You are on page 1of 16

Construction and Building Materials 249 (2020) 118743

Contents lists available at ScienceDirect

Construction and Building Materials


journal homepage: www.elsevier.com/locate/conbuildmat

Urgency of repair of building elements: Prediction and influencing


factors in façade renders
Clara Pereira ⇑, Ana Silva, Jorge de Brito, José D. Silvestre
CERIS, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

h i g h l i g h t s

 In the context of building inspection, the urgency of repair is analysed.


 The study is based on a sample of 52 rendered surfaces with 94 detected defects.
 Descriptive statistics, principal components and cluster analyses are applied.
 A multiple linear regression model is built to predict the urgency of repair.
 Affected area and type of defect and façade are the model’s independent variables.

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

Article history: Within building inspection, determining the urgency of repair depends on varying factors (e.g. risk, pro-
Received 20 May 2019 gression conditions and façade importance). This study analyses the urgency of repair of defects in ren-
Received in revised form 16 January 2020 dered façades based on the inspection of 52 surfaces where 94 defects were detected, defining a
Accepted 12 March 2020
methodology to predict the repair urgency of defects and building elements. Descriptive statistics are
applied, and principal components and cluster analyses are performed. Through multiple linear regres-
sion, ‘‘repair urgency” is predicted as a function of the defects’ characteristics, which are, in the best pre-
Keywords:
diction model, ‘‘defect percentage area”, ‘‘type of defect” and ‘‘type of façade”.
Building pathology
Building assessment systems
Ó 2020 Elsevier Ltd. All rights reserved.
Cluster analysis
Deterioration
Façades
Multiple linear regression analysis
Principal components analysis
Repair urgency

1. Introduction system implied the harmonisation of the classification of defects,


probable causes, diagnosis methods and repair techniques. Accord-
At Instituto Superior Técnico (IST) - University of Lisbon (UL), a ingly, files of defects, diagnosis methods and repair techniques
global inspection system of non-structural elements of the building were also harmonised, as was the inspection form. To assess the
envelope is under development. This system embraces the inspec- general operability of the preliminary version of the unified inspec-
tion methodologies applied to several building elements and mate- tion system, an inspections’ campaign was performed in current
rials, namely: external claddings of pitched roofs [1]; flat roofs [2]; buildings, in which data on 12 types of building elements and
adhesive ceramic tiling [3]; natural stone claddings [4]; wood materials were gathered, totalling 225 elements/surfaces. The pre-
floorings [5]; door and window frames [6]; epoxy resin floorings sent study uses part of the sample of this inspections’ campaign,
[7]; vinyl and linoleum floorings [8]; wall renders [9]; external only comprising inspected wall renders (52 surfaces). From their
thermal insulation composite systems (ETICS) [10]; painted analysis, 94 defects were detected (average of 1.81 defects per sur-
façades [11]; and architectural concrete surfaces [12]. The unifica- face). The present study uses a broader sample than that of Pereira
tion of these partial inspection systems in a global inspection et al. [13].
The compiled data, and the comprehensiveness of the
selected variables on degradation, allowed performing statistical
⇑ Corresponding author.
processing and analysis. In this paper, the analysis focuses
E-mail addresses: clareira@sapo.pt (C. Pereira), jb@civil.ist.utl.pt (J. de Brito),
jose.silvestre@tecnico.ulisboa.pt (J.D. Silvestre). specifically on repair urgency of the detected anomalies. The

https://doi.org/10.1016/j.conbuildmat.2020.118743
0950-0618/Ó 2020 Elsevier Ltd. All rights reserved.
2 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

urgency to repair a defect may be distinguished from its sever- building elements, and that concept (Sw) is included in the pre-
ity, or seriousness [14–16]. While severity only refers to the sent study, but the slight disparity of concepts should be high-
degradation level of the phenomenon [17–20], repair urgency, lighted. In simpler terms, severity may refer to the level of
in this case, includes the concept of aesthetical value, consider- intervention whilst repair urgency to requiring swift action
ing that the same defect in different surfaces, e.g. one more vis- [22], at times including the notion of risk [23].
ible or important to the building than the other, should have Accurately assessing the repair urgency may be used to update
different levels of repair urgency—the defect in a more important maintenance plans [24,25], as the initially scheduled actions may
surface should be repaired first. Silva et al. [21] use the concept need to be anticipated or delayed and decision-making requires
of severity of degradation (Table 1) to predict the service life of data to prioritise interventions [20].

Table 1
Summary of selected variables.

Variable
Group Name Code Unit Type Description
Characteristics of Type of defect D – categorical Type of deterioration of the rendered surface, according to the predetermined
the defect variable classification of defects (Table 2).
Defect area DAr m2 continuous Area affected by the defect.
variable
Defect DPctAr % continuous Percentage of surface area affected by the defect.
percentage variable
area
Average ThickCr mm continuous Average thickness of cracks (only in crack-type defects).
thickness of variable
cracks
Diagnosis of the Repair ReUrg months discrete The urgency of repair is expressed as the admissible number of months until the defect
defect urgency variable is repaired. 1
Direct causes DiC – categorical Main causes of the deterioration process. For each defect, 1 to 4 direct causes were
variable appointed. 2
Indirect InC – categorical Secondary causes related with the first stages of the deterioration process. For each
causes variable defect, 1 to 5 indirect causes were appointed. 3
Diagnosis DiagM – categorical In situ diagnosis methods that may be useful to deepen the knowledge about the defect.
methods variable For each defect, 0 to 3 diagnosis methods were appointed (optional field). 4
Repair ReT – categorical Techniques advised to eliminate the causes and repair the defects. For each defect, 1 to
techniques variable 3 repair techniques were appointed. 5
Characteristics of Surface code SurCode – categorical Identification of the surface with the detected defect, combining the building number
the rendered variable and the number of the rendered surface. 6
surface Surface area SurAr m2 continuous Area of a rendered surface with homogeneous characteristics (e.g. colour, protection).
variable
Severity Sw % continuous Numerical index to quantify a surface overall degradation, considering the ratio
ofdegradation variable between the area affected by defects, their severity and impact in the overall condition
and the highest theoretical degradation level [21].
Age of the SurA years discrete Age of the surface at the time of inspection. If major repairs to the surface took place,
surface variable which should be the considered date. 7
Type of façade FType – categorical Type of façade, according to its position in relation to the main entrance of the building
variable (e.g. front façade), where the rendered surface is located.
Façade FOr – categorical Position of the façade, according to four cardinal directions and four intercardinal
orientation variable directions, where the rendered surface is located.
Colour ColB % continuous Percentage obtained from the average of RGB values of the surface colour, in which
brightness variable 100% corresponds to white (255, 255, 255) and 0% corresponds to black (0, 0, 0).
Characteristics of Age of the BuildA years discrete Age of the building at the time of inspection. 7
the building building variable
Exposure ExpoPollu – ordinal The building’s exposure to polluting agents [38].
topolluting variable
agents
Aerodynamic Roug – ordinal Categories of aerodynamic roughness of the building surroundings, according to
roughness variable Eurocode 1 (NP EN 1991–1-4: 2010 [39]).
Distance from SeaDist km continuous Shortest distance from the building to the sea or river shore.
the sea variable
Municipality Mun – categorical Location of the building according to the second-level administrative subdivision of
variable Portugal.
Summer SuTemp °C continuous The highest temperature in the summer in the surrounding area of the building,
temperature variable according to Eurocode 1 (NP EN 1991–1-5: 2009 [40]).8
(max.)
Wind action Wind – categorical The area where the building is located, according to the wind action zoning from
variable Eurocode 1 (NP EN 1991–1-4: 2010 [39]).
1
Occurrences are grouped in five categories: imminent danger, contingency measures needed (1 month); repairs needed immediately (6 months); repairs needed in the
short-term (12 months); repairs needed in the long-term (24 months); no urgent repairs needed, assess in the next inspection (30 months) [20].
2
Considered indispensable for the development of the defect [41]. Categories according to the classification of probable causes of defects for a global inspection system.
3
Categories according to the classification of probable causes of defects for a global inspection system.
4
Categories according to the classification of diagnosis methods for a global inspection system.
5
Categories according to the classification of repair techniques for a global inspection system.
6
B for building, a number, an underscore, R for render, and a number; e.g. B1_R1, B7_R52.
7
Although the concept of age is a continuous variable, in this case, age is converted into a discrete variable, as the surface’s age is counted in years.
8
Results are grouped in temperature categories (45 °C, 40 °C and 30 °C) and areas of the administrative subdivision of Portugal.
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 3

The purpose of this study is to explore a set of statistical analy- refers to defects specific of a single type of building element or sin-
ses (descriptive statistics, principal components and factor analy- gularities, combining a varied set of phenomena. Each defect corre-
ses, cluster analysis and multiple regression analysis) to appraise sponds to a code: capital letter A, a hyphen, a capital letter
the possibility of predicting the urgency of repair of defects in identifying the category (A, B, C and D) and sequential numbering
façade renders using a multiple regression analysis model, based within each category (A–A1, A–A2, . . ., A–D15).
on the data of an inspection’s campaign. Defects listed in the global classification should match a visually
In this paper, a brief review on building pathology is given, then identifiable phenomenon, excluding issues like insufficient ther-
the research methodology is presented, the selected variables are mal insulation or insufficient ventilation.
contextualised, the statistical procedures to model and classify From the global classification list of defects, only some apply to
information are presented and, finally, the multiple regression wall renders. Those were the ones considered during fieldwork, for
analysis model is discussed. identification purposes, namely (relevant for the selected variable
‘‘type of defect” in Table 1): A–A1, A–A2, A–A3, A–A4, A–A6, A–
B1, A–B2, A–B3, A–B4, A–C1, A–C2, A–C6, and A–C9.
1.1. Building pathology
‘‘A–A1 Leakage damp” occurs in localised areas due to inci-
dental causes associated with construction errors, human or
The global inspection system of non-structural elements of the
equipment failure, accidents and lack of maintenance. Frequently
building envelope under development includes a harmonised clas-
leakages are associated with plumbing, rainwater drainage sys-
sification list of defects (Table 2), based on which each defect is
tems and sewage failure. In façades specifically, it is frequently
identified during fieldwork. This classification list is based on a
due to clogging of rainwater hoppers, gutters and downpipes.
review of the literature [26–37], including information on building
Additionally, deficiencies in the tail-ends of the roofing system
technology to provide guidance on issues associated with con-
with parapet walls or in the coping of parapet walls are also fre-
struction and design. The classification list of defects should be dis-
quent causes of leakage in façade renders. If the origin of the
tinguished from a classification list of causes of defects, not
leakage is associated with rainwater, then the manifestation of
attributing the same value to what constitutes a cause and what
the defect varies according to seasons of the year. Moreover,
constitutes a defect, as an effect (defect) and its origin (cause)
some singularities on the façade may also be associated with
are different and may be observed and identified as different ele-
leakage damp, like windows, parapet walls and railings of bal-
ments within a degradation mechanism.
conies [30,42,43].
Considering the set of twelve individual building inspection
‘‘A–A2 Surface moisture” in wall renders may be associated
system, dedicated to a single type of building element or material,
with varied phenomena related to damp [44]: construction mois-
included in the global inspection system, their classification lists of
ture, rainwater, rising damp, condensation and hygroscopicity.
defects were the basis for the creation of the global classification of
The materials used to build the walls are the origin of construction
defects. Those lists were merged, adjusting the level of detail, elim-
moisture [45]. Besides the water used in mortars and concrete,
inating repetitions and aiming at conciseness. The global list was
there is also water absorbed by the materials before or in the
built to be as short as possible, considering a balanced level of
moment of application. Superficial water evaporates rapidly, but
detail, not eliminating any defect that may occur in any of the
a portion of water inside the materials’ pores evaporates slowly.
building elements taken into account. The list is organised in four
The migration of this portion of water out to the surface of the ren-
categories to improve its readability: defects of physical, chemical
der may lead to the appearance of moisture stains. Rainwater
and mechanical nature and other defects. The first three categories
[46,47], combined with wind, intersects the façade directly,
group the defects according to the type of aggressive actions that
enabling the penetration of water in cracks and open joints.
affect building elements, and the last category (other defects)

Table 2
Harmonised classification list of defects of the global inspection system.

Code Denomination Code Denomination


A-A Defects of physical nature
A-A1 Leakage damp* A-A4 Colour changes*
A-A2 Surface moisture* A-A5 Spalling/peeling/exfoliation and pop–outs
A-A3 Dirt and accumulation of debris* A-A6 Cohesion loss/disaggregation and chalking*
A-B Defects of chemical nature
A-B1 Biodeterioration/biological growth* A-B4 Blistering/bulging*
A-B2 Vegetation growth* A-B5 Corrosion on the current surface
A-B3 Efflorescence/cryptoflorescence and carbonation* A-B6 Corrosion in metallic fastening or tail–end elements
A-C Defects of mechanical nature
A-C1 Mapped cracking* A-C7 Warpage, swelling, deformation and other flatness deficiencies
A-C2 Oriented cracking on the current surface* A-C8 Material gap/puncture
A-C3 Fracture or splintering on the current surface A-C9 Detachment*
A-C4 Cracking and/or splintering adjacent to joints/edges A-C10 Loss of adhesion
A-C5 Wear or scaling of the finishing coat A-C11 Bending and rupture of metallic fastening elements
A-C6 Scratches/grooves and deep wear* – –
A-D Other defects
A-D1 Flaws in tail-end elements A-D9 Insufficient or excessive overlap of the claddings elements in roofs
A-D2 Misalignment of cladding elements A-D10 Clearances/gaps in door and window frames
A-D3 Finishing defects/discontinuities in architectural concrete surfaces A-D11 Absent or damaged hinges or locks in door and window frames
A-D4 Finishing colour flaws in painted façades A-D12 Ponding/insufficient or excessive slope in roofs
A-D5 Finishing texture flaws in painted façades A-D13 Inadequate operation of elements of the rainwater drainage system
A-D6 Degradation of the filling material of current joints A-D14 Deficient capping adjacent to flat roofs
A-D7 Absence/loss of filling material in connecting elements or current joints A-D15 Incorrect or deficient interventions in claddings of pitched roofs
A-D8 Inadequate operation of expansion joints in flat roofs – –
*
Defects from the global classification that apply to wall renders
4 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

Additionally, a rainwater surface is formed on the façade sur- windowsills), according to the volumetric configuration of the
face, which penetrates the wall progressively due to capillarity building (presence of protruding elements). A more porous and
phenomena and wind overpressure (water runoff). This type of rougher wall render hinders that washing process, thus presenting
moisture is seen in stains, usually disappearing after raining peri- a more uniform dirt layer. Differentiated dirt stains may be associ-
ods. However, if the exposure period is long, the stains may pro- ated with water runoff and thermophoresis phenomena, previ-
gress and other issues may start to arise (dirt deposition, ously mentioned.
biological growth or efflorescence). Rising damp [48,49] occurs in ‘‘A–A4 Colour changes” in wall renders [29,56] is associated
rendered façades in direct contact with the ground, and when with changes or discolouration of the colour of the finishing coat
there is water in the soil. Stains may develop due to capillarity phe- and with corrosion stains. If the wall render is painted, then the
nomena in the lower areas of the walls when no waterproofing defect is attributed to the paint finish. Still, in the case of one-
barrier was provided while building the wall. If rising damp is coat renders, where the pigment is incorporated in the mortar,
caused by groundwater, the effects do not show significant the defect is attributed to the wall render. This defect is highly
changes throughout the whole year and the height of the stains associated with the natural ageing process of the render, due
is similar in every wall. If rising damp is caused by superficial to the continuous exposure to aggressive agents over time. Solar
water, there are variations in the effects throughout the year, more radiation (ultraviolet radiation) and the weather (wind and rain)
intense in rainy seasons, and different walls show stains with dif- are factors that speed up the process of superficial degradation
ferent intensity or height, according to the distance to the superfi- of the cladding, becoming responsible for colour and appearance
cial water source. Rising damp may be particularly harmful if changes of the wall render, like fainted colours, saturation loss,
associated with water transporting dissolved salts, leading to the gloss changes and yellowing. Façades oriented south, particularly
occurrence of efflorescence. Surface moisture associated with con- those with dark or highly saturated colours, usually show a
densation [50,51] is originated by the water vapour generated higher incidence of this defect, due to a longer exposure to sun-
inside buildings or released by building materials. In façades, it is light. As for corrosion stains [57], they may occur due to the cor-
associated with thermophoresis phenomena [28,44] that lead to rosion of metallic elements embedded in the wall render or in
the appearance of staining with the shape of building materials its substrate (not visible) or to the corrosion of metallic elements
(bricks, for instance). It results from the combination of moisture fixed on the façade. Those corrosion phenomena originate oxides
and temperature differentials. The wall render is applied over a that may migrate to the surface of the cladding originating
heterogeneous substrate composed by mortar joints, bricks and orange stains (rust), or runoff through the façade, in the case
bearing elements (like columns and beams). Thus, during drying of exposed elements. Embedded elements where this is frequent
periods, the wall has differential temperatures along those sub- are the reinforcement of concrete bearing structures and metallic
strate materials, because, while the water evaporates, the superfi- meshes or profiles embedded in the wall render. The corrosion
cial temperature decreases, and, additionally, the substrate of embedded elements also causes increases in volume that
materials do not have the same water absorption coefficient, dry- may cause cracking, or, in more serious cases, the loss of cohe-
ing at different rates. This occurs frequently when there is no ther- sion or disaggregation of the render. The corrosion phenomena
mal insulation or it is insufficient, or when the render is too thin. are frequent in clothes racks, metal railings, bearing structures
Due to thermophoresis, dirt particles are deposited on the façade for equipment (e.g. air conditioning), equipment, flashings and
facing. The same phenomenon may be associated with water rainwater drainage elements.
migrating inside the walls, dragging impurities that remain on ‘‘A–A6 Cohesion loss/disaggregation and chalking” in wall ren-
the façade surface after the water evaporates. As for hygroscopicity ders [43–45,58,59] refers to the loss of connection elements within
phenomena [43,45,52], they are visible in moisture stains and the render’s mortar, followed by a significant loss of particles. This
cause the degradation of the render in areas with a high concentra- process turns the render into a fragile material, easy to disaggre-
tion of salts. Hygroscopicity is a property associated with some gate. A–A6 may result in powderiness, consisting of the disaggre-
salts that have the ability of dissolving when the relative humidity gation of several components of the mortar, which crumbles and
is higher than 70%. The salts crystallise again when the relative leads to the detachment of the materials as dust or small grains.
humidity decreases. Recrystallisation is accompanied by a consid- A–A6 may also result in granular disaggregation, which consists
erable increase in volume, which results in the degradation of the of the loss or washing of fine mortar particles leading to the easy
renders. This phenomenon occurs in renders showing a porous detachment of sand-size particles even with low-intensity
structure, with the capability of storing water. mechanical loads. This phenomenon is usually associated with rain
‘‘A–A3 Dirt and accumulation of debris” refers to stains caused and wind, exposing aggregates and the cladding’s substrate. Ero-
by vandalism/graffiti, dirt and deposits of particles. Non- sion is also included in A–A6. It corresponds to the localised loss
authorised graffiti [53,54] is associated with vandalism and occurs of mass of the surface of the material due to the action of weather
more frequently in urban areas in the lower areas of façades in agents. A–A6 tends to be more frequent in old renders or in renders
direct contact with public streets. Degraded and non-watched with low cement content. It may also be a consequence of detach-
urban areas are usually more affected, as well as those close to ment of the outer layers of the wall render, exposing underlying
schools or places where youngsters gather. The occurrence mech- layers, which are more sensitive to weather agents.
anism of graffiti is directly associated with human action, including ‘‘A–B1 Biodeterioration/biological growth” [51,60–63] corre-
several types of paintings and markings on surfaces, as well as var- sponds to the appearance of stains resulting from the proliferation
ious types of paint, which are absorbed due to the porosity of the of microorganisms on the wall render’s surface, such as mould and
render. As for dirt and the deposition of particles [43,55], it may fungi. A–B1 usually occurs in environments with a relative humid-
be uniform or differentiated. Uniform dirt on façades is a common ity higher than 70%, mild temperatures, adequate pH and lighting,
defect in urban and industrial areas, resulting in the deposition of oxygen and an organic medium. Hence, it is usual to detect biolog-
soil particles, dust, soot and other polluting particles. Rain and ical growth simultaneously with other defects also associated with
wind transport these elements up to the façades, where they accu- the prolonged presence of water and dirt. A–B1 affects the visual
mulate. The texture of the wall render influences the accumulation qualities of wall renders, but it also gradually destroys the aggre-
rate of dirt, as a smooth finish eases washing the facing with rain- gates of hardened mortars through organic acids from microorgan-
water. In those cases, dirt may become more easily detectable due isms. The stains are usually black, grey, brown, greenish, whitish or
to areas that are not reached by rainwater (e.g. below balconies or yellowish.
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 5

‘‘A–B2 Vegetation growth” [60,61,64] refers to the spreading of cussion test. The phenomenon evolves to a situation visible to the
algae, lichens, moss and plants over the surface of the wall render. human eye, as a convex shape is formed.
It is a defect similar to biological growth in terms of the need of ‘‘A–C1 Mapped cracking” [43,76] in wall renders is usually asso-
moisture for its occurrence as well as some other causes. Still, veg- ciated with shrinkage phenomena, which occur due to the drying
etation growth may have more severe consequences as the size of and hardening processes of the mortar. They are inevitable and
the vegetation may lead to the destruction of the render by the may occur up to 28 days after executing the render. The occurrence
roots of some plants. of cracking results from tension induced in the render due to the
‘‘A–B3 Efflorescence/cryptoflorescence and carbonation” resistance of the substrate to shrinkage. Mapped cracking is the
[27,30,43,60,64–75] includes three similar manifestations. development of a pattern of cracks, without a specific direction,
Efflorescence/cryptoflorescence (Fig. 1) is the exudation of water- which progressively open as tensile stress equals the strength of
soluble mineral salts. These effects occur in the presence of salts the material. The network of cracks is larger and the cracks thicker
and water inside the walls. The salts may be dissolved in the water and deeper, the more rigid or thicker the render is. In terms of
that is inside the walls or may have origin in building materials of thickness, mapped cracks are usually thinner than 0.2 mm, but in
the wall, being dissolved in migrating water. Once in solution, the grave situations, according to the type of render, the thickness
crystallisation of the salts may occur inside the pores of the ren- may reach up to 2 mm. In current renders, mapped cracking does
der’s mortar (cryptoflorescence) or on the surface of the wall ren- not usually go deeper than 1 mm of the cladding.
der (efflorescence). The appearance of efflorescence is that of ‘‘A–C2 Oriented cracking on the current surface” [43,77] in wall
whitish stains, with varying extent and shape. Locally, the surface renders is a defect that, although directly affecting them, is not
may be covered with crystalline flakes with a powdery appearance, usually caused within the render layer of a wall, as the most com-
and soluble in water, or with pellicles with a vitreous appearance mon cause is the differential movement of the wall components
hardly soluble in water. These stains may also appear as runoff, (e.g. bricks and bearing structure). Linear cracks occur isolated,
usually from cracks. As for cryptoflorescence, whose crystallisation sometimes with a zigzag appearance, with a significant thickness
process increases the volume of the material, it may result in cohe- (at times above 2 mm) and deep, generally going through the
sion loss and disaggregation and, ultimately, detachment of the whole cladding layer. The orientation and slope of these cracks
render. The use of impermeable paint in the finishing coat stops are generally well defined, providing information about the proba-
the water evaporation process and favours the occurrence of ble origin of the cracking phenomenon. If the substrate causes the
cryptoflorescence. Carbonation is the appearance of whitish cracks, they generally follow the direction of joints (e.g. mortar
incrustations of salts of calcium carbonate on the surface of the joints or joints between brick panes and the bearing structure).
wall render. They result from the dissolution of calcium hydroxide This is due to the different behaviour of different materials in
of cementitious products in water. As the solution migrates to the terms of thermo-hygrometry, shrinkage, creep and deformation
surface of the cladding, it is exposed to carbon dioxide and trans- under loads. If the substrate is homogeneous, it is important to
formed into calcium carbonate (insoluble). The main difference ensure that the render layer is compatible with the substrate in
between efflorescence and carbonation is the fact that salts of terms of Young’s modulus and coefficient of thermal expansion.
efflorescence are usually water-soluble. Calcium carbonate salts It is also frequent to detect oriented cracking next to openings on
may be easily identified through the reaction with a phenolph- the façade (e.g. windows), which usually begin at a corner of the
thalein solution. A–B3 occurrences progress over time and are dif- opening and follow a 45° angle, approximately.
ficult to detect in recent wall renders. After five to ten years, the ‘‘A–C6 Scratches/grooves and deep wear” in wall renders [43] is
render tends to present these effects more clearly. essentially translated solely into scratches and grooves. They result
‘‘A–B4 Blistering/bulging” [30,43] is the first part of an evolutive from human action and consist of changes in the cladding caused
process that may result in the detachment of the render if no by mechanical actions. Those actions may be impacts, blows and
action is taken. The loss of adhesion between the render and the abrasion, of accidental origin, or not. Considering the intensity
substrate consists of the loss of mechanical connections, as the ren- and the type of action, the result in the wall render may go from
der layer moves away from the substrate. This defect may not be slight scratches or grooves, without the loss of material, up to sig-
evident in a merely visual inspection, but it is detectable by a per- nificant deformations and losses of material. Most frequently,

Fig. 1. Cases of efflorescence/cryptoflorescence.


6 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

these occurrences result from accidents associated with the occu- For quantitative variables, descriptive statistics measures were
pation, circulation and normal use of people, like when transport- applied to better understand their features (Table 3). An area of
ing large objects. 1,713.00 m2 wall renders was inspected, of which 281.37 m2 were
‘‘A–C9 Detachment” [30,43,58,78–80] is a defect usually pre- affected by defects. Additionally, the variable average thickness of
ceded by ‘‘A–B4 Blistering/bulging”. A–C9 consists of the definitive cracks has less valid cases, since valid cases for this variable corre-
separation of the render from the substrate, resulting in the spond only to defect-types A–C1 mapped cracking, A–C2 oriented
detachment of regions of the cladding. The falling of the material cracking on the current surface and A–C6 scratches/grooves and deep
is caused by the loss of integrity of the rigid layer of the wall render wear, representing 48 cases of the sample. In other words, the vari-
resulting from deformations of the cladding (blistering and bul- able average thickness of cracks was not indicated in all cases, as it is
ging) associated with the effect of gravity (the weight of the clad- only applicable to some types of defects.
ding layer). The mode for repair urgency, the admissible period until defect
repair, is ‘‘24 months”, which was recorded in 78 instances (out of
2. Methodology 94 cases), meaning that the sample is mainly composed of non-
urgent cases (the sample only comprises observations in repair
Data were collected in-situ in 52 rendered surfaces in seven urgency with values ‘‘6 months”, ‘‘12 months” and ‘‘24 months”;
buildings in Continental Portugal, in Leiria, Lisbon and Alcobaça values ‘‘1 month” and ‘‘30 months” were not observed in the sam-
municipalities. The standardised gathering of data was based on ple) (Fig. 2). In the wall renders sample of Sá et al. [9], level 2 (the
the inspection form model of the mentioned global inspection sys- least urgent) was the most detected level of urgency of repair,
tem. To summarise, the inspection form includes data on: weather which is consistent with the sample’s results. The mode of summer
conditions at the moment of inspection; general and construction temperature (max.) also has a high frequency (79) for value 40 °C,
data; environmental exposure characteristics; maintenance data associated with the location of the buildings included in the
of the building; location and construction characteristics of the sample.
inspected surface; detected defects and respective repair urgency; The skewness of Sw is higher than that of DPctAr, proving that,
direct and indirect probable causes of each defect; advised diagno- even though the extent of defects is important to assess degrada-
sis methods; adequate repair techniques for the defect and causes; tion, there are other relevant criteria to quantify degradation. In
a photographic survey of the building, inspected surfaces and the histogram of defect percentage area (Fig. 3), modal class corre-
defects. Information was gathered on site by visual observation sponding to defects reaching 0–20% of the rendered surface area
of the rendered surfaces, with the aid of some basic instruments, is prominent. With a maximum value of 90% and a minimum of
such as camera, tape-measure and crack-width ruler. No in-situ 0.1% (Table 3), this variable has a high coefficient of variation, only
or laboratory tests were performed. surpassed by variable defect area. In the histogram of the severity of
The data were arranged in an MS Access database. From the col- degradation (Fig. 3), modal class 0–20% is also dominant, corre-
lection of data, the variables were selected, and some basic statis- sponding to a low degradation level of the sample, in accordance
tics were applied to the data, including correlation analysis. In view with the repair urgency, with only seven cases in which repair up
of a better reading of the information structure, a principal compo- to six months was required.
nent analysis was performed. Cases were further studied through In categorical variables, no detailed descriptive statistics may be
cluster analysis. Finally, with the purpose of predicting the urgency applied. Still, some frequency distributions may be extracted from
of repair of each defect, a multiple linear regression analysis was data, such as the distribution of frequency of types of defects
performed. Throughout this sequence of steps, software Statistica detected in the sample (Fig. 4). A–C2 oriented cracking on the cur-
(Version 8.0) [81] was used, complemented with the use of MS rent surface is the most frequently detected defect, representing
Excel. 33% of the sample. Cause C-E1 inexistent or inadequate maintenance
was the indirect probable cause most associated with detected
defects, identified in 40% of cases. C-C10 stress concentration within
3. Preliminary analysis
the substrate was the direct cause most frequently linked to
detected defects, also in 40% of the sample.
3.1. Selected variables, descriptive statistics and correlation of
Diagnosis method D-A5 crack width ruler and crack-measuring
variables
microscope is prominent, as it was advised in 48% of the observa-
tions, following the high frequency of cracking defects. Simultane-
Considering detected defects as the observations in the sample,
ously, repair techniques R-A12 application of a new (adequate)
the variables without missing data and with different values from
cladding/finishing coat over the existing one/replacement, R-A11
case to case were selected (Table 1). With a sample size of 94 cases,
replacement or reapplication of the cladding/glazing (partially or com-
considering a single group t-test with a 5% two-sided significance
pletely) and R-A14 treatment of cracks or other holes in the cladding
level, the sample has 80% power to detect an effect size of 0.292.

Table 3
Summary of basic descriptive statistics of quantitative variables.

Variable Valid N Mean Median Sum Min. Max. Lower Quartile Upper Quartile Std.Dev. Coef.Var. Skewness
DAr 94 2.99 0.52 281.37 0.0025 103.61 0.14 1.28 11.42219 381.5938 7.74626
DPctAr 94 11.6% 3.5% 1086.4% 0.1% 90.0% 1.0% 11.0% 0.18629 161.1935 2.58745
ThickCr 48 0.84 0.65 40.40 0.30 2.50 0.42 1.10 0.54756 65.0567 1.58686
ReUrg 94 21.51 24 2022 6 24 24 24 5.66207 26.3222 1.97673
Sw 94 11.0% 5.8% 1034.2% 0.1% 90.0% 2.6% 13.1% 0.14078 127.9500 2.76169
SurA 94 18.22 13 1713 1 34 8 31 11.17135 61.3022 0.30912
BuildA 94 28.87 14 2714 8 66 8 58 24.21509 83.8695 0.63113
SurAr 94 26.08 11.40 2451.36 0.20 133.63 5.40 30.00 34.39589 131.8948 2.15660
ColB 94 72.9% 89.8% 6852.3% 21.6% 98.0% 27.1% 98.0% 0.33005 45.2766 0.77727
SeaDist 94 7.97 11.90 749.00 0.30 12.20 4.00 11.90 4.67370 58.6553 0.45978
SuTemp 94 39.20 40 3685 35 40 40 40 1.84087 4.6958 1.88946
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 7

Fig. 2. Examples of cases with different levels of repair urgency: (a) detachment of the render below the eaves at a side façade (observed level of repair urgency: 6 months); (b)
mapped cracking (highlighted in orange) on the front façade (observed level of repair urgency: 12 months); and (c) oriented cracking on the rear façade (observed level of
repair urgency: 24 months).

Histogram: DPctAr Histogram: Sw


90 90 79
78
80 80
70 70
60 60
50 50
40 40
30 30
20 9 20 12
10 3 2 2 10 2 0 1
0 0
0-20% 21 -40% 41-60% 61-80% 81-100% 0-20% 21 -40% 41-60% 61-80% 81-100%

Fig. 3. Histograms of defect percentage area and severity of degradation.

A-A1 leakage damp; A-A2 surface moisture; A-A6 cohesion


loss/disaggregation and chalking; A-B1 biodeterioration/biological
growth; A-B3 efflorescence/cryptoflorescence and carbonation; A-B4
blistering/bulging; A-C1 mapped cracking; A-C2 oriented cracking on
the current surface, A-C6 scratches/grooves and deep wear; A-C9
detachment.

Fig. 4. Frequency of observations by type of defect and a case of a render without finishing coat and with high severity of degradation (greater than20%), showing mapped
cracking, efflorescence and surface damp.

were those most frequently prescribed, in 51%, 46% and 45% of ated with surface age if the sample includes surfaces that have not
detected defects, respectively. been submitted to maintenance actions. Defect percentage area is
40% of defects were detected in front façades of buildings, and part of the calculations to determine the severity of degradation,
17% of defects were found in façades oriented southeast. Most and distance from the sea may influence the maximum summer
defects (62%) were detected in buildings at zone A, i.e. in continen- temperatures recorded in the regions where the buildings in the
tal areas more than 5 km away from the sea or rivers, and at an sample were located.
altitude below 600 m. Spearman’s correlation coefficient allows widening the corre-
There is a high linear correlation between (Table 4): defect area lation analysis to ordinal and categorical variables. A Spearman
and defect percentage area; severity of degradation and defect per- correlation coefficient of –1.00 is determined between variables
centage area; building age and surface age; and summer temperature ExpoPollu and Mun. Even though these variables should not rep-
(max.) and distance from the sea. Building age is necessarily associ- resent any correlation at all, in the sample, all the buildings in
8 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

Table 4
Linear correlation matrix between variable pairs; underlined correlations are significant at p < 0.05 (N = 94).

Variable DAr DPctAr ReUrg Sw SurA BuildA SurAr ColB SeaDist SuTemp
DAr 1.00 – – – – – – – – –
DPctAr 0.52 1.00 – – – – – – – –
ReUrg 0.18 0.26 1.00 – – – – – – –
Sw 0.17 0.66 0.27 1.00 – – – – – –
SurA 0.05 0.06 0.15 0.03 1.00 – – – – –
BuildA 0.00 0.05 0.08 0.08 0.80 1.00 – – – –
SurAr 0.43 0.00 0.07 0.07 0.21 0.22 1.00 – – –
ColB 0.15 0.06 0.01 0.21 0.30 0.15 0.27 1.00 – –
SeaDist 0.24 0.10 0.04 0.01 0.30 0.46 0.47 0.37 1.00
SuTemp 0.29 0.07 0.07 0.05 0.13 0.25 0.38 0.33 0.70 1.00

the same municipality showed the same type of exposure to pol- malised factor loadings across variables for each factor. It corre-
luting agents. Hence, these data are redundant and, for this rea- sponds to maximising the variances in columns (Table 5).
son, variable Mun is not considered in further analyses. Factor loadings of three variables (DAr, ReUrg and ColB) repre-
sent low statistical significance, as the proportion of explained
variance by factors is low [83], even though a tendency of associa-
3.2. Principal components analysis tion with one factor may be detected. DAr tends to be more associ-
ated with Factor 3, ReUrg to Factor 3 and ColB to Factor 4.
The use of numerical models allows summarising the studied The first factor, explaining 29% of the variance, represents five
data. The principal components analysis summarises the maxi- variables. So, incrementing Factor 1 corresponds to a smaller ren-
mum level of the original information (variance) in a minimal dered surface, a greater distance from the sea, higher exposure to
number of factors. So, this analysis is used to better understand polluting agents, higher temperatures in the summertime and
the collected data and highlight relationships between variables. higher aerodynamic roughness. Factor 1, is thus associated with
To perform a principal components analysis, quantitative data the surrounding conditions of the building and to characteristics
were used, including ThickCr. Since this variable did not include of the building element. The second factor refers to parameters
records for the whole sample, the average was used to replace that may influence the progression of defects, as it is positively cor-
missing data, to include all the cases in the sample in the analysis. related to SurA, BuildA and Roug. Ageing of materials and the
At a preliminary stage, values were standardised to eliminate the aggressiveness of the environment influence the progression and
variables’ dispersion effect. Consequently, variables became severity of defects [84]. The third factor refers to degradation, as
dimensionless and identical importance is attributed to each to it progresses negatively with increments in variables defect per-
build factors—standardised variables have a mean of 0 and a stan- centage area and severity of degradation. Finally, Factor 4 refers to
dard deviation of 1. the thickness of detected cracks, and the thicker a crack, the lower
To assess the number of retained factors, based on the selected the Factor 4 value.
variables and standardised data, eigenvalues, cumulative percent- Entities may be projected in planes formed by pairs of factors
age of explained variance and the eigenvalues diagram were con- (Fig. 5), based on the factor scores of each entity according to each
sidered. Four factors were retained in this analysis (Table 5), factor. In Fig. 5, Factors 1, 2 and 4 form planes paired with Factor 3,
using varimax normalised rotation of factors [82]. In factor analy- which is highlighted due to referring to degradation. As variable re-
sis, the varimax is a typical rotational strategy. Rotational strate- pair urgency is not expressed enough in a single factor, and because
gies intend to obtain a clear pattern of factor loadings, it is the object of the present study, the entities may be grouped
corresponding to factors with clearly high loadings for some vari- according to this variable’s values. It is observed that entities that
ables and low loadings for others. The goal of a varimax normalised need to be repaired in six months are all in the negative half of Fac-
rotation of factors is to maximise the variances of the squared nor- tor 3. In other words, the most urgent detected defects affect larger
areas of the rendered surface and have higher severity of degrada-
tion indexes. The entities are scattered without a clear trend in the
Table 5
Factor loadings of principal components; underlined loadings are greater than 0.60. direction of the remaining factors.

Variable Factor 1 Factor 2 Factor 3 Factor 4


3.3. Cluster analysis
DAr 0.4109 0.0041 0.5768 0.1610
DPctAr 0.0850 0.0191 0.8858 0.0001
ThickCr 0.1493 0.1371 0.0932 0.8372
Cluster analysis classifies a great number of cases in operative
ReUrg 0.1090 0.1073 0.5365 0.1965 groups, allowing to determine case-types with a minimal loss of
Sw 0.0747 0.1032 0.8071 0.0403 information, according to similarities and differences between
SurA 0.1914 0.7966 0.0653 0.3846 entities.
BuildA 0.2074 0.9569 0.0025 0.0655 In cluster analysis, as in principal components analysis, data
SurAr 0.6492 0.2026 0.0174 0.1475 were standardised, and any missing data were replaced by the cor-
ColB 0.4808 0.0706 0.0997 0.5164 responding variable’s mean (simple imputation method), which is
SeaDist 0.9129 0.2231 0.0130 0.0094 the most common method applied to deal with missing values
ExpoPollu 0.9468 0.0662 0.0156 0.0249
[85]. Hypotheses using hierarchical and non-hierarchical methods
SuTemp 0.8505 0.4744 0.0394 0.1143
were tested, and the best results are presented.
Roug 0.6546 0.7069 0.0465 0.1468
Using the joining (tree-clustering) and the complete linkage
Expl.Var 3.8300 2.4165 2.0839 1.2430
method [83], all variables were considered, except those which
Prp.Totl 29.5% 18.6% 16.0% 9.6%
allowed multiple responses (DiC, InC, DiagM and ReT), and Euclid-
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 9

Factor 1 vs. Factor 3


0.1

-
-0.4 -0.3 -0.2 -0.1 0.1 0.2
-0.1

-0.2

-0.3

-0.4

-0.5

6 12 24

Factor 2 vs. Factor 3


0.1

-
-3.5 -2.5 -1.5 -0.5 0.5 1.5
-0.1

-0.2

-0.3

-0.4

-0.5

6 12 24

Factor 4 vs. Factor 3


0.1

-
-3.5 -2.5 -1.5 -0.5 0.5 1.5
-0.1

-0.2

-0.3

-0.4

-0.5

6 12 24

Fig. 5. Bidimensional projection of cases in factor planes (Factor 3 in y-axes, Factors 1, 2 and 4 in x-axes).

ian distances were used. It was found that: (i) six was the adequate tion must result in the minimal Euclidean distance between enti-
amount of clusters, and two clusters were one-unit groups refer- ties inside the cluster and the maximum Euclidian distance
ring to outliers; (ii) the remaining clusters were built approxi- between clusters. Variables DAr, DPctAr, ThickCr, ReUrg, Sw, SurA,
mately based on the buildings’ location, surrounding area and BuildA, SurAr, ColB, SeaDist, ExpoPollu, SuTemp and Roug were used.
the building elements’ characteristics; (iii) one cluster only The best solution includes three clusters, defined by the buildings’
includes entities detected up to 600 m from the sea; (iv) another characteristics.
cluster only includes entities at 4.0 km from the sea; (v) and In the first cluster, entities up to 5.8 km from the sea/river
another cluster only includes entities at 12 km from the sea; (vi) are grouped. In the second cluster, only entities at about
one clusters mixes-up entities based on the distance from the sea 12 km from the sea are included. In the third cluster, only enti-
(12 and 4 km away) but refers only to entities with Sw  6.2%; ties with a high percentage of affected area are included (Fig. 6).
(vii) in the tree diagram, entities 4 and 12 km away from the sea Observing the plot of means of the clusters, clusters are more set
end up being grouped in further steps of the diagram. Different apart in what refers to the distance from the sea, exposure to pol-
steps of entities grouping using the complete linkage method have luting agents and summer temperature (max.). The third cluster
a clear path. simultaneously has the largest, the most severe and the most
The k-means non-hierarchical method was also used. According urgent defects. The second cluster refers to entities in younger
to this method, a predefined number of clusters must be inputted buildings and surfaces, but no lower degradation is observed,
in the software [83]. The method works iteratively, and the solu- compared with the first cluster.
10 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

3.0
2.5
2.0
1.5
1.0
Cluster 1
0.5
Cluster 2
0.0
Cluster 3
-0.5
-1.0
-1.5

Fig. 6. Plot of means of each cluster (k-means method).

Table 6
Conversion of variable D into dummy variables.

D A-A1 A-A2 A-A6 A-B1 A-B3 A-B4 A-C1 A-C2 A-C6


A-A1 1 0 0 0 0 0 0 0 0
A-A2 0 1 0 0 0 0 0 0 0
A-A6 0 0 1 0 0 0 0 0 0
A-B1 0 0 0 1 0 0 0 0 0
A-B3 0 0 0 0 1 0 0 0 0
A-B4 0 0 0 0 0 1 0 0 0
A-C1 0 0 0 0 0 0 1 0 0
A-C2 0 0 0 0 0 0 0 1 0
A-C6 0 0 0 0 0 0 0 0 1
A-C9 0 0 0 0 0 0 0 0 0

4. Multiple linear regression model The studied multiple linear regression models included several
iterations of standard and stepwise models, considering a signifi-
Analysing data through a multiple linear regression model cance level of 5%. Models with six quantitative candidate variables
enables the prediction of a dependent variable based on the known and four qualitative candidate variables, converted into dummy
values of independent variables. In this case, the dependent vari- variables, were tested.
able is ReUrg. Data for ReUrg were collected from the observation
of the characteristics of each defect, based on parameters that deli-
5. Results and discussion
mit each level of repair urgency for each type of defect. In every
type of defect, some parameters are considered, namely: the rela-
The best model was obtained with quantitative variable DPctAr
tive extent of the defect; location of the defect; and the phe-
and the qualitative variables D and FType, which were converted
nomenon progression conditions. Then, in every defect, there are
into dummy variables (Tables 6 and 7). The model’s selection of
varying parameters, such as the presence of moisture (applicable
qualitative variables is consistent with: (i) the varying parameters
only to biological growth, vegetation growth, oriented cracking
to determine ReUrg in fieldwork according to the type of defect; and
and scratches/grooves and deep wear) or the exposure of the sub-
(ii) the association of type of façade with the concept of aesthetical
strate (applicable only to the render’s detachment). ReUrg is a vari-
value according to the façade’s importance within the building—a
able that depends on other variables, which are the basis of the
parameter used in every type of defect to define the repair urgency
determination of the maximum admissible period until the defect
level.
is repaired, in categories of months. For instance, a defect A–C2 ori-
Tables 8 and 9 and Fig. 7 show the results of the multiple linear
ented cracking on the current surface is assessed to need immediate
regression model and the regression equation is as follows:
intervention (up to six months) if the affected area is of high aes-
thetical value, if there are conditions for the phenomenon to pro- ReUrg ¼ 11:39  A  C1 þ 12:71  A  A2 þ 9:92  A  C2
gress, if cracking reaches both the cladding and the substrate and
if moisture is present. þ 12:36  A  A1 þ 11:20  A  B3  13:38  DPctAr
With the multiple linear regression analysis, a prediction of þ 11:06  A  B1 þ 14:74  A  C6 þ 14:01  FType2
ReUrg close to reality is sought. That means avoiding negative
þ 14:59  FType3 þ 11:82  FType1  4:78  A  B4 ð1Þ
and obtaining low positive prediction residual values. For instance,
if the prediction points out 22 months until repair when a repair up The model defines the number of months until repair according
to six months was advised, the prediction residual value is to the entity’s type of defect and type of façade, depending on the
16 months. This hypothetical scenario may question the building defect’s extent. For each percentage point of relative area affected
element’s safety, as the situation would be graver than predicted, by a defect, its repair is anticipated 0.13 months (about four days),
requiring short-term action. considering the remaining variables constant. This way, according
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 11

Table 7  Only the occurrence of A–B4 (blistering/bulging) and DPctAr


Conversion of variable FType into dummy variables. anticipate repairs; nevertheless, A–B4 has less influence on the
FType FType1 FType2 FType3 determination of ReUrg, as, in practical terms, the anticipation
Front 1 0 0 of repair actions is more dependent on the defect’s relative
Side 0 1 0 extent;
Rear 0 0 1  As for the type of façade, the cases occurring in front façades
Wall 0 0 0 (FType1) represent a lower delay of repairs, namely about
12 months, if remaining variables are constant; this is consis-
tent with the fact that a defect in a front façade empirically
Table 8 implies higher repair urgency, when compared with other
Summary statistics of the multiple linear
façades, due to the aesthetic perception of the façade by users
regression model for dependent variable
ReUrg (forward stepwise method). and owners, who tend to promote maintenance actions in order
to restore an adequate level of performance concerning the
Statistic Value
façades’ appearance; this notion comprises the concept of pri-
Multiple R 0.9847 oritisation of repair (or maintenance) works based on economic,
Multiple R2 0.9697
aesthetic, psychological and cultural aspects [25];
Adjusted R2 0.9652
F(12,82) 218.5820  By the previous conclusion, according to the model, if the
p 0.0000 defect occurs in a rear façade (FType3), repair actions may
Std.Err. of Estimate 4.1451 be delayed for about three months, when compared with a
front façade;
 If A–A2 (surface moisture) occurs, the largest delay of repairs is
to the model, a defect affecting 50% of the rendered surface should obtained (considering D dummy variables), namely about
be repaired six and a half months before a sound surface. From the 13 months, if the remaining variables are constant; this means
best model, the following may be concluded: that, in the sample, surface damp defects tend to have lower
repair urgency;
 Dummy variable A-C1 (mapped cracking) has an important cor-  If the detected anomaly is A–C9 (detachment), the urgency of
relation with ReUrg, as it was the first selected variable to be repair of that case will depend exclusively on its extent and
included in the model; i.e. in the sample, a mapped cracking on the type of façade in which it is detected, according to the
defect has a distinguishing repair urgency; conversion of variable type of defect into dummy variables;

Table 9
Regression summary of the multiple linear regression model for dependent variable ReUrg (forward stepwise method).

N = 94 Beta Std.Err. of Beta B Std.Err. of B t(82) p-level


A-C1 0.198 0.0311 11.387 1.7952 6.343 0.00000
A-A2 0.276 0.0380 12.708 1.7449 7.283 0.00000
A-C2 0.256 0.0388 9.918 1.5042 6.594 0.00000
A-A1 0.140 0.0248 12.364 2.1816 5.668 0.00000
A-B3 0.127 0.0246 11.202 2.1686 5.165 0.00000
DPctAr 0.131 0.0257 13.385 2.6128 5.123 0.00000
A-B1 0.073 0.0214 11.060 3.2563 3.397 0.00105
A-C6 0.118 0.0206 14.739 2.5607 5.756 0.00000
FType2 0.385 0.0390 14.012 1.4202 9.866 0.00000
FType3 0.295 0.0317 14.589 1.5697 9.294 0.00000
FType1 0.338 0.0430 11.817 1.5051 7.852 0.00000
A-B4 0.031 0.0214 4.777 3.2568 1.467 0.14624

35
10
30
5
25
residuals

20 0
months

15
-5
10
-10
5

0 -15
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94

Obs ReUrg Pred. ReUrg Residuals Mean residuals

Fig. 7. Column plot of observed, predicted and residual values of variable ReUrg (multiple linear regression analysis).
12 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

Normal probability plot of residuals


3

Expected normal value 1

-1

-2

-3
-16 -11 -6 -1 4 9
Residuals

Distribution of raw residuals


Expected normal
35

30

25

20

15

10

0
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14

Fig. 8. Analysis of residuals: normal probability plot of residuals and distribution of raw residuals.

 Defect type A–A6 (cohesion loss/disaggregation and chalking) same affected surface extent and type of façade. Case number 36
was not chosen by the model as an independent variable; this (Fig. 9) refers to surface damp on a façade, including the effects
can be explained by the sample’s specificity, as it is the only of thermophoresis and water runoff. In fieldwork, the defect was
defect type that was detected only once. considered to need repair in twelve months (maximum), as the
façade is of high aesthetical value (confining with the street), this
Additionally, it should be noted that the interval of residual val- case of surface damp has conditions to progress and the affected
ues ranges from 15 to 10 (Fig. 7), which is a relatively large inter- surface is above 30%. Hence, the parameters to check for the value
val. However, the lowest residual value, which is less favourable, of urgency of repair ‘‘12 months” in wall renders when surface
corresponds to an isolated case, as the next low residual value is damp is detected, in the inspection system used in the fieldwork,
of 8. The highest residual value (10) is observed in three cases, are only partially associated with physical degradation, and still,
and a fourth case has a residual value of 9. Still, as mentioned, pos- the mechanisms of deterioration [27,44] are not thoroughly anal-
itive residual values are not so worrying, as they correspond to tak- ysed at that stage. Applying the regression equation, the model
ing action before damage affects the safety of users. indicates that the defect needs to be repaired in no more than
Moreover, data on residuals reveal that the multiple linear 13.6 months.
regression model obtained can accurately explain the variability Case number 61 (Fig. 4) also refers to surface damp. In field-
of the dependent variable. Analysing charts in Fig. 8, the normal work, the defect was considered to need repair within 24 months.
probability plot of residuals and the distribution of raw residuals Applying the regression equation, the model predicts that the
show that residuals are approximately normally distributed. Fur- defect needs to be repaired in no more than 23.7 months. However,
thermore, the mean and the median of residuals are roughly zero based only in visual observation and using the surveyor’s empirical
(0.07 and 0.26, respectively). knowledge to apply the predefined parameters, it may be consid-
The model can be assessed using examples in the sample. As ered that the defect needed much more urgent intervention.
mentioned, repair urgency may not always match an empirical This model should only be transferred to other situations (sam-
perception of severity in terms of physical degradation, as other ples) with great caution, as the sample does not include every type
factors are involved when repair urgency is addressed. For of defect that may occur in rendered façades, namely: A–A3 dirt and
instance, Gaspar and de Brito [80] attribute identical weighting accumulation of debris, A–A4 colour changes and A–B2 vegetation
coefficients to mapped and oriented cracking to establish the growth. As the vast majority of inspected rendered surfaces were
severity of degradation of a rendered surface in terms of physical painted, as opposed to being one-coat pigmented renders, defects
deterioration [76]. In turn, the application of the current multiple A–A3 and A–A4 were linked to the paint coating. So, the model does
regression model leads to results that delay the repair of mapped not represent the general population, although it is a starting point
cracking when compared with oriented cracking, considering the of the research and may represent samples with similar character-
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 13

Fig. 9. Case 36 in the sample, referring to surface damp.

istics. Still, the methodology can be transferred to other samples,  The principal components analysis summarises the characteris-
obtaining a different multiple regression model. If using samples tics of the sample in four factors, allowing to understand the
with similar characteristics, researchers need to inspect façade sample at a glance; Factor 1 represents the surrounding condi-
renders considering the classification of defects presented in tions of the building and the characteristics of the building ele-
§1.1. Then, the characterisation of defects needs to include the ment; Factor 2 represents age and the surrounding aggressive
affected area (percentage of affected area according to the surface conditions; Factor 3 represents the degradation characteristics;
area of the cladding) and the type of façade in which the defect was and Factor 4 the thickness of cracking defects;
detected. Hereafter, researchers may apply the presented multiple  The representation of the cases in the sample in planes com-
linear regression model to predict the urgency of repair of defects posed by pairs of factors allows detecting some outliers in the
in their sample. sample, but also the influence of the parameters summarised
To transfer the methodology to other samples, a similar in the factors in the urgency of repair of the cases;
approach to the sample should be taken, i.e. a classification list  Still, the principal components analysis does not always have a
of defects should be adopted, and parameters to determine levels direct or easy interpretation and requires some attentive read-
of urgency of repair should be established, accordingly. Based on ing of the results in order to draw conclusions;
the sample’s results, a multiple linear regression analysis may be  The cluster analysis allows forming groups within the sample to
performed to build a prediction model based on the sample’s char- understand its particulars better. For instance, in this sample,
acteristics and the established urgency of repair parameters. the third cluster clearly showed that it included those cases
According to the regression model statistics (Table 8), the coef- with the highest average Sw and area affected by the defect as
ficient of determination is close to the adjusted coefficient of deter- well as the cases with the lowest value of ReUrg (less time until
mination, and they are both high. Thus, the explanatory power of the repair is required). This way, the cluster analysis validates
the regression equation is not conditioned by the high number of the soundness of the collected data;
independent variables in the model.  However, depending on the method used to perform the cluster
The results that are obtained through multiple linear regression analysis (in this research the joining [tree-clustering] with the
analysis, applying the preceding methodology, may be used to complete linkage method and the k-means non-hierarchical
sense the general gravity of defects in the sample, analysing differ- method were used, with Euclidian distances), it may be sensi-
ences between their B parameters. If the model is tested in very tive to outliers distorting the results; thus, the application of
large samples, the automation of repair urgency calculation may cluster analysis should go through an iterative process where
also be considered, including machine learning possibilities. the researcher tries to find balance between the applicability
As for taking the leap to the determination of a surface’s repair of the method and the logic expressed in the results;
urgency, as opposed to defect-case determination, several  Both the principal components and the cluster analyses pre-
approaches can be considered: (i) using statistics to build a com- sented in this research could benefit from a sample of cases
pletely different model, with inspected surfaces as the sample’s detected in a larger amount of buildings located in places with
entities; (ii) adapting the defect-case approach model to determine more diverse characteristics;
a surface’s urgency of repair; and (iii) making a qualitative deci-  The multiple linear regression analysis provides a valid model
sion, considering that the surface’s repair urgency is that of the to predict the urgency of repair of defects in this or in similar
most urgent defect detected in it. The latter may be justified in samples;
practical terms, as it would be more pragmatic to the owner and  However, it would be important to enlarge the sample to
comfortable to the users of a building that when, for instance, a increase the validity of the model, provided that more types
severe crack is repaired, the damp stains on the same surface of renders are inspected (one-coat renders) and more diverse
would be repaired as well, even if far apart. However, this approach types of defects are detected. That way, the results could be
may not always be financially possible. transferred to the population of wall renders;
Finally, some pros and cons of the results of this research should  Lastly, the presented methodology can be used in other samples
be addressed: and may be advantageous to compare different results in differ-
ent samples. Still, some other issues may be raised: how
 The preliminary analysis on variables, descriptive statistics and detailed does the prediction of the urgency of repair need to
correlation of variables allows having a sense of the type of be for decision-makers and maintenance planners? Should the
sample that one is dealing with (main detected defects, affected prediction go into the detail of days, or one month can be con-
area, and main correlations, for instance); sidered as the maximum acceptable unit?
14 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

6. Conclusions e a Tecnologia (FCT) project PTDC/ECI-CON/29286/2017 ‘‘Buildings’


Envelope SLP-based Maintenance: reducing the risks and costs for
In the context of inspection of rendered façades and trying to owners” and FCT PhD Scholarship SFRH/BD/131113/2017.
better understand the repair urgency of the detected defects, this
study involved, in a first stage, the application of descriptive statis-
tics techniques, including the correlation of variables, trying to References
characterise in detail the sample under analysis. Principal compo-
nents and factor analyses were also applied, in view of condensing [1] N. Garcez, N. Lopes, J. de Brito, G. Sá, Pathology, diagnosis and repair of pitched
information in summary-variables, in which the most relevant roofs with ceramic tiles: Statistical characterisation and lessons learned from
inspections, Constr. Build. Mater. 36 (2012) 807–819, https://doi.org/10.1016/
characteristics of information were highlighted. Cluster analysis j.conbuildmat.2012.06.049.
techniques were also applied to the data, which led to understand- [2] J. Conceição, B. Poça, J. de Brito, I. Flores-Colen, A. Castelo, Data analysis of
ing the weight of the variables that characterise the buildings with inspection, diagnosis, and rehabilitation of flat roofs, J. Perform. Constr. Facil.
33 (2019) 04018100, https://doi.org/10.1061/(ASCE)CF.1943-5509.0001252.
detected defects (entities). From those variables, the distance from [3] J.D. Silvestre, J. de Brito, Ceramic tiling in building façades: Inspection and
the sea is highlighted. pathological characterization using an expert system, Constr. Build. Mater. 25
As for the prediction of the maximum admissible period until (2011) 1560–1571, https://doi.org/10.1016/j.conbuildmat.2010.09.039.
[4] N. Neto, J. de Brito, Validation of an inspection and diagnosis system for
the defect is repaired, i.e. repair urgency, during fieldwork, it is anomalies in natural stone cladding (NSC), Constr. Build. Mater. 30 (2012)
done according to data collected considering a conservative and 224–236, https://doi.org/10.1016/j.conbuildmat.2011.12.032.
approximate estimate. This piece of data was determined accord- [5] A. Delgado, C. Pereira, J. de Brito, J.D. Silvestre, Defect characterization,
diagnosis and repair of wood flooring based on a field survey, Mater.
ing to well-defined parameters for each type of defect. Still, it is
Construcción. 68 (2018) 1–13, https://doi.org/10.3989/mc.2018.01817.
a quantitative fact distributed in five categories (1, 6, 12, 24 and [6] A. Santos, M. Vicente, J. de Brito, I. Flores-Colen, A. Castelo, Inspection,
30 months), of which only three were found in the sample (6, 12 diagnosis, and rehabilitation system of door and window frames, J. Perform.
Constr. Facil. 31 (2017) 04016118, https://doi.org/10.1061/(ASCE)CF.1943-
and 24 months). The repair urgency predicted by the multiple linear
5509.0000992.
regression model is more differentiated than the observed repair [7] J. Garcia, J. de Brito, Inspection and diagnosis of epoxy resin industrial floor
urgency. In other words, predictions, unlike observations, are not coatings, J. Mater. Civ. Eng. 20 (2008) 128–136. doi:10.10617(ASCE)0899-1561
grouped in categories. Even though results show a large interval (2008)20:2(128).
[8] C. Carvalho, J. de Brito, I. Flores-Colen, C. Pereira, Inspection, diagnosis, and
of residual values, the model is considered to present good data rehabilitation system for vinyl and linoleum floorings in health infrastructures,
prediction, even if a larger sample could be beneficial. J. Perform. Constr. Facil. 32 (2018) 04018078, https://doi.org/10.1061/(ASCE)
Being able to determine a reliable methodology to specify the CF.1943-5509.0001229.
[9] G. Sá, J. Sá, J. de Brito, B. Amaro, Statistical survey on inspection, diagnosis and
urgency of repair while surveying buildings allows developing repair of wall renderings, J. Civ. Eng. Manag. 21 (2015) 623–636, https://doi.
information to update maintenance plans. Furthermore, setting a org/10.3846/13923730.2014.890666.
concise number of parameters to determine the urgency of repair [10] B. Amaro, D. Saraiva, J. de Brito, I. Flores-Colen, Statistical survey of the
pathology, diagnosis and rehabilitation of ETICS in walls, J. Civ. Eng. Manag. 20
contributes to improving the surveyor’s procedures in fieldwork, (2014) 511–526, https://doi.org/10.3846/13923730.2013.801923.
simultaneously increasing the simplicity and robustness of data. [11] R. Pires, J. de Brito, B. Amaro, Statistical survey of the inspection, diagnosis and
This research constitutes a proposal that may make way to the repair of painted rendered façades, Struct. Infrastruct. Eng. 11 (2015) 605–618,
https://doi.org/10.1080/15732479.2014.890233.
future use of a semi-automatic methodology that may improve
[12] C. da Silva, F. Coelho, J. de Brito, J. Silvestre, C. Pereira, Statistical survey on
maintenance procedures and lead to considerable economic inspection, diagnosis and repair of architectural concrete surfaces, J. Perform.
sustainability. Constr. Facil. 31 (2017) 04017097, https://doi.org/10.1061/(ASCE)CF.1943-
5509.0001092.
Future studies may enlarge the rendered surfaces sample, so that
[13] C. Pereira, A. Silva, J. de Brito, J.D. Silvestre, Previsão e factores de influência da
it can represent the general population and enables the transference urgência de reparação de anomalias em rebocos pintados na envolvente de
of the model to other situations (samples), and apply the same edifícios (Prediction and influencing factors of the urgency of repair of defects
methodology to other building materials or elements, such as in painted renders in the building envelope), in: Construção 2018, Porto,
Portugal, 2018: pp. 830–840.
painted façades, architectural concrete surfaces or adhesive ceramic [14] J. Douglas, E.A. Noy, Building surveys and reports, 4th ed., Wiley-Blackwell,
tiling. In future developments, the scope of the repair urgency pre- Chichester, United Kingdom, 2011. doi:10.1002/9780470759462.
diction could be broadened and be applied to the whole surface or [15] P. Glover, Building surveys, 7th ed., Butterworth-Heinemann, Oxford, United
Kingdom, 2009.
building element, as opposed to a single defect approach. [16] S.D. Foltz, D.T. Mckay, Condition assessment aspects of an asset management
program, US Army Corps of Engineers, Washington DC, USA, 2008. http://
www.dtic.mil/get-tr-doc/pdf?AD=ADA476675.
CRediT authorship contribution statement [17] J.-S. Lee, Value engineering for defect prevention on building façade, J. Constr.
Eng. Manag. 144 (2018) 04018069, https://doi.org/10.1061/(ASCE)CO.1943-
Clara Pereira: Conceptualization, Methodology, Formal analy- 7862.0001500.
[18] J. Georgiou, P.E.D. Love, J. Smith, A comparison of defects in houses constructed
sis, Investigation, Writing - original draft, Visualization. Ana Silva: by owners and registered builders in the Australian State of Victoria, Struct.
Conceptualization, Formal analysis, Writing - review & editing. Surv. 17 (1999) 160–169, https://doi.org/10.1108/02630809910291343.
Jorge de Brito: Conceptualization, Methodology, Writing - review [19] J. Layzell, S. Ledbetter, FMEA applied to cladding systems - Reducing the risk of
failure, Build. Res. Inf. 26 (1998) 351–357, https://doi.org/10.1080/
& editing, Project administration. José D. Silvestre: Conceptualiza- 096132198369689.
tion, Methodology, Writing - review & editing. [20] F. Ruiz, A. Aguado, C. Serrat, J.R. Casas, Optimal metric for condition rating of
existing buildings: is five the right number?, Struct. Infrastruct. Eng. (2019),
https://doi.org/10.1080/15732479.2018.1557702.
Declaration of Competing Interest [21] A. Silva, J. de Brito, P.L. Gaspar, Methodologies for service life prediction of
buildings. With a focus on façade claddings, Springer, Switzerland, 2016.
doi:10.1007/978-3-319-33290-1.
The authors declare that they have no known competing finan- [22] G. Crevello, N. Hudson, P. Noyce, Corrosion condition evaluations of historic
cial interests or personal relationships that could have appeared concrete icons, Case Stud. Constr. Mater. 2 (2015) 2–10, https://doi.org/
to influence the work reported in this paper. 10.1016/j.cscm.2014.12.005.
[23] M. Noya, A.L.T.S. da Motta, M. Moura, R.C. de Souza, B. Barzellay, Risk analysis:
a methodology applicable to the building inspection, J. Civ. Eng. Archit. 10
Acknowledgements (2016) 1052–1058, https://doi.org/10.17265/1934-7359/2016.09.008.
[24] V. Hutsebaut-Buysse, Maintenance in historic buildings in Belgium and
Portugal, Universidade de Lisboa (2016).
The authors gratefully acknowledge the support of CERIS from [25] I. Flores-Colen, J. de Brito, V. Freitas, Discussion of criteria for prioritization of
IST-UL. This work was supported by the Fundação para a Ciência predictive maintenance of building façades: Survey of 30 experts, J. Perform.
C. Pereira et al. / Construction and Building Materials 249 (2020) 118743 15

Constr. Facil. 24 (2010) 337–344, https://doi.org/10.1061/(ASCE)CF.1943- historical buildings, Stud. Conserv. (2019) 600–607, https://doi.org/10.1080/
5509.0000104. 00393630.2019.1638666.
[26] Agence Qualité Construction, Fiche pathologie bâtiment - Infiltrations par les [53] A. Moura, I. Flores-Colen, J. de Brito, Study of the effect of three anti-graffiti
liaisons menuiserie extérieure/gros œuvre (Building pathology file - Leakages products on the physical properties of different substrates, Constr. Build.
through external carpentry connections/substructure), (2013). http://www. Mater. 107 (2016) 157–164, https://doi.org/10.1016/
qualiteconstruction.com/node/1277 (accessed 24 August 2018). j.conbuildmat.2015.12.181.
[27] D. Camuffo, Microclimate for cultural heritage: Conservation, restoration, and [54] A. Moura, I. Flores-Colen, J. de Brito, A. Dionisio, Study of the cleaning
maintenance of indoor and outdoor monuments, 2nd ed., The Netherlands; effectiveness of limestone and lime-based mortar substrates protected with
Boston, MA USA, Elsevier, Amsterdam, 2014. anti-graffiti products, J. Cult. Herit. 24 (2017) 31–44, https://doi.org/10.1016/j.
[28] C.M. Junior, H. Carasek, Relationship between the deterioration of multi story culher.2016.04.004.
buildings facades and the driving rain, Rev. La Construcción. 13 (2014) 64–73 [55] P.L. Gaspar, J. de Brito, Minimum required performance level for rendered
(accessed 15 February 2017) http://rdlc.alerta.cl/index.php/rdlc/article/view/ facades, in: 11DBMC Int. Conf. Durab. Build. Mater. Components, Istanbul,
251/15. Turkey, 2008. doi:10.13140/RG.2.1.2212.1685.
[29] M. Gil, J. Aguiar, A. Seruya, R. Veiga, L. Carvalho, H. Vargas, J. Mirão, A. Candeias, [56] A. López, G.A. Guzmán, A.R. Di Sarli, Color stability in mortars and concretes.
Colour assays: An inside look into Alentejo traditional limewash paintings and Part 1: Study on architectural mortars, Constr. Build. Mater. 120 (2016) 617–
coloured lime mortars, Color Res. Appl. 36 (2011) 61–71, https://doi.org/ 622, https://doi.org/10.1016/J.CONBUILDMAT.2016.05.133.
10.1002/col.20584. [57] J. Lahdensivu, H. Tietäväinen, P. Pirinen, Corrosion of Reinforcement in Existing
[30] I. Flores-Colen, J. de Brito, V.P. de Freitas, Stains in facades’ rendering - Concrete Façades, in: Int. Conf. Durab. Build. Mater. Components, XII DBMC,
Diagnosis and maintenance techniques’ classification, Constr. Build. Mater. 22 Porto, Portugal, 2011: pp. 253–274. doi:10.1007/978-3-642-37475-3_10.
(2008) 211–221, https://doi.org/10.1016/j.conbuildmat.2006.08.023. [58] J.M. Dias, M.J. Henriques, L. Matias, M.S. Ribeiro, S. Raposo, NDT techniques for
[31] S. Sazedj, A. Morais, M.T. Pinheiro-Alves, R. Silva, Pathologies—Incompatibility the analysis of anomalies related with durability - heritage buildings with
of materials and human intervention in a historic building of Elvas, in: J.M.P.Q. masonry walls and confining concrete elements, in: Encontro Nac (Ed.), BETÃO
Delgado (Ed.), New Approaches to Build. Pathol. Durab., Springer, Singapore, ESTRUTURAL - BE2018, Portugal, Lisboa, 2018, pp. 1767–1776.
2016. doi:10.1007/978-981-10-0648-7. [59] I. Flores Colen, J. de Brito, V.P. de Fretias, Assessment of in use performance
[32] W.H. Ransom, Building failures. Diagnosis and avoidance, 2nd ed.,., Spon Press, parameters of rendering façades, in: V.P. de Freitas (Ed.), A State-of-the-Art
London, UK, 1987. Rep. Build. Pathol., CIB International Council for Research and Innovation in
[33] A.M. Waldum, Restoration of masonry facades, renders and final coats in a Building and Construction, Porto, Portugal, 2013: pp. 65–72. http://
severe climate, Build. Res. Inf. 21 (1993) 51–55, https://doi.org/10.1080/ site.cibworld.nl/dl/publications/pub_393.pdf.
09613219308727255. [60] F. Wilson, Building materials evaluation handbook, Van Nostrand Reinhold
[34] J. Douglas, B. Ransom, Understanding building failures, 3rd ed., Taylor & Company Inc., New York, USA, 1984. doi:10.1007/ 978-1-4684-6650-8.
Francis, London, United Kingdom, 2007. [61] E. Di Giuseppe, Analytical and experimental methods for the assessment of
[35] N.N.O. Bakri, M.A.O. Mydin, General building defects: Causes, symptoms and the biological proliferation in buildings, in: Nearly Zero Energy Build. Prolif.
remedial work, Eur. J. Technol. Des. 3 (2014) 4–17. Microorg. A Curr. Issue Highly Insul. Airtight Build. Envel., Springer,
[36] R. Bonshor, L. Bonshor, R. Sadgrove, Cracking in buildings, 2nd ed., IHS, Dordrecht, The Netherlands, 2013: pp. 37–58. doi:10.1007/978-3-319-
Berkshire, UK, 2016. 02356-4_5.
[37] K. Gaspar, M. Casals, M. Gangolells, Classifying system for façades and [62] F.L. Guerra, E.G. da Cunha, F. Galli, Analysis of microclimate in a historical
anomalies, J. Perform. Constr. Facil. 30 (2016) 04014187, https://doi.org/ building to assess the probability of recurrence of filamentous fungi, in: J.M.P.
10.1061/(ASCE)CF.1943-5509.0000693. Q. Delgado (Ed.), Recent Dev. Build. Diagnosis Tech., Springer, Singapore, 2016:
[38] P.L. Gaspar, J. de Brito, Tipos de vida útil das construções (Types of service life pp. 195–213. doi:10.1007/978-981-10-0466-7_11.
of constructions), in: 3.o Encontro Nac. Sobre Patol. e Reabil. Edifícios Patorreb [63] R.M.S. Almeida, E. Barreira, Monte Carlo simulation to evaluate mould growth
2009, Porto, Portugal, 2009: pp. 301–306. doi:10.13140/RG.2.1.4886.1208. in walls: the effect of insulation, orientation, and finishing coating, Adv. Civ.
[39] Instituto Português da Qualidade, NP EN 1991-1:2010. Eurocódigo 1: Acções Eng. 2018 (2018) 1–12, https://doi.org/10.1155/2018/8532167.
em estruturas. Parte 1-4: Acções gerais. Acções do vento (Eurocode 1: Actions [64] D.S. Watt, Buiding pathology. Principles and practice, 2nd ed., Blackwell
on structures. Part 1-4: General actions. Wind actions), Instituto Português da Publishing, Oxford, United Kingdom, 2007.
Qualidade, Caparica, Portugal, 2010. [65] H. Morillas, M. Maguregui, J. Trebolazabala, J.M. Madariaga, Nature and origin
[40] Instituto Português da Qualidade, NP EN 1991-1-5:2009. Eurocódigo 1: Acções of white efflorescence on bricks, artificial stones, and joint mortars of modern
em estruturas. Parte 1-5: Acções gerais. Acções térmicas (Eurocode 1: Actions houses evaluated by portable Raman spectroscopy and laboratory analyses,
on structures. Part 1-4: General actions. Thermal actions), Instituto Português Spectrochim, Acta Part A Mol. Biomol. Spectrosc. 136 (2015) 1195–1203,
da Qualidade, Caparica, Portugal, 2009. https://doi.org/10.1016/j.saa.2014.10.006.
[41] J. de Brito, F.A. Branco, M. Ibañez, Knowledge-based concrete bridge inspection [66] R. Cruzan, Manager’s guide to preventive building maintenance, The Fairmont
system, Concr. Int. 16 (1994) 29–63. Press Inc, Lilburn, GA USA, 2009.
[42] A. Menezes, M. Glória Gomes, I. Flores-Colen, In-situ assessment of physical [67] E.A.-L. Teo, N. Harikrishna, A quantitative model for efficient maintenance of
performance and degradation analysis of rendering walls, Constr. Build. Mater. plastered and painted façades, Constr. Manag. Econ. 24 (2006) 1283–1293,
75 (2015) 283–292, https://doi.org/10.1016/j.conbuildmat.2014.11.039. https://doi.org/10.1080/01446190600987704.
[43] I. Flores-Colen, J. de Brito, Renders, in: M.C. Gonçalves, F. Margarido (Eds.), [68] T. Yates, Mechanisms of air pollution damage to brick, concrete and mortar, in:
Mater. Constr. Civ. Eng. Sci. Process. Des., Springer, Switzerland, 2015: pp. 53– P. Brimblecombe (Ed.), Eff, Imperial College Press, London, United Kingdom,
122. doi:10.1007/978-3-319-08236-3. Air Pollut. Built Environ., 2003, pp. 107–132.
[44] C. Pereira, J. de Brito, J.D. Silvestre, Contribution of humidity to the degradation [69] B. de Vekey, Deterioration and conservation of masonry, in: P. Domone, J.
of façade claddings in current buildings, Eng. Fail. Anal. 90 (2018) 103–115, Illston (Eds.), Constr. Mater. Their Nat. Behav., 4th ed., Spon Press, London,
https://doi.org/10.1016/j.engfailanal.2018.03.028. United Kingdom, 2010: pp. 291–299.
[45] J.V. Paiva, J. Aguiar, A. Pinho, Guia técnico de reabilitação habitacional [70] S. Madureira, I. Flores-Colen, J. de Brito, C. Pereira, Maintenance planning of
(Technical guide for housing rehabilitation), Instituto Nacional de Habitação facades in current buildings, Constr. Build. Mater. 147 (2017) 790–802, https://
(INH); Laboratório Nacional de Engenharia Civil (LNEC), Lisbon, Portugal, 2006. doi.org/10.1016/j.conbuildmat.2017.04.195.
[46] B. Blocken, J. Carmeliet, Overview of three state-of-the-art wind-driven rain [71] I.H. Seeley, Building maintenance, 2nd ed., Palgrave, Hampshire, United
assessment models and comparison based on model theory, Build. Environ. 45 Kingdom, 1987.
(2010) 691–703, https://doi.org/10.1016/j.buildenv.2009.08.007. [72] W.-K. Chong, S.-P. Low, Assessment of defects at construction and occupancy
[47] B. Blocken, D. Derome, J. Carmeliet, Rainwater runoff from building facades: A stages, J. Perform. Constr. Facil. 19 (2005) 283–289, https://doi.org/10.1061/
review, Build. Environ. 60 (2013) 339–361, https://doi.org/10.1016/j. (ASCE)0887-3828(2005)19:4(283).
buildenv.2012.10.008. [73] A.E. Grimmer, Keeping it clean: Removing exterior dirt, paint, stains and
[48] M.J. Carretero-Ayuso, A. Moreno-Cansado, J. de Brito, Study of the prevalence graffiti from historic masonry buildings, U.S. Department of the Interior,
of critical and conflict-prone points in facades, Eng. Fail. Anal. 75 (2017) 15–25, National Park Service Cultural Resources, Heritage Preservation Services,
https://doi.org/10.1016/j.engfailanal.2016.12.009. Washington DC, USA, 1988.
[49] J. Hinks, G. Cook, The technology of building defects, Spon Press, Oxford, [74] R.O. Heckroodt, Guide to the deterioration and failure of building materials,
United Kingdom, 1997. Thomas Telford, London, United Kingdom, 2002.
[50] BRE Housing Defects Prevention Unit, Domestic draughtproofing: balancing [75] J. Tuna, J. Feiteira, I. Flores-Colen, M.F.C. Pereira, J. de Brito, In situ
ventilation needs against heat losses. Defect Action Sheet 136, Housing characterization of damaging soluble salts in wall construction materials, J.
Defectis Prevention Unit, Building Research Establishment, Department of the Perform. Constr. Facil. 29 (2015) 04014127, https://doi.org/10.1061/(asce)
Environment, Watford, United Kingdom, 1989. cf.1943-5509.0000616.
[51] W. Zillig, K. Lenz, K. Sedlbauer, M. Krus, Condensation on façades - influences [76] C. Pereira, A. Silva, J. de Brito, I. Flores-Colen, J.D. Silvestre, Contribution of
of construction type and orientation, in: J. Carmeliet, H. Hens, G. Vermeir cracking and spalling to the degradation of façade claddings in current
(Eds.), Res. Build. Phys., A.A. Balkema Publishers, Leuven, 2003: pp. 437–444. buildings, in: CIB World Build. Congr. 2019, Hong Kong SAR, China, 2019: p.
https://www.irbnet.de/daten/iconda/CIB2387.pdf (accessed 15 February WC0020.
2017). [77] T. Esteves, I. Flores-Colen, C.M. Silva, Inspection and numerical modeling of
[52] C. Ferreira, V.P. Freitas, J.M.P.Q. Delgado, The influence of hygroscopic cracking in existing nonbearing walls, J. Perform. Constr. Facil. 32 (2018)
materials in the fluctuation of relative humidity in museums located in 04018033, https://doi.org/10.1061/(ASCE)CF.1943-5509.0001169.
16 C. Pereira et al. / Construction and Building Materials 249 (2020) 118743

[78] E. Bauer, C.B. Piazzarollo, J.S. de Souza, D.G. dos Santos, Relative importance of [82] H.F. Kaiser, The varimax criterion for analytic rotation in factor analysis,
pathologies in the severity of facade degradation, J. Build. Pathol. Rehabil. 5 Psychometrika. 23 (1958) 187–200, https://doi.org/10.1007/BF02289233.
(2020) 1–10, https://doi.org/10.1007/s41024-020-0072-6. [83] J.F. Hair, W.C. Black, B.J. Babin, R.E. Anderson, Multivariate data analysis, 6th
[79] C.P. Santos, L. Matias, A.C. Magalhães, M.R. Veiga, Application of thermography Editio, Pearson Prentice Hall, New Jersey, USA, 2006.
and ultra-sounds for wall anomalies diagnosis a laboratory research study, in: [84] M. Azzalin, M. Lauria, Building pathology database and maintenance approach
NDT-CE 2003 - Int. Symp. Non-Destructive Test. Civ. Eng., Berlin, Germany, in a well defined contex : Calabrian historical centres, in: 10DBMC Int. Conf.
2003. https://www.ndt.net/article/ndtce03/papers/v082/v082.htm. Durab. Build. Mater. Components, Lyon, France, 2005.
[80] P.L. Gaspar, J. de Brito, Limit states and service life of cement renders on [85] J.C. Wayman, Methods for handling missing data: What is it and how can I use
façades, J. Mater. Civ. Eng. 23 (2011) 1396–1404, https://doi.org/10.1061/ it?, in: Annu. Meet. Am. Educ. Res. Assoc., Chicago, IL USA, 2003: pp. 1–16.
(ASCE)MT.1943-5533.0000312.
[81] StatSoft Inc., STATISTICA (data analysis software system), (2007). www.
statsoft.com.

You might also like