International Journal of Fatigue 113 (2018) 98–112

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Prognosis of low-strain fatigue induced damage in reinforced concrete T

structures using embedded piezo-transducers
⁎
Suresh Bhalla , Naveet Kaur1
Department of Civil Engineering, Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110016, India

A R T I C LE I N FO A B S T R A C T

Keywords: Fatigue induced damage under low-strain conditions is a commonly encountered phenomenon in reinforced
Fatigue concrete (RC) structures subjected to continuous vibrations, such as bridges under vehicular loads, buildings
Residual stiffness under wind and structures directly supporting or in the vicinity of vibration emitting machinery. This type of
Reinforced concrete (RC) fatigue tends to weaken the structure silently, and quite often, its timely cognizance is missed out in the life-cycle
Low-strain
management of the structure. Among the various structural health monitoring (SHM) frontiers, very scarce
Smart materials
Electro-mechanical impedance (EMI) technique
research has been devoted to this kind of damage in RC structures. This paper covers a long-term experimental
Concrete vibration sensor (CVS) study related to low-strain fatigue damage monitoring encompassing a real-life sized RC structure. The ex-
Piezoelectric ceramic lead zirconate titanate perimental specimen was subjected to over eight million cycles of flexural loading with the maximum bending
(PZT) strain values restricted to 50 µm/m. The structure was instrumented with piezo-based composite concrete vi-
bration sensors (CVS), which were embedded inside the beam near the surface. The CVS operated in dual mode,
acting as sensor for the global vibration technique as well as the local electro-mechanical impedance (EMI)
technique. As EMI sensors, they facilitated the determination of the equivalent stiffness parameter (ESP), and
were found to be very expedient for damage detection as well as localization during the incipient stages of
fatigue damage, coinciding with the appearance of first few cracks. The ESP served to represent the diminishing
trend to actual residual stiffness fairly well during the initial stages. Acting in the global mode, determination of
the overall stiffness of the structure by the same CVS provided an alternate damage measure, in terms of a
realistic estimate of the overall residual flexural stiffness. This proved useful parameter during moderate to
severe damage conditions, and in particular near the final failure of the structure. The monitoring paradigms
presented in the paper pave way for effective prognosis of fatigue induced damage in real-life RC structures.

1. Introduction cement-mortar matrix, the phenomenon of fatigue manifests somewhat

‘Fatigue’, a weakened condition in structures caused by repeated
loading, ultimately resulting in fracture at a stress much lower than that
necessary to cause fracture in a single application [1], is a well un-
derstood phenomenon in metals. It begins as a well-demarcated crack,
progressively advancing in three phases – (a) initiation/nucleation, (b)
slow sustained growth, and finally, (c) culminating in abrupt failure
[2]. Fatigue in metals can be represented in terms of the plot of the
maximum induced stress versus the number of load cycles to failure,
commonly called as the SN or the Wohler curve [3]. These curves can
be used for making some degree estimation about the service life of a
metallic component when subjected to a known amplitude of cyclic
loads. Most metals have SN curves asymptotic to a definite stress value,
called as the ‘fatigue limit’ [4]. However, in the context of concrete,
which is a composite material consisting of rocky aggregate in a
differently. Here, the beginning of fatigue damage is marked by the
development of micro-cracks, which, in contrast to metals, are wide-
spread throughout the matrix. Further damage progression leads to the
appearance of larger cracks, but not as localized as in the case of the
metals. Further, the overall fatigue performance of concrete depends
upon whether it is subjected to high or low-strains, with the component
failing in relatively small (< 10,000) or large (> 10,00) number of
loading cycles respectively [5]. Accordingly, the fatigue is designated as
low-cycle (high strain) or high-cycle (low-strain) respectively. In addi-
tion, in contrast to metals, the fatigue limit cannot be explicitly defined
for concrete. Rather, fatigue limit is expressed as the strength after
undergoing a specified number of cycles, say a million [4]. Another
distinguishing feature of concrete fatigue, as opposed to metals, is the
phenomenon of temporary recovery during the rest periods [6].
A number of studies can be found in the literature dovetailed to

⁎
Corresponding author.
E-mail addresses: sbhalla@civil.iitd.ac.in (S. Bhalla), nkaur.crri@nic.in (N. Kaur).
1
Presently Scientist, CSIR-Central Road Research Institute (CRRI), Mathura Road, New Delhi 110025, India.

https://doi.org/10.1016/j.ijfatigue.2018.04.002
Received 4 November 2017; Received in revised form 11 March 2018; Accepted 2 April 2018
Available online 06 April 2018
0142-1123/ © 2018 Elsevier Ltd. All rights reserved.
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

Vo
+
_
PZT PATCH CVS D’
+
_

(a)

1 MORTAR

EPOXY
ADHESIVE 0.3 mm
PZT PATCH
LAYER

(b)
Fig. 1. Concrete vibration sensor (CVS). (a) Embedded installation in an RC structure. (b) Optical image along section xx (Courtesey: Prof. Richard Ball, University of
Bath, UK).

curve between the normalized secant stiffness and the normalized

Fig. 2. 2D effective impedance approach for modelling PZT-structure electro-
mechanical interaction [39].
fatigue damage in plain concrete subjected to compression or flexure.
Mu et al. [7] experimentally found the damage evolution phase in plain
concrete under compressive fatigue to be represented by an S-shaped
number of cycles. Three distinct regions of the curve were identified.
The first region, which is small (spanning about 5% of cycles), is
marked by a significant drop in the axial stiffness. The next region,
which forms the major part, is characterized by a gradual and linear
stiffness loss, typically spanning up to 90% of the loading cycles. This is
followed by the third and the final region of rapid loss of stiffness,
eventually culminating in failure. During the experiments, the total
number of cycles to failure ranged between 1300 and 15,000, corre-
sponding to the upper limit of the loading equal to 90–76% of the
average static compressive strength of concrete respectively. In addi-
tion, the slope of the second region was found to provide a reasonable
indication of the fatigue life of the specimen. Medeiros et al. [8] found
that the loading frequency has pronounced effect on fatigue life of plain
concrete under compression. A higher frequency implies longer fatigue
life. Oh [3] generated SN curves of plain concrete subjected to com-
pressive fatigue. Nieto et al. [9] proposed a simplified approximate
method for obtaining SN curve of plain concrete under high-cycle fa-
tigue, employing vibration response for estimating the accumulated
damage and the logarithmic slope parameter of the curve. Oneschkow

99
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

Function Generator
(Agilent 33210A)
Computer
Digital Multimeter
(Agilent 3441A) LCR meter
Oscilloscope
(TDS 2004B)

Location of embedded CVS Amplifier

flushing at top (typical). (LDL PA 500L)
Identical CVS flusher with Inertial Type
bottom at same locations Dynamic Shaker Shaker
(LDS V406) 7

6 5 4 3 2 1
19 18 17 16 15 14 13 12 11 10 9 8

Notch after filling concrete

Number of CVS (Typical)

(a)

Embedded Piezo Patch

Mortar Layer

Adhesive Layer 2 Nos. 16 mm Dia

Mortar Layer

2 Nos. 12 mm Dia

Right Half of RC Beam

Notch for damage

Concrete Vibration Sensor (CVS)

(Typical)
(b)
Fig. 3. (a) Experimental set-up and specimen details (b) See through 3D model of the symmetric half of the RC beam showing main reinforcement, notch and location
of CVS.

[10] investigated low-cycle compressive fatigue behaviour of high-
strength concrete with respect to the maximum stress level, the loading
frequency and the waveform. Kolluru et al. [11] studied the fatigue
behaviour of concrete in flexure under low-cycle loading. Flexural crack
growth was found to comprise of two phases: (a) a deceleration stage,
characterized by a decrease in crack growth rate; and (b) an accelera-
tion stage, marked by a steady rate of crack growth. The damage evo-
lution, in terms of stiffness, was again found to follow an S-curve with
three distinct regions as in the case of low-cycle compressive fatigue
[7]. Several other studies involving low-cycle fatigue in flexure mode
can be found [12–14]. However, studies involving high-cycle (low-
strain) fatigue phenomenon of plain concrete, both compression as well
as flexure, are scarce in literature and only a handful of studies are
available [15,16].
Compared to plain concrete, the fatigue behaviour of reinforced
concrete (RC) is somewhat more complex, owing to the composite ac-
tion of two contrastingly different materials – ductile steel and brittle
concrete. Here, the fatigue process could occur in concrete, steel, or the
bond between the two materials. Only limited investigations are re-
ported on fatigue damage in RC structures. Stroeven [6] studied the
behaviour of RC specimens under compression induced fatigue. Chang
and Chai [4] were among the first to report on high-cycle fatigue per-
formance of RC beams under flexure. They applied fatigue loads ran-
ging from 50% to 80% of the static load capacity. The corresponding

100
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

Table 1 fracture through concrete. Based on the test results, empirical re-
Properties of PZT patch [45,46]. lationships were derived between the fatigue stress in tensile/shear
Property Unit Value reinforcement bars and the number of cycles to failure. Makita and
Bruhwiler [17] studied the tensile fatigue behaviour of ultra-high per-
formance fibre and steel reinforced concrete specimens.
2
Plan size mm 10 × 10
Thickness mm 0.3
In spite of the known vulnerability of RC structures to fatigue, not
Piezoelectric strain coefficient, d31 m/V −2.100 × 10−10
Young’s modulus, YE N/m2 6.667 × 1010
much research has been devoted to systematic prognosis of fatigue in-
Piezoelectric strain coefficient, d33 m/V 5.000 × 10−10 duced damage. Fatigue of RC structural members caused by vibrations
T
Electric permittivity, ε33 Farad/m 1.7785 × 10−8 from generator sets was one of the causes for sudden collapse of an RC
2 YE
2d31 N/V2 5.35 × 10−9 framed garment factory in Bangladesh in 2013 [18], which had a death
1−ν toll of 1134 associated with it. Given the high degree of fatigue vul-
Mechanical loss factor, η 0.0325
Dielectric loss factor, δ 0.0224
nerability coupled with the low level of scientific understanding of the
Poisson’s ratio, ν 0.3 process in RC structures, it is imperative for the structural health
monitoring (SHM) community to develop specialized sensors and
prognosis approach for effectively monitoring this kind of damage,
8 which could be lethal for RC or prestressed concrete structures. This
7 paper presents a new fatigue prognosis approach covering in-
6 strumentation and monitoring of RC structures subjected to low-strain
Voltage (Volts)

5 (high-cycle) fatigue using piezo-based embeddable concrete vibration

4 sensors (CVS), a kind of ‘smart’ sensor. The main novelty of the prog-
3 nosis approach is that it employs both low and high frequency vibration
2 signals from the embedded CVS sensors, covering fatigue damage in-
1 itiation, localization, growth and final collapse. The experimental study
0 has been performed on a real-life sized prototype structure that was
0 5 10 15 20 25 30 subjected to over eight million cycles of low-strain loading. Low-strain
Frequency (Hz) loading on the structure further adds to the novelty of the research as
this type of loading is the most realistic form of fatigue loading emu-
Fig. 4. Typical voltages versus frequency plot from impact test.
lated in the lab environment. The forthcoming sections of the paper
provide a brief introduction to CVS, the main sensors deployed for fa-
Table 2 tigue damage monitoring, covering their working as global vibration
Fatigue load summary. and local electro-mechanical impedance (EMI) sensors, followed by the
State Cycles Cycle ratio (CR) Natural frequency experimental study involving the real-life sized RC beam, results, dis-
cussions and conclusions.
1 0 0.000 17.2
2 20,000 0.002 17.1
3 30,000 0.004 17 2. Sensors and systems for monitoring fatigue
4 50,000 0.006 16.9
5 70,000 0.008 16.9 2.1. Concrete vibration sensor (CVS)
6 100,000 0.012 16.9
7 150,000 0.018 16.9
8 200,000 0.025 16.9
CVS is a packaged and ready-to-use lead zirconate titatnate (PZT)-
9 250,000 0.031 16.8 cement-mortar composite sensor especially designed for RC structures
10 300,000 0.037 16.8 [19]. Fig. 1(a) shows an external view of the CVS and Fig. 1(b) an
11 500,000 0.062 16.6 optical image at a vertical section, highlighting the internal structure,
12 600,000 0.075 16.5
comprising of the PZT element, the surrounding hardened epoxy ad-
13 700,000 0.087 16.5
14 800,000 0.100 16.5 hesive layer and the outermost mortar covering. Owing to this special
15 900,000 0.112 16.5 fabrication, CVS has sound encapsulation to protect the PZT element
16 1,000,000 0.124 16.5 from deterioration by water inside concrete. It can withstand the hard
17 1,100,000 0.137 16.5 conditions generally encountered in RC structures during casting and
18 1,200,000 0.149 16.5
19 1,300,000 0.162 16.5
achieves near perfect bonding with the surrounding concrete material
20 1,400,000 0.174 16.4 after curing, thereby becoming a permanent part of the structure. As an
21 1,700,000 0.211 16.2 embedded sensor, the CVS has been shown to be equally competent for
22 1,800,000 0.223 16 global as well as localized monitoring of RC structures [20]. The au-
23 2,200,000 0.273 15.8
thors have recently demonstrated its mantle for combined structural
24 2,450,000 0.304 15.7
25 2,650,000 0.329 16.6 health monitoring (SHM) and energy harvesting [21] as well as impact
26 3,000,000 0.373 16 and fatigue induced damage in plain concrete [22]. Following subsec-
28 5,000,000 0.622 16 tions cover brief theoretical aspects related to CVS as a dynamic strain
30 7,000,000 0.871 15.8 sensor and as an EMI sensor.
31 8,000,000 0.995 13.2
32 8,020,000 0.998 11.26
34 8,030,000 0.999 10.7 2.2. CVS as dynamic strain sensor for global SHM
35 8,035,000 0.999 10.2
36 8,040,000 1.000 9.9 The piezoelectric materials, in particular PZT, belong to the cate-
gory of the ‘smart’ materials, owing to the ‘stimulus-response’ char-
acteristics exhibited by them. When subjected to mechanical stresses, a
number of the cycles of loading endured by the specimens ranged from
PZT patch undergoes the development surface charges, a phenomenon
400,000 to over 11million respectively. Two modes of failure were
referred to as the ‘direct effect’. Similarly, when subjected to an electric
observed – (a) flexural failure, marked by the failure of the tensile re-
field, it undergoes mechanical strains, the phenomenon being called the
inforcement; and (b) shear failure, marked by diagonal shear-crack
‘converse effect’. The application as dynamic strain sensor is based on

101
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

Along one third length of the beam Near the Notch (beam top)

D C

50 mm

30 mm
50 mm 30 mm

C
19 18 17 16 15 14 13 12 11 10 8 6 5 4 3 2 1
B A
D 9
7

Near the mid-span of the beam Near the Notch (Neam fron )
50 mm

50 mm
50 mm
B 50 mm A

Fig. 5. Fatigue cracks developed on RC beam between 20k and 900k cycles.

the direct effect. Once the CVS is embedded inside the RC structure can be conveniently employed in identifying the location and severity
during casting (Fig. 1a), the PZT element functions as a dynamic strain of damage, as demonstrated by the authors [21,23].
sensor. The voltage V measured across the terminals of the patch can be
expressed in terms of the axial strain S1 experienced by it as [23]
2.3. CVS as electro-mechanical impedance (EMI) sensors for local SHM
d h YE ⎞
V = ⎜⎛ 31 ⎟ S1 = k p S1 Apart from acting as dynamic strain sensors as described above, the
T
⎝ ε33 ⎠ (1) PZT patches can also serve as “damage sensors” in the EMI technique,
which has established its mantle for high resolution localized SHM of a
where h is the thickness of the patch, d31 (refer Fig. 1b for directions of
wide variety of engineering structures and systems [26–35]. Several
axes “1” and “3”) the piezoelectric strain coefficient, Y E = Y E (1 + ηj )
new low-cost variants of the technique have also emerged [36,37]. In
the complex Young’s modulus of elasticity of the patch at constant
principle, the EMI technique is similar to the conventional global vi-
electric field and ε33 T T
= ε33 (1−δj ) its complex electric permittivity at
bration techniques for SHM but differs in few practical aspects. The
constant stress, η and δ respectively representing the mechanical and
major distinction is with respect to the excitation frequencies em-
the dielectric loss factors. kp is the overall strain-to-voltage calibration
ployed, being 30–400 kHz in the case of the EMI technique against a
factor of the patch. The output voltage can be easily measured by any
few Hz for the global vibration techniques [29]. Secondly, the PZT
standard oscilloscope or a high resolution digital multimeters and
patch serves as both sensor and actuator simultaneously in the EMI
converted into strain with the aid of Eq. (1).
technique, in contrast to being sensor alone in the case of the global
For a structural component undergoing flexural vibrations, a pair of
vibration techniques. Actuation is achieved in the EMI technique by
CVS instrumented at the extreme fibres of a section (Fig. 1a) can serve
electrically exciting the PZT patch through an LCR meter. The patch,
as a dynamic curvature sensor. Here, the polarities of the terminals of
housed inside the CVS, induces deformations in the host structure,
the two patches are so combined as to result into an addition of the
whose response is then acquired by the same patch acting as a sensor,
respective voltages when the member vibrates under flexure. For such a
captured by the same LCR meter as the current response [20]. The final
combination, the curvature of the flexural member can be expressed in
outcome of this process is the collocated admittance of the patch,
terms of the top and the bottom piezo strains (Stop and Sbott respectively)
consisting of the conductance (real part) and the susceptance (ima-
and hence the net voltage Vo generated across the terminals of the pair
ginary part). When plotted as function of frequency, these measure-
can be expressed as [24]
ments constitute a unique “signature” of the structure, to be altered
Stop + Sbott Vo only by the occurrence of any physical change in the structure, such as a
ϕ= =
D′ kp D′ (2) damage. Hence, at any future point of time, whenever the condition of
the structure is desired to be assessed, the signature can be acquired
where D′ (slightly less than the overall depth of the section, see Fig. 1(a) again and compared with the baseline signature. Consistency of the
is the distance between the centrelines of the top and bottom piezo signature indicates the well being of the structure whereas any change
elements. Employing the standard procedure of experimental modal in the signature indicates an occurrence of damage. Thus, the signature
analysis [25], such curvature sensors can be directly employed to ob- virtually acts as a finger print of the structure. Due to the use of high
tain the curvature mode shape of the flexural component, cir- frequencies of the order of kilohertz (and the consequent invoking of
cumventing any numerical computation as in the case of accelerometers the local modes of vibration), the EMI technique has damage sensitivity
[23]. Curvature is recognized as a sensitive parameter for damage de- far greater than the conventional global vibration techniques. It is
tection in the global dynamic techniques. The curvature mode shape especially proven to be expedient for detecting occurrence of localized

102
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

30 mm
30 mm

30 mm
30 mm

(a) (b)

20 mm
20 mm

(c)
Fig. 6. Fatigue cracks developed on beam after 8.0 million cycles near the notch area at the (a) front side, (b) back side and (c) bottom of the RC beam.

damages in concrete at the very incipient stage, as demonstrated by effective velocity, u̇eff .
authors through numerous studies on lab sized/real-life sized structures The effective drive point impedance of the host structure, Zs,eff, can
[28,29]. also be defined on similar lines, by applying a distributed force on the
Liang et al. [38] proposed the impedance approach to model the surface of the host structure along the boundary of the proposed loca-
PZT-structure electro-mechanical interaction in 1D structures. Bhalla tion of the PZT patch. Solving the governing 2D wave equation and
and Soh [39] extended the formulations to 2D structures by introducing introducing new correction factors, Bhalla and Soh [40] derived a
the concept of ‘effective impedance’. The related physical model is modified expression for the complex electro-mechanical admittance Y
shown in Fig. 2 for a square PZT patch of length ‘2l’ and thickness ‘h’. In as
the effective impedance approach, a distributed force ‘f’ per unit length,
2 2
varying harmonically with time, is considered to act along the l 2 ⎡ T 2d31 YE 2d31 YE ⎛ Za,eff ⎞ ⎤
Y = G + Bj = 4ωj ⎢ε33 − + ⎜ ⎟T
boundary of the patch, as opposed to a concentrated force in Liang’s h (1−ν ) (1−ν ) ⎝ Zs,eff + Za,eff ⎠ ⎥ (4)
⎣ ⎦
approach [38]. This distributed force causes harmonic planar de-
formations in the PZT patch. The ‘effective mechanical impedance’ of where ν is the Poisson’s ratio and ω is the angular frequency. The term
the patch can be defined as T is complex tangent ratio, ideally equal to {tan(κl)/κl}, with
κ = ω ρ (1−ν 2)/ Y E being the 2D wave number. Correction factors C1
→
̂
∮ f ·nds and C2 were introduced [40] to realistically model the behaviour of the
S F
Za,eff = = PZT patch, modifying the expression for T as
̇
ueff ̇
ueff (3)
1 ⎛ tan(C1 κl) tan(C2 κl) ⎞
where n ̂ is a unit vector normal to the boundary and ‘F’ represents the T = ⎜ + ⎟

2 ⎝ C1 κl C2 κl ⎠ (5)
effective force causing deformation of the PZT patch. ueff = δA/po is the
‘effective displacement’, with δA denoting the change in the surface The correction factors can be determined from the signatures of the
area of the patch and po its perimeter in the undeformed condition. PZT patch in ‘free-free’ condition prior to its bonding on the host
Differentiation of effective displacement with respect to time yields the structure. Further, the corrected mechanical impedance of the PZT

103
S. Bhalla, N. Kaur
RMSD (%)
6
3
S2 S5 S6 S7
16
14
12
10
RMSD (%)
0
S2 S5 S6 S7
30K
Fig. 7. RMSD plot in 50–250 kHz range along beam length. (a) At 20k cycles (b) At various stages between 30k and 250k cycles.
patch was derived as [39,40]
2h Y E
Za,eff =
jω (1−ν ) T
As can be observed from Eq. (4), any damage to the structure (i.e.
any change in mechanical impedance ‘Zs,eff’) will manifest itself as a
deviation in the admittance value, thereby providing an indication of
the damage. Next subsection summarizes the literature related to the
application of the EMI technique for monitoring fatigue.
2.4. EMI technique for fatigue monitoring
Giurgiutiu et al. [41] reported the first ever application of the EMI
technique to monitor spot welded single lap joints under fatigue. They
tested a miniature lap joint made of aluminium plates,
25.4 × 167 × 2 mm in size, and reported good sensitivity as well as
localization of damage using the surface-bonded PZT patches acting as
EMI sensors. They employed a wide frequency range of 200–1000 kHz,
relied on the real part of the electro-mechanical impedance (inverse of
the admittance Y , see Eq. (4)) and computed the statistical root mean
square deviation (RMSD) as a measure of the fatigue induced damage.
On application of cyclic loading to the joint specimen, the RMSD index
underwent an initial rise followed by a plateau region, thereafter pro-
viding very limited additional information on progression of damage. In
addition, the use of frequencies in excess of 500 kHz possibly rendered
the sensing zone of the patches limited to their very vicinity [27]. Af-
shari et al. [42] observed that the changes occurring in conductance
signatures of the PZT patches bonded to components undergoing fa-
tigue cracks exhibited non-linear trends. They introduced non-linear
feature extraction to analyse conductance signatures for fatigue related
International Journal of Fatigue 113 (2018) 98–112
S9 S11 S12 S14 S15 S16 S18
(a)
S9 S11 S12 S14 S15 S16 S18
50K 70K 100K 150K 200k 250k
(b)
damage diagnosis. They successfully demonstrated the advantages of
the non-linear feature extraction to monitor fatigue crack growth in
aluminium beam specimens, 304.8 × 25.4 × 3.2 mm in size, achieving
(6) a detection of fatigue cracks much prior to the conventional approach.
Lim and Soh [43] demonstrated fatigue life estimation of a 1D alumi-
nium beam using the EMI technique. Based on the experimental results,
they proposed a semi-analytical damage model by incorporating line-
arly elastic fracture mechanics into finite element modelling. All the
above experimental investigations, however, were largely confined to
very small structures.
Bhalla et al. [44] extended the EMI technique to monitor fatigue
damage in large real-life steel joints using an altogether different ap-
proach. They showed that the use of statistical indices such as RMSD
produces inconsistent conclusions. In general, fatigue damage can be
best monitored by observing changes in specimen’s stiffness as a func-
tion of the number of loading cycles. This however is difficult without
removing the component from the structure. To circumvent this pro-
blem, they proposed an equivalent stiffness parameter (ESP) based
approach for quantifying fatigue induced damage and estimating the
residual service life. The ESP could be obtained from the EMI signatures
of the bonded patch through signature decomposition based computa-
tional procedure [40]. They demonstrated the ESP based approach on
three real-life sized bolted steel double-lap joints, without warranting
in-situ measurement of the actual residual stiffness of the component.
The approach based on the ESP was a clear improvement over statistical
indicators, such as RMSD. All the three specimens exhibited reasonably
good correlation between the loss of the absolute specimen stiffness and
the loss of the PZT identified ESP, justifying the use of the latter for
remaining life assessment. The empirical correlations were derived
between the remaining life of specimen and the loss of ESP in non-
104
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

100
90
80
70

RMSD (%)
60
50
40
30
20
10
0
S2 S5 S6 S7 S9 S11 S12 S14 S15 S17 s18
(a)

21
19
17
15
RMSD (%)

13
11
9
7
5
3
S2 S5 S6 S7 S9 S11 S12 S14 S15 S17 s18
30K 50K 70K 100K 200k 250k

(b)
Fig. 8. RMSD plot in 1–1000 kHz range along beam length. (a) At 20 k cycles (b) At various stages between 30k and 250k cycles.

dimensional terms. However, the global dynamic vibration capability of
PZT patches was not touched upon. Dixit and Bhalla [22] extended the
equivalent parameter approach for fatigue damage assessment of plain
concrete specimens, 150 × 150 × 150 mm in size, subjected to fatigue
damage in laboratory conditions.
This paper investigates the possibility of the ESP based approach in
conjunction with PZT based global dynamic measurements for low-
strain (high-cycle) fatigue damage prognosis of RC structures utilizing
embedded CVS. As pointed out before, the fatigue process is somewhat
different and more complex in a heterogeneous material like concrete.
This is the first comprehensive fatigue related study on a large sized RC
structure. The following sections of the paper cover details of the ma-
terial and instrumentation details of the real-life sized RC specimen, its
testing under low-strain fatigue and the results of the damage prognosis
approach based on overall global stiffness and the piezo-identified ESP.
3. Specimen fabrication, instrumentation and experimental
details
3.1. Fabrication and instrumentation
The fatigue investigations covered in this paper were performed on
a real-life sized RC beam shown in Fig. 3(a). Fig. 3(b) provides a see
through model of the symmetrical right half of the beam for clarity. The
beam had a cross-section of 210 × 160 mm and an overall length of 4 m
(3.8 m measured centre to centre of supports). The beam was cast using
M 40 grade self-compacting concrete with 30% fly ash. The beam had
two 16 mm diameter bars on top and two 12 mm diameter bars on
bottom of high yield deformed (HYD) steel of characteristic strength
415 MPa.
The beam was cast with two layers of CVS flushing with its top and
bottom surfaces, totalling 19 pairs, equally spaced along the length (see
Fig. 3b). Originally, the beam had been the test specimen for study
encompassing combined energy harvesting and SHM by the authors
[21]. Table 1 lists the key parameters of the PZT patches used in the
CVS [45,46]. In the earlier study [21], a notch was created between
CVS locations 7 and 8 (See Fig. 3a) at the time of casting (later filled up
with concrete) so as to ensure a controlled damage location. The da-
mage induced in the earlier study involved three stages: (i) chipping off
concrete from the notch; (ii) cutting a bottom reinforcement bar; and
(iii) cutting the other reinforcement bar. After the investigations were
over, the reinforcement bars were welded together and the notch again
filled with concrete, so that it was ready for the fatigue related in-
vestigations reported in this paper. Entire measurement-cum-fatigue
damage induction process is described in detail in the following.
For the global dynamic technique, the natural frequency of the
beam at any point of time during the test was measured by an impact
test, employing the CVS 16 (top) as the dynamic strain sensor (Fig. 3a).
This CVS was connected to Agilent 34411A digital multimeter in turn
controlled by a personal computer operating in VEE PRO 9.3 interface.
The dynamic response of the CVS, acquired at a sampling rate of 1 milli
second, was transformed into frequency domain automatically in the
VEE PRO environment, from which the first natural frequency was
identified as the peak of the curve. Fig. 4 shows a typical frequency
domain plot. The first natural frequency of the retrofitted beam before
the commencement of the test was measured as 17.2 Hz.
The beam was then excited using LDS V406 series portable shaker
specially configured to operate in the inertial mode [47]. Sinusoidal
signals were generated using Agilent 33210A function generator in turn
amplified using LDS PA500L power amplifier before being fed into the
shaker. The gain of the amplifier was so adjusted as to achieve an ac-
celeration of 3.14 m s−2 at the centre of the beam, which corresponded

105
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

0.005

CR=0.62
(State 28)

0.004
(State 1)
CR=0
Conductance (S) CR=0.16
0.003 (State 19)

CR=0.10
(State 14) CR=0.37
(State 26)
0.002
CR=1.00
(State 36)
CR=0.04
0.001 (State 10)

0.000
50 70 90 110 130 150 170 190 210 230 250
Frequency (kHz)
(a)

0.014

0.012

CR=0
0.010
CR=0.37,0.62
Conductance (S)

CR=0.16
0.008

(State 36)
0.006 CR=1.00
CR=0.04, 0.10
(State 10, 14)
0.004

0.002

0.000
50 150 250 350 450 550 650 750 850 950 1050
Frequency (kHz)
(b)
Fig. 9. Variation of conductance signatures for CVS 7 with damage progression for frequency range (a) 50–250 kHz and (b) 50–1000 kHz (CR = Cycle ratio).

to a force of 75 N by the shaker in accordance with predetermined
computations [46]. The frequency of excitation was maintained equal
to the natural frequency of the beam (17.2 Hz) for maximum impact.
Measurement of strain by the pair of CVS located at number 8 position
(see Fig. 3) using Eq. (1) revealed a peak strain of slightly less than
50 µm/m. The vibrations were halted at various stages during the test,
as listed in detail in Table 2 for measurement of damage related signals
from the PZT patches. The intermediate states are referred to as State
1–36, with the cumulative number of cycles of loading totalling to
8.04 million. At each of the States 1–36, the current natural frequency
of the beam was determined by impact test as described before and this
served as the excitation frequency for the next loading. This measure-
ment served to employ the CVS as dynamic strain sensor.
For the EMI technique, the admittance signatures (comprising of
conductance and susceptance) of all the bottom layer CVS, were ac-
quired at all intermediate states (see Table 2). in two different ranges:
(a) 50–250 kHz @ an interval of 100 Hz; and (b) 1–1000 kHz @ an
interval of 1 kHz, using Agilent 4980A LCR meter. These served to carry
out local SHM using the EMI technique. In general, there is no hard and
fast rule about choosing the frequency range and the associated fre-
quency interval. A range of 30–400 kHz is in general preferred by the
EMI community based on experience [27]. As can be seen, the first
range conformed to this guideline. However, the second range was
additionally chosen for comparative purposes, however, with a larger
interval so as to ensure a reasonable signature acquisition time. During
each halt (States 1–36), the sides and the top and the bottom surfaces of

106
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

18 after this state, when very widespread cracks suddenly appeared at

16 8 million cycles (State 31), with the natural frequency sharply reducing
14 to 13.2 Hz, the highest observed rate of drop with regard to the number
RMSD (%)

12 of cycles since the commencement of the experiment. The new cracks at

10 this stage were found to be mainly confined to the notch area. Fig. 6
8 shows the view of the beam from point of view of cracks at various
6 locations after 8 million cycles (State 31). A very widespread and in-
4 tense crack pattern can be observed at this stage, consisting of a large
2 number of secondary cracks offshooting from the previously developed
0 cracks at the notch location (See Fig. 5, zoom ‘A’), which is the hub of
0.0 0.2 0.4 0.6 0.8 1.0 1.2
the fatigue damage culmination. The load cycles were continued till
CR
8.04 million cycles (State 37) and the frequency dropped to 9.88 Hz at
(a) this stage. In between, marginal increment in frequency was observed
25 at some instances during the halts (at night times), again indicating
temporary recovery. The beam had undergone a very severe level of
20 damage till this point and further application of the load cycles was
therefore discontinued. The rebars were found intact till this stage. The
RMSD (%)

15 shaker assembly was dismantled from the beam. Attempts by a suc-

10
5 from the point of the EMI signatures and damage prognosis from a
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
CR
(b)
Fig. 10. Variation of RMSD with loading cycles for CVS 7. (a) Frequency range
50–250 kHz. (b) Frequency range 1–1000 kHz.
the beam were carefully inspected for the occurrence of any cracks,
which were marked using a permanent marker and identified by a
number tag indicating the cumulative number of cycles (in thousands,
using symbol ‘k’) endured by the specimen till that crack length.
3.2. Experimental observations
Fig. 5 presents a view of cracks observed on the beam up to 900k
cycles. The first fatigue crack was observed after 20k cycles (State 2),
emanating from the notch area (zoom ‘A’ of Fig. 5). The figure also
presents further development of cracks during the test, up to 900k cy-
cles (State 15). The very first crack in the notch area continued to grow
further, albeit in an inclined fashion, till it reached the top of the beam.
An additional crack was observed to develop from the bottom at 20k
cycles at a location near the shaker, growing up to the mid depth of the
beam till 900k cycles (see zoom ‘B’ of Fig. 5). Some cracks were also
observed on the top of the beam at the notch location, commencing 50k
cycles (State 4) as visible in zoom ‘C’ of Fig. 5. Another isolated crack
appeared on the front face of the beam after 120k cycles (intermediate
of States 6 and 7), as visible in zoom ‘D’ of Fig. 5. As can be observed
from Table 2, the first natural frequency of the beam marginally re-
duced from 17.2 Hz to 16.5 Hz till 900 k cycles (State 15). The top and
the bottom peak strains at the level of the CVS at location 7 (see Fig. 3)
were consistently found to be under 50 µm/m, as determined from the
measured CVS voltage using Eq. (1).
No other major crack was observed after 900k cycles until 7000k
cycles (State 30). In between, the natural frequency was observed to
reduce in a linear fashion from 16.5 Hz (900k cycles) to 15.7 Hz (2450k
cycles i.e. State 24). After State 24, the test was halted for a week. After
the rest period, an unusual observation was that the natural frequency
increased to 16.6 Hz at State 25. This observation shows fatigue damage
recovery capability of concrete, a fact earlier reported by Stroeven [6].
In general, after State 26, the natural frequency continued to decline at
a very small rate (See Table 2), the value after 7 million cycles (State
30) being 15.8 Hz. Dramatic events started happening immediately
ceeding group to carry out a modal analysis of the beam by impact test
led to rupture of the bottom reinforcement bars at the location of the
weld. The forthcoming sections present the observations of the CVS
combination of the global dynamic and the local EMI techniques.
4. Results and discussion
The admittance signatures of the CVS residing in the bottom layer
were acquired using Agilent 4980A LCR meter before commencing the
experiment and also at all the intermediate states during the test (see
Table 2). As mentioned before, two different frequency ranges were
chosen: (a) 50–250 kHz at an interval of 100 Hz; and (b) 1–1000 kHz at
an interval of 1 kHz. In order to quantify the changes in signature re-
lative to the baseline, root mean square deviation (RMSD) index was
computed at each intermediate state for all bottom CVS using [28],
N
∑i = 1 (Gi1−Gi0 )2
RMSD (%) = × 100
N
∑i = 1 (Gi0 )2 (7)
where Gi1 is the post-damage conductance at the ith measurement point
and Gi0 is the corresponding pre-damage value. Fig. 7(a) shows the plot
of the RMSD index along the length of the beam for the 50–250 kHz
range after 20 k cycles (State 2), when the first crack was observed (see
Fig. 5). The purpose of such histogram is to identify damage location
[28], the RMSD value is expected to be highest for the CVS located in
the vicinity of the damage. On observing together with Fig. 3, the
maximum RMSD value is exhibited by CVS 7, which is the location of
the notch. In addition, CVS 12 and CVS 14 also exhibit comparable high
values, somewhat before the corresponding crack actually became
visible at State 7 (See Fig. 5 zoom ‘D’). Fig. 7(b) presents a compilation
of the RMSD plots at other cycles, varying from 30k cycles to 250k
cycles. It may be noted from the figure that damage localization
somewhat appears to be widespread now (not just confined to the notch
region), a fact which can be corroborated with Fig. 6. Fig. 8(a) and (b)
similarly present the RMSD histograms for the other range, that is,
1–1000 kHz. Here, it can be observed that CVS 6, just one sensor away
from exact damage location, is exhibiting the maximum value. For
higher cycles (Fig. 8b), the damage localization is very accurate and
better than the corresponding plot for the other frequency range (see
Fig. 7). In addition, the RMSD magnitude is also higher than the other
range (see Fig. 6). Hence, the 1–1000 kHz range proves to be better
from damage sensitivity as well as damage localization considerations.
It should however be noted that few of the sensors (such as CVS 8) were
not found operational and could not be interrogated for the sake of the
above analysis.
Fig. 9 shows the variation of the conductance signatures for CVS 7,
the sensor closest to the main damage location (the notch region, see

107
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

State 1

State 25

State 30

State 15

Normalized Stiffness
State 24
State 31

State 37
I II
III

Cycles (in Millions)

CR
(a)
Normalized Stiffness

IA
IB
IC
0.6
Cycles (in Millions)

CR

(b)
Fig. 11. Variation of normalized stiffness with cycles. (a) Complete plot for 8.04 million cycles. (b) Initial region for 4 million cycles.

Figs. 3, 5 and 6), in the two frequency ranges. The host material being
concrete, the signatures are somewhat dominated by the PZT patches’
own peaks rather than the structural peaks [31]. In both the frequency
ranges, discernible shifts in signatures can be clearly observed. In
overall, it can be concluded that the CVS, operating as EMI sensors,
facilitated a fair level of fatigue damage localization, especially at in-
cipient level of damage.
The RMSD or the raw signature based damage analysis present
herein, however, does not provide any idea about the severity of da-
mage nor any solid information about the residual stiffness so that some
idea about remaining life can be made. This is exemplified by Fig. 10,
which shows the variation of the RMSD index of CVS 7 (closest to da-
mage location) with increasing number of load cycles, indicated in the
graphs non-dimensionally as cycle ratio CR, the number of cycles ap-
plied up to a particular damage state normalized with respect to the
total number of cycles (8.04 million). Except for the first 10–15% of the
load cycles (CR = 0.1–0.15), the index does not seem to follow a con-
sistent trend. It falls down thereafter and remains more or less constant
in magnitude in both the frequency ranges. Hence, it can at the best
used for capturing incipient level damage and the corresponding da-
mage localization. These observations match with those observed
during earlier fatigue tests on steel joints [44]. The next sub-section
presents a deeper analysis based on global stiffness estimated by the
same set of CVS operating as global vibration sensor, in conjunction
with piezo-identified stiffness parameter, designated as ESP.
4.1. Fatigue damage prognosis using stiffness parameters
As mentioned before, in addition to EMI signature acquisition, the
embedded CVS were employed to obtain the first natural frequency of
the test beam through the impact test. From the first natural frequency,
the flexural rigidity, EI, was worked out using [48]
4ρAf12 L4
EI =
π2 (8)
where ρ is the density of the beam, A the cross-sectional area, f1 the
measured first natural frequency and L the length of the beam. Fig. 11
shows the observed variation of EI, normalized with respect to the in-
itial value, as a function of the loading cycles (which are shown in the x-
axis in absolute terms as well as in normalized form, as cycle ratio, CR).
An overall assessment of the plot (Fig. 11a) shows three distinct re-
gions: Region I of high rate of stiffness degradation (up to 2.5 million
cycles or CR equal to 0.31); followed by Region II of very low rate of
stiffness degradation (up to seven million cycles, or CR equal to 0.87);

108
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

50 0
Experimental (CR=0.04)
45
Equivalent (CR=0.04)
Equivalent (CR=0)
40 Experimental (CR=1) -5
Equivalent (CR=1) Experimental (CR=0)
35

30 -10 Equivalent (CR=1)

x (Ns/m)

y (Ns/m)
25 Experimental (CR=1)

20 -15
Equivalent (CR=0)
15
Experimental (CR=0)
10 -20
Equivalent (CR=0.04)
5
Experimental (CR=0.04)
0 -25
140 150 160 170 180 190 200 140 150 160 170 180 190 200
Frequency (kHz) Frequency (kHz)
(b) (c)
100 0

90
Equivalent (CR=1)
-10
Experimental (CR=1)
80
-20
70
Equivalent (CR=0)
-30
60 Experimental (CR=1)
y (Ns/m)

Equivalent (CR=0)
x (Ns/m)

50 Experimental (CR=1) -40

40 -50 Equivalent (CR=0)

30 Experimental (CR=0)
-60
20
-70
10

0 -80
200 210 220 230 240 250 200 210 220 230 240 250
Frequency (kHz) Frequency (kHz)
(d) (e)
Fig. 12. Identification of system parameters by CVS. (a) Identified structural system (series k-m-c system). (b) x vs frequency (140–200 kHz). (c) y vs frequency
(140–200 kHz). (d) x vs frequency (200–250 kHz). (e) y vs frequency (200–250 kHz).

and finally, a very steep zone represented by Region III. The overall
average slopes (in terms of normalized stiffness to normalized cycles) of
the first, second and the third regions were worked out to be 1, 0.07 and
4.67, respectively. On further closer look (Fig. 11b), one can find that
the Region I is in turn composed of three distinct sub-regions: IA,
having steep slope; IB, with almost zero slope; and IC, having an in-
termediate slope. The slope of Region IA was determined to be 2.05
where as that of Region IC to be 1.19. Further, the stiffness recovery
during the rest period of over one week (between States 24 and 25) can
also be observed in the curves as a distinct peak. The relative stiffness
loss at the end of the S-curve is about 90% in contrast to 20% reported
by Mu et al. [7] for plain concrete under compression fatigue. This
shows much higher fatigue ductility of RC as compared to plain con-
crete, which is as expected.
The paper further explores the possibility of employing the CVS
identified stiffness parameter, namely ESP, and its relation with the
overall stiffness (as covered above) for fatigue induced damage as
previously demonstrated on prototype steel joints [44]. For this
purpose, the real and the imaginary components of the EMI based ad-
mittance signature were further processed to obtain the real (x) and the
imaginary (y) components of the drive point mechanical impedance of
the structure (based on Eq. (4)), using the computational procedure
outlined in detailed in the related publication [44]. After this exercise,
within the frequency range of 50–250 kHz, a close observation of the
trend of the components, x and y, indicated that the system behaviour
matched that of a series spring-mass-damper system, shown in
Fig. 12(a). For such a system, typically, the plot of y verses frequency is
straight line with negative slope [24], which was indeed the case
(Fig. 12c and e). The identified series spring-mass-damper system is in
contrast to the parallel spring-damper system observed for the steel
joints in the earlier investigations [44]. Following expressions hold true
for x and y in terms of the parameters k, c and m for such a scenario [24]
c −1
x=
c −2 + (ω/ k−1/ ωm)2 (9)

109
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

Region III
Region IA

(a)

(b)
Fig. 13. Variation of normalized ESP with cycles. (a) Complete plot for 8.04 million cycles. (b) Initial region for 1 million cycles.
K/K (Based on actual flexural stiffness)

4.0 Solving Eqs. (9)–(11) simultaneously, system parameters, c, m and K

(which is indeed the ESP), can be derived at any frequency as [22]
3.5
x 2 + y2
3.0 c=
x (12)
2.5
cy (ω2−ωo2)
2.0 m=
(x −c ) ωωo2 (13)
1.5
cy (ω2−ωo2)
1.0 K=
(x −c ) ω (14)
0.5
Fig. 12(b)–(e) shows the comparison between the experimental
0.0 plots of x and y with the equivalent analytical values worked out using
0 5 10 15 20 25 30 the above equations for CVS 7.
K/K (Identified by CVS) Satisfactory agreement can be observed between the experimental
and the equivalent analytical curves. As pointed out above, the para-
Fig. 14. Correlation between loss of absolute stiffness with loss of PZT identi-
fied stiffness. meter “K” in Eq. (14) is the ESP. Fig. 13 shows a plot of the normalized
ESP as a function of the number of load cycles. On comparing with
Fig. 11, we find that the initial and the final trends of the curve are
−(ω/ k−1/ ωm)
y= similar to regions IA and III respectively of the plot of normalized actual
c −2 + (ω/ k−1/ ωm)2 (10) (global) stiffness (see Fig. 11), obtained through the impact test using
the same CVS as dynamic strain sensor. The normalized slope of Region
In addition, at frequency ω = ωo, at which y = 0 (that is the y plot IA was computed here to be 4.35 against 2.05 for the actual global
crosses the x-axis), is given by [22] stiffness, reflecting same order of magnitude. However the intermediate
region shows a high degree of scatter. This is in contrast to the ob-
k = mωo2 (11) servations on steel joints [44] where ESP showed a consistent behaviour

110
S. Bhalla, N. Kaur International Journal of Fatigue 113 (2018) 98–112

45
14
40
35 12

30 10

S = K/K
S = K/K 25 8
20
6
15
4
10
5 2

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
L=1-N/N0 L=1-N/N0
(a) (b)
30
14
25
12

10 20
S = K/K

S = K/K
8
15
6
10
4
5
2

0 0
0.00 0.10 0.20 0.30 0.40 0.50 96 97 98 99 100
L=1-N/N0 L=1-N/N0
(c) (d)
Fig. 15. Variation of stiffness loss with non-dimensional remaining life for. (a) Actual flexural stiffness. (b) CVS identified stiffness (for L = 1–100). (c) CVS identified
stiffness (for L = 0–0.44). (d) CVS identified stiffness (for L = 96.23–100).

throughout the duration of the test. Non-dimensional form of the de-
gradation of ESP was determined as
δK
δS =
K (15)
Similarly, the non-dimensional form of absolute stiffness can be
obtained by replacing K in the above equation by EI obtained by using
Eq. (8). Fig. 14 shows the plot of the non-dimensional stiffness para-
meter based on actual stiffness loss versus the same determined using
the ESP for Region IA. Fair degree of correlation can be observed be-
tween the two parameters. The non-dimensional ESP based stiffness is
about 7 times higher than the actual non-dimensional flexural stiffness.
This provides an empirical conversion factor between the ESP and the
actual stiffness. Fig. 15(a) plots the non-dimensional actual stiffness loss
as a function of non-dimensional remaining service life, expressed as
N This paper has presented a new and effective prognosis paradigm for
L = 1−
No
where N represents the number of load cycles to a particular damage
state and No (=8.04 million) the total number of cycles to failure. The
curve has a typical S-shaped trend very similar to the S curve of
Fig. 11(a). Fig. 15(b) shows the corresponding curve based on ESP. It
can be observed that a large number of outliers can be observed if we
attempt to fit an S shape out of the data points based on ESP. However,
the initial and the final positions of the curve show a linear trend, as
shown in Fig. 15(c) and (d) respectively. Hence, on overall, the ESP
provides fair degree of representation to the actual stiffness in the initial
and the final stages of fatigue, but not in the intermediate stages. Unlike
steel joints [44], ESP alone cannot be relied upon for damage prognosis.
On the other hand, the overall stiffness of the structure, determined
using the same CVS as global dynamic sensor, provides a fair degree of
damage prognosis, especially for moderate to severe fatigue damage
level. Unlike steel joints, determination of EI is quite practical and ex-
pedient for RC structures such as bridges, since it entails only the
measurement of frequency, which is straight forward. The EMI tech-
nique, as a compliment, is expedient in damage localization, especially
during the incipient stages of fatigue. It also provides warning of col-
lapse in terms of steep decline of ESP in the final stages of the fatigue
damage. Hence, operating in the dual mode (both global as well as local
EMI sensors), the CVS can facilitate sound prognosis of fatigue induced
damage in RC structures under flexure.
5. Conclusions
low-strain (high-cycle) fatigue damage of RC structures using em-
(16)
bedded CVS, utilizing them in both global as well as local vibration
modes. In the local mode, the CVS, operating as EMI sensor, are very
good in detecting the occurrence as well as the location of fatigue in-
duced damage at the incipient stage. Of the two frequency ranges in-
vestigated, namely 50–250 kHz and 1–1000 kHz, the second range
tends to be more sensitive as well as facilitating a more realistic damage
localization. The same CVS, functioning as dynamic strain sensor, is
found to be very expedient in fatigue damage prognosis, relating the
overall stiffness loss to the remaining service life. The equivalent stiff-
ness, identified by the CVS in the EMI mode (in terms of ESP) is found to
correlate well with the actual stiffness in the first and the third region of
the fatigue S-curve. Hence, it can be employed in the beginning and the

111
S. Bhalla, N. Kaur
end stages, that is, for detecting fatigue initiation and predicting the
imminent collapse. However, it is not found to perform very well during
the intermediate stage, that is, the second part of the S-curve. Hence,
both approaches need to be operationalized in synergy for optimal fa-
tigue damage prognosis. Determination of the natural frequency alone
is often not too difficult for RC structures like bridges, where moving
loads themselves provide the necessary excitation. There is further
scope for exploring the efficacy of the prescribed paradigm under low-
cycle fatigue. Also, further experimental investigations can be done to
evaluate PZT patches bonded directly to the reinforcement bars.
Acknowledgements
This research did not receive any special grant from funding agen-
cies in the public, commercial or not for profit sectors. The authors
gratefully acknowledge the assistance provided by the undergraduate
students, namely Abhimanyu, Abhinav Bindal and Bhrigu Joshi during
the performance of some of the experiments. Thanks are also due
Anjaneya Dixit, for assistance in literature review and preparing Fig. 2,
as well as to Pranjal Khandelwal, Teaching Assistant, for assistance in
editing manuscript and some of the graphs.
References
[1] Encyclopaedia Britannica. < https://www.britannica.com/ > ; 2018 [accessed 08
March 2018].
[2] Ellyin F. Fatigue damage crack growth and life prediction. London: Chapman and
Hall; 1997.
[3] Oh BH. Fatigue analysis of plain concrete in flexure. J Struct Eng-ASCE
1986;112(2):273–88.
[4] Chang D, Chai WK. A study on the fatigue strength behaviour of reinforced concrete
structures. Int J Pres Ves Pip 1989;40:51–75.
[5] Wikipedia. < https://en.wikipedia.org/wiki/Fatigue_(material) > ; 2018. [ac-
cessed 08 March 2018].
[6] Stroeven P. Low cycle compression fatigue of reinforced concrete structures.
Procedia Eng 2010;2:309–14.
[7] Mu B, Subramaniam KV, Shah SP. Failure mechanism of concrete under fatigue
compressive load. J Mater Civil Eng, ASCE 2004;16:566–72.
[8] Medeiros A, Zhang X, Ruiz G, et al. Effect of the loading frequency on the com-
pressive fatigue behaviour of plain and fiber reinforced concrete. Int J Fatigue
2015;70:342–50.
[9] Nieto AJ, Chicharro JM, Pintado P. An approximated methodology for fatigue tests
and fatigue monitoring of concrete specimens. Int J Fatigue 2006;28:835–42.
[10] Oneschkow N. Fatigue behaviour of high-strength concrete with respect to strain.
Int J Fatigue 2016;87:38–49.
[11] Kolluru SV, O’ Neil EF, Popovics JS, et al. Crack propagation in flexural fatigue of
concrete. J Eng Mech-ASCE 2000;126(9):891–8.
[12] Raithby KD. Flexural fatigue behaviour of plain concrete. Fatigue Eng Mater
1979;1979(2):269–78.
[13] Gaedicke C, Roesler J, Shah SP. Fatigue crack growth prediction in concrete slabs.
Int J Fatigue 2009;31:1309–17.
[14] Simon KM, Kishen JMC. A multiscale approach for modeling fatigue crack growth in
concrete. Int J Fatigue 2017;98:1–13.
[15] Jadallah C, Bagni C, Askes H, Susmel L. Microstructural length scale parameters to
model the high-cycle fatigue behaviour of notched plain concrete. Int J Fatigue
2016;82:708–20.
[16] Mai SH, Le-Corre F, Foret G, Nedjar B. A continuum damage modelling of quasi-
static fatigue strength of plain concrete. Int J Fatigue 2012;37:78–85.
[17] Makita T, Bruhwiler E. Tensile fatigue behaviour of ultra-high performance fiber
reinforced concrete combined with steel rebars (R-UHPFRC). Int J Fatigue
2014;59:145–52.
[18] Wikipedia. 2013 Savar building collapse, < https://en.wikipedia.org/wiki/2013_
Savar_building_collapse > [accessed 09 April 2018].
[19] Bhalla S, Gupta A. A novel vibration sensor for concrete structures. Patent
Application no. 1011/DEL/2011, Invention Disclosure no. FT/IPR/CE/SB/2007/
112
International Journal of Fatigue 113 (2018) 98–112
0570, Foundation for Innovation and Technology Transfer (FITT), IIT Delhi, India;
2007.
[20] Shanker R, Bhalla S, Gupta A. Integration of electro-mechanical impedance and
global dynamic technique for improved structural health monitoring. J Intel Mater
Syst Struct 2010;21(3):285–95.
[21] Kaur N, Bhalla S. Combined energy harvesting and structural health monitoring
potential of embedded piezo concrete vibration sensors. J Energy Eng-ASCE
2015;141(4):1–18. (D4014001).
[22] Dixit A, Bhalla S. Prognosis of fatigue and impact induced damage in concrete using
embedded piezo-transducers. Sens Actuators A: Phys 2018;274:116–31.
[23] Shanker R, Bhalla S, Gupta A. Dual use of PZT patches as sensors in global dynamic
and local EMI techniques for structural health monitoring. J Intel Mater Syst Struct
2011;22(16):1841–56.
[24] Bhalla S, Moharana S, Talakokula V, Kaur N. Piezoelectric materials: applications in
SHM, energy harvesting and bio-mechanics. London: Wiley-Athena; 2017.
[25] Ewins DJ. Modal testing: theory, practice, and application. 2nd ed. England:
Hertfordshire, Research Studies Press Ltd; 2000.
[26] Ayres JW, Lalande F, Chaudhry Z, et al. Qualitative impedance-based health
monitoring of civil infrastructures. Smart Mater Struct 1998;7(5):599–605.
[27] Park G, Cudney HH, Inman DJ. Impedance-based health monitoring of civil struc-
tural components. J Infrastruct Syst, ASCE 2000;6(4):153–60.
[28] Soh CK, Tseng KKH, Bhalla S, et al. Performance of smart piezoceramic patches in
health monitoring of a RC bridge. Smart Mater Struct 2000;9(4):533–42.
[29] Bhalla S, Soh CK. Structural impedance based damage diagnosis by piezo trans-
ducers. Earthq Eng Struct Dyn 2003;32:1897–916.
[30] Giurgiutiu V, Zagrai A, Bao JJ. Damage identification in aging aircraft structures
with piezoelectric wafer active sensors. J Intel Mater Syst Struct 2004;15:673–87.
[31] Soh CK, Bhalla S. Calibration of piezo-impedance transducers for strength predic-
tion and damage assessment of concrete. Smart Mater Struct 2005;14(4):671–84.
[32] Lim YY, Bhalla S, Soh CK. Structural identification and damage diagnosis using self-
sensing piezo impedance transducers. Smart Mater Struct 2006;15(4):987–95.
[33] Yang YW, Bhalla S, Wang C, et al. Monitoring of rocks using smart sensors. Tunn
Undergr Sp Technol 2007;22(2):206–21.
[34] Moharana S, Bhalla S. A continuum based modeling approach for adhesively
bonded piezo-transducers for EMI technique. Int J Solids Struct
2014;51(6):1299–310.
[35] Talakokula V, Bhalla S, Ball RJ, et al. Diagnosis of carbonation induced corrosion
initiation and progression in reinforced concrete structures using piezo-impedance
transducers. Sensor Actuat A-Phys 2016;241:79–91.
[36] Peairs DM, Park G, Inman DJ. Improving accessibility of the impedance-based
structural health monitoring method. J Intel Mater Syst Struct 2004;15(2):129–39.
[37] Bhalla S, Gupta A, Bansal S, et al. Ultra low-cost adaptations of electro-mechanical
impedance technique for structural health monitoring. J Intel Mater Syst Struct
2009;20(8):991–9.
[38] Liang C, Sun FP, Rogers CA. Coupled electro-mechanical analysis of adaptive ma-
terial systems-determination of the actuator power consumption and system energy
transfer. J Intel Mater Syst Struct 1994;5(1):12–20.
[39] Bhalla S, Soh CK. Structural health monitoring by piezo-impedance transducers:
modeling. J Aerosp Eng-ASCE 2004;17(4):154–65.
[40] Bhalla S, Soh CK. Structural health monitoring by piezo-impedance transducers:
applications. J Aerosp Eng-ASCE 2004;17(4):166–75.
[41] Giurgiutiu V, Reynolds A, Rogers CA. Experimental investigation of E/M impedance
health monitoring for spot-welded structural joints. J Intel Mater Syst Struct
1999;10(10):802–12.
[42] Afshari M, Park S, Inman DJ. The early stage crack detection using non-linear
feature extraction of self-sensing piezoelectric impedance measurements. In: Proc
7th Int Workshop on Str Health Monit, Stanford University, 9–11 September 2009.
p. 1870–1885.
[43] Lim YY, Soh CK. Fatigue life estimation of a 1D aluminum beam under mode-1
loading using the electro-mechanical impedance technique. Smart Mater Struct
2011;20(12):125001.
[44] Bhalla S, Vittal APR, Veljkovic M. Piezo-impedance transducers for residual fatigue
life assessment of bolted steel joints. Struct Health Monit 2012;11(6):733–50.
[45] PI Ceramic. < https://www.pieramic.com > ; 2018 [accessed on 08 March 2018].
[46] Bhalla S. A mechanical impedance approach for structural identification, health
monitoring and non-destructive evaluation using piezo-impedance transducers [Ph.
D. Thesis]. Singapore: School of Civil and Environmental Engineering, Nanyang
Technological University; 2004.
[47] Kaur N. Integrated structural health monitoring and energy harvesting potential of
adhesively bonded thin piezo patches operating in d31 mode [Ph.D. Thesis]. India:
Department of Civil Engineering, Indian Institute of Technology Delhi; 2015.
[48] Chopra AK. Dynamics of structures. India: Prentice Hall of India Limited; 2017.