Engineering Structures 89 (2015) 111–129

Contents lists available at ScienceDirect

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

Sensitivity study on tunnelling induced damage to a masonry façade

Giorgia Giardina a,⇑, Max A.N. Hendriks b,c, Jan G. Rots b
a
Department of Engineering, University of Cambridge, Trumpington Street, CB2 1PZ Cambridge, UK
b
Department of Structural Engineering, Norwegian University of Science and Technology, Rich. Birkelandsvei 1A, 7491 Trondheim, Norway
c
Department of Structural Engineering, Delft University of Technology, Stevinweg 1, 2628 CN Delft, The Netherlands

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

Article history: Assessing the potential damage caused by soil subsidence to masonry structures is essential for the suc-
Received 19 March 2014 cessful realisation of underground projects in urban areas. The damage assessment procedures need to
Revised 20 January 2015 take into account the highly non-linear behaviour of the structural materials, characterised by brittle
Accepted 22 January 2015
cracking and consequent stress redistribution, and the important effect of soil–structure interaction. This
Available online 24 February 2015
paper presents the results of a sensitivity study performed on a 2D finite element model that was
validated through comparison with experimental results. The study investigates the effect of openings,
Keywords:
material properties, building weight, initial damage, normal and shear behaviour of the base interface
Damage assessment
Masonry façade
and applied settlement profile. The results assess quantitatively the major role played by the normal
Parametric analysis stiffness of the soil–structure interaction and by the material parameters defining the quasi-brittle
Settlement masonry behaviour.
Tunnelling Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction different solutions to preserve the prototype relationship between

This paper investigates the effect of selected structural and
geotechnical parameters on the response of masonry buildings
subject to tunnelling-induced settlements.
Tunnelling activity in urban areas is growing, with a consequent
higher risk of induced ground settlements that can cause damage
on existing surface structures. The prediction of damage risk and
the design of mitigation measures to protect the building, when
necessary, requires a complete understanding of the interaction
between the structure and the subsiding soil. The mutual influence
of building and ground responses is made more complex by the
fact that both the masonry and the soil exhibit a non-linear rela-
tion between the applied displacements and the induced
deformations.
Observations deduced from measurements of soil and structure
displacements in actual tunnelling projects [1] has been essential
to validate and improve empirical-analytical [2,3] and numerical
based [4] assessment methods. These methods have been included
in complete staged procedures that represent the most common
approaches currently used in practice to predict the potential
building damage [5,6].
The simplification of the problem to physical models that
can be tested under controlled laboratory conditions is another
fundamental method of investigation. Researchers adopted
material strains and stresses in scaled model testing: the use of
geotechnical centrifuges [7], the controlled reduction of material
strength [8] or the amplification of static loads [9]. In addition to
offering a deeper insight into the simulated mechanisms,
experimental tests can be used to validate numerical models
[6,10]. By including non-linear constitutive laws to describe the
material behaviour, numerical models enable to reproduce the
response of the structure and its interaction with the ground
[6,10–15].
In this research, a numerical model previously validated
through comparison with experimental results is used to investi-
gate a wide range of possible scenarios. A sensitivity study per-
formed on a 2D finite element model of a façade subject to
differential settlements aims to assess quantitatively the influence
of openings, material parameters, loading conditions, interaction
properties and settlement troughs (Fig. 1). An important outcome
of this research is the quantitative assessment of the structural
damage in terms of cracking and distortions, as a function of the
parameter variations. These results can be used to develop a com-
prehensive damage classification system that correlates the main
building characteristics with the risk of being damaged by a certain
ground deformation [16].
2. Numerical model

⇑ Corresponding author. The adopted 2D numerical model reproduces the experiment

E-mail address: gg376@eng.cam.ac.uk (G. Giardina). described in [9] and it is shown in Fig. 2. In the laboratory test,

http://dx.doi.org/10.1016/j.engstruct.2015.01.042
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
112 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

a 1/10th scaled masonry façade was subjected to self-weight and

Physical problem
to a vertical load of 12.2 kN, distributed in 35 application points,
Tunnelling induced damage which replicate the prototype stress field. Single bricks were
on masonry buildings removed at the location of the applied loads and wooden lintels
Simplification were inserted above the openings. The façade was supported by
a H-shaped steel profile with two hinges and one fixed constraint.
Experimental testing A controlled hogging deformation was applied to the structure by
Laboratory model of a pulling downwards the left end of the steel profile. A rubber layer
simulating the soil–structure interaction was inserted between the
Interpretation

masonry façade
Validation

subjected to settlement façade and the steel beam.

In the finite element model, the façade was modelled by 8-node
Idealization
plane stress elements with 3  3 points Gaussian integration
Numerical analysis schemes. A coaxial rotating crack model was applied to the mason-
ry and a linear tension softening relation was assumed after crack-
Finite element model of a
ing. The tension softening law was defined by the tensile strength
masonry façade subjected
to settlement f t , the fracture energy Gf and the crack bandwidth h, which is relat-
ed to the element size and equal to 11.8 mm. No damage criterion
Generalization was assumed in compression and 6-node line interface elements
characterised by no tension, assigned stiffness in compression
Sensitivity study and negligible stiffness in shear were inserted between the façade
Parametric analysis on and the steel profile. The material parameters are listed in Table 1.
the numerical model
•Amount of openings 2.1. Model validation
•Building material
•Building weight Figs. 3 and 4 shows the results of the model validation in terms of
•Initial damage
displacements of selected points and in terms of crack pattern. The
•Soil-structure interface
parameters numerical and experimental curves match well throughout the
•Settlement profile entire simulation, suggesting the model ability to capture the settle-
ment reduction due to the interface effect. Furthermore, the model is
Application able to reproduce all the main cracks leading to the failure mechan-
isms described along with the experimental results. Further details
Operational tool about the numerical model and its validation can be found in [10].
Vulnerability assessment
framework 2.2. Mesh and step size dependency
Fig. 1. Paper content (in red) and research overview. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of To evaluate the influence of the model discretization on the
this article.) response, a mesh refinement study was performed. The reference

1428
50

F1 F2 F2 F1 F1 F2 F2 F1

non-linear
plane stress element

F3 F4 F4 F3 F3 F4 F4 F3

13.3

10.5
1186
F6 F7 F7 F8 F6 F7 F7
F5 F5

lintel

interface F9 F9
F 10 F 10 F 10 F 10 F 10 F 10 F 10 F 10
A
S
beam
32
96 1074 415 115
1700

Fig. 2. Model of the scaled masonry façade. All dimensions are in mm.
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 113

Table 1 material (low Gf and substantial amount of elastic energy storage).

Material parameters for the numerical model.
In the reference case, assuming a convergence criterion based on
Material Parameter Value internal energy with a tolerance of 1  104 and a maximum of
Masonry Young’s modulus 9
E ¼ 3  10 N=m 2 1  103 iterations, the convergence is reached in 22 out of 23 load-
Density q ¼ 1:9  103 kg=m3 ing steps. In order to investigate the influence of load step size, the
Poisson’s ratio m ¼ 0:2 displacement was first applied in 23 steps of 0.5 mm and then in
Tensile strength f t ¼ 0:1  106 N=m2 46 steps of 0.25 mm; no load step size dependency was observed
Fracture energy Gf ¼ 10 N=m
(Fig. 5).
Lintel wood Young’s modulus Ew ¼ 11  109 N=m2
Density qw ¼ 0:5  103 kg=m3 2.3. Scale effect
Poisson’s ratio mw ¼ 0:15
Beam steel Young’s modulus Es ¼ 2:1  1011 N=m2 To verify the possibility to extend to real buildings the numer-
Density qs ¼ 7:85  103 kg=m3 ical results derived from the scaled model, the true-scaled numer-
Poisson’s ratio ms ¼ 0:3 ical analysis was performed, now excluding the amplification loads
Base interface Normal stiffness kn ¼ 0:7  109 N=m3 replicating the self weight. The outcomes in terms of displace-
Tangent stiffness kt ¼ 1 N=m3 ments of control points and maximum principal strains show a
Tensile strength f t;b ¼ 0 N=m2 substantial agreement between the two models (Figs. 7 and 8).

mesh, made by 8-node plane stress elements with dimensions
10:5  13:3 mm2 (Fig. 2) was compared with a finer mesh of 8-n-
ode elements with dimensions 5:25  6:65 mm2 and a mesh of 4-
node elements with dimensions 10:5  13:3 mm2 . Fig. 5 shows
the displacements of selected points for the three different models,
while Fig. 6 compares the maximum principal strain distributions.
No significant differences were detected in the façade response,
proving that the original mesh refinement was sufficient.
A critical aspect of the incremental–iterative scheme applied to
cracking models is represented by convergence difficulties, which
especially arise in case of large-scale structures of very brittle
3. Identification and selection of parameters and variations
This section describes the parameters analysed in the 2D sensi-
tivity study.
3.1. Opening ratio
The opening ratio as defined as the area of apertures over the
total area of the façade was experimentally recognised as influenc-
ing the final damage and therefore varied in the numerical tests.
Three values of percentage of openings with respect to the wall

14 14
B
12 12
C
Displacement C (mm)
Displacement B (mm)

10 10

8 8
A A
6 6

4 4
experimental experimental
2 numerical 2 numerical

0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14

Applied displacement A (mm) Applied displacement A (mm)

(a) (b)
Fig. 3. (a) Vertical displacement of point B and (b) horizontal displacement of point C as a function of the applied settlement: comparison between numerical and
experimental results.

(a) (b)
Fig. 4. Comparison between (a) deformed mesh, multiplied by a factor of five and (b) experimental crack pattern at 10 mm of applied vertical displacement.
114 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

14 14
B
12 12 C

Displacement C (mm)
Displacement B (mm)
10 10

8 8 A
A
6 6
reference model reference model
4 4
finer mesh finer mesh
linear elements linear elements
2 Δ s = 0.5 mm 2 Δ s = 0.5 mm
Δ s = 0.25 mm Δ s = 0.25 mm
0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Applied displacement A (mm) Applied displacement A (mm)
(a) (b)
Fig. 5. (a) Vertical displacement of point B and (b) horizontal displacement of point C as a function of applied settlement: influence of the mesh and step size.

compression
elastic tension
partially open crack
fully open crack

(a) reference model (b) finer mesh (c) linear elements

Fig. 6. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different mesh sizes.

0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140

14 140 14 140
B
12 120 12 C 120
Displacement C (mm)
Displacement B (mm)

10 100 10 100

8 80 8 A 80
A
6 60 6 60

4 40 4 40

2 1/10th scale 20 2 1/10th scale 20

real scale real scale
0 0 0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Applied displacement A (mm) Applied displacement A (mm)
(a) (b)
Fig. 7. (a) Vertical displacement of point B and (b) horizontal displacement of point C as a function of applied settlement: scale effect.

1m
compression
elastic tension
1m partially open crack
fully open crack

(a) 1/10th scale (b) real scale

Fig. 8. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement: scale effect.
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 115

elastic E = 1000 MPa 0.9 ft = 0.1 MPa *

Stress (MPa)

Stress (MPa)

Stress (MPa)
Gf = 50 N/m E = 3000 MPa * ft = 0.3 MPa
0.1 Gf = 10 N/m * 0.1 E = 9000 MPa
ft = 0.9 MPa

0.3

0.1
0 0.017 0.085 0 0.017 0.0171 0 0.017 0.05 0.15
Strain Strain Strain
(a) fracture energy (b) Young’s modulus (c) tensile strength

Fig. 9. Assumed stress–strain relationships for the variations of masonry parameters (⁄ indicates the reference case).

0 5

Horizontal displacement (mm)

hogging
Vertical displacement (mm)

sagging
mixed
−5

−10 0

−15
hogging
sagging
mixed
−20 −5
0 0.5 1 1.5 0 0.5 1 1.5
Façade horizontal coordinate (m) Façade horizontal coordinate (m)
(a) vertical displacements (b) horizontal displacements

Fig. 10. Hogging, sagging and mixed hogging–sagging displacement profiles applied to the façade in the sensitivity study.

surface were considered: 0%, 10% and 30%. The last one is in agree- settlement profile. The assumed sagging profile is illustrated in
ment with the experiment described in [9] (reference case). Fig. 10a (sagging curve).

3.2. Material parameters 3.5. Soil–structure interaction

To simulate the different types of masonry, a range of realistic The sensitivity study also investigated the effect of the normal
values was assumed for the parameters defining the material prop- and shear behaviour of the interface as a function of the foundation
erties: Young’s modulus, tensile strength and fracture energy. The and soil type and conditions. In particular, first the effect of the
combinations under analysis are summarised in Fig. 9. The selected interface normal stiffness was investigated. With reference to the
stiffness assumed in the experimental model, three different values
values cover a wide range of scenarios, from historic masonry to
new constructions [17]. An exception is the highest value of f t , were considered: kn ¼ 0:7  107 N=m3 ; kn ¼ 0:7  108 N=m3 and
which is unrealistic for masonry and virtually represents a material kn ¼ 0:7  109 N=m3 (reference case). The reference value corre-
with no failure in tension. When varying E and f t ; Gf was varied sponds to the normal stiffness equivalent to a Dutch wooden pile
accordingly (Fig. 9b and c), in order to keep constant the slope of foundation [18].
the softening branch. The background is that the slope of the soft- The influence of the interface behaviour in the tangential direc-
ening branch modifies the strain localisation rate. In this way, the tion was also analysed. For this purpose, the reference case with no
effect of changing E and f t is corrected for this influence. horizontal displacements applied was compared with two other
cases where the model was subjected to the horizontal displace-
ment profile shown in Fig. 10b (hogging curve). Two different val-
3.3. Building weight ues of tangential stiffness were assumed for the interface:
kt ¼ 0:7  108 N=m3 and kt ¼ 0:7  109 N=m3 ; in both cases a Cou-
In the tested reference case, vertical amplified loads were
lomb friction criterion was adopted, with cohesion c ¼ 0 MPa, fric-
applied to the façade before imposing the hogging settlement pro- tion angle / ¼ 30 and dilatancy angle w ¼ 0 (non-associated
files. In order to take into account the effect of initial loading on the plasticity).
settlement-induced damage, the vertical load amplification was
varied: the reference case (self-weight plus vertical loads) was
compared to the situation where no load was applied before the 3.6. Settlement profile
settlement and to the case with self-weight only. As for the refer-
ence case, the initial loads were kept constant during the settle- Finally, the sensitivity study was extended to different types of
ment application. settlement profiles; this simulates the tunnelling-induced settle-
ment troughs that affect structures in different locations with
respect to the tunnel (Fig. 10). First, the hogging situation is consid-
3.4. Initial damage ered, where the eccentricity between the building and the tunnel is
sufficient to locate the building outside the inflexion point of the
To analyse the effect of possible initial damage of the structure, greenfield profile. Second, the building is ideally placed in a sym-
the reference hogging profile was applied to the façade model that metric position right above the tunnel centreline (zero eccentricity,
was preliminary damaged by doubled vertical loads or by a sagging sagging profile). Finally, an intermediate location between the first
116 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

two cases is chosen, resulting in a mixed profile that includes the Table 2
point of inflection. Fig. 10a reproduces the vertical displacements Damage classification of masonry buildings as a function of the maximum crack
width.
of the soil surface induced by the tunnelling in the three selected
locations, while Fig. 10b shows the corresponding horizontal dis- Damage level Damage class Crack width (mm) Tensile strain (%)
placements. For both the sagging and mixed settlement troughs, 1 Negligible up to 0.1 0–0.05
the same variations of interface horizontal behaviour previously 2 Very slight up to 1 0.05–0.075
3 Slight up to 5 0.075–0.15
applied to the hogging profile were considered: kt ¼ 0 N=m3 (no
4 Moderate 5–15 0.15–0.3
horizontal displacements), kt ¼ 0:7  108 N=m3 and 5 Severe 15–25 >0.3
9 3 6 Very severe >25 >0.3
kt ¼ 0:7  10 N=m .

4. Results each single variation. The comparison is visualised as the ratio

between the numerical and LTSM damage levels (damage level
For all the examined variations, the settlement-induced struc- ratio). For values of damage level ratio <1 the LTSM is less conser-
tural response was evaluated by comparing the global deforma- vative than the numerical analysis. It should be emphasised that
tions and the crack patterns resulting from the parametric the classification of damage for real buildings is a complex evalua-
analysis. The horizontal strain, angular distortion and maximum tion that requires judgement and experience; the crack width
crack width as a function of the progressive applied deflection ratio therefore represents only one factor in assessing the final building
D=L were used to quantify the structural damage. The definition of damage.
each damage indicator is reported below, and illustrated in Fig. 11 Horizontal strains were included in the LTSM only in hogging
[6]. cases with horizontal displacements applied to the structure. Dif-
Top horizontal strain: ferent values for the equivalent stiffness ratio E=G in the LTSM
DxD  DxC were assumed in the variations of opening percentage, extrapolat-
eh;top ¼ ð1Þ ing them from Burland and Wroth’s guidelines [2] as reported by
L
Son and Cording [19]:
Base horizontal strain:
DxA  DxB E=G ¼ 2:6 for the wall with 0% of openings;
eh;base ¼ ð2Þ
E=G ¼ 8 for the façade with 10% of openings;
L
E=G ¼ 11 for the façade with 30% of openings.
Slope:
DyA  DyB Alternative ways to report the damage have been proposed by
s¼ ð3Þ
L different authors. For example, in their case studies Clarke and Lae-
Tilt: fer [20] used the amount of loose, exposed and repaired brickwork
as indicators of the building physical conditions [21], while
DxC  DxB Nghiem et al. [22] adopted the total length of opened cracks to
h¼ ð4Þ
H quantify the damage in wood physical models.
Angular distortion:
4.1. Opening ratio
b¼sh ð5Þ
Note that the deflection ratio D=L refers to the settlement pro- Fig. 12 shows the maximum principal strain distribution and
file applied to the lower side of the base interface, while the angu- the deformed configuration at the maximum applied displacement
lar distortion b indicates the consequent structural deformation, of 11.5 mm (end of the experimental test) for the considered val-
measured on the façade. ues of opening percentage. The small rectangular holes in all three
The maximum crack width was derived from the relative dis- models indicate the additional vertical load applications, used in
placements between two nodes located outside the opposite mar- the scaled experiment to amplify the gravity: both in the experi-
gins of the most cracked area. Furthermore, the final damage in ment and in the model they work as imperfections in the façade.
terms of maximum crack width was translated into the corre- The contour plots indicate a strong localisation of the damage at
sponding damage class, according to [34] (Table 2). This allowed the corner of the openings or at the imperfections, where the
for a direct comparison of the final assessment with the result of cracks that define the failure mechanism are concentrated. This
the application of the Limiting Tensile Strain Method (LTSM) to is consistent with previous results [6,12].

Fig. 11. Damage indicators used to compare the numerical results.

 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 117

compression
elastic tension
partially open crack
fully open crack

(a) 0% (b) 10% (c) 30%

Fig. 12. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different values of opening percentage.

−4 −3
x 10 x 10
0.01 6 8
no openings no openings no openings
10% openings 10% openings 6 10% openings
Base horizontal strain
Top horizontal strain

0.008 30% openings 30% openings 30% openings

Angular distortion
4
4
0.006 2
2
0.004 0

−2
0
0.002
−4

0 −2 −6
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10
(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

8 6
no openings 5
Maximum crack width (mm)

no openings
10% openings 6 10% openings 4
30% openings 3
Damage level ratio

6 30% openings
5
Damage level

2
4
4 1
3

2 1/2
2 no openings
1/3
1 1/4 10% openings
1/5 30% openings
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10
(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 13. Variation of opening percentage: damage indicators as a function of the applied deflection ratio.

In the reference case, the first bending crack arises at the top of sible to identify the brittle crack initiation (compare with 13d).
the façade, and progressively crosses the entire section in the ver- Around D=L ¼ 103 for the 10% façade opening ratio and
tical direction (Fig. 12c). Conversely, in the blind wall the increased D=L ¼ 2  103 for the blind wall, a main horizontal crack arises,
structural stiffness reduces the initial bending, and the main crack leading to an accelerated anticlockwise rotation of the façade
develops horizontally, near the base (Fig. 12a). In the intermediate upper part, and to a simultaneous lowering of the right-bottom
case, the failure mechanism presents both the horizontal and ver- corner part of the façade. This explains the rapid variation of angu-
tical cracks, but limited to the area around the largest window at lar distortion in the two curves (Fig. 13c) and the consequent nega-
the ground floor (Fig. 12b). tive values assumed by b.
The damage indicators increase with the increase of openings Fig. 12 shows how the relatively high stiffness of the blind wall
(Fig. 13). In particular, the top horizontal strain (Fig. 13a) is the and the wall with the small openings leads to gapping in the no-
deformation parameter that clearly indicates the bending failure tension interface, while in the reference case the façade follows
mechanism in a hogging area, and it reflects the maximum crack more closely the applied settlement trough. The damage level cor-
width evolution (Fig. 13d). Fig. 13c shows that for the reference responding to the maximum crack width growth (Fig. 13e) con-
case also the angular distortion b is larger than in the other two firms that for the analysed situation a façade with a larger
models; the structure adapts to the imposed deformation due to opening ratio is more prone to the damage induced by the hogging
the initial flexibility induced by the larger opening ratio and to a settlement. The increased structural vulnerability due to the crack
further reduction of stiffness after the crack initiation. The evolu- localisation and the reduced shear section has a much stronger
tion of the angular distortion b in the other two cases makes it pos- effect than the increased bending flexibility given by more
118 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

openings. As shown in Fig. 13f, the LTSM only takes into account and fracture energy of 6000 MPa, 0.3 Mpa and 50 N/m, respective-
the latter effect, leading to a substantially higher damage level pre- ly. In the next sections the important influence of material proper-
diction based on the numerical results compared to the LTSM ties on the final damage will be highlighted.
prediction. When comparing the numerical results of this study with the
These results are consistent with the outcomes of previous LTSM damage assessment, it has to be considered that the damage
studies. Son and Cording [19] performed numerical tests to evalu- distribution and level also depends on the specific profile applied
ate the variation of equivalent bending and shear stiffness with the to the numerical model. Fig. 14 shows the vertical displacement,
increase in opening percentage. Depending on the masonry prop- local deflection ratio and local curvature of the hogging profile.
erties, they observed a decrease in the equivalent shear stiffness The comparison with the Gaussian curve used as a reference [9]
between 45% and 77% when the percentage of windows openings highlights that the local curvature beneath and to the right of the
increases from 0 to 30, while the equivalent bending stiffness right-hand door is higher than for the Gaussian curve. Fig. 14c in
decreases only 20–26%. As a consequence of this, the E=G ratio particular suggests that the crack pattern is strongly dependent
could vary between 11 and 52 for an opening percentage of 30 on both the position of the windows, determining the weakest
and different material properties. In the framework of the relative cross section, and the hinge section (Fig. 2), defining the location
stiffness approach, which will be described in relation to the soil– of the maximum bending moment [9].
structure interaction mechanism (Section 4.5), Melis and Rodri-
guez Ortiz [23] concluded that the presence of openings decreases 4.2. Material parameters
the overall bending stiffness of the structure. More specifically,
they determined a reduction of up to 90% in the global bending In the adopted continuum model, the tension behaviour after
stiffness for a building with openings from 25% to 40% and length cracking is defined by the Young’s modulus E, the tensile strength
L shorter than their height H. Other authors observed that the win- f t and the fracture energy Gf .
dow shape does not have a major impact on the structural respon-
se [24]. 4.2.1. Fracture energy
Netzel [25] also analysed the different behaviour of a massive Fig. 15 compares the deformations of the reference model, char-
wall and a façade with openings subject to tunnelling, by using a acterised by a rotating crack model for masonry with Gf ¼ 10 N/m,
smeared crack model for the masonry. He also compared the with two other cases: a rotating crack model with Gf ¼ 50 N/m, i.e.
numerical results with the corresponding LTSM prediction, observ- higher material ductility, and a linear elastic model. In the non-lin-
ing a general agreement between the two assessment methods for ear models, the increased value of Gf leads to a slight reduction of
both the structure types in hogging area. While the results for the the maximum crack width, but it does not alter the failure
wall are in line with his findings, the façade shows in general a mechanism (Fig. 15a and b). On the contrary, the elastic model
higher level of damage with respect to the LTSM prediction. A rea- exhibits a significantly stiffer behaviour, causing a tension-induced
son for this can be seen in the higher values assumed in his study gap at the base interface and suggesting a dissimilar failure
for the masonry parameters: a Young’s modulus, tensile strength mechanism, with a different distribution of the maximum

−3 −3
x 10 x 10
2 3.5 7
Vertical displacement (mm)

Numerical
Applied curvature (1/m)
Applied deflection ratio

0 3 Analytical 6

−2 2.5 5

−4 2 4

−6 1.5 3

−8 1 2

−10 Numerical 0.5 1 Numerical

Analytical Analytical
−12 0 0
0 0.5 1 0 0.5 1 0 0.5 1
Horizontal coordinate (m) Horizontal coordinate (m) Horizontal coordinate (m)
(a) Vertical displacement (b) Local deflection ratio (c) Local curvature

Fig. 14. Comparison between the applied numerical and analytical (Gaussian) settlement profile.

compression
elastic tension
partially open crack
fully open crack

(a) G f = 10 N/m (b) G f = 50 N/m (c) elastic model

Fig. 15. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different values of fracture energy Gf .
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 119

principal strain (Fig. 15c). As a consequence, in the elastic model, elastic model is not adequately reproducing the response of
the localisation of the maximum tensile strains, which indicates masonry structures to soil settlements, since such a response is
the expected location of cracks, differs from the actual crack loca- governed by cracking. Similar conclusions have been reached by
tion observed in the experiment. This result emphasises the need Netzel [25].
to use a non-linear model accounting for the post-crack stress Fig. 16f shows that the use of the LTSM would predict a less sev-
redistribution. ere structural damage, if compared to the cracking model (damage
The maximum damage induced by a hogging deformation level ratio >1). Due to the basic assumption of the LTSM (linear
(Fig. 16d) is well represented by the top lateral strain (Fig. 16a). elastic model strain compared with limiting strain) the analytical
As expected, the angular distortion b of the elastic model is not procedure cannot consider the redistribution and reserve by slow
increasing with the applied deflection ratio (Fig. 16c), while for cracking propagation. The results of numerical analyses could be
the cracking models the curve slope increases in correspondence used to adapt the tensile limiting strain values indicated by the
with the main crack opening (see Fig. 16d). The higher vul- LTSM to include the effect of material cracking parameters on
nerability of more brittle structures is shown by the increase of b the quasi-brittle structural response.
for reduced Gf (Fig. 16c) and it is confirmed by the damage level
comparison plotted in Fig. 16e. This tendency agrees with the sen- 4.2.2. Young’s modulus
sitivity study made by Boonpichetvong and Rots [12], who Figs. 17 and 18 show that the model is not particularly sensitive
observed an increase in the crack propagation rate with the to the variation of E, in terms neither of failure mechanism nor of
decrease in fracture energy and so emphasised the importance of global damage. The development of top lateral strain (Fig. 18a),
tensile softening when using a smeared crack model for the angular distortion (Fig. 18c) and maximum crack width (Fig. 18d)
masonry. Son and Cording [6] also showed numerically that an as a function of the applied settlement consistently indicates a

−4 −3
x 10 x 10
0.01 0 8
elastic elastic
Gf=50N/m Gf=50N/m
Base horizontal strain
Top horizontal strain

0.008 6

Angular distortion
Gf=10N/m −0.5 Gf=10N/m

0.006 4
−1
0.004 2

−1.5 elastic
0.002 Gf=50N/m 0
Gf=10N/m
0 −2 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

8 6
5
Maximum crack width (mm)

elastic
Gf=50N/m 6 4
3
Damage level ratio

6 Gf=10N/m
5
Damage level

2
elastic
4
4 Gf=50N/m 1
3 Gf=10N/m

2 1/2 elastic
2
1/3 Gf=50N/m
1 1/4 Gf=10N/m
1/5
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 16. Variation of masonry fracture energy: damage indicators as a function of the applied deflection ratio.

compression
elastic tension
partially open crack
fully open crack

(a) E = 1000 Mpa (b) E = 3000 Mpa (c) E = 9000 Mpa

Fig. 17. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different values of masonry Young’s modulus E.
120 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

−4 −3
x 10 x 10
0.012 0 8
E=1000MPa E=1000MPa
0.01 E=3000MPa E=3000MPa

Base horizontal strain

Top horizontal strain

E=9000MPa 6 E=9000MPa

Angular distortion
−0.5
0.008
4
0.006 −1
2
0.004
−1.5
E=1000MPa 0
0.002 E=3000MPa
E=9000MPa
0 −2 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

10 6
E=1000MPa 5
Maximum crack width (mm)

E=3000MPa 4
6
8 3

Damage level ratio

E=9000MPa
5
Damage level

2
6 4
1
4 3

2 1/2
2 E=1000MPa 1/3 E=1000MPa
1 E=3000MPa 1/4 E=3000MPa
E=9000MPa 1/5 E=9000MPa
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 18. Variation of masonry Young’s modulus: damage indicators as a function of the applied deflection ratio.

higher vulnerability of the more flexible structure (E ¼ 1000 MPa), but only takes into account different structural types (e.g. façades,
which is more apt to follow the imposed deformation. Fig. 18b bearing walls and frame structures).
shows the corresponding development of compressive strain at
the façade base, increasing with the increase in structural 4.2.3. Tensile strength
flexibility. The sensitivity analysis performed by varying the masonry ten-
Locally, the reduced Young’s modulus should lead to crack ini- sile strength indicates that f t has a strong influence on the final
tiation at higher values of deformation (Fig. 9b); however, for the damage. An increased strength leads to a structural response simi-
assumed range of E values, this effect does not compensate the lar to the one of an elastic façade (Fig. 19). This is confirmed by the
larger angular distortion of the structure. The difference in terms graphs in Fig. 20 and it is consistent with the results obtained by
of global damage induced by the tensile strain is not significant, Boonpichetvong and Rots [12]. As for the reference case
and it does not considerably affect the final damage classification (f t ¼ 0:1 MPa), the evolution of the horizontal strains eh;top ; eh;base
(Fig. 18e). The relatively little effect of the Young’s modulus varia- and the maximum crack width wmax for f t ¼ 0:3 MPa evidences a
tion can be explained by the assumption of elastic-softening beha- strong discontinuity at the crack initiation (Fig. 20a, b, and d). This
viour for the masonry. The influence of a lower Young’s modulus is change of slope occurs at higher values of applied D=L. In the
in fact negligible, if compared with the global stiffness reduction extreme case of f t ¼ 0:9 MPa, there is no angular distortion b
induced by the cracking. Son and Cording [6] also observed that (Fig. 20c). The structure is not affected by tensile strain nor damage
the effect of cracking is more relevant than the initial stiffness of (Fig. 20d) and the compressive deformation (Fig. 20b) reaches
the structure. The results are in agreement with the assumptions an asymptotic value that corresponds to the beginning of the
of the LTSM, which does not contemplate different values for the interface gapping. These results are consistent with the work of
equivalent beam E=G ratio depending on the type of masonry, Boonpichetvong and Rots [12] who reported a decrease of the

compression
elastic tension
partially open crack
fully open crack

(a) ft = 0.1 MPa (b) ft = 0.3 MPa (c) ft = 0.9 MPa

Fig. 19. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different values of masonry tensile strength f t .
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 121

−4 −3
x 10 x 10
0.01 0 8
ft=0.1MPa ft=0.1MPa

Base horizontal strain

ft=0.3MPa ft=0.3MPa
Top horizontal strain

0.008 6

Angular distortion
ft=0.9MPa −0.5 ft=0.9MPa
0.006 4
−1
0.004 2
ft=0.1MPa
−1.5
0.002 ft=0.3MPa 0
ft=0.9MPa
0 −2 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10
(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

8 6
5
Maximum crack width (mm)

ft=0.1MPa
6 4
ft=0.3MPa 3

Damage level ratio

6 ft=0.9MPa 5
Damage level
2
ft=0.1MPa
4
4 ft=0.3MPa 1
3 ft=0.9MPa
2 1/2 ft=0.1MPa
2
1/3 ft=0.3MPa
1 1/4 ft=0.9MPa
1/5
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10
(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 20. Variation of masonry tensile strength: damage indicators as a function of the applied deflection ratio.

critical angular distortion with the decrease in tensile strength, prediction. It also emphasises the need for establishing a proper
while the cracking onset is delayed if a higher tensile strength is load amplification in scaled experiments.
assumed. Several authors investigated the effect of weight on the struc-
The damage classification is strongly dependent on the f t varia- tural response to settlement, taking into account different geome-
tion (Fig. 20e). As for the fracture energy, the LTSM takes one lim- tries and building typologies. Franzius et al. [26] performed
iting strain value for all masonry types. For poor quality masonry, numerical analysis on several building models of different weight
characterised by low values of f t , the analytical method is much and stiffness. By comparing the effect of the variation in weight
less conservative than the numerical model (Fig. 20f). The differ- with the effect to the variation in stiffness, he concluded that for
ence decreases for materials with higher strength, where the struc- realistic values of both, the influence of weight was negligible with
tural response approximates the elastic one. In the same way as for respect to the stiffness effect. Differently, Farrell [7] experimentally
the Gf variations, these results confirm the need to include a crack tested the response to tunnelling of building beam models of dif-
model when evaluating the structural damage. ferent weight and stiffness, observing the development of a gap
between thee soil and the structure and suggesting its correlation
4.3. Building weight with both the effects of building weight and stiffness. Recent
numerical tests also showed that the relation between the building
Fig. 21 highlights how the initial loading conditions influence stiffness and the tunnelling-induced deformation depends on the
the final damage. The response of the model with no initial load building weight [27]. Other authors also highlighted the impor-
applied (Fig. 21a) and the façade with self-weight only (Fig. 21b) tance of modelling the building weight [11,28].
are compared with the reference case, which includes both the
self-weight and the amplified vertical loads (Fig. 21c).
If no amplified loads are applied, the imposed settlement leads 4.4. Initial damage
to an extensive gap from both ends of the no-tension interface
(Fig. 21a and b). In absence of any vertical load, the façade is sub- In order to explore the effects of initial damage, two other cases
jected to a rotation about a point at the hogging support beam were investigated where actions prior to the settlement applica-
(Fig. 21a), while if the structure is subjected to self-weight the con- tion caused significant damage to the structure. Fig. 23 presents
tact area is extended, due to the redistribution of interface stresses the failure mechanisms as a consequence of the reference hogging
(Fig. 21b). In this latter case, the maximum tensile strain localisa- settlement for a façade subjected to extra initial loading (Fig. 23b)
tion in the façade occurs at the central section located above the or to an initial sagging settlement trough (Fig. 23c). The results
contact area. In both cases, the structure tilts without being affect- show the significant influence of the initial damage. The doubled
ed by internal distortion (Fig. 22c) and no damage arises (Fig. 22d). initial loading, induced by the same type of vertical loads with
The two variations of initial loading do not represent realistic respect to the reference case, makes the structure more vulnerable,
cases, and therefore the comparison with the LTSM assessment causing an earlier damage localisation, without altering the failure
(Fig. 22f) does not give indications usable in practice. Note that mechanism. Conversely, the initial sagging profile changes the
the amplified vertical loads comprise 90% of the total vertical loads. façade sensitivity to the subsequent hogging displacement. The
However, it emphasises the need to take into account the building sagging deformation induces compressive stresses in areas that
weight in order to obtain a reliable settlement-induced damage subsequently are subjected to tension, and therefore delays the
122 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

compression
elastic tension
partially open crack
fully open crack

(a) no initial loading (b) self-weight only (c) self-weight and amplified vertical
loads (reference case)

Fig. 21. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different initial loading.

−4 −3
x 10 x 10
0.01 0 8
no load no load
dead load only dead load only
Base horizontal strain
Top horizontal strain

0.008 dead load + live loads −0.2 6 dead load + live loads

Angular distortion
0.006 −0.4 no load 4
dead load only
0.004 −0.6 dead load + live loads 2

0.002 −0.8 0

0 −1 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

8 6
5
Maximum crack width (mm)

no load
dead load only 6 4
3
Damage level ratio

6 dead load + live loads

5
Damage level

2
4 no load
4 dead load only 1
3 dead load + live loads

2 1/2
2 no load
1/3
1 1/4 dead load only
1/5 dead load + live loads
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 22. Initial loading variation: damage indicators as a function of the applied deflection ratio.

compression
elastic tension
partially open crack
fully open crack

(a) reference case (b) doubled initial loading (c) initial sagging settlement

Fig. 23. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for different initial damage.
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 123

−3 −3
x 10 x 10
0.012 2 10
1 x live loads 1 x live loads
2 x live loads 2 x live loads

Base horizontal strain

0.01 8
Top horizontal strain

initial sagging 0 initial sagging

Angular distortion
0.008 6
−2
0.006 4
−4
0.004 2

−6 1 x live loads
0.002 2 x live loads 0
initial sagging
0 −8 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

10 6
5
Maximum crack width (mm)

1 x live loads
4
2 x live loads 6
8 initial sagging 3

Damage level ratio

5
Damage level
2
6 4
1
4 3

2 1/2
2 1 x live loads 1/3 1 x live loads
1 2 x live loads 2 x live loads
1/4
initial sagging 1/5 initial sagging
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10
(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 24. Initial damage variation: damage indicators as a function of the applied deflection ratio.

hogging type failure mechanism (Fig. 24a). The maximum crack 4.5. Soil–structure interaction
width (Fig. 24d), which is used here to assess the damage level
(Fig. 24e) is significantly reduced. In this case, the activation of a The availability of experimental data corresponding to a specific
different failure mechanism makes the initially applied settlement value of the base interface normal stiffness, which accounts for the
beneficial to the structure, under the assumption of neglecting interaction between the building, the soil and the foundation,
stress relaxation. However, it is expected that if the building expe- allowed for the validation of the semi-coupled numerical model.
riences in its history a differential settlement similar to the one Therefore, the reference continuum model represents a solid base
induced by tunnelling, an increased vulnerability can be expected. for the evaluation of the interface effect on the structural response.
The initial conditions of the building, both in terms of loading
and damage, are not explicitly included in the LTSM. Consequently, 4.5.1. Bedding stiffness
in the comparison of Figs. 22f and 24f, only the numerator of the
For kn ¼ 0:7  107 N=m3 , thus two orders of magnitude smaller
damage level ratio varies with the variation of the initial condi-
than the reference stiffness, the interface leads to a strong redistri-
tions. This parameter could therefore be usefully included in a fur-
bution of the base displacements, and therefore the façade stresses
ther refinement of the preliminary assessment, as indicated by the
are significantly smaller than the applied ones (Fig. 25). In this
Dutch Foundation for Building Research (SBR) [29]. The SBR sug-
extreme case, the maximum principal stresses are everywhere
gests to take into account the increased vulnerability of buildings
lower than the masonry tensile strength, and no damage arises.
as a consequence of the initial conditions by reducing the tolerable
strain limits in the Burland’s classification [34]. Depending on the For a stiffness of 0:7  108 N=m3 , the initial deformations are simi-
outcome of preliminary inspections, the strain limits can be lar to those in the lower interface stiffness case (Fig. 26a). However,
decreased of 20–30% for buildings in moderate conditions and at D=L ¼ 2:5  103 the reduced distribution of displacements
55–75% for buildings in poor conditions. leads to the façade cracking (Fig. 26d), which develops according
The use of observations from building monitoring to refine the to a failure mechanism different from the reference one
damage prediction has been applied during several underground (Fig. 25c). At this level of applied deformation, the façade displace-
projects in urban environment. In the design of the Porto and Turin ments start to follow the ones related to the reference value with
metro lines, building conditions, including existing defects, were kn ¼ 0:7  109 N=m3 , and the amount of damage of the two cases
used to define the intrinsic vulnerability of the structure and to becomes comparable. This tendency is confirmed by the compar-
adjust the damage category accordingly [30]. Clarke and Laefer ison between the applied and resulting settlements at the façade
[20] extended the traditional damage prediction method by includ- base for different values of kn (Fig. 27).
ing a vulnerability assessment procedure that takes into account The outcome is in line with the extensions of the LTSM method
the current physical conditions of the building and they applied based on the concept of relative stiffness [4,7,32,33]. Depending on
it to a case area in Dublin. For the Crossrail project in London, Dev- its magnitude with respect to the soil stiffness, the stiffness of the
riendt et al. [31] refined Burland’s approach [34] by combining the structure mitigates the greenfield deformations and therefore the
damage assessment with an evaluation of the structural sensitivity, strains induced to the building. The ratio between the overall
building conditions and heritage sensitivity derived by the building building stiffness and the soil stiffness has been introduced by
inspection. Potts and Addenbrooke [4] to quantify the difference between
124 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

compression
elastic tension
partially open crack
fully open crack

(a) kn = 0.7 × 107 N/ m3 (b) kn = 0.7 × 108 N/ m3 (c) kn = 0.7 × 109 N/ m3 (refer-
ence case)

Fig. 25. Maximum principal strain distribution, deformed configuration and base stress distribution at 11.5 mm of applied displacement, for different values of interface
normal stiffness kn .

the greenfield deformation of the soil and the deformation current- for higher values of relative stiffness. In agreement with the dam-
ly experienced by the surface building. Several authors [7,32,33] age patterns shown in Fig. 25 and the graphs of Fig. 26, Netzel [25]
proposed further refinements of the initial definition of the relative also reported an increased damage sensitivity for increasing values
stiffness; all these studies demonstrated that structures relatively of soil stiffness.
flexible with respect to the soil tends to follow the greenfield dis- Fig. 26e shows the effect of the interface stiffness variation on
placements, while stiffer structures exhibit a lower deflection and the final classification. At the same level of applied deflection ratio,
tend to flatten the shoulder of the settlement profile. the difference between the assumed value of interface stiffness can
In the semi-coupled model presented in this study, for a con- lead to a difference of up to five damage levels. When comparing
stant value of the façade stiffness, an increase in the normal stiff- the numerically derived damage level with the result of the analy-
ness of the interface representing the interaction with the soil tical assessment, the original formulation of the LTSM is adopted,
can be translated in a decrease in relative stiffness, and vice versa. with no evaluation of the soil–structure interaction effect. There-
Consistently with the relative stiffness approach, for the highest fore, the difference between the three curves in Fig. 26f is propor-
value of interface stiffness and thus lowest relative stiffness, the tional to the damage reduction due to the decreasing interface
structure follows closely the applied deformation (Figs. 25c and stiffness. The modification factors introduced by [4] and modified
27), while the settlement profile at the base is significantly flatter by [32] would lead to a less conservative analytical assessment,

−4 −3
x 10 x 10
0.01 2 8
7 3 7 3
kn=0.7x10 N/m kn=0.7x10 N/m kn=0.7x107N/m3
Base horizontal strain

1.5
Top horizontal strain

8 3 8 3
0.008 kn=0.7x10 N/m kn=0.7x10 N/m 6 kn=0.7x108N/m3
Angular distortion

9 3 9 3
kn=0.7x10 N/m 1 kn=0.7x10 N/m kn=0.7x109N/m3
0.006 4
0.5
0.004 2
0

0.002 0
−0.5

0 −1 −2
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

12 6
5
Maximum crack width (mm)

kn=0.7x107N/m3 4
10 6
kn=0.7x108N/m3 3
Damage level ratio

5
Damage level

9 3 2
8 kn=0.7x10 N/m
kn=0.7x107N/m3
4 8 3
6 kn=0.7x10 N/m 1
3
kn=0.7x109N/m3
4 kn=0.7x107N/m3
2 1/2
1/3 kn=0.7x108N/m3
2 1 1/4 kn=0.7x109N/m3
1/5
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 26. Variation of interface normal stiffness: damage indicators as a function of the applied deflection ratio.
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 125

0 0

Vertical displacement (mm)

Vertical displacement (mm)

−2 −2

−4 −4

−6 applied −6 applied
k n=0.7x107N/m3 k n=0.7x107N/m3

−8 k n=0.7x108N/m3 −8 k n=0.7x108N/m3
k n=0.7x109N/m3 k n=0.7x109N/m3
−10 −10
0 400 800 1200 0 400 800 1200

Horizontal coordinate (mm) Horizontal coordinate (mm)

(a) (b)

0 0
Vertical displacement (mm)

Vertical displacement (mm)

−2 −2

−4 −4

−6 applied −6 applied
k n=0.7x107N/m3 k n=0.7x107N/m3

−8 k n=0.7x108N/m3 −8 k n=0.7x108N/m3
3
9
k n=0.7x10 N/m k n=0.7x109N/m3
−10 −10
0 400 800 1200 0 400 800 1200
Horizontal coordinate (mm) Horizontal coordinate (mm)
(c) (d)
Fig. 27. Comparison between numerical and experimental settlements at the façade base for different values of kn at applied vertical displacement of (a) 2.5 mm, (b) 5 mm,
(c) 7.5 mm, and (d) 10 mm.

compression
elastic tension
partially open crack
fully open crack

(a) no horizontal displacements (b) kt = 0.7 × 108 N/ m3 (c) kt = 0.7 × 109 N/ m3

Fig. 28. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, for the variation of the interface shear behaviour.

depending on the ratio between the soil and the structure stiffness.
In the analysed cases, the use of the modification factors would
reduce the discrepancy between cracking model and LTSM damage
levels for kn ¼ 0:7  107 N=m3 , where the LTSM predicts a higher
damage level (Fig. 26f).
The interface normal stress distribution for the three analysed
cases (Fig. 25) confirm a strong interdependency between the
soil–structure interaction stiffness and the final damage, not only
in terms of maximum deformation and crack width, but also in
type of failure mechanism. The relation is also affected by the
amount and location of openings. Note that the façade deforma-
tions in Fig. 25 exclude the deformations resulting from the initial
loading. Therefore the distance between the supporting beam and
the interface base in Fig. 25a and 25b only indicate a relative dis-
placement; no gap is arising in the interface, as proven by the cor-
responding distribution plots of the interface normal stresses,
which are all compressive.
4.5.2. Horizontal displacements
In the reference case, corresponding to the experimental condi-
tions, no horizontal displacements are considered. The selection of
this case takes into account the field measurements reported by
Burland et al. [34], who observed that the soil generally does not
transmit horizontal strain to the structure. When applying to the
reference model the horizontal component of the hogging profile
shown in Fig. 10b, neither the failure mechanism nor the final
damage is significantly modified (Fig. 28). The horizontal displace-
ments lead to an earlier crack initiation due to the horizontal strain
(Fig. 29d and e). For increasing values of applied deflection ratio,
the horizontal strains at the façade base act as a restraint of the
126 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

−3 −3
x 10 x 10
0.01 4 4
no horizontal displacements no horizontal displacements no horizontal displacements
kt=0.7x108N/m3 kt=0.7x108N/m3 kt=0.7x108N/m3

Base horizontal strain

Top horizontal strain

0.008 3 3

Angular distortion
kt=0.7x109N/m3 kt=0.7x109N/m3 kt=0.7x109N/m3
0.006 2 2

0.004 1 1

0.002 0 0

0 −1 −1
0 1 2 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

12 6
5
Maximum crack width (mm)

no horizontal displacements
4
10 kt=0.7x108N/m3 6
3

Damage level ratio

kt=0.7x109N/m3 5
Damage level

8 2
4
6 1
3
4 no horizontal displacements 1/2 no horizontal displacements
2
kt=0.7x108N/m3 1/3 kt=0.7x108N/m3
2 1
kt=0.7x109N/m3 1/4 kt=0.7x109N/m3
1/5
0 1/6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 29. Variation of interface shear behaviour: damage indicators as a function of the applied deflection ratio.

hogging failure mechanism, which is typically characterised by a
bending crack developing from the wall top. Finally, the two rotat-
ing blocks defined by the vertical crack are sufficiently disconnect-
ed to overrule this effect, and the horizontal dragging contributes
to a further separation. This conclusion is supported by the evolu-
tion of the top horizontal strain eh;top and the maximum crack
width wmax (Fig. 29a and d). The horizontal displacements reduce
the façade D=L ¼ 0:6  103 and
deformation
3
D=L ¼ 1:5  10 . For higher D=L’s, amongst the two models with
horizontal displacements applied, the most vulnerable is the one
with the lowest interface shear stiffness.
Fig. 29e points out the higher vulnerability induced by the
horizontal component of tunnelling-induced displacements in a
hogging zone. In this graph only the effect of earlier crack initiation
is visible, because for D=L > 1:5  103 all the three cases fall in the
highest damage level. For this variation, the damage level ratio in
Fig. 29f it is not proportional to the damage level in Fig. 29e. The
reason is that for buildings in a hogging zone also the LTSM takes
into account the increased vulnerability due to horizontal strains.
In the analytical evaluation, the horizontal displacement is
assumed to be transmitted fully to the structure, and therefore
the increase in damage level is conservative. This leads to a
reduced difference between the numerical and the analytical out-
puts. The results of these analyses confirm the conclusions from
[25]: horizontal ground movements in a hogging zone have a nega-
tive effect on surface structures. However, for practical use of this
recommendation, it has to be taken into account that field observa-
tions revealed little transmission of horizontal deformations
through continuous footings [34].
4.6. Settlement profile
In case of buildings subjected to sagging deformation (Fig. 10a),
the horizontal component of ground displacements has a inverse
effect on the structural response, with respect to the hogging case.
If no horizontal displacements are applied, the façade develops a
typical failure mechanism with a main central crack initiating from
the base (Fig. 30a). The base horizontal displacements point
towards the middle of the façade (Fig. 10b), resisting the global
deformation caused by the vertical displacements (Fig. 30b). In
the non-realistic case of a very high shear stiffness, the induced
stiffening of the façade would lead to the base interface gapping
between (Fig. 30c).
In case of sagging, the deformation parameters that reflect best
the failure mechanism and the maximum crack growth (Fig. 31d)
are the base lateral strain and the angular distortion (Fig. 31b
and c). The damage level and the damage level ratio (Fig. 31e
and f) show a consistent progression with the increase of the
applied deformation, because in the LTSM calculation the effect
of horizontal strain is not taken into account. This assumption,
commonly adopted in practice, is conservative, because it excludes
the beneficial effect of the restraint.
Contrary to what is reported in literature [1], at the same level
of applied deflection ratio, the façade in a sagging area without
horizontal displacements applied exhibits more damage than the
one subjected to hogging deformation. A possible explanation lies
in the choice for the symmetric sagging deformation, which corre-
sponds to a building located above the tunnel axis, and is therefore
not characteristic for the statistical distribution in reality. Further-
more, the specific location of openings significantly reduces the
cross section in the façade subjected to the larger sagging deforma-
tion, facilitating the crack initiation and brittle growth. Lastly, in
real buildings a small part of the horizontal strains is transmitted
from the ground to the structure, acting in a sagging zone as a
restraint of the failure mechanism, as described above.
In addition to the pure hogging and pure sagging deformations,
a mixed situation, where the middle point of the façade is located
above the inflection point of the tunnelling-induced settlement
trough, is considered (Fig. 10). In this case, the first crack develops
from the top of the façade in the section subjected to the hogging-
induced maximum bending strain (Fig. 32a).
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129 127

compression
elastic tension
partially open crack
fully open crack

(a) no horizontal displacements (b) kn = 0.7 × 108 N/ m3 (c) kn = 0.7 × 109 N/ m3

Fig. 30. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, sagging profile.

−4 −3 −3
x 10 x 10 x 10
1 20 5
no horizontal displacements no horizontal displ.
kt=0.7x108N/m3 kt=0.7x108N/m3
Base horizontal strain
Top horizontal strain

15

Angular distortion
0.5 0
kt=0.7x109N/m3 kt=0.7x109N/m3
10
0 −5
5 no horizontal
displacements
−0.5 −10
0 kt=0.7x108N/m3
kt=0.7x109N/m3
−1 −5 −15
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

20 6
5
Maximum crack width (mm)

no horizontal displacements
4
kt=0.7x108N/m3 6
3
Damage level ratio

15 kt=0.7x109N/m3 5
Damage level

2
4
10 1
3
no horizontal
2 displacements 1/2 no horizontal displacements
5
kt=0.7x108N/m3
kt=0.7x108N/m3 1/3
1 1/4 kt=0.7x109N/m3
kt=0.7x109N/m3 1/5
0 1/6
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
−3 −3 −3
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSM damage level ratio

Fig. 31. Sagging profile: damage indicators as a function of the applied deflection ratio.

compression
elastic tension
partially open crack
fully open crack

(a) no horizontal displacements (b) kn = 0.7 × 108 N/ m3 (c) kn = 0.7 × 109 N/ m3

Fig. 32. Maximum principal strain distribution and deformed configuration at 11.5 mm of applied displacement, hogging–sagging mixed profile.
128 G. Giardina et al. / Engineering Structures 89 (2015) 111–129

−3 −4 −4
x 10 x 10 x 10
1.5 12 5
no horizontal displacements no horizontal displacements
kt=0.7x108N/m3 10 kt=0.7x108N/m3

Base horizontal strain

Top horizontal strain

Angular distortion
0
kt=0.7x109N/m3 8 kt=0.7x109N/m3
1
6
−5
4
0.5
2 kt=0.7x108N/m3
−10
0 kt=0.7x109N/m3
no horizontal displ.
0 −2 −15
0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
−4 −4 −4
Applied deflection ratio x 10 Applied deflection ratio x 10 Applied deflection ratio x 10

(a) Top horizontal strain (b) Base horizontal strain (c) Angular distortion

1.5 6
5
Maximum crack width (mm)

no horizontal displacements no horizontal displacements

4
kt=0.7x108N/m3 6 kt=0.7x108N/m3 3

Damage level ratio

kt=0.7x109N/m3 5 kt=0.7x109N/m3
Damage level

1 2
4
1
3
0.5 1/2 no horizontal displacements
2
1/3 kt=0.7x108N/m3
1 1/4 kt=0.7x109N/m3
1/5
0 1/6
0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
−4 −4 −4
x 10 x 10 x 10
Applied deflection ratio Applied deflection ratio Applied deflection ratio
(d) Maximum crack width (e) Numerical damage level (f) Numerical/LTSMdamagelevelratio

Fig. 33. Hogging–sagging mixed profile: damage indicators as a function of the applied deflection ratio.

The horizontal displacements increase the structural vul- of the masonry and the large opening ratio that induce crack
nerability, leading to cracks in the lower part of the façade subject- localisation.
ed to hogging deformations. The final damage increases with the Initial loading and initial cracking also strongly influence the
increase of the interface shear stiffness (Fig. 33b and d). With structural vulnerability, especially if they damage the structure
respect to the previous cases in pure hogging and sagging area, with the same mechanism subsequently leading to failure. The
the structure is subjected to an applied deflection ratio of almost effect of horizontal displacements, although not always found in
one order of magnitude smaller. This explains the reduced value real cases, was detected as beneficial for sagging deformations
of maximum crack width (Fig. 33d) and the corresponding lower and detrimental for the hogging type.
damage level (Fig. 33e and f). The quantitative results of the 2D variational study have been
used to set the framework of an alternative damage classification
system [16].
5. Conclusions

In this paper, the 1/10th scaled 2D finite element semi-coupled References

model validated in [10] was used to perform a sensitivity analysis
[1] Burland JB, Standing JR, Jardine FM. Building response to tunnelling: case
of different factors on the response of a masonry façade due to tun- studies from construction of the Jubilee Line Extension, London. CIRIA special
nelling. Geometrical characteristics, material parameters, bound- publication series. London: Thomas Telford; 2001.
ary conditions and initial loading and damage were included in [2] Burland JB, Wroth CP. Settlement of buildings and associated damage. In:
Proceedings of conference on settlement of structures. Cambridge: Pentech
the parametric study. The outputs were used to form a solid frame- Press; 1974. p. 611–54.
work of evaluation and comparison with the analytical LTSM. [3] Boscardin MD, Cording EJ. Building response to excavation-induced
The results underlined the high dependency of the final damage settlement. J Geotech Eng 1989;115(1):1–21.
[4] Potts DM, Addenbrooke TI. A structure’s influence on tunnelling-induced
on the material cracking and the soil–structure interaction, which ground movements. Proc Inst Civ Eng Geotech Eng 1997;125(2):109–25.
should therefore be included in the structural response evaluation. [5] Mair RJ, Taylor RN, Burland JB. Prediction of ground movements and
Due to the great influence of stress redistribution, the use of an assessment of risk of building damage due to bored tunnelling. In: Mair RJ,
Taylor RN, editors. Geotechnical aspects of underground construction in soft
elastic constitutive model or an inaccurate evaluation of the soil– ground. Proceedings of the international symposium, Balkema, Rotterdam;
structure interaction can lead to the prediction of an incorrect fail- 1996. p. 713–8.
ure mechanism. [6] Son M, Cording EJ. Estimation of building damage due to excavation-induced
ground movements. J Geotech Geoenviron Eng 2005;131(2):162–77.
The conclusion was confirmed by the comparison between [7] Farrell RP. Tunnelling in sands and the response of buildings. Ph.D. thesis,
damage classifications based on analytically calculated tensile University of Cambridge; 2010.
strain and numerically derived crack width; the comparison also [8] Laefer DF, Hong LT, Erkal A, Long JH, Cording EJ. Manufacturing, assembly, and
testing of scaled, historic masonry for one-gravity, pseudo-static, soil–
emphasised the need to explicitly take into account the crack
structure experiments. Constr Build Mater 2011;25(12):4362–73.
localisation induced by the openings. These parameters are the [9] Giardina G, Marini A, Hendriks MAN, Rots JG, Rizzardini F, Giuriani E.
ones causing a strong discrepancy between the numerical and Experimental analysis of a masonry façade subject to tunnelling-induced
the analytical damage classification in the reference case, i.e. the settlement. Eng Struct 2012;45:421–34.
[10] Giardina G, Graaf Avd, Hendriks MAN, Rots JG, Marini A. Numerical analysis of
model of the laboratory test. The higher vulnerability of the a masonry façade subject to tunnelling-induced settlement. Eng Struct
experimentally validated case is mainly due to the poor quality 2013;54:234–47.
 G. Giardina et al. / Engineering Structures 89 (2015) 111–129
[11] 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.
[12] Boonpichetvong M, Rots JG. Settlement damage of masonry buildings in soft-
ground tunnelling. Struct Eng 2005;83(1).
[13] DeJong MJ, Hendriks MAN, Rots JG. Sequentially linear analysis of fracture
under non-proportional loading. Eng Fract Mech 2008;75(18):5042–56.
[14] Netzel HD, Kaalberg FJ. Numerical damage risk assessment studies on adjacent
buildings due to TBM-tunnelling in Amsterdam. In: Proceedings of the
GeoEng2000 conference, Melbourne; 2000.
[15] Amorosi A, Boldini D, De Felice G, Malena M, Sebastianelli M. Tunnelling-
induced deformation and damage on historical masonry structures.
Geotechnique 2014;64(2):118–30.
[16] Giardina G, Hendriks MAN, Rots JG. Damage functions for the vulnerability
assessment of masonry buildings subjected to tunneling. J Struct Eng 2014.
[17] Rots JG. Structural masonry: an experimental/numerical basis for practical
design rules. CUR report series, Balkema; 1997.
[18] Rots JG. Settlement damage predictions for masonry. In: Verhoef L, Wittman F,
editors. Maintenance and restrengthening of materials and structures – brick
and brickwork. Proceedings of the international workshop on urban heritage
and building maintenance, Aedificatio, Freiburg; 2000. p. 47–62.
[19] Son M, Cording EJ. Evaluation of building stiffness for building response
analysis to excavation-induced ground movements. J Geotech Geoenviron Eng
2007;133(8):995–1002.
[20] Clarke J, Laefer DF. Evaluation of risk assessment procedures for buildings
adjacent to tunnelling works. Tunnel Undergr Space Technol 2014;40:333–42.
[21] Laefer DF, Conry B, Murphy D, Ceribasi S. A new multi-parameter condition
assessment scale for tunnel risk estimation. In: ISMGGE, development of urban
areas and geotechnical engineering, vol. 2. St. Petersburg; 2008. p. 571–576.
[22] Nghiem HL, Al Heib M, Emeriault F. Physical model for damage prediction in
structures due to underground excavations. Tunnel Undergr Constr
2014:155–64.
[23] Melis M, Rodriguez Ortiz J. Consideration of the stiffness of buildings in the
estimation of subsidence damage by EPB tunnelling in the Madrid subway. In:
Response of buildings to excavation induced ground movements conference,
London; 2001.
129
[24] Truong-Hong L, Laefer DF. Impact of modeling architectural detailing for
predicting unreinforced masonry response to subsidence. Automat Constr
2013;30:191–204.
[25] Netzel HD. Building response due to ground movements. Ph.D. thesis, Delft
University of Technology, Delft; 2009.
[26] Franzius JN, Potts DM, Addenbrooke TI, Burland JB. The influence of building
weight on tunnelling-induced ground and building deformation. Soil Found
2004;44(1):25–38.
[27] Giardina G, DeJong MJ, Mair RJ. Important aspects when modelling the
interaction between surface structures and tunnelling in sand. In: Yoo, Park,
Kim, Ban, editors. Geotechnical aspects of underground construction in soft
ground. Korean geotechnical society. Seoul, Korea: Taylor and Francis; 2014.
[28] Rampello S, Callisto L, Viggiani G, Soccodato FM. Evaluating the effects of
tunnelling on historical buildings: the example of a new subway in Rome/
Auswertung der Auswirkungen des Tunnelbaus auf historische Gebäude am
Beispiel einer neuen U-Bahnlinie in Rom. Geomech Tunnel 2012;5(3):275–99.
[29] SBR. Over het bepalen en bewaken van de kwaliteit van gebouwen. Stichting
Bouwresearch Rotterdam; 1998 [In Dutch].
[30] Guglielmetti V, Grasso P, Mahtab A, Xu S, editors. Mechanized tunnelling in
urban areas: design methodology and construction control. Taylor and Francis;
2008.
[31] Devriendt M, Palmer E, Hill R, Lazarus D. Historic and non-historic building
impact assessment methodology for major tunnelling infrastructure projects.
In: Bilotta, Flora, Lirer, Viggiani, editors. Geotechnical engineering for the
preservation of monuments and historic sites. London: Taylor and Francis
group; 2013. p. 335–41.
[32] Franzius JN. Behaviour of buildings due to tunnel induced subsidence. Ph.D.
thesis, Imperial College, London; 2003.
[33] Goh KH, Mair RJ. Building damage assessment for deep excavations in
Singapore and the influence of building stiffness. Geotech Eng J SEAGS
AGSSEA 2011;42(3).
[34] Burland JB, Mair RJ, Standing RN. Ground performance and building response
due to tunnelling. In: Jardine RJ, Potts DM, Higgins KG, editors. Conference on
advances in geotechnical engineering, vol. 1. Institution of Civil Engineers;
2004. p. 291–342.