See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273958526

Behavior of curved and skewed bridges with integral abutments

Article  in  Journal of Constructional Steel Research · June 2015

DOI: 10.1016/j.jcsr.2015.03.003

CITATIONS READS

8 540

5 authors, including:

Yaohua Deng
Iowa State University
20 PUBLICATIONS   143 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Yaohua Deng on 23 February 2016.

The user has requested enhancement of the downloaded file.

 Journal of Constructional Steel Research 109 (2015) 115–136

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Behavior of curved and skewed bridges with integral abutments

Yaohua Deng a,⁎, Brent M. Phares a, Lowell Greimann a, Gus L. Shryack b, Jerad J. Hoffman c
a
Iowa State University, Institute for Transportation, Bridge Engineering Center, 2711 South Loop Drive, Suite 4700, Ames, IA 50010, United Sates
b
7821 Pennsylvania Ave, Kansas City MO 64114, United Sates
c
47582 258th Street, Sioux Falls, SD 57104, United Sates

a r t i c l e i n f o a b s t r a c t

Article history: Horizontally curved steel girder bridges are being designed with integral abutments due to their merits in terms
Received 9 January 2015 of cost, constructability and maintenance. This type of bridges is relatively new in the United States and their
Accepted 3 March 2015 performance evaluation is deemed difficult. Especially, the behavior of such bridges under thermal load
Available online 22 March 2015
conditions is not well understood and relevant code guidelines are not sufficiently provided. The purpose of
this study was to investigate the behavior of a curved and skewed bridge with integral abutments through a
Keywords:
Finite element analysis
numerical analysis and field monitoring-oriented program. A newly constructed, one lane, three span,
Temperature horizontally-curved, integral abutment bridge was instrumented using a field monitoring system to capture
Curved bridges the bridge behavior under change in ambient conditions and was tested using a dump truck traveling across
Skewed bridges the bridge at a walk pace. A three dimensional FE model was established and its predictions were reasonably
Integral abutments compared with the collected data. Subsequently, a parametric study was performed to investigate the influence
of curvature and skew on the bridge behavior under design loading conditions. It was found that the stresses in
girders varied with changes in skew and curvature significantly. With a 10° skew and 0.06 radians arc span length
to radius ratio, the curved and skewed integral abutment bridges can be designed as a straight bridge if a 10%
increase is applied to the total induced stresses. Thermal stress and its magnitude relative to the maximum stress
can reach up to 3 ksi and 15% of the maximum stress, respectively.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
The National Cooperative Highway Research Program (NCHRP)
raised concerns regarding the design, fabrication, and erection of hori-
zontally curved steel girder bridges [1]. These concerns centered around
difficult-to-predict girder displacements, fit-up issues, and unintended
locked-in stresses, including thermal stresses. The major reason for
the concerns is that up to one-quarter of the steel girder bridges in the
United States incorporate curvature in their design, and thus having a
better understanding of actual behavior – and therefore having better
design methodologies – is of notable importance. However, additional
complications arise when analyzing and designing I-girders in curved
bridges compared to straight girder bridges and range from the individ-
ual plates to the constructed girder as a whole [2].
To complicate matters even more, horizontally curved steel girder
bridges are being designed with integral abutments, given that integral
abutments are less expensive and easier to construct, simple to detail
in design, and require less maintenance along with elimination or
reduction of deck expansion joints [3]. Although the combined use of
⁎ Corresponding author.
E-mail addresses: jimdeng@iastate.edu (Y. Deng), bphares@iastate.edu (B.M. Phares),
greimann@iastate.edu (L. Greimann), gshryack@gmail.com (G.L. Shryack),
joff04@hotmail.com (J.J. Hoffman).
horizontally curved steel girder bridges with integral abutments stands
to be a promising design, this combination is relatively new to the
United States. Further, this combination can be more difficult to under-
stand and analyze as compared to an equivalent straight girder bridge
[4]. Importantly, the behavior of horizontally curved bridges with
integral abutments during thermal loading is not well understood.
Unfortunately, the current version of the AASHTO LRFD Bridge Design
Specifications [5] does not provide sufficient guidelines for thermal
behavior of bridges with integral abutments.
Thermal behavior of horizontally-curved bridges was studied by
Moorty and Roeder [6] and Hall et al. [1]. Thermal behavior of straight
IABs was also investigated by Abendroth and Greimann [7], Kim
and Laman [8], and Shah [9] through field monitoring program and finite
element (FE) simulations. As indicated by Kim and Laman [8], the
difficulties in predicting the performance of integral abutment bridges
are generally due to the complexity of daily and seasonal temperature var-
iations, nonlinear soil structure interaction, and time dependent effects.
Recent studies on thermal behavior of curved integral-abutment
bridges were completed by Thanasattayawibul [10], Doust [4], and
Kalayci et al. [11]. Thanasattayawibul [10] performed a parametric
study using a three-dimensional FE model to investigate the effect
that different parameters (i.e., bridge length, temperature, soil
profile type, span length, radius, and pile type) would have on the be-
havior of horizontally curved, steel-girder, integral-abutment bridges.

http://dx.doi.org/10.1016/j.jcsr.2015.03.003
0143-974X/© 2015 Elsevier Ltd. All rights reserved.
116 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

319'-0"
N
149'-0"
85'-0" 85'-0"
(FP) Girder A Girder B (FP)
Base Line

2'-6"
(IAB) (IAB)

26'-0"
5'-2" Girder D Girder C

15° 15°

15° S. Pier CL Brg. Radial Line N. Pier CL Brg. 15°

S. Abut. CL Brg. N. Abut. CL Brg.
(a) Plan view

1'-7"

26'-0"
1'-6"

Strain
gauge

Diaphragm
3'-1" Girder D Girder C Girder B
Girder A
Four Girders Spaced @ 7'-8''

(b) Cross-section
Fig. 1. Bridge 309 (1 ft = 0.30 m; 1 in. = 25.4 mm).

Conclusions and recommendations were made for future research and
for the design of this type of bridges. To investigate the thermal behav-
ior of horizontally-curved, steel-girder, integral-abutment bridges,
Doust [4] established multiple bridge models with varying horizontal
curvatures and total bridge lengths, considering different loading condi-
tions including gravity loads, lateral loads, temperature effects, concrete
shrinkage, and earth pressure. The author found that for bridges longer
than a specific length, the internal forces due to expansion are smaller in
a horizontally-curved bridge than in a straight bridge of similar length.
Kalayci et al. [11] established a rigorous three dimensional FE model to
evaluate the thermal behavior of curved integral abutment bridges. The
effects of bridge curvature, abutment backfill soil type, degree of lateral
restraint from the U-shaped wing walls on the seasonal response of two
span curved IABs were investigated. They found that: (1) as the curva-
ture increased, longitudinal displacements, earth pressures and weak
axis bending moment on piles decreased, and the lateral displacements

Table 1
Dimensions and section properties of instrumented sections.

tft (in.) bft (in.) tw (in.) hw (in.) tfb (in.) bfb (in.)

Girder A North/South span 1 20 7/16 48 1 20

Center span 7/8 20 7/16 48 1.375 20
Girder D North/South span 7/8 18 7/16 48 1 18
Center span 3/4 18 7/16 48 1.375 18

Note: tft — top flange thickness; bft — top flange width; tw — web thickness; hw — web height; tfb — bottom flange thickness; bfb — bottom flange width; 1 in. = 25.4 mm.

Fig. 2. Fixed pier bearing.

 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136
1'-7"
8'-71 8" 9'-9"
3'-3" Pressure cell
9"
Piles spaced @5'-3''
Fig. 3. Front elevation of integral abutment (1 ft = 0.30 m; 1 in. = 25.4 mm).
increased even for a moderate curvature; (2) loose sand backfill re-
duced the backfill pressures and relieved top of pile moments under
the positive temperature change; and (3) U-shaped wing walls partially
reduce strong axis bending on abutment piles.
The purpose of this study is to investigate thermal behavior of a
curved and skewed bridge with integral abutments through a numerical
analysis and field monitoring-oriented program. A field monitoring
system was designed and applied to a new horizontally curved bridge
with integral abutments. A finite element model was developed for
one monitored bridge and the predictions from the model were com-
pared with the measured data obtained in the live load and thermal
load conditions. Subsequently, a parametric study was performed
using the developed FE model to investigate the sensitivity of design
calculations to curvature and skew.
2. Field monitoring program
During the re-alignment of the intersection of Interstates I-35, I-80
and, I-235 at the northeast corner of Des Moines, Iowa, six new steel-
girder bridges with integral abutments were constructed at the inter-
section. Five of those bridges were selected for the field monitoring
program for assessing their long-term thermal behavior. The bridges
were instrumented using a field monitoring system that records strains,
temperatures, back wall pressures, etc. The data were collected once per
hour under varying thermal conditions. Of the five monitored bridges,
Bridge 309, a horizontally curved and skewed integral abutment bridge,
will be the focus of this study. This is due to the fact that Bridge 309 was
the most heavily instrumented bridge and the data from the field
monitoring system can most likely be utilized to validate the FE model.
2.1. Description of Bridge 309
Bridge 309 is a one lane, three span, horizontally-curved, integral
abutment bridge with a 289.6 m (950 ft) horizontal baseline curvature
85'-0"
(FP)
(IAB)
309SAPrW
309SAPrE
Girder Gauges
15°
15°
Fig. 4. Bridge 309 instrumentation (1 ft = 0.30 m; 1 in. = 25.4 mm).
117
29'-2"
1'-7"
4'-5"
Backwall
Maskwall
(beyond)
9'-9" Pile Cap
9" S3×7.5
9"
radius, a width of 7.9 m (26 ft), and spans of 25.9 m (85 ft), 45.4 m
(149 ft), and 25.9 m (85 ft), as shown in Fig. 1(a). The baseline is located
0.76 m (2.5 ft) from the west exterior girder, measured perpendicular to
the roadway. The abutments and piers are skewed at 15° left ahead.
Bridge 309 was constructed with four welded, I-shaped, plate girders
spaced at 2.34 m (7 ft–8 in.) and nineteen non-composite bent plate
diaphragms spaced at 5.72 m (18 ft–9 in.) and 6.02 m (19 ft–9 in.) in
the south/north spans and the center span, respectively. Note that the
supports align with the bridge skew and that the intermediate dia-
phragms align with the radius. The cross-section of Bridge 309 is
shown in Fig. 1(b). In this study the girders were labeled A, B, C, and D
from the outside of the curve to the inside of the curve.
The girder dimensions of the subsequently described instrumenta-
tion locations are shown in Table 1. The girder sections include a con-
crete deck made composite with the steel girders with welded shear
studs. The slab thickness is 203 mm (8 in.) on average. The fixed pier
bearing (see Fig. 2) was utilized for both the south and north piers.
For this type of bearing, a curved sole plate with a pintle was welded
to the girder bottom flange. The curved sole plate rests on a masonry
plate, which is attached to the top of the pier cap. A fixed pier is
designed to allow only rotation about an axis perpendicular to the
longitudinal direction of the girder and restrain translation in all direc-
tions of the girder at the pier location. Each of the bridge 309 integral
abutments has a width of 8.89 m (29 ft–2 in.) in the radial direction,
each mask wall has a width of 0.48 m (1 ft–7 in.) as shown in Fig. 3.
Each girder bears on the abutment pile cap (on a short length of an
S3 × 7.5) and the entire abutment is supported by vertical piles spaced
at 1.54 m (5 ft–0.8 in.). The eight HP 10 × 57 piles were aligned with
the weak axes that coincided with the abutment centerline. Note that
for clarity the reinforcing steel in the reinforced concrete back wall,
mask wall, and pile cap is not shown. Some of the vertical reinforcing
bars extend from the pile cap into the abutment back wall and mask
walls to form a composite section between the pile cap and the
abutment walls.
319'-0"
149'-0"
Girder A 85'-0"
309WTemp
Girder B (FP)
(IAB)
309NAPrW
Girder C 309AirTemp
Girder D 309NAPrE
309ETemp
15°
15°
118 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

SA-HP1
NA-HP1

NA-HP2

SA-HP2
NA-HP3
SA-HP3
NA-HP4

SA-HP4
NA-HP5

SA-HP5 NA-HP6

SA-HP6

(a) South Abutment (b) North Abutment

Fig. 5. Piles in south and north abutments.

2.2. Bridge instrumentation
Vibrating wire strain gages (VWSGs) are commonly utilized to
record long-term strain due to static loading and environmental effects
[12]. VWSGs measure the slowly-changing strain and filter out the
dynamic strain due to moving live load. For the bridge instrumentation,
Geokon Model 4150 VWSG [13] was utilized to measure the load-
induced strain on steel members. The fundamental of VWSG is that
the strain change in the wire is associated with the change of its
vibration frequency which can be expressed by the tension, length
and mass of the wire. The strain in the mounted steel measured by
VWSG is mechanical strain (i.e., load induced strain) induced by the
restraint to bridge components due to the temperature change. The
thermal strain due to free expansion is excluded automatically because
the steel vibrating wire of the VWSG has the same thermal expansion
coefficient as the mounted steel. The load induced strains are directly
measured without the necessity of temperature correction. Additionally,
the strain response measured by VWSGs was mostly induced by the

E
Mlc

Mlt

2 1 yt
x-axis Neutral
P axis
Mx
yb

3 4

xb xt

y-axis
Outer of the Inner of the
Mlb
curvature curvature

Girder curvature
Fig. 6. Strains, Forces and Moments in a Girder Cross-section.
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 119

thermal effects in this monitoring program. Note that the effects of
concrete shrinkage and creep on the girder strains were small during
the monitoring period, because the strain data were not collected until
more than one year after the construction of Bridge 309.
When sunlight impinges directly on the VWSG, the temperature in
the wire will be different from the steel which the gage is mounted to,
reducing the reliability of measured data [13]. For this study, it was
found that an unusual amount of outliers were noticed during initial
evaluation of thermal strains collected from the VWSGs mounted on
the girders. It was found that the cover used to protect the strain
gages did not provide enough thermal protection when exposed to
direct sunlight. The sunlight raised the temperatures of the VWSGs
significantly and resulted in what were determined to be erroneous
readings. To remedy this, only the girder strain gage data collected
between 9 PM to 6 AM were considered in this paper.
As shown in Fig. 4, VWSGs were attached to the exterior girders
(Girder A and Girder D) at the mid-length of each span of Bridge 309.
For each monitored girder section, four VWSGs were attached to the
inside face of the top and bottom flanges as shown in Fig. 1(b) and
Fig. 6. The gages were placed 1 in. from the flange tips and oriented to
measure the longitudinal strains. In addition to strain, steel tempera-
tures were recorded using these VWSGs with thermistors inside. Each
VWSG was covered by the protective coatings after being mounted to
the steel surface.
A total of six piles was cast into each abutment pile cap of Bridge
309. The VWSGs were only attached to the inside faces of both flanges
of three piles at both the south and north abutments and were located
23 cm (9 in.) below the abutment pile cap as shown in Fig. 3. And at
each instrumented cross-section, two gages were mounted on each
flange with a distance of 2.54 cm (1 in.) from the flange tips. At the
south abutment (SA), the west most pile, outside the curve, was
labeled SA-HP1, and the labeling continued east with the east most
pile, inside the curve, labeled SA-HP6, as shown in Fig. 5(a). At the
north abutment (NA), the west most pile, outside the curve, was
labeled NA-HP1, and the labeling continued east with the east most
pile, inside the curve, labeled NA-HP6, as shown in Fig. 5(b). At each
abutment of Bridge 309 piles HP1, HP4, and HP6 were instrumented
with strain gages.
Temperature gages 309ETemp and 309WTemp were placed at mid-
depth of the bridge decks on the east and west side of the north pier to
measure the internal concrete temperature as shown in Fig. 4. And
temperature gage 309AirTemp was hung below the deck at the middle
of the north pier to measure the ambient air temperature as shown
in Fig. 4. Two pairs of pressure cells 309SAPrW and 309SAPrE, and

(a) Temperature in west of top flange

(b) Temperature in east of top flange

Fig. 7. Seasonal and daily changes of monitored temperatures in Girder D at mid-center span (1 °F = 5/9 °C).
120 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

(c) Temperature in west of bottom flange

(d) Temperature change in east of bottom flange

Fig. 7 (continued).

309NAPrW and 309NAPrE were utilized to measure soil pressure
behind the south and north abutments, respectively. These pressure
cells were mounted at the third points of each abutment width as
shown in Fig. 4 and approximately at the mid-height of the distance
from the top of the slab to the bottom of the pile cap as shown in
Fig. 3. Other details related to bridge instrumentation can be found in
the final report by Greimann et al. [14].
the flange tip on the outside of the curve is considered positive.
It should be noted that these forces are chosen to align with the
AASHTO LRFD Specifications' codified approach for calculating lateral
forces [5].
Based on the elementary beam theory and the specific girder cross
sectional properties, the internal strain components, εa, εx, εlt, and εlb,
can be directly expressed by the internal forces, P, Mx, Mlt, and Mlb,
respectively, as given in Eq. (1).

2.3. Formulation of girder internal forces and strains 2 3

P
6 ðAEÞeff 7
The internal forces and strains (i.e., traditional design values — 2 6
3 6 7
7
four girder forces and four internal strain components) at each girder εa 6 M x yb 7
6 εx 7 6 ðEI x Þeff 7
cross section can be derived based on the field measurements due to 6 7¼6 7
ð1Þ
4 εylt 5 6
6 M lb xb
7
7
ambient temperature changes, which were described in detail by
εylb 6 7
6 Es I y f t 7
Shryack et al. [15]. As shown in Fig. 6, a girder coordinate system 6 7
4 M lb xb 5
was established on a composite girder cross-section and different in-
Es I y f b
ternal forces in the cross-section are also illustrated. These internal
forces include axial force (P, tension is positive), strong axis bending
(Mx, positive when the top flange is in compression), lateral bending where, εa = internal axial strain; εx = strong axis bending strain; εlt =
of the concrete deck (Mlc), and lateral bending of the top and bottom lateral bending strain in the top flange; εlb = lateral bending strain in
flanges of the steel girder (Mlt and Mlb). For lateral bending, tension in the bottom flange; (EA)eff = effective axial rigidity; (EIx)eff = the
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 121

effective flexural axial rigidity for x-axis bending; xi = distance 2.4. Variations of measured strains and strain components
from neutral axis to strain gage i along the x-axis (i = 1, 2, 3, 4);
yi = distance from neutral axis to strain gage i along the y-axis; The monitored data from 01/17/2011 to 11/19/2011 were utilized
xb = distance from neutral axis to the top flange, equals x2, x3, − x1, for analysis in this study. All data in the time interval from 9 pm to
or −x4; yt = distance from neutral axis to the bottom flange, equals 6 am for each day are processed to eliminate the sunlight influence as
x3, x4, − x1, or −x2. mentioned previously. And the data were processed with respect to
Further, the measured strains in each steel girder cross-section the reference data (i.e., zero data) at the time instant of 1/20/2011
can be expressed by the four internal strain components due to the 6 AM. For demonstration purposes, the data obtained at the center
internal forces. A set of matrices are developed as shown in Eq. (2) span of Girder D were used as an example to describe the seasonal
describing the relationship between the four measured strains and daily variations in temperature and strains in the steel girders.
at the strain gage locations and the four desired internal strain For Girder D at mid-center span, the time histories of monitored
components: temperatures in the west and east of the top flange are shown in
Fig. 7(a) and (b) respectively, and the time histories of monitored
2 3−1 temperatures in the west and east of the bottom flange are illustrated
2 3 yt 2 3 in Fig. 7(c) and (d) respectively. As shown in Fig. 7, the light gray dots
1 −1 0
εa 6 yb 7 ε1
6 εx 7 6 yt 7 6 ε2 7 represent the seasonal variations in temperatures and the various
6 7 61 0 7 6 7
4 εylt 5 ¼ 6
6 yb
1 7
7 4 ε3 5 ð2Þ types of dark dots represent the daily variations in temperature for
ε ylb 4 1 −1 1 5 different months at different locations of the Girder D cross-section.
0 ε4
1 −1 0 −1 Time histories of monitored strains in the west and east of the top flange
and the west and east of the bottom flange of Girder D at mid-center
span are also illustrated in Fig. 8(a), (b), (c) and (d) respectively.
where, εi = strain reading at gage i. Similarly, as shown in Fig. 8, the light gray dots represent the seasonal

(a) Strain in west of top flange

(b) Strain in east of top flange

Fig. 8. Seasonal and daily changes of monitored strains in Girder D at mid-center Span.
122 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

(c) Strain in west of bottom flange

(d) Strain in east of top flange

Fig. 8 (continued).

variations in strains and the various types of dark dots represent the
daily variations in strains for different months at different locations of
the Girder D cross-section. Figs. 7 and 8 indicate that the patterns of
these seasonal and daily variations in temperature and strain are similar
at different locations of the Girder D cross-section. Based on Eq. (2), the
four internal strain components can be derived by the four measured
strains at the strain gage locations and are plotted in Fig. 9. Fig. 9(a),
(b), (c), and (d) shows the internal axial strain versus time, strong
axis bending strain versus time, top flange lateral bending versus time,
and bottom flange lateral bending versus time, respectively. Likewise
in each figure, the light gray data shows the strain from the life of the
project and the black data shows the daily strain cycle from specific
days of the project. Fig. 9 indicates that the annual cycle range is larger
than the daily cycle range for axial strain, and in the case of major axis
bending and lateral flange bending in both the top and bottom flanges
the daily cycle range is comparable to the annual cycle range.
2.5 . Equivalent cantilever length of piles
The HP 10 × 57 bearing piles were driven through a 0.91 m (3 ft) di-
ameter predrilled hole that extended 4.11 m (13.5 ft) below the bottom
of the pile cap. The predrilled hole was filled with bentonite clay. The
soil immediately below the bottom of the predrilled hole is embank-
ment fill (i.e., stiff clay) from the previously constructed and now
removed bridge. The piles are oriented to bend about their weak axes
as the bridge expands and contracts. Using axial and lateral springs to
represent the soil behavior would be ideal. However, these springs
would have essentially zero stiffness within the predrilled hole because
of the almost fluid properties of the bentonite clay. Below the predrilled
hole, it is not at all reasonable to assess the spring properties without a
much more thorough investigation of the soil properties, which was
not conducted in this study. Consequently, the effects of piles on the
superstructure response were characterized by the equivalent cantile-
ver length model and the interaction between the pile and soil compo-
nents was not extensively investigated in this study. The background,
development, formulation and limitations of this model were docu-
mented in the research work by Greimann et al. [16], Greimann et al.
[17], and Abendroth and Greimann [7].
For the equivalent cantilever length model, the steel pile is idealized
as a column with an equivalent length, Le, and rotationally fixed ends,
shown in Fig. 10. Based on the maximum moment in the equivalent
cantilever system which is equal to that in the actual system for a
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 123

given horizontal displacement, the following equation was derived by piles of Bridge 309 were utilized. An analytical relationship between
Greimann et al. [16]: the pile weak axis bending strain and the horizontal pile displacement
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi was developed based on the equivalent cantilever analytical model
u    3  2  
u128 lu 4 þ 128pffiffiffi 2 llu þ 96 llu þ 24 2 u þ 6
pffiffiffi l proposed by Abendroth and Greimann [7], as shown in Fig. 11.
u The relationship between the horizontal displacement, Δ, and the
ls u l c c c lc l
¼u  2 pffiffiffi lu
  − u ð3Þ
lc t lu lc resulting end moment, Me, can be expressed by:
128 l þ 64 2 þ 16
c lc
6Es IΔ
Me ¼ ð6Þ
where, ls = the equivalent length below the stiff clay; lu = the length L2e
from the abutment bottom to the stiff clay (4.11 m (13.5 ft)); lc =
critical length, which can be determined by: where Es = steel elastic modulus; I = moment inertia of the pile cross-
section. Assuming a linear moment diagram for the pile, the end
sffiffiffiffiffiffiffiffi
moment, Me, is related to the moment at the strain gages, Mg, by:
4 Es I y
lc ¼ 4 ð4Þ
kh M e Lg
Mg ¼ ð7Þ
Le
where, Es = steel elastic modulus (200 GPa (29,000 ksi)); Iy = weak
axis moment inertia of the pile (42,039,374 mm4 (101 in.4)); kh = where Lg = Le minus twice the distance from the top of the pile to the
horizontal stiffness of the stiff clay (14.4 N/mm2 (300 ksf) per Greimann strain gage location. The relationship between the weak axis bending
et al. [16]). Based on Eqs. (3) and (4), the equivalent cantilever length moment, Mg, and the weak axis bending strain at the location of the
(le) can be derived by Eq. (5) and is equal to around 18 ft (5.49 m). strain gages, εg, is given by:.

le ¼ lu þ ls ð5Þ Mg c
εg ¼ ð8Þ
EI
To further validate the equivalent cantilever length model, the mea-
sured weak axis bending strain and displacement for the instrumented where, c = distance from the neutral axis to the pile gage location.

(a) Axial strains

(b) Vertical bending strains

Fig. 9. Seasonal and daily changes of monitored internal strain components in Girder D at mid-center span.
124 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

(c) Lateral bending strains in top flange

(d) Lateral bending strains in top flange

Fig. 9 (continued).

Using Eqs. (6) through (8), the weak axis bending strain εg, can be Abutment
calculated for any horizontal displacement Δ. The solid, diagonal line
in Fig. 12 represents this relationship indicating that the predictions
using an equivalent cantilever length of 5.49 m (18 ft) compare
reasonably with the test results. Note that the details of measured
survey data were described by Greimann et al. [14] and typical results
are shown in Fig. 12. Each rectangle in the figure represents a survey
date. For pile SAHP1, the horizontal sides of the rectangle represent
the 95% confidence interval of pile displacement, at the top of the Predrilled lu
pile, for a given survey date. The vertical sides of the rectangle repre-
hole
sent the change in the measured microstrain in the pile that occurred le
during survey.

3 . Finite element modeling and predictions

Refined analysis methods are required for analyzing the behavior of Stiff clay
Bridge 309 because the geometry of Bridge 309 does not satisfy the
ls
criteria set by AASHTO [5] for curved bridges in terms of bridge skew,
cross sections, and radius to arc span length ratios. A three dimensional
FE model of Bridge 309 was established using a commercial FE analysis
software package — ANSYS 14.0 [18] and its predictions were com-
pared with the girder strain data measured from the field monitoring
system. Fig. 10. Equivalent cantilever pile model.
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136
Δ deck centerline. A jointless connection was also modeled between the
Me
Me Mg
Lg
Mg
Me Me
Fig. 11. Equivalent Cantilever Pile Model with Horizontal Displacement and Bending
Moment.
3.1. Finite element modeling
The FE model utilized the as-built bridge dimensions taken from the
plan sheets of Bridge 309. The abutment, deck, and web of the steel plate
girders were modeled using SHELL63 — a 3-D elastic shell element that
has both bending and membrane capabilities. This element has four
nodes and six degrees of freedom at each node: translations in x, y,
and z directions and rotations about x, y, and z axes. The abutment
piles, pier caps, pier columns, and flanges of the steel plate girders
were modeled using BEAM4 — a uniaxial 3-D elastic beam element
with tension, compression, torsion, and bending capabilities. At each
node the element has six degrees of freedom: translations in x, y, and
z directions and rotations about x, y, and z axes.
Mesh attributes with respect to the longitudinal length of the bridge
rather than the number of elements per girder cross section have signif-
icant influence on the analytical results [2]. They further suggested that
the element length should roughly equal 2% of the span length. Approx-
imate element lengths (around 30 cm (12 in.)) for the girders were
selected accordingly. The diaphragms were directly attached to the
girders along their full depth. Namely, diaphragms were connected
to the girder by sharing common nodes at the connection locations.
The mesh size of the deck was adjusted to align the deck nodes with
those of girders. Rigid links were used to connect the girders to the
deck considering the elevation difference of the girder top and the
Microstrain
-0.5 -0.4 -0.3 -0.2
Fig. 12. SAHP1 Weak Axis Bending Strain versus Displacement (1 in. =25.4 mm).
125
concrete parapets and the deck by sharing common nodes. Both abut-
ments were directly attached to the girders, deck, and abutment piles
through sharing common nodes. The abutment piles were modeled
with an equivalent cantilever length of 5.49 m (18 ft) as previously
described, and all rotation and translation degrees of freedom were
restrained at the base of the piles as discussed in the aforementioned
Le section. The piles were oriented such that their weak axes coincided
with the abutment centerline. Pier columns and pier cap components
were modeled about their centerlines. Piers were fixed at the base
of the modeled columns. Constraint equations were defined for the
nodes at the connection location of girders and pier cap components,
such that only rotation about an axis perpendicular to the longitudinal
direction of the girder was allowed and translation in all directions of
the girder was restrained at the connection location by the piers.
Concrete material properties for the piers, abutments, deck, and par-
apets, are summarized in Table 2, including the modulus of elasticity
(E), Poisson ratio, ν, density, ρ, and coefficient of thermal expansion,
α. Oesterle et al. [19] reported the concrete thermal expansion coeffi-
cient with a mean of 8.7 × 106 °C (4.8 × 106 °F) and a range of 4.1 to
14.6 × 106 °C (2.3 to 8.1 × 106 °F) according to the findings from over
available 200 sources. For this study, the mean value of the concrete
thermal expansion coefficient was utilized in the FE model. It should
be noted that no consistent value of thermal expansion coefficient was
utilized for the FE model in the literature, such as 10.8 × 106 °C
(6.0 × 106 °F) used by Kim and Laman [20], 9.9 × 106 °C (5.5 × 106 °F)
used by Kalayci et al. [11], 19.4 × 106 °C (10.8 × 106 °F) used by Fennema
et al. [21]. And the steel material properties, for the girders, diaphragms,
and abutment piles, are also listed in Table 2. Note that the densities
of the materials are used to apply the dead load of components in
combination with gravity. The established FE model and its components
are illustrated in Fig. 13.
3.2. FE model predictions under live load conditions
The predicted results from the developed FE model were compared
with test data measured during field live load testing as reported by
Greimann et al. [14] and Hoffman and Phares [22]. The test instrumen-
tation consisted of forty strain transducers at two separate radial bridge
cross-sections located in the bridge north span as shown in Fig. 14. Each
of the girders was instrumented with four strain transducers at each
radial cross-section. The cross sections, Section 1 and Section 2, were
located at the location in the middle of the unbraced length between
two diaphragms and at the location containing diaphragms. The two
adjacent sections are both parallel to the radius of curvature. Live load
300
Survey Data
200
100
Equivalent Cantilever Pile Model
0
-0.1 0 0.1 0.2 0.3 0.4 0.5
-100
-200
-300
Δx (in.)
126 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136
Table 2
Concrete and steel material properties.
Material E (psi) ν ρ (lb/in.−3) α (10−6/°F)
Concrete 3,824,000 0.2 0.08681 4.8
Steel 29,000,000 0.3 0.2836 6.5
Note: 1 °F = 5/9 °C; 1 psi = 0.006895 MPa; 1 lb/in.−3 = 27,680 kg/m3.
testing was conducted by placing a loaded tandem axle dump truck at
three load paths: Load Path 1 (LP1) — located towards the inside of
the curve 0.61 m (2 ft) from the face of the guardrail; Load Path 2
(LP2) — centered on the bridge deck; and Load Path 3 (LP3) — located
towards the outside of the curve 0.61 m (2 ft) from the face of the guard-
rail). To obtain the static strain envelops due to the live loads, the truck
traveled along each load path across the bridge at a crawling speed.
Concentrated forces representing the truck live loading was applied
to the nodes of the model taking into account the truck axle weights,
wheel spacing, and transverse axle spacing of the dump truck used
during field testing. The girder strains were extracted from the FE
model at bridge Sections 1 and 2 and then compared to the measured
strains acquired in the field testing. Girder bottom flange strains were
of specific interest for the comparison purposes due to their large mag-
nitudes relative to the top flange strains.
The measured and predicted bottom flange strains of two example
girder cross-sections, Girder B at Section 1 and Girder A of Section 2
due to truck loading at the center path (LP2) and the outside path
(LP3), are compared in Fig. 15(a) and (b) respectively. Fig. 15 indicated
that the strains predicted using the FE model follow the same order as
those measured in the field testing. Further, the maximum peak girder
strains in the bottom flanges at all girder locations for all three load
paths are compared with measured data from the field testing, as
shown in Table 3. Table 3 indicates that the predicted results are slightly
larger than the measured results at all girder locations for all three load
paths. The magnitude of difference was least when the truck was cen-
tered and at Bridge Section 1. Since Bridge Section 1 is located between
two diaphragm sets, results may indicate that the behavior of the
modeled girder–diaphragm interaction diverges from the actual field
condition. In sum, the FE model can reasonably predict the bridge
response according to the comparison with the strain data measured
under live load conditions.
3.3. FE model predictions under thermal conditions
Furthermore, the predictions from the FE model were also compared
with the monitoring results for Bridge 309 under different thermal
Deck
Parapet
Abutment
Girder
Pile
Fig. 13. 3D view of the FE model of Bridge 309.
conditions. The purpose of this comparison is to verify that the FE
model can reasonably predict the thermal behavior of the horizontally
curved IAB — Bridge 309. As previously introduced, the field monitoring
system recorded the temperatures in the steel girders and the concrete
deck, the strains in steel girders and back wall pressures at abutments.
The measured field data at certain time instants from 01/17/2011 to
11/19/2011 were selected. With respect to the reference time instant —
1/20/2011 6:00 AM, the measured temperature changes at the four se-
lected time instants are tabulated in Table 4. It should be noted that the
temperature in the top or bottom flange is obtained by averaging the
temperature measured in the west and east of each flange due to their
negligible difference. And similarly, the deck temperature is also calcu-
lated by averaging the temperature measured in the west and east of
the deck. Table 4 indicates that the measured temperatures in the top
(or bottom) flanges of different girders at different spans are almost
identical. This is probably due to the fact the temperatures were mea-
sured at 6:00 AM in the morning and temperature is evenly distributed
longitudinally and transversely in the girders. Accordingly, the temper-
ature changes can be further averaged for all the measurements in the
top or bottom flanges respectively as given in Table 5. The strain chang-
es in girders were mainly induced by the change of temperature in the
concrete deck and steel girders and back wall pressures at abutments
as summarized in Table 5. Accordingly, for each time instant of
Table 5, a temperature gradient based on the temperature in the top
and bottom flanges was applied to all the steel sections of the super-
structures in the FE model. A uniform temperature change was applied
to the concrete deck and parapet as given in Table 5. The soil behind
each abutment was assumed to be homogenous and the stress distribu-
tion increases linearly from zero at the surface downward [23]. And
thus, the back wall pressures with a triangular profile at south and
north abutments (the average values given in Table 5) were applied to
the FE model since the pressures were measured at the mid-height
of the distance from the top of slab to the bottom of the pile cap as
shown in Fig. 3.
Recall that strains were measured in the cross-sections of Girder A
and Girder D at the mid-lengths of the south span, north span and
center span. Based on the four strains in the top and bottom flanges of
each cross-section, axial strains in the sections of Girders A and D at
mid-spans were derived using Eq. (2) at the selected time instants.
The strain changes at the selected time instants gave the changes of in-
ternal axial strain (εa), strong axis bending strain (εx); lateral bending
strain in the top flange (εlt), and lateral bending strain in the bottom
flange (εlb) in the cross-sections of girders A and D at different locations
as shown in Table 6. Likewise, these strain changes were also calculated
based on the four strains at the bottom and top flanges obtained using
Pier
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 127

the FE model and summarized in Table 6. Table 6 indicates that
the strain components predicted using FE model show reasonably
agreement with the measured results especially for the axial strain
components. The discrepancy is mainly due to the possible temperature
difference among different girders and among different girder locations
from girder mid-spans through pier locations and abutments, tempera-
ture gradient in deck, etc., which were not accurately measured in the
field monitoring program and not taken into account in the FE model.
To present further comparison with monitored data, the strain compo-
nents in girder D at mid-center span (as an example cross-section)
predicted using FE model are compared with the measured results at
nine time instants during the monitoring period as shown in Fig. 16.
Fig. 16 indicates that the FE predictions capture the seasonal variation
of strain components and are in good agreement with the measured
results. In sum, the FE model can reasonably predict the bridge response
according to the comparison with the monitored strain data due to
different thermal conditions.
4. Parametric study under design loading conditions
4.1. Design load conditions
After the predictions using the FE model were compared with the
monitored data for Bridge 309 under both live load and thermal load
conditions, design loading conditions were subsequently investigated
and utilized for the parametric study. Note, that in formulating design-
type values, the parapets were excluded for the FE model. The loading
conditions selected were based on the design limit states in accordance
with the AASHTO LRFD Specifications [5]. For this study, the Strength I
load combination was selected for detailed evaluation. For the scope

North 85'-0" Pier No.

2 to North Abut. Traffic Flow
Brg.

Pier No. 2 Intermediate

Diaphragm
North Abut. Brg.

Girder A

Girder B

Girder C
4 Spa. @ 18'-9
" Section
1 Section
2 Girder D

a) Location of Two Instrumented Sections (1 ft = 0.3 m; 1 in. =25.4 mm)

Center of Curvature

Strain
Transducer

Girder D Girder C Girder B Girder A

b) Section 1 – Location of Strain Transducers

Center of Curvature

Strain
Transducer

Diaphragm
Girder D Girder C Girder B Girder A

c) Section 2 – Location of Strain Transducers

Fig. 14. Live Load Testing Instrumentation of Bridge 309.
128 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

a) Section 1 Girder B - Center Path (LP2)

b) Section 2 Girder A - Outside path (LP3)

Fig. 15. Comparison of Measured and Predicted Strains in Bottom Flanges of Two Example Girder Cross-sections (1 ft = 0.3 m).

Table 3
Comparison of measured and predicted strains in girder bottom flanges for different load paths.

Gage location Outside load path Center load path Inner load path

εT (με) εFE (με) RFE/T εT (με) εFE (με) RFE/T εT (με) εFE (με) RFE/T

Section 1 Girder A 51 63 1.24 26 33 1.27 8 11 1.38

Girder B 48 54 1.13 42 44 1.05 20 25 1.25
Girder C 20 25 1.25 42 43 1.02 44 53 1.20
Girder D 6 8 1.33 20 30 1.50 52 62 1.19
Section 2 Girder A 45 63 1.40 24 34 1.42 8 12 1.50
Girder B 40 48 1.20 36 38 1.06 20 27 1.35
Girder C 22 28 1.27 37 37 1.00 37 43 1.16
Girder D 7 8 1.14 19 29 1.53 47 57 1.21

Note: εT — strain response measured during testing; εFE — strain response predicted from the FE model; RFE/T — RAtio of prediction to measured strain response.
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 129

Table 4
Temperature changes in air and different locations of Bridge 309.

Time instant Girder A — Girder D — Girder A — Girder D — Girder A — Girder D — Air (°F) Concrete (°F)
south span (°F) south span (°F) center span (°F) center span (°F) north span (°F) north span (°F)

TF BF TF BF TF BF TF BF TF BF TF BF

1/20/2011 6:00 AM 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2/21/2011 6:00 AM 24.3 21.2 23.8 21.2 24.0 20.9 23.6 21.0 23.4 21.2 23.5 21.2 20.6 24.5
3/21/2011 6:00 AM 37.2 31.4 36.7 31.3 37.4 31.5 38.1 31.8 37.2 31.5 37.0 31.5 30.8 36.0
4/21/2011 6:00 AM 30.4 27.9 30.0 27.7 30.3 28.1 30.0 27.6 29.5 27.7 30.4 28.0 28.5 29.4
5/21/2011 6:00 AM 54.9 54.1 54.6 54.0 55.0 54.0 54.6 54.1 53.8 54.1 54.5 54.0 54.1 54.9

Note: TF — top flange; BF — bottom flange; 1 °F = 5/9 °C.

of this study, only the dead, live, and thermal loads were considered.
The Strength I load combination considered can be expressed as:
Strength I ¼ γ p DC þ 1:75LLð1 þ IM=100Þ þ γ TU TU þ γ TG TG
where, γp = factor of dead load; DC = dead load; LL = live load; IM =
impact factor; γTU = factor of uniform temperature; TU = uniform
temperature; γTU = factor of gradient temperature; TG = gradient
temperature.
Dead loads due to the self-weight of the deck, steel girders, and steel
diaphragms were included by density and gravity acceleration. And the
dead loads due to the self-weight of parapets were calculated and
applied over the length of the bridge deck in place of the footprint of
the parapets. Maximum and minimum load factors, γp, of 1.25 and
0.90 were applied for the Strength I load combination. The minimum
load factor was used when the force results from the live load were
not additive to the dead load results.
Vehicular live loading was considered to be the HL-93 load case and
the combination of the design truck with the design lane load was
assumed to govern for this study. The number of design lanes equals
ð8Þ two because Bridge 309 has a 7.92 m (26.0 ft) roadway width and the
width of design lanes is 3.66 m (12.0 ft). Two design trucks were placed
side by side at increments of 10.0 ft along the bridge centerline, produc-
ing maximum responses to live load. Load factors for Strength I load
combinations were taken as 1.75 and a vehicular dynamic load allow-
ance of 33% was applied to the static live load application. Note that
only the design truck is subject to the IM.
The design loads also considered two thermal loads, uniform tem-
perature and temperature gradient [5]. Uniform temperature is applied
to the entire depth of the superstructure. Temperature ranges were
based on the classification of the bridge due to material type and loca-
tion. Bridge 309 classifies as a steel or aluminum structure located in a
cold climate. The thermal analysis assumed two lock-in temperatures
for a positive (+TU) and negative (− TU) temperature change. Lock-
in temperatures were based on acceptable temperatures for placing

Table 5
Final temperature changes for concrete and steel for FE model.

Time instant Temperature Back wall pressure

Concrete (°F) Steel top (°F) Steel bottom (°F) South abutment (psi) North abutment (psi)

1/20/2011 6:00 AM 0 0 0 0 0
2/21/2011 6:00 AM 24.5 23.8 21.1 0.3 0.2
3/21/2011 6:00 AM 36.0 37.3 31.5 0.1 0.0
4/21/2011 6:00 AM 29.4 30.1 27.8 0.2 0.2
5/21/2011 6:00 AM 54.9 54.6 54.1 14.1 11.0

Note: 1 °F = 5/9 °C; 1 psi = 0.006895 MPa.

Table 6
Comparison of measured and predicted strain changes for Girders A and D at three spans.

Time Instant Girder A — south Girder D — south Girder A — center Girder D — center Girder A — north Girder D — north
span (με) span (με) span (με) span (με) span (με) span (με)

Δεa Δεx Δεlt Δεlb Δεa Δεx Δεlt Δεlb Δεa Δεx Δεlt Δεlb Δεa Δεx Δεlt Δεlb Δεa Δεx Δεlt Δεlb Δεa Δεx Δεlt Δεlb

2/21/11 Measured −40 23 −1 3 −31 5 1 4 −33 8 5 5 −39 12 0 0 −37 9 −2 2 −50 21 −19 −5

6:00 AM data
FE results −24 21 0 −1 −27 22 1 0 −28 10 6 −2 −32 13 1 1 −26 21 0 −1 −25 21 1 0
Ratio 0.61 0.88 0.85 0.83 0.71 0.51
3/21/11 Measured −65 33 −4 6 −50 10 1 8 −56 19 9 7 −72 25 −2 2 −62 21 −8 −3 −78 32 −22 −2
6:00 AM data
FE results −46 42 0 −2 −49 44 2 1 −51 24 11 −4 −58 29 1 2 −48 42 −1 −2 −48 43 2 1
Ratio 0.70 0.98 0.91 0.80 0.76 0.61
4/21/11 Measured −47 18 −2 3 −41 8 −3 6 −47 2 4 8 −51 12 −3 3 −49 8 −6 1 −67 25 −17 3
6:00 AM data
FE results −36 22 0 −1 −38 23 1 0 −40 5 9 −4 −45 11 1 2 −38 22 0 −1 −37 21 1 0
Ratio 0.76 0.94 0.86 0.88 0.77 0.55
5/21/11 Measured −88 −3 −1 −1 −68 −26 6 11 −83 −3 5 9 −95 −4 −3 3 −88 −13 −9 −1 −102 19 −23 6
6:00 AM data
FE results −60 −5 1 3 −116 −11 4 0 −84 −5 15 −10 −102 10 2 5 −108 −14 −2 2 −55 0 1 −2
Ratio 0.68 1.70 1.02 1.07 1.23 0.53
130 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

concrete during construction. For a positive uniform temperature
change, the lock-in temperature was assumed to be 4.4 °C (40 °F);
yielding +TU equal to 26.7 °C (80 °F) to reach the upper temperature
range of 48.8 °C (120 °F). For a negative uniform temperature change,
the lock in temperature was assumed to be 26.7 °C (80 °F); yielding
− TU equal to − 78.9 °C (− 110 °F) to reach the lower range of
− 34.4 °C (− 30 °F). The temperature gradient, TG, applies three
temperatures throughout the depth of the superstructure in addition
to the uniform temperature values. For the positive temperature gradi-
ent, Temperature 1 (T1 = 7.78 °C (46 °F)) at the top of the deck,
Temperature 2 (T2 = −11.1 °C (12 °F)) in the middle of the deck, and
Temperature 3 (T3 = − 13.3 °C (8 °F)) constant for the depth of the
girder, were selected based on the geographic bridge location. Negative
temperature gradient values were obtained by multiplying the positive
values by −0.30 for plain concrete decks with no asphalt overlay.
parameters of the bridge models were kept the same as the original
Bridge 309 model. A total of twelve FE models were established with
the range of values for curvature and skew as tabulated in Table 7.
Note that Bridge 309 (which was the baseline for this study) has a
curvature radius of 289.6 m (950 ft) and a skew of 15°. All bridge
models with a curvature radius of 6386 m (20,950 ft) (very high curva-
ture radius arbitrarily selected) were deemed to essentially represent
the straight bridge case. For the entire parametric study, the Strength I
load combination was considered. Given the extreme condition
monitored from the field monitoring program, an average back wall
pressure of 0.081 MPa (11.7 psi) was applied to both the south and
north abutments when considering positive temperature change effects.
5. Results and discussion

The parametric study focused on results near the approximate

4.2. FE models for parametric study
A parametric study was performed to investigate the influence of
bridge curvature and skew on the stresses induced in the girders of
Bridge 309. Further, the impact of having two fixed piers (as opposed
to one fixed pier and one expansion pier) on the design of these curved
girder bridges was investigated.
Different bridge models were established by systematically changing
the degree of curvature and skew. To allow for comparison, all other
maximum positive and negative moment regions. Strains were extract-
ed for each of the four girders at seven cross sections along the bridge
length. Specifically, three mid-span cross sections representing positive
moment regions and four support cross sections representing the nega-
tive moment regions were considered. Given the different curvatures
and skew, the mid-span cross sections were selected to be parallel to
the bridge skew. The FE results indicated that the mid-center span sec-
tion and the north pier section were representative of the locations
where the most significant stresses exist in the bridge. And the stress

a) Axial strains

b) Vertical bending strains

Fig. 16. Comparison of Monitored Strains and Predicted Strains using FE Model for Girder D at Mid-center Span.
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 131

c) Lateral bending strains in top flange

d) Lateral bending strains in top flange

Fig. 16 (continued).

results for Girders A and D were the focus of the sensitivity study. As the
stresses at Points 3 and 4 were larger than at Points 1 and 2, Points 3 and
4 are the focus here (see Fig. 6).
The maximum stresses in the mid-center span and north pier
sections of Girders A and D are plotted in Figs. 17 and 18, respectively.
Maximum stress in this discussion referred to the largest sum of stresses
(vertical bending plus lateral bending plus axial) in each cross
section. Fig. 17 shows that at the mid-center span section the stress in
Girder A increases and the stress in Girder D decreases with an increase
of curvature for all the skews; further, the stresses in Girder A and
Girder D both decrease with an increase of skew for all the curvatures.
Fig. 18 shows that for the north pier section, the stresses in Girder A
and Girder D generally decrease with an increase of curvature, except
for girder D at a 0° skew; and the stresses in Girder A and Girder D
increase with an increase of skew. In sum, the stresses in the girders
of integral abutment bridges significantly vary with changes in skew
and curvature.
According to the AASHTO LRFD Specifications [5], the effect of
curvature, when calculating strong axis bending moments, may be ig-
nored when the following criteria are satisfied: (1) concentric girders;
(2) skew less than 10° from radial; (3) equal girder stiffness; (4) arc
span length to radius ratio less than 0.06 radians (corresponding to
737 m (2417 ft) curvature radius for Bridge 309). In addition to showing
the actual maximum stresses, Fig. 17 and Fig. 18 show the percentage of
6386 m (20,950 ft) curvature radius (i.e., the baseline condition). For
the bridge with 10° skew and 0.06 radians arc span length to radius
ratio, the maximum stresses in Girder A at the mid-center span, Girder
D at the mid-center span, Girder A at the north pier, and Girder D at
the north pier were 107%, 97%, 99%, 110% of those in the straight bridge,
respectively, as shown in Figs. 17(a), (b) and 18(a), and (b), respective-
ly. Thus, it can be concluded that with a 10° skew and 0.06 radians arc
span length to radius ratio (i.e., meeting the geometrical requirements
to ignore curvature for strong axis bending), the curved and skewed
integral abutment bridges can be designed as a straight bridge if a 10%
increase is applied to both the calculated axial forces and bending
moments (or the total induced stress). If a designer cannot accept
the 10% increase in force/stress estimation (even when meeting the
geometrical requirements outlined in AASHTO [5]), the designer is
advised to use a more discretized modeling approach.
Table 7
Variable values of the curvature and skew.
Skew, degrees Curvature radius, ft (curvature, 1/ft)
0 350 (0.00303) 550 (0.00182) 950 (0.00105) 20,950 (0.00005)
15 350 (0.00303) 550 (0.00182) 950 (0.00105) 20,950 (0.00005)
30 350 (0.00303) 550 (0.00182) 950 (0.00105) 20,950 (0.00005)

the maximum stress obtained for the bridge with the zero skew and Note: 1 ft = 0.30 m.
132 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

a) Girder A

b) Girder D
Fig. 17. Maximum Stresses in Girder A and D at Mid-center Span with Varying Skew and Curvature (1 ksi = 6.895 MPa; 1 ft = 0.3 m).

Fig. 17 indicates that the largest maximum stress occurs at the mid-
center span section of Girder A. Additionally, the sensitivity study
results indicated that the maximum lateral bending stress occurs at
the north pier section of Girder D. As a result, these two critical sections
were selected for further study under design thermal conditions.
The behavior of the bridge models was further studied under the
conditions due to positive temperature change (T+) (sum of uniform
temperature and gradient temperature changes), negative temperature
change (T−) (sum of uniform temperature and gradient temperature
changes), dead load (DL), or live load (LL), respectively. The stresses,
due to different loads for the mid-center span section of Girder A and
the north pier section of Girder D, are illustrated in Figs. 19 and 20,
respectively. As shown in Fig. 19(a), for the mid-center span section of
Girder A, thermal stresses due to the positive and negative temperature
changes, which have positive and negative contribution to the maxi-
mum stress respectively, slightly vary along with curvature and skew.
And the stresses due to dead and live loads which both have positive
and negative contributions to the maximum stress respectively, signifi-
cantly vary along with the curvature and skew. As shown in Fig. 19(b),
the percentages of all these stresses to the maximum stresses do not
vary significantly mainly due to the fact that the live and dead loads
have major contributions to the maximum stresses and the variations
of their contributions follow the same pattern. Fig. 19 also indicates
that the thermal stresses in the mid-center span section of Girder A
can reach up to 3 ksi and contributes to up to 10% of the maximum
stress.
As shown in Fig. 20(a), for the north pier section of Girder D, thermal
stresses due to the positive temperature change have positive contribu-
tion to the maximum stress, vary along with curvature and skew
observably. The negative temperature change sometimes has a positive
contribution to the maximum stress which is smaller than that due to
the positive temperature change and also varies along with curvature
and skew observably. And the stresses due to dead and live loads
which both have positive and negative contributions to the maximum
stress respectively, markedly vary along with curvature and skew. As
shown in Fig. 20(b), the percentages of all these stresses to the maxi-
mum stresses do not vary significantly within 5% except the stresses
due to the negative temperature change. Fig. 20 also indicates that the
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 133

a) Girder A

b) Girder D
Fig. 18. Maximum Stresses in Girder A and D at North Pier with Varying Skew and Curvature (1 ksi = 6.895 MPa; 1 ft = 0.3 m).

thermal stresses in the north pier section of Girder D can reach up to
20.7 MPa (3 ksi) and contributes to up to 15% of the maximum stress.
In sum, thermal stress and its percentage to the maximum stress
vary along with curvature and skew and can reach up to 20.7 MPa
(3 ksi) and 15% of the maximum stress, respectively. The thermal stress
need to be considered by the analytical approach used to design curved
and skewed IABs.
It should be noted that the above findings can only be applied to
the curved and skewed IAB bridges with parameters/conditions
investigated in this study including: (1) skew of no more than 30°;
(2) three spans; (3) two fixed pier bearings; (4) consideration of
the amplification factors due to the second order effects based on
AASHTO specifications [5].
6 . Summary and conclusions
To monitor the thermal behavior of horizontally-curved, steel-girder
bridges with integral abutments, a newly constructed, one lane, three
span, horizontally-curved, integral abutment bridge (Bridge 309) was
instrumented and monitored under ambient environmental conditions
with varying thermal conditions using a field monitoring system. The
bridge was monitored over a period of approximately fifteen months.
Strains were measured in the exterior girders and six piles at abutments.
Steel temperatures and internal concrete temperatures were also
monitored. Soil pressures behind the south and north abutments are
also recorded. Utilizing elementary beam theory and the specific girder
cross sectional properties, the four force-related strains (i.e., axial strain,
strong axis bending strain, lateral bending strain in the top flange, and
lateral bending strain in the bottom flange) were derived from the
four measured strains at each girder cross section. The equivalent
cantilever length of the piles was calculated for piles using the equiva-
lent cantilever analytical model and verified against the monitored
and survey data.
A three dimensional FE model was established to further study the
thermal behavior of Bridge 309. The predictions of the FE model were
compared with monitored data under live load conditions and different
thermal conditions. The ranges of internal strain components predicted
using the FE model compared well with these calculated based on the
monitored strain data. Subsequently, a parametric study using the FE
model was subsequently performed to investigate the influence of
bridge curvature and skew on behavior of Bridge 309, under design
loading conditions (i.e., the Strength I load combination). A total of
134 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136

a) Different Stresses

b) Percentage of Different Stress to Maximum Stress

Fig. 19. Stresses in Girder A at Mid-Center Span with Varying Skew and Curvature Due to Different Loads (1 ksi = 6.895 MPa; 1 ft = 0.3 m).

twelve bridge models were established by only changing the degree of
curvature and skew. The following conclusions can be drawn:
• When vibrating wire strain gages are exposed to direct sunlight and
the cover does not provide enough thermal protection, measured
data are not reliable.
• The stresses in the girders of integral abutment bridges significantly
varied with changes in skew and curvature.
• With a 10° skew and 0.06 radians arc span length to radius ratio
(i.e., meeting the geometrical requirements to ignore curvature for
strong axis bending), the curved and skewed integral abutment
bridges could be designed as a straight bridge if a 10% increase is
applied to both the calculated axial forces and bending moments
(or the total induced stresses). If a designer cannot accept the 10%
increase in force/stress estimation (even when meeting the geometri-
cal requirements outlined in AASHTO [5]), the designer is advised to
use a more discretized modeling approach.
• The largest maximum stress occurred at the mid-center span section
and that the largest lateral bending stress occurred at the north pier
section.
• Thermal stress and its percentage to the maximum stress varied along
with curvature and skew and can reach up to 20.7 MPa (3 ksi) and 15%
of the maximum stress, respectively. The thermal stress should be
considered in the design of curved and skew integral abutment
bridges.
The above findings can only be applied to the curved and skewed
IAB bridges with parameters/conditions investigated in this study
 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136 135

a) Different Stresses

b) Percentage of Different Stresses to Maximum Stress

Fig. 20. Stresses in Girder D at North Pier with Varying Skew and Curvature Due to Different Loads (1 ksi = 6.895 MPa; 1 ft = 0.3 m).

including: (1) skew of no more than 30°; (2) three spans; (3) two fixed the paper reflect the conclusions and opinions of the authors and do
pier bearings; (4) consideration of the amplification factors due to the not necessarily express the views of the funding agency.
second order effects based on AASHTO specifications [5].

Acknowledgments References

The authors would like to acknowledge the Federal Highway
Administration for sponsoring this transportation pooled fund study
with Grant No.: TPF-5(169). The authors would also like to thank the
state pooled fund Department of Transportation (DOT) partners for
their support: Iowa DOT (IADOT — lead state), Ohio DOT (ODOT), Penn-
sylvania DOT (PennDOT), Wisconsin DOT (WisDOT). The contents of
[1] Hall DH, Grubb MA, Yoo CH. Improved design specifications for horizontally curved
steel girder highway bridges. NCHRP Report 424. Washington, DC: Transportation
Research Board, National Research Council; 1999.
[2] Lydzinski J, Baber T. Finite element analysis of the wolf creek multispan curved
girder bridge. Charlottesville, Virginia: Virginia Transportation Research Council,
FHWA/VTRC 08-CR8; 2008.
[3] Mistry V. Integral abutment and jointless bridges. Baltimore, USA: Conference on
high performance steel bridges of the National Bridge Research Organization; 2000.
136 Y. Deng et al. / Journal of Constructional Steel Research 109 (2015) 115–136
[4] Doust S. Extending integral concepts to curved bridge systems. University of
Nebraska; 2011(Ph.D. dissertation).
[5] AASHTO. LRFD bridge design specifications. American Association of State Highway
and Transportation Officials: Washington, D.C.; 2010.
[6] Moorty S, Roeder CW. Temperature-dependent bridge movements. J Struct Eng
1992;118:1090–105.
[7] Abendroth ER, Greimann LF. Field testing of integral abutments. Final Report No.
HR-399. Ames, IA: Center for Transportation Research and Education. Iowa State
University; 2005 (http://www.ctre.iastate.edu/reports/hr399.pdfN).
[8] Kim W, Laman JA. Integral abutment bridge response under thermal loading. Eng
Struct 2010;32(6):1495–508.
[9] Shah B. 3D finite element analysis of integral abutment bridges subjected to thermal
loading. Kansas State University; 2011(MS thesis).
[10] Thanasattayawibul N. Curved integral abutment bridges. (University of Maryland);
2006PhD dissertation.
[11] Kalayci E, Civjan SA, Breña SF. Parametric study on the thermal response of curved
integral abutment bridges. Eng Struct 2012;43:129–38.
[12] McBride KC. Thermal stresses in the superstructure of integral abutment bridges.
West Virginia University; 2005(M.S. thesis).
[13] Geokon. Instruction manual model VK-4100/4150 vibrating wire strain gages.
Lebanon, NH: Geokon; 2009.
[14] Greimann L, Phares BM, Deng Y, Shryack G, Hoffman J. Field monitoring of
curved girder bridges with integral abutments. Report No. InTrans Project 08-323.
View publication stats
Ames, IA: Iowa State University; 2014 (http://www.intrans.iastate.edu/research/
documents/research-reports/curved_girder_integral_abutments_w_cvr.pdf,).
[15] Shryack G. Field monitoring and evaluation of curved girder bridges with integral
abutments. M.S. Thesis: Iowa State University; 2012.
[16] Greimann LF, Abendroth RE, Johnson DE, Ebner PB. Pile design and tests for integral
abutment bridges. Final Report, Iowa DOT Project HR-273. Ames, IA: Iowa State
University; 1987 (http://trid.trb.org/view.aspx?id = 275298).
[17] Greimann LF, Yang PS, Edmunds SK, Wolde-Tinsae AM. Design of piles for integral
abutment bridges. Final report, Iowa DOT project HR-252. Ames, IA: Iowa State
University; 1984 (http://ntl.bts.gov/lib/42000/42500/42500/hr252.pdf).
[18] ANSYS. Release 14.0 documentation for ANSYS, ANSYS help; 2011.
[19] Oesterle RG, Refai TM, Volz JS, Scanlon A, Weiss WJ. Jointless and integral abutment
bridges analytical research and proposed design procedures. Report DTFH61-92-C-
00154. FHWA, US Department of Transportation; 1998.
[20] Kim W, Laman JA. Seven-year field monitoring of four integral abutment bridges. J
Perform Constr Facil 2011;26(1):54–64.
[21] Fennema JL, Laman JA, Linzell DG. Predicted and measured response of an integral
abutment bridge. J Bridg Eng 2005;10(6):666–77.
[22] Hoffman J, Phares B. Thermal load design philosophies for horizontally curved girder
bridges with integral abutments. J Bridg Eng 2014;19(5):04014008.
[23] Coduto Donald P. Foundation design principles and practices. Second Edition. New
Jersey: Prentice Hall; 2001.