Professional Documents
Culture Documents
www.elsevier.com/locate/engstruct
Received 22 July 2002; received in revised form 15 May 2003; accepted 16 May 2003
Abstract
In the present paper, the methodology, suggested in FEMA 273 guidelines for the evaluation of the seismic capacity of existing
structures, is applied for the case of unreinforced masonry-type plane frames. Results from comparative pushover analyses of
masonry plane frames are presented. Calculations are carried out by the SAP2000 Nonlinear and CAST3M programs. For the
analysis by the SAP2000 Nonlinear program, linear elements with appropriate plastic hinges are used for the modelling. For the
analysis by the CAST3M program, two different types of models are examined, using either homogenous plane elements or discrete
brick and joint elements. From the analytical results, useful conclusions are drawn on the strength and inelastic behaviour of masonry
frames under three different distributions of the seismic lateral loads.
2003 Elsevier Ltd. All rights reserved.
Keywords: Unreinforced masonry walls; Profile of lateral loads; Pushover analysis; Frame elements; Homogenous elements; Discrete brick elements
0141-0296/03/$ - see front matter 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0141-0296(03)00118-4
1516 T. Salonikios et al. / Engineering Structures 25 (2003) 1515–1523
앫 Diagonal cracking shear failure involving the bed and bay (1B) and a seven-bay (7B) model (Fig. 1). The
head joints and/or the bricks depending on their rela- choice of the models is dictated from the fact that URM
tive strength. buildings in earthquake prone Southern Europe countries
앫 Shear sliding failure along a bed joint at one end of such as Greece are usually low-rise (in most cases one
the structural element (only for piers). or two storeys high), but may be smaller or larger
depending on the use (private or collective). In addition
This is fully coherent with what have been observed to the self-weight of the masonry, extra masses are con-
during several damage surveys of masonry buildings sidered at the floor levels. For the 1B frame, a uniformly
subjected to earthquake excitations. Two main types of distributed mass of 6 tons/m is assumed for the first
cracks have been observed: either horizontal at the ends floor, and of 4 tons/m for the second floor. The corre-
of the pier (or vertical at the ends of the spandrels), or sponding values for the 7B frame are assumed to be 3
diagonal in both element types. and 2 tons/m, respectively, so as to obtain the same mass
When an URM building is subjected to a seismic base (and associated axial load) on the piers of the 1B frame
excitation, both bending and shear resisting mechanisms and on the internal piers of the 7B frame (the mass on
may be activated and failure will first appear at the the two external piers of the 7B is smaller).
location with the smallest strength. To reliably describe For the analysis, three types of modelling are investi-
the inelastic behaviour of an URM building, a model gated. In the first, the linear frame elements of the
should allow both flexural and shear mechanisms to be SAP2000 Nonlinear program are used, while in the two
activated. others, a continuous and discrete modelling is applied
Methods of displacement-based or pushover inelastic using suitable plane and joint elements of the
analysis for the seismic evaluation of URM buildings CAST3M program.
possess the advantage that loads are applied in a static The basic mechanical characteristics of the masonry
way and hence the results can be easily verified. Also, material are not a parameter for this study and are there-
structural elements are checked using deformation and fore the same for all computations (thickness of the
displacement values, which are the actual result of the walls t = 0.60 m, volumetric mass r = 2 t / m3, Young
earthquake excitation on the building [4]. Irrespective of modulus E = 1650 MPa, Poisson ratio n = 0.2, tension
the design methodology to be used (displacement-based strength f t = 0.1 MPa, compression strength f c = 3.0
or simplified pushover analysis), there is always a need MPa). Additional parameters required for a specific
for inelastic modelling of the building and for appropri- modelling option have been chosen on the basis of exist-
ate distribution of lateral loads along the height. Among ing guidelines [5], engineering judgment and/or experi-
the models suggested in the literature for the simulation ence acquired from previous use of the model.
of URM walls, most prominent are the discrete brick and
joint [7], the continuum [8] and the frame element [9] 2.1. Linear (S) model (SAP2000)
models. A comparative study using all three proposed
models is presented in this paper. In the computations, The structures are considered as an assemblage of lin-
three different lateral load distributions are also assumed, ear elements (beams) available in the SAP2000 Nonlin-
since in URM buildings the appropriate lateral loads pro- ear program [9]. These beams have an elastic–plastic
file is not always obvious. This problem arises from the behaviour based on the plastic hinge concept. In this
fact that, in such buildings, an important fraction of the case, a necessary prerequisite is the definition of the
total mass is distributed along the wall height and in possible locations and types of the plastic hinges that
many cases there are flexible wooden floors that do not might develop along the span of each element. These
provide sufficient diaphragm constrain at their level. plastic hinges should be able to accurately describe the
possible failure mechanisms observed in the actual struc-
tures (flexural or shear). Here, a suitable moment–
2. Modelling assumptions rotation (M–⌰) plastic hinge is provided at both ends
of each element (flexural mechanism), while a shear–
The resistance of URM buildings to seismic exci- displacement (V–d) plastic hinge is located at the mid
tations relies on the in-plane and out-of-plane resistance span of the element (shear mechanism), as shown in
of several load-bearing masonry walls arranged in Fig. 2.
orthogonal planes. Out-of-plane failure may be pre- The constitutive laws of the plastic hinges are defined
vented by an appropriate connection between perpen- according to the guidelines of FEMA 273, Table 7.2 [5]:
dicular walls. In-plane failure is governed by the geo- the initial linear elastic branch is followed by a horizon-
metric arrangement of the openings that defines a series tal one after the first cracking. At a higher inelastic
of piers and spandrels (plane frame) expected to sustain deformation, a sudden drop of the strength to 60% of
the gravity and lateral loads. In the present work, two its maximum value is assumed, and then kept constant
different, two storey frames are examined, namely a one- afterwards. For the case of structural components with
T. Salonikios et al. / Engineering Structures 25 (2003) 1515–1523 1517
Fig. 4. Continuous (Mazars) model: (a) meshes and (b) uniaxial stress–strain curves of the material in tension and compression.
cohesion 0.09 MPa. The tension/compression strengths such a force-controlled loading makes inaccessible the
of the masonry (0.1 and 3.0 MPa) have been affected to possible softening branch of the response. In the com-
the tension/compression cut-offs of the interface. The parative analyses, three different lateral load distri-
post-peak behaviour was again chosen very brittle in ten- butions were used:
sion and more ductile in compression and shear (Fig. 5).
Furthermore, in order to maintain a reasonable size of 앫 Load case ACC: The lateral force applied at each
finite element problem, especially for the 7B frame, the node is proportional to the mass tributary to that node.
dimensions of the bricks have been overvalued (0.50 × When all nodes have the same mass, this load case
0.25 m instead of 0.20 × 0.10 m). results in a uniform distribution of the lateral loads.
앫 Load case LOAD: The lateral force applied at each
2.4. Distribution of lateral loading node is proportional to the product of the height of
the node times the mass tributary to that node. When
The loading is imposed on the structures in a two-step all nodes have the same mass, this load case results in
sequence. In the first step, the vertical (permanent) loads an inverse triangular distribution of the lateral loads.
are applied and, in the second step, the lateral loads are 앫 Load case MODE: The lateral force applied at each
imposed. The values of the latter are then increased mon- node is proportional to the product of the displace-
otonically step by step [13]. The analysis is carried out ment of the node in the first mode shape times the
until a prescribed displacement value at a prescribed mass tributary to that node. This load case results in
(control) node is achieved or until any further increment lateral loads proportional to the inertial forces applied
of the lateral load becomes impossible. In particular, to the structure when vibrating freely according to its
T. Salonikios et al. / Engineering Structures 25 (2003) 1515–1523 1519
Fig. 5. Discrete (elastic bricks, nonlinear joints) model: (a) meshes and (b) failure criterion of the joints with the corresponding stress–displacement
curves in tension, shear and compression.
first mode shape. When the first mode shape is shear Each of the 18 analysis cases described above leads
or even bending dominated as in the case of the build- to the pushover curve, i.e. the base shear versus displace-
ings studied here, this load case results very close to ment at the control point (middle point at the top of the
the previous one (i.e. LOAD) because the lateral building) curve, as well as to the failure mechanism cor-
forces have a fairly small vertical component even in responding to the maximum shear achieved (peak or
a bending mode. plateau). As mentioned earlier, the probable softening
branch could not be obtained because the loading was
force-controlled.
3. Comparative presentation of the numerical The base shear–displacement curves obtained by the
results simplified (S) analyses are presented in Fig. 6, for each
frame and each load case considered. As can be easily
A preliminary analysis (computation of the first mode seen, the lateral load distribution does not seem to gre-
shapes and frequencies) was necessary in order to define atly affect the pushover curves of the S models (same
the distribution of the lateral forces for the MODE load ultimate strength). The only difference seems to be a
case. This was also an opportunity to check the possible higher stiffness of the frames under the ACC load case,
differences already existing between the different model- which is clearly due to the distribution of the lateral
ling strategies in the linear elastic domain. All three loads: independently of the load case considered, the
modelling approaches (S, C and D) have been found to piers of the first storey are subjected to the same shear
yield the same fundamental frequency and very similar (i.e. the total base shear V), whereas the shear applied
mode shapes. to the piers of the second storey is much lower in the
1520 T. Salonikios et al. / Engineering Structures 25 (2003) 1515–1523
Fig. 6. Pushover curves for the linear model: (a) 1B frame and (b)
7B frame.
Fig. 8. (a) Pushover curves and (b) failure patterns for the 1B con-
tinuous model.
4. Conclusions