Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Engineering Structures 28 (2006) 1852–1863

www.elsevier.com/locate/engstruct

FEM modelling of unbonded post-tensioned segmental beams with dry


joints
J. Turmo a,∗ , G. Ramos b,1 , A.C. Aparicio b,1
a Civil Engineering School, University of Castilla- La Mancha UCLM, Avda. Camilo José Cela s/n, 13071 - Ciudad Real, Spain
b Department of Construction Engineering, Technical University of Catalonia UPC, C/ Jordi Girona 1-3, 08034 - Barcelona, Spain

Received 19 September 2005; received in revised form 27 February 2006; accepted 14 March 2006
Available online 19 May 2006

Abstract

This work presents a finite element method (FEM) study of the structural behavior of segmental concrete structures with external pre-stressing,
focusing on the response of these structures under combined shear and flexure. Some FEM models have been validated using experimental shear
tests conducted on 7 m long segmental beams. Three types of models with different levels of complexity are presented. The first one attempts
to reproduce the behaviour of the beams modelling the joint with its real geometry. The second model refines the previous one, introducing the
cracking observed in the experimental tests by means of a discrete crack. The third model evaluates the possibility of reproducing the experimental
test results using a flat joint model, simplifying the modelling by not reproducing the interlocking geometry of the keys.
Since this work aims to shed light on the shear flow mechanism in these types of structures, a theoretical study of the shear transference is
carried out, analyzing the stress flow that develops in the web of these types of structures after the joint opens. The arch effect is identified as
the structural response mechanism that carries the actuating vertical loads. Interesting conclusions are also extracted from the study regarding the
reinforcement of these structures.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Segmental bridges; Castellated joints; Shear strength; Dry joints; Unbonded post-tensioning

1. Introduction passive. The first example application constructed was the


Long Key Bridge [1]. A more recent example can be found in
Segmental concrete bridges with external pre-stressing and Bangkok [2].
dry joints are associated with a span-by-span construction For the Serviceability Limit State, these bridges are designed
process that is thought to be the fastest of its type. For the considering that the limit state of decompression must not
construction of each of the spans, the concrete segments are be reached; hence a minimum compression σn = 0.5 MPa
placed one next to the other, suspended from a beam or is maintained in all sections and the joints remain closed.
arranged in a mobile falsework, and are assembled by means When the overload increases up to the Ultimate Limit State,
of external pre-stressing. Generally, it is not necessary to apply
joints open ostensibly (Fig. 1a), and the structure rapidly loses
any epoxy resin between the joint faces of the segments. It
stiffness and reaches a considerable deflection. The fact that
is precisely the subject matter of the present work to study
there is no passive reinforcement crossing the joints causes
externally pre-stressed segmental bridges with resin-free dry
bending moments to be carried by more active reinforcement
joints. Its more remarkable characteristic is the nonexistence
or by higher initial stresses in the steel tendons.
of bond reinforcement crossing the joints, neither active nor
The shear transference across open joints is a more complex
matter. The universally accepted theory for evaluating the shear
∗ Corresponding author. Tel.: +34 926 295300x3293; fax: +34 926 295391.
strength of resin-free dry joints with shear keys considers
E-mail addresses: jose.turmo@uclm.es (J. Turmo),
gonzalo.ramos@upc.edu (G. Ramos), angel.carlos.aparicio@upc.edu the shear stresses to be transferred across the joint through
(A.C. Aparicio). two qualitatively and quantitatively different mechanisms. The
1 Tel.: +34 93 4017356, fax: +34 93 401 1036. first mechanism represents the friction resistance that arises

c 2006 Elsevier Ltd. All rights reserved.


0141-0296/$ - see front matter
doi:10.1016/j.engstruct.2006.03.028
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1853

Three types of models with different levels of complexity


are presented herein. The first one attempts to reproduce
the behaviour of the beams by modelling the joint with its
actual geometry. The second one refines the previous model,
introducing the cracking observed in the experimental tests
by means of a discrete crack. The third model evaluates the
possibility of reproducing the experimental test results using a
flat joint model, simplifying the modelling by not considering
the interlocking geometry of the keys.
The most important feature of the FEM models validated in
this paper is that they have been calibrated with experimental
Fig. 1. Beam tests. Open Joint at ULS. (a) Flexural failure and (b) shear failure. test setups that were specifically designed for the elements to
fail in shear. Other authors have presented two-dimensional
(2D) [7] and three-dimensional (3D) [8] FEM studies of the
shear transfer in this type of bridge. However, their models
either were not calibrated or were calibrated with tests where
a flexure type of failure is induced.
In the opinion of the authors, this is the first time that a
model of these characteristics has been validated and applied
to understand the response of this type of structure under
combined flexure and shear. Furthermore, a model where
the shear cracks within the segments are taken into account
is validated for the first time. From the application of the
models, innovative conclusions have been reached regarding
the reinforcing criteria and the resistant mechanisms of these
structures once the opening of the joint takes place.
Fig. 2. Crack patterns in the vicinity of open joints and hanger
reinforcement [5].
2. Experimental program
when two flat and compressed surfaces attempt to slide against
A brief summary of the experimental tests modelled in this
the other; this resistance is proportional to the actuating
paper is outlined in the present section. More information about
compression, and the corresponding proportionality factor is
these tests and other similar shear tests performed at UPC
the friction coefficient, µ1 . The second mechanism considers
are available in [6,9]. Shear tests used in this paper for the
the support effect of the castellated shear keys of the web
validation of the models involve a total of four segmental beams
(Fig. 1b). These keys permit shear transfer when they are in
of 7.60 m in length, with an I-shaped cross section (Fig. 3a).
contact, behaving like small plain concrete corbels (the small
Two of the beams were cast of normal conventional concrete
dimensions of the keys do not permit us to place conventional
(PC) and the other two of steel fiber reinforced concrete
reinforcement). The shear strength of the keys by surface area is
(SFRC). The beams were provided with interlocked dry joints,
called cohesion, c. If compression stresses exist, then the keys
with three shear keys of 90 mm in height in each joint.
turn out to be small pre-stressed concrete corbels, increasing
These tests involved simply supported beams with a span
the ultimate shear capacity as compression stresses increase.
of 7.20 m consisting of three segments of varying length
Following the terminology of Fouré [3], the corresponding
assembled together with external pre-stressing, to which an
proportionality factor is called internal friction, µ2 . external load Q1 was applied up to failure (Fig. 3b). The
An issue that causes controversy among researchers is the following nomenclature was used in these test series:
reinforcement of the segment in the proximity of the open joint.
Some authors support the theory indicating that it is necessary (1) V1-PC-35: PC beam of f cm = 34 MPa with an axial pre-
to provide reinforcement to hang the shear force in the zone stressing force of 350 kN
next to the open joint [4,5]. Fig. 2 sketches the justification for (2) V1-PC-70: PC beam of f cm = 34 MPa with an axial pre-
this hanger reinforcement graphically. stressing force of 650 kN
This paper presents a study carried out to validate some (3) V1-SFRC-35: SFRC beam of f cm = 34 MPa with an axial
FEM (finite element method) models for the analysis of the pre-stressing force of 290 kN
behaviour of bridges with unbonded external pre-stressing. For (4) V1-SFRC-70. SFRC beam of f cm = 34 MPa with an axial
these validations, the four experimental shear tests on pre- pre-stressing force of 650 kN.
stressed concrete beams presented in [6] have been used. The The objective of V1-PC tests was to verify the behavior
most remarkable result of these experimental tests was that of an open joint when subjected to shear loading, to quantify
beams with shear reinforcement failed at the same load level the influence of the axial pre-stressing force on the capacity
as the beams without shear reinforcement. of the joint and the structure, and to study the efficiency of
1854 J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863

Fig. 3. V1 test: (a) cross section; (b) configuration and setup (dimensions in mm).

the shear reinforcement near the open joint to determine if it M by Eq. (1):
is necessary to include the hanger reinforcement proposed by
∂ M(x)
some authors [4,5]. With V1-SFRC tests, which complement V (x) = . (1)
V1-PC tests, the objective was to study the possibility ∂x
of replacing the conventional shear reinforcement by steel After flexural cracking, the exterior bending moment M in
fibres when using SFRC. With this aim, the conventional each section x of the beam is compensated by a pair of axial
reinforcement placed in V1-PC beams (Φ8 mm stirrups every forces separated by a lever arm z, including a compression and
30 cm) was completely removed in these beams (except for a tension axial force, Nc and Ns , respectively (Eq. (2)):
the reinforcement at anchorage and deviator zones, and a few
longitudinal bars placed to avoid the premature flexural failure M(x) = Ns · z = Nc · z = N · z. (2)
of the longest segment). Then, Eq. (1) can be transformed in the following manner:
The steel fibers (Dramix RC 65/35 BN) content was
60 kg/m3 (volume content of 0.75%). The BBR system was ∂ M(x) ∂(N · z) ∂(N ) ∂(z)
V (x) = = =z· +N· . (3)
used for pre-stressing by means of a small hydraulic tensioning ∂x ∂x ∂x ∂x
jack. V1-75 beams were provided with eight pre-stressing In conventional concrete beams, the lever arm, z, remains
tendons and V1-35 beams with four. The pre-stressing steel approximately constant between two contiguous sections, at
grade was Y 1860 S7, with a nominal tensile strength of 1860 least during the initial loading stages. Then,
MPa, identical to grade 270 ksi proposed by ASTM A 416. The
nominal diameter of the strands was 15.24 mm (0.6 in.). Hot- ∂(z)
=0 (4)
rolled deformed steel bars of B500S quality were utilized as ∂x
passive reinforcement. This steel grade has a minimum yield and
strength of 500 MPa, a minimum tensile strength of 550 MPa,
and a minimum elongation of 12% for a five diameter gauge ∂(N )
V (x) = z · . (5)
length. ∂x
Eq. (5) represents what is traditionally known as the beam
3. Theoretical studies effect and, after cracking takes place, leads to a distribution of
tangential stresses along the length and width of the transversal
The response of conventional concrete beams under
tangential stresses cannot be studied at a sectional level. The section, as in the one shown in Fig. 4a. Hence, it is necessary
formation of a strut-and-tie mechanism for shear transmission to place stirrups to carry the tensile stresses when the web
before failure obligates us to carry out a spatial study of the of the beam cracks (strut-and-tie analogy). If, due to any
shear response. In the same manner, the structural response circumstance, the bond of the longitudinal reinforcement with
of segmental concrete beams with dry joints and external the surrounding concrete is lost, the reinforcement is unable to
pre-stressing confines us to considering just a pure sectional vary its stress from one section of the beam to another, which
analysis to evaluate the shear response. implies that:
As is known, the shear load V actuating in a section x of a ∂(N )
beam is mathematically related to the exterior bending moment =0 (6)
∂x
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1855

Fig. 4. Normal and shear stresses in a conventional concrete beam (a) and in an externally post-tensioned segmental concrete beam.

and
∂(z)
V (x) = N . . (7)
∂x Fig. 5. Model for V1 tests: one single joint and two segments.
Eq. (7) is known as the arch effect, and means that the shear
force is resisted by an inclination of the compressive axial force. at the slabs, 0.10 m at the webs, and 0.35 m at the web–flange
Generally, these two mechanisms superimpose before the beam transition zone. To avoid masking the trajectory of stresses, a
fails under shear. width of 0.10 m was assigned to the elements at anchorages
In a segmental structure with external pre-stressing, where and deviator zones, at the same time increasing the modulus of
there is no passive reinforcement connecting the segments and elasticity of the material to simulate its stiffness.
where the pre-stressing tendons contact the concrete only at The joints were modelled by two-node interface elements, to
anchor blocks and deviators, the axial force remains constant which different constitutive equations were assigned according
in every cross section of the beam. Thus, the transmission of to the model considered. Taking into account that the
the shear force relies on the arch effect. In this manner, an experimental and numerical tests pointed out that the limit
indissoluble association of longitudinal compressive stresses state of decompression was not reached at the joint between
and tangential stresses takes place. The distribution of normal segments D1 and D2 (Fig. 3), the model used for the analysis
and tangential stresses in a section of such structures is was simplified and provided with two segments instead of three
represented in Fig. 4b. and one joint instead of two (Fig. 5). It is worth noting that
carrying out an analysis with so many sources of non-linearity
4. Description of the models implies prioritizing between these sources and choosing the
most significant ones if convergence is to be achieved.
To study the spatial mechanism of the shear response, The beam was positioned over two point-supports that
several second-order effect analyses were carried out with the did not allow vertical displacements. One of the supports
finite element commercial code Diana 8.1. With the same also impeded horizontal displacements. Loads were introduced
aim, several 2D models were prepared with bi-dimensional punctually at the upper slab. The initial pre-stressing was
elements. introduced by applying external loads actuating at anchorages
The small width of the slabs and the data obtained from and deviators. To simulate the evolution of the pre-stressing
the embedded strain gages placed at the upper flange in V1- along the development of the test, the cables were modelled
PC tests permit the ensuring of a uniform behaviour along the by bar-type elements. In this way, each pre-stressing cable
width of the slab [6,9]. Therefore, the beams were modelled was modelled by three bar-type elements, reproducing the
using bi-dimensional elements consisting of four nodes and polygonal path of the pre-stressing force (Fig. 6). The
eight degrees of freedom. These elements had a width of 0.60 m inconvenience of this model is that it simulates the existence
1856 J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863

Fig. 6. Mesh for beam V1, including cable-type elements (in black).

of infinite friction at the deviator zones, which does not permit


the cables to slip. Experimentally, it has been proved that, when
the joints open, the loss of balance between the forces at each
side of the deviators makes the cable slip, and the increments
in deformation are distributed along the entire length of the
cable [6,9]. A modulus of elasticity of E 0p = 119 000 MPa
was assigned to the pre-stressing steel, with the objective of Fig. 7. Interface element.
modelling the free slipping of the cables at the deviators in a
simple way. This fictitious value of the modulus of elasticity
(E 0p ) can be deduced from Eq. (8), relating the cable length
between deviators crossing the joint L 1 , the total length of the
equivalent middle cable L T , and the modulus of elasticity of
the pre-stressing steel E p :
L1 4800 m
E 0p = · Ep = · 190 000 MPa = 119 543 MPa. (8)
LT 7629 m

In this manner, the stiffness of the bar-type element, EA/L,


which takes into account all the deformation increments after
the joint opening, is equal to the stiffness of the entire cable.
This method gives good results, since the increment in axial
force in the element affected by the opening of the joint
adequately approximates the increment of pre-stressing force Fig. 8. Non-cohesive Coulombian friction model.
measured at the deviators, though it has the inconvenience that
the stress increment in the cables is not distributed along its Table 1
Values of the constants used to model the behaviour of the concrete joint
total length. To avoid this problem, once the increase in pre-
stressing, 1Pi , is known at the central part of the cable and kn 50 000 MN/m3
just before failure, 1Pi is distributed along the total length kt 50 000 MN/m3
tan φ 0, 60
of the cable. In order to do so, the structure is loaded with a c 0, 0
self-equilibrated system of forces actuating at anchorages and tan ϕ 0, 0
deviators, with a magnitude equal to 1Pi and a direction equal
to that of each lateral part of the cable, as is described in [10].
This last step avoids masking the effects that the deviation stiffness coefficients (kn and kt ), the friction angle (φ), the
forces at the anchorages have in the stress distribution of the cohesion (c), and the dilatancy angle (ϕ).
structure. Usually, important penalty values were assigned to the
The concrete material was modelled by a linear elastic normal and tangential stiffness coefficients (kn and kt ), which
behaviour. Only the elements of the upper slab were relate the normal and tangential stresses (tn and tt ) of the
modelled by a plastic behaviour in compression, considering joint in the elastic state to the relative normal and tangential
a characteristic compressive strength f ck = 35 MPa. The displacements between the joint faces (1u n and 1u t ). These
stress–strain constitutive law is the one proposed in [11]. penalty values are intended to simulate the initial continuous
geometry before the decompression. Lower stiffness values
4.1. Joint model with the real geometry were proven to simulate the contact of two concrete surfaces
better in the case of panel tests [6,9]. When two concrete
The singular nature of the union between segments is surfaces make contact, the connection is materialized through
modelled by two-node interface elements (Fig. 7). Since there the contact of the different irregularities of their surfaces.
is a frictional type of behaviour between the faces of the joint, For growing compression stresses, these irregularities deform
the constitutive law considered for these interface elements (crush), imposing a relative approximation between the two
when the joint was modelled with its actual geometry was a concrete faces. The value of tan φ is taken to be equal to
Coulombian friction law without cohesion (Fig. 8). the friction coefficient between two concrete surfaces, µ1 ,
The values adopted for the characterization of the joint are proposed by AASHTO [13]. Once the maximum shear stress
presented in Table 1, where the regular notation for this kind is reached, the faces of the joint slide, so the value of the angle
of constitutive law has been used [12]. This table includes the ϕ, or dilatancy angle, is taken to be equal to zero. The cohesion
values adopted in the analysis for the normal and tangential value is equal to zero. Hence, if the joint is not compressed, it
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1857

Table 2
Values of the constants used in the flat joint model

Units Flange Web


kn MN/m3 50 000 50 000
kt MN/m3 50 000 50 000
tan φ 0, 60 1, 07
c MPa 0, 0 4, 54
tan ϕ 0, 0 0, 0

is not capable of transmitting either tensile or shear stresses.


Beyond this stage, when the limit state of decompression is
reached, the displacement between both sides of the joint are
not directly linked anymore.
Fig. 9. Cohesive Coulombian friction model.
4.2. Flat joint model
Table 3
When modelling the keys with their real geometry, the main Values of the constants used to model the behaviour of the discrete cracking
issue is that the meshing must be adequately refined in the kn 107 MN/m3
surrounding area of the joint in order to reproduce the real kt 107 MN/m3
geometry. Then, in a real bridge, with segments 2–3 m in depth f ctm 2.6 MPa
and 40–50 m in span, with joints every 2–4 m with 30–40 keys Gf 0.105 N mm/mm2
per joint, the required numerical capacity can be significant. β 0, 0
The advantage of a flat joint model is that the geometry of the
segments can be reproduced by a mesh with a smaller number counterpart are not directly linked. Practically, this implies
of elements, which implies a significant reduction in degrees of discarding the contribution of the shear keys in contact under
freedom, which at the same time implies important savings in a neutral axis, which is in line with the present standards. More
computation time. Furthermore, rectangular elements and the information on the joint model can be found in [6,9].
regularity of the mesh favour numerical convergence.
The flat joint model is composed of two-node interface 4.3. Cracking model
elements, with different constitutive laws depending on
the geometry that the elements were reproducing. These Since the cracking observed in the experimental tests [6,
constitutive laws were derived from the formulation proposed 9] corresponded to the development of one single crack that
by AASHTO [13] to evaluate the shear capacity of dry joints; accumulates all the deformations (Fig. 1b), a discrete crack
a formulation that, in [14], has proven to be the more accurate approach was selected for the model. The chosen model is
than others [3,5,15]. materialized by two-node interface elements, with a constitutive
A Coulombian-type friction law with no cohesion was equation based on the theory of total deformations, which
assigned to the interface elements modelling the joints at relates the stress vector to the total relative displacements
the upper and bottom flanges (Fig. 8). On the other hand, a between the crack faces. In the elastic state, as after cracking,
Coulombian-type friction law with cohesion was considered for the normal and tangential responses of the joint are decoupled.
the interface elements modelling joints at the webs (Fig. 9). The constitutive equation of the discrete crack is defined
These friction laws are defined numerically in Table 2. The numerically in Table 3, where the regular notation for this kind
AASHTO formulation differentiates the strength provided by of constitutive law is used [12].
the area of the key, Ak , from the strength of the smooth area In the model, the terms kn and kt of the elastic stiffness
between the keys, Asm . To have a simple and practical model, matrix had a sufficiently high value to model the continuous
it seems useful to combine these formulae and assign a mean geometry before displacements but, at the same time, small
constitutive law to the entire web. In this manner, cohesion c0 enough to avoid numerical problems during the analysis.
and friction µ02 , which define the mean behaviour of the entire The tensile strength f ctm has been calculated from the mean
web, are determined from Eqs. (9) and (10): concrete compressive strength ( f cm = 34 MPa) according to
EC2 [16]. The value of fracture energy G f has been obtained
c · Ak
c0 = (9) considering the mean concrete compressive strength and the
Ak + Asm maximum size of the aggregate (dmax = 25 mm), following the
µ2 · Ak + µ1 · Asm recommendations of the Model Code [11]. The normal stress
µ02 = . (10)
Ak + Asm post-peak behaviour is governed by a linear tension softening
function, though the important crack opening observed in the
The constitutive law governs the behaviour of the interface experimental tests shows that the post-peak contribution of the
elements only when compression exists. If this does not material is practically irrelevant. Due to this fact, common
occur, the displacements of a node and of its corresponding functions have been chosen for conventional and steel fibre
1858 J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863

Fig. 10. Sensibility to geometrically nonlinear analysis. Fig. 11. Load–deflection curves from V1-35 experimental tests (V1-PC-35 and
V1-SFRC-35) and numerical analysis (V1-35-FEM-B and V1-35-FEM-A).
reinforced concrete. After cracking, the shear stiffness of the
hardly any influence in terms of deformation, but notably
crack is reduced to zero (β = 0). The large crack openings
smoothens the stress field in the immediate vicinity of the joint.
observed in the experimental tests allow us to suppose that the
The values of the axial stiffness and tangential stiffness of
tangential stresses, generated due to the relative displacement
the interface elements, kn and kt , were fundamental to model
between the crack faces, are equal to zero (Fig. 17).
correctly the load–deformation and load–stress behaviour of
shear tests on panels [6,9]. Changing the values of kn from
5. Results and discussion a penalty value of 10 000 000 MN/m3 to lower values of
50 000 MN/m3 increased deformations at failure by 100%
5.1. Parametric analysis (Section 4.1). However, values adopted for kn and kt hardly
influence the deformation of the beam, which is fundamentally
Analyzing the influence of the different parameters, it can governed by the crack opening and not by the support between
be observed that the model is extremely geometrically non- keys or the slipping of the joint. When concrete plasticity is
linear. The inclusion of a geometrical non-linearity increases introduced in the slab, there is no influence of kn and kt .
the vertical displacements of the point-load of the model by The tensile cracking of the bottom slab does not affect the
around 50% for the test maximum load. Fig. 10 shows the modelling of the tests either. Though the tensile strength is
difference between performing the analysis using the second- reached in part of the bottom flange, the zone is quite limited;
order theory (V1-PC-35-NLG) or not (V1-PC-35). hence, it appears that the modelling of the crack does not
The structural behaviour is extremely sensitive to the value affect the global response of the structure. This negligible
of initial pre-stressing. It affects both strength and flexibility. effect would not influence actual structures made of multiple
As it was found that this was the main parameter affecting not segments at all, as the joint opening impedes flexural tensile
only the behaviour of the model but also the actual structures, stresses arising [10].
a complete sensibility analysis was performed in [10]. The
influence of the pre-stessing level can be summarized in five 5.2. Numerical analysis of beams V1 modelling the real
main points: (a) the joint opening load is directly influenced geometry of the joint
by the level of pre-stressing—the lower the initial pre-stressing, In this series of numerical tests, the behaviour of beams
the lower the joint opening load; (b) the stiffness of the structure V1 has been studied, focusing attention on the non-linearity
is not affected by the pre-stressing level while joints are closed; produced by the presence of dry joints and its effect on the
(c) the stiffness of the structure is strongly influenced by the beam response. Keys and joints were modelled using their
pre-stressing force once the joints open; the higher the pre- real geometry, provided with two-node interface elements. The
stressing force, the stiffer the structure; this is to say, for mesh used to carry out the analysis can be seen in Fig. 6.
the same external load actuating in the structure, the higher Fig. 11 shows the experimental load–displacement response
the pre-stressing level, the smaller the deflection; (d) the for beam V1-PC-35 and beam V1-SFRC-35. For comparison,
maximum deflection at failure is also influenced by the pre- the same figure shows the curves obtained from the numerical
stressing force—the higher the pre-stressing force, the bigger analysis; V1-35-FEM-A with an initial pre-stressing force P0 =
the deflection; and (e) the increase in the initial pre-stressing 0.350 MN, and V1-35-FEM-B with an initial pre-stressing
force increases the strength of the structure. force P0 = 0.290 MN. The load-joint opening responses for
The combination of interface elements with the linear elastic V1-35 tests can be seen in Fig. 12.
material of the slabs adequately approximates the experimental It is worth noting how the numerical curve V1-35-FEM-A
test results in terms of deformations. However, it generates very reasonably approaches the experimental response of V1-SFRC-
high compression stresses around the joint that have no physical 35 and how the behaviour of V1-PC-35 is reproduced by V1-
sense. The inclusion of a plastic material in compression has 35-FEM-B.
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1859

For V1-70, numerical results do not approximate experimen-


tal tests as sharply as in the case of V1-35. The cracking of the
segment, which goes across the entire height of the beam, sig-
nificantly modifies the deformation behaviour of the structure.
This effect can be observed clearly from Fig. 13, which presents
the experimental and numerical (FEM) deformations of struc-
ture V1-PC-70 for different load levels (0.140 MN, 0.176 MN,
and 0.199 MN). Though, in the elastic phase, the model per-
fectly reproduces the deformation of the structure, the opening
of the joint and the diagonal cracking of the segment provoke
a peak in the deformation which the model is not capable of
reproducing. To model the behaviour of the beam adequately, it
is necessary to give an appropriate treatment to the cracking in
Fig. 12. Load-joint opening curves from V1-35 experimental tests (V1-PC-35
the web of the segment.
and V1-SFRC-35) and numerical analysis (V1-35-FEM-B and V1-35-FEM-A). A selection of stress fields obtained from the analysis of
V1-35-FEM-B is presented in Fig. 14 for the hypothesis of
maximum load. Fig. 14 includes the vector field of principal
compressive stresses generated in the structure for an absolute
value higher than 2.5 MPa and lower than 35 MPa (Fig. 14a),
and for an absolute value higher than 4.5 MPa (Fig. 14b).
The same figure also presents a detail of the joint area for a
decompression stress higher than 2.5 MPa (Fig. 14c) and tensile
stresses higher than 2.5 MPa (Fig. 14d).
From Fig. 14a, it can be observed how the resistant
mechanism can be assimilated to two superposed arches: one
arch arises from an anchorage (A), goes across the open joint
(C), through the load application point (D), and finally ends at
the other anchorage (F); the other arch starts developing at the
deviator (B 0 ), goes across the open joint (C), through the load
application point, and finally ends at the second deviator (E 0 ).
Fig. 13. Deformation scheme of beam V1-PC-70 for different load levels Fig. 14b shows how higher principal compressive stresses
(0.140 MN, 0.176 MN, and 0.199 MN); experimental tests and numerical are generated in the zone where both arches superpose. In
analysis (FEM). this example, the contribution of the arches to the transference
of load is very similar. The first arch (between anchorages)
The initial pre-stressing force significantly influences the transmits 51% of the load, while the second arch (between
decompression load, the evolution of the vertical displacement deviators) transmits the remaining 49%. Calculation of the
and the crack opening; as the initial pre-stressing force contribution of each arch to the load transference has been
decreases, the decompression load decreases and the deflection carried out, verifying the amount of vertical load that each arch
and final joint opening increase. The model satisfactorily compensates, as a function of the vertical reactions of each arch.
reproduces the deflection of the beam up to a load Q = The vertical reactions are provided by the supports and forces
0.11 MN. At this stage, the crack development in the segment, at deviators in the first and second arches, respectively.
which is not considered in the model, makes the experimental From Fig. 14c, it can be observed how the principal
and numerical curves diverge from each other. However, this compressive stresses incline due to the effect of the tangential
deviation is not large since, as mentioned, cracking only affects stresses in the joint area at the compression zone.
a limited part of the segment [6,9]. Fig. 14d shows a detail of the field of tensile stresses, in the
Also, it is worth remarking that the increments in the pre- perpendicular direction with respect to the cracking observed in
stressing force, measured by the load cells when the maximum the experimental tests (Fig. 1b). As can be seen, tensile stresses
load is reached, are also well reproduced by the numerical only appear in the immediate vicinity of the open joint and at
tests. The values of these axial force increments obtained in the the anchorage zone, due to the deviation of the isostatics when
numerical tests are just a 7% higher than those measured in the the pre-stressing force is introduced. The absence of tensile
experimental tests. stresses in the web of the beam strengthens the idea of an arch-
type resistant mechanism.
The numerical study of V1-PC-70 and V1-SFRC-70 tests
was carried out in the same manner as for the V1-35 tests 5.3. Numerical analysis of the beams modelling the cracking in
analyzed above. The same bi-dimensional model was used, the web of the segment
varying the initial pre-stressing force. The initial pre-stressing
force P0 = 0.650 MN, introduced to perform the simulation, The experimental tests on series V1-70, where a main
was equal for both tests. diagonal crack across the segment concentrating large rotations
1860 J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863

Fig. 14. Stress field under maximum load from test V1-35-FEM-B. (a) Principal compressive stresses |σII | > 2.5 MPa. (b) Principal compressive stresses
|σII | > 4.5 MPa. (c) Detail of the joint area; principal compressive stresses |σII | > 2.5 MPa. (d) Detail of the joint area; principal tensile stresses |σI | > 2.5 MPa.

Fig. 15. Detail of the model used for the simulation of test V1-PC-70 (a) and test V1-SFRC-70 (b). (Dry joint and discrete cracking in black.)

in its plane was observed [6,9], cannot be modelled without each potential crack (Fig. 1b), and were placed horizontally and
considering the effect of cracking. Therefore, the mesh used for vertically, bridging the cracks located where the longitudinal
the analysis of beams V1 was modified, introducing interface and transverse reinforcement was placed in the real segment.
elements that reproduced the geometry of the crack observed The area of each of these elements corresponds to the area of
in tests V1-PC-70 and V1-SFRC-70. The characteristics of the the real reinforcement, providing them with an elastic–perfectly
mesh can be observed in Fig. 15a and b, showing the dry joint plastic behaviour, with a yielding stress of 500 MPa. It is worth
and potential cracks. remarking that a model with such an amount of non-linearity
As can be seen from Fig. 15, not only new interface elements sources is numerically quite unstable.
had to be included; three-node triangular elements with six
degrees of freedom were also incorporated in that zone. In For V1-PC-70-FEM, the model was not able to reproduce
this manner, the complicated geometry of the segment in the the behaviour up to failure of test V1-PC-70, since, for a
proximity of the joint is well covered with a dense mesh of load level Q 1 of approximately 80% of the ultimate load, the
elements. calculation program diverged or was not capable of inverting
In the model simulating the behaviour of beam V1-PC-70, it the stiffness matrix. Though the number and size of load
has been necessary to include elements capable of reproducing increments was varied repeatedly to reach the ultimate load,
the reinforcement of the segment. Thus, two-node linear the problem could not be avoided; the analysis was interrupted
elements with four degrees of freedom were incorporated. shortly after achieving cracking of the segment, even though the
These elements connected the nodes of the mesh at both sides of concrete of the upper slab had a principal compressive stress of
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1861

and b, respectively, and compared with the results of the


numerical simulation; V1-PC-70-FEM (Fig. 17c) and V1-
SFRC-70-FEM (Fig. 17d). The same figure shows clearly how
the opening of the joint progresses along the web of the
segment, and how the keys situated above the initiation point
of the crack remain in contact.
The principal stress fields obtained from the analysis of
V1-SFRC-70-FEM are presented in Fig. 18. Fig. 18a shows
the vector field of principal compressive stresses higher than
2.5 MPa, from which it can be seen how practically the
entire beam is under compression, except in the immediate
vicinity of the load, where the isostatics are located in the
compression flange and acquire the shape of an arch. The
Fig. 16. Deformation scheme of beam V1-SFRC-70 for different load levels;
notorious inclination of the principal compressive stresses in
0.165 MN, 0.185 MN, and 0.198 MN, from experimental tests and numerical the compressed zone of the webs shows how, in this area, the
analysis (FEM). webs are responsible for the transference of tangential stresses.
On the other hand, once the crack opens, the tensile stresses
23.9 MPa and the stress of the stirrup next to the joint was just shown in Fig. 14d are freed. This extreme can be observed from
317 MPa. Fig. 18b, which shows the principal tensile stresses higher than
In test V1-SFRC-70-FEM, the ultimate load of the 2.5 MPa. As can be seen, tensile stresses are practically absent
experimental test was finally reached by applying successive outside the tension fields generated to regulate the introduction
variations to the size of the load increments, which were of the pre-stressing force at anchorages and deviators.
arbitrarily chosen to avoid divergence, since reducing the
value of these increments did not necessarily imply achieving 5.4. Numerical analysis of the beams using a flat joint model
convergence.
Fig. 16 presents the deformed configuration of the structure A detail of the mesh utilized in this analysis can be observed
for different load levels (0.165 MN, 0.185 MN, and 0.198 MN), in Fig. 19, with a total of 1327 elements and 1422 nodes, in
from experimental tests and numerical modelling (FEM). The which a flat joint model has been introduced.
peak that appears at the immediate vicinity of the joint in the Fig. 20 shows the load-deflection responses obtained with
deformation scheme, which the model presented in 5.2 was the different models, in which the joint between segments
unable to reproduce (Fig. 13), is know satisfactorily modelled. is modelled with its real geometry (V1-35-FEM-B) or with
A detail of the deformed geometry obtained under maximum a flat joint (V1-35-FEM-B-P). The analysis of these figures
load in tests V1-PC-70 and V1-SFRC-70 is shown in Fig. 17a allows us to appreciate the adequate behaviour of the flat

Fig. 17. Detail of the deformation scheme under maximum load. Experimental tests V1-PC-70 (a) and V1-SFRC-70 (b). Numerical modelling V1-PC-70-FEM (c)
and V1-SFRC-70-FEM (d).
1862 J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863

Fig. 18. Field of principal stresses from test V1-SFRC-70-FEM-B under maximum load; compressive stresses |σII | > 2.5 MPa (a) and tensile stresses
|σI | > 2.5 MPa (b).

Fig. 19. Detail of the finite element mesh for the analysis of V1-FEM-P (flat Fig. 20. Load–deflection curves from V1-35. Comparison of the results from
joint in black). different models.

joint model; slight differences are observed only in the joint cracking development. It is also necessary to include the non-
opening phase, however they do not blur the overall good linear compressive behaviour of the concrete in the model,
results. Very good agreement can also be observed between the and carry out the analysis considering the second-order effects
stress fields of both models, apart from some little differences theory.
in the vicinity of the joint mainly due to the different nature A flat joint model is proposed that permits the analysis of
of the contact between the two models of joints. This does complete bridges without the need for modelling the joints with
not affect the overall behaviour of either model. Basically, the its real geometry, which implies much simpler models, with a
fact that the two models give such similar results implies that consequent reduction in computer time.
the deformation and stress behaviour of the structure is, to a The theoretical studies and the analysis carried out confirm
great extent, governed by the opening of the joint instead of by that the resisting mechanism of these types of beam is spatial,
the support system between keys. The possibility of modelling not sectional. In the presence of external vertical loads, the
the geometry of the keys by means of a flat joint allows us to structure responds with an arch-type resistant mechanism. Two
simplify the models used to study real bridges significantly. arches arise to carry the loads: one arch materializes between
the supports of the beam and another between the deviators,
6. Conclusions which actuate like intermediate supports.
A strut model can be drawn in the concrete beam, capable of
A numerical model for the analysis of segmental concrete resisting the actuating loads at the Ultimate Limit State without
beams with dry joint and external pre-stressing has been the need for tension ties, which are materialized by the pre-
proposed. The model has satisfactorily reproduced the results stressing cables. In other words, a resistant mechanism that
obtained in large-scale experimental tests. allows the formation of a flexure- and shear-resistant scheme
The opening of the joint can be simulated by using interface is generated in the beam, without the need for conventional
elements, to which a Coulombian-type friction behaviour is reinforcement. This reinforcement can be useful to control
assigned. If there is an important cracking in the web of the cracking in service, but its efficiency in resisting shear is quite
segment, it becomes necessary to model the web accordingly, limited, especially in the immediate vicinity of the open joint.
since the stiffness of the structure is greatly influenced by the Though tensile stresses truly appear in this zone (Fig. 14d), the
J. Turmo et al. / Engineering Structures 28 (2006) 1852–1863 1863

effect is local and does not contribute to the global resisting AFPC external prestressing in strucutres. Saint-Rémy-lès-Chevreuse.
mechanism of the structure. In fact, when the tensile strength is June 1993. p. 449–66.
[5] ATEP design and construction of bridges and structures with external
reached and the beam cracks, the tensile stresses are freed and
prestressing. Madrid. September 1996 [in Spanish].
the arch continues to support the load (Fig. 18b). The stirrups [6] Turmo J. Flexure and shear behavior of segmental concrete bridges
have to be placed to control cracking, but its presence does not with external prestressing and dry joints. Ph.D thesis. Directed by:
imply substantial increments in shear capacity. Prof. Dr. Aparicio AC and Prof. Dr. Ramos G. ETSICCP de
Barcelona. Dept. Ing. de la Construcción. July 2003 [in Spanish].
Acknowledgments http://www.tdx.cesca.es/TDX-1030103-090157/.
[7] Huang J, Eibl J. Design of segmental bridges under combined bending,
shear and torsion FE-study. In: Proceedings of the workshop AFPC
Partial funding of the research by the Spanish Ministry of external prestressing in strucutres. Saint-Rémy-lès-Chevreuse. June 1993.
Science and Technology (MAT2002-00849 Project), and by the p. 335–47.
Spanish Ministry of Public Works (Project: “Theoretical and [8] Rombach G. Precast segmental box girder bridges with external
Experimental Study of the Shear Transference in Segmental prestressing-design and construction. Rennes: INSA; February 2002.
p. 1–15.
Steel Fiber Reinforced Concrete Beams, with External
[9] Turmo J. Study of the structural behaviour under flexure and shear of
Prestressing and Dry Joints”), is greatly appreciated. One of segmental concrete bridges with external prestressing and dry joints.
the authors benefited from a scholarship from the Spanish Madrid: ACHE; 2006.
Ministry of Education and Culture from the year 2000 until [10] Turmo J, Ramos G, Aparicio AC. FEM study on the structural behaviour
2003 (Scholarship for the Formation of University Professors). of segmental concrete bridges with unbonded prestressing and dry joints:
simply supported bridges. Engineering Structures 2005;27(11):1652–61.
The valuable comments of the anonymous reviewers of the
doi:10.1016/j.engstruct.2005.04.011.
paper are also acknowledged. [11] Model Code CEB-FIP 1990. Colegio de Ingenieros de Caminos. Madrid;
1995.
References [12] DIANA 8.1.User’s Manual. TNO. Delft; 2002.
[13] A.A.S.H.T.O. Guide specifications for design and construction of
[1] Muller J. Construction of the long key bridge. Journal of the Prestressed segmental concrete bridges. Washington; 1998.
Concrete Institute 1980;25(6):97–111. [14] Turmo J, Ramos G, Aparicio AC. Shear strength of dry joints
[2] Brockmann C, Rogenhofer H. Bang Na expressway, Bangkok, Thailand— of concrete panels with and without steel fibres. Application to
world’s longest bridge and largest precasting operation. Journal of the precast segmental bridges. Engineering Structures 2006;28(1):23–33.
Prestressed Concrete Institute 2000;45(1):26–38. doi:10.1016/j.engstruct.2005.07.001.
[3] Foure B. Shear test on keyed joints between precast segments. In: [15] Bakhoum MM. Shear behaviour and design of joints in precast
Proceedings of the workshop AFPC external prestressing in strucutres. concrete segmental bridges. Ph.D dissertation. Massachussetts Institute of
Saint-Rémy-lès-Chevreuse. June 1993. p. 297–319. Technology; 1991.
[4] Virlogeux M. et al. Some elements for a codification of external [16] UNE-ENV 1992-1-1:1992 Eurocode 2: Design of concrete structures.
prestressing and of precast segments. In: Proceedings of the workshop Part 1-1. General rules and rules for building. AENOR; 1992.

You might also like