Professional Documents
Culture Documents
DOI: 10.1002/stc.2042
RESEARCH ARTICLE
1
Department of Civil Engineering, Wasit
Summary
University, Wasit, Iraq
2
Department of Civil Engineering,
The structural performance of bridge structures is temporal and is mainly con-
Gaziantep University, Gaziantep, Turkey trolled by the types of the applied loads. To continuously observe the structural
performance of bridges, structural health monitoring sensors that include among
Correspondence
Sallal R. Abid, Department of Civil many temperature sensors are used. The impact of nonuniform temperature distri-
Engineering, Wasit University, Wasit, Iraq. butions in bridge girders due to the environment thermal loads has been recog-
Email: sallal@uowasit.edu.iq
nized by former researchers and bridge design codes. To evaluate these and
other effects on the structural behavior of bridge structures, many field and exper-
imental structural health monitoring studies were carried out. However, more
researches are required to investigate the temperature distributions in other girder
configurations. This work is directed to investigate the impact of air temperature
and solar radiation on temperature gradient distributions in concrete‐encased
composite girders. For this purpose, an experimental concrete‐encased steel girder
segment was instrumented with thermocouples and other sensors. The experimen-
tal data recording continued for 6 months during the hot and cold seasons. Further-
more, a thermal finite element (FE) parametric study was conducted to investigate
the effect of the girder size. The test results showed that the vertical and lateral
temperature gradient distributions and the variation of the temperature gradients
with time are controlled by the amount and location of the received solar radia-
tions. The FE analysis showed that the daily temperature variations are higher in
smaller girders, whereas the temperature gradients are smaller than in larger
girders. Moreover, the FE results showed that the thickness of the girder's concrete
members has an important impact on temperature gradients and temperature
distributions.
K EY WO R D S
composite girder, solar radiation, temperature distribution, temperature gradient, thermal strain
1 | INTRODUCTION
As bridges are generally built in open areas and subjected to the daily and seasonally fluctuation of the environment thermal
loads, they are subjected to undesirable deformations and stresses.[1–3] The solar radiation causes the temperature to vary
nonlinearly through the depth and the width of the girders. Such variation leads to additional stresses (equilibrating stresses)
to prevent the nonlinear deformation of the girder cross section. Moreover, the average temperature of the whole structure
changes during the day and night cycles causing longitudinal movements or continuity stresses based on the end conditions.[4]
Struct Control Health Monit. 2017;e2042. wileyonlinelibrary.com/journal/stc Copyright © 2017 John Wiley & Sons, Ltd. 1 of 23
https://doi.org/10.1002/stc.2042
2 of 23 ABID ET AL.
According to previous studies, the stresses that can be referred to as thermal equilibrating stresses can be comparable to those
caused by gravity loads and can result in concrete cracking.[4–6]
Many field studies[7–14] were conducted during the last 30 years to better understand the temperature distributions and their
structural effects on concrete, steel, and composite bridge structures. Generally, field studies are sponsored by the transportation
agencies as parts of complete long‐term structural health monitoring (SHM) projects. Due to their importance, bridges became a
primary focus of SHM techniques and damage detection methods.[15,16] The long‐term SHM systems include the instrumenta-
tion of parts of the bridge, including the superstructure, by many measurement sensors and data acquisition systems. Due to the
high cost of bridge instrumentation and the enormous sensors required for such measurements, smaller size experimental girder
segments were used by many previous studies[17–24] to evaluate the effect of the environmental thermal loads on the temperature
distributions, structural performance, and on the reliability of the obtained data from the SHM systems. Other studies[6,19,25–34]
used analytical, numerical, and statistical techniques for this purpose.
In previous studies,[20,21,33] a full‐scale experimental concrete box‐girder segment was instrumented with 64 thermocouples
and a weather station to analyze the temperature distributions in such superstructures. The thermocouples were distributed in
different locations along the webs and flanges of the girder to evaluate the vertical and lateral temperature gradients. Based
on the analysis of the collected data, design gradients and equations were introduced. Kulprapha and Warnitchai[18] constructed
and tested experimental 8.0 m length continuous prestressed concrete bridge girders. The tested girders have two spans each of
4.0 m. The cross‐sectional dimensions were scaled down with approximately 1/10 to 1/20 scale factors where the section depth
was 300 mm only. Electrical resistance strain gages were used to measure the strain in concrete, reinforcing bars, and
prestressing tendons. Type K thermocouples were used to monitor the temperature distributions over the experimental girders.
Lee and Kalkan[17] measured the temperature variation in a 1.6 m depth and approximately 1.5 m length experimental
prestressed girder segment using 18 thermocouples distributed along the web and the flanges. On the basis of the recorded tem-
perature measurements and thermal loads data, they proposed formulas for maximum vertical and lateral temperature differen-
tials. Wang et al.[22] used experimental model of a box‐girder arch bridge to evaluate the effect of nonlinear temperature
gradients on such bridges. The experimental data of temperature from 1‐day measurements were used to verify finite element
(FE) studies. Xia et al.[23] conducted an experimental work on a simply supported slab specimen to investigate the effect of tem-
perature gradients on the structural vibration characteristics. The slab was 3.0 m span, 0.12 m thickness, and 0.8 m wide and
was instrumented by 11 thermocouples and 14 accelerometers. Following researchers[34] used the experimental data of Xia
et al.[23] to numerically study the impact of environmental thermal loads on the structural frequencies of T‐beam and box girder
bridge structures. Liu et al.[24] carried out 3‐day temperature measurements on an H‐beam steel specimen under solar radiation.
The web depth and flanges width of the specimen were 200 mm, whereas the length of the specimen was 500 mm. The spec-
imen was instrumented with10 temperature sensors.
The current bridge design codes recognize the thermal actions in bridge girders and deck slabs. To overcome such actions,
these codes provide design procedures using temperature gradient models along the depth of the superstructure. Despite that
these codes do not explicitly specify the depth limitations for the use of these gradient, the details of these design gradients have
some restrictions that limit the minimum applicable depth of the section, whereas no values are specified for the upper limit of
the superstructure depth. The AASHTO LRFD Bridge Design Specifications[35] considers the proposed design temperature gra-
dient for members as shallow as 400 mm. The British standard BS 5400‐2[36] recognizes concrete girders from composite ones.
For concrete girders, the design gradient is applicable for sections as shallow as 350 mm; whereas for composite girders, the
design gradient is applicable for sections not shallower than 600 mm. The European standard EN 1991‐1‐5[37] introduces
two procedures to evaluate temperature gradients across the depth of the bridge superstructure. The first assumes a linear tem-
perature difference between the top and bottom surfaces of the section. This procedure does not mention any limitation for the
size of the girder. On the other hand, the second procedure is that of the BS 5400‐2.[36] The New Zealand Bridge Design Manual
SP/M/022[38] introduces a nonlinear gradient model for superstructures with depth not less than 1.4 m. However, it also gives
recommendation for the application of this gradient model to sections shallower than 1.4 m.
The above literature review shows that different size and shape of experimental specimens have been used as experimental
SHM girders or decks to evaluate the effect of solar radiation and air temperature on the distributions of temperature and its
consequent effects on the structural behavior of different types of girders. However, no previous study has investigated temper-
ature gradient distributions in concrete‐encased steel girders. Also, it is obvious in the introduced review that the proposed tem-
perature gradient models are considered applicable for wide range of girder sizes, which is quite questionable.
In this research, a composite concrete‐encased steel girder segment was instrumented with several thermocouples and other
sensors to evaluate the temperature gradient distributions in composite girders. The variation of these gradients with the overall
size of the girder, the depth of the girder, and the thickness of the flanges and the web were numerically studied using a FE
thermal model, which was verified with the experimental temperature measurements. Moreover and as discussed earlier, as
ABID ET AL. 3 of 23
temperature effects are dependent of the time of the day and the year, SHM system was used to observe the long‐term impact of
temperature variation. On the other hand, the reliability of the different data acquired from the SHM acquisition systems is also
affected by this temperature variation. To estimate and correct this effect, studies that use analytical and numerical techniques,
which are verified with field or experimental data, were carried out and are still required. Therefore, the experimental data pre-
sented in this study can also be used as a verification tool for future diagnostic studies.
2 | E X P E R I M E N TA L P RO G R A M
To study the influence of the ambient thermal loads on composite girders, an experimental work was directed using an exper-
imental composite girder segment. The time‐dependent ambient thermal loads include primarily the thermal solar radiation in
addition to the temperature and the speed of the contacting air. In this section, the preparation, geometry, and instrumentation of
the experimental girder segment are explained.
FIGURE 2 The experimental girder segment: (a) the formwork and the distributed sensors of the top flange and (b) the final view of girder and the
insulation boards
thus at 250 mm from the insulated edges. The coordinates of the thermocouples are listed in Table 1; in addition, Figure 2a
shows the thermocouples distributed in the top flange. In Table 1, TC refers to the thermocouples embedded in the concrete,
whereas TS refers to the thermocouples that measure the temperature of the steel section. The origin point (0, 0) in Table 1
was considered as the center of the bottom surface of the girder. The third group of sensors consisted of two vibrating wire strain
gages to measure the thermal strains of the concrete. The strain gages were placed on the reinforcing mesh in two perpendicular
directions as shown in Figure 2a.
After concrete casting, all wires from thermocouples and strain gages were grouped and connected to the data acquisition
system, which composed of the data logger, multiplexers, and vibrating wire strain analyzer from Campbell Scientific. The data
from all sensors were collected from the first of July until the end of December 2015 at time intervals of 30 min. However,
during this period, there were some faults that interrupted the data collecting for discrete intervals.
3 | E X P E R I M E N TA L T E M P E R AT U R E S A N D STR A I N S
Three main factors affect the temperature and hence the temperature gradients in bridge girders. These are the daily maximum
and minimum air temperatures, the solar radiation intensity, and the movement of the sun from rise to set. Based on these influ-
ential factors, the distributions of vertical and lateral temperature gradients are different in summer than in winter. These distri-
butions are studied in this section. To optimize the analysis of the experimental records, 10 days from summer and 10 days from
winter were selected to visualize the thermal effects of the environmental thermal loads on the experimental composite girder.
From July 1 to 10 were selected as the summer representative days, whereas from December 21 to 30 were selected to represent
winter. The daily variations of the temperature gradients and thermal strains in addition to the distributions of the maximum
vertical and lateral temperature gradients are discussed in this section for the two periods.
The maximum temperature gradient in this paper refers to the difference between the girder's maximum and minimum tem-
peratures along a specified section. All sections were considered along the position of thermocouples at the mid‐length cross
section. The vertical temperature gradient was calculated from the concrete and steel thermocouples passing through the vertical
centerline of the cross section, hence all thermocouples having zero x‐coordinates as listed in Table 1. These thermocouples,
from top to bottom, are TC1 to TC3, TS1 to TS3, and TC10. The top‐flange lateral temperature gradient was calculated from
ABID ET AL. 5 of 23
the thermocouples distributed horizontally along the centerline of the top flange, thus, TC3 to TC7. On the other hand, the web
lateral temperature gradient was calculated at the central depth of the cross section from thermocouples TC11, TS2, and TC12.
3.1 | Solar radiation and air temperature during the selected periods
Figure 3a shows the hourly variations of air temperature and solar radiation from July 1 to 10, whereas Figure 3b shows these
variations from December 21 to 30. Figure 3a shows that the daily maximum temperatures were close during the first week of
July with an edge of approximately 35 °C, whereas it decreased in July 8 and 9 to less than 32 °C and increased in July 10 to
slightly more than 36 °C. Another notice is that the difference between the daily maximum and minimum was high during the
first 5 days of July, whereas it was low during the following 4 days. The variation of the daily maximum solar radiation was
minimal during this period where values around 1,000 W/m2 were recorded for the 10 days.
Figure 3b shows that solar radiation curves are not smooth during some of the 10 days from December 21 to 30, which
reflects the existence of a variable cloud cover during these days. In general, the daily maximum hourly solar radiation in this
period ranged from approximately 510 W/m2 to approximately 580 W/m2, except in December 29, where it was approximately
300 W/m2. The figure also shows that the daily minimum air temperature during this period was mostly less than 0 °C, whereas
the daily maximum temperature exhibited higher fluctuation.
FIGURE 3 Ten‐day variation of the environmental thermal loads: (a) July 1 to 10 and (b) December 21 to 30.
6 of 23 ABID ET AL.
FIGURE 4 Hourly gradients from July 1 to 10: (a) vertical gradient; (b) lateral top‐flange gradients; and (c) lateral web gradients
about 7:00 in the morning, after which the gradient increases sharply (quick rising) until reaching the maxima around 13:00.
The temperature gradient is the highest from midday to about 14:00 and starts the sharp descending phase (dropping region)
after 15:00. The gradient continues to decrease after sunset until starting the next day's loop at the sunrise.
After sunrise, the solar radiation is low and is almost perpendicular to vertical surfaces and thus the radiation reaching the
top surface is low to cause serious increase in the gradient. Therefore, the gradient increases slowly during this period as shown
in the first ascending phase in Figure 4a. After about 2 hr, the altitude of sunrays becomes higher, leading to an effective
worming of the top surface of the girder. This altitude increases and solar radiation increases sharply after 9:00, leading to much
higher vertical temperature gradients, which explains the second ascending phase of the gradient. The sun movement reverses
after midday, and the sunrays altitude starts to decrease showing a significant decrease in the worming budget after 15:00.
ABID ET AL. 7 of 23
Within the sunset, the cooling phase starts by two means. The first is the reradiation of heat from the girder to the surrounding
ambient by long waves, whereas the second is the convection from the surrounding cold air. The cooling phase continues
decreasing the temperature of the top surface and hence decreasing the vertical temperature gradient until the next day's
morning.
The lateral temperature gradients measured at the northern edges (Figure 4b,c) show different behavior from that of
the vertical temperature gradient shown in Figure 4a, whereas those measured at the southern edges show an agreement
with the behavior of the vertical gradient. Lateral temperature gradients at the northern edges exhibit two periods of
rising and dropping during the single day: major rising after sunrise and minor rising before sunset. Mentioning that
in summer, the sun rises from the northeast and sets at the northwest would explain this behavior. During these two
periods, the altitude angle of sunrays is too low. Thus, solar radiation is concentrated on the northern edge of the top
flange and the northern surface of the web leading to the rising of the temperature of these surfaces compared to the
cold interiors, hence, increases the lateral gradients there. As shown in Figure 4a,b, as temperature gradients at the north-
ern surfaces start dropping, those at the south surfaces start rising. This is because the sun moved from the sunrise posi-
tion toward the south.
In general, the recorded values of lateral temperature gradients were much lower than those of vertical gradient in summer.
The recorded maximum lateral temperature gradient along the top slab for the period from July 1 to 10 was 5.9 °C. Similarly,
the maximum lateral temperature gradient across the web was only 4.9 °C, whereas the recorded maximum vertical gradient was
15.4 °C.
FIGURE 5 Hourly gradients from December 21 to 30: (a) vertical gradient; (b) lateral top‐flange gradients; and (c) lateral web gradients
during the early morning hours, whereas it is completely shaded during the mid‐day hours at which the solar radiation is
maxima. Similarly, the lower surface of the girder is almost completely shaded along the whole day hours. However, reflected
radiations from the ground led to a slight increase in its temperature compared to thermocouple TS3 (within the bottom flange),
at which the temperature was the lowest, and hence the vertical gradient was zero.
In December 25, vertical surfaces receive considerable amounts of solar energy during the whole day, which leads to the
increase of the temperature of the central thermocouple. However, due to the low conductivity of concrete, the temperature there
still slightly less than the temperature of the top surface thermocouple. As a result, the winter maximum gradient distribution
has three inflection points, whereas it is only one in summer as shown in Figure 6.
ABID ET AL. 9 of 23
FIGURE 7 Distributions of vertical temperature gradients at specific times (a) in July 5 and (b) in December 25
10 of 23 ABID ET AL.
as shown in the gradient distribution at 12:00. As the sun moves from its highest position in the sky, the temperature of the top
surface starts decreasing showing lower temperature gradients with time as shown in the distributions at 15:00 and 18:00. Dur-
ing the night hours, the temperature of the top surface decreases whereas the interiors of the girder keep the most of the heat that
was conducted during the day. As a result, negative vertical gradients form as shown from the gradient distribution at 21:00.
The bottom surface also shows positive and negative gradients as shown in Figure 7a. This is due to the heating from the
reflected radiations from the ground and the direct convection during the cold hours. The central thermocouple (TS2) showed
the zero gradients at all time steps except at those with high positive temperature gradients (12:00 and 15:00) where the zero
gradients occurred within the bottom flange (TS3).
In December 25 and as shown in Figure 7b, similar changes occurred in the vertical temperature gradients but with some
differences. In December 25, the values of the negative gradients are comparable to the positive gradient values; whereas in July
5, the negative gradients are too small compared to positive gradients. Another difference is that in December 25, the gradients
(positive and negative) at the bottom surface of the girder are not small compared to those at the top surface. On the other hand,
in July 5, the positive gradients at the bottom surface are too small compared to the top surface's gradients, whereas they are
almost equal for negative gradients. The shape of the positive vertical temperature gradient in December 25 is also different
from that of July 5 as discussed in the previous section.
FIGURE 8 Distributions of lateral temperature gradients along the top flange at specific times (a) in July 5 and (b) in December 25
ABID ET AL. 11 of 23
midday hours, the temperatures of the top surface thermocouples TC6 and TC7 become higher than edge temperatures, whereas
the temperature of the central thermocouple TC3 is lower because of the continuity with the colder concrete of the web. As a
result, the distribution at 15:00 forms. This distribution reverses during the night hours due to the convection cooling at the
exposed surfaces and the long‐wave reradiation as shown in the distributions at 21:00.
In December 25, the distributions of lateral temperature gradients along the top flange are too steady compared to those in
July 5 as shown in Figure 8b. As the sun rises from the southeast and continues moving to the south of the equatorial plane, the
temperature of the southern edge of the top flange is always higher than that of the northern edge during the day. Moreover, the
significant changes of the gradient take place only at edge thermocouples; whereas in contrary to the distributions in July 5, TC5
and TC6 showed no significant variations compared to the central thermocouple TC3.
The lateral distributions across the web shown in Figure 9 are simple and similar to those along the top flange in December
25. It is obvious in Figure 9a that the highest temperature gradients in July 5 occur at the northern surface, whereas it is clear in
Figure 9b that the gradients at the southern surface are higher in December 25.
FIGURE 9 Distributions of lateral temperature gradients across the web at specific times (a) in July 5 and (b) in December 25
12 of 23 ABID ET AL.
FIGURE 10 Ten‐day variation of thermal strains: (a) July 1 to 10 and (b) December 21 to 30
The nonlinear temperature distributions shown in Figure 6 cause free nonlinear thermal strains that try to deform the
section nonlinearly. However, the actual thermal strain distribution is linear because the section should keep plane after
deformation. As a result, equilibrating thermal strains form. These thermal strains are compression along the hotter regions
and tension along the colder regions. As shown in Figure 10a,b, the strains increased from zero at the starting time to tensile
strains during the cooling hours where the vertical temperature gradient was negative. Starting from the first shining hours,
the thermal strains changed from tension toward compression as the temperature of the top surface increased. The
compression strains increased during the day hours reaching maximum daily values during the midday hours. After which,
the thermal strains reversed toward zero strain as the temperature of the top flange decreased. This cycle is repeated for
the next 9 days, but with different strain values depending on the temperature gradient distributions of these days compared
to the starting time.
A maximum compression strain of approximately 170 microstrains and a maximum tensile strain of approximately 40
microstrains were recorded during the first 10 days of July. During the last 10 days of December, the recorded maximum
compression and tensile strains were approximately 110 and 40 microstrains, respectively. It is worth to remind that these strains
are change of strains from the initial strain values.
4 | F I N I T E E L E M E N T TH E R M A L M O D E L L I NG
2
∂ T ∂2 T ∂2 T ∂T
k þ þ ¼ ρC p : (1)
∂x2 ∂y2 ∂z2 ∂t
ABID ET AL. 13 of 23
In the above Fourier equation, T refers to the temperature at any point in the girder at any time, t, whereas k is the thermal
conductivity coefficient of the material (concrete and steel) in W/m°C, which is equal in all directions for isotropic materials, ρ is
the density of material in kg/m3, and Cp is the specific heat in J/kg°C.
The boundary loads on the surfaces of the girder can be given as
∂T ∂T ∂T
k nx þ k ny þ k nz þ q ¼ 0: (2)
∂x ∂y ∂z
The terms nx, ny, and nz are the direction cosines of the normal vectors,[39] whereas the term q represents the sum of all ther-
mal loads on the boundaries in W/m2, which includes the convection qc, the total solar radiation qs, and the long‐wave reradi-
ation qre, which are given by
q ¼ qc þ qs þ qre : (3)
The convection between the atmospheric air and the exposed surfaces of the girder is represented in Equation 4 and is
dependent of air temperature Ta, surface temperature Ts, and the convection coefficient hc, which is given in Equation 5 and
is dependent of the wind speed w.
qc ¼ hc ðT s −T a Þ; (4)
where
The total solar radiation qs refers to the sum of three components. The direct solar radiation qsr (beam and diffused),
which is represented by Equation 7, the ground reflected radiation qgr, and the surface‐to‐surface radiation among the girder
surfaces qss.
qsr ¼ aI s ; (7)
qgr ¼ aI g ; (8)
where Is is the solar radiation received by any horizontal surface on the ground, whereas a is the absorptivity of the
girder surfaces. Ig is that fraction of the solar radiation reaching and reflected by the ground and other objects around
the girder.
1−cos δ
I g ¼ ag I s ; (9)
2
where ag is the reflection coefficient of the ground (albedo), whereas δ is the surface inclination angle.
The term qre describes the cooling of the girder via heat emitting by long‐wave radiation. This term changes nonlinearly
based on the difference between the quartic temperatures of the air and the girder surfaces.
qre ¼ ϵβs T 4s −T 4a (10)
In which, ϵ is the emissivity of girder surfaces and βs is the Stefan–Boltzmann constant, which equals 5.67 × 10−8W/m2K4.
The size of the elements was chosen to be fine enough to capture accurate temperatures. In the mesh of the girder, the total
number of tetrahedral elements was approximately 21,000 elements, whereas the triangular surface elements were approxi-
mately 6,000. The hemi‐cube method of COMSOL was used to calculate the mutual view factors between the surfaces of
the girder to evaluate the surface‐to‐surface radiation. This method is accurate because it considers the shadowing effect of
the different parts of the girder.
Depending on the literature,[33] the material thermal properties were selected carefully to efficiently simulate the thermal
FE analysis of the composite girder. Table 2 lists the material properties that were used. In addition, the used emissivity
and the absorptivity of concrete were 0.85 and 0.5, respectively. The initial time of the FE thermal analysis of the girder
was selected to be the midnight, at which the temperature of the whole girder was considered equal to air temperature at
that time. This time was selected because the temperature differences between all parts of the girder are small after
several cooling hours. The analysis was started 48 hr before the target day to eliminate the effect of the supposed uniform
temperature at the initial time.
5 | S I Z E E F F E C T O N T E M P E R AT U R E A N D T E M P E R AT U R E G R A D I E N T S
As reviewed in the introduction section, the design temperature gradients of the bridge design codes are considered appli-
cable for wide range of section sizes. However, it was disclosed in previous researches[2,3,41] that the size of the section
matters and that the available design gradients may result in underestimated or overestimated stresses for large sections or
special configuration sections. From this review, it can be said that the depth of the girder and the thickness of the
flanges and web should have an influence on the temperature gradient distributions. In this section, a parametric study
is carried out to investigate the effect of the size of the girder using the verified FE model. This parametric study is
subdivided into three studies. In the first, the overall section sizing is studied so that both section depth and member
thicknesses are increased in similar rates keeping the same aspect ratios. This parametric study investigates the combined
effect of the girder depth and the thickness of its flanges and web. In the second parametric study, the effect of the girder
depth is studied using fixed thickness of web and flanges, whereas the third investigates the effect of the thickness of the
web and flanges for a fixed depth section. In all parts of the FE parametric study, the environmental data of July 5, 2015,
were used.
TABLE 2 Concrete and steel thermal properties used in the finite element thermal analysis
FIGURE 11 Finite element (FE) versus experimental 24‐hr variation of temperature of specific thermocouples
Dimensions (mm) A B C D E a b c
FIGURE 13 Size effect on 24‐hr temperature distributions at (a) top; (b) bottom; (c) mid‐height; and (d) north edge of the composite girder
ABID ET AL. 17 of 23
FIGURE 15 Twenty‐four‐hour temperature distributions for different girder depths with 150 mm thick web and flanges at (a) top; (b) bottom; (c)
mid‐height; and (d) north edge of the composite girder
ABID ET AL. 19 of 23
of the steel web of the experimental, double, and triple size girders at the time of maximum gradient (13:00) are 35, 31.3, and
29.9 °C, respectively. This vertical temperature distribution leads to the increase of the vertical temperature gradient as the size
of the girder increase as shown in Figure 14.
FIGURE 16 Vertical temperature gradients for different girder depths with 150 mm thick web and flanges
20 of 23 ABID ET AL.
FIGURE 17 Twenty‐four‐hour temperature distributions for different web and flanges thicknesses at (a) top; (b) bottom; (c) mid‐height: and
(d) north edge of the composite girder
ABID ET AL. 21 of 23
FIGURE 18 Vertical temperature gradients for different thicknesses of web and flanges
temperature. The results also showed that the thickness of the flanges and web has more significant effect than that of
the girder depth.
6 | CONCLUSIONS
As one of the temporal loads that affect the long‐term structural performance, temperature variation was generally one of the
major foci of bridges SHM programs. Aiming to investigate the temperature distributions in concrete‐encased steel girders
and the effect of the girder size on these distributions, an experimental work was carried out and a FE parametric study was
conducted. The vertical and lateral temperature gradient distributions were determined from thermocouples instrumented in dif-
ferent locations in an experimental concrete‐encased steel girder segment. The thermocouple records in addition to strains and
other measurements were continued from the start of July to the end of December. The FE thermal analysis was used to inves-
tigate the effect of the depth of the girder, the thickness of the web and flanges, and the combined effect of them on temperature
and temperature gradient distributions.
The experimental results showed that the variations and the magnitudes of the maximum vertical and lateral temper-
ature gradients are highly dependent on sun movement and its altitude in the sky, and hence on the season of the year
and on the location of the structural element with respect to sun movement. High vertical temperature gradients reaching
15 °C were recorded in July, whereas the recorded lateral temperature gradients were less than 6 °C. In contrary, during
December, lateral temperature gradients were more significant than vertical gradients. The recorded maximum lateral tem-
perature gradient in December was more than 12 °C, whereas the recorded maximum vertical temperature gradient was
only 3.2 °C. The distributions of vertical and lateral temperature gradients along the web, along the top flange, and across
the web were also found significantly different between July and December. The variation of thermal strains with time
from compression to tension is directly related to the distribution of the vertical temperature gradient and its variation
with time.
The FE model of the girder showed good capability to simulate the heat conduction, convection, and radiation in the girder
and with the surrounding environment. For the 15 experimental thermocouples, the maximum absolute error with the FE tem-
peratures and along the 24 hr was 3.2 °C in July 5 and 2.0 °C in December 25.
To study the effect of the girder size and using the verified FE model, the girder size was doubled and tripled compared to
the experimental girder keeping the same geometrical aspect ratios. The comparisons between the three girders showed that
smaller girders suffer higher daily temperature fluctuation. At the central depth of the composite girder within the steel section,
the daily maximum temperature difference between the three girders was 7.8 °C. On the other hand, the maximum vertical tem-
perature gradient was higher for larger girders. The maximum vertical temperature gradients of the experimental, double, and
triple size girders were 14.0, 16.7, and 18.5 °C, respectively.
Another FE parametric study was directed to investigate the particular effect of the girder depth. For girders with depths of
500, 1,000, and 1,500 mm, but with constant thickness of web and flanges of 150 mm, the maximum daily temperature vari-
ation was only 2.1 °C, whereas the maximum difference between the maximum vertical temperature gradients of the three
girders was less than 0.7 °C.
22 of 23 ABID ET AL.
A third parametric study using 1,500 mm deep sections with different web and flanges thickness of 150, 200, 250, and
300 mm showed that the largest share of the temperature and temperature gradient variations can be attributed to the thickness
of the web and flanges of the girder. Thus, the effect of the thickness of the web and flanges on temperature variation and tem-
perature gradient distributions is more significant than the effect of the girder depth.
R E F E RENC E S
[1] F. A. Branco, P. A. Mendes, J. Struct. Eng. 1993, 119(8), 2313.
[2] D. Li, M. A. Maes, W. H. Dilger, Can. J. Civ. Eng. 2004, 31(5), 813.
[3] Z. Song, J. Xiao, L. Shen, Adv. Struct. Eng. 2012, 15(3), 399.
[4] R. A. Imbsen, D. E. Vandershaf, R. A. Schamber, R. V. Nutt, Thermal effects in concrete bridge superstructures, in National cooperative highway
research program report 276, Transportation Research Board, Washington, DC, USA 1985.
[5] M. Elbadry, A. Ghali, ACI J. 1986, 83(6), 1001.
[6] F. Giussani, Eng. Struct. 2009, 31(10), 2392.
[7] W. Dilger, J. C. Beauchamp, M. S. Cheung, A. Ghali, J. Struct. Div. 1981, 107(ST11), 2147.
[8] K.‐Y. Wong, Struct. Control Health Monit. 2004, 11(2), 91.
[9] G. W. William, S. N. Shoukry, M. Y. Riad, Bridge Struct. 2005, 1(2), 103.
[10] Q. Chen, Effects of thermal loads on Texas steel bridges. PhD Thesis, the University of Texas at Austin, Texas, USA, 2008.
[11] S. N. Shoukry, M. Y. Riad, G. W. William, Eng. Struct. 2009, 31(12), 2954.
[12] Y. Xu, B. Chen, C. Ng, K. Wong, W. Chan, Struct. Control Health Monit. 2010, 17(6), 632.
[13] K. Koo, J. Brownjohn, D. List, R. Cole, Struct. Control Health Monit. 2013, 20(4), 609.
[14] Y. Xia, B. Chen, X.‐Q. Zhou, Y.‐L. Xu, Struct. Control Health Monit. 2013, 20(4), 560.
[15] E. Obrien, C. Carey, J. Keenahan, Struct. Control Health Monit. 2015, 22(12), 1396.
[16] J. Gu, M. Gul, X. Wu, Struct. Control Health Monit. 2017, https://doi.org/10.1002/stc.1998.
[17] J.‐H. Lee, I. Kalkan, Adv. Struct. Eng. 2012, 15(3), 447.
[18] N. Kulprapha, P. Warnitchai, Eng. Struct. 2012, 40, 20.
[19] R. Kromanis, P. Kripakaran, Comput. Struct. 2014, 136, 64.
[20] S. R. Abid, N. Tayşi, M. Özakça, Constr. Build. Mater. 2016, 106, 523.
[21] S. R. Abid, N. Tayşi, M. Özakça, Experimental measurements on temperature gradients in concrete box‐girder bridge under environmental ther-
mal loads. In Proceedings of the Istanbul Bridge Conference, Istanbul, Turkey, 2014.
[22] Y. Wang, Y. Zhan, R. Zhao, Adv. Struct. Eng. 2016, 19(7), 1043.
[23] Y. Xia, Y.‐L. Xu, Z.‐L. Wei, H.‐P. Zhu, X.‐Q. Zhao, Eng. Struct. 2011, 33, 146.
[24] H. Liu, Z. Chen, T. Zhou, Appl. Therm. Eng. 2012, 37, 329.
[25] C. Berwanger, J. Struct. Eng. 1983, 109(10), 2325.
[26] J.‐M. Lucas, A. Berred, C. Louis, Struct. Build. 2003, 156(2), 175.
[27] O. Larsson, Struct. Concr. 2009, 10(4), 193.
[28] H. Li, S. Li, J. Ou, H. Li, Struct. Control Health Monit. 2010, 17(5), 495.
[29] O. Larsson, R. Karoumi, Struct. Eng. Int. 2011, 21(1), 74.
[30] B. Chen, Z. Chen, Y. Sun, S. Zhao, Math. Probl. Eng. 2013, 2013(11), 1.
[31] D. Bortoluzzi, S. Casciati, L. Faravelli, J. Vib. Control 2013, 19(15), 2301.
[32] B. Gu, Z. Chen, X. Chen, J. Cent. South Univ. 2014, 21(3), 1227.
[33] N. Tayşi, S. R. Abid, Adv. Struct. Eng. 2015, 18(4), 469.
[34] H. Nandan, M. P. Singh, Eng. Struct. 2014, 81, 480.
[35] AASHTO, American association of state highway and transportation officials, AASHTO LRFD bridge design specifications. Washington DC,
USA 2012.
[36] British Standards Institution, BS 5400: Part2. Steel, concrete and composite bridges, specifications for loads. London, UK, 2006.
[37] European Committee for Standardization, EN 1991‐1‐5:2003. Eurocode 1: actions on structures‐part 1–5: general actions‐thermal actions.
Brussels, Belgium, 2009.
[38] NZ Transport Agency, Bridge manual SP/M/022. Section 3: design loading. Wellington, New Zealand, 2013.
ABID ET AL. 23 of 23
[39] A. Ghali, R. Favre, M. Elbadry, Concrete structures: Stresses and deformation, 3rd ed., E & FN Spon, London 2002.
[40] COMSOL Multiphysics v 4.3, COMSOL multiphysics user's guide. Stockholm, Sweden, 2012.
[41] E. Mirambell, A. Aguado, J. Struct. Eng. 1990, 116, 2388.
How to cite this article: Abid SR, Mussa F, Tayşi N, Özakça M. Experimental and finite element investigation of
temperature distributions in concrete‐encased steel girders. Struct Control Health Monit. 2017;e2042. https://doi.org/
10.1002/stc.2042