs
0.1
0.05
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
bm
b
Figure 5.10 Values of the effective flange with coefficient bm/b, for
the given values of b/li (Figure 4.6.2.6.2-2 from the AASHTO LRFD
Bridge Design Specifications [9])

The suggested method in the AASHTO LRFD Bridge Design Specification [9] for
determining the diffusion of post-tensioning forces is shown in Figure 5.11. The point loads are
assumed to diffuse into the cross section within a 60o cone. This means that for each distance b
away from the point load, the diffusion has spread approximately to width b. The method makes
calculations simple for determining the distance from the point load to a section of full diffusion.
The method will not predict the actual stresses in the cross section at points closer to the point
load. Figure 5.12 shows the actual distribution of stresses from two point loads on a double tee
[44]. The distribution is nonlinear and produces significant transverse compressive and tensile
stresses. The use of the AASHTO method for diffusion of point loads located away from the
girder webs in the flanges or in a diaphragm also needs verification.

293
 b1 b2 b2 b1
bn0 bn0
bn bn bn bn

Section A - A

bn0 bn0
A A
bn bn bn bn

30o 30o

Plan

Figure 5.11 Effective flange widths, bn, for normal forces

(Figure 4.6.2.6.2-4 from the AASHTO LRFD Bridge Design
Specifications [9])

294
 0.5 0.35 0.15

4.0

Post-Tensioning

x=8m Compression
2.95 1 5.6 1 2.8
Cross-Section x=4m

x=2m

x=1m

x = 0.5 m Tension

x=0m 1 kPa

10 kN

10 kN x=1m Tension
Longitudinal
Normal Stresses Compression
x = 0.5 m

1 kPa
Transverse
Splitting Stresses
Figure 5.12 Example of the effects of the diffusion of post-tensioning forces (after
Kristek [44])

5.3.3 Previous measurements

Measurements by Roberts [10] on the box girders of the San Antonio Y were very similar
to the measurements taken at U.S. 183. The proportions of the girders tested at San Antonio were
also similar to those at U.S. 183. Measured stress distributions across flange widths were
compared to stresses calculated using the AASHTO effective flange width method, beam theory
with plane sections remaining plane, and a finite element computer program called SHLAG
developed by Song and Scordelis [45]. The SHLAG program computes stress distributions with
due consideration for shear lag in a generalized box girder cross-section. Roberts found that using
the beam theory approach would not predict the peak box girder stresses. On average the stresses
calculated using beam theory were only 47% of the measured stresses. The AASHTO effective

295
flange width method predicted stresses 72% of peak measured stresses on average. Roberts found
the method needlessly complex and suggested another method that uses a stepped flange instead
of the tapered flange of AASHTO. The SHLAG program gave the best prediction of peak
stresses, at 84% of the measured peak stresses. In general the measured stress distributions near
the anchorage zones were highly irregular. Measured deflections of the girders tested were
adequately predicted using the section properties from the AASHTO effective flange method.

5.4 DATA COLLECTION, ANALYSIS AND RESULTS

Measured strains, converted to stresses, and deflections were compared to results
calculated using the AASHTO effective flange width method. The girders tested varied greatly in
span to width ratio, so the effect of shear lag also varied greatly.

5.4.1 Mainlane girder

Three-span continuous mainlane unit D2 was tested during the application of post-
tensioning, dead load and live load. The central span of the unit, span D5, was instrumented with
strain gauges at two cross sections. One section was near the anchorage segment, and the other
section was at midspan. Deflections were measured in all three spans of the unit.

Post-tensioning and dead load response

The post-tensioning tendon locations in span D5 are shown in Figure 5.13. Tendons were
stressed immediately following the assembly of the segments on the erection trusses. Stressing
began with the deviated external 19 - 15mm diameter strand tendons 1 through 6. These tendons
were stressed in symmetrical pairs. Internal wing tendons 15 through 18 were stressed next to
help control top flange tension during the stressing of the bottom slab tendons. The bottom slab
internal tendons, also 19 - 15mm diameter strand tendons, were stressed last in symmetrical pairs.
Measurements were taken at intervals during the entire post-tensioning sequence, including
deflections.

296
 3 and 7 -15mm dia strand wing tendons (15 - 18)
19 - 15mm dia. strand external deviated tendons (1 - 6)
19 - 15mm dia. strand bottom slab tendons (7 - 14)
A

D5-9 D5-16
Deflection
measurement Up Station A
locations

18 17 16 15

6 5 4 3 2 1

1413 121110 9 8 7

View A-A
Figure 5.13 Mainlane girder D5 tendon locations

The measured longitudinal stresses in segment D5-16 from the post-tensioning of tendons 1
through 6, at the section immediately beyond the anchorage diaphragm, are shown in Figure 5.14.
Strain gauges were located across the entire section, but are all plotted on the half section in the
figure. Therefore, two stress changes may be shown at the same gauge location in the half section.
Since the numerous jacks supporting the span continuously along its length were not adjusted
during stressing, the stresses plotted include post-tensioning stresses and some dead load stresses.
The measured top flange stresses in Figure 5.14 indicate that little diffusion of the post-tensioning
forces has occurred at this cross section, which was 1525mm from the anchor plates. Tendons 1
and 6 caused higher stresses toward the wing of the girder, and tendons 3 and 4 caused higher
stresses toward the center of the girder, as would be expected. The heavy end diaphragm
containing the anchorages was able to distribute compressive force to the middle of the bottom
slab. Stress changes were small in the bottom slab where the slab meets the web. From inspection
of the lower plot in Figure 5.14, plane sections did not remain plane through the height of the web.

297
 2

(+ compression, - tension)
1.5
Stress (MPa)

0.5

-0.5

3 2 1 Segment D5-16
Stressing External Deviated
19 - 15mm dia. strand tendons

Legend
(+ compression, - tension)

1 Tendons 3 and 4
Tendons 2 and 5
Stress (MPa)

Tendons 1 and 6
0
Strain gauge

-1
-1
1

Stress (MPa)
(+ compression, - tension)

Figure 5.14 Longitudinal stresses in segment D5-16 from the tensioning of external
tendons 1 through 6
298
 Figure 5.15 shows the measured longitudinal stresses in midspan segment D5-9 from the
tensioning of external tendons 1 through 6. The stresses are much more uniform across the width
of the flanges at this section than in segment D5-16. The lower plot in Figure 5.15 also indicates
that plane sections remained plane through the height of the web. The magnitude of the measured
stresses was slightly greater when tendons 1 and 6 were stressed than when tendons 3 and 4 were
stressed. Tendons 1 and 6 were located closer to the girder webs, and therefore had a shorter
distance to full diffusion of their post-tensioning force into the cross section. As a result, tendons
1 and 6 were more effective in producing negative bending moments in the girder to balance
positive dead load bending moments.

299
 (+ compression, - tension)
0.5
Stress (MPa)
0

-0.5

3 2 1
Segment D5-9
Stressing External Deviated
19 - 15mm dia. strand tendons
(+ compression, - tension)

2 Legend
Tendons 3 and 4
Stress (MPa)

Tendons 2 and 5
Tendons 1 and 6
1
Strain gauge

0
2

Stress (MPa)
(+ compression, - tension)

Figure 5.15 Longitudinal stresses in segment D5-9 from the tensioning of

external tendons 1 through 6

300
 Measured cross sectional stress changes in segment D5-16 from the stressing of the eight
bottom slab internal tendons are plotted in Figure 5.16. The plots adjacent to the half sections in
the figure show stresses measured on both sides of the cross section. The bottom slab internal
tendons were stressed symmetrically in pairs. The gauges that measured the strains were located
1690mm from the anchorage plates, and 75mm beyond the heavy end diaphragm in the typical
cross section of the girder. The stress changes in the top flange were small compared to the stress
changes in the bottom slab, as would be expected given the location of the tendons. The top
flange stresses were higher over the webs, and lower in the center of the top slab and toward the
wing tips. The measured stress changes in the bottom slab were as high as 6.8MPa at the center of
the bottom slab. The stresses' changes caused by the stressing of each pair of tendons was not
uniform across the width of the bottom slab. The heavy end diaphragm had a large access hole
located at the center of the cross section, as can be seen in Figure 2.40. The diaphragm was more
effective in distributing stresses across the width of the bottom flange when tendons away from
the access hole, such as 7 and 14, were stressed. The stress gradient across the width of the
bottom slab from the stressing of tendons 10 and 11 was quite large at 4.9MPa. The stresses
plotted at the bottom of Figure 5.16 down the depth of the webs indicate that plane sections were
not remaining plane at this cross section. The stresses down the web depth were much closer to
linear when tendons 7 and 14 were stressed than the other tendons. Tendons 7 and 14 were
located near the bottom of the web.

301
 (+ compression, - tension)
1
0.5
Stress (MPa)

0
-0.5
-1

Segment D5-16
Stressing Straight Internal
19 - 15mm dia. strand tendons
10 9 8 7

7
Legend
Tendons 10 and 11
6 Tendons 9 and 12
(+ compression, - tension)

Tendons 8 and 13
5
Tendons 7 and 14
Stress (MPa)

4
Strain gauge

0
-1
4

Stress (MPa)
(+ compression, - tension)

Figure 5.16 Longitudinal stresses in segment D5-16 from the tensioning of

internal bottom flange tendons 7 through 14
302
 Measured cross sectional stress changes in segment D5-9 from the stressing of the eight
bottom slab internal tendons are plotted in Figure 5.17. These stresses in segment D5-9 near
midspan include some dead load stresses, as well as the post-tensioning stresses. The plots in the
figure show that the post-tensioning stresses were well distributed, and linearly distributed down
the depth of the webs. A small stress gradient remained across the width of the bottom slab.
Based on inspection of all three plots in Figure 5.17, all the tendons appear to be equally effective
in producing negative moment at midspan.

303
 (+ compression, - tension) 0.5
Stress (MPa)

-0.5

Segment D5-9
Stressing Straight Internal
19 - 15mm dia. strand tendons

Legend
10 9 8 7
3 Tendons 10 and 11
Tendons 9 and 12
(+ compression, - tension)

Tendons 8 and 13
2
Stress (MPa)

Tendons 7 and 14
Strain gauge

0
3

Stress (MPa)
(+ compression, - tension)

Figure 5.17 Longitudinal stresses in segment D5-9 from the

tensioning of internal bottom flange tendons 7 through 14
304
 Figure 5.18 gives a summary of the measured post-tensioning stresses in segment D5-16 at the
instrumented cross section. The stresses are plotted by tendon group, including the internal wing
tendons. The total stresses in the top flange were dominated by the stresses from the external
deviated tendons 1 through 6. Bottom flange stresses were dominated by the stresses produced
from the internal bottom slab tendons. Compressive stresses at the center of the bottom slab were
very high at 49% of the design compressive strength of the concrete. A high stress gradient was
measured across the width of the bottom flange. No tension was indicated at this cross section
from post-tensioning forces, although, without the wing tendons, tension would have occurred in
the wings. The application of dead load would change the measured stresses in Figure 5.18 very
little, since the dead load moment was small so close to the girder bearings. A summary of the
stresses in segment D5-9 produced by all the tendons is shown in Figure 5.19. The post-
tensioning stresses appear to be well distributed across the cross section, and nearly linear down
the depth of the webs. No tension was indicated at any point in the cross section, even though the
girder had not been released from the erection truss. Tension might have been expected in the top
flange near midspan since the girder dead load was not applied incrementally as tendons were
stressed.

305
 (+ compression, - tension) 6
4
Stress (MPa)

2
0
-2

16 15

3 2 1 Segment D5-16
Stressing of All Tendons

10 9 8 7

20 Legend
18 Tendons 1 - 6
(+ compression, - tension)

16
Tendons 15 - 18
14
Tendons 7 - 14
Stress (MPa)

12
10 Sum of All Tendons
8 Strain gauge
6
4
2
0
-2
14

12

10

-2
8

Stress (MPa)
(+ compression, - tension)

Figure 5.18 Longitudinal stresses in segment D5-16 from the tensioning of bottom
flange, wing and external tendons 1 through 18

306
 (+ compression, - tension)
Stress (MPa)
3
2
1
0
-1

16 15

Segment D5-9
Stressing of All Tendons
3 2 1

10 9 8 7
Legend
18
Tendons 1 - 6
16
(+ compression, - tension)

14
Tendons 15 - 18
12 Tendons 7 - 14
Stress (MPa)

10 Sum of All Tendons

8
Strain gauge
6
4
2
0
18

16

14

12

10

Stress (MPa)
(+ compression, - tension)

Figure 5.19 Longitudinal stresses in segment D5-9 from the tensioning of bottom
flange, wing and external tendons 1 through 18

307
 Measured stresses from post-tensioning forces and the full dead load moment for segment D5-
16 are plotted in Figure 5.20. These measurements were taken several hours after the completion
of post-tensioning operations. The measured stresses in the top flange indicate that some tension
might have existed at points over the webs. The large stress gradient over the width of the bottom
slab remained unchanged although the magnitude of the bottom flange stresses decreased about
1MPa. Dead load and post-tensioning stresses were calculated using the AASHTO [8] effective
flange width method and 30o diffusion of post-tensioning force method. The results from this
analysis are plotted as the solid lines in Figure 5.20. Post-tensioning forces were only assumed to
be active within a 60o cone propagating from the anchorage plate. Therefore, the plotted stresses
look like a step function. A designer could have assumed a smoother distribution of post-
tensioning stresses based on these calculated results. In general, the calculated results indicated
less diffusion of post-tensioning forces than was measured. The presence of the heavy end
diaphragm may have had an influence on the diffusion. The overall shape of the stress distribution
in the top flange was predicted well by the calculation, although the magnitude of the peak-
calculated stresses was much too high. The measured stress gradient across the width of the
bottom slab was poorly predicted by the 30o diffusion method.

308
 20
18
(+ compression, - tension) 16
14
Stress (MPa)

12
10
8
6
4
2
0
-2

16 15

3 2 1 Segment D5-16
Dead Load Plus Prestress

10 9 8 7

32 Legend
30
28 Measured DL+PS
(+ compression, - tension)

26
Calculated DL+PS
24
Stress (MPa)

22 Strain gauge
20
18
16
14
12
10
8
6
4
2
0

Figure 5.20 Measured and calculated longitudinal stresses in

segment D5-16 from post-tensioning and dead load

The measured and calculated post-tensioning and dead load stresses in segment D5-9
near midspan are shown in Figure 5.21. The AASHTO effective flange width method used in the
calculation predicted the measured stresses quite accurately, although the measurements show
more compression in the top flange and less compression in the bottom flange. The neutral axis of
the girder was probably slightly lower in the cross section than predicted by the calculation.

309
 (+ compression, - tension)
Stress (MPa)
6
4
2
0

16 15

Segment D5-9
Dead Load Plus Prestress
3 2 1

Legend
10 9 8 7
18
Measured DL+PS
16
(+ compression, - tension)

14 Calculated DL+PS
12
Stress (MPa)

Strain gauge
10
8
6
4
2
0

Figure 5.21 Measured and calculated longitudinal stresses in segment D5-9 from post-
tensioning and dead load

Measured deflections from the construction of the three spans of unit D2 are plotted in Figure
5.22. The camber from prestressing gives an indication of the effectiveness of each post-
tensioning tendon. Based on the measured deflections in Figure 5.22, the tendons located closer to
the webs were slightly more effective in cambering the girders. Some variation in deflection
occurred among the three spans from the application of post-tensioning and dead load. This
variation could be expected since the force in the numerous hydraulic jacks supporting the
segments on the truss were not monitored or adjusted during post-tensioning. Therefore the
percentage of dead load moment present during the different post-tensioning stages was not
known, and not easily predicted. The change in deflection from the point in time at which all
tendons were stressed with the girder remaining partly on the truss, and on April 7,1995 when all
dead load had been applied indicates that dead load was not applied consistently during post-

310
tensioning. Span D6 showed the greatest deflection change, an upward camber, and span D4
showed the least. Since the total estimated dead load deflection was 36mm, or 1/1050 of the span
length, the percentage of the total dead load applied immediately following post-tensioning was
between 21% and 33% for the instrumented spans. The bridge design did not specifically call for
the control of the application of dead load during post-tensioning of these spans. The erection
trusses were much less stiff than the concrete girder, so the trusses deflected upward during post-
tensioning without much loss in the dead load moment they were carrying. The hydraulic jacks
located on top of the truss, three supporting each segment, were locked off prior to post-tensioning
so their force could not be monitored or adjusted. Camber in the girders decreased between April
7, 1995 and October 4, 1995 by about 25%. At noon on a hot day the following summer, July 11,
1996, the camber had increased 40% from the October 4, 1995 measured camber because of a
positive thermal gradient.
60

50
Midspan Deflection in mm

40 Span D4

Span D5

30 Span D6

20

10

0
(All dead load applied)
Tendons 10 and 11
Tendons 1 and 6

October 4,1995

July 11,1996
Tendons 9 and 12
Tendons 3 and 4

Tendons 8 and 13

Tendons 7 and 14
Tendons 2 and 5

Tendons 15 - 18

All tendons
(on truss)

April 7,1995

Figure 5.22 Deflections from post-tensioning and dead load

(+ deflection is camber)

311
Live load test
A live load test was performed on the mainlane unit D2 using six dump trucks. The trucks
were loaded, such that when they were placed back to back at the center of a span, the peak live
load moment produced was identical to that from an HS20-44 truck. The weights and dimensions
of these trucks are given in Table 5.1. The trucks were placed both symmetrically and
unsymmetrically with respect to the centerline of the bridge in three lanes. The plan view of the
six live-load cases for the mainlane are shown in Figures 5.23 and 5.24.

Table 5.1 Axle weights and spacing for live load test trucks on Unit D2
Truck Weight of Rear Weight of Front Total Weight Axle Spacing
Axles (kN) Axles (kN) (kN) (mm)
1 122.5 43.1 165.6 5030
2 132.0 39.6 171.6 4850
3 119.6 44.4 164.0 4890
4 119.1 40.8 159.9 4090
5 135.1 39.0 174.1 4720
6 119.7 44.1 163.8 4360

312
 Front Axle Rear Axle
Weight Weight

Axle Spacing

Live Load Case 2 Typical Truck

38.90m 38.90m 38.90m

Span D4 Span D5 Span D6

17070mm
3650mm (Typ.) 6 2
3 1
3650mm (Typ.)
4 5

D4 D5 D6 D7
CL
3650mm Spacing between
rear axles
Live Load Case 3

6 2
3 1
4 5

D4 D5 D6 CL
D7
3650mm

Live Load Case 5

6 2
3 1
4 5

D4 D5 D6 D7
14325mm 16150mm

Live Load Case 6

6 2
3 1
4 5

D4 CL D5 D6 D7
3650mm

Figure 5.23 Live load cases–Mainlane girders D4, D5 and D6

313
 Live Load Case 8

38.90m 38.90m 38.90m

Span D4 Span D5 Span D6

17070mm
3650mm (Typ.) 6 2
3650mm (Typ.) 3 1
1830mm 4 5
D4 D5 D6 D7
CL
3650mm Spacing between
rear axles

Live Load Case 9

4 1 3
1830mm 5 6 2
D4 D5 D6 D7

3660mm 3660mm
12500mm 14330mm
Figure 5.24 Live load cases–Mainlane girders D4, D5 and D6 (continued)

The measured and calculated stresses in segment D5-9 for live load case 2 are given in Figure
5.25. Live load case 2 was a symmetric loading on span D5 intended to produce a maximum
positive moment. The measured stresses are lower than the stresses predicted using the AASHTO
effective flange width method. The measured stresses were calculated from the measured strains
using the elastic moduli found from the test prisms described in Chapter 2. These elastic moduli
were consistently found to be less than that of the actual moduli of the concrete in the girders. The
measured stresses down the girder web in Figure 5.25 were nearly linear, with a neutral axis
slightly below that predicted by the calculation. Also the distribution of measured stresses across
the width of the top flange was nearly uniform. Since integration of the measured stresses over the
cross sectional area must result in a live load moment equal to that applied, the modulus used to
find the measured stresses from the measured strains must have been too small. The measured
stresses were also less than the calculated stresses because the stiffness from the external deviated
tendons was not included in the analysis. Stresses from load case 8 in segment D5-9 are shown in
Figure 5.26. The measured stresses were nearly identical to those measured during load case 2
even though the load case was unsymmetric. Stresses measured in the right web, the web most

314
directly under the live load, and the right top and bottom flanges were only 10% to 15% larger
than those measured in load case 2. The calculated stresses shown for load case 8 are the same as
for the symmetrical load case 2. An accurate calculation for the longitudinal cross sectional
stresses in the girder for the unsymmetrical load case would be very difficult. The actual web
shear from torsion bending was effected by the connection of the top flange of the loaded girder to
the adjacent girder flanges, and the stiffness of the elastomeric bearings. The torsional moment
may have been taken by all three spans of the unit. Therefore, for purposes of comparison, the
calculated stresses for load case 2 were plotted in Figure 5.26.

315
 (+ compression, - tension)
1.5
Stress (MPa)

0.5

Segment D5-9
Live Load Case 2

0
Legend
Measured
-0.5
(+ compression, - tension)

Calculated
-1
Strain gauge
Stress (MPa)

-1.5

-2

-2.5

-3
-0.5

-1.5

-2.5

-3.5
1.5

0.5

-1

-2

-3
1

Stress (MPa)
(+ compression, - tension)

Figure 5.25 Longitudinal stresses from load case 2

316
 (+ compression, - tension) 1.5
Stress (MPa)

0.5

Segment D5-9
Live Load Case 8

-0.5 Legend
(+ compression, - tension)

-1
Measured from Load Case 8
Stress (MPa)

Calculated from Load Case 2

-1.5
Strain gauge
-2

-2.5

-3
-0.5

-1.5

-2.5

-3.5
1.5

0.5

-1

-2

-3
1

Stress (MPa)
(+ compression, - tension)

Figure 5.26 Longitudinal stresses from load case 8

317
 Deflections were measured during all live load cases using the taunt wire baseline deflection
measuring system described in Chapter 2. The deflection of the bearings was not included in the
presented deflections in the following figures. Figure 5.27 shows a plot of the measured
deflections and deflections calculated using the AASHTO effective flange width method for live
load case 2. The calculated deflection was only 10% greater than the measured deflection. The
difference between the calculated and measured stresses in Figure 5.25 would have predicted
greater error in the deflection calculation than 10%. The deflections for load case 3 shown in
Figure 5.28 gave nearly identical results. Load case 2 and load case 3 were maximum positive
moment cases for spans D5 and D6 respectively. Deflections in the unloaded spans during live
load cases 2 and 3 were essentially negligible, at less than 0.5mm at all but one point. The
maximum measured midspan deflections in cases 2 and 3 were only 1/6300 of the span length.

-1
Deflection in mm

-2 Legend

-3 Measured

-4
Calculated

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.27 Deflections from load case 2

318
 1

-1

-2
Deflection in mm

Legend
-3
Measured
-4
Calculated
-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.28 Deflections from load case 3

The test truck arrangement for load case 5 was intended to produce a negative moment couple
over pier D6 between the cast-in-place deck joint and the elastomeric bearings. The deflections
for load case 5 are plotted in Figure 5.29. The calculated and measured deflections were nearly
identical. The deflection calculation assumed that each of the loaded spans acted as a simple span.
Based on the differences between the calculated and measured deflections in load cases 2 and 3,
the measured deflections for load case 5 appear to be somewhat high. The deflected shape of the
spans also appears to be almost triangular. The presence of the front wheels of the test trucks
immediately adjacent to the deflection measuring plates at midspan may have increased the
measured midspan deflections over the actual average midspan deflections of the girders.

319
 1

-1
Deflection in mm

-2

-3
Legend
-4
Measured
-5 Calculated

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.29 Deflections from load case 5

Live load case 6 was similar to load cases 2 and 3, and was designed to produce a maximum
positive moment in span D4. The measured deflections for this span, shown in Figure 5.30, were
much smaller than the calculated deflections at only 1/10500 of the span length. The calculated
deflections were for a simple span, but based on the positive measured deflections in unloaded
span D5, some negative moment must have been developed in the superstructure over pier D5. In
this case a moment couple developed between the cast-in-place deck joint and the cast-in-place
fixity block between the bottom flanges of girders D4 and D5. The performance of the cast-in-
place joints is discussed in detail in Chapter 7.

320
 1

-1
Deflection in mm

-2

-3
Legend
-4 Measured
Calculated
-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.30 Deflections from load case 6

The deflections for live load case 8 are shown in Figure 5.31. Load case 8 was similar to load
case 2, but unsymmetrical with respect to the longitudinal centerline of the bridge. Load case 8
was intended to create maximum bending moment and torsional moment. The measured midspan
deflection for load case 8 was actually smaller that the measured midspan deflection for load case
2 by 5%, and 15% smaller than the deflection calculated for the symmetrical load case plotted in
the figure. The slight decrease in midspan deflection over load case 2 was probably due to the
high torsional rigidity of the girder combined with the deck continuity to the girder wings of the
adjacent spans. The measured deflection for load case 8 was 1/6700 of the span length. Live load
case 9 was also an unsymmetrical loading intended to create transverse bending in pier D6, and
negative superstructure moments over pier D6 if possible. Judging by the results of load case 5 in
Figure 5.29, no negative moment was developed. In load case 5 the calculated and measured
deflections were quite similar, but in load case 9 the measured deflections were 20% smaller than
the calculated deflections for symmetrically loaded simple spans (shown in Figure 5.32). The

321
cast-in-place deck joint may have stiffened the structure for this torsional case, and provided some
load transfer to span D4 through bending of the girder wings.

-1
Deflection (mm)

-2

Legend
-3
Measured
-4
Calculated
-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.31 Deflections from load case 8

322
 1

-1
Deflection (mm)

-2
Legend
-3
Measured
-4 Calculated

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 5.32 Deflections from load case 9

5.4.2 Ramp P girder

The balanced cantilever construction of ramp P provided the opportunity for many dead
load and post-tensioning force test cases. Measurements were taken at all stages of construction
and during a live load test. Strain changes were measured in segment P16-2 adjacent to the heavy
end diaphragm at pier P16, deviator segment P16-10 at the quarter point of the span, and segment
P16-17 at midspan.

Temporary post-tensioning
The temporary stress across a segment joint required during epoxying is 0.28Mpa, as
specified in the AASHTO Segmental Guide Specification [8]. The stress across the face of a
segment when epoxied during balanced cantilever construction is provided by temporary or
permanent post-tensioning and dead load moment from the weight of the segment itself. The
temporary force for epoxying the ramp P segments was provided by 5 short lengths of post-
tensioning bars anchored in blisters within the core of the girder (see Figure 5.33). Using this
method, all epoxying and temporary post-tensioning could be accomplished from the deck and the

323
core of the girder. Also, the short lengths of external post-tensioning bars easily accommodated
the curvature of the structure. To avoid passing temporary post-tensioning through the congested
heavy pier segment diaphragm, the cantilevering tendon ducts in the top flange over the webs were
used to assemble the first two segments in balanced cantilever onto the pier segment. The straight
temporary post-tensioning bars did not work well in the curved alignment of the internal
cantilevering post-tensioning ducts.

2.5
(+ compression, - tension)

2.0
1.5
Stress (MPa)

1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0

Temporary Post-
Tensioning Blisters
(+ compression, - tension)

2.0
Stress (MPa)

1.5
1.0
0.5
0.0
-0.5

Dead Load
& Temporary
Post-Tensioning
Stress Required
Across Joint
(0.28 MPa)

Concrete
2 1 0 Strain Gauge
Stress (MPa)
(+ compression, - tension)

Figure 5.33 Diffusion of force from temporary post-tensioning in segment P16-10

324
 When a segment was epoxied and stressed into place, the crane released the segment creating a
negative dead load moment from the weight of the segment. This negative moment was not
enough to provide adequate squeeze of the epoxy on the bottom flange joint, so a temporary post-
tensioning blister was provided on the bottom slab. Furthermore, the constructors soon realized
that the bottom flange temporary post-tensioning bar must be stressed first to nearly the required
force, before the top temporary post-tensioning bars were fully stressed, in order to close the joint
fully at the bottom flange. The measured stresses in segment P16-10 from the five temporary
post-tensioning bars and self-weight are shown in Figure 5.33. The measured stresses exceeded
the required joint stress for epoxying at all points except the wing tips and at one point in the
bottom flange. During casting of a segment, a thermal gradient is developed in the plastic newly
cast segment because of the heat of hydration generated by the match casting segment. This
gradient is present when the segment takes a set. When the segment cools to uniform temperature,
the segment takes a curved shape and no longer exactly matches the segment against which it was
cast. The segments tend to touch first at the wing tips. This “banana” shaped segment behavior is
more important for wider girders, such as the mainlane girders. Epoxy could be seen slipping
down the joint at the center of the top flange of the mainlane girders in some instances, indicating
no joint squeezing force was present at that location on the top flange. The distribution of
temporary joint stresses in the top flange of segment P16-10 in Figure 5.33 did not indicate
presence of a banana shape, which would have revealed itself as a stress increase near the
wingtips, as seen on the mainlane. Segment P16-10 could have benefited from an increase in
temporary joint force across the wing tips, and on the bottom flange.

Diffusion of cantilever post-tensioning

A plan and elevation of the half span cantilever P16 is given in Figure 5.34. Construction
began with the installation of the two piece pier segment, which was stressed onto the pier P16 to
provide a moment connection for the balanced cantilever construction. Construction proceeded
with the epoxying and temporary post-tensioning of segment P16-1 (not shown in Figure 5.34).
The epoxying and temporary post-tensioning of P16-2 balanced the bending moments in pier P16.
The pair of 12 strand 15mm diameter cantilever tendons T201 were then stressed to create the
positive bending moments required to balance the negative moments from the dead load of the
next two segments placed in cantilever. Construction proceeded by this method until midspan
segment P16-17 was epoxied into place.

325
 1.45 m 9 @ 2.87 m

T201 T202 T203 T204 T205 T206 T301 T302

P16-2 P16-4 P16-6 P16-8 P16-10 P16-12 P16-15 P16-16 P16-17

Elevation View (Horizontal:Vertical Scale = 1:2)

Center Line
of Pier P16

7.92 m
T201 T202 T203
( Before T204 T205 T206 T301
Barriers T302
Cast )

Radius of Girder at Center Line = 221 m

Plan View (Horizontal:Vertical Scale = 1:1)

Figure 5.34 Layout of segments and cantilever tendons in the P16 upstation cantilever

Figure 5.35 shows the measured and calculated stress distribution in segment P16-10
from the stressing of cantilevering tendons T205. The instrumented section is 2410mm away from
the anchor plates. The AASHTO Segmental Guide Specification method for the diffusion of post-
tensioning forces is used to predict the stress at this cross section. The method assumes diffusion
of the force within a 60o cone. The post-tensioning stresses are also calculated using full section
properties. The measured stresses indicate that the post-tensioning force is poorly diffused at this
cross section, with a nonlinear stress distribution down the depth of the web. Tension was
measured across the full width of the bottom flange. Peak stresses were predicted quite well using
the AASHTO 30o diffusion method, but the shape predicted by this method tends to overestimate
stresses at the center of the top flange and within much of the height of the webs. The AASHTO
30o diffusion method would tend to overestimate the negative bending capacity of this section for
service level stresses.

326
 2.5

(+ compression, - tension)
2.0
1.5
Stress (MPa)

1.0 Right Tendon

0.5 Stresses
0.0
-0.5
-1.0 Left Tendon
Stresses
-1.5
-2.0

Shaded area represents the

portion of the cross-section
assumed effective for T205 Cantilever
calculated stresses. Tendon
(+ compression, - tension)

Anchorages
2.0
Stress (MPa)

1.5
1.0
0.5
0.0
-0.5

Measured
Calculated -
Effective Flange
Width
Calculated -
Full Section

Concrete
Strain Gauge
2 1 0

Stress (MPa)
(+ compression, - tension)

Figure 5.35 Diffusion of post-tensioning force from the cantilever tendons in segment P16-10

Cantilevering final stresses

The measured and calculated stresses from all cantilevering dead load and cantilever
post-tensioning at the instrumented section in segment P16-2 are shown in Figure 5.36. The
instrumented section in segment P16-2 is 450mm from the face of the heavy diaphragm in the pier

327
segment. Full section properties were used to calculate the stresses at this section. From
inspection of the plots of measured stresses in the top and bottom flanges, no tension existed in the
cross section at the completion of the cantilever construction. Measured stress distribution was
nearly linear from both dead loads and post-tensioning forces. The calculated stresses were
somewhat unconservative since top flange compression from post-tensioning was overestimated,
and bottom flange compression from dead load was overestimated.

25
(+ compression, - tension)

20
15
Stress (MPa)

10
5
0
-5
-10
-15

30
(+ compression, - tension)

25
20
Stress (MPa)

15
10
5
0
-5
-10

Dead Load
Calculated
Dead Load
Prestress
Calculated
Prestress

Concrete
20 10 0 -10 Strain Gauge
Stress (MPa)
(+ compression, - tension)
Figure 5.36 Longitudinal stresses in segment P16-2 after completion of the
P16 upstation cantilever
328
 The cumulative stress changes from three gauges in segment P16-2 are plotted in Figure 5.37 as
a function of construction sequence. A small amount of tension was present in the bottom flange
at gauge C638 during the early part of the construction sequence. Tension in the bottom flange
over the pier was of little consequence since this section would be in substantial compression by
the end of construction. Top flange tensile stresses were not measured at any time at gauge
locations C604 or C607. The balance between tensile stress changes from dead load and
compressive stress changes from cantilever post-tensioning was achieved during construction,
with a nearly constant residual compressive stress present during most of the construction
sequence.

C604 C607

35
C604 C638
30 C607
25 C638
Stress (MPa)

20

15

10

0
P16-4

P16-6

P16-8

P16-10

P16-12

P16-15

P16-16

P16-17
T201

T202

T203

T204

T205

T206

T301

T302

-5

Construction Event
(Tendon Stressed or Segment Erected)
Figure 5.37 Longitudinal stresses from selected strain gauges in segment
P16-2 over the course of the cantilever construction sequence

Continuity post-tensioning stresses

The stresses in ramp P following balanced cantilever construction were altered to those
needed for a continuous structure with internal and external continuity post-tensioning. The
location of these tendons is shown in Figure 5.38. Bottom slab internal continuity tendons T21
329
and T22 were stressed in span P16 following the casting of the closure at midspan. The pair of
tendons T21 and the pair of tendons T22 were 9 strand 15mm diameter tendons. External deviated
tendons T1, T2 and T3 were then stressed, providing an increase in both positive moment capacity
at midspan and negative moment capacity over the piers. These tendons were 19 strand 15mm
diameter tendons, the same size tendons used on the mainlane girders.

1.45 m 9 @ 2.87 m

P16-2 P16-4 P16-6 P16-8 P16-10 P16-12 P16-15 P16-16 P16-17

T1, T2, & T3

T22 T21 Vertical Deviator Horizontal Deviator

Elevation View (Horizontal:Vertical Scale = 1:2)
Center Line
of Pier P16

T1
T2
8.53 m

T3

Radius of Girder at Center Line = 221 m

Plan View (Horizontal:Vertical Scale = 1:1)

Figure 5.38 Layout of internal and external continuity tendons in span P16

Figures 5.39 through 5.41 show plots of the measured and calculated cross sectional
stresses produced by stressing the continuity tendons, and the final measured stresses at the
completion of construction. The measured stresses from the continuity tendons in segment P16-2
near the pier segment (shown in Figure 5.39) indicate that the location of the anchorages for
tendons T1 and T2 had influence on the distribution of stresses in the top flange. The anchorages
for these tendons were located on the other side of the heavy pier segment diaphragm from the
instrumented section. The forces in these anchorages did not diffuse into the wings over this short
distance of 3350mm. Shear lag response of the continuous girder to the stressing of tendon pair
T3 may also have had influence in the distribution of stresses in the top flange. The measured

330
stresses were linear down the depth of the girder webs, as can be seen in the bottom plot of Figure
5.39. The stresses calculated using the AASHTO effective flange width method and 30o diffusion
method overestimated the stresses in both the top and bottom flanges of the girder in segment P16-
2. The final cross sectional stresses in segment P16-2 are all compressive, with a minimum of
7MPa residual compression in the top flange for withstanding live load and thermal gradient
tensile stresses.

25
(+ compression, - tension)

20
Stress (MPa)

15
10
5
0
-5
-10
-15

30
(+ compression, - tension)

25
Stress (MPa)

20
15
10
5
0
-5
-10

Measured Stress
from Continuity
Tendons
Calculated Stress
from Continuity
Tendons
Final Measured
Stress Distribution

20 10 0 -10 Concrete
Strain Gauge
Stress (MPa)
(+ compression, - tension)
Figure 5.39 Longitudinal stresses in segment P16-2 after stressing of the
continuity tendons for span P16

331
 35
30
25
(+ compression, - tension)

20
15
Stress (MPa)

10
5
0
-5
-10
-15
(+ compression, - tension)

30
25
20
Stress (MPa)

15
10
5
0
-5
-10
Measured Stress
from Continuity
Tendons
Calculated Stress
from Continuity
Tendons
Final Measured
Stress Distribution

Concrete
Strain Gauge
20 10 0 -10
Stress (MPa)
(+ compression, - tension)

Figure 5.40 Longitudinal stresses in segment P16-10 after stressing of the

continuity tendons for span P16

332
 25

(+ compression, - tension)
20
15
Stress (MPa)

10
5
0
-5
-10
-15

(+ compression, - tension)

30
25
20
Stress (MPa)

15
10
5
0
-5
-10
Measured Stress
from Continuity
Tendons
Calculated Stress
from Continuity
Tendons

Concrete
Strain Gauge

20 10 0 -10
Stress (MPa)
(+ compression, - tension)

Figure 5.41 Longitudinal stresses in segment P16-17 after stressing of the continuity
tendons for span P16

Measured and calculated stresses from the continuity tendons in segment P16-10 are
shown in Figure 5.40. The calculated values accurately predicted the measured stresses from post-
tensioning in the top flange except over the right girder web in the top flange. The high measured
strains from the gauges located over the webs would indicate plastic behavior in the girder
concrete. This may have been the case since the strain gauges were located in the most congested
location in the segment. The gauges giving the high strain readings were located very close to the
333
large stirrup bars and the cantilever post-tensioning ducts. Segment P16-10 was also a deviator
segment with its associated vertical load at the base of the web diffusing into the segment. Bottom
flange stresses were overestimated by the calculation. Measured stresses appear to be linear down
the depth of the webs except near the gauge near the top fiber of the girder. The final measured
stress distribution in segment P16-10 plotted in Figure 5.40 did not reveal tension stresses at any
location in the cross section. Segment P16-10 was located at the quarter point of the span, so both
small negative moments from live load and somewhat larger positive moments from live loads and
thermal loads could be expected. The final stress distribution in the segment had slightly more
residual compressive stress in the bottom flange than the top flange to resist tensile stresses from
the design live and thermal loads.

The measured and calculated stress distribution from the continuity post-tensioning
forces in segment P16-17 (shown in Figure 5.41) was very similar to that in segment P16-10.
Once again, several strain gauges in the top flange over the girder web indicated higher than
expected strains. Also, the calculated stress was too large in the bottom flange by 50%, which
may have led to tensile stresses under live load in an unconservative design. At 10MPa, the
bottom flange residual compressive stress was high when compared to stress changes during the
live load tests.

Figures 5.42, 5.43 and 5.44 show plots of the actual concrete stress at various gauge
location in segments P16-2, P16-10 and P16-17 respectively. In general, the continuity tendons
produced higher stresses in the top flanges at points over the webs, as can be seen by comparing
gauges C604 and C607 in Figure 5.42 or C704 and C707 in Figure 5.43. The output from gauge
C717 in Figure 5.43 is similar to that of gauge C704, but much larger in magnitude. This may
indicate that the gauge was working correctly. The stress changes at all the gauge locations shown
in Figures 5.42 through 5.44 were gradual and predictable as tensioning of each of the tendons
proceeded.

334
 C604 C607

35
C604
30 C607
C638
25
Stress (MPa)
C638
20
15
10
5
0
-5
P16-17

T21

T22

T1 (right)

T2 (right)

T3 (right)
T1 (left)

T2 (left)

T3 (left)
Figure 5.42 Stresses from selected strain gauges from segment P16-2
over the course of the continuity stressing sequence
C704 C707 C717

C704
C707
35
C738
C717 30 C738
25
Stress (MPa)

20
15
10
5
0
T1 (right)

T2 (right)

T3 (right)
T1 (left)

T2 (left)

T3 (left)
P16-17

T21

T22

-5

Figure 5.43 Stresses from selected strain gauges from segment P16-10 over
the course of the continuity stressing sequence

335
 C804 C807

35
30 C804
C838 C807
25
C838
Stress (MPa)

20
15
10
5
0
-5
P16-17

T21

T22

T1 (left)

T1 (right)

T2 (left)

T2 (right)

T3 (right)

T3 (left)
Figure 5.44 Stresses from selected strain gauges from segment P16-17
over the course of the continuity stressing sequence
Live load test
A live load test was performed on ramp P shortly after the completion of construction, but
before the 50mm of blacktop was placed. Four dump trucks were placed in pairs back to back to
simulate positive moments produced by two lanes of HS20-44 truck live loading. The weights
and dimensions of each truck are given in Table 5.2. The five load cases shown in Figure 5.45
produced positive and negative bending moments in the continuous structure, as well as torsional
moments. Load cases 1, 3 and 5 were intended to produce maximum positive moments in spans
P14, P15 and P16 respectively. Since all strain measurements were taken in span P16, loads were
placed on other spans of the five span continuous unit to study moment distribution.

Table 5.2 Axle weights and spacing for live load test trucks on Ramp P
Truck Weight of Rear Weight of Front Total Weight Axle Spacing
Axles (kN) Axles (kN) (kN) (mm)
1 121.1 52.5 173.6 4720
2 113.2 47.2 160.4 4700
3 139.5 52.5 192.0 5210
4 120.9 52.6 173.5 5000

336
 1.8 m 1.8m
(6’) (6’)

Arrows indicate directions of trucks.

Center line 1.8 m (6’)

of ramp 1.8 m (6’)

27.4 m
(90’)

Live Load Case 1

1 2
3 4

17.4 m 21.9 m
(57’) (72’) (Measured to rear
axles of trucks)
Live Load Case 2
1 2
3 4

21.7 m
(71.2’)

Live Load Case 3

1 2
3 4

21.7 m 27.4 m
(71.2’) (90’) Outside of
Horizontal Curve
Live Load Case 4
1 2
3 4

27.4 m Inside of
(90’) Horizontal Curve
Live Load Case 5
1 2
3 4 Span P16

38.1 m 54.9 m 43.3 m 54.9 m 38.1 m

(125’) (180’) (142.4’) (180’) (125’)
P13 P14 P15 P16 P17 P18
Figure 5.45 Live load test cases for the ramp P girder

337
 The measured stresses in segment P16-2 from all load cases are shown in Figure 5.46. The
gauges in segment P16-2 were located 1900mm upstation from the centerline of pier P16. Load
cases 1 through 4 produced a negative bending moment at segment P16-2, and load case 5
produced a small positive moment. Load case 1 caused the greatest bending moment at P16-2, but
the measured stress changes were small with a maximum stress change in the bottom flange of
only 1.75MPa or 9% of the allowable service level compressive stress. Load case 3 was similar to
load case 1 except the trucks were placed on span P15. Negative moment from load case 3 was
distributed both to pier P16 and span P16. The measured stresses at segment P16-2 were less than
half those measured during load case 1. The moment connection of the superstructure to the pier
appears to have been working as a monolithic connection. The stress changes in segment P16-2
from load case 5 were very small, even though the load was only one and a half spans away from
the instruments. The moments produced by the truckloads from case 5 were distributed both to
the downstation portion of the superstructure and to piers P15 and P16.

338
 (+ compression, - tension)
0.5
Stress (MPa)
0.0

-0.5

-1.0

-1.5

*Note scale is much smaller

than dead load, prestress,
and thermal stress plots.
(+ compression, - tension)

2.0

1.5
Stress (MPa)

1.0

0.5

0.0

-0.5

Load Case 1
Load Case 2
Load Case 3
Load Case 4
Load Case 5

Concrete
Strain Gauge

2 1 0 -1
Stress (MPa)
(+ compression, - tension)

Figure 5.46 Measured live load longitudinal stresses in segment P16-2

Measured and calculated stresses for load cases 1 and 2 at segment P16-2 are shown in
Figure 5.47. The AASHTO effective flange method was used in the calculations and required a
small reduction in wing tip length at segment P16-2. Peak stresses in the top flange were
conservatively estimated in the calculation, although the measured stresses were probably lower
than the actual stresses because of the low assumed modulus of elasticity of the concrete. Shear
lag is evident in the measured stresses in the girder wings. At other locations in the cross section

339
the measured stress distributions appear to be uniform, with a nearly linear stress gradient down
the depth of the web.

(+ compression, - tension)
Stress (MPa)

0.0

-0.5

-1.0

-1.5

*Note scale is smaller

than dead load, prestress,
and thermal stress plots.
(+ compression, - tension)

2.5

2.0
Stress (MPa)

1.5

1.0

0.5

0.0

Load Case 1
Calculated
for Case 1
Load Case 2
Calculated
for Case 2

Concrete
2 1 0 -1 Strain Gauge
Stress (MPa)
(+ compression, - tension)

Figure 5.47 Measured and calculated live load longitudinal stresses in

segment P16-2

Measured and calculated stresses for load cases 1 and 2 at segment P16-10 are shown in Figure
5.48. The measured stress changes at this cross section were very small, since this section is close
to the inflection point of the span for both load cases. Maximum stress change in the bottom
flange for load case 1 was only 0.5MPa in tension. The calculated stresses for the bottom flange
predicted the measured stresses well on average. Much more significant stress changes were

340
measured in the midspan segment P16-17. Measured and calculated stresses for load cases 1 and
2 at segment P16-17 are shown in Figure 5.49. Once again, the calculated stresses exceeded the
measured stresses in the bottom flange by about 40% in the bottom flange and 20% in the top
flange. The peak measured stress change in the bottom flange from load case 1 was 2.2MPa in
tension at midspan. The final girder stress in the bottom flange at midspan following continuity
post-tensioning was about 10MPa in compression, so the live load stresses change did not threaten
to put the bottom flange in tension. In general, the stress changes from the live loads were small
in all test cases.
(+ compression, - tension)
Stress (MPa)

0.5

0.0

-0.5

*Note scale is smaller

than dead load, prestress,
and thermal stress plots.
(+ compression, - tension)

0.5
Stress (MPa)

0.0

-0.5

-1.0

Load Case 1
Calculated
for Case 1
Load Case 2
Calculated
for Case 2

Concrete
1 0 -1
Strain Gauge
Stress (MPa)
(+ compression, - tension)

Figure 5.48 Measured and calculated live load longitudinal stresses in

segment P16-10
341
 (+ compression, - tension)
Stress (MPa) 1.5

1.0

0.5

0.0

*Note scale is smaller

than dead load, prestress,
and thermal stress plots.

0.0
(+ compression, - tension)

-0.5
Stress (MPa)

-1.0

-1.5

-2.0

-2.5

-3.0

Load Case 1
Calculated
for Case 1
Load Case 2
Calculated
for Case 2

Concrete
Strain Gauge
2 1 0 -1 -2 -3
Stress (MPa)
(+ compression, - tension)

Figure 5.49 Measured and calculated live load longitudinal stresses in

segment P16-17

5.4.3 Two-cell girder–Unit C15, L2

Ramp L and mainlane girder C were joined by a cast-in-place closure pour as shown in
Figure 5.50. The width of the gore closure varied from 4390mm in span C41 to 6680mm in span
C42. The slab was post-tensioned transversely for live loads. Both the ramp L and the mainlane
C girders had two bearings at the individual piers that supported the girders. The girders were
constructed as simple spans with continuity to the adjacent spans consisting only of the cast-in-

342
place deck joint between top flanges of adjacent girders. These girders were only tested under live
load, and deflections were the only measurements taken. The purpose of the measurements was to
determine the degree of live load bending moment sharing between the two transversely connected
girders.

Ramp L Mainlane C

2050mm 3820mm Varies 5474mm 5798mm

place slab
Cast-in-
556mm

Piano wire Gauge mounting

plate
Cross Section
Unit C15, L2
Figure 5.50 Cross section of Unit C15, L2

Live load test

The six dump trucks used in the live load test were loaded similarly to the trucks used in
the other live load tests. The truck were placed back to back to produce a bending moment equal
to that produced by three lanes of HS20-44 trucks. The trucks were also separated to try to
produce a negative bending moment in the semi-continuous girder connection over the piers. The
weights and dimensions of the trucks used in the test are given in Table 5.3.

Table 5.3 Axle weights and spacing for live load test trucks on Units C15 and L2
Truck Weight of Rear Weight of Front Total Weight (kN) Axle Spacing
Axles (kN) Axles (kN) (mm)
1 133.9 55.9 189.8 5030
2 128.7 44.0 172.7 4090
3 109.5 41.5 151.0 4100
4 121.3 55.8 177.1 5000
5 137.6 42.5 180.1 5000
6 127.4 34.4 161.8 4700

Unit C15, L2 was a twin three-span semi-continuous unit with expansion bearings at
piers C41, L4, C44 and L7. A plan view of the three-span unit is shown in Figure 5.51. The gore
343
closure was terminated within span C43, L6, so the live load test cases were all located in the other
spans. The live load cases are shown in Figures 5.51a and 5.51b. All load cases except case 10
were designed to determine the degree of moment sharing between the two girders of unequal
section properties. Load cases 12 and 13 were drastically unsymmetrical with respect to the
longitudinal centerline of the bridge deck.

Front Axle Rear Axle

Weight Weight

Axle Spacing

Live Load Case 9 Typical Truck

Pier L7
Pier L6
Pier L5
Pier L4

Span L6
Span L4 Span L5
39.94m
34.09m 40.20m

1 4 3650mm (Typ.)
2 5 3650mm (Typ.)
3 6
3650mm

C
L
Span C41 Span C42 Span C43
34.14m
Pier C43

Pier C44
40.20m 40.07m
Pier C41

Pier C42

Live Load Case 10

1 4
2 5
6
3
12.80m 14.94m

(a) Live load cases for Unit C15, L2

344
Live Load Case 11

Pier L7
Pier L6
Pier L5
Pier L4
Span L6
Span L5 39.94m
Span L4
40.20m
34.09m

1 4
3650mm (Typ.) 2 5
3650mm (Typ.) 3 6

3650mm Spacing between rear axles T

( yp.)

CL
Span C41 Span C42 Span C43

Pier C43

Pier C44
Pier C41

34.14m 40.20m 40.07m

Pier C42

Live Load Case 12

1830mm (Typ.)
3 6
3650mm (Typ.) 2 5
3650mm (Typ.) 1 4

3650mm Spacing between rear axles

CL

Live Load Case 13

3 6
3650mm (Typ.) 2 5
3650mm (Typ.) 1 4
1830mm (Typ.) 3650mm Spacing between rear axles

CL

(b) Live load cases for Unit C15, L2

Figure 5.51 (a) and (b) Live load cases for Unit C15, L2

345
 The deflections measured during load case 9 are shown in Figure 5.52. The trucks for
load case 9 were placed symmetrically with respect to the longitudinal and transverse centerline of
the bridge deck of span C41, L4. The measured deflections reveal that the mainlane girder took
more of total bending moment than did the ramp girder. Assuming that the cast-in-place top
flange was shared equally by both the mainlane C girder and the ramp L girder, the ramp L girder
was calculated to be only 56% as stiff in bending as the mainlane C girder. Since the mainlane
girder deflection was 3.0mm and the ramp girder deflection was 2.2mm, calculations reveal that
71% of the total live load bending moment was taken by the mainlane girder. The deflection of
the mainlane C girder was only 1/11000 of the span length. The calculated deflections in Figure
5.52 assumed that no transverse continuity existed between the mainlane side of the bridge and the
ramp side of the bridge. A fictitious longitudinal cut was taken down the center of the cast-in-
place top slab. Wheel loads placed on the mainlane side of the cut were assumed to be taken
entirely by the mainlane girder, and wheel loads placed on the ramp side of the cut were assumed
to be taken entirely by the ramp girder. This way the results from the calculated deflections could
be easily interpreted and compared to the measured results. The AASHTO effective flange width
method was used in all calculations. The calculations for load case 9 assumed that the mainlane
carried 2 lanes of the trucks and the ramp girder carried 1 lane of trucks. This obviously means
that the mainlane girder would carry 67% of the live load moment. The measurements revealed
that the mainlane girder carried slightly more at 71% of the total live load moment. The
significantly stiffer mainlane girder was only able to attract 4% of additional live load moment for
the symmetrical load case 9.

346
 1

-1
Deflection (mm)

-2

Legend
-3 Ramp L Measured
Mainlane C Measured
-4 Ramp L Calculated
Mainlane C Calculated

-5

-6

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m

Figure 5.52 Deflections from load case 9

Load case 10 was similar to load case 9 in that the trucks were placed symmetrically with
respect to the longitudinal centerline of the bridge deck, but both spans C41, L4 and C42, L5 were
loaded. The measured and calculated deflections for load case 10 are plotted in Figure 5.53. The
calculations assumed that the mainlane girder carried 2 lanes of trucks and the ramp girder carried
1 lane of trucks. The measured deflection in span C41, the mainlane girder, was 1.9mm, and the
deflection in span L4 was 0.9mm. Taking into account the relative stiffnesses of the mainlane and
ramp girders, the mainlane girder carried 80% of the total live load moment. This was a higher
percentage of the live load moment than seen in load case 9. The measured deflection for load
case 10 in span C42 was 2.4mm, and the deflection in span L5 was 1.3mm. From calculation the
mainlane girder carried 77% of the total live load moment for span C42, L5.

347
 1

-1
Deflection (mm)

-2

-3
Legend
Ramp L Measured
-4
Mainlane C Measured
Ramp L Calculated
-5 Mainlane C Calculated

-6

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m

Figure 5.53 Deflections from load case 10

Load case 11 was identical to load case 9, except the trucks were placed on span C42, L5.
The measured and calculated deflection for load case 11 are shown in Figure 5.54. The deflection
calculation assumed that the mainlane girder carried 2 lanes of trucks, and the ramp girder carried
1 lane of trucks. The measured deflection of mainlane girder C42 was 4.9mm, and the measured
deflection of the ramp girder L5 was 3.6mm. The deflection of the mainlane girder was 1/8000 of
the span length, which was larger than the deflection of mainlane girder C41 in load case 9. The
span length of mainlane girder C42 was significantly longer than C41. From calculation, the
mainlane girder carried 71% of the total live load moment, identical to the percentage taken by the
mainlane girder in load case 9.

348
 1

-1
Deflection (mm)

-2

-3

-4 Legend
Ramp L Measured
Mainlane C Measured
-5
Ramp L Calculated
Mainlane C Calculated
-6

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m

Figure 5.54 Deflections from load case 11

The truck placement for load case 12 placed 2.5 lanes of trucks on the ramp girder, and
0.5 lanes of trucks on the mainlane girder for deflection calculation purposes. The ramp girder
would carry 83% of the total live load moment if the girders had no longitudinal connection. The
measured deflections in Figure 5.55 indicate that the mainlane girder carried significantly more
load than 17% of the total moment. The measured deflection of the mainlane girder was 3.3mm,
and the measured deflection in the ramp girder was 6.4mm or 1/6100 of the span length. From
calculation the mainlane girder carried 48% of the total live load moment for load case 12, almost
three times the moment predicted by the simple calculation method. The measured and calculated
deflections for load case 13 are plotted in Figure 5.56. The truck locations for load case 13 placed
all 3 lanes of trucks on the mainlane girder for deflection calculation purposes. The measured
deflection for mainlane girder C42 was 5.4mm or 1/7200 of the span length, and the measured

349
deflection for the ramp girder L5 was 2.4mm. From calculation, the mainlane girder carried 80%
of the total live load moment, and the ramp carried 20% of the moment even though no truck
wheel loads were located beyond the webline of the mainlane girder.

-1

-2

-3

-4
Deflection (mm)

-5

-6

-7

-8

-9 Legend
Ramp L Measured
-10
Mainlane C Measured

-11
Ramp L Calculated
Mainlane C Calculated
-12

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m

Figure 5.55 Deflections from load case 12

350
 1

-1

-2
Deflection (mm)

-3

-4

-5

Legend
-6
Ramp L Measured
Mainlane C Measured
-7
Ramp L Calculated
Mainlane C Calculated
-8

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m

Figure 5.56 Deflections from load case 13

5.4.4 Three-cell girder - Unit C13

The three-cell girders of two-span unit C13 were cast-in-place as simple spans. A cross
section of the unit C13 girder is shown in Figure 5.57. Since the outer two cells and wings were
originally designed to be precast, all the longitudinal post-tensioning was located in these two
cells. Also, the bearings supporting the entire span were only located beneath these two cells, with
two bearings at each pier. The central top and bottom flanges varied in width from 4800mm in
span C36 to 5300mm in span C37. The top flange was post-tensioned transversely. Deflections
were measured at the center of each of the three cells. The live load response of these girders
351
would be expected to differ from that of the twin single-cell girders of Unit C15, L2 because two
slabs connect the outer two cells in unit C13, increasing the transverse bending stiffness, and
substantially increasing torsional stiffness of the central cell.

From Ramp L From Mainlane C

Varies 1994mm Varies 1994mm 5930mm

Piano wire Gauge mounting

plate
Cross Section
Unit C13
Figure 5.57 Cross section of Unit C13

Live load test

The live load test was the only test on unit C13. Six dump trucks were placed back to
back as in the other live load tests. The dimensions and weights of the dump trucks are shown in
Table 5.4. The five load cases shown in Figures 5.58a and 5.58b are similar to the five load cases
tested on unit C15, L2. These load cases were intended to reveal the amount of sharing of bending
moments between the two halves of the girders. The three-cell girders were treated as two half
girders, or twin single-cell girders, for calculation purposes because bearing reactions were only
located under the outer two cells, and the outer two cells were much more compact than the central
cell.

Table 5.4 Axle weights and spacing for live load test trucks on Unit C13
Truck Weight of Rear Weight of Front Total Weight Axle Spacing
Axles (kN) Axles (kN) (kN) (mm)
1 119.2 36.7 155.9 4900
2 117.5 43.9 161.4 4720
3 112.0 45.4 157.4 4900
4 119.0 35.1 154.1 4700
5 113.6 30.6 144.2 4700
6 122.7 41.0 163.7 4090

352
 Front Axle Rear Axle
Weight Weight

Live Load Case 2

Axle Spacing
Pier C36

Pier C37

Pier C38
Typical Truck

Span C36 Span C37

40.84m 40.84m

1 4
3650mm (Typ.)
2 5
3650mm (Typ.)
3 6

CL
3650mm Spacing between
rear axles

Live Load Case 3

1 4
2 5
3 6

15.24m 15.24m
Pier C37

Pier C38
Pier C36

(a) Live load cases for Unit C13

Figure 5.58 (a) and (b) Live load cases for Unit C13

353
 Live Load Case 4

Pier C36

Pier C37

Pier C38
Span C36 Span C37
40.84m 40.84m

1 4
3650mm (Typ.)
2 5
3650mm (Typ.)
3 6

CL
Spacing between 3650mm
rear axles
Live Load Case 5

3650mm
1830mm

4 2
3 5
1 6

CL

Live Load Case 6

3650mm
1830mm

4 2
3 5
1 6

CL
(b) Live load cases for Unit C13
Figure 5.58 (a) and (b) Live load cases for Unit C13 (con’t)

Deflection calculations for the load cases to follow were performed using the same
assumptions as for the deflection calculation for unit C15, L2, including the AASHTO effective
flange width method. A longitudinal cut was taken down the center of both the top and bottom
354
central flanges, and wheel loads falling on each side of the cut would be carried by the half girder
on that particular side of the cut. The calculated girder stiffnesses used to determine the
percentage of live load moment taken by each half of the girder were also calculated by assuming
a longitudinal cut down the centerline of the central top and bottom slabs. This way the measured
deflections could be compared to calculated deflections with well-defined boundary conditions
and easily calculated bending stiffnesses.

The measured and calculated deflections from load case 2 are plotted in Figure 5.59. The
maximum measured deflection was in the center cell of the girder because of the deflection of the
central top slab, and to a lesser extent the torsional rotation of the outer two cells. The calculated
deflections assumed that 1.75 lanes of trucks were carried by the right side of the girder, and 1.25
lanes of trucks were carried by the left side of the girder. The cross section of span C36 was not
symmetric with respect with the centerline of the bridge deck, so the symmetrically placed load
case 2 was carried more by the right side of the girder that had the longer wing. One line of truck
wheels fell exactly on the fictitious longitudinal cut between girder halves. Therefore, calculated
deflections assumed that the right side of the girder carried 58% of the total live load moment, and
the left side of the girder carried 42% of the moment. The measured deflection of the right cell
was 4.6mm or 1/8900 of the span length, and the measured deflection of the left cell was 4.2mm.
The bending stiffness of the left half of the girder was 96% of the bending stiffness of the right
half of the girder. Using these values in a calculation, the right half of the girder carried 53% of
the load and the left half of the girder carried 47% of the load. These percentages indicate that
some of the live load moment was shared between girders, as would be expected.

355
 1

-1
Deflection (mm)

-2

-3

Legend
-4 Left Cell Measured
Middle Cell Measured
Right Cell Measured
-5
Left Cell Calculated
Right Cell Calculated
-6

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 5.59 Deflections from load case 2

Load case 3 was designed to produce a negative moment over the pier, if possible, but
also provided a load case on two spans where the center of gravity of the load was not at midspan.
The measured and calculated deflections for load case 3 are shown in Figure 5.60. The calculated
deflections assumed that the right and left sides of the girder shared one line of wheels, so the right
girder carried 1.75 lanes and the left carried 1.25 lanes as in case 2. The measured deflections in
span C36 appear to be effected by the presence of wheel loads near the gauge mounting plates so
no conclusions can be drawn from that span. The measured deflections in span C37 from load
case 3 indicate a midspan deflection of 1.7mm in the right cell and 1.4mm in the left cell. The

356
bending stiffness of the left cell was 92% of the bending stiffness of the right cell in span C37.
From a calculation using the measured deflections and the stiffness ratio, the right side of the
girder carried 56% of the total bending moment and the left side of the girder carried 44% of the
total moment. The calculated percentages assuming the longitudinal flange cuts down the
centerline of the top and bottom flange were 58% for the right side of the girder and 42% for the
left side of the girder.

-1
Deflection (mm)

-2

-3
Legend
Left Cell Measured
-4 Middle Cell Measured
Right Cell Measured
-5 Left Cell Calculated
Right Cell Calculated
-6

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 5.60 Deflections from load case 3

Load case 4 was nearly identical to load case 2, except the live load was placed on span
C37. The measured and calculated deflection for load case 4 is plotted in Figure 5.61. The
measured deflection of the right side of the girder was 4.2mm, and the measured deflection of the

357
left side was 4.0mm. The right side deflection was only 1/9700 of the span length, which was
smaller than the deflection in span C36 from load case 2. From calculation, the right side of the
girder carried 53% of the total live load moment, and the left side of the girder carried 47% of the
moment. These are identical to the results from load case 2, even though the relative stiffnesses of
the girder halves were different. The calculated deflection calculations assumed that 58% of the
total live load moment was carried by the right girder.

-1
Deflection (mm)

-2

-3 Legend
Left Cell Measured
-4 Middle Cell Measured
Right Cell Measured
Left Cell Calculated
-5
Right Cell Calculated

-6

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 5.61 Deflections from load case 4

Load case 5 was highly unsymmetrical with all three-load lanes placed on the right girder
in span C36. The measured and calculated deflections for load case 5 are plotted in Figure 5.62.

358
The calculated deflection for the right cell is almost twice the measured deflections, indicating
significant sharing of the total live load moment. The measured deflection of the right cell was
5.3mm, and the measured deflection of the left cell was 4.1mm. From calculation, the right cell
was carrying only 57% of the total live load moment, so the left cell was able to share a high
percentage of the total live load bending moment on the girder.

-1

-2
Deflection (mm)

-3

-4

-5

Legend
-6
Left Cell Measured
Middle Cell Measured
-7
Right Cell Measured

-8
Right Cell Calculated
No Calculated Deflection
in Left Cell
-9

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 5.62 Deflections from load case 5

359
 Load case 6 was similar to load case 5 but the load trucks were placed on span C37. All
three-load lanes were located on the right half of the girder. The measured and calculated
deflections for load case 6 are plotted in Figure 5.63. The measured deflection of the right cell
was 5.0mm, and the measured deflection of the left cell was 4.1mm. Using the proper ratio of
bending stiffnesses for span C37, calculations reveal that just under 57% of the total live load
moment was carried by the right cell. This was nearly identical to the results from load case 5.
The results from load cases 5 and 6 show that the design truck live load for the span produces
nearly the same bending moments in two outer cells regardless of the location of the trucks on the
deck. The maximum measured deflections from load cases 5 and 6 were only 1/7700 of the span
length.

360
 1

-1

-2
Deflection (mm)

-3

-4 Legend
Left Cell Measured
-5 Middle Cell Measured
Right Cell Measured
Left Cell Calculated
-6
Right Cell Calculated

-7

-8

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 5.63 Deflections from load case 6

5.4.5 Large ramp pier

The large ramp pier P16 was a bending and compression member, with the greatest
moments occurring during the unbalanced condition from the superstructure construction. Live
load moments on the ramp P superstructure created bending moments in the pier P16 because the
moment connection to the superstructure was not released after superstructure construction was
completed. The live load moments in the pier were smaller than the construction load moments.

361
Strain measurements were taken during all construction stages and during all live load tests, but
the performance of the pier could best be evaluated by studying the load cases providing the
largest strain changes.

Response of large ramp Pier P16 to superstructure construction loads

The Ramp P superstructure was constructed in balanced cantilever with an unbalanced
moment on the piers no larger than that from the weight of one superstructure segment. Since the
contractor had trouble adequately stressing the superstructure to pier tie down bars on pier P16,
this unbalanced moment was taken partially or completely by the crane used to lift the segment
into place. This was possible since two ground based cranes were used to construct the five span
continuous girder on Ramp P. The complicated construction sequence and limited ground access
dictated the use of two cranes. Unfortunately, the maximum unbalanced moment on pier P16, as
predicted in the pier’s design, never occurred. Instead, the maximum moment in the pier from the
superstructure construction occurred when segment P16-17 was placed. This was the final
segment placed in cantilever near the midspan of span P16. Most importantly, this segment was
placed after continuity had been made in the superstructure span P15. The cantilevering moment
from segment PC16-17 was distributed to both pier P16 and to the superstructure span P15 and
pier P15. The response of pier P16 was measured for this load case.

Measurements were taken by strain gauges oriented vertically at several locations along
the height of pier P16. The bending moment in the pier would have been essentially constant
along the height of the pier during balanced cantilevering, but since the pier was part of an
indeterminate frame when segment PC16-17 was placed, the moment in the pier changed over its
height. Four planes of gauges were selected for study. The first plane of gauges was in segment
PC16-1 located 2.057m above the top of the footing. The second set of gauges was in segment
PC16-5 located 12.167m above the top of the footing. The third set of gauges was in segment
PC16-7 located 16.085m above the top of the footing. The fourth set of gauges was located on the
16 vertical tie down bars located near the top of the pier capital segment PC16-8, 19.810m above
the top of the footing. The length from the top of the footing to the center of gravity of the
superstructure box girder was 21.886m. The 285kN load from the placement of superstructure
segment P16-17 had a cantilever arm of 25.845m to the center of gravity of the pier, yielding a
bending moment of -7366kN-m to be distributed to the structure. Because of the horizontal
curvature of the superstructure, a torque of 435kN-m also had to be distributed to the structure.

362
 Figures 5.64 through 5.67 show the measured concrete stresses at the gauge locations and
calculated stresses along the north-south axis. The superstructure centerline geometry was exactly
parallel to the pier’s north-south axis at pier P16, as well as the other ramp piers. For the purpose
of comparison, a plane frame analysis was performed, ignoring torque effects, and is plotted in
Figures 5.64 through 5.67. The solid line assumed fixity at the top of the footing. The dashed line
assumed the point of fixity was 4.2m below the top of the footing to account for flexibility of the
footing and four 1.070m drilled shafts. The 4.2m pier height increase was chosen to calibrate the
model to the measured results of gauges C400, C402, C410 and C412 in Figure 5.64 along the
north-south axis of the pier. These gauges in segment PC16-1 were located near the point of
inflection in the pier. Distribution of the cantilever moment of -7366kN-m was -3263kN-m to pier
P16 and -4103kN-m to span P15 when fixity was assumed at the top of the footing. Distribution
was -3041kN-m to pier P16 and -4324kN-m to span P15 when the point of fixity was moved down
4.2m.

Stresses calculated with fixity

N condition at the top of the footing
Stresses calculated with fixity
Up Sta
condition 4.2m below top of the footing
C400
C400
C402

C403
C404

C418
C403
C419

C402
C419 C404
C417
C405

C415 C417 C407 C405

C415
C412 C407
C414

C414 C409
C412
500 C410

C413 C408
2000
1500
1000

-1500
-500
-1000

-2000
0

C410

Concrete Stress in kPa

Figure 5.64 Stresses in pier segment PC16-1 from placement of
superstructure segment P16-17

363
 Stresses calculated with fixity
N condition at the top of the footing

Up Sta Stresses calculated with fixity

condition 4.2m below top of the footing

C430

C430
C432

C439
C441
C439 C441 C435 C433

C433
C438

C436
2000
1500
1000

-1500
500

-500
-1000

-2000
0
C436

Concrete Stress in kPa

Figure 5.65 Stresses in pier segment PC16-5 from placement of
superstructure segment P16-17

Stresses calculated with fixity

N condition at the top of the footing
Stresses calculated with fixity
Up Sta condition 4.2m below top of the footing
C450
C450
C468

C468 C453
C454

C452
C453

C452
C469 C454
C469
C467
C465

C465 C467 C457 C455

C464
2000 C460 C459

C463

C464 C459
C462

C462
C458

C463 C458
1500
1000

-1500
500

-500
-1000

-2000
-2500
0

C460

Concrete Stress in kPa

Figure 5.66 Stresses in pier segment PC16-7 from placement of
superstructure segment P16-17

364
 Stresses calculated with fixity
condition at the top of the footing
N Stresses calculated with fixity
condition 4.2m below top of the footing
Up Sta

S538
S535

S522
S521
S520
S537
S538 S535 S522 S517

S517
S537 S536 S521 S520

S534
S529

S528
S523
S534 S529 S528 S523
S531 S530 S527 S524

S527
S524
S530
S531
2000
1500
1000

-1500
500

-500
-1000

-2000
0
Concrete Stress in kPa
Figure 5.67 Stresses in pier segment PC16-8 tie-down bars from placement
of superstructure segment P16-17

In plan view, the torque from the superstructure would tend to cause tension in the west
face of the pier. This is poorly reflected by the gauges in segment PC16-1 in Figure 5.64. This
trend is consistently evident as indicated by the gauge measurements in Figures 5.65 through 5.67.
If the entire torque were taken by pier P16, the change in the calculated stresses in the extreme
fibers along the east-west axis, plotted in Figures 5.64 through 5.67, would be about 400kPa. The
horizontal radius of the superstructure was only 218.3m, so torsional effects were included in the
design of this pier.

Pier P16 was compact and regular in shape over much of its height, and appears to have
behaved in a very predictable way, with stress change linear across the section. The accuracy of
the design of this pier would rely more on the accuracy of the analysis, not on the sectional
behavior of the pier. Accurate modeling of the foundation stiffness would be part of a good
design of this pier. The post-tensioned connection of the superstructure to the pier also behaved in
a predictable manner. The calculated stresses plotted in Figure 5.67 assumed that plane sections
remained plane in the grout pad on top of the pier capital, and therefore the full moment of inertia
of the grout pad could be used. This assumption appears to have been valid, since the measured

365
stresses plotted in Figure 5.67 fall on either side of the calculated stresses. No tension in the
connection was evident.

5.5 DISCUSSION OF RESULTS

The following discussion compares the measured results for the different structural
elements that were instrumented. The results calculated using the design code methods are also
evaluated.

5.5.1 Diffusion of post-tensioning

The first post-tensioning force to be diffused into a segmental concrete is often the
temporary force required to squeeze the segment joints during epoxying. The AASHTO
Segmental Guide Specification requires that a minimum of 0.28MPa of stress must exist at all
points on the segment joint during epoxying to squeeze the epoxy out of the joint so that concrete
surfaces are in contact, and presumably to close any geometrical inconsistencies between match
cast joint faces. From visual inspection, this temporary force was not sufficient to close the gap
that existed between the center of the top flanges of the mainlane segments in all instances. Epoxy
was able to slide down the joint after the temporary force had been fully applied. Additional force
should be applied to close these gaps in segmental bridges wide enough to be susceptible to
significant thermal induced warping during casting and curing.

The temporary epoxying stresses measured in the ramp P segment (shown in Figure 5.33)
were just adequate to squeeze the epoxy at all points of the segment joint. But, the measured
stresses did not meet the code specification at the wing tips and at one point of the bottom flange.
The stress plot shows that the temporary joint force was poorly diffused into the section.
Designers should give some consideration to the diffusion of the temporary joint forces from the
temporary anchorage blisters or other anchorage location. The ramp P segments had most of the
temporary joint force anchored at two blisters located at the tops of the webs inside the core of the
box girder. This was an excellent location for the blisters from a construction standpoint, but the
force was too low or too remote to squeeze the wing tips. Also, because of the lack of diffusion of
the temporary force into the wings, the bottom flange was put into tension from the stressing of
the top four temporary force post-tensioning bars. The contractor had to stress the bottom bar first
in order to close the joint across the bottom flange. As the top bars were stressed, the force in the
bottom bar changed little because its elastic modulus was much larger than the modulus of
elasticity of the concrete, and strain changes were small.
366
 Indications from strain measurements in the mainlane girder anchor segment were that
the post-tensioning force from the 6 deviated tendons and 14 bottom slab tendons was diffused to
a greater extent than would be found using the AASHTO 30o diffusion method. The shape of the
diffused stresses was predicted poorly by the calculation. These tendons were located within or
immediately next to the heavy end diaphragm in the anchor segment. The AASHTO 30o diffusion
method predicted slightly more diffusion than was measured to occur for the cantilever tendons in
segment P16-10, shown in Figure 5.35. The girder wings took more stress than predicted by the
calculation, and the webs took less stress. Given the simplicity of the 30o diffusion method,
stresses were adequately predicted at the instrumented section in segment P16-10. No diaphragm
existed at this location.

Strain measurements at the midspan section of the mainlane girder indicated that the
post-tensioning forces had diffused to a nearly uniform distribution (see Figure 5.21), even though
the AASHTO effective flange width method would predict otherwise by a small amount. Using
the 30o diffusion method, post-tensioning stresses should have been uniform at this cross section
of the girder. The cantilever post-tensioning tendon forces did not diffuse to a uniform
distribution in segment P16-2 (see Figure 5.36) because the anchorage locations were spaced
along the span at each segment joint. Therefore, some of the anchorages were located quite close
to the instrumented section in segment P16-2. The prestressing forces and bending moments from
the continuity tendons in ramp P generated well-distributed stresses across the girder cross-
sections with the exception of the instrumented section in P16-2 (see Figure 5.39). The
anchorages for two sets of the deviated external tendons were too close to that section to be fully
diffused.

In general, at cross sections where the AASHTO 30o diffusion method would predict full
diffusion of a post-tensioning force, the measurements revealed a nearly uniform distribution of
stresses from that prestressing force. Since neither the ramp P girder nor the mainlane girder
required a significant reduction in wing or flange length for the AASHTO effective flange length
calculation, the effect of shear lag from primary prestressing moments may not have been seen in
the measurements. The bf curve in Figure 5.10 should be used for the entire girder to account for
shear lag effects from primary prestressing moments. For secondary prestressing moments, the
appropriate curve bs or bf should be selected depending on the location on the span, as specified by
AASHTO, since the secondary moments are generated by the pier reactions. If the exact diffusion

367
of post-tensioning forces must be known at points close the anchorages, the AASHTO 30o
diffusion method is not sufficiently accurate for the calculation.

5.5.2 Response to dead loads and live loads

Use of the cross section properties calculated using the AASHTO effective flange width
method adequately predicted the dead load plus prestress final stress distribution for the mainlane
girder (shown in Figure 5.21), although an insufficient number of measured data points existed to
determine the exact shape of the stress distribution in the flanges. The live load stresses calculated
for mainlane live load cases 2 and 8 predicted the location of the neutral axis of the girder quite
accurately, as seen in Figures 5.25 and 5.26, when the AASHTO effective flange width method
was used. The location of the neutral axis was also predicted very accurately for live load cases 1
and 2 in ramp P (shown in Figures 5.47, 5.48 and 5.49), although the effective flange for these
cases was nearly the actual length. The differences between the measured and calculated stresses
in the flanges were mostly due to an underestimation of the modulus of elasticity of the concrete.
The maximum measured live load stress changes for these girders were only 15% of the maximum
allowable service level stress. Deflections measured during the live load cases were similarly
small at 1/6500 of the span length or less for the simulated HS20-44 truckloads. Dead load
deflection for the mainlane girder was 1/1050 of the span length, so the live load moments were
about 16% of the dead load moments.

The AASHTO effective flange width method provided section properties that gave
realistic calculated stresses. The major problem with the method is that the moment of inertia of
the girder must be calculated at several sections on every span. Using a stepped transition for the
flange widths near the piers to the flange widths near midspan would only simplify the calculation
in some instances. In a continuous structure, the bottom slab thickness is often modified to
provide some extra moment of inertia near the interior piers, so section properties would be
varying anyway. The moment of inertia at random sections within the AASHTO tapered
transition can be found with sufficient accuracy by interpolation, so little time is saved by using a
stepped transition. The greatest realization of time lost is when standard box girder sections are
being used with tabulated section properties. Cutting the flanges to their effective width
potentially requires minor recalculation of the moment of inertia and neutral axis location for
many cross section locations.

368
 A faster method would modify the moments or stresses found from a normal analysis
with the full section properties. For example, the moment diagram could be modified by a
function of the absolute value of the shear diagram and the girder shape. Peak stresses would be
calculated based on the modified moment diagram. The function could be calibrated to results
using the AASHTO effective flange width method over a range of b/li (from Figures 5.8 and 5.9)
of 0.05 to 0.20. This range would cover the flange widths most used in segmental box girder
bridges.

5.5.3 Behavior of multiple-cell box girders

Deflections measured during the live load tests revealed that significant sharing of live
load moments was occurring in the multiple-cell girders, especially for the highly unsymmetrical
load cases. A summary of the test results for these girders is given in Table 5.5. Column (3) in
the table is a record of the measured deflections at midspan. Column (4) in the table contains the
percentage of total bending stiffness for each girder or girder half assuming the shared central
flanges were cut longitudinally down the center. Column (5) gives the calculated percentage of
the total live load moment carried by each girder or girder half based on the data in columns (3)
and (4). Column (6) gives the percentage of the live load applied to the girder deck, assuming a
longitudinal cut down the centerline of the central girder flanges. The percentage of the total
moment carried by the mainlane C girders in load cases 9, 10, 11 and 12 indicates that the
mainlane girders carried more than their expected share of the total moment. This is attributable
to the higher torsional stiffness of the mainlane girders. Both the mainlane girders and the ramp
girders were set on relatively soft elastomeric bearings. The distance between bearings at each
pier for the mainlane girders was 3350mm, and the distance between bearings for the ramp girders
was 1830mm.

369
 Table 5.5 Summary of live load test results for multiple-cell girders
(1) (2) (3) (4) (5) (6)
Load Girder Measured % of Total % of Total Live % of Live Load
Case Deflection Bending Load Moment Applied To Girder
(mm) Stiffness Carried Deck
9 C41 3.0 64% 71% 67%
9 L4 2.2 36% 29% 33%
10 C41 1.9 64% 80% 67%
10 L4 0.9 36% 20% 33%
10 C42 2.4 64% 77% 67%
10 L5 1.3 36% 23% 33%
11 C42 4.9 64% 71% 67%
11 L5 3.6 36% 29% 33%
12 C42 3.3 64% 48% 17%
12 L5 6.4 36% 52% 83%
13 C42 5.4 64% 80% 100%
13 L5 2.4 36% 20% 0%
2 C36 Right 4.6 51% 53% 58%
2 C36 Left 4.2 49% 47% 42%
3 C37 Right 1.7 52% 56% 58%
3 C37 Left 1.4 48% 44% 42%
4 C37 Right 4.2 52% 53% 58%
4 C37 Left 4.0 48% 47% 42%
5 C36 Right 5.3 51% 57% 100%
5 C36 Left 4.1 49% 43% 0%
6 C37 Right 5.0 52% 57% 100%
6 C37 Left 4.1 48% 43% 0%

Similar behavior was seen in unsymmetrical load cases 12 and 13. In load case 12, 83%
of the truckload was placed on the ramp girder L5, yet the mainlane girder C42 carried 48% of the
total live load bending moment. When 100% of the total live load was placed on the mainlane
370
girder C42 in load case 13, only 20% of the total live load moment was carried by the ramp girder.
Therefore, the ramp girder can more easily distribute its live load moments to the more torsionally
rigid mainlane girder. The stiffness of the bearings must be included in an analysis calculating the
load sharing between these two single-cell girders. The maximum measured live load deflection
for the ramp L girder was very small at only 1/6300 of the span length.

The behavior of the three-cell unit C13 was quite different from that of unit C15, L2. A
heavy diaphragm was cast the full width of the three-cell girders over the centerline of bearing, so
the torsional rigidity of the individual cells was unaffected by the bearing stiffnesses. In
symmetrical live load cases 2, 3 and 4 the slightly right-of-center live load (58% on the right
girder half) was carried by both girders almost evenly (53% to 56% of the total live load moment
carried by the right girder half). For the highly unsymmetrical load cases 5 and 6, where the entire
load was placed on one girder half, only 57% of the total live load moment was carried by the
loaded girder half, and 43% of the moment was distributed to the other girder half. The presence
of the heavy pier diaphragm would have great influence on live load moment and stress
calculations for this girder, simplifying the calculation. Maximum live load deflections of unit
C13 were similar to the maximum deflections measured during the other live load tests at only
1/7800 of the span length.

5.6 CONCLUSIONS
The following conclusions have been made based on comparisons of the measured data
with the calculated results:

5.6.1 Temporary post-tensioning for epoxying segments

1. An engineer should determine if the segments under design will take a “banana” shape
during casting and curing, estimate the warping deflection that will occur, and include the
deflection in the design of the temporary post-tensioning to insure full closure of the gap.

2. The diffusion of the temporary post-tensioning force from the anchorages or blisters should
be estimated, and stresses calculated at the extremities of the cross section away from the
anchorage points. Assuming that plane sections will remain plane, and that a linear stress
gradient will pass through the center of gravity of the section may lead to an inaccurate
estimate of the actual stress distribution. The temporary post-tensioning forces and locations
should be designed to adequately stress the entire cross section considering diffusion.

371
 Furthermore, the dead load of the segment constructed in balanced cantilever will not
produce a predictable stress distribution during the epoxying and temporary post-tensioning
process, and should not be relied upon as a source of bottom flange squeezing stress.

3. The sequence of temporary post-tensioning should be considered, especially if anchorages

are not well distributed throughout the cross section.

5.6.2 Diffusion of post-tensioning forces

1. The AASHTO 30o diffusion method tended to underestimate the amount of diffusion of post-
tensioning forces from anchorages in or immediately adjacent to anchorage diaphragms.

2. The AASHTO 30o diffusion method is not sufficiently accurate for calculating stresses in the
vicinity of post-tensioning anchorages.

3. The AASHTO 30o diffusion method was sufficiently accurate for predicting the distance
from an anchorage to the point of full diffusion into the cross section.

4. At sections where the post-tensioning force is calculated to be fully diffused, shear lag in the
cross section from primary post-tensioning moments can be compensated for by using the
AASHTO effective flange width method. Only the bf width should be calculated and used
over the entire span since the pier reactions are not included in the calculation of primary
moments or the stresses resulting from these moments.

5. Use the AASHTO effective flange width method for predicting cross sectional stresses from
secondary moments.

5.6.3 Girder response to dead loads and live loads

1. The measured stresses and deflections from the simulated HS20-44 truckload (no impact)
were quite small when compared to girder dead load. Live loads on the mainlane girder
produced deflections only 1/6500 of the span length, while calculated dead load deflection
was 1/1050 of the span length.

2. The AASHTO effective flange width method gave sufficiently accurate results for the
calculation of stresses and deflections from dead loads and live loads, although the girders
tested did not experience significant shear lag at sections of high moment and stress gradient.

3. The AASHTO effective flange width method requires considerable section property
calculation for girders that may only experience a small amount of shear lag, such as most

372
 common segmental girders and nearly all simple span segmental girders. Another method
should be developed for these girders that uses unmodified section properties.

5.6.4 Performance of multiple-cell girders under live load

1. The twin single-cell girders tested shared as much as 30% of the applied live load moment.

2. The amount of live load moment sharing between the twin single-cell girders was sensitive to
the torsional stiffness of each girder. The stiffness of the bearings must be included in the
design of these girders.

3. The three-cell girders tested consistently shared as much as 43% of the applied live load
moment.

4. The amount of live load moment sharing between sides of the three-cell girders was not
sensitive to the stiffness of the bearings because of the presence of the pier diaphragm that
was cast full width of the three girder cells.

5. Live load deflections were quite small in the multiple-cell girders with a maximum measured
deflection of only 1/6300 of the span length or smaller.

5.6.5 Performance of the segmental pier under bending

1. The cross sectional behavior of the segmental pier was entirely predictable, but the stiffness
of the drilled shaft foundation had to be included in the analysis.

2. The moment connection of the pier to the balanced cantilever ramp superstructure using a
cast-in-place grout pad and 16 post-tensioning bars performed as would a monolithic
connection.

373
374
 CHAPTER 6

BEHAVIOR OF DISCONTINUITY (OR D-) ZONES

6.1 PROBLEM AND TECHNICAL BACKGROUND

The design of a post-tensioned segmental box girder or pier must include the design of
several discontinuity or D-zones. A D-zone can be described as a region of the structure in the
immediate vicinity of an applied concentrated force or drastic change in structural geometry. In
such regions a plane section before loading generally does not remain plane after loading. Since
the design of the structure in regions of constant or gradually changing cross sectional properties
(B-zones where plane sections remain plane) is relatively straight forward, many of the problems
encountered in post-tensioned segmental structures occur because of design deficiencies or
uncertainties in the D-zones. The improper or unconservative design of D-zones may create
serviceability problems that overshadow the basic structural soundness of the entire bridge, and
lead to skepticism regarding the safety of post-tensioned segmental bridges by engineers and the
public sector alike.
Because of its basic nature, a post-tensioned structure has very large applied concentrated
forces at post-tensioning tendon anchorages. Thus, the design of several anchorage zones is
inevitable. The locations of the anchorage plates frequently do not lie within the confines of the
typical cross section, requiring anchorage diaphragms or blisters. Furthermore, external tendons
require deviators at points along the superstructure that must transfer large vertical forces to the
girder. Another source of concentrated force is the reaction of the pier on the superstructure.
Segmental box girders generally have one and occasionally two cells. Therefore, the girder webs
carry large shear forces that are usually eccentric to the pier or bearing reactions because of the
inclination of the webs or bearing locations. The segments located over the piers must be
designed to carry the superstructure loads to the bearing reactions, as well as to provide anchorage
zones for many of the post-tensioning tendons.
The design of a post-tensioned box girder or pier generally includes two distinct design
limit states. The first case is the service load state. The cross section of the girder or pier is
usually post-tensioned in such a manner that little or no tension occurs in the concrete, so cracking
is not expected and elastic analysis is usually valid. The second design limit state is the factored
load limit state, where the ultimate strength of the girder is checked. Plastic analysis is used for
determining resistance at the factored load limit state, as cracking of the concrete and yielding of
375
the tendons and of the mild steel reinforcement may be expected. The design of D-zones must
also include these two limit states, but the proper method of analysis may not be readily evident to
the engineer. For example, in the design of anchorage zones and deviators, the difference between
the applied service level forces and the ultimate forces from the tendons is small, since at service
load conditions the tendons are generally stressed to 70% or 80% of their ultimate strength. If the
D-zone is designed to be post-tensioned, an elastic analysis may often be valid for the service level
limit state and possibly the ultimate limit state. The more common design approach uses plastic
methods to design the D-zone at the ultimate limit state as a reinforced concrete structural element.
Then the proper distribution of mild reinforcement, and sometimes the addition of post-tensioning
are chosen to ensure adequate performance at the service limit state.
Strut-and-tie modeling (STM) is increasingly becoming the most common method for
plastic analysis of D-zones. The method allows the engineer to visualize the flow of forces in the
D-zone, as well as providing a conservative equilibrium model of the D-zone. Strut-and-tie
modeling can trace its origin to 1899 with William Ritter’s [46] introduction of the truss model for
shear design of reinforced concrete beams and Mörsch’s [47] introduction of the truss analogy for
design of web reinforcement in 1902. Strut-and-tie modeling is a generalized application of the
truss analogy. Marti [48] and Mueller[49] created the strut-and-tie model’s scientific basis for a
rational application, working with Thürliman [50] at the Swiss Federal Institute of Technology,
using the theory of plasticity. Leonhardt had advanced the practical use of STM at the University
of Stuttgart and in his consulting practice [51]. Schlaich et al further developed STM as a
consistent method by which structural concrete can be designed. Their landmark PCI Journal
paper was the first major introduction of STM into US literature [52]. A thorough history of strut-
and-tie modeling can be found in Bergmeister [53].
Schlaich et al give numerous useful examples of strut-and-tie models in the PCI Journal
paper [52], such as the basic models in Figure 6.1. These models can be used to represent the
diffusion of concentrated forces. Some of the strut-and-tie models presented in the PCI Journal
paper [52] were intended for box girder structures, such as the models shown in Figure 6.2. Many
of these models were reproduced with permission, and included in the AASHTO Guide
Specification for the Design and Construction of Segmental Concrete Bridges [8]. The models
shown in Figure 6.2 can be used to calculate the tension forces generated by the transfer of the
web shear forces to the single bearing. A possible arrangement for post-tensioning bars is shown
in part c) of the figure. In the PCI Journal paper, Schlaich et al give the guidelines and criteria
needed to develop an efficient model that will provide good service level performance. Proper

376
assembly of a model requires an understanding of the basic components (struts, ties and nodes).
The elements of the model occupy much of the physical space of the structural element under
design, which differs substantially from the design of an actual truss with essentially one-
dimensional members. Schlaich et al recommend three types of struts, shown in Figure 6.3, that
differ in failure criterion because of their differences in transverse tensile stresses. The
determination of tie forces may also require careful consideration since the tie force may be shared
by an array of bars and uncracked concrete instead of one or several larger bars. Nodes are
defined as singular or smeared nodes. Of paramount importance in nodes with discontinuous or
deviated tension ties is adequate anchorage of the tie. Several models of nodal regions are given
in Figure 6.4.

377
Figure 6.1 Typical strut-and-tie models [52]

378
Figure 6.2 Typical strut-and-tie models for segmental box girder design [52]

Figure 6.3 Strut types [52]

379
Figure 6.4 Anchorage details at nodes [52]

380
 Strut-and-tie modeling is widely used by engineers to design D-zones for the ultimate
limit state, but the effectiveness or accuracy of the design at the ultimate design load level has
been tested only in a few laboratory tests. On the other hand, all D-zones designed by the strut-
and-tie method are subjected to service limit state loads. Therefore, the adequacy of the design of
the D-zone at the service load level is evaluated in every structure, if only visually, providing
engineers with broad experience base. This experience is important because strut-and-tie models
and reinforcing details that do not perform well at the service load level must not be used in the
future. Also, without a past performance record of various types of D-zones, it is often suggested
that elastic analysis be used to locate struts and ties for the ultimate design model. This dual
method of analysis takes great time and effort by an engineer, and eliminates most practical
advantage of the use of strut-and-tie modeling. Strut-and-tie modeling is intended to allow the
engineer to visualize the flow of forces at ultimate load. Adequate service level load performance
is achieved by following proper detailing guidelines based largely on studying the performance of
similar D-zones designed and constructed in the past.
Seven D-zones were instrumented on U. S. 183. The instrumentation was intended to
provide strain measurements in the tension and compression fields of the D-zones. Combined
with the locations of cracks found in the D-zones, the strain measurements provided an indication
of the flow of forces at service load. The instrumented D-zones included two pier capitals, two
pier segments with anchorage diaphragms, two deviators, and one anchorage blister.

6.2 PREVIOUS STUDIES

Few relevant studies have been performed on full scale pier capitals, but many studies
have been performed on superstructure D-zones such as anchorage diaphragms, blisters and
deviators. NCHRP Report 356 [54] titled Anchorage Zone Reinforcement for Post-tensioned
Concrete Girders summarizes a large number of anchorage zone tests, and includes suitable strut-
and-tie models for various anchorage geometries. The NCHRP Report 356 [54] and the Schlaich
et al paper in the PCI Journal [52] should be studied by engineers before designing anchorage
zones by the strut-and-tie method.
Anchorage diaphragms and deviators were instrumented and analyzed by Roberts et al
[7] at the San Antonio Y. The anchorage diaphragm and deviator designs were similar in some
cases to the anchorage diaphragms and deviators on U. S. 183. The D-zones for both projects
were designed by TxDOT engineers. Design improvements were utilized on U. S. 183 based on
the knowledge gained at the San Antonio Y. The strut-and-tie models for a San Antonio Y

381
anchorage diaphragm segment shown in Figure 6.5 give an indication of the numerous possible
locations of struts and ties in a complex D-zone. The strut and tie locations have great influence
on the variability of the total vertical and horizontal tension forces. Model IV at the bottom of
Figure 6.5 was selected as the best model of the four. Model IV assumes a 30o diffusion angle of
the tendon forces to the far face of the anchorage diaphragm from the anchorage plate. The model
also tries to achieve statical determinacy so that the stiffnesses of the struts and ties have little
influence on the forces in the struts and ties. The design of the diaphragm proved to be adequate,
except for localized spalling of the deck concrete in the anchor segment above and in front of the
top tendon anchorages. Reinforcing ties were added in this area on the U. S. 183 anchorage
segments and successfully prevented this type of spalling.

382
Figure 6.5 Strut-and-tie model of a diaphragm from the San Antonio Y [7]

383
 Deviators were also studied by Roberts [7] for the San Antonio Y. Of the various
deviators studied, a beam type deviator shown in Figure 6.6 was instrumented at the San Antonio
Y. This was the same type of deviator used at U. S. 183. Because a large vertical component of
force is applied by the external tendons at a deviator, the optimal type of deviator from a design
standpoint would be a diaphragm type deviator connected full height to the girder web. This type
of deviator provides a nearly direct compressive load path from the tendon deviations to the girder
web. However, from a constructability standpoint, the full height deviator requires that the core
form for the segment be substantially modified from that used for a typical segment. For this
reason, a partial height blister or beam type deviator is preferable to constructors, and will perform
adequately if designed correctly. The instrumented deviators at San Antonio performed well with
little cracking. A strut-and-tie model was developed for the beam type deviator, shown in Figure
6.6, that verified the low measured tensile and compressive stresses. The D-zone forces were
diffused into the girder cross section by web and beam bending.

Figure 6.6 Strut-and-tie model of a deviator from the San Antonio Y (after Roberts et al [7])

384
 The NCHRP Report 356 [54] presents the results of eight half scale intermediate
anchorages or blisters of the types shown in Figure 6.7. The blister in d) of Figure 6.7 most
closely represents the instrumented blister on U. S. 183. The specimens were tested to failure
using an oversized tendon. Cracking generally occurred perpendicular to the centerline of the
tendon near the anchor plate, and parallel to the tendon in front of the blister. Compressive failure
occurred immediately in front of the spiral confinement and was explosive in nature.

Figure 6.7 Types of anchorage blisters [54]

The design and performance of blister type deviators were studied by Beaupre et al [55].
Various reinforcement details were used in the deviators. The deviators had either two or three
tendons that were deviated vertically as well as horizontally. The measured limit states are shown
in Table 6.1. The tendon force D is divided by the nominal design jacking force D0 (80% of

385
tendon ultimate tensile strength) to give the ratios shown in Table 6.1. Oversize tendons were
used to develop the forces needed to test the deviators to the ultimate limit state. The test results
revealed that visible cracking generally occurred when the sum of the tendon forces D were close
to the total design jacking force D0, but varied from this in some test cases because of poor or
unnecessarily conservative reinforcement details. Yielding of the reinforcement occurred at about
1.3 times the tendon force needed to crack the concrete on average. The ultimate limit state of
deviators was exceeded when the tendon force was greater than twice the design jacking force in
most cases. Ultimate strengths were lower for the deviators with tendons deviated vertically as
well as horizontally away from the girder web.

Table 6.1 Summary of tendon forces at various limit states for deviators
tested by Beaupre et al [55]

D/D0 at Each Limit State

Test Microcracking Visible First Yield Ultimate Type of
Cracking Deviator
1A 1.30 1.60 2.33 3 Tendon*
1B 1.23 1.82 2.05 2.41 2 Tendon**
2A 0.87 0.87 1.05 1.45 2 Tendon**
2B 0.96 0.96 1.03 1.27 2 Tendon**
3A 1.03 1.73 2.79 3 Tendon*
3B 1.12 1.33 2.08 3.16 2 Tendon**
4A 0.98 0.98 1.33 2.24 2 Tendon**
4B 1.66 2.03 2.03 2.40 2 Tendon*
5A 0.78 0.78 1.06 2.53 2 Tendon**
5B 1.32 1.49 1.93 2.64 2 Tendon*
Average 1.12 1.26 1.59 2.32
* Tendons deviated vertically and toward web
** Tendons deviated vertically and away from web

6.3 DATA COLLECTION, ANALYSIS, AND DISCUSSION OF RESULTS

The most important instrumentation in the D-zones under study in the U. S. 183 project
were strain gauges. Strain gauges were installed directly on reinforcing bars, as well as on post-
tensioning bars and on the steel pipe ties of the mainlane Y-pier. Concrete gauges (C-gauges)
were also installed to measure strain changes in uncracked concrete. Cracking of the concrete
could be expected at a tensile strain of about 170με, assuming fc=83Mpa. The mild steel
reinforcement could be expected to yield at about 2070με, assuming fy=414Mpa. The measured

386
strains were compared to these two limit state strains, that is the cracking and yielding limit states.
For the anchorage zones and deviators, the measured strains were used to calculate D/D0 at the
cracking limit state. This was done so that the test results could be compared to the results of the
tests by Beaupre et al [55]. D is the sum of the forces from all tendons, and D0 is the sum of the
forces from all tendons if stressed to 80% of ultimate.

6.3.1 Mainlane pier

The mainlane pier, shown in elevation in Figure 6.8, was designed to visually reveal the
flow of forces from the superstructure bearing reactions through the pier capital, and into the
foundation. A solid pier capital with conventional reinforcement was replaced with a Y-shaped
capital and transverse structural steel pipe tension ties across the top of the pier to enhance this
visual effect. The removal of the unessential concrete created the appearance of a strut-and-tie
model bridge pier. Actually, the pier was designed using a frame model, although the proportions
of the capital make such an analysis questionable. The initial intent of the designers was to
control cracking in the pier capital by using post-tensioning. However, the post-tensioning was
later eliminated. Bending in the top concrete elements of the capital was controlled by increasing
the cross sectional area of the steel pipes over that needed for strength. The design was successful
as no cracks were noted anywhere on the pier.

387
 Pipe Gauges (See Figure 6.11)

654 mm 654 mm 654 mm

S105
(S106) C133 C129 C130

(C131) (C132) (C128)

S103
(S104) C122 C126 C124
(C125)(C123) (C127)
S101 C116 C117 C118
(S102) (C119) (C120) (C121)

152 mm
C110 C111 C112 C115

4420 mm
C104 C105 C106 C109
2286 mm
152mm

C101 C102
C103

LEGEND
C-gauge
Strain Gage
( ) Indicates Instrument
on Transverse C L Pier

Figure 6.8 Mainlane pier D6 details

388
 Instrumentation was placed in one quadrant of the pier, as shown in Figure 6.8, as well as
on the structural steel pipe ties. A simple strut-and-tie model of the pier capital was developed,
with the location of the vertical struts made to coincide with the resultant measured compressive
force from loads on the bearings. This model is shown in Figure 6.9. A frame model similar to
the one used by the designers was also developed, and is shown in Figure 6.10. The results
calculated from these two models could then be compared to the measured results, especially the
tension in the pipes, to check the accuracy of the design models. For the STM model in Figure
6.9, the total tie force is 36% of the total bearing load on one side of the pier. For the frame model
in Figure 6.10, the total tie force is 31% of the total bearing load on one side of the pier. The
frame model predicts one-sixth less force in the pipe ties because the bending stiffness of the
concrete members is included in the analysis. This is not significantly less than the STM
prediction because the pipe ties were very stiff and hence lowly stressed. For the measured
results, a 1με strain change in a pipe indicates a force change in each pipe of 1.71kN.

θ = 70ο

Compressive Force

Measured Strain
Distribution

Figure 6.9 Simple strut-and-tie model for mainlane pier D6

389
 6 5 4
5 4
3
Node
3
Element
2
2
1
1
Model

Figure 6.10 Frame model for mainlane pier D6

The greatest application of load on the pier capital occurred when the superstructure was
set on its bearings. The dead load from span D5 was placed eccentrically on the pier capital. This
load case and the strains measured in the pier capital are shown in Figure 6.11. The measured
concrete strains indicated that the stresses from the superstructure dead load were not uniform,
even near the base of the capital, as would be expected because of the bending moment with
respect to the transverse centerline of the pier. For instance, gauges C133 and C129 located on the
exterior face of the concrete showed compressive strains of -72 and -80 με, while the adjacent
gauges C131 and C132 located on the transverse centerline of the pier showed compressive strains
of only -15 and -27 με. Similarly, the average of the strains measured at the base of the capital by
gauges C116, C117 and C118 was -81με, while the average of C119, C120, and C121 located on
the centerline of the pier was only -29με. The strain measurements in the concrete also indicated
that bending moment was present, with respect to the longitudinal centerline of the bridge, in the
capital concrete where it met the column concrete. For example, gauges C119, C120 and C121
gave strains of -17με, -29με and -40με, indicating a nearly linear strain gradient across the width
of the member. No tension was measured anywhere in the concrete by the gauges.

390
 + Tension
Gauge με
2215kN 2215kN
3353mm S105 -
C133 -72
A A C129 -80
C130 -82
S106 -23
C131 -15
C132 -27
S105 C133 C129 C130 C128 -27
(S106)
(C131) (C132) (C128)
S103 -57
S103 C122 C126 C124 C122 -59
(S104) C126 -82
(C125)(C123) (C127)
C124 -76
S101 C116 C117 C118
(S102) (C119) (C120) (C121)
S104 -27
C125 -19
C123 -32
C127 -67

S101 -
C116 -72
C117 -86
C118 -86
S102 -
LEGEND C119 -17
C-gauge C120 -29
C121 -40
Strain Gage
( ) Indicates Instrument
on Transverse CL Pier + Tension, εy=1200με
Gauges με

CL S107 & S108 -3

S109 & S110 12
S111 & S112 46
S113 to S116 94
572mm

S125
S126
S117 & S118 130
S127 S123 S121 S119
S128 S124 S122 S120 S119 & S120 -10
S129
S121 & S122 11
S130
2215kN S123 & S124 32
S117 S125 to S128 100
S118 S129 & S130 116
S113 S111 S109 S107
S114 S112 S110 S108
S115
S116

View A-A

Figure 6.11 Measured strains in pier D6 capital from the superstructure dead load

391
 Strain measurements taken by the gauges on the structural steel pipes indicated that the
pipes, with their rough galvanized surface and intermediate plates, bonded well to the concrete.
Gauges S113 to S116 gave a tension strain measurement of only 94με, 36με less than gauges
S117 and S118, even though they were located only about 50mm from the exterior face of the
concrete. This was a reduction in force of 28%. Similar behavior was measured by gauges S125
to S130, but with a reduction in force of only 14%. In order to directly compare the forces
measured in the pipes from the unsymmetrical dead load case in Figure 6.11 to the results
predicted by the two dimensional models, it is assumed that the strain changes in the four pipes
varied linearly along the longitudinal centerline of the bridge. Therefore, since the strain
measured in the exterior pipe by gauges S117 and S118 was 130με, and the strain measured in the
interior pipe by gauges S129 and S130 was 116με, the strain changes in the other two pipes are
assumed to have been 102με and 88με. Using these assumed strains and applying additional
2215kN loads to the remaining two bearings, the strain in all the pipes would be 218με. The total
measured force in the four pipes would then be 1491kN under the dead load of spans D5 and D6,
or 34% of the applied bearing forces on one side of the pier. This measured pipe force was close
to the mid range average of the forces calculated using the two models from Figures 6.9 and 6.10.
The lower bound strut-and-tie model prediction of 36% was shown to be slightly conservative as
would be expected. The actual load would be about 10% greater than the elastic analysis
prediction. The largest measured concrete strain of -86με at gauges C117 and C118 was well
below the 610με strain that would exist at the usual service level maximum allowable stress of
0.4fc′.
Strain measurements were also taken during a symmetrically placed live load case and
are given in Figure 6.12. The measured strain in gauges S129 and S130 was 17με. Using this
strain for all four pipes, the total pipe force is calculated to have been 116kN, or 40% of the
bearing load on one side of the pier. This force in the pipes exceeded even that predicted by the
strut-and-tie model by around 10%, although strain measurements from gauges S117 and S118
might have had an influence on this result if they had been working. The strain gauges located
within the concrete on the pipes continued to show that the pipes were adhering to the concrete,
and that tension stresses from the pipes were diffused over a short length. In general the pier
behaved as designed, with excellent service level behavior in every respect. The frame model was
suitable for determining bending moments in the capital for the design of reinforcement, but the

392
strut-and-tie model was more accurate for determining the forces in the pipes and would be
improved if concrete tension and concrete creep were considered.

290kN 290kN + Tension

3353mm Gauge με

A A S105 -
C133 -4
C129 -6
C130 -2
S106 0
S105 C133 C129 C130 C131 -2
(S106) C132 -4
(C131) (C132) (C128)
C128 0
S103 C122 C126 C124
(S104) S103 -2
(C125)(C123) (C127)
S101 C116 C117 C118
C122 -4
(S102) (C119) (C120) (C121) C126 -6
C124 -
S104 -
C125 -4
C123 -6
C127 0

S101 -
C116 -6
C117 -6
LEGEND C118 -6
C-gauge S102 -
Strain Gage C119 -2
C120 -4
( ) Indicates Instrument
C121 -
on Transverse CL Pier
CL + Tension, εy=1200με
572mm

Gauges με

143kN S107 & S108 1

S125
S109 & S110 1
S126
S127 S123 S121 S119
S111 & S112 6
S128 S124 S122 S120 S113 to S116 11
S129 S117 & S118 -
S130
147kN
572mm

S117 S119 & S120 0

S118 S121 & S122 1
S113 S111 S109 S107
S123 & S124 1
S114 S112 S110 S108
S115 S125 to S128 12
S116 View A-A S129 & S130 17

Figure 6.12 Measured strains in pier D6 capital from live load case 5

393
6.3.2 Mainlane deviator segment
The mainlane deviator was instrumented with 18 strain gauges installed directly on the
mild reinforcing bars in the deviator itself. The location of the gauges is shown in Figure 6.13.
The instrumentation consisted of three identical planes of six gauges shown in Section A-A of
Figure 6.13. The 19 - 15mm diameter strand tendons each produced a vertical deviation force of
about 447kN, and a horizontal friction force of 145kN. The deviator pipes used were bent on a
radius of only 2m, much smaller than the 7.5m radius drawn in the plans, which concentrated the
deviation forces from the tendon over a very short length of the deviator. This is evident in the
measurements, since gauges S7 to S12 gave strain measurements much larger than the other
gauges located more toward the ends of the pipes. Because of the small radius bend, this deviator
should not be considered typical of deviators in general. The deviator did have significantly large
cracks, and the mild reinforcement did undergo large elastic strains in some cases approaching
half the yield point strain. Ideally, the deviator pipes would be smoothly bent with constant radius
over the full length of the deviator. Even with the proper radius bend, the deviation force may not
be uniform along the length of the deviator because of misalignment between duct and tendon.

394
 A

102mm to
356mm
Strain gauge

140mm
S7 S9
S8 S10 S11S12

22mm bars (typ.)

13mm bars (typ.)

279mm
Tendon 6
A 16mm bars (typ.)
Tendon 5
Tendon 4
Strain Measurements in με, + Tension, εy=2070με
Vertical Reaction Each Tendon = 447kN
External Tendons Stressed DL+PS
Gauge 3 & 4 2 & 5 1 & 6 1 thru 6 (24 hr.)
S1 44 135 57 243 298
S2 17 101 226 369 412
S3 23 32 0 57 78
S4 10 72 186 283 310
914mm S5 -4 6 68 72 95
S6 21 17 17 55 63

S7 129 129 118 395 587

S8 125 441 276 863 1047
S9 2 76 -4 80 144
S10 17 447 384 866 996
S11 -2 19 428 450 511
Gauges S1-S6 Gauges S13-S18 S12 25 29 44 99 148
Gauges S7-S12
S13 32 72 27 137 182
Horizontal Reaction Each Tendon = 145kN S14 10 6 4 21 21
S15 17 17 21 57 74
S16 6 15 25 48 38
SECTION A-A 2 4 34 36
S17 36
S18 15 11 8 34 27

Figure 6.13 Measured strains in the deviator reinforcing bars from superstructure post-
tensioning forces and dead load

Cracking of the segment D5-12 deviator most likely began following the stressing of
tendons 3 and 4. Gauges S7 and S8, located on the top and side of the deviator pipe for tendon 3,
indicated fairly large strain changes, given in Figure 6.13, that could be associated with the
formation of cracks close to the tendon ducts. Cracking in the concrete could be expected at
strains of about 170με, which would be realized during stressing of tendons 2 and 5. As the
remaining tendons were stressed, gauges S7 and S8 continued to show large strain changes, with

395
the final stress in gauge S8 being halfway to the yield point. Using simple linear interpolation,
D/D0 at the yielding limit state can be calculated to be 1.72, which is slightly more conservative
than the average D/D0 at yielding of 1.59 measured by Beaupre et al [55]. Large strain changes in
gauges S10 and S11 after the stressing of tendons 3 and 4 indicated that an unseen horizontal
cracked plane probably existed between deviator pipes, and between the deviator pipe for tendon 1
and the girder web main stirrup bars. D/D0 for the D5-12 deviator is calculated to be 0.58 at the
cracking limit state. This is well below the average D/D0 at cracking of 1.26 measured in the tests
by Beaupre et al. The inclined stirrup bars at the interface of the deviator and the web did not
undergo appreciable tensile strain changes, as indicated by strain changes in gauges S6, S12 and
S18, because of the lack of cracks in the concrete near the web. Strain measured at gauge S12 was
99με at the end of stressing tendons 1 through 6. This strain was below the strain needed to crack
the surrounding concrete in tension. The strain measured at gauge S12 24 hours after the
conclusion of prestressing, and after the girder had been lowered from the truss onto its bearings,
is given in the last column of the table in Figure 6.13 titled DL+PS (24 hr.). This strain was
somewhat higher, at 148με, than the strain measured on the previous day, but still below the strain
needed to crack the concrete. The fine distribution of inclined bars may have limited cracking in
the concrete between tendon 1 and the web. The 24 hour strains at all gauge locations were larger
in general than the strains measured during the stressing operation.
The location of cracks on the mainlane segment D5-9 deviator are shown in Figure 6.14.
Since tendons 3 and 4 were stressed first, the predominant longitudinal cracks occurred over these
ducts and tendons. These cracks were wide open and effectively reduced the bending stiffness of
the deviator beam. The largest cracks were the transverse cracks over the deviator pipes, probably
caused by the concentration of vertical force over the short radius bend in the pipes. This
concentrated force at the bend caused a splitting force that could not be restrained by the few
longitudinal bars on the top of the deviator beam. Large transverse cracks of this type usually do
not occur in deviators with properly bent deviator pipes. Because of the large internal horizontal
crack and the large transverse crack, most of the vertical force from the deviation of the tendons
had to be transferred by the hoop bars to the bottom slab of the girder section, and then diffused
through the bottom slab to the web. The average force measured in the three 16mm bars
instrumented by gauges S8, S10 and S11 was 35kN, so the 18 legs of hoop bars immediately
adjacent to the sharp bend radius of the ducts carried about 634kN of the total 1341kN vertical

396
force. This put high demand on these nine bars, considering that 30 of these bars were included in
the design of the deviator for this purpose.

A A

VIEW LOOKING UPSTATION

VIEW A-A

VIEW LOOKING DOWNSTATION

Figure 6.14 Crack patterns in the segment D5-12 deviator

With some of the mainlane deviator bars at one half yield stress, and considering that 18
legs of bars were taking one half of the vertical thrust from the deviated tendons, the change in
stress in these bars from a simulated HS20-44 live load at midspan was of interest. The strain
397
measurements and forces on the deviator from the live load are presented in Figure 6.15. Given
the end rotation of the girder, the deviated portion of the tendons did not undergo any significant
force change from the live load case. For this reason, the vertical reaction from the tendon was
very small at only 0.3kN, and produced insignificant strain changes in the deviator bars. From a
fatigue standpoint, the highly stressed deviator bars should not degrade. The large cracks in the
deviator were filled with an epoxy to protect these bars from corrosion, although the integrity of
the epoxy was questionable as it never appeared to fully harden.

102mm to
356mm
Strain gauge

140mm
S1 S3
S2 S4 S5 S6

22mm bars (typ.)

13mm bars (typ.)
279mm

Tendon 6
A 16mm bars (typ.)
Tendon 5
Tendon 4

Vertical Reaction Each Tendon = 0.3kN + Tension

Gauge Strain in με
S1 0
S2 0
S3 0
S4 -2
914mm S5 0
S6 -4

S7 0
S8 2
S9 0
S10 0
S11 -2
Gauges S1-S6 Gauges S13-S18 S12 -6
Gauges S7-S12
S13 2
Horizontal Reaction Each Tendon = 3.1kN S14 2
S15 0
S16 4
SECTION A-A 0
S17
S18 2

Figure 6.15 Measured strains in the deviator reinforcing bars from live load case 2

398
 To improve analysis and design of future deviators, general observations about the
structural system and loading need to be made. First, if tendons are stressed in symmetry, the
transverse loading on the girder from the deviator forces is symmetrical. Therefore, no shear will
be present in the central portion of the deviator beam or in the top slab. Shear forces from the
deviator will diffuse only toward the girder web. Second, the compressive stiffness of the girder
web will predominately dictate the flow of forces from the deviator, so a two dimensional analysis
of the deviator and surrounding girder can be used to determine moments, shears and axial forces
at points away from the deviator. Section properties of the deviator beam, girder web and top slab
can be calculated assuming the forces from the deviator diffuse longitudinally into the girder at a
30o angle with respect to the transverse section through the centerline of the deviator. The
reaction to the vertical components of the deviated tendon forces can be assumed to be located at
the center of gravity of the girder in the web.
The moment couple created between the deviator forces and the web reaction will be
resisted by an axial force couple between the top slab and deviator beam, and transverse bending
of the girder. The amount of transverse bending calculated in the deviator beam is somewhat
dependent on whether cracked or uncracked section properties are used in the analysis. Cracks
will most likely be present near the deviator pipes even at service load, so a reduction in cross
sectional stiffness may be warranted in this area. The results of an analysis for service level forces
from the D5-12 deviator are shown as applied loads in Figure 6.16. The deviation forces from the
tendons were easily calculated since the deviation angles were known, and changed little during
the stressing of the tendons. When calculating the deviation forces for the girder at ultimate, the
forces in the tendons and the deviation angles are not as easily calculated. It is unlikely that the
external tendons will ever reach their ultimate stress because of the partially bonded nature of
external tendons and the imposed limits on the girder's sectional ductility at ultimate moment.
To be very conservative, the ultimate force for the tendon can be used, but the deviation
force depends largely on the deviation angle at the ultimate load level. Plastic rotations can be
calculated and the deviation angles at ultimate can be determined. Also of utmost importance is
the calculation of horizontal force on the deviator from changes in tendon force at ultimate. The
force changes in the inclined and horizontal legs of the tendon on either side of the deviator will
not be equal, and depend greatly on whether the tendon is assumed to be fully bonded to the
deviator, or more realistically allowed to slip. For this reason, a longitudinal strut-and-tie model
should be developed for the ultimate load design of the deviator, bottom flange, and web.

399
 1905kN
429kN

447kN

440kN
154kN

422kN
S7 S9
1362kN 78kN 149kN S12
S8 S10 S11
1825kN

HALF SECTION THROUGH CENTERLINE OF DEVIATOR

0kN
Legend

kN
2

42 9
-20

Tension Tie

Compression Strut
kN
6kN

674
2
-22

1516kN 1528kN 1299kN 1134kN -248kN

kN
120

kN
kN

-44

1362kN
04
284kN

406kN
338kN

281kN

-642kN
4

830
68

1k
5 kN
-1
- 11

kN 43
-395kN -657kN -1367kN -1539 5k
N N
6k
627kN

kN

73
908kN

1825kN
0
kN

-2
81

kN
34 1
1
-1

-1168kN -458kN 298kN

CALCULATED MAJOR STRUT AND TIE FORCES

Figure 6.16 Strut-and-tie model for the segment D5-12 deviator

400
 The service level analysis results for the D5-12 deviator, shown in Figure 6.16, revealed
that about half of the moment from the deviator force and web reaction was taken by transverse
bending of the girder. The axial force in the deviator beam and top slab must have been 1037kN
to resist the entire moment. From Figure 6.16, the axial force in the deviator beam was calculated
to be 1825kN compression minus 1362kN tension, or 463kN in compression. Full elastic section
properties were used in the analysis, so the moment calculated in the deviator beam would be
conservative, and shear and bending in the web would be unconservative. Because of the high
shear in the deviator between the deviator pipes and the web, the negative moment in the deviator
beam becomes a positive moment at the juncture of the deviator with the web. This places a point
of contraflexure between the deviator pipes for tendons 1 and 6 and the girder web. This should
be considered when selecting struts and ties during the deviator design. Since the forces and
moments near the deviator have been determined for the service load case, a strut-and-tie design
can proceed once the limits of the D-zone have been determined. Since only horizontal struts and
ties can be used in the central portion of the deviator beam because of the lack of shear forces, the
exact limit of the D-zone within the central portion of the deviator beam is unimportant to the
design. The limit of the D-zone within the girder web has been selected as one deviator length, or
914mm, up the web from the web to deviator beam juncture.
The strut-and-tie model shown in Figure 6.16 was developed based on the calculated
forces and moments, and on the assumption that horizontal cracks were present between deviator
pipes near the sharp radius bend, and at the locations shown in Figure 6.14. The model assumed
that shear forces in the deviator beam near the pipes were taken to the web only in the portion of
the deviator beam beneath the deviator pipes because of the extensive cracking. For this reason
the model would also be suitable under ultimate load condition. The vertical tendon forces are
transferred to the deviator only through the vertical reinforcing. Because of shear forces beneath
the pipes, the tensile forces in the vertical ties are largest below the pipes. The flow of transverse
tensile force above the pipes is deviated to the bottom of the girder by compression struts near the
point of inflection. The main tension tie and the main compression struts intersect orthogonally
near the point of inflection, indicating high shear and low moment. The locations of tension ties in
the model were chosen such that reinforcement could be designed directly from the calculated
tensile values. Failure criteria for the struts and ties must be checked during the ultimate load
design, as well as the proper anchorage of reinforcement. The D5-12 deviator had heavy
congestion of bars because of the web stirrups, and web to bottom flange fillet reinforcement.
Modification of the typical web and fillet reinforcement details would have been advisable over

401
the length of the deviator so that the designed tie reinforcement could be placed. The 13mm C-
shaped bars placed on an angle, seen in Figure 6.13 at the junction of the deviator and the web,
were difficult to install because of the heavy web and fillet reinforcement running perpendicular to
the direction of the actual tensile stress field. The short 90o bend of the 13mm angled bars was
also inadequate to develop the forces calculated in the strut-and-tie model, and could not be relied
upon to develop the forces the model would predict at an ultimate load state. Forces would
redistribute as anchorage failure of the 13mm bars proceeded.
Forces in the instrumented reinforcing bars could be calculated based on the strain
measurements, and estimated for the adjacent reinforcement. Reinforcing bar forces were most
accurately calculated at crack locations, for example, at gauge S8 shown in the top section of
Figure 6.16. Using the average of the 24 hour strains measured by the gauges S2, S8 and S14, and
multiplying this average strain by the total reinforcing bar area (20 legs of 16mm bars) and the
modulus of elasticity of the steel, the tie force was calculated to be 408kN. This was about 20%
higher than the result of 338kN predicted by the STM. The correlation between these calculated
and measured values is good, considering the crudeness of the calculation of measured force in the
reinforcement. A similar calculation of the tie force at gauge location S10, shown in the top
section of Figure 6.16, gives a measured force of 371kN. This is 32% higher than the 281kN tie
force predicted by the STM. The same calculation of measured force at the tie with S11 proved to
be inaccurate because of the lack of cracking at gauge location S5 and S17. The force was
calculated to be only 89kN. This assumed that no tension was taken by the concrete. The STM
predicted 406kN in the tie at gauge S11. The actual state of stress in the concrete at gauges S5,
S11 and S17 was more complex than at the other vertical bar gauges because of their close
proximity to the intersection of several major struts and ties. This apparently resulted in fewer
cracks at the gauge locations.
Calculation of the tie forces at gauges S7 and S9, shown at the top of Figure 6.16, proved
to be inaccurate because the strain gauges were actually placed on the 180o bends of the vertical
tie reinforcement. Measurements would have been more easily interpreted if gauges had been
placed on the heavy transverse bars as well. Even with the presence of a large crack at S7, the
estimated reinforcing bar force based on the average of gauges S1, S7 and S13 was only 318kN.
The STM predicted 1516kN. The STM force was conservative because the bending moment in
the deviator calculated in the preliminary analysis used full uncracked section properties for the
deviator. The large crack at S7 would have reduced the bending moment and increased the axial
compressive force in the deviator, effectively reducing the tension at S7. Tension strains

402
measured at gauge S12 and at the adjacent gauges S6 and S18 were less than the strain required to
crack the concrete. Therefore, the tension force in the major inclined tension tie, calculated to be
1205kN by the STM at the bottom of Figure 6.16, was not accurately calculated based on the
measured strains since the tensile contribution of the concrete was neglected. Also, because of the
poor placement of the inclined bars, gauges S6, S12 and S18 were actually located within the web,
and very close to the heavy web stirrup reinforcement.
The lack of cracking at some gauge locations made the comparison between the
measured strains and calculated forces, and the STM forces difficult except at the vertical ties
gauged by S8 and S10. At these two locations the STM was unconservative, at least in the upper
portion of these vertical ties. As seen at the bottom of Figure 6.16, the forces in the vertical tie
bars are drastically different in their top and bottom halves because of the shear carried in the
lower portion of the deviator. Such a drastic force change predicted by the STM in the short tie
bars was probably unrealistic. The actual force distribution in the vertical tie bars was probably
more constant, resulting in less force in the bottom of the bars and more at the top. The vertical
bars should be sized based on the forces calculated in the bottom of these ties using the STM,
giving a conservative design.

6.3.3 Mainlane anchor segment

The mainlane anchor segment D5-16 was instrumented with strain gauges on the
reinforcing bars in the heavy end diaphragm. The location of these gauges is shown in Figure
6.17. Gauges were placed on bars at locations where cracks were expected to occur. The
measurement of strain at a crack location allows the total tensile force present to be calculated
accurately in the instrumented bar. The reinforcement in the diaphragm was designed to resist
tensile forces caused by the post-tensioning that act adjacent to the far face of the diaphragm and
away from the tendon anchorages. The gauges were located on bars on this face of the diaphragm,
as is indicated in Section A-A of Figure 6.17. Tensile strains expected to occur in the top flange
were measured by gauges S25, S26 and S27. Tensile strains expected to occur in the bottom
flange were measured by gauges S22, S23 and S24, with gauge S22 actually placed on the lower
horizontal leg of a closed tie bar. The remainder of the gauges were placed to measured tensile
strains in the vertical direction.

403
 Deviated External Tendons
13mm bar (typ.) (3850kN Each Tendon)

T4 T5 T6

25mm bar (typ.)

16mm bar (typ.)

35mm bars Future tendons

Strain Gauges

T11 T12 T13 T14 Bottom Slab Internal Tendons

(3850kN Each Tendon)
HALF SECTION
Looking Upstation
CL Pier
1448mm 152mm
extension A
S27 S28 S35 S40 S45
457mm

S26
T1 - T6
S25 S34 S39 S44

S38 S43
1067mm

S31
S33

S30
S32 S37 S42
S24
610mm

T7 - T14 S19 S20 S23 S29 S36 S41

S21

230mm S22
A
Gauges located Gauge Locations
on bars in this plane (Looking Upstation)
SECTION A-A

Figure 6.17 Mainlane anchor segment D5-16 details

Visual inspection of the crack patterns in the anchor segment shown in Figure 6.18
indicated that tensile forces were present on most of the downstation face of the diaphragm.

404
Cracks did not propagate into the top flange because of residual compression from the transverse
prestressing. From Section A-A of Figure 6.18, it can be seen that vertical tensile forces were
present over most of the length of the segment in the top of the diaphragm. Cracks from tensile
forces in the bottom of the diaphragm propagated only about 250mm along the length of the
segment. The bearing reactions from girder dead load were not present when these cracks were
formed. The girder was supported by jacks under the wings near the webs until all tendons were
stressed. Additional cracks were noted in the bottom flange of segments D5-15 and D5-14, as
shown in Figure 6.19. These cracks were located directly beneath the internal bottom slab tendon
ducts, and did not propagate to the top surface of the bottom slab.

CL Pier

HALF VIEW
SECTION A-A Looking Upstation

Figure 6.18 Crack patterns in segment D5-16 diaphragm

405
 152mm
1616mm 2515mm 2515mm

A
B

D5-16 D5-15 D5-14

A

SECTION A-A BOTTOM VIEW (SOFFIT)

Figure 6.19 Crack patterns in the bottom flange of segments D5-16, D5-15 and D5-14
Table 6.2 contains the strains measured during the post-tensioning of the six deviated
external tendons and the wing tendons. Since the stressing operation occurred over several hours,
the column titled "Sum T1 to T6" includes strain changes measured over the entire period, and
does not always match the sum of the strains from the three columns of data to the left. The
deviated external tendons were stressed in pairs beginning with tendons T3 and T4. The most
significant tensile strain changes during the post-tensioning of tendons T1-T6 were measured by
gauges S31 and S33, although the strains were only 36% and 17% of the nominal yield point
strain of the bars at 2070με. D/D0 at cracking was quite small since cracking initiated during the
stressing of Tendon T3 and T4. The strain at S31 was 219με following the stressing of T3 and T4,
which exceeded the approximate cracking strain of 170με. The tensile strains measured by gauges
S30 and S32 toward the bottom of the diaphragm were much less in magnitude, even though a

406
crack existed at this location. The strain changes measured by gauges S30 to S32 for each pair of
deviated external tendons were generally similar, regardless of the location of the tendon pair.
This may be due in part to the propagation of cracks at these locations as stressing
progressed. Strains measured by the horizontal line of gauges S31, S33, S38 and S43 indicated
that vertical tensile strains in the diaphragm decreased toward the web. The strain measured by
gauge S31 was 745με from the stressing of tendons T1 to T6, while the strain measured by gauge
S43 was only 70με. A similar decrease across the width of the diaphragm was measured by the
horizontal line of gauges S30, S32 and S37 with measured strains of 125με, 67με and -23με
(compression) respectively, with gauge S42 located near a crack between the two empty future
post-tensioning ducts measuring 114με. The line of gauges S29, S36 and S41 did not measure
significant tensile strains, with gauge S41 measuring compression. Gauges S34, S39 and S44,
located directly in front of the anchorage plates but on the opposite face of the diaphragm,
measured about 14με each from the stressing of each pair of tendons regardless of the pair.
Gauges S28, S35, S40 and S45 recorded tensile strains of about 18με when the tendon nearest
each gauge was stressed, and less strain when other tendons were stressed.
In general, the vertical strains measured at each corresponding gauge location were
similar, regardless of the pair of deviated external tendons stressed. Once again, this may be due
to the propagation of cracks as stressing progressed, which would relieve tension in the concrete
and increase tension in the gauged steel bars.

407
 Table 6.2 Measured strains in anchor segment D5-16 reinforcing bars from
stressing post-tensioning tendons in microstrain

Strains in microstrain, + Tension

Deviated External Tendon Stressed Wing Tendons
Gauge T3, T4 T2, T5 T1, T6 Sum T1 to T6 T15 - T18
S19 8 2 6 11 -2
S20 8 11 8 27 2
S21 10 4 6 23 -2
S22 21 17 13 67 4
S23 27 19 8 80 42
S24 29 17 4 68 29
S25 36 13 -8 51 13
S26 6 -6 -10 2 13
S27 13 0 -15 4 -11
S28 23 19 10 53 6
S29 2 8 6 15 2
S30 44 55 27 125 -6
S31 219 276 264 745 -11
S32 15 34 15 67 -4
S33 67 112 152 342 6
S34 17 17 13 48 -4
S36 -4 -10 -6 -11 -2
S37 4 -21 -4 -23 0
S39 13 13 2 23 -10
S40 -49 15 17 -10 -8
S41 -8 -6 -11 -25 2
S42 2 68 19 114 -8
S43 25 21 27 70 -8
S44 17 11 11 38 0
S45 11 13 17 44 8

Gauges S25, S26 and S27 measured transverse strains in the top flange from the
tensioning of external deviated tendons T1 to T6. Strains in these three gauges averaged 18με
when T3 and T4 were stressed, 2με when T2 and T5 were stressed, and -11με when T1 and T6
were stressed. The strains in these gauges also indicated that small values of transverse positive
bending moment were present. Horizontal strains measured in the bottom flange by gauges S22,
S23 and S24 averaged 26με from tendons T3 and T4, 18με from tendons T2 and T5, and 8με from
tendons T1 and T6. The strains from these gauges clearly indicated that the response of the girder
near the diaphragm was different for each pair of deviated external tendons. Strain changes in the
diaphragm from the tensioning of the four internal wing tendons were small, but the horizontal

408
strain changes in the top and bottom flanges were larger. The average strain change in gauges
S25, S26 and S27 from the wing tendons was 5με, which is very small, but with positive bending
indicated in the top flange. The average strain change in gauges S22, S23 and S24 was 25με from
the wing tendons, with negative bending indicated in the bottom flange.
The strains measured in all gauges from the stressing of the eight bottom slab internal
tendons T7 to T14 are given in Table 6.3. In general, the change in strain at each gauge location
was highly dependent on which pair of bottom slab tendons was stressed. Gauges S19, S20, S21,
S22, S29, S36 and S41 were located directly in front of the anchor plates for the deviated external
bottom slab tendons, and all measured similar tensile strains averaging 130με for all eight tendons.
The gauges on the vertical bars near the top of the diaphragm and in the top flange measured very
little strain change.
Gauges S31 and S33 measured more substantial tensile strain changes from the stressing
of tendons T10 and T11, at 51με and 112με respectively, but measured progressively more
compressive strain changes as the other tendon pairs were stressed. Strains measured by S31 and
S33 were -44με and -30με from the post-tensioning of tendons T7 and T14. These strain changes
in S31 and S33 were small, considering the gauges were located at a crack created during the
stressing of the external tendons. Gauge S43 gave similar results to S31 and S33, but with smaller
magnitude. Gauges S30 and S32 measured strain changes from each pair of tendons that were not
easily interpreted, but appear to be influenced by the proximity of the tendon anchorage to the
strain gauge. S37 gave progressively more compressive strain changes as the tendons closer to the
webs were stressed, while S42 gave progressively more tensile strain changes as the tendons
closer to the webs were stressed.
Strains in gauges S25, S26 and S27 were quite small for all stressed bottom slab tendons,
averaging near zero. Horizontal strains measured in the bottom flange by gauges S22, S23 and
S24 were relatively large, and averaged 47με from tendons T10and T11, 68με from tendons T9
and T12, 38με from tendons T8 and T13, and -16με from tendons T7 and T14. The trend for
transverse strains in the bottom flange was toward compression as tendons closer to the webs were
stressed. The strain response to all eight bottom flange tendons as a group was only notable in the
bars located in or near the bottom flange, with the exception of the bar gauged by S33.

409
 Table 6.3 Measured strains in anchor segment D5-16 reinforcing bars from
stressing post-tensioning tendons (continued) in microstrain

Strains in microstrain, + Tension

Internal Bottom Slab Tendon Stressed
Gauge T10, T11 T9, T12 T8, T13 T7, T14 Sum T7 All Dead Load
to T14 Tendons & Prestress
S19 68 34 27 15 143 150 143
S20 59 32 25 13 133 163 156
S21 46 32 25 15 120 141 122
S22 32 42 34 -13 101 182 114
S23 57 91 46 -21 186 317 258
S24 53 72 34 -15 154 258 201
S25 2 4 4 0 4 74 -2
S26 -2 0 -2 -2 -11 8 70
S27 8 4 2 -2 10 -6 -27
S28 0 0 0 -2 -4 57 67
S29 53 68 40 19 182 198 181
S30 0 -6 -6 29 15 129 133
S31 51 11 8 -44 15 730 842
S32 11 8 17 8 2 65 68
S33 112 108 25 -30 230 578 642
S34 -105 -2 2 -2 -106 -63 -25
S36 23 25 36 27 125 112 116
S37 28 29 19 -10 65 38 63
S39 15 13 4 -4 30 38 53
S40 -6 15 -8 -10 -30 -48 -57
S41 27 29 30 25 108 84 87
S42 -13 -10 4 6 -8 108 65
S43 10 8 0 -13 0 55 68
S44 4 0 0 -6 0 36 46
S45 0 0 -2 -4 -4 53 46

From the column titled "All Tendons" in Table 6.3, most of the gauges recorded tensile
strains at the end of stressing. As would be expected, the measured strains were larger at the
gauge locations where cracks were present, and smaller where few or no cracks were present, such
as near the webs and in the horizontal bars in the top flange. When the girder was set on its
bearings, the strains in the last column of Table 6.3 were recorded. These strains do not differ
greatly from those measured at the completion of stressing, except in the vertical bars close to the
bearing reaction, and in the transverse bars in the bottom flange.
When developing a strut-and-tie model for the anchor segment, the limits of the D-zone
and the forces on the D-zone must be determined. The D-zone for the anchor segment ends in the

410
girder at point where plane sections remain plane from the applied prestressing forces. Using the
assumption of a 30o diffusion, the force from tendon group T7 to T10 or T11 to T14 diffuses fully
to the edge of the wing tip defined by the effective flange width method at a section 11.4m from
the anchor plates. This distance may be reduced for the service load case after inspection of the
compressive force paths emanating from the anchor plates, including those for the wing tendons,
through sections approaching the B-zone boundary. At the chosen section the linear service load
stress distribution from dead load and prestress can be resolved into numerous concentrated loads
to be applied to the strut-and-tie model, as shown in Figure 6.20.
The D-zone boundary for the service load case in Figure 6.20 was selected to be at the
centerline of the deviator to simplify calculations. Other loads on the D-zone include the prestress
loads at the anchor plates, dead load, and bearing reactions. If symmetry is used and the girder is
split down it's longitudinal centerline, the transverse axial force distribution and bending moments
from the top and bottom flanges must be included as self-equilibrating reactions. The transverse
forces in the top and bottom flanges must provide a moment to balance the eccentricity between
the longitudinal post-tensioning forces at the anchor segment and the diffused cross sectional post-
tensioning and dead load forces. This moment has been calculated, and is given in Figure 6.20 for
the service load case. The distribution of the transverse forces and moments in the top and bottom
flange must be determined using equilibrium of the girder in cross section, and by assuming a
transverse stress distribution along the length of the girder, as shown in Figure 6.20. A shorter
assumed diffusion length will result in larger calculated transverse forces. A longer assumed
diffusion length will result in smaller calculated transverse tensile forces at sections away from the
anchor segment.
A strut-and-tie model can be used to conveniently solve for these transverse diffusion
forces, but the model should make provision for dead load forces since post-tensioning and dead
load were applied simultaneously. Assuming symmetrical stressing of tendons during jacking is
required, all prestressing should be assumed to be applied simultaneously to reduce calculation
effort.

411
 Resolution of linear stress
567kN distribution into equivalent
1322kN
concentrated loads
3561kN

10m 3561kN
3561kN
mm
N/
0 9k 3588kN
= 0 .1
lo ad 3588kN
e ad
rd 3588kN
de 434kN
lf gir 3588kN
e ha
On
567kN

1322kN

Stressing
3 @ 3588kN
forces
Assumed Longitudinal
Stress Distribution,
Sum of Top and
4 @ 3588kN C Bottom Flanges
T
Moment Couple = 36500kN-m

2063kN Dead Load Reaction

(on bearing or temporary supports)

Approximate Compressive Force Paths (Dead load and Prestress)

Top and bottom slab forces to be determined by equilibrium using actual support conditions

Figure 6.20 Post-tensioning and dead load compressive force paths during
construction of span D5
The distribution of dead load force to the support jacks under the girder wings was
difficult to determine during post-tensioning of the girder. The girder was initially supported
uniformly along its length by the truss jacks under the girder wings, as shown at the top of Figure
6.21. As the girder was post-tensioned, some of the girder dead load was carried by the girder
itself in bending, loading the wing support jacks near the piers and unloading the wing support
jacks near midspan. This case is shown in the middle view of Figure 6.21. The deflection
measurements taken during post-tensioning and thereafter indicated that the full dead load of the
girder was not transferred entirely to the few wing jacks near the girder ends at the completion of
post-tensioning. The support reactions during post-tensioning could not be accurately calculated.

412
 69kN-m
2063kN Girder Dead Load

206kN

2063kN
138kN-m
Reaction at Jacks
(Distributed Along Span) 206kN

Initial Support Condition on Trusses

Partial Girder Dead Load

Carried by Girder Bending
Partial Girder Dead Load on Jacks

2063kN
Reaction at Jacks
(Concentrated Nearer Girder Ends)
Intermediate Support Condition on Trusses

2063kN Girder Dead Load

2063kN
Final Support Condition on Bearings Reaction at Bearings

Figure 6.21 Dead load support conditions during construction of span D5

One of the difficulties in comparing measured girder behavior and reinforcement strains
with STM calculations is that the STM is a lower bound plasticity model most suitable for the
ultimate limit state. Measurements are obtained on the bridge for the service limit state and are
not directly comparable. For the calculated ultimate limit state, where the bottom slab internal

413
tendons and perhaps the deviated external tendons are assumed to be at ultimate stress, the
location of the boundary of the D-zone could be chosen from service load stresses to reduce
calculation effort. Since the STM is a lower bound calculation procedure, considerable freedom is
given to model selection. If the limit of the D-zone is chosen too close to the anchor segment,
reinforcement needed to control tensile stresses in the actual D-zone may not be included in the
design.
To illustrate the process, the full plastic moment capacity of the section has been assumed
to develop at a section near the deviator, as shown in Figure 6.22. Therefore, the ultimate tendon
forces are balanced by compression across the full width of the top flange. The forces from the
tendons in Figure 6.22 are treated as external loads on the concrete girder. The compressive force
paths from the tendon loads are shown as heavy lines in the figure. The compressive force paths
are deviated from the top flange to the web by shear forces. A transverse moment couple of
97,400kN-m is required to offset the moment from diffusion forces.
4791kN
4791kN
4827kN
4827kN
10m 4827kN
ad
e Lo 4827kN
Li v 4791kN
lus
adp
o
dL Shear
ea 1778kN
e dD
tor (at Deviator)
F ac

4 @ 413kN
(Typ. Development)

3 @ 4827kN

Assumed Longitudinal
Stress Distribution, Sum
4 @ 3588kN C of Top and Bottom Flanges
T
Moment Couple = 97,400kN-m
Reaction

Figure 6.22 Span D5 compressive force paths at ultimate load

414
 A strut-and-tie model can be used to calculate the transverse forces and vertical forces in
the girder and diaphragm. Struts and ties for the model should be located on each face of the
diaphragm. A STM of forces on the diaphragm interior face is shown in Figure 6.23(a). Struts
and ties should also accommodate bearing reaction forces, as shown in Figure 6.23(b). The strut-
and-tie model for the D5-16 anchor segment is complex, so strut locations should be limited to
those along the compression force paths, and those required for equilibrium. Ties should be
located and oriented where reinforcement can be effectively placed. Transverse stresses caused by
internal spreading within the compressive struts must also be considered, especially near
anchorage plates and in the highly stressed bottom flange.

To Web
Tendon Anchorage

Tendon Force Path at

Face of Diaphragm
Tension Tie at Face
of Diaphragm

(a) Ties at Interior Face of Diaphragm

(from tendon forces)

From Web

(b) Struts and Ties in Diaphragm

(from shear force and bearing reaction)

Figure 6.23 Segment D5-16 diaphragm tie locations at ultimate load

415
 The bottom flange cracks shown in Figure 6.19 extended 5.3m along the span indicating
that the assumed 1.6m length for the design D-zone for the ultimate load case was unrealistically
short for the service load case. The nonuniform distribution of bottom flange stresses from post-
tensioning forces shown in Figure 6.24 also indicated that significant transverse tensile may have
existed well beyond the designed D-zone. Transverse tensile strains in the bottom flange, at the
section shown in Figure 6.24, could have been as high as 100με from the Poisson effect alone.
Transverse tensile diffusion strains from the bottom flange tendons and the deviated external
tendons would add to the Poisson’s strains.
From inspection of the compression strut locations in the bottom flange at service load,
shown in Figure 6.20, compressive stress in the bottom flange was high at locations between the
anchor segment and the deviator. In fact at the instrumented section 1700mm from the anchor
plates, shown in Figure 6.24, maximum measured bottom flange strains converted to stresses
exceeded the allowable compressive service level stress of 15.2 MPa by 30%. The combination of
high longitudinal compressive stress and transverse tensile stress resulted in longitudinal cracking
of the bottom flange. The tensile strains measured in the anchor segment diaphragm itself
indicated that the diaphragm design was quite conservative, with the highest measured strain of
842με at gauge S31 being only 41% of the yield strain of the bar. Based on this steel strain, first
yield would be realized at D/D0 of 2.44, which is high.

416
 16 15

3 2 1

10 9 8 7

SECTION B-B
-20 Allowable Compressive
-18 Stress (-15.2MPa)
Bottom Flange Stress (MPa)

-16 Legend
-14 Tendons 1 - 6
-12
(+ tension)

Tendons 15 - 18
-10
Tendons 7 - 14
-8
-6 Sum of All Tendons
-4 Strain gauge
-2
0
2

Figure 6.24 Measured bottom flange longitudinal stresses from post-tensioning forces

6.3.4 Large ramp pier capital

The large ramp pier capital segment PC16-8 was designed as an anchorage zone for the
pair of U-shaped vertical pier post-tensioning tendons, and as the anchorage zone for the 16
threadbars that were used to create a moment connection with the superstructure. In addition, the
solid pier capital had to transfer moments from balanced cantilever construction to the voided
typical pier segments beneath it. Concrete strain gauges were located as shown in Figure 6.25 to
measure the distribution of vertical strain changes throughout the pier capital. Gauges C541 to
C544 were installed to measure strains between the anchorage plates for the groups of threadbars
and the main pier post-tensioning tendons. Gauges at Sections C-C and D-D were placed to
measure the diffusion of vertical forces near the interface of the capital and top pier segment.
Strain gauges were also installed directly on the reinforcing bars and threadbars in the capital. The
locations of these gauges are shown in Figure 6.26. The gauged bars in Sections B-B and C-C in
Figure 6.26 were designed as tie bars to resist tensile forces in the capital from the various applied
loads.

417
 Up Sta B
C539 C533
C540 C534

C541
C544
Top View of Large Ramp
Pier Capital

C543

C542
Segment PC16-8
A A
C536
C538
C537 C535
DYWIDAG Tie-Down Bars
for superstructure moment B
connection
Tendon
C542 Centerline
(C543) C541
anchorage
Compressive Struts
(C544)
914mm

C533 C543 C542

(C539) (C544) (C541)
C535
(C537)
830mm

D D
C536 C534
(C538) (C540)
C C
75mm

Section B-B Section A-A

Post-Tension
Up Sta Up Sta
Tendons
C501 C517
C509 C525
C508
C502 C524 C518
C516 C532
C510 Hidden Line for C526
the Voided Core of
Segment PC16-7 C523 C519
C507 C515 C511 C503 C531 C527

C514 C512 C528

C506 C504 C530 C520
C513 C522
C529
C505 C521
2390mm 2510mm

Section C-C Section D-D

Figure 6.25 Large ramp pier capital segment PC16-8 strut gauges

418
 S530 S527
S531 S529 S528 S524
A S536 S521 A B
S534 S523
S537 S535 S522 S520 Up STA
S538 S517
S538 S517
S539 S518
S540 S519
C C
S532 S525 S535 S522
S539 S518
S537 S536 S521 S520
S533 S526
S540 S519 S534 S529 S528 S523
A S530S527 A
B B S531 S524
S532 S525
S533 S526
Elevation of Segment PC16-8 View A-A B
(Looking Up Station)

35mm bar

P.T. bar tie-downs Up Sta

Up Sta
S501
32mm bar S544 S541

S508 S502

S509
S510 S545
S549
S514 S515

S546
S548

S552

S550
S507 S513 S516 S503

S551
S511
S547
S512
S506 S504
Post-Tension
S505 Tendons S543 S542
16mm bar

Section B-B Section C-C

Figure 6.26 Large ramp pier capital segment PC16-8 tie gauges
The measured strain changes from post-tensioning of the main pier tendons are shown in
Table 6.4. The strains measured by gauges C541 to C544 between anchor plates and strains
measured at the bottom of the threadbars were similar at -92με and -86με respectively. This
would correspond to a maximum compressive stress of 3.9MPa or 8% fc′. Compressive vertical
strains decreased rapidly toward the outer edge of the pier capital, with strains at gauges C533,

419
C535, C537 and C539 averaging only -11με. Compressive vertical strains also decreased rapidly
toward the top of the capital, as seen by the strain measurements on the threadbars, even though
these gauges were located very close to the main pier tendon anchor plates. Vertical concrete
strains measured at the base of the capital were difficult to interpret because of the large number of
damaged gauges. Gauges C501 and C505 at the bottom of the capital measured -76με while
gauges C503 and C507 measured -51με, indicating that the post-tensioning force had not fully
diffused. The tendon anchorages were located more toward the longitudinal centerline of the
bridge, as can be seen in the Top View in Figure 6.25.
Many of the gauges located on the tie reinforcing carried appreciable tension, such as
S548, S552, S550 and S546 located on longitudinal bars at the top of the pier at 159με. This
strain approaches the 170με required to crack the concrete, but is well below the 2070με yield
point of the reinforcement. Gauges S541 to S544 located near the anchorage hook for these same
bars did not go into appreciable tension. The transverse bars gauged by S545, S549, S551 and
S547 at the top of the capital were actually within a compression strut and measured -15με. At the
bottom of the capital, both the longitudinal and transverse bars gauged by S509 to S516 were in
tension.
The transverse bars averaged twice the tensile strain of the longitudinal bars, at 66με
versus 31με respectively. Tension was present in the circumferential bars gauged by S501, S505,
S503 and S507, with the longitudinally oriented gauges measuring twice the strain of the
transverse gauges at 100με and 52με respectively. These strains were well below the 170με
required to crack the concrete, and no cracking was noticed beyond those cracks that formed
during curing of the segment.

420
 Table 6.4 Measured strain changes in segment PC16-8 gauges from post-tensioning forces

+ Tension + Tension
Strut Gauges Microstrain Tie Gauges Microstrain
C501,5 -76 S501,5 52
C503,7 -51 S503,7 100
C502,4,6,8 6 S502,4,6,8 40
C509,13 - S509,10,11,12 66
C511,15 -34 S513,14,15,16 31
C510,12,14,16 -78 Threadbar Top -5
C517,21 - Threadbar Mid. -49
C519,23 -11 Threadbar Bot. -86
C518,20,22,24 -56 S541,42,43,44 7
C525,29 - S545,49,51,47 -15
C527,31 -29 S548,52,50,46 159
C526,28,30,32 -34
C533,35,37,39 -11
C534,36,38,40 -78
C541,42,43,44 -92

The measured strain changes from the unbalanced placement of the farthest
superstructure segment P16-17 are given in Table 6.5. Segment P16-17 was placed on the
upstation side of the pier, and produced a substantial moment of 7366kN-m and compressive force
of 285kN. Stresses at the top of the capital were nearly linear, as can be seen in Figure 5.67, and
also at a section in the pier 610mm below the bottom of the capital, as seen in Figure 5.66. Strains
in each group of threadbars were nearly uniform along their height, but with the gauges at the tops
of the threadbars taking less strain than those at the middle or bottom, be it tensile or compressive.
For example, gauges S517, S520, S521 and S522 averaged -17με change at the top of the
threadbar group, while gauges S518 and S519 measured -21με and -19με at the middle and
bottom respectively. A simple P/A+Mc/I type stress calculation predicted that the strain change at
the top of the threadbars should have been about -33με, indicating that the threadbars were not yet
grouted.
The tie bars, both longitudinal and transverse, located at the top of the capital did not
measure significant strain changes. Gauges S545, S549, S551 and S547 on the transverse bars
measured tension at 4με, while gauges S548, S552, S550 and S546 measured compression at -6με.
Strain changes at the bottom of the capital in the tie bars gauged by S509 to S515 measured
essentially no strain change. The circumferential bars gauged by S501 to S508 did measure strain
changes with a trend toward tension on the compression side of the pier, the upstation side, and

421
compression on the tensile side of the pier. These measured strains were as high as 17με at gauge
location S502. This compares to the calculated Poisson’s strain of 14με expansion at S502 that
would have occurred at the edge of the grout pad on top of the capital. A strut-and-tie model for
this pier capital should include the superstructure anchor segment, and the ultimate forces and
moments carried to the pier, because of the monolithic connection. This STM will be developed
in conjunction with the anchor segment in Section 6.3.5.

Table 6.5 Measured strain changes in segment PC16-8 gauges from placement of
superstructure segment P16-17

+ Tension + Tension
Strut Gauges Microstrain Tie Gauges Microstrain
C501 -38 S501 4
C503 -11 S502 17
C505 32 S503 11
C506 19 S504 -2
C508 -15 S505 -10
C515 -2 S506 -4
C523 4 S507 2
C524 -15 S509,10,11,12 -1
C527 -6 S513,14,15,16 0
C531 -2 S517,20,21,22 -17
C535 23 S518 -21
C536 15 S519 -19
C538 13 S523,24,27,28 3
C539 -17 S525 8
C541 -17 S526 10
C542 4 S529,30,31,34 17
S532 15
S533 27
S535,36,37,38 -8
S539 -10
S540 -10
S541,44 -4
S542,43 -8
S545,49,51,47 4
S548,52,50,46 -6

422
6.3.5 Ramp interior anchor segment
The ramp P anchor segment P16-1, actually cast in halves then post-tensioned together,
was designed as an anchorage zone for tendons T1, T2 and T4, as well as to carry moments and
shears to pier P16. Details for segment P16-1 are given in Figure 6.27. The segment was
prestressed in three dimensions, with top slab transverse prestressing, 16 vertical post-tensioned
threadbars anchored at deck level and within the pier capital, and by longitudinal post-tensioning
for the superstructure. Gauges were installed on the heavy tie reinforcement on the upstation face
of the anchor segment, as seen in Figure 6.27. Significant tensile stresses were not expected in the
anchor segment because of the prestressing, but cracks were seen on the upstation and downstation
faces of the segment, as seen in Figure 6.28. Cracks were located mostly in the upper part of the
diaphragm, as was the case for the mainlane anchor segment. The cracks in the ramp P anchor
segment did not propagate any significant distance into the access passage through the diaphragm,
with the longest crack being only 75mm. These cracks were presumably terminated when they
entered the vertical compression field from the 16 threadbars. The location of these vertical
threadbars is shown in Figure 6.27.

423
 CL Pier P16
787mm 660mm CL Tiedown
Threadbars

B A
Tendon T4
Tendon T3

Tendon T1
Tendon T2

Segment P16-1

Up Station
Segment P16-3 Segment P16-2
B A
ELEVATION
3960mm

29mm closed bars Tendon T4 Anchorage

Cantilever tendon ducts
22mm bars
S631

Transverse S628
S629 S630

S634
2130mm

prestressing Tendon T3
S627
S633
Tendon T1
S626
S632 Tendon T2
S625

749mm 749mm 19mm closed bar

CL Tiedown Threadbars (Typ.)
SECTION A-A
(Segment P16-1 Up Station Face)
Tendon T4

Tendon T3

Tendon T1
Anchorage
Tendon T2
Anchorage

SECTION B-B
(Segment P16-1 Down Station Face)

424
 Figure 6.27 Ramp P girder anchor segment P16-1 details

100mm
depth

75mm
depth
Post-tension Bar Anchorage
Protection Blocks (Typ.)

Segment P16-1 Up Station Face

75mm
depth
Tension 25mm
depth

Compression

Segment P16-1 Down Station Face

Figure 6.28 Crack patterns in segment P16-1

Strain changes measured in the anchor segment from the unbalanced placement of the
most distant superstructure segment P16-17 are given in Table 6.6. Strain changes in the vertical
bars were compressive, as shear forces were transferred to the bearing surface. The anchor
segment was made monolithic with the pier capital with a grout pad that extended nearly the full
width and length of the anchor segment bearing plinth. The flow of shear force from the webs to
the single bearing surface caused tension in gauge S625 of 11με. Compression would normally be
expected at this location, with tension in the top flange, as demonstrated by the simple STM at the

425
bottom of Figure 6.28. Compression may have existed across the bottom of the diaphragm at
locations more toward the centerline of the pier. The gauge S625 measurement may have been
influenced by Poisson's type strains from the vertical compression on the upstation face of the
anchor segment. The Poisson’s strain is calculated to be about 11μe (expansion) at a point
halfway between the bottom of the access opening and the bottom of the bearing plinth, using a
simple P/A+Mc/I calculation with a Poisson’s ratio of 0.2.
Measured compressive strains on the vertical bars tended to increase toward the bottom
of the diaphragm, with the highest compressive strain measured by gauge S632 at -44με. This
strain at S632 calculated using P/A+Mc/I, assuming the entire diaphragm was an effective part of
the cross section, would give a strain of -71με. The P/A+Mc/I calculated stress is inaccurate
because the state of stress is actually quite complicated within the diaphragm at the junction of the
superstructure and the pier capital. The strain measured at S632 was lower because S632 was
essentially outside the compression field between the girder web and the pier capital, thus plane
sections through the diaphragm were not plane after the loading.

Table 6.6 Measured strain changes in segment P16-1 gauges from placement of
superstructure segment P16-17

+ Tension
Gauge Microstrain
S625 11
S627 -2
S628 -13
S629 -23
S630 -17
S631 -15
S632 -44
S633 -23

After completion of the superstructure closure pours, continuity post-tensioning was

stressed, including external tendons T1, T2, T3 and T4. Tendons T1, T2 and T4 were anchored on
the segment P16-1 diaphragm, as shown in Figure 6.27, while tendon T3 passed continuously
through the anchor segment. The strain changes measured by the gauges in segment P16-1 from
the post-tensioning of these tendons is given in Table 6.7. Gauge S625 measured tension from the
stressing of the tendons anchored on the downstation face of the diaphragm, and measured
compression when T4 was stressed and anchored on the upstation face of the diaphragm. Greater
tensile stress was recorded by S625 when T2 was stressed than when T1 was stressed. T1 was

426
located closer to the web. The response of the vertical gauges was very similar from the stressing
of tendons T1 and T2, with somewhat higher strain changes caused by the stressing of T2.
Tendon T3 caused tension in most of the instrumented vertical bars, even though this tendon
passed continuously through the segment producing a downward deviation force.
The instrumented face of the diaphragm and the interior of the diaphragm may have had
different responses from the T3 deviation force, especially if the tendon did not contact its duct
near the face of the diaphragm. Tendon T4 was anchored very near the plane of strain gauges,
with compression indicated by the gauges closest to the anchor plate. Gauge S628 and S631
measured -11με and -29με respectively. All other gauges measured a small amount of tension
from the stressing of T4, with the exception of S625 on the bottom flange transverse bar. The
greatest tensile change from the stressing of all continuity tendons was measured by S630 at 71με.
With an assumed cracking strain of 170με, D/D0 at first cracking from the longitudinal external
post-tensioning alone would be high at 2.7.

Table 6.7 Measured strain changes in segment P16-1 gauges from superstructure post-
tensioning forces in microstrain

Strain in microstrain, + Tension

Tendons Stressed
Gauge T1 T2 T3 T4 All Tendons
S625 6 10 4 -6 14
S627 10 8 13 11 42
S628 23 19 -13 -11 18
S629 10 15 10 8 43
S630 23 25 13 10 71
S631 19 25 11 -29 26
S632 13 15 10 2 40
S633 13 11 2 11 37

An elevation of a strut-and-tie model for the ramp P girder under ultimate load in span
P16 is shown in Figure 6.29. AASHTO [23] factored dead load and live loads have been placed to
maximize the moment at the top of the capital. The superstructure moments, shears, and axial
forces from dead load, live load and prestress have been resolved into concentrated forces and
applied to the model. The discontinuous post-tensioning in the girder and pier from both multi-
strand tendons and post-tensioning bars makes this model quite complicated. Ties have been
placed at locations of primary reinforcement, including at three levels in the pier capital. The
vertical tie-down bars are the only reinforcement passing through the horizontal plane between the

427
bottom of the anchor segment plinth and the top of the capital. The selected struts and ties
crossing this plane reflect this limitation. The vertical tie down bars on the tension side of the
capital have reached ultimate tensile capacity at 7922kN. Another important tie location was
member 8 located within the anchor segment. High moment from span P16 must be distributed to
the pier, requiring a continuous flow of tension force from the top of the P16 girder near the pier,
through the diaphragm, and down the upstation side of pier P16. Other important tie locations
were at the top of the capital at members 21 and 22, between tiedown bar anchorages at members
28 and 36, and at the bottom of the capital. Based on the strain measurements, the circumferential
bars at the base of the capital were at important strut and tie locations, depending on the load
applied. This would be expected since the voided pier can only react on the capital along its top
perimeter. Also, to maintain the integrity of the voided pier section, all diffusion of forces from
the superstructure and post-tensioning tendons and bars should be designed to occur within the
limits of the capital without excessive strains or large cracks at the base of the capital. Additional
transverse reinforcement could have been added in the voided pier segments where drastic
changes in pier cross-sectional geometry occurred, such as at the top and bottom of the voided
section, to resist diffusion forces occurring outside the designed D-zones. No cracks were noted at
the top or bottom of the voided pier section from service level forces.
A strut-and-tie model taken through a section of the compression face of the anchor
segment and pier capital is shown in Figure 6.30. The eccentricity of the web shear forces to the
anchor segment plinth reaction causes a tensile force to occur in the top flange. This transverse
tensile force was partially resisted by the transverse deck post-tensioning. Other important tie
locations were at members 3 and 4, and between tiedown bar anchorages at member 11. The
three-directional prestressing of the anchor segment resulted in excellent service load level
performance. The factored loads applied to the STM in Figures 6.29 and 6.30 were AASHTO
HS25-44 factored live loads and factored dead load placed in the pattern needed to produce a
maximum ultimate moment on the span P16 face of the P16-1 anchor segment. These loads were
much lower than the loads required to cause a plastic moment in the superstructure, or even cause
top fiber tension in the anchor segment. The ultimate factored load was nearly enough to cause an
ultimate moment at the top of the capital, and was assumed to do so in the STM in Figure 6.29 to
create the largest possible forces in the D-zone. Based on the results of the STM in Figure 6.29
and 6.30, minimal plasticity would be expected in the anchor segment at factored ultimate load
because of the effectiveness of the three-dimensional prestressing. The mild steel in the anchor

428
segment diaphragm was many times more than necessary for taking the calculated tension tie
forces.

429
 A
5547kN + Tension
Cantilever tendons
3461kN ElementForce (kN)
30860kN 1 -3158
31340kN 2 -7170
4395kN 3 -6774
External tendon 4 2384
13920kN profiles 5 7922
1059kN 6 -32920
3656kN
7 169
7922kN
19110kN 8 3763
28890kN 9 -1041
6966kN 6966kN 10 -2967
Shear and 11 -2566
compression at joint 13 -4297
14 -1686
15 -4132
16 -10448
17 -27128
7922kN 18 -9875
19 -898
5547kN
20 -7757
21 1072
22 7726
23 -3269
24 -3874
25 -8798
26 -17103
734kN 27 -4897
28 6668
A 29 -2958
6009kN 12640kN 30 -8847
31 -4946
32 -5177
5547kN
33 2268
3461kN
34 -996
30860kN 2 5 6
31340kN 1 14 35 -12792
6 2 7 9
13920kN 1 4395kN 36 676
4 10
3 8 15
1059kN 3656kN 4 3 11 13
7922kN
19110kN 7 11 28890kN 16 17
5
5658kN
18 19
6966kN 21 22 20
8 10 1308kN
13
9 25 26 27
14 24 36
18
23 28
7922kN 29
30 31
5547kN
33
12 16

32 34
35
734kN 15 37
17
6009kN
12640kN
Nodes, Loads and Reactions Elements

430
 Figure 6.29 Strut-and-tie model at pier P16 - Side view

5547kN 5547kN
2357kN 2357kN
Shear at saddle Shear at saddle

1543kN Transverse
prestressing
External tendons 1543kN

Diaphragm longitudinal
prestressing
2357kN Web Shear 2357kN
Web Shear
3483kN 3483kN
+ Tension
Element Force (kN)
1 356
2 -6748
3 583
4 1521
5 -1361
6 -2064
7 -2220
5547kN 5547kN 8 -9339
9 -987
10 -3621
11 2455
12 -7655
13 -1148

5547kN Section A-A

2357kN
Shear at saddle Sym. 3
Sym.
1543kN Transverse 1
1 4
2 prestressing
3 4
2 5
3483kN
6
2357kN 7
5 6 Web Shear
9
7 8
8 10
11
9 10

5547kN 12
13
11 12

7569kN
Nodes, Loads and Reactions Elements

Figure 6.30 Strut-and-tie model at pier P16 - Looking down station

431
6.3.6 Ramp deviator segment
The deviator in segment P16-10 of ramp P, shown in Figure 6.31, was designed as an
inverted T-beam type deviator similar to those in the mainlane girders. The primary difference
between the ramp P deviator and the mainlane deviators was the width. The six 19-15mm
diameter strand tendons passing through the ramp P deviator occupied nearly the full width of the
deviator. The primary reinforcement in the ramp P deviator is shown in Figure 6.31. The
instrumentation in the ramp P deviator is also shown in Figure 6.31. The instrumentation
locations were similar to those in the mainlane deviator, but additional gauges were added to
measure strains at points away from the deviator pipes. Gauges S719, S720 and S721 were added
to measure tensile strains at the top center of the deviator beam, but unfortunately none of the
gauges survived the casting process of the segment.
Gauges S722 and S723 were installed on the bottom flange longitudinal reinforcement
near the bottom flange to web fillet juncture. Also, gauges S724 and S725 were installed on the
bottom flange longitudinal reinforcement at the centerline of the segment. Concrete strain gauges
C749 to C755 were installed 40mm from the interior surface of the web at the locations shown in
Figure 6.31. Tendons T1 to T3 produced both horizontal and vertical deviation forces that were
not symmetrical on both sides of the deviator because of the horizontal curvature of the bridge.
The deviation forces are given in Figure 6.32. The instrumentation was located on the left side of
the deviator looking upstation, and therefore was close to tendons T1 left, T2 left and T3 left.

432
 S713 S716
203mm
(S714) (S717)
((S715)) ((S718)) 2@229mm

12 - 16mm
3 - 29mm bars
bars A

S701
(S702)
((S703))
S704 Tendon T1R
(S705) S719 Tendon T2R
((S706)) S707 S710 A (S720) Tendon T3R
(S708) (S711)
((S721))
((S709)) ((S712))

SECTION AT P16-10
Looking Up Station

C751 C752 C753

C750 C754
Web guages

S716 S717 S718

S719 (S713) S720 (S714) S721 (S715)
C749 C755
914mm x 356mm
Deviator beam
29mm bar
Tendon duct 16mm closed bar

S724 S725
(S722) (S723)

S710 S711 S712

(S707) (S708) (S709)
((S704)) ((S705)) ((S706))
(((S701))) (((S702))) (((S703)))

SECTION A-A

Figure 6.31 Ramp P girder segment P16-10 deviator details

433
 Vertical Reactions from Post-tensioning Horizontal Reactions from Post-tensioning

Tendon T1 Left = 305kN Tendon T1 Left = 107kN

Tendon T1 Right = 469kN Tendon T1 Right = 165kN
Tendon T2 Left = 275kN Tendon T2 Left = 93kN
Tendon T2 Right = 426kN Tendon T2 Right = 151kN
Tendon T3 Left = 299kN Tendon T3 Left = 102kN
Tendon T3 Right = 465kN Tendon T3 Right = 165kN

Figure 6.32 Vertical and horizontal forces from tendons T1, T2 and T3
The tendons were not stressed in pairs, and because of the balanced cantilever
construction sequence they were not stressed on the same day or even during the same week. The
strains presented in Table 6.8 were the strain changes that occurred during the actual stressing
operation of the tendon indicated. The last column of Table 6.8 titled "Final" includes strain
changes measured over the entire construction period, and therefore include time dependent
strains. Gauges S701 to S703 were located on the closed bars intended to act as shear
reinforcement for the deviator beam. These gauges were located on the plane at the top surface of
the web fillet, as shown in Figure 6.31. From Table 6.8 it can be seen that the measured strain
changes in these bars were similar and very small, regardless of whether tendons on the right side
or left side were stressed, at about an average of 9με for tendons T1 and an average of 4με for
tendons T3. The total average instantaneous strain in gauges S701 to S703 was 38με from the
stressing of all tendons T1, T2 and T3. Longitudinally oriented gauges C749 and C755 measured
a total instantaneous strain change from the stressing of all tendons of -152με and -159με
respectively at an elevation in the girder slightly closer to the neutral axis of the section than that
of gauges S701 to S703. Gauges S722 and S723 measured instantaneous strains from all tendons

434
of -182με and -205με at an elevation lower than gauges S701 to S703. Using a linear
interpolation of these measured compressive strains, the longitudinal strain near S701 to S703
would be about -170με. Using a Poisson's ratio of 0.2 gives a vertical strain at gauges S701 to
S703 of about 34με. This is only slightly less than the measured total instantaneous strain of
38με, indicating that most of the measured strain change was not due to tensile forces from the
tendon deviations. Gauges S701, S702 and S703 may have been located too close to the girder
web to measure appreciable tensile strains from the shear force.

Table 6.8 Measured strain changes in segment P16-10 gauges from post-
tensioning forces in microstrain

Strain in microstrain, + Tension

Tendon stressed
Gauge T1L T1R T2L T2R T3L T3R Sum of Final
Instant.
S701 10 10 8 10 0 6 42 78
S702 8 8 6 6 4 8 38 80
S703 8 8 6 6 4 4 34 46
S704 0 -2 -4 4 0 0 -2 93
S705 11 2 11 2 2 4 32 163
S707 6 0 4 0 6 2 17 61
S708 10 0 10 8 2 0 29 84
S709 8 0 34 -6 0 -2 34 84
S710 -2 -4 0 8 40 23 65 239
S711 2 0 8 11 19 4 44 106
S713 8 6 19 10 -2 8 47 84
S714 8 4 23 -2 6 6 44 0
S716 8 8 19 21 53 8 116 13
S722 -32 -34 -32 -40 -11 -32 -182 -287
S723 -36 -32 -34 -40 -27 -36 -205 -300
S724 -32 -34 -32 -40 -10 -25 -173 -243
C749 -30 -23 -29 -30 -13 -27 -152 -172
C750 -27 -15 -27 -25 -19 -25 -137 -203
C751 0 4 0 -2 -6 -4 -8 4
C752 -4 2 -4 -2 -6 -6 -19 -13
C753 0 4 -2 -2 -2 -2 -4 -2
C754 -2 -13 -19 -17 -11 -17 -97 -101
C755 -30 -25 -27 -29 -19 -29 -159 -201

Gauges S704, S705 and S706 were located on the vertical reinforcing immediately
adjacent to the duct for tendon T1 left, as seen in Figure 6.31. From Table 6.8 it can be seen that
the strain changes from the stressing of T1 left were quite different at gauges S704 and S705, at

435
0με and 11με respectively. Gauge S706 was not working. This difference in strain indicates that
the tendon was not bearing evenly on the entire surface of the duct. Gauges S704 and S705 also
recorded different strains when T2 left was stressed, at -4με and 11με. The sum of the total
instantaneous strains for gauge S704 was essentially zero, while the sum of total instantaneous
strains for gauge S705 was 32με. The final long term strain at gauges S704 and S705 indicated
much more substantial tension, at 93με and 163με respectively. Continued cracking of the
deviator over time probably was the cause of the increase in tensile strain at these gauge locations.
The strain at S705 was close to the cracking strain of 170με. Gauges S707, S708 and S709,
located adjacent to the tendon duct for T2 left, also measured quite different strain changes from
each other as each tendon was stressed. This was especially apparent when tendon T2 left was
stressed. Gauges S707, S708 and S709 measured 4με, 10με and 34με respectively, indicating that
the tendon was bearing more on the upstation side of the deviation duct.
The long term final strains for gauges S707, S708 and S709 were somewhat more
uniform than the sum of instantaneous strain measurements, at 61με, 84με and 84με. The long
term strain measurements at gauges S710 and S711, located adjacent to the tendon duct for T3 left,
differed quite a bit at 239με and 106με respectively. Gauge S710 was located at a horizontal
crack, as seen in Figure 6.33. Gauge S716, located above tendon T3 left, was also located at a
crack and indicated a tensile strain of 53με when T3 left was stressed. The sum of the
instantaneous measured strain changes at gauge S716 was 116με, but had declined to only 13με
over the long term as seen in Table 6.8. The strain behavior at gauge S714 above tendon T2 left
was similar, with a sum of the instantaneous strains of 44με declining to zero strain over the long
term. Gauges S714 and S716 were located on the very short length top leg of the vertical stirrup
bars. The tensile stress in this short leg may have decreased because of concrete cracking or creep
near the top of these bars. This stress would have been taken by the 29mm full width top bars,
shown in Figure 6.31. Unfortunately gauges S719 to S721 were damaged and could not measure
the strain in these top bars.
As seen in the mainlane deviator, the largest strain readings were taken on the vertical
bars adjacent to the tendons closest to the centerline of the girder, even though the stressing order
was reversed. Cracking was much less substantial in the Ramp P deviator, initiating during the
stressing of the T3 tendons. This relates to a D/D0 at cracking of about 0.88. Deviator reinforcing
bars were stressed to 12% of the yield stress or less. An estimated D/D0 at first yield would be

436
unrealistically large at 7.6. In actuality yielding would occur earlier than this, since the concrete in
tension would rupture placing more force on the steel reinforcement.

A A

VIEW LOOKING UPSTATION

VIEW A-A

VIEW LOOKING DOWNSTATION

Figure 6.33 Crack patterns in segment P16-10

437
 Gauges S722, S723 and S724 gave similar strain changes from the stressing of each of
the tendons, except when T3L was stressed. The strain changes at gauge S723 were higher than
the strain changes at S722 and S724 during the stressing of tendons T3 left and T3 right. This is
opposite of what would be expected because the horizontal friction force acted toward gauges
S722 and S724. A greater number of gauges at this location may have shown otherwise, but
gauges C749 and C755 gave a similar result. The total instantaneous strain change was -152με at
C749 and -159με at C755, very close to the cracking strain. The strain measurements at
diagonally placed gauges C750 and C754 included the local compressive strain from the tendon
deviation forces, the Poisson's strain from longitudinal prestressing forces, and the shear strain
from the reaction at the deviator. Gauge C750 measured a higher total instantaneous strain change
than gauge C754, at -137με and -97με respectively, because of the difference in shear strain on
each side of the deviator and the orientation of the gauges. Gauges C751, C752 and C753
measured very little compressive strain change, but may have been located near the neutral
bending and compression axis of the web.
Development of a transverse strut-and-tie model for service or ultimate strand forces for
the ramp P deviator used the same method as in the mainlane deviator. The service load forces on
the tendon ducts within the deviator are given in Figure 6.32. A strut-and-tie model for the
deviator is shown in Figure 6.34. The model assumes that significant cracking between deviator
pipes would occur at ultimate load. Therefore vertical forces from the tendons are assumed to be
transferred to the concrete below the ducts, and then carried in shear to the web. The location of
the tension ties was chosen to directly allow selection of reinforcement for the region of the tie.
The service level live load forces on the deviator had little influence on the reinforcement in the
deviator. The strain changes on the bars in the deviator for the maximum positive live load
moment in span P16, given in Table 6.9, were less than 1% of εy showing negligible live load
changes.
For the ultimate load case, tendon force changes and deviation angle changes might result
in 35% or larger vertical force changes on the deviator, and substantial horizontal force changes on
the deviator when compared to the service load case. Assumptions made about the slippage or
lack of slippage of the tendons within the deviator will have an influence on the horizontal force
changes. The service level performance of the P16-10 deviator was far superior to that of the
mainlane deviator, with no extensive cracking seen or anticipated based on the measurements.
Measured strains were only marginally higher than the strains necessary to crack the concrete at a

438
few locations. The distribution of force from the tendons to the deviator was not uniform along
the length of the ducts.
An approximate calculation of tie forces based on the measured strains gave the forces
shown in the top section of Figure 6.34. The STM assumed that a horizontal cracked plane existed
between deviator pipes, as might be the case at ultimate load. Judging by the measured strains,
this was not the case at the service load level except near the T3 deviator pipe. The calculated tie
force based on the measured strains from gauges S710 and S711 was 343kN. The STM predicted
238kN at this location. The estimated tie force at S704 and S705 based on the measured strains
was also higher than the force predicted by the STM at 254kN versus 161kN. The estimated tie
force, based on the measured strains at S707, S708 and S709, was lower than predicted by the
STM at 152kN versus 243kN. As was the case for the mainlane deviator, the forces in these ties
in the top portion of the deviator were probably different from the STM prediction because the
actual flow of forces in the deviator differed somewhat from the STM. The forces would differ
because of cracking at some location and not others, and also because of the crude method used to
calculate forces from the measured strains. Vertical ties designed by using the STM tie forces
from the lower half of the deviator would be conservative for a fully cracked deviator at ultimate
load levels. This deviator must also be designed for longitudinal forces from the tendons at
ultimate load.

439
 331

126
kN

2kN
3kN

304kN

291kN
269kN
1272kN

24kN 56kN
68kN 70kN
254kN 1769kN
152kN
343kN

HALF SECTION THROUGH CENTERLINE OF DEVIATOR

-124
331

Legend
0kN
kN

Tension Tie
285

Compression Strut
-11
kN

62k

Gauge Location
N

350kN 1421kN 1512kN 1478kN 1398kN 1272kN

N
190

-99
994k
k

243kN
161kN

238kN
-1
-143

221kN
kN

8k
18
-105k

-167
N
1k

1kN
N

-1548kN -1078kN -884kN

N
N

5k

17
48
10

5 kN
-6
-9

-2
5k

-396kN

57

75
30

703kN

1769kN
94
460kN

kN
N

kN
kN

kN

388kN -221kN -691kN

CALCULATED MAJOR STRUT AND TIE FORCES

Figure 6.34 Strut-and-tie model for the segment P16-10 deviator

440
 Table 6.9 Measured strain changes in segment P16-10 gauges from live load case 2

Gauge Microstrain, + Tension

S701 -2
S702 -2
S703 -2
S704 0
S705 0
S707 0
S708 0
S709 0
S710 -2
S711 -2
S713 0
S714 0
S716 -2
S722 13
S723 17
S724 15
C749 8
C750 15
C751 0
C752 0
C753 -2
C754 -4
C755 8

6.3.7 Ramp anchorage blister segment

The anchorage blister in segment P16-4 was instrumented with both strain gauges
installed directly on the mild reinforcing bars, and with concrete strain gauges. The location of
these gauges and details of the blister are given in Figure 6.35. Tendon T22, anchored in the
blister, was a bottom slab continuity tendon located within the length of span P16, passing through
the cast-in-place closure near the center of the span. Two 9 - 15mm diameter strand tendons T22
were anchored in blisters in segment P16-4, one located on each side of the girder. The
measurements presented in this section were taken only during the stressing of tendon T22 left.
The main reinforcement in the blister consisted of 16mm closed bars, located around the tendon
duct and extending into the bottom flange of the girder, and 90o bend 16mm bars anchored in the
web and bottom flange of the girder. This reinforcing, shown in Figure 6.35, was distributed over
the full length of the blister, but only the bars near the anchor plate were instrumented. Additional

441
confinement reinforcing was placed within the larger 16mm closed bars shown in Figure 6.35.
Longitudinal 13mm bars located in the bottom flange and web were instrumented with gauges
S618 to S622. Concrete strain gauges C649 to C655 were placed in the bottom flange and web,
but were oriented both longitudinally and transversely. Tendon T22 was only deviated in the
vertical plane within segment P16-4, so the tendon applied a distributed vertical force on the top of
its duct as well as a compressive force of 1800kN on the anchor plate.

S618
16mm bar (Typ.)
C655 13mm (Typ.)
S619
C654 S609
S610
S611
S620
S621
S612
S622 S613 S614
S623

C653
C652

C650
C649
C651 16mm bar

S624
S615
S616 S617

9 - 15mm dia. strand tendon

356mm high x 2046mm long
(1800kN jacking force)
anchorage blister

Figure 6.35 Ramp P girder segment P16-4 anchorage blister details

During the cantilever erection process this segment experiences substantial negative
moment so that the lower flange has high compressive stresses. Measured strains in the blister
from the stressing of tendon T2 left are given in Table 6.10. No cracks were seen on the blister,
and all values of measured strain were well below the cracking strain for the concrete of 170με.
Thus, the concrete can be assumed uncracked so that the measured strains in the reinforcing bars
can be assumed to exist in the concrete surrounding the bars. This greatly increased the area of
material available to take tensile stresses over that of a cracked blister, and reduced the magnitude
of the measurements from those for cracked concrete. Based on the maximum measured tensile
strain of 40με in the blister, D/D0 at first cracking would be large at about 4.3. In plan view, the

442
blister can be visualized as a corbel extending from the girder web. The measurements from
gauges S609, S610 and S611substantiate the corbel action, with strain changes of 11με, -2με and -
4με respectively. The tensile strain at gauge S609 quickly changes to a compressive strain at
gauges S610 and S611. The distance between gauges S609 and S611 was only 400mm, and the
distance from S609 to the edge of the blister was 75mm.
Viewing the blister in elevation, corbel action may also be anticipated, but the
measurements by gauges S624, S615, S616 and S617 did not indicate corbel type behavior. The
measured strain changes at gauges S624 and S615 averaged -37με, the strain at S616 was -30με,
and at S617 was -6με. These measurements indicated a trend of compression changing to tension
along the length of the blister from the applied loads. In elevation, the tendon anchorage applied
force to the blister at an angle of 14o with respect to the horizontal. The horizontal component of
the anchorage force was, therefore, 1750kN, and the vertical component was 435kN downward.
Also, the deviation of the tendon within the blister along the 6m radius duct provided an upward
distributed force totaling 427kN, and a horizontal frictional force totaling 107kN.

Table 6.10 Measured strain changes in segment P16-4 anchorage blister gauges from post-
tensioning of tendon T22

Gauge Microstrain, + Tension

S609 11
S610 -2
S611 -4
S612 40
S614 25
S615 -44
S616 -30
S617 -6
S618 8
S619 21
S620 38
S621 38
S622 40
S623 0
S624 -30
C649 -4
C650 -2
C651 -6
C652 34
C653 38
C655 13

443
 Gauges S612 and S614, located above the anchorage plate, measured strain changes of
40με and 25με respectively, indicating that transverse tension was present in the immediate
vicinity of the concentrated load from the anchor plate. This should be considered a local zone
force from the anchorage. Gauge S623, located on a closed 16mm bar near the face of the blister
and the bottom flange to web fillet, indicated no strain change. Gauges S618 to S622 measured
tensile strains behind the blister in the bottom flange and web.
Wollman [54] recommended that 25% of the unfactored stressing force on an
intermediate anchorage be tied back by mild reinforcing steel stressed no greater than 60% of its
yield strength. Gauges S620 to S622 were located most directly behind the anchorage, and
measured nearly equal strain changes at 39με. This tensile strain was insufficient to crack the
concrete, and produced negligible tensile stress changes in the longitudinal bottom flange mild
reinforcement. This tension decreased quickly further up the web, with gauge S619 measuring a
strain of 21με, and gauge S618 measuring a strain of only 8με. The concrete strain gauges C652,
C653 and C655 measured nearly the same strain changes as the S-gauges located behind the
blister at 34με, 38με and 13με respectively. The longitudinal tensile stresses behind the
anchorage would be balanced by compression stresses at other points in the cross section.
Transversely oriented concrete strain gauges C649, C650 and C651 indicated a small amount of
compressive strain, averaging -4με. A strut-and-tie model for the anchorage blister must include
ties to deviate the compression force from the anchor plate and curved duct to the bottom slab and
web, as well as to control splitting forces from the concentrated load within the blister. Tensile
forces behind the blister will also result in diffusion of some of the compressive force from the
tendon. Calculation of general zone stresses in front of and behind the blister from the stressing of
the tendon anchored in the blister will help define the flow of compressive forces.
The response of the girder from force changes in tendon T22 is different for the service
load case and the ultimate load case. For the service load case, the tensioning of partial span
length tendon T22 produces compression and negative bending in the central portion of span P16.
The side spans P15 and P17 and the portion of the girder over the piers are put in positive bending
from the secondary or boundary condition effects of stressing T22. This changes the moment
diagram from that of simple post-tensioned cantilevers to that more closely resembling a post-
tensioned continuous girder. The shear diagram changes little in span P16 from the stressing of
T22. Therefore, the diffusion of the compression force in front of the T22 anchorage, and the

444
tension force immediately behind the blister are affected only by forces in the local zone. The
compression in the bottom flange continues to be large on average at the blister location after T22
is stressed, but decreases when the external tendons are stressed. Tension may exist in the
concrete behind a blister at the service load level, so it is not advisable to locate the anchorage end
of a blister on the same transverse plane as a segment joint. A strut-and-tie model for the
anchorage blister is shown in Figure 6.36 at the service load level.

445
 1800kN
Jacking Force 435kN
1750kN

427kN

107kN
Friction

15kN
10 3 1k N
10 Forces on Blister
20
30
3 40 15kN
9 50
13 19
3 9 5k N 23 29 60 64 6 k N
33 39
8 10 49
2 18 53 59
2 80 k N 7 12 28 63
6 17 27 38
35 6 k N 1 16 22
5 11 48
4 26 58
15 21 4 39 k N
41kN 14
25 46 47 52 57
24 56 62
34kN 2 60 k N
35kN 34 51 55 61
44
Blister Elements 75kN 3 0 6k N
69kN 54
Element End Nodes 45kN
64 16 and 13 16kN
65 16 and 14 Element Locations for STM of Blister D-Zone
66 16 and 15 Some Elements Omitted for Clarity
67 16 and 18
68 16 and 19
69 16 and 20 15
70 16 and 21 20
71 16 and 22
72 21 and 15 26
16
73 21 and 20 21
74 21 and 13
14 19
75 21 and 18
22
76 21 and 26
77 21 and 25 13 25
78 21 and 24
79 22 and 21 18
80 22 and 20 24
81 22 and 19
Node Locations for STM of Blister
82 22 and 18
Some Elements Omitted for Clarity
83 24 and 26

Figure 6.36 Strut-and-Tie model for segment P16-4 anchorage blister for
service load forces from Tendon T22

446
 The STM was only subjected to the force from the tendon anchored in the blister, and the
resultant reactions. Cantilevering dead load and prestressing stresses were present, but not
included in the model. The cantilevering stresses were easily calculated within the D-zone near
the blister, and should be added to the stresses found from the STM. The STM in Figure 6.36 was
designed to predict tensile forces in the blister at points away from the local zone, as well as to
predict the tension in the bottom flange behind the anchorage. Boundary conditions were
carefully applied so that the bottom flange and web could be modeled as plates, since the flow of
axial forces was of primary interest. The limits for the D-zone were chosen using the 30o diffusion
method. Full diffusion was assumed to occur when the full height of the girder web was included
in the 60o diffusion cone. Moments, axial forces and shears were calculated at the limits of the D-
zone, and resolved into and applied as concentrated forces on the STM. Positive bending of the
girder created tension in the bottom slab between the pier and the blister.
Using a plane sections remain plane type analysis, including no local effects from the
anchorage, a tensile stress of 1.83MPa was calculated to exist at gauge location S622. Using the
measured strain at S622 of 40με, the actual stress was calculated to be about 1.60MPa in tension,
or within 12% of that predicted by the simple analysis. Tie forces in members 24, 26 and 28
calculated using the STM were 396kN, 198kN and 144kN, as shown in Table 6.11. This predicted
the trend measured by gauges S622, S621, S620 and S619, with decreasing strains of 40με, 38με,
21με and 8με. The stress at gauge S622 predicted using the STM would be about 1.92MPa, which
is 20% larger than the measured strain converted to stress of 1.60MPa.
Comparison of the measured and calculated tension behind the anchor plate indicated that
the bottom flange longitudinal stress distribution was not substantially influenced by the presence
of the point load. The recommended 25% of the point load, at 99kN, would not need to have been
added to the bottom flange force calculated using the plane sections analysis in order to get
sufficiently accurate bottom flange stresses. The positive bending moment, caused by continuity
of the span, essentially produced the proper tensile stress distribution behind the blister.

447
 Table 6.11 Element forces from P16-4 blister STM
Tension (+)
Element Force in kN Element Force in kN Element Force in kN
1 42 29 -44 57 -19
2 98 30 -783 58 -431
3 107 31 52 59 -10
4 356 32 77 60 -646
5 18 33 140 61 1
6 264 34 304 62 11
7 -3 35 91 63 18
8 397 36 125 64 -287
9 -182 37 38 65 -78
10 -898 38 32 66 -26
11 -42 39 -190 67 -535
12 -62 40 -653 68 -542
13 79 41 -109 69 -332
14 372 42 -46 70 -391
15 26 43 -1 71 -108
16 238 44 -105 72 65
17 35 45 -82 73 152
18 236 46 -282 74 177
19 -115 47 71 75 158
20 -814 48 -420 76 -36
21 -46 49 1 77 -70
22 -51 50 -654 78 -94
23 30 51 58 79 -66
24 396 52 -9 80 -81
25 75 53 6 81 290
26 198 54 -262 82 287
27 10 55 -47 83 45
28 144 56 -244

Blister member 64 was in compression, at 287kN, confirming the measured compressive

strain at gauges S615 and S624 that averaged 37με compression. Reactions at nodes 9 and 13
indicated slight compression, averaging about 0.15MPa or 4με compression. This compared well
to measured strains at gauges C649, C650 and C651 that averaged 4με compression. Slight
compression was calculated using the STM in member 66, which indicated that significant corbel
action was not occurring. The small measured strains at gauges S609, S610 and S611 would tend
to verify this calculation. Significant tension ties within the blister were the members that tied
node 21 to the bottom flange and web, namely members 72, 73, 74 and 75. The greatest tensile
forces were associated with the upward thrust of the tendon on the curved section of duct inside

448
the blister. Since all this thrust was concentrated at node 22 in the STM, all the tension force was
taken by members 81 and 82. Steel designed to take this tensile force would need to distributed
along the length of the duct curvature, as was done in the actual design of this blister.
Forces within the blister changed little with the application of live loads since the blister
was located near the point of inflection. At the ultimate load level, large shear force changes in
span P16 and plastic behavior of the girder will deviate the compressive force in front of the T22
anchorage, as well as development forces from the tendon, to the top flange of the girder as shown
in the strut-and-tie model in Figure 6.37. Forces in the blister change because of longitudinal
stress changes in the bottom flange and webs, although transverse and vertical forces within the
blister itself change little since the ultimate force in the tendon within the blister can only be
moderately larger than the force after jacking. Since the service load level performance of the T22
blister was excellent, the ultimate load level performance can also be expected to be good since
the forces within the blister will change little.

Legend
Tie
Strut

To Pier To Midspan

Figure 6.37 Struts and ties near anchorage blister at ultimate

449
6.4 RECOMMENDATIONS AND CONCLUSIONS
The service level performance of the D-zones under study ranged from inadequate when
unanticipated details like sharply bent deviator pipes were used, to excellent when designs were
very conservative. The following recommendations and conclusions were drawn from
observations of the measured data, visual inspections, and review of the designs and details.

6.4.1 Pier capitals

The mainlane Y-pier was designed as a concrete-steel composite structure. The force in
the transverse steel pipe ties was reasonably predicted using either a frame analysis, at 31% of the
vertical applied vertical force, or a simple strut-and-tie model, at 36% of the applied vertical force.
As would be expected the STM was more conservative compared to the measured pipe force of
34% of the applied vertical force. The anchorage of the pipe ties in the concrete was excellent,
with a fully bonded condition existing at the time of testing under full superstructure dead load.
The stiffness of the pipe ties limited the bending moment in the concrete Y compression struts
such that no cracking occurred at the service load level. Given the low stress levels in the pipes
and in the concrete, the pier should behave very well with an ultimate load placed on the
superstructure.
The pier P16 capital segment also essentially remained uncracked from service load level
forces, although the capital was initially cracked from thermal gradient forces during curing.
These cracks had little impact on the service level performance of the capital, based on visual
inspection. Most measured tensile strain changes were small during the monitored live load test
cases, at less than 5% of the yield strain for the instrumented reinforcement. Heavy bars placed
longitudinally and transversely at the top and bottom of the pier capital carried tensile force as
expected in most cases, but measured strain changes in the circumferential gauged bars at the
bottom of the pier were equal to or larger than the magnitude of the strain changes in the designed
heavy tie bars. At the ultimate load condition a strut-and-tie model for this solid section should
include circumferential ties at the bottom to maintain similar geometry to the voided section
below.

6.4.2 Deviators
The instrumented mainlane and ramp deviators were beam type deviators similar in
dimension and reinforcement, with the exception of the width. The vertical tendon loads on the
450
ramp deviator were only about 85% of those on the mainlane deviator. The deviation force on the
tendon ducts did not appear to be well distributed along the length of the ducts in the ramp
deviator. Bar strains near the center and end of the deviator pipes varied by as much as 200%.
However, no large cracks developed and no exceptionally large strains were measured in the bars.
The ramp deviator had good service level performance, with small crack widths and maximum
reinforcement tensile strains of only 10% the yield strain. The mainlane deviator ducts did not
follow the design drawing and were sharply bent, concentrating the entire deviation force near the
center of the deviator. This concentration of force caused extensive open cracks, and high stresses
(50% of fy) in the reinforcing bars adjacent to the sharp bend in the duct.
Deviation ducts should be smoothly radiused over the full length of the deviator to
improve service level performance. A sharp bend may also increase friction between tendon,
grout and duct, increasing horizontal loads on the deviator at ultimate. The reinforcement details
for both deviators were easily constructed at points away from the girder webs. Near the web to
deviator intersection, the heavy web stirrup bars and other bars transitioning the web to the bottom
flange remained unchanged from that for a typical section. This left little room for the deviator
top bars or shear reinforcement.
Priority should be given to the placement and development of the bars connecting the
deviator to the webs, since this is the critical force path for developing the ultimate moment
capacity of the girder. Figure 6.38 is a photograph of the actual reinforcement connecting the
deviator to the web of the ramp P girder. This demonstrates that the heavy concentration of web
to bottom slab fillet typical reinforcement, as well as the main web reinforcement, makes it
difficult to include properly located and anchored inclined reinforcement. The excellent
performance (εsmax<10%εy) dictates that some reinforcement might be removed. Prime candidates
for removal would be those bars placed parallel to the compression field. Further model tests
could confirm the result of reducing such reinforcement.

451
 Figure 6.38 Ramp P girder segment P16-10 deviator to web detail

6.4.3 Pier segments

The mainlane pier segment, with heavy anchorage diaphragm, was not post-tensioned
vertically to control tensile stresses from the diffusion of post-tensioning forces. Regardless of
this, the mainlane pier segment performed well at service load levels. The highest measured
reinforcing bar stress was 35% of the yield stress, and occurred in the heavy vertical bar
immediately adjacent to the access passage through the diaphragm at a crack location. The heavy
end diaphragm was designed to take all the transverse and vertical tensile stresses from the
anchorage zone at the ultimate load condition. The transverse stiffness of the anchorage
diaphragm was insufficient to attract all the transverse tension in the bottom flange of the girder at
the service level load condition, since the post-tensioning forces were poorly diffused over the
short length of the anchor segment.
The bottom flange cracked longitudinally directly beneath the bottom flange internal
tendon ducts. The designed D-zone should have included a length of typical section beyond the
anchorage diaphragm, since the bottom flange was highly stressed over this length by the bottom
flange internal post-tensioning tendons. The area of transverse bottom flange steel could then
have been calculated and included in the design to control cracking. The pier segment

452
instrumented on ramp P was post-tensioned vertically to make a moment connection with the pier.
As a result little cracking occurred in the heavy anchorage diaphragm, and strain measurements
were small.

6.4.4 Anchorage blister

The instrumented anchorage blister was uncracked following post-tensioning and
thereafter. A small amount of induced tension was measured behind the anchorage, but since the
bottom flange of the girder was in significant compression from cantilevering dead load, no cracks
were found. Very small tension was also measured in the bars that connected the blister to the
web, indicating some corbel action. No corbel action was indicated between the blister and the
bottom slab. Instead, compression was measured in the vertical bars adjacent to the anchor plate
where the blister joined the bottom slab. Tension was measured in the vertical bars further along
the blister because of the vertical force generated by the radial deviation of the tendon over the
length of the blister. The design of this blister was very conservative. Tension forces behind the
blister were predicted conservatively using a STM.

453
 CHAPTER 7
BEHAVIOR OF A SEMI-CONTINUOUS UNIT

7.1 INTRODUCTION
The mainlane and ramp girders on U.S. 183 constructed by the span-by-span method
were designed as three span semi-continuous units. The box girders were constructed as isolated
simple spans, and then lowered off the erection trusses onto four independent elastomeric
bearings. At a later stage in construction the individual simple spans were linked into semi-
continuous units by cast-in-place deck slab closures. Construction in simple spans allowed the
erection sequence to advance more quickly than for continuous girders with multiple span tendons
and possible closure pours. TxDOT engineers decided to design the structure in simple spans
based on the construction experiences on the San Antonio Y.

The San Antonio Y had multiple span fully continuous units requiring complex tendon
layouts, involved stressing sequences, and difficult to build pier segments with deviated ducts and
varying anchorage locations. Even though the final girder configuration was continuous, the
maximum moments in the girder were the positive moments from self weight during construction
either as a simple span for the first unit erected after an expansion joint, or as a span with one end
continuous with the preceding span for subsequent units erected. Economy dictates that span-by-
span erected segmental bridges be constructed using a single span erection truss. Thus, a
continuous structure erected span-by-span does not have the moment diagram for dead load of a
continuous girder constructed and tensioned entirely on falsework.

In addition, continuous spans must be designed for thermal gradient induced moments
over interior piers that often are opposite in sign to the live load moments. Post-tensioning
provided to prevent tension from thermal gradients would act opposite to that provided for live
load. This results in a substantial amount of prestressing in continuous girders across the pier,
with little eccentricity to the centroid of the girder cross-section. In order to avoid such
contradictory conditions, the U.S. 183 girders, except for the five span continuous girder built by
the balanced cantilever method on Ramp P, were all designed and constructed as simple spans.
The external tendon profiles had no curvatures at the anchor segments (Figure 7.1), and the
internal tendons only had substantial curvature near the dead end anchorage.

453
 See Detail A

3 and 7 -15mm dia strand tendons

19 - 15mm dia. strand tendons

Elevation - Girder D5

Rams stressing external tendons

Notch for joint

Detail A
Figure 7.1 Post-tensioning tendons on the mainlane girder

454
 Since providing an expansion joint at every pier would have proven to be costly, would
have increased maintenance, and potentially would have given a poor ride quality, most spans
were connected longitudinally at the deck into three span units. Longitudinal expansions and
contractions from temperature changes required that deck finger joints be provided at least at
every third span since the girders were supported on elastomeric bearings. TxDOT has been
designing deck slabs to be cast continuously over simple span I-girders for many years. The
durability and ride quality of these slabs has proven to be excellent when properly detailed. Since
the deck of the box girder was precast monolithic with the girder webs and bottom flange, the
closure at the deck between the ends of each simple span was cast-in-place on U.S. 183. The
original details in the contract plans provided for a concrete drop-in panel that would later be
connected to the box girders with a closure pour (Figure 7.2).

Precast drop-in panel

Cast-in-place
concrete
(each side) Epoxy coated reinforcing

Bituminous fiberboard
pad (Typ. each side)

Anchor
Elastomeric segment (Typ.)
bearing (Typ.)

Cast-in-place fixity block

with encased pipe at fixed
bearing locations only Pier capital

Longitudinal Section
Figure 7.2 Original joint details from contract plans for U. S. 183
455
 The contractor decided to provide temporary construction access along the length of the
bridge deck by spanning the gaps between spans with heavy timbers covered by steel plates
instead of using the concrete drop-in panel. The joints were then fully cast-in-place three at a time
well after construction of the main spans. The construction sequence of the joints is shown in
Figure 7.3. All reinforcing was epoxy coated, including the 90o splice bars in the anchor segment.

Epoxy coated splice bars

a). Notch provided for deck joint

b). Epoxy coated reinforcing steel in place

c). Deck joint concreting

Figure 7.3 Actual cast-in-place joint construction
456
 The primary design case for the joint slab was to withstand wheel loads placed directly
on the joint. A secondary design consideration was to maintain the bridge alignment, since guided
bearings and fixity blocks were provided at the minimum required number of locations. One
guided bearing was provided at each pier, and one fixity block was provided for each bridge unit.
Because of this secondary design consideration, the joint to anchor segment connection would
have to be capable of carrying tension and compression caused by thermal strain changes in the
girders. Deterioration of the deck-to-girder connection would be detrimental to the performance
of the bridge unit, and would require the replacement of the joint.

The ride quality of these cast-in-place deck joints, topped with 5mm of asphalt placed
continuously over the entire unit, is exceptional. The only indication to the motorist that the joint
exists occurs on those few spans where the camber from prestress and dead load differed from the
design vertical alignment because of creep or casting geometry errors. Even the worst spans on
the project have very good ride quality.

7.2 PROBLEM AND TECHNICAL BACKGROUND

Once the closures had been cast, the length of joint slab had the potential to carry tension,
compression, and bending forces. The overlap of hooked reinforcement from the anchor segment
and the joint slab could possibly carry moment caused by girder end rotations, although the girder
end rotations from live load were extremely small. TxDOT design engineers were interested in
the magnitude and effects of negative moment developed in the box girders, not the joint itself,
over the piers from live loads. Potentially, a live load induced moment couple could develop
between the joint slab in tension and the restraint from the bearings and pier capital as a
compression force. TxDOT engineers did not intend or desire this moment couple as a
performance characteristic. The closure slab was designed as a bending member for wheel loads
and girder end rotations, seen in Figure 7.4, with two layers of reinforcement. The slab bars were
spliced to 90o bars in the notch of the girder, providing a moment connection even if the closure
concrete did not totally bond to the girder concrete. Some of the 90o bars in the anchor segment
notch, shown in Detail A of Figure 7.1, were of insufficient length, and were supplemented with
additional straight bars drilled and epoxied into the anchor segment. TxDOT designers controlled
the effective longitudinal length of the slab between girder wings with bituminous fiberboard
pads, shown in the bottom section of Figure 7.4. The actual 150mm gap between the wings was
457
effectively lengthened to 380mm by the bituminous pads. The angle change at the end of one box
girder from positive girder bending, as well as differential wing tip deflection from live load,
otherwise would have induced large bending moments in the closure slab with high shear forces
because of the short slab span lengths. The fiberboard pads reduced the effective bending stiffness
of the slab and therefore reduced the shear forces for a given angle change at the end of the box
girder. An additional reduction in bending stiffness was caused by shrinkage cracks at the cold
joint.

Epoxy Coated Reinforcement (Typ.) Wheel load

on joint
Cast-in-place Concrete

Anchor
Segment
760mm (Typ.)
End rotation
13mm x 50mm Preformed
Bituminous Fiber Material
(each side)

Longitudinal Section
Through Centerline of Superstructure

Epoxy Coated Reinforcement (Typ.) Wheel load

Cast-in-place Concrete on wing

13mm x 380mm
150mm Preformed Bituminous
Fiber Material
Anchor Segment Wing (Typ.)
End rotation
Longitudinal Section
Through Superstructure Wing

Figure 7.4 Longitudinal sections–cast-in-place joint

458
 Another source of force in the slab comes from thermal loading, both from uniform
temperature changes, and from thermal gradients in the box girders. A uniform temperature
change causes the length of the box girders to change, with the finger joints accommodating this
deflection. The three span semi-continuous unit has one longitudinally fixed set of elastomeric
bearings, fixity being provided by a fixity block, at an interior pier and three other sets of
longitudinal expansion elastomeric bearings at the remaining piers of the unit. Expansion of the
central span had to be carried across one end span and two sets of elastomeric bearings. The shear
force required to longitudinally deflect these bearings is taken by the closure slab, and changes on
both a daily basis and on a seasonal basis. To provide some residual compression in the slab
concrete, the adjacent spans were jacked apart slightly using threadbars and held in place until the
joint concrete had cured. Regardless of this procedure, shrinkage cracks formed at the cold joints.
The overlap of the 180o joint bars and the 90o anchor segment bars, shown in Figure 7.4, must be
strong enough to carry any tensile forces developed across the joint. An initial assessment of the
strength of this detail, based on inspection of Figure 7.4, is that the longitudinal tensile capacity
would be minimal.

Thermal gradients in the box girder also produce forces in the continuity slab. The
thermal gradient causes a rotation at the girder end, producing resultant bending and axial forces in
the slab. Also, since the box girders under study have some shear lag in the top flange in response
to thermal gradients, the continuity slab acts to stiffen the top flange and change the behavior of
the box girders to some extent. Stress changes in the slab give an indication of the validity of
assumptions made during the thermal gradient analysis of the girder, such as plane sections
remaining plane.

7.3 PREVIOUS STUDIES

Extensive research was conducted and reported by Roberts, et al. [7] on a cast-in-place
continuity joint similar to that in use on the U.S. 183 segmental box girders. Details of the joint
studied by Roberts, et al. [7] on the San Antonio Y are shown in Figure 7.5. One difference
between the dimensions of this joint and the joints used on U.S. 183 was the distance between
anchor segments. The length of the slab between girder webs was 970mm (seen in the top section
of Figure 7.5), compared to 760mm on U.S. 183 (seen in the top section of Figure 7.4). The gap
between wings of the San Antonio Y girders was only 50mm, effectively lengthened to 250mm by
a bituminous pad, as seen in the bottom section of Figure 7.5. This was somewhat shorter than the

459
380mm bituminous pad length on U.S. 183 (see the bottom section of Figure 7.4). Another
difference was that the reinforcement detail connecting the joint to the anchor segments was far
more substantial on the San Antonio Y, and would be expected to provide greater moment fixity
and tensile capacity between anchor segment and joint slab than the detail used at U. S. 183.

The most important difference between the joint details of the San Antonio Y and U.S.
183 was the type of bearing used on each project. The San Antonio Y girders were supported on
50mm thick fabric reinforced pads, with a teflon sliding surface at expansion joints. In order for
the bearing to allow longitudinal movement from girder expansion, contraction and end rotation,
the friction between the teflon pad and the steel plate attached to the fabric reinforced pad must be
overcome. The bearings on U.S. 183 were 70mm thick steel shim reinforced 50 durometer
elastomeric pads at the locations of the continuity joints. Longitudinal movement of the girders
was accommodated by horizontal shear deformation of the elastomeric pads.

Anchor Segment (Typ.)

Cast-in-place Concrete
Reinforcing Bars (Typ.)

970mm

Longitudinal Section
Through Centerline of Superstructure

Superstructure Wing (Typ.)

Cast-in-place Concrete
Reinforcing Bars (Typ.)

50mm 13mm x 250mm Preformed

Bituminous Material

Longitudinal Section
Through Superstructure Wing

Figure 7.5 Joint details on the San Antonio Y

460
 Figure 7.6 shows a typical live load test result from Roberts [7]. Significant continuity
was developed across the joint as evidenced by the deflection of the adjacent span. Roberts was
able to accurately predict this deflection using a frame model that treated the bearings as bending
and compression members. Figure 7.6 shows the large degree of continuity that was developed,
even though the bearings at Pier A44 were expansion bearings with a teflon-stainless steel sliding
surface. The rotational or bending stiffness of the 50mm fiber reinforced pads, and friction
between the teflon and stainless steel plates contributed greatly to the negative moment capacity of
the system. This moment continuity was not a desired performance characteristic. Roberts
predicted that elastomeric pads, with their lower shear stiffness, would develop substantially less
negative moment continuity across the joint.

1.2m
1.2m
10.6m

10.6m
3.6m
2.4m
3.6m
Span A43 Span A44

Pier A44
2
Deflection in mm

1
0
-1 Measured
-2
-3
-4
-5
211 kN
204 kN
158 kN
205 kN
199 kN
148 kN

Calculated

Span A43 Span A44

2
Deflection in mm

1
0
-1 Measured
-2
-3
-4
-5 Calculated

Elevation
Figure 7.6 Deflections during live load test, from Roberts, et al. [7]
461
7.4 DATA COLLECTION AND ANALYSIS
Strain gauges were cast within the cast-in-situ joint concrete between spans D5 and D6 of
bridge unit D2, and located as shown in Figure 7.7. Gauges were placed in two layers to measure
bending strains in the slab, as well as tension and compression strains. The transverse location of
the strain gauges was identical to gauges located in the box girder anchor segment. The continuity
slab on bridge unit D2 was the only slab to be instrumented with strain gauges.

17,680mm
3@610mm 3@610mm
2184mm 2184mm 2121mm 2121mm 2184mm 2184mm

229mm
C74 C72 C70C68 C66 C64 (typ.) C62 C60 C58 C56 C54 C52 C50
C75 C73 C71C69 C67 C65 C63 C61 C59 C57 C55 C53 C51

3@51mm (typ.)
51mm clr. (typ.)

C- Gauge

Cast-in-place joint

Section Through Anchor Segment

Near Centerline of Bearing
Figure 7.7 Strain gauge locations in cast-in-place joint–Unit D2

7.4.1 Live load tests

Live load tests were conducted on three different bridge units that had the cast-in-place
continuity slabs. Load cases were selected to maximize the negative moment carried by the joint-
bearing couple, using the trucks available during the test. The trucks used in the test were actually
designed to produce a positive moment, when placed back to back at midspan, similar to an HS20
design truckload.

The first live load test of a joint was conducted on bridge unit D2. Strains in the joint and
girder concrete were recorded electronically, and deflections were measured in all three spans of
the unit. The second test of the continuity slab detail was conducted on Unit C13, a three-cell
cast-in-place two-span unit, providing the transition in roadway width from the mainlane girder to

462
both a ramp and mainlane girder. The third live load test of the joint detail was conducted on two
three-span girder units, C13 and L2, connected longitudinally at their wing tips by a gore closure.

Live load test on Mainlane Unit D2

The live load cases of interest and the span geometry for Unit D2 are shown in Figure
7.8. The truck weights are shown in Table 7.1.

Table 7.1 Axle weights and spacing for live load test trucks on Unit D2
Truck Weight of Rear Weight of Front Total Weight (kN) Axle Spacing
Axles (kN) Axles (kN) (mm)
1 122.5 43.1 165.6 5030
2 132.0 39.6 171.6 4850
3 119.6 44.4 164.0 4890
4 119.1 40.8 159.9 4090
5 135.1 39.0 174.1 4720
6 119.7 44.1 163.8 4360

463
 Front Axle Rear Axle
Weight Weight

Axle Spacing

Live Load Case 2 Typical Truck

38.90m 38.90m 38.90m

Span D4 Span D5 Span D6

17070mm
6 2
3650mm (Typ.)
3 1
3650mm (Typ.) 5
4

D4 D5 D6 D7
CL
3650mm Spacing between
rear axles
Live Load Case 3

6 2
3 1
4 5

D4 D5 D6 CL
D7
3650mm

Live Load Case 5

6 2
3 1
4 5

D4 D5 D6 D7
14325mm 16150mm

Live Load Case 6

6 2
3 1
4 5

D4 CL D5 D6 D7
3650mm

Figure 7.8 Live load cases for Unit D2

464
 The measured strains in the instrumented joint between spans D5 and D6 for the various
load cases have been converted to concrete stresses, and are shown in Figures 7.9 through 7.12.
The top and bottom fiber stresses have been linearly extrapolated from the concrete strains
measured by the two gauges at a given transverse location in the slab.

1.5
Concrete Stress in Mpa (+ Tension)

0.5

-0.5

-1

-1.5
Top Fiber
-2
Top Gauge
-2.5 Bottom Gauge

-3 Bottom Fiber

C- Gauge

Cast-in-place joint

Section Through Cast-in-place Joint

at Pier D6
Figure 7.9 Cast-in-place joint stresses–load case 2

465
 1.5

Concrete Stress in Mpa (+ Tension)

1

0.5

-0.5

-1

-1.5 Top Fiber

-2 Top Gauge
Bottom Gauge
-2.5
Bottom Fiber
-3

C- Gauge
Cast-in-place joint

Section Through Cast-in-place Joint at Pier D6

Figure 7.10 Cast-in-place joint stresses–load case 3

1.5
Concrete Stress in Mpa (+ Tension)

0.5

-0.5

-1

-1.5 Top Fiber

-2 Top Gauge
Bottom Gauge
-2.5
Bottom Fiber
-3

C- Gauge
Cast-in-place joint

Section Through Cast-in-place Joint at Pier D6

Figure 7.11 Cast-in-place joint stresses–load case 5
466
 1.5

Concrete Stress in Mpa (+ Tension)

1

0.5

-0.5

-1

-1.5 Top Fiber

-2 Top Gauge
Bottom Gauge
-2.5
Bottom Fiber
-3

C- Gauge
Cast-in-place joint

Section Through Cast-in-place Joint at Pier D6

Figure 7.12 Cast-in-place joint stresses–load case 6

Integrating the slab stresses over the slab area for the load cases shown in Figures 7.9
through 7.12, the net axial force in the slab was close to zero. Therefore, moment carried by the
slab-bearing couple should have been close to zero. Figure 7.9 shows that the stresses in the joint
were not uniform. The measured stresses tended to peak just inside of the girder webs where the
continuity slab changed width. The continuity slab had a longer bearing area at this location
because of the geometry of the notch in the anchor segment, as can be seen at the very bottom of
Figure 7.1. This caused greater fixity at the anchor segments where the slab changed width, and
therefore increased the bending stiffness of the slab. The stresses shown in Figure 7.9 were
primarily slab bending stresses. Also interesting to note was the apparent difference in bending
stiffness of the slab near the wingtips. The left portion of the slab had greater bending stresses
than the right portion of the slab because of the difference in location of the vertical expansion
cuts in the adjacent parapets. The right parapet had an expansion cut very near the centerline of
the joint which reduced the bending stiffness of the joint and parapet. In load cases 2, 3 and 5, the
average stress at the joint in the wing was compressive. This was due to tensile stresses in the
parapets.

467
 Also interesting to note was the difference in the magnitudes of the joint stresses between
load case 2 (see Figure 7.9) and case 3 (see Figure 7.10). The measured stresses were much
higher in load case 2. The moment connection of the slab to the anchor segment for span D6 must
have been less effective than the connection to the span D5 anchor segment. Shrinkage cracks
between the slab and the precast concrete anchor segments were noted along the entire length of
the interface, and may have been different on each side of the slab. The stresses measured in load
case 6 (see Figure 7.12) indicated that the slab was in positive bending. Therefore, the joint over
pier D5 must have been capable of carrying some negative moment in the joint slab itself. The
reinforcement detail connecting the joint to the notch in the anchor segment would allow a
moment to be developed in the joint because of end rotations of the box girder. The joint was
specifically designed to carry small amounts of moment caused by end rotations of the girders.
The presence of shrinkage cracks at the cold joint would tend to reduce moments in the joint
caused by end rotations of the girders. This would have negligible influence on the behavior of
the box girders, and no influence on the intended function of the joint itself.

The deflection plots shown in Figures 7.13 through 7.16 show that little continuity was
developed across the joint at Pier D6. The magnitudes of the maximum deflections from load
cases 2 and 3, shown in Figures 7.13 and 7.14, were nearly identical at 6mm, or only 1/6500 of the
span length. They were both substantially different from the 4mm maximum deflection in span
D4 from load case 6, shown in Figure 7.16, indicating some continuity of the joint over pier D5.
A negative moment was developed at the down station joint of the three span unit at pier D5
during load case 6, based on the measured upward deflection of span D5 shown in Figure 7.16.
The cast-in-place fixity block (see Figure 7.2), located between the bottom flanges of the box
girders at pier D5, provided a substantially stiffer compression strut when compared to the
elastomeric bearings alone at pier D6.

The negative moment developed was still small and unpredictable because of gaps
between the bottom flanges and the fixity block caused by temperature changes in the girders, and
possibly shrinkage of the block itself. Very small negative moment was developed by the joint at
pier D5 during load case 2, as seen from the very small deflections in span D4 shown in Figure
7.13. The fixity block was also eccentric to the longitudinal centerline of the girder. The joint at
pier D6 was selected for instrumentation because this fixity block, cast around a heavy steel pipe
emanating from the pier capital, was not present.

468
 1

0
Deflection in mm -1

-2

-3

-4

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 7.13 Deflections from live load test on Unit D2–load case 2

-1

-2
Deflection in mm

-3

-4

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 7.14 Deflections from live load test on Unit D2–load case 3
469
 1

Deflection in mm -1

-2

-3

-4

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 7.15 Deflections from live load test on Unit D2–load case 5

-1
Deflection in mm

-2

-3

-4

-5

-6

-7

Span D4 Span D5 Span D6

D4 D5 D6 D7

38.90m 38.90m 38.90m

Figure 7.16 Deflections from live load test on Unit D2–load case 6

470
Live load test on Unit C13
The span geometry for Unit C13 and the load cases of interest are shown in Figure 7.17.
The axle weights for the trucks used are given in Table 7.2. The girder on Unit C13 was originally
designed to be precast and erected as two single celled girders, with cast-in-place top and bottom
slabs between the two girders. Since these girders would have required a special casting machine
and erection equipment, the contractor decided to cast these spans in place. The original girder
shape was retained, as was the cast-in-place continuity slab between spans. Details for the
continuity slab were similar to those for Unit D2.

Table 7.2 Axle weights and spacing for live load test trucks on Unit C13
Truck Weight of Rear Weight of Front Total Weight (kN) Axle
Axles (kN) Axles (kN) Spacing
(mm)
1 119.2 36.7 155.9 4900
2 117.5 43.9 161.4 4720
3 112.0 45.4 157.4 4900
4 119.0 35.1 154.1 4700
5 113.6 30.6 144.2 4700
6 122.7 41.0 163.7 4090

471
 Front Axle Rear Axle
Weight Weight

Live Load Case 2

Axle Spacing
Pier C36

Pier C37

Pier C38
Typical Truck

Span C36 Span C37

40.84m 40.84m
A

1 4
3650mm (Typ.)
2 5
3650mm (Typ.)
3 6

CL A
3650mm Spacing between
rear axles

Live Load Case 4

A

1 4
2 5
3 6 Pier C38
Pier C37

A
Pier C36

CL
3650mm

Varies 1994mm Varies 1994mm 5930mm

Piano wire Gauge mounting

plate
SECTION A-A
Figure 7.17 Live load cases–Unit C13

472
 Span deflections were the only measurements taken during the live load test on Unit C13.
Deflections were measured at the quarter points of each cell in each span. Continuity provided by
the joint for live load was small, as shown in Figures 7.18 and 7.19, even with fixity blocks on
each of the two piers at C37. Maximum deflections were approximately 20% smaller in span C37
than in span C36 because the C37 girder was the wider of the two. The maximum deflection in
span C36 was very small at 1/7900 of the span length. Deflections in the unloaded span in Figures
7.18 and 7.19 averaged to about zero.
1

-1
Deflection (mm)

-2

-3

Legend
-4 Left Cell Measured
Middle Cell Measured
Right Cell Measured
-5

-6

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 7.18 Deflections from live load test on Unit C13–load case 2

473
 1

-1
Deflection (mm)

-2

-3 Legend
Left Cell Measured
-4 Middle Cell Measured
Right Cell Measured

-5

-6

Span C36 Span C37

C36 C37 C38

40.84m 40.84m

Figure 7.19 Deflections from live load test on Unit C13–load case 4

Live load test on Units C15 and L2

The span geometry for the twin single-celled girders and the live load cases of interest are
shown in Figure 7.20. The girders were constructed using the span-by-span method on erection
trusses. The mainlane girder on line C and the ramp girder on line L both had one variable width
wing. The wing tips were joined with a cast-in-place closure strip. The axle weights for the trucks
used in this live load test are shown in Table 7.3. A fixity block was only located at pier C42.

474
Live Load Case 9

Pier L7
Pier L6
Pier L5
Pier L4
Span L6
Span L4 Span L5
39.94m
34.09m 40.20m
A

1 4
2 5
3 6

3650mm
CL
A
Span C41 Span C42 Span C43
34.14m

Pier C43

Pier C44
40.20m 40.07m
Pier C41

Pier C42

Live Load Case 11

Pier L7
Pier L6
Pier L5
Pier L4

1 4
3650mm (Typ.)
2 5
3650mm (Typ.) 6
3

3650mm Spacing between rear axles

Pier C41

Pier C42

Pier C43

Pier C44

CL
A

2050mm 3820mm Varies 5474mm 5798mm

556mm

Piano wire Gauge mounting plate

Section A-A
Figure 7.20 Live load cases–Units C15 and L2
475
 Table 7.3 Axle weights and spacing for live load test trucks on Units C15 and L2
Truck Weight of Rear Weight of Front Total Weight Axle Spacing
Axles (kN) Axles (kN) (kN) (mm)
1 133.9 55.9 189.8 5030
2 128.7 44.0 172.7 4090
3 109.5 41.5 151.0 4100
4 121.3 55.8 177.1 5000
5 137.6 42.5 180.1 5000
6 127.4 34.4 161.8 4700

Deflections were measured down the center of each of the two girder cells for each load
case, and are shown in Figures 7.21 and 7.22. The live load in the side span C41 in load case 9
actually caused the middle span C42 to deflect downward, as seen in Figure 7.21. The end
rotation of the anchor segment for span C41 probably caused uplift forces to be carried through the
joint to the anchor segment for span C42. This uplift force would have been eccentric to the span
C42 bearing reactions, resulting in a positive moment applied to the end of span C42. Figure 7.22
indicates that a negative moment was carried by the joint over pier C42, with the fixity block,
from live loads in the central span, although the moment was probably quite small based on the
small measured upward deflections of spans C41 and L4. Maximum downward deflections in the
loaded spans were quite small at 3mm for span C41 during load case 9 (1/11400 of the span
length), and 4.9mm for span C42 during load case 11 (1/8200 of the span length). The deflection
in span C42 was larger than the deflection in span C41 because span C42 was the longer of the
two spans.

476
 1

-1
Deflection (mm)

-2

Legend
-3 Ramp L Measured
Mainlane C Measured
-4

-5

-6

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m
Figure 7.21 Deflections from live load test on Units C15, L2–load case 9
1

-1
Deflection (mm)

-2

-3

-4
Legend
-5 Ramp L Measured
Mainlane C Measured
-6

Span C41, L4 Span C42, L5 Span C43, L6

C41 C42 C43 C44

34.14m 40.21m 40.08m
Figure 7.22 Deflections from live load test on Units C15, L2–load case 11

477
7.4.2 Temperature gradient effects
To study the effects of temperature gradients in the girders on the joints, two positive
temperature gradient cases were selected. The first gradient occurred on June 4, 1995 before the
continuity joint had been cast, and the second gradient occurred on August 20, 1996 after the joint
and blacktop were in place. These positive gradient cases were selected for their similarity. The
temperature changes measured by the thermocouples in mainlane segment D5-9 are shown in
Figure 7.23 for each case. The magnitude of the August 20, 1996 case was slightly larger than the
June 4, 1995 case, but the two gradient shapes were almost exactly alike.

478
 Temperature Change in Celcius
12

10

Segment D5-9
Under positive gradient conditions
Temperature Change in Celcius

12

10 Temperature Changes
8 June 4, 1995 - No Joint
6 August 20, 1996 - Joint in Place
4
Thermocouple Location
2

-2 0 2 4 8 10 12
6
Temperature Change in Celcius
Figure 7.23 Measured temperatures for thermal gradient load case

479
 Longitudinal cross sectional strains caused by the thermal gradients were measured in
segment D5-9 near midspan, and in segment D5-16 immediately adjacent to the heavy end
diaphragm. The measured strains in segment D5-9 are shown in Figure 7.24. Because of the large
number of damaged gauges in the top flange, no conclusions could be made by comparing the
strain changes from the two gradient cases in this part of the girder. The temperature change in
the wingtip thermocouple was larger in the August 20, 1996 case, and therefore produced higher
strain changes, as shown at the top of Figure 7.24. Strain changes in the web and bottom flange
gauges were similar for each gradient case, making it difficult to identify any influence that the
joint may have had on the girder's structural response to thermal gradients.
Microstrain (+ Tension)

50

-50

-100

Segment D5-9
Microstrain (+ Tension)

Under positive gradient conditions

Temperature Changes
August 20, 1996 - Joint in Place
0
June 4, 1995 - No Joint
-50 Concrete Strain Gauge Location
-100

100
-50

50
0

Microstrain (+ Tension)
Figure 7.24 Measured strains in segment D5-9 from thermal gradient load case

480
 The measured strains in segment D5-16 and the cast-in-place continuity joint are shown
in Figure 7.25. The measured strains from the segment D5-16 gauges in the top flange, webs and
bottom flange were nearly identical for both thermal load cases. The stiffening effect of the joint
on the top flange was apparently small, and did not significantly change the girder's response to
the thermal gradient loading. The positive thermal gradients in the girders did produce bending
and axial strains in the continuity slab. The rotation at the girder ends produced positive bending
in the slab, as can be seen by comparing the top and bottom slab gauge strains at the top of Figure
7.25. These strains tended to peak where the slab changed width, as in the live load cases. The
transverse distribution of joint strains also revealed that the joint was acting to stiffen the top
flange of the girders. The joint strains were more compressive in the wings and tensile between
webs of the box girders, indicating that self-equilibrating stresses were acting across the joint
width because of warping of the girder flanges.

481
 300

250

200
Microstrain (+ Tension)

150

100

50

-50

-100

-150

Segment D5-16 and Cast-in-Place Joint

Microstrain (+ Tension)

Under positive gradient conditions

Temperature Changes
August 20, 1996 - Joint in Place
0 June 4, 1995 - No Joint
Joint Top Gauges on August 20,1996
-50
Joint Bottom Gauges on August 20, 1996
Concrete Strain Gauge Location
-100

-50

50
0

Microstrain (+ Tension)

Figure 7.25 Measured strains in segment D5-16 and cast-in-place joint from
thermal gradient load case
482
7.5 CONCLUSIONS AND RECOMMENDATIONS
The cast-in-place continuity joints on the U.S. 183 Elevated, in most respects, performed
as designed. The joints were intended to provide a low cost riding surface between simple span
box girders. The joints were not designed to carry superstructure moments across the joint, but
were designed to carry axial loads produced by shear forces in the elastomeric bearings from
thermal expansion of the box girders. The joints had a semi-rigid moment connection to the
anchor segments of the box girders and were designed for moments from truck wheels, and shear
forces from the end rotation of the box girders.

7.5.1 Construction procedures

The reinforcing bars for the cast-in-place continuity joints were spliced to 90o bend bars
in the anchor segments, seen in Figure 7.1, that passed through the construction joint seat for the
cast-in-place slab. The 90o bend bars were difficult to properly locate when the anchor segments
were cast, resulting in the addition of many extra bars that were drilled into and bonded to the
anchor segment. Also, these 90o bend bars often intruded into the horizontal surface intended for
the fiberboard material used to tune the effective length of the joint. Widening the notch for the
cast-in-place joint would have reduced the potential for inadequate splice lengths in the 90o bend
anchor segment bars, and provided more room for the fiberboards and stay-in-place forms. The
short overlap of the 180o bars in the joint and 90o anchor segment notch bars provided adequate
tensile strength across the deck joint during the test period, but long term strength of this spice
detail is questionable and was not tested in this study.

The fiberboards used under the joint between girder wings were also used as the stay-in-
place form for the joint itself. A steel stay-in-place form was eventually used in the wider portion
of the joint between girder webs. The use of these stay-in-place forms was ideal because of the
lack of access to the underside of the joint once it was cast, and the speed at which the forms could
be placed.

In order to control the movement of the end of the box girders from thermal expansion
while the joint concrete hardened, adjacent box girders were locked together with the used of post-
tensioning bars. These bars passed through the future post-tensioning ducts located in the anchor
segment diaphragms. Only one joint of a span was locked at any one time. The joint was locked
at a time when the girders were cool so that some residual compression would be applied to the
joint after the joint was unlocked. The system seemed to work quite well, although the shrinkage

483
of the joint concrete eventually caused a crack to open along the entire length of the cold joint,
reducing the fixity of the joint to the girders. Shrinkage cracks were also noted to propagate from
the reentrant corner of the joint where the joint slab changed width. Softening the reentrant corner
of the joint with a 45o angle would be nearly as easy to form in the surface of the anchor segment,
and would lessen the stress concentration effect of the 90o corner. Fiberboards with a 45o cut
could be used under the joint slab in this area if the 90o corner is retained beneath the joint.

7.5.2 Cast-in-place joint behavior under live loads and thermal loads
The continuity joint did not substantially alter the structural performance of the adjacent
box girders. The large negative moment couple that was measured on the San Antonio Y [7] did
not develop in the U.S. 183 semi-continuous three-span or two span units. The relatively stiff
fabric reinforced pads with Teflon-stainless steel sliding surface developed a substantial negative
moment in the San Antonio Y girders. Such a couple should be avoided unless this effect is
specifically included in the design of the girders and the joint. The elastomeric bearings used on
the U.S. 183 girders were very weak in shear, as this type of bearing is intended to be, and did not
allow the formation of any appreciable negative moment in the live load tests. The joint itself
experienced local bending moment from the rotation of the ends of the girders from both live loads
and thermal loads. The bending strains in the joint were much increased at the reentrant corner of
the joint where the joint suddenly changed width. The joint width change was provided between
girder webs to allow access to the tendon ducts. A 45o transition where the joint changed width
would have lessened the stresses in the joint, while not influencing access to the tendon ducts or
the interior of the box girder.

The cast-in-place fixity block between box girder bottom flanges was responsible for the
development of some moment continuity across joints where the blocks were located. The
negative moments developed ranged from insignificant to large. They did not develop in every
case, depending on the fit of the block to the girder flanges. The fixity blocks were also eccentric
to the centerline of the girder. In order to eliminate the formation of this negative moment, the
fixity block should be moved or redesigned to eliminate a moment couple with the continuity
joint.

484
 CHAPTER 8
SUMMARY, RECOMMENDATIONS AND CONCLUSIONS

8.1 PROJECT SUMMARY

The U.S. 183 Segmental Bridge is an urban viaduct constructed with both precast and
cast-in-place structural elements. The mainlane girders were precast segmentally and constructed
using the span-by-span method. One three-span semi-continuous (top slab continuity only)
mainlane unit was selected as one of the units for study during precasting, erection and under live
load. The mainlane girders rested on innovative Y-shaped piers with a steel pipe tension tie across
the top of the “Y”. One Y-pier was instrumented in conjunction with the mainlane three-span unit
under study. Most of the ramp girders were also constructed using the span-by-span method.
However, five of the spans of Ramp P were constructed in balanced cantilever. This five span
girder constructed in balanced cantilever was also selected for detailed study, along with one
precast segmental pier located beneath the most heavily instrumented ramp span. A limited study,
only for live load behavior, was carried out on a two span semi-continuous three-cell girder that
was cast in-place, as well as on three spans of twin single-cell girders with gore closures at the
wing tips. These single-cell girders were constructed by the span-by-span method, and provided
the transition in cross section needed where the ramp girders merged into the mainlane girders.

One of the topics studied in detail on the project was post-tensioning tendon force losses,
which included friction losses, elastic shortening losses, and long term losses. Several different
post-tensioning tendon profiles were studied on both the mainlane and ramp girders, along with a
U-shaped tendon in the segmental pier. Another topic studied was thermal gradients and their
effects. Thermal gradients were measured in the mainlane and ramp girder, and in the segmental
voided pier. Temperatures were measured every hour beginning immediately after construction,
and measurements continued for years. The measured gradients and measured thermal induced
strains and related stresses were compared to design code recommended gradients, and stresses
calculated using common analysis methods. The third topic studied was the general response of
the girders and piers to applied loads, including dead loads, prestressing loads, and live loads.
Measured strain distributions, related stress profiles, and deflections were compared to results
calculated using methods recommended in design specifications, such as the AASHTO effective

485
flange width method. Another topic under study was the performance of post-tensioned anchorage
zones and other D-zones (discontinuity zones). The service level performances of seven D-zones
were evaluated using strain measurements and visual inspections. The D-zones studied included
two tendon deviators, two anchor segments with heavy anchorage diaphragms, two pier capitals,
and one anchorage blister. The final topic under study was the performance of the cast-in-place
deck slab joint that provided the riding surface between simple span girders. Live load tests were
performed to determine if a moment couple developed between this deck joint and the bearings.

8.2 GENERAL CONCLUSIONS

The implementation of sound and proven design and construction practices resulted in a
highly successful TxDOT bridge project for the City of Austin. High quality piers and
superstructure were constructed. The design proved to be stiff, amply strong, as well as attractive.
The project was completed on schedule with few problems. The bridge design was performed by
the TxDOT Design Division staff. These engineers drew on their experiences from the
construction of the San Antonio Y segmental viaduct when designing the U. S. 183 segmental
project. The primary contractor was Martin K. Eby Construction Company in a joint venture with
Flatiron Structures Company. The experienced construction team produced high quality precast
segments with few rejections. The erection scheme for the piers and superstructure was
straightforward, resulting in few delays once initial spans were successfully erected. The designer
allowed the constructor to use flexible and consistent construction cycles.

8.2.1 Post-tensioning losses

The post-tensioning tendons used on the project were of common design, and provided an
opportunity for routine friction and elongation tests. The friction losses were found to be higher
than expected in the deviated external tendons, and lower than expected in the internal straight
tendons. The elongation measurements could not always be counted on to accurately predict the
measured friction losses because of small differences in tendon stiffness.

8.2.2 Thermal gradients

The positive design thermal gradients recommended for central Texas by NCHRP 276
[39] were found to be excessive and temperature measurements indicated the magnitudes at deck
level could be reduced from 26oC to 16oC for the no blacktop case, and from 20oC to 15oC with
50mm of blacktop. However, from comparison of measured stresses with the results of the

486
conventional analysis procedure indicated that further development of this unproved analytical
technique is urgently required. In the interim the NCHRP values and the current analysis
procedure give results close to the measured structural response.

The negative design thermal gradient recommended by NCHRP 276 was found to be
excessive and temperature measurements indicated the magnitudes at deck level could be reduced
from peak values of -12.8oC to –7oC without blacktop and –10oC to –5oC with 50mm of blacktop.
However, they were found to produce calculated strains less than those actually measured that
occurred with less severe measured negative thermal gradients. Again, improved calculation
methods are urgently needed. The AASHTO LRFD [9] design negative gradient shape was found
to be inaccurate when compared with measured gradients. The magnitude of positive and negative
thermal gradients through the thicknesses of the girder webs and flanges indicated that the thermal
gradient design case should also be considered during transverse design of the cross section. The
transverse thermal design gradients should be linear. Significant positive and negative thermal
gradients were measured in the voided segmental pier, yet produced tensile stresses important only
during certain construction load cases.

8.2.3 Load response of box girders

The structural response of the various elements under study to dead load, prestressing,
and live loads was found to be predictable using common design methods. The diffusion of post-
tensioning forces assuming the spread of force within a 60o cone, and the prediction of shear lag
by the effective flange width method were found to be adequate, with some limitations. The
diffusion of temporary post-tensioning forces needs to be carefully considered when determining
the compressive stresses actually present during curing of epoxy joints. The measured deflections
of the box girders under live loads indicated high stiffness, with the multi-cell girders able to
significantly distribute or share moments from unsymmetrically placed live loads.

8.2.4 Behavior of D-zones

The discontinuity zones studied all proved to be very conservatively designed, but had
different service level performances. One D-zone had large cracks with bars stressed half way to
the yield point because of sharply bent deviator pipes, while others had no service level cracking.
Strut-and-tie modeling, with due consideration for the load paths to the boundaries of the D-zones,
is encouraged.

487
8.2.5 Cast-in-place joint behavior
The instrumented cast-in-place joint, providing the riding surface between simple spans
of the three-span semi-continuous bridge units, generally performed as designed. No live load
moment was carried across the joint because of the lack of shear stiffness in the elastomeric
bearings. The details of the joint could be modified to improve constructability, especially where
the cast-in-place joint splices to the precast anchor segment.

8.3 SPECIFIC CONCLUSIONS AND RECOMMENDATIONS

The following specific conclusions and recommendations are based on the evaluation of
the measurements, review of other field and laboratory data, and based on visual observations
made over the course of the project.

8.3.1 Post-tensioning losses

1.) Measurements of live end anchorage zone friction losses in the laboratory bench test and in-
place friction test indicated that an assumed design loss of 2% would be sufficient, unless
actual live end losses are known from previous measurements. This includes the friction
losses within the trumpet, through the anchorage hardware, and from the stressing equipment.

2.) Measured wedge seating losses were slightly less than the design value of 6mm. The current
design value is adequate.

3.) The bench test proved to be of little value to all parties except the researchers, other than as a
basic calibration trial of the various pieces of the stressing system. The modulus of elasticity
determined in the bench test did not prove to be representative of most of the tendons used in
the structure, presumably because of slight variations in strand area. The in-place friction test
was much more useful for providing information to the engineers and constructors. Accurate
elongation calculations must be based on the results of an in-place friction test, otherwise the
measured elongation tolerance may not be easily met.

4.) Wobble friction in the straight internal ducts of the structure was quite small. These ducts
were effectively held in position during concrete placement by inflatable mandrels. Friction
coefficients for internal tendons in ducts constructed using inflatable or rigid mandrels can
conservatively be selected as μ=0.16 and K=0.0007m-1. For draped internal ducts, friction

488
 coefficients are μ=0.16 and K=0.0013m-1 for monolithic girders and K=0.0016m-1 for
segmental girders, based on other studies [19] [22].

5.) The friction coefficient for external tendons in smoothly bent deviator pipes with consistent
radius can be chosen as μ=0.25. The friction coefficient in the sharply bent deviator pipes
used in some of the U. S. 183 girders, at about a 2m radius, generated a coefficient of friction
of about μ=0.35. The sharp radius bend also caused large cracks in the deviator concrete, and
should be avoided.

6.) The additional wobble angle β=0.04 radians suggested by Roberts [7] was found to be
sufficient when applied at each deviator of the mainlane girders, if the proper coefficient of
friction was used in the calculation. The additional wobble angle β=0.04 radians was
recommended based on studies of girders constructed span-by-span with straight or large
radius horizontal geometry. The additional β angles measured in Ramp P, with a horizontal
curvature of 221m, were higher at β=0.11 radians when using an assumed friction coefficient
of μ=0.25. The horizontal curvature of the girder makes accurate deviator pipe placement
more difficult, thereby warranting a higher design β angle. The β angle should be applied at
all deviators and saddles. The use of a diabolo, or double trumpet bell shaped deviator pipe,
would help reduce the β angles on curved structures. The diabolo style deviator pipe was not
necessary for the mainlane girders, based on the friction tests.

489
Table 10-2 in the AASHTO Segmental Guide Specification [8] should be modified as follows:

Friction Wobble
Coefficient (μ) Coefficient
1. For draped strand tendons in galvanized metal sheathing 0.16-0.25* 0.0016m-1
(segmental construction)

2. For straight strand tendons in galvanized metal sheathing 0.16 0.0007m-1

(segmental construction using mandrels)

6. For strand tendons in rigid steel pipe deviators 0.25-0.35** 0.04rad***

* A friction coefficient of 0.25 is appropriate for duct curvatures with radii between 6m and 15m.
** A friction coefficient of 0.35 is appropriate for duct curvatures with radii of 2m or less.
*** This additional wobble angle is applied at each deviator and saddle, and may be higher for
girders with horizontal curvature. It can be reduced to 0.02rad if a diabolo type deviator pipe is
provided.

7.) Anchorage details for the long 150m (3 span) external tendons in ramp P adjacent to a
deviation saddle proved to be unacceptable. The large elongations caused entangled tendons
to be drawn close to the back of the anchor head where they broke. Straight anchorage
geometry would have allowed the 19 strand tendons to untangle to some extent in the long
distance between the anchorage and the deviator. No strand breakages of this type occurred
in any of the 14 tendons in each of 162 spans of mainlane girders. If deviation saddles are
required adjacent to a live end anchorage where a first pull must be made, the length of
elongation may need to limited, requiring stressing from both ends of the tendon. Proper
support of unstressed tendons over their deviated length would help reduce the total
elongation substantially by reducing the slack length.

8.) Elastic shortening loss calculations for the external deviated tendons were found to be
inaccurate if slippage was not assumed to occur at the deviators. The measured values fell
between the cases calculated using a deviator with infinite friction and zero friction. The
more conservative loss from these two cases should be used for design unless an extremely
complex calculation is performed.

9.) Long term losses were found to be small when compared to other losses for the girders under
study. The segments were well aged before they were erected and prestressed.

490
8.3.2 Thermal gradients
The following recommendations and conclusions have been made based on the measured
gradients in the various structural elements under study, as well as the measured response of the
structure both to these gradients and to the current AASHTO LRFD design thermal gradients. The
recommended gradients based on this study apply specifically to central Texas.

8.3.2.1 Superstructure gradients for longitudinal design

Positive thermal gradients

1.) The 95% fractile values for measured deck level magnitudes of the positive gradients were
similar for both the cases without and with 50mm of blacktop, at 16oC without blacktop and
15oC with blacktop, although thermal data without blacktop were limited.

2.) The NCHRP 276 or AASHTO LRFD recommended design positive gradients for girders with
50mm of blacktop at T1=20oC, and T1=25.6oC without blacktop, were unrealistically high
when compared to the 95% fractile T1,meas values, and also considerably larger than the
absolute maximum T1,meas values. The distribution of positive gradient magnitudes over time
indicated that the thermal gradient case without blacktop deserves a higher design gradient
magnitude than the case with 50mm of blacktop. Measured peak positive gradient
magnitudes at deck level decreased when the instrumented bridge units were opened to traffic,
presumably because of cooling from the increased airflow (see Figure A.1).

3.) The AASHTO LRFD recommended positive gradient shape more accurately represented the
measured positive gradient shape of the ramp P girder, and the NCHRP 276 recommended
positive gradient shape better represented the shape measured on the mainlane, with the
exception of the deck level temperature. In either case, the temperature gradient in the bottom
slab should be considered, with the soffit level temperature at 3oC for central Texas.

4.) Calculated stresses using the AASHTO LRFD recommended design positive thermal gradient
compared well to stresses measured on the mainlane girder, but compared poorly and
unconservatively to stresses measured on the ramp P girder, even though the design gradient
was larger than the measured gradient. Evidence of sectional distortion or warping was
measured in every thermal gradient case. Also, soon into the life of the girders, high strains
were measured in response to thermal gradients in the top slab over the webs. These high
strains were measured in both the mainlane girder and the ramp P girder, and would indicate

491
 plastic behavior in the concrete. Improved analysis techniques for thermal gradients are
urgently requirted.

5.) Based on the measured positive gradients alone, a reduction in the magnitude of the design
positive thermal gradient T1 values would be warranted. However, analytical study of the
structural response to thermal loads of a wide variety of girder cross sections needs to be
performed before any reduction in the design positive thermal gradients can be implemented.
The effects of cross sectional shape, diaphragms, continuity, and potential plasticity should be
considered in this study.

Negative thermal gradients

1.) The shape of the AASHTO LRFD recommended negative thermal gradient did not compare
well to measured negative thermal gradient shapes in either the mainlane girder or the ramp P
girder. The negative gradient shape recommended by NCHRP 276 better represented the
actual shape of the negative gradient, especially in the lower part of the cross section.

2.) Based on the measured negative thermal gradients, the peak top fiber gradient temperatures
recommended by AASHTO LRFD or NCHRP 276 were too extreme for Central Texas. The
maximum T1,meas values were closer to -0.3 times the NCHRP 276 recommended positive
gradient temperatures. Measured 95% fractile T1,meas values were –7oC without blacktop and
–5oC with 50mm of blacktop at deck level. These 95% fractile values are substantially below
the AASHTO LRFD recommended peak negative design gradient temperatures of T1=–13oC
and T1=–10oC without and with blacktop respectively. All points of the negative gradient
other than the top fiber temperature would be represented fairly accurately with the NCHRP
276 negative gradient shape.

3.) Based on the unconservative calculated stresses in the ramp P girder when compared to the
measured stresses, no change to the current recommended design negative gradient from
NCHRP 276 can be recommended, pending further study of box girder response to thermal
gradients.

492
8.3.2.2 Thermal gradients for transverse design

1.) Measured stresses from both positive and negative thermal gradients through the thicknesses
of the top slab, webs and bottom slab were large enough to warrant a design thermal gradient
for transverse design.

2.) Based on the measured temperatures, a positive thermal gradient should only be applied to the
top flange, and the gradient shape should be linear. For Central Texas, a positive thermal
gradient for no blacktop should have a peak deck level temperature of 18oC, and decrease at
-0.072oC/mm for top slabs less than 250mm thick. For thicker top slabs the 18oC gradient can
be assumed to decrease linearly through the thickness of the slab to 0oC at the bottom fiber.
Similarly, a positive thermal gradient for 50mm blacktop should have a peak deck level
temperature of 17oC, and decrease at -0.068oC/mm for top slabs less than 250mm thick. For
thicker top slabs the 17oC gradient can be assumed to decrease linearly through the thickness
of the slab to 0oC at the bottom fiber

3.) Based on the measured temperatures, significant negative thermal gradients occurred
simultaneously in the top flange, webs, and bottom flange. For Central Texas, a negative
thermal gradient for a top flange with no blacktop should have a peak deck level temperature
of -10oC, and increase at 0.040oC/mm for top slabs less than 250mm thick. For thicker top
slabs the -10oC gradient magnitude can be assumed to increase linearly through the thickness
of the slab to 0oC at the bottom fiber. Similarly, a thermal gradient for 50mm blacktop should
have a peak deck level temperature of -5oC, and increase at 0.020oC/mm for top slabs less
than 250mm thick. For thicker top slabs the -5oC gradient can be assumed to decrease
linearly through the thickness of the slab to 0oC at the bottom fiber. The negative gradients
occurring in the webs and bottom flange should also be assumed to be linear, and can be
obtained by multiplying the recommended top flange negative gradient peak temperature and
slope for no blacktop by 0.75. This gives a peak surface temperature of –7.5oC increasing at
0.030oC/mm.

493
8.3.2.3 Thermal gradients for the design of piers

1.) Significant thermal gradients and thermal induced stresses were measured in the voided
segmental pier P16 and the solid mainlane pier D5.

2.) A positive thermal gradient for the design of voided piers based on the measurements can be
derived by multiplying recommended positive gradient shape for box girders from Section
8.3.2.1 with no blacktop by 0.75. The peak magnitude of the gradient is 12oC. The far fiber
temperature should remain at 3oC.

3.) A negative thermal gradient for the design of voided piers based on the measurements is
identical to that in Section 8.3.2.1 for superstructures, with a peak gradient value of –7oC.

4.) Given the inaccuracy of the analysis method commonly used, the proposed gradients should
not be used until better analysis methods have been developed. Pending further research into
a better analysis method, the NCHRP 276 recommended positive thermal gradient shape with
no blacktop multiplied by 75% can be used for the design of voided piers. The extreme fiber
temperature should remain at 3oC. The NCHRP 276 recommended negative thermal gradient
shape with no blacktop can also be used for the design of voided piers. The peak negative
gradient temperature at the extreme fiber should be reduced to -0.3 times the recommended
positive gradient temperature with no blacktop, and the rest of the gradient shape should be
used unmodified.

5.) Although little data were taken to define the shape of thermal gradients in solid pier sections,
the magnitude of the thermal gradients were measured to be similar to those of the voided
pier. The recommended thermal gradients for the design of voided piers in 4). above are
recommended for the design of solid piers until further studies can be done.

6.) The significant stresses produced by the thermal gradients in the piers were mostly
inconsequential for pier design because the dead load axial stresses prevented tension, except
for one construction load case during the construction of the balanced cantilever
superstructure of ramp P. The decision to use a thermal gradient load case for the design of
piers should be made by the engineer.

The following text should be added to Section 7.4.4 in the AASHTO Segmental Guide
Specification: At the discretion of the Engineer, voided segmental piers may be designed for

494
thermal gradients. A positive thermal gradient for the design of voided piers can be derived by
multiplying the NCHRP 276 recommended positive gradient shape for box girders with no
blacktop by 0.75, with the exception of the far fiber temperature which should remain at 3oC. The
NCHRP 276 recommended negative thermal gradient shape with no blacktop can be used for the
design of voided piers. The peak negative gradient temperature at the extreme fiber should be
reduced to -0.3 times the recommended positive gradient temperature with no blacktop.

7.) Daily thermal induced stress changes in the piers were measured to be of the same magnitude
as those produced by the superstructure dead load. In order to control surface stresses and
concrete fatigue cracking, a nominal amount of transverse steel should be selected for a pier
based on both the concrete volume and surface area, such as by the AASHTO LRFD equation
5.10.8.2-1.

8.) Negative thermal gradients that occurred during curing of the pier P16 segments were large
enough to crack the concrete segments while in the form. Negative thermal gradient
o
magnitudes were measured as high as -35 C shortly after removal from the form. An area of
transverse steel calculated by the AASHTO LRFD Equation 5.10.8.2-1 would not have been
enough to prevent the cracking, since the transverse area of steel actually in the pier segments
exceeded the amount found by this equation. A designer should consider the negative
gradient produced in higher strength concrete elements during curing. Transverse steel should
be increased to handle the thermal stresses, or provisions should be made to reduce the heat of
hydration.

8.3.3 Load response of box girders

The following conclusions have been made based on comparisons of the measured data
with the calculated results:

8.3.3.1 Temporary post-tensioning for epoxying segments

1.) An engineer should determine whether or not the segments under design will take a "banana"
shape during casting and curing, estimate the warping deflection that will occur, and include
the deflection in the design of the temporary post-tensioning to insure full closure of the gap.

2.) The diffusion of the temporary post-tensioning force from the anchorages or blisters should be
estimated using a 30o diffusion angle, and stresses calculated at the extremities of the cross

495
 section away from the anchorage points. Assuming that plane sections will remain plane, and
that a linear stress gradient will pass through the center of gravity of the section may lead to
an inaccurate estimate of the actual stress distribution. The temporary post-tensioning forces
and locations should be designed to adequately stress the entire cross-section considering
diffusion. Furthermore, depending on temporary support methods, the dead load of the
segment constructed in balanced cantilever may not produce a predictable stress distribution
during the epoxying and temporary post-tensioning process, and should not be relied upon as
a source of bottom flange epoxy squeezing stress.

3.) The sequence of temporary post-tensioning should be considered, especially if anchorages are
not well distributed throughout the cross section.

8.3.3.2 Diffusion of post-tensioning forces

1.) The AASHTO 30o diffusion method assumes a post-tensioning force or other point load is
fully diffused in a concrete member within a 60o cone emanating from the point of force
application. This method tended to underestimate the amount of diffusion of post-tensioning
forces from anchorages in or immediately adjacent to anchorage diaphragms. The strut-and-
tie method or other continuum method should be used to predict stresses near these
diaphragms.

2.) The AASHTO 30o diffusion method is not sufficiently accurate for calculating stresses in the
vicinity of post-tensioning anchorages. These local zone stresses can be designed based on
the recommendations from other studies such as NCHRP Report 356 [54].

3.) The AASHTO 30o diffusion method was sufficiently accurate for predicting the distance from
an anchorage to the point of full diffusion into the cross section.

4.) At sections where the post-tensioning force is calculated to be fully diffused, shear lag in the
cross section from primary post-tensioning moments can be compensated for by using the
AASHTO effective flange width method. Only the bf width should be calculated and used
over the entire span since the pier reactions are not included in the calculation of primary
moments or the stresses resulting from these moments.

5.) Use the AASHTO effective flange width method for predicting cross sectional stresses from
secondary moments.

496
8.3.3.3 Girder response to dead loads and live loads

1.) The measured stresses and deflections from the simulated HS20-44 truck load (no impact)
were small when compared to girder dead load. Live loads on the mainlane girder produced
deflections only 1/6500 of the span length, while calculated dead load deflection was 1/1050
of the span length.

2.) The AASHTO effective flange width method gave sufficiently accurate results for the
calculation of stresses and deflections from dead loads and live loads, although the girders
tested did not experience significant shear lag at sections of high moment and stress gradient.

3.) The AASHTO effective flange width method requires considerable section property
calculation for girders that may only experience a small amount of shear lag, such as most
common segmental girders and nearly all simple span segmental girders. Another method
should be developed for such girders that uses unmodified section properties.

8.3.3.4 Performance of multiple-cell girders under live load

1.) The twin single-cell girders tested shared as much as 30% of the applied live load moment.

2.) The amount of live load moment sharing between the twin single-cell girders was sensitive to
the torsional stiffness of each girder. The stiffness of the bearings must be included in the
design of these girders.

3.) The three-cell girders tested consistently shared as much as 43% of the applied live load
moment.

4.) The amount of live load moment sharing between sides of the three-cell girders was not
sensitive to the stiffness of the bearings because of the presence of the pier diaphragm that
was cast full width of the three girder cells.

5.) Live load deflections were small in the multiple-cell girders with a maximum measured
deflection of only 1/6300 of the span length or smaller.

8.3.3.5 Performance of the segmental pier under bending

1.) The cross sectional behavior of the segmental pier was entirely predictable, but the stiffness of
the drilled shaft foundation had to be included in the analysis.

497
2.) The moment connection of the pier to the balanced cantilever ramp superstructure using a
cast-in-place grout pad and 16 post-tensioning bars performed as would a monolithic
connection.

8.3.4 Behavior of D-zones

The service level performance of the D-zones under study ranged from inadequate when
unanticipated details were used, like sharply bent deviator pipes, to excellent when designs were
conservative. The following recommendations and conclusions were drawn from observations of
the measured data, visual inspections, and review of the designs and details.

8.3.4.1 Mainlane Y-pier

1.) The mainlane Y-pier was designed as a concrete-steel composite structure. The force in the
transverse steel pipe ties was reasonably but unconservatively predicted using a frame
analysis, but was more conservatively predicted by a simple strut-and-tie model.

2.) The anchorage of the mainlane Y-pier pipe tension ties in the concrete was excellent, with a
fully bonded condition existing at the time of testing under full superstructure dead load.

3.) The stiffness of the pipe ties limited the bending moment in the concrete Y compression and
bending struts such that no cracking occurred at the service load level. Given the low stress
levels, the pier should behave adequately with an ultimate load placed on the superstructure.

8.3.4.2 Segmental pier P16 capital

1.) The pier P16 capital segment essentially remained uncracked from service load level forces,
although the capital was initially cracked from thermal gradient forces during curing. These
cracks had little impact on the service level performance of the capital, based on visual
inspection. Measured strains from service loads were small when compared to the strains
needed to crack the concrete.

2.) The heavy bars placed longitudinally and transversely at the top and bottom of the pier capital
carried tensile force as expected in most cases, but the circumferential gauged bars at the
bottom of the pier measured strains of equal or larger magnitude to the designed heavy tie

498
 bars. A strut-and-tie model for this solid section should include circumferential ties at the
bottom to maintain similar geometry to the voided section beneath at the ultimate load
condition.

8.3.4.3 Deviators

1.) The instrumented mainlane and ramp deviators were beam type deviators similar in
dimension and reinforcement, with the exception of the width. The vertical tendon loads on
the ramp deviator were only about 85% of those on the mainlane deviator. The deviation
force on the tendon ducts did not appear to be well distributed along the length of the duct in
the ramp deviator, but no large cracks developed and no exceptionally large strains were
measured in the bars. The ramp deviator had good service level performance. The mainlane
deviator ducts did not meet the design requirement for smoothly bent pipes. The actual pipes
used were sharply bent, concentrating the entire deviation force at the center of the deviator.
This concentration of force caused extensive open cracks, and high stresses in the reinforcing
bars adjacent to the sharp bend in the duct. Deviation ducts should be smoothly bent over the
full length of the deviator to improve service level performance. A sharp bend also increases
friction between tendon, grout and duct, increasing horizontal loads on the deviator at
ultimate. Such sharp bends should not be permitted.

2.) The reinforcement details for both deviators were easily constructed at points away from the
girder webs. Near the web to deviator intersection, the heavy web stirrup bars and other bars
transitioning the web to the bottom flange remained unchanged from that for a typical section.
This left little room for the deviator top bars or shear reinforcement. Priority should be given
to the placement and development of the bars connecting the deviator to the webs, since this is
critical force path for developing the ultimate moment capacity of the girder.

8.3.4.4 Mainlane anchor segment D5-16

1.) The pier segment, with heavy anchorage diaphragm, for the mainlane was not post-tensioned
vertically to control tensile stresses from the diffusion of post-tensioning forces. Regardless
of this fact, the mainlane pier segment performed well at service load levels. The highest
measured reinforcing bar stress was 35% of the yield stress, and occurred in the heavy vertical
bar immediately adjacent to the access passage through the diaphragm at a crack location.

499
2.) The heavy end diaphragm was designed to take all the transverse and vertical tensile stresses
from the anchorage zone at the ultimate load condition. At the service load level, the bottom
flange of the girder near the anchor segment remained under high compression from the
internal bottom slab tendons. The bottom flange cracked longitudinally directly beneath the
bottom flange internal tendon ducts. The designed D-zone should have included a length of
typical section, or a distribution of transverse service level stresses should have been
estimated, and additional steel provided to control cracking.

8.3.4.5 Ramp P anchor segment P16-1

The pier segment instrumented on ramp P was post-tensioned vertically to make a moment
connection with the pier. The top flange of the girder was also post-tensioned transversely. As a
result little cracking occurred in the heavy anchorage diaphragm, and strain measurements were
small.

8.3.4.6 Anchorage blister

1.) The instrumented anchorage blister was uncracked following post-tensioning and thereafter,
indicating a very conservative design. A significant tensile stress change was measured
behind the anchorage, at 1.7MPa, primarily from secondary prestressing moment. The tensile
force change in the bottom flange was about 70% of the force from the post-tensioning tendon
in the blister, based on the measured strains. Since the bottom flange of the girder was in
significant compression from cantilevering dead load, no cracks were found. Very minor
tension was also measured in the bars in front of the anchor plate that connected the blister to
the web, indicating little corbel action. No corbel action was indicated between the blister and
the bottom slab. Instead, compression was measured in the vertical bars adjacent to the
anchor plate where the blister joined the bottom slab.

2.) Tension was measured in the vertical bars along the length of the blister because of the
vertical force generated by the radial deviation of the tendon over the length of the blister, and
from diffusion of the tendon force into the blister.

500
8.3.5 Cast-in-place joint behavior

8.3.5.1 Construction Procedures

1.) The reinforcing bars for the cast-in-place continuity joints were spliced to 90o bend bars in the
anchor segments that passed through the construction joint seat for the cast-in-place slab. The
90o bend bars were difficult to properly locate when the anchor segments were cast, resulting
in the addition of many extra bars that were drilled into and bonded to the anchor segment.
Also, these 90o bend bars often intruded into the horizontal surface intended for the fiberboard
material used to tune the effective length of the joint. Widening the notch for the cast-in-
place joint would have reduced the potential for inadequate splice lengths in the 90o bend
anchor segment bars, and provided more room for the fiberboards and stay-in-place forms.
2.) The fiberboards used under the joint between girder wings were also used as the stay-in-place
form for the joint itself. A steel stay-in-place form was eventually used in the wider portion
of the joint between girder webs. The use of these stay-in-place forms was ideal because of
the lack of access to the underside of the joint once it was cast, and the speed at which the
forms could be placed.
3.) In order to control the movement of the end of the box girders from thermal expansion while
the joint concrete hardened, adjacent box girders were locked together with the used of post-
tensioning bars. These bars passed through the future post-tensioning ducts located in the
anchor segment diaphragms. Only one joint of a span was locked at any one time. The joint
was locked at a time when the girders were cool so that some residual compression would be
applied to the joint after the joint was unlocked. The system seemed to work quite well,
although the shrinkage of the joint concrete eventually caused a crack to open along the entire
length of the cold joint, reducing the fixity of the joint to the girders. Shrinkage cracks were
also noted to propagate from the reentrant corner of the joint where the joint slab changed
width. Softening the reentrant corner of the joint with a 45o angle would be nearly as easy to
form in the surface of the anchor segment, and would lessen the stress concentration effect of
the 90o corner. Fiberboards with a 45o cut could be used under the joint slab in this area if the
90o corner is retained beneath the joint in the anchor segment.

501
8.3.5.2 Cast-in-place joint behavior under live loads and thermal loads

1.) The continuity joint did not substantially alter the structural performance of the adjacent box
girders. The large negative moment couple that was measured on the San Antonio Y [7] did
not develop in the U. S. 183 semi-continuous three-span or two span units. The relatively stiff
fabric reinforced pads with Teflon-stainless steel sliding surface developed a substantial
negative moment in the San Antonio Y girders. Such a couple should be avoided unless this
effect is specifically included in the design of the girders and the joint. The elastomeric
bearings used on the U. S. 183 girders were very weak in shear, as this type of bearing is
intended to be, and did not allow the formation of any appreciable negative moment in the
live load tests.
2.) The cast-in-place continuity joint experienced local bending moment from the rotation of the
ends of the girders from both live loads and thermal loads. The bending strains in the joint
were much increased at the reentrant corner of the joint where the joint suddenly changed
width. The joint width change was provided between girder webs to allow access to the
tendon ducts. A 45o transition where the joint changed width would have lessened the
stresses in the joint, while not influencing access to the tendon ducts or the interior of the box
girder.
3.) The cast-in-place fixity block between box girder bottom flanges was responsible for the
development of some moment continuity across joints where the blocks were located. The
negative moments developed ranged from insignificant to large. They did not develop in
every case, depending on the fit of the block to the girder flanges. The fixity blocks were also
eccentric to the centerline of the girder. In order to eliminate the formation of this negative
moment, the fixity block should be moved or redesigned to eliminate a moment couple with
the continuity joint.

8.4 OBSERVATIONS ON CONSTRUCTABILITY AND DESIGN

The following observations are intended to point out the elements of design and methods
of construction that proved to be excellent, as well as those that were less than optimal.

8.4.1 Mainlane Y-pier

1.) Since most of the superstructure was to be precast, creating the necessity for a large
precasting facility, all the piers on the project were initially designed to be precast and post-

502
 tensioned. Considering the construction timeline, the time required to construct the precasting
facility and bring casting machines into operation, and the availability of space at the site, the
contractor elected to cast most of the piers in place. Designers should have considered
including a cast-in-place substructure option in the contract plans as a means of potentially
reducing bid prices, and reducing the amount of design engineering to be performed during
the course of the construction period.

2.) The cast-in-place construction of the mainlane piers proceeded without difficulty, with the
exception of the fabrication of the capital reinforcement bar cage. The complex shape of the
capital formwork, the configuration of the bars, and the exact fit of the tension tie pipes to the
forms made it desirable that the capital bar cages be tied in the form. The contractor in
electing to cast-in-place should have designed the cage to be tied on a jig, thus speeding
construction of the cage and eliminating bar cage fabrication time from the capital
construction time. Based on observation, this would also have greatly improved the working
condition for the ironworkers.

3.) The wires intended to deter roosting of pigeons on the flat surfaces of the pier have proved to
be ineffective.

8.4.2 Mainlane girder

1.) The simple span design of the mainlane girders allowed the superstructure construction to
proceed uninterrupted on a pace of a span every two days for each set of trusses. Reducing
the number of critical construction steps was important since many of the time consuming
tasks could only be performed at night. The simple span construction also eliminated
continuity induced thermal gradient stresses.

2.) The proportions chosen for the mainlane girder cross section could have been more efficient
structurally if soffit width was increased. However the light soaring appearance was a more
important consideration for project acceptance by the public.

3.) The twin triangular erection trusses designed for U. S. 183 worked well at all locations except
near the straddle bents. The trusses selected by the contractor could not fit in the space below
the girder wing and above the top surface of the straddle bents. This should have been
considered in the construction engineering.

503
4.) The casting procedure for the typical mainlane segments was quickly refined and perfected,
producing very high quality segments. The complicated shape of the anchor segments,
combined with the congestion of heavy bars with their lap splices initially made construction
of these segments difficult. To reduce congestion designers should consider using post-
tensioning bars or T-headed bars to replace heavy mild reinforcement bars with conventional
anchorage hooks and lap splices.

5.) Many of the casting machines had bulkheads for transverse pretensioning of the top flange of
the segments. This was very efficient given the number of segments cast in each of these
machines. In general, the foundation conditions were suitable for rock anchors that tied the
heel of the bulkheads to the ground.

6.) The ride quality on the three-span semi-continuous bridge units was excellent. The asphalt
topping was placed continuously across the cast-in-place deck joints and the ballasted future
location of the finger joints. The asphalt was later cut, and the finger joint and surrounding
cast-in-place concrete installed flush with the top of the asphalt.

7.) The duct boot clamps, that provided the water tight seal between the steel deviator and
diaphragm pipes and the polyethylene ducts, were removed immediately following grouting.
This eliminated the air tight seal for the grout. Grout allowed to dry out readily shrinks and
cracks. These clamps should be left in place.

8.) The bottom slab tendon anchorages were deviated at one end of the girder to facilitate tendon
feeding. Tendons were prefabricated and installed by pulling through from the girder dead
end. The deviation was not necessary, and required that the anchor segments on the live and
dead end of the girder have different blockout locations.

9.) The temporary post-tensioning blisters proved very effective and efficient during span
construction, although forming and reinforcing these external blisters added some complexity
to the precasting process.

10.) The 25mm drain holes through the bottom slab and deviator beams were too small in
diameter, and immediately became clogged with debris and eventually grout. Rain frequently
entered the girders through the storm drains and other openings before all the storm drain
piping was installed, flooding the girders.

504
11.) The access holes located in the bottom flange near one anchor segment in each span were
frequently blocked by the storm drain pipes.

12.) Storm drains should be located only near pier segments, in the first adjacent segment, to
minimize the length of pipe located inside the core of the box girder. Also, routing the storm
drain pipes through the center of the anchor segment access passage destroys the usefulness of
the access passage. The storm drain pipe should be routed against the side of large size access
passages, or routed through the diaphragm of anchor segments with small access passages.

8.4.3 Segmental Pier P16

1.) The precasting and erection procedures for the segmental piers were very efficient.

2.) Casting the voided typical column segments was very simple compared to the casting
procedure and complexity for some of the segments on the project, but a systematic twist was
cast in. Fortunately this twist was accommodated at the ramp anchor segments that were cast
later. Otherwise, the ducts for the tie down bars would not have aligned.

3.) Vertical thermal cracks formed in the voided pier segments during curing. These cracks were
located adjacent to the large drain pipes cast within the cross section. Also, drain pipes were
installed where they were not necessary in some piers used for cantilever construction of the
superstructure. This probably weakened these piers torsionally.

4.) The capital segment reinforcement was tied in two separate pieces because the heavy bars at
the top of the cage could not be supported by the lower portion of the cage, requiring
additional time for ironwork inside the form. A one-piece cage would have reduced
production time

5.) The capital segment generated too much heat during curing, and cracked in many places from
tensile stresses caused by the huge negative thermal gradient.

6.) The tie down bars and anchorage plates were difficult to locate at the required compound
angle within the central portion of the pier capital segment. The tie down bars were also
difficult to stress at deck level because of close spacing and a small recess. A U-shaped
strand tendon with oversize duct would have allowed for more alignment error, and been
easier to stress within the recess at deck level.

505
7.) The cast-in-place base around and under the first pier segment above the footing, placed on a
stay-in-place steel stand, appears to have been an effective method of connecting the precast
elements to the footing. Unfortunately, it was impossible to check for voids and collection of
bleed water on the underside of the first segment.

8.) The small amount of transverse steel originally designed for the pier was doubled for actual
construction. Shear capacity of the pier should be sufficient to develop a ductile failure mode
for the entire five span frame.

9.) Fabrication of the reinforcement cage for the typical voided pier segments was simple. A
segmental post-tensioned cage is lightweight with small size vertical bars. The effectiveness
of the short bars, with a 180o bend on one end and 90o bend on the other end, placed on a grid
through the thickness of the pier segment walls should be evaluated. These bars were
intended to provide a restraining force on the small vertical bars to prevent their buckling
when the local concrete became plastic under an ultimate load, and to restrain the transverse
reinforcement so it could act as concrete confinement. The grid spacing was much too great
to prevent buckling of the small size vertical bars, or provide confinement, especially in light
of the fact that these short bars had an ineffective 90o bend on one end. In a post-tensioned
pier the vertical bars are intended to reinforce the concrete against tensile stresses from
thermal forces, shrinkage forces and other forces not associated with the develop of a plastic
moment. These vertical bars are not the main tension bars, nor are they significant
compression reinforcement or confinement reinforcement.

10.) Four internal post-tension bars were used to construct the segmental pier. Two of these bars
were terminated at the top of the last typical voided pier segment. The blockout for the
anchorage created two substantial voids beneath the pier capital that were not filled with
epoxy or grout. The contract plans should specify that the voids be filled at the time the
capital is set in place.

8.4.4 Ramp P constructed in balanced cantilever

1.) The quality of the precast segments was excellent. They were constructed similarly to the
mainlane segments using the short line method.

506
2.) The overall construction scheme of the ramp P five span continuous girder was excellent, and
allowed construction to proceed from one or both ends of the five span girder. Efficient
construction required two ground based cranes.

3.) The web bar detail at the top slab, with 90o bends for development, created significant
congestion. The ducts for the cantilevering tendons had to be deviated from outside the plane
of web bars to the anchorage plate at the centerline of the web. Closed stirrup type web bars
should be used where possible.

4.) The two-piece anchor segment was difficult to assemble and move into final surveyed
position on top of the pier capital. This type of assembly would have been greatly simplified
if the halves were epoxied together on the ground then lifted into place, weight permitting.
The anchor segment should have been designed to be of similar weight to the other segments
on the project so that it could have been cast and erected in one piece.

5.) Final positioning on top of the pier may have been simplified with the use of a precision
match cast concrete or high strength grout template at the locations with permanent fixity.
The use of fiber reinforced elastomeric pads as temporary bearings worked well once the
anchor segment halves were assembled.

6.) Small adjustments of the ramp geometry were easily accomplished because of the flexibility
of the piers. Substantial adjustments up to 150mm were easily made when the permanent
expansion bearings were installed on piers P14 and P17.

7.) Access to the core of the box girder was poor until the girder was essentially completed. The
access passage through the top flange should have been located near the anchor segments for
use during balanced cantilever construction and post-tensioning of the continuity tendons.

8.) The external temporary post-tensioning anchorage blisters were easy to use, although the
location of bottom flange blisters interfered with movement of workers and with the
alignment of the external tendons. The use of the top flange cantilever tendon ducts for
temporary post-tensioning of the first two typical segments to the anchor segments worked
poorly. The post-tension bars were nearly impossible to remove because the size of the bar
couplers was only marginally smaller that the size of the duct, and because of the horizontal
curvature of the bridge. The temporary post-tensioning bars should have passed through the
anchor segment diaphragm, regardless of the congestion.

507
9.) The 19-strand external continuity tendons used in the ramp girder were the same size as most
of the tendons used on the rest of the project, but unfortunately required the use of a very
large ram inside the completed box girder. Clearance problems were immediately evident. A
combination of the large amount of slack from the lack of support of the unstressed external
tendons, the large number of strands in the tendons, the long lengths of the tendons and
resultant long elongations, the location of a saddle immediately adjacent to the stressing end
anchorage, and the eventual entanglement of the strands resulted in several broken tendons. It
can be assumed that some similar entanglement also existed behind the anchor heads of the
tendons that did not break.

10.) The complex three-dimensional geometry of Ramp P made proper alignment of the deviator
and saddle pipes very difficult in the short-line casting machine. Pipes with bell-shaped ends,
or “diabolos”, would have greatly reduced the impact of a misalignment.

11.) The drain holes were too small, as in the mainlane girder, and immediately became clogged.

8.4.5 Transition spans

1.) The contractor elected to cast the three-cell transition spans in place because of two reasons.
First, there were relatively few transition spans so the cost of two additional casting machines
was not warranted. Second, the erection trusses used on the project could not be used to erect
the precast single-wing, single-cell girders without major modification. The decision to cast
these girders in place was later regretted. The merger of a ramp girder into the main girder is
a problem often encountered in precast segmental construction. Provisions should be made in
the design so that the standard casting machines can construct the special segments required at
the transition locations. Also, construction methods should be conceived, considering pier
and superstructure shapes, so that the standard erection trusses can be used to erect the special
segments. For example, on the U. S. 183 transition spans the full width anchor segment
diaphragm could have been cast between precast anchor segments, and the girders constructed
as torsional cantilevers using only one erection truss under each of the single wings.

2.) The location of the drain pipes was poor, as for the mainlane girder. The combination of the
three external tendons and the drain pipes in the small outer two cells of the three-cell girder
made passage in these cells nearly impossible.

508
8.4.6 Modified spans
1.) Construction of the modified spans proceeded without major incident, and the gore closures
between the two girder wings were easily formed and cast from deck level.

2.) The drain pipes were poorly located and blocked the bottom flange access openings.

8.5 RECOMMENDATIONS FOR FUTURE STUDIES

8.5.1 Instrumentation
The Campbell 21X dataloggers and AM416 Multiplexers used on the project worked
exceptionally well without error. The greatest limitation of this datalogger system was the
maximum time interval of one hour. Current Campbell dataloggers may have overcome this
limitation. When using the AM416 multiplexers, it is important not to switch power to the
instruments because damage to the precision low resistance switches can be the result. Future use
of a multiplexer used for such a purpose could cause errant readings because of the increased
resistance of the switches. The quarter bridge wiring scheme used on this project with the strain
gauges worked very well for the short term measurements, and allowed a maximum number of
channels to be read for each multiplexer. The long term use of this wiring system is questionable.
Since switching occurs within the bridge, changes in resistance anywhere in the system are
interpreted as strain changes when the output voltage is read.

If this system is to be used for long term readings, eliminating the task of rewiring the
system after the short term measurements have been completed, multiple dummy gauges should
be read by the system to help calibrate the strain gauges in the structure over time. Also, the
interval used for the short term measurements should not exceed 30 seconds, unless preliminary
tests prove otherwise, so that temperature induced resistance changes do not occur. For the same
reason, the system should be allowed to run at the minimum time interval for a period of time to
let the system reach a stable temperature before testing begins. A better wiring scheme for long
term strain gauge measurements with the 21X or similar datalogger would use an independent full
bridge for each strain gauge, switching only the output voltage. Excitation voltage could be
provided directly by the datalogger either intermittently or continuously. Battery life would be
reduced if the gauges are excited continuously. No problems should be expected from the wiring
schemes used for other gauges used on this project.

509
 The strain gauges used on this project, whether C-type or S-type, performed well over the
short term unless damaged during construction. Damage usually occurred from vibration or direct
impact. The gauges also tended to fail rapidly after traffic was allowed on the bridge. To limit
damage to the C-type gauges, installed on short steel rods and embedded in the concrete, larger
rods should be used than the 5mm rods used on this project. The necessary increase in length of
the larger rods would also speed field installation. The 5mm rods used on this project for the C-
gauges worked exceptionally well in the laboratory, but were easily bent in the field. For long
term measurements weld-on type strain gauges should be used when steel elements are
instrumented, especially if field installation does not allow temperature and humidity control, or if
vibration is expected. Some vibrating wire gauges should be used for long term concrete strain
measurements. High redundancy of strain gauges is recommended, since some are bound to fail
and all gauges do not have to be used. The shielded and teflon coated thermocouples used on this
project worked extremely well.

Gauges for each instrumented segment should be wired to an independent datalogger

system that can be tested, and calibrated to temperature changes, well before erection of the
segment begins. Demec strain measurements can only be relied upon for very short test periods
on an actual structure, since temperatures and temperature gradients change continuously
throughout the day. A substantial number of temperature measurements, combined with time
consuming analysis, is required to eliminate thermal strains from the Demec strain data. The
speed at which Demec strains can be read almost precludes their use in any number during short
term tests. The use of more internal or external thermally compensated strain gauges is preferable
to the use of the Demec for all but backup duty, unless actual strain changes are required. Reliable
strain gauge data from the instrumented tendons proved the most difficult of measurements.
These gauges were field installed with the glue least sensitive to temperature, yet often did not
stay bonded to the tendons after they were stressed, given the large strain change induced shear
force in the glue. Other methods should be considered for instrumentation of tendons. However,
weld-on strain gauges do not seem feasible.

The number of gauges of all types used on this project was formidable, but denser grids
of strain gauges should have been used in the cross sections of the girders to measure, for
example, strain changes across the width of the webs. More cross sections along the length of the
girders should also have been instrumented to evaluate the effects of the D-zone forces.

510
8.5.2 Future research topics
Friction losses were not entirely predictable on this project. A large database of friction
data must exist, particularly with the proprietary prestress hardware suppliers and state agencies.
This data should be evaluated, and future research needs recommended, if necessary. Friction
between tendon and deviator duct following stressing should be studied for both ungrouted and
grouted tendons. This friction has important design implications on elastic shortening loss
calculations, ungrouted tendon force changes from construction loads, and tendon force changes at
the ultimate load state for various allowed plastic rotations.

An extensive thermal record was obtained for various structural elements on U. S. 183.
The thermal record was largely limited to the years after the application of asphalt, but the thermal
gradients from NCHRP 276 appear to be much larger than actually occur. The commonly used
design method for calculating thermal gradient stresses does not appear to be accurate, based on
the measured strains. The common design method assumes plane sections remain plane, and this
was definitely not the case. An analytical study needs to be performed to evaluate the current
design method. The study should consider different box girder shapes and girder continuity, as
well as the influence of changes in the design gradient shape on calculated stresses. The effect of
local concrete plasticity in areas of the cross section that have heavy reinforcement should also be
considered. The analysis results could be tabulated as a design aid for standard box girder
sections. The reinforcing steel in the pier segments, especially the solid capital, was inadequate
for controlling thermal induced tensile stresses during curing. Given the importance of high one-
day concrete strengths for precast segments, with the resultant high heat of hydration, a study
should be performed to determine an adequate percentage of surface reinforcement for control of
thermal stresses for segmental elements of different shapes, thicknesses, and concrete mixtures.

The box girders studied at U. S. 183 behaved in a predictable manner to dead loads, live
load, and prestressing forces. Shear lag in the box girders did not have great influence on the
design. Using the AASHTO Effective Flange Width Method [8] resulted in substantial cross
sectional reduction of the mainlane simple span girder only near the supports, where moment was
low. Also, the ramp P girder was compact and did not suffer greatly from shear lag induced
stiffness reduction. Nevertheless, the stress calculations for these girders were as complex as for
girders with substantial shear lag. A new method for calculating shear lag induced stress increases

511
in box girders of common proportion should be developed and tabulated for standard box girder
cross sections.

512
 APPENDIX

SUMMARY OF MEASURED THERMAL GRADIENTS

The following Figures A.1, A.2 and A.3 were prepared by M. Keith Thompson for CTR
Report 0-1820 [56]. The report presents a summary of all the long term thermal data taken by
University of Texas researchers at both the U. S. 183 Segmental in Austin, and at the San Antonio
Y Project. Report 0-1820 also gives a statistical analysis of the data beyond that performed for
this dissertation. From Figures A.1 and A.2 it can be seen that the daily maximum negative
thermal gradients decreased markedly after the blacktop was applied to these superstructure box
girders. Also from Figure A.1, the maximum daily positive gradient appears to have decreased
somewhat after the bridge was opened to traffic. Little change in the maximum daily positive or
negative thermal gradients occurred in the segmental pier column as a result of the construction of
the superstructure, judging by the regularity of the data over time in Figure A.3.

Daily Peak Positive Gradients

Daily Peak Negative Gradients
18 Bridge Open to Traffic (8/22/96)
16
Gradient Magnitude (oC)

14
12
10
8
6
4
2
0
-2
-4
-6
-8 Blacktop in Place (5/27/96)
-10
Sept.

Sept.

Sept.
June

June

June

June
Aug.

Nov.
Dec.

Aug.

Nov.
Dec.

Aug.

Nov.
Dec.

Aug.
Mar.

Feb.
Mar.

Feb.
Mar.

Feb.
Mar.
May

May

May

May
Jan.

Jan.

Jan.
Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Apr.
July

July

July

July

1995 1996 1997 1998

Figure A.1 US 183 Mainlane box girder daily maximum thermal gradients

513
 Daily Peak Positive Gradients
Daily Peak Negative Gradients
18
16 Bridge Open to Traffic (4/4/97)
Gradient Magnitude (oC)

14
12
10
8
6
4
2
0
-2
-4
-6
-8 Blacktop in Place (3/7/97)
-10

Sept.
June

June
Nov.
Dec.

Nov.
Dec.
Aug.

Aug.
Feb.
Mar.

Feb.
Mar.
May

May
Jan.

Jan.
Apr.

Oct.

Apr.
July

July
1996 1997 1998
Figure A.2 US 183 Ramp P box girder daily maximum thermal gradients

Daily Peak Positive Gradients

Superstructure Erection Daily Peak Negative Gradients
18 (9/96 - 11/96)
16
Gradient Magnitude (oC)

14
12
10
8
6
4
2
0
-2
-4
-6
-8
-10
-12
Sept.

Sept.
June

June

June
Aug.

Aug.
Nov.
Dec.

Nov.
Dec.
Feb.
Mar.

Feb.
Mar.
Jan.

Jan.
May

May

May
Apr.

Oct.

Apr.

Oct.

Apr.
July

July

1996 1997 1998

Figure A.3 US 183 segmental box pier P16 daily peak thermal gradients

514
 REFERENCES
1. Billington, Sarah L., “Improving Standard Bridges Through Aesthetic Guidelines and
Attractive, Efficient Concrete Substructures,” Ph.D. Dissertation, The University of Texas at
Austin, December 1997.

2. Koseki, K., Breen, J. E., “Exploratory Study of Shear Strength of Joints for Precast Segmental
Bridges,” Center for Transportation Research Report No. 248-1, Austin, Texas, September
1983.

3. MacGregor, R. J. G., Kreger, M. E., Breen, J. E., “Strength and Ductility of a Three-Span
Externally Post-Tensioned Segmental Box Girder Bridge Model,” Center for Transportation
Research Report No. 365-3F, Austin, Texas, January 1989.

4. Taylor, A. W., Rowell, R. B., Breen, J. E., “Design and Behavior of Thin Walls in Hollow
Concrete Bridge Piers and Pylons,” Center for Transportation Research Report No. 1180-1F,
Austin, Texas, November 1990.

5. Beaupre, R. J., Powell, L. C., Breen, J. E., “Deviation Saddle Behavior and Design for
Externally Post-Tensioned Bridges,” Center for Transportation Research Report No. 365-2,
Austin, Texas, July 1988.

6. Wollman, G. P., Yates, D. L., Breen, J. E., “Deviation Saddle Behavior and Design for
Externally Post-Tensioned Bridges,” Center for Transportation Research Report No. 465-2F,
Austin, Texas, November 1988.

7. Roberts, C. L., Breen, J. E., Kreger, M. E., “Measurement Based Revisions for Segmental
Bridge Design and Construction Criteria,” Center for Transportation Research Report No.
1234-3F, Austin, Texas, August 1993.

8. Guide Specifications for Design and Construction of Segmental Concrete Bridges, American
Association of State Highway and Transportation Officials, Washington, D. C., 1989.

9. AASHTO LRFD Bridge Design Specifications, 1st Edition, American Association of State
Highway and Transportation Officials, Washington, D.C., 1994.

10. Roberts, Carin L., “Measurement Based Revisions for Segmental Bridge Design and
Construction Criteria,” Ph.D. Dissertation, The University of Texas at Austin, December
1993.

11. Arrellaga A., J. A., “Instrumentation Systems for Post-Tensioned Segmental Box Girder
Bridges,” Master’s Thesis, The University of Texas at Austin, 1991.

12. Stone, W. C., “Design Criteria for Post-Tensioned Anchorage Zone Tensile Stresses,” Ph.D.
Dissertation, The University of Texas at Austin, May 1980.

13. Andres, Valerie A., “Verification of Force Distribution in an Innovative Bridge Pier,”
Master’s Thesis, The University of Texas at Austin, 1995.

515
14. Chang, W. R., Etsion, I., Bogy, D. B., “Static Friction Coefficient Model for Metallic Rough
Surfaces,” Journal of Tribology, Transactions of the American Society of Mechanical
Engineers, Vol. 110, January 1988.

15. Mindess, Young, Concrete, Prentice-Hall, Englewood Cliffs, New Jersey, 1981.

16. Bezouska,T. J., “Friction Loss in Post-Tensioned Prestressing Steel Units,” Report SSR-3-66,
Division of Highways, Department of Public Works, California, September 1966.

17. Bezouska, T. J., “Friction Losses in Rigid Post-Tensioning Ducts,” Division of Highways,
California Department of Public Works, California, February 1971.

18. Dywidag Systems International, USA, Inc., “Methods for Reducing Friction in Post-
Tensioning Tendons, “Dywidag Report, Lincoln Park, New Jersey, February 1988.

19. Tran, T. T., “Reducing Friction Loss for Post-Tensioning Tendons in Monolithic Girders,”
Master’s Thesis, The University of Texas at Austin, December 1992.

20. Harstead, G. A., Kummerle, E. R., Archer, J. C., Porat, M. M., “Testing Large Curved
Prestressing Tendons,” Journal of the Power Division, Proceedings of the American Society
of Civil Engineers, March 1971.

21. Yasuno, H., Kondo, S., Tadano, N., Mogami, T., Sotomura, K., “Friction Problems with
Multi-Strand Tendons,” Proceedings of the 9th Congress of the Federation Internationale de la
Precontrainte, Volume 2, Wexham Springs, England June 1982.

22. Davis, R. T., Tran, T. T., Breen, J. E., Frank, K. H., “Reducing Friction Losses in Monolithic
and Segmental Bridge Tendons,” Center for Transportation Research Report No. 1264-2,
Austin, Texas, October 1993.

23. Standard Specifications for Highway Bridges, Sixteenth Edition, American Association of
State Highway and Transportation Officials, Washington, D.C., 1996.

24. Comite Euro-International du Beton-Federation Internationale de la Precontrainte, CEB-FIP

Model Code for Concrete Structures, Switzerland, 1990.

25. ACI Committee 318, Building Code Requirements for Reinforced Concrete (ACI-318-95),
American Concrete Institute, Detroit, 1995.

26. PTI, Precast Segmental Box Girder Bridge Manual, Prestressed Concrete Institute and Post-
Tensioning Institute, 1977.

27. ACI-ASCE Committee 323, “Tentative Recommendations for Prestressed Concrete,” Journal
of the American Concrete Institute, Proceedings V. 54, January 1958.

28. Zia, P., Preston, H. K., Scott, N. L., and Workman, E. B., “Estimating Prestress Losses,”
Concrete International: Design and Construction, V. 1, June 1979, pp. 32-38.

29. ACI Committee 209, “Prediction of Creep, Shrinkage, and Temperature Effects in Concrete
Structures,” ACI-209R-82, American Concrete Institute, Detroit, Michigan, 1982.

516
30. ASTM A416-74, from Post-Tensioning Manual, Third Edition, Post-Tensioning Institute,
Phoenix, Arizona, 1982.

31. Tassin, D., Dodson, B., Takebayashi, T., Deeprasertwong, K., Leung, Y. W., “Computer
Analysis and Full Scale Test of the Ultimate Capacity of a Precast Segmental Box Girder
Bridge,” The American Segmental Bridge Institute.

32. Hoffman, P. C., McClure, R. M., and West, H. H., “Temperature Studies for an Experimental
Bridge,” Research Project 75-3 Interim Report, Pennsylvania State University, June 1980.

33. Hirst, M. J. S. and Dilger, W. H., “Prediction of Bridge Temperatures,” IABSE Proceedings
P138/89, August 1989.

34. Shiu, K. N., “Seasonal and Diurnal Behavior of Concrete Box-Girder Bridges,”
Transportation Research Record 982.

35. Priestly, M. J. N., “Design of Concrete Bridges for Thermal Gradients,” ACI Journal, May
1978, pp. 209-217.

36. Arockiasamy, M., and Reddy, D. V., “Thermal Response of Florida Bridges,” Florida Atlantic
University, Boca Raton, Florida, June 1992.

37. Pentas, H. A., Avent, R. R., Gopu, K. A., and Rebello, K. J., “Field Study of Bridge
Temperatures in Composite Bridges,” Transportation Research Record No. 1460, pp. 45-52.

38. Potgieter, I. C., and Gamble, W. L., “Response of Highway Bridges to Nonlinear Temperature
Distributions,” Report No. FHWA/IL/UI-201, University of Illinois at Urbana-Champaign,
April 1983.

39. Imbsen, R. A., Vandershof, D. E., Schamber, R. A., and Nutt, R. V., “Thermal Effects in
Concrete Bridge Superstructures,” NCHRP 276, Transportation Research Board, Washington,
D. C., September 1985.

40. British Standards Institution, “Steel, Concrete and Composite Bridges, Part I, General
Statement,” British Standard BS 5400, Crowthorne, Berkshire, England 1978, pp. 43.

41. Ryals, K. K., “Fretting Fatique of Tendons in Deviator Ducts of Externally Post-Tensioned
Segmental Box Girders,” Master’s Thesis, The University of Texas at Austin, 1992.

42. Beer, F. P., Johnson, E. R., Mechanics of Materials, McGraw-Hill, New York, 1981.

43. 1983 Ontario Highway Bridge Design Code, Highway Engineering Division, Toronto,
Ontario, 1983.

44. Kristek, V., Theory of Box Girders, John Wiley and Sons, New York, 1979.

45. Song, Q., “Shear Lag Analysis of Simple and Continuous T, I, and Box Beams,” Report No.
UCB/SESM-84/10, University of California at Berkeley, June 1984.

46. Ritter, W., “Die Bauweise Hennebique (The Hennebique System),” Schweizerische
Bauzeitung, Bd. XXXIII, No. 7, Zurich, 1899.

517
47. Morsch, E., Der Eisenbetonbau, Seine Theorie und Anwendung (Reinforced Concrete, Theory
and Application), Stuttgard, 1902.

48. Marti, P., “Truss Models in Detailing,” Concrete International, V. 7, No. 12, 1985, pp. 66-73.

49. Muller, P., “Plastic Analysis of Reinforced Walls and Beams,” Report No. 83, Institute of
Structural Engineering, ETH Zurich, 1978, 160pp.

50. Thurlimann, B., “Plastic Analysis of Reinforced Concrete Beams,” Introductory Report,
IABSE Colloquium on Plasticity in Reinforced Concrete (Copenhagen, 1979), International
Association for Bridge and Structural Engineering, Zurich, V. 28, 1978, pp. 71-90.

51. Leonhardt, F., Monnig, E., “Lectures on Reinforced Concrete Structures, Part 2: Special
Cases of Calculations for Reinforced Concrete Construction,” 3rd edition, Springer Verlog
Publishers, Berlin, 1986.

52. Schlaich, J. et al, “Toward a Consistent Design of Structural Concrete,” PCI Journal, Volume
32, No. 3, Prestressed Concrete Institute, May/June 1987.

53. Bergmeister, K. et al, “Detailing for Structural Concrete,” Center for Transportation Research
Report 1127-3F, Austin, Texas, May 1993.

54. Breen, J.E., Burdet, O., Roberts, C., Sanders, D., Wollman, G., “Anchorage Zone
Reinforcement for Post-Tensioned Concrete Girders,” NCHRP Report 356, Transportation
Research Board, National Academy Press, Washington, D. C., 1994.

55. Beaupre, R. J., Powell, L. C., Breen, J. E., Kreger, M. E., “Deviation Saddle Behavior and
Design for Externally Post-tensioned Bridges,” Center for Transportation Research Report
365-2, Austin, Texas, July 1988.

56. Thompson, M. K., Breen, J. E., Kreger, M. E., “Indications About Thermal Gradient
Magnitudes From Studies of Concrete Box Girder Bridges,” Center for Transportation
Research Report 0-1820, Austin, Texas, October 1998.

518