Sunteți pe pagina 1din 8

Engineering Structures 32 (2010) 10381045

Contents lists available at ScienceDirect

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

Time-dependent analyses of segmentally constructed balanced


cantilever bridges
Richard Malm , Hkan Sundquist
Department of Civil and Architectural Engineering, Royal Institute of Technology (KTH), SE-100 44 Stockholm, Sweden

article

info

Article history:
Received 21 January 2009
Received in revised form
24 July 2009
Accepted 22 December 2009
Available online 15 January 2010
Keywords:
Balanced cantilever
Segmental construction
Cast-in-place
Creep
Shrinkage
Deflection

abstract
Segmentally constructed concrete cantilever bridges often exhibit larger deflections than those predicted
by the design calculations. The slender and long spans in combination with the fact that permanent loads
are only partially compensated for by prestressing are reasons for the large deflections that increase
during the life time of the bridge, although at a decreasing rate. The rate of drying shrinkage may be one
reason for the accelerating displacement of cast-in-place bridges. The construction of continuous spans
instead of introducing joints has both comfort and durability advantages. The continuous span is however
more complicated to design, and secondary restraint moments due to creep, shrinkage and thermal
effects develop at the connection. The results of analyses of the stepwise cast-in-place construction of
a balanced cantilever bridge with time-dependent material properties show both higher deflection than
those originally assumed in the design calculations and high stresses in the webs due to stressing of the
tendons in the bottom flange. The analyses show significant effects of creep during cantilevering and of a
non-uniform drying shrinkage rate on the continuous bridge.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
Prestressed segmentally constructed concrete bridges are
sensitive to a long-term increase in deflection and are often subjected to an increasing long-term deflection. The total vertical
displacement of such bridges is a result of a large downward displacement due to the dead load, live loads and a large upward
displacement due to prestress. The long-term increase in displacements is of great importance for the serviceability, durability and
reliability. Due to this, it is important to be able to obtain accurate
predictions of the deformation of these bridges during construction and their service life. Several bridges have been closed or repaired due to excessive deflection before the end of their initially
assumed service life. The cost of a reduced service life is tremendous for society, the owners and users.
Box-girder bridges are traditionally analysed according to
theory of bending where the cross-sections are assumed to remain
plane. This theory is, however, too simplified to capture the
deformation of boxgirder bridges accurately. The main deficiency
of this theory is that it cannot capture the shear lag effect in the
slabs due to the dead weight and the prestress. The shear lag
causes a nonlinear distribution of normal stresses over the top
and bottom flanges in the cross-section. Neglecting the shear lag
effect may lead to a considerable underestimation of the long-term

Corresponding author. Tel.: +46 8 790 8585; fax: +46 8 216949.


E-mail address: Richard.Malm@byv.kth.se (R. Malm).

0141-0296/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.12.030

deflections of boxgirder bridges. A 3D finite element (FE) model


consisting of shell or solid elements can automatically capture the
effect of shear lag and can also capture the effects of differential
shrinkage and drying creep if suitable material descriptions are
used.
Few examples where time-dependent effects have resulted in
cracking in cast-in-place balanced cantilever bridges are found
in the literature. The literature regarding time-dependent effects
in this type of bridge mainly focuses on large long-term deflections [15]. In the study of Kristek and Vrablik [6], a program to
optimize the tendon layout to counteract the increasing long-term
deflections is presented. Previous stepwise analyses of balanced
cantilever bridges include visco-elastic creep, but do not include a
non-uniform shrinkage rate, in studies of precast concrete [7] and
with the prestressing implemented as equivalent nodal forces [8].
After only a few years of service, two similar bridges in Sweden, both segmentally constructed with the balanced cantilever
technique, had to be closed to traffic due to extensive cracking in
the webs. The hypothesis is that the cracks in these bridges are
due to the stressing of the tendons in the bottom flange in combination with the fact that there are no tendons in the web. The
purpose of this paper is to report a study of the influence of timedependent effects in the construction stage. It is particularly important to study which effects are likely to cause cracking and must
therefore be included in order to create an accurate model that
can describe the cracking. This study is based on a finite element
analysis of a segmentally constructed balanced cantilever bridge
that describes the stepwise construction with the nonlinear timedependent development of the material properties.

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

0
H1

V1

12 13

1039

10

6
34

34

33

33

32

9
74

120

70

54

Fig. 1. Elevation of the Grndal bridge in the construction stage.

4
4
4
4
4
4
4
4
4
2
2
2
2

4
4
4
4
4
4
4
4
4
2
2
2
2
2
2
2
2
2
2
2
4

Fig. 2. Elevation of the Grndal bridge with the tendon arrangement and the extent of cracking on the web facing south.

1.1. Description of the bridges


The Grndal bridge and the Alvik bridge had to be closed to traffic due to extensive cracking in the web of their boxgirder sections. These cracks were first found only a few years after service.
The bridges are parts of the light-rail commuter line in Stockholm,
Sweden. The inclined web cracks were first observed in an inspection only 2 years after completion. Subsequent bridge inspections
showed that the cracks were increasing both in number and in
size. The largest cracks were observed near the quarter-point of
the webs at the inside of the boxgirder section and they were up
to 0.6 mm wide. Since the webs were more cracked on the inside
of the boxgirder section it was considered probable that thermal
effects, under summer conditions, might be one of the factors causing the cracks. The designers feared that a shear failure might be
imminent unless the bridges were closed to traffic. The inclined
web cracks were initially assumed to jeopardise the ultimate limit
safety. During a temporary closure, the bridges were strengthened.
Information regarding the strengthening using a combination of
carbon-fibre laminates and vertical Dywidag tendons can be found
in the literature [911]. A previous investigation [12] suggested
that the cracking was due to inadequate shear reinforcement in the
webs in the serviceability limit state.
Both the Grndal bridge and the Alvik bridge are prestressed
continuous hollow boxgirder bridges. The Grndal bridge consists
of 11 spans with a total length of 430 m. Fig. 1 shows the
elevation of this bridge. The main and the two adjacent spans were
constructed with the balanced cantilever construction technique
while the side spans were erected span by span on a supporting
scaffold. Ten of the twelve piers of the Grndal bridge have a rock
foundation while the remaining two piers are built on piles. The
highest pier on the Grndal bridge is 34 m.
The cross-sectional height of the superstructure is approximately 7.50 m above the piers and about 2.75 m in the mid-span.
The webs are relatively slender with a thickness of 0.35 m and have
a rather low amount of reinforcement: horizontal reinforcement
with a diameter of 12 mm and 200 mm spacing and vertical web
reinforcement with a diameter of 16 mm and 200 mm spacing. The

amount of reinforcement is increased in the mid-span to a diameter of 20 mm in the horizontal bars. Prestressing cables are provided in the upper flange as they are necessary in the construction
stage, and the cables in the bottom flange are post-tensioned after
the completion of the superstructure when the centre segment is
cast. The tendon arrangement for the main-span is shown in Fig. 2
together with a sketch of the extent of cracking in the web.
1.2. Balanced cantilever construction
The principle of the free cantilever construction method is that a
previously cast segment serves as the work basis for the execution
of the next segment. A form traveller is attached to the previously
cast segment and carries the form work for the new segment that
is to be cast. An illustration of a form traveller is shown in Fig. 4.
According to Hewson [13], the weight of the traveller used for insitu construction with the balanced cantilever technique is usually
40120 tonnes for spans between 50 and 200 m. This interval in
the weight of traveller is slightly smaller according to Takcs [3]
where it typically weights 500900 kN. After a segment is poured
the traveller remains as a support for the newly cast segment
until it has reached sufficient strength and can be stressed to
the existing cantilever arm with post-tensioned tendons anchored
in the new segment [14]. To compensate for the long-term
deflections, an upward displacement during cantilevering occurs
due to tensioning of the tendons. These planned displacements are
commonly referred to as camber.
The main-span of the Grndal bridge was symmetrically cast
from piers seven and eight, see Fig. 1. The cantilever arms consist
of 13 segments, each 4 m long, from the piers, and where two
adjacent cantilevers meet they are joined with one 1.4 m long
centre segment to close the structure. The segments were cast with
a travelling form at intervals of 1 week.
2. Finite element analysis
The analyses presented in this paper have been performed with
the finite element (FE) software Abaqus/Standard 6.7 [15]. The
modelling approach used is a three-dimensional model using shell

1040

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

Fig. 3. FE model of the bridge.

A
SECTION

A-A

Fig. 4. Illustration of a form traveller.

elements. Thereby, the FE model automatically captures the effect


of shear lag. A numerical analysis with the program is divided into
steps, each corresponding to a load change from one magnitude to
another. In this case, a step represents the casting of one segment.
The segmental casting has been modelled in separate steps where
new elements have been introduced into the model in each step.
The newly introduced elements are given material properties that
develop over time to describe that the concrete cures. The FE model
used for the analysis consists of shell elements in the main and the
two adjacent spans. The remaining spans and all piers have been
defined as beam elements, as shown in Fig. 3.
2.1. Evolution of material properties
There are several material properties and phenomena that have
some effect on the response of the structure. The evolution of
material parameters such as the elastic modulus, creep, relaxation
and shrinkage has been described according to the methods in the
design codes CEB-FIP Model Code 1990 [16] and Eurocode 2 [17].
One major problem with most concrete material models used
to describe creep is that they cannot be combined with the
material models used to describe cracking. This is the case with the
visco-elastic material model, visco in Abaqus, which cannot be
combined with the material model suitable for describing concrete
cracking, concrete damaged plasticity in Abaqus. This
means that to analyse the effects of creep and cracking, some
strategy to compensate for this has to be adopted. The study
presented in this paper focuses on identifying the time-dependent
effects that have to be included to accurately describe the cracking
that occurred in the Grndal bridge.
2.1.1. Elastic modulus
In this construction process, the concrete is loaded at an early
age, where the concrete has to carry a load at a low degree of
maturity. Concrete increases in strength and stiffness as a result
of curing. At an early age, the strength and stiffness increase
quickly and the increase then gradually stagnates but does not stop
completely.

In the development of the material in the finite element


analyses, the evolution of the elastic modulus was implemented
according to the CEB-FIP Model Code 1990, [16]. The material
properties corresponding to a concrete age of 28 days can be
calculated based on the compressive strength according to
Ec = E

1/3

fcm

(1)

fcm0

where Ec is the modulus of elasticity of concrete at an age of 28


days (MPa), E = 2.15 104 (MPa), fcm is the compressive strength
of concrete at an age of 28 days (MPa) and fcm0 = 10 (MPa).
When an elastic analysis is performed, a lower value of the
modulus of elasticity should be used to take into account the initial
plastic cracking due to the plastic shrinkage. It is suggested that
this is done by decreasing the elastic modulus according to Ecs =
0.85Ec , where Ecs is the secant modulus of elasticity in the elastic
range for concrete.
To take into account concrete of an arbitrary age, the timedependent function may be used:

r
Ec ( t ) =


s 1

exp

28
t

Ecs

(2)

where t is the age of the concrete in days and s is a coefficient


depending on the cement type and is equal to 0.20 for rapidly
hardening cement for high strength concrete, 0.25 for normal and
rapidly hardening cement and 0.38 for slowly hardening cement.
The implemented development of the elastic modulus is illustrated in Fig. 5(a). The calculation is based on average values of the
compressive strength of 10 specimens made from the concrete mix
used in the Grndal bridge. To include the increase in elastic modulus in the finite element analysis, a field variable was introduced
that described the evolution. The field variable was defined to correspond to the time after casting for each segment in the FE analysis. After a new segment had been introduced, the material was
given an initial elastic modulus which increased as the total time
in the analysis increased, as shown in Fig. 5(a).

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

1041

Fig. 5. Development of (a) elastic modulus, (b) creep coefficient, (c) shrinkage strain and (d) non-uniform shrinkage rate.

2.1.2. Shrinkage
Shrinkage has been implemented according to Eurocode 2 [17]
where the total shrinkage is the sum of the autogenous and the
drying shrinkage. The drying shrinkage is defined as

cd = ds (t , t0 )kh cd,0

(3)

with

ds (t , t0 ) =

(4)

(t ts ) + 0.04 h3
ds2 ffcm
c0

(5)

with

RH = 1.55 1

100

3 !
(6)

where RH is the relative humidity of the ambient environment (%),


h = 2Ac /u is the notional size of the structural member (mm),
Ac is the area of the cross-section (mm2 ), u is the perimeter of
the cross-section in contact with the atmosphere (mm), fcm is the
average compressive strength of concrete at an age of 28 days
(MPa), fcm0 = 10 (MPa), ds1 and ds2 are coefficients depending
on the cement type and are equal to 6 and 0.11 respectively for
rapidly hardening high strength cement, 4 and 0.12 respectively for
normal and rapidly hardening cement and 3 and 0.13 respectively
for slowly hardening cement.
The autogenous shrinkage strain develops due to chemical
reactions during hardening in the early age concrete. Autogenous

(8)

and

cd,0 = 0.85 106 RH (200 + 110ds1 ) exp

(7)

where

as = 1 exp0.2

and the basic drying shrinkage strain is calculated as

RH

ca (t ) = as (t )ca ()
ca () = 2.5(fck 10) 106

t ts

shrinkage is of special importance when younger concrete is cast


against older already hardened concrete [3,17]. It can, according
to [17], be expressed as

(9)

The shrinkage can be described by applying an external thermal


load in a FE analysis. When shrinkage is introduced in the FE
analyses, differential shrinkage between the segments is always
considered. The most common option when including shrinkage
is to assume that the shrinkage is constant over the cross-section.
In this case, the bottom flange is at most five times thicker than
the top flange. This will have a considerable effect on the time the
drying shrinkage occurs, but the drying creep will not be affected
as much, according to Bazant and Bajewa [18]. Kristek et al. [4]
describe a case where the bottom flange is almost twice as high
as that in the Grndal bridge and the top flange is of comparable
thickness. The difference in drying creep for these two flange
thicknesses is less than 10%.
In the following analyses, two approaches have been made
to study the influence of shrinkage. In the first approach, each
segment is assigned a single shrinkage curve for the whole crosssection, i.e. one notational size for the whole cross-section of each
segment, as shown in Fig. 5(c). In the second approach, referred
to as a non-uniform shrinkage rate, the webs, the top and bottom

1042

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

flanges of each segment are assigned different shrinkage curves, as


shown in Fig. 5(d). The notational size is calculated separately for
these parts in the cross-section to include the effect of thickness
dependence on the drying shrinkage. A similar approach was
made in [3], with satisfactory accuracy compared to the more
advanced creep and shrinkage model B3 developed in [18]. The
shrinkage strain is represented by a corresponding temperature
and introduced into the finite element model, where negative
temperatures are introduced to describe shrinkage.
2.1.3. Creep
Creep is accounted for by using the description in CEB-FIP
Model Code 1990, [16]. Model Code was also used by the company
designing the bridges. The creep coefficient is calculated as

(t , t0 ) = 0 c (t t0 )

0 = RH (fcm )(t0 )

(11)

with
1 RH /100

(12)

0.46(h/100)1/3

(fcm ) =

5.3

(13)

fcm /10

(t0 ) =

1
0.1 +

(14)

t0

The time-development function is described by

c (t t0 ) =

0.3

t t0

(15)

H + t t0

with

H = 150 1 + 1.2

RH
100

18 !

E0
3(1 2)

h
100

+ 250 1500.

E0
2(1 + )

KR ( t )
K0

E0
2(1+)(1+(t ,t0 ))
E0
2(1+)
E0
3(12)(1+(t ,t0 ))
E0
3(12)

(19)

(1 + (t , t0 ))
1

(1 + (t , t0 ))

(20)

2.1.4. Relaxation
The relaxation of the prestressing tendons can be defined with
the same visco-elastic material model that is used to define creep
in concrete. The relaxation has been implemented according to
Eurocode 2 [17] where, for low relaxation wire and strands, the
relaxation loss can be calculated according to

1pr
= 0.66 105 1000 exp9.1
pi

kR (t ) =

(17)

(18)

There are four ways of defining the visco-elastic relaxation


parameters in Abaqus: by direct specification of the prony series,
by inclusion of creep data, by inclusion of relaxation test data or by
inclusion of frequency-dependent data obtained from sinusoidal
oscillation experiments. In the present case, the relaxation test
data were used to specify the visco-elastic behaviour. The
normalised shear and bulk moduli, gR (t ) and kR (t ), were defined
as functions of the creep coefficient:

t
1000

0.75(1)
(21)

where 1pr is the absolute value of the relaxation losses of the


prestress, pi is the absolute value of the initial prestress for posttensioning, t is the time after tensioning (in h), = pi /fpk where
fpk = 1770 (MPa) is the characteristic value of the tensile strength
of the prestressing steel, and 1000 = 2.5% is the value of relaxation
loss at 1000 h after tensioning and at an average temperature of
+ 20 C.
Relaxation is implemented as relaxation test data, with the
normalized shear and bulk moduli, gR (t ) and kR (t ), defined as
functions of the relaxation loss

(16)

The instantaneous shear modulus is defined as


G0 =

G0

The creep coefficient used to describe the visco-elastic material


behaviour in the FE analysis is shown for one segment in Fig. 5(b).
The creep is calculated based on a notional size for each segment.
The creep behaviour is included in the material definition of
the finite element model. This will lead to each element having
different creep strains depending on when introduced into the
model and their stress level. This will result in different amount
of creep strains over the cross-section of the segments.

gR (t ) =

Creep-related problems can be described in Abaqus with two


different material models, either the material model creep or by
using a visco-elastic description, visco. According to [19], the
material model creep in Abaqus is not suitable for the analysis
of concrete if the stresses vary and especially not if it involves
unloading. Because of this, creep has in this study been included
in the model with the visco-elastic material definition visco
and a quasi-static numerical integration. The visco-elastic material
has been defined assuming a constant bulk modulus, i.e. a timeindependent dilatational response according to the expression:
K0 =

kR (t ) =

GR (t )

(10)

where, 0 is the notional creep coefficient and c (t t0 ) is the


time function describing the development of creep with time. The
notional creep coefficient is estimated as

RH = 1 +

gR (t ) =

GR (t )
G0
KR ( t )
K0

=1

(22)

= 1 .

(23)

The relaxation is included in the material properties for the tendons. The prestress loss may be higher than predicted by textbook
formulas that is why the total loss of prestress is calculated in the
analysis where it is dependent on the creep and shrinkage of the
concrete.
2.2. Segmental construction phase
The segmental construction has been performed in the finite
element analyses where the whole structure was initially modelled
with the geometry of the planned final structure according to the
construction drawings. As a first step, all elements in the cantilever
arms were deactivated, i.e. removed from the calculation. In
subsequent steps, the casting sequence in the cantilevering process
was simulated, the corresponding segments being activated in
the finite element model. The elements are activated with the
Abaqus command with strain so that they are added at a
zero state in a smooth conjunction with previously cast segments.
Each calculation step corresponds to 1 week to simulate the
casting sequence and the development of the material properties.
The elements are introduced in the beginning of each step with
their dead weight and a low value of the elastic modulus. As the
calculation of the step progresses, the concrete cures and as a result
the elastic modulus, creep and shrinkage increase. In the beginning
of the next step, this segment is post-tensioned and a new segment

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

1043

is activated as stress free. In reality, the post-tensioning of the


segment is performed after approximately 5 days of curing, but in
the model it has been introduced instantaneously in the beginning
of the subsequent step. This reduces the number of steps needed
in the analysis. During this second step, the first cast segment
develops its material properties from 7 to 14 days and the newly
cast segment develops from young concrete to the properties that
corresponds to 7 days of curing. This procedure is then continued
throughout the cantilevering process. After the completion of the
structure and the cast centre segment, the tendons in the bottom
flange are post-tensioned.
2.3. Post-tensioning
After each segment is cast, two tendons in the upper flange
are post-tensioned symmetrically above each web. A sketch of the
tensioning scheme is shown in Fig. 2. Each tendon is tensioned with
a force of 2377 kN except for the tendons anchored closest to the
main-span which are tensioned with a force of 2327 kN. Each wire
consists of 12 strands with a diameter of 16 mm.
All tendons have been modelled as separate truss elements. If
prestress is defined, and unless it is held fixed, it will be allowed to
change during an equilibrating static analysis step. This is a result
of the straining of the structure as the self-equilibrating stress state
establishes itself. An example is the pretension type of concrete
prestressing in which reinforcing tendons are initially stretched
to the desired tension before being covered by concrete. After
the concrete cures and bonds to the tendon, release of the initial
tension transfers the load and introduces compressive stresses in
the concrete. The resulting deformation in the concrete reduces
the stress in the tendon. The post-tensioning effect of concrete
prestressing where the tendons are allowed to slide through the
concrete in conduits and where the prestress loading is maintained
by an external source, the prestressing jacks. In order to simulate
this in the FE analysis, the command prestress hold is used
to indicate that the initial stress in the tendons should remain
constant during the initial equilibrium calculation, [15].

Fig. 6. Calculated camber of the cantilevers 8V and 8H at various construction


stages in relation to the weight of the travelling form.

3. Results
To study the influence of the different time-dependent material properties and effects, several different models have been performed. It is of interest to study their impact on the deflections and
the stresses in the cantilevers during the construction process. It is
especially of interest to see if the built-in stresses from tensioning, the bottom tendons could be sufficient to initiate cracking in
the webs. The development of the elastic modulus, creep, weight of
the form traveller, relaxation and shrinkage have all been studied
separately as well as their combined effect.
3.1. Camber
The calculated camber of the two cantilevers on each side of
pier 8 is shown in Fig. 6 for a visco-elastic model, with creep and
shrinkage included, depending on the weight of the traveller. The
weight of the newly cast segment is included in all the calculations.
Hence, the total load shown in the legend of the figure is the load
of the traveller and does not include the weight of the newly cast
concrete. The interval of the load of the traveller shown in the
figure is between 200 and 1000 kN. In the figure, the displacement
measured in-situ is also shown. It can be seen that some of the
segments of the cantilever 8V08V13 cast in-situ deviated from
an ideal curve, and hence they had to be corrected to get back on
track. The index 8V08V13 refers to the cantilever segments 013
cast from pier 8 towards the main-span, i.e. on the left-hand side
of pier 8 in Fig. 2.

Fig. 7. Calculated camber of the cantilevers 8V and 8H at various construction


stages depending on material properties included in the analysis.

The difference in the maximum camber is almost 60% if the


traveller load is 200 instead of 1000 kN, which corresponds to a
change in the displacement of 30 mm. The difference in camber
is less than 10 mm in Fig. 6 for loads of 400 and 600 kN. The
weight of the travellers used for construction of the Grndal bridge
correspond to a load of 480 kN according to the design company,
and this value is used in all the following analyses.
Fig. 7 shows the effect of camber depending on the material
properties included in the analysis. All models include the evolution of the elastic modulus, and creep, shrinkage and relaxation
are also considered. It can be seen in the figure that shrinkage
has a rather small impact on the displacement during cantilevering, regardless of whether non-uniform shrinkage over the crosssection is considered. This was expected since the restraint from
shrinkage has a small impact while the cantilevers have a free end.
Shrinkage will have a larger impact when the centre segment of the
cantilevers is cast. The relaxation of the tendons has also a small
impact on the camber. A comparison of the camber in analyses
with and without creep shows that creep plays an important role

1044

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

Fig. 8. Elevation of the bridge after closing the structure. The zero level corresponds
to the elevation according to the construction drawings.

in the construction process. This agrees with the conclusion of [7]


that creep has a large effect during cantilevering. At the time when
the centre segments are cast, the first cast segments have reached
approximately 60% of their final creep, because a substantial part
of the load originates from the dead weight. The models including
creep receive a higher camber and, for the configuration shown in
the figure, the camber decreases by approximately 30% if the creep
is neglected.
3.2. Construction phase
The tensile stresses are relatively low in the webs during the
casting of the cantilever arm, except for the upper part of the
web where large shear stresses are present due to the prestressing
forces. After casting the centre segment and thereby changing the
system to a statically indeterminate one, restraints are introduced
which lead to high stresses that would initiate bending cracks
in mid-span. After this, the tendons in the bottom flange are
post-tensioned which reduces the stresses in these regions as the
structure moves upward. After completion of the superstructure,
rather high stresses occur in the webs near the location of the
anchorage blocks as the tendons in the bottom flange are stressed.
In this model, the actual dimensions of the anchorage blocks are
not included. Henceforth the stresses in the bottom flange are
higher than would actually occur. It can be seen that the posttensioning of the bottom tendons increases the stresses in the webs
in segments 410, i.e. those near the quarter-point. The principal
tensile stresses in these regions are inclined, and this would initiate
flexural-shear cracks. The principal compressive stress direction
is about 25 in relation to the horizontal plane at the centre of
gravity in the web of these segments after the bottom tendons
have been tensioned. According to [9], observations showed that
the inclination of the cracks in the Grndal bridge was between
2030 near the quarter-point of the span. In the crack pattern of
the Grndal bridge, several of the cracks begin at the transition
between segments and on the anchorage blocks at the bottom
flange. This indicates that the cracks have initiated as a result of
tensioning the bottom tendons.
The calculated displacement of the bridge just after the completion of the superstructure by casting the centre segment is
shown in Fig. 8. The zero level deflection is defined as the elevation

Fig. 9. Elevation of the bridge after stressing the tendons in the bottom flange. The
zero level corresponds to the elevation according to the construction drawings.

according to the design drawings. All models shown in Fig. 5(a)


include the development of the elastic modulus. The model
including creep, shrinkage and relaxation has a mid-span displacement that is about twice as high as that without creep and shrinkage. The models that includes either creep or shrinkage show a
displacement that is intermediate between the models with and
without these two combined effects. The separate studies of creep
and shrinkage yield almost the same deflection at this construction
stage. The curves denoted shrinkage include non-uniform shrinkage between the segments, and the curves denoted non-uniform
shrinkage rate also include different shrinkage rates for the top
flange, web and bottom flange in each segment. It can be seen
in Figs. 89 that there is only a small difference between uniform shrinkage for the whole cross-section and the case where
the shrinkage rate is dependent on the thickness of each part in
the cross-section. After the tendons in the bottom flange are posttensioned, the bridge moves upward, as shown in Fig. 9. The models that include creep will be most affected by the prestressing and
thereby deflect more than the models that neglect creep. The upward displacement will later decrease as the ballast is applied and
due to creep and shrinkage, as shown in Fig. 10. The model that
includes the combined effect of creep and shrinkage has a larger
deflection than would be obtained by superposition of the two separate effects. This shows that a simple superposition of the two
effects will not give an accurate estimation of the deflection. The
combined effect of shrinkage and creep also gives a larger deflection after the tendons in the bottom flange are post-tensioned than
that given by the two effects separately.As seen in the figure, this
can result in quite a large difference since the size of the creep is
dependent on the stress level. Despite this, it is often be neglected
in design calculations.
Non-uniform shrinkage due to a large difference in the shrinkage rates of the members in the cross-section is, according to [4],
the main reason for the increasing displacement of the segmentally cast boxgirder bridges. It can be seen in Fig. 10 that the nonuniform drying shrinkage has a large impact on the deflection after
2 years. Directly after completing the structure, the difference is
rather small but it increases with time until all the members have
reached their final shrinkage. It is important to remember that the
shrinkage still has not ended after 2 years, especially not in the
thick bottom flange which has only reached approximately half its
final shrinkage.

R. Malm, H. Sundquist / Engineering Structures 32 (2010) 10381045

1045

S, Max . Principal (MPa)


Multiple section points
+3.00
(Avg: 75%)

+2.75
+2.50
+2.25
+2.00
+1.75
+1.50
+1.25
+1.00
+0.75
+0.50
+0

Y
Z

Fig. 11. Tensile stresses in the cantilevers after stressing the bottom tendons.

Fig. 10. Elevation of the bridge after 2 years including the load of ballast. The zero
level corresponds to the elevation according to the construction drawings.

4. Discussions and conclusions


The results show that to create a finite element model of
the Grndal bridge that can accurately describe the cracking in
the webs, the time-dependent effects have to be taken into account. Both creep and shrinkage have a large impact during
the construction process and models that neglect these effects
underestimate the cracking. Creep is important already during cantilevering and, in the case of the Grndal bridge, neglecting this
effect resulted in an underestimation of the deflections of the cantilever of approximately 30%. Creep is of even greater importance
for the long-term displacements. After 2 years of service, which
was the age of the Grndal bridge when the cracks were first observed, the displacement is three times higher in the mid-span if
creep is included. Experimental results that shows significant effect of creep, a short time after casting, can also be found in the
literature [20]. In their experiments, two prototypes of two-span
boxgirder bridges were studied. In their study, the measured displacements increased by 50% due to creep after only 27 days.
Due to the time it takes for the drying shrinkage to occur, the
effect of the non-uniform shrinkage rate is significant, especially
after 2 years of service. The difference between an uniform
shrinkage rate and a non-uniform shrinkage rate is initially small
but it increases with time. According to Kristek et al. [4], the
deflection of some cast-in-place bridges has continued to increase
over very long periods, and the deflection curves do not level of
even after 30 years.
In the Grndal bridge it was mainly the lower part of the webs
that cracked. The FE model of the construction process showed
that after post-tensioning the tendons in the bottom flange, large
stresses appear in these regions, see Fig. 11. This indicates that the
post-tensioning of the bottom flange had a large impact on the
cracking of the webs. A good tendon arrangement is crucial, not
only to benefit the stress state, but also to counteract the increasing
long-term displacement and to avoid cracking in the webs.
5. Further research
One large problem with most FE material models for describing
creep is that they cannot be combined with models which describe
cracking. Some sort of approximation has to be made to combine
these two effects in a single analysis. The subject for future research

will be different methods to combine the visco-elastic effect of


creep and the cracking of the webs. One approach might be to
assign different properties to different regions, where the regions
subjected to high tensile stresses are allowed to crack while the
other regions are allowed to creep. One other approach might be
to describe the cracks with a discrete crack model, resulting in that
the cracks are described as displacements between the elements.
References
[1] Chiu C, Chern J, Chang K. Long-term deflection control in cantilever prestressed
concrete bridges I: Control method. J Eng Mech 1996;122(6):48994.
[2] Chiu C, Chern J, Chang K. Long-term deflection control in cantilever prestressed
concrete bridges II: Experimental verification. J Eng Mech 1996;122(6):
495501.
[3] Takcs P. Deformation in concrete cantilever bridges: Observations and
theoretical modelling. Ph.D. thesis. Trondheim : The Norwegian University of
Science and Technology (NTNU); 2002.
[4] Kristek V, Bazant Z, Zich M, Kohoutkova A. Box girder bridge deflectionsWhy
is the initial trend deceptive? Concr Int 2006;28(1):5563.
[5] Bishara A, Papakonstantinou N. Analysis of cast-in-place concrete segmental
bridges. J Struct Eng 1996;116(5):124768.
[6] Kristek V, Vrablik L. Optimisation of tendon layout to avoid excessive
deflections of long-span prestressed concrete bridges. Concr Eng Int 2007;
11(1):304.
[7] Hedjazi S, Rahai A, Sennah K. Evaluation of creep effects on the timedependent deflections and stresses in prestressed concrete bridges. Bridge
Struct 2007;3(2):11932.
[8] Cruz P, Mari A, Roca P. Nonlinear time-dependent analysis of segmentally
constructed structures. J Struct Eng 1998;124(3):27887.
[9] Malm R. Shear cracks in concrete structures subjected to in-plane stresses. Lic.
thesis. Stockholm: Royal Institute of Technology (KTH); 2006.
[10] Tljsten B, Carolin A. Strengthening two large concrete bridges in Sweden for
shear using CFRP laminates. In: Structural Faults and Repair, London. 2003.
[11] Carolin A. Carbon fibre reinforced polymers for strengthening of structural
elements. Ph.D. thesis. Lule: Lule University of Technology; 2003.
[12] Malm R, James G, Sundquist H. Monitoring and evaluation of shear crack
initiation and propagation in webs of concrete boxgirder sections. In:
International conference on bridge engineeringChallenges in the 21st
Century. 2006. p. 14754.
[13] Hewson N. Prestressed concrete bridges: Design and construction. London:
Thomas Telford Ltd; 2003.
[14] Mondorf P, Kuprenas J, Kordahi E. Segmental cantilever bridge construction
case study. J Constr Eng Manag 1997;79(1):7984.
[15] Hibbitt H, Karlsson B, Sorensen E. ABAQUS version 6.7 finite element
programStandard users manual. Pawtucket: Hibbitt, Karlsson and Sorensen
Inc.; 2007.
[16] MC 90, CEB-FIP model code 1990. 6th ed. London: Thomas Telford; 1993.
[17] EC 2. Eurocode 2: Design of concrete structuresPart 1-1: General rules and
rules for buildings. Brussels: EN 1992-1-1, CEN; 2004.
[18] Bazant Z, Baweja S. Creep and shrinkage prediction model for analysis and
design of concrete structuresModel B3. Mater Struct 1995;28(6):35765.
[19] Canovic S, Goncalves J. Modelling of the response of the New Svinesund
BridgeFE Analysis of the arch launching. M.Sc. thesis. Gothenburg: Chalmers
University of Technology; 2005.
[20] Scordelis A, Elfgren L, Larsen P. Time-dependent behavior of concrete box
girder bridges. ACI J 1979;76(9):15977.

S-ar putea să vă placă și