Advanced Engineering Informatics 30 (2016) 54–64

Contents lists available at ScienceDirect

Advanced Engineering Informatics

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

Information requirements for earthquake damage assessment

of structural walls
Engin Burak Anil a,⇑, Burcu Akinci a,1, James H. Garrett a,2, Ozgur Kurc b,3
a
Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA 15213, USA
b
Middle East Technical University, Ankara 06800, Turkey

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

Article history: The engineering analysis for determining the remaining seismic capacity of buildings following earth-
Received 22 February 2015 quakes requires performing structural calculations, observations of the actual damage, and applying
Received in revised form 28 October 2015 extensive engineering judgment. Additionally, the analysis should often be performed under stringent
Accepted 30 December 2015
time requirements. This study identifies the information requirements for representing the damage infor-
Available online 18 January 2016
mation and performing the visual damage assessment of structural walls. The damage descriptions for
seven common damage modes of structural walls were studied by employing the affinity diagramming
Keywords:
method. The study showed that the information required to represent the damaged conditions can be
Damage assessment
Information
grouped under five broad categories and using seventeen damage parameters. A sensitivity analysis
BIM showed that the damage parameters have varying degrees of importance. The results of the study can
Earthquake be used to develop formal representation of damage information in information models and potentially
FEMA allow better allocation of data collection time in the field.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction components by observing the indications of damage, performing

structural computations, and synthesizing the structural proper-
The earthquake regions around the world contain a large stock ties and construction details of a building [5].
of buildings. For example, only in Los Angeles, CA, there is a stock Manual approaches for evaluating structures have been criti-
of estimated 40,000 particularly vulnerable RC buildings, due to cized for being error-prone, slow, and subjective [6–9]. Therefore,
old design practices [1,2]. The total number of buildings, which will several researchers have been studying ways to automate the dam-
need to be assessed following a future earthquake is much higher age assessment process [9–28]. Studies on computer vision tech-
given that it would be necessary to assess not just vulnerable RC niques show promising advancements for capturing and
buildings, but all buildings that got impacted from the earthquake. identifying damage indicators, such as cracking, spalling, and dis-
Historical evidence and research studies show that there is a press- placements, using machine vision methods [9,14,16–22,26,29–33
ing need to assess the damage severities of buildings objectively, ]. On the other hand, automation of the actual damage assessment
accurately, and rapidly, and quantify the effects of the earthquake procedures using the captured damage indicators is not a well-
on the structural properties of the components [3,4]. studied subject. The studies on automating the damage assessment
Current methods for damage assessment are manual and rely procedures include using augmented reality along with the resid-
heavily on the structural engineering expertise of the inspectors ual story drift as the damage metric, and using damage indicators
and their engineering judgment [5]. Inspectors are expected to (e.g., cracking, spalling) in a fragility analysis to identify the after-
have a good level of understanding and experience in assessing shock vulnerability of buildings [10,11]. To the best knowledge of
the effects of the ground motion on the seismic performance of the authors, however, none of the previous studies address the
automation of the engineering analyses for damage assessment,
based on the FEMA 306 guideline, which practicing engineers have
⇑ Corresponding author. Tel.: +1 (412) 482 2013. to follow [34].
E-mail addresses: [email protected] (E.B. Anil), [email protected] (B. Akinci),
This study focuses on the FEMA 306 manual for ‘‘Evaluation of
[email protected] (J.H. Garrett), [email protected] (O. Kurc). Earthquake Damaged Concrete and Masonry Wall Buildings” [34].
1
Tel.: +1 (412) 268 2959. We picked this guideline to focus on for three reasons. First, it is
2
Tel.: +1 (412) 412 268 5090. the standard document, which the engineers have to follow when
3
Tel.: +90 312 210 2447.

http://dx.doi.org/10.1016/j.aei.2015.12.002
1474-0346/Ó 2016 Elsevier Ltd. All rights reserved.
 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64 55

assessing the performance of damaged buildings [34]. FEMA 306 is of the details of individual damage modes. Through this way, the
referenced by the latest version of ASCE/SEI 41-13 (2013) for the abstraction process can support all of the damage modes for rein-
evaluation and repair of earthquake damage, component type forced concrete walls. This, in fact, requires bottom-to-top discov-
identification, visual condition assessment, and non-destructive ery of the hierarchical structure of information requirements.
testing. Second, the guideline builds on a thorough investigation Therefore, we need a structured way of studying the damage
of existing body of research and presents a procedure that can used assessment guidelines, sort them into hierarchies of damage infor-
for a wide variety of damage modes and component types [34]. mation, and determine the abstraction of damage information.
Finally, it is being adopted by countries, such as New Zealand, thus Using the results of such an information requirement identification
reaching wider application [35]. study, a representation to support various tasks associated with
According to FEMA 306, engineering analyses are performed on damage assessment can be developed.
those earthquake damaged buildings, which are tagged for further This paper studies the information items regarding all of the
analysis after a rapid assessment [34]. The current practice of dam- damage parameters, such as cracking, spalling, crushing, reinforce-
age assessment requires the quantification of the degrading effects ment bar damage, and residual displacements. The identified infor-
of the earthquake by determining the damage modes and damage mation requirements will be used for developing a BIM schema for
severities of structural components, which is called the engineer- representing damage conditions and for automatically assessing
ing analysis [34]. The damage mode is the dominant behavior of the damage modes and severities of RC walls in a future study.
a structural component (e.g., ductile flexure, shear, etc.) and the The proposed approach builds on the FEMA 306 guideline for the
damage severity is a measure of how much the damage has pro- ‘‘Evaluation of Earthquake Damaged Concrete and Masonry Wall
gressed in the dominant mode. The damage mode and severity Buildings” in identifying the information requirements associated
determines the reduction in strength, stiffness, and displacement with the damage parameters stated above.
as a result of earthquake damage. The research approach utilizes affinity-diagramming method
The damage modes and damage severities of structural compo- for identifying the overarching patterns of information require-
nents are identified using a combination of strength calculations ments for the visual assessment and groups the information
and visual assessment. In strength calculations, the lateral forces requirements [44]. The affinity diagrams suggested an initial list
which would generate particular damage modes are calculated of eighteen parameters. A sensitivity analysis was performed to
using strength equations. The actual damage behavior of compo- validate the results and analyze the relative importance of the
nents can differ from theoretical behavior determined by strength damage parameters. The sensitivity analysis showed that damage
calculations, due to several reasons, such as differences in material parameters that occur at high levels of damage and those that
strengths between design and actual, and interactions between are specific to certain types of damage behavior has more impact
building components and loading conditions. The visual observa- on the assessment results. Given the fact that engineers have only
tions allow the engineer to determine the actual damage mode a limited amount of time to collect data in the field, the results of
of the components by assessing how the damage indicators are the sensitivity analysis can potentially be used to prioritize collect-
formed (e.g., cracking, spalling, crushing, rebar damage, and resid- ing damage parameters.
ual displacements). However, the visual assessment is not ade-
quate alone to determine the damage mode and severity. 2. Damage assessment procedures based on FEMA 306
Therefore, a combination of strength analysis and visual assess-
ment is employed. FEMA 306 presents a very detailed analysis of Following earthquakes, a three-step procedure is applied
the visual damage indicators for different damage modes. [34,45,46]. First, within few days after an earthquake, experts per-
Previous research has found seven common damage modes for form rapid assessment to classify whether the buildings in the
reinforced concrete wall components (i.e., piers and spandrels), impacted area are safe, unsafe, or require restrictions in their usage
including ductile flexure, pre-emptive shear, diagonal compres- [45]. Safe buildings can be used without any restrictions. Unsafe
sion, boundary compression, sliding shear, and pier rocking [34]. buildings should not be entered under any circumstance. Some
Reinforced concrete frames generally exhibit either flexure or restrictions on the usage of a building can be time limits or by
shear type of failure. Compared to RC frames, distinguishing the location. For example, an inspector might restrict the access
between damage modes can be especially challenging for wall to certain parts of a building, which contain potential hazards.
components, considering that the damage modes can look alike Engineers generally have less than 30 min to perform a rapid
at low severities and calculations are required to determine the assessment per building [45]. Second, a detailed assessment is
governing mode [34]. Therefore, this study focuses on structural performed on those buildings that were tagged for restricted
walls. usage [46]. Detailed assessment is similar to the rapid assess-
Building Information Models (BIM) can potentially support rep- ment in execution and function, but it is more thorough. Finally,
resenting the damaged conditions of the buildings for visual within the weeks after an earthquake, engineering analyses
assessment and structural information for engineering analysis, are performed on safe and restricted buildings, in order to quantify
such as the configuration, reinforcement details, and finite element possible effects of damage on the structural properties of struc-
models [12,36–40]. Hence, BIMs can be used to support the engi- tural components and to design retrofitting measures [34]. In all
neering analyses for strength analysis and visual assessment of these stages, collection of the damage data and analysis of the
[36,38]. However, current BIMs are not developed to represent visual damage information using engineering knowledge are key
damaged conditions of buildings [12,32,41]. Therefore, represent- factors for accurate assessment of the damage levels [34].
ing the damage information in BIMs need to be investigated fur- The engineering analysis has more complex requirements than
ther. This study focuses on the visual assessment aspect of the rapid and detailed assessments. The goal of an engineering analysis
analysis and develops the information requirements of visual is to determine the remaining strength, stiffness, and displacement
assessment. The results of this study is combined with a formaliza- capacity of structural components [34]. In order to determine the
tion of the strength analysis for a complete engineering analysis in remaining capacities of damaged components, the source of the
[42] current damage and its nonlinear mechanism must be determined.
A representation requires abstraction of information, which This is not a straightforward task since the loads on the components
depicts the common structure of all of the features required for and the relative stiffness of connected components influence the
damage assessment [43]. The abstraction should also support all damage mode and severity.
56 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64

The engineering analysis procedure of FEMA 306 has a 5 step by having access to accurate and complete information on existing
process for the identification of the damage modes and severities structural conditions of the building.
using the visual damage information and capacity calculations For each component of the structural system, the engineer clas-
[34]. The first step involves assembling information about the sifies the damage according to behavior mode and severity. The
earthquake and a building, which involves identifying the compo- engineer also categorizes the severity of damage for each type of
nents making up the lateral and vertical load bearing system. The damage encountered within any component. FEMA 306 guideline
second step involves identifying types of components by determin- classifies damage for reinforced concrete walls into common, less
ing the governing inelastic behavior of each element. FEMA 306 common or uncommon modes [34]. This research focuses only
defines 5 types of components: stronger pier, weaker spandrel, on the common modes. Only unreinforced masonry is considered
stronger spandrel, weaker pier, and pier-spandrel panel zone. This for infilled frames, since reinforced infills are not very common
step requires calculations for comparing the relative strengths and in reinforced concrete construction [34]. The damage modes that
stiffness of components. The third step involves documenting the are considered in this study are as following (the letters in paren-
damage. The fourth step involves the classification of the observed thesis indicate the codes of the damage modes):
component damage by determining the damage modes and dam-
age severities using strength calculations and visual observations. 1. Ductile Flexure (A): Moment strength governs the strength of
Damage patterns observed on the components are used to deter- the component.
mine the damage severity for the determined damage mode of a 2. Flexure/Diagonal tension (B): Initially, the moment strength
component. The fifth and last step involves the verification of the governs. High-ductility shear strength is less than the moment
results of the analysis by comparing the results of the visual assess- strength. Low-ductility shear strength is higher than the
ment and the strength calculations for the components. moment strength.
The main goals of the information assembly step (the first step 3. Flexure/Diagonal compression (C): Initially, the moment
in the above process) are: (1) gathering information about the strength governs. High ductility web crushing strength is less
damaging earthquake and generating a capacity curve for the than the moment strength.
building by conducting structural analysis and (2) collecting infor- 4. Flexure/Sliding shear (D): Initially, the moment strength gov-
mation about the structural design of a building and the configura- erns. Sliding shear strength is less than the moment strength.
tion of the load bearing system. The information about the 5. Flexure/Boundary compression (E): Moment strength is higher
damaging ground motion is generally made available shortly after than all other modes even after possible degradation.
the earthquake. The procedures for analyzing structural models to 6. Pre-emptive diagonal tension (H): Low-ductility shear strength
generate capacity curves and estimating maximum global dis- is less than the moment strength.
placements are standardized according to accepted practices [47]. 7. Global or individual pier foundation rocking (M & N): The lat-
Gathering building information entails reviewing construction eral force, which would impose rocking should be less than all
and structural drawings, previous structural computations, con- other strengths.
struction records, and previous inspection reports. Such informa-
tion is useful for identifying structural components by The results of the strength analysis and visual assessment
performing strength calculations, and for determining the scope should corroborate. If the results of the two analyses do not corrob-
of the field inspections and testing programs. For example, if the orate, the potential error sources need to be checked and the anal-
locations and sizes of reinforcing bars are known accurately, ysis repeated. FEMA 306 lists several potential error sources. The
expensive and intrusive tests may not be necessary [34]. Inaccu- distribution of the lateral forces, which was used in the structural
racy of the information gathered in this step might reflect to the analysis, may be different from the actual. The strength of the com-
overall results as discrepancies between calculations and visual ponents may differ from the actual. The intensity of the damaging
observations. Additionally, in the evaluation stage, expected ground motion may differ from that assumed in the structural
strengths can be penalized up to 25% depending on the confidence analysis.
on the accuracy and completeness of information, which would There can be several problems in the engineering analyses. The
increase the retrofitting costs [47]. design and construction documents can become outdated as a
The engineers are expected to identify the types of components result of retrofitting or remodeling of the building. The calculation
(e.g., weaker spandrel, stronger pier, etc.) by anticipating the of strength of components and structural analysis are performed
inelastic lateral mechanism for each element. An element entails by interpreting the information on the construction drawings
a vertical or horizontal portion of a building, which resist lateral and design documents, which might be obsolete. The current FEMA
and vertical loads. An element is an assembly of components, such 306 process requires manual re-entry of information into com-
as wall piers, beams, columns, and slabs. For example, a structural puter programs, which is error-prone and redundant. Manual doc-
wall with openings can be made up of components, such as piers umentation of damage can lack important details about the
and beams. The components in such a configuration will also observations, the inspection may not be as thorough as required
include the piers on a different axis, which make up the flanged by the assessment guidelines, and the results may be subjective
sections. The identification of these components is central to the and incorrect.
overall process and requires calculation of strengths of the compo-
nents and a thought process by the engineer to anticipate the
lateral behavior of the components using the visual damage 3. Background research
patterns.
Intrusive and non-intrusive tests are performed to document 3.1. Methods for identification of information requirements
the damage, material strengths, and relative ages of cracks. Visual
observation is the most important non-intrusive test to document Developing a schema for representing the damaged conditions
the earthquake damage. Some of the special techniques include of buildings and supporting the visual assessment tasks, requires
selective removal of material, material sampling, sounding, petrog- the information requirements of such a schema. Methods for iden-
raphy, spectral analysis, and penetrating radar. Tests, which tifying information requirements for developing representations
require specialized equipment and experienced personnel to per- and reasoning mechanisms can be grouped into two: exhaustive
form, are generally expensive. Such expensive tests can be reduced listing and structured methods.
 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64 57

3.1.1. Exhaustive listing of shear crack widths”. This allows us to define the relevant enti-
If the problem being studied is relatively small, the information ties, in this case ‘‘shear cracks”, and their attributes, ‘‘widths”,
requirements can be identified by studying the process, which the and constraints that apply to these attributes. In designing a repre-
representation will support, and listing all the entities or concepts, sentation schema such abstraction of entities, attributes, and rela-
which are involved in the solution of the problem [38,40,48,49]. In tionships are essential.
this approach, studying standards, mathematical equations, which In the first step, information was collected about the visual
are involved in the process, documents and research studies, are damage assessment practice. In total, 278 statements were
common. Mathematical models and computational methods can extracted. In some occasions, further interpretation of the state-
also drive the information identification. For example, for finite ments was required to deduce the meaning of the statements.
element procedures, the finite element model and the solution For example, some of the individual statements are meaningful
method determines the structure and the content of the represen- only in the context of a given damage mode. The statement ‘‘Rein-
tation method [50–52]. forcement has fractured” in the ductile flexural (A) damage mode
A commonality among these methods is that they are not well refers to the fracturing of the flexural reinforcement, not just any
structured. The information requirements for designing the infor- reinforcement [5]. This can be deduced from the fact that this type
mation model or decision system are not collected, studied, and of behavior can occur if a component has sufficient reinforcement
evaluated in a formal way. For well-defined problems, such as a to prevent shear failure and flexural reinforcement is not heavy [5].
mathematical computation or small-scale problems, unstructured Hence, moment demand exceeds the shear demand. Such state-
methods may be easy to apply and yield direct results rapidly. How- ments were modified to depict the exact meaning. Each statement
ever, a limitation of this approach is that as the number of entities was written on a note card along with its damage mode, an iden-
increases and the problem becomes more complex the organization tification number for the statement, and the damage severity, at
and classification of information becomes more difficult [53]. which the statement is observed.
The statements constitute the first level of the affinity diagram.
The second step included building up a structure from particular to
3.1.2. Affinity diagramming method
general, starting from the first level. The process is as follows. First,
Affinity diagramming is used for discovering common themes
a note containing a single statement is put on the board. Other
and issues among individual examples of a work practice [54].
statements are browsed to see which statements have an affinity
Affinity diagrams are built by induction to create a representation
with the first statement. Two notes are said to have an affinity if
of a work practice starting from its details and moving towards
they say similar things [54]. Two statements are identified as hav-
general concepts by grouping details and generating new insights
ing an affinity if they help differentiating between different dam-
about overarching patterns of data. Since the structure of an affin-
age modes. Notes having an affinity are put together.
ity diagram is built from detail, differences between individual
The goal of the affinity diagram is building a hierarchy of state-
examples can be accommodated.
ments in such a way that the grouping in the hierarchy help in dif-
A common process to build an affinity diagram starts with gath-
ferentiating between different damage modes. In order to illustrate
ering statements or requirements by studying the target work
the process of grouping, consider the following statements about
practice (e.g., user interviews) and follows as grouping these state-
cracking (Table 1). It can be seen that the statements are about dif-
ments by their affinities (i.e., statements which say similar things)
ferent types of cracks (i.e., flexural, shear, or any type of crack
and building a hierarchical structure by creating super-groups that
regardless of type), there are width comparisons, and some of
have affinities in several levels (generally 3 or 4 groups) [54].
the widths are recurring (e.g., 0.8 mm, 1.6 mm, etc.). One way of
Considering the complexity of the damage assessment tasks,
grouping these statements is by highlighting the crack type as
using a structured approach, such as affinity diagramming
the overarching pattern of the statements and group by recurring
approach is appropriate. The affinity diagram can be built for the
crack widths:
visual damage assessment process by first exhaustively listing all
of the statements describing the damage modes and treating each

Level 2 – Type of cracks, which are narrower than 3.2 mm.
damage mode as examples of the same work practice. The affinities
– Level 1 – Flexural crack width <3.2 mm.
between statements in the descriptions of the damage modes can
– Level 1 – Shear crack width <3.2 mm.
be discovered through an induction process: the statements are
grouped in multiple levels to reveal new insights about the pat-
This type of grouping ignores the fact that crack widths are only
terns of features, which engineers look for during the damage
meaningful if the type of the crack is known (e.g., flexural, shear),
assessment process. In this study, the described process was fol-
which is one of the primary indicator of the damage mode. For a
lowed to identify the information requirements for developing a
certain damage mode, we expect to see cracks of a certain type
schema for damage assessment of structural walls.
to be wider than the other [34]. For example, wide flexural crack-
ing and much narrower shear cracks or no shear cracks at all may
4. Approach be an indication of ductile flexural behavior. On the other hand, X
shaped cracking extending along the diagonals of a wall with little
The affinity diagram was built in two steps: First, statements or no flexural cracking may indicate a brittle shear type of behav-
that make up the damage descriptions in the FEMA 306 guidelines ior. In other words, crack width is secondary to the type of crack.
were extracted. Each statement was recorded on note cards. Sec- Better way of grouping the same statements is by highlighting
ond, affinities were identified between these statements by induc- the crack width as the overarching pattern and grouping by crack
tion. The interpretation of the statements and potential affinities type:
are determined by considering the domain knowledge.
The note cards were grouped, regrouped, and organized to
Level 2 – Width of flexural cracks.
reveal patterns in the information in multiple levels. The levels – Level 1 – Flexural cracks <3.2 mm.
represents the constituting elements of the notes in a group at – Level 1 – Flexural cracks <4.8 mm.
an abstract level. In other words, each level is a description of what
the note cards contain or describe. For example, a series of defini- This grouping recognizes that the crack type should be known
tions about the limits of shear cracking may be grouped as ‘‘Limits before setting a limit on the crack width, for determining the
58 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64

Table 1 Table 2
A selected subset of statements about crack types and widths. The statements contain Other statements extracted from the guideline about cracking.
crack types and crack widths.
# Statement
# Statement
1 No significant spalling or vertical cracking
1 Flexural crack width <3.2 mm 2 No significant spalling or vertical cracking
2 Flexural crack width <6.4 mm 3 No vertical cracking or spalling
3 Flexural crack width <6.4 mm 4 Spalling or vertical cracking at toe regions in plastic hinge zone
4 Flexural crack width <4.8 mm 5 Spalling or vertical cracking at toe regions in plastic hinge zone
5 Shear crack width >4.8 mm & <9.6 mm 6 Vertical cracks at the extreme fibers of the plastic hinge zone
6 Shear crack width <3.2 mm 7 One or more wide shear cracks begin to form
7 Shear crack width <1.6 mm 8 Wide shear cracks concentrated in a single crack
8 No shear cracks 9 Wide flexural cracking and spalling should be concentrated in the
9 Shear crack width >1.6 mm plastic zone
10 Shear crack width <3.2 mm 10 Minor flexural cracking may extend beyond the plastic hinge zone
11 No shear crack width exceeds 3.2 mm 11 Flexural cracks beyond plastic hinge zone don’t exceed 1/8”
12 Shear crack width >3.2 mm & <9.6 mm 12 Flexural cracking and spalling concentrated at the hinge zone
13 Shear crack width <3.2 mm 13 Flexural cracking may extend beyond p. hinge zone
14 No shear crack width exceeds 3.2 mm
15 Shear crack width may exceed 1/8 but cannot exceed 9.6 mm
16 Shear crack width <3.2 mm
17 Crack width <9.6 mm
18 Crack width <3.2 mm Complementing patterns as well as differences between rein-
19 Crack width <1.6 mm forced concrete and masonry are identified. The emerging pattern
20 Crack width <1.6 mm
is that cracking is the main damage indicator. Out of 278 state-
21 Crack width does not exceed 9.6 mm
22 All crack widths <6.4 mm ments, 156 statements are about cracking. This corresponds to
23 Crack widths does not exceed 9.6 mm 56% of the total count. The remaining statements are distributed
24 No crack width exceeds 4.8 mm among the other damage indicators (spalling, crushing, buckling
and fracture of reinforcement, and residual displacement). Out of
the 5 Level 3 groups of reinforced concrete, 3 groups are about
cracking and crack properties (Table 1). The other two groups are
damage mode. Thus, the crack width is put at a lower level than the
about location and quantity of other damage indicators. For
crack type. When the affinity diagram is built this way we can
masonry components 4 out of 7 and for infilled frames 2 of 5 level
obtain the patterns in the statements, which help differentiating
3 groups are about cracking (Table 2). Similar to reinforced con-
between different damage modes, while keeping the details at
crete, damage location emerges as an important feature for dam-
the lower levels, which help determining the damage severities
age mode identification. Additionally, separation of masonry
for the damage modes.
from infills and partial or total walking type of damage are defined
There are other statements, which talk about cracking (Table 2).
for masonry.
These statements seem like they may fit into the hierarchy, which
The damage indicators of both masonry infilled frames and
were identified in the previous paragraph. However, considering
reinforced concrete components have largely the same overarching
the meaning of the statements and how these statements are used
patterns of information requirements at Level 3. The only differ-
to distinguish between different behavior modes, one realizes that
ence is the movement mode of infills, which constitutes about 7%
there are stronger affinities among the statements in Table 2. The
of the statements on masonry damage, and 4% of the overall
first three statements are used to distinguishing the initiation of
statements.
core-crushing. Statements 4–6 are used to locate damage for core
The main observations from the affinity diagram agree with the
crushing. The location of crushing helps differentiate between a
experimental studies on behavior of structural systems. Generally,
shear dominated mode or boundary crushing mode. The remaining
damage initiates as cracking [55,56]. Other damage indicators add
statements are important for distinguishing between flexure and
onto cracking as damage progresses [57]. Often existence of other
pre-emptive shear type of behavior. Similar to the previous exam-
damage indicators marks the transition of the damage severity to
ple, these other statements in Table 2 were studied to understand
an upper level (e.g., moderate to heavy) or the failure of compo-
the precedence of the information for identifying damage modes.
nent. Therefore, it can be said that cracking is always present
The next section presents and discusses the final division.
regardless of other damage indicators.
The grouping was repeated at the third and fourth levels to
The type and character of cracks depends on damage behavior
build a hierarchy. The statements for reinforced concrete walls,
of the component [5]. For example, in a ductile flexural behavior
unreinforced masonry walls, and infilled frames were considered
we primarily expect to see cracks in transverse direction to the
separately. Therefore, the study resulted in three affinity diagrams.
component axis (i.e., flexural cracks). A vertical crack along the
The next section will present the final diagrams and discuss the
edge of a concrete column may indicate lap-splice splitting.
findings.
The same cannot be said for all of the other damage indicators.
For example, residual displacement alone may not be enough to
5. Results and discussion gain insights into the cause of the damage behavior. Therefore,
cracks can be attributed as having special importance in damage
All 278 statements were used to create a structure for the iden- detection. The patterns in the affinity diagram agree with these
tification of damage modes (Tables 3–5). Reinforced concrete and insights.
masonry have different mechanisms for failure and, hence, differ- The affinity diagrams show that there are several important
ent damage modes. Therefore, the statements for reinforced con- properties of cracks both for concrete and masonry components.
crete, masonry, and additional modes for infilled frames were First and foremost properties are the type and width. Crack type
considered separately. Due to the small number of statements for depends on the alignment of crack with the axis of component
the additional modes of infilled frames (22 statements), they are and loading direction. Possible crack types are flexural, shear,
only considered in three levels versus four levels in reinforced con- and vertical cracks. Width is generally not constant along cracks.
crete and masonry (Table 5). Therefore, crack width is considered as a point property of cracks.
 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64 59

Table 3
The affinity diagram for reinforced concrete components. Level 1 contains the individual statements extracted from the FEMA 306 guideline and is hidden for brevity. Cracking
and crack properties emerge as the most important feature in the diagram, followed by the location and quantity of the other damage indicators.

Concrete Statement Count Percent

Level 4 Level 3 Level 2 83 (Total)
Crack properties 47 57% of total
Crack point property 25 53% of Level 4
Flexural crack size 5
Shear crack size 12
Any/all crack size 8
Crack path property 10 21% of Level 4
Continuity of flexural cracks along component 4
Degradation of concrete along cracks 2
Amount lateral offset along sliding plane 4
Crack pattern property 12 26% of Level 4
Type of concentrated single crack 7
Concentration of flexural cracks in the plastic hinge zone 5
Properties of other damage indicators 36 43% of total
Damage location 15 42% of Level 4
Spalling or crushing location and amount 7
Location of buckled rebars 3
Spalling or vertical cracking location 5
Damage quantity 21 58% of Level 4
Spalling or vertical cracking 6
Residual displacement amount 7
Buckled or fractured bars 8

Table 4
Cracking is the main damage indicator in masonry, similar to reinforced concrete. The affinity diagram shows other features in addition to size and type of cracks, such as opening
of head joints and sliding of bed joints.

Masonry Statement Count Percent

Level 4 Level 3 Level 2 173 (Total)
Crack properties 96 56% of total
Crack point property 23 24% of Level 4
Size of cracks 5
Size of opening of head joints 5
Sliding/opening of bed joints 13
Crack path property 44 46% of Level 4
Spalling and unit damage along cracks 9
Cracks go through units 9
Sliding off supporting bricks along crack 3
Amount of offset along cracks 13
Crack pattern property 25 26% of Level 4
Continuity of horizontal cracks 2
Continuity of diagonal cracks 2
Patterns of cracks 10
Location of concentrated cracks 3
Type of concentrated cracks 2
Type and number of cracks in a region 4
Amount of cracked courses in units 6
Properties of other damage indicators 77 44% of total
Damage location 60 78% of Level 4
Location of brick crushing 12
Location of mortar crushing 3
Location of cracking of bed joints and spalling 13
Location of cracks with various orientations 20
Location of regional crushing 5
Location of falling masonry 3
Location of spalling of face shells 2
Mortar separation around perimeter 2
Crushing of mortar and bricks around perimeter 3
Location of movement of face shells 2
Movement modes of components 12 15% of Level 4
Type and existence of total walking of piers 4
Existence of total movement of spandrels 5
Existence of partial out-of-plane movement of walls 1
Existence of partial rotation of component 2

Maximum crack width is measured and recorded for each The affinity diagramming study also discovered patterns along the
crack [34]. Similarly, for masonry components, opening of head crack paths, such as degradation and sliding along cracks, and con-
joints, and sliding of bed joints are considered as point properties. tinuity of cracking.
60 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64

Table 5 Table 6
Affinity diagram for additional modes of infilled frames. Due to the low number of The damage modes and severities for every damage mode was implemented
statements, only three levels were considered. according to the descriptions in FEMA 306. Table shows the rules as given in FEMA
306 and the corresponding damage parameters for the Insignificant severity of the
Infilled Frames Statement Count Percentage ductile flexure mode (A). In order to avoid false positives, damage parameters, which
Level 4 Level 3 Level 2 22 (Total) do not belong to the insignificant A damage, have been set accordingly.

Crack properties 13 59% of total Insignificant A (FEMA Corresponding Additional rules for
Crack location 7 306) implementation insignificant A
Size and extent of X-cracking 6
No shear width maxFlex 6 4.8 mm isBoundaryRebarDamage = false
Properties of other damage indicators 9 41% of total exceeds 4.8 mm
Spalling extent and location 6 No shear crack exceed maxShear 6 3.2 mm isRebar = false
Fractured rebar location 1 3.2 mm
Crushing location 2 No significant spalling maxVert = 0 displacementSignificance < 2
or vertical cracking
isSpalled = FALSE isShearConcentraated = false
isFlexuralConcentrated = false
Patterns of cracks and properties of crack patterns can be isBoundaryCrushed = false
observed as a composite property of cracks (Tables 4 and 5). Con-
centration and number of cracks, location, type, and amount of
cracked courses in units are the main properties of crack patterns.
Especially, in masonry, the features, which define the patterns are
much richer compared to concrete. This can be attributed to the has other damage indicators that do not belong to ductile flexure
construction of masonry components using two different materials (e.g., concentrated shear crack).
(i.e., masonry units, and mortar), which have different material In order to avoid false positives, the damage modes were imple-
properties. The relative strength of units and mortar, for instance, mented in a way that considers the exclusion of the damage indi-
determines whether the units will crack along the crack paths cre- cators that do not belong to the damage modes (Table 6). For
ating smoother crack patterns similar to concrete, or whether the example, for the ductile flexural damage mode (A), all of the dam-
cracks will be stair-shaped with intact units. age severities exclude concentrated shear cracks, since such shear
cracking cannot exist for a ductile flexural mode.
The accuracy of the rule-based system was verified using an
6. Validation actual damaged building. The example building is a T-shaped rein-
forced concrete two-story building [58]. The building was designed
The affinity diagramming study identified the hierarchy of and constructed in the late 1950s. The floors and the roof are con-
information items, which visually characterize the damage modes structed with waffle slabs. Reinforced concrete walls in both direc-
for reinforced concrete and masonry structural walls. However, the tions resist lateral forces. Several walls have door openings in the
affinity diagrams do not provide insights into the usefulness or middle of the wall. The building was damaged during an earth-
value of the information items. In order to validate the information quake and it was professionally inspected and seismically evalu-
identification study, a sensitivity analysis was performed. For the ated at that time. The building contains damage indicators which
analysis a rule-based system was implemented using the results can be seen in ductile flexure (A), pre-emptive shear (H), flexure/-
of the affinity diagramming study and the FEMA 306 damage diagonal tension (B), and pier-wall rocking (N & M) mechanisms.
classifications. Additionally, the building contains flanged and non-flanged solid
The validation focused on reinforced concrete components walls, piers, and spandrels, which are the main component types
(piers and spandrels). Inside each Level 2 group for reinforced con- according to FEMA 306. Eighteen components were divided into
crete walls, there are 18 information items that could be used to 42 piers, spandrels, and solid walls and analyzed. In this case study,
distinguish the damage modes and severities (Table 7). The combi- we use the documentation that was generated at the time of
nations of these information items and the values they take deter- inspection and evaluation.
mine the damage mode and the severity of a component. A rule- The 18 damage parameters for each of the 42 wall components
based system was implemented, which takes the 18 parameters of the building composed of reinforced concrete piers and span-
as an input and outputs the damage severities for all possible dam- drels were generated semi-automatically from hand-sketches.
age modes. The rule-based system follows the damage descriptions Twelve of the forty two components were coupling beams and
for reinforced concrete walls in the Chapter 5 of the FEMA 306. failed in pre-emptive shear with severities ranging from Insignifi-
Every damage mode is implemented according to the descriptions cant to Heavy [58]. The rest of the thirty components have insignif-
in the guideline. The damage modes and damage severities are icant damage as a result of overturning. Per the analogy explained
considered independently. Hence, given 18 parameters, which in FEMA 307, the prototype used the ductile flexural mode’s
describe the damage condition of a component, the rule engine description to assess the severity of the overturning behavior visu-
returns an answer for each damage severity of each damage mode. ally. The prototype implementation correctly identified the dam-
In order to better filter the damage indicators, the damage clas- age modes and severities of all of the forty-two components
sifications were combined to consider the damage parameters correctly, which verified the accuracy of the implementation.
which are not a part of each damage mode. The damage descrip- The identified information items for reinforced concrete were
tions in the FEMA 306 result in false positives, if they are imple- validated using a sensitivity analysis. The sensitivity analysis mea-
mented as they are given. For example, at insignificant severity sured the value of information as the ratio of error in the damage
of the ductile flexural damage mode, shear and flexural crack width assessment to error in the value of each parameter. Base configura-
limits are given along with the fact that there should be ‘‘no signif- tions for insignificant, slight, moderate, heavy, and extreme cases
icant residual displacement”. When ductile flexure damage mode were generated for every damage mode based on FEMA 306
(A) is implemented by defining only these three cases, every case (Table 9). The ranges of values of the damage parameters were
that contains these three conditions (shear crack width is smaller determined based on the values in FEMA 306. The maximum crack
than the limit; flexural crack width is smaller than the limit; and sizes for the baseline configurations were selected arbitrarily
no residual displacement), the rule set returns true, even if the case within FEMA’s range of values.
 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64 61

Table 7
Eighteen parameters have been identified based on the affinity diagrams and the damage descriptions, that help determine the damage mode and severity.

Affinity diagram Description Parameter name Range of error

group
Crack point property
Maximum flexural crack width maxFlex 2.0 to +2.0 mm with 0.5 mm increments
Maximum vertical crack width maxShear 2.0 to +2.0 mm with 0.5 mm increments
Maximum flexural crack width maxVert 2.0 to +2.0 mm with 0.5 mm increments
Crack path property
Existence of a sliding crack and the severity of offset on the sliding plane slidingSignificance 0: no sliding crack
1: slight offset along the sliding crack
2: significant offset along the sliding crack
Crack pattern property
Existence of a concentrated shear crack isShearConcentrated T/F
Existence of a concentrated flexural crack isFlexuralConcentrated T/F
Damage location
Maximum shear crack width in the toe region maxShearInToe 2.0 to +2.0 mm with 0.5 mm increments
Maximum shear crack width in the web region maxShearInWeb 2.0 to +2.0 mm with 0.5 mm increments
Existence of vertical cracks in the toe isToeVerticallyCracked T/F
Existence of spalling in the toe region isToeSpalled T/F
Existence of spalling in the web region isWebSpalled T/F
Severity of spalling in the web region spallingWebSeverity 0: no spalling in web
1: slight spalling in web
2: significant spalling in web
Severity of the spalling in the toe region spallingToeSeverity 0: no spalling in toe
1: slight spalling in toe
2: significant spalling in toe
Existence of rebar damage in the boundary region isBoundaryRebarDamaged T/F
Existence of boundary crushing isBoundaryCrushed T/F
Damage quantity T/F
Existence of spalling isSpalled T/F
Existence of rebar damage isRebarDamaged T/F
Severity of residual displacement displacementSignificance 0: no residual displacement
1: slight residual displacement
2: significant residual displacement

Table 8 Table 9
The assessment results for the base configuration of insignificant ductile flexural Base configurations used in the simulations for ductile flexure (A).
damage (RC1A). The base configuration is checked against all damage modes and
severities. It can be seen that the base configuration also conforms insignificant D and Damage parameter A (Ductile Flexural)
E type of damage. Insignificant Slight Moderate Extreme

Damage mode/severity Assessment Damage mode/severity Assessment maxShear 3.0 3.0 4.0 4.0
maxFlex 4.0 6.0 6.0 8.0
A (ductile flexural) & M/N (pier/wall D (flexure/sliding shear)
maxVert 0.0 0.0 1.0 2.0
rocking)
slidingSignificance 0.0 0.0 0.0 0.0
Insignificant True Insignificant True
isShearConcentrated FALSE FALSE FALSE FALSE
Slight False Slight False
isFlexuralConcentrated FALSE FALSE FALSE TRUE
Moderate False Moderate NotUsed
maxShearinToe 0.0 0.0 0.0 0.0
Heavy NotUsed Heavy False
maxShearinWeb 0.0 3.0 4.0 4.0
Extreme False Extreme False
isToeVerticallyCracked FALSE FALSE TRUE TRUE
B (flexure/diagonal tension) E (flexure/boundary-zone isToeSpalled FALSE FALSE TRUE TRUE
compression) isWebSpalled FALSE FALSE FALSE TRUE
Insignificant False Insignificant True spallingWebSeverity 0.0 0.0 0.0 1.0
Slight NotUsed Slight False spallingToeSeverity 0.0 0.0 1.0 2.0
Moderate False Moderate False isBoundaryRebarDamaged FALSE FALSE FALSE TRUE
Heavy False Heavy False isBoundaryCrushed FALSE FALSE FALSE FALSE
Extreme False Extreme False isSpalled FALSE FALSE TRUE TRUE
isRebarDamaged FALSE FALSE FALSE TRUE
C (flexure/diagonal compression) H (preemptive shear)
displacementSignificance 0.0 0.0 1.0 2.0
Insignificant False Insignificant False
Slight NotUsed Slight NotUsed
Moderate False Moderate False
Heavy False Heavy False
Extreme False Extreme False spallingWebSeverity), the error is represented with an integer value
within the 2, +2 range. When changing the value of a parameter,
other affected parameters are also changed. For example, setting
the ‘‘existence of spalling in the toe region” to TRUE, also sets the
In total, 25 base configurations were created. In each run, a ‘‘existence of spalling” to TRUE. The opposite is not true, since
known amount of error was introduced to one of the 18 parame- spalling can occur at other regions other than the toe region. The
ters. For crack widths, the error is measured in millimeter 18 parameters are varied one at a time for every baseline configura-
from the baseline value. For binary values (e.g., isSpalled), tion. The variations were fed into the verified rule-based system.
introducing an error converts True to False and vice versa. For The value of the information items were calculated for every
the severity type of parameters (e.g., displacementSignificance, baseline configuration and averaged over all damage modes.
62 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64

Table 10 severity assessment for the error in parameter values. For example,
For every variation in the configuration, the change in the results are recorded in a in an insignificant ductile flexure damage, where the actual flexu-
results vector. Zero indicates no change in severity from the base configuration as a
result of the variation.
ral crack width is 4.0 mm and there is an error of 1.0 mm, then,
all other parameters kept constant, it may indicate a moderate
Damage mode maxFlex Result pre-emptive shear behavior and no change in the other damage
Base conf. Test case (error) modes (Table 10).
4.0 mm 1.0 mm The sensitivity of the damage assessment to the change in the
A Insignificant Insignificant 0 value of a parameter is calculated as the error in the parameter value
B No Damage No Damage 0 divided by the magnitude of the severity change. The change in dam-
C No Damage No Damage 0
age severity was quantified in a 0–5 scale, where 0 represents no
D Insignificant Insignificant 0
E Insignificant Insignificant 0 change in the given mode and 5 represents a jump from none to
H No Damage Moderate j1j
¼ 0:33 extreme severity or from extreme to none. In the example in
j3j
Table 10, the severity for H (pre-emptive shear) changed from no-
damage to moderate, jumping over insignificant and slight severi-
For every configuration, a results vector was computed, which ties, which corresponds to a change in severity of magnitude 3.
indicates the change in the results from the base configuration The values in all 25 variations are then averaged and aggregated
(Tables 8 and 10). The vector quantifies the error in the damage into a single table (Table 11). Each column in the table represents a

Table 11
Sensitivity of the damage modes to errors in the damage parameters. Higher the number, higher the sensitivity of the damage mode to misclassification of the damage parameter.
The analysis shows that maximum vertical crack width has no effect on the classification of the damage and severity. The other parameters have varying degree of impact on the
classification.

Average Change in Severity Assessment

Parameter Size of Error A B C D E H
-2.0 0.00 0.48 0.06 0.00 0.00 0.04
-1.5 0.00 0.64 0.08 0.00 0.00 0.05
-1.0 0.00 0.24 0.00 0.00 0.00 0.08
maxShear -0.5 0.00 0.00 0.00 0.00 0.00 0.00
0.5 0.24 1.36 0.64 0.24 0.24 0.16
+1.0 0.36 0.72 0.36 0.36 0.36 0.08
+1.5 0.24 0.48 0.24 0.24 0.24 0.05
+2.0 0.22 0.46 0.20 0.22 0.22 0.12
-2.0 0.30 0.30 0.00 0.06 0.30 0.30
-1.5 0.48 0.40 0.00 0.08 0.48 0.40
-1.0 0.12 0.00 0.00 0.00 0.12 0.60
maxFlex -0.5 0.24 0.00 0.00 0.00 0.24 0.00
0.5 1.44 1.20 0.48 0.48 1.44 0.32
+1.0 0.92 1.04 0.32 0.44 0.92 0.16
+1.5 0.61 0.69 0.21 0.29 0.61 0.11
+2.0 0.50 0.56 0.20 0.26 0.50 0.08
maxVert -2.0 - +2.0 0.00 0.00 0.00 0.00 0.00 0.00
isSpalled 1 0.52 0.16 0.16 0.68 0.52 0.12
isToeVerticallyCracked 1 0.40 0.16 0.16 0.52 0.40 0.12
-2 0.00 0.26 0.22 0.08 0.00 0.00
spallingWebSeverity -1 0.00 0.52 0.44 0.16 0.00 0.00
1 0.52 0.68 0.32 0.52 0.52 0.12
2 0.26 0.34 1.20 0.26 0.26 0.06
-2 0.04 0.00 0.00 0.08 0.04 0.00
spallingToeSeverity -1 0.08 0.00 0.00 0.16 0.08 0.00
1 0.76 0.16 0.16 0.52 0.76 0.12
2 0.38 0.08 0.08 0.26 0.38 0.06
-2 0.00 0.00 0.10 0.10 0.00 0.00
slidingSignificance -1 0.00 0.00 0.00 0.20 0.00 0.00
1 1.40 1.28 0.60 1.88 1.56 0.68
2 0.70 0.64 0.30 1.56 0.78 0.34
isToeSpalled 1 0.64 0.16 0.16 0.56 0.64 0.12
isBoundaryRebarDamaged 1 1.40 1.48 0.60 0.52 1.56 0.88
isBoundaryCrushed 1 1.40 1.28 0.60 0.72 1.56 0.68
isRebarDamaged 1 1.40 1.04 0.80 0.68 1.56 0.44
-2 0.20 0.00 0.00 0.00 0.20 0.00
displacementSignificance -1 0.40 0.00 0.00 0.00 0.40 0.00
1 0.48 0.40 0.16 0.00 0.64 0.16
2 0.50 0.54 0.16 0.26 0.58 0.24
maxShearinToe -2 0.00 0.00 0.00 0.08 0.00 0.00
-1.5 - +2.0 0.00 0.00 0.00 0.00 0.00 0.00
-2.0 0.00 0.18 0.00 0.00 0.00 0.00
maxShearinWeb -1.5 0.00 0.24 0.00 0.00 0.00 0.00
-1.0 - +2.0 0.00 0.00 0.00 0.00 0.00 0.00
isWebSpalled 1 0.52 0.68 1.28 0.52 0.52 0.12
isShearConcentrated 1 1.40 1.48 0.60 0.72 1.56 2.96
isFlexuralConcentrated 1 1.40 1.28 0.60 0.68 1.40 0.68
 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64 63

damage mode, such that the sensitivity of the damage assessment of the sensitivity analysis for each parameter across all damage
to perturbations in the values of each parameter can be observed modes we can obtain the mean effect of each parameter on the
for each mode individually. The rows in the table correspond to damage assessment (Table 12). The mean values clearly show the
the errors in the parameter values. Thus, each cell represents the relative importance of the parameters. Considering the time con-
sensitivity of the severity assessment for a damage mode for given straints in the field for data collection, the results can potentially
amount of error in the parameter value (e.g., maximum shear crack be used to drive the data collection towards on more impacting
size). Zero means that the assessment is insensitive to errors. Lar- parameters to make efficient use of resources.
ger values represent higher impact to the assessment.
The results show that some of the parameters had minimal or
no effect on the classification of the damage severity. Maximum 7. Conclusions
vertical crack width (maxVert) has no effect on the classification
of damage. Maximum width of the shear cracks in the toe region This paper identifies the information requirements for deter-
had minimal effect if the width of the cracks is underestimated mining the damage modes and severities of earthquake damaged
by 2 mm. Maximum width of the shear cracks in the web regions reinforced concrete walls and coupling beams. The research
also had small effect. However small, maxShearInToe and maxS- approach uses the affinity-diagramming method for identifying
hearInWeb has value for determining the damage severities. On an abstraction and hierarchy of the information items, which are
the other hand, since maxVert has no value for the damage classi- required for developing a representation. The study is validate by
fication, it can be eliminated. performing a sensitivity analysis on the damage modes and sever-
The damage modes have varying sensitivities for the damage ities. The following conclusions can be drawn from the study:
parameters. For example, flexure/diagonal tension (B) mode is For both reinforced concrete and masonry walls, cracking is the
much more sensitive to errors in the maximum width of shear most important information item for determining the damage
cracks compared to pre-emptive shear (H) mode. Flexure/diagonal mode and severity. Crack width, type, and location are important
tension develops shear cracks more steadily compared to the pre- parameters associated with cracks. For other damage indicators,
emptive mode. Shear cracks can open up to about 0.8 mm for such as spalling, crushing, and rebar damage, the important
insignificant severity and up to 1.6 and 3.2 mm for moderate and parameters are location and severity. Location of damage is defined
heavy severities, correspondingly. On the other hand, the pre- based on the region on the components (e.g., hinge, web, and toe).
emptive shear mode develops a concentrated shear crack suddenly, The information items can be grouped under five broader cate-
as the heavy damage severity is reached. Thus, higher sensitivity of gories, three of which belong to crack properties and the other two
the H mode to isShearConcentrated can be observed in Table 11. describe properties of the other damage indicators, such as spal-
The validation study showed that the damage state of a rein- ling, crushing, residual displacement, and rebar damage. The same
forced concrete wall can be represented using 17 damage parame- hierarchy also applier to masonry wall components. Thus, it can be
ters. The damage parameters include the widths of shear and said that although the appearance and the specific parameters for
flexural cracks, locations and/or severities of spalling and crushing, defining damage modes and severities for masonry is different
rebar damage and residual displacements. By averaging the results from reinforced concrete, similar patterns of information can be
expected.
For reinforced concrete wall components (piers and spandrels),
the damaged condition can be represented using 17 parameters.
Table 12 These parameters include, the maximum widths of shear and flex-
By averaging the change of damage assessment across all of the damage modes, the ural cracks, existence and/or severity of spalling, crushing, and
mean effect of each damage parameter on the assessment can be obtained. rebar damage, severities of residual displacements, sliding on a lat-
eral plane, and concentration of shear and flexural cracks.
Average change across all damage The damage modes of reinforced concrete wall components
Parameter modes
have varying degrees of sensitivity to errors in the 17 damage
maxShear 0.20 parameters. The assessment is least sensitive to the width of shear
maxFlex 0.38 cracks on the web region. Most effective parameters are the con-
maxVert 0.00 centration of cracks, particularly shear cracks, and damages in
the boundary region, such as rebar damage and crushing.
isSpalled 0.36
The hierarchies and the damage parameters, which were iden-
isToeVerticallyCracked 0.29
tified in this study, elucidate three aspects of the visual damage
spallingWebSeverity 0.28 assessment problem. First, the results of the study describe the
spallingToeSeverity 0.18 information content of a potential information model. The objects
slidingSignificance 0.51 and the relationship between the objects, which form the informa-
isToeSpalled 0.38 tion model, can be derived using the affinity diagrams. Second, the
types and details of the queries needed to extract the necessary
isBoundaryRebarDamaged 1.07
information for performing the damage assessment tasks can be
isBoundaryCrushed 1.04 developed using the findings of the study. Third, the collection of
isRebarDamaged 0.99 damage information can be prioritized to focus on the most
displacementSignificance 0.22 impacting parameters in the field to efficiently use time.
maxShearInToe 0.01 Several research directions are suggested. First, developing a
BIM based representation of damage for structural walls and inte-
maxShearInWeb 0.02
grating the visual damage assessment with strength analysis by
isWebSpalled 0.61 building on the damaged representation should be studies. Second,
isShearConcentrated 1.45 the extensibility and generality of the findings of this paper to
other structural systems, such as reinforced concrete frames,
isFlexuralConcentrated 1.01
should be studies. Third, the findings of this study can potentially
be combined with rapid assessment methods.
64 E.B. Anil et al. / Advanced Engineering Informatics 30 (2016) 54–64

References [28] Z. Zhu, I. Brilakis, G. Parra-Montesinos, Real time concrete damage visual
assessment for first responders, Presented at the Proceedings of the ASCE
Construction Research Congress, Seattle, WA, 2009.
[1] S. Bernstein, How Risky Are Older Concrete Buildings?, in: Los Angeles Times,
[29] Y. Tsai, C. Jiang, Y. Huang, Multiscale crack fundamental element model for
2005.
real-world pavement crack classification, J. Comput. Civ. Eng. (2014), http://
[2] E. Venere, Study identifies quake-prone concrete buildings in Los Angeles area,
dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000271, 04014012.
WEST LAFAYETTE, Ind.: Network for Earthquake Engineering Simulation
[30] A.B.D.H.N. Koutsopoulos, Primitive-based classification of pavement cracking
(NEES), 2014.
images, J. Transp. Eng. 119 (1993) 402–418.
[3] E. Taciroglu, P. Khalili-Tehrani, M7.8 Southern San Andreas Fault Earthquake
[31] S.-N. Yu, J.-H. Jang, C.-S. Han, Auto inspection system using a mobile robot for
Scenario: Non-ductile Reinforced Concrete Building Stock, United States
detecting concrete cracks in a tulle, Automat. Constr. 16 (2007) 255–261.
Geological Survey (USGS) and California Geological Survey (CGS), 2008.
[32] E.B. Anil, B. Akinci, J.H. Garrett, O. Kurc, Representation of damage information
[4] S. Kato, Y. Sato, 2011 Japan: Economic & Social Impacts of the Earthquake – Update.
for post-earthquake damage assessment of reinforced concrete frames,
Eco Focus (Credit Agricole S.A). <http://economic-research.credit-agricole.com/
Presented at the ASCE International Workshop on Computing in Civil
site/japan-economic-and-social-impacts-of-the-earthquake-update-1319.html>.
Engineering, Los Angeles, 2013.
[5] FEMA, FEMA 306 – Evaluation of Earthquake Damaged Concrete and Masonry
[33] K.T. Chang, E.H. Wang, Developing procedures for post-earthquake structural
Wall Buildings: Basic Procedures Manual, FEMA, Washington, DC, 1998.
evaluation by laser scanning techniques, Insight 54 (2012) 562–567.
[6] B.J.P. Graybeal, B.M. Phares, D.D. Rolander, M. Moore, Visual inspection of
[34] FEMA, FEMA 306 – Evaluation of Earthquake Damaged Concrete and Masonry
highway bridges, J. Nondestruct. Testing 21 (2002) 67–83.
Wall Buildings: Basic Procedures Manual, FEMA, Washington, DC, 1998.
[7] B.M. Phares, D.D. Rolander, B.A. Graybeal, G. Washer, Reliability of visual
[35] N.Z.S.f.E. Engineering, Building Safety Evaluation During a State of Emergency
bridge inspections, Public Roads 64 (2001).
– Guidelines for Territorial Authorities, Department of Building and Housing,
[8] C.L. Menches, A.B. Markman, R.J. Jones, Innovative Method for Investigating the
New Zealand, 2009.
Facility Damage Assessment Process, Denver, Colorado, 2008.
[36] A. Yasaka, H. Kataoka, K. Kasima, M. Takeda, M. Usami, N. Yabuki, et al., The
[9] Z. Zhu, S. German, I. Brilakis, Visual retrieval of concrete crack properties for
development of the reinforced-concrete structural model on IFC specification,
automated post-earthquake structural safety evaluation, Automat. Constr. 20
Presented at the Joint International Conference on Computing and Decision
(2011) 874–883.
Making in Civil and Building Engineering, Montreal, Canada, 2006.
[10] S. German, J.-S. Jeon, Z. Zhu, C. Bearman, I. Brilakis, R. DesRoches, et al.,
[37] M.K. Weise, Peter Liebich, Thomas Sherer, J. Raimar, Structural analysis
Machine vision enhanced post-earthquake inspection, J. Comput. Civil Eng. 27
extension of the IFC modeling framework, ITCon 8 (2003).
(2013).
[38] E.B. Anil, G. Unal, O. Kurc, Information requirements for design and detailing of
[11] V.R. Kamat, S. El-Tawil, Evaluation of augmented reality for rapid assessment
reinforced concrete frames in multiuser environments, J. Comput. Civil Eng. 26
of earthquake-induced building damage, J. Comput. Civil Eng. 21 (2007) 303–
(2012) 465–477.
310.
[39] BuildingSmart, Industry Foundation Classes (IFC), 2013.
[12] R. Zeibak-Shini, R. Sacks, S. Filin, Toward generation of a building information
[40] R. Barak, Y.S. Jeong, R. Sacks, C.M. Eastman, Unique requirements of building
model of a deformed structure using laser scanning technology, Presented at
information modeling for cast-in-place reinforced concrete, J. Comput. Civil
the ICCCBE, Russia, 2012.
Eng. 23 (2009) 64–74.
[13] M.E. Venugopal, M. Charles, R. Sacks, J. Teizer, Semantics of model views for
[41] E.B. Anil, P. Tang, B. Akinci, D. Huber, Deviation analysis method for the
information exchanges using the industry foundation class schema, Adv. Eng.
assessment of the quality of the as-is Building Information Models generated
Inform. 26 (2012) 412–428.
from point cloud data, Automat. Constr. 35 (2013) 507–516.
[14] Q.C. Zou, Yu Li, Qingquan Mao, Song Qingzhou Wang, CrackTree: automatic
[42] E.B. Anil, B. Akinci, O. Kurc, J.H. Garrett, Building information modeling based
crack detection from pavement images, Pattern Recognition Images 33 (2012)
earthquake damage assessment for reinforced concrete walls, J. Comput. Civil
227–238.
Eng. (2015). vol. 10.1061/(ASCE)CP.1943-5487.0000551.
[15] N.T.A. Sy, M. Avila, S. Begot, J.C. Bardet, Detection of defects in road surface by a
[43] J. Rumbaugh, Object-Oriented Modeling and Design, Prentice Hall, 1991.
vision system, Presented at the 14th IEEE Mediterranean Electrotechnical
[44] H. Beyer, K. Holzblatt, Contextual Design: Defining Customer Centered
Conference, France, 2008.
Systems, first ed., Morgan Kaufmann, 1997.
[16] M. Torok, M. Golparvar-Fard, K. Kochersberger, Image-based automated 3D
[45] ATC-20, Procedures for postearthquake Safety Evaluations of Buildings,
crack detection for post-disaster building assessment, J. Comput. Civ.
Applied Technology Council (ATC), Redwood City, CA, 1989.
Eng. (2014), http://dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000334,
[46] ATC-21, Rapid Visual Screening of Buildings for Potential Seismic Hazards: A
A4014004.
Handbook, Applied Technology Council (ATC), Redwood City, CA, 2002.
[17] E.B. Anil, B. Akinci, J.H. Garrett, O. Kurc, Characterization of laser scanners for
[47] ASCE, Seismic rehabilitation of existing buildings, American Society of Civil
detecting cracks for post-earthquake damage inspection, Presented at the
Engineers, Reston, Virginia, 2006.
ISARC 2013, Montreal, Canada, 2013.
[48] E.B. Anil, B. Akinci, D. Huber, Representation requirements of as-is building
[18] I. Abdel-Qader, O. Abudayyeh, M.E. Kelly, Analysis of edge-detection
information models generated from laser scanned point cloud data, Presented
techniques for crack identification in bridges, J. Comput. Civil Eng. 17 (2003)
at the ISARC 2011, Seoul, S. Korea, 2011.
255–263.
[49] O. Melchor-Lucero, C. Ferregut, Earthquake damage assessment of reinforced
[19] L.-C. Chen, Y.-C. Shao, H.-H. Jan, C.-W. Huang, Y.-M. Tien, Measuring system for
concrete members using an expert system, Presented at the Eleventh World
cracks in concrete using multitemporal images, J. Surv Civil Eng 132 (2006)
Conference on Earthquake Engineering, Acapulco, Mexico, 1996.
77–82.
[50] J. Mackerle, Object-oriented programming in FEM and BEM: a bibliography
[20] Z.Q. Chen, T.C. Hutchinson, Image-based framework for concrete surface crack
(1990–2003), Adv. Eng. Softw. 35 (2004) 325–336.
monitoring, Adv. Civil Eng. (2010) 18.
[51] W. Celes, G.H. Paulino, R. Espinha, A compact adjacency-based topological data
[21] D. Lecompte, J. Vantomme, H. Sol, Crack detection in a concrete beam using
structure for finite element mesh representation, Int. J. Numer. Meth. Eng. 64
two different camera techniques, Struct. Health Monit. 5 (2006) 59–68.
(2005) 1529–1556.
[22] T. Yamaguchi, S. Nakamura, R. Saegusa, S. Hashimoto, Image-based crack
[52] B.W.R. Forde, R.O. Foschi, S.F. Stiemer, Object-oriented finite element analysis,
detection for real concrete surfaces, Trans. Electr. Electron. Eng. IEEJ 3 (2008)
Comput. Struct. 34 (1990) 355–374.
128–135.
[53] Z. Turk, T. Isakovic, M. Fischinger, Object oriented modeling of design system
[23] D.F. Laefer, M. Fitzgerald, M. Eoghan, C. David, D. Lennon, S.W. Morrish, Lateral
for RC buildings, J. Comput. Civil Eng. 4 (1994) 436–453.
image degradation in terrestrial laser scanning, Struct. Eng. Int. 19 (2009) 184–
[54] H.H. Beyer, Karen, Contextual Design: Defining Customer-Centered Systems,
189.
Morgan Kaufmann Publishers, San Francisco, CA, 1998.
[24] D. MacKinnon, J.A. Beraldin, L. Cournoyer, B. Francois, Evaluating laser range
[55] A.S. Mehrabi, Benson Schuller, Michael Noland, L. James, Experimental
scanner lateral resolution in 3D metrology (2009).
evaluation of masonry-infilled RC frames, J. Struct. Eng. 122 (1996) 228–237.
[25] M.J. Olsen, F. Kuester, B.J. Chang, T.C. Hutchinson, Terrestrial laser scanning-
[56] K.E. Pilakoutas, Amr, Cyclic behavior of reinforced concrete cantilever walls –
based structural damage assessment, J. Comput. Civil Eng. 24 (2009) 264–272.
part I - experimental results, ACI Struct. J. 92 (1995) 271–281.
[26] D.F. Laefer, J. Gannon, E. Deely, Reliability of crack detection methods for
[57] R. Perera, Performance evaluation of masonry-infilled RC frames using loading
baseline condition assessments, J. Infrastructure Syst. 16 (2010) 129.
based on damage mechanics, Eng. Struct. 27 (2005) 1278–1288.
[27] P. Tang, D. Huber, B. Akinci, Characterization of laser scanners and algorithms
[58] FEMA, FEMA 307 – Evaluation of Earthquake Damaged Concrete and Masonry
for detecting flatness defects on concrete surfaces, J. Comput. Civil Eng. 25
Wall Buildings: Technical Resources, FEMA, Washington, DC, 1998.
(2011) 31.