You are on page 1of 14

Engineering Structures 201 (2019) 109803

Contents lists available at ScienceDirect

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

Jointed Masonry Model: A constitutive law for 3D soil-structure interaction T


analysis
Wanda G. Lasciarreaa, Angelo Amorosib, Daniela Boldinic, Gianmarco de Felicea, ,

Marialaura Malenaa
a
Department of Engineering, Roma Tre University, Rome, Italy
b
Department of Structural and Geotechnical Engineering, Sapienza University of Rome, Rome, Italy
c
Department of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Bologna, Italy

ARTICLE INFO ABSTRACT

Keywords: Classical approaches to soil-structure interaction are often characterised by relatively simple constitutive as-
Masonry sumptions for either one or both components of the problem. Such simplified assumptions prove to be appro-
Soil-structure interaction priate for simple soil-foundation cases, while showing all their limits when tackling more complex problems, as
Numerical modelling those involving excavation in the vicinity or beneath historical masonry structures. In such cases, the need for
Constitutive modelling
reliable prediction of the potential damage induced by construction activities on surface structures justifies the
Tunnelling
adoption of more advanced numerical approaches, possibly based on realistic constitutive assumptions for both
soils and masonries, together with an accurate modelling schematisation of the excavation process. In recent
years the Authors have adopted an advanced numerical approach to investigate this issue in the two-dimensional
domain, accounting for the non-linearity and irreversibility of the soil behaviour and schematising the block
masonry structure as a homogenised anisotropic medium. This study extends this approach to three dimensional
conditions, to more realistically account for a number of features, including the possible different relative or-
ientations between the structure and an underground tunnel under construction. The focus in this contribution is
on the modelling of the masonry, here described by a modified version of the Jointed Rock model, an anisotropic
elastic perfectly plastic constitutive model based on a simplified multilaminate approach and implemented in the
commercial code Plaxis 3D. This model takes into account the directional properties of the medium, identifying
the orientation of three planes along which the Mohr-Coulomb yield criterion applies. Here we first describe the
modification introduced in the original model and then illustrate some benchmark numerical examples to va-
lidate it. This is followed by the illustration of a 3D analysis of an idealised tunnelling-structure interaction
problem, aimed at highlighting some of the features of the proposed masonry model.

1. Introduction traction [3,4], or incorporating the discontinuities induced by mortar


joints [5,6] and the complex interaction among the masonry units,
Preserving our monuments is a matter of vital importance since each through a homogenization technique, resulting in an overall anisotropic
one of them is the historical and artistic witness of the evolution of the behaviour [7–11]. Most of these approaches are developed under plane
man, society and territory. Particularly nowadays, this issue should also stress or generalised plane strain conditions, resulting in a negligible
consider the possible interaction with underground infrastructures that influence on the macroscopic elastic behaviour [12], but affecting its
are increasingly constructed in the historical city centres. Consequently, non-linear response [13]. More complex three-dimensional models
it is crucial to implement a methodology for the analysis of coupled have been proposed in the framework of either continuum [14], or
geotechnical and structural problems in order to accurately predict the discrete limit analysis [15–17], mainly addressed to earthquake en-
possible damage induced on historical structures by underground con- gineering applications [18–20]. However, only few approaches con-
structions [1] or excavations [2]. sider the interaction of masonry structures with geotechnical infra-
A considerable number of nonlinear models have been developed structures accounting for both soil nonlinear behaviour and masonry
for masonry, aiming at capturing the essential no-tensile strength under cracking phenomena in 2D [21–23] or 3D analyses [24–27]. The 3D


Corresponding author.
E-mail address: defelice@uniroma3.it (G. de Felice).

https://doi.org/10.1016/j.engstruct.2019.109803
Received 13 October 2018; Received in revised form 13 October 2019; Accepted 14 October 2019
Available online 30 October 2019
0141-0296/ © 2019 Elsevier Ltd. All rights reserved.
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

modelling appears to be a relevant ingredient for investigating the re-


sponse of masonry subjected to the displacement field induced by the
construction of a nearby geotechnical infrastructure. This is especially
true in the case of tunnels, where the progressive excavation induces
subsidence troughs which evolve as the tunnel face advances.
In this work a 3D nonlinear constitutive model is proposed for the
analysis of masonry structures. The model, called Jointed Masonry Model
(JMM), is a modified version of the well-known Jointed Rock Model
(JRM), an anisotropic elastic perfectly plastic constitutive law, based on
a simplified multilaminate approach, originally developed for the de-
scription of the mechanical behaviour of fractured rock-mass [28,29].
The JRM assumes different plane orientations along which a local
Mohr-Coulomb yield criterion with a tension cut-off holds, to represent
the rock-mass discontinuities planes. Similarly to fractured rocks, an-
cient masonry structures are characterised by high strength units, such
as stone blocks or bricks, with weak joints, either dry joints or lime Fig. 1. Global and local coordinate systems in 2D conditions.
mortar joints, that represent the possible discontinuities where the
cracks tend to develop. When subjected to soil settlements or earth- cut-off and a corresponding versatile flow rule, which can be either
quake loading, the behaviour of ancient masonry structures strongly associated or not. The Mohr-Coulomb and tensile-cut off yield functions
depends on the effective arrangement of block units, rather than on the are respectively defined as:
mechanical properties of the blocks themselves, and therefore, similarly
to fractured rocks, the crack pattern is driven by the discontinuity fiC = | s, i| + n, i tan i c0, i for i = 1. .np0 (2.1)
planes represented by the joints.
However, contrary to fractured rocks, the interlocking among ma-
fit = for i = 1. .np0 (2.2)
sonry units deriving from staggered head joints, provides a further n, i t 0, i

strength against the opening of vertical cracks.


Accordingly, the JMM model coincides with the JRM for the treat- where i = 1, ..,np0 with np0 3 is the specific orientation considered.
ment of bed joints that can be simply regarded as discontinuous planes Along each orientation i, n, i and s, i are the normal and shear stresses
with a local Mohr-Coulomb yield condition, while a more refined model (Fig. 1), while c0, i, i and t 0, i are the cohesion, friction angle and tensile
needs to be developed for the head joints to account for the specific strength, respectively, with t 0, i c0, icot i .
features of masonry, stemming from the interlocking of the units. The original JRM can be directly adopted for block masonry bed
As discussed before, more sophisticated material models exist in the joints, represented by the plane orientation 2 in Fig. 2 [2], but not of
literature to describe the nonlinear behaviour of masonry, at the cost of head joints, since those latter are typically arranged in a staggered way
a relatively large number of material parameters and state variables, (Fig. 2 – plane orientation 1), resulting in an increased overall tensile
both hard to be detected or initialised in engineering practice. The and shear strength.
advantages of the JMM consist in the reduced number of material Let us consider the portion of the wall shown in Fig. 3(a), made of
parameters, whose calibration procedure is well-defined and straight- blocks with height a and width b, subjected to a vertical compressive
forward, and in the low computational cost stemming from a balanced stress 2 and to a horizontal tensile stress 1. Each block is subjected to a
compromise between accuracy and simplicity. The constitutive model compressive stress n,2 = 2 acting perpendicularly to the bed joints, a
was implemented in a geotechnical Finite Element code, called Plaxis tensile stress n,1 1 on the head ones, and a tangential stress s,2 on

3D, aiming at analysing soil-structure interaction problems without the bed joints as shown in Fig. 3b, where this latter arises from the
introducing classical oversimplifications for the soil deposits (e.g. friction among the blocks and have opposite sign in the two half sides of
Winkler-based approximations) usually adopted in structural-oriented each single block.
finite element codes. Aiming at incorporating the interlocking effect in a continuous
The paper is organised as follows. In Section 2, the formulation of equivalent head joint and assuming it can be described by the yield
JMM is provided starting from that of the JRM. Then the model is va- conditions (2.1) and (2.2), the macroscopic tensile strength t,1 on the
lidated through comparison with some selected benchmark problems macroscopic head joint, i.e. the ultimate value of n,1, can be evaluated
related to masonry panels under either tensile or shear loading condi- as:
tions, for which analytical solutions or experimental results are avail-
able. Finally, the simulations of the interaction between a tunnel and a
3D masonry building under different geometrical configurations are
presented and discussed to highlight the performance of the proposed
approach.

2. Formulation of JMM constitutive model

2.1. Strength conditions

The Jointed Masonry Model (JMM) was developed as a modified


version of the Jointed Rock Model (JRM), an anisotropic elastic per-
fectly plastic constitutive model developed for the description of frac-
tured rock-masses and already implemented in Plaxis 3D [29]. More
specifically, the JRM is a simplified multi-laminate model characterised
by isotropic elasticity and anisotropic yielding. Its formulation takes
into account a maximum of three different rock-mass discontinuity
planes, along which a Mohr-Coulomb yield criterion holds, with tension Fig. 2. Definition of plane 1 and plane 2 in the JMM.

2
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

σ2
σn,2
τs,2

h σ1 σ1 σn,1 σn,1 a

τs,2

σn,2
σ2
b

Fig. 3. Stress state acting on: (a) a portion of the masonry wall; (b) the single block.

n b
= + (c0,2 n,2tan 2 ) = + c0,2
t ,1 t 0,1
h 2 (2.3) t ,1 t 0,1 n,2
tan 2 (2.9)
where h is the height of the portion of wall while n is the corresponding
number of bed joints. c1 = c0,1 n,2 c0,2 tan 1
Writing the block height as:
tan 2 (2.10)

h It is worth noting that the geometry of the blocks is retained in the


a=
(2.4) model through the parameter. However, given the continuum for-
n
mulation, the failure mechanisms can occur at any point of the domain,
and substituting into Eq. (2.3), it becomes: provided that the strength is mobilised along a direction corresponding
to that selected for the definition of the strength parameters.
b b
= + c0,2 n,2tan 2 No modifications are introduced for the strength parameters of
t ,1 t 0,1
2a 2a (2.5)
joints aligned along direction 2 (i.e. for the bed joints), not interested by
The first term in Eq. (2.5) is the contribution of tensile strength t0,1 the interlocking effect. Remarkably, this direction results as the weakest
on the brick head joints, the second one accounts for the cohesive one, consistently with experimental observations.
contribution c0,2 of the bed joints and the third term refers to the fric- Modern masonry structures can benefit from some tensile and co-
tional contribution tan φ2 along the bed joints due to the compressive hesive strength at the joints, while historical ones are often dry, i.e.
stress n,2. characterised by the absence of mortar, or build up using fragile mor-
Without considering the interlocking of the bricks, i.e. adopting the tars, leading in both cases to null or negligible cohesive strength con-
original Jointed Rock Model, the corresponding ultimate horizontal tribution at the joints. In such circumstances, this latter strength in-
stress would have simply been equal to: gredient should be disregarded, consistently with what proposed in
classical theory of masonry behaviour [4]. Accordingly, Eq. (2.9)
t ,1 = t 0,1 (2.6) simply reduces to:
thus, uniquely corresponding to the tensile strength t0,1 along the t,1 = n,2 (2.11)
head joints.
The increment in the tensile strength of the macroscopic head joints while Eq. (2.10) becomes:
implies a corresponding increment in the available cohesion: c1 = n,2tan 1 (2.12)
b b
c1 = c0,1 n,2tan 2 c0,2 tan
2a 2a 1
(2.7) 2.2. Elastic properties

characterised by a translation of the strength domain along a di-


According to [16], the macroscopic elastic properties are derived
rection parallel to the original one, as illustrated in Fig. 4.
via homogenisation, starting from the elastic properties of blocks
In summary, in the JMM constitutive formulation, the influence of
(Eb, b) and joints (Em, m) , as well as from the brick geometry (a, b) and
the block interlocking is accounted for by a new parameter defined as:
the thickness tm of the mortar joints:
b 1 1 4a 1
= tan = + 2 +
2
2a (2.8) G aKt b Kn + 4abKt µb (2.13)
Accordingly, the Eqs. (2.5) and (2.7) are rewritten as follows:
12 21 b
= =
E1 E2 2(3µb b
+ 2µb2 ) (2.14)

1 4a 1 b + 2µb
= + +
E1 4abKn + b2Kt 4µb 4(3µb b + 2µb2) (2.15)

1 1 1 b + 2µb
= + +
E2 aKn 4µb 4(3µb b + 2µb2) (2.16)
where
Eb
µb =
2(1 + b)

Fig. 4. JMM: Mohr plane representation of the modified strength domain for b Eb
b =
macroscopic head joints (along direction 1). (1 2 b)(1 + b)

3
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Eb Em Table 1
Kn =
tb (Eb Em) Parameters of the JMM for the tensile test, from [30].

1 1 Material properties
K t = µb µ m
t m µb µ m γ (kN/m3) 15
G (kPa) 477,093
Kn and Kt being the normal and the shear joint stiffness, respec- (-) 0.12
tively. (-) 0.687
An equivalent isotropic behaviour can also be adopted for the JMM, c0,1 2 (kPa) 5
with elastic modulus E equal to the average between E1 and E2, and the 1 2 (°) 31
shear modulus G derived from Eq. (2.13). 1 2 (°) 31
t0,1 2 (kPa) 8.32

3. Benchmark numerical examples


Table 2
The examples described in the present section discuss the predictive
Parameters of the JMM for tilting tests, from [30].
capabilities of the proposed model under either tensile or shear loading
conditions. The tests were selected in order to reproduce typical Material properties

benchmark problems for which experimental results, analytical solu- Texture A Texture B
tions or well-established numerical outcomes are available. γ (kN/m3) 15 15
G (kPa) 511,490 477,093
3.1. Horizontal tensile test (-) 0.14 0.12
(-) 1.372 0.687
c0,1 2 (kPa) 0.0 0.0
Here we consider an ideal masonry panel subjected to self-weight γ,
1 2 (°) 31 31
a vertical pressure q applied on top of the wall and a prescribed in- 31 31
1 2 (°)
creasingly horizontal displacement imposed at one side. The panel has a t0,1 2 (kPa) 0.0 0.0
dimension of 1.50 × 0.98 m2 and an out-of-plane thickness of 0.12 m. It
is constituted by 14 courses of bricks, each brick with dimensions of
16 × 7 × 12 cm3 (Fig. 5). Boundary conditions are represented in 32 × 16 × 7 cm3 arranged in two different textures A (a = 7 cm;
Fig. 5. b = 16 cm) and B (a = 7 cm; b = 32 cm), corresponding to factors
The self-weight and the vertical pressure q = 10 kPa are first ap- equal to 1.372 and 0.686 respectively (Table 2) with friction angle
plied, followed by the application of a uniformly increasing horizontal equal to 31°. The panel is subjected to its self-weight, when tilting its
displacement ud . base up to the occurrence of failure. The test aims at representing the
The values of the mechanical parameters adopted in the numerical lateral strength of a wall when subjected to horizontal acceleration,
analysis, considered appropriate for a historical structure, are sum- simulating the effect of in-plane earthquake action. The load factor is
marised in Table 1. The tensile strength on the sliding planes was as- defined as the ratio between the components of the body forces along
sumed to be equal to c0, icot i . direction 2 and 1 (Fig. 2).
The problem admits an analytical solution for the ultimate hor- The ultimate behaviour of the system is analysed for six different
izontal resultant tensile force, according to both static and the kine- wall aspect ratios , defined as ratio between height H and length L of
matic approaches of the limit analysis, as discussed in [30]: the wall, all referring to the same wall area equal to 16 m2. The me-
chanical characteristics of the joints are summarised in Table 2.
b H b b
Flim = H t0 + c0 + tan + q tan Fig. 6 compares the analytical upper bound and the homogenisa-
2a 2 2a 2a (3.1)
tion-based numerical solution provided in [30] to the numerical results,
In this case, it provides a value of Flim = 25.4 kN/m. The corre- as obtained by the JMM. The maximum load factors attained in the
sponding value obtained by the numerical analysis performed with the numerical analyses at failure, max , are summarised in Table 3, as-
JMM is equal to 25.3 kN/m. suming a tolerance of 10−3 in terms of unbalanced forces: a difference
generally lower than 3% emerges comparing the results of the two
3.2. In-plane tilting test different numerical models. Fig. 7a shows the JMM predictions of the
principal tensile strain pattern at failure, which nicely compare to the
This example is inspired by a set of experimental tests described in crack pattern of the reference experimental tests (Fig. 7b).
[31]. The masonry panel is assumed as made by blocks with dimensions As already discussed in [30], the max value decreases with in-
creasing values of for both the textures. For decreasing values of the
value of λ at failure tends toward a constant value. This can be related
to the observed concentration of the tensile strains in a limited portion
of the wall, the upper right sector (Fig. 7), such that the actual width of
the wall itself does no longer influence the failure mechanism. This
latter feature does not hold for larger values of , for which the entire
width of the wall is involved in the strain localisation mechanism as
shown in Fig. 8. The role of the factor can be assessed by comparing
the tensile strain localisation pattern shown in Fig. 8: the smaller the
value of , the more concentrated the tensile strain localisation in a
narrower zone.
The analytical and numerical solutions match satisfactorily in the
case of = 0.687 while tend to diverge for low values of ρ for = 1.37
(Fig. 7). This should be related to the different assumptions on which
the solutions relay: the analytical upper-bound solution was obtained
assuming a rigid block failure mechanism, characterised by a rotation
Fig. 5. Geometrical characteristics of the wall and boundary conditions.

4
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 6. Maximum load factors max attained at failure for in plane tilting test with different aspect ratio.

Table 3 concentrated tensile strain localisation zones. The more concentrated


Maximum load factors max attained at failure for in-plane tilting test. the localised zone, the closer the numerical solution is to the analytical
Wall aspect ratio (ρ) 0.5 1 1.5 2 2.5 3 one: this justifies the accordance of the results for relatively large values
of for both values of , although the better match is observed for the
Texture A Homogenised 0.477 0.461 0.422 0.367 0.312 0.270 = 0.687 case, characterised by a more concentrated localisation
model pattern (Fig. 8).
JMM 0.492 0.466 0.420 0.361 0.310 0.271
Deviation −3% −1% 1% 2% 1% 0%
Texture B Homogenised 0.410 0.410 0.391 0.348 0.301 0.262
3.3. Out-of-plane tilting test
model
JMM 0.401 0.417 0.408 0.362 0.332 0.271
Deviation 2% −2% −4% −4% −10% −3% In this section the capability of the proposed model in reproducing
out-of-plane deformative and failure mechanisms is investigated. The
numerical simulation refers to one of the tests, named S10, conducted
around a point at the base of the wall and the related development of a during the experimental campaign performed in [32], where a series of
discontinuity plane, while the FE solution is based on a continuum static 1:5 scaled dry-stone masonry walls were subjected to tilting. The
approach characterised by the development of either diffuse or more walls were constructed with blocks having dimensions

Fig. 7. Principal tensile strain pattern of the JMM and corresponding experimental results for the tilting test with texture B and = 0.5 (left), = 1(right).

5
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Table 4
Parameters of the JMM for tilting tests [32].
Material properties

γ (kN/m3) 26.8
G (kPa) 461,490
(–) 0.25
(–) 0.93
c0,1 2 (kPa) 0.0
1 2 (°) 35
1 2 (°) 35
t0,1 2 (kPa) 0.0

with the experimental ones obtained in [33] (Fig. 11). The specimen
has dimensions 122 × 122 × 0.48 cm3. The average ratio height b
versus base a of the blocks is equal to 3. The diagonal load was trans-
ferred to the specimen through two steel elements placed on the two
diagonally opposite corners of the panel and modelled as an elastic
material. In the experiments, the strain was calculated through the
measurements of displacements at the four reference points along the
two diagonals (Fig. 11), whose coordinates are reported in Table 5.
During the numerical test, the displacements at the base of the lower
steel element were constrained and the specimen was first subjected to
Fig. 8. Principal tensile strain pattern of the JMM for slender wall ( = 2.5) its weight and then to a displacement orthogonal to the upper steel
having texture A (left) and texture B (right). element equal to 20.00 mm (Fig. 11).
Two families of plane orientations were activated: the first one was
30 × 80 × 40 mm3. The prototype masonry structure (Fig. 9) was inclined at 45° to represent the head joints, and the second one inclined
composed by a main wall, 21 bricks high and 12 bricks wide, and two at −45° to represent the bed joints. The material properties used in the
orthogonal partition walls, 21 bricks high and 10 bricks wide. The numerical analysis are the same as those proposed by the Authors of the
mechanical characteristics of the joints, derived from [32], are sum- experiments (Table 6).
marised in Table 4. Fig. 11 compares the total strain distribution obtained in the nu-
As in the previous section, numerical simulations of the tilting test merical analysis and the crack pattern detected in the experiment,
were carried out constraining the walls at the base, applying gravity showing an overall good agreement, despite in the experimental test a
first followed by a horizontal body force proportional to the self-weight. fully diagonal crack was observed in the direction of the two steel
The load factor attained in the numerical analysis at failure, max , is elements, while in the numerical simulation the strain concentration
0.213, while the experimental one was equal to 0.231. Fig. 10(a) shows pattern is more vertical. This feature was probably due to the activa-
the top view of the deformed shape at failure on half model revealing tion, in the numerical simulation, of both discontinuities along planes 1
the presence of torsional rotations at the base edges of the two ortho- and 2, while the real masonry is made of irregular blocks with no clear
gonal connected walls, as it occurred in experiments. Fig. 10(b)–(c) discontinuity directions.
illustrates the principal tensile strain pattern in the partition walls and The results are illustrated in terms of shear stress-strain ( )
in the main wall, respectively. The figures show that in-plane tensile curve as shown in Fig. 12. Both, the stress and strain components were
strains concentrate on the upper edge of the partition walls where they calculated as average values considering the four points mentioned
start detaching from the main wall, while this latter is affected by before. The results indicate a reasonable agreement between numerical
tensile strains due to out-of-plane bending. and experimental data, though the peak shear strength prediction is
underestimated of about 10%.

3.4. Diagonal compression test


3.5. Shear wall tests
The predictive capability of the proposed model for a diagonal
compression test was investigated by comparing the numerical results As a benchmark example of a shear wall test, the two panels

Fig. 9. Geometrical configuration and failure mechanism for experimental test S10 [32].

6
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 10. JMM simulation of test S10: deformed shape and tensile principal strain pattern.

Fig. 11. Crack pattern obtained in the diagonal compression experimental tests [33] (left) and distribution of total strain in the numerical analysis performed with
the JMM (right).

Table 5
Coordinates of the control points [33].
Point Node x (m) z (m)

1 38,075 0.22 0.24


2 2874 0.94 0.96
3 20,336 0.94 0.26
4 25,415 0.23 0.96

Table 6
Parameters of the JMM for diagonal compression test
[33].
Material properties

γ (kN/m3) 19
G (kPa) 20,000
(–) 0.25
(–) 0.94
c 0,1 2 (kPa) 55
1 2 (°) 32
1 2 (°) 32
t0,1 2 (kPa) 60

(Fig. 13) that were tested in [34] are considered. The panels are 0.99 m
Fig. 12. Shear stress-strain ( ) curve of diagonal compression test.
wide, 1.00 m high, and 0.10 m thick, and are made of wire-cut solid
clay bricks with dimensions 210 × 52 × 100 mm3. The opening, if
present, has a dimension of 400 × 235 mm2. imposing a vertical pressure q equal to 300 kPa and finally applying a
The tests were simulated by first activating the self-weight, then uniform horizontal displacement at the top of the wall (shear condi-
tions), precluding any rotation of its upper edge.

7
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 13. Geometrical characteristics of the analysed walls [34].

The described experimental tests were simulated in [35] through a and, as such, cannot reproduce the final decay of strength. Nonetheless,
discrete model in which the joints were modelled as interface elements the JMM formulation, which is far less complex than the discrete
allowing for crack, slip or crushing, adopting a Mohr-Coulomb friction model, provides a reasonable estimate of the wall lateral strength for
with tension cut-off and a cap for compressive failure. practical applications.
Aiming at providing a direct comparison with the numerical results
discussed in [35], the JMM simulations were carried out adopting 4. Numerical analyses of boundary value problems
constitutive parameters directly related to those assumed in [35], as
summarised in Table 7 where the elastic shear modulus G derives from In the following sections, the response of several ideal masonry
the homogenisation procedure discussed in Section 2.2. structural configurations to shallow tunnelling is analysed. The studied
The comparison of the failure pattern at the target displacement of cases resemble those often occurring during the construction of real
2 mm for the wall without opening, is shown in Fig. 14: the shear strain metro-lines in the centre of historical towns. They include isolated 3D
at the end of the analysis is concentred along the diagonal, in good façades, with or without openings, and a complete 3D building, all
agreement with the crack pattern obtained in [35]. described through the JMM.
In the case of the panel with opening (Fig. 15) at the target dis-
placement of 25 mm, the JMM model reproduces quite reasonably the
4.1. Modelling of tunnel excavation
shear strain localization along the diagonal obtained in [35] with the
panel accurately discretized with interfaces at the joint locations. The
The soil domain, adopted for all the numerical analyses, has di-
JMM model also back-predicts the horizontal cracks that originate from
mensions of 80 × 80 × 20 m3 (Fig. 17). The coordinate system is de-
the vertical sides of the wall, both at the top left pier next to the opening
fined such that x is the distance from the tunnel axis in the horizontal
and at the bottom right side.
direction, z is the depth below the ground surface and y is the direction
Fig. 16 shows the total horizontal shear at the base of the wall as a
of the tunnel axis.
function of the displacement applied at the top of the wall. The curve
The geotechnical conditions and tunnel characteristics are inspired
derived from the JMM-based numerical simulations are compared with
by the case-history of the Milan metro line 5 discussed in [36,37]. The
those observed in the experiments and those resulting by the simula-
soil is a sandy material, described by a linear elastic-perfectly plastic
tions reported in [35]. All the inelastic phenomena occur at the inter-
Mohr-Coulomb model, characterised by the following parameters:
face elements, where the degradation of stiffness and strength is also
γ = 17 kN/m3, E′ = 75 MPa, = 0.3, c′ = 1 kPa (a value slightly larger
accounted for, in contrast to the perfectly plastic formulation of the
than zero is adopted to avoid numerical problems), ' = 31°, = 0°.
JMM, that cannot account for any decay of strength along the joints
The ground-water level is at the ground surface and the pore pressure
distribution is hydrostatic. Drained conditions were assumed for the soil
Table 7
during the whole analysis due to its high permeability.
Parameters of the JMM for shear wall tests, directly re- The tunnel is supposed to be constructed by an EPB-TBM machine,
lated to those of [35]. such that modest induced-subsidence and low impacts on the sur-
rounding structures are expected. These goals are obtained thanks to
Material properties
the pressure applied at the face, the presence of the shield and the
γ (kN/m3) 15 immediate installation of a pre-casted lining ring backfilled with quick-
G (kPa) 1,404,494 setting grouting. The tunnel has a diameter D equal to 8 m and an axis
(-) 0.06 depth of z0 = 9 m below the ground surface (Fig. 17).
(-) 1.514
To reduce the computational cost, for each analysis the tunnel is
c0,1 2 (kPa) 350
assumed as already excavated and lined for the first 20 m. The ex-
1 2 (°) 36.9
36.9 cavation of the following 40 m is simulated step-by-step considering
1 2 (°)
t0,1 2 (kPa) 250 face advancements of 1.5 m, this latter length corresponding to the
longitudinal dimension of the lining ring. In detail, in each calculation

8
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

[
*10-
3]
1,
00

-
0,32

-
1,63

-
2,95

-
4,26

-
5,58

-
6,89

-
8,21

-
10,
84

-
11,
50
Fig. 14. Crack pattern at an applied displacement of 2 mm obtained in [35] (left) and corresponding distribution of total shear strain calculated with the JMM (right).

step the following processes are simulated: excavation of the soil for a that obtained by a Gaussian distribution curve characterised by a width
1.5 m tunnel length, setting of a null pore water pressure in the ex- parameter K equal to 0.25, appropriate for the considered geotechnical
cavated elements, application of a pressure equal to 90 kPa at the new conditions. This latter represents the best empirical approximation of
tunnel face, advancement of the EPB shield of 1.5 m, application of a the expected free-field settlement distribution. The accordance between
pressure simulating the grouting action at the back of the shield and the numerical results and the empirical predictions indicates that the
installation of a new lining ring. This procedure, similar to that dis- relatively simple constitutive assumptions for the soil and the simula-
cussed in [36,37], is aimed at reproducing in a simplified, though tion technique adopted for the tunnel excavation are adequate to the
realistic, way all the 3D complex processes that occur during EPB purpose of the present work.
tunnelling. The generated subsidence is controlled by imposing a value
of surface contraction to the EPB shield equal to 0.5%. For all the
analyses, a more refined mesh was used for tunnel elements as com- 4.3. Masonry wall without openings
pared to the soil ones.
The masonry wall introduced in the tunnel-soil-structure interaction
analysis is 8 m high, 20 m wide and 0.5 m thick. The foundation is
4.2. Free field analysis modelled as a strip footing characterised by a width of 1 m, thickness
equal to 0.5 m and 1 m embedded depth, as suggested in [27] (Fig. 19).
Fig. 18 shows the free-field surface settlements along a transversal The structure develops perpendicularly to the tunnel longitudinal axis
section of the tunnel at a distance of 31.25 m from the origin of the and is located at a distance of 31.25 m from the side of the domain
tunnel axis. The surface volume loss is equal to 0.4%, typical of well (Fig. 17). The mechanical parameters adopted for the masonry wall and
performing tunnelling activity. The settlement profile is compared to foundation are the same of the example in Section 3.1.

Fig. 15. Crack pattern at an applied displacement of 25 mm obtained in [35] (left) and corresponding distribution of total shear strain calculated with the JMM
(right).

9
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 16. Comparison between experimental and numerical results: horizontal reaction versus horizontal displacement (left) wall without opening, (right) wall with
opening.

In the numerical analysis, the gravity of the masonry façade was


activated before the simulation of the tunnel excavation. At the final
stage of tunnel excavation (Fig. 20), the maximum extensional total
strain in the masonry façade reaches a value equal to 1.9%, which
corresponds to a severe damage according to [38]. The extensional
strain distribution in the masonry wall clearly resembles the pattern
described in literature as “short settlement” applied in the middle of the
façade [39]. In this case, in fact, the cracks are inclined at about 45°
towards the centre of the wall, with a higher concentration in the lower
portion of the structure.
The distribution of the vertical displacements at the foundation
level differs significantly from that obtained under free field conditions
(Fig. 21). In particular, while the maximum settlement is nearly the
same, the volume loss is significantly larger, as a consequence of the
key factors affecting this interaction process, namely the weight of the Fig. 18. Comparison between numerical and empirical settlement profiles at
wall and the relative stiffness of the wall-soil system. the ground surface (volume loss of 0.4%).

4.4. Masonry wall with openings [24,25], is shown in Fig. 22.


Lintels are included above all the window and door openings to
The second example discussed is related to a masonry façade with avoid tensile failure in the masonry due to local sagging effects. The
openings. The dimensions of the structure and of the foundation are the mechanical parameters of the wall and the foundation are those
same of the previous example, while the layout of the openings, re- adopted in the previous example, while the lintels are modelled as
producing the scheme of an ideal case proposed in the literature

Fig. 17. Sketch of the soil domain, tunnel direction and wall position.

10
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 19. Dimensions of masonry façade and foundation’s details [27].

Fig. 20. Deformed mesh and extensional total strain distribution in the masonry façade without openings.

the wall without openings, the gravity of the masonry façade was ac-
tivated before the simulation of the tunnel excavation.

4.4.1. Analysis without eccentricity


In this first analysis, the position of the masonry façade is centred
with respect to the tunnel axis (Fig. 23). Due to the presence of the
openings in the façade, the settlement profile at the base of the foun-
dation shows a stepped profile (Fig. 21) if compared to that obtained for
the wall without openings. Apart from this peculiar shape, its average
profile is in agreement with that of the previous example, however it is
characterised by a lower volume loss possibly due to the lower weight
of this structure. The extensional total strains induced in the façade are
shown in Fig. 23: they reach a maximum value of 11.7%, a larger figure
as compared to that observed in the full wall case due to the presence of
the openings. Extensional strains concentrate around the openings, with
a global pattern that is consistent with that observed in the full wall
described before. These results highlight the importance of accounting
Fig. 21. Vertical displacements at foundation level.
for openings when investigating the pattern and magnitude of exten-
sional strains distribution developed in a façade.
elastic elements (G = 3.75E6 kPa, ν = 0.2).
As in the previous case, the structure is oriented perpendicularly to
4.4.2. Analysis with eccentricity
the tunnel axis at a distance of 31.25 m from the origin of the tunnel
A second analysis was carried out considering an eccentric position
axis. It was first located in a symmetrical position with respect to the
of the structure, with its left corner aligned with the tunnel axis
tunnel axis, then an eccentric position was considered. As in the case of
(Fig. 24a). Fig. 24 shows the deformed configuration and the

Fig. 22. Dimensions of the masonry façade with openings and foundation’s details [24,25].

11
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 23. Deformed mesh and extensional total strains distribution in a masonry façade with openings.

extensional strains distribution at the end of the analysis. As a con-


sequence of the tunnel excavation, the wall seems to be subdivided into
three portions, the first one on the left being characterised by an almost
rigid rotation towards the tunnel. In particular, a maximum strain value
of 5.4% is reached at the upper right corner of this portion, about at the
inflection point of the subsidence trough as evaluated at the foundation
intrados (Fig. 21). This is an expected outcome since the deformation
distribution resembles the typical damage observed in terminal settle-
ment foundation problems [39].
The same analysis was repeated assuming a linear elastic model for
the masonry façade (γ = 15 kN/m3, G = 477.7 MPa, I ½ = 0.12) to
evaluate the influence of the constitutive assumptions on the computed
results. In Fig. 26, a comparison between the extensional total strains Fig. 25. Complete mesh of the tunnelling analysis with the 3D masonry
pattern obtained adopting the two constitutive assumptions is pro- building.
posed. At the end of the analysis, when an elastic behaviour is adopted,
the total strain distribution seems to be widespread throughout a large considered conservative.
part of the façade; on the contrary, a very localised total strain pattern
characterises the analysis in which the JMM is employed. The max- 5. Coupled analysis with 3D masonry building
imum value of total strain for the elastic model is equal to 0.06%, two
order of magnitude smaller than that for the case of JMM. The response of a fully 3D masonry building to tunnelling is finally
Accordingly, Fig. 21 highlights that, in the case of the elastic model, presented (Fig. 25). The building is composed by a front and a rear
the profile is characterised by a significantly reduced maximum set- façade and two end walls [24,25].
tlement and deflection ratios in sagging and hogging as compared to The two facades and the foundations have the same geometrical
that observed when the non-linear behaviour of the masonry is taken characteristics of the masonry façade with openings analysed in the
into account. These results highlight that the usually adopted partially previous section, while the end walls have no openings and are char-
coupled approach, in which the soil-structure interaction problem is acterised by the same thickness of the façades and a length equal to
computed schematising the structure as an elastic medium to, then, 10 m (Fig. 25). Mechanical parameters of the walls are the same
carry out a more refined non-linear structural analysis imposing the adopted in the previous examples.
previously obtained displacement field, can highly underestimate the To emphasise the capability of the adopted 3D modelling approach,
deformation and damage of the structure and as such in cannot be the building was set eccentric and not perpendicular to the tunnel, as

Fig. 24. Deformed configuration (a) and extensional total strains distribution at the end of the tunnelling analysis (b) performed with JMM (left) and Linear Elastic
Model (right).

12
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

Fig. 26. Extensional total strains distribution at the end of the tunnelling analysis performed with JMM (left) and Linear Elastic Model (right).

Fig. 27. Extensional total strains distribution in the front (left) and the rear (right) façade performed with JMM.

shown in Fig. 25: the tunnel axis passes by the midpoint of the left end
wall, forming an angle of 30°.
After the activation of the gravity of the masonry façade, the si-
mulation of the tunnel excavation was carried out. Fig. 26 shows the
extensional strain distribution on the four sides of the building at the
end of tunnelling. More details are provided in Fig. 27 for the front and
the rear façade, where damage concentrates. As shown in Fig. 27,
tensile strains tend to localise at different locations along the façades,
shifting from the left part of the front façade to the central part of the
rear one, at similar distances from the tunnel axis.
A maximum tensile strain equal to 4.5% was found on the front
façade. This value is slightly smaller than that obtained for the same
masonry façade modelled perpendicular to tunnel axis in an eccentric
position (Figs. 23 and 24).
Also in this case, a further analysis was carried out describing the
masonry building as an elastic continuum (γ = 15kN/m3,
G = 477.7 MPa, 12 = 0.12). A maximum value of total strain equal to
0.05% was obtained in this case on the front façade at the end of the
analysis. As in the previous example, at the end of the tunnel excavation Fig. 28. Vertical displacements at foundation level of the front façade.
the total strain distribution on the façades results diffused and wide-
spread while, for the nonlinear masonry modelling case, a far more
presented in this paper. The model captures the essential no-tensile
localised total strain patter was obtained.
strength and anisotropy of masonry deriving from the shape and ar-
Fig. 28 highlights that the settlement profiles along the two facades
rangement of the masonry units, with a limited number of material
are significantly affected by the adopted constitutive hypotheses for the
parameters whose calibration strategy is straightforward. The pre-
masonry. Linear elasticity is associated to considerably lower maximum
dictive capabilities of the constitutive model were validated under
settlements and practically null deflection ratios in sagging and hog-
tensile and shear loading conditions by numerical simulations of se-
ging. On the contrary, the JMM-based analyses lead to more complex
lected benchmark problems and experiments.
subsidence profiles due to the effect of the localised deformations.
The overall performance of the model is satisfactory, given its re-
latively simple formulation. However, some features of the mechanics
6. Conclusions of masonry structures, as for example their out of plane behaviour, are
only poorly accounted for by JMM and, as such, deserve further pos-
A nonlinear constitutive model for the analysis of masonry struc- sible developments. Along this line, it is also worth noting that the
tures, called Jointed Masonry Model (JMM), was developed and

13
W.G. Lasciarrea, et al. Engineering Structures 201 (2019) 109803

model was developed referring to a single leaf masonry panel with [12] Anthoine A. Homogenization of periodic masonry: plane stress, generalized plane
running bond pattern. Therefore, at the present stage the JMM is not strain or 3D modelling? Commun Numer Methods Eng 1997;13(5):319–26.
[13] Massart T, Peerlings R, Geers M, Gottcheiner S. Mesoscopic modeling of failure in
able to describe the behaviour of multi-leaf walls, since splitting of the brick masonry accounting for three-dimensional effects. Eng Fract Mech
two leaves is not accounted for. This limitation could be addressed in 2005;72(8):1238–53.
future, extending the number of active failure planes from two to three [14] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry walls. Part
I: failure surfaces. Comput Struct 2006;84(3):166–80.
and consistently orienting the third plane along the inter-leaf surface. [15] Livesley RK. Limit analysis of structures formed from rigid blocks. Int J Numer
The non-linear behaviour of the block units is not considered in this Methods Eng 1978;12(12):1853–71. https://doi.org/10.1002/nme.1620121207.
work, as mainly focused on the discontinuities induced by the joints. [16] Portioli F, Casapulla C, Gilbert M, Cascini L. Limit analysis of 3D masonry block
structures with non-associative frictional joints using cone programming. Comput
This assumption is appropriate for structures having weak mortar joints Struct 2014;143:108–21.
and subjected to low compressive stresses. In such a case, the behaviour [17] Malena M, Portioli F, Gagliardo R, Tomaselli G, Cascini L, de Felice G. Collapse
of the units can be considered as elastic, being the joints the main mechanism analysis of historic masonry structures subjected to lateral loads: a
comparison between continuous and discrete models. Comput Struct
source of non-linearity. However, as a further step the model could be
2019;220:14–31.
further improved to include an isotropic strength criterion accounting [18] Noor-E-Khuda S, Dhanasekar M, Thambiratnam DP. An explicit finite element
for the possible crushing of masonry blocks under compression. modelling method for masonry walls under out-of-plane loading. Eng Struct
The JMM model was implemented in a geotechnical Finite Element 2016;113:103–20.
[19] de Felice G, De Santis S, Lourenço P, Mendes N. Methods and challenges for the
code and used to investigate soil-structure interaction problems, aiming seismic assessment of historic masonry structures. Int J Archit Herit
at overcoming either the classical Winkler-based approximations for 2017;11(1):143–60.
soils, when modelling masonry structures, or the equivalent plate ap- [20] Roselli I, Malena M, Mongelli M, Cavalagli N, Gioffrè M, De Canio G, et al. Health
assessment and ambient vibration testing of the “Ponte delle Torri” of Spoleto
proximations for buildings, when modelling soil excavations. The effect during the 2016–2017 Central Italy seismic sequence. J Civil Struct Health Monitor
of tunnel excavation under different masonry walls configurations was 2018;8(2):199–216.
investigated, taking into account the decrease in stiffness induced by [21] Liu G, Houlsby GT, Augarde CE. 2-dimensional analysis of settlement damage to
masonry buildings caused by tunnelling. Struct Eng 2000;79(1):19–25.
openings and damage and their effects on the settlement profiles. The [22] Amorosi A, Boldini D, De Felice G, Malena M. Tunnelling-induced deformation on a
proposed approach allowed to directly evaluate the tunnelling-induced masonry structure: a numerical approach. Geotechnical aspects of underground
damage occurring in the structures, which can result as significantly construction in soft ground Proceedings of the 7th international symposium on
geotechnical aspects of underground construction in soft ground. 2012. p. 353–9.
larger than that predicted adopting a linear elastic constitutive model [23] Giardina G, Van de Graaf AV, Hendriks MAN, Rots JG, Marini A. Numerical analysis
for the masonry, this latter being the customary assumption in en- of a masonry façade subject to tunnelling-induced settlements. Eng Struct
gineering practice to tackle similar soil-structure interaction problems. 2013;54:234–47.
[24] Burd HJ, Houlsby GT, Augarde CE, Liu G. Modelling tunnelling-induced settlement
of masonry buildings. Proc Inst Civ Engr – Geotech Eng, 2000;143(1):17–29, http://
Declaration of Competing Interest dx.doi.org/10.1680/geng.2000.143.1.17.
[25] Pickhaver JA, Burd HJ, Houlsby GT. An equivalent beam method to model masonry
The authors declare that they have no known competing financial buildings in 3D finite element analysis. Comput Struct 2010;88:1049–63.
[26] Giardina G, Ritter S, DeJong MJ, Mair RJ. Modelling the 3D brittle response of
interests or personal relationships that could have appeared to influ- masonry buildings to tunnelling. Structural analysis of Historical Constructions:
ence the work reported in this paper. Anamnesis, diagnosis, therapy, controls - Proceedings of the 10th international
conference on structural analysis of historical constructions SAHC. 2016. p. 481–8.
[27] Yiu WN, Burd HJ, Martin CM. Finite element modelling for the assessment of
References tunnel-induced damage to a masonry building. Géotechnique 2017;67(9):780–94.
[28] Godman RE, Taylor RL, Brekke TL. A model for the mechanics of jointed rock. J Soil
[1] Amorosi A, Boldini D, de Felice G, Malena M, Sebastianelli M. Tunnelling-induced Mech Found Divis 1968;94(3):637–60.
deformation and damage on historical masonry structures. Géotechnique [29] Binkgreve RBJ, Engin E, Swolfs WM. Plaxis 3D Reference Manual, Delft, The
2014;64(2):118–30. Neaderlands, Plaxis bv; 2013.
[2] Amorosi A, Boldini D, de Felice G, Lasciarrea WG, Malena M. Analisi geotecnica e [30] de Felice G, Amorosi A, Malena M. Elasto-plastic analysis of block structures
strutturale del Ninfeo di Genazzano. Rivista Italiana Di Geotecnica through a homogenization method. Int J Numer Analyt Methods Geomech
2015;49(1):29–44. 2010;34(3):221–47.
[3] Heyman J. The stone skeleton. Int J Solids Struct 1966;2(2):249–79. https://doi. [31] Ceradini V. Modellazione e sperimentazione per lo studio della struttura muraria
org/10.1016/0020-7683(66)90018-7. storica. Ph.D. Thesis, University Rome ‘La Sapienza’; 1992 [in Italian].
[4] Angelillo M. Constitutive relations for no-tension materials. Meccanica [32] Restrepo LF, Magenes G, Griffith MC. Dry stone masonry walls in bending Part I:
1993;28(3):195–202. static tests. Int J Archit Herit 2014;8(1):1–28.
[5] Nazir S, Dhanasekar M. Modelling the failure of thin layered mortar joints in ma- [33] Borri A, Corradi M, Galano L, Vignoli A. Analisi Sperimentali e numeriche per la
sonry. Eng Struct 2013;49:615–27. valutazione della resistenza a taglio delle murature. Ingegneria Sismica
[6] Nazir S, Dhanasekar M. A non-linear interface element model for thin layer high 2004;3:50–68.
adhesive mortared masonry. Comput Struct 2014;144:23–39. [34] Raijmakers TMJ, Vermeltfoort ATh. Deformation controlled tests in masonry shear
[7] de Buhan P, de Felice G. A homogenization approach to the ultimate strength of walls. Research report TNO-Bouw 1992; Report B-92-1156 [in Dutch].
brick masonry. J Mech Phys Solids 1997;45(7):1085–104. [35] Lourenco PB. Computational strategies for masonry structures PhD thesis Delft
[8] Cecchi A, Sab K. A multi-parameter homogenization study for modeling elastic University Press; 1996.
masonry. Eur J Mech – A/Solids 2002;21(2):249–68. https://doi.org/10.1016/ [36] Fargnoli V, Boldini D, Amorosi A. Twin tunnel excavation in coarse grained soils:
S0997-7538(01)01195-0. observations and numerical back-predictions under free field conditions and in
[9] Zucchini A, Lourenço PB. A micro-mechanical model for the homogenisation of presence of a surface structure. Tunn Undergr Space Technol 2015;49:454–69.
masonry. Int J Solids Struct 2002;39(12):3233–55. https://doi.org/10.1016/S0020- [37] Fargnoli V, Gragnano CG, Boldini D, Amorosi A. 3D numerical modelling of soil–-
7683(02)00230-5. structure interaction during EPB tunnelling. Géotechnique 2015;65(1):23–7.
[10] Marfia S, Sacco E. Multiscale damage contact-friction model for periodic masonry [38] Boscardin MD, Cording EJ. Building response to excavation-induced settlement. J
walls. Comput Methods Appl Mech Eng 2012;205–208(1):189–203. Geotech Eng 1989;115(1):1–21.
[11] de Felice G, Malena M. Failure pattern prediction in masonry. J Mech Mater Struct [39] Mastrodicasa S. Dissesti statici delle strutture edilizie. Milan, Italy: Hoepli Editore;
2019. https://doi.org/10.2140/jomms.2019.14.620. 1993. in Italian.

14

You might also like