You are on page 1of 13

Engineering Structures 90 (2015) 117–129

Contents lists available at ScienceDirect

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

A finite-discrete element model for dry stone masonry structures


strengthened with steel clamps and bolts
Hrvoje Smoljanović, Željana Nikolić ⇑, Nikolina Živaljić
University of Split, Faculty of Civil Engineering, Architecture and Geodesy, Split, Croatia

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

Article history: This paper presents a new robust finite-discrete element numerical model for analysis and prediction of
Received 16 November 2013 the collapse of dry stone masonry structures strengthened with steel clamps and bolts. The model
Revised 19 December 2014 includes fracture and fragmentation of the blocks as well as a cyclic behaviour, yielding, stiffness
Accepted 5 February 2015
degradation, failure and the influence of pulling out of the clamps and the bolts from the stone block.
Available online 6 March 2015
The developed model can be used for the estimation of the seismic resistance of historical dry stone
masonry structures reinforced with steel clamps and steel bolts, which is very important for the
Keywords:
structures classified as cultural heritage.
Dry stone masonry structures
Steel clamps
Ó 2015 Elsevier Ltd. All rights reserved.
Steel bolts
Finite-discrete element method
Incremental dynamic analysis

1. Introduction displacement and rotation including complete detachment such as


recognizing new contacts. To overcome this limitation some finite
A large part of cultural heritage all over the world are historical element formulations with large displacements [13,14] and con-
structures built as dry stone masonry. However, some of them tact detection have been developed [15].
which were originally built with mortar joints have experienced It is noted that other attractive tools for modelling of dry stone
a significant loss of mortar during time and the behavior of these masonry structures are based on a discrete element method [16–
structures becomes similar to those made of dry stone masonry. 23]. The common idea in different applications of the discrete ele-
Most of these structures have been damaged due to seismic activ- ment method to masonry structures is the idealization of the mate-
ity [1]. With the aim of increasing their resistance, many of dry rial as a discontinuum where joints are modelled as contact
stone historical structures were further strengthened by steel surfaces between different blocks. This approach is suitable for
clamps and bolts. modelling different types of non-linear behaviour including large
In order to evaluate the resistance of these structures and to be displacements and rotation with complete detachment of blocks.
able to preserve the cultural heritage it is necessary to develop a In recent times an increasing number of models attempted to
numerical model which could take into account all the effects combine the advantages of finite and discrete element methods
occurring in dry stone masonry structures including the fragmen- [24–28]. The most advanced and most often used numerical meth-
tation of the blocks and non-linear behaviour of steel clamps and ods which combine the advantages of the finite and discrete ele-
bolts during dynamic loading. ment method are Discontinuous Deformation Analysis (DDA)
The most commonly used numerical tool for the analysis of [29] and Combined Finite-Discrete Element Method (FEM/DEM)
masonry structures is the finite element method where the mate- [30–32]. These methods are designed to handle contact situations
rial is regarded as a fictitious homogeneous orthotropic continuum in which transition from continua to discontinua can appear. DDA
[2–6]. These models encounter a significant limitation to simulate is more suitable for static problems, while FEM/DEM is more suit-
strong discontinuities between different blocks of the masonry. For able for problems involving transient dynamics until the state of
overcoming these limitations joint interface elements were devel- rest or steady state is achieved.
oped to model the discontinuities [7–12]. Most of these models Within the framework of the FEM/DEM method the blocks are
cannot take into account the mutual mechanical interaction, finite discretized by constant strain triangular finite elements. Material
non-linearity, including fracture and fragmentation of discrete ele-
ments as well as cyclic behaviour during dynamic load [33], is con-
⇑ Corresponding author.
sidered through contact elements which are implemented within a
E-mail address: zeljana.nikolic@gradst.hr (Ž. Nikolić).

http://dx.doi.org/10.1016/j.engstruct.2015.02.004
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
118 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

finite element mesh [34,35]. The interaction between discrete ele-


ments is considered through the contact interaction algorithm
based on the principle of potential contact forces [36,37] and the
Coulomb-type law for friction [38]. The method uses an explicit
numerical integration of the equation of motion.
The FEM/DEM method was successfully applied in the analysis
of dry stone masonry structures and showed good agreement with
experimental results [39]. This model is capable of predicting the
collapse mechanism of dry stone masonry structures under seismic
loads as well as for determining the safety of the structure with
regard to the occurrence of the collapse load. Due to this reason
the combined finite-discrete element method is used as a basis
for developing a new numerical model for steel clamps and steel
bolts which is presented in this paper. This model can be useful
in making the right decisions regarding the restoration of dry stone
masonry structures which have experienced deterioration over
time. It can be important because the most part of these structures
are classified as cultural heritage.

2. Types of steel clamps and bolts


Fig. 2. Steel clamps and bolts.

Historical dry stone masonry structures are commonly


strengthened with two types of steel clamps (Fig. 1). Steel clamps
model for steel clamps and bolts has been implemented into a
of type I (Fig. 1a) are inserted on the lateral surface of the structure
Y2D computer program which will be presented below.
into the previously made holes that are subsequently backfilled.
Schematic presentations of steel clamps type I and II and steel
These types of clamps are most commonly used in strengthening
bolts modelled in this paper are shown in Fig. 2.
of dry stone walls. Considering the dynamic response of structure
in plane, extracting of this type of clamps from stone block cannot
occur. Steel clamps of type II (Fig. 1b) are inserted on the top side of 3. Discretization of a dry stone structure with steel clamps and
stone blocks into the previously made holes. They are often used in bolts
the construction of dry stone arches. Unlike clamps type I, in the
case of dynamic response of structures in the plane, clamps type In this numerical model each stone block is modelled as a dis-
II can extract from the previously made holes due to the relative crete element which is discretized by constant triangular finite ele-
displacements of the stone blocks. In that case, clamps type II lose ments. Contact interaction between stone blocks is considered
its carrying purpose which is necessary to take into account in through the contact interaction algorithm based on the principle
numerical modelling. Both types of steel clamps have only tension of potential contact forces [36,37] which include the Coulomb-type
bearing capacity. The steel bolts are most commonly used when law for friction [38]. Material non-linearity, fracture and fragmen-
connecting the capitals and columns or capitals and upper beams tation are considered through the contact elements which are
(Fig. 1c) and they dominantly have shear bearing capacity. implemented within the finite element mesh of each block.
Due to the presence of many parameters which affect on beha- The steel clamps type I and II and steel bolts were modelled
viour of clamps and bolts in dry stone masonry structures such as with one-dimensional elements which can be placed in arbitrary
the elastic properties of stone and steel, the width and depth of the positions inside the stone finite elements.
hole into which the clamps and bolts are inserted, elastic proper- Discretization of dry stone masonry structure with embedded
ties of infill material, geometry of clamps and bolts, etc., it is very steel clamps types I and II and bolts is shown in Fig. 3.
difficult to develop numerical model which can take into account
all types of failure mechanisms and especially the influence of local 4. Material model of the stone block
interaction between the bolt or clamp on one side and masonry
block on another. The most important features that characterize the non-linear
In order to be able to analyse such dry stone masonry structures behaviour of dry stone masonry structures due to seismic activity
by the finite-discrete element method, a new robust numerical are mutually sliding of blocks along contact surfaces, the rotation

Fig. 1. Steel clamps and bolts: (a) steel clamp inserted on the lateral face of the structure; (b) steel clamp inserted on the top side of stone blocks; (c) steel bolts [40].
H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 119

Fig. 3. Discretization of dry stone structure with steel clamps and bolts.

of a single block or one part of the structure, mechanical interac-


tion between the blocks, the presence of asperities in contact sur-
face of stone blocks which exert influence on the normal and
tangential stiffness of the structure, etc. Material nonlinearity
accompanied with fracture and fragmentation of stone blocks is
not so present in dry stone masonry [39,41], however it became
noticeable in the case of large compressive stress, or in the case
of the progressive collapse of the structure.
Due to the previously mentioned facts, in the model presented
in this paper the behaviour of a constant triangular finite element
is linear viscoelastic in compression and tension. The relation
between stress and strain is given by Hooke’s law according to:
Fig. 4. Strain softening curve defined in terms of separations.
E ^ E ^
T¼ Ed þ Es þ l
D ð1Þ
1þt 1  2t

where T is the Cauchy stress tensor, E is the modulus of elasticity, t implementation, it is enforced through the penalty function
^ ^
is Poisson ratio, Ed is the shape changing part and Es is the volume method [36].
changing part of Green-St. Venant strain tensor, l  is the damping For separation d 6 dt the bonding stress is given by:
"  2 #
coefficient and D is the rate of the deformation tensor [30]. The last
2d d
term in Eq. (1) presents the energy dissipation due to the propaga- r¼  ft ð2Þ
dt dt
tion of elastic waves occurred by contact interaction of two adjacent
blocks. where dt = 2hft/p0 is the separation corresponding to the bonding
The presented numerical material model in finite element can stress being equal to the tensile strength ft, h is the size of particular
be additionally improved by introducing the diffuse dissipative element (average length of side of triangle) and p0 is penalty term.
mechanisms (micro-cracks) accompanied by strain hardening in In that way the separation of adjacent elements before reaching the
pre-peak fracture process [42–44] which would be interesting for tensile strength is normalized by the element size.
analyses of dry stone masonry structures under high compressive After reaching tensile strength ft stress decreases with an
loads. increasing separation d and at d = dc bonding stress tends to zero.
Material nonlinearity in post-peak fracture process including In literature this phenomenon is known as the strain softening.
fracture and fragmentation are realized by a combined single and The stress–crack opening curve which describes the strain soften-
a smeared crack model in contact elements [34]. ing behaviour of the stone at the fracture process zone after peak,
The cracks are assumed to coincide with the finite element depends primarily on the type of stone and its structure, which is
edges, which are achieved in advance through the topology of adja- characterized with the shape, size and orientation of minerals in it.
cent elements being described by different nodes. The separation The curve used in presented numerical model represents an
of these edges induces a bonding stress which is taken to be a func- approximation of the experimental stress–separation curves on
tion of the size of separation d (Fig. 4). The area under stress– concrete specimens [45], however, it proved to be in good correla-
separation curve represents the energy release rate Gf = 2c, where tion with experimental results describing fracture process in other
c is the surface energy, i.e. the energy needed to extend the crack quasi-brittle materials that have a disordered internal structure
surface by unit area. like brick [46] and stone [39]. For separation dt < d < dc the bonding
In theory the separation, d = dt = 0 coincides with the bonding stress is given by:
stress being equal to the tensile strength ft, i.e., no separation
occurs before the tensile strength is reached. In the actual r ¼ yft ð3Þ
120 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

where y is a heuristic scaling function representing an approxima- expected place, or by improving presented numerical model by
tion of the experimental stress–separation curves taken according introducing embedded discontinuity [42–44] which is proved to
to Hordijk [45] provide a mesh-invariant response.

y ¼ ½1 þ ðc1 DðdÞÞ3 ec2 DðdÞ  DðdÞð1 þ c31 Þec2 ð4Þ


5. Numerical model of steel clamps
where c1 = 3 and c2 = 6.93, while the damage parameter D(d) is
determined according to the following expression: 5.1. Steel material model

ðd  dt Þ=ðdc  dt Þ; if dt < d < dc ; The stress–strain relationship for a monotonically increasing
DðdÞ ¼ ð5Þ
1; if d > dc loading in steel is shown in Fig. 5a. The hysteresis behaviour of a
steel is enforced through Kato’s stress–strain model [48] shown
As an alternative to the described curve, one can use the curve
in Fig. 5b. From the given strain, stress is calculated by the follow-
proposed by Hillerborg et al. [47].
ing expressions:
The edges of two adjacent elements are held together by shear
stress calculated by using the penalty function method [36]. After
(1) during unloading (Fig. 5b, curve (1)):
reaching shear strength fs, which coincides with sliding t = ts, the
stress decreases with an increasing sliding t and at t = tc shear rsc ¼ f y  Es ðesh  esc Þ ð8Þ
stress tends to zero. For sliding ts < |t| < tc shear stress is given by:
where Es is Young’s Modulus of steel.
s ¼ ð1  DðtÞÞf s ð6Þ (2) during negative loading (Fig. 5b, curve (2)):
" ,( ! )#
where D(t) is the damage parameter given by: EB

rsc ¼ f y a  faða  1Þg  ðesc  esh þ ey Þ þ a  1
fy
ðt  ts Þ=ðtc  ts Þ; if t s < t < t c ;
DðtÞ ¼ ð7Þ
1; if t > t c ; ð9Þ

In presented numerical model crack pattern is pre-determined where EB ¼ ðEs =6Þ log 10ðesh  ey Þ, a ¼ Es =ðEs  EB Þ;
with finite element mesh. In cases when the cracking of stone (3) during reloading–unloading (Fig. 5b, curve (3)):
blocks are expecting, this problem can be avoid using relatively
rsc ¼ rpm þ Es ðesc  epm Þ ð10Þ
small meshes and consequently moderate CPU power [35], adopt-
ing large finite element mesh which can provide cracks on where rpm is the minimum value of rsc in its loading history;

(a) (b)
Fig. 5. Stress–strain model of steel: (a) monotonic loading; (b) cyclic loading (Kato).

Fig. 6. Steel clamp type I in initial and current configuration.


H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 121

(4) during reloading (Fig. 5b, curve (4)): Coordinates of point P1 in current configuration can be obtained in
" ,( ! )# similar way.
EB
rsc ¼ f y þ rpm þ f y a  faða  1Þg  ðey  esc þ epm Þ þ a  1 Strain of steel clamp is given by
fy
ð11Þ lc  li
esc ¼ ð14Þ
li
5.2. Steel clamps type I where lc and li are lengths of a steel clamp in current and initial con-
figurations respectively.
The steel clamp type I was defined by its first P0 and end P1 The force in steel clamp (Fig. 7a) is given by
points (Fig. 6). The strain of a steel clamp in any time step can be
obtained from coordinates of these points in current configuration. jf 0sc j ¼ jf 1sc j ¼ Asc rsc ð15Þ
Coordinates of point P0 in current configuration are obtained
according to where Asc is the cross-sectional area and rsc is the stress of the steel
clamp obtained from the steel material model. Forces f0sc and f1sc
x0c ¼ a0 þ a1 x0i þ a2 y0i acting in points P0 and P1 are distributed into the nodes of the par-
ð12Þ
y0c ¼ b0 þ b1 x0i þ b2 y0i ent stone triangular finite element in the form of equivalent nodal
forces (Fig. 7b).
where x0i i y0i are coordinates of point P0 in initial configuration.
Unknown coefficients ai, bi i = 0, 1, 2 can be obtained from the
5.3. Steel clamps type II
known coordinates of a parent triangular finite element in current
configuration which leads to
The steel clamp type II was defined by its first point P0, the end
xic ðxji yki  xki yji Þ þ xii ðxkc yji  xjc yki Þ þ yii ðxjc xki  xji xkc Þ point P1 and the anchorage length lk (Fig. 8). The intersection
a0 ¼
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ between the blocks edges and line segment P0 P1 gives the referents
xic ðyji  yki Þ þ xjc ðyki  yii Þ þ xkc ðyii  yji Þ points R0 and R1 (Fig. 8).
a1 ¼ The coordinates of points P0, P1, R0 and R1 in the current con-
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ
figuration are obtained in a similar way as it was shown in the
xic ðxji  xki Þ þ xii ðxkc  xjc Þ þ xjc xki  xji xkc model of steel clamps type I. The strain of steel clamps is given
a2 ¼ 
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ by (14) where lc and li are the lengths of the steel clamp type II
ð13Þ
yic ðxji yki  xki yji Þ þ xii ðykc yji  yjc yki Þ þ yii ðyjc xki  xji ykc Þ in current and initial configurations respectively (Fig. 8). The stress
b0 ¼ in the steel clamp type II rsc is obtained from the stress–strain rela-
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ
tion defined by the steel material model.
yic ðyji  yki Þ þ yjc ðyki  yii Þ þ ykc ðyii  yji Þ
b1 ¼ The influence of tangential separation p is approximately taken
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ
into account through a reduction of stress r0sc given by
yic ðxji  xki Þ þ xii ðykc  yjc Þ þ yjc xki  xji ykc
b2 ¼ 
xii ðyji  yki Þ þ xji ðyki  yii Þ þ xki ðyii  yji Þ
r0sc ¼ zrsc ð16Þ

(a) (b)
Fig. 7. Steel clamp type I forces: (a) force in steel clamp; (b) equivalent nodal forces.

Fig. 8. Steel clamp type II in initial and current configuration.


122 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

where z is the scaling function. The scaling function is equal to one


when there is no shear separation while it is equal to zero when the
shear separation is equal to lk. For shear separation 0 < p < lk the
scaling function depends on the elastic properties of stone and a
steel clamp, the width of the hole in which the steel clamp is
embedded, the elastic properties of the infill material, the cross sec-
tion area of the clamp, etc. In this numerical model the scaling func-
tion is assumed as [49]
ea D
z ¼ 1:0  ð17Þ Fig. 11. Stress field in clamp and in stone around the clamp.
1:0 þ ðea  1:0ÞD
where variable D = D(p) is determined according to following
load compressive in stone linearly decreases to a depth of three
expression
8 thicknesses (Fig. 11) of clamp, local crushing of stone would not
< 0:0
> for p ¼ 0:0; appear if it is
DðpÞ ¼ 1:0 for p P lk ; ð18Þ
>
: f u hb 6 3hbð3:3f c Þ=2 ð20Þ
p=lk other;
which leads to
The shape of function z for different values of shape parameter a is
shown Fig. 9. f c P 0:2f u ð21Þ
In absence of experimental results the shape parameter a = ac
where b is width and h is the thickness of clamp.
for steel clamps can be initially set to zero which leads to a linearly
Adopting the value of tensile strength of steel fu being equal
decreasing function. If experimental results exist, parameter ac can
500 MPa, the compressive strength of stone should be higher than
be chosen to best fit experimental data.
100 MPa in order to avoid local crushing of stone. Taking into
The force in a steel clamp (Fig. 10a) is given by
account that the average compressive strength of stone ranges
jf 0sc j ¼ jf 1sc j ¼ Asc r0sc ð19Þ between 80 MPa and 250 MPa [51] and that clamps are usually
made of copper with tensile strength significantly smaller then
Forces f0sc and f1sc acting in points P0 and P1 are distributed into the
500 MPa, it can be concluded that adopted assumption is justified.
nodes of the parent stone triangular finite element in the form of
Due to these reasons, it is expected that local nonlinearities which
equivalent nodal forces (Fig. 10b).
can appear around clamp have no significant influence on the glob-
In actual implementation it was assumed that local crushing of
al dynamic behaviour of the structure.
stone around clamp type I and II would not appear. Taking into
account that the local compressive strength of stone is 3.3 times
5.4. Behaviour of clamps type II under monotonic loading
higher than the global compressive strength fc [50] and assuming
that in the case of local crushing of stone around the clamp the
The model described above has been implemented in the Y2D
code which is based on the combined finite-discrete element
method [31,52].
The behaviour of clamps type II under monotonic loading was
performed on two rigid triangles connected with a steel clamp
(Fig. 12). The material characteristics of the steel clamp are shown
in Table 1.
The monotonically increasing load was performed in terms of
constant velocity vx and vy in point B. Velocity vx was equal
vx = 0.2 m/s while velocity vy was varied with values of 0.0 m/s,
0.2 m/s, 0.3 m/s and 0.4 m/s.
The stress–strain relation in a steel clamp for all cases of loading
is shown in Fig. 12. It can be seen that for velocities vy different
from zero, the reduction in stress occurs due to the extracting of
Fig. 9. Reduction factor for different values of shape parameter a. the steel clamp from the stone block. Extracting of the steel clamp

(a) (b)
Fig. 10. Steel clamp type II forces: (a) force in steel clamp; (b) equivalent nodal forces.
H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 123

Fig. 12. Stress–strain relation in steel clamps for different cases of loading.

Table 1
Material characteristics of steel.

Modulus of elasticity Yield stress Ultimate stress Cross section area Strain (end of yield) Ultimate strain Break strain
Es (MPa) fy (MPa) fu (MPa) As (m2) esh eu ebr
183,000 446 640 0.05 0.005 0.100 0.120

from the block increases with the increase of the velocity vy, which Tangential separation s induces shear stress ssb in the steel bolt
leads to the additional reduction of the stress in the steel clamp. and at separation s = spb the shear stress reach its maximum fsu
(Fig. 14a). With increasing tangential separation s > spb shear stress
decreases and at separation s > stb it drops to zero and the bolt is
6. Numerical model of steel bolts assumed to be broken.
Values of fsu, spb, stb and shape of functions, which define the
The steel bolt was defined by its first point P0 and the end point relation between shear separation and shear stress, depends on
P1. The intersection between the blocks edges and line segment elastic properties of stone and a steel bolt, the width of the hole
P0 P1 gives the referents points R0 and R1 (Fig. 13). The coordinates in which the steel bolt is embedded, the elastic properties of the
of points P0, P1, R0 and R1 in current configuration are obtained in a infill material, the cross section area of the bolt, etc., and need to
similar way as it was shown in model of the steel clamps type I. be determined experimentally for each particular case.

Fig. 13. Steel bolt in initial and current configuration.

(a) (b)
Fig. 14. Material model in the steel bolt: (a) shear stresses versus shear separation; (b) cyclic behaviour.
124 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

8   
The maximum value of shear stress fsu in steel bolt is limited < 2 jsj  jsj 2 f z for jsj 6 s ;
spb spb su pb
with shear strength of bolt material fsb which can be written as ssb ¼ ð29Þ
:
zfsu for jsj 6 spb ;
f su 6 f sb ð22Þ
The influence of normal separation o is approximately taken
and with shear stress in bolt which cause local crushing of stone into account through a reduction of stress s0sb given by
around the bolt. The shear strength of steel fsb can be determined
from the tensile strength fu by applying the Von Mises yield criteri- s0sb ¼ zssb ð30Þ
on which leads to where z is the scaling function. The scaling function is equal to one
pffiffiffiffiffiffiffiffi when there is no normal separation while it is equal to zero when
f sb ¼ 1=3f u ð23Þ
the normal separation is equal to l/2 where l is the steel bolt length.
Taking into account that the local compressive strength of stone For normal separation 0 < o < l/2 the scaling function depends of
is 3.3 times higher than the global compressive strength fc [50] and elastic properties of stone and a steel bolt, the width of the hole
assuming that in the case of local crushing of stone around the bolt in which the steel bolt is embedded, the elastic properties of the
the compressive stress in bolt linearly decreases to a depth of three infill material, the cross section area of the bolt, etc. In this numer-
diameters of bolt, it can be written ical model the scaling function is assumed according to relation (17)
2 where the variable D = D(o) is determined according to following
d p 2 expression
f su 6 3d ð3:3f c Þ=2 ð24Þ
4
D ¼ DðoÞ ¼ 2o=l ð31Þ
which leads to
Cyclic behaviour of bolt (Fig. 13b) is assumed as
f su 6 6:3f c ð25Þ
Dðstb  spb Þ þ spb
where d is diameter of bolt. z ¼ zðDmax Þ ð32Þ
Dmax ðstb  spb Þ þ spb
In actual implementation for separation 0 6 |s| < spb shear stress
is given by where Dmax is the maximum value of D(s) in its loading history.
 2 ! Shear force in the steel bolt (Fig. 15a) is given by
s s
ssb ¼ 2  f su ð26Þ jf 0sb j ¼ jf 1sb j ¼ Asb ssb ð33Þ
spb spb
where Asb is cross-section area of the bolt.
where value of spb is treated as input parameter. For separation spb -
Forces f0sb and f1sb which are assumed to act in the centre of the
6 |s| < stb shear stress is assumed as
bolt anchored in the stone block, are distributed into the nodes of
ssb ¼ zfsu ð27Þ the parent stone triangular finite element in the form of equivalent
nodal forces (Fig. 15b).
where z is the scaling function defined with (17) where
8
< 0:0
> for jsj < spb ; 6.1. Behaviour of bolts under monotonic loading
D ¼ DðsÞ ¼ 1:0 for s P stb ; ð28Þ
>
: The model described above has been implemented in the Y2D
ðjsj  spb Þ=ðstb  spb Þ other;
code which is based on the combined finite-discrete element
The value of the shape parameter a = abs needs to be chosen to method [31,52].
best fit experimental data. In the absence of experimental data it The behaviour of the model of bolts under monotonic loading
can be set to initial value abs = 0 (Fig. 9). However, collapse of the was performed on two rigid triangles connected with steel bolt
stone masonry structures usually occurs due to the loss of the glob- (Fig. 16). The parameters applied in the numerical analysis are
al stability and this parameter has no influence on the global struc- shown in Table 2.
tural behaviour. If the collapse of the structure caused by the The monotonically increasing load was performed in terms of
breaking of the bolts, parameter abs has only influence on the shape constant velocity v = 0.2 m/s in point B. Initial normal separation
of the load–displacement curve in softening phase, but not to the o in numerical analysis was varied with values of 0.0 m, 0.05 m
value of collapse load. and 0.075 m.
The complete relationship for the shear stress as the function of Shear stress–shear separation relations in steel bolts for differ-
shear separation can be written as ent initial normal separation are shown in Fig. 16. It can be seen

(a) (b)
Fig. 15. Steel bolt: (a) force in steel bolt; (b) equivalent nodal forces.
H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 125

Fig. 16. Shear stress–shear separation relations in steel bolt for different initial normal separation o.

Table 2 granite columns with Dorian capitals are located at the entrance
Numerical parameter of steel bolt.
of the Prothyron. The capitals support the broad gable with an arch
Diameter Tangential Ultimate tangential Break tangential in the middle. The structure was originally built of dry stone blocks
strength separation separation with steel bolts embedded between columns, capitals and the
D (mm) ftb (MPa) spb (mm) stb (mm)
upper beams. Along the upper edge of broad gable the structure
50 369.5 0.05 3.00 was supported by wooden beams in direction perpendicular to
the plane of the structure, which makes it suitable for analysis with
2D numerical models. Throughout history deformations of the
stone blocks which constitute the broad gable have occurred with
that increasing the initial normal separation, the reduction in shear
the movement of the central columns. Due to the movement of the
stresses also increases.
blocks, restoration of the structure was performed using copper
clamps during the period of the Austro-Hungarian Empire.
7. Numerical application In order to evaluate the dynamic response of the structure, the
incremental dynamic analysis of the original geometry of the
7.1. Seismic analysis of the Prothyron in Split – Croatia structure before the restoration was applied. The analyses were
performed for three cases: (a) structure without steel bolts and
The following example shows the application of the developed copper clamps, (b) structure with steel bolts embedded between
model (FEM/DEM method) for the simulation of the dynamic columns, capitals and the upper beams (original structure before
response of the structure of the Prothyron in Split (Fig. 17) to seis- restoration), (c) structure with steel bolts and copper clamps
mic load. (structure after restoration).
The Prothyron is located on the south side of the Peristyle (the For all three cases the structure was exposed to horizontal and
square in front of the Cathedral of St. Domnius). Four massive vertical ground acceleration (Fig. 18) which was recorded on
15.4.1979 in Dubrovnik on rock soil during an earthquake with
the epicentre in Petrovac (Montenegro). The accelelogram was
firstly scaled on peak ground acceleration ag = 0.22 g which is valid
for Split. After that the acceleration was gradually increased to the
collapse of the structure.
Figs. 19 and 20 respectively show the geometry and discretiza-
tion of structure.
Young’s Modulus of clamps and stones and tensile strength of
clamps used in the numerical analysis are taken from the literature
[53] and shown in Table 3. Bolts were made of steel with known
tensile strength. Shear strength of steel was obtained from steel
tensile strength fst by applying the Von Mises yield criterion given
by (23) and it was adopted as 239 MPa. Assuming the value of
compression strength of stone being equal 100 MPa, shear stress
in bolt which leads to the crushing of the stone can be obtained
Fig. 17. The Prothyron entrance at the Peristyle in Split. according to (25) and is equal 630 MPa. Since the shear strength

0.8 0.8
0.6 0.6
2

2
acceleration / m/s

acceleration / m/s

0.4 0.4
0.2 0.2
0.0 0.0
-0.2 0 10 20 30 -0.2 0 10 20 30
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
time / s time / s
(a) (b)
Fig. 18. Ground acceleration recorded during the Petrovac earthquake (1979. Montenegro): (a) horizontal direction; (b) vertical direction.
126 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

Fig. 19. Geometry of the Prothyron structure.

Table 4
Material characteristics of stone.

Tensile Shear Fracture energy in Fracture energy in


strength strength tension shear
ft (MPa) fs (MPa) Gft (N/m) Gfs (N/m)
10 20 720 1440

Dynamic analysis of the structure for the acceleration of


ag = 0.22 g showed that the movement of the blocks, i.e. opening
of the joints appeared for a dry stone structure (Fig. 21a) and a
structure with steel bolts (Fig. 21b). For a dry stone structure with
steel bolts and copper clamps (Fig. 21c) there was no opening of
the joints.
Dynamic analysis showed that the collapse of dry stone struc-
ture occurred for the acceleration of ag = 0.45 g (Fig. 22a). For the
same acceleration the structure with steel bolts also sustained total
collapse (Fig. 22b). Dry stone structure with steel bolts and copper
Fig. 20. Finite element mesh of the Prothyron structure.
clamps has experienced the movement of the blocks i.e. opening of
the joints (Fig. 22c).
The collapse mechanism for dry stone structure with steel bolts
of steel is smaller than the shear stress in bolt which leads to and peak ground acceleration of ag = 0.45 g over time is shown in
crushing of the stone, shear strength of steel was adopted as the Fig. 23. It can be seen that the collapse of the structure with steel
shear strength of the bolt. bolts occur due to the spacing of central columns followed by col-
In the performed analyses the damping coefficient was assumed lapse of central arc. It can be concluded that the contribution of
and equal 35  106 which was proven to provide a realistic predic- bolts in terms of increasing seismic resistance is insignificant and
tion of behaviour in dry stone masonry [39]. The value of sliding that the bolts had only a constructive purpose in the construction
friction was supposed and it was adopted as 0.6. phase.
Material characteristics of stone which were used in numerical The behaviour of the structure with steel bolts and copper
analysis and which values were not available were assumed on the clamps for an acceleration ag = 1.2 g over time is shown in
basis of data from literature [51] and shown in Table 4. Fig. 24. The figure shows that the broad gable with embedded cop-
per clamps behaves as one body.
Normal stress in the copper clamp C1, shear stress in the steel
bolt B1 and displacement of the top of the structure in time for
Table 3
ag = 1.2 g, are shown in Fig. 25.
Material characteristics of steel and stone [53].
The performed analysis shows that the original structure before
Sample Young’s Modulus Tensile strength its restoration with the copper clamps had low seismic resistance
E (MPa) fst (MPa)
because the significant displacement of the central block occurred
Bolt 181,000 414 at a design acceleration ag = 0.22 g. The influence of embedded
Clamp 65,000 125 copper clamps for seismic resistance of structure is significant
Stone 48,400 –
since the collapse of the structure appears for ag = 1.2 g.
H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 127

Fig. 21. The enlarged central section between the two central columns of Prothyron for acceleration ag = 0.22 g: (a) dry stone structure; (b) dry stone structure with steel bolts
embedded between columns, capitals and the upper beams; (c) dry stone structure with steel bolts and copper clamps.

Fig. 22. The enlarged central section between the two central columns of Prothyron for acceleration ag = 0.45 g: (a) dry stone structure; (b) dry stone structure with steel bolts
embedded between columns, capitals and the upper beams; (c) dry stone structure with steel bolts and copper clamps.

Fig. 23. Collapse mechanism of dry stone structure with steel bolts in time: (a) t = 0.0 s; (b) t = 11.8 s; (c) t = 17.2 s; (d) t = 17.9 s; (e) t = 18.7 s; (f) t = 19.6 s.
128 H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129

Fig. 24. Dry stone structure with steel bolts and copper clamps in time: (a) t = 9.8 s; (b) t = 12.0 s; (c) t = 15.1 s; (d) t = 23.5 s; (e) t = 27.6 s; (f) t = 32.2 s.

1.0
displacement / m

0.5

0.0
0 5 10 15 20 25 30 35 40
-0.5

-1.0
t /s
(a) (b) (c)
Fig. 25. Response of dry stone structure with steel bolts and copper clamps in time for ag = 1.2 g: (a) normal stress in the copper clamp C1; (b) shear stress in the steel bolt B1;
(c) displacement of the top of the structure.

The purpose of this analysis was to show the possibility of the describe the behaviour of clamps and bolts which are often used
presented model in seismic analyses of historic dry stone masonry in strengthening dry stone masonry structures. These are:
structures. Relevant conclusions for the seismic resistance of this
structure can be carried out on the basis of an analysis for a set  an embedded model of steel clamps type I which are inserted on
of seven accelelograms compatible with the soil type [54]. the lateral surface of the structure;
 an embedded model of clamps type II which are inserted on the
8. Conclusions top side of stone blocks;
 cyclic behaviour, yielding and failure of the steel for clamps
This paper presents a new robust numerical model for analysis type I and II, as well as the influence of pulling out of the clamps
and prediction of the collapse of dry stone masonry structures type II from the stone block;
strengthened with steel clamps and bolts. The model is based on  an embedded model of steel bolts which are inserted on the top
a combined finite-discrete element method with the possibility side of stone blocks;
of the modelling of contact interaction, energy dissipation during  cyclic behaviour, stiffness degradation and failure of the bolt
impact, block deformability, fracture and fragmentation which and the influence of pulling out the bolt from the stone block.
are important especially in the analyses of structures under high
compressive loads. The developed model can be used for the estimation of the seis-
Several numerical algorithms were developed and implement- mic resistance of historical dry stone masonry structures rein-
ed in a combined finite-discrete element code to realistically forced with steel clamps and steel bolts, which is very important
H. Smoljanović et al. / Engineering Structures 90 (2015) 117–129 129

for the structures classified as cultural heritage. The model can also [25] Hart RD, Cundall PA, Lemos V. Formulation of a three-dimensional distinct
element model – Part II: mechanical calculations. Int J Rock Mech Min Sci
help to make the right decisions regarding the restoration of dry
1988;25(3):117–25.
stone masonry structures which have experienced deterioration [26] Barbosa BE. Discontinuous structural analysis. In: Proceedings of the 11th
over time. world conference on earthquake engineering. Elsevier; 1996. p. 830.
[27] Petrinic N. Aspects of discrete element modelling involving facet-to-facet
contact detection and interaction. PhD dissertation. UK: University of Wales;
Acknowledgements 1996.
[28] Shi GH, Goodman RE. Discontinuous deformation analysis – a new method for
The partial financial support provided by the Ministry of computing stress, strain and sliding of block systems. In: Key questions in rock
mechanics. Rotterdam: Balkema; 1988. p. 381–93.
Science, Education and Sports of the Republic of Croatia for the [29] Pearce CJ, Thavalingam A, Liao Z, Bićanić N. Computational aspects of the
Non-linear Dynamic Analysis of Three-dimensional Reinforced discontinuous deformation analysis framework for modelling concrete
Concrete Structures Project, Grant. No. 083-0831541-1532 is grate- fracture. Eng Fract Mech 2000;65:283–98.
[30] Munjiza A, Owen DRJ, Bicanic N. A combined finite-discrete element method in
fully acknowledged. transient dynamics of fracturing solids. Eng Comput 1995;12:145–74.
[31] Munjiza A. The combined finite-discrete element method. 1st ed. UK: John
References Wiley & Sons; 2004.
[32] Munjiza A, Knight EE, Rouiger E. Computational mechanics of discontinua. 1st
ed. UK: John Wiley & Sons; 2012.
[1] Parisi F, Augenti N. Earthquake damages to cultural heritage constructions and
[33] Živaljić N, Smoljanović H, Nikolić Ž. A combined finite-discrete element model
simplified assessment of artworks. Eng Failure Anal 2013;34:735–60.
for RC structures under dynamic loading. Eng Comput 2013;30(7):982–1010.
[2] Lourenço PB. Computational strategies for masonry structures. PhD
[34] Munjiza A, Andrews KRF, White JK. Combined single and smeared crack model
dissertation. Delft, The Netherlands: University of Technology; 1996.
in combined finite-discrete element method. Int J Numer Meth Eng
[3] Lourenço PB, Rots JG, Blaauwendraad J. Continuum model for masonry:
1999;44:41–57.
parameter estimation and validation. J Struct Eng ASCE 1998;1(6):642–52.
[35] Munjiza A, John NWM. Mesh size sensitivity of the combined FEM/DEM
[4] Berto L, Saetta A, Scotta R, Vitaliani R. An orthotropic damage model for
fracture and fragmentation algorithms. Eng Fract Mech 2002;69:281–95.
masonry structures. Int J Numer Methods Eng 2002;55:127–57.
[36] Munjiza A, Andrews KRF, White JK. Penalty function method for combined
[5] Radnić J, Harapin A, Smilović N, Grgić N, Glibić M. Static and dynamic analysis
- finite-discrete element system comprising large number of separate bodies.
of the old stone bridge in Mostar. Gradevinar 2012;64(8):655–65.
Int J Numer Meth Eng 2000;49:1377–96.
[6] Casolo S. Macroscopic modelling of structured materials: relationship between
[37] Munjiza A, Andrews KRF, White JK. NBS contact detection algorithm for bodies
orthotropic Cosserat continuum and rigid elements. Int J Solids Struct
of similar size. Int J Numer Meth Eng 1998;43:131–49.
2006;43(3–4):475–96.
[38] Xiang J, Munjiza A, Latham JP, Guises R. On the validation of DEM and FEM/
[7] Trovalusci P, Masiani R. Non-linear micropolar and classical continua for
DEM models in 2D and 3D. Eng Comput 2009;26:673–87.
anisotropic discontinuous materials. Int J Solids Struct 2003;50(5):1281–97.
[39] Smoljanović H, Živaljić N, Nikolić Ž. A combined finite-discrete element
[8] Radnić J, Harapin A, Matešan D, Trogrlić B, Smilović M, Grgić N, Baloević G.
analysis of dry stone masonry structures. Eng Struct 2013;52:89–100.
Numerical model for static and dynamic analysis of masonry structures.
- [40] Gotovac B. Ponovna izgradnja starog mosta. Ceste Mostovi 2004;50(7–
Gradevinar 2011;63(6):529–46.
9):23–33.
[9] Lofti HR, Shing PB. Interface model applied to fracture of masonry structures. J
[41] Senthivel R, Lourenço PB. Finite element modelling of deformation
Struct Eng ASCE 1994;120:63–80.
characteristics of historical stone masonry shear walls. Eng Struct
[10] Pegon P, Pinto AV. Seismic study of monumental structures – structural
2009;31:1930–43.
analysis, modelling and definition of experimental model. Ispra, Italy, Report:
[42] Brancherie D, Ibrahimbegovic A. Novel anisotropic continuum-discrete
EUR 16387 EN 1996, JRC.
damage model capable of representing localized failure of massive
[11] Lourenço PB, Rots JG. A multi-surface interface model for the analysis of
structures. Part I: theoretical formulation and numerical implementation. Int
masonry structures. J Eng Mech ASCE 1997;123:660–8.
J Eng Comput 2009;26:100–27.
[12] Carol I, López C, Roa O. Micromechanical analysis of quasi-brittle materials
[43] Ibrahimbegovic A, Melnyk S. Embedded discontinuity finite element method
using fracture-based interface elements. Int J Numer Methods Eng
for modelling of localized failure in heterogeneous materials with structured
2001;52(2):193–215.
mesh: an alternative to extended finite element method. Comput Mech
[13] Kozar I, Ibrahimbegovic A. Finite element formulation of the finite rotation
2007;40:149–55.
solid element. Finite Elem Anal Des 1995;20(2):101–26.
[44] Ibrahimbegovic A, Brancherie D. Combined hardening and softening
[14] Ibrahimbegovic A, Kozar I. Non-linear Wilson’s brick element for finite elastic
constitutive model for plasticity: precursor to shear slip line failure. Comput
deformations of the three-dimensional solids. Commun Numer Methods Eng
Mech 2003;31:88–100.
1995;11(8):655–64.
[45] Hordijk DA. Tensile and tensile fatigue behaviour of concrete – experiments,
[15] Burnett S, Gilbert M, Molyneaux T, Beattie G, Hobbs B. The performance of
modelling and analyses. Heron 1992;37(1):3–79.
unreinforced masonry walls subjected to low-velocity impacts: finite element
[46] Oliveira DV, Lourenço PB. Implementation and validation of constitutive
analysis. Int J Impact Eng 2007;34:1433–50.
model for cyclic behaviour of interface elements. Comput Struct
[16] Cundall PA. A computer model for simulating progressive large scale
2004;82:1451–61.
movements in blocky rock systems. In: Proceedings of the symposium on
[47] Hillerborg A, Modeer M, Petersson PE. Analysis of crack formation and crack
rock fracture (ISRM). Nancy, France; 1971. p. 1–8.
growth in concrete by means of fracture mechanics and finite elements. Cem
[17] Cundall PA, Hart RD. Numerical modelling of discontinua. Eng Comput
Concr Res 1976;6:773–82.
1992;9:101–13.
[48] Kato B. Mechanical properties of steel under load cycles idealizing seismic
[18] Pagnoni T. Seismic analysis of masonry and block structures with the discrete
action. Structural concrete under seismic actions. AICAP-CEB Symp
element method. In: Proc 10th European conference on earthquake
1979;132:7–27.
engineering; 1994. p. 1674–94.
[49] Carol I, Prat P, López CM. Normal/shear cracking model: application to discrete
[19] Lemos JV. Discrete element modelling of the seismic behaviour of stone
crack analysis. J Eng Mech 1997;123(8):765–73.
masonry arches. In: Computer methods in structural masonry. London; 1998.
[50] Comite European de Normalization (CEN). Eurocode 2: design of concrete
p. 220–7.
structures, EN 1992-1-1, Brussels; 2004.
[20] Baggio C, Trovalusci P. Stone assemblies under in-plane actions-comparison
[51] Crnkovic B, Saric LJ. Building with natural stone. 2nd ed. Croatia: IGH; 2003 [in
between nonlinear discrete approaches. Comput Methods Struct Masonry
Croatian].
1995;3:184–93.
[52] Mahabadi OK, Grasselli G, Munjiza A. Y-GUI: a graphical user interface and
[21] Baggio C, Trovalusci P. Collapse behaviour of three-dimensional brick-block
pre-processor for the combined finite-discrete element code, Y2D,
systems using non-linear programming. Struct Eng Mech 2000;10:181–95.
incorporating material heterogeneity. Comput Geosci 2010;36(2):241–52.
[22] Nappi A, Tin-Loi F. A numerical model for masonry implemented in the
[53] Rak M, Duvnjak I, Krolo J, Meštrović M. Report of measurement forces in
framework of a discrete formulation. Struct Eng Mech 2001;11(2):171–84.
copper clamps on Protiron – Peristil Diocletian’s Palace in Split. Report 180-12/
[23] Gilbert M, Casapulla C, Ahmed HM. Limit analysis of masonry block structures
2012. Zagreb: Faculty of Civil Engineering, University of Zagreb; 2012 [in
with non-associative frictional joints using linear programming. Comput
Croatian]).
Struct 2006;84(13):873–87.
[54] Comite European de Normalization (CEN). Eurocode 8: design of structures for
[24] Cundall PA. Formulation of a three-dimensional distinct element model – Part
earthquake resistance, EN 2004-1-1, Brussels; 2004.
I: a scheme to detect and represent contacts in a system composed of many
polyhedral blocks. Int J Rock Mech Min Sci 1988;25(3):107–16.