Engineering Structures 42 (2012) 18

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Behavior of precast prestressed concrete bridge girders involving thermal effects

and initial imperfections during construction
Jong-Han Lee
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA

a r t i c l e

i n f o

Article history:
Received 10 December 2010
Revised 20 January 2012
Accepted 8 April 2012

Keywords:
Precast prestressed concrete
AASHTO-PCI standard girder
Thermal effects
Elastomeric bearing
Nonlinear nite element

a b s t r a c t
The instability of precast prestressed concrete bridge girders during construction have been of particular
concern to bridge engineers. After they are installed on bearing supports, prestressed concrete girders are
immediately subjected to environmental thermal loads that may be exacerbated by fabrication and construction errors. Thus, in this research, the environmental thermal loads, which cause extremes in thermal deformations in precast prestressed concrete girders, were determined. Then a three-dimensional
nonlinear nite element sequential analysis procedure was developed to evaluate the behavior of a precast prestressed concrete girder subjected to both thermal loads and geometry and support imperfections
during each construction stage. This analysis indicated instability in a 30-m long prestressed concrete BT1600 girder when total lateral deformation in the middle height of the girder at mid-span exceeded about
25 cm.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Since the advent of precast prestressed concrete girders in
bridge design and construction, the demand for a more expansive
girder span, which would reduce construction costs and improve
bridge esthetics, has been increasing. However, the lengthening
of girders with deeper precast sections and high-strength concrete
increases the likelihood that the girders will destabilize. Such failure, one of which is illustrated in Fig. 1, have led to considerable
apprehension about the behavior of precast prestressed concrete
girders during construction, specically before the addition of the
slab and bracing. One investigation into the collapse of the girders,
Oesterle et al. [1] indicated that a combination of several factors,
including the initial sweep (or lateral deformation), the thermal
sweep, and the support slope, could cause lateral instability of
the girders during construction.
The initial sweep of the girder occurs during fabrication, shipping, and handling. During fabrication, the eccentricity of prestressing strands can create an error that leads to unexpected
initial sweep in the girder. Then shipping and handling can subject
the girder to unaccounted loads or boundary conditions that also
affect the initial sweep. When placed on supports that are not level,
the girder can also experience sweep. In addition, while prestressed concrete girders are resting on a bearing support, thermal
environmental effects can produce additional sweep that may contribute to the instability of the girders prior to the placement of a
Tel.: +1 404 894 2278; fax: +1 404 894 2201.
E-mail address: jonghan.lee@gatech.edu
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.04.003

bridge deck and diaphragms. For the initial sweep in the girder, the
PCI Bridge Design Manual [2] provides a tolerance of 3 mm per 3 m
(1/8 in per 10 ft) of member length. Nevertheless, in practice, this
small tolerance value has not been adhered to because of the many
unexpected conditions that occur during fabrication and handling,
as mentioned previously. However, no study that evaluates the lateral deformation of the girders has been carried out, especially during construction when the girders are subjected to the combined
effects of thermal response, initial sweep, and support slope.
Some relevant initial research was conducted by Mast [3,4],
who calculated the stability of a girder suspended from lifting devices and transported on elastic supports. Mast proposed a method
based on the ratio of a resisting moment at the support to an overturning moment induced by the girder sweep and support slope.
The method was adopted in the PCI Bridge Design Manual [2] for
the evaluation of how safe a girder is from rollover (or overturning)
during shipping and lifting. However, the manual provided no specic method or guideline that analyzes the lateral stability of the
girders placed on elastic bearing supports during construction.
The AASHTO LRFD Bridge Design Specications [5] and the AASHTO
LRFD Bridge Construction Specications [6] simply addressed the
importance of considering the safety of precast members during
all construction stages, but they did not provide any specic guidelines related to the stability of precast prestressed concrete girders
during construction.
Thus, as an initial study, this research evaluated the behavior of
a precast prestressed concrete girder subjected to the combined effects of the initial sweep, the thermal response, and the support
slope during construction. For the largest vertical and lateral

J.-H. Lee / Engineering Structures 42 (2012) 18

and the position of the sun. The specic calculation procedures

are outlined in Dufe and Beckman [9].
Variations in the air temperature were calculated using the
sinusoidal Kreith and Kreider equation [10]:

T air t

Fig. 1. Stability failure of precast prestressed AASHTO Type-V girders during the
construction of the Red Mountain Freeway in Arizona [1].

thermal response, the thermal effects of seasonal variations and

bridge orientations on vertical and lateral thermal gradients were
evaluated for four AASHTO-PCI standard girder sections in Atlanta,
Georgia. For a 30-m long prestressed BT-1600 girder, which
showed the largest vertical and lateral thermal gradients in the
summer and the winter, respectively, the behavior of the girder
subjected to the thermal response, the initial sweep, and the support slope was evaluated using three-dimensional (3D) nonlinear
nite element sequential analysis.
2. Determination of thermal loads
The primary environmental parameters causing temperature
variations in bridges are solar radiation, air temperature, and wind
speed. These environmental values were determined from 30-year
(from 1961 to 1990) monthly averaged daily solar radiation and
climatic data provided by the National Renewable Energy Laboratory [7] for Atlanta. To account for seasonal variations in environmental conditions, this study chose the daily solar radiation values
of 21.9, 29.4, 22.4, and 11.9 MJ/m2 for March, June, September, and
December, respectively. The months of March, June, September,
and December were dened as representative months of the
spring, summer, fall, and winter. The daily maximum and minimum air temperatures for each season were determined from
the record maximum temperature and the average daily minimum
temperature of the 30-year climatic values, respectively, for the
representative months. The average daily minimum temperature
was used rather than the record daily minimum temperature since
it is highly unlikely that the record maximum and minimum temperatures occurred on the same day. However, since wind speed
was minimal on the days when the largest vertical and lateral thermal gradients occurred, this study neglected the effect of wind
speed on girder temperatures.
The temporal variation in the daily solar radiation was calculated using the Liu and Jordan equation [8]:

It H

p
24


a b  cos w 

cos w  cos ws
sin ws  ws cos ws

h
1
1
pi
T max T min  T max  T min  sin t  9
2
2
12

in which Tair(t) is the air temperature as a function of time t, and

Tmax and Tmin are the daily maximum and minimum air temperatures, respectively. In the calculations, the lengths of the 21st of
March, June, September, and December represented the length of
the days in each season.
With the dened seasonal environmental conditions, a twodimensional heat transfer analysis was conducted on four AASHTO-PCI standard girder sections: Type-I, Type-IV, Type-V, and
BT-1600 [2]. The heat transfer mechanisms involved in this study
are heat irradiation from the sun, heat radiation to the surroundings, heat convection between the surroundings and the concrete
surface, and heat conduction in the concrete. In the calculation of
heat gain from the sun and heat loss to the surroundings, the value
of solar absorptivity and surface emissivity of concrete was selected to be 0.50 and 0.85, respectively [11]. The thermal conductivity and specic heat of concrete, which denes the heat ow
within the body of the concrete girder, were taken as 1.50 W/m K
and 1000 J/kg K, respectively, based on a previous study on the
temperature prediction of concrete pavement [12].
The inuence of changes in the girder orientations on temperature distributions was also evaluated for eastwest (EW), south
north (SN), southwestnortheast (SWNE), and southeast
northwest (SENW) orientations. The largest vertical thermal
gradients, calculated from the largest temperature difference
between the highest and lowest temperatures along the depth of
the sections, were found in the summer and in the EW orientation. The largest lateral thermal gradients across the middle of
the top ange, the web, and the bottom ange were found in the
EW orientation in the winter because of the greater exposure of
the vertical surfaces of the girder to the sun in the EW orientation.
Among the four sections, the deeper and wider Type-V and BT1600 sections exhibited larger vertical and lateral thermal gradients. The largest vertical thermal gradients in the summer were
26 C in the Type-V section and 25 C in the BT-1600 section. The
largest lateral thermal gradients of the Type-V and BT-1600
sections in the winter were about 20 C in the top ange, 15 C
in the web, and 25 C in the bottom ange. Thus, for the BT-1600
girder in an EW orientation, the thermal loads in this study were
determined from the summer and winter environmental conditions in Atlanta.


1

in which I(t) is the solar radiation at time t on a horizontal surface, H

the total daily solar radiation, w the solar hour angle, and ws the
sunrise hour angle; a = 0.409 + 0.5016  sin(ws  1.047), and
b = 0.66090.4767  sin(ws  1.047). For other inclined surfaces of
the girder, irradiation was estimated according to the location and
orientation of the girder, the geometry and shadow of the girder,

Fig. 2. Arrangement of the prestressing strands in the BT-1600.

J.-H. Lee / Engineering Structures 42 (2012) 18

3. 3D nonlinear nite element thermal response analysis

3.1. 3D nite element model
Based on the preliminary design reference given in the PCI
Bridge Design Manual [2], the BT-1600 girder was designed to be
30 m long with a concrete compressive strength of 48 MPa, girder
spacing of 1.8 m, and 24 low relaxation strands of 12.7 mm in
diameter. As shown in Fig. 2, four strands were placed in the top
ange and twenty strands in the bottom ange. The strands placed
in the top and bottom anges were each prestressed to 44.5 kN and
150.4 kN, respectively.
The 30-m long BT-1600 girder was modeled with a total of
350,400 linear solid elements of approximately 2.5 cm by 2.5 cm
with 5 cm in the longitudinal direction of the girder per element
in the nite element program Abaqus [13]. The prestressing
strands were modeled using 3D two-node truss elements and dened as embedded elements in the solid concrete elements. The
embedded element technique used in this study constrains the
translational degrees of freedom of the embedded elements to
the interpolated values of the corresponding degrees of freedom
of the host solid elements.
The modulus of elasticity of the concrete used in this study was
calculated to be 28,500 MPa using the ACI Committee 435 [14] for
high-strength concrete. Since the maximum tensile stress of concrete due to thermal effects
p did not exceed the allowable tensile
stress of concrete, 0:63 fci , in which fci represents the compressive
strength of concrete, the concrete materials used in this thermal
response analysis were modeled to be linear elastic. The coefcient
of thermal expansion pertaining to the thermal movement of the
girder was taken as 10.8  106/C [5].
The material properties of the 12.7 mm diameter strands were
those listed in the PCI Bridge Design Manual. The design yield
strength of the strands was 1689 MPa. After the yielding of the
strands, or 0.0086 yield strain, the stress and strain relationship
was dened to be perfectly plastic. The coefcient of the thermal
expansion of the strands was dened to be the same as that of
the concrete.

3.2. Support conditions

Precast prestressed concrete bridge girders are generally supported by steel-reinforced elastomeric bearing pads that provide
vertical support in compression and minimum horizontal resistance to the girder due to friction. According to a recent study on
the elastomeric bearing stiffness of standard precast prestressed
concrete girders [15], the vertical stiffness calculated from the
AASHTO LRFD Bridge Design Specications [5] provided good agreement with that obtained from the nite element analysis. This
study also recommended that the effects of the horizontal bearing
restraint be ignored in the design of AASHTO precast concrete
bridge girders. Thus, the vertical stiffness of the bearing pad was
evaluated using the AASHTO specications. This analysis assumed
that the horizontal stiffness of the bearing pad was zero, but included the lateral resistance to the girder, provided by the dowel
bar in the middle of the pad.
The size of the bearing pad for the BT-1600 girder, based on the
Georgia Department of Transportation manual [16], was determined to be 25 cm long and 50 cm wide. A 7.5 cm diameter hole
in the middle of the pad was also designed for the dowel bar.
The hole of the bearing pad at the other end of the girder was slotted so that the girder could expand longitudinally. Thus, the dowel
bar, located in the middle of the bearing pad, was modeled with
both lateral and longitudinal restraints at the one end and only a
lateral restraint at the other end.

With the dened geometry of the bearing pad, the compressive

modulus of bearing pad was calculated using the stress and strain
curve given in the AASHTO LRFD Bridge Design Specications [5].
Since elastomeric bearings become stiffer as strain increases, the
stiffness of the vertical bearings was dened as a bi-linear relationship with one inection point at 0.028 compressive strain. That is,
the compressive modulus of the elastomeric bearing pad, Es, is initially 148 MPa up to 0.028 compressive strain, and thereafter increases to 230 MPa. The vertical stiffness of bearing pad, k, was
then derived as follows:

Es  A
H

in which A is the area of the bearing pad and H the thickness of the
bearing.
The vertical bearing stiffness was modeled using a series of nonlinear spring elements, which provide restraint only when compressed. The compressive stiffness of individual springs was
calculated using the tributary area of the springs categorized as
corner, edge, and center spring elements according to the location
of the springs within the bearing surface. Fig. 3 shows the force and
displacement relationship dened from the calculated compressive
stiffness, in which k1 represents compressive strain up to 0.028 and
k2 represents that greater than 0.028. The vertical spring elements
used to model the bearing pads are illustrated in Fig. 4. The arrows
shown in this gure represent the restrained directions at the both
ends due to the dowel bars located in the middle of the pad.
3.3. Thermal responses
The 3D nonlinear nite element thermal response analysis is
composed of a static analysis and its subsequent nonlinear thermal
stress analysis. First, the 3D static analysis is performed to introduce camber and stresses induced by prestressing forces to the girder. The prestressing forces were dened as the initial stress
conditions and applied uniformly along the strand. The values of
the initial stresses, assigned to the top and bottom strands, were
482 MPa and 1578 MPa, respectively, calculated by dividing the
prestressing forces by a nominal area of the strand. The support
boundary condition in this analysis was dened as a simply supported condition at the location of the dowel bars. The camber
and stresses obtained from the rst static analysis provide the initial conditions for the start of the subsequent thermal stress analysis. Then the 3D thermal stress analysis employed the girder
temperatures obtained from heat transfer analysis as sequential
thermal loads to determine the thermal response of the girder.
Since the heat transfer analysis is carried out on a 2D cross-section
of the girder, the temperature distributions are transferred to the

Fig. 3. Relationship between the force and displacement of the spring element.

J.-H. Lee / Engineering Structures 42 (2012) 18

(a) One end of the girder

(b) Other end of the girder

Fig. 4. Finite element model with spring elements for elastomeric bearing pads and restrained conditions for dowel bars.

Fig. 5. Flowchart of the 3D nite element thermal response analysis.

3D nite element model with a constant temperature variation

along the length of the girder. The support conditions involved in
this analysis were dened to be elastomeric bearing pad conditions
shown in Fig. 4. For the purpose of comparison, the analysis included boundary conditions that dened the bearing pads as rigid
vertical restraints. The detailed process of the 3D nite element
thermal response analysis is given in the owchart shown in Fig. 5.
Fig. 6 shows variations in the vertical and lateral thermal movements at the mid-span of the 30-m long prestressed BT-1600 girder obtained from the 3D thermal response analysis of the
summer and winter environmental conditions in Atlanta. The initial vertical camber induced by the prestressing strands was
3.28 cm, and it decreased to 0.81 cm due to the self-weight of
the girder. The environmental thermal loads increased the vertical
displacement to 2.67 cm at 3 p.m. in the summer and 1.91 cm at
2 p.m. in the winter under elastomeric bearing pad conditions.
Thus, the total vertical thermal movement was 1.86 cm in the summer and 1.10 cm in the winter. In the rigid support conditions, the
total vertical thermal movement was 1.07 cm in the summer and
0.46 cm in the winter. In contrast to the vertical thermal movement, lateral thermal movement at mid-span was greater in the
winter due to larger lateral thermal gradients. The lateral thermal
movement was 1.02 cm in the summer and 1.96 cm in the winter.
The support conditions of the elastomeric bearing pads did not affect the lateral thermal movements, and the differences in the lateral movements along the depth of the girder were minimal.
Nonlinear thermal strains induced by environmental thermal
loads cause self-equilibrium stresses in the girder due to the strain
difference between the nonlinear thermal strains and the nal

Fig. 6. Variations in the vertical and lateral thermal movements at the mid-span of the 30-m long prestressed BT-1600 girder under elastomeric bearing pad conditions in the
summer and the winter in Atlanta.

J.-H. Lee / Engineering Structures 42 (2012) 18

linear strains. The BT-1600 girder initially exhibited longitudinal

compressive stresses of 5723 kPa on the top surface and
7584 kPa on the bottom surface at mid-span due to prestressing
forces and self-weight. In the summer, when the largest vertical
thermal gradients occurred, the concrete longitudinal compressive
stresses increased to 7791 kPa (by 36%) on the top surface and
9170 kPa (by 21%) on the bottom surface. In the winter, the compressive stresses on the top and bottom surfaces only slightly
changed due to the smallest vertical thermal gradients. However,
the middle of the web showed higher longitudinal compressive
stresses due to larger thermal gradients in the web in the winter.
The tensile stresses of the top and bottom strands only slightly
changed (increased by less than 41 MPa) due to environmental
thermal effects.

Similarly, lateral thermal movement was calculated using the

lateral thermal gradients over the cross-section. However, the lateral thermal gradients vary from the top to the bottom anges, so
the lateral curvature of the girder, ux, was obtained from three lateral thermal gradients in the middle of the top ange, the web, and
the bottom ange:

X R

ux

E  a  DTx  hx  xdx
E  Iy

in which DT(x) is the lateral thermal gradient at width x, h(x) the

depth of the girder, Iy the moment of inertia of the cross-section
with respect to the weak y-axis, and i (=1, 2, 3) the lateral thermal
gradients of the top ange, the web, and the bottom ange of the
girder, respectively.

4. Beam model for the calculation of thermal deformations

4.2. Comparison of the beam model with the 3D nite element analysis
4.1. Development of the beam model
Thermal strain distributions induced by environmental thermal
effects are nonlinear. Because of the nonlinear thermal gradients in
the girders, their thermal movements are basically calculated using
3D numerical analysis. Thus, based on beam theory, this study derived an analytical method that could calculate the vertical and
thermal deformations from the nonlinear thermal gradients.
Fig. 7 illustrates vertical strain distributions along the depth of a
prestressed concrete girder section caused by a nonlinear vertical
thermal gradient. The unrestrained plane section tends to expand
in accordance with the vertical thermal gradient shown in
Fig. 7b. However, according to the NavierBernoulli hypothesis,
the nal strain prole is linear, illustrated in Fig. 7c. Thus, the vertical thermal deformation can be obtained by integrating the curvature, uy, over the length of the girder. For a simple span, the
vertical deformation at mid-span, dy, is

dy

uy  L 2

in which L is the length of the girder. The curvature of the girder can
be calculated using the following equation:

E  a  DTy  wy  ydy
E  Ix

in which DT(y) is the vertical thermal gradient at depth y, w(y) the

width of the section, E the concrete modulus of elasticity, a the coefcient of thermal expansion of concrete, and Ix the moment of inertia of the cross-section with respect to the strong x-axis.

(a) Cross-section

For the BT-1600 girder, which showed the largest vertical and
lateral thermal gradients among the four AASHTO-PCI standard
girder sections, this study calculated the vertical and lateral
thermal movements in the summer and winter environmental
conditions in Atlanta using the proposed model. The length of
the BT-1600 girder was 30 m, and the material properties of concrete were the same as used previously.
The vertical and lateral thermal movements calculated using
the beam model were compared with those obtained from the
3D nite element thermal stress analysis for simply supported
boundary conditions. Fig. 8 shows that the thermal movements
calculated from the beam model correlate well with those obtained
from the 3D nite element analysis. Differences between the maximum vertical and lateral thermal movements of the beam model
and those of the 3D nite element analysis were less than 0.08 cm
(5.3%).
4.3. Equations for calculating maximum thermal movements in
AASHTO-PCI standard girders
The beam model was further used to propose simple equations
that are capable of calculating the vertical and lateral thermal
movements with the design span of precast prestressed concrete
bridge girders. For the four AASHTO-PCI standard girder sections,
Type-I, Type-IV, Type-V, and BT-1600 sections, Table 1 summarizes
the equations in terms of the span length of the girder. The equations can be used to calculate the maximum vertical and lateral
thermal movements of the girders located in Atlanta. The thermal

(b) Thermal strain

(c) Final strain

Fig. 7. Strain distributions induced by a nonlinear vertical thermal gradient in a simply supported prestressed concrete girder section.

J.-H. Lee / Engineering Structures 42 (2012) 18

Fig. 8. Comparisons of the vertical and lateral thermal movements calculated using the beam model with those obtained from the 3D nite element analysis.

Table 1
Maximum vertical and lateral thermal movements of the four AASHTO-PCI standard girder sections in the summer and the winter (Units:  2.73  105 m).
AASHTO-PCI standard sections

Type-I
Type-IV
Type-V
BT-1600
a

Max. vertical movement

Max. lateral movement

Max. span (m) [2]

Summer

Winter

Summer

Winter

117L2
39L2
59L2
55L2

71L2
23L2
26L2
20L2

53L2
123L2
40L2
40L2

63L2
153L2
82L2
83L2

15
37
44
40

L is the span of the girder in meters.

loads were based on summer and winter environmental conditions

that yield the largest vertical and lateral thermal movements,
respectively.
Since the behavior of the girders mainly depends on the moment of inertia of the cross-section, the Type-I section, which has
the smallest moment of inertia with respect to the strong x-axis,
has the largest vertical thermal movement. For a Type-I girder designed to be a maximum span of 15 m, the maximum vertical thermal movement was calculated to be 0.71 cm in the summer in
Atlanta. The largest lateral thermal movements occurred in the
Type-V and BT-1600 sections, which have the largest lateral thermal gradients and a small amount of moment of inertia with respect to the weak y-axis.

5. Behavior of a precast prestressed concrete girder involving

thermal effects with geometry and support imperfections
5.1. 3D nonlinear nite element sequential analysis procedure
The environmental thermal effects are combined with fabrication and construction errors to affect the behavior of precast prestressed concrete girders during construction, especially prior to
the placement of cross bracing and the deck slab. This study found
that the main imperfection was initial lateral deformation in the
girder at mid-span and the bearing slope in the lateral direction.
To analyze the combined effects of the initial sweep, the bearing
support slope, and thermal loads on the prestressed concrete girder, a 3D nonlinear nite element sequential analysis procedure
was developed which could update the geometry and stresses of
the girder during each construction state. The owchart of the
sequential analysis procedure is depicted in Fig. 9.
The rst static analysis shown in Fig. 9 was performed to generate the sweep in the girder and camber and stresses due to prestressing forces. As mentioned previously, the camber and
stresses are obtained from prestressing forces, dened as the initial
stress conditions. The shape of the initial sweep, dened in the rst
static analysis, is obtained from another previous static analysis

Fig. 9. Flowchart of the 3D nonlinear nite element sequential analysis.

with the self-weight of the girder applying to the lateral direction.

The magnitude of the initial sweep is specied by scaling the

J.-H. Lee / Engineering Structures 42 (2012) 18

maximum lateral deformation obtained from the static analysis to

a target sweep value. Since the sweep and the camber occur prior
to the placement of the girder on the bearing supports, the support
boundary condition in this rst static analysis was assumed to be a
simply supported condition.
The initial sweep in the girder and the camber and stresses induced by prestressing forces are employed to update the geometry
and stress states of the girder for the next analysis. Then the slope
of the bearing support is produced by applying displacement
boundary conditions corresponding to the support slope to the upgraded 3D nite element model. Since the support slope is a stressfree behavior, the stress states in the concrete and prestressing
strands are the same as those dened in the rst static analysis.
Finally, the prestressed concrete girder updated from the previous analysesthe rst static analysis for the initial sweep of the
girder and the effects of prestressing forces and the second static
analysis for the support slope of the girderprovides a reference
conguration of the next 3D nonlinear nite element analysis.
The self-weight of the girder and the thermal loads obtained from
the 2D heat transfer analysis are applied to assess the behavior of
the prestressed concrete girder during construction. This analysis
accounted for the nonlinearity of the geometry and the nonlinear
behavior of the elastomeric bearing pads.

The initial evaluation of the support slope on the behavior of the

prestressed BT-1600 girder was conducted with no initial sweep.
The support slopes chosen for this study were 0, 2.5, and 5.
The angle of 0 represents a perfectly at condition between the
girder and the supports. The maximum value of the support slope
of 5 was chosen based on the maximum measured support slope
of 0.0079 rad. (4.5) on the collapsed girders in Arizona [1].
For a support slope of 5 at both ends of the girder, this study
examined the vertical and lateral responses of the prestressed
BT-1600 girder with increases in initial sweep at 3, 6, 9, 11, and
13 cm. The initial sweep of 3 cm was the sweep tolerance of
3 mm per 3 m length of prestressed concrete beam provided in
the PCI Bridge Design Manual [2].
5.3. Vertical behavior of the prestressed concrete girder

Since the stability of prestressed concrete girders during construction mainly depends on their lateral behavior, the thermal
load involved in this 3D sequential nonlinear analysis was based
on the winter environmental conditions showing the largest lateral
thermal gradients in Atlanta. Among the four AASHTO-PCI sections,
this study selected the 30-m long prestressed BT-1600 girder.

For the geometrically perfect structure, or the 30-m long BT1600 girder with no initial sweep, Fig. 10a shows variations in
the vertical movements due to thermal loads and self-weight for
the support slope of 0, 2.5, and 5. Fig. 10b shows the variations
in the vertical movements of the girder with increases in the initial
sweep from 3 cm to 13 cm with a constant 5 support slope. From
Fig. 10, we can see that the girder, after being installed on the bearing supports, underwent slight increases in vertical movement
with increases in the support slope or the initial sweep. In particular, for the initial sweep of 11 and 13 cm, the 3D nonlinear nite
element analysis stopped at 2 p.m. and at 10 a.m., respectively, as
shown in Fig. 10b. The failure of this numerical analysis was due
to the larger increases in lateral movements with increases in the
initial sweep in the sloped girder. The lateral behavior of the girder
with increases in the initial sweep and support slope will be discussed below.

Fig. 10. Variations in the vertical movements of the 30-m long BT-1600 girder at
mid-span with increases in (a) support slope with no initial sweep and (b) initial
sweep for a 5 support slope.

Fig. 11. Variations in the lateral movements at the mid-height of the 30-m long BT1600 girder web at mid-span with increases in (a) support slope with no initial
sweep and (b) initial sweep for a 5 support slope.

5.2. Structural analyses with the support slope and the initial sweep

J.-H. Lee / Engineering Structures 42 (2012) 18

5.4. Lateral behavior of the prestressed concrete girder

For the lateral behavior of the 30-m long prestressed BT-1600
girder during construction, Fig. 11 exhibit variations in the lateral
movements at the middle height of the girder web with increases
in the support slope and the initial sweep. Increases in the support
slope with no initial sweep, illustrated in Fig. 11a, only slightly
changed the lateral deformations due to the combination of thermal effects and self-weight, as was found in the vertical behavior
of the girder. However, with increases in the initial sweep for a
constant support slope of 5, Fig. 11b reveals increases in the lateral movements at the middle height of the web at mid-span. For
the 30-m long prestressed BT-1600 girder with an initial sweep
of 11 and 13 cm, the lateral movements were about 12 cm immediately after the girder was installed on the sloped bearing support.
The combined thermal effects then increased the lateral movements, including the initial sweep of 11 cm and 13 cm, to
25.7 cm at 1 p.m. and 25.9 cm at 9 a.m., respectively, at the middle
height of the girder web at mid-span, shown in Fig. 11b. After that,
as mentioned previously, the 3D nite element analyses halted due
to error messages of largest increment of displacement in the lateral direction at the top ange of the girder at mid-span and
excessive distortion in solid concrete elements. According to
the messages, the failure of this numerical solution to converge
is an indication of instability in the structure due to the large increase in lateral deformations.

6. Conclusions and discussions

Using the 3D nonlinear nite element analysis, this initial study
evaluated the behavior of a precast prestressed concrete bridge girder during construction, particularly before the placement of the
slab deck and bracing. This analysis included initial sweep in the
girder and the slope of the support, which occur during fabrication,
handling, and construction, and environmental thermal loads to
the girders immediately after resting on bearing supports.
To determine thermal loads, this study rst evaluated vertical
and lateral thermal gradients using seasonal variations in environmental conditions from 30-year (from 1961 to 1990) monthly daily
climatic values for Atlanta. The study then examined the inuence
of girder orientations on the thermal gradients for four AASHTOPCI standard girder sections. The maximum vertical thermal gradients were found in the summer in an eastwest orientation, and
the maximum lateral thermal gradients were found in the winter
in an eastwest orientation. Among the four AASHTO-PCI sections,
the deeper and wider Type-V and BT-1600 sections exhibited the
largest vertical and lateral thermal gradients. Thus, the thermal
loads involved in this study were determined from the summer
and winter environmental conditions for Atlanta.
For a 30-m long prestressed BT-1600 girder, the 3D nite element thermal response analysis was performed which consisted
of two sequential analyses: A 3D static analysis with a simply supported boundary condition and a 3D thermal stress analysis with
elastomeric bearing pads. The bearing pads were modeled as nonlinear springs with the effective vertical stiffness of the bearing
pads. Results of the analyses showed no instability in the girder
due to the combined effects of self-weight and thermal loads without any initial sweep or support rotation.

In the following analysis, to combine thermal effects with fabrication and construction errors, this study developed a 3D nite element sequential nonlinear analysis procedure that accounted for
the changes in the geometry and stress states of the girder. This
analysis included the nonlinear behavior of the girder and the elastomeric bearing pads. The analyses indicated possible instability in
the 30-m BT-1600 girder when lateral deformations due to the
combination of thermal effects, the initial sweep, and the support
slope exceeded about 25 cm at the middle height of the girder,
close to the centroid of the girder cross-section.
In addition, this study proposed a beam model to calculate environmentally-induced vertical and lateral thermal deformations for
simply supported girders. The vertical deformation is based on a
vertical thermal gradient along the depth of the cross-section,
and because the lateral thermal gradients vary from the top to
the bottom anges, the lateral deformation is dened from three
lateral temperature gradients of the middle of the top ange, the
web, and the bottom ange. A comparison showed that the vertical
and lateral movements of the beam were within 0.08 cm (6%) of
those determined by the 3D nite element analysis. Furthermore,
this study proposed simple equations for calculating the maximum
vertical and lateral thermal movements in terms of the span length
of the girders for simply supported four AASHTO-PCI standard
girders located in Atlanta: Type-I, Type-IV, Type-V, and BT-1600
girders.
To provide more generalized conclusions on the lateral behavior
of precast prestressed concrete girders, including the initial sweep
and the support slope during conduction, the development of the
3D and simplied models should include more girders under various environmental conditions.
References
[1] Oesterle RG, Sheehan MJ, Lot HR, Corley WG, Roller JJ. Investigation of red
mountain freeway bridge girder collapse. CTL Group Project No. 262291. CTL
Group, Skokie, IL; 2007.
[2] Precast/Prestressed Concrete Institute. PCI bridge design manual. PCI, Chicago,
IL; 2003.
[3] Mast RF. Lateral stability of long prestressed concrete beams Part 1. PCI J
1989;34(1):3453.
[4] Mast RF. Lateral stability of long prestressed concrete beams Part 2. PCI J
1993;34(1):3453.
[5] American Association of State Highway and Transportation Ofcials. AASHTO
LRFD bridge design specications: 2008 interim revision. AASHTO,
Washington, DC; 2007.
[6] American Association of State Highway and Transportation Ofcials. AASHTO
bridge construction specications: 2006 and 2007 interim revisions. AASHTO,
Washington, DC; 2004.
[7] National Renewable Energy Laboratory. Solar radiation manual for at-plate
and concentrating collectors. NREL, Golden, CO; 1994.
[8] Liu BYH, Jordan RC. The interrelationship and characteristic distribution of
direct, diffuse, and total solar radiation. Sol Energy 1960;4(1):119.
[9] Dufe JA, Beckman WA. Solar engineering of thermal processes. John Wiley &
Sons; 1960.
[10] Kreith F, Kreider JF. Principles of solar engineering. Macmillan Company; 1978.
[11] Jeong JH, Zollinger DG. Finite element modeling and calibration of temperature
prediction of hydrating Portland cement concrete pavements. J Mater Civil Eng
2006;18(3):31724.
[12] Branco FA, Mendes PA. Thermal actions for concrete bridge design. J Struct Eng
1993;199(9):231331.
[13] Abaqus. Dassault Systems Simulia, Providence, RI; 2008.
[14] ACI Committee 435. Control of deection in concrete structures. ACI
Committee Report, Farmington Hills, MI; 1995.
[15] Yazdani N, Eddy SM, Cai CS. Validation of AASHTO bearing stiffness for
standard precast concrete bridge girders. ACI Struct J 2000;97(3):43643.
[16] Georgia Department of Transportation. GDOT bridge and structures design
policy manual: revised July 2009. GDOT, Atlanta, GA; 2005.