Вы находитесь на странице: 1из 11

Available online at www.sciencedirect.

com

Construction
and Building

MATERIALS

Construction and Building Materials 22 (2008) 200210

www.elsevier.com/locate/conbuildmat

Inuence of the geometry and the abutments movement on the


collapse of stone arch bridges
G.A. Drosopoulos a, G.E. Stavroulakis
b

b,c,*

, C.V. Massalas

a
Department of Material Science and Technology, University of Ioannina, GR-45100 Ioannina, Greece
Department of Production Engineering and Management, Technical University of Crete, GR-73132, Chania, Greece
c
Department of Civil Engineering, Technical University of Braunschweig, Braunschweig, Germany

Received 10 May 2006; received in revised form 8 September 2006; accepted 22 September 2006
Available online 7 November 2006

Abstract
The ultimate failure load of stone arch bridges is calculated in this paper. The nite element model consists of contact interfaces which
simulate potential cracks. A parametric investigation demonstrates the inuence of the geometry on the mechanical behavior. A reduction of the rise of the arch (below the initial, real geometry) generally causes an increase of the limit load, until a shallow, at arch. The
results are in agreement with the ndings of Heyman. Further reduction of the rise leads to a reduction of the limit load. Furthermore,
deep arches fail following the four hinges collapse mechanism, while compressive failure arises in shallow arches. Finally, for settlement
of the supports, in accordance with Heymans statements, three hinges are developed before collapse of the arch.
2006 Elsevier Ltd. All rights reserved.
Keywords: Unilateral contact-friction; Stone arch bridges; Arch geometry; Abutments movement

1. Introduction
Two dierent issues concerning the mechanical behavior
of stone arch bridges are considered in this study. First, a
parametric investigation of the impact of the geometry of
the stone arch to the ultimate load of the structure, is conducted. Then, the behavior of the arch in which a movement of the abutments takes place, is examined. In all of
these cases a unilateral contact-friction model is used, in
order to depict the non-linear behavior of the structure.
A stone arch bridge consists of stone blocks and mortar
joints. Blocks have high strength in compression and low
strength in tension while mortar has generally low strength.
Thus, a safe assumption of a no tension material can be
adopted at least for the purpose of limit analysis. To simu*

Corresponding author. Address: Department of Production Engineering and Management, Technical University of Crete, GR-73132, Chania,
Greece. Tel.: +30 28210 37418; fax: +30 28210 69410.
E-mail addresses: me01122@cc.uoi.gr (G.A. Drosopoulos), gestavr@
dpem.tuc.gr (G.E. Stavroulakis), cmasalas@cc.uoi.gr (C.V. Massalas).
0950-0618/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2006.09.001

late this behavior and the ultimate load of the arch, a discretized model has been developed. In particular, the
bridge is divided with a number of interfaces perpendicular
to the center line of the arch, uniformly distributed along
the length of the arch. Opening or sliding of the interfaces
denotes crack initiation.
Unilateral contact law governs the behavior in the normal direction of an interface, indicating that no tension
forces can be transmitted and a gap may appear if the compressive stresses in this direction become zero. For the
behavior in the tangential direction a Coulomb friction
model including stick-slip eects is taken into account.
The either-or decisions incorporated in the unilateral contact and friction mechanisms make the whole mechanical
model highly nonlinear. Due to the presence of non-dierentiable functions within these models, they are characterized as non-smooth mechanics models. For practical
applications it is important to use carefully tuned pathfollowing iterative techniques for the reliable numerical
solution of the problem. Furthermore, the limit analysis
problem is related to the solvability of the underlying

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

mechanical problem using analogous theoretical results


concerning the solvability of variational inequalities and
complementarity problems [1].
One of the most signicant researchers of stone arch
bridges in the 20th century is Heyman [2]. Based on earlier
studies, he established the four hinges collapse mechanism
as the tool for the determination of the limit load of these
structures. In this framework, he developed a limit analysis
method which is briey described later in the present work.
The limit analysis of discrete structures with a frictional
contact interface law oers an interesting aspect in the
study of masonry bridges. Drucker [3] rst underlined the
problem of applying the bound theorems of plasticity to
frictional problems. Moreover, Livesley [4] demonstrated
that the adoption of a simplied associated constitutive
law for frictional materials may yield an overestimation
of the true collapse load. Melbourne and Gilbert [57]
studied the limit analysis of stone arches as a linear programming formulation in which the upper bound theorem
was included. Ferris and Tin-Loi [8] computed the collapse
load of discrete rigid block systems with frictional contact
interfaces as a special constrained optimization problem.
Orduna and Lourenco [911] developed two and three
dimensional models of discrete structures (like stone
arches) and they took into account torsion failure mode.
They also included in their study reinforcement elements.
From another point of view, Ng and Faireld [12] proposed a modication of the collapse mechanism method
by considering the interaction between the arch deections
and the backll pressures.
The present work is focused on the parametric investigation of the eects of the shape and the settlement
of the abutments on the collapse load and mechanisms
of stone arches. In addition, since for low-rise arches
compressive failure is critical for the failure of the structure, this issue has been added in the previously developed, unilateral contact model with friction. The
applicability of the unilateral contact model for the limit
analysis and the failure model predictions was studied in
[13]. The results were comparable with the predictions of
the classical Heymans method and with available experimental measurements.
The eect of the geometrical shape of the stone arch on
its ultimate strength and especially the determination of
the optimal shape is a non trivial optimization problem
with respect to the limit load involving unilaterally constrained structures. The solution of this general problem
could be obtained by the methods of the structural shape
optimization for non-smooth problems, which is a nonclassical task (see, among others, [14]). In this paper we perform a parametric investigation, using suitable nonlinear
nite element models, in order to have some rst results
useful for engineers. The structural optimization approach
is under investigation and the results will be reported in a
future work. For the solution of the ultimate load of each
structure, the proposed unilateral contact-friction model is
used. Hierarchically, the proposed method is divided into

201

three levels: (i) the unilateral contact-friction model is


developed; (ii) the failure analysis is performed for the
investigation of the ultimate load of each structure; in this
framework, the maximum load for which equilibrium can
be maintained is the limit load; (iii) the optimal shape is
estimated from the results of a parametric investigation
with respect to the rise of the arch using a total of 19
dierent models (bridges of dierent shapes, mainly rise).
2. The unilateral contact-friction model
2.1. Introduction to the frictional-contact mechanics
The behavior in the normal direction of an interface is
described by the unilateral contact model. In particular,
let us consider the boundary of an elastic body which
comes in contact with a rigid wall. Let u be the single
degree of freedom of the system, which describes the movement of the elastic body towards the rigid wall, g be the initial opening between the body and the wall and tn be the
corresponding contact pressure in case contact occurs.
The basic unilateral contact law is described by the set of
inequalities (1), (2) and by the complementarity relation
(3) [1517],
hug 60)h60
 tn P 0

1
2

tn u  g 0

Inequality (1) represents the non-penetration relation, relation (2) implements the requirement that only compressive
stresses (contact pressures) are allowed and Eq. (3) is the
complementarity relation according to which either separation with zero contact stress occurs or contact is realized
with possibly non-zero contact stress.
The behavior in the tangential direction is dened by a
static version of the Coulomb friction model. Two contacting surfaces start sliding when the shear stress at the interface reaches a maximum critical value equal to
tt scr ljtn j

where tt, tn are the shear stress and the contact pressure at a
given point of the contact surfaces respectively and l is the
friction coecient. There are two possible directions of
sliding along an interface, therefore tt can be positive or
negative depending on that direction. Furthermore, there
is no sliding if jttj < ljtnj (stick conditions). The stick-slip
relations of the frictional mechanism can be mathematically described with two sets of inequalities and complementarity relations, similar to (1)(3), by using
appropriate slack variables [15].
2.2. Formulation and solution of the unilateral
contact-friction problem
For the frictional-contact problem the Virtual Work
equation is rst written in a general form

202

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

q : de dV

Z
Z
du  t dS du  f dV
du  tn dS 0
S
V
S0
Z

du  tt dS 0
5
S0

where tn and tt are the normal and tangential traction vectors on the actual contact boundary S 0 , q is the stress tensor, de is the virtual strain tensor, du is the virtual
displacement vector and t, f are the surface and body force
vectors, respectively.
The nonlinearity in the unilateral contact problem is
introduced by the variational inequality [16]
du  tn 6 0

which represents the Principle of Virtual Work in a variational form. The contact constraint is enforced with Lagrange multipliers representing the contact pressures.
The Virtual Work equation can be properly rewritten by
taking into account the discontinuity imposed by the contact constraint. In particular, relation (6) leads to the following inequality
Z
du  tn dS 0 6 0
7
S0

By substituting inequality (7) in Eq. (5), the following variational inequality problem is obtained
Z
Z
Z
Z
q : de dV  du  t dS  du  f dV 
du  tt dS 0 6 0
V

S0

8
The stick-slip nonlinearity of the frictional problem is
introduced by the variational inequality
dut  tt 6 maxdut  ttcr ;

dut  ttcr

where ttcr is the vector of the critical shear stresses scr in the
tangential direction of the interfaces. As it has been previously outlined, for a two-dimensional problem there are
two possible directions of sliding along an interface. The
critical shear stresses for sliding in the one direction are represented by the vector ttcr , while the ones for sliding in the
other direction by the vector ttcr . Relation (9) implies that
no slip occurs when jttj < scr = ljtnj while slip starts when
tt = scr. Lagrange multipliers are also used in the Principle
of the Virtual Work to enforce sticking conditions. The
set of the nonlinear relations is nally solved by a suitable
NewtonRaphson incremental iterative procedure.
Let us clarify further the relation with classical linear
elastostatics by considering a frictionless unilateral contact
problem, written in matrix form as follows:
T

Ku N r P0 kP
Nu  g 6 0

10
11

rP0

12
T

Nu  g r 0

13

Eq. (10) expresses the equilibrium equations of the unilateral contact problem, where for simplicity frictional terms

are omitted. K is the stiness matrix and u is the displacement vector. Po denotes the self-weight of the structure and
P represents the concentrated live load. N is an appropriate
geometric transformation matrix and vector g contains the
initial gaps for the description of the unilateral contact
joints. Relations (11)(13) represent the constraints of the
unilateral contact problem for the whole discretized structure and are based on the local description given by relations (1)(3). The enforcement of the constraints is
achieved by using Lagrange multipliers. Thus, r is the vector of Lagrange multipliers corresponding to the inequality
constraints and is equal to the corresponding contact pressure (tn). It is noted the usual assumption of mechanics
that tensile stresses are considered to be positive (e.g. compressive stresses in one interface are considered to be negative: tn 6 0 (or tn P 0)) and the opposite assumption
of optimization communities that active Lagrange multipliers are positive.
The problem described above is a non-smooth parametric linear complementarity problem (LCP) [18] parametrized by the one-dimensional load parameter k. All
required quantities can be calculated by using nite element
techniques. Using path-following the solution of the problem can be calculated in the interval 0 6 k 6 kfailure, where
kfailure is the value of the loading factor for which the unilateral contact problem does not have a solution. This is
the limit analysis load.
A schematised algorithm can be proposed to summarize the used model. The loading of the arch is divided
into two steps: the self-weight is initially enforced while
live load (in our case concentrated, but in general of
arbitrary form) is then applied. For a well-designed arch,
when the loading includes only the self-weight the contact-friction model does not lead to failure, as the selfweight increases the stability of the structure by inducing
compression on the whole length of the arch. At the end
of this step the live load is applied incrementally, up to
collapse. During this procedure the solvability of the
problem is checked. In particular, in a structural problem
where unilateral contact-friction constraints exist, parts
of the structure between the frictional contact interfaces
may lose contact and develop rigid body displacements.
In this case a solution exists, if the boundary conditions
imposed by the interfaces and the self-weight of the
structure can suciently equilibrate the live load. When
this equilibrium cannot be achieved, the structure
collapses.
2.3. Model for the compressive failure of the arch
For the compressive failure of the arch the Drucker
Prager plasticity model, in which a cap yield surface has
been added [19,20] is used (Fig. 1). The addition of the
cap yield surface to the DruckerPrager shear failure surface bounds the yield surface in hydrostatic compression,
thus providing an inelastic hardening mechanism to represent plastic compaction.

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

203

Fig. 1. DruckerPrager Cap model plasticity: (a) in the p-t plane; (b) in the deviatoric plane.

The DruckerPrager shear failure surface is written as


F s t  p tan b  d 0

14

where b and d represent the angle of internal friction and


the cohesion of the material respectively, t is a deviatoric
stress measure and p is the equivalent pressure stress.
The cap yield surface is
s

2
Rt
2
 Rd pa tan b 0
F c p  pa
1 a  a=cosb
15
where R is a material parameter that controls the shape of
the cap, a is a small number used to dene a transition yield
surface for a smooth intersection between the cap and
shear failure surfaces (for the sake of simplicity the transition surface is not written explicitly here) and pa is an evolution parameter, which is obtained by substituting (p, t)
with (pb, 0) in Eq. (15)
pa

pb  Rd
:
1 R tan b

16

In relation (16), pb represents the hydrostatic compression


yield stress and is dened by a bilinear hardening law (with
an almost horizontal branch), relating pb with the plastic
strain.
The model described above uses associated ow rule in
the cap region and non-associated ow in the shear failure
and transition regions. However, yielding in the shear (and
in the transition) surface is prevented because the contactfriction model stands for the shear failure of the masonry
arch. This is implemented by giving high enough values
in the parameters that mainly dene the shape of the shear
failure surface (b and d), see Fig. 1. In particular, b is considered to be equal to 70 and d equal to 1 GPa. Finally,
the material parameters R and a are taken to be equal to
0.005 and 0.03, respectively.

which denes the resultant force in a cross-section of


the arch). When the thrust line in a cross-section is adjacent to the ring of the arch (which is dened by the
intrados and the extrados curve of the arch), the eccentricity e of the normal force PN, from the center line
of the arch at the given cross-section, becomes maximum. As a result, a bending moment M equal to PNe
is developed around the center line of the arch and a
hinge occurs, on the assumption that the arch does not
develop any tensile strength. Since a three-pin arch is a
statically determinate structure, opening of a fourth
hinge converts the structure into a mechanism and collapse occurs. Therefore, the four hinges collapse mechanism is the collapse mode for an arch loaded with a
vertical concentrated force in the quarter-span of the
arch. The same loading position (together with the selfweight) is considered both in this study and in the experimental research on stone arches referred herein, as the
quarter-span is probably the worst position of the live
load according to [2]. In addition, no compressive failure
for the masonry is usually expected [2].
Concerning the inuence of the geometry of the arch on
the limit load, Heyman mentioned in his classical study
that the at arch is an innitely strong structure (a perfect structure) within the framework of three fundamental
assumptions: (i) the material has no tensile strength; (ii) no
sliding mode failure is possible for the arch; (iii) no compressive failure occurs. He explained that there is no pattern of hinges that lead to a four hinges collapse
mechanism and for this reason the at arch can be very
strong. The unilateral contact-friction model adopted for
the analysis of the stone arch, leads to similar results as it
will be shown later in this study.
4. Inuence of the geometry of the stone arch on the ultimate
load
4.1. General

3. A brief description of Heymans method


The classical collapse mechanism method of Heyman
[2], has been proposed for the determination of the load
carrying capacity of stone arch bridges. It is based on
the estimation of the thrust line (a funicular polygon

From the engineering point of view, it is interesting to


know the optimal shape of an arch and the inuence of different values of the span to rise ratio on the limit load and
collapse mechanism of the structure. Consequently, a parametric investigation of the inuence of the geometry of the

204

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

stone arch on the mechanical behavior of the structure is


presented in this section.
A number of models is considered, in which the span,
the out of plane width and the thickness of the ring (e.g.
the distance between the intrados and the extrados curve
of the arch) remain constant while the rise of the bridge
is gradually decreased. The rst model, which is the one
with the largest rise, has a span to rise ratio equal to 1.65
and a rise equal to 5.70 m. A reference geometry e.g. the
initial geometry of the arch bridge, is the one presented
in Fig. 2a with a span to rise ratio equal to 3.15 [21]. The
rise corresponding to this geometry, is equal to 3.00 m.
The geometry with the lowest rise has a span to rise ratio
equal to 32.5 and a rise equal to 0.30 m. The rise of each
model is 30 cm smaller than the previous one, thus a total
of nineteen models are analyzed. For the simulation of the
behavior of each structure, the unilateral contact-friction
model, which has been described previously, is adopted.
Fourteen interfaces are considered for each model. The
friction coecient is equal to 0.6 and the position of the
load is in the quarter-span of the bridge.
For the validation of the proposed method, which is
based on the contact eects, the ultimate failure load of
each model is recalculated by using a modern implementation of the classical collapse mechanism method based on
linear programming [5,6]. This formulation is included in
a software called Ring [7]. In particular, the structure is
divided into a number of rigid blocks while interfaces
between them allow opening or sliding. A linear programming formulation related to the upper bound theorem is
then used for the solution of the limit load analysis (e.g.
estimation of the thrust line and the limit load of the arch).
The associated ow rule is assumed in case frictional sliding along interfaces occurs.
In the nite element unilateral contact-friction model
the supports are horizontal and the two blocks adjacent
to the supports are considered xed to the ground
(Fig. 2a), for the whole range of the geometries. The same
conditions are applied to the Ring software up to a specic
shallow geometry, where the span to rise ratio is equal to
6.33 and the rise to 1.50 m, respectively. Below that geometry no result is received for horizontal supports with their
adjacent blocks xed to the ground, because numerical

instabilities arise in the linear programming formulation.


In order to obtain a limit load for these models, an inclination is given to the supports, see Fig. 2b. In addition, xed
boundary conditions are removed from the blocks adjacent
to the supports, thus an opening of a hinge or sliding is permitted to the interface of these blocks adjacent to the supports. The inclination (angle) given to the supports, is
properly chosen so as to lead to the greatest limit load, in
comparison with the limit load obtained for other values
of angles. Finally, the contact model is reconsidered for
these geometries. The same angle is applied to the supports
while the assumption that the two blocks adjacent to the
supports are xed to them is removed.
4.2. Results of the parametric investigation of the geometry
Two are the main conclusions which are obtained from
the investigation of the geometry of the stone arch. According to the rst one, reduction of the rise (e.g. increase of the
span to rise ratio) causes an increase of the ultimate load. It
seems that this continues until a particular geometry of the
arch which is called in this work: optimum geometry. If
the rise is further reduced then the limit load will be
reduced, as well. The analysis demonstrates that the optimum geometry is shallower than the initial geometry,
where initial geometry means the shape of the arch near
to the shape of existing, real structure. The rise corresponding to this optimum geometry is small enough compared to
the initial geometry of the structure, e.g. the bridge is close
to the ground (as it will be shown later). Accordingly, Heyman [2] mentioned that the at arch is an innitely strong
structure (a perfect structure) if no compressive failure of
the arch is possible to occur. The physical meaning of the
rst conclusion (e.g. reduction of the rise causes an increase
of the ultimate load) is related with the second conclusion.
As it will be shown in the next paragraph, the failure mode
of the arch is changed as the rise is reduced. In this alteration of the failure mode can be mainly attributed the fact
that when the rise is reduced the limit load is increased.
This could be also justied by another remark which is
mentioned in the following paragraphs and which indicates
that: as the rise is reduced the position of the hinges is
changing (in particular the pattern of hinges is transferred

Fig. 2. (a) Initial (reference) geometry with horizontal-xed conditions in supports; (b) A shallow geometry with an inclination and without xed
conditions in supports.

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

205

to the left side of the structure where the concentrated live


load is applied, as it will be explained later). Obviously this
remark only concerns shallow arches which collapse by the
four hinges mechanism.
The second conclusion of the present study, is related
with the collapse mode for the whole range of the geometries. In particular, up to a rise below the one of the initial
structure, the four hinges collapse mechanism is the failure
mode. But if the rise is further reduced indicating that the
shape of the arch approaches the at arch, then no four
hinges arise from the contact model. Similarly, Heyman
[2] underlined that there is no pattern of four hinges for
the at arch. From the present study it is concluded that
the collapse mode of the shallow arch is the compressive
failure of the masonry. Thus, one of the Heymans assumptions needs to be removed in order to depict the real
mechanical behavior of the arch.
The diagrams of Fig. 3 show the limit load and the corresponding span to rise ratio, obtained both by the contact
and the linear programming formulation. These diagrams
contain a range of the span to rise ratio up to the value
6.33 in which a shallow geometry with a rise equal to
1.50 m corresponds. Until this value of the rise, the boundary conditions of the Fig. 2a are applied to both models.
From the diagrams of Fig. 3 it is obvious that the two
methods result in an increase of the limit load when the
span to rise ratio is also increased (e.g. when the rise is
reduced). This increase of the limit load continues to take
place below the initial geometry (with a rise of 3.00 m). If
innite compressive strength is considered for the linear
programming formulation, the two diagrams are almost
identical and no divergence exists between them.
Moreover, for the contact model the compressive failure
of the arch starts in the shallow geometry with a span to
rise ratio equal to 6.33. If a compressive strength equal to
10 MPa is considered for the linear programming formulation, the compressive failure mode begins to become
important when the span to rise ratio is equal to 5.27
and the rise is 1.80 m, respectively. The geometry in which
the compressive failure becomes signicant in both models,
is more shallow in comparison with the initial geometry of
the structure.
The diagrams of Fig. 4 continue until the value 32.5 of
the span to rise ratio (rise 0.30 m). In these diagrams results
obtained only by the unilateral contact model are repre-

sented. The linear programming formulation presents


numerical instabilities for these shallow geometries, provided that the boundary conditions are like the ones
described in Fig. 2a. Consequently, no limit load is
obtained by the linear programming, for these boundary
conditions. Fig. 4 shows that there is a shallow geometry
where the limit load becomes maximum. After this geometry, the limit load is reduced. The optimal geometry is very
close to the at arch, indicating that the results of this
study are really close with the statements of Heyman about
the at arch. In particular, according to the contact model
the optimal geometry is received when the span to rise ratio
becomes 15.97 (Fig. 4). The rise of this bridge is equal to
0.60 m. The same optimal geometry arises if an innite
compressive strength is considered for the masonry or the
model with the cap yield surface which has been previously
described, is used for the compressive failure.
In the diagrams of Fig. 5 the failure load vs the span to
rise ratio are also represented, for a range of the span to
rise ratio between 7.92 and 15.97, similar to Fig. 4. The
value 32.5 of the span to rise ratio gives the limit load.
The results for this value are not included in Fig. 5, since
numerical instabilities arise for this shape in the linear programming model (i.e. following the classical collapse load
theory). It must be emphasized that the proposed method
remains stable in this case. Moreover, in these diagrams
the boundary conditions are as it is described in Fig. 2b,
e.g. an inclination is given to the supports while the block
adjacent to its support is not xed to it. The same boundary conditions have been applied both to the contact and
the linear programming formulation. From both models,
an identical optimum geometry is received. The span to rise

Fig. 3. Failure load-Span to rise ratio diagrams up to the value 6.33 of the
span to rise ratio.

Fig. 5. Failure load-Span to rise ratio diagrams up to the value 15.97 of


the span to rise ratio for the boundary conditions of Fig. 2b.

Fig. 4. Failure load-Span to rise ratio diagrams up to the value 32.5 of the
span to rise ratio.

206

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

ratio which corresponds to this geometry is equal to 10.60


(rise equal to 0.90 m), thus the optimum geometry for the
boundary conditions of Fig. 2b is 0.30 m deeper from the
optimum geometry obtained for the boundary conditions
of Fig. 2a and the contact model. This divergence seems
to be insignicant indicating that an optimum shape with
a rise between 0.60 m and 0.90 m arises, from both models
and for the variation of the boundary conditions described
above. It has to be mentioned here that in Figs. 4 and 5, the
diagrams show signicantly high values of the ultimate
load of shallow arches. This can be proven to be unrealistic, if one takes into account the behavior of the abutments.
Thus, failure of the abutments could lead to collapse before
the previously mentioned high values of the ultimate load
are attained.
Concerning the failure mode, both models result in the
four hinges mechanism up to a shallow geometry. In particular, for the contact model with horizontal supports and
xed boundary conditions (Fig. 2a), the four hinges collapse mechanism appears up to the value 7.92 of the span
to rise ratio (a rise equal to 1.20 m). For further reduction
of the rise, the compressive failure without the four hinges
is the collapse mode. Moreover, compressive failure initially appears in the shallow shape where the span to rise
ratio is 6.33 and the corresponding rise is 1.50 m. This
shape is below the initial geometry of the arch bridge (with
a rise equal to 3.00 m), indicating that compressive failure
of the masonry arch is an issue of concern for shallow
arches. At geometries where the rise is equal to 1.50 m
and 1.20 m respectively, compressive failure is accompanied with the four hinges collapse mechanism. In the linear
programming formulation with the same boundary conditions, the four hinges mechanism appears up to the value
6.33 of the span to rise ratio (a rise equal to 1.50 m). Below
that rise no limit load is obtained by the linear programming for the boundary conditions of the Fig. 2a, due to
numerical instability. Compressive failure begins to

become signicant when the span to rise ratio is equal to


5.27 and the rise is 1.80 m, respectively.
In Figs. 6 and 7 these two collapse modes depicted by
the proposed contact-friction model, are shown. Fig. 6a
represents the four hinges mechanism for the deepest arch
of the investigation while in Fig. 6b the same mechanism is
shown for another deep arch, with a rise equal to 3.90 m,
e.g. 90 cm higher than the one of the initial geometry. In
Fig. 7a the four hinges mechanism is shown for a shallow
arch with rise equal to 1.50 m. In this geometry the compressive failure mode starts, as it has been previously mentioned and it is represented in this Figure with the grey
color. In Fig. 7b the optimum geometry, with a rise equal
to 0.60 m, is represented. From this gure it is made clear
that the collapse mode is the compressive failure (grey
color), without any pattern of four hinges. The collapse
shapes of Figs. 6 and 7 are obtained from the contact
model by applying the boundary conditions of Fig. 2a.
In Figs. 8a and b the four hinges collapse mechanism for
the deepest arch and a shallow arch, both obtained by the
linear programming formulation, are shown. The boundary conditions of the Fig. 2a have also been applied to
the supports. By observing Figs. 6a and 8a it is clear that
the two methods result in identical patterns with four
hinges. Figs. 7a and 8b show the same collapse mechanism,
as well. Except these two representative shapes, similar collapse mechanisms obtained by the two methods, almost for
the total range of the geometries. Another comment can be
made by examining the Figures of the collapse modes. In
particular, by Figs. 6a, (b) and 7a it seems that while the
rise of the arch is reduced, the pattern of hinges is transferred to the left side of the structure where the concentrated live load is applied. The same remark is conrmed
by the results obtained by the linear programming formulation itself (see Figs. 8a and (b)). The force-displacements
diagrams for two deep and two shallow arches are shown
in Figs. 9a and (b), respectively.

Fig. 6. Unilateral contact-friction model: (a) Four hinges collapse mechanism for the model with rise equal to 5.70 m (the deepest arch); (b) Four hinges
collapse mechanism for the model with rise equal to 3.90m (deep arch).

Fig. 7. Unilateral contact-friction model: (a) Four hinges collapse mechanism for the model with rise equal to 1.50 m (shallow arch); (b) Compressive
failure of the masonry for the model with rise equal to 0.60 m (the optimum geometry, for the boundary conditions of Fig. 2a).

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

207

Fig. 8. Linear programming formulation with boundary conditions as represented in Fig. 2a: (a) Four hinges collapse mechanism for the model with rise
equal to 5.70 m (the deepest arch); (b) Four hinges collapse mechanism for the model with rise equal to 1.50 m (shallow arch).

Fig. 9. Force-displacement diagrams from the contact model with the boundary conditions of Figure 2a: (a) for two deep arches; (b) for two shallow
arches.

Finally, some comments related with the results


obtained from both the contact and the linear programming formulation where boundary conditions are similar
to the ones described in Fig. 2b, follow. The four hinges
collapse mechanism arises up to the value 7.92 of the span
to rise ratio (rise 1.20 m) for both models. The same conclusion arises for the boundary conditions of the Fig. 2a.
Moreover, compressive failure occurs in these models as
well, since they represent shallow geometries with span to
rise ratio between 7.92 and 15.97 (rise between 1.20 m
and 0.60 m, respectively). For the shallow geometry with
rise equal to 0.60 m, the block adjacent to the right abutment is sliding, indicating that sliding in the springing
opposite to the loading could be an issue if the supports
are not horizontal and sliding is permitted between the
block adjacent to the support and the support itself, see
Figs. 10a, (b). For these Figures, it seems that a hinge
opens in the left abutment denoting (together with the sliding in the right abutment) that each block adjacent to the
support is not xed to it. It should be mentioned here, that
the linear programming formulation takes into account the
sliding failure mode by using associated ow rule, see references [57]. This means that sliding in the tangential direction of an interface is accompanied with an opening in the
normal direction, indicating that the dilatancy angle is not

zero (in particular is equal to the friction angle) and the


associated ow rule is applied. For this reason, in the linear
programming formulation of Fig. 10b the sliding of the
block adjacent to the right abutment is accompanied with
an opening.
5. Settlement of supports of the stone arch
A nal application of the unilateral contact-friction
model, which is related with the movement of the supports
of the stone arch, is presented in this section. Heyman mentioned in his studies, that if the abutments spread for a reason, the arch could accommodate itself to the increased
span by forming three hinges, one at the crown in the extrados, and one at each abutment in the intrados. He
also mentioned, that if the abutments are too close,
three hinges have again been formed to accommodate
the decreased span. In this case, one hinge will be at the
crown in the intrados and one at each abutment in the
extrados (unlike the previous case where abutments had
spread).
Movement of one or two of the supports of a bridge
(with single span) could be an issue during the construction
of the structure or in case an earthquake would shake the
arch. A vertical movement of the supports could also take

Fig. 10. Collapse modes for the boundary conditions of the Fig. 2b: (a) the contact model; (b) the linear programming formulation.

208

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

place if the ground below the abutments is weak. Finally


the presence of water, like in rivers, may cause movements
of the supports. Three types of abutments movement may
generally occur. In particular, an abutment may move in
the horizontal direction outwards (in the direction of the
extrados of the arch), inwards (in the direction of the intrados of the arch) or vertically. It should be mentioned that
settlement of the abutments may become more dangerous
than an earthquake for the same bridge, as it has been
shown with an example in [17].
The behavior of the structure up to collapse after the
movement of the abutments is examined by the implementation of the unilateral contact - friction model, which has
been previously described. The analysis takes place in three
steps. In the rst step only the self-weight is applied while
in the second step a movement of supports is enforced.
During the third step, the concentrated live load is applied
either in the quarter span or in the middle span of the arch.
Movement of the left, of the right and of both of the supports are also considered. The geometry of the model
shown in Fig. 2a is used, with fourteen interfaces and a friction coecient equal to 0.6.
From the results, a rst comment can be made about the
value of the limit load of the structure, when a vertical or
an horizontal (outward or inward) movement of supports
takes place, in the presence of a quarter span load on the
arch. In particular, the limit load is exactly the same with
the one obtained from the model without any abutments
movement. This can be shown by the force-displacement
diagrams of Fig. 11, where the left support has been
enforced with a vertical or an horizontal movement equal
to 5 cm. The same remark can generally be made for a middle span loading. The diagrams of the Fig. 11 which represent the outward and the inward abutment movement,
have an irregularity indicating a temporary degradation

of the stiness. A possible explanation for this result will


be given later, by examining the failure mode of these
structures.
For the case of outward or inward movement of the left
abutment, a formation of three hinges takes place after
enforcing the movement, when the only load in the structure is the self-weight. As it is shown in Fig. 12, the position
of the three hinges coincides with Heymans statements.
The structure is statically determinate and when a fourth
hinge is created, the arch is about to collapse. In particular,
when the live load is applied at the quarter span of the arch
(in the left side of the structure), the fourth hinge does not
appear immediately, but an alteration in the positions of
the three hinges takes place. The way in which the hinges
are moving, is dierent in case of outward or inward or vertical displacement of the support, as well as in case a quarter or a middle span loading is applied. For outward
displacement of the left support and for a quarter span
load (at the left side of the arch), the hinges in the crown
and in the right abutment are moving counterclockwise
and a fourth hinge opens in the right abutment. The same
mechanism is developed, if the right support has an outward movement. Respectively, for an inward displacement
of the left support and for the same loading conditions, the
hinges in the left support and in the crown are moving
clockwise and the fourth hinge opens in the left support.
The same mechanism is also developed, if the right support
has an inward movement. The irregularity of the force displacements diagrams of the Fig. 11 could be attributed
to this alteration of the pattern of the three hinges until a
fourth hinge opens.
In case a vertical displacement is applied to the left support of an arch, two instead of three hinges are formed
when only the self-weight is present as it is shown in
Fig. 13.

Fig. 11. Forcedisplacement diagrams for the arch with and without
abutments movement.

Fig. 13. Vertical movement 5 cm of the left support when only the selfweight is applied to the arch.

Fig. 12. (a) Outward movement 5 cm of the left support when only the self-weight is applied to the arch; (b) Inward movement 5 cm of the left support
when only the self-weight is applied to the arch.

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

Fig. 14. Five hinges collapse mechanism in case a middle span loading is
applied to the arch.

If a middle span instead of a quarter span loading is


applied to the arch, the arch collapses with a ve hinges
mechanism, see Fig. 14. The fth hinge opens due to the
symmetry in both the geometry and in the loading conditions. The mechanism of the three hinges, for outward
and inward movement of abutments, and of the two hinges,
for vertical movement of the abutment, is not changed
when only the self-weight is present. However, the way
the hinges are moving when the live load is applied
changes. In particular, for an outward movement of the left
support, the two hinges of the springings (Fig. 12a) are
moving simultaneously (due to symmetry) towards the
middle of the arch. In their position two new hinges open,
as is shown in Fig. 14. Respectively, for an inward movement of the left support, the two hinges in the springings
are not moving, as they have the same position when the
structure collapses (see Figs. 12b and 14). Instead, two
new hinges open, each of them in the area between the
springing and the middle of the arch (Fig. 14). The hinge
in the middle opens at the end of the analysis. In fact
two activated hinges around the middle point are shown
in the gure. This is misleading due to the fact that only
a nite number of interfaces are available in the model. If
one uses a more detailed model, with higher number of
interfaces, the two active ones are coming closer and closer,
so that at the limit of a continuum model (e.g. a no-tension
material) one should expect to have only one hinge. This
assumption is in accordance with similar observations
based on Heymans theory.
Some further remarks are mentioned here. If an inward
displacement is simultaneously applied to both supports of
the arch and the load position is in the quarter span, then
everything is similar with the case of the arch in which only
one abutment has an inward movement. This is also the
case for a movement of both supports outward.

6. Conclusions
In the present study two dierent issues related with the
mechanical behavior of the masonry arch are presented. A
unilateral contact-friction nite element model is proposed
for the simulation of the arch and the calculation of the
limit load.
The rst part of this study is dedicated to the numerical investigation of the inuence of the geometry of the

209

stone arch on the mechanical behavior of the structure.


For the validation of the results obtained, a comparison
with a linear programming formulation based on the classical collapse mechanism method is made. Two are the
main conclusions which are obtained from this investigation. According to the rst one, reduction of the rise (e.g.
increase of the span to rise ratio) causes an increase of the
ultimate load up to a specic geometry, the optimum
geometry. If the rise is further reduced then the limit
load will be reduced, as well. The optimum geometry
which is obtained from this study has a small rise in comparison with the initial shape of the arch, indicating that
the shallow arch is generally stronger than the initial one.
Similarly, Heyman mentioned in his work that a at arch
can be a very strong structure. The second conclusion, is
related with the collapse mode for the whole range of the
geometries. In particular, up to a rise below the one of the
initial structure, the four hinges collapse mechanism is
the failure mode. But if the rise is further reduced then
no four hinges arise from the contact model. The collapse
mode in this case is the compressive failure of the
masonry.
The second part of the present work is devoted to the
inuence of the abutments movement on the masonry arch.
Results show that the limit load of a structure where movement of abutments takes place, does not change in comparison with the structure with xed supports. However, the
self-weight causes the formation of hinges in the arch, in
contrast with the structure having xed abutments. Heyman has also mentioned the development of hinges in the
arch if movement of the supports occurs. The corresponding force-displacement diagram, has an irregularity indicating a temporary degradation of the stiness. This is
attributed to the fact that when the live load is applied to
the arch, the hinges which have been created due to settlement of supports are moving until the failure of the
structure.
Further research need to be done concerning the
investigation of the behavior of shallow arches. As the
compressive collapse of these structures is the main failure mode, a more accurate failure model needs to be
adopted instead of the cap model with the hardening
law which is used in this study. For instance, a model
with a softening law could be considered. Moreover,
the inuence of the backll in the investigation of the
optimum shape would be interesting, as well. Finally,
as shallow arches seem to be stronger in comparison
with deep arches, large values of forces are going to be
developed in the abutments of shallow arches. As a
result, an investigation of the limit behavior of these
abutments should be conducted.
It should be mentioned here that the proposed method is
general and can be applied on multiple-span masonry
bridges, walls and other structures of more complicated
shapes. The reason for considering single-span bridges
has been the availability of published experimental and theoretical results for comparison.

210

G.A. Drosopoulos et al. / Construction and Building Materials 22 (2008) 200210

References
[1] Stavroulakis GE, Panagiotopoulos PD, Al-Fahed AM. On the rigid
body displacements and rotations in unilateral contact problems and
applications. Comput Struct 1991;40:599614.
[2] Heyman J. The masonry arch. Ellis Horwood series in engineering
science. England; 1982.
[3] Drucker DC. Coulomb friction, plasticity and limit loads. J Appl
Mech 1953;21:714.
[4] Livesley RK. Limit analysis of structures formed from rigid blocks.
Int J Numer Methods Eng 1978;12:185371.
[5] Gilbert M, Melbourne C. Rigid-block analysis of masonry structures.
Struct Eng 1994;72:35660.
[6] Melbourne C, Gilbert M. The behaviour of multiring brickwork arch
bridges. Struct Eng 1995;73:3947.
[7] Gilbert M. RING home page. Available from: http://www.ring.shef.ac.uk/. Internet, 2001.
[8] Ferris MC, Tin-Loi F. Limit analysis of frictional block assemblies as
a mathematical program with complementarity constraints. Int J
Mech Sci 2001;43:20924.
[9] Orduna A. Seismic assessment of ancient masonry structures by rigid
blocks limit analysis. Ph.D. Thesis, Department of Civil Engineering,
University of Minho, 2003.
[10] Orduna A, Lourenco PB. Three-dimensional limit analysis of rigid
blocks assemblages. Part I: Torsion failure on frictional interfaces and
limit analysis formulation. Int J Solids Struct 2005;42:514060.
[11] Orduna A, Lourenco PB. Three-dimensional limit analysis of rigid
blocks assemblages. Part II: Load-path following solution procedure
and validation. Int J Solids Struct 2005;42:516180.

[12] Ng K-H, Faireld CA. Modifying the mechanism method of masonry


arch bridge analysis. Construct Build Mater 2004;18:917.
[13] Drosopoulos GA, Stavroulakis GE, Massalas CV. Limit analysis of a
single span masonry bridge with unilateral frictional contact interfaces. Engineering Structures 2006;28:186473.
[14] Haslinger J, Makinen RAF. Introduction to shape optimization:
Theory, approximation and computation. Philadelphia: SIAM; 2003.
[15] Mistakidis ES, Stavroulakis GE. Nonconvex optimization in mechanics. Smooth and nonsmooth algorithms, heuristics and engineering
applications. Dordrecht (The Netherlands): Kluwer Academic Publishers; 1998.
[16] Panagiotopoulos PD. Inequality problems in mechanics and applications. Convex and nonconvex energy functions. Boston, Basel,
Stuttgart: Birkhauser Verlag; 1985.
[17] Leftheris BP, Stavroulaki ME, Sapounaki AK, Stavroulakis GE.
Computational mechanics for heritage structures. Southampton
(UK): WIT Press; 2006.
[18] Stavroulakis GE. Optimization for modeling of nonlinear interactions
in mechanics. In: Pardalos PM, Resende GC, editors. Handbook of
applied optimization. New York: Oxford University Press; 2002. p.
97891.
[19] Hofstetter G, Mang HA. Computational mechanics of reinforced
concrete structures. Braunschweig (Germany): Friedr. Vieweg und
Sohn Verlag; 1995.
[20] Orduna A, Lourenco P. Cap model for limit analysis and strengthening of masonry structures. J Struct Eng 2003;129:136775.
[21] Page J. Load tests to collapse on two arch bridges at Strathmashie
and Barlae. Department of Transport, TRRL Research Report 201.
Crowthorne (England): TRL; 1989.

Вам также может понравиться