Professional Documents
Culture Documents
37-53, 1998
© 1997 Elsevier Science Ltd
All rights reserved. Printed in Great Britain
PII: S0141-0296(97)00042-4 0141-0296/98 $19.00 + 0.00
ELSEVIER
Paola Ronca
Politecnico di Milano, Department of Structural Engineering, 20133 Milano, Italy
(Received June 1995; revised version accepted October 1996)
37
38 Numerical analysis of old masonry buildings: F. Genna et al.
I
However, models which take into account the actual frac- analyzed wall
turing behaviour, with instantaneous energy dissipation at
a nonzero load level, at the present state of development
usually fail in simulating the effect of confinement in terms
of ductility and, as already said, each model has usually
been developed and tested for particular applications and
in conjunction with special numerical procedures, thus
making it difficult to foresee their applications in a gen-
eral situation.
In this work some of these models are applied, with pos-
sible suitable modifications, within a homogeneous numeri-
cal environment, sharing the same basic rate formulation
and the same basic integration procedures. Thus, a first aim
has been that of unifying the numerical treatment of differ-
ent constitutive models. A second goal has been the com-
parison of the numerical predictions with the observed
actual state of the structure. Such comparison may be useful
to assess the sensitivity and appropriateness of each model,
and relevant data, when applied to a problem whose size
allows only a global discretization, but in which every
structural detail may play an essential role on the global Figure2 Three-dimensional discretization of the San Faus-
tino wall
behaviour.
(i) by comparison with the three-dimensional results, a ary condition thereafter adopted. Finally, with the nonlinear
plane model was sufficiently accurate as far as the stress analyses in view, the three structural elements (wall, arches
and strain state in the wall was concerned and (ii) the high- and diaphragms) have been physically separated in the
est significant stress values could be found in the zones numerical model, and suitable finite elements have been
close to the supporting diaphragms where the arches have introduced at their interfaces, to allow the simulation of the
the longest span. Also, the left end side of the wall seems behaviour of the joints between wall, arches and dia-
to be the most severely loaded. In fact, the left end arch phragms, made by either simple support with interposed
has one of the longest spans and, at the same time, it has mortar (wall on arches and wall on diaphragms) or by
no restraint whatsoever to its horizontal displacement at the clamping, brick by brick (lateral joints between wall and
left abutment. This consideration, together with the necess- diaphragms). The numerical model whose results in the
ity of further refining the mesh within the limits of the nonlinear range will be discussed in Section 5 is, therefore,
available hardware, actually motivated the subsequent defined as follows. The geometry is shown in Figure 4. The
decision of analyzing, in some detail, the behaviour only wall is discretized by means of 845 4-node plane stress
of the part of the wall resting above the two arches at the elements, whose stiffness matrix is integrated using a 2 x 2
left end extremity of the 'friars' cells corridor'. Gauss grid. The stresses are computed and the constitutive
Two nonlinear analyses of the whole plane model of law is enforced only at the centroid of each element. The
Figure 3 have also been performed, the results of which thickness of these elements is 30 cm. The two arches are
are reported elsewhere 8. These analyses used the elastic- discretized by means of 204 4-node elements as used for
plastic 'no tension' constitutive law and a cracking model, the wall; their thickness is 60 cm. The stress-strain law for
whose characteristics will be explained in the next section. these elements can be either linear elastic or elastic-plastic
Their results indicated that (i) the full plane model required or cracking, depending on the constitutive model adopted.
far too much computing power than available to be properly The three diaphragms are discretized by means of 127 4-
analyzed in the nonlinear range and (ii)the 'no tension' node elements of the same type as already defined. Their
model is utterly unrealistic for such analysis. This con- thickness is 3.81 m for the bottom part and 1.40 m for the
clusion will be strengthened by the results discussed in top portion, from the arches abutments to the top of the
detail in the next sections. However, both the linear elastic diaphragms. These elements are always considered as lineal"
and the nonlinear results for the full model of the wall con- elastic, since their dominant stress components are the out-
firmed the opportunity of pursuing a detailed study of the of-plane ones, not taken into account by this model for the
stress state in the left end part of the wall. With this goal, sake of memory saving. This is one of the strongest limi-
a finer discretization of the first two arches at the left end tations of the analyses reported herein, since the cracking
of the structure and of the above-placed wall has been of the diaphragms is expected to play an important role on
made. Such a model is shown in Figure 4. The thickness the overall behaviour of the structure. However, since we
of the three supporting diaphragms is of 3.81 m for the have used a plane model of both wall and diaphragms, with
lower part of the diaphragms, i.e. from the boundaries up the purpose of studying the in-plane behaviour of the wall
to the arch abutments. It reduces to 1.40 m for the upper and the arches only, we have been forced to neglect any
parts, going from the arch abutments to the corridor level, nonlinearity in the diaphragm in-plane behaviour, which
at the lower end of the wall (i.e. the level of the extrados would have been meaningless. Therefore, the diaphragms
at the keystone of the arches). have been considered essentially as elastic supports. Only
The problem created by the necessity of defining bound- a three-dimensional model could take into full account the
ary conditions for the right end vertical edge of the mesh cracking of the diaphragms.
was tackled by means of a comparison with the results Finally, the joints between the wall, arches and dia-
given by the full plane model. The boundary condition for phragms are modeled by means of 89 interface 2-node
the mesh of Figure 4 which gives the minimum error, in a elements, with two active stress components: normal and
least-square sense, is the one in which only the horizontal shear stress. The geometry of these elements is defined by
displacements are set to zero. This is, therefore, the bound- the position of their nodes, whose relative displacement in
the direction connecting the two nodes governs the normal
stress behaviour, and whose relative displacement in the
orthogonal direction governs the shear stress behaviour.
Their initial distance has been set equal to 1 cm. The inter-
face element behaviour is formulated, within the finite
element code, in terms of stresses and strains (as opposite
to common 'spring' elements, described in terms of forces
and displacements). The stress-strain law of the interfaces
can be both linear elastic and nonlinear, as explained in the
following sections.
The model, therefore, uses a total of 1265 finite elements
for 2734 unconstrained degrees-of-freedom. The bandwidth
of the stiffness matrix, after optimization, is 118. The
adopted material data will be described in better detail, case
by case, in Section 5. The boundary conditions at the right
edge of the wall imply no horizontal displacement; the
bases of the three diaphragms are fixed.
[] Finally, the loading conditions are as follows.
Figure4 Adopted finite element mesh of the left end portion
of the San Faustino wall (1) Self weight and accidental load due to snow on the
40 Numerical analysis of old masonry buildings: F. Genna et al.
central top roof, corresponding to 10800 Newton per This nonlinear model is piecewiselinear, thus enabling
linear meter of the wall, placed on the top edge of the its direct formulation in terms of the Mathematical Pro-
mesh. When performing the collapse analysis the load gramming framework introduced by Maier ~. This model
due to the snow has been considered as a permanent is elastic-plastic and, therefore, it does not allow the simul-
load, and only the collapse load factor of the accidental ation of fracture. However, since it is applied to interface
load on the floor slab has been computed. Only a elements, whose widths are very small, it can be thought
shakedown analysis could have tackled the problem of of as a fair approximation of a cracking model, in that the
simultaneously variable snow load and accidental load developed plastic strains are actually localized into zones
on the floor slab; this analysis has not been performed. of small width, as far as the interface elements are con-
(2) Self weight of the roof slab in the central part, corre- cerned. A different situation occurs when dealing with con-
sponding to 3200 Newton per linear meter of the wall tinuum elements.
placed on a line 16.00 m above the basis of the dia- The constitutive laws utilized to simulate the nonlinear
phragms. behaviour of the wall, discretized by means of continuum
(3) Self weight and snow load (assumed as permanent in elements, are briefly described in the sequel.
the collapse analysis) on the side roof at the east side,
corresponding to 6500 Newton per linear meter of the 3.1. The Galileo-Rankine elastic-plastic model
wall placed on a line 15.00 m above the basis of the According to this constitutive law the principal stresses
diaphragms. must not exceed a limiting value in tension, o-,, and a limit-
(4) Self weight of the roof slab, cell side at the east side ing value in compression, o-~. The G a l i l e o - R a n k i n e con-
of the wall, corresponding to 2600 Newton per linear dition, in a plane stress case referred to a cartesian ortho-
meter of the wall placed on a line 13.75 m above the gonal frame x - y , can be written as follows
basis of the diaphragms.
(5) Self weight of the floor slab of the cells cocridor, corre- ~r~ = ~ , . - ~r,~r,. + ~rt(~r, + o~,) - ~ --< 0 (2)
sponding to 7500 Newton per linear meter of the wall
placed on a line 9.00 m above the basis of the dia- o3, = ~ , . - o:,cr,.- o-~.(o-,+ o~,.) - ~ --< 0 (3)
phragms.
(6) Accidental load on the same floor slab, considered as where the yield stress in compression, (r~, has been taken in
12500 Newton per linear meter of the wall placed on absolute value. Equations (2) and (3) represent two convex,
a line 9.00 m above the basis of the diaphragms. smooth surfaces in the stress space. Their intersection, how-
(7) Self weight of the masonry, estimated as 20000 New- ever, is a line in which the gradient is not defined. The flow
ton per cubic meter. rule is associated and, in the intersection between the ten-
sile and the compressive surfaces, it follows Koiter's gen-
No actions other than the mechanical ones listed above eralization.
have been considered. Further work might as well be
devoted to the investigation of the effects of thermal load- 3.2. The no tension elastic-plastic model
ing (possibly significant on such a large structure) and of From the analytical viewpoint this is a straightforward spe-
prescribed settlements of the diaphragm bases. The model cialization of the preceding model to the case o-~ ---* zc and
described here has been analyzed by means of the computer o-,---* 0. Thus, equations (2) and (3) reduce to the single
code STRUPL-2 9, which can perform nonlinear analyses of constraint
plane problems for several constitutive models.
9
~,- ~r,o~,,-< 0 (4)
3. The adopted constitutive models This equation defines a cone in the cartesian stress space,
The interface elements have been described either as linear with a singularity at the origin of the space also unsafe with
elastic or in terms of a M o h r - C o u l o m b friction law with respect to the yield condition; the adopted flow-rule is, as
cohesion, which reads always in this work, associated. In practice, the no tension
model has often been claimed to be a useful tool for the
analysis of masonry structures. The argument is that the
_+T-o-tan0-k <-0 (1)
presence of fractures causes the inability of transmitting
tensile normal stresses, therefore causing a global behav-
in which ~- is the shear stress, cr is the normal stress, & is iour similar to that of a linear elastic material unable to
the friction angle and k is the cohesion. The direction of stand tensile principal stresses. A recent paper ~2, devoted
incremental plastic flow here is orthogonal to the active to the numerical integration of this model, shows clearly,
yield plane, and follows Koiter's rule ~° if the stress point by means of numerical examples, that the no tension consti-
reaches the corner generated by the two yield planes at the tutive law is unable to represent realistically the nonlinear
value 7 = 0. Although this description is rather poor in behaviour of brittle structures. This conclusion will be
terms of kinematics, owing to the excessive dilatancy intro- strengthened by the results obtained in the present paper.
duced, it leads to a much simpler numerical formulation
than a non-associated one and, for the loading levels con- 3.3. The Drucker-Prager elastic-plastic constitutive
sidered, it should not lead to significant errors on stresses law
and displacements. Only the collapse analysis is strongly
The yield condition, in the elastic-perfectly plastic three-
affected by the assumption of normality, in that it will only
dimensional case, is defined by the following equation
provide an upper bound to the actual collapse load factor,
together with a possibly poor description of the collapse , t
Parameters c and a are related to the friction angle 4, and in the direction associated to the maximum principal stress)
the cohesion k of the considered material. The symbol p and the crack opening w. The model is governed by the
identifies the hydrostatic stress and J; the second invariant fracture energy in tension of the material, %, assumed to
of the deviatoric part of the stress tensor. Several relation- be a constant material parameter, and is defined by the fol-
ships can be adopted to relate the cohesion k and the fric- lowing relationship
tion angle 4, of a geomaterial to the strength parameters a
and c of D r u c k e r - P r a g e r ' s model, depending, for instance, 2%
o-,, = o5 ~ w (7)
on the desired match with a Coulomb type criterion. In the
calculations reported herein we have adopted the follow-
ing relationships In the finite element formulation the crack opening w is
considered as 'smeared' within a finite zone of the element
2 sin 4, 6k cos & and it is, therefore, related to the strain by means of some
r-
• C --
'characteristic' length ~4. The crack is fully opened only
\/3(3 - sin4,) \;3(3 - sin4,) when the normal stress has reduced to zero; the behaviour
C C of the cracked point from then on is holonomic.
O-t= • O-c=
(1/\'3 + a) (1/,]3 - cQ 3.5. A modified De Felice elastic-plastic model
A recent work by De Felice ~ describes a constitutive law
where o-, and O-~ are the yield stresses in uniaxial tension applicable to masonry panels made of rigid blocks and lay-
and compression, respectively• In the plane stress case, ers of mortar obeying a friction law of the Coulomb type
D r u c k e r - P r a g e r ' s constitutive law reduces to a single, con- with cohesion. The De Felice model takes into account, as
tinuous yield surface, whose equation reads parameters governing the continuum, viewed as the result
of a homogeneization process, the cohesion k and friction
angle 4' of the mortar, as well as the ratio m between the
height and width of the single block. In a reference system
x - y aligned with the direction of the layers the equations
cKol, + O-,3 - c --< 0 (6) describing this model are as follows
The values of the cohesion and the friction angle can be %. + oi,, tan 4) - k --< 0 (8)
chosen so as to obtain a good match with uniaxial strength
in tension and compression; in this case, however, the pre- - % + o5, tan 4 , - k --< 0 (9)
dicted behaviour for most stress states of biaxial com- 2mo-, + (2m tan 4' + 1 )r,, +
pression will be elastic, as shown in Figure 5, which refers
to the material data used in analyses 6 and 7 of Section 5.
kP= NA (16)
X ~ 0 (173
(2) rate equilibrium where Ke~ is the assembled elastic stiffness matrix. Matrix
A is defined as follows
f 13)
VB~& dV = f¢ A = -N~(DBK~I I B~D - D ) N (23)
in which o" is the vector of total stresses at the inte- Equations (21) express a so-called Linear Complementarity
gration points of the finite element mesh. The inte- Problem (LCP). Its solution can be obtained on the basis
gration symbolically represents also the assembly oper- of an algorithm, described for instance in Franchi and
ation, which allows to relate the rate of 'internal forces', Genna t6, which takes advantage of the physical meaning of
at the left-hand side of (13), to the rate of the prescribed all the terms involved and which essentially amounts to the
external actions, defined by vector F; inversion of matrix A, plus a few other steps required only
when the inversion of A furnishes a solution A which does
(3) elastic-pe .rfectly plastic constitutive law not comply with the constraint (21c). In the implemented
algorithm matrix A is constructed one row and column at
= D£~ = D(t~ - t~P) (14) a time, following the integration scheme and, therefore, its
inversion, at every integration step, is quite inexpensive.
which represents the elastic part of the constitutive law, Also, the computer storage is kept limited to the minimum
governed by the matrix of the elastic moduli, D; necessary, since the dimension of the LCP is only that cot-
Numerical analysis of old masonry buildings: F. Genna et al. 43
plastic interfaces with the Coulomb normal stress- (3) there are stress concentrations in the connection zones
shear stress constitutive law, with zero cohesion. Its between wall and vertical diaphragms. Such concen-
execution time was 6000 times that of the elastic analy- trations suggest detachment, which can be taken into
sis; account by the nonlinear models;
(7) elastic-plastic analysis with the De Felice constitutive (4) there are strong compressive stresses in the zones
model for the wall and arches, and elastic-plastic inter- where the wall is supported by the diaphragms, which
faces with the Coulomb normal stress-shear stress would be stronger if the detachment mentioned before
constitutive law, with zero cohesion. Its execution time takes place. Both the wall and the diaphragms (whose
was only 190 times that of the elastic analysis; the actual stress state is not computed in this analysis)
computing time reduction is entirely due to the piece- should be carefully examined in this respect. Also, the
wiselinearity of the yield condition for the whole left end diaphragm must carry the entire thrust of the
model; left arch, thus undergoing bending out of its plane.
(8) collapse analysis with the De Felice constitutive model Such bending, indicated by the displacement pattern
for the wall and arches, and elastic-plastic interfaces provided by all the analyses (see, for instance,
with the Coulomb normal stress-shear stress constitut- Figure 14), is very dangerous and suggests the need of
ive law, with zero cohesion. Its execution time was exploring the global stability of this first diaphragm.
12000 times that of the elastic analysis.
The overall picture is rather uncertain since, already judg-
All the results shown in this section, figures and tables ing from the linear elastic results, one might conclude that
included, will be presented in Newton and millimeters the wall cannot withstand the service loads without crack-
units. ing, even if localized. Therefore, further nonlinear analyses
appear mandatory.
5.1. Analysis number 1: elastic analysis
The material data used for describing the elastic properties 5.2. Analyses number 2, 3, 4 and 5: nonlinear analyses
of the masonry are as follows: Young's modulus E = 5000 with linear elastic interface elements
MPa and Poisson's ratio v = 0.20. The material has been
Here we analyze results given by different constitutive
considered as homogeneous and isotropic; the numerical
models with comparable values of strength parameters,
values have been chosen on the basis of compression tests
whose choice has been guided both by the few suggestions
performed on wall specimens. The results of the elastic
found in literature about old masonry 23 and by the linear
analysis have been used only as a guideline for judging the
elastic results. Both these guidelines indicate that quite low
goodness of the discretization and for choosing the strength
values of resistance need to be used, for characterizing the
parameters needed by the nonlinear analyses. The peak ten-
masonry as a homogeneous material, in order to correctly
sile stress is here of about 1 MPa, but the meaningful
predict its nonlinear behaviour. Therefore, the adopted
maximum values are about 0.4 MPa. In the most part of
material data are as follows:
the wall the tensile stresses are within the range
0.03 + 0.15 MPa. The compressive values reach peaks of
about - 2 MPa. Their meaningful maximum values are of • for the Galileo-Rankine model: uniaxial strength in ten-
-1.3 MPa and the most part of the wall is subjected to com- sion ~rt=0.3 MPa; uniaxial strength in compression
pressive principal stresses whose absolute value is lower ~r~ = 1.0 MPa;
than 0.5 MPa. The peak values of all the stress, strain and • for the Drucker-Prager model: cohesion k = 0.2 MPa;
displacement components, for all the analyses herein friction angle 4~ = 35 °. These values correspond, accord-
reported, are summarized in Table 1 and visualized, for ing to the Drucker-Prager criterion, to uniaxial strength
ease of comparison, in Figure 10. The elastic results pro- parameters o-t = 0.28 MPa and o-~ = 0.77 MPa, similar to
vide relatively small stress values, if one neglects the peak those used for the Galileo-Rankine condition;
values at the corners of the openings, site of localized • for the cracking model: uniaxial strength in tension
cracking in reality. Despite such a low stress state, the wall ~rt = 0.3 MPa; fracture energy % = 0.02 N/mm. This last
exhibits actual cracking, whose typical pattern in one por- value has just been guessed, in the absence of other evi-
tion of the wall above a long-span arch is depicted in dence, as about 1/5 of that suggested for plain concrete.
Figure 11, taken from Ronca and Castiglioni 7. Therefore, The choice of a low value for the fracture energy,
the strength parameters required by global models for a together with the large size of the problem, should mag-
nonlinear analysis need to be quite low, if one wants to nify possible local instability phenomena due to sudden
detect 'yielding'. Also, the elastic analysis pinpoints the energy release when cracking occurs.
weakness of the left portion of the structure, supported by
an arch lacking a horizontal thrust at its left abutment. Figure 12 shows the principal plastic strains for the four
For the purpose of this work, however, the elastic analy- different constitutive models adopted. It is immediately
sis gives the following indications: apparent that three of the four plots illustrated by Figure 12
are similar to each other, whereas the fourth one
(1) the extension of the zone undergoing tension suggests (Figure 12c), referring to the results of the no tension
the possibility of cracking, even if, in a global sense, model, corresponds to a completely different situation, in
the wall seems subjected to a stress state which can which the plastic zone practically encompasses the whole
remain elastic; structure. The other models indicate only a small set of
(2) the tensile stress peaks at the corners of the openings plastic elements. The largest plastic strains, for the no ten-
will correspond, in the nonlinear range, to sure sion model, seem to develop in a vertical band close to the
localized 'yielding' (cracking), which should have no left edge of the wall, as well as into the whole opening zone
influence on the global behaviour; and into the totality of the two arches. The displacements
46 Numerical analysis of old masonry buildings: F. Genna et al.
.~_.o_
X XO X X XO X X XO X X XO X
tu .__~" 0 0 ~
• . ~ ":,~ . "7 "7 '-:,~ . '7. . .
o ~ O ~ o O ~ OoOoOoOd OoOoOoOo~r-.
I I I I I I I I I I I I
9 999 99 ~99 99
t~
X X x X x X x X x X x X x X x
OoOoOoO~ o OoOoOo O d O o O 6 O o ~
I I I I I I I I I I I
9
0 0 o ~9O o 0' o °~ ~ o O ~ o
~ X~ ~ ~ X X X~ X
X X X~ X X x X X~ X
,,~'1=
~_~._
OoO~ooo~ o O o O o O o O o O o O o O o
I I I i l l I [ I I I I I
v ~ o~E~oOOo~
X X XxX X x X x X x X x
0
• ~ 8 ~
0 ooo~OoO~ d o odd o d o d o d o d o ~
(,.) I I I I I
c
g
c'~) X X XxX x X x X x X x X X
~ 0 ~ 0 ~
• ~ . ~ ~ = ~~ .
~ .
~
t- °d°d ~ o o OoOo O o O o O o O o ~
O I I I I I I I I I I
r-
oo0ooooo oooooo0 ,
•o X X X X X X X X X X X X X X X X
5 ,~m-~.=_ o~o~ooo~
I I I I
o 9 o 9 o I ; o o ,. ooooo,
I I
oo
I I
N
.~_ ~ 0 ~ o O~0 ~ ~ ~' vo O o ~~ o' ~ '~
I o ¢0
X x X x X x X x X x X x X x X X
. ~ o ~ o
OoO~OoO~ OoOoOoOo o d o d o o O o ~
t49n'-~ u I I I I I I I I i I I I
'' o~oo0
o X x X x X x X X
~ ~ ,
ffl
~ o o ~ o o ~ ddOdOdOdO~
w to I I I I
> O.
0,.
E E
13_
~.~.~x oo
• . o~.E x ¢--
0.0o
~. c -IZ
~o EEaa2_
Numerical analysis of old masonry buildings: F. Genna et al. 47
[] []
(a) (b)
[!iii!! !!-i;iiiii [] []
_1
(c) (d)
Figure 12 Principal plastic strain directions and magnitudes (with different scales) as computed in analyses 2 (Galileo-Rankine,
Figure 12a), 3 (Drucker-Prager, Figure 12b), 4 (no tension, Figure 12c) and 5 (tension cracking, Figure 12d)
shear fracturing behaviour should be considered and relatively high shear stresses in the presence of compressive
the possibility of pre-cracking ductility should be normal stress. The wall itself has been described by means
taken into account. All of these topics are the subject of the Drucker-Prager model. Following the suggestions
of work in progress. given by the previous results, the strength parameters have
been slightly modified with respect to the first set of analy-
5.3. Analyses number 6 and 7: nonlinear analyses with ses and are: cohesion k = 0 . 2 3 MPa and friction angle
nonlinear inte.rface elements ~b=56 ° . Such values correspond, according to the
These analyses try and take into account the suggestions Drucker-Prager model, to uniaxial resistances
given by the previous results and attempt at confining into o-~= 0.2 MPa and ~r,. = 1.5 MPa, i.e. to the yield domain
the interface elements the plastic strains caused by the high shown in Figure 5.
shear stresses in the connection zones. The interface Figure 13 shows the principal plastic strains thus
elements have, therefore, been modeled by a Coulomb fric- obtained, extremely localized around the corners of the
tion law with no cohesion. The adopted friction angle has openings and at the intrados of the keystone of the left
been cb = 50 °, in such a way as to allow the transmission of arch. As expected, yielding is not predicted anymore in the
Numerical analysis o f old m a s o n r y buildings: F. Genna e t al. 49
x / ///..-.- . . . . . 11111111-
J Figure 75a Isostatic lines for stresses (i.e. lines connecting the
Figure 13 Principal plastic strain directions and magnitudes as
computed in analysis number 6 (Drucker-Prager for the plane directions of principal stresses) as computed in analysis num-
elements and Coulomb with no cohesion for the interface ber 6 (see Figure 13). Isostatic lines of tension, only where ~, _> 0
elements)
i///~--\\\N I
.~ X X .~ I
X X X X ~k
xl
Ii 0.2
Prll
0.17
0,14
0.11
0.0!
0.0!
o0.Ot
Figure 16a Dithering of m a x i m u m principal stresses as computed in analysis number 6 (see Figure 13)
win Prln
l ii:
-o:l
-0.I
Figure 16b Dithering of minimum principal stresses as computed in analysis number 6 (see Figure 13)
The obtained results, represented in Figures 17 20, are 5.4. Analysis number 8: collapse analysis
quite similar to the previous ones, thus confirming the A final analysis has been performed with the purpose of
soundness of the homogeneization procedure employed to finding the collapse load multiplier for the uniformly dis-
derive the De Felice equations. Here the plastic zone is tributed load applied to the floor slab at the friars' cells
slightly increased, with larger plastic strains at the keystone corridor, and by considering all the other actions as dead
intrados of the left arch and at the abutments of the right load. The adopted constitutive law for the wall has been
arch. The peak displacements are slightly lower than in the the De Felice one; the strength parameters are the same as
previous analysis. The pattern of the isostatic lines in the in the last analysis.
right part of the wall seems more oriented towards that of The collapse load multiplier, thus obtained, is cq~ ~ 12.
a panel with load applied on the top edge. This could be The corresponding plastic zone is depicted in Figure 21,
explained by the sort of 'intermediate' behaviour, between which shows the complete collapse of the left arch. Note
arch and beam, of the right arch, which also shows signs that such collapse occurs without implying any failure of
of yielding at the abutments. the supporting left diaphragm, considered as linear elastic.
Numerical analysis o f old masonry buildings: F. Genna et al. 51
N
presence of some degree of confinement which, as said,
adds ductility to the average structural behaviour. There-
fore, it is possible to conclude that the elastic-plastic
results are useful for giving a quantitative evaluation of the
stress state of the San Faustino wall. The critical points are,
according to the obtained results, the opening corners (not
relevant to the global behaviour), the keystone of the left
arch (not inspectionable on the real structure) and, in gen-
eral, all the connections between wall and elements below
it. The pattern of the isostatic lines in the wall suggests also,
without predicting it in a quantitative way, the possibility of
Figure 18 Displaced mesh as c o m p u t e d in analysis number 7 important crack openings in the direction of the com-
(see Figure 17). Displacements are magnified by a factor of 600
pression lines in the zone above the openings. Such
phenomena could be caused also by a damage occurring to
the diaphragms, which cannot be taken into account by the
t/ . . . . . _ .~\\ .~ I I I I~l
I""-~ - ~ \ \ k ///// ....... //-~-'~ ....
I I/~\\\ ~\
~"-~ frill gill L.~'-"" I E ' ~ \ % ~ ' 1 - - ~ 9 . / . . ' I / ~ I~ IIIll \\\\\\
/ I I I i l / t \ \ \.k\.\
_- ~ ~ . ~\,x~\i11L.,l~i(rl I U- /.//.e~./ " ~ . N , x ~ l l l l l l l L /. / . / I \ \ \ \ N .\
II
II
II
II
II
II
II
II
II
Figure 19a Isostatic lines for stresses as c o m p u t e d in analysis II
II
number 7 (see Figure 17). Isostatic lines of tension, only
where m ~> O
Figure 19b Isostatic lines of compression, only where m~ < 0
52 N u m e r i c a l analysis o f old m a s o n r y buildings: F. Genna et al.
I ligRzPrls
mE 0.2
m- 0.17
0.14
~ - 0.11
D~ - 0.08
0.0e
B t oT M
Figure 20a Dithering of maximum principal stresses as computed in analysis number 7 (see Figure 17)
Mitt Pril
Ii ii l
lit
l -OA
Figure20b Dithering of m i n i m u m principal stresses as computed in analysis number 7 (see Figure 17)
plane stress model, utilized here for computer memory elastic-plastic results be considered as reliable. An accurate
reasons. cracking analysis, however, without predefinition of the
Owing to the lack of reliable strength parameter values cracks themselves, for a structure of this size is an awesome
for the masonry as a homogeneous material, quite low numerical task. Also, a good crack opening detection func-
values need to be adopted with respect to those obtained tion needs to be available. All these considerations are the
from simple 'uniaxial' tests or from tests on the compo- motivation for work in progress, in which an accurate
nents alone. This effect may be linked with the so-called cracking model is being implemented in a parallel comput-
'size effect', predicted by fracture mechanics theory. The ing environment together with an element-by-element for-
problem of the choice of strength parameters remains, how- mulation which would allow remeshing in the zone
ever, an open subject. Finally, we wish to stress, once more, undergoing strain localization. This appears to be the most
the importance of performing analyses which take into promising way for analyzing brittle structures of large
account the actual cracking characteristics of the problem. size numerically.
Only when instability phenomena can be ruled out, can the
Numerical analysis of old masonry buildings: F. Genna et al. 53
: . . . :. :.: .
2 Alessandri, C. and Brebbia, C. A. 'Strength of masonry walls under
static horizontal loads: boundary element analysis and experimental
.. F-I:: : ! El.: tests', Engng Anal, 1987, 43, 118-134
3 Lotfi, H. R. and Benson Shing, P. 'Interface model applied to fracture
of masonry structures', ASCE J. Struct. Engng, 1994, 120(1), 63-80
• . . . . . . . . . . .
4 Sekender Ali, Sk. and Page, A. W. 'Finite element model for masonry
• . . . , . . . . . . . . . . . ,