Engineering Structures 117 (2016) 86–100

Contents lists available at ScienceDirect

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

Collapse analysis of slender masonry barrel vaults

G. Ramaglia, G.P. Lignola ⇑, A. Prota
Department of Structures for Engineering and Architecture, University of Naples Federico II, Via Claudio 21, Naples 80125, Italy

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

Article history: Significant part of the world cultural heritage is represented by masonry buildings. Many of them are
Received 20 July 2015 characterised by a complex architecture like as the religious buildings and need to be preserved, from
Revised 2 March 2016 a structural point of view, especially in high seismic risk areas. The recent earthquakes showed that
Accepted 7 March 2016
the Historical and Monumental buildings are often characterised by high seismic vulnerability. The low-
Available online 21 March 2016
est seismic performances are associated with the presence of thrusting elements like as arches and vaults.
In this background, the numerical analyses and experimental simulations provide important information
Keywords:
about the structural behaviour of such elements. The structural behaviour of curved elements can be
Shaking table test
Simplified modelling
complex to simulate and to predict exclusively by numerical analyses. Indeed, the experimental tests
Graphical method can provide an efficient contribution to the calibration and interpretation of the numerical models.
Barrel vaults The present work focuses on a particular typology of vaults generally used as roofs in religious build-
Arch ings. These vaults typically do not include any backfill and are slender. These typologies of masonry
Experimental validation vaults cannot be analysed with classical approaches where no-tension is assumed. In fact, the tensile
strength must be included in order to assess the seismic capacity of these masonry elements. A simplified
analytical model, which includes the tensile strength, was proposed. Validation of the analytical model is
provided by comparing predictions of the load capacity and the failure mode with those obtained from
previous shaking table test series on two full scale masonry vaults. The proposed method represents a
useful modelling tool in order to design dynamic tests on masonry vaults and to assess their vulnerability.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction inated often poorly built structures and characterised by strong

Masonry structures represent a significant part of international
architectural heritage. Although they are among the oldest struc-
tures, their knowledge is still limited and it is usually difficult to
investigate their structural behaviour. These difficulties are often
due to the heterogeneity of the materials and of building tech-
niques. Both the design techniques and building techniques from
age to age greatly evolved. The design techniques were previously
based on purely geometric rules. In the past the structural compo-
nents were designed using a geometric approach without investi-
gating the actual structural behaviour. This design approach, in
some cases, has led to robust structures. Indeed, many of these
structures: historic buildings, churches, railway bridges, etc. were
subject to several seismic events without serious damage. How-
ever, this feature is typical of prestigious and strategic structures
where same special structural measures were implemented.
Instead, the geometrical rules applied on ordinary buildings, orig-
⇑ Corresponding author.
E-mail addresses: giancarlo.ramaglia@unina.it (G. Ramaglia), glignola@unina.it
(G.P. Lignola), aprota@unina.it (A. Prota).
vulnerability. Only in recent decades designers become aware of
how to analyse masonry structures.
The computing capacity of tools now available allows the devel-
opment of complex numerical models that can be used to evaluate
many structural aspects. The heterogeneity of the masonry often
makes these numerical models very complex and unreliable when
applied to structures different from those originally used to
develop them. For these structures and their components (individ-
ual walls, vaults, etc.) the experimental support is necessary both
to assess the mechanical behaviour and to validate models by
means of numerical and experimental comparisons.
In this context, the development of simplified modelling
approaches, manageable and convenient for seismic capacity
assessment is certainly valuable. Particular attention should be
paid to some structural components, typically found in the monu-
mental buildings and churches, like as the vaults.
These structural elements, during the seismic event, often influ-
ence the behaviour of the entire building by means of the interac-
tions which they have with the adjacent structural components.
The recent seismic events (L’Aquila, 2009 and Emilia Romagna,
2012 earthquakes) showed how the vaulted structures are vulner-

http://dx.doi.org/10.1016/j.engstruct.2016.03.016
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
Fig. 1. San Martino, Buonacompra and San Francesco D’Assisi respectively, Emilia Romagna (Italy), damages after 2012-Earthquake.
able. Fig. 1 shows some examples of the vaulted structures, as roof-
ing, included in this study.
Knowledge of ultimate seismic capacity of these structural com-
ponents and the collapse mechanism are key aspects both for the
structural analysis and for the design of the strengthening inter-
ventions. The use of refined numerical FEM model can’t be always
adopted in practical structural problems without an experimental
support [1]. For this reason, the application of simplified methods
is interesting especially for ordinary engineering applications.
This work focuses on masonry vaults with significant slender-
ness ratio (span/thickness of the cross section ratio) typically used
as roofs in the churches. For this type of structures the use of clas-
sical approaches (e.g. Heyman’s theory [2]) to evaluate the seismic
capacity could not be applicable or would produce inaccurate
results. For particular geometrical values of span, rise and cross
section, the tensile strength of the masonry can’t be neglected as
it will be clarified in the following.
In a first part the critical issues of the classical analysis methods
are briefly analysed. Then the proposed method of analysis for
masonry curved elements with great slenderness is explained in
detail. The analytical method will be discussed accounting for var-
ious assumptions on the materials behaviour (linear-elastic,
cracked and plastic with stress-block).
The proposed analytical method is validated by means of exper-
imental and numerical comparisons. The experimental results
have been obtained from several experimental tests conducted in
the laboratory of Department of Structures for Engineering and
Architecture of the University of Naples, Federico II. In these tests
two masonry vaults in solid facing clay bricks were tested by
means of uniaxial dynamic tests on the shaking table. The vaults
are typical of churches built in Italy. The vaults were tested by
using several increasing signals in order to obtain an increasing
damage level. More details of the experimental program can be
found in Giamundo et al. [3]. The focus of the present paper is on
the analytical simulation of the collapse of slender masonry barrel
vaults and the role of the tensile strength by means of a simplified
model.
2. State of the art
An analysis of the principal analytical models and experimental
tests on masonry vaults has been conducted, along with the design
techniques used for many masonry arches in the past. A first pur-
pose is to investigate the classical theoretical approaches which
can be used to assess the seismic capacity of curved masonry’s ele-
ment and point out their limitations. After such analysis the novel-
ties of the proposed method are introduced as improvements of
87
the classical approaches, in particular accounting for the effect of
tensile strength of masonry.
2.1. Design by geometrical rules
Structural analyses by using numerical models and prescrip-
tions on technical codes have been applied to masonry buildings
only recently. The first constructions were designed according to
geometrical rules. The geometrical rules, in few cases (i.e. presti-
gious buildings), were coupled to other structural improvements
like as steel ties, regular walls, elimination of thrust, etc. Such
design rules were mainly based on past experiences, hence chan-
ged significantly during the years [4–6]. Generally, for a curved ele-
ment, given the span and radius of the arch, its thickness could be
easily evaluated.
In this work the focus is on curved elements made of masonry
as arches and barrel vaults, yet including simple arch elements
up to the major civil works as arch bridges, viaducts, etc. A deep
awareness of their actual safety level is still lacking.
In Brencich et al. [7] the empirical rules were analysed accord-
ing to modern structural principles. They showed that, in the
majority of real cases designed through empiric rules, the struc-
tural performances are satisfactory. In fact, many of these struc-
tures are still in service although there are both a severe
environmental degradation and inadequate maintenance. In De
Santis et al. [8] a representative sample of 34 Italian railway arch
bridges made of masonry has been assessed. The bridges date back
to XIX and XX Centuries and differ both in terms of geographical
position and geometrical properties. Surveying the geometrical
characteristics of the historical bridges (referring to the vault thick-
ness vs. span for deep arches and to pier top width vs. span and
vault thickness for shallow arches), the majority of the existing
scrutinised bridges were designed according to the empirical rules.
2.2. Classic theoretical approach
The basic assumptions refer to the material behaviour and the
no-tension is the main assumption usually adopted in the engi-
neering problems for masonry structures. No-tension means that
material has no tensile strength, hence even if tensile strains arise
for strain compatibility, corresponding tensile stresses are always
zero, masonry cracks and the effective cross-section reduces; this
assumption is on safe side and justified by the low tensile strength
of typical masonries.
Different behaviours for material can be considered, namely
linear-elastic, cracking and plastic. For each behaviour, it is well
known how to evaluate the failure surfaces [9–14] and how they
change accordingly.
88 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
In linear-elasticity, the limit surface is defined by a linear
expression (Fig. 2), in particular by two straight lines. The axial
force value at maximum bending moment is equal to 50% of the
axial plastic force (centred compressive force). The failure surface
is governed by the most stressed fibre, whose stress level can’t
be greater than the compressive strength, r0 or lower than zero,
considering that the cross section never cracks.
The cracking behaviour takes into account that a portion of the
cross section may not react in tension, while the reacting portion is
always in compression (stress lower than the compressive
strength). The failure surface is described by means of a non-
linear curve (Fig. 2). The cracking curve ends at a compressive force
equal to 50% of the pure axial strength; afterwards, the cracking
condition cannot exist (cross section is fully compressed).
Finally, under perfect plasticity assumption in compression, the
plastic failure surface can be calculated. This failure condition is
defined by a quadratic function (Fig. 2). Introducing the normalised
variables, axial force n and flexural moment m, for a section with
unit depth:
6M 6M
m¼ ¼ ð1Þ
s  P 0 s2  r0
and
P P
n¼ ¼ ð2Þ
P0 s  r0
where P and M are the normal force and bending moment, respec-
tively, acting on the cross section, s is the thickness of barrel vault or
better the height of rectangular cross section and P0 is the compres-
sive axial capacity (with a uniform compressive strength r0 in the
cross section). The failure surface can be described and plotted in
a normalised form [15], as shown in Fig. 2.
The different failure surfaces, shown in Fig. 2, by means of the
no-tension assumption, start from the origin of axes. Indeed, a pure
traction condition for the cross section (pure traction), is not com-
patible whit the basic assumptions. Finally, all the failure surface
functions are symmetric to the n axis (hence not plotted here on
the negative m side).
2.3. Heyman’s method
About the totality of engineering applications on masonry are
conducted under no-tension assumption, according to best knowl-
edge of authors. The main effects of tensile strength on the
performance of the masonry were studied in the course of time,
σ0
σ0
<σ0
Fig. 2. Normalised interaction diagram m—n for rectangular cross section: no-tension assumption ðrt ¼ 0Þ.
e.g. [16–18]. Most of those studies have been performed by using
detailed approaches as micro-mechanical modelling. However,
both numerical and analytical models accounting for tensile
strength and based on a simplified approach, like the present
one, are sporadic. In particular, for vaulted masonry structures
the Heyman’s theory [2] is commonly adopted.
The maximum compressive stresses are, on average, typically
lower than the compressive strength due to the significant dimen-
sions of the cross sections usually associated to the geometric
rules. With reduced contact surfaces only (i.e. localised interaction
between brick and mortar due to cracking effects), the normal
stresses could achieve high values.
The analytical model for the design of masonry vaults loaded by
a generic pattern is based on the following assumptions:
– No tensile strength of materials;
– compressive strength is infinite;
– no sliding failure mode.
In this way the hinge mechanism is the only failure mode.
The first assumption is conservative and common in the analy-
ses of masonry structures.
Since the crushing load is usually higher than the load which
activates the hinges [19], especially for unreinforced masonry
vaults, the crushing failure is rare.
Sliding failure mode occurs when the shear stress, at the inter-
facial surface between mortar and brick, reaches the shear
strength. In such conditions the sliding between two contacting
surfaces is activated leading to the failure of the vaulted structure
[20]. This mechanism could be activated when the angle between
the thrust line and the axis of the arch is larger than the friction
angle. In each cross section the thrust line is defined by the envel-
ope of resultant’s of the acting compressive stress distribution, also
in cracked condition. The thrust line should be inside the arch pro-
file, hence its angle with the axis of the arch is usually small.
According to the limit analysis [21–24], for a generic load pat-
tern, if the thrust line is contained entirely within the arch bound-
aries, all the blocks are able to carry the load through compressive
stresses only. However, there is not only one thrust line to guaran-
tee stability of the curved element. Indeed, according to the lower
bound theorem, any thrust line which is located within the arch
boundaries, corresponds to an equilibrium configuration for the
curved structure [25,26]. The structure is stable under a generic
load if and only if a thrust line entirely contained in the boundaries
of arch can be found.
σ0
σ0
σ0
σ0
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
Accounting for the no-tension assumption in cracked conditions
the maximum eccentricity of the axial load is equal to s=2 and
reduces at increasing axial load. Therefore, the thrust line must
be always entirely contained in the geometry of the curved ele-
ment (i.e. eccentricity of axial load contained in the range
ðs=2; þs=2Þ). The collapse condition of the curved element is rig-
orously related to the thrust line’s configuration. The collapse can
be activated if the number of contact points between the thrust
line and boundaries of curved element (thrust line becomes tan-
gent to the arch boundaries) is enough to generate a hinge mech-
anism (Fig. 3). Variations of the thrust line configuration can be
induced by variations of the horizontal and/or vertical loads. The
thrust line configuration, during the entire load history, depends
both on load pattern and on restraint conditions of the curved
element.
2.4. Drawbacks of Heyman’s method
The validity of the Heyman’s method is recognised in most of
engineering applications. Nevertheless, for some cases it could pro-
vide inaccurate solutions. Masonry vaults whit high Span=Thickness
ratios cannot be analysed by the Heyman’s method. The reduced
thickness of the curved element requires a minimum of tensile
stresses already under a gravitational load. Fig. 4 shows a masonry
arch that, in no-tension assumption, even under gravity loads only
has a load multiplier lower than 1, hence under gravity accelera-
tion the plastic compatibility conditions cannot be respected. This
example shows that, only for a vertical load lower than gravita-
tional, a thrust line entirely contained into the boundary of the
arch can exist. Such an arch is supposed to collapse, even if in real-
ity it stands up since many years and there is no evidence of any
structural problems. The failure condition is given when no thrust
line is contained in the arch boundaries and this can be seen, from
a mechanical point of view, in terms of failure domain, if plastic
compatibility conditions cannot be respected: some P—M stress
points lie outside the failure surface (or in other words the eccen-
tricity of compressive resultant P is higher than limit values given
by the M=P ratios on failure surface).
Therefore, for some structures a minimum tensile strength is
crucial, especially for masonry vaults without backfill where the
axial load is very small for gravity loads, hence eccentricities and
their fluctuations along the arch profile could be very large. This
feature is characteristic of the vaulted roofing adopted in religious
buildings. Present work focuses on this critical aspect.
Dynamic experimental tests on masonry vaulted structures
were designed by using the proposed method. This simplified
method has provided valuable basic information on the vault’s
seismic capacity and its failure mode.
Fig. 3. Classic Heyman’s theory applied to a masonry arch with fixed restraints (at
collapse with fixed gravity load and variable horizontal load).
89
Fig. 4. Masonry arch where the no-tension assumption cannot be applied (gravity
load only).
2.5. Experimental tests
The structural analyses exclusively based on theoretical and
numerical models are often not generalizable if not supported by
experimental tests. Especially for masonry structures, where some
aspects are complex, the experimental tests represent an interest-
ing support and validation. In scientific literature dynamic experi-
mental tests on masonry vaults are scarce. In most cases pseudo-
static approach was used in the lab, point wise loading the arches
by means of hydraulic jacks.
In the experimental tests, the assessment of the dynamic beha-
viour is not the only goal, but also the retrofit effects. After the
specimen has been tested, repair and retrofit interventions can
be performed according to different retrofitting techniques, and
the specimen can be tested again.
Several experimental tests investigated the structural perfor-
mance of retrofit systems with Fibre Reinforced Cementitious
Matrix (FRCM) on vaulted masonry, after it was successfully
adopted for masonry walls (e.g. [27]). Jasienko et al. [28] tested
seven semi-circular brick arches which underwent repair and ret-
rofit with FRCM system. The experimental results showed that in
all the cases the use of this retrofit technique enhances signifi-
cantly the load carrying capacity of the arches. Furthermore the
cementitious matrix ensures a great stress distribution over the
masonry substrate. Girardello et al. [29] provided another contri-
bution to the study of FRCM on arches. In such study the experi-
mental tests were performed on masonry curved elements
subjected to both monotonic and cyclic vertical loads. The results
showed that the FRCM system increases the load capacity and
the ductility of the specimen. Briccoli Bati et al. [30] provided a
comparison between the experimental performances of Fibre Rein-
forced Polymers (FRP) and FRCM systems. The tests have been car-
ried out on two arches showing the different effects of the two
techniques. The first specimen was retrofitted by means of CFRP
(Carbon FRP) strips. The second specimen was retrofitted by means
of FRCM with glass fibres. The results showed that, under cyclic
loads, the specimen with FRCM performed better. The increase of
horizontal load capacity, compared to the unreinforced specimen,
was equal to 78% for the FRCM system, while for the CFRP system
it was equal to 21%. Furthermore, the specimen retrofitted with
FRCM showed a failure mechanism similar to the unreinforced
specimen, the retrofitting with FRCM does not radically change
the structural behaviour. Instead, the specimen with CFRP system
showed a radical change in structural behaviour.
These research works are an important contribution to the
knowledge of the seismic performances of the different retrofitting
systems on masonry arches. Nevertheless, in such studies the
masonry arches are usually subjected to monotonic or cyclic
90 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
pseudo-static loads. Such experimental approaches could not be
able to realistically simulate the seismic behaviour. Furthermore,
in the static tests the external loads are usually applied by means
of hydraulic jacks. Due to concentrated load, a sliding failure can
been observed at the locations in which the load is applied to the
vault [31]. As previously described, it is well known that masonry
vaults are usually prone to the flexural failure by hinge mechanism
and the sliding failure is often neglected [2]. The point load could
induce a premature damage to the retrofit system. Such issues
could lead to an improper interpretation of the seismic perfor-
mances of the masonry vaults and their retrofitting.
The shaking tables system is able to simulate realistically the
structural behaviour under seismic actions [32]. An accurate simu-
lation of the actual seismic load conditions by means of a base
excitation is a crucial aspect. In Giamundo et al. [33,34] a study
aimed to the assessment of the structural performances of a
masonry vault and its retrofitting with an IGM system (Inorganic
Matrix Grid) has been performed. Several shaking table tests on a
masonry vault have been conducted. The IGM performances have
been assessed by means of comparison between the damages
detected prior and after retrofitting. The experimental results
showed that IMG provides both a significant increase of seismic
capacity and ductility.
3. Proposed analytical model
In experimental tests the support of numerical modelling is fun-
damental. The use of complex numerical models can be useless in
the preliminary phase of an experimental program. Especially dur-
ing the planning and design phases of the experimental tests the
use of simplified models is valuable. In order to plan a dynamic test
by means of a shaking table system, the signals and their intensi-
ties should be selected by using preliminary calculations [35].
Many structural problems concerning masonry structures are usu-
ally analysed under no-tension assumptions. The structural analy-
sis of masonry vaults through the Heyman’s model can be
performed, however this model cannot be applied on particularly
slender vaults, like as roofing typically used in religious buildings.
These structures have both a high Span=Thickness ratio and a
backfill is not present over the vault. In these cases the Heyman’s
model cannot be adopted because the hinge mechanism can be
activated already under gravity loads and the tensile strength
should be included.
3.1. Tensile strength effects on the failure surface
The tensile strength provides different effects on the failure
surface.
As previously shown the no-tension assumption provides sev-
eral failure surfaces according to the assumptions on the cross sec-
tion behaviour (linear-elastic, cracking and plastic behaviour). The
low axial force values in such masonry arches at collapse justify
the approximated overlapping of the yielding surface and cracking
surface. Under this assumption, s=2 is the maximum eccentricity
for the thrust line whatever the (low) axial load value. The elastic
failure surface is compatible with the brittle behaviour of the
masonry material, but is close also to the ideal plastic behaviour.
In any case the maximum eccentricity is independent on the axial
force achieved during the load history, and this makes the Hey-
man’s model easily implementable in a graphical approach.
Conversely, if the tensile strength is included, a dependency of
the maximum eccentricity to the axial force values occurs.
Nevertheless, as it will be explained in the following, the axial
force is low and much lower than its ultimate value. Therefore,
the maximum eccentricity can be assumed almost independent
on axial load variations until the hinge mechanism is activated.
This assumption facilitates a graphical approach; however it is
not essential in the proposed approach and can be avoided.
The tensile strength effect on the failure surface is generally an
expansion of the boundaries. The maximum thrust line’s eccentric-
ity is dependent on axial force whatever the behaviour of materi-
als. Once the internal axial load is known, the maximum
eccentricity value according to the P—M interaction diagram can
be calculated as M=P. The model is proposed for the three different
behaviours (linear-elastic, cracking and plastic/stress block). The
math expressions of the maximum eccentricity values are provided
in closed form. The no-tension assumption (i.e. classical approach)
can be considered as a particular case of the general proposed
method where the tensile strength rt is zero.
3.1.1. Linear elastic behaviour
The failure surface is evaluated in terms of P—M interaction
domain assuming a unit depth for cross section. With rt the tensile
strength value, the relationship between bending moment and
axial force is assessed according to equilibrium and stress compat-
ibility Eq. (3), in terms of maximum and minimum stresses, rmax
and rmin.
rmax 6 r0
ð3Þ
rmin P rt
In linear elastic assumption the limit condition is obtained by
means of the compressive or tensile strength achievement in the
most stressed fibre. In particular, the strength can be achieved only
on the external fibres. The Fig. 5 shows the internal stresses
according to the elastic limit state for several points of the P—M
interaction domain.
The point ðeÞ ¼ ðM bal ; P bal Þ identifies the balanced condition for
the cross section, where both the compressive strength and the
tensile strength were achieved. Assigned a normal force value,
the boundary of the P—M interaction domain can be obtained by
using the equilibrium equations.
Defined the parameter a ¼ rt =r0 , the neutral axis, xbal, at the
balanced condition can be expressed as follows.
s
xbal ¼ ð4Þ
aþ1
The maximum bending moment M el ðPÞ depends on the axial
force. The axial force values vary from the pure traction condition
Pt to the pure compressive state P0 . These terms, assuming an uni-
tary depth for the cross section, are expressed as P t ¼ rt  s and
P0 ¼ r0  s, respectively.
The envelope of each stress point M el ðPÞ provides the interac-
tion surface. According to previous Eqs. (1) and (2) the failure sur-
face can be provided in a normalised form (Fig. 6).
The failure surface shifts to the left (i.e. traction side) and the
maximum bending moment, Mel;max , increases corresponding to a
lower balanced axial force P bal . The math expressions of P bal and
Mel;max , and their normalised counterparts nbal and mbal, can be
obtained in normalised form by previous math assumptions (1)
and (2).
ð1  aÞ
nbal ¼ ð5Þ
2
ð1 þ aÞ
mbal ¼ ð6Þ
2
Similarly, the relation between bending moment and axial force
which provide the failure surface, in normalised form can be
obtained as mel(n):
mel;1 ¼ a þ n for n 6 nbal ð7Þ
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100 91

Fig. 5. Generic failure conditions for the cross section with linear-elastic behaviour.

σ0

σ0
(e)
-σt σ0
σ0
(f) >-σt
<σ0
(g)
σt σ0
(h) <σ0
(c) (d) -σt
-σt
-σt (b) -σt
(a) (i)

Fig. 6. P—M interaction surface with linear-elastic behaviour.

mel;2 ¼ 1  n for n > nbal
and the no-tension assumption is simply given by a ¼ 0. The anal-
ysis of the P—M interaction domain shows that the maximum value
of eccentricity eel ðPÞ of the thrust line depends on normalised axial
force (Fig. 6). Its math expression is:
s
a 
eel;1 ¼ þ 1 for n 6 nbal
6 n
 
s 1
eel;2 ¼  1 for n > nbal
6 n
It is observed that for a ¼ 0 (i.e. classical equation without ten-
sile strength) eccentricity is independent on axial load value. For
a > 0 and 0 < n 6 nbal the eccentricity of the thrust line is always
greater than s=6 (maximum value with no-tension and linear-
elastic assumption).
3.1.2. Cracking behaviour
According to the proposed model the tensile strength rt can be
achieved but not exceeded (while in classical assumption, no-
tensile stresses are accepted). In a brittle condition, where stresses
are virtually higher than tensile strength, that area is excluded
from calculations, hence it is assumed to crack. Therefore, two dif-
ferent parts can be identified in the cross section:
– a first part of the cross section where the tensile strength has
been exceeded causing the cracking of material. The portion
which is under the fibre at rt value doesn’t provide any con-
tribution to carrying capacity;
– a second portion where the stress is lower than the tensile
strength or in compression (but lower than compressive
strength), hence providing a contribution to capacity.
ð8Þ In cracking assumption the limit condition is obtained by means
of the simultaneous achievement of compressive and tensile
strengths on a section with reduced thickness s0 < s. The structural
issue can be solved by using an approach similar to the linear-
elastic behaviour, but conducted on a reduced cross section with
height equal to s0 . When the strength values r0 and rt on the outer-
most fibres of the reduced section are achieved, the failure condi-
ð9Þ
tion is triggered. The Fig. 7 shows the internal stress state
according to the proposed cracking’s model where a tensile
strength is included.
ð10Þ
The linear elastic behaviour is always suitable, but in a reduced
cross section. The s0 height value can be obtained as a balanced fail-
ure condition of reduced cross section, hence including the crack-
ing behaviour. This parameter can be expressed, in normalised
form, as follows.
2n
s0 ¼ s for 0 6 a < 1 and n 6 nbal ð11Þ
1a
Due to the linear behaviour of material, the neutral axis is
expressed revising Eq. (4):
s0
xcr ¼ ð12Þ
aþ1
The maximum bending moment value M cr ðPÞ is dependent on
axial force. The axial forces can vary from zero to Pbal according
to the cracking model. Indeed, it can be demonstrated that for
P < 0 the balanced failure cannot be achieved if rt < r0 as sup-
posed. In fact, a net traction cannot be achieved if tensile strength
is lower than compressive strength (and both must be reached).
Furthermore, with P < 0 even assuming a cracking condition with-
out a balanced failure (i.e. stress in compression is lower than com-
92 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100

Fig. 7. Generic failure conditions for the cross section with cracking behaviour.

pressive strength), the associated bending moment, M cr , is lower
than M el . However, it is noted that this is not conforming to the
given assumptions on cracking failure condition.
The envelope of each M cr ðPÞ provides the interaction surface
(Fig. 8) with a cracking behaviour. Even if stress distribution is lin-
ear, due to cracking, the curve is a quadratic function of axial force
and bending value is dependent on both the axial force and a
parameter.
The relation between bending moment mcr and axial force,
which provides the cracking failure surface, in normalised form is:
 
6  n2 aþ1
mcr ¼ 3  n þ þ 1 for 0 6 a < 1 and n 6 nbal
a  1 3ða  1Þ
The maximum eccentricity value ecr ðPÞ of the thrust line
depends on the axial force. The equation of the eccentricity in nor-
malised form is:
 
s sn aþ1
ecr ¼ þ þ 1 for 0 6 a < 1 and n 6 nbal
2 a  1 3ða  1Þ
It is observed that for a ¼ 0, the previous equations fall into the
classical equation where no-tension is assumed, and eccentricity is
s=2 exactly in pure bending (i.e. n = 0) and approximately at low
axial load values.
The Fig. 8 shows the main effects of tensile strength assumption
in cracking conditions. Between the linear elastic and cracking fail-
ure surfaces, depending on the a value, an intersection point can
exist.
If cracking state occurs or not, it depends on a and, by compar-
ing Eqs. (7) and (13), the limit values for a can be assessed.
For a P 1=3 the cracking failure surface is totally below the
elastic failure surface, while for a < 1=3 the surface is partially
above the elastic failure surface (totally above in the case of a ¼ 0).
3.1.3. Plastic behaviour (stress-block model)
A perfect plastic behaviour can be used for the compressive
stress under the no-tension assumption. When the tensile strength
is assumed, a perfect plastic behaviour cannot be adopted for the
tensile stress also. Therefore, a stress-block based model [36] was
adopted in order to assess the post-elastic behaviour of the
masonry. The stress-block model was adopted both for the com-
pressive and tensile behaviour. The failure surface is very close to
the yielding surface obtained under the no-tension assumption.
In fact, the tensile strength has an extremely low value for the
ð13Þ masonry material. On average, the tensile strength rt is approxi-
mately equal to 1=10 of the compressive strength r0 .
Interaction between bending moment and normal force
depends on internal stresses as shown in Fig. 9.
In particular, the entire boundary of the P—M interaction
domain is obtained through different equations. When n ranges
from a to a=2 (i.e. points a00 and b00 respectively), the behaviour
ð14Þ does not differ from the linear elastic one (i.e. there is no compres-
sion, the behaviour is governed by the tensile strength and the
height of the plastic zones in tension is multiplied by 0.5). In this
range the P—M interaction domain is provided by the previous
Eq. (7).
When n P 0:5 some fibres are compressed and according to
the stress-block model the height of the plastic zones is the neutral
axis, xp, multiplied by w (usually assumed equal to 0.8), hence it is
w  xp and the distance of application point of resultant force from
neutral axis is k  xp (usually assuming k equal to 0.4). The neutral
axis with plastic behaviour is provided by the following equation.

σ0 σ0

(c’) (d’)
-σt -σt

σ0

(b’)
-σt

σ0

(a’) -σt

Fig. 8. P—M interaction surface with cracking behaviour.

 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100 93

Fig. 9. Generic failure conditions for the cross section by using stress-block model.

 
2nþa a the neutral axis (external to the cross section). The axial force val-
xp;1 ¼ s for 0 6 a < 1 and  6 n 6 w ð15Þ
aþ2w 2 ues range from a compression of w  P0 up to a pure compression
value of P 0 .
The previous equation provides the neutral axis up to the limit
The relationship between bending moment and axial force in
condition when the cross section is totally compressed (i.e. point e00
normalised form is:
shown in Fig. 9). This condition is reached when the neutral axis is
s  w. The axial force values range from a traction of 0:5  Pt to a  
1 k
compression value of w  P 0 . The envelope of each M p ðPÞ value is mp;2 ¼ 6n n for 0 6 a < 1 and w 6 n 6 1 ð18Þ
2 w
provided by a quadratic function. In Fig. 10 the P—M interaction
domain is shown in a normalised form, mp(n). The maximum of The maximum eccentricity value ep ðPÞ of the thrust line can be
the bending moment M p ðPÞ is obtained at the balanced axial load obtained as follows:
nbal . The latter value is provided through the Eq. (5).
 
The failure surface is always outside the interaction surfaces 1 k
ep;2 ¼ s n for 0 6 a < 1 and w 6 n 6 1 ð19Þ
with linear-elastic and cracking behaviour. The relation between 2 w
bending moment and axial force in normalised form is:
The Eqs. (7), (8), (13), (16) and (18) provide the stress compat-
! ibility equations that, starting from the gravity load up to collapse
a að2n þ aÞ2 k  wð2n þ aÞ2 að2n þ aÞ wð2n þ aÞ condition, must be satisfied.
mp;1 ¼ 6    þ þ
12 6ða þ 2wÞ2 ða þ 2wÞ2 12ða þ 2wÞ 2ða þ 2wÞ If the tensile strength is included, even for low axial force val-
a ues, the eccentricity of the thrust line is not constant with axial
for 0 6 a < 1 and  6 n 6 w ð16Þ load force.
2
The proposed model showed that, including tensile strength,
The previous equation can be applied for n values up to n ¼ w.
the maximum eccentricity of the thrust line can be greater com-
The math expression of the maximum eccentricity ep ðPÞ of the
pared to the classical approach (no-tension assumption). Indeed,
thrust line is:
the thrust line can be external to the geometrical boundaries of
! the curved element. These maximum eccentricities are provided
s a að2n þ aÞ2 k  wð2n þ aÞ2 að2n þ aÞ wð2n þ aÞ in closed form by the Eqs. (9), (10), (14), (17) and (19).
ep;1 ¼   þ þ
n 12 6ða þ 2wÞ2 ða þ 2wÞ2 12ða þ 2wÞ 2ða þ 2wÞ
a
for 0 6 a < 1 and  6 n 6 w ð17Þ 3.2. Graphical interpretation: fictitious thickness
2
For n P w the failure surface falls into the classical region with For a generic load pattern (vertical and/or horizontal load), the
no-tension assumption. In fact, starting from a section totally com- masonry arch is in safe condition if the hinge mechanism has not
pressed, the eventual tensile strength does not impact on the fail- been activated (most likely mechanism). The collapse condition
ure surface. The maximum bending moment is independent from is activated when the thrust line reaches the maximum eccentric-

σ0 σ0

-σt
(d’’) (e’’)
σ0
σ0

-σt
(c’’)
-σt
-σt (b’’) (f’’)
(a’’)

Fig. 10. P—M interaction surface with plastic behaviour (stress-block model).
94 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100

ity in a sufficient number of points. However, the maximum eccen-
tricity depends on the material’s behaviour and axial load value.
The previous Eqs. (9), (10), (14), (17) and (19) provide the max-
imum eccentricities eel , ecr and ep of the thrust line according to the
linear, cracking and plastic behaviour, respectively. The fictitious
thickness, which depends on the axial force value achieved during
the load history, can be assessed using these equations.
Fixed a generic load step, the axial force value is known and the
maximum eccentricity value can be calculated for any section
(Fig. 11). The fictitious geometry of the masonry arch evolves with
the axial loads.
3.2.1. Dependence of the fictitious thickness on the axial force and
simplified approach aimed at the graphical method application
For each load step, the thrust line must be contained in bound-
aries that are altered during the entire load history until the col-
lapse condition. This aspect dictates some limits to the
applicability as a graphical method. However, the fictitious geom-
etry could be assumed as fixed, assuming a negligible variation of
the axial forces. With this additional simplification an approach
similar to Heyman’s method can be performed (collapse multiplier
and failure mode). However, it is noted that increasing the axial
load, the fictitious thickness reduces. For this reason, the expected
increases of axial load should be carefully noted and estimated.
On safe side, the maximum expected axial load value in each
section could be considered to evaluate the linked eccentricity.
The maximum expected axial load value could be related to the
axial load values under gravity conditions.
In Fig. 12 a masonry vault loaded by a generic vertical and hor-
izontal load pattern is shown.
Starting from the gravitational load, a thrust line variation is
evaluated due to horizontal loads. The thrust line can be tangent
to the boundary of fictitious geometry of the masonry arch during
the loading history. The maximum numbers of tangent points
depends on the restraint conditions. The collapse condition is
achieved by means of the hinge mechanism (i.e. sufficient number
of tangent points or hinge formations).
The solving equations are based on a discrete approach. The
curved element can be modelled by means of a finite number of
elements (finite number of beams nd and nodes ndþ1 ). In a generic
control section Si , according to the assumed material behaviour,
the plastic hinge can be activated if the maximum bending
moment is exceeded.
The restraint devices can be replaced by their reactions. Assum-
ing a fixed arch, defining with A and B the two sections at imposts,
for each of them three reactions occurs. HA , HB and V A , V B are the
horizontal thrusts and vertical reactions at the impost sections
respectively, whereas M A and M B are the bending moments acting
in the X; Y plane. On each discretized element there is a discretized
external load V i and Hi where i ¼ 1; 2; . . . ; nd is the generic element
of the arch.
In a Cartesian system x; y; z are defined in the barycentre of the
generic arch’s element. The reaction values HB ; V B ; M B at the arch’s
impost B can be written based on the HA ; V A ; M A counterparts as
follows:
X
nd

Therefore, the graphical method can be applied on a curved

V B ¼ V A þ Vi
0
masonry element where the actual thickness s is increased to a fic-
X
nd
titious value equal to twice the eccentricities eel , ecr and ep , point HB ¼ HA þ Hi ð20Þ
wise evaluated according to previous equations. 0
X
nd X
nd
MB ¼ M A þ V A  l  V i  ðl  xi Þ þ Hi  ðyi Þ
3.3. Solving algorithm 0 0

where l is the arch’s span and xi ; yi identify the generic element’s

In order to apply the proposed analytical model, a solver algo-
location in the local coordinate system.
rithm is provided by using a synthetic form. A masonry arch loaded
Similarly the internal forces within the generic element j of the
by a generic load pattern can be analysed by means of a discrete
arch, according to Fig. 12, can be expressed as follows:
approach. The vertical and horizontal load patterns can be inde-
pendent and the goal is to assess the maximum load leading to a X
j

collapse mechanism. This load value can be easily linked to a lat- V j ¼ V A þ Vi

eral acceleration. 0

Usually, the vertical load is assumed constant throughout the X

j

entire load history. The collapse condition can be achieved by using

Hj ¼ HA þ Hi
0
an increasing horizontal load pattern. The horizontal load’s shape
X
j X
j
can be chosen proportional to the masses or to the first natural Mj ¼ M A þ V A  xj þ HA  yj  V i  ðxj  xi Þ  Hi  ðyj  yi Þ
mode (as classical pushover analyses require). For some curved 0 0
elements, the horizontal shaking motion yields to additional not ð21Þ
negligible vertical accelerations [37].

emax(n2)
emax(n1) emax(n3)

emax(n3)
emax(n2)
emax(n1)

n1 n2 n3

Fig. 11. The evolution of the fictitious geometry varying the normal stress values.
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
Fig. 12. Discrete model of a masonry arch under a generic load pattern.
where the index i identifies all the elements starting from the
impost’s section A to the control section Sj .
The internal axial and shear stresses, Pj and Tj, can be calculated
as:
Pj ¼ Hj  cos aj  V j  sin aj
ð22Þ
T j ¼ ðHj  sin aj þ V j  cos aj Þ
where aj is the angle of the axis normal to the control section Sj
with respect to the horizontal axis x (local tangent at the curved
element).
The bending moment can be calculated by using the third equa-
tion of the system (21).
On each element a horizontal force is applied, expressed as
Hi ¼ k  V i , where the natural number k is the multiplier of the hor-
izontal load, for a generic load step. The k factor can be increased
up to the collapse condition (hinge mechanism), hence related to
the collapse’s acceleration.
The internal forces M j ; Pj ratio provides the eccentricity of the
thrust line at any load step. Denoting with M 0 ðPj Þ the maximum
bending moment according to the behaviour assumed for the
masonry material (i.e. linear-elastic, cracking or plastic), at each
section the additional compatibility condition, written as follows,
must be satisfied.
M j M 0 ðPj Þ
6 that is ej 6 ej;0 ðPj Þ ð23Þ
Pj Pj
where ej;0 is the maximum eccentricity previously discussed. In all
control sections, hinges can be activated.
In order to apply an incremental analysis aimed at the collapse
activation, in the generic load step where the hinge is activated the
0
structure must be updated. If j is the cross section where the hinge
is activated, it is M j 6 M 0 ðPÞ.
0 Architecture, University of Naples Federico II.
Indeed, the compatibility equations for any step after the first
hinge activation must be satisfied. The structure is updated during
the loading history up to the ultimate collapse mechanism.
The incremental analysis provides both the collapse’s accelera-
tion and the evolution of the hinge mechanism.
95
If the assessment of the collapse acceleration is the goal, the
solution is even easier by using the Limit Analysis Theorems.
According to the framework of the lower bound theorem, each
possible thrust line lying within the fictitious arch boundaries, cor-
responds to an equilibrium configuration of the curved element
where the tensile strength is assumed. The solver algorithm can
be implemented as discussed below.
The structure must be transformed into its equivalent isostatic
structure by the elimination of the restraint devices. They must be
replaced by the related unknown reactions X k with k ¼ 1; 2; . . . ; K
and K is the number of statically indeterminate reactions. The
internal force diagrams are evaluated (bending moment and axial
force) on the equivalent isostatic structure due to the external
gravity loads V i (M V j and PV j ), the horizontal loads Hi ¼ k  V i (M Hj
and P Hj ) and the statically indeterminate reactions (M X K and P X K ).
The global internal forces include the different effects as shown:
P j ¼ P V j þ P Hj þ X K  P X K
ð24Þ
M j ¼ M V j þ M Hj þ X K  M X K
Finally the compatibility Eq. (23) must be imposed.
In order to determine the static multiplier k, a numerical proce-
dure can be conducted that transforms the structural problem into
a linear programming problem optimising the system of equations
given by (24).
4. Method validation by means of experimental dynamic
investigations
The proposed method has been validated by means of previous
experimental shaking table tests discussed in Giamundo et al. [3].
Several experimental tests on two full scale masonry vaults were
performed at the Department of Structures for Engineering and
The specimens have been tested in two different periods, but
similar materials and workmanships were adopted. The goal was
to assess the seismic capacity of masonry vaults and the retrofit
effects after damage. Indeed, the tests have been carried out in
two parts: in a first part the tests have been performed on the
96 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
unreinforced specimens; in a second step the dynamic tests have
been performed again on the specimen retrofitted by means of
an IGM system and repaired by means of mortar joint repointing
and grout injections. For both specimens a similar approach has
been used. This work focuses on the analysis of unreinforced spec-
imens only.
4.1. Specimens characteristics
Several experimental tests on shaking table have been con-
ducted. The two full scale masonry vaults have different geometri-
cal characteristics. Both the specimens, are made of solid facing
clay bricks (255.512) cm3 and pozzolana-based mortar joints
(10 mm thick). The vaults have been built up on a rigid steel beams
system which is part of the test setup. The geometry of the speci-
mens and their structural characteristics are representative of the
vaulted roofs commonly adopted in the historical religious build-
ings (Fig. 13), yet respecting some lab constraints.
In-plane geometries of the curved element are the same for
both specimens. The first specimen is characterised by a segmental
arch profile which is less than a semicircle. The span, rise and
depth values of the vault are 298 cm, 114 cm and 220 cm respec-
tively. Further details about the geometry of the first specimen
are provided in Fig. 14a. Conversely, the second specimen has a dif-
ferent depth both for the base and its curved element. The arch was
built over two masonry walls. By means of these panels the vault’s
imposts have a height equal to 103 cm with respect to the bases as
shown in Fig. 14b. These masonry panels are made in two brick
rows and they have a greater cross section compared to the curved
element. Moreover the masonry lateral walls have been raised up
to 234 cm height. The depth of the vaulted structure is reduced
to 116 cm to fulfil payload of the shaking tables system. A wooden
truss is added to simulate the condition of a real vault, with a tri-
angular truss as roofing, in the phase of repairing and retrofitting,
and it is simply supported on the walls, not providing at all a tying
beneficial effect.
The backfill, usually acting on traditional vaults, provides a ben-
eficial effect on the structure [38–40]. However, this effect is not
present in vaulted roofing as shown in Fig. 13. Therefore, in these
experimental studies, in order to simulate such vault typology,
both the specimens have been tested without any backfill. The
specimens were constrained on the shaking table and such bound-
ary conditions simulate the action of the ties frequently adopted in
the retrofit of historical vaulted structures. In a multiscale
approach, perfect fixing of the imposts represents the first step
to focus on the vault itself, while different boundary conditions
Fig. 13. San Biagio D’Amiterno Church, L’Aquila (Italy), damages after 2009-Earthquake.
would have added further variables to the system resulting in a
more complex problem interpretation. The addition of the masonry
lateral walls, even if with reduced height so far, added the effect of
sub-structure deformability, yet conforming to lab constraints, and
the model is able also to simulate them. Rationale and limits of
previous shaking table tests are discussed elsewhere (e.g. [3]),
while the focus of present work is the simplified modelling, hence
the validation by comparison with available experimental tests
and configurations. More complex tests are foreseen for the future
and the proposed model will be able to assist in the design of the
future tests accounting for more slender walls or different struc-
tural geometries.
The unreinforced specimens have been tested by using several
signals (i.e. replicas of natural and artificial accelerograms, with a
frequency content compatible with the tested specimens) repeated
increasing the peak accelerations, as commonly done in shaking
test programs. The effect of replicas is expected to yield to strength
decay and, to account for tensile strength uncertainty, parametric
analyses can be performed on a range of possible strength values.
At the end of each test, visual surveys on the specimen have been
conducted. The dynamic behaviour of the two specimens was sig-
nificantly different.
4.2. Experimental results
The first unreinforced specimen has been tested by means of
seven artificial signals with increasing intensity. Until to the fifth
test no substantial damages have been detected. After the last sig-
nal, in several locations, crack openings at the interfaces between
mortar and brick have been detected. The crack openings denote
the hinges activation and have been observed both at the intrados
and the extrados of the masonry vault. Cracking at the intrados
occurred along almost the entire depth of the joint at about 1/6
of the span length. At the extrados the cracking has involved a lar-
ger number of joints. The formation of all the described cracks
occurred at the interfaces only by means of a mechanism of ‘‘crack
opening and closing”.
The cracks occurred at almost symmetrical locations at the
inversion of the shaking direction. In order to highlight such mech-
anism, in Fig. 15a a still image taken from a video record of the
tests shows the crack tips highlighted by a red circle (lateral loads
acting from left to right). The crack opening remarks the triggering
of a hinge mechanism in the vault. The vault’s collapse was pre-
vented by stopping the test. However, the specimen went very
close to the hinge mechanism (i.e. four hinges activated). The
fourth hinge activated at an acceleration close to 0.48 g.
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100 97

12
30 31 32 31 30 29 28

(a) 23 24
25 26
27 28 29 27 26
25
24 23
5,5
Specimen 1 19
20
21
22 22
21
20
19
18 18
17 17
16 16
15 15
14 14

114
13 13
12 12

133
11 11
10 10

152
9 9
8 8
7 7
6 6
5 5
4 4
3 Structural model 3
2 2
1 1
38

38 298 38 220

25
Wooden truss 25 116
36
Specimen 2 36
35 35
34 34
33 33
32 32
31 31

12
30 30 31 32 31 30 29 28 30
29
23 24
25 26
27 28 29 27 26
25
24 23
5,5 29
28 22 22 28
(b) 27
26 18
19
20
21 21
20
19
18
27
26
17 17
25 16 16 25
24 15 15 24
14 14
23 13 13 23
22 12 12 22
21 11 11 21
133

10 10
20 20
9 9

234
19 8 8 19
18 7 7 18

265,5
17 6 6 17
16 5 5 16
15 4 4 15
14 3 3 14
13 2 2 13
1 1
12 12
11 11
10 10
9 9
8 8
7 7
6 6
5 Structural model 5
4
3
4
3 58,5
2 323 2
1 1
31,5

38 298 38 233

Fig. 14. Specimen geometry and structural model adopted: (a) first and (b) second tested vault, respectively.

The second unreinforced specimen has been tested by means of tical load pattern has been considered. The internal force variations
five increasing natural signals up to the collapse. At the end of each are due to horizontal load increase only, by means of a monotoni-
test, visual surveys on the specimen have been conducted. Till the cally increasing horizontal acceleration. The horizontal load pat-
last signal having a Peak Ground Acceleration (PGA) equal to tern is proportional to the masses of the structure, while vertical
0.22 g, visual survey did not show any evident damage. The col- load is gravity.
lapse occurred by activation of four plastic hinges almost instantly The masonry vaults have been modelled by means of a homoge-
at peak acceleration of the last signal, as shown in Fig. 15b, when neous material and fixed at the base.
loads acted from left to right. The elastic failure surface has been chosen for the simulation.
Both the experimental results confirm the hinge mechanism to According to this assumption the interaction P—M surface is pro-
be the limiting failure mode in masonry vaults compared to sliding vided by Eqs. (7) and (8). Furthermore, the maximum eccentricity
or crushing failure [3]. values for the thrust line are provided by Eqs. (9) and (10).
The structural analyses take into account variability of the
mechanical properties also. Starting from mechanical parameters
4.3. Analytical modelling and experimental comparison
estimated by experimental tests, a reasonable range of mechanical
characteristics has been provided in order to evaluate the collapse
The two masonry vaults have been analysed by means of the
accelerations for the two unreinforced specimens. Therefore, by
proposed analytical model. For the structural analysis a fixed ver-

(a) (b)

Fig. 15. (a) Last signal for the first unreinforced specimen (hinge mechanism activated without collapse) and (b) last signal for the second unreinforced specimen (collapse
occurred).
98 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100

Table 1 The collapse multiplier k has been calculated by means of the

Expected ranges of mechanical parameters. lower bound theorem. The analytical model takes into account
Mechanical parameter Minimum Mean Maximum for the axial load variations. This last aspect produces a continuous
Young modulus (GPa) 1.2 1.5 1.8 update of the fictitious geometry within which the thrust line must
Compressive strength (MPa) 2.4 3.2 4 be contained.
Tensile strength (MPa) 0.16 0.32 0.64 By the knowledge of k, the PGA values can be assessed at the
collapse condition for the two specimens. In Table 2 the results
are shown in terms of PGA values at collapse condition for the
two specimens.
using the proposed simplified model, several sensitivity analyses
The vault was divided into 64 elements and a specific weight of
have been performed.
18 kN/m3 was assumed for the masonry. The maximum horizontal
The expected ranges of the mechanical characteristics are
load multiplier was evaluated having the thrust line inside the fic-
shown in Table 1 and, for each set of parameters, the collapse
titious boundaries of the arch. A sensitivity analysis was conducted
acceleration has been calculated for both the masonry specimens.
on the number of elements for the discretization and its impact is
The tensile strength has the higher impact and its mean value
less significant than the tensile strength. The present choice repre-
has been estimated by means of an additional static vertical load
sents a compromise between accuracy and efforts, where the most
test on the first vault at the end of the shaking test sequence
relevant parameter in such slender walls without backfill is the
(rt = 0.32 MPa).
tensile strength. In Fig. 16 the thrust line configuration is shown

Table 2
PGA values at mechanism activation.

Specimen rt = 0.16 MPa rt = 0.32 MPa estimated value rt = 0.64 MPa Experimental value
PGAspecimen 1 (g) 0.33 0.54 0.88 0.48
PGAspecimen 2 (g) 0.16 0.23 0.41 0.22

Hinge 4 Hinge 4
Hinge 1

Hinge 3

Hinge 1

Hinge 2
Hinge 2 Hinge 3

Fig. 16. Theoretical thrust line configuration for the first specimen (hinge mechanism activated) and P—M points in the elastic failure surface.

Hinge 4

Hinge 4

Hinge 2

Hinge 2

Fig. 17. Theoretical thrust line configuration for the second specimen (hinge mechanism activated) and P—M points in the elastic failure surface (curved element).
 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100
for the first specimen at the theoretical collapse condition (i.e.
k = 0.54 g) and the internal forces M j ; Pj are shown as hollow signs
inside the interaction P—M surface, in solid line. M j ; P j points
clearly reach the elastic surface in four locations representing the
hinges formation. The hinge position corresponds to the section
where the internal forces M j ; Pj intersect the interaction surface
or, similarly, corresponds to the tangent point of the thrust line
to the fictitious boundaries of the arch. It is worth noting that
experimental test was stopped at 0.48 g to prevent collapse, how-
ever the formation of the fourth hinge almost occurred experimen-
tally. These numerical results have been shown only for the
average values of mechanical parameters and they are in satisfac-
tory agreement with hinge location (a comparison of experimental
result in Fig. 15a and theoretical outcome is in Fig. 16) and corre-
sponding experimental load.
Similarly in the following figures, the graphical representations
of the numerical results are provided for the second specimen. The
method and the average values of mechanical parameters were the
same because the vaults were similar in terms of materials. The
arch and walls were divided into 68 elements according to the
specimen geometry. The curved portion was discretized in the
same number of elements as previous case, while the walls were
divided into two elements only, because potential locations for
hinges were at their ends. However, increasing the number of ele-
ments in the lateral walls yields to the same results. A uniform hor-
izontal load was assumed involving both walls and curved portion
(i.e. the arch). Fig. 17 shows the configuration of the thrust line and
the internal forces at collapse condition (i.e. k = 0.23 g) for the
curved element only. The internal forces M j ; P j reach the elastic
surface in two points only, representing the hinges 2 and 4 forma-
tion on the arch.
Conversely, Fig. 18 provides the thrust line configuration for the
lateral masonry walls at the same load. It is worth noting that due
to very low axial load values, the maximum eccentricity provided
by equations (9) and (10) achieves extremely high values, espe-
cially at the top of the walls. On the other hand, at the intersection
of the arch with the walls, the concentrated force due to the arch
yields to a prompt increase of the axial load, hence a reduction of
the fictitious thickness in the lower portion of the walls. The inter-
nal forces Mj ; Pj reach the elastic surface at the bases only, repre-
senting the hinges 1 and 3 formation at the bases of the lateral
walls. The tying at the wood truss location was not included in
the model because, as expected, wood truss was free to slip during
the experimental shakings.
Hinge 3
Fig. 18. Theoretical thrust line configuration for the second specimen (hinge mechanism activated) and P—M points in the elastic failure surface (masonry walls).
99
The agreement in terms of both activation multiplier and failure
mode (i.e. hinge mechanism pattern, comparing experimental
locations in Fig. 15b and theoretical outcomes in Figs. 17 and 18)
is very satisfactory.
5. Conclusions
The paper discussed the main aspects related to the analysis of
slender masonry vaults without any backfill and with high
Span=Thickness ratios, as those typically used in religious buildings
as roofing. Their analysis by using the Heyman’s Theory cannot be
performed due to the very low axial force values, typically
achieved. Indeed, if no-tension is assumed, the collapse is
expected, by means of theoretical analysis, even under gravita-
tional load (even if the real structure shows no evidence of any
structural problem) hence a tensile strength value should be
assumed. The equations in closed form are provided to account
for three possible material behaviours (linear-elastic, cracking
and plastic with stress block). The proposed model is easily imple-
mentable by means of a solving algorithm.
Through the numerical and experimental comparison the pro-
posed analytical method has been validated. The analytical model
provided reliable results and the hinge locations were correctly
predicted. The model is sensitive to tensile strength of masonry
and the impact of parameter variations was evaluated. The number
of elements for the discretiziation has a reduced impact on the
results compared to tensile strength. For the first specimen accord-
ing to the proposed approach the collapse acceleration is 0.54 g
yielding to four hinges formation. For the same experimental pro-
gram, by assuming the same average mechanical parameters, the
analytical model provided the actual collapse of the second unrein-
forced masonry vault at an acceleration of 0.23 g. Hence the model
is able to account also for systems with different configurations
(i.e. curved portion alone or resting on lateral walls). By assuming
a constant safe value of axial force, it can be applied as a graphical
method too, providing a fictitious geometry for the arch.
The proposed method is valuable to evaluate the ultimate
capacity of the masonry vaults, by using a simplified approach.
Similarly, the hinges location is satisfactorily predicted, hence
the location of potential retrofit interventions is estimated. Finally,
the model can be extended to the FRP, FRCM or IGM retrofit sys-
tems, being able to increase the load carrying capacity of the
vaults. This capacity increase is due to the expansion of the failure
surface which moves far away from unreinforced failure condition.
Hinge 1
Hinge 1 Hinge 3
100 G. Ramaglia et al. / Engineering Structures 117 (2016) 86–100

A further validation is required with experimentally tested retro- [17] Fanning JP, Boothby TE, Roberts BJ. Longitudinal and transverse effects in
masonry arch assessment. Constr Build Mater 2001;15(1):54–60. http://dx.
fitted vaults.
doi.org/10.1016/S0950-0618(00)00069-6.
[18] De Buhan P, de Felice G. A homogenization approach to the ultimate strength
of brick masonry. J Mech Phys Solids 1997;45(7):1085–104. http://dx.doi.org/
Acknowledgements 10.1016/S0022-5096(97)00002-1.
[19] Foraboschi P. Strengthening of masonry arches with fiber-reinforced polymer
The authors acknowledge the financial support provided by strips. J Compos Constr 2004;8:191–202. http://dx.doi.org/10.1061/(ASCE)
1090-0268(2004) 8:3(191).
Rete dei Laboratori Universitari di Ingegneria Sismica – ReLUIS [20] Drosopoulos GA, Stavroulakis GE, Massalas CV. FRP reinforcement of stone
for the research program funded by the Department of Civil Protec- arch bridges: unilateral contact models and limit analysis. Compos B Eng
tion – Executive Project 2014–2018. 2007;38:144–51. http://dx.doi.org/10.1016/j.compositesb.2006.08.004.
[21] Fortunato A, Fraternali F, Angelillo A. Structural capacity of masonry walls
under horizontal loads. Ingegneria Sismica/Int J Earthq Eng 2014;31:41–51.
References [22] Angelillo M, Fortunato A, Lippiello M, Montanino A. Singular stress fields in
masonry structures: derand was right. Meccanica 2014;49:1243–62. http://dx.
doi.org/10.1007/s11012-014-9880-6.
[1] Lignola GP, Prota A, Manfredi G. Numerical investigation on the influence of
[23] Angelillo M, Babilio E, Fortunato A. Singular stress fields for masonry-like
FRP retrofit layout and geometry on the in-plane behavior of masonry walls.
vaults. Continuum Mech Thermodyn 2013;25:423–41. http://dx.doi.org/
ASCE J Compos Constr 2012;16(6):712–23. http://dx.doi.org/10.1061/(ASCE)
10.1007/s00161-012-0270-9.
CC.1943-5614.0000297.
[24] Angelillo M, Fortunato A. Equilibrium of masonry vaults. In: Frémond M,
[2] Heyman J. The stone skeleton: structural engineering of masonry
Maceri F, editors. Nov Approaches Civ Eng SE – 6, vol. 14. Berlin
architecture. Cambridge: Cambridge University Press; 1995.
Heidelberg: Springer; 2004. http://dx.doi.org/10.1007/978-3-540-45287-4.
[3] Giamundo V, Lignola GP, Maddaloni G, Iovinella I, Balsamo A, Prota A, Manfredi
[25] Moseley H. On the theory of the arch. In: Weale J, editor. The theory, practice
G. Full scale clay brick un-reinforced masonry vault: a shaking table test. In:
and architecture of bridges of stone, iron, timber and wire with examples on
Proceedings of the fifteenth international conference on civil, structural and
the principle of suspension. London: J. Weale; 1843.
environmental engineering computing; 2015. doi: 10.4203/ccp.108.115.
[26] Milankovitch M. Theorie der Druckkurven. Zeitschrift Für Math Und Phys
[4] Katalin B. When Heyman’s Safe Theorem of rigid block systems fails: non-
1907;55:1–27.
Heymanian collapse modes of masonry structures. Int J Solids Struct 2014;51
[27] Lignola GP, Prota A, Manfredi G. Nonlinear analyses of tuff masonry walls
(14):2696–705. http://dx.doi.org/10.1016/j.ijsolstr.2014.03.041.
strengthened with cementitious matrix-grid composites. ASCE J Compos
[5] Oppenheim IJ. The masonry arch as a four-link mechanism under base motion.
Constr 2009;13(4):243–51. http://dx.doi.org/10.1061/(ASCE)CC.1943-
Earthquake Eng Struct Dynam 1992;21(11):1005–17. http://dx.doi.org/
5614.0000007.
10.1002/eqe.4290211105.
[28] Jasienko J, Di Tommaso A, Bednarz L. Experimental investigation into collapse
[6] Dimitri R, De Lorenzis L, Zavarise G. Numerical study on the dynamic behavior
of masonry arches reinforced using different compatible technologies. Proc
of masonry columns and arches on buttresses with the discrete element
MuRiCo3 Mech Masonry Struct Strength Compos Mater 2009.
method. Eng Struct 2011;33(12):3172–88. http://dx.doi.org/10.1016/j.
[29] Girardello P, Pappas A, da Porto F, Valluzzi RM. Experimental testing and
engstruct.2011.08.018.
numerical modelling of masonry vaults. Proc Int Confer Rehabilit Restor Struct
[7] Brencich A, Morbiducci R. Masonry arches: historical rules and modern
2013.
mechanics. Int J Arch Herit 2007;1:165–89. http://dx.doi.org/10.1080/
[30] Briccoli Bati S, Rovero L, Tonietti U. Strengthening masonry arches with
15583050701312926.
composite materials. J Compos Constr 2007;11:33–41. http://dx.doi.org/
[8] De Santis S, de Felice G, Peluso D. An overview on Italian railway masonry
10.1061/(ASCE)1090-0268(2007) 11:1(33).
bridges with load-carrying capability estimate. Structural Analysis of
[31] Garmendia L, San-José JT, García D, Larrinaga P. Rehabilitation of masonry
Historical Constructions. Wroclae, Poland; 2012. ISSN 0860-2395H. ISBN
arches with compatible advanced composite material. Constr Build Mater
978-83-7125-216-7.
2011;25:4374–85. http://dx.doi.org/10.1016/j.conbuildmat.2011.03.065.
[9] Giamundo V, Lignola GP, Prota A, Manfredi G. Nonlinear analyses of adobe
[32] Duarte RT, Ritto-Corrêa M, Vaz CT, Campos-Costa A. Shaking table testing of
masonry walls reinforced with fiberglass mesh. Polymers (Basel)
structures. Proc Earthq Eng Tenth World Conf 1994. ISBN: 9054100605.
2014;6:464–78. http://dx.doi.org/10.3390/polym6020464.
[33] Giamundo V, Lignola GP, Maddaloni G, da Porto F, Prota A, Manfredi G. Shaking
[10] Prota A, Marcari G, Fabbrocino G, Manfredi G, Aldea C. Experimental in-plane
table tests on a full-scale unreinforced and IMG-retrofitted clay brick masonry
behavior of tuff masonry strengthened with cementitious matrix-grid
barrel vault. Bull Earthq Eng 2016. http://dx.doi.org/10.1007/s10518-016-
composites. J Compos Constr 2006;10:223–33. http://dx.doi.org/10.1061/
9886-7.
(ASCE)1090-0268(2006) 10:3(223).
[34] Giamundo V, Lignola GP, Maddaloni G, Balsamo A, Prota A, Manfredi G.
[11] Borri A, Casadei P, Castori G, Hammond J. Strengthening of
Experimental investigation of the seismic performances of IMG reinforcement
brick masonry arches with externally bonded steel reinforced composites. J
on curved masonry elements. Compos B Eng 2015;70:53–63. http://dx.doi.org/
Compos Constr 2009;13(6):468–75. http://dx.doi.org/10.1061/(ASCE)CC.1943-
10.1016/j.compositesb.2014.10.039.
5614.0000030.
[35] Ramaglia G, Lignola GP, Prota A. A simplified approach to evaluate retrofit
[12] Parisi F, Lignola G, Augenti N, Prota A, Manfredi G. Nonlinear behavior of a
effects on curved masonry structures. In: Proceedings of third international
masonry subassemblage before and after strengthening with inorganic
conference on smart monitoring, assessment and rehabilitation of civil
matrix-grid composites. J Compos Constr 2011;15:821–32. http://dx.doi.org/
structures SMAR 2015. Antalya, Turkey; 2015. ISBN: 9783905594652.
10.1061/(ASCE)CC.1943-5614.0000203.
[36] Whitney CS. Design of reinforced concrete members under flexure and
[13] Augenti N, Parisi F, Prota A, Manfredi G. In-plane lateral response of a full-scale
combined flexure and direct compression. ACI J 1937;33:483–98.
masonry subassemblage with and without an inorganic matrix-grid
[37] De Santis S, de Felice G. A fibre beam-based approach for the evaluation of the
strengthening system. J Compos Constr 2010;15:578–90. http://dx.doi.org/
seismic capacity of masonry arches. Earthquake Eng Struct Dynam
10.1061/(ASCE)CC.1943-5614.0000193.
2014;43:1661–81. http://dx.doi.org/10.1002/eqe.2416.
[14] Parisi F, Iovinella I, Balsamo A, Augenti N, Prota A. In-plane behaviour of tuff
[38] Melbourne C, Gilbert M. The behaviour of multiring brickwork arch bridges. J
masonry strengthened with inorganic matrix-grid composites. Compos B Eng
Inst Struct Eng 1995;73:39–47.
2013;45:1657–66. http://dx.doi.org/10.1016/j.compositesb2012.09.068.
[39] Callaway P, Gilbert M, Smith CC. Influence of backfill on the capacity of
[15] Lignola GP, Flora A, Manfredi G. Simple method for the design of jet grouted
masonry arch bridges. Proc ICE – Bridge Eng 2012;165:147–57. http://dx.doi.
umbrellas in tunneling. J Geotechn Geoenviron Eng 2008;134(12):1778–90.
org/10.1680/bren.11.00038.
http://dx.doi.org/10.1061/(ASCE)1090-0241(2008)134:12(1778).
[40] Gago AS, Alfaiate J, Lamas A. The effect of the infill in arched structures:
[16] Fanning PJ, Boothby TE. Three-dimensional modelling and full-scale testing of
analytical and numerical modelling. Eng Struct 2011;33:1450–8. http://dx.doi.
stone arch bridges. Comput Struct 2001;79:2645–62. http://dx.doi.org/
org/10.1016/j.engstruct.2010.12.037.
10.1016/S0045-7949(01)00109-2.