Experimental Mechanics (2013) 53:1607–1619

DOI 10.1007/s11340-013-9769-7

Damage Assessment of Reinforced Concrete Structures Using Fractal

Analysis of Residual Crack Patterns
A. Farhidzadeh & E. Dehghan-Niri & A. Moustafa &
S. Salamone & A. Whittaker

Received: 19 October 2012 / Accepted: 27 May 2013 / Published online: 15 June 2013
# Society for Experimental Mechanics 2013

Abstract Currently, assessing the performance and safety of Abbreviations

reinforced concrete structures relies on routine-based visual in- DG Damage grade
spection (VI). Cracks width measurements are commonly used FD Fractal dimension
as a convenient indicator of damage; however other factors, such DI Damage index
as distribution and pattern of the cracks should be considered HD High definition
equally important in measuring the extent of damage present in IAEA International atomic energy agency
the structure. As a result, condition assessed by VI is subjective in LS Load step
nature and depends on the experience, knowledge, expertise, and NDE Non-destructive evaluation
judgment of the inspector carrying out the assessment. A new RC Reinforced concrete
approach based on the fractal analysis of residual crack patterns is RT Transition box size
proposed in this paper to assess the structural integrity of RO Object box size
reinforced concrete elements. A new damage index is presented RS Structure box size
to quantitatively perform a damage classification. The method- RD Discretization size
ology is validated through experimental studies on two large- RCSW Reinforced concrete shear wall
scale reinforced concrete shear walls subjected to a displacement RSL Relative stiffness loss
controlled reversed cyclic loading. Damage grades are also iden- SHM Structural health monitoring
tified based on width of cracks and proposed damage index (DI). SW Shear wall
The results demonstrate a more accurate estimation of damage VI Visual inspection
grades using DI. Furthermore, it is demonstrated that the DI can
estimate the relative stiffness loss of the specimens with accept-
able accuracy.
Introduction
Keywords Fractal dimension . Damage index . Visual
inspection . Cracks pattern quantification . Reinforced Civil structures such as buildings, bridges, nuclear power
concrete structures plants or dams include reinforced concrete (RC) structures
whose performance and functionality is essential. Unfortu-
A. Farhidzadeh : E. Dehghan-Niri : A. Moustafa : nately, many of these structures are facing an increasing
S. Salamone (*) number of challenges that can jeopardize their safety and
Smart Structures Research Laboratory, Department of Civil, serviceability. Some of these challenges, such as aging and
Structural, and Environmental Engineering, University at Buffalo,
maintenance, are life-cycle related and expected. Other chal-
The State University of New York, 212 Ketter Hall, Buffalo
NY 14260, USA lenges such as man-made hazards or natural events (i.e.,
e-mail: [email protected] earthquakes, hurricanes and tsunami) are not expected and
can weaken and destabilize a structure. Currently, assessing
A. Whittaker
the performance and safety of civil structures relies on routine-
Department of Civil, Structural, and Environmental Engineering,
University at Buffalo, The State University of New York, based visual inspection (VI); unfortunately, condition assessed
212 Ketter Hall, Buffalo, NY 14260, USA by VI is subjective in nature and depends on the experience,
1608 Exp Mech (2013) 53:1607–1619

knowledge, expertise, and judgment of the inspector carrying theories cannot accurately differentiate damage states. In this
out the assessment. As a result there is a potential that in- paper a different approach based on the fractal analysis of
dications of structural degradation could be missed [1, 2]. crack patterns is proposed to assess the structural integrity of
Notable examples of such unanticipated failures that have complex structures.
exposed shortcomings in condition assessment methodologies Fractal analysis is a new scientific paradigm that has been
can be found worldwide: 1) August 2009, the viaduct of the used successfully in many applications, including micro-
Belfast-Dublin railway line collapsed, although less than crack characterization in brittle materials [17], fracture of
a week before an engineer inspected the viaduct and rocks [18, 19], cracking in concrete material [14, 20–25]
found “no visible structural issues” [3]; 2) the Canter- and crack propagation in reinforced concrete structures [26,
bury television building in Christchurch, New Zealand, 27]. However, its application in the structural health moni-
collapsed during the magnitude 6.3 earthquake on Feb- toring (SHM) community has not been fully exploited. This
ruary 2011; the building was inspected by engineers after paper is organized as follows: A brief introduction on the
a magnitude 7.1 earthquake on September 2010 and after theory of fractal analysis is first presented and subsequently,
the 4.9 magnitude aftershock on December 2010 [4, 5]. a damage index developed on the basis of a fractal dimension
However, it collapsed after the 6.3 magnitude earthquake is proposed to elaborate, facilitate, and enhance interpreta-
on February 2011. On both occasions, the building was tion of damage states. Experimental results are also
declared safe, having suffered only superficial damage; presented with application to the damage assessment of
the 115 who died in that building alone accounted for two large-scale reinforced concrete shear walls (RCSWs)
nearly two-thirds of the 181 victims [5]. under lateral cyclic loading.
Various condition-rating scales have been developed
worldwide to overcome the qualitative nature of VI-based
condition assessment such as the International Atomic Ener- Fractal Analysis
gy Agency guidebook [1], National Bridge Inspection Stan-
dards [6], and The Federal Highway Administration code Background
[7]. These scales usually provide quantified condition in
terms of visible distress attributes such as cracks, surface The term “Fractal” was first introduced by Mandelbrot [28] to
pitting, spalling, delamination, wear or abrasion, concrete indicate objects whose complex geometry cannot be charac-
color change, etc. In particular, cracks are of large interest terized by an integer dimension. A classical example to illus-
in RC structures, since their properties, such as width, reflect trate this technique is the “length” of a coastline [16]. When
not only condition of concrete as material but also the con- measured at a given spatial scale d, the total length of a
dition of the entire system at structural level. Crack width is crooked coastline L(d) is estimated as a set of N straight line
commonly used as a convenient indicator of damage to RC segments of length d. Because small details of the coastline
elements, but it should be noted that recent standards and not recognized at lower spatial resolutions become apparent at
studies list other factors, such as distribution and pattern of higher spatial resolutions, the measured length L(d) increases
the cracks to be equally important in measuring the extent of as the scale of measurement d increases. Therefore, in fractal
damage present in the structure [8–10]. The pattern of crack- geometry the Euclidean concept of “length” becomes a pro-
ing at any point in time provides a window into the flow of cess rather than an event, and this process is controlled by a
forces at lower load levels, whether a member exhibits flex- constant parameter called fractal dimension (FD). The FD can
ure or shear dominated behavior [11, 12], the change in be a non-integer number varying depending on the complexity
directions of loading after cracking, the effectiveness of steel of an object. For example, the FD for a curve will lie between
reinforcement, stress in the reinforcements, if there are an- 1 and 2, depending on how much area it fills [29]. Similarly,
chorage problems, and other forms of information [8, 13]. To the FD of surfaces lies between 2 and 3 depending on the
reduce the subjectivity and variability associated with man- roughness of surface [14]. The complexity of two curves or
ual VI, automatic vision-based surface crack detection two surfaces can then be easily compared, as the values of FD
methods have become a center of several researches in recent are not anymore restricted to the topological dimensions
years [9, 10, 14–16]. Although these methods can success- of 1 and 2. Several algorithms have been proposed for
fully detect cracks on concrete surfaces, very few if any have the calculation of the FD, including the box-counting
the ability of assessing the impact of these cracks on the algorithm, Hurst R/S analysis, fractional Brownian mo-
overall capacity of the structure. Since the state of cracks tion, and the power spectrum method [28, 30–32].
contains an intrinsic complexity due to erratic damage mech- Among them, the box-counting algorithm is the most
anism of concrete, conventional mathematical or statistical popular [32].
Exp Mech (2013) 53:1607–1619 1609

D=0 covered with a collection of square boxes, and the number of

elements of a given size r is counted to see how many of
them are necessary to completely cover the curve. As the size
of the area element approaches zero, the total area covered by
the area elements will converge to the measure of the fractal
dimension [14, 17, 33–37]. This can be expressed mathe-
log (N(r))

1 2 3 4 5 matically as:
 
logN ðrÞ
D ¼ lim ð1Þ
r→0 logð1=rÞ

where N(r) is the total number of boxes of size r required to

D=0 completely cover the curve and D is the fractal dimension of
RO RS RT RD the curve. In practice, D is estimated by fitting a straight line
to the log-log plot of N(r) versus 1/r over a range of box
log(1/r) sizes. This can be expressed as:
Fig. 1 Variation of FD versus box sizes logðN ðrÞÞ ¼ Dlogð1=rÞ þ C ð2Þ

where C is a constant. The slope of the least square fit line is

Box Counting Algorithm taken as an estimator of the fractal dimension (D) of the
curve. The selection of the range of box sizes is crucial.
The box counting method is motivated by the approach of Figure 1 illustrates a typical structure of a graph of log(N(r))
considering the space filling properties of the curve as an vs. log(1/r) for the fractal analysis of a 2D object. It can be
indication of its complexity. In this approach, the curve is observed that, five possible regions can be identified: (1) if r

(a)
15 15 15 15

10 10 10 10

5 5 5 y 5
y

0 0 0 0

-5 -5 -5 -5

-10 -10 -10 -10

-10 0 10 -10 0 10 -10 0 10 -10 0 10
x x x x
(b)
15 15 15 15

10 10 10 10

5 5 5 5
y

0 0 0 0

-5 -5 -5 -5

-10 -10 -10 -10

-10 0 10 -10 0 10 -10 0 10 -10 0 10
x x x x
(c)
2 2 2 2
D=1.00 D=1.1775 D=1.2115 D=1.4203
log(N(r))

log(N(r))

log(N(r))

log(N(r))

1.5 1.5 1.5 1.5

1 1 1 1

0.5 0.5 0.5 0.5

-1.5 -1 -0.5 0 -1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log(1/r) log(1/r) log(1/r) log(1/r)

Fig. 2 Fractal dimension for simulated cracks: (a) large box sizes, (b) small box sizes, (c) FD estimation
1610 Exp Mech (2013) 53:1607–1619

(a) larger than any cracks in a picture, then the boxes required to
cover the cracks will fill the whole surface) thus D=2 as for an
area; (3) this is the region from which the box sizes should be
selected to give the best estimation of fractal dimension; (4) in
this region the transition from the Fractal to Euclidean regime
occurs. The size of r in this region is called RT in this paper. In
fact, the fractal (disordered) regime at the small scales transits
towards a Euclidean regime at large scales [16, 38]. By defini-
tion, the scale of an image is the ratio of a distance on the
image to the corresponding distance on the specimen (e.g., an
overall view photo of an object has a smaller scale than a
zoomed photo); (5) finally if r is smaller than the discretization
size, RD (e.g., sampling frequency) D=0 as is the dimen-
sion of a point.
Figure 2 shows variations of fractal dimension (D) for
different simulated crack patterns. In particular, Fig. 2(a)
(b) 3050 mm
depicts simulated cracks patterns with coarse mesh size.
450 mm The gray boxes are the ones which contain at least one
segment of a crack. Fig. 2(b) shows how N(r), changes for
different box sizes and Fig. 2(c) illustrates the calculation of
D for each simulated pattern. As expected, the fractal dimen-
Loading 1650 mm sion (D) of a line (left column) is equal to one. Interestingly,
actuators
D for four parallel lines (D=1.17) is less than the same lines
grouped into a diamond shape (D=1.21). When a cross is
915 mm
added, it produces more total length and more surface cov-
erage. Therefore, it causes a jump in D from 1.21 to 1.42. As
AE sensors a result, one may infer that crack patterns with relatively
470 mm
larger values of D may correspond to structural elements
significantly damaged.

Proposed Damage Index

(c)
In order to include the proposed approach in a Structural
Health Monitoring (SHM) strategy for estimating the level of
damage in RC structural elements, a new damage index (DI)
is defined as follows:

Di −D1
DI ¼ ; 0 ≤ DI < 1 ð3Þ
2−D1

where Di is the fractal dimension of the current status

of visible cracks, (e.g., in the ith inspection); D1 is the
fractal dimension computed during the first inspection
(e.g., once the cracks become visible for the first time).
Fig. 3 Experimental setup; (a) north face of SW1 [12]; (b) north face of DI varies between 0 and 1 and describes the difference
SW2; (c) south face kept clean for crack mapping
between the current status of crack patterns and the
baseline D1. The constant “2” in the denominator is
is much larger than the size of the object, RO, (i.e., the box the maximum value of fractal dimension for surface
always cover the object so N(r) =1) then D=0; (2) if r is very cracks (i.e., when cracks cover the whole area of the
larger than the size of structure in the object, RS, (e.g., if r is concrete, D tends to 2).
Exp Mech (2013) 53:1607–1619 1611

Table 1 Experimental details The lateral loading protocols for SW1 and SW2 are illus-
Characteristics SW1 SW2 trated in Fig. 4. It consisted of 10 load steps (LS) for SW1
and 11 load steps for SW2. The first load step had three
Height [mm] 3300 2100 cycles and the subsequent load steps had two cycles. The
Width [mm] 3050 3050 first load step was primarily applied to verify the function-
Thickness [mm] 200 200 ality of the experiment setup and it was not included in the
Reinforcement ratio [%] 0. 67 1 damage assessment results since no damage was observed.
Concrete strength f′c (28th day) [MPa] 21 35 The last load step for SW2 failed to be completed because of
Concrete strength f′c (test day) [MPa] 25 48.2 a malfunctioning which resulted out-of-plane rotation of the
Reinforcing bars yield strength [MPa] 464 434 wall. The loading cycles are intended to simulate the effects
Reinforcing bars ultimate strength [MPa] 708 460 of earthquake shaking. The loading rate was set at around
0.6 cm/s for SW1 and 0.13 cm/s for SW2. More details on
the design of these walls can be found in [42].
Experimental Setup Axial load was not applied on these specimens because axial
stresses in low aspect ratio shear walls are typically very small,
The test specimens were two large scale rectangular as measured by a fraction of the product of Ag ×fc′ (in which Ag
shear walls, named, SW1 and SW2, designed based on is the gross section area and f′c is the compressive strength). The
the ACI 318-08, chapter 21, earthquake-resistance struc- force-displacement hysteresis loops and the corresponding
tures [39] with a height to width ratio of 0.94 and 0.54, backbone curves are illustrated in Fig. 5. SW1 and SW2 had
respectively [40, 41]. The experimental setup is shown a nonlinear response at LS7 and LS8, respectively, and their
in Fig. 3 and specimens’ characteristics are summarized ultimate strengths were obtained at LS9 and LS10. After
in Table 1. reaching the peak load in every cycle, the actuators’ load was
Each specimen was rigidly connected to the kept constant for approximately 20 min to allow for cracks to be
laboratory’s foundation using 14 Dywidag bars with a mapped and pictures to be taken. The north face of each wall
1.5-in nominal diameter and subjected to a displacement was instrumented and the south face was kept clear for crack
controlled reversed cyclic loading in the plane of its mapping, see Fig. 3(c). Photos were taken with an HD optic
web. The lateral load was applied by two horizontally camera. Cracks were mapped manually using Adobe
inclined high force capacity actuators. The actuators Photoshop Element software [43] and the results were used as
were actively controlled by master–slave method to input to the proposed fractal analysis algorithm.
ensure a synchronized and symmetric lateral loading
on the wall. The two actuators were inclined by an
angle of 9° to prevent out of plane deformation. The Experimental Results
actuators were attached to the specimen through custom
made brackets and plates that were post-tensioned to the Visual Inspection
sides of the specimen [2]. The centerline of loading was
45 cm below the top of the walls. More details on This section presents the results of visual observations and
design of these walls can be found in [42]. crack width measurements along with identified damage

(a) (b)
60 40
SW1 SW2
30
Displacement[mm]

40
Displacement[mm]

20
20
10
0 0

-20 -10
-20
-40
-30
-60 -40
0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 11
Load Step Load Step

Fig. 4 Load protocol for: (a) SW1 [12], (b) SW2 [2]
1612 Exp Mech (2013) 53:1607–1619

(a) (b)
1500 3000 10
1st cycles 8 9 1st cycles 9 11
7 10 8
1000 2nd cycles 6 2000 2nd cycles 7
5
SW1 SW2
500 1000
Force [kN]

Force [kN]
0 0

-500 -1000
5
7
-1000 6 -2000
8
10 9 8 7 10 9
-1500 -3000
-60 -40 -20 0 20 40 60 80 -30 -20 -10 0 10 20 30 40
Displacement [mm] Displacement [mm]

Fig. 5 Force-displacement hysteresis loops for: (a) SW1 [12]; (b) SW2 [2]

Fig. 6 Residual cracks pattern for SW1 by increasing load step [42]
Exp Mech (2013) 53:1607–1619 1613

LS2 LS3 LS4

LS5 LS6 LS7

LS8 LS9 LS10

Fig. 7 Residual cracks pattern for SW2 by increasing load step [42]

grades (DGs). Considering that specimens used in this study appearing most frequently) crack widths are indicated in each
are typically being used for nuclear safety related structures, picture with a square dot (□) and with a circle (○), respectively.
the IAEA guidebook on non-destructive testing of concrete Figure 8 shows the maximum and modal crack width
structures [1] was used to define the DGs. According to this measured on the surface of the wall when it was at the peak
guideline, three DGs can be defined (i.e., I, II, and III), based on deformation of each load step. Crack width thresholds based
the width of cracks. In particular, if the maximum width is less on the IAEA guidelines and DG boundaries are also indicat-
than 0.2 mm, the damage is classified as grade I, between 0.2 mm ed by horizontal and vertical dashed lines, respectively. It can
and 1 mm as grade II, and above 1 mm as grade III. In general, be observed that, DG II initiates in LS2 for SW1 and LS4 for
non-structural or hairline cracks are treated as grade I, structural SW2. Instead the critical damage stage (III) initiates in LS7
cracks (or medium cracks) are considered as moderate damage or and LS9, respectively. Considering that nonlinear mechani-
grade II, and critical damage or wide cracks as grade III. No cal behavior was observed during these load steps (see
repair is needed for grade I, whereas appropriate rehabilitation Fig. 5), they will be assumed as the load steps in which
strategy is necessary for grades II and III [1]. Residual cracks “critical damage” occurred.
developments in each load step are illustrated in Figs. 6 and 7 for Figure 9 illustrates cracks width measured once loading
SW1 and SW2, respectively. The maximum and modal (i.e., was ceased and cracks were closed (i.e., residual cracks). It

(a) (b)
10.0 4.0
Crack width at peak deformation [mm]

Crack width at peak deformation [mm]

Maximum width Maximum width

9.0
Modal width 3.5
8.0 Modal width
SW1 3.0
7.0 SW2
6.0 2.5
I II III
5.0 I II III 2.0
4.0 1.5
3.0
1.0
2.0
1.0 0.5

0.0 0.0
0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Load step Load step

Fig. 8 Width of cracks at peak displacement vs. load steps and damage grades for: (a) SW1, (b) SW2
1614 Exp Mech (2013) 53:1607–1619

(a) (b)
4.0 4.0
Maximum width SW1 Maximum width
3.5 3.5

Residual crack width [mm]

Residual crack width [mm]
Modal width Modal width
3.0 3.0
SW1 SW2
2.5 2.5 I II III
I II III
2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Load step Load step

Fig. 9 Width of residual cracks vs. load steps and damage grades for: (a) SW1, (b) SW2

can be seen that, based on residual cracks width DG, II The box-counting algorithm was used to compute the
initiates in LS7 for both walls whereas DG III initiates in fractal dimension (D) for each image. Box sizes were
LS8 for SW1 and LS10 for SW2. Therefore, condition selected in a range between RS and RT. Then, a grid of
assessment based on residual cracks width would provide square boxes was superimposed over each picture
very inaccurate results. It should be mentioned that residual (Fig. 11(a)). Next the number of boxes that intersect
cracks width are used currently by engineers to determine the cracks, N(r), was counted. The method was repeated with
type of repair required for a damaged component (e.g., after a denser and denser grid to define the number of boxes as
an earthquake). a function of the grid spacing, r. Therefore, the estimated
D was the slope of least square fit line from RS to RT in
Fractal Dimension Results the log-log plot of N(r) versus 1/r. This procedure is
illustrated in Fig. 11 for the SW1 at LS5.
This section presents the results of damage classification Figure 12 shows the evolution of the fractal dimen-
using the fractal dimension of residual cracks. At this aim, sion in each load step, for the two specimens SW1 and
pictures were taken at the end of each load step in the south SW2. In general, the fractal dimension increases as the
face of the wall. These pictures were imported in AutoCad surface damage increases; this is consistent with the
[44], and then cracks were manually mapped, as shown in damage evolution observed for the simulated cracks in
Fig. 10 for LS5 as an example. Fig. 2. In addition, three regions can be identified: 1) an

LS5 (a) LS5 (b)

Fig. 10 Cracks style for SW1-LS5: (a) photo of the wall, (b) extracted cracks pattern
Exp Mech (2013) 53:1607–1619 1615

(a) (b) (c)

2.5
RT FD=1.63
RS
2

log(N(r))
1.5

1
-2 -1.5 -1
log(1/r)

Fig. 11 Steps of FD calculation on residual cracks for SW1-LS5 as an example; (a) coarse mesh, (b) fine mesh, (c) estimation of FD

initial region in which the fractal dimension sharply increases; 1.73 (an increment of about 0.1). This increase is mostly due
2) a plateau region in which the fractal dimension slowly to the accumulation of more diagonal cracks on the middle area
changes, and 3) a third region in which the fractal dimension and horizontal cracks on the bottom area of the wall. From LS8
further increases. These observations can be explained as to LS10 few more cracks are formed; therefore small variations
follows, by comparing Fig. 12 with the residual cracks pat- in the FD are observed. Similarly, in SW2, the FD sharply
terns shown in Figs. 6 and 7. increases up to 1.63 in LS3 (region 1), then gradually increases
In SW1, when the specimen is loaded for the first time, there until LS8 (FD=1.77) with small increments of about 0.03
is an appreciable change on the surface of the specimen caused (region 2). After this load step (onset of plastic response), the
by the formation of horizontal (flexural) and diagonal (shear) FD increases up to 1.87 with an increment of about 0.1 (region
macro-cracks, as can be observed in Fig. 6 (i.e., LS2 to LS5). 3). Afterward, the FD slowly increases until failure of the
As a result, the FD increases up to a value of 1.63 between LS2 specimen (LS10).
and LS5. This phase is called “region 1”. Then the FD slowly Figure 13 illustrates how the selection of RT may
changes (by less than 0.01 increments) until LS7 in which the affect the fractal dimension results. A general decrease
onset of plastic response was observed (see Fig. 5); this phase is can be observed using finer values of RT; however, the
characterized by no significant changes in the crack patterns overall trend remains mostly unchanged during the
(see Fig. 6). This region of dormancy is called “region 2”. loading process and the aforementioned regions can
Interestingly after this step (region 3), the FD jumps up to be still identified.

(a) (b)
2.00 2.00
1.90 1.90
1 2 3
1.80 1.80
Fractal dimension
Fractal dimension

1.70 1.70
1.60 1.60
1.50 1.50
1.40 1.40
1.30
1.30 1 2 3
1.20 1.20
1.10 SW2
1.10 SW1
1.00 1.00
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Load step Load step

Fig. 12 Variation of fractal dimension versus increasing load steps (RT =11 cm)
1616 Exp Mech (2013) 53:1607–1619

Fig. 13 Trend of FD vs. load steps using a range of RT’s

Damage Index Results wall. Furthermore, it is valuable indicator of damage in cases

where applied load and stiffness loss (as a sign of damage)
In this section the results of the proposed damage index (DI) are not available. Table 2 summarizes the walls’ stiffness,
are presented and compared with the relative stiffness loss RSL, D, DI, and true DG classification (i.e., based on cracks
(RSL) defined as: width measured at peak displacement) for each load step. In
this work two threshold levels, δj, are proposed in order to
Ki discriminate damage grades (i.e., j=1 for DG II and j=2 for
RSL ¼ 1− ð4Þ
K1 DG III). In particular, a δ1 =0.4 is defined for DG II and δ2 =
0.7 for the onset of DGIII.
where Ki is the lateral secant stiffness of the wall at the ith To assess how this methodology could enhance a regular
load step (i.e., i=1,2,…10) and K1 is the initial stiffness of visual inspection, Fig. 15 summarizes the final results of
the wall. Figure 14 shows the results of RSL and DI versus damage grade identification using crack width measurement
drift ratio for both walls. Interestingly, DI can generally track at peak deformation and at zero deformation (i.e., residual
the relative stiffness loss with a slight error. Therefore, DI cracks) versus proposed damage index. As shown in this
can be exploited to infer the remnant lateral stiffness of the figure, damage grade assessment based on residual crack

(a) (b)
1.0 1.0 1.0 1.0
0.9 SW1 0.9 0.9 SW2 0.9
Relative stiffness loss (RSL)

0.8 0.8 0.8 0.8

Relative stiffness loss (RSL)
Damage index (DI)

0.7 0.7 0.7 0.7

Damage index (DI)

0.6 0.6 0.6 0.6

0.5 0.5 0.5 0.5
0.4 0.4 0.4 0.4
0.3 0.3 0.3 0.3
0.2 0.2 0.2 0.2
DI DI
0.1 0.1 0.1 0.1
RSL RSL
0.0 0.0 0.0 0.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Drift ratio [%] Drift ratio [%]

Fig. 14 Comparison of DI and relative stiffness loss for (a) SW1, (b) SW2
Exp Mech (2013) 53:1607–1619 1617

Table 2 Specimens’ lateral stiffness and fractal dimension results

Load Step SW1 SW2

Stiffness [kN/m] D RSL DI True DG Stiffness [kN/m] D RSL DI True DG

1 1588 – – – I 5190 – – – I
2 1370 1.05 0.14 0.00 II 4101 1.38 0.21 0.00 I
3 926 1.40 0.42 0.41 II 3607 1.62 0.31 0.39 I
4 860 1.52 0.46 0.54 II 2991 1.64 0.42 0.42 II
5 794 1.63 0.50 0.65 II 2787 1.68 0.46 0.48 II
6 733 1.63 0.54 0.66 II 2758 1.73 0.47 0.57 II
7 572 1.64 0.64 0.66 III 2570 1.77 0.50 0.63 II
8 435 1.73 0.73 0.74 III 1861 1.78 0.64 0.65 II
9 302 1.73 0.81 0.75 III 1163 1.87 0.78 0.79 III
10 209 1.75 0.87 0.76 III 470 1.93 0.91 0.88 III

width has the maximum error with respect to other methods in tors. Worldwide, various condition-rating grades have been
both walls. However, DI compensates for this error signifi- developed to overcome the qualitative nature of VI-based
cantly in both walls. For SW1, it could extensively enhance condition assessment. These scales usually quantify damage
the erroneous estimation of onset of DG II by residual cracks conditions in terms of crack’s width. However, in the elastic
from LS6 to LS2 (4 load steps improvement). For SW2, it regime cracks can be totally or partially closed upon load
could accurately signify LS3 as the onset of DG II. Further- removal. Thus, width of cracks may leads to erroneous re-
more, DI could slightly improve onset of DG III in SW2. sults. In this paper a novel approach based on the fractal
dimension (FD) of crack patterns was proposed to assess the
structural integrity of RC structures. The approach was val-
Conclusions idated through large-scale tests on two RC shear walls
subjected to lateral reversed cyclic loading. It was observed
It is known that RC structures are subjected to deterioration that in general the fractal dimension increases as the surface
due to aging, increased load, and natural hazards. To mini- damage increases. Further, a damage index (DI) based on
mize the maintenance costs and to increase the operation fractal dimension of visible residual cracks was introduced to
lifetime, researchers and practitioners are increasingly inter- discriminate different damage grades. It was shown that the
ested in improving current nondestructive evaluation (NDE) proposed DI could be exploited to estimate the remnant
technologies. Conventionally, the assessment of RC struc- lateral stiffness of the wall. The proposed index could have
tures relies on visual inspection (VI) which is time- a role in post-disaster assessment and planning by estimating
consuming and depends heavily on the skills of the inspec- the vulnerability of buildings exposed to natural hazards

(a) (b)
SW1 Grade I SW2 Grade I
Grade II Grade II
Grade III Grade III
Method

DI
Method

DI

Residual crack width Residual crack width

Crack width at peak deformation Crack width at peak deformation

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Load step Load step

Fig. 15 Damage grade classification by different methods; (a) SW1, (b) SW2
1618 Exp Mech (2013) 53:1607–1619

(storms, earthquakes, etc.) and enabling authorities to decide 15. Jahanshahi MR, Masri SF (2012) Adaptive vision-based crack detec-
tion using 3D scene reconstruction for condition assessment of struc-
whether or not a building can be safely re-occupied. More
tures. Autom Constr 22:567–576. doi:10.1016/j.autcon.2011.11.018
formal tests need to be carried out to verify the robustness of 16. Carpinteri A, Chiaia B, Nemati KM (1997) Complex fracture
the approach. In addition, more specimens with different energy dissipation in concrete under different loading conditions.
shape and size should be examined. Mech Mater 26(2):93–108
17. Carpinteri A, Yang GP (1996) Fractal dimension evolution of
microcrack net in disordered materials. Theor Appl Fract Mec
Acknowledgments The authors acknowledge National Science
25(1):73–81
Foundation (NSF) for providing the financial support under Grant No.
18. Lee YH, Carr JR, Barr DJ, Haas CJ (1990) The fractal dimension as
CMMI-0829978. The experiments presented herein could not have
a measure of the roughness of rock discontinuity profiles. Int J
been completed without contributions from the staff of the Structural
Rock Mech Min 27(6):453–464
Engineering and Earthquake Simulation Laboratory (SEESL) of the
19. Kulatilake PHSW, Fiedler R, Panda BB (1997) Box fractal dimen-
State University of New York at Buffalo. The financial support and
sion as a measure of statistical homogeneity of jointed rock masses.
work of the SEESL staff are gratefully acknowledged. The authors also
Eng Geol 48(3–4):217–229
acknowledge the advice and help provided by the technical staff at the
20. Carpinteri A, Lacidogna G, Niccolini G (2009) Fractal analysis of
NEES Equipment Site at the University at Buffalo.
damage detected in concrete structural elements under loading.
Chaos Soliton Fract 42(4):2047–2056
21. Carpinteri A, Corrado M, Lacidogna G (2012) Three different
approaches for damage domain characterization in disordered ma-
References terials: Fractal energy density, b-value statistics, renormalization
group theory. Mech Mater 53:15–28
22. Chiaia B, van Mier JGM, Vervuurt A (1998) Crack growth mech-
1. IAEA (2002) Guidebook on non-destructive testing of concrete anisms in four different concretes: microscopic observations and
structures. International Atomic Energy Agency, Vienna fractal analysis. Cem Concr Res 28(1):103–114
2. Farhidzadeh A, Salamone S, Luna B, Whittaker A (2013) Acoustic 23. Saouma VE, Barton CC (1994) Fractals, fractures, and size effects
emission monitoring of a reinforced concrete shear wall by b-value in concrete. J Eng Mech-ASCE 120(4):835–854
based outlier analysis. J Struct Health Monit Int J 12(1):3–13. 24. Peng J, Wu Z, Zhao G (1997) Fractal analysis of fracture in
doi:10.1177/1475921712461162 concrete. Theor Appl Fract Mec 27(997):135–140
3. RAIU (2010) Malahide viaduct collapse on the Dublin to Belfast line, 25. Carpinteri A, Cornetti P (2011) Size effects on concrete ten-
on the 21st August 2009. Railway Accident Investigation Unit, Dublin sile fracture properties: an interpretation of the fractal ap-
4. Comerio M, Elwood K, Berkowitz R et al (2011) The M 6.3 proach based on the aggregate grading. J Mech Behav Mater
Christchurch, New Zealand, Earthquake of February 22, 2011. 13(3–4):233–246
EERI special earthquake report. Earthquake Engineering Research 26. Sun H-Q, Ding J, Guo J, Fu DL (2011) Fractal research on cracks of
Institute (EERI), Oakland reinforced concrete beams with different aggregates sizes. Adv Mat
5. DBH (2011) Christchurch CBD Buildings 22 February 2011 after- Res 250–253:1818–1822
shock stage 1 expert panel report. New Zealand Department of 27. Cao M, Ren Q, Asce M, Qiao P (2006) Nondestructive assessment
Building and Housing, Wellington of reinforced concrete structures based on fractal damage charac-
6. NBIS (1996) Code of federal regulations, No. 23CFR650. National teristic factors. J Eng Mech-ASCE 132(9):924–931
Bridge Inspection Standards, Washington, DC 28. Mandelbrot BB (1982) The fractal geometry of nature. W. H.
7. FHWA (1995) Recording and coding guide for the structure inven- Freeman, New York
tory and appraisal of the nation’s bridges. Report No. FHWA-PD- 29. Dubuc B, Quiniou JF, Roques-Carmes C, Tricot C, Zucker SW
96-001. Federal Highway Administration, Washington, D.C (1989) Evaluating the fractal dimension of profiles. Phys Rev A 39:
8. ATC-43 (1998) FEMA 306. Evaluation of earthquake damaged 30. Taylor CC, Taylor SJ (1991) Estimating the dimension of a fractal. J
concrete and masonry wall buildings. The Applied Technology Roy Stat Soc B Met 53(2):353–364
Council, Redwood City 31. Esteller R, Vachtsevanos G, Echauz J, Litt B (2001) A comparison
9. Chen Z, Hutchinson TC (2010) Image-based framework for con- of waveform fractal dimension algorithms. IEEE T Circuits-I
crete surface crack monitoring and quantification. Adv Civ Eng 48(2):177–183
2010:1–18. doi:10.1155/2010/215295 32. Theiler J (1990) Estimating fractal dimension. J Opt Soc Am A
10. Sohn H-G, Lim Y-M, Yun K-H, Kim G-H (2005) Monitoring 7(6):1055–1073
crack changes in concrete structures. Comput-Aided Civ Inf 33. Moustafa A, Salamone S (2012) Fractal dimension-based Lamb
20(1):52–61 wave tomography algorithm for damage detection in plate-like
11. Farhidzadeh A, Salamone S, Singla S (2013) A probabilistic ap- structures. J Intel Mat Syst Str 23(11):1269–1276
proach for damage identification and crack mode classification in 34. Raghavendra BS, Narayana Dutt D (2010) Computing fractal di-
reinforced concrete structures. J Intel Mat Syst Str. doi:10.1177/ mension of signals using multiresolution box-counting method. Int
1045389X13484101 J Inf Math Sci 6(1):50–65
12. Farhidzadeh A, Dehghan Niri E, Salamone S, Luna B, Whittaker A 35. Shoupeng S, Peiwen Q (2007) A fractal-dimension-based signal-
(2012) Monitoring crack propagation in reinforced concrete shear processing technique and its use for nondestructive testing. Russ J
walls by acoustic emission. ASCE J Struct Eng. doi:10.1061/(ASCE) Nondestruct 43(4):270–280
ST.1943-541X.0000781, First published on 01 December 2012 36. Hadjileontiadis LJ, Douka E (2007) Crack detection in plates using
13. ATC (1998) FEMA 308. Repair of earthquake damaged concrete fractal dimension. Eng Struct 29(7):1612–1625
and masonry wall buildings. The Applied Technology Council, 37. Long QY, Suqin L, Lung CW (1991) Studies on the fractal dimen-
Redwood City sion of a fracture surface formed by slow stable crack propagation. J
14. Issa MA, Issa MA, Islam MS, Chudnovsky A (2003) Fractal Phys D Appl Phys 24:602–607
dimension - a measure of fracture roughness and toughness of 38. Mandelbrot BB (1985) Self-affine fractals and fractal dimension.
concrete. Eng Fract Mech 70:125–137 Phys Scr 32(4):257
Exp Mech (2013) 53:1607–1619 1619

39. ACI Committee 318 (2008) Building code requirements for struc- concrete shear walls by acoustic emission. 54th Acoustic Emission
tural concrete and commentary (ACI 318-08). American Concrete Working Group Meeting, Princeton, NJ, USA (Student paper
Institute (ACI), Farmington Hills. ISBN 9780870312649 award), pp 55–57
40. Farhidzadeh A, Salamone S, Dehghan-Niri E, Luna B, Whittaker 42. Rocks JF (2012) Large scale testing of low aspect ratio reinforced
AS (2012) Damage assessment of reinforced concrete shear walls concrete walls. M.Sc. Thesis, University at Buffalo, NY
by acoustic emission. NDE/NDT for Highways and Bridges: 43. Adobe Photoshop Element, version 9, adobe company, Copyright
Structural Materials Technology (SMT), NY, pp 74–81 © 2010 Adobe Systems Incorporated. http://www.adobe.com
41. Farhidzadeh A, Salamone S (2012) Introducing sifted b-value 44. Autodesk AutoCAD Civil 3D, student version 2011, Copyright ©
analysis and a new crack classification for monitoring reinforced 2012 Autodesk, Inc. http://students.autodesk.com/