NONLINEAR ANALYSIS OF INTEGRAL BRIDGES:

FINITE-ELEMENT MODEL
By Susan Faraji,1 John M. Ting,2 Members, ASCE, Daniel S. Crovo,3 and Helmut Ernst4

ABSTRACT: Integral abutment bridges (IABs) are jointless bridges where the deck is continuous and connected
monolithically with the abutment walls. The biggest uncertainty in the design of these bridges is the reaction of
the soil behind the abutments and next to the foundation piles, especially during thermal expansion. This lateral
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

soil reaction is inherently nonlinear and is a function of the magnitude and nature of the wall displacement.
Handling the soil-structure interaction in the design of IABs has always been problematic, usually requiring
iterative, equivalent linear analysis. This paper describes the implementation of a full 3D finite-element model
of an IAB system which explicitly incorporates the nonlinear soil response. This paper also presents the results
from a small parametric study on a sample bridge where the soil compaction levels in the cohesionless soils
behind the wall and adjacent to the piles were varied. These results show that the level of compaction in the
granular backfill strongly dominates the overall soil reaction, and that this reaction greatly impacts the overall
structural response of the bridge system.

INTRODUCTION bridges with skew and nonskew alignments, and to evaluate

of the performance of these bridges during seismic excitation,
Integral abutment bridges (IABs) are jointless bridges where a new, more integrated analysis and design tool is necessary.
the deck is continuous and connected monolithically with the This paper describes the implementation of a full 3D finite-
abutment wall with a moment-resisting connection (Fig. 1). A element (FE) model of an integrated abutment bridge system
line of vertical piles beneath the abutment wall is often used which explicitly incorporates the nonlinear soil response be-
to carry the vertical bridge loads. The IAB concept has become hind the abutment walls and adjacent to the supporting piles.
increasingly popular in recent years due to reduced mainte- Using commercially available finite-element code GTSTRUDL
nance costs associated with the expansion joints and abutment (1991), the nonlinear soil behavior is handled using nonlinear
bearings, reduced corrosion and material degradation at the springs at the abutment wall and pile nodes. The nonlinear soil
joints, and better overall structural performance, particularly springs behind the abutment walls are the force-deflection de-
under seismic loading. sign curves recommended in the National Cooperative High-
The biggest uncertainty in the analysis and design of an ways Research Program (NCHRP) (1991) design manual. The
integral abutment bridge is the reaction of the soil behind the nonlinear p-y design curves recommended by the American
abutment walls and next to the foundation piles. The magni- Petroleum Institute (API) (1993) are used adjacent to the piles.
tude of these soil forces can become substantial during thermal This FE implementation streamlines the analysis and design
expansion of the bridge system, and can greatly affect the of IAB systems and is capable of analyzing both skew and
overall structural design of the bridge-wall-pile system. This nonskew bridge orientations, and should be able to evaluate
lateral soil reaction is inherently nonlinear and depends on the seismic response.
magnitude and nature of the wall displacement, which can A full 3D finite-element model was made of a sample high-
both translate and rotate. This is a soil-structure interaction way bridge in Fitchburg, Mass. This 45.7 m (150 ft) long,
problem, where the magnitudes and nature of the soil and three-span continuous bridge was modeled using over 1,000
structural deformations and stresses are interdependent. The beam, beam-column, and slab members and elements for the
left-hand-side of Fig. 1 shows a typical nonlinear lateral soil deck, wall, pier, and pile systems, and over 350 nonlinear sup-
force reaction as a function of wall deflection, while the right- port springs for the nonlinear soil behavior. The bridge model
hand-side of this figure shows a typical soil reaction pressure was subjected to a thermal loading increment of ⌬T = 44.4⬚C
distribution at one level of wall and pile deformation. (80⬚F). This paper presents this FE model together with the
Handling the soil-structure interaction in the analysis and results from a small parametric study where the soil compac-
design of integral abutment bridges has always been problem- tion levels in the cohesionless soils behind the wall and ad-
atic, usually requiring iterative analysis wherein the soil re- jacent to the piles were varied.
actions are manually adjusted depending on the deformation
level behind the abutment wall and adjacent to each supporting CONVENTIONAL ANALYSIS AND DESIGN
pile. Such an iterative, equivalent linear approach is time-con-
suming and approximate, and does not lend itself to dynamic The integral abutment bridge is a classic example of a soil-
analysis. With the desire to design and construct longer span structure interaction system. The soil reaction pressure distri-
bution on the wall is inherently nonlinear, and varies with
1
Prof., Dept. of Civ. and Envir. Engrg., Univ. of Massachusetts, Lowell, depth, amount, and mode of wall displacement. Depending on
MA 01854. E-mail: susan㛭[email protected] the relative flexural stiffness of the composite bridge deck,
2
Prof. and Head, Dept. of Civ. and Envir. Engrg., Univ. of Massachu- abutment wall, foundation piles, and lateral stiffness of the soil
setts, Lowell, MA 01854. E-mail: john㛭[email protected]
3
Asst. Struct. Engr., Massachusetts Hwy. Dept., 10 Park Plaza, Boston,
behind the wall and next to the piles, the amount and mode
MA 02116. E-mail: [email protected] of deformation of the wall varies. This in turn affects the level
4
Asst. Geotech. Engr., Massachusetts Hwy. Dept., 10 Park Plaza, Bos- and amount of the soil reaction pressure behind the wall. This
ton, MA 02116. E-mail: [email protected] soil-structure interaction also applies to the lateral soil pressure
Note. Discussion open until October 1, 2001. To extend the closing acting on the foundation piles.
date one month, a written request must be filed with the ASCE Manager The lateral soil resistance both behind the abutment wall
of Journals. The manuscript for this paper was submitted for review and
possible publication on July 23, 1999; revised November 30, 2000. This
and next to the laterally loaded pile can be thought of as a
paper is part of the Journal of Geotechnical and Geoenvironmental series of uncoupled ‘‘Winkler’’ springs, where the deflection
Engineering, Vol. 127, No. 5, May, 2001. 䉷ASCE, ISSN 1090-0241/01/ at one level of the wall is not presumed to affect the value of
0005-0454–0461/$8.00 ⫹ $.50 per page. Paper No. 21501. the reaction force at another level. Behind the abutment wall,
454 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1. Typical Soil Response Adjacent to Abutment and Piles

the lateral soil reaction to the superstructure expansion/con- bution is typically nonlinear, with a greater pressure closer to
traction can vary between the limits of the active (soil moving the ground surface, and fairly low pressures near the bottom
outward in the direction of the deck) to the passive (wall mov- of the wall (where the soil deformation is lower). Nevertheless,
ing into the soil) states. In general, these springs are nonlinear, in the typical simplified method, a linear soil pressure distri-
with some ultimate resistance dependent upon confining stress bution is assumed, and some average K is used to estimate the
level, soil type, level of soil compaction, soil strength, and magnitude and location of the resultant soil force acting behind
mode of wall deformation. Depending on the loading case, soil the abutment wall. This resultant force and its location are
type, and effective confining stress in the soil, and whether the used in the simplified analysis to estimate the deformation in
loading is due to a retaining wall or a laterally moving pile, the wall, which in turn may be used to refine the assumed
different equivalent-linear secant modulus values of the Wink- pressure distribution and average K value.
ler soil stiffness may be chosen for input into a frame or FE After an analysis of the bridge-wall system with simplified
structural analysis. soil response, the computed displacement levels are compared
At present, the design of integral abutment bridges usually with the assumed values, and the lateral soil pressure distri-
goes through an iterative process wherein the soil-structure bution and average soil K behind the wall and equivalent pile
interaction portion of the problem is greatly simplified. First, length are adjusted as necessary. The analyses are then redone
the magnitude of the lateral loads (or displacements) which until the assumed and calculated deflections match. Note that
must be transferred from structure to soil system are estimated this iteration is time-consuming, usually requiring human in-
—both behind the abutment wall and at the top of the vertical tervention at each step, and as a result is not always performed.
line of piles. These loads (or displacements) are used to esti- Also, since the analysis is equivalent linear rather than non-
mate the lateral soil stiffness behind the wall, and translational linear, the results of such an analysis are not necessarily ac-
and rotational stiffnesses of the soil-pile system. curate. Likewise, both the equivalent length of the pile-soil
Often, the soil-pile system is modeled as an equivalent system and the assumed lateral soil distribution and average
length of horizontally unsupported cantilever beam-column. K behind the wall can vary depending on the loading case, so
This latter approach may be accomplished using fairly these simplifying estimates need to be repeated each time a
straightforward analyses (CALTRANS 1981; Greiman et al. new load case is analyzed.
1987; Husain and Bagnariol 1996), or with the assistance of To streamline this iterative process and increase the accu-
computer program such as LPILE (Reese and Wang 1993), or racy of the equivalent-linear analysis, it was decided to im-
COM624 (Wang and Reese 1993). This reduces the nonlinear plement a full 3D finite-element analysis with nonlinear soil
soil-structure system to an equivalent linear structural system reactions built into the FE model. For the soil reactions behind
which is valid only at the assumed loading and displacement the abutment walls, the recommendations of the NCHRP
level. The resulting equivalent structural system is then ana- (1991) design manual are used, while the nonlinear p-y design
lyzed for different loading cases using 2D or 3D finite-element curves recommended by the API (1993) are used adjacent to
analyses. the piles.
For the soil behind the walls, the effective horizontal normal
stress ␴⬘h is usually related to the effective vertical normal FINITE-ELEMENT MODEL
stress ␴⬘␯ using a lateral earth pressure coefficient K = ␴⬘h /␴⬘␯ ,
3D Nonlinear Finite-Element Bridge Model
where this value of K is chosen depending on the level of
anticipated displacement. When the same value of K is as- The nonlinear soil behavior behind the abutment walls and
sumed for all depths, then this presumes a triangular distri- next to the vertical piles have been incorporated into a 3D FE
bution of horizontal earth pressure, with a resultant reaction bridge model using commercially available computer package
force located at H/3 above the base of the wall. The magnitude GTSTRUDL. Nonlinear analysis capabilities in GTSTRUDL
of this resultant would be F = (1/2)K␥H 2, where ␥ is the unit include (1) small-strain geometric; (2) material nonlinearity for
weight of the soil. plane and space truss members; and (3) material nonlinearity
In general, however, the lateral soil pressure distribution is for support reactions. Solutions for nonlinear systems are
not triangular, especially if there is some degree of wall ro- based on a direct iteration technique using secant stiffnesses.
tation about the base. For these cases, the soil pressure distri- In the current study, the structural elements of the bridge were
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001 / 455

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

 modeled as linear elements while the soil reactions adjacent modeled with nonlinear soil springs to a depth of 8.84 m be-
to the piles and behind the abutment walls were modeled as low ground surface. The 2.44 m-high abutment walls are mod-
nonlinear support springs. eled as plate elements with an overall effective height of 1.83
A full 3D FE model was made of the Bemis Road Bridge m. These were connected to the bridge superstructure with
over the Nashua River in Fitchburg, Mass. This is a typical fixed connections at the end of each girder. The soil response
short span, continuous steel stringer bridge with integral abut- behind the abutment walls are modeled using uncoupled non-
ments (Fig. 2). The superstructure consists of a concrete slab linear springs. Fig. 3 plots the FE mesh for one of the two
acting in composite with steel stringers. The deck slab is mod- abutments.
eled using bending and stretching plate elements while the The seven HP piles are fixed into each abutment wall, al-
steel stringers and diaphragms are modeled as beam elements. lowing full moment transfer. Each pile is modeled using beam
In initial analyses, the superstructure was modeled without re- elements with the top node of the pile the same as the bottom
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

gard for composite action for simplicity. Later analyses used node of the wall. Each abutment HP pile and drilled pier shaft
rigid links between the stringers and deck slab to ensure strain is modeled with 15 0.61 m-long beam elements down to a
compatibility and shear transfer between the deck slab and depth of 8.84. The p-y soil response next to each pile is mod-
girder elements. Details of the bridge construction and finite- eled with 15 nonlinear soil springs, as shown in Fig. 3 for the
element model, including the GTSTRUDL input file, are found north abutment wall. The values used in these springs are de-
in Ting and Faraji (1998). scribed in the next section. Fig. 4 plots a perspective view of
The pier caps at the two intermediate piers are modeled as the full bridge FE mesh in the undeformed state.
beam elements. The three reinforced concrete columns/drilled One load case that is of major concern for integral abutment
shafts at each bent are modeled with beam elements with in- bridges is the loading associated with thermal expansion. A
termediate nodes. In the current study, the 3D system was uniform temperature increase of 44.4⬚C (80⬚F) was applied to

FIG. 2. Elevation View Schematic of Bemis Road Bridge, Fitchburg, Mass.

FIG. 3. Elements, Nodes, Members, and Soil Springs for North Abutment Wall and HP Piles

456 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

FIG. 4. GTSTRUDL Finite-Element Model of Fitchburg Bridge in Un-

deformed State

the composite deck in this analysis. A coefficient of thermal

expansion of 1.2 ⫻ 10⫺5/⬚C (6.5 ⫻ 10⫺6/⬚F) is used.
For this preliminary study, the soil conditions behind the
abutment wall and next to the HP piles were varied as follows:
loose/dense, loose/loose, dense/loose, and dense/dense, respec- FIG. 5. Comparison between Design Curves for Lateral Earth Pressure
tively. For comparison, the bridge was analyzed as a full 3D K during Horizontal Translation and Bottom Wall Rotation
system as well as an equivalent 2D system. For the 2D system,
an equivalent ‘‘slice’’ of the bridge used was equivalent to one
HP pile, one deck girder, and a 2.74 m (9 ft) wide strip of
abutment wall and deck slab. As a simplification, the piers
were treated as (1) a simple pin; (2) a simple roller; or (3) a
full substructure consisting of rigid pier cap, column/drilled
shaft, and nonlinear soil springs.

Force-Deflection Response of Soil behind Rigid

Abutment Wall
To implement the appropriate nonlinear soil response curves
behind the rigid abutment wall, an extensive literature review
was first conducted of the theoretical and numerical analyses,
experimental data, and design curves currently available as
applied to integral abutment bridges. Since the backfill behind
the abutment wall is also under the approach slab to the bridge,
this backfill is usually a compacted cohesionless granular fill.
There are two predominant sets of design curves used for
estimating the nonlinear force-deflection relation behind a FIG. 6. Force-Deflection Relations for Main Inside Nodes from
rigid retaining wall. These curves typically include both active NCHRP (1991) Design Curves for Dense Sands
and passive regimes, and are usually used for both pure hor-
izontal translation as well as pure rotation about the base. One each wall element was computed using the dimensions shown
commonly used set of curves, for loose, medium dense and in Fig. 3. Typical interior panels in the FE model are of width
dense cohesionless granular materials, are based on finite-el- w = 1.37 m and height h = 0.61 m. This area is multiplied by
ement analyses by Clough and Duncan (1971), and are found the effective vertical normal stress ␴⬘␯ for a given panel depth
in several design manuals such as Clough and Duncan (1991), z and by the lateral earth pressure coefficient K for a given
NCHRP (1991), and Husain and Bagnariol (1996). A second deflection to yield a lateral force-deflection curve for a given
common set of design curves are found in U.S. Department node
of the Navy (1971) and the Canadian Geotechnical Society
F = K␴⬘wh
␯ (2)
(CGS) (1992). As can be seen in Fig. 5, these curves are simi-
lar, except the NCHRP curve for dense materials is much Since the wall actually extends an additional 0.30 m above
lower than the corresponding Canadian FEM curve. The and below the top and bottom lines of nodes, respectively,
NCHRP design curves were used in the current study. these interior nodes all have the same panel width and effec-
The general form of the NCHRP lateral earth pressure K tive height. The nodes at the ends and over the outside line of
versus deflection design curves was digitized for loose, me- HP piles have different effective widths. These details were all
dium dense, and dense soils. These curves relate the horizontal input into the spreadsheet program, and the resulting nodal
normal stress ␴⬘h to the vertical effective normal stress ␴⬘␯ ac- force-deflection curves were computed for each line of nodes.
cording to Fig. 6 plots the ‘‘typical’’ force-deflection relations for these
inside nodes for dense sand at different depths.
␴⬘h = K␴⬘␯ (1)
Lateral Force-Deflection Response of Soil Adjacent
where for a uniform density dry soil, ␴⬘␯ = ␥z, where ␥ = dry
to Vertical Piles
density of the soil.
To calculate the effective soil spring resistance for input into Most bridge abutments utilize either spread footings or pile
the finite-element bridge model, the effective panel size of foundations at the bottom of the wall to support the bridge
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001 / 457

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

 loads. Piles are often favored over spread footings because of ultimate soil resistance pus and pud can be made. The lower of
erosion and vertical deformation considerations. For integral these two values is used as the ultimate resistance, pu at a given
abutment bridges, it is commonly thought that a foundation depth. The initial stiffness ki is chosen for a given angle of
that is rotationally and translationally stiff results in larger su- internal friction ␸. These are used to calculate the soil resis-
perstructure loads during thermal loading. In order to increase tance p (units force/length) as a function of deflection y. These
the lateral and rotational flexibility of the pile head, piles for values of p are then multiplied by the height of each beam
an IAB foundation are often driven into preaugered 3 m (10 element to obtain the soil spring force as a function of lateral
ft) deep holes which are then backfilled with loose sand. Also, displacement at each depth.
the piles are often installed with their weak axis of bending Fig. 7 plots the resulting force-deflection curves adjacent to
parallel to the bridge centerline. the piles for loose sand conditions (␸ assumed as 30⬚) adjacent
Most analysis and design methods for the lateral loading of to a 0.30 m (12 in.) wide pile. Spacing depth between curves
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

vertical piles involve modeling the pile as an elastic beam- is 0.61 m (two pile diameters or 24 in.), and the pressure has
column, and the soil as a series of uncoupled soil springs. Most been converted to equivalent spring force as a function of lat-
current design methods for integral abutment bridges assume eral deflection. The force-deflection plots are shown at differ-
that these soil springs exhibit equivalent linear response even ent depths below the ground surface, with the depths expressed
though it is widely known to be nonlinear and hysteretic. The as in pile diameters. Similar curves were generated for medium
most sophisticated nonlinear soil modeling for lateral pile dense and dense sands. In the actual construction of the inte-
loading is currently done by the oil industry. Many of the gral abutment bridge, the top 3 m of each pile is normally
current computer programs for modeling lateral pile behavior preaugered and backfilled with loose granular material. Con-
such as LPILEplus (Reese and Wang 1993) and COM624P sequently, one could assume that the lateral soil resistance dur-
(Wang and Reese 1993) are based on the API (1993) design ing the first cycle of thermal loading should be relatively soft.
guidelines for fixed offshore platforms. However, it is anticipated that this backfill would compact and
At a given depth, z, the p-y design curve recommended by densify naturally over time.
the API (1993) is a continuous hyperbolic tangent curve
RESULTS
k1 ⭈ z 
p = A ⭈ pu tanh  ⭈y (3)
A ⭈ pu  The 3D finite-element model of the bridge was subjected to
a thermal increment of ⌬T = 44.4⬚C (80⬚F). Soil properties
where pu = estimated ultimate resistance calculated based on were varied behind the wall and adjacent to the piles resulting
approximate failure analysis with units of force/length; k1 = in four test combinations: loose (behind the abutment)/loose
soil stiffness (units of force/length3); and A = empirical factor (next to the pile), loose/dense, dense/loose, and dense/dense.
accounting for cyclic or static loading. All three terms are pre- Of these, the dense/loose case is expected to most closely
sumed to vary with depth. match the typical IAB construction where the backfill behind
The current study used the API recommendations to gen- the wall is compacted and the HP pile is installed into a
erate the nonlinear soil pressure-pile deflection p-y curves as preaugered hole with initially loose sand.
a function of depth for each pile. To generate the specific p-y Fig. 8 shows the exaggerated deflected shape of the FE
curves adjacent to the HP piles beneath the abutment walls, mesh during thermal expansion. Fig. 9 plots the lateral deflec-
the equations for this method were implemented in a spread- tion of the abutment wall and connecting HP piles as a func-
sheet program. Then, for a given depth, soil density, and soil tion of depth for each of the four soil compaction cases (‘‘LD’’
angle of internal friction ␸, estimates of the shallow and deep refers to loose next to wall and dense next to pile). Fig. 10
plots the average normal stress behind the abutment wall,
while Fig. 11 plots the moment in the HP piles. Note that the
level of compaction in the backfill behind the wall dictates the
pile top deflection as well as the magnitude and location of
the peak moments in the piles. Also, note the moments in the
HP piles are at a peak near the top of the piles for all four
load cases, owing to the increased restraint from rotation at
the bottom of the wall.
The moments in the superstructure were also computed. Fig.
12 plots the composite moments in the center girder of the

FIG. 7. Force-Deflection Relations for Pile Nodes at Varying Depths

from API (1993) Design Curves for Loose Sands (Depths from 1 to 29 FIG. 8. Deformed Shape of FE Mesh after Thermal Loading (Deflec-
Pile Diameters below Ground Surface; Nodes Spaced 0.6 m Apart) tions Exaggerated)

458 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

 modeled using an equivalent section equal to approximately
1/7 of the cross section of the bridge, since there are seven
lines of stringers and seven HP piles. The deck and slab are
modeled using beam elements equivalent to one stringer plus
a 2.74 m (9 ft) wide strip of the bridge deck slab. The flexural
stiffness EI of the section accounting for composite action is
1.89 ⫻ 109 N-m2 (6.6 ⫻ 108 kips-in.2).
The seven HP piles are reduced to a single pile in the 2D
model. There are only three piers beneath each pier cap. As a
simplification, each of the two piers may be treated as (1) a
roller support since the bent is very flexible; or (2) a full sub-
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

structure with a rigid pier cap, equivalent 2D pier (with 3/7 of

the flexural stiffness of a single pier), and nonlinear soil
springs. The joints between the superstructure and abutment
walls are assumed to be moment connections, allowing full

FIG. 9. Deflection of Abutment Wall during Thermal Loading for Dif-

ferent Soil Conditions (Expansion to Left) (DL = Dense Next to Wall,
Loose Next to Pile)

FIG. 11. Moment in HP Piles under Abutment Subject to Thermal

Loading of ⌬T = 44.4⬚C (80⬚F)

FIG. 10. Average Horizontal Normal Stress ␴h behind Abutment Wall

during Thermal Loading (Expansion to Left)

superstructure for each of the soil compaction levels for the

thermal loading case.
In an integral abutment bridge, the axial forces in the girders
cannot be ignored. The results show that the magnitude of the
axial forces in the girders varies considerably depending on
the compaction level behind the abutment walls. For the dense
compaction case, maximum axial force in the girder was be-
tween 1.04 MN (234 kips) and 1.14 MN (257 kips), depending
on the compaction next to the piles, and between 445 kN (100
kips) and 596 kN (134 kips) for loose compaction behind the FIG. 12. Composite Moments in Girder Plus 2.74 m (9 ft) Wide Deck
abutment walls. Slab for Different Soil Conditions at Thermal Loading ⌬T = 44.4⬚C
Equivalent 2D analyses were performed with the bridge (80⬚F)

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001 / 459

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

 moment transfer. Similarly, the joints between the abutment atic. A full 3D finite-element model of an integral abutment
walls and HP piles are assumed to be moment connections. bridge system has been constructed which automatically in-
The results from these analyses indicate that the 2D roller sup- corporates the nonlinear soil response behind the abutment
port and full bent models exhibit nearly identical results and walls and adjacent to the supporting piles, and can model both
compare very favorably with the full 3D analysis results. More skew and nonskew bridge orientations. Using GTSTRUDL, the
details on these results are found in Ting and Faraji (1998). nonlinear soil behavior is handled using nonlinear springs at
the abutment wall and pile nodes. The nonlinear force-deflec-
DISCUSSION tion design curves recommended in NCHRP (1991) were used
to represent the soil response behind the abutment wall while
The results from these various nonlinear FE analyses show the nonlinear p-y curves currently recommended by API
that it is crucial that care be taken in the modeling of the (1993) were used for the lateral soil response next to the ver-
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

superstructure. In particular, the composite action of the slab tical piles.

and stringers must be properly modeled. Failure to do so alters A 45 m-long, three-span sample bridge was modeled using
the relative stiffness of the deck compared with the wall-pile- over 1,000 beam, beam-column, and slab members and ele-
soil system, which can result in dramatically different struc- ments for the deck, wall, pier, and pile systems, and over 350
tural response, especially in the moments in the composite nonlinear support springs for the nonlinear soil behavior. The
bridge deck and HP pile system. bridge was subjected to a thermal loading increment of ⌬T =
Based on the 3D analyses where soil compaction levels 44.4⬚C. The compaction levels behind the walls and adjacent
were varied, the results show that the level of soil compaction to the piles were varied between the loose and dense condi-
behind the abutment wall is a vitally important factor affecting tions. For this loading, the program converged fairly rapidly
the overall bridge behavior. The level of compaction behind with a convergence tolerance of 0.01 to a stable equilibrium
the wall greatly affects the axial forces and moments in the result. Results indicate that care must be taken to properly
deck, increasing both by more than twice in peak value when model the composite action of the superstructure.
the compaction is varied from the loose to dense soil com- Results from this study show that the level of soil compac-
paction range. The magnitude of the peak moments in the HP tion behind the abutment wall is a vitally important factor
piles decreases by a factor of 2 when the compaction is varied affecting the overall bridge behavior. This degree of soil com-
from loose to dense. In all cases, the peak moment in the pile paction greatly affects the axial forces and moments in the
occurs at the wall/pile interface. deck, increasingly by more than twice in peak value when the
For the 45 m-long sample bridge, the thermal load caused compaction is varied from the loose to dense soil compaction
a deck level horizontal displacement of about 1.1 cm at each range. The level of soil compaction adjacent to the HP piles
abutment. The full passive soil resistance is almost achieved does not significantly influence abutment wall or superstruc-
near ground surface. At greater depths behind the wall, the soil ture deflections or moments, but does impact the moments in
pressure was about 50% of the passive value even though hor- the HP piles.
izontal displacements were relatively small. The variation in Results from this study and companion research conducted
the soil pressure with depth behind the wall is slightly nonlin- at the University of Massachusetts–Amherst (Thomson and
ear (Fig. 10). For longer bridges, one would expect the soil Lutenegger 1998) are being used to refine the design procedure
pressures behind the wall to approach full passive values and for integral abutment bridges in Massachusetts (‘‘Mass-
be considerably nonlinear. Highway’’ 1999). Future work to be done includes assessing
The level of soil compaction adjacent to the HP piles does the impact of varying the nonlinear force-deflection design
not significantly influence abutment wall or bridge deck de- curve for the backfill. In particular, the force-deflection design
flections or moments. However, the moment in the HP pile is curves in CGS (1992) and ‘‘MassHighway’’ (1999) need to be
considerably affected by the relative density of the soil both implemented and assessed. Full-scale passive wall loading data
next to the pile and behind the abutment. For dense soil behind from a Massachusetts Highway Department-funded research
the abutment, variation from loose to dense sand next to the program at the University of Massachusetts–Amherst (Thom-
piles increases the peak moment in the pile from 28 to 50 kN- son and Lutenegger 1998) will be evaluated and compared. A
m (Fig. 11). This is attributed to the (approximately same) nonlinear FE model will be built to simulate the University of
lateral deflection of the pile head being carried over a shorter Massachusetts–Amherst experimental wall setup using nonlin-
effective length of pile in the dense soil-pile system. ear design curves and actual force-deflection data.
When dense backfill exists behind the abutment wall, the To help evaluate actual performance of a full-scale bridge
moments in the HP piles decrease by a factor of 2 regardless system, a 3D FE model will be developed for a new fully
of the density adjacent to the piles. This is due to the reduced instrumented bridge currently under construction in Orange,
deflection levels at the base of the wall when the abutment Mass. As part of an ongoing research project funded by the
backfill is dense. This directly results in reduced pile head Massachusetts Highway Department, field data will be contin-
displacements, and hence reduced pile moments. uously collected and reduced. This field data includes lateral
Results from equivalent 2D models of the sample bridge soil pressures behind the abutment walls and flexural strains
constructed using nonlinear soil springs behind the walls and in the HP piles. This reliability of the FE model will be as-
adjacent to the piles indicate that for nonskew alignments, the sessed by comparing FE results with observed bridge behavior.
2D analysis can be quite good compared with the full 3D With increasing confidence in the nonlinear 3D FE analysis
analysis. Nearly identical stringer moments and HP pile de- for IABs, the impact of design and construction details should
flections and moments were obtained in each case. Modeling be examined. Topics that should be considered include the
soil nonlinearity at piers is probably unnecessary if the pier is impact of preaugering holes for the HP piles and backfilling
very flexible. This is the case in the sample bridge, which with loose granular fill, the impact of skew alignments on the
possessed three drilled shafts at each bent connected to the forces and moments at the abutment/superstructure joint, and
superstructure via a roller connection. the seismic response of long-span skew and nonskew IABs.
These studies should lead to a better theoretical understanding
CONCLUSIONS of the behavior of skew and nonskew, long-span integral abut-
ment bridges during thermal, gravity, and seismic loadings.
The handling of soil-structure interaction in the analysis and This work should help streamline the design process for IABs
design of integral abutment bridges has always been problem- in general.
460 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001

J. Geotech. Geoenviron. Eng. 2001.127:454-461.

 ACKNOWLEDGMENTS National Cooperative Highway Research Program (NCHRP). (1991).
Manuals for the design of bridge foundations, R. M. Barker, J. M.
This work was funded by the Massachusetts Highway Department and Duncan, K. B. Rojiani, P. S. K. Ooi, C. K. Tan, and S. G. Kim, eds.
the U.S. Department of Transportation, Federal Highway Administration Rep. 343, Transportation Research Board, Washington, D.C.
through the University of Massachusetts Transportation Research Center. Reese, L. C., and Wang, S.-T. (1993). ‘‘LPILEPLUS —Stress and defor-
Their support is gratefully acknowledged. mation analysis of piles under lateral loading.’’ Ensoft, Inc., Austin,
Tex.
REFERENCES Thomson, T. A., and Lutenegger, A. J. (1998). ‘‘Passive earth pressure
tests on an integral bridge abutment.’’ Proc., 4th Int. Conf. Case His-
American Petroleum Institute (API). (1993). ‘‘Recommended practice for tories in Geotech. Engrg.
planning, designing, and constructing fixed offshore platforms—Work- Ting, J. M., and Faraji, S. (1998). ‘‘Streamlined analysis and design of
ing stress design.’’ 20th Ed., API RP2A-WSD, Washington, D.C. integral abutment bridges.’’ Rep. UTMC 97-13, University of Massa-
California Department of Transportation (CALTRANS). (1981). Bridge chusetts Transportation Center, Amherst, Mass.
Downloaded from ascelibrary.org by KMUTT KING MONGKUT'S UNIV TECH on 10/19/14. Copyright ASCE. For personal use only; all rights reserved.

design practice manual, Sacramento, Calif. U.S. Department of the Navy. (1971). ‘‘Design manual—Soil mechanics,
Canadian Geotechnical Society (CGS). (1992). Canadian foundation en- foundations, and earth structures.’’ NAVFAC DM-7, Naval Facilities
gineering manual, 3rd Ed., Toronto. Engineering Command, Alexandria, Va.
Clough, G. W., and Duncan, J. M. (1971). ‘‘Finite-element analyses of Wang, S. T., and Reese, L. C. (1993). ‘‘COM624P—Laterally loaded pile
retaining wall behavior.’’ J. Soil Mech. and Found. Div., ASCE, 97(12), analysis for the microcomputer, ver. 2.0.’’ FHWA-SA-91-0048, Spring-
1657–1673. field, Va.
Clough, G. W., and Duncan, J. M. (1991). ‘‘Earth Pressures.’’ Chapter 6,
Foundation engineering manual, 2nd Ed., H. Y. Fang, ed., Van Nos-
trand Reinhold, New York, 223–235. NOTATION
Faraji, S. (1997). ‘‘Behavior of integral abutment bridges in Massachu-
setts—Year II.’’ Rep., University of Massachusetts Transportation Cen- The following symbols are used in this paper:
ter, Amherst, Mass.
Greiman, L. F., Abendroth, R. E., Johnson, D. E., and Ebner, P. B. (1987).
‘‘Pile design and tests for integral abutment bridges.’’ Final Rep., Iowa F = total lateral wall force;
State University, Ames, Iowa. H = height of wall;
GTSTRUDL user’s manual; vol. 1, rev. M. (1991). GTICES Sys. Lab., K = lateral earth pressure coefficient ␴h /␴␯ ; also computed using
Georgia Institute of Technology, Atlanta. F = (1/2)K␥H 2;
Husain, I., and Bagnariol, D. (1996). ‘‘Integral abutment bridges.’’ Rep.
No. SO-96-01, Struct. Ofc., Ministry of Transportation of Ontario,
⌬ = lateral wall displacement;
Toronto, Canada. ␥ = unit weight of soil;
‘‘MassHighway Bridge Manual.’’ (1999). Design guidelines and standard ␴⬘h = horizontal normal stress;
details for integral abutment bridges, Draft, Massachusetts Highway ␴⬘␯ = vertical normal stress; and
Department, Boston. ␸ = angle of internal friction.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2001 / 461

J. Geotech. Geoenviron. Eng. 2001.127:454-461.