Professional Documents
Culture Documents
31 47 65
Composite Full-scale Precast,
elastic behavior experimental prestressed
of precast testing of UHPC concrete
concrete insulated in deck bulb-tee girder spans
wall panels girder connections using 0.7 in. strand
YOUR TRUSTED SOURCE FOR
CONCRETE LIFTING SYSTEMS
AND ACCESSORIES
www.conacweb.com
31
Investigating UHPC in Deck Bulb-Tee Girder Connections, 47
Part 2: Full-Scale Experimental Testing
Abdullah Haroon, Eric Steinberg, Richard Miller, Bahram Shahrooz, and Waleed Hamid
Index of advertisers
ALP ............................................................. 8, 9 PCI...............................................6, 16, 26, 88
www.alpsupply.com www.pci.org
CONAC........................Inside Front Cover Prestress Supply Inc ............ Back Cover
www.conacweb.com www.prestresssupply.com
Hamilton Form ......................................... 20 Tindall ........................... Inside Back Cover
www.hamiltonform.com www.tindallcorp.com
JVI ..................................................................... 1 Tuckers........................................................... 4
www.jvi-inc.com www.tuckerbilt.com
July–August 2023
Volume 68, Number 4
Departments
On the cover
Precast/Prestressed Concrete Institute
Chairman’s Message 5
Situated on a triangular
block in downtown San Your PCI
Francisco, Calif., the Serif
Residences and the Line Slabs, Panels, and Hollow-core
President’s Message 7
Hotel use glass-fiber- Plan to participate in Precast Days
reinforced precast concrete
panels to achieve a three-
dimensional faceted facade.
31
Composite
elastic behavior
47
Full-scale
experimental
of precast testing of UHPC
65
Precast,
prestressed
concrete
From PCI Headquarters 10
concrete insulated in deck bulb-tee girder spans
Our Members 17
In the News 21
Industry Calendar 23
Project Spotlight 24
Research Corner 27
PCI Directories 89
Board of Directors and
Technical Activities Council 89
Regional Offices 91
JOURNAL EDITORIAL ADVISORY COMMITTEE Coming Ahead 91
Chair Richard A. Miller Adel ElSafty Chungwook Sim
Vice Chair Pinar Okumus Amir Fam Sri Sritharan Meet Frank Nadeau 92
Secretary Collin Moriarty Alexander G. Mihaylov
Staff Liaison Tom Klemens Stephen J. Seguirant
Laura Vidale Copy Editor Postmaster: Please send address changes to PCI Journal, 8770 W. Bryn Mawr Ave., Suite 1150, Chicago, IL 60631.
Laura Bedolla Technical Activities Program Manager Periodicals postage rates paid at Chicago and additional mailing offices.
Your PCI
R ecently, at a regional event, producer members were asked what the perceived benefits are
of belonging to PCI. The answers were quite positive and varied. Returning from the trip,
I had time for reflection on the answers and to consider what the correlation in the answers
may be.
The strength of PCI comes directly from the active involvement of volunteer members. PCI
has 107 committees and councils with 2645 volunteers. That is an amazing level of involvement.
Those who are on the committees are responsible for progress and steering their respective com-
mittee. The councils are in charge of the committees and have seats on the board. Those with
seats on the board steer PCI.
The correlation in the answers was the level of involvement in PCI. Those who are involved
are getting more benefit. It’s not rocket science, but it’s certainly noteworthy. PCI is enjoying
another successful year thanks to those volunteer members and PCI staff. As there is more salt
than pepper in the hair of those on the board, we must constantly look to involve the next
generation.
I thought about what my answer to the question would be. It was not any of the given
answers. The phrasing of the question did not lend itself to one of the biggest benefits of PCI:
the people. Getting involved with an intelligent group of motivated people is highly rewarding
on a personal and professional level. There is great joy in attaining common goals and develop-
ing lifelong friendships. If you had told young me that committee and board work would be so
rewarding, I would not have believed you, and I would have been wrong. When attending meet-
ings, wherever they are, I feel like I’m home with my friends.
This is your PCI. Make it so. J
Matt Ballain
2023 PCI Board Chair
Vice President and General Manager
Coreslab Structures (INDIANAPOLIS) Inc.
Indianapolis, Ind.
Sponsored by:
Plan to participate
in Precast Days
W hile summer is in full swing, we are planning some exciting in-person events this fall at
PCI. We hope you can be a part of all of them.
Registration is now open for Committee Days, which is October 4–7 in Tampa, Fla., the first
time the meeting will be held outside of Chicago, Ill. We are very excited about the opportunity
to spread our wings in these new locations. The J. W. Marriott is a new hotel, and the location
in downtown Tampa is outstanding. In addition to the usual committee meetings and social
gatherings, we are planning a special event to support the PCI Foundation and also celebrate the
retirement of PCI managing director of quality programs, Mike Kesselmayer, at the end of the
year. (Rumor has it there may even be a bit of live entertainment provided by a PCI member!)
Also on the agenda this fall is the Architectural Precast Concrete Production Workshop
September 6–8 in Hartford, Conn. Hosted by the Architectural Precast Committee, this two-
day event for operations personnel of architectural producers includes presentations and a plant
tour of a PCI producer member.
Most exciting is PCI’s full participation in Precast Days as we celebrate and showcase the pre-
cast concrete industry October 9–20. During this period, precast concrete producer members are
invited to open their doors and host a local Precast Days event that is customized to your com-
pany and its values and products. PCI joins the Canadian Precast/Prestressed Concrete Institute
and the National Precast Concrete Association in this North American showcase of precast con-
crete from coast to coast. Precast Days also coincides with Manufacturing Day.
Precast Days is a great opportunity for you to help grow the awareness of precast concrete.
Activities might include plant tours, education sessions, job fairs, and product demonstrations.
Many producers who have participated in the past have used the event for a variety of audiences.
Opening to the local community can show local residents what precasters contribute to their
built environment. It can also showcase the benefits of building with precast and precast, pre-
stressed concrete to members of the design community. In addition, local students get to see the
possibilities of building with precast concrete and the opportunities available with a career in the
precast concrete industry. It’s also a great opportunity to invite local politicians and members of
Congress to tour the plant and see how precast concrete benefits the community.
Whatever the audience, we encourage all PCI members to consider planning an open house
as part of Precast Days this fall. More information about Precast Days and a link to marketing
materials for promoting the event are available under Marketing Resources in the Members sec-
tion of the PCI website, PCI.org. Please reach out to PCI marketing staff if you have any ques-
tions, and please let us know if you plan to participate.
Have a great summer and we’ll see you in the fall! J
Bob Risser, PE
PCI President and CEO
2023/24 Leadership PCI class organization, has spent more than 40 years as a management
consultant, executive manager, engineer, and educator in the
kicks off with 21 members utility industry. Collaboration Unlimited helps companies and
groups prepare for continued success in an environment of
The 2023/24 Leadership PCI class commenced in February at the 2023 PCI Convention at the Precast Show in Columbus, Ohio.
Industry Calendar
Event details are subject to change.
BEI-2023 “Sustainability in Bridge Engineering”
July 17–20, 2023
Roma Eventi-Fontana di Fontana Trevi, Rome, Italy
CACRCS 2022
September 12–15, 2023
Parma, Italy
Greenbuild
September 26–29, 2023
Walter E. Washington Convention Center, Washington, D.C.
PTI 2023 Committee Days
October 3–6, 2023
Cancun, Mexico
ACI Concrete Convention
October 29–
Boston Convention Center and Westin Boston Waterfront,
November 2, 2023
Boston, Mass.
AASHTO Annual Meeting
November 12–16, 2023
J. W. Marriott Indianapolis, Indianapolis, Ind.
ASBI 35th Annual Convention
November 5–8, 2023
Westin La Paloma Resort and Spa, Tucson, Ariz.
World of Concrete
January 22–25, 2024
Las Vegas Convention Center, Las Vegas, Nev.
ACI Concrete Convention
March 24–28, 2024
Hyatt Regency New Orleans, New Orleans, La.
PTI Convention
April 14–17, 2024
Indianapolis, Ind.
ACI/RILEM International Conference on Cementitious Materials
and Alternative Binders for Sustainable Concrete June 23–26, 2004
Toulouse, France
fib International Conference on Concrete Sustainability
September 11–13, 2024
Guimarães, Portugal
Photo courtesy of
USC/Gus Ruelas.
PCI-funded research
on insulated wall panels
shows their strength
Andrew Osborn
P
recast concrete insulated wall panels are rapidly
gaining popularity because of their light weight,
thermal efficiency, and economy. Typical precast
concrete insulated wall panels consist of two layers of con-
crete separated by a layer of insulation and steel or fiber-re-
inforced-polymer connectors, which provide some level of
composite behavior. Designs using such walls are popular
for warehouse structures, cold storage, data centers, and
other buildings that require large open spaces. In addition,
they are gaining popularity in the nonwarehouse commercial
market and can integrate architectural features. Architects
like the versatility and thermal efficiency of precast concrete
insulated wall panels.
There are multiple shear connectors available on the market. SNC = section modulus for the noncomposite section
Their geometries and materials vary considerably. Further,
some connectors rely on the bond of the insulation to carry SFC = section modulus for the fully composite section
some horizontal shear, but others do not. In addition, connec-
tors can be composed of various fiber-reinforced composites ⎛ I − I NC ⎞
or unfilled polymer or they can be made of steel. The variety I PC = 100 ⎜ PC ⎟ (2)
of connectors has made it challenging in the past to establish a ⎝ I FC − I NC ⎠
uniform design and analysis process. where
Several contemporary methods for predicting behavior of wall IPC = moment of inertia of a partially composite wythe
panels rely on shear load versus shear displacement data for
precast concrete insulated wall panel wythe connectors.15,17–19 INC = moment of inertia of a noncomposite wythe
In the elastic range, which is where many precast concrete
insulated wall panels are designed, the initial elastic stiffness IFC = moment of inertia of a fully composite wythe
K0.5 is used for predicting behavior.20 Researchers usually use
double shear tests to estimate this value from the load deflec- This percentage is provided through design tables or submittal
tion plot (Fig. 2), and the stiffness K0.5 is reported in kip/in. documents to the engineer. Using this percentage, the engi-
This value and others from the load-versus-deflection plot are neer would back calculate the section properties of interest
used in various analytical methods for predicting full-scale (Ig or S) for use in designing the panel as if it were any solid
panel behavior, including the iterative sandwich beam theory panel for flexural slender wall design (Eq. [3] and [4]).
(ISBT) that was used in the research reported in this paper.
M
Over the past 20 years, precast concrete insulated wall panel σt = (3)
design has most often been accomplished with a percent S PC
composite approach that estimates the degree of compos- where
ite action.20 This approach is demonstrated by comparing
partially composite stress σPC in the diagram labeled “Actual σt = maximum tensile stress at the outer fiber
behavior” with σPC in the diagram labeled “Percent compos-
ite” in Fig. 1, where the approach matches the stresses (or M = applied moment
deflections) of an advanced analysis or experimental results
at cracking. Although this approach has not been codified
5wL4
or applied uniformly across the industry, wythe connector σ midspan = (4)
suppliers have typically used proprietary methods based on 384Ec I PC
testing or finite element analysis to estimate the apparent where
composite action.5,15,19,22 This approach has also been used to
some extent in previous research.23,24 δmidspan = midspan deflection
w = applied uniform load The horizontal shear design is a completely separate limit state
that is not covered in the approach here; it is typically handled
L = span length by shear flow or similar methods.8,9,27 For a given panel, it is ex-
pected that the percentages of composite action will be different
Ec = modulus of elasticity for cracking and deflection because the mechanics are different
There was no mathematical way to balance these mostly The ISBT method was used to simulate 1 million panels
qualitative objectives across the various models considered in R 4.2.038 using random combinations of the variables in
during development. Fortunately, among the various statis- Table 1. Variable ranges were selected to represent most
tical models considered by the authors, the simplified linear experimental precast concrete insulated wall panel configu-
approximation presented in this paper was the most accurate rations discussed in the “Experimental Results” section, as
and easiest to implement. well as the vast majority of the precast concrete insulated
wall panel configurations used in practice. Each variable was
The development of these simplified approximations re- simulated on a discrete scale, with the jump between possible
quires many observations of percent composite estimates values indicated by the increment column in Table 1. These
(for both cracking and deflection) across a wide range of simulated observations were used to train (that is, fit) simple
different precast concrete insulated wall panel configurations. statistical models to predict percent composite strength. These
Unfortunately, the small number of actual experiments on simulated observations are referred to hereafter as training
precast concrete insulated wall panels makes it impossible data. Attempts to expand the simulated ranges of these vari-
to rely solely on observed measurements when developing a ables began to overwhelm the training data with impractical
simplified approximation. Further, experimental precast con- variable combinations, which compromised efforts to find a
crete insulated wall panels tend to be smaller than the walls simplified approach that would work for typical panels. Given
used in actual design, and the current lack of an engineering this constraint, wall configurations with variables outside the
design standard for precast concrete insulated wall panels specified ranges should not use the simplified linear approach
creates high variability in the strength measurements taken by presented in this paper.
the different research groups performing the experiments. For
these reasons, the simplified linear approximation described While the variable ranges in Table 1 reasonably encapsulate
in this paper was developed using the random generation of typical precast concrete insulated wall panel configurations,
wall configurations and composite action output from the it is possible to generate combinations of parameters not
analytical ISBT approach, which has previously been shown practical for actual use. The model-fitting process assumes
to provide high-fidelity approximations of the cracking and that most of these unreasonable variable combinations wash
deflection behavior of precast concrete insulated wall panels.18 out when modeling averages; however, a few combinations
Note: 1 in. = 25.4 mm; 1 ft = 0.305 m; 1 kip = 4.448 kN; 1 ksi = 6.895 MPa.
*Variables were also used in the simplified linear approximation.
lead to numerical issues with the ISBT predictions (that a = horizontal asymptote
is, excessively high shear stiffness, grossly different wythe
thicknesses). For this reason, only observations where the x = some data input
ISBT predicted percent composite action values between 0%
and 100% for both cracking and deflection were retained. It is reasonable to assume in this context that a is less than
As a result, 0.5% of the 1 million observations in the master or equal to 100. To preserve the simplicity of the simplified
dataset were removed. Of the remaining observations, 80% modeling approach, the choice was made to set a equal to
were used to train the simplified linear approximation, while 100 and use a two-stage modeling approach to adjust the
the other 20% of remaining observations were used for slope and curvature of the asymptotic regression line for
model testing. different precast concrete insulated wall panel configura-
tions. Equations (6) and (7) present this two-stage modeling
The trends observed in Fig. 4 are the motivation for the final approach for estimating percent composite action (β̂I for
modeling approach, which shows the strong, asymptotic re- cracking or β̂Sm for deflection).
lationship that connector stiffness k and wall length (or span)
l share with percent composite strength for both cracking ⎛ ck⎞ ⎛cl⎞ ⎛ 1 ⎞ ⎛ 1 ⎞
and deflection on the training dataset. The number of obser- b̂w = b0 + b1 ⎜ 1 ⎟ + b2 ⎜ 2 ⎟ + b3 ⎜ ⎟ + b4 ⎜ ⎟ (6)
vations in the training dataset is large enough that it would ⎝ 100 ⎠ ⎝ 100 ⎠ ⎝ c2 wa ⎠ ⎝ c3 Ec ⎠
be impossible to differentiate between individual points in a where
traditional scatterplot. For this reason, observations are binned
into a hexagonal lattice, where colors represent the number b̂w = estimated asymptotic regression coefficient
of observations that fall within each cell of the lattice.39,40 The
orange line represents a locally smoothed trend line, which is b0 = regression coefficient
fit as a single variable generalized additive model, which fits a
smoothed trend line through data without assuming a specific b1 = regression coefficient
model form.41 The smoothed trend line mathematically bal-
ances the degree of smoothness in such a way that the general c1 = constant for unit conversion (the constant equals
trend is captured without trying to represent random variation 1 when the imperial units indicated in Table 1 are
in the data. used)
The shape of the trend lines in Fig. 4 inspires the use of an b2 = regression coefficient
asymptotic regression model with the mathematical form in
Eq. [5]). c2 = constant for unit conversion (the constant equals
1 when the imperial units indicated in Table 1 are
used)
a (5)
ŷ = a −
x +1 b3 = regression coefficient
where
wa = average wythe layer thickness
ŷ = the predicted output (in this case, percent
composite) b4 = regression coefficient
c3 = constant for unit conversion (the constant equals ed coefficients for the stage one model for cracking cr are
1 when the imperial units indicated in Table 1 are -0.25, -0.01, -0.01, 1.23, 936, respectively. For deflection,
used) the estimated coefficients are -0.06, -0.003, -0.004, 0.32,
340, respectively. The use of three digits is required to ensure
⎛ ⎞ necessary precision due to the large numbers computed in the
⎜ ⎟ final calculation.
⎜ 1 ⎟
β̂1 or β̂ S = 100 ⎜ 1− 2 ⎟
(7)
⎜ b̂ ⎛ c1k ⎞ ⎛ c2 l ⎞ + 1⎟ Example application
m
w⎜
⎜⎝ ⎝ 100 ⎟⎠ ⎜⎝ 100 ⎟⎠ ⎟⎠
To illustrate the use of the function, a realistic full-scale
The coefficients b0, b1, b2, b3, and b4 are fit using ordinary panel from Salmon et al.42 was used. This panel has a span of
least squares regression on the training dataset. The estimat- 29 ft (8.8 m) and width of 8 ft (2.4 m) with a configuration
) ( )( ) ⎥ + ( −0.01) ⎢ ( )100
⎡ 1 73 ⎤ ⎡ 1 ( 348) ⎤ ⎡ approximation
1 ⎤ ⎡on the1 test ⎤set that was not part of the training
( ) ( ) (
b̂w cr = −0.25 + −0.01 ⎢ ⎥ + 1.23 ⎢ data. This
(1)( 2.75)
⎥ +test
936set
⎢ contained ⎥ = 0.362
(1)( 4524) nearly 200,000 observations ran-
⎢⎣ 100 ⎥⎦ ⎢⎣ ⎢⎣ ⎥⎦⎥⎦ ⎢⎣ ⎥
domly generated using the⎦ISBT approach. Like Fig. 4, Fig. 5
( −0.01) ⎢ ( )( )
⎡ 1 348 ⎤ ⎡ 1 ⎤ ⎡ 1 ⎤ bins the observations within a hexagonal lattice with colors
⎥ + 1.23 ⎢ ⎥ + 936 ⎢ ⎥ = 0.362
⎢⎣ 100 ⎥⎦ ( )(
⎢⎣ 1 2.75 ⎥⎦ ) ⎢⎣ 1 4524 ⎥⎦ ( )( ) representing the number of points in each bin. A smoothed
trend line generated with a generalized additive model shows
the average ratio of the true cracking and deflection strength
⎧ ⎫ against the estimated values obtained using the percent com-
⎪ ⎪ posite predictions calculated using the linear approach. The
⎪⎪ 1 ⎪⎪ accuracy ratios are made by using actual estimates of crack-
β̂ S = 100 ⎨1− ⎬ = 76%
) ( )( ) ( )( )
2
m
⎪ ⎡ 1 73 ⎤ ⎡ 1 348 ⎤ ⎪ ing and deflection as calculated from the percent composite
⎪ (
0.362 ⎢ ⎥⎢ ⎥ + 1⎪ predictions. This allows the accuracy ratios to account for the
⎪⎩ ⎢⎣ 100 ⎥⎦ ⎢⎣ 100 ⎥⎦ ⎪⎭ difference between fully composite and noncomposite actions
and avoids the inflated variance in accuracy ratios that comes
) ( )( ) ( ) ( )( )
⎡ 1 73 ⎤ ⎡ 1 348 ⎤ ⎡ comparing
when 1 ⎤ ⎡
near-zero ⎤ composite predictions.
1 percent
( ) (
b̂w def = −0.06 + −0.003 ⎢ ) ( ⎥ + −0.004 ⎢ ⎥ + 0.32 ⎢ ⎥ + 340 ⎢
( )( )
⎥ = 12
( )( )
⎢⎣ 100 ⎥⎦ ⎢⎣ 100 ⎥⎦ ⎢⎣ 1 2.75
Figure 5 shows ⎢⎣ 1 4524
⎥⎦ agreement ⎥⎦
between the linear approximation
Considerations for machine learning While the results on the ISBT test dataset show the effec-
tiveness of the simplified linear approximation, it was still
The implementation of more-complex machine learning ap- necessary to determine the effectiveness of the approximation
proaches, namely regression trees43 and random forests were on real observations. For this evaluation, the authors used a
dataset of 69 precast concrete insulated wall panel experi- Table 2. Panels in the experimental dataset
ments conducted by multiple researchers (Table 2).11,16,27,42,46–53
Load
Reference Number of panels
configuration
Of these 69 experiments, 25 used two-point loading rather
than the uniform loading that the simplified approximation 46 6 Four-point loads
was designed to estimate. Additionally, the protocol for
construction and testing experimental panels varied from 47 6 Four-point loads
researcher to researcher, which makes it difficult to separate 27 5 Uniform
variability in estimation error due to modeling error from
error due to differences in wall panel testing or construction. 48 9 Uniform
For this reason, accuracy results were grouped according to 16 5 Uniform
the paper publishing the experimental unit.54 Results are only
shown for papers using uniform loading with at least five 49 5 Two-point loads
experimental units. 50 4 Two-point loads
The second category is iterative methods, which have more deflection. At worst, the simplified linear approach slightly
flexibility and can capture different geometries and any underestimates experimental cracking and deflection but
connector spacing; these methods are known as the simpli- performs no worse in prediction than the lower-performing
fied model,15 mechanics-based model,17 and the ISBT.18 The industry standard methods. Notably, the simplified approach’s
simplified method is an iterative method that adds the force deflection predictions did not perform well when compared
couple created by the connectors to the noncomposite resis- to data from Gombeda et al.,27 which investigated panels with
tance after multiple iterative loops to enforce equilibrium connectors concentrated at the ends or in a hybrid configu-
through calculation of the slip profile. A further simplifica- ration, as opposed to a uniform configuration. The method
tion in that method is the consideration of only the rotational presented herein should only be applied to situations with
slip component, neglecting the axial slip component of the uniformly spaced connectors.
aforementioned methods. The ISBT, as described previously,
treats each wythe as a continuous beam element, whereas It should be noted that most of the previously established
the mechanics-based model takes similar mechanics but methods tend to perform poorly in estimating the true values
discretizes each wythe into finite length sections. A similar of percent composite action for at least one of the experimen-
iterative scheme is then implemented to enforce force equi- tal data sets cited in this paper. Table 3 summarizes the bias
librium, and curvatures are integrated to obtain deformations and COV for the uniform loading results (Fig. 6 and 7) as
following convergence. The third category is the matrix well as the experimental results subject to two-point loading.
analysis method, termed the beam-spring model.5 Each of The results in this table show that cracking loads tend to, on
these three methods uses reference-reported geometry, mate- average, be slightly underestimated for experimental panels
rial properties, and double shear elastic stiffness to make its subject to uniform loading and underestimated for experi-
predictions.56 mental panels subject to two-point loading. The bias in the
simplified linear approximation is reduced in both uniform
These results show that, at best, the simplified linear approx- and two-point loading scenarios when considering deflection,
imation presented in this paper does as well or better than the though all methods tend to have a higher COV for deflection
more complicated industry-standard methods for cracking and predictions. Table 3 results reveal that the new approach
Table 3. Summary of the biases and coefficients of variation for the experimental data
Holmberg
Al-Rubaye Allen Al-Rubaye Jensen et Gombeda
New and Plem
et al. (2021) (1969) et al. (2019) al. (2020) et al. (2017)
(1965)
Two-point Bias 1.33 (1.13) 1.11 (1.22) 0.91 (1.28) 0.91 (1.28) 1.11 (1.04) 1.15 (1.26) 0.97 (1.16)
loads COV 0.27 (0.47) 0.21 (0.44) 0.27 (0.48) 0.27 (0.48) 0.21 (0.32) 0.16 (0.42) 0.37 (0.49)
Uniform Bias 0.88 (0.90) 0.97 (1.06) 1.06 (1.14) 1.06 (1.14) 0.98 (1.07) 0.97 (1.03) 0.97 (1.09)
loads COV 0.21 (0.33) 0.16 (0.23) 0.32 (0.3) 0.32 (0.3) 0.16 (0.2) 0.18 (0.19) 0.2 (0.36)
Note: Values outside of parentheses are for cracking composite. Values in parentheses are for deflection composite. COV = coefficient of variation.
tends to have a lower COV than the methods from Allen55 and as accurate as any other method considered in this paper in
Holmberg and Plem12 but a higher COV than the ISBT18 and estimating percent composite action.
mechanics-based model.17 The best methods for predicting
percent composite action are, as expected, better than the Conclusion
simplified approach. However, the simplified approach is
reasonably accurate in predicting percent composite action This paper has outlined a new simplified method for predicting
for the experimental data, and, in many cases, it has a smaller percent composite and thus the elastic moment of inertia and
COV than existing approaches in the literature. Further, when the elastic section modulus for precast concrete insulated wall
the assumptions behind the adapted ISBT approach are met panels. One million simulations from the popular ISBT method
(namely, uniform placement of the connectors), as is the case were used to statistically develop this simplified approach.
with the data from Cox et al.,47 the simplified approach is Furthermore, the new method was compared with the limited
This paper has not explored applications of the simplified 10. Pfeifer, D. W., and J. A. Hanson. 1965. “Precast Concrete
linear approximation for complex panels (for example, panels Wall Panels: Flexural Stiffness of Sandwich Panels.” ACI
with openings and nonuniform connectors). A rigorous ex- symposium publication vol. 11: 67–86. https://doi.org
ploration of how the simplified method might be adapted for /10.14359/16692.
application to such panels is a topic for future work.
11. Huang, J., Q. Jiang, X. Chong, X. Ye, and D. Wang. 2020.
Acknowledgments “Experimental Study on Precast Concrete Sandwich
Panel with Cross-Shaped GFRP Connectors.” Magazine
This research was funded in part by a Peak Summer Research of Concrete Research 72 (3): 1–49. https://doi.org
Fellowship provided by Utah State University’s Office of /10.1680/jmacr.18.00258.
Research.
18. Al-Rubaye, S., T. Sorensen T, and M. Maguire. 2021. 28. Granholm, H. 1949. Om Sammansatta Balkar Och Pelare
“Iterative and Simplified Sandwich Beam Theory for Med Särskild Hänsyn till Spikade Träkonstruktioner [On
Partially Composite Concrete Sandwich Wall Panels.” Composite Beams and Columns with Particular Regard
Journal of Structural Engineering 147 (10): 04021143. to Nailed Timber Structures]. Gothenborg, Switzerland:
https://doi.org/10.1061/(ASCE)ST.1943-541X.0003116. Gumpert.
19. Olsen, J., S. Al-Rubaye, T. Sorensen, and M. Maguire. 29. Teixeira, N., D. G. Tomlinson, and A. Fam. 2016.
2017. “Developing a General Methodology for “Precast Concrete Sandwich Wall Panels with Bolted
Evaluating Composite Action in Insulation Wall Panels.” Angle Connections Tested in Flexure under Simulated
Master of science thesis, University of Utah. https:/ Wind Pressure and Suction.” PCI Journal 61 (4): 65–83.
/digitalcommons.usu.edu/cgi/viewcontent.cgi?article= https://doi.org/10.15554/pcij61.4-02.
7769&context=etd.
30. Olsen, J., and M. Maguire. 2016. “Shear Testing of
20. ICC (International Code Council) Evaluation Service. Precast Concrete Sandwich Wall Panel Composite Shear
2010. Acceptance Criteria for Semi-continuous Fiber- Connectors.” In Proceedings: PCI Convention and
Reinforced Grid Connectors Used in Combination National Bridge Conference. Chicago, IL: PCI.
with Rigid Insulation in Concrete Sandwich Panel
Construction. ICC-ES Acceptance Criteria AC422. 31. ACI (American Concrete Institute). 2019. Building Code
Country Club, IL: ICC. Requirements for Structural Concrete (ACI 318-19) and
Commentary (ACI 318R-19). ACI 318-19. Farmington
21. Maguire, M., and F. F. Pozo-Lora. 2020. “Partially Hills, MI: ACI.
Composite Concrete Sandwich Wall Panels: What Is
‘Percent Composite’”? Concrete International 42 (10): 32. Maguire, M., M. Chang, W. N. Collins, and Y. Sun. 2017.
36. AASHTO (American Association of State Highway 48. Naito C. J., J. M. Hoemann, J. S. Shull, A. Saucier, H.
and Transportation Officials). 2020. AASHTO LRFD A. Salim, B. Bewick, M. I .Hammons. 2011. Precast/
Bridge Design Specifications. 9th ed. Washington, DC: Prestressed Concrete Experiments Performance on Non-
AASHTO. Load Bearing Sandwich Wall Panels. Tyndall Air Force
Base, FL: Air Force Research Laboratory Materials and
37. Drysdale, R. G., A. A. Hamid, and L. R. Baker. 1994. Manufacturing Directorate. https://apps.dtic.mil/sti/pdfs
Masonry Structures: Behavior and Design. New York, /ADA545204.pdf.
NY: Prentice Hall.
49. Luebke, J. 2021. “Out-of-Plane Behavior of Concrete
38. Core, R. 2022. “R: A Language and Environment for Insulated Wall Panels with 2-inch, 8-inch, and 10-inch
Statistical Computing.” Vienna, Austria: R Foundation for Insulation.” Master’s thesis, University of Nebraska–
Statistical Computing. https://www.r-project.org Lincoln. https://digitalcommons.unl.edu/archengdiss/67.
/foundation/.
50. Jiang, H., Z. Guo, and J. Liu. 2018. “Composite Behavior
39. Wickham, H. 2016. “Data Analysis.” In Ggplot2, of Sandwich Panels with W-Shaped SGFRP Connectors.”
189–201. New York, NY: Springer. KSCE Journal of Civil Engineering 22 (5): 1889–1899.
https://doi.org/10.1007/s12205-017-2050-3.
40. Neuwirth, E. 2014. “RColorBrewer: ColorBrewer
Palettes.” R Package version 1.1-2. 51. Maguire, M., and S. Al-Rubaye. 2022. Tilt-Up Partially
Composite Insulated Wall Panels. Lincoln, NE:
41. Wood, S. N. 2011. “Fast Stable Restricted Maximum University of Nebraska. https://doi.org/10.32873/unl
Likelihood and Marginal Likelihood Estimation of .dc.oth.014.
Semiparametric Generalized Linear Models.” Journal
of the Royal Statistical Society Series B: Statistical 52. Huang, J. 2019. “Structural Performance of Precast
Methodology 73 (1): 3–36. https://doi.org/10.1111 Reinforced Geopolymer Concrete Sandwich Panels
/J.1467-9868.2010.00749.X. Enabled by FRP Connectors.” PhD diss, Hong Kong
Polytechnic University. https://theses.lib.polyu.edu.hk
42. Salmon, D. C., A. Einea, M. K. Tadros, and T. D. Culp. /handle/200/10015.
1997. “Full Scale Testing of Precast Concrete Sandwich
Panels.” ACI Structural Journal 94 (4): 354–362. https:// 53. Zhi, Q. 2017. “Experimental Evaluation of Precast
doi.org/10.14359/486. Concrete Sandwich Wall Panels with Steel-Glass Fiber-
Reinforced Polymer Shear Connectors.” Advances in
43. Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Structural Engineering 20 (10): 1476–1492. https://doi
Stone. 1984. Classification and Regression Trees. New .org/10.1177/1369433216683198.
York, NY: Routledge. https://doi.org/10
.1201/9781315139470. 54. Wickham, H., M. Averick, J. Bryan, et al. 2019.
“Welcome to the Tidyverse.” Journal of Open Source
c3 = constant for unit conversion σt = maximum tensile stress at the outer fiber
fr = modulus of rupture
h = height (width)
k = connector stiffness
ℓ = wall length
L = span length
Keywords
P
refabricated bridge elements and systems are an
essential part of accelerated bridge construction
because they offer significant time and cost savings,
improved safety, and convenience for public travel. Precast,
prestressed concrete deck bulb-tee girders are one such
system because the top flange of the precast concrete girder
section can act as the deck of the bridge, which eliminates
the need for time-consuming forming, casting, and curing
of a cast-in-place deck. These girders are manufactured
■ The research presented in this paper is the ex- in precast concrete plants under closely controlled and
perimental testing portion of a larger study per- monitored conditions, transported to the construction site,
formed under National Highway Research Program and erected so that flanges of adjacent units join together.
project 18-18. To enable wider adoption of deck The wide upper flange of deck bulb-tee girders results in
bulb-tee girder systems, design guidelines and a very efficient section for prestressing purposes because
standard details must be developed for joints used it provides a large cross-sectional area that can resist the
in these systems to ensure that these joints perform prestressing force of many strands. The casting of the deck
adequately and are useful for accelerated bridge as part of the girder also allows for a variation of the deck
construction. thickness, resulting in a more efficient transverse design of
the deck.
■ This paper, part 2, presents full-scale experimental
testing performed as a follow-up to the analytical Despite the many benefits of deck bulb-tee girder bridge
investigation that appeared in part 1. Longitudinal systems, their use has been limited and most existing bridges
joints of a deck bulb-tee girder system grouted with of this type have relatively short spans with low traffic
ultra-high-performance concrete were tested under volumes. Concerns regarding the long-term performance
thermal and fatigue loading. The longitudinal joint of longitudinal and transverse connections between indi-
system was also subjected to camber differential vidual girders partially explain why these systems have not
between adjacent girders. In addition, a continu- been widely used. In addition, variable camber profiles of
ity joint composed of conventional concrete and the girders can cause bridge deck profile problems as the
ultra-high-performance concrete was tested for posi- cambers in adjacent girders do not align, making it difficult
tive and negative moments at the pier. to make the connection between flanges. Forces are induced
6" 6"
2"
6 5
83
4"
4
83
4"
3
83
4"
2
83
4"
1
Vibrating
Vibrating wirewire
gagegauge
Thermistor
Thermistor
Figure 1. Girder details for longitudinal joint tests. Note: no. 5 = 16M; 1” = 1 in. = 25.4 mm.
Figure 3 shows the girder assembly on the laboratory floor. To generate a thermal gradient in the girders, the heat lamps
To apply the temperature load, a temperature gradient along and circulation fans were turned on initially. The heaters
the girder depth was to be developed. For this purpose, a were ramped up gradually to heat the top of the surface. The
closed environmental chamber or “insulated box” was built on heating equipment was placed to heat the entire top of the
top of the girder deck. This methodology was adapted from girder system as uniformly as possible. The AASHTO LRFD
a paper by Shi et al.16 A wooden frame was constructed on specifications stipulate a temperature gradient T1 of 38°F
top of the girders and then covered with 2 in. thick (51 mm) (21°C) to 54°F (30°C) between the top of the girder and a
expanded polystyrene foam sheets. The top view of the com- position 16 in. (406 mm) deep and a gradient T2 of 9°F (5°C)
pleted insulated box in shown in Fig. 3. Eighteen 250-watt to 14°F (8°C) between the depths of 4 in. (100 mm) and 16 in.
heat lamps, eight high-capacity heaters, and six circulation (Fig. 3). In this artificially simulated heating environment, it
fans were installed inside the insulated box. The heat lamps, was not possible to control the temperatures in both locations
heaters, and fans were uniformly distributed inside the insu- simultaneously. The temperature gradient within the girder
lated box and were regulated to generate thermal gradient in was critical to the performance of the joints, so the gradient
the girders. between 4 and 16 in. (gradient T2 for zone 1) was the key
parameter that was monitored during a typical heat cycle. This
Thermal loading of open joints The thermal properties temperature gradient value marked the peak of the heating,
of concrete vary with the mixture proportions. Therefore, an and once this gradient was around 14°F for all the girders, the
initial establishment of thermal properties of the girders was heat was turned off.
Figure 2. Instrument location and labels for thermal and fatigue testing of longitudinal joints. Note: LVDT = linear variable dis-
placement transducer; VW = vibrating wire.
Figure 3. Test setup for thermal and fatigue load test for longitudinal joints. Note: AASHTO = American Association of State
Highway and Transportation Officials, T1 = thermal gradient between girder surface and 16 in. depth; T2 = thermal gradient be-
tween 4 and 16 in. depths. 1 in. = 25.4 mm; °F = 1.8(°C).
Figure 4 shows the camber of the girders alongside the throughout the girder depth. It took around 8 hours to reach
gradient at a depth of 4 in. (100 mm) during a typical heating the heating peak (defined by a 14°F [8°C] gradient at a 4 in.
cycle. As expected, all the girders cambered up with an [100 mm] depth); at that point, the surface temperature was
increase in the girder temperature gradient and cambered around 100°F (37.8°C) and the temperature at a depth of
down as the girders cooled. Girders 1 and 2 had almost the 20 in. (510 mm) was almost same as at the beginning of the
same maximum camber of approximately 0.13 in. (3.3 mm), heating cycle. After 4 hours of cooling, the surface tempera-
whereas the maximum camber for girder 3 was about 0.2 in. ture dropped faster than the temperature inside the girder
(5 mm). This variation in the cambers of the girders was ex- flange. Also, the girder web became slightly warmer than it
pected given the concrete variability. Figure 4 shows the strain was at the peak heating.
across the tops and bottoms of the joints alongside the average
girder temperature gradient before grouting. Both the tops Figure 4 compares the girder temperature gradient with the
and bottoms of the joints showed a compressive strain as the gradient band recommended in the AASHTO LRFD speci-
girder temperature gradient, indicating that the girder flanges fications. The gradient at 4 in. (100 mm) depth T2 was in the
were expanding and the joint was closing. The tops of the range of 15°F (8°C) to 16.5°F (9°C) for the three girders. This
joint (gauges J12-3 and J23-3) developed higher compressive temperature gradient is above the recommended value of 14°F
strain than the bottoms of the joint (gauges J12-B and J23-B). (8°C) for T2 in the most extreme AASHTO zone 1 condition.
The temperature gradient at the surface of the girders T1 was
Figure 4 shows the temperature profiles through the depth between 33°F (18°C) and 40°F (22°C), which was slightly
of girder 2 at various times during a typical heat cycle. At below the range of 38°F (21°C) to 54°F (30°C) recommended
the start of heating, the temperature was almost uniform in the AASHTO LRFD specifications. However, as discussed,
Temperature Gradient, °F
Temperature Gradient, °F
0.2 -500
10 10
Microstrain
Camber, in.
0.15 -1000
0
0.1 5
-1500
-10
0.05
-2000 0
0 -20 0 2 4 6 8
0 10 20 30
Time, hours
Time, hours
Girder camber on primary axis and gradient Strain across joints on primary axis and
at 4 in. depth on secondary axis average girder gradient on secondary axis
0 10 20 30 40 50 60
60 70 80 90 100 110
0
0
4
4
8
12
8
16
20 AASHTO Zone 1
12 AASHTO Zone 4
24
8 am (Start of heating) Girder 1
28
16 Girder 2
32 4 pm (P eak heating)
Girder 3
36 8 pm (After 4 hours of cool ing) 20
40
Temperature Gradient, °F
Temperature, °F
Temperature variation along the depth of girder 2 Peak temperature gradient in girders compared
at various times during a typical heat cycle with AASHTO gradient for a typical heat cycle
Figure 4. Thermal testing of the girder system with open joints before grouting the joints with ultra-high-performance concrete.
Note: AASHTO = American Association of State Highway and Transportation Officials. 1 in. = 25.4 mm; °F = 1.8(°C).
under laboratory conditions, it was not possible to control the mixing; the manufacturer indicated that neither the age
both T1 and T2 simultaneously and because T2 was critical for of the material nor the additional water put the UHPC out of
the performance of the joints, it was selected as a control for specification.
the heating cycle.
To make the UHPC flow along the joints, the joints were
UHPC placement in joints After testing on open joints covered with formwork both at the bottom and the top. Two
was completed, formwork was placed at the top and bottom openings were made on each joint, and “chimneys” were
of the joints, and the joints were grouted with UHPC. A installed. The UHPC was placed into the joint through these
proprietary UHPC mixture was used for this project. Table chimneys. To simulate field conditions, the girders were
1 provides some of the properties of the material used. This heated the day before UHPC placement and again in the
commercially available product has been used in numerous morning of placement to keep the girders as warm as possible
projects throughout the United States and was used by the during the placement process. When the crew was ready to
authors in several previous projects. It should also be noted grout the joints, the ceiling of the insulation box was removed
that due to delays caused by the COVID-19 shutdown, the to allow easy access of the joints to place the UHPC. Just
UHPC sat in the laboratory for approximately 1 year before before the UHPC was placed into the joints, the top forms
the experiment was conducted. The UHPC seemed to have were removed and a fine water mist was applied to wet each
a slightly more doughy consistency during placement than joint; then the top forms were screwed back into place. The
the authors were accustomed to. Table 2 presents the UHPC UHPC was transferred from the mixer in buckets into the
mixture proportions used for this project and the fresh UHPC chimneys, which were placed approximately at third points
properties. Slightly higher amounts of water were used on the joints (Fig. 5). This arrangement allowed the UHPC to
because the premixture was approximately 1 year old. A rep- flow in each direction. A hydraulic head was maintained in the
resentative of the manufacturer was present and supervised chimneys to ensure that the entire joint was filled.
7 to 10 in. diameter without Thermal loading after grouting Starting on the next day
Flow visible sign of fiber segrega- after the UHPC placement, thermal loading as described in the
tion previous section was applied to the girder system. Fourteen
heating cycles were completed. The girders were heated
Approximately 120 minutes/ until the temperature gradient between 4 and 16 in. (100 and
Working time/set time
15 to 18 hours 406 mm) below the girder surface reached approximately 14°F
>14 ksi† at 4 days‡ (8°C), and then they were allowed to cool until the gradient
Compressive strength* fell at least below 6°F (3°C). Ideally, the temperature gradient
>21 ksi at 28 days should have been reduced to 0°F (0°C) before the next cycle
Tensile strength§ >725 psi at 28 days but cooling the girders to this extent in the enclosed laboratory
setup with the heat box on top would have taken a very long
Modulus of elasticity >6500 ksi at 28 days time. Also, Miller et al.17 observed during field monitoring of
Long-term shrinkage <800 microstrain at 28 days
temperature profiles in box girders that girders hold some heat
overnight, so the authors decided to limit the lower tempera-
<250 coulomb (very low) at ture gradient to 6°F or less.
Chloride ion penetrability
56 days
Data were recorded from instruments during every thermal
Resistance to freezing and
>96% RDM at 300 cycles cycle. Results of various measurements during the first and
thawing
fifth heating cycles are discussed herein. The first cycle was
Note: Field results may differ depending on mixing/test methods, critical because the joints were gaining strength during this
equipment used, temperature, and site/curing conditions. 1 in. = time. During the fifth cycle, the joints had gained sufficient
25.4 mm; 1 lb/ft3 = 16.01 kg/m3; 1 psi = 6.895 kPa; 1 ksi = 6.895 MPa. strength and joint formwork was removed. No significant
* Compression tests are performed on 3 × 6 in. cylinders with ends differences in the measurements were observed between the
ground flush before testing. fifth and final thermal cycles.
† 14 ksi is the typical minimum compressive strength required before ap-
plication of design live load for most closure pour applications; consult
During the first cycle, the girder cambers were slightly lower
the engineer or project specifications to verify.
than the cambers recorded before grouting of the joints
(Fig. 4). This finding was expected because the joint form-
‡ Four days or less is typical when the ambient curing temperature is
work was still in place during this cycle and would have pro-
greater than 60°F (16°C). For colder temperatures, an accelerating
vided some restraint to the movement of the girders. During
admixture may be required to obtain 14 ksi in four days. For 14 ksi
the fifth cycle, when the formwork was removed, the girder
compressive strength in 12 to 36 hours, consider using a rapid-set
camber values were the same as the camber values before
product.
grouting of the joints. However, the fifth-cycle variation in the
§
This test measures the sustainable postcracking direct tension strength camber values of different girders was much lower than varia-
of a mixture with 2% (by volume) steel fibers. tion before grouting of the joints, indicating that the joints had
gained strength and the girder system was acting as a unit. For
After sufficient UHPC was placed into the joints, the insulat- both joints, the strains inside the joints were found to increase
ed box was closed, and girders were heated again to regain from 150 to 250 microstrains during first thermal cycle to 250
the lost temperature gradient. Three days later, the forms were to 350 microstrains during fifth thermal cycle; this difference
removed from the joint. Both joints turned out well. Figure 5 can be attributed to joint formwork being present during the
shows the top of the joint between girders 2 and 3 at the west first cycle and absent during the fifth. No obvious signs of
end. The sealer material on each side of the joint was used to reinforcing bar slippage or cracking were observed.
ensure that the UHPC did not leak out of the joint and gave
an approximately 0.125 in. (3.2 mm) height above the joint. Figure 6 compares the strain across joints 1–2 and 2–3 during
UHPC is often placed approximately 0.125 to 0.375 in. (3.2 the first and fifth thermal cycles. During the first cycle, both
to 9.5 mm) higher than the adjoining members in the field and joints developed a compressive strain in the range of 300
ground flush. to 400 microstrains upon heating. Upon cooling, the strains
reduced in magnitude but remained compressive. However,
During the UHPC placement, 3 × 6 in. (76 × 150 mm) during the fifth cycle, the joints developed a compressive
cylinders were cast for compression testing (Table 3). The strain in the range of 80 to 120 microstrains upon heating and
28-day (thermal) results were for cylinders stored within the then, upon cooling, the strain became tensile in the range of
insulated box built on the girders. Several thermal cycles 0 to 50 microstrains. These findings indicate that during the
were performed while these cylinders were inside the box. first cycle, when the UHPC joints were still gaining strength,
The purpose of these cylinders was to match the conditions they allowed the girder flanges to expand and therefore the
Laboratory temperature, °F 51 51
Note: UHPC = ultra-high-performance concrete. 1 in. =25.4 mm; 1 lb/yd3 = 0.59 kg/m3; °F = 1.8(°C) + 32.
joints developed higher compressive stresses (300 to 400 mi- The first 100,000 cycles were applied without thermal
crostrains). However, during the fifth cycle, when the joints loading to monitor any immediate effect of live load after
had gained strength, they resisted the movement of girder thermal cycles. After the initial 100,000 cycles were com-
flanges and therefore the compressive stresses were reduced pleted, thermal loading was also applied during the cyclic
(80 to 120 microstrains). live loading (Fig. 7). During each thermal cycle, 100,000
cycles of live load were applied. At the end of each thermal
Combined thermal and fatigue loading After the cycle, cyclic loading was stopped and a 70 kip (310 kN)
thermal cycles were completed, cyclic live loading was applied load was applied statically to monitor potential degrada-
to the girder system. Figure 7 illustrates the load setup without tion of connection integrity in terms of load distribution
the insulating box in place. A cyclic load between 0 and 70 kip or deflection of individual girders. These pauses in the
(0 and 310 kN) was applied at a frequency of 2 Hz. A total of cyclic live-loading process were necessary to read instru-
1 million live-load cycles were applied. mentation, which could not be read during the 2 Hz cyclic
“Chimneys” at third points used to pour Top view of the UHPC joint between girders
UHPC into the joints 2 and 3 after removing the formwork
Joint flooding and inspection After the combined Joint 2–3 between girders 2 and 3 had a few small leaks. The
thermal and fatigue loading was completed, the joints were leakage occurred both at the interface between the girder and
visually analyzed. No cracking was visible. However, micro- the UHPC and within the UHPC.
cracking can be difficult to visually identify; therefore, joints
were flooded with water to inspect for any leaks (Fig. 8). Several factors, either individually or in combination, could
Dams were constructed and sealed around the top surface of have caused the cracking. The interface could have dried out
the joints. These dams were then flooded with water. Flooding during delays in placing the UHPC. However, cracking was
revealed severe cracking in joint 1–2 between girders 1 and 2. found within the UHPC in addition to along the interface.
0
0 5 10 15 20 25
-100 0
Microstrains
Microstrains
0 5 10 15 20 25
-200
-300 -100
-400
Cycle 1 Cycle 5
-500 -200
Time, hours
Time, hours
Joint 1-2 after first and fifth loading cycles
0
0 5 10 15 20 25
-100 0
Microstrains
Microstrains
0 5 10 15 20 25
-200
-300 -100
-400 Cycle 1 Cycle 5
-500 -200
Time, hours Time, hours
Joint 2-3 after first and fifth loading cycles
Figure 6. Strain across the joints during thermal loading of the girder system after placement of ultra-high-performance con-
crete in the longitudinal joints.
Elevation
Elevation
view view
Schematics of fatigue loading setup without insulation box Static loading after every 100,000 cycles
Figure 7. Combined thermal and fatigue load setup and loading protocol. Note: 1 kip = 4.448 kN.
Shrinkage of the UHPC during curing could have resulted in mal testing could now be used. Girder 4 was placed in between
cracking along the interface and within the UHPC. Changes girders 1 and 3 to create a new setup with two longitudinal
in strain measurements during thermal loading may indicate joints. As noted earlier, girder 4 had six additional prestressing
that cracking occurred during the thermal cycle. The thermal strands to ensure that its camber differed from the camber of
cycle loading in this experiment was severe and may not be girders 1 and 3. The measured girder cambers at midspan for
typically experienced during UHPC placement. Lastly, the girders 1, 4, and 3 were 0.875, 1.375, and 1 in. (22.2, 34.9,
UHPC was more than 1 year old when placed. Flow of the and 25.4 mm), respectively. Thus, the differential cambers
UHPC did appear to slow down for the joint that experienced were 0.5 in. (12.7 mm) between girders 1 and 4 and 0.375 in.
more cracking. (9.5 mm) between girders 4 and 3.
Differential camber testing The instrumentation used to monitor the behavior of the joints
of longitudinal joints was same as the instrumentation used in the thermal and
fatigue tests, except that the vibrating-wire gauges were now
Deck bulb-tee girders may have variable camber profiles used to record strain only. Because the girder assembly was
if their material properties vary or if they are fabricated on changed for the differential camber testing, a new scheme of
different dates. When these girders are erected next to each instrumentation labels (Fig. 9) was used to be consistent with
other, the system may have a camber differential in the ad- the changed girder labels and to avoid confusion with the
jacent beams. Many contractors employ some type of level- previous scheme of instrumentation.
ing procedure to minimize the camber differential between
adjacent units. The girders are either pulled up or pushed Leveling of camber and UHPC placement The dif-
down and jacked in place while the joints are grouted and ferential camber between girders was removed by loading the
cured. Any procedure used to remove or reduce the camber middle girder (girder 4) to approximately 90 kip (400 kN).
differential introduces additional forces in the joints. Tests This loading resulted in a downward deflection of 0.5 in.
were performed to evaluate the effect of camber differential (12.7 mm) for girder 4. The hydraulic cylinder was locked off
between the girders. to hold the girder in position. The loading slightly decreased
from seepage of the hydraulic pressure, creep of the girder, and
Test setup and instrumentation After the thermal and settlement of the system. However, the downward deflection
fatigue load testing was completed, the girder assembly was remained at approximately 0.5 in. to match flange heights
cut along the joint lines to disconnect the three girders. Girder (Fig. 10). Formwork was then attached to the girders and the
2, the middle girder during thermal and fatigue tests, was now UHPC was placed. Table 4 presents the UHPC mixture pro-
removed from the laboratory floor. Girders 1 and 3 swapped portions and fresh properties, and Table 5 provides the com-
positions so that their exterior sides of flanges not used in ther- pressive strength of the UHPC.
Figure 9. Instrument location and labels for differential camber testing of longitudinal joints. Note: LVDT = linear variable dis-
placement transducer; VW = vibrating wire.
Load release and joint inspection Three days after before the UHPC was placed, rose by 0.2 in. (5.1 mm).
UHPC placement, the top forms were removed. Dams were The midspan of girders 1 and 3 rose approximately 0.12 in.
erected around the joints and waterproofed. On the fourth day (3.0 mm). Since the cambers were measured at the bottom,
after UHPC placement, the dams were filled with water for more this finding implies a slight bending of the flange, a slight ro-
than 30 minutes. No cracking was observed from leakage of the tation of the outside girders, or a combination of slight flange
joint. The load used to remove the differential camber was then bending and slight girder rotation.
released gradually in 15 kip (67 kN) increments. Upon removal
of the first load increment, leakage and cracking occurred in Fatigue loading After the load was fully released from the
joint 3–4 between girders 3 and 4. The cracking occurred within system, cyclic testing of 70 kip (310 kN) was performed. A
the UHPC as well as within the interface between the UHPC and total of 1 million cycles were applied to the system. At various
the girder flanges. The joint 4–1 between girders 4 and 1 also periods throughout the cyclic loading, the cyclic loading was
had minor cracking. Most of the cracking occurred on joint 3–4 stopped and a static 70 kip load was applied to investigate any
and mostly in the region where UHPC was placed at the end. deterioration of the joints due to live load. Several of the stat-
ic loadings also included flooding the joints. No evidence of
Upon removal of the camber leveling load, girder 4, which new cracks due to cyclic loading was found. The few existing
had been pushed down by 0.5 in. (12.7 mm) at the midspan cracks grew by less than 1 in. (25.4 mm).
Figure 10. Differential camber testing of longitudinal joints. Note: 1 in. = 25.4 mm.
Figure 11 plots the camber data from static load tests before ity joint detail tested. UHPC was only used in the upper portion
the start of the load cycles and after completion of 1 million of the diaphragm to save on material costs and deal with the
cycles of loading. The cambers at the start and end are nearly higher negative moment. The lower portion of the diaphragm
identical, indicating that the system did not degrade over the was conventional concrete. The continuity connection had a
loading cycles. Figure 11 plots the support reaction data from positive moment design capacity of 78 kip-ft (106 kN-m) and a
the beginning of the experiment and after 1 million cycles of negative moment design capacity of 1152 kip-ft (1562 kN-m).
loading. The reactions are also nearly identical, indicating that
the system did not degrade. Test setup and instrumentation To test the perfor-
mance of a continuity joint made of UHPC, two 20 ft (6 m)
Continuity joint testing long precast concrete economical fabrication girders were
constructed. The girders were placed end to end, 7.5 in.
The continuity joint design was based on the AASHTO LRFD (191 mm) apart. A 1 ft wide (0.3 m) diaphragm was construct-
specifications15 and results of analytical modeling presented in ed to make the girders continuous. The girders were embedded
the first part of this article series.14 Figure 12 shows the continu- approximately 2.25 in. (57.2 mm) into the diaphragm. The
Table 4. UHPC mixture constituents and fresh UHPC properties for differential camber testing of the
longitudinal joints
Laboratory temperature, °F 75 79
Note: UHPC = ultra-high-performance concrete. 1 in. =25.4 mm; 1 lb/yd = 0.59 kg/m ; °F = 1.8(°C) + 32.
3 3
G3 G4 G1 Load G3 G4 G1 Load
0 0 0 0
-0.04 10 -0.04 10
-0.08 20 -0.08 20
Camber, in.
30 30
Camber, in
Load, kip
Load, kip
-0.12 -0.12
40 40
-0.16 -0.16
50 50
-0.2 -0.2
60 60
-0.24 70 -0.24 70
Before fatigue load After 1 million cycles
-0.28 80 -0.28 80
5 7 9 11 13 15 2 7 12
Time, minutes Time, minutes
Girder camber and total static load
14 14
12 12
10 10
8 8
6 6
4 4
2 Before fatigue load 2 After 1 million cycles
0 0
5 7 9 11 13 15 2 4 6 8 10 12
Time, minutes Time, minutes
Support reactions
Figure 11. Results of static load tests at the beginning and end of the live load cycles for differential camber test. Note: 1 in. =
25.4 mm; 1 kip = 4.448 kN.
Figure 12. Continuity joint test setup details. Note: UHPC = ultra-high-performance concrete; VW = vibrating-wire. 1” = 1 in. =
25.4 mm; 1’ = 1 ft = 0.305 m.
Negative moment testing To test the joint for negative load of 98 kip (436 kN) (third-to-last load increment), the
moment, two frames were erected at 12 ft (3.7 m) on either strains in both bars at the middle of the joint (W3 and E3)
side of the diaphragm. The east frame was hinged to the abruptly increased and indicated yielding. The 98 kip load
top flange of the girder to prevent vertical movement. On corresponds to a total moment of 1357 kip-ft (1840 kN-m),
the west frame, a hydraulic cylinder was attached to load including the self-weight. This moment exceeded the cal-
the diaphragm. Figure 14 shows the schematics of the test. culated moment capacity of the joint by 18%. The strain in
The temporary support on the west end of the girder was the reinforcement W2 yielded at the next load increment,
then removed to create a cantilever. The self-weight of the but the remaining bars did not indicate yielding even on the
west girder resulted in a negative moment of approximately final load increment.
230 kip-ft (310 kN-m) at the joint. An additional 114 kip
(507 kN) load was applied through the hydraulic cylinder to Conclusion
create another 1311 kip-ft (1777 kN-m) negative moment at
the joint. Tensile cracks formed at the top of the continuity Based on the experimental results, the following observations
joint due to negative moment. Most of the cracking occurred and general conclusions were made:
either in the girder flanges or at the interface of the connec-
tion. The UHPC did not crack. • Although cracking in the system tested for combined
thermal and fatigue load was not fully discovered until
Figure 14 shows strains in the gauges attached to the the end of cyclic testing, the authors believe that the
top-flange reinforcement and embedded in the joint for cracking occurred early in the thermal cycles. The
the east and west girder reinforcement, respectively. At a thermal gradient applied to the system was slight-
56
4 46
36
0 26
-4 16
6
-8 -4
0 5 10 15 20
Time, minutes
Vibrating-wire gauge reading and positive moments at the joint
Figure 13. Positive moment testing of the continuity diaphragm. Note: 1’ = 1 ft = 0.305 m; 1 kip-ft = 1.356 kN-m.
ly higher than that specified in the AASHTO LRFD • Though cracking occurred during removal of the differen-
specifications and started immediately after UHPC tial camber leveling load, crack growth during 1 million
placement. Significant cracking occurred in the second load cycles was 1 in. (25.4 mm) or less. In addition, new
joint pour and minimal cracking occurred in the first cracks were not formed during the load cycles.
joint placement. This finding may be explained by
the UHPC gaining sufficient strength in the first joint • The use of UHPC in the longitudinal joints showed prom-
before thermal-generated strains developed but not ising results as it resulted in simpler joint detailing and
gaining enough strength in the second joint before the continued to transfer load even after cracking. However,
thermal-generated strains. The flow of UHPC during more research is required to mitigate thermal cracking of
placement in the second joint was slightly lower than the joints and subsequent leakage.
during placement the first joint, which could have
affected performance. Although cracking may have • The 7.5 in. (191 mm) continuity joint consisting of
existed in the joints, load transfer continued to occur UHPC in the top 6 in. (150 mm), and conventional con-
between girders under combined static and thermal crete in the remaining height of the diaphragm did not
load. show any distress under a positive moment expected from
continuity being formed at an age of 28 days.
• In the system tested for differential camber, cracking
was discovered right after the first loading increment • The top reinforcement in the continuity joint began to
was removed from the differential camber leveling. The yield at a negative moment more than the calculated
cracking occurred primarily in the east end of the joints. negative moment capacity. Additional negative moment
The UHPC was placed in the east ends last and seemed to caused more bars to yield and resulted in a capacity far
have slower flow. exceeding the calculated capacity.
E1 E3 E4 W1 W2 W3
E5 Moment W4 W5 Moment
12000 1600 12000 1600
*Gauge E2 Negative moment, kip-ft
Microstrain
6000 800 6000 800
0 0 0 0
0 30 60 90 120 0 50 100
Time, minutes Time, minutes
Top-flange reinforcement strain for east girder Top-flange reinforcement strain for west girder
Figure 14. Negative moment testing of the continuity diaphragm. Note: 1’ = 1 ft = 0.305 m; 1 kip-ft = 1.356 kN-m.
7. Miller, R. A., G. M. Hlavacs, T. Long, and A. Greuel. 17. Miller, R., B. Shahrooz, A. Haroon, E. Steinberg, W.
1999. “Full-Scale Testing of Shear Keys for Adjacent Hamid, A. Chlosta, and C. Slyh. Forthcoming. Proposed
Box Girder Bridges.” PCI Journal 44 (6): 80–90. https:// AASHTO Guidelines for Adjacent Precast Concrete Box
doi.org/10.15554/pcij.11011999.80.90. Beam Bridge Systems. NCHRP research report 1026.
Washington DC: Transportation Research Board.
8. Sharpe, G. P. 2007. “Reflective Cracking of Shear Keys
in Multi-Beam Bridges.” Master’s thesis, Texas A&M Notation
University. http://oaktrust.library.tamu.edu/bitstream
/handle/1969.1/ETD-TAMU-1912/SHARPE-THESIS T1 = thermal gradient between girder surface and 16 in.
.pdf?sequence=1. depth
9. Grace, N. F., E. A. Jensen, and M. R. Bebawy. 2012. T2 = thermal gradient between 4 in. and 16 in. depths
“Transverse Post-Tensioning Arrangement for Side-by-
Side Box-Beam Bridges.” PCI Journal 57 (2): 48–63.
https://doi.org/10.15554/PCIJ.03012012.48.63.
Reader comments
Kent A. Harries, Bahram M. Shahrooz, Payne D. Ball, TianQiao Liu, Venkata S.S.P.
Sathiraju, Abdullah Alabdulkarim, Richard A. Miller, and Reid W. Castrodale
I
n the United States, seven-wire prestressing strands
conforming to ASTM A4161 and AASHTO M 2032 are
used to pretension concrete bridge components. Typically,
Grade 270 (1860 MPa) low-relaxation strand is used for
bonded pretensioning (referred to in this context as prestress-
ing strand or strand). For many years, the standard in the
bridge industry was 0.5 in. (13 mm) diameter strand. Research
conducted in the 1990s supported the use of 0.6 in. (15.2 mm)
diameter strand. This size of strand is now commonly used
as a means of increasing available pretensioning force, which
makes it possible to extend spans, increase girder spacing,
and decrease structural depth. Currently, seven-wire, 0.7 in.
(17.8 mm) diameter Grade 270 (1860 MPa) low-relaxation
strands are primarily used as cable or strand roof anchors in
the mining and tunneling industries; however, these strands,
which conform to ASTM A416, may provide benefits similar
to those associated with 0.6 in. strand. Table 1 compares the
properties of 0.5, 0.6, and 0.7 in. strands.
■ An extensive analytical study is presented to as- The use of 0.7 in. (17.8 mm) strands in pretensioned bridge
sess the maximum girder span lengths that can components is not explicitly permitted in any known
be achieved when using 0.6 and 0.7 in. (15.2 and international design standard.3 The American Association
17.8 mm) strands. of State Highway and Transportation Officials’ AASHTO
LRFD Bridge Design Specifications4 includes 0.7 in. strand
■ Girder span increases of up to 22% were achieved by reference to AASHTO M 2032 but is otherwise silent
using 0.7 in. strand in place of 0.6 in. strand. on the use of 0.7 in. strand. Internationally, ASTM A4161
seems to be the only available specification for seven-wire,
■ The larger pretension forces affected end-region de- Grade 270 (1860 MPa) 0.7 in. strand. (The European
tailing and increased congestion, although all result- standard EN 101385 identifies seven-wire, Grade 250
ing requirements were constructable. [1720 MPa] 0.7 in. strand products, but these are not used in
bridge construction.) Nonetheless, Grade 270 0.7 in. strands • The total number of girders in a bridge could be
have been evaluated as pretensioning reinforcement in several reduced by using individual girders that have greater
experimental projects.6–8 pretension force. Fewer girders may shorten the
construction time, cut construction costs, and reduce
This study was conducted as part of National Cooperative overall energy consumption for fabrication and girder
Highway Research Program (NCHRP) project 12-109. transportation.
Given the limitations on the length of this paper, the reader is
referred to the NCHRP report3 and appendixes9 for additional • Longer spans could be achieved by using girders with
study details. greater pretension force. Longer spans may reduce the
number of piers required for a new bridge or permit
Motivation for using 0.7 in. strands the elimination of the central pier in typical two-span
bridges. In bridge replacement projects, particularly in
Record-breaking span lengths have recently been achieved by congested urban areas, eliminating the central piers on
using 0.6 in. (15.2 mm) strands. Examples include 62.5 m long large thoroughfares or interstate crossings may permit
(205 ft) girders in the Alaskan Way viaduct in Seattle, Wash.;10 more efficient expansion of the roadways beneath the
63.7 m long (209 ft) girders in the U.S. 17-92 interchange bridge, eliminate the hazards associated with piers
at State Road 436 in Casselberry, Fla.;11 64.0 m long (210 located close to the roadways, and minimize the impact
ft) girders in the Deerfoot Trail extension near Calgary, AB, of the span on environmentally sensitive habitats.
Canada;12 and 68.0 m long (223 ft) modified wide-flange girders Nonetheless, there are practical upper limits on girder
for a high-occupancy vehicle extension in Tacoma, Wash.13 length based on size and weight limitations associated
Given these accomplishments, a logical question is whether the with shipping and handling.14
precast concrete industry needs 0.7 in. diameter strands.
• Shallower girders that have greater pretension force could
Table 1 shows that the area of a 0.7 in. (17.8 mm) strand is be used for the same span length. This benefit becomes
92% greater than that of a 0.5 in. (13 mm) strand and 36% particularly important in replacement projects that must
greater than a 0.6 in. (15.2 mm) strand. The larger area of increase existing clearances or expand the hydraulic
0.7 in. strands, in conjunction with higher-strength concrete, opening beneath the bridge.
has the potential to offer the following benefits:3
Objectives of reported study
• The number of strands required in a girder for the same
girder span could be reduced. Fewer strands would alle- The objectives of the analytical study presented in this paper
viate congestion in heavily reinforced pretensioned ele- were to assess maximum achievable pretensioned girder
ments and may be economically advantageous. However, span lengths when using 0.6 and 0.7 in. (15.2 and 17.8 mm)
simply reducing the number of strands in a girder by strands and examine the influence of girder shape and size
replacing 0.6 in. strand with fewer 0.7 in. strands that on the potential benefits of using 0.7 in. strands. The study
provide the same reinforcement area has little, if any, also examined the impacts of using 0.7 in. strands on girder
structural advantage, such as increasing span length or end region detailing requirements, prestress transfer, and the
allowing for fewer girders (by increasing girder spacing). handling and erection stability of long-span girders.
Reducing the required number of strands could be poten-
tially beneficial in situations where filling all (or most) This study did not specifically address many additional
strand locations in a section with 0.6 in. strand does not factors that should be considered regarding the use of 0.7 in.
provide sufficient pretensioning force. Replacing 0.6 in. (17.8 mm) strands. Some of the issues that require consider-
strand with 0.7 in. strand on a one-to-one basis using the ation involve the handling of the heavier and stiffer strand and
same 51 mm (2.0 in.) grid spacing is one way to achieve larger strand forces and the potential need to retool existing
a greater pretension force. stressing beds and hardware.
Three normalweight concrete compressive strengths f c′ (69, 103, • S = 3660 mm (144 in.) and tf = 203 mm (8.00 in.)
and 124 MPa [10, 15, 18 ksi]) and one lightweight
concrete compressive strength f c′(69 MPa) were considered. • S = 4270 mm (168 in.) and tf = 203 mm (8.00 in.)
The unit weights of concrete wc were 2400, 2480, 2530, and
2000 kg/m3 (150, 155, 158, 125 lb/ft3), respectively. The • S = 4880 mm (192 in.) and tf = 229 mm (9.02 in.)
concrete strength at strand release f ci′ was assumed to be
0.6 f c′ for cases where f c′ exceeded 69 MPa, and 0.8 f c′ for cases Girder cross sections and key properties for all shapes are
where f c′ was equal to or less than 69 MPa.6 A normalweight provided in NCHRP Web-Only Document 315 appendix A.9
concrete composite slab with a compressive strength f c′ of
31 MPa (4.5 ksi) and unit weight wc of 2320 kg/m3 (145 lb/ft3) The designs assumed a simple, nonskewed span, and the
was included in all designs. A nominal 80 kg/m3 (5 lb/ft3) allow- designs were performed using a spreadsheet developed by the
ance for reinforcing steel was added to all concrete unit weights. authors, which was benchmarked and validated against LEAP
1100 NU1100
1220 BIV48
1520 UF60G5
1600 NU1600
1625 BT63
1880 WF74G
2000 NU2000
2440 FIB96
2540 WF100G
Note: FIB = Florida I-beam; NEXT = northeast extreme tee; NU = University of Nebraska I-girder; WF = wide flange. 1 mm = 0.0394 in.
Figure 1. Maximum span length chart for PCI BT72 with different concrete strengths. The percentages shown indicate increas-
es in achievable span length when 0.6 in. diameter strands (lower curves with dashed lines) are replaced with 0.7 in. diameter
strands (upper curves with solid lines). Note: BT72 = 72 in. bulb tee; LWC = lightweight concrete; NWC = normalweight concrete.
1 in. = 25.4 mm; 1 m = 3.28 ft; 1 MPa = 0.145 ksi.
One measure of the efficiency of a prestressed concrete calculated for all sections considered herein and normalized
section is the lever arm of the internal couple.15 This value is with respect to girder depth h. Figure 3 plots the relationship
represented as e + kt (Fig. 3), where e is the distance between between the achievable span increase and (e + kt)/h for all the
the centroid of the cross section and the centroid of pre- sections considered. Less-efficient girders—that is, those with
stressing steel and kt is the distance between the centroid of a smaller (e + kt)/h (such as Washington U girders [UF] and
the cross section and the top kern point. The top kern point PCI BT72 girders)—tend to benefit more from using 0.7 in.
is defined as the uppermost location in the cross section at (17.8 mm) strands.
which the compression resultant may be placed such that the
condition of zero tension is maintained at the bottom face of In addition to the girder geometry, the achievable increase
the girder. The greater the distance e + kt is, the more effi- in span length using 0.7 in. (17.8 mm) strands generally
cient the section is, and so less prestressing force is required increased as a function of girder depth (Fig. 3). Even among
to carry a given load over a given span. The value e + kt was relatively optimized shapes, deeper girders had a greater
Comparison of observations
with previous related study Figure 2. Comparison of maximum achievable span lengths
using 0.6 in. and 0.7 in. diameter strands. Note: 1 in. =
Salazar et al.16 published a parametric study on the use of 25.4 mm; 1 m = 3.28 ft.
0.7 in. (17.8 mm) strands that was similar to the study reported
herein. Using a methodology like that adopted here, Salazar et study achieved longer spans when 0.7 in. strands were used,
al. considered AASHTO Type IV, Type V, and Type VI girders, not all did. Contrary to our findings, Salazar et al. reported
Texas bulb-tee girders, Texas spread-box girders, and Texas U that shallower cross sections benefitted more from the use of
girders. They concluded that I-girder and bulb-tee girder spans larger strand diameters. Also, somewhat contrary to the find-
could be increased up to 3 m (10 ft) by using 0.7 in. strand ings of this study, Salazar et al. concluded that there was little
instead of 0.6 in. (15.2 mm) strand. This increase generally advantage to using 0.7 in. strand to increase girder spacing
required concrete strength at prestress transfer f ci′ greater than and that any advantage appeared at a girder spacing so large
69 MPa (10 ksi) and the use of harping or other methods to as to be impractical. Nonetheless, in general, the conclusions
control end-region stresses. Salazar et al. concluded that use of Salazar et al. are consistent with the findings of our study,
of 0.7 in. strand in U girders and box girders did not result in with one notable exception; Salazar et al. found no advan-
greater span lengths than could be achieved with 0.6 or 0.5 in. tage to using 0.7 in. strand in box girders, whereas our study
strands. They concluded that for I-girders and bulb-tee girders did indicate a potential advantage. Differences in box-girder
of a given span length, the required depth of some of the configuration may explain this discrepancy. Salazar et al. only
girder shapes could be decreased if 0.7 in. strands were used in considered Texas box shapes used in a spread configuration,
conjunction with higher concrete strength at prestress transfer. whereas this study considered AASHTO box shapes in an
Although many I-girders and bulb-tee girders reported in that adjacent configuration.
Figure 3. Achievable span length increases as function of girder geometry. Note: BT72 = 72 in. (1800 mm) bulb tee; e = distance
between the centroid of the cross section and the centroid of prestressing steel; kt = distance between the centroid of prestress-
ing steel and the top kern point; NU = University of Nebraska I-girder; UF = Washington U girder; WF100G = Washington wide-
flange 100 in. (2540 mm) I-girder. 1 mm = 0.0394 in.
To validate the approach used in the parametric design For each of the four cases (each girder shape using 0.6 or
study reported herein, investigators used the 2016 release 0.7 in. [15.2 or 17.8 mm] strands), a girder spacing S equal
of Cervenka Consulting’s ATENA finite element software to 2440 mm (96.1 in.) was used; deck tributary area and live
to develop full-girder models of selected cases. The anal- load distribution factors were calculated on this basis. It was
yses were based on the extensive modeling and validation assumed that each girder was fabricated 450 mm (18 in.)
studies presented in Shahrooz et al.6 Material properties longer than its span length L and was supported on 450 mm
used for the models were consistent with those used for long, full-width neoprene bearings. The distribution factor
design (as discussed previously). Initial prestressing force for flexure of interior girders gM,int was determined from
fpi was taken as 1396 MPa (202.5 ksi), and transfer length AASHTO LRFD specifications Table 4.6.2.2.2b-1 for cross
was consistent with the AASHTO LRFD specifications.4 section type (k). Table 5 summarizes the details of the PCI
Transfer length Lt is equal to 60db (where db is the strand BT72 and NU2000 models.
diameter) or 915 mm (36.00 in.) for 0.6 in. (15.2 mm)
strand and 1067 mm (42.01 in.) for 0.7 in. (17.8 mm) Table 6 summarizes the analysis results, and Fig. 4 shows
strand. Prestress losses upon tendon release were deter- selected stress and crack distributions. The finite element
mined within the finite element model based on the bond results accurately mirrored the design requirements, and
slip model; these losses were approximately 10% of the each girder mostly met the concrete stress requirements
initial prestressing force but varied somewhat based on the of the AASHTO LRFD specifications. At prestress trans-
girder cross section.6 Long-term losses were calculated to fer, the concrete tension stresses for the PCI BT72 girders
result in an effective prestressing force of 0.56fpu (where fpu fell between the AASHTO LRFD specification limits of
is the tensile strength of the strand)6 and were used at all 0.25 f ci' (0.095 f ci' in ksi units) and 0.63 f ci' (0.24 f ci'
subsequent steps (Table 4). Only the critical flexural load in ksi units), indicating a need for nonprestressed rein-
case was considered in steps 3, 4, and 5. The overstrength forcement in the region of tensile stress. The tensile stress-
factor Ω was found by increasing the axle loads LLtruck es in the NU2000 girders at prestress transfer fell below
(only) until failure of the girder. 0.25 f ci' , so no additional reinforcement would be required.
In all cases except PCI BT72-6 (where 6 indicates 0.6 in.
Two I-girder shapes were considered, and comparative [15.2 mm] diameter strand), the maximum compression
designs with both 0.6 and 0.7 in. (15.2 and 17.8 mm) strands stresses at prestress transfer exceeded the AASHTO LRFD
were modeled. PCI BT72 girders benefitted the most (in specifications limit of 0.60 f ci′ . The high compressive stress-
terms of potential span length increase) from replacing 0.6 in. es were localized near the girder ends. This discrepancy
strands with 0.7 in. strands. In contrast, 2000 mm (79 in.) between the finite element model and section-based design
fci = 0.6fc = 62
1 Release tendons Girder self-weight only n/a ≈ 0.9(0.75fpu) = 182
Eci = 41,000
fc = 103
2 Place deck slab Girder and slab self-weight n/a 0.56fpu = 151
Ec = 49,000
fc = 103 fc = 31
4 Strength 1.25DC + 1.50DW +1.75(LL + IM) 0.56fpu = 151
Ec = 49,000 Ec = 30,000
Note: DC = weight of components (barrier walls and appurtenances); DW = weight of wearing surface; Ec = modulus of elasticity of concrete corre-
sponding to fc ; Eci = modulus of elasticity of concrete corresponding to fci′ ; fc = concrete compressive strength; fci′ = concrete strength at strand release;
fpu = tensile strength of prestressing strand; IM = impact factor; LL = HL93 live load; LLlane = lane load component of HL93 live load; LLtruck = truck axle
load component of HL93 live load; n/a = not applicable; Ω = overstrength factor. 1 MPa = 0.145 ksi.
Girder
40.4 47.7 55.4 61.8
length, m
Girder span
40.0 47.3 54.9 61.3
L, m
Strands at
midspan
(harped
strands
shown in
32 straight 32 straight 56 straight 46 straight
box)
2 harped 2 harped 4 harped 14 harped
6 debonded
Top straight
Four strands stressed to 67 kN at y = 1778 mm Four strands stressed to 67 kN at y = 1948 mm
strands
0.04Apsfpi,
267 298 467 636
kN
Web rein-
forcement Thirteen pairs no. 5 at
Five pairs no. 5 at 102 mm Six pairs no. 5 at 76 mm Ten pairs no. 5 at 51 mm
at girder 51 mm
end
Web rein-
forcement Pairs no. 4 at 356 mm Pairs no. 4 at 457 mm Pairs no. 4 at 356 mm Pairs no. 4 at 356 mm
over span
Bulb tie
512 391 867 1561
force, kN
Bulb rein-
forcement Nine no. 3 hoops at Nine no. 4 hoops at Fifteen no. 4 hoops at
No. 3 hoops at 152 mm
at girder 114 mm 121 mm 64 mm
end
Bulb rein-
forcement No. 3 hoops at 152 mm No. 3 hoops at 152 mm No. 3 hoops at 152 mm No. 3 hoops at 152 mm
over span
Note: Aps = area of prestressing strand; BT72-6 = 72 in. bulb tee with 0.6 in. strand; BT72-7 = 72 in. bulb tee with 0.7 in. strand; fpi = initial prestressing
force; gM.int = distribution factor for flexure of interior girders; NU200-6 = 2000 mm University of Nebraska I-girder with 0.6 in. strand; NU2000-7 =
2000 mm University of Nebraska (NU) I-girder with 0.7 in. strand; x = distance measured from end of girder; y = vertical location of prestressing strand
measured from bottom of girder. No. 3 = 10M; no. 4 = 13M; no. 5 = 16M. 1 mm = 0.039 in.; 1 in. = 25.4 mm; 1 m = 3.28 ft; 1 kN = 0.225 kip.
Note: BT72-6 = 72 in. bulb tee with 0.6 in. strand; BT72-7 = 72 in. bulb tee with 0.7 in. strand; fc,max = maximum concrete compression stress; ft,max = max-
imum concrete tension stress; NU200-6 = 2000 mm University of Nebraska I-girder with 0.6 in. strand; NU2000-7 = 2000 mm University of Nebraska
(NU) I-girder with 0.7 in. strand; wcr,max = maximum predicted crack width; Ω = overstrength factor (see Table 4). 1 mm = 0.0394 in.; 1 MPa = 0.145 ksi.
procedure is expected and has been identified previously.6 In these comparisons, the number of strands for each girder
The design procedure used gross section properties to assess type was the same and, thus, the total prestressing force was
the stress condition (that is, P/A ± Pe/I, where A is the gross approximately 35% greater for the models having 0.7 in.
section area, P is the total prestressing force, and I is the (17.8 mm) strands (Table 1). Nonetheless, the spans increased
gross moment of inertia of the section). Near the girder only 18% and 12% for the PCI BT72 and NU2000 girders,
end, where the prestressing force is introduced primarily respectively. This combination manifested as greater precom-
in the flange, the entire section area A was not engaged for pression near the midspan, resulting in a higher decompres-
some distance into the beam. The prestressing force near the sion load and less cracking. Because the span increase for the
girder end is, in effect, resisted over a smaller area, resulting NU girder was proportionally less, the counteracting effects
in a higher stress. The spreading of the compression force of the applied moment were less significant. The improved
over the depth of the beam can be visualized at the left end cracking behavior of NU2000-7 (2000 mm [79 in.] University
of the stress plots (Fig. 4). This effect was more pronounced of Nebraska girder with 0.7 in. diameter strand) compared
for girders having large prestressing forces, and in deeper with NU2000-6 (with 0.6 in. [15.2 mm] diameter strand) at
girders with uniform bottom flanges (such as BT and NU Strength limit I is evident in Fig. 4.
girders). The highly stressed bottom flange represents a
smaller portion of the gross section area for a deeper girder, Failure of the finite element models occurred due to an
which leads to a greater discrepancy between actual local inability of the models to (mathematically) converge at loads
stresses and those calculated based on the gross area. This greater than the Ω values (Table 6). This lack of convergence
effect can be mitigated by additional debonding to introduce was associated with the relatively conservative linear bond
the prestressing force more gradually along the span. slip model used. (This model was calibrated for an AASHTO
LRFD specifications-compliant transfer length Lt of 60db.)
At the Service I limit state, tensile stress was not observed As expected, the finite element model predicted extensive
to exceed 0.50 (0.19 in ksi units) and at the Service III limit cracking near the girder midspan (Fig. 4). Although the
state, compressive stress did not exceed 0.60 f c′ in any case. stresses in the strands were approaching rupture, none of the
All designs were governed by the Strength I limit state. models predicted that the strand would rupture at the failure
Consistent with the design goal of maximizing girder length, load attained; rather, they predicted that the degree of crack-
all girders met, but demonstrated relatively little reserve ing would lead to a bond slip failure between closely spaced
capacity above, the Strength I limit state. As may be expect- cracks. The 31 MPa (4.5 ksi) slab experienced considerable
ed for long girders, relatively significant decompression and damage in the PCI BT72 models. A stronger and, therefore,
cracking were observed in the midspan regions, particularly stiffer slab would result in a relatively minor improvement in
for the PCI BT72 girders (Fig. 4). behavior. Based on the finite element model behavior reflect-
BT72-6
failure (Ω = 1.9)
BT72-7
Strength I
BT72-7
failure (Ω = 2.5)
NU2000-6
Strength I
NU2000-6
failure (Ω = 2.5)
NU2000-7
Strength I
NU2000-7
failure (Ω = 3.7)
Figure 4. Finite-element-predicted longitudinal stress contours and crack patterns. A half span is shown with support at left
and midspan at right. Note: BT72-6 = 72 in. bulb tee with 0.6 in. strand; BT72-7 = 72 in. bulb tee with 0.7 in. strand; NU200-6 =
2000 mm University of Nebraska I-girder with 0.6 in. strand; NU2000-7 = 2000 mm University of Nebraska (NU) I-girder with
0.7 in. strand; Ω = overstrength factor. 1 mm = 0.0394 in.; 1 in. = 25.4 mm.
ing the design objectives, the design approach for the para- shows the associated cracking, which was expected to become
metric study was validated. more significant but was not expected to propagate along
the girder at the Strength I limit state. However, the cracking
Effects at girder ends extended beyond the h/4 distance over which the concentrated
reinforcement was provided. This result supports providing
The finite element models provided some insight into control of the required splitting reinforcement over a longer length, as
web-splitting cracks expected to occur due to the large pre- is permitted in Washington state17 and elsewhere, or as pro-
stressing forces near the girder ends. The models included the posed by Tuan et al.18 The Washington State Department of
vertical web reinforcement, which was arranged over the initial Transportation limits the splitting reinforcement to pairs of no. 5
h/4 length of the girder (Table 5), to resist splitting. Figure 5 bars (16M) at 2.25 in. (57.2 mm) spacing but permits this detail
Prestress transfer
Strength I
Figure 5. Finite-element-predicted crack patterns over first 1.5 m (4.9 ft) of girder ends (support at left). Note: BT72-6 = 72 in.
bulb tee with 0.6 in. strand; BT72-7 = 72 in. bulb tee with 0.7 in. strand; NU200-6 = 2000 mm University of Nebraska I-girder with
0.6 in. strand; NU2000-7 = 2000 mm University of Nebraska (NU) I-girder with 0.7 in. strand. 1 mm = 0.0394 in.; 1 in. = 25.4 mm.
to extend beyond h/4 to accommodate all required bars. Tuan et dation, ties should be no. 5 bars (16M) or smaller, and spacing
al. proposed that one-half of required splitting reinforcement be should not be less than 51 mm (2.0 in.). The greatest tie-force
located within h/8, with the remainder being extended to h/2. requirement observed was 1.7 kN/m (9.8 kip/in.) predicted for
an NU2000 girder with 0.7 in. (17.8 mm) strand and a span
Transverse bulb confinement of 56.4 m (185 ft), where f c′ was 103 MPa (15 ksi) and S was
reinforcement 3.05 m (10.0 ft). This tie force could be resisted by no. 3 (10M)
hoops spaced at 32 mm (1.3 in.), which violated bar-spacing
Article 5.9.4.4.2 of the AASHTO LRFD specifications requirements; no. 4 (13M) hoops spaced at 57 mm (2.2 in.); or
requires minimum confinement of the bulb of single-web no. 5 (16M) hoops spaced at 95 mm (3.7 in.). Though congest-
sections consisting of at least no. 3 (10M) hoops spaced ed, all options are feasible (Fig. 6). Table 5 presents examples
at 6.0 in. (150 mm) over a length 1.5h from the end of the of the required bulb confinement reinforcement for the finite
beam. For heavily loaded or long-span components (as element modeled PCI BT72 and NU2000 girders.
considered herein), this requirement has been inadequate in
some cases.3,6,19 Shahrooz et al.6 identified the development Despite the greater total prestressing force present when 0.7 in.
of tension oriented transversely across the bulb of single-web (17.8 mm) strands are used, the tie force is only marginally
sections as a potential failure mode requiring tie reinforce- affected.19 Using 0.7 in. strands may result in fewer strands and,
ment across the bulb width to control associated longitudinal thus, it is easier to use a preferential strand pattern to mini-
cracking at the Strength I limit state. The magnitude of tie mize tie forces.19 The greater strand debonding required when
forces is affected by girder geometry, in particular the ratio larger prestressing forces are present also reduces the tie force
of bearing width to flange depth bb/hb (Fig. 6). Girders with because this reduces anchorage stresses at the girder end. On
wide, flat bulbs are most susceptible to developing large tie the other hand, the use of harped strands can increase tie forces
forces. The strut-and-tie modeling approach proposed by because there are fewer strands in the bulb at the girder end
Harries et al.19 (Fig. 6) was used to design bulb confinement and the harped strands are aligned with the web. Thus, debond-
reinforcement intended to mitigate lateral splitting failures at ing was given preference over harping in this study. Allowing
the ultimate limit state for all 448 single-web design cases in debonding ratios greater than 0.25, as proposed by Shahrooz
the parametric study; details are reported in Harries et al.19 et al.6 and adopted in the AASHTO LRFD specifications,4 also
affects reduced confinement-tie requirements.
All 448 single-web design cases generated for the parametric
study exhibited transverse tie forces that could be resisted Web-splitting reinforcement
without requiring unreasonably large tie bars or violating
bar spacing requirements in most cases. Based on practical Article 5.9.4.4.1 of the AASHTO LRFD specifications
considerations regarding tie placement and concrete consoli- requires that vertical splitting reinforcement (Fig. 6) be
Figure 6. Confinement requirement for single-web girder (bulb-tee shape shown). Note: bb = bearing width; hb = flange depth;
V = shear force; α = ratio of tie force to shear force.
provided in the girder web to resist a force equal to 4% of stressed concrete girders is considered in terms of the poten-
the total prestressing force: 0.04Aps fpi (where Aps is the area tial for rollover and susceptibility to excessive deformations
of the prestressing strand). Splitting forces (and the resulting that would cause concrete stress limits to be exceeded.
reinforcement requirement) are proportional only to pre- Mast20,21 noted that prestressed girders are stiff in torsion,
stressing force. Only debonding (not harping) can mitigate so lateral torsional buckling is not usually a consideration.
the splitting force (since only bonded strands are included in Rollover, which is the rigid body rotation of the girder, may
Aps). Once again, allowing debonding ratios greater than 0.25 control long girder design.22 Bracing is a relatively straight-
will mitigate splitting forces and the resulting reinforcement forward way to mitigate rollover at all handling, trans-
congestion. portation, and erection stages; this is commonly done and
represents good practice. Girders are checked for stability
The greatest splitting force requirement observed in the for the processes of moving (lifting) the girders in the plant,
parametric study was 1.45 kN/m (8.3 kip/in.), which was storage (on dunnage), transportation to the site, and lifting,
observed for multiple University of Nebraska 900 mm (35 in.) as well as in their final erected geometry prior to bracing.
(NU900) girders with 0.6 in. (15.2 mm) strands. (Harping, Girders are also checked for cracking and failure, which can
rather than debonding, was more effective at increasing occur due to excessive lateral deflection. In general, a factor
span length for this type of section.) These cases require of safety of 1.0 is used for cracking and 1.5 for failure.
bundled pairs of no. 5 (16M) hoops spaced at 76 mm (3.0 in.). Girders are also checked for rollover using a factor of safety
Congestion could be mitigated by extending the region over of 1.5.
which splitting reinforcement was placed. When splitting
reinforcement was permitted to be extended over h/3, rather In this study, investigators used the prestressed girder stability
than h/4, the splitting reinforcement requirement for the analysis approach prescribed by PCI.23 Fundamentally, the
NU900 girders could be met using single no. 5 hoops spaced stability analysis calculates factors of safety FS and stipulates
at 51 mm (2.0 in.), bundled pairs of no. 5 hoops spaced at acceptance criteria for conditions causing cracking (FScr >
102 mm (4.0 in.), or pairs of no. 4 (13M) hoops spaced at 1.0), failure (FS' > 1.5), and rollover (FSroll > 1.5). The anal-
51 mm. Table 5 includes examples of the required splitting yses are rigorous, considering girder geometry and material
reinforcement for the finite element modeled PCI BT72 and properties at each stage as well as other factors affecting
NU2000 girders. stability such as camber, prestressing force, lateral wind pres-
sure, centrifugal force during transportation, etc. The analysis
Long-span girder stability for this study was done using the Girder Stability Analysis
Excel spreadsheet created by PCI and revised by the research
This paper has shown that span lengths of existing girder team to address several programming errors found in the orig-
shapes may, theoretically, be increased as much as 22% inal version. The revised spreadsheet was validated using the
when 0.7 in. (17.8 mm) strand is used. As girders become well-documented 68 m long (223 ft) WF100G described by
longer, stability considerations during lifting and handling West;13 the validation and example calculations are presented
can begin to control aspects of design. Stability of pre- in Alabdulkarim24 and Shahrooz et al.9
Washington 74 in.
wide-flange I-gird- 124 3.05 45.8 55.2 1.21
er (WF74G)
Florida 96 in.
I-beam girder 124 2.44 63.1 68.0 1.13
(FIB96)
University of Ne-
braska 2000 mm 124 1.83 59.8 67.1 1.12
I-girder (NU2000)
Note: fc = concrete compressive strength; L0.6 = maximum achievable span for pretensioned girder with 0.6 in. diameter strand; L0.7 = maximum achiev-
able span for pretensioned girder with 0.7 in. diameter strand. 1 in. = 25.4 mm; 1 in. = 25.4 mm; 1 m = 3.28 ft; 1 MPa = 0.145 ksi.
Stability analysis of parametric design its final position, the value a of 152 mm (6.0 in.) was selected
cases in all analyses.
This study focused on stability of those cross-section and Table 10 presents the results of stability analyses in terms
span combinations that had the greatest achievable increases of the three factors of safety prescribed by PCI.23 In a few
in span length when 0.6 in. (15.2 mm) strands were replaced analyses (Table 10), additional revisions to assumptions
with 0.7 in. (17.8 mm) strands (Table 7). These cases poten- were necessary to achieve adequate factors of safety for the
tially represented the most efficient use of 0.7 in. strands, but long spans. A complete set of sample calculations for the
they also introduced the greatest potential impacts on girder NU2000 case is provided in appendix E of NCHRP Web-Only
stability. The study only analyzed stability of the longer Document 315.9
0.7 in. strand-reinforced girders.
Despite the long spans, adequate stability could be achieved
Table 8 lists girder-dependent input parameters required with all cross sections (Table 10). When the girder factor
for PCI stability analysis.23 Girder unit weight was assumed of safety for rollover FSroll was less than 1.5, the girder
to be 2400 kg/m3 (150 lb/ft3) in all cases. Table 9 reports simply requires bracing to be placed at its ends. This use of
other input parameters required for each step in the stability bracing should be standard practice for all such long girders,
analysis. Complete reporting of all cases presented, includ- and it was required for all of the evaluated girders, except
ing sample calculations, is available in NCHRP Web-Only the Washington 74 in. (1880 mm) wide-flange I-girder
Document 3159 appendix E. (WF74G), when the girders were placed in their final in situ
position.
In the analyses conducted, symmetric girder support was
located a distance a from each girder end; this location was The sections with the lowest ratio of weak- to strong-axis
varied to maximize the calculated factors of safety. An initial moment of inertia Iy/Ix, WF100G and Florida 96 in. I-beam
assumption of a equal to 0.1L was made, and the analyses girder (FIB96), failed multiple stability checks for the girder
were then revised until adequate (or maximum) factors of in transportation and when placed on dunnage. West13 offers
safety were achieved. Table 10 reports the resulting values of a simple remedy: increase the width of the top flange. In
a used. The value of a during transportation may be limited the WF100G and FIB96 cases, this approach worked. The
by interaction of the vehicle and roadway geometry—specif- top flanges of the WF100G mod and FIB96 mod sections
ically, the arc swept out by the overhanging end of the girder. were increased 457 and 305 mm (18.0 and 12.0 in.), respec-
A maximum value a of 6.1 m (20 ft) was used for the trans- tively, to achieve stability at all construction stages. West13
portation stage in this study. When the girder was placed into reported that increasing the top flange of a 68.0 m (223 ft)
A, mm2 6.99 × 105 5.32 × 105 4.95 × 105 7.50 × 105 7.59 × 105 5.83 × 105
Girder
Ix, mm4 6.35 × 1011 3.06 × 1011 2.27 × 1011 3.51 × 1011 6.09 × 1011 3.29 × 1011
geometry
Iy, mm4 2.86 × 1010 3.00 × 1010 1.71 × 1010 4.34 × 1010 3.21 × 1010 2.53 × 1010
J, mm4 3.56 × 109 2.73 × 109 2.57 × 109 4.75 × 109 4.60 × 109 3.01 × 109
Straight
46 46 32 57 66 52
strands
216 (mid) 216 (mid) 152 (mid) 254 (mid) 361 (mid) 279 (mid)
cgsharped, mm
2324 (end) 2324 (end) 1778 (end) 1765 (end) 2324 (end) 1796 (end)
long WF100G by 305 mm addressed stability issues during approximately 50% was required to satisfy FScr > 1.0 when
transportation of these particular girders. The more-slender the girder was placed on dunnage.
WF100G and FIB96 girders also tended to require stiffer
supports. For example, the WF100G mod girder (WF100G Conclusion
girder with increased top-flange width) initially failed the
cracking check for the transportation stage. A value for An extensive analytical study assessed the maximum achiev-
FScr of 1 was achieved by increasing the hauling rig stiff- able girder span lengths when 0.6 and 0.7 in. (15.2 and
ness Kqtrans 43% from 9260 kN m/rad (6830 kip ft/rad) to 17.8 mm) strands were used. Investigators conducted a para-
13,200 kN m/rad (9750 kip ft/rad). The original stiffness metric design study with 584 cases to examine the influence
was selected based on recommendations in stability analysis of girder shape and size on the potential benefits of using
guidelines23 and in consultation with practicing engineers 0.7 in. strands. A detailed finite element evaluation of some
making these calculations. The authors have not investi- of the longer spans achieved was also conducted. The impacts
gated whether the rig stiffnesses required are achievable in of using 0.7 in. strands on end-region detailing requirements,
practice. prestress transfer, and handling and erection stability of long-
span girders were examined. Based on the results and dis-
The PCI BT72 section had a thinner bottom flange width b cussion reported in this paper, we offer the following general
of 660 mm (26 in.) than the other sections. (All other sec- conclusions and observations:
tions had bottom flange widths of approximately 1000 mm
[39 in.].) The thinner flange width resulted in a significant- • The design case studies show that one-to-one replace-
ly lower bearing rotational stiffness Kqseat calculated to be ment of 0.6 in. strands by 0.7 in. strands was not possible
2560 kN m/rad (1890 kip ft/rad), which is approximately 30% given the presence of many other design constraints;
of the bearing rotational stiffness of the wider WF100G.25 among those constraints, stress limits at pretensioning
Increasing the rotational stiffness of the PCI BT72 section by release are a critical concern.
Lateral wind force at lifting from bed or in field wwind.lift 0.22 kN/m
Note: 1 mm = 0.0394 in; 1 m = 3.28 ft; 1 kN = 0.225 kip; 1 kN m = 0.738 kip ft; 1 km/h = 0.621 mph.
WF100G FIB96
WF100G WF74G PCI BT72 OHWF72 FIB96 NU2000
modified modified
Note: bold entries are below acceptance criteria. a = distance from girder end for symmetric girder support; fc = concrete compressive strength; = con-
crete compressive strength fpeff = effective prestressing stress; fpu = tensile strength of prestressing strands; FS' = factor of safety for failure; FScr = factor
of safety for cracking; FSroll = factor of safety for rollover; Ix = moment of inertia about strong axis; Iy = moment of inertia about weak axis; Kqseat = bearing
rotational stiffness; Kqtrans = hauling rig stiffness; n/a = not applicable; wwind,trans = lateral wind force during transportation. 1 m = 3.28 ft.
* Wind speeds during lifts may be limited in some cases to achieve factor of safety shown.
• When designs with 0.6 in. and 0.7 in. strand were com- • When larger-diameter strand was used to maximize preten-
pared, up to 22% increases in girder span length were sion force, greater relief of initial stresses at the girder ends
achievable in the designs with 0.7 in. strands. was required. When the harping of strands is not permitted
(as in the case of Texas U girders) or when the degree of
• The span length of existing girder shapes optimized for strand debonding is limited, the potential benefits of using
0.6 in. strands (such as NU and WF) was not appreciably the larger 0.7 in. strands cannot be fully realized.
increased when 0.7 in. strands were used. Less-efficient
shapes (such as PCI bulb tees) exhibited greater potential • The increase in achievable span length associated with
increases in their spans when 0.7 in. strand was used. 0.7 in. strand was generally proportional to girder depth.
• The increased prestressing force from the use of 0.7 in. 2. AASHTO (American Association of Highway and
strands resulted in greater splitting forces, leading to Transportation Officials). 2020. Standard Specification
potentially more congested web reinforcing steel re- for Steel Strand, Low-Relaxation Uncoated Seven-Wire
quirements at the beam ends. Nevertheless, the required for Concrete Reinforcement. AASHTO M 203M/M 203.
reinforcement was met for all the cases considered, and Washington, DC: AASHTO.
no constructability issues were anticipated.
3. Shahrooz, B. M., R. A. Miller, K. A. Harries, and R.
• The finite element models demonstrated that cracking ex- Castrodale. 2022. Use of 0.7 in. Diameter Strands in
tended beyond the h/4 distance over which concentrated Precast Pretensioned Girders. NCHRP report 994.
splitting reinforcement was provided. This result supports Washington, DC: TRB (Transportation Research Board).
designs in which the required splitting reinforcement is https://doi.org/10.17226/26677.
provided over a longer length, as is permitted by some
states. 4. AASHTO. 2020. AASHTO LRFD Bridge Design
Specifications. 9th ed. Washington, DC: AASHTO.
• The potential for longer achievable spans increases the
susceptibility of girders to instabilities. As is required 5. Technical Committee ECISS/TC 19 SC2. 2000.
for much shorter girders than those considered here, end Prestressing Steels—Part 1: General Requirements. Draft
braces must be installed immediately upon placement prEN 10138-1. Brussels, Belgium: European Committee
on bearings to provide safety against rollover. For other for Standardization.
conditions, the following measures can improve safety
against stability effects: 6. Shahrooz, B. M., R. A. Miller, K. A. Harries, Q. Yu, and
H. G. Russell. 2017. Strand Debonding in Pretensioned
— Refining lift points and symmetric dunnage support Girders. NCHRP report 849. Washington, DC: TRB.
locations a can optimize resistance to stability https://doi.org/10.17226/24813.
effects. The value of a may be practically limited
during transportation based on vehicle geometry 7. Schuler, G. 2009. “Producer’s Experience with 10,000 psi
and routes chosen. Concrete and 0.7 in. Diameter Strands.” HPC Bridge
Views, no. 54, 7–8. http://concretebridgeviews
— Increasing the width of the top flange of a girder .com/2009/03/producers-experience-with-10000-psi
thereby increasing Iy/Ix has a pronounced effect on -concrete-and-0-7-in-diameter-strands.
improving stability.
8. Nebraska Department of Roads. 2015. The Roadrunner,
— Providing stiffer transportation or dunnage February/March.
support—assuming that this is possible—improves
stability. 9. Shahrooz, B. M., R. A. Miller, K. A. Harries, and R.
Castrodale. 2022. Details of the Study on the Use of 0.7
— Girders with relatively thin bottom flanges (bulb- in. Diameter Strands in Precast Pretensioned Girders.
tee sections in this study) are more susceptible to NCHRP Web-Only Document 315. Washington, DC:
rollover while supported on dunnage or in transpor- TRB. https://doi.org/10.17226/26676.
tation.
10. Concrete Technology Corp. “Alaskan Way Viaduct.”
The parametric study was intended to be illustrative and to Accessed February 25, 2023. http://www.concretetech
identify trends. Many assumptions were made, and the results .com/projects/alaskan-way-viaduct-.
presented are constrained by these.
12. Canadian Precast/Prestressed Concrete Institute. 2002. 23. PCI Bridges Committee. 2016. Recommended Practice
“Record-Breaking Precast NU Girders Installed in for Lateral Stability of Precast, Prestressed Concrete
Alberta.” https://www.cpci.ca/en/about_us/project Bridge Girders. CB-02-16. Chicago, IL: PCI. https://doi
_month/april_2002/. .org/10.15554/CB-02-16.
13. West, C. 2019. “Prestressed Concrete Girders Achieve 24. Alabdulkarim, A. 2021. “Use of 0.7 in. Diameter
Record Lengths.” Aspire 13 (4): 56–57. https://www Prestressing Strand in Bridge Girders: Bond Behavior
.aspirebridge.com/magazine/2019Fall/CCC and Girder Stability.” PhD diss., University of Pittsburgh.
-PrestressedConcreteGirders.pdf. https://d-scholarship.pitt.edu/40487.
14. Castrodale, R. W., and C. D. White. 2004. Extending 25. Stanton, J. F., C. W. Roeder, P. MacKenzie-Helnwein,
Span Ranges of Precast Prestressed Concrete Girders. C. White, C. Kuester, and B. Craig. 2008. Rotation
NCHRP report 517. Washington, DC: TRB. https://doi Limits for Elastomeric Bearings. NCHRP report 596.
.org/10.17226/23375. Washington, DC: TRB. https://doi.org/10.17226/23131.
17. Khaleghi, B. 2006. “Design Memorandum: Splitting Aps = area of prestressing strand
Resistance of Pretensioned Anchorage Zones.”
December 23, 2006. Washington State Department of b = bottom flange width
Transportation. https://wsdot.wa.gov/eesc/bridge
/designmemos/11-2006.htm. bb = bearing width
18. Tuan, C. Y., S. A. Yehia, N. Jongpitaksseel, and M. K. bbot flange = bottom flange width
Tadros. 2004. “End Zone Reinforcement for Pretensioned
Concrete Girders.” PCI Journal 49 (3): 68–82. https://doi btop flange = top flange width
.org/10.15554/pcij.05012004.68.82.
cgsharped = center of gravity of harped prestressing strand
19. Harries, K. A., B. M. Shahrooz, B. E. Ross, P. Ball,
and H. R. Hamilton. 2019. “Modeling and Detailing cgsstraight = center of gravity of straight prestressing strand
Pretensioned Concrete Bridge Girder End Regions
Using the Strut and Tie Approach.” Journal of Bridge db = nominal diameter of prestressing strand
Engineering 24 (3). https://doi.org/10.1061/(ASCE)
BE.1943-5592.0001354. dr = debonding ratio
20. Mast, R. F. 1989. “Lateral Stability of Long Prestressed DC = weight of components (barrier walls and appurte-
Beams Part 1.” PCI Journal 34 (1): 34–53. https://doi nances)
.org/10.15554/pcij.01011989.34.53.
DW = weight of wearing surface
21. Mast, R. F. 1993. “Lateral Stability of Long Prestressed
Beams Part 2.” PCI Journal 38 (1): 70–88. https://doi e = distance between the centroid of the cross section
.org/10.15554/pcij.01011993.70.88. and the centroid of prestressing steel
22. Zureick, A. H., L. Kahn, K. M. Will, I. Kalkan, J. Hurff, ebrg.seat = bearing tolerance from centerline of girder to cen-
and J. H. Lee. 2009. Stability of Precast Prestressed terline of support
Eci = modulus of elasticity of concrete corresponding to LLlane = lane load component of HL93 live load
fc = concrete compressive strength LLtruck = truck axle load component of HL93 live load
f ci′ = concrete strength at strand release L0.6 = maximum achievable span for pretensioned girder
with 0.6 in. (15 mm) diameter strand
fc,max = maximum concrete compressive stress
L0.7 = maximum achievable span for pretensioned girder
fpeff = effective prestressing stress with 0.7 in. (18 mm) diameter strand
fpu = tensile strength of prestressing strand radiustrans = turn radius for adverse cross slope
FScr = factor of safety for cracking Veltrans = hauling rig velocity in turn
FSroll = factor of safety for rollover w = girder weight per unit length
gM,int = distribution factor for flexure of interior girders wbrg.seat = plan dimension of the bearing
hbrg.seat = height of bearing wwind.lift = lateral wind force at lifting from bed or in field
hroll.seat = height of roll center from bearing seat wwind.seat = lateral wind force
hroll.trans = height of roll center above roadway wwind.trans = lateral wind force
Ω = overstrength factor
Abstract
Payne D. Ball is a structural It has been proposed that 0.7 in. (17.8 mm) diame-
engineer with Oldcastle ter prestressing strand be permitted for use in bridge
Infrastructure and a former girders. If 0.6 in. (15.2 mm) diameter strand is re-
graduate student at the University placed on a one-to-one basis with 0.7 in. strand, the
of Cincinnati. pretensioning force can be increased by 35%. When
designs use 0.7 in. strands as well as high concrete
strengths, longer-span prestressed concrete girders may
TianQiao Liu, PhD, is an associate be achieved. An extensive analytical study is presented
professor at Beijing University of to assess the maximum girder span lengths that can be
Technology in China and a former achieved when using 0.6 and 0.7 in. strands. A para-
postdoctoral researcher at the metric design study with 584 cases was conducted to
University of Pittsburgh. examine the influence of girder shape and size on the
potential benefits of using 0.7 in. strands. A detailed
finite element analysis of some of the longer spans
Venkata S.S.P. Sathiraju is a achieved was also conducted. The impacts of using
bridge engineer with Gresham 0.7 in. strands on end-region detailing requirements,
Smith and a former graduate prestress transfer, and handling and erection stability of
student at the University of long-span girders were examined. Girder span increas-
Cincinnati. es of up to 22% were achieved using 0.7 in. strand in
place of 0.6 in. strand. The larger pretension forces
affected end-region detailing and increased congestion,
Abdullah Alabdulkarim, PhD, is though all resulting requirements were constructible.
an assistant professor at King The longer spans affected girder stability calculations,
Saud University in Riyadh, Saudi and some girder types required a wider top flange to
Arabia. meet stability-related limit states.
Keywords
Publishing details
Reader comments
Board of Directors
Matt Ballain, Chair, Coreslab Lloyd Kennedy, Institute Program Patty Peterson, Institute Program
Structures (INDIANAPOLIS) Director, Educational Activities, Director, Business Performance,
Jim Fabinski, Vice Chair, EnCon Finfrock Industries LLC Tindall Corp.
Colorado Brent Koch, Producer Member Richard Potts, Producer Member
Carlos Cerna, Secretary-Treasurer, Director, PCI West, Con-Fab Director, Georgia/Carolinas PCI,
Manco Structures Ltd. California LLC Standard Concrete Products Inc.
J. Matt DeVoss, Immediate Past Cheryl Lang, At-Large Member, Tindall Jim Renda, Associate Member
Chair, Jackson Precast Inc. Corp. Director, Supplier, Cresset Chemical
Bob Risser, President and CEO, PCI Matt Mahonski, Producer Member Co.
Director, PCI Central, High Concrete Lenny Salvo, Producer Member
Dusty Andrews, Producer Member Group Director, Florida Chapter, Coreslab
Director, PCI Washington/Oregon, David Malaer, Producer Member Structures (ORLANDO) Inc.
Knife River Corp.–Northwest Director, Texas, Oklahoma, New Peter Simoneau, Producer Member
Dennis Cilley, Associate Member Mexico, Valley Prestress Products Inc. Director, PCI Northeast, Dailey
Director, Erector, American Steel & Jane Martin, Institute Program Director, Precast
Precast Erectors Marketing, Gate Precast Co. Ben Spruill, Producer Member
Ned Cleland, Professional Member Brian Miller, Associate Member Director, PCI Gulf South, Gulf Coast
Director, Blue Ridge Design Inc. Director, Supplier, GCP Applied Prestress Partners
Todd Culp, Producer Member Technologies Inc. Kimberly Wacker, At-Large Member,
Director, PCI Midwest, Coreslab James Miller, Producer Member Wells
Structures (OMAHA) Inc. Director, PCI of Illinois & Wisconsin Lee Wegner, Producer Member
Brandon Farley, Producer Member Richard A. Miller, Institute Program Director, PCI Mountain States,
Director, PCI Midwest, EnCon Field Director, Technical Activities, Forterra Structural Precast
Services LLC University of Cincinnati Mike Wolff, Institute Program
Evan Fink, Producer Member Christopher Mosley, Professional Director, Quality Activities,
Director, PCI Mid-Atlantic, Member Director, The Consulting Mid-States Concrete Industries LLC
Northeast Prestressed Products Engineers Group Inc. Diep Tu, Regional Council
LLC Andrew Osborn, Institute Representative, Nonvoting, Florida
Troy Jenkins, Institute Program Program Director, Research and Prestressed Concrete Association
Director, Transportation Activities, Development, Wiss, Janney, Elstner Colin Van Kampen, Membership
Northeast Prestressed Products Associates Inc. Council Representative, Nonvoting,
Structures Inc.
Edith Gallandorm (312) 360-3219 egallandorm@pci.org Codes and standards managing director
Bob Risser (312) 360-3204 brisser@pci.org President and chief executive officer
Beth Taylor (312) 583-6780 btaylor@pci.org Chief financial and administrative officer
Information technology and events
Cindi Ward (312) 360-3214 cward@pci.org
coordinator
Randy Wilson (312) 428-4940 rwilson@pci.org Architectural precast systems director
Coming ahead
Materials
• Review of the effects of supplementary cementitious
materials on age-dependent concrete properties
affecting precast construction
• Axial load behavior of reinforced concrete columns
with high-strength steel coiled strips as confinement
• Alkaline activation and mechanical properties of lake
sediment geopolymers
Also
• Numerical and digital image correlation study of
the flexural behavior of prestressed ultra-high-
performance concrete beams made from locally
available materials
• Meet Amir Fam
A lifetime
of learning and giving
William Atkinson
Photograph of the
completed construction
To bring this 163,000 -square-foot total precast tower ending with the installation of the T-SLABs. This process
to life on a constricted, metropolitan job site, Tindall provided early access to each floor for other trades,
developed nearly 700 precast elements that could be dramatically accelerating the construction schedule.
quickly erected with just a single crane and a small
In addition to providing long, clear spans for each
on-site crew. These elements included load-bearing
floor, the T-SLABs also concealed MEP throughout the
architectural wall panels, stair and elevator towers,
structure. This was possible thanks to T-SLAB’s unique
and the T-SLAB ® floor system.
design, which allows the internal lightweight concrete
Unlike most precast structures, which are fully blocks to be omitted to accommodate bathroom
constructed in sections, The Icon was erected installations, additional reinforcement, or mechanical,
sequentially by level. The precast elements for each electrical, and plumbing items, further accelerating the
floor were installed in a “back-to-front” construction construction schedule. In the end, The Icon was fully
method, beginning with the insulated wall panels and erected in just under 16 weeks.
TINDALLCORP.COM