buildings
Article
Seismic Vulnerability Assessment of Portuguese
Adobe Buildings
Samar Momin 1 , Holger Lovon 1 , Vitor Silva 1,2 , Tiago Miguel Ferreira 3, *






Citation: Momin, S.; Lovon, H.;
Silva, V.; Ferreira, T.M.; Vicente, R.
Seismic Vulnerability Assessment of
Portuguese Adobe Buildings.
Buildings 2021, 11, 200. https://
doi.org/10.3390/buildings11050200
Academic Editors: Rita Bento and
Ana Simões
Received: 6 April 2021
Accepted: 7 May 2021
Published: 10 May 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Buildings 2021, 11, 200. https://doi.org/10.3390/buildings11050200
and Romeu Vicente 1
1 RISCO, Research Centre for Risks and Sustainability in Construction, Civil Engineering Department,
Campus Universitário de Santiago, University of Aveiro, 3810-193 Aveiro, Portugal; samar@ua.pt (S.M.);
holger.lovonq@ua.pt (H.L.); vitorsilva@ufp.edu.pt (V.S.); romvic@ua.pt (R.V.)
2 Universidade Fernando Pessoa, Praça de 9 de Abril 349, 4249-004 Porto, Portugal
3 Department of Civil Engineering, Campus de Azurém, University of Minho, ISISE,
4800-058 Guimarães, Portugal
* Correspondence: tmferreira@civil.uminho.pt
Abstract: Adobe construction represents 5.3% of the total Portuguese building stock according to
the latest National Housing Census. The distribution of these adobe buildings is scattered across
the country, with higher density in the central region and in Algarve in the south, where the seismic
hazard is highest. A large proportion of these buildings are still in use for residential and commercial
purposes and are of historical significance, contributing to the cultural heritage of the country. Adobe
buildings are known to exhibit low seismic resistance due to their brittle behavior, thus making them
vulnerable to ground shaking and more prone to structural damage that can potentially cause human
fatalities. Three buildings with one-story, two-stories, and two-stories plus an attic were numerically
modeled using solid and contact elements. Calibration and validation of material properties were
carried out following experimental results. A set of 30 ground motion records with bi-directional
components were selected, and non-linear time-history analyses were performed until complete
collapse occurred. Two novel engineering demand parameters (EDPs) were used, and damage
thresholds were proposed. Finally, fragility and fatality vulnerability functions were derived. These
functions can be used directly in seismic risk assessment studies.
Keywords: adobe buildings; seismic vulnerability assessment; fragility functions; physical damage
estimation; fatality vulnerability functions; indoor fatalities; LS-DYNA; Portugal
1. Introduction
In this study, the seismic vulnerability assessment of three Portuguese adobe buildings
was conducted to derive fragility and fatality vulnerability functions, which can be used
to estimate losses due to earthquakes. Over the past centuries, Portugal has experienced
numerous earthquakes; a lesser-known event from the 18th century, the 1722 Algarve
earthquake, occurred 33 years before the well-known great 1755 Lisbon earthquake and
tsunami, and caused widespread damage in Algarve. More recent events in Portugal such
as the 1909 Mw 6.0 Benavente earthquake, the 1969 Mw 8.0 Algarve earthquake, and the
1980 Mw 6.8 Azores earthquake are examples of events where significant damage was
observed. The 1909 Benavente earthquake, despite its recorded moderate magnitude (Mw
6.0), is known to be the largest crustal earthquake in the Iberian Peninsula [1]. It occurred
around a period during which the Art Nouveau movement was gaining popularity in the
central region, and adobe constructions were in vogue. According to a post-earthquake
survey report of the 1909 Benavente earthquake [2], 879 buildings were damaged: 20%
reported light damage, 40% reported moderate damage, and the remaining 40% were
completely damaged. Furthermore, there were severe casualties that resulted in 46 deaths,
out of which 30 were from the village of Benavente. The direct and indirect consequences
of this event played an important role in influencing the local seismic building culture.
https://www.mdpi.com/journal/buildings
Buildings 2021, 11, 200 2 of 22

The majority of the past studies regarding the seismic vulnerability of adobe buildings
were focused on the South American building stock. Tarque et al. [3] developed analytical
fragility functions following the displacement-based earthquake loss assessment (DBELA)
methodology [4] for commonly found one story buildings from Peru. Ahmad et al. [5]
broadly proposed fragility functions for a range of commonly found buildings, including
adobe structures in Pakistan. A study focused broadly on the South American building
stock [6] developed fragility functions for unreinforced masonry with adobe blocks, using
single-degree-of-freedom (SDoF) oscillators to model the response of the buildings. Re-
cently, Martins and Silva [7] developed analytical fragility and vulnerability functions using
censored cloud regression analysis, for one, two, and three-story adobe buildings. Finally,
Sumerente et al. [8] developed fragility functions, combining in-plane and out-of-plane
loading conditions, for typical one- and two-story adobe buildings in Peru.
According to Abeling and Ingham [9], the building volume loss is considered a better
damage descriptor for estimating risk to occupants as compared to traditional damage
states because it can be directly correlated with earthquake fatalities. Such data is essential
to better estimate earthquake-related risk and losses, thus minimizing economic losses
and mitigating fatalities. The fragility and vulnerability functions of the Portuguese adobe
buildings with one-story, two-stories, and two-stories plus an attic were derived following
the steps listed below.
1. Building surveys that contained well-documented evidence such as building schemat-
ics and descriptions of the buildings were examined. However, due to the scarcity
of a large dataset specific to adobe buildings, three buildings with one archetype per
building class of one-story, two-stories, and two-stories plus an attic were selected for
the case study presented herein.
2. These selected buildings were numerically modeled using the advanced finite ele-
ment method-based software LS-DYNA [10]. These models were tested by following
experimental validation and a series of sensitivity analyses to check the influence of
various parameters.
3. A set of 30 bi-directional ground motion records were selected based on the peak
ground acceleration (PGA) as the intensity measure (IM), and used to perform several
non-linear time-history analyses until the complete collapse of the buildings.
4. To derive the fragility functions, two novel engineering demand parameters (EDPs)—
namely the crack propagation ratio (CPR) and volume loss ratio (VLR)—were selected,
and damage thresholds were proposed such that they can be correlated with the
damage classifications of the European Macroseismic Scale (EMS-98) [11].
5. The fragility functions were derived using the well-established cloud analysis [12] to
quantify physical structural damage due to a given IM.
6. Furthermore, a fatality vulnerability function was derived to estimate indoor fatalities.

2. Characterization of Adobe Buildings in Portugal

According to the 2011 Population and Housing Census of Portugal (INE) [13], the total
building stock of Portugal comprises 3,353,610 buildings, out of which the adobe building
stock amounts to 178,422 buildings, i.e., about 5.32% of the whole Portuguese building
stock (Figure 1).
 Buildings 2021, 11, x FOR PEER REVIEW 3 of 23
Buildings 2021, 11, 200 3 of 22

Figure1.1.Geographical
Figure Geographicaldistribution
distributionofofadobe
adobebuildings
buildingsininPortugal,
Portugal,asasreported inin
reported the INE
the 2011
INE [13].
2011
[13].
In Portugal, earthen materials have been used to construct loadbearing walls in the
form of In adobe or rammed
Portugal, earthen earth for the
materials haveconstruction
been usedof tobuildings,
construct especially
loadbearing in walls
the central
in the
and southern regions of the country. Locally, earth-based building
form of adobe or rammed earth for the construction of buildings, especially in the materials are divided
central
into
andthree different
southern types
regions of of
thebuilding
country.techniques: rammed earth,
Locally, earth-based known
building as “taipa”;
materials wattle-
are divided
and-daub, “tabique”; and adobe [14]. Currently, there is a lack of reliable
into three different types of building techniques: rammed earth, known as “taipa”; wattle- vulnerability
models for adobe buildings in Portugal. In this study, the advances in computational
and-daub, “tabique”; and adobe [14]. Currently, there is a lack of reliable vulnerability
modeling capabilities were utilized for the development of sophisticated numerical models
models for adobe buildings in Portugal. In this study, the advances in computational mod-
that can simulate the complete collapse of a full-scale building subjected to bi-directional
eling capabilities were utilized for the development of sophisticated numerical models
loading and are capable of predicting crack propagation and volume loss. As previously
that can simulate the complete collapse of a full-scale building subjected to bi-directional
mentioned, there are many adobe buildings in the central and southern part of the country,
loading and are capable of predicting crack propagation and volume loss. As previously
due to the existence of sandy soils and the presence of lime, which has been reported to
mentioned, there are many adobe buildings in the central and southern part of the coun-
be used as a stabilizing agent. It can be observed that the central region of Portugal has
try, due to the existence of sandy soils and the presence of lime, which has been reported
the highest density of adobe buildings, since it was the prevailing construction system
to be used as a stabilizing agent. It can be observed that the central region of Portugal has
adopted for the first half of the 20th century [15] and the legacy from the patrimony built
the highest density of adobe buildings, since it was the prevailing construction system
in that period is still significant. As part of the endeavor towards the preservation of the
adopted forand
inheritance the cultural
first halfheritage
of the 20th century
of the [15] and
country, fromthethelegacy frompast
historical the patrimony
to the present, built
in thatbuildings
adobe period isirrespective
still significant.
of theAs part of the
grandeur endeavorare
or simplicity towards the preservation
extremely important. Inoffact, the
inheritance and cultural heritage of the country, from the historical
many patrimonial buildings of high historical, cultural, and architectural value [16] are past to the present,
adobe
still in abuildings
reasonable irrespective
conservation of the grandeur
state [17]. Theorfactor
simplicity aretoextremely
that led the upsurge important.
in adobeIn
fact, many patrimonial buildings of high historical, cultural, and architectural
as a construction material of choice, especially in the central region, is credited to the Art value [16]
are still in a reasonable conservation state [17]. The factor that
Nouveau movement that spread across Europe in various forms. This movement had led to the upsurge in adobe
aassignificant
a construction material
influence of choice,
in some especially
Portuguese in[18],
cities the central region,inisAveiro,
particularly credited to theitArt
where is
estimated that 30–40% of the buildings are still adobe buildings [19]. The Southern regiona
Nouveau movement that spread across Europe in various forms. This movement had
significant
of Algarve, influence
where theinseismic
some Portuguese cities [18],
hazard is relatively particularly
high [20], has ainhigh
Aveiro, where
density it is es-
of adobe
timated that 30–40% of
buildings, as shown in Figure 1.the buildings are still adobe buildings [19]. The Southern region
of Algarve, where the seismic hazard is relatively high [20], has a high density of adobe
Database
buildings, of Adobe Buildings
as shown Information
in Figure 1.
An appropriate selection of the buildings to be modeled depends largely on the avail-
ability of building survey data containing documented evidence with building schematics,
dimensions, and detailing with photographs that can be consulted to infer information
 Database of Adobe Buildings Information
Buildings 2021, 11, 200
An appropriate selection of the buildings to be modeled depends largely on the avail- 4 of 22
ability of building survey data containing documented evidence with building schemat-
ics, dimensions, and detailing with photographs that can be consulted to infer information
about
aboutthetheexterior
exteriorwalls,
walls,interior
interiorwalls,
walls,openings,
openings,lintel
lintelbeams,
beams,roofing
roofingsystems,
systems,attic,
attic,and
and
features such as gable end
features such as gable end walls. walls.
Three
Threebuilding
buildingsurveys
surveyspertaining
pertainingtotoearth-based
earth-basedbuildings
buildingsinincontinental
continentalPortugal
Portugal
were
wereexamined
examinedininthis thisstudy,
study,namely,
namely,thetheAlgarve
Algarvesurvey
surveyreport
report[21],
[21],the
theAlentejo
Alentejosurvey
survey
report
report[22],
[22],and
andthethesurvey
survey of façade
of façadewalls in Aveiro
walls in Aveiro[19].[19].
FromFromthesethese
surveys, it wasitcon-
surveys, was
cluded that itthat
concluded wasitessential to segregate
was essential these earthen
to segregate these buildings accordingaccording
earthen buildings to the primary,
to the
secondary, and tertiary
primary, secondary, andconstruction material material
tertiary construction that wasthat usedwasto used
buildtothe external
build walls
the external
and then
walls thethen
and internal
the walls,
internal respectively. The surveysThe
walls, respectively. reported
surveys a vast amount
reported a of
vastbuildings
amount
of buildings
constructed constructed
using usingtheir
taipa as both taipa as bothand
primary their primaryconstruction
secondary and secondary construction
material, along
material,
with some along
buildingswiththatsomeused buildings
adobe and thattabique
used adobe
as theand tabiqueand
secondary as the secondary
tertiary and
construc-
tertiary
tion construction
materials materials
for internal walls.for internal
From thesewalls. From
surveys, wethese surveys,
selected we selected
a one-story a one-
building
story
with building
gable with gable
end walls in which endboth
wallsthe
in primary
which both andthe primaryconstruction
secondary and secondary construction
material used
material
was adobe.used was adobe.
The selection The selection
of buildings with two of and
buildings with two
more stories was and
carriedmoreoutstories
followingwas
carried out following studies [23] that were archived at the University
studies [23] that were archived at the University of Aveiro’s database, which documented of Aveiro’s database,
which documented
detailed building drawings detailed building drawings
constructed with adobeconstructed
as the primarywithconstruction
adobe as thematerial.
primary
construction material. It is noteworthy to mention that these buildings,
It is noteworthy to mention that these buildings, located in the municipality of Ílhavo, located inare
the
municipality
currently in use, Ílhavo,
of as shown areincurrently in use, as shown in Figure 2a,b.
Figure 2a,b.

(a) (b)
Figure
Figure2.2.Photographs of the
Photographs main
of the façades:
main (a) two-story
façades: (Building
(a) two-story 2); (b)2);
(Building 2-story plus anplus
(b) 2-story attican attic
(Building
(Building3).3).

3.3.Case
CaseStudy
StudyBuildings—Geometrical
Buildings—Geometricaland andMaterial
MaterialProperties
Properties
Theinformation
The informationobtained
obtainedfrom
fromthe
thedocumented
documentedevidence
evidencediscussed
discussedininSection
Section22was
was
studied in detail and incorporated in the LS-DYNA software environment
studied in detail and incorporated in the LS-DYNA software environment to build the to build the
numerical models and perform the vulnerability assessment of the three selected
numerical models and perform the vulnerability assessment of the three selected case case study
buildings,
study namely
buildings, Buildings
namely 1–3. 1–3.
Buildings
3.1. Geometrical Properties
3.1. Geometrical Properties
The selected buildings are unique buildings due to their geometrical properties such
The selected buildings are unique buildings due to their geometrical properties such
as exterior wall thickness, story or inter-story heights, total area, percentage of openings,
as exterior wall thickness, story or inter-story heights, total area, percentage of openings,
and building features such as lintel beams and gable-end walls. Each case study building
and building features such as lintel beams and gable-end walls. Each case study building
is described below.
is described below.
3.1.1. Building 1
3.1.1. Building 1
This one story building with gable-end walls, was located in the county Castro Marim
This onewas
in Algarve story building
utilized forwith gable-end
residential walls, was
purposes, located
but was in the county
abandoned at theCastro Ma-
time of the
rim in Algarve was utilized for residential purposes, but was abandoned at the time of
survey. Its floorplan is rectangular, and it is divided into four compartments. The primary the
and secondary construction material is adobe, and the external walls are 30 cm thick,
remaining constant along the height of the building. The thickness of the walls was modeled
by adopting a “two-parts” scheme in order to avoid computational instabilities, as shown
in Figure 3c. The internal walls are 15 cm thick and divide the various compartments of the
Buildings 2021, 11, 200 5 of 22

building. However, for simplifying the numerical models and reducing the computational
cost, these internal walls were not modeled in this study. The main façade has a door
and two windows on either side. Each visible multicoloured block is termed as a ‘part’,
and these different parts constitute the prisms that are numbered from 1–7, as depicted
in Figure 3a. The interlocking of blocks was modeled by arranging them in a staggered
pattern, and the addition of constraints at the intersections simulate the interlocking of two
orthogonal walls, as shown in Figure 3c. Since buildings are built using lintel beams, these
elements were included above the openings in the numerical models. These are elastic
members constituted by a set of blocks whose nodes are constrained at the edges in the
x, y, and z directions, as shown in Figure 3a. These are structural components that are
important to model, as they play an important role in preventing the onset of premature
collapse mechanisms due to seismic loading. The gable-end walls compose both the left
and right façades of the building and do not have any openings, as shown in Figure 3b.
The roof is composed of simple wood beams and trusses, which support the ceramic
roofing tiles. The beams of the roofing system were modeled using a discrete element
with compression-only springs added in the two orthogonal directions with equal spacing,
as shown in Figure 3a,c. The mechanical properties of the wooden beams of the roofing
system were defined according to prescriptions from Eurocode 5 [24]. The dimensions of
the building are summarized in Table 1, and the geometrical details and features that have
been modeled in the LS-DYNA environment are shown in Figure 3a–c. Further details
pertaining to numerical modeling are discussed in Section 4.

Table 1. Building characteristics and dimensions.

Building Characteristics Building 1 Building 2 Building 3

Total no. of stories 1 2 2 + Attic
X-direction 6.30 10.80 8.20
Length (m)
Y-direction 7.80 9.30 12.00
Area (m2 ) 49.14 100.44 98.40
1st Story 2.85 3.30 2.40
Height (m) 2nd Story - 3.30 3.60
Attic - - 3.00
Total height (m) 2.85 6.60 9.00
Total no. of external walls 4 4 4
External wall thickness (cm) 30 60 40
Gable-end walls Yes No Yes
Lintel beams Yes Yes Yes
Total no. of window openings 4 15 16
Total no. of door openings 2 3 4
Total percentage of openings (%) 10 20 15
Buildings 2021, 11, x FOR PEER REVIEW 6 of 23
Buildings 2021, 11, 200 6 of 22

(a) (b)

(c)
Figure 3.
Figure 3. Dimensions
Dimensions of
of Building
Building 11 and
and details
details of
of the
the modeling
modeling inin the
the LS-DYNA
LS-DYNA environment:
environment: (a)
(a) elevation
elevation view
view showing
showing
the front (main façade), (inset right) close-up of a lintel beam; (b) elevation view showing the symmetric left and right
the front (main façade), (inset right) close-up of a lintel beam; (b) elevation view showing the symmetric left and right gable
gable end walls; (c) plan view of the building and position of compression-only springs: (left inset) wall thickness detail
end walls; (c) plan view of the building and position of compression-only springs: (left inset) wall thickness detail (right
(right inset) close-up view of a corner intersection, constraints, and interlocking.
inset) close-up view of a corner intersection, constraints, and interlocking.

3.1.2. Building 2
3.1.2. Building 2
The two-story residential building is located in the municipality of Ílhavo and is cur-
The two-story residential building is located in the municipality of Ílhavo and is
rently in use (see Figure 2a). Its floorplan is rectangular and composed of a ground and
currently in use (see Figure 2a). Its floorplan is rectangular and composed of a ground
first floor with equal inter-story heights. The primary construction material is adobe,
and first floor with equal inter-story heights. The primary construction material is adobe,
while the external walls are 60 cm thick as a constant throughout the entire height of the
while the external walls are 60 cm thick as a constant throughout the entire height of the
building. The interior walls are 15 cm thick and were built using a secondary construction
building. The interior walls are 15 cm thick and were built using a secondary construction
material (tabique). These thin tabique internal walls have a much lower structural capacity
material (tabique). These thin tabique internal walls have a much lower structural capacity
when compared to the thick adobe external walls. For this reason, as well as to reduce the
when compared to the thick adobe external walls. For this reason, as well as to reduce
computational demand of the numerical models, these elements were not modeled in this
the computational demand of the numerical models, these elements were not modeled in
study. Nonetheless, a recent study by Battaglia et al. (2020) [25] indicates that tabique
this study. Nonetheless, a recent study by Battaglia et al. (2020) [25] indicates that tabique
walls can lead to a slight reduction in the structural capacity. Each of the façades has mul-
walls can lead to a slight reduction in the structural capacity. Each of the façades has
tiple openings, as presented in Figure 4a,b. The lintel beams were modeled above each of
multiple openings, as presented in Figure 4a,b. The lintel beams were modeled above each
the openings,
of the asas
openings, previously
previouslydescribed.
described.The
Theroof
roofisiscomposed
composedofofsimple
simple wood
wood beams and
beams and
trusses, which support the ceramic roofing tiles. The beams of the roofing system
trusses, which support the ceramic roofing tiles. The beams of the roofing system were were
modeled using a discrete element with compression-only springs in one direction, as
 Buildings 2021, 11, 200 7 of 22
Buildings 2021, 11, x FOR PEER REVIEW 7 of 23

modeled using a discrete element with compression-only springs in one direction, as shown
shown in Figure
in Figure 4c.mechanical
4c. The The mechanical properties
properties of theofwooden
the wooden
beamsbeams of roofing
of the the roofing system
system were
were defined according to prescriptions from Eurocode 5 [24]. The dimensions
defined according to prescriptions from Eurocode 5 [24]. The dimensions of the building of the
building are summarized
are summarized in Tablein1,Table 1, and
and the the geometrical
geometrical details
details are shownareinshown
Figurein4a–c.
Figure 4a–
Further
c. details
Furtherpertaining
details pertaining to numerical
to numerical modelingmodeling are discussed
are discussed in Sectionin4.Section 4.

(a) (b)

(c)
Figure
Figure4.4.
Dimensions ofof
Dimensions Building
Building 2 and details
2 and ofof
details the modeling
the modelinginin
the LS-DYNA
the LS-DYNA environment:
environment:(a)(a)
elevation view
elevation showing
view showing
the front (main façade); (b) elevation view showing the left and right walls; (c) plan view of the building and position
the front (main façade); (b) elevation view showing the left and right walls; (c) plan view of the building and position ofof
compression-only springs (left inset) wall thickness detail.
compression-only springs (left inset) wall thickness detail.

3.1.3. Building
3.1.3. Building 33
The
The two-story
two-story plus attic
plus with
attic gable-end
with gable-end walls is aisresidential
walls building
a residential alsoalso
building located in
located
the municipality of Ílhavo. Similar to Building 2, this one is also currently
in the municipality of Ílhavo. Similar to Building 2, this one is also currently in use (seein use (see Fig-
ure 2b). Its
Figure floorplan
2b). is rectangular
Its floorplan and is and
is rectangular composed of a basement
is composed which serves
of a basement whichasserves
a stor-as
age and a wine-cellar with a lower height than the other two stories, a
a storage and a wine-cellar with a lower height than the other two stories, a habitable habitable first floor
with
firstmultiple
floor withcompartments, and an attic and
multiple compartments, at the
antop.
atticThe primary
at the construction
top. The material is
primary construction
adobe. Theisexternal
material adobe. walls are 40 cm
The external thick,
walls arewhich
40 cmisthick,
constantwhichthroughout thethroughout
is constant entire height the
ofentire
the building.
height ofThetheinterior walls
building. Theareinterior
15 cm thick
wallsandare were
15 cmbuilt
thickusing
and a secondary
were con- a
built using
struction
secondary material (tabique);material
construction these walls were these
(tabique); not numerically
walls were modeled for the reasons
not numerically modeled de-
for
scribed for Building 2. Each of the façades has multiple openings, as shown in Figure 5a,b.
The lintel beams were modeled above each of the openings, as mentioned previously. In
Buildings 2021, 11, 200 8 of 22
Buildings 2021, 11, x FOR PEER REVIEW 8 of 23

the reasons described for Building 2. Each of the façades has multiple openings, as shown
in Figure 5a,b. The lintel beams were modeled above each of the openings, as mentioned
the basementInlevel,
previously. there are windows
the basement level, therethatareare fixed and
windows thatunopenable, serving
are fixed and only for
unopenable,
lighting, whereas
serving only the above
for lighting, story and
whereas theattic
abovehave openable
story windows
and attic that are used
have openable for light-
windows that
ing and ventilation. The flooring system consists of wooden beams that
are used for lighting and ventilation. The flooring system consists of wooden beams are equally spaced
that
at
are40equally
cm andspaced
the roof that
at 40 cmconsists
and theofroof
wooden trusses. of
that consists The beams trusses.
wooden of the roofing system
The beams of
were modeled
the roofing using
system a discrete
were modeled element
using awith compression-only
discrete springs added in the
element with compression-only two
springs
orthogonal
added in the directions with equal
two orthogonal spacingwith
directions as shown in Figureas5a,c.
equal spacing Theinmechanical
shown Figure 5a,c.prop-
The
erties of the wooden beams of the roofing system were defined according
mechanical properties of the wooden beams of the roofing system were defined according to prescriptions
from Eurocode 5from
to prescriptions [24]. Eurocode
The dimensions of the
5 [24]. The building are
dimensions summarized
of the in Table
building are 1, and the
summarized in
geometrical details are shown in Figure 5a–c. Further details pertaining to
Table 1, and the geometrical details are shown in Figure 5a–c. Further details pertaining numerical mod-to
eling are discussed
numerical modelinginare Section 4. in Section 4.
discussed

(a) (b)

(c)
Figure
Figure 5.
5. Dimensions
Dimensions of
of Building
Building 33 and
and details
details of
of the
the modeling
modeling in
in the
the LS-DYNA
LS-DYNA environment:
environment: (a)
(a) elevation
elevation view
view showing
showing
the front (main façade) and back (rear façade) showing the gable end walls; (b) elevation view showing the left and
the front (main façade) and back (rear façade) showing the gable end walls; (b) elevation view showing the left and right right
walls; (c) plan view of the building and position of compression-only springs (left inset) wall thickness detail.
walls; (c) plan view of the building and position of compression-only springs (left inset) wall thickness detail.

3.2. Material Properties

The study of adobe mechanical properties is of paramount importance to improve
the reliability of numerical models, which can then be used to derive fragility and vulner-
ability functions. Several studies and experimental campaigns have been conducted at the
Civil Engineering Department of the University of Aveiro since 2005, providing crucial
Buildings 2021, 11, 200 9 of 22

3.2. Material Properties

Buildings 2021, 11, x FOR PEER
The REVIEW
study of adobe mechanical properties is of paramount importance to improve the 9
reliability of numerical models, which can then be used to derive fragility and vulnerability
functions. Several studies and experimental campaigns have been conducted at the Civil
Engineering Department
insights andof the University
information of Aveiro
to sustain moresince 2005,numerical
robust providingmodels.
crucial insights
Adobe has also
and information to sustainstudied
extensively more robust numerical
in Portugal, models.on
but mainly Adobe has also been extensively
the characterization of its mechanical p
studied in Portugal, but mainly
erties [26,27]. on the characterization
The experimental campaigns of on
its in-plane
mechanical properties
cyclic tests on[26,27].
full-scale doub
The experimental campaigns on in-plane cyclic tests on full-scale double-T
shaped adobe walls carried out by Silveira et al. [28] provided the value shaped adobe
of Young’s m
walls carried ulus
out by
(E)Silveira
adoptedet al. [28] provided
herein. the value
The material of Young’s
properties modulus
reported (E) adopted
by Silveira et al. [28] wer
herein. The material
corporatedproperties
in a casereported by Silveira
study by Sarchi etetal.
al.[29]
[28]towere incorporated
create two finite in a case models,
element
study by Sarchiwereetthen
al. [29]
usedtotocreate
developtwofragility
finite element models, that were then used to
functions.
develop fragility functions.
3.2.1. Testing the Modeling Platform
3.2.1. Testing the Modeling Platform
In the context of this study, it was pertinent to first test the modeling platform, h
In the context of this study, it was pertinent to first test the modeling platform, herein
LS-DYNA, then check if the results of the experimental tests were reproducible and
LS-DYNA, then check if the results of the experimental tests were reproducible and
applicable. A numerical model of the adobe double-T wall experiment reported in Sil
thus applicable. A numerical model of the adobe double-T wall experiment reported
et al. [28] was recreated in the LS-DYNA environment as shown in Figure 6a, and a
in Silveira et al. [28] was recreated in the LS-DYNA environment as shown in Figure 6a,
range of values of Young’s modulus was applied and tested in the linear range. Ea
and a trial range of values of Young’s modulus was applied and tested in the linear range.
the eigenmodes and their corresponding eigenfrequencies were checked, and it was fo
Each of the eigenmodes and their corresponding eigenfrequencies were checked, and it
was found that that
at aatYoung
a Young modulus
modulus of 738
of 738 MPa,MPa, the corresponding
the corresponding frequency
frequency was
was 23 Hz,23asHz, as sh
in Figure 6b, henceforth providing a match to the experimental
shown in Figure 6b, henceforth providing a match to the experimental results. This value results. This value
was corroborated by another study [30] that suggested that the Young modulus should be shoul
corroborated by another study [30] that suggested that the Young modulus
lower than 1 lower
GPa for than
this1kind
GPa of
formaterial.
this kind of material.

(a) (b)
Figure 6. Recreated
Figure single adobe single
6. Recreated double-T walldouble-T
adobe in the LS-DYNA environment
wall in the (a) dimensions
LS-DYNA environment (a)and detailing;and
dimensions (b) Top and
isometric view of the in-plane
detailing; (b) Topmode of vibration
and isometric withofitsthe
view corresponding
in-plane modematched frequency
of vibration withf =its23corresponding
Hz.
matched frequency f = 23 Hz.
3.2.2. Sensitivity Analysis
3.2.2. Sensitivity Analysis
A sensitivity analysis was performed in which the various parameters involve
A sensitivity analysis was
the numerical performed
modeling suchinaswhich
tensilethe variousnormal
strength, parameters involved
energy, in theenergy re
and shear
numerical modeling
rates, andsuch as tensile strength,
coefficients of frictionnormal energy,by
were tested and shear
using energy proposals
different release rates,from the li
and coefficients of friction
ture. were tested by
It is a well-known using
fact fromdifferent proposals
the scientific from the
literature literature.
that It is a
adobe masonry exhib
well-known fact from the scientific literature that adobe masonry exhibits a brittle
brittle behavior, owing to its low compressive, tensile, and shear strength [31–35]. Fobehavior,
owing to its low compressive,
reason, tensile,toand
when subjected shear
cyclic strength
action, adobe[31–35].
masonry For this reason,
usually when
presents a low resist
subjected to cyclic action, adobe masonry usually presents a low resistance capacity.
capacity. The tensile strength was assumed to be 0.05 MPa [31], and the recommenda The
tensile strength wasLourenço
from assumed andto bePereira
0.05 MPa [31],
[36] andfollowed
were the recommendations
to define the from Lourenço
initial range of param
and Pereira [36]
suchwere followed
as the normal to energy
define the initialrate
release range
andofshear
parameters
energysuch as the
release normal
rate. According to
Shawa et al. [37], it is acceptable to assume equal shear and tensile parameters for
sonry. The parameters that influenced the structural response the most were the no
and shear energy release rates. Thus, the final values of normal energy and shear en
release rate adopted in this study for buildings 1, 2, and 3 are 10 N/m, 30 N/m, an
Buildings 2021, 11, 200 10 of 22

energy
Buildings 2021, 11, x FOR PEER REVIEW release rate and shear energy release rate. According to AlShawa et al. [37], it 23
10 of is
acceptable to assume equal shear and tensile parameters for masonry. The parameters
that influenced the structural response the most were the normal and shear energy release
rates. Thus, the final values of normal energy and shear energy release rate adopted in
N/m, respectively. From the literature, there was no clear recommendation for the coeffi-
this study for buildings 1, 2, and 3 are 10 N/m, 30 N/m, and 20 N/m, respectively. From
cient of friction. Thus, a static coefficient of 0.4 and a dynamic coefficient of 0.3 was as-
the literature, there was no clear recommendation for the coefficient of friction. Thus, a
sumed. A summary of all the mechanical properties and input parameters adopted for the
static coefficient of 0.4 and a dynamic coefficient of 0.3 was assumed. A summary of all
numerical models of the adobe buildings are given in Table 2.
the mechanical properties and input parameters adopted for the numerical models of the
adobe buildings are given in Table 2.
Table 2. Mechanical properties and input parameters of the numerical models.

Element Table 2. Mechanical properties and input

Mechanical parameters of the numericalValue
Properties models. Units
Young’s modulus 0.74 GPa
Element Mechanical Properties Value Units
SOLID elements Poison’s ratio 0.30 -
Young’s modulus 0.74 GPa 3
Density 1500 kg/m
SOLID elements Poison’s ratio 0.30 -
Static
Densitycoeff. of friction 1500 0.4 kg/m-3
Dynamic coeff.
Static coeff. of friction
of friction 0.4
0.3 -
-
Scale factor for segment
Dynamic coeff. of friction penalty stiffness 0.3 1.0 - -
Cohesive elements
Normal
Scale factor and shear
for segment penaltyfailure stress
stiffness 1.0 0.05 -MPa
Cohesive elements
Normal
Normal andand
shear failure
shear stressrelease rate
energy 0.05 10, 30, 20 MPa
N/m
Normal and shear energy release rate 10, 30, 20 N/m
Normal (CN) and tangential stiffness 0.74 GPa
Normal (CN) and tangential stiffness 0.74 GPa
Timber elasticity modulus 7.00 GPa
Timber elasticity modulus 7.00 GPa
Timber elasticity modulus (5%) 4.70 GPa
Springs Timber elasticity modulus (5%) 4.70 GPa
Springs Design compressive
Design compressive strengthstrength 16.00 16.00 MPa
MPa
DesignDesign
bending bending
strengthstrength 14.00 14.00 MPa
MPa

4.
4. Framework
Frameworkfor forVulnerability
Vulnerability Assessment
Assessment
The
Theframework
frameworkfor forvulnerability
vulnerability assessment
assessmentof the adobe
of the buildings
adobe is shown
buildings in Fig-
is shown in
ure 7, and
Figure eacheach
7, and of these steps
of these is discussed
steps in detail
is discussed in the
in detail following
in the subsections.
following subsections.

Figure7.
Figure Frameworkfor
7.Framework forvulnerability
vulnerabilityassessment
assessment of
of adobe
adobe buildings.
buildings.

4.1. Numerical Modeling

4.1. Numerical Modeling
LS-DYNA, a general-purpose finite element software capable of simulating complex
LS-DYNA, a general-purpose finite element software capable of simulating complex
real-world problems, was used in this study. A shared memory parallel (SMP) double
real-world problems, was used in this study. A shared memory parallel (SMP) double
precision solver was utilized to improve the convergence of the non-linear analyses, and a
precision solver was utilized to improve the convergence of the non-linear analyses, and
combined finite-discrete element [38] strategy was adopted to simulate various stages of
a combined finite-discrete element [38] strategy was adopted to simulate various stages
damage, such as cracking of walls and volume loss. The development of the numerical mod-
of damage, such as cracking of walls and volume loss. The development of the numerical
els in this study followed the recommendations from Karanikoloudis and Lourenço [31],
models in this study followed the recommendations from Karanikoloudis and Lourenço
and Lourenço and Pereira [36].
[31], and Lourenço and Pereira [36].
The pre-processing, processing, and post-processing of the numerical models can be
The categorised
broadly pre-processing,
intoprocessing,
three steps,and post-processing
as shown in Figure of
8. the numerical
Firstly, models
the detailed can be
building
broadly categorised into three steps, as shown in Figure 8. Firstly, the detailed
drawings are used to define the geometry in the software environment (LS-PrePost [39]) building
drawings
Secondly,aretheused to define
assembled the geometry
models in the
have to be software
processed andenvironment (LS-PrePost
the data stored. Finally,[39])
the
Secondly, the assembled models have to be processed and the data stored. Finally, the
relevant data is extracted and further post-processed to obtain results such as the maxi-
mum displacements or fraction of cracked walls. To improve the computational perfor-
mance and accuracy of these steps, automatic processes have been implemented in the
Buildings 2021, 11, 200 11 of 22

Buildings 2021, 11, x FOR PEER REVIEW 11 of 23

relevant data is extracted and further post-processed to obtain results such as the maximum
displacements or fraction of cracked walls. To improve the computational performance and
accuracy of these steps, automatic processes have been implemented in the MATLAB [40]
MATLAB [40] routines
routines (1–3). For each(1–3).
of theFor each
three of the
steps three steps
mentioned mentioned
above, above,
a schematic of athese
schematic of
routines
these routines
is shown is shown
in Figure in Figure
8. Routine 8. Routine
1 was 1 was
used for used forthe
assembling assembling
required the required
geometry of ge-
the
ometry
buildingofbased
the building based on input
on a simplified a simplified input
protocol. protocol. Itgenerates
It efficiently efficientlythe generates
nodes andthe
nodes andelements
the solid the solid and
elements and connects
connects the segments
the segments to produce to produce the complete
the complete building building
model.
RoutineRoutine
model. 2 was used
2 wastoused
execute the second
to execute step ofstep
the second thisofprocess, that is,that
this process, running the static
is, running the
and dynamic
static and dynamicanalyses, which
analyses, are performed
which are performedin a in
stepwise manner.
a stepwise manner.It generates
It generatesa set of
a set
instructions
of instructionsthat cancan
that be be
runrun
from thethe
from system console
system thatthat
console controls the LS-DYNA
controls the LS-DYNA solver and
solver
can run
and can multiple analyses
run multiple simultaneously
analyses while efficiently
simultaneously segregating
while efficiently and storing
segregating and the data
storing
generated.
the Routine Routine
data generated. 3 was used to execute
3 was used tothe third the
execute stepthird
of this process:
step of this the extraction
process: of
the ex-
the relevant
traction of thedata fromdata
relevant the analyses, which, inwhich,
from the analyses, the present
in thestudy,
presentmainly
study,relates
mainlytorelates
crack
propagation
to and volume
crack propagation and loss.
volume loss.

Figure
Figure 8.
8. Procedure
Procedure utilizing
utilizing the
the automated algorithms in LS-DYNA for pre-processing,
pre-processing, processing,
pro-
cessing, and post-processing
and post-processing of the building
of the building models. models.

4.1.1. Solids
4.1.1. Solids Elements
Elements forfor Walls
Walls
To suit
To suit this
this study’s
study’s objectives,
objectives, itit is
is pertinent
pertinent to to model
model the
the collapse
collapse mechanism
mechanism of of the
the
buildings. This process allows quantifying the crack propagation and
buildings. This process allows quantifying the crack propagation and the volume loss of the volume loss of
material. According to a classification given by D’Altri et al. [41], this modeling
material. According to a classification given by D’Altri et al. [41], this modeling strategy strategy
can be
can be considered
considered aa block-based
block-based approach,
approach, and and aa similar
similar novel
novel computational modeling
computational modeling
strategy was
strategy was proposed
proposed by by Pulatsu
Pulatsu etet al.
al. [42].
[42]. To
To this
this end,
end, the
the building
building was
was discretized
discretized
uniformly into blocks, as shown in Figure 9. In this study, the buildings
uniformly into blocks, as shown in Figure 9. In this study, the buildings were modeled were modeled
utilizing three different mesh size discretization to reproduce the required
utilizing three different mesh size discretization to reproduce the required wall thickness. wall thickness.
The thickness
The thickness of of the
the walls
walls isis modeled
modeled adopting
adopting aa “two-parts”
“two-parts” scheme,
scheme, in
in order
order to
to avoid
avoid
computational instabilities. Thus, to reproduce a wall of 30 cm thickness,
computational instabilities. Thus, to reproduce a wall of 30 cm thickness, we used two we used two
blocks of 15 cm each. Furthermore, each block is constituted of two elements
blocks of 15 cm each. Furthermore, each block is constituted of two elements with a mesh with a mesh
size of
size of 7.5
7.5 cm
cm each,
each, as
as shown
shown in in Figure
Figure 3c.3c. The
The same
same approach
approach has has been
been applied
applied toto the
the
other two buildings as shown in Figures 4C and 5C. Finally, the computational cost plays
other two buildings as shown in Figures 4c and 5c. Finally, the computational cost plays
an important role, and larger mesh sizes help in reducing the cost significantly. A summary
an important role, and larger mesh sizes help in reducing the cost significantly. A sum-
of the components of the models is given in Table 3.
mary of the components of the models is given in Table 3.
Table 3. Numerical modeling characteristics.
Table 3. Numerical modeling characteristics.
Building 1 Building 2 Building 3
Building 1 Building 2 Building 3
Mesh
Meshsize (cm)
size (cm) 7.5 7.5 15 15 1010
Block size (cm) 15 30 20
Block size (cm) 15 30 20
No. of nodes 141,408 123,156 300,654
No. ofNo. of elements
solid nodes 42,223141,408 37,040123,156 300,654
89,962
No. No.
of solid elements
of parts 5162 42,223 4433 37,040 89,962
10,932
Total volume of blocks of parts the walls (m3 )
No.composing 17.80 5162 124.90 4433 89.90
10,932
Total volume of blocks Roofcomposing the walls (m3) Spring-based
17.80 Spring-based
124.90 Spring-based
89.90
Roof Spring-based Spring-based Spring-based
Buildings 2021, 11, x FOR PEER REVIEW 12 of 23
Buildings 2021, 11, 200 12 of 22

9. Procedure
Figure 9. Procedure utilizing
utilizing the
the automated
automatedalgorithms
algorithmsin
inLS-DYNA
LS-DYNAfor
forpre-processing,
pre-processing,processing,
pro-
cessing, and post-processing of the building
and post-processing of the building models. models.

The elements used in our models were LS-DYNA’s standard three-dimensional solid
The elements used in our models were LS-DYNA’s standard three-dimensional solid
elements formed by eight nodes (hexahedron) ELFORM_TYPE_1, based on linear shape
elements formed by eight nodes (hexahedron) ELFORM_TYPE_1, based on linear shape
functions using one-point integration with constant stress [10]. The material properties
functions using one-point integration with constant stress [10]. The material properties
attributed to these blocks is *MAT_001, an isotropic elastic material model [10]. This combi-
attributed to these blocks is *MAT_001, an isotropic elastic material model [10]. This com-
nation of solid elements and associated material properties proved to be efficient, enabling
bination of solid elements and associated material properties proved to be efficient, ena-
multiple non-linear time history analyses. However, while working with solid elements, it
bling multiple non-linear time history analyses. However, while working with solid ele-
is possible to develop Hourglassing (HG). This is a pathology that can potentially cause
ments, it isinconsistencies
numerical possible to develop
due Hourglassing (HG). This
to the development is a pathology
of nonphysical, that
zero can potentially
energy modes of
cause numerical
deformation inconsistencies
that produce dueand
zero strain to no
thestress
development of nonphysical,
in the solids. To control thiszero energy
numerical
modes of deformation that produce zero strain and no stress in the solids. To control
error, an HG control (IHQ4) that follows a stiffness form of type 2 Flanagan-Belytschko has this
numerical error, an HG control (IHQ4) that follows a stiffness
been utilized, with a Hourglass coefficient (QH) of 0.05 [37]. form of type 2 Flanagan-
Belytschko has been utilized, with a Hourglass coefficient (QH) of 0.05 [37].
4.1.2. Segment Contact Interface Elements and Cohesive Mixed-Mode Material Model
4.1.2.The
Segment Contact Interface
solid elements Elements
were connected andzero
using Cohesive Mixed-Mode
thickness master andMaterial Model A
slave interfaces.
penaltyThestiffness
solid elements
approach were
wasconnected usingthe
used to model zero thicknessimpact,
separation, master and
andsliding
slave interfaces.
with fric-
A penalty
tion between stiffness
blocks.approach was used to model the separation, impact, and sliding with
A *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE_TIEBREAK
friction
was usedbetween
as these blocks.
automatic A contacts
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE_TIE-
are non-oriented, and can detect penetration coming
BREAK
from wasside
either usedofas
thethese automatic
master contacts
and slave are non-oriented,
segments. The selected and canalgorithm
contact detect penetration
is based
coming
on from either[10],
the *MAT_138 sidewhich
of the has
master and slave
a Discrete segments.
Crack Model,The
andselected contact algorithm
it was assumed that the
tensile
is basedbehavior is governed
on the *MAT_138 by which
[10], a bilinear
hastraction separation
a Discrete law with
Crack Model, andlinear
it wassoftening,
assumed
as shown
that in Figure
the tensile 10a. This
behavior materialbymodel
is governed has atraction
a bilinear pre-defined failurelaw
separation criterion, and sof-
with linear the
Buildings 2021, 11, x FOR PEER REVIEW 13 of 23
cohesive
tening, aselements
shown in were removed
Figure aftermaterial
10a. This meetingmodel
the criteria.
has a No damage was
pre-defined assumed
failure in
criterion,
the
andinitial state ofelements
the cohesive the buildings. The schematic
were removed of a master
after meeting and slave
the criteria. Nonode is shown
damage was as-in
Figure
sumed10b, andinitial
in the a summary
state ofofthe
the buildings.
mechanicalThe properties andof
schematic input parameters
a master adopted
and slave nodefor is
the
shownnumerical
in Figuremodels are given
10b, and in Table
a summary of 2.
the mechanical properties and input parameters
adopted for the numerical models are given in Table 2.

(a) (b)
Figure
Figure 10.
10. Material
Material models
models in in LS-DYNA:
LS-DYNA: (a)
(a) Mixed-mode
Mixed-mode traction-separation
traction-separation law
law adapted
adapted from
from [10];
[10]; (b)
(b) assessment
assessment of
of
projection distance for a free slave node adapted from [43].
projection distance for a free slave node adapted from [43].

4.1.3. Optimization
Due to the large deformations of the solid elements from their original position, an
element may become so distorted that its volume becomes negative. This behavior in-
creases processing time and eventually causes premature termination of the analysis. A
function to remove such elements was incorporated using the *DEFINE_ELE-
Buildings 2021, 11, 200 13 of 22

4.1.3. Optimization
Due to the large deformations of the solid elements from their original position, an element
may become so distorted that its volume becomes negative. This behavior increases processing
time and eventually causes premature termination of the analysis. A function to remove such
elements was incorporated using the *DEFINE_ELEMENT_DEATH_SOLID_SET [10] option.
This functionality tracks solid blocks that fall below the base of the building and removes them
from the calculation. This optimization significantly improved the computational performance
and required storage, while maintaining a steady iterative timestep.

4.2. Eigenvalue Analysis

Eigenvalue analysis is fundamental towards understanding the fundamental period of
Buildings 2021, 11, x FOR PEER REVIEW 14 of 23
vibration of the buildings. The modes of vibration corresponding to the X and Y direction
of the building models with significant mass participations are shown in Figure 11a–f. The
local modes with corresponding low values of mass participation are not shown for the
sake of brevity.

X-direction Y-direction

T = 0.17 s T = 0.12 s
(a) (b)

T = 0.24 s T = 0.28 s
(c) (d)

T = 0.44 s T = 0.43 s
(e) (f)
Figure11.
Figure 11.Fundamental
Fundamentalperiods
periodsof ofvibration
vibrationand
andthe
thecorresponding
correspondingmodes
modesfor
fortranslation
translationin
inX-direction
X-directionand
andY-direction:
Y-direction:
(a,b)building
(a,b) building1;1;(c,d)
(c,d)building
building2;2;(e,f)
(e,f)building
building3.3.

4.3. Hazard Demand

A set of 30 bi-directional ground motions records were selected based on the local
tectonic regime in Portugal and a seismic hazard disaggregation study for the most pop-
ulated regions [44]. The IM considered was PGA, and the records were segregated into
five bins between 0.05 g and 1.05 g. Some of these records were scaled to cover the range
Buildings 2021, 11, 200 14 of 22

4.3. Hazard Demand

A set of 30 bi-directional ground motions records were selected based on the local tec-
tonic regime in Portugal and a seismic hazard disaggregation study for the most populated
regions [44]. The IM considered was PGA, and the records were segregated into five bins
between 0.05 g and 1.05 g. Some of these records were scaled to cover the range of PGA
required to induce collapse. A limit was set for the duration of the records, since it was
not computationally feasible to use ground motion records that exceeded 60 s in duration.
Buildings 2021, 11, x FOR PEER REVIEW 15 of 23
Figure 12 illustrates the acceleration spectra for the selected ground motion records along
with the median spectrum and the 16th and 84th percentile spectra.

(a) (b)
Figure
Figure 12.12.Response
Responsespectra
spectraof
of the
the bi-directional
bi-directional components
components ofofthe
theselected
selectedground
groundmotion
motionrecords: (a) (a)
records: horizontal com-
horizontal
ponent in X-direction; (b) horizontal component in Y-direction.
component in X-direction; (b) horizontal component in Y-direction.

4.4.
4.4. Non-Linear
Non-Linear TimeTime History
History Analysis
Analysis
Non-linear
Non-linear time
time history
history analyses
analyses were
were performed
performed onon
allall three
three building
building models
models using
using
the
the setofofselected
set selectedground
ground motion
motion records
recordsin inthe
thetwo twoorthogonal
orthogonal directions.
directions.A gravity
A gravityanal-
ysis prior
analysis priorto to
a non-linear
a non-linear time history
time historyanalysis
analysis ensures
ensuresthatthat
thethe
model
modelis preloaded
is preloaded prior
to applying
prior to applying dynamic
dynamic loads. In this
loads. manner,
In this manner, it isitpossible
is possible to clearly
to clearlydistinguish
distinguishthethepre-
loading phase from the transient phase [45]. The pre-loading
preloading phase from the transient phase [45]. The pre-loading was performed using a was performed using a high
damping
high damping in in
orderordertotogetgetaasoft
soft response
response (pseudo-static)
(pseudo-static) to to avoid
avoidanyanydynamic
dynamiceffects,
effects,
whereas
whereas forfor
thethe non-linear
non-linear time
time histories,
histories, a damping
a damping of of
3%3%was was assumed
assumed [31].
[31].
InIn theseanalyses,
these analyses,ititwas was expected
expected that depending
depending on onthetheintensity
intensityofof the ground
the ground mo-
tion records,
motion records,the the contact
contact interfaces develop cracks cracks due duetotothetheenergy
energyinput.
input.This
This effect
effect
causes
causes thethe solid
solid elements
elements of the
of the walls
walls get get displaced
displaced from from
theirtheir original
original position,
position, and trig-
and trigger
ger a combined collapse mechanism. The failure modes
a combined collapse mechanism. The failure modes of these buildings under seismic of these buildings under seismic
loading
loading depend
depend onon thethe mechanical
mechanical properties
properties andand ononthethe overall
overall geometric
geometric configuration
configuration
ofof
thethe building
building [17,46].
[17,46]. Examples
Examples ofof thethe end
end stage
stage of of
thethe non-linear
non-linear time
time history
history analyses
analyses
according
according to to
thethe different
different assumed
assumed damage
damage states
states areare shown
shown in in Figure
Figure 13a–o.
13a–o.
 Buildings 2021, 11, x FOR PEER REVIEW 16 of 23
Buildings 2021, 11, 200 15 of 22

Building 1 Building 2 Building 3

DS 1

(a) (b) (c)

DS 2

(d) (e) (f)

DS 3

(g) (h) (i)

DS 4

(j) (k) (l)

Figure 13. Cont.

Buildings 2021, 11, x FOR PEER REVIEW 17 of 23
Buildings 2021, 11, 200 16 of 22
DS 5

(m) (n) (o)

Figure
Figure 13. Crack
13. Crack formation
formation and volume
and volume loss shown
loss shown at theatend
thestage
end stage
of theof the non-linear
non-linear time history
time history analyses
analyses sustained
sustained by
by buildings 1–3
buildings 1–3 for the 5 damage states: (a,d,g,j,m) building 1; (b,e,h,k,n) building 2;
for the 5 damage states: (a,d,g,j,m) building 1; (b,e,h,k,n) building 2; (c,f,i,l,o) building 3. (c,f,i,l,o) building 3.

4.5.
4.5.EDPs
EDPsand andDSs
DSs
ToTo compute the
compute probability of
the probability ofexceeding
exceedinga asetset ofof damage
damage states,
states, it isitnecessary
is necessary to
to estab-
establish clear criteria relating each damage state with an engineering threshold
lish clear criteria relating each damage state with an engineering threshold (i.e., engineering (i.e., en-
gineering demand parameter).
demand parameter). We adopted Wethe
adopted
damage the damage
states fromstates from theMacrosceismic
the European European Macro- Scale
sceismic
(EMS 98)Scale[11],(EMS 98) [11],
and defined and defined
a damage a damage
criteria based oncriteria based(1)on
two EDPs: thetwo EDPs:
crack (1) the
propagation
crack
ratio,propagation
which refersratio,
to thewhich
cracks refers to the cracks
that develop on the that
wallsdevelop
prior to on
thethe
onsetwalls prior to theof
of detachment
onset of detachment of the blocks, and (2) the volume loss ratio, which
the blocks, and (2) the volume loss ratio, which refers to the post-failure movements refers to the post-
of the
failure movements of the blocks that have the potential to cause fatalities.
blocks that have the potential to cause fatalities. The two novel EDPs are sub-categorized The two novel
EDPs are sub-categorized
into the five damage states into(DS1
the five damage
to DS5) states4.(DS1
in Table Thetothresholds
DS5) in Table 4. The
for the thresh-
novel EDPs
olds for the novel EDPs were defined following the post-earthquake
were defined following the post-earthquake damage observations from So and Pomonis damage observations
from
(2012)So[47],
andinPomonis (2012) volume
which typical [47], in which
loss fortypical
collapse volume loss formasonry
of European collapsebuildings
of European were
masonry
indicated. buildings
The datawere indicated.
generated from The data generated
the non-linear from the
time history non-linear
analyses time history
pertaining to crack
analyses
propagationpertaining to crack
and volume propagation
loss and volume
were extracted using the loss were extracted
automated MATLAB using the auto-
routines and
mated MATLAB routines
segregated for further analysis.and segregated for further analysis.

Table
Table4.4.EDPs
EDPsand
andDSs
DSsand
andthe
thecorresponding
correspondingdamage
damagethresholds.
thresholds.
EDPs DSs
EDPs DSs
Damage Description
Damage Description
Threshold
Threshold
DS 1 Negligible to slight 15%
Crack propagation ratio DS 1 Negligible to slight 15%
Crack propagation ratioDS 2 Moderate 25%
DS 2 Moderate 25%
DS 3 Substantial to heavy 10%
DS 3 Substantial to heavy 10%
Volume loss ratio Volume loss ratio
DS 4 DS 4 Very heavy
Very heavy 25%
25%
DS 5 DS 5 Destruction
Destruction 40%
40%

4.6.
4.6.Cloud
CloudAnalysis
Analysis
Fragility
Fragilityassessment
assessmentisisan animportant
importantstep stepininseismic
seismicrisk
riskassessment
assessment[12]. [12].Fragility
Fragility
functions define the conditional probability of exceeding a damage
functions define the conditional probability of exceeding a damage state for state for a given
a givenIM.
IM.
Incremental Dynamic Analysis (IDA) [48] and Multiple Stripe Analysis
Incremental Dynamic Analysis (IDA) [48] and Multiple Stripe Analysis (MSA) [49] are (MSA) [49] are
commonly
commonlyused used in in
fragility studies.
fragility However,
studies. However, thesethese
methods usually
methods requirerequire
usually a large acom-
large
putational
computational effort due to the need to scale the ground motion records, and aperform
effort due to the need to scale the ground motion records, and perform large
number
a large of dynamic
number of analysis.
dynamic The issue of
analysis. Thecomputational effort is particularly
issue of computational important
effort is particularly
inimportant
the current study, as the detailed numerical models are computationally-demanding.
in the current study, as the detailed numerical models are computationally-
Thus, for the sake
demanding. Thus,offorsimplicity,
the sake ofthesimplicity,
Cloud analysis
the Cloudprocedure
analysiswas adoptedwas
procedure [12]. This pro-
adopted [12].
cedure has been used
This procedure in recent
has been usedstructural
in recentfragility
structuralstudies and itstudies
fragility is capable
andof it considering
is capable of
the record-to-record
considering variability and variability
the record-to-record other sources andofother
uncertainty
sources related to structural
of uncertainty mod-to
related
eling. The steps followed for the fragility assessment are described below,
structural modeling. The steps followed for the fragility assessment are described below, and illustrated
inand
Figure 14.
illustrated in Figure 14.
Buildings 2021, 11, x FOR PEER REVIEW 18 of 23

Buildings 2021, 11, 200 17 of 22

1. The results extracted from the non-linear time history analysis were plotted (IM vs.
EDP) in a logarithmic scale, and a linear regression was performed to generate a
1. trendline thatextracted
The results best fits the data.
from the non-linear time history analysis were plotted (IM vs.
EDP)
2. The in a logarithmic
following equation wasscale,used
andto
a linear regression
calculate was performed
the expected to dependent
value of the generate a
trendline
variable thatgiven
(EDP) best fits
an the
IM.data.
A homogeneity of variance was assumed for the IM-EDP
2. random
The following equation
variables; hence thewas used can
model to calculate the expected
be described as shownvalue of the dependent
in Equation (1).
variable (EDP) given an IM. A homogeneity of variance was assumed for the IM-EDP
random variables; 𝐸 hence
log 𝐸𝐷𝑃the |𝐼𝑀
model 𝑏can log 𝐼𝑀 logas𝑎 shown in Equation (1).
be described

∑ ( EDP
log)|𝐸𝐷𝑃 𝐸 log 𝐸𝐷𝑃 |𝐼𝑀 (1)
𝜎
E[log r I M ] = b × log( I M ) + log( a)
∑in=1 (log𝑛( EDP2i )−E[log(EDP)| I Mi ])2 (1)
σrec−rec = n −2
where 𝐸 log 𝐸𝐷𝑃 |𝐼𝑀 stands for the expected logarithm of EDP given an IM, b and
log(a)where
are theE[regression
log( EDP)|parameters,
I M ] stands for𝜎 the expected
is the record-to-record
logarithm of EDPvariability,
given an IM, n is the
b and
no. oflog(a)
records,
are the 𝐸𝐷𝑃 corresponds
andregression parameters,to σthe
rec−i-EDP
rec is thevalue obtained from
record-to-record the non-linear
variability, n is the
analysis
no.for the corresponding
of records, 𝐼𝑀 .
and EDPi corresponds to the i-EDP value obtained from the non-linear
3. Theanalysis for thefragility
structural corresponding
functions I Mobtained
i. from the probabilistic model can be ex-
3. pressed
The structural fragility
using Equation (2). functions obtained from the probabilistic model can be ex-
pressed using Equation (2).
𝐸𝐷𝑃 𝐸 log 𝐸𝐷𝑃|𝐼𝑀
𝑃 𝐸𝐷𝑃 𝑑𝑠 |𝐼𝑀 1 Φ 
EDP𝜎dsi − E[log( EDP| I M )]
 (2)
P[ EDP ≥ dsi | I M ] = 1 − Φ (2)
σrec−rec
where Φ is the cumulative normal standard distribution and the 𝐸𝐷𝑃 is the damage
where
threshold Φ is15%)
(e.g., the cumulative
correspondingnormal
to astandard
given damage distribution and the
state (e.g., DS1).EDPdsi is the damage
threshold (e.g., 15%) corresponding to a given damage state (e.g., DS1).

(a) (b)
Figure
Figure 14.
14. A
Aschematic
schematic ofof the
thefragility
fragility assessment
assessment procedure:
procedure: (a)
(a) probability
probability of
of exceedance
exceedance for
for three
three IMLs
IMLs given
given aa damage
damage
threshold; (b) Fragility function for a given damage state.
threshold; (b) Fragility function for a given damage state.

4.7.
4.7. Fatality
FatalityVulnerability
VulnerabilityFunctions
Functions
The
The method to
method toestimate
estimatefatality
fatality rates
rates in
inthis
thisstudy
studyfollows
followsthethemethodology
methodology proposed
proposed
by
by Abeling and Ingham [9], which uses a relationship between the volume loss of
Abeling and Ingham [9], which uses a relationship between the volume loss of the
the
building and an expected fatality ratio. The buildings volume loss is a damage
building and an expected fatality ratio. The buildings volume loss is a damage descriptor descriptor
that
that measures
measures thethe ratio
ratio between
between the the volume
volume of debris (V
of debris (Vd)) and the space capacity (Vc)
d and the space capacity (Vc )
measured
measured up-to a height of 2 m from the floor level, as shown in Figure
up-to a height of 2 m from the floor level, as shown in Figure 15.
15. The
The volume
volume
loss can be calculated using Equation
loss can be calculated using Equation (3). (3).

Vd
Volume loss = (3)
Vc
Buildings 2021, 11, x FOR PEER REVIEW 19 of 23

Buildings 2021, 11, 200 18 of 22

𝑉
𝑉𝑜𝑙𝑢𝑚𝑒 𝑙𝑜𝑠𝑠 (3)
𝑉
A cloud of fatality ratio (FR)—IM data was produced by multiplying the volume loss
A cloud of fatality ratio (FR)—IM data was produced by multiplying the volume loss
obtained from each of the structural analysis by the corresponding FR proposed by Abeling
obtained from each of the structural analysis by the corresponding FR proposed by Abel-
and Ingham [9]. This cloud of points was then fitted to a lognormal cumulative density
ing and Ingham [9]. This cloud of points was then fitted to a lognormal cumulative density
function, as defined in Equation (4)
function, as defined in Equation (4)
( I M) −𝜃 θ
 
Fatality ratio = 0.52 × Φ lnln𝐼𝑀 (4)
𝐹𝑎𝑡𝑎𝑙𝑖𝑡𝑦 𝑟𝑎𝑡𝑖𝑜 0.52 Φ (4)
𝛽β
where
where 0.52
0.52 is the maximum
maximum fatality
fatalityratio
ratiocorresponding
correspondingtoto100%
100%volume
volume loss
loss asas
perper Φ
[9],[9],
Φis the cumulative normal standard distribution, θ is the logarithmic mean, and β is
is the cumulative normal standard distribution, θ is the logarithmic mean, and β is the
the
logarithmic
logarithmicstandard
standarddeviation,
deviation,as
aspresented
presentedininTable
Table5.5.

Figure15.
Figure 15. Illustration
Illustration of
of volume
volume loss
lossusing
usingBuilding
Building22(adapted
(adaptedfrom
fromOkada
Okada[50]).
[50]).

5. Results
Table 5. Fatality vulnerability parameters for adobe buildings.
The results of this study for the three adobe buildings are presented in terms of the
Building Class θ* β **
derived fragility and fatality vulnerability functions, as shown in Figure 16a–f with com-
parisons to1-story
the recent study by Martins and 1.06
Silva (2020) [7]. The corresponding0.88
fragility
2-story 0.45 0.75
parameters (i.e., µ—logarithmic
2-story plus attic
mean and σ—logarithmic
0.45
standard deviation)
0.77
are pre-
sented in Table 6 and the fatality vulnerability
* θ—logarithmic mean, ** β—logarithmic standard deviation.
parameters are presented in Table 5. It can
be inferred from the values of dispersion (i.e., σ) in Table 6, that there are two different
5. Results
values for the dispersion, one value corresponding to DS1-DS2 and another for DS3-DS5.
The Thedifference
resultsarises due
of this to thefor
study twotheEDPs
threeselected to predict are
adobe buildings the different
presentedphenomenon.
in terms of
It isderived
the intuitive to understand
fragility and fatalitythat vulnerability
a building can develop as
functions, cracks
shown after
in which
Figurevolume
16a–f withloss
occurs. The 1-story building is less vulnerable in the first damage
comparisons to the recent study by Martins and Silva (2020) [7]. The corresponding states, due to the absence
of a story
fragility above and(i.e.,
parameters the µ—logarithmic
low height. It becomesmean and vulnerable in the last
σ—logarithmic damage
standard states, since
deviation) are
post-cracking,
presented the walls
in Table 6 andlose their load
the fatality bearing capacity,
vulnerability and these
parameters blocks arein
are presented immediately
Table 5. It
displaced
can fromfrom
be inferred theirthe
original
valuesposition. Also,(i.e.,
of dispersion the elastic period
σ) in Table of the
6, that buildings
there are twohas an in-
different
fluencefor
values onthe
DS1 and DS2, one
dispersion, while DS3-DS5
value are in the to
corresponding non-linear
DS1-DS2range. According
and another to Ginell
for DS3-DS5.
and difference
The Tolles [32]arises
the post-elastic
due to thebehavior
two EDPs ofselected
adobe isto different
predict from ductile building
the different phenomenon. mate-
Itrials,
is intuitive to understand that a building can develop cracks after
since it is a brittle material characterized by low compressive, tensile, and shear which volume loss
occurs.
strength. TheIt 1-story
is worthbuilding
noting thatis less
allvulnerable in the
the buildings arefirst damage states,
unreinforced, anddue
theirtoload
the absence
bearing
of a story above
capacities and the
rests solely onlow
theheight.
integrity It becomes vulnerable in the last damage states, since
of the walls.
post-cracking, the walls lose their load bearing capacity, and these blocks are immediately
displaced from their original position. Also, the elastic period of the buildings has an
influence on DS1 and DS2, while DS3-DS5 are in the non-linear range. According to
Ginell and Tolles [32] the post-elastic behavior of adobe is different from ductile building
materials, since it is a brittle material characterized by low compressive, tensile, and shear
strength. It is worth noting that all the buildings are unreinforced, and their load bearing
capacities rests solely on the integrity of the walls.
 Buildings 2021, 11, x FOR PEER REVIEW 20 of 23
Buildings 2021, 11, 200 19 of 22

Fragility function Fatality vulnerability function

(a) (b)

(c) (d)

(e) (f)
Figure
Figure 16.16. Fragility
Fragility andand fatality
fatality vulnerability
vulnerability function
function for each
for each of theofbuilding
the building classes,
classes, respectively
respectively (a,b) 1-story;
(a,b) 1-story; (c,d) 2-
(c,d) 2-story;
story; (e,f) 2-story plus an attic.
(e,f) 2-story plus an attic.
Buildings 2021, 11, 200 20 of 22

Table 6. Fragility parameters for adobe buildings.

DS 1 DS 2 DS 3 DS 4 DS 5
Building Class IM
µ* σ ** µ σ µ σ µ σ µ σ
1-story PGA −0.89 0.39 −0.67 0.39 −0.43 0.37 −0.20 0.37 −0.08 0.37
2-story PGA −1.38 0.43 −1.00 0.43 −0.38 0.50 −0.05 0.50 0.12 0.50
2-story plus attic PGA −1.39 0.42 −0.96 0.42 −0.46 0.49 −0.17 0.49 −0.02 0.49
* µ—logarithmic mean, ** σ—logarithmic standard deviation.

6. Conclusions
This study presented the development of fragility and vulnerability functions for
typical adobe buildings in Portugal, one of the most vulnerable types of construction in
the country. The selection of the buildings to numerically model was carried out based
on survey data comprising building drawings, photographs, and information regarding
the construction practice and material properties. The sensitivity of mechanical properties
and modeling options was assessed, and calibration and validation of material properties
were carried out following the results from a full-scale experimental in-plane cyclic tests
on adobe walls. A set of 30 ground motion records with bi-directional components were
selected, and non-linear time-history analyses were performed until collapse was reached.
Two novel EDPs were adopted and damage thresholds were proposed such that they can
be correlated with damage classifications in agreement with the European Macroseismic
Scale (EMS-98) [11]. Fragility functions were then derived following the cloud analysis
methodology [12] and compared to a recent study [7].
The fragility functions proposed herein can be used directly in the assessment of
damage due to earthquake scenarios, or in probabilistic risk analysis to identify regions
in the country where the probability of collapse of these structures is particularly high.
Similarly, the vulnerability functions in terms of fatality rates can be employed to assess
human losses, which is particularly important for the development of risk management
measures.

Author Contributions: Conceptualization, V.S., R.V. and T.M.F.; methodology, S.M., H.L., V.S., R.V.
and T.M.F.; software, H.L., S.M.; validation, S.M., H.L., V.S., T.M.F. and R.V.; investigation, S.M., H.L.;
resources, R.V.; data curation, H.L., S.M.; writing—original draft preparation, S.M.; writing—review
and editing, T.M.F., R.V., H.L. and V.S.; visualization, S.M., H.L.; supervision, V.S., R.V. and T.M.F.;
project administration, R.V.; funding acquisition, R.V., V.S. and T.M.F. All authors have read and
agreed to the published version of the manuscript.
Funding: This work was funded by the Foundation for Science and Technology (FCT)—Aveiro
Research Centre for Risks and Sustainability in Construction (RISCO), Universidade de Aveiro,
Portugal in the framework of research project PTDC/ECI-EST/31865/2017 MitRisk—Framework for
seismic risk reduction resorting to cost-effective retrofitting solutions.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.

References
1. Sousa, G.; Gallaecia, E.S. Historical Seismicity in Portugal. In Seismic Retrofitting: Learning from Vernacular Architecture; Taylor &
Francis: Abingdon, UK, 2015.
2. Choffat, P. Le tremblement de terre du 23 Avril 1909 dans le Ribatejo. In Revista de Obras Pu´blicas e Minas, Tomo XLIII; Imprensa
Nacional: Lisboa, Spain, 1912. (In French)
Buildings 2021, 11, 200 21 of 22

3. Tarque, N.; Crowley, H.; Pinho, R.; Varum, H. Displacement-Based Fragility Curves for Seismic Assessment of Adobe Buildings
in Cusco, Peru. Earthq. Spectra 2012, 28, 759–794. [CrossRef]
4. Crowley, H.; Pinho, R.; Bommer, J.J. A Probabilistic Displacement-based Vulnerability Assessment Procedure for Earthquake Loss
Estimation. Bull. Earthq. Eng. 2004, 2, 173–219. [CrossRef]
5. Ahmad, N.; Ali, Q.; Crowley, H.; Pinho, R. Earthquake loss estimation of residential buildings in Pakistan. Nat. Hazards 2014, 73,
1889–1955. [CrossRef]
6. Villar-Vega, M.; Silva, V.; Crowley, H.; Yepes, C.; Tarque, N.; Acevedo, A.B.; Hube, M.A.; Gustavo, C.D.; María, H.S. Development
of a Fragility Model for the Residential Building Stock in South America. Earthq. Spectra 2017, 33, 581–604. [CrossRef]
7. Martins, L.; Silva, V. Development of a fragility and vulnerability model for global seismic risk analyses. Bull. Earthq. Eng. 2020,
2020, 1–27. [CrossRef]
8. Sumerente, G.; Lovon, H.; Tarque, N.; Chácara, C. Assessment of Combined In-Plane and Out-of-Plane Fragility Functions for
Adobe Masonry Buildings in the Peruvian Andes. Front. Built Environ. 2020, 6, 52. [CrossRef]
9. Abeling, S.; Ingham, J.M. Volume loss fatality model for as-built and retrofitted clay brick unreinforced masonry buildings
damaged in the 2010/11 Canterbury earthquakes. Structures 2020, 24, 940–954. [CrossRef]
10. Hallquist, J.O. Livermore Software Technology Corporation (Lstc), Ls-Dyna Keyword User’s Manual Volume I; 2018; Volume II. Available
online: https://ftp.lstc.com/anonymous/outgoing/jday/manuals/LS-DYNA_Manual_Volume_I_R11.pdf (accessed on 15
March 2021).
11. Comisión Sismológica Europea. Escala Macro Sísmica Europea EMS-98; 1998; Volume 15. Available online: http://media.gfz-
potsdam.de/gfz/sec26/resources/documents/PDF/EMS-98_Original_englisch.pdf (accessed on 15 March 2021).
12. Jalayer, F.; De Risi, R.; Manfredi, G. Bayesian Cloud Analysis: Efficient structural fragility assessment using linear regression. Bull.
Earthq. Eng. 2015, 13, 1183–1203. [CrossRef]
13. De Estatística, I.N. Census Data 2011. Lisbon. 2011. Available online: http://censos.ine.pt/ (accessed on 15 March 2021).
14. Costa, C.; Arduin, D.; Rocha, F.; Velosa, A. Adobe Blocks in the Center of Portugal: Main Characteristics. Int. J. Arch. Herit. 2019,
15, 467–478. [CrossRef]
15. Correia, M.; Carlos, G.; Merten, J.; Viana, D.; Rocha, S. Vernacular Heritage and Earthen Architecture; Taylor & Francis Group:
London, UK, 2013; pp. 111–116. [CrossRef]
16. Varum, H.; Figueiredo, A.; Silveira, D.; Martins, T.; Costa, A. Investigaciones realizadas en la Universidad de Aveiro sobre
caracterización mecánica de las construcciones existentes en adobe en Portugal y propuestas de rehabilitación y refuerzo.
Resultados alcanzados. Inf. Constr. 2011, 63, 127–142. [CrossRef]
17. Varum, H.; Costa, A.; Fonseca, J.; Furtado, A. Behaviour Characterization and Rehabilitation of Adobe Construction. Procedia Eng.
2015, 114, 714–721. [CrossRef]
18. Varum, H.; Costa, A.; Velosa, A.; Martins, T.; Pereira, H.; Almeida, J. Caracterização mecânica e patológica das construções em
Adobe no distrito de Aveiro como suporte em intervenções de reabilitação. In Houses and Cities Built with Earth: Conservation,
Significance and Urban Quality; CULTURA 2000—Education and Culture, European Union Program; ARGUMEMTUM: Monsaraz,
Portugal, 2006; pp. 41–45.
19. Silveira, D.; Varum, H.; Costa, A.; Neto, C. Survey of the Facade Walls of Existing Adobe Buildings. Int. J. Arch. Herit. 2016, 10,
867–886. [CrossRef]
20. IPQ. NP EN 1998-1. Eurocódigo 8: Projecto de Estruturas Para Resistência Aos Sismos. In Parte 1: Regras Gerais, Acções Sísmicas e
Regras Para Edifícios; Instituto Português da Qualidade: Caparica, Portugal, 2010; p. 230. (In Portuguese)
21. Braga, A.M.; Estêvão, J.M.C. Os Sismos E a Construção Em Taipa No Algarve. In Sísmica 2010–8◦ Congr. Sismol. e Eng. Sísmica;
University of Aveiro: Aveiro, Portugal, 2010; pp. 1–13.
22. Correia, M.R. Rammed Earth in Alentejo; Argumentum: Lisbon, Portugal, 2007.
23. Cancela, D.C.P. Comportamento Higrotérmico e Monitorização de Construções em Adobe. Master’s Thesis, Civil Engineering,
Univeristy of Aveiro, Aveiro, Portugal, 2013.
24. EN 1995-1-2:2004. Part 1–2: General—Structural fire design. In Eurocode 5: Design of Timber Structures; CEN: Brussels, Belgium,
2004.
25. Battaglia, L.; Ferreira, T.M.; Lourenço, P.B. Seismic fragility assessment of masonry building aggregates: A case study in the old
city Centre of Seixal, Portugal. Earthq. Eng. Struct. Dyn. 2021, 50, 1358–1377. [CrossRef]
26. Silveira, D.; Varum, H.; Costa, A.; Martins, T.; Pereira, H.; Almeida, J. Mechanical properties of adobe bricks in ancient
constructions. Constr. Build. Mater. 2012, 28, 36–44. [CrossRef]
27. Silveira, D.; Varum, H.; Costa, A. Influence of the testing procedures in the mechanical characterization of adobe bricks. Constr.
Build. Mater. 2013, 40, 719–728. [CrossRef]
28. Silveira, D.; Varum, H.; Costa, A.; Pereira, H.; Sarchi, L.; Monteiro, R. Seismic behavior of two Portuguese adobe buildings: Part
I—in-plane cyclic testing of a full-scale adobe wall. Int. J. Arch. Herit. 2018, 12, 922–935. [CrossRef]
29. Sarchi, L.; Varum, H.; Monteiro, R.; Silveira, D. Seismic behavior of two Portuguese adobe buildings: Part II—numerical modeling
and fragility assessment. Int. J. Arch. Herit. 2018, 12, 936–950. [CrossRef]
30. NIKER. New Integrated Knowledge Based Approaches to the Protection of Cultural Heritage from Earthquake-Induced Risk; Research
Report; University of Padova: Padova, Italy, 2010.
Buildings 2021, 11, 200 22 of 22

31. Karanikoloudis, G.; Lourenço, P.B. Structural assessment and seismic vulnerability of earthen historic structures. Application of
sophisticated numerical and simple analytical models. Eng. Struct. 2018, 160, 488–509. [CrossRef]
32. Ginell, W.S.; Tolles, E.L. Seismic Stabilization of Historic Adobe Structures. J. Am. Inst. Conserv. 2000, 39, 147–163. [CrossRef]
33. Vargas, J.; Bariola, J.; Blondet, M.; Mehta, P.K. Seismic strength of adobe masonry. Mater. Struct. 1986, 19, 253–258. [CrossRef]
34. Blondet, M.; Ginocchio, F.; Marsh, C.; Ottazzi, G.; Villa-Garcia, G.; Yep, J. Shaking Table Test of Improved Adobe Masonry Houses.
In Proceedings of the 9th World Conference on Earthquake Engineering, Kyoto, Tokyo, Japan, 2–6 August 1988.
35. Illampas, R.; Ioannou, I.; Charmpis, D.C. Adobe bricks under compression: Experimental investigation and derivation of
stress–strain equation. Constr. Build. Mater. 2014, 53, 83–90. [CrossRef]
36. Lourenço, P.B.; Pereira, J.M. Seismic Retrofitting Project Recommendations for Advanced Modeling of Historic Earthen Sites; Getty
Conservation Institute: Los Angeles, CA, USA; TecMinho–University of Minho: Guimarães, Portugal, 2018.
37. Alshawa, O.; Sorrentino, L.; Liberatore, D. Simulation Of Shake Table Tests on Out-of-Plane Masonry Buildings. Part (II):
Combined Finite-Discrete Elements. Int. J. Arch. Herit. 2016, 11, 1–15. [CrossRef]
38. Hazay, M.; Munjiza, A. Introduction to the Combined Finite-Discrete Element Method. In Computational Modeling of Masonry
Structures Using the Discrete Element Method; IGI Global: Hershey, PA, USA, 2016; pp. 123–145. ISBN 1522502319.
39. Dis, C. LS-DYNA/LS-PrePost Ex. 0. Introduction 2011, 31 (Suppl. S1), 1–74.
40. T.M. Inc. MATLAB; T.M. Inc.: Natick, MA, USA, 2018.
41. D’Altri, A.M.; Sarhosis, V.; Milani, G.; Rots, J.; Cattari, S.; Lagomarsino, S.; Sacco, E.; Tralli, A.; Castellazzi, G.; De Miranda,
S. Modeling Strategies for the Computational Analysis of Unreinforced Masonry Structures: Review and Classification. Arch.
Comput. Methods Eng. 2020, 27, 1153–1185. [CrossRef]
42. Pulatsu, B.; Erdogmus, E.; Lourenço, P.B.; Lemos, J.V.; Tuncay, K. Simulation of the in-plane structural behavior of unreinforced
masonry walls and buildings using DEM. Structure 2020, 27, 2274–2287. [CrossRef]
43. Bala, S. Tie-Break Contacts in LS-DYNA; Livemore Software: Livermore, CA, USA, 2007.
44. Sousa, L.; Marques, M.; Silva, V.; Varum, H. Hazard Disaggregation and Record Selection for Fragility Analysis and Earthquake
Loss Estimation. Earthq. Spectra 2017, 33, 529–549. [CrossRef]
45. Hallquist, J.O. LS Dyna Theory Manual; Livemore Software: Livermore, CA, USA, 2006.
46. Costa, A.; Humberto, V.; João, M.G. Structural Rehabilitation of Old Buildings; Springer: Berlin/Heidelberg, Germany, 2014.
47. Pomonis, A. Derivation of Globally Applicable Casualty Rates for use in Earthquake Loss Estimation Models. In Proceedings of
the 15th World Conference Earthquake Engineering, Lisbon, Portugal, 24–28 September 2012.
48. Vamvatsikos, D.; Cornell, C.A. Incremental dynamic analysis. Earthq. Eng. Struct. Dyn. 2002, 31, 491–514. [CrossRef]
49. Vamvatsikos, D.; Cornell, C.A. Direct Estimation of Seismic Demand and Capacity of Multidegree-of-Freedom Systems through
Incremental Dynamic Analysis of Single Degree of Freedom Approximation. J. Struct. Eng. 2005, 131, 589–599. [CrossRef]
50. Okada, S. Description for indoor space damage degree of building in earthquake. In Proceedings of the Eleventh World
Conference on Earthquake Engineering, Acapulco, Mexico, 23–28 June 1996.