M
ost medium-span bridges in the United States are
constructed with precast, prestressed concrete
girders that support a cast-in-place concrete
deck. These girders usually have an upward deflection
(camber) that results from the combined effects of the ec-
centric axial prestressing force and the gravity loads. If the
camber is less than expected, the contractor must increase
the amount of concrete needed to construct the cast-in-
place concrete pad between the top of the girder and the
deck. If the camber is greater than expected, the girder
may interfere with placement of the deck reinforcement.
Differences between expected and actual cambers are even
more critical if the bridge owner rejects the girder for be-
ing out of compliance with the specifications.
■ Two practical methods were also developed using 164 Thermal camber typically has the most important conse-
observations from 24 girders. quences during construction. At that time, the girders are
47.2 in.
8 in.
T3 8 in.
T = 2.7°F
Figure 1. Design temperature gradient. T = temperature at top surface of deck; ∆T = temperature difference over height of girder; ∆T1 = temperature change at top
surface of girder; ∆T2 = temperature change 4 in. below top surface of girder; ∆T3 = temperature change at bottom of girder. 1 in. = 25.4 mm; ˚C = (˚F – 32)/1.8.
of 11 girders.9 Three of these girders were monitored on girders are typically cast at approximately noon on one day
two separate days, resulting in a total of 14 girder-obser- and released the next morning. Weekend girders are typical-
vation days. The camber of the monitored girders varied ly cast on a Friday and released on the following Monday
during each day by 0.70 to 0.95 in. (18 to 24 mm) for morning. Typically, the concrete for these weekend girders
girders ranging in length from 119 to 132 ft (36.3 to 40.2 is more mature at release, even though the curing tempera-
m). These field observations are used later in this paper ture may have been lower, on average, than for a weekday
to calibrate the proposed simple methods for estimating girder. All of the measurements discussed in this paper
thermal camber. were taken approximately two months after the prestress-
ing force had been applied to the girder concrete, by which
Experimental program time the effects of creep, shrinkage, and relaxation were
expected to be constant over the course of any one day.
In this study, two large prestressed concrete girders were
instrumented to quantify daily temperature variations, de- The internal temperatures of the two girders were
velop an understanding of the effects of these variations on measured at a single cross section near midspan at one-
camber, and calibrate theoretical and more practical models minute intervals using fifteen type J thermocouples. The
for predicting thermal camber. The results are summarized changes in camber at midspan of the two girders were
here. Nguyen gives details.10 measured continuously with a deflection sensor in the
fabricator’s storage yard. Figure 2 shows the locations
Table 1 lists key properties for the monitored girders. Both of the thermocouples. More thermocouples were placed
girders were Washington State WF100G girders with an in the top flange than in the web, both because the top
I-shaped cross section and a depth of 100 in. (2500 mm) flange was expected to experience the largest variation in
(appendix 5.6-A4-10, Washington State Department of temperature during the day and because the thin flange
Transportation Bridge Design Manual (LRFD).11 Weekday tips had the largest surface/volume ratio, which increases
Girder Girder mark Cast date Measurement date Depth, in. Length, ft f c' at release, psi f c' at 28 days, psi
Weekend girder H6B May 18, 2012 July 26, 2012 100 172.7 9215 11,260
Weekday girder H8A May 29, 2012 July 25 and 26, 2012 100 164.0 8250 12,430
Note: f ' = compressive strength of concrete. 1 in. = 25.4 mm; 1 ft = 0.305 m; 1 psi = 6.895 kPa.
c
UW
MW
2 ft 2 in.
LW
6 in. 1 ft 10 in.
BC1
BL
BC2 BR
5 in.
Figure 2. Thermocouple locations. Note: B = bottom of girder; C = center of girder; L = left side of girder; LW = lower web; R = right side of girder; T = top of girder;
UW = upper web.
the possibility of transverse (as well as vertical) variations day, when sunlight was strongest, was higher at the bottom
in temperature. left sensor than at the bottom right or bottom center sensor
locations.
Measured temperature
histories In the late afternoon, the internal temperatures in the top
of the girder dropped rapidly as the solar radiation became
Figure 3 shows the measured internal temperature histo- weaker and the air temperature decreased, whereas the
ries for girder H8A for 22 hours and for girder H6B for 9 temperatures in the web and bottom portions in the girders
hours. A problem with electrical power supply prevented cooled more slowly. This difference can be explained by
the data logger from monitoring girder H6B for a longer the fact that the bottom region of the girders was affected
time. by heat from the ground and was less exposed to wind.
During the day, the ground was heated by the sun. How-
The data follow expected trends. For both girders, the tem- ever, the ground changes temperature more slowly than
perature profile over the height of girders was nearly uni- does the air, so the bottom flange continued to warm in the
form until about 10:00 a.m. During the late morning and evening due to heat from the ground.
afternoon, the temperature in the top of the flange exposed
directly to sunlight increased rapidly until it reached a peak Figure 4 shows the corresponding vertical temperature
at about 4:00 p.m. At this time, the temperature change in profiles of the girders at various times. For each elevation
the top flange was approximately equal to 40˚F (22˚C). In that had more than one thermocouple (for example, the
contrast, the temperatures in the web and bottom flange flanges), the average temperature is plotted at that eleva-
changed less because these locations were shaded by the tion. The temperatures varied most over the height of the
top flange and by adjacent girders in the storage yard. This girder at approximately 4:00 p.m.
situation is similar to that which a girder would experience
during installation, at which time the flange, adjacent gird- Calculation of thermal
ers, and deck formwork would similarly shadow the web camber changes
and bottom flange. from temperature profiles
The left side of the web of girder H6B was also exposed Thermal camber can be predicted from the measured in-
to direct sunlight, which might be similar to the exposure ternal temperature profile using the principles of mechan-
condition for an exterior girder in a bridge. Figure 3 plots ics and an assumed value of the coefficient of thermal
that exposure data, in which the temperature during the expansion. The total strain at any point will consist of
Girder H6B
Figure 3. Internal temperature histories. Note: B = bottom of girder; C = center of girder; L = left side of girder; LW = lower web; R = right side of girder; T = top of
girder; UW = upper web. ˚C = (˚F – 32)/1.8.
a thermal and a stress-related component (Fig. 5). The cur- Stresses will develop over the depth of the member if the
vature at each point along the girder can be computed by thermal strain gradient is nonlinear. The total strain εtot is
finding the profile of total strain that satisfies the following the sum of the free environmental strain εe, which consists
assumptions: here of just the thermal strain but more generally may also
include other components, such as shrinkage, plus the
• Plane sections remain plane. mechanical strain εm, which is given by stress σ divided by
Young’s modulus E:
• The free thermal strain is constant across the width at
any elevation.
(3) (5)
(4)
(7)
For a statically determinate girder, axial-force equilibrium
is determined by Eq. (5). Again this equation can be rearranged to give Eq. (8).
These equations can be combined to give Eq. (9). Once ε0 and ϕ are known, the total strain field can be cal-
culated from Eq. (3), and the stresses from Eq. (4).
(15)
(10)
where
(11)
L = length of the girder
where The negative sign in Eq. (15) results from the fact that the
calculated deflection is negative, but in practice, camber is
Atr = area of the transformed section taken as positive upwards.
Itr = moment of inertia of the transformed section The measured temperature profiles were used to predict
camber changes in girders H6B and H8A over the period
If different materials are involved, the centroid is that of that they were monitored. For simplicity, the gross (rather
the transformed section. The matrix equations decouple to than the transformed) section properties were used in the
give Eq. (12) and (13). calculations. According to the Federal Highway Adminis-
tration, the coefficient of thermal expansion of portland ce-
ment concrete ranges from about 4.4 × 10-6 to 6.6 × 10-6/˚F
(12) (7.9 × 10-6 to 12.0 × 10-6/˚C).12 Wight and MacGregor
found that the coefficient of thermal expansion for normal-
weight concretes made with siliceous aggregates ranges
from 5.0 × 10-6 to 7.0 × 10-6/˚F (9.0 × 10-6 to
12.6 × 10-6/˚C).13 In their calculations, Barr et al. assumed
(13)
a value of 6.5 × 10-6/˚F (11.7 × 10-6/˚C) for this coeffi-
cient.14
The integrals on the right side of Eq. (12) and (13) are
most conveniently evaluated numerically because in most Figure 6 compares the measured camber histories with the
cases the girder width b(y) cannot be defined with a single values calculated using Eq. (12) to (15) and the measured
equation. thermal profiles. Two calculated curves are shown, cor-
responding to coefficients of thermal expansion of 5.5 ×
The calculations simplify in two special cases. First, if E is 10-6 ˚F (9.9 × 10-6/˚C) and 6.5 × 10-6/˚F (11.7 × 10-6/˚C).
the same throughout the girder cross section, Eq. (12) and The calculated and measured cambers rise and fall at
(13) can be divided through by E, which then no longer similar rates and peak at similar times. The calculated
influences the calculated camber. Second, if the environ- curves matched the measured ones most closely when the
mental strain gradient is linear, the right side of Eq. (13) coefficient of thermal expansion of the concrete was taken
simplifies to -EItrα∆T/h, where α is the coefficient of ther- as 5.5 × 10-6/˚F. The small discrepancies can be attributed
mal expansion and ∆T is the temperature difference over to the following:
the height of the girder. Then Eq. (13) reduces to Eq. (14).
• the assumed value for the coefficient of thermal expansion
Figure 6. Comparison between calculated and measured thermal cambers. Note: α = coefficient of thermal expansion. 1 in. = 25.4 mm.
• measurement errors For the effective strain profile shown in Fig. 7, the corre-
sponding thermal curvature is Eq. (16).
• variations in E over the cross section
The temperature history camber model is based on the as- To calculate the camber at a particular time during the
sumption that the temperature difference between the top day, ∆Teff (t) is needed. In the temperature history camber
and bottom of the girders can be related to variations in the
ambient temperature. Site-specific data should be used when
available. In their absence, the ambient temperature histories
from a nearby meteorological station should be used. εTH,eff
(20)
(17)
The measured camber values Mi were compared with the A second, simpler model, called the peak temperature
calculated camber values Ci that were obtained using camber model, was developed. This method does not
Girder Girder mark Girder type Location Depth, in. Length, ft Length/depth
Note: Tmax = maximum temperature during a day; Tmin = minimum temperature during a day. 1 in. = 25.4 mm; ˚C = (˚F – 32)/1.8.
(22)
Figure 8. Optimized values of calibration factor A0 of all girders.
Figure 12. Peak temperature camber model: comparison of measured and calculated thermal camber changes for Minnesota and Georgia State girders.
Note: 1 in. = 25.4 mm.
5. Shushkewich, K. W. 1998. “Design of Segmental Errorrms = root mean square camber error over time of a
Bridge for Thermal Gradient.” PCI Journal 43 (4): girder
120–137.
f c' = compressive strength of concrete
6. Barr, P. J. 2000. “Consistent Crudeness in Prestressed
Concrete Girder Design.” PhD diss., University of h = depth of the girder
Washington, Seattle, WA.
ha = asphalt thickness
7. AASHTO. 1994. AASHTO LRFD Bridge Design
Specifications. 1st ed. Washington, DC: AASHTO. Itr = moment of inertia of the transformed section
Tmax = maximum temperature during a day ∆Teff = effective temperature difference between top
and bottom of girder
Tmin = minimum temperature during a day
∆Teff(t) = instantaneous effective temperature difference
y = vertical coordinate, measured positive down- between top and bottom of girder
wards from origin
ε0 = total strain at origin
Y = vertical coordinate, measured upwards from
a point located Y0 below the concrete deck εe = free environmental strain
surface
εm = mechanical strain
Y0 = distance from top surface at which temperature
change is zero εtot = total strain
John Stanton, PhD, PE, is a This paper was reviewed in accordance with the Precast/
professor of civil engineering at the Prestressed Concrete Institute’s peer-review process.
University of Washington. He
holds degrees from the University Reader comments
of Cambridge in the United
Kingdom, Cornell University in Please address reader comments to journal@pci.org or
Ithaca, N.Y., and University of Precast/Prestressed Concrete Institute, c/o PCI Journal,
California, Berkeley, and has worked in practice in 200 W. Adams St., Suite 2100, Chicago, IL 60606. J