Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Review
Structural Health Monitoring

2022, Vol. 21(2) 2 28–249
A review of structural health The Author(s) 2021

Article reuse guidelines:
monitoring of bonded structures using sagepub.com/journals-permissions

DOI: 10.1177/1475921721993419
journals.sagepub.com/home/shm
electromechanical impedance
spectroscopy
A Francisco G Tenreiro1 , António M Lopes2 and Lucas FM da Silva2
Abstract
The article presents a literature review of electromechanical impedance spectroscopy for structural health monitoring,
with emphasis in adhesively bonded joints. The concept behind electromechanical impedance spectroscopy is to use vari-
able high-frequency structural vibrations with piezoelectric elements to monitor the local area of a structure for changes
in mechanical impedance that may indicate imminent damage. Various mathematical models that correlate the structural
impedance with the electric response of the piezoelectric sensors are presented. Several algorithms and metrics are
introduced to detect, localize, and characterize damage when using electromechanical impedance spectroscopy.
Applications of electromechanical impedance spectroscopy to study adhesive joints are described. Research and devel-
opment of alternative hardware for electromechanical impedance spectroscopy is presented. The article ends by pre-
senting future prospects and research of electromechanical impedance spectroscopy–based structural health monitoring,
and, while advances have been made in algorithms for damage detection, localization, and characterization, this technol-
ogy is not mature enough for real-world applications.
Keywords
Structural health monitoring, electromechanical impedance spectroscopy, piezoelectric wafer active sensors, damage
algorithms, adhesive joints
Introduction appears in a progressive or sudden manner, such as

fatigue solicitations or impact conditions. If no action
Analyzing and monitoring the performance of a given is taken, damage propagation may occur until failure
structure, without using destructive evaluation meth- is attained and the structural system can no longer
ods, are the underlying ideas of structural health moni- operate.1 However, one cannot conclude that all
toring (SHM). In this field, non-destructive testing material defects are damage in the structure, as some
(NDT) tools are used to in situ continuously monitor a material imperfections may not negatively impact its
structure and the measured signals are processed by a performance.2
damage identification algorithm to determine the pres- In general, an SHM setup includes a set of sensors
ence of a given fault. The implemented algorithm has and actuators. The actuating elements are connected to
the objective of detecting the existence of damage, its a function generator that creates the input signal
location, type, and the degree of severity to the struc- excitation, while the sensors are connected to a
ture’s well-being.1–6 SHM is a vast field of study with
many applications, such as aeronautic,7,8 aerospace,9 1
xão em Engenharia Mecânica e Engenharia
Instituto de Ciência e Inovac
naval,10 and civil structures.11–13 Industrial (INEGI), Porto, Portugal
When studying damage identification and charac- 2
Departamento de Engenharia Mecânica, Faculdade de Engenharia
terization, one must know what constitutes a damage (FEUP), Universidade do Porto, Porto, Portugal
for the monitoring setup. In this context, damage is
the change of material and/or geometric properties in Corresponding author:
António M Lopes, Departamento de Engenharia Mecânica, Faculdade de
the observed system, which may negatively impact the Engenharia (FEUP), Universidade do Porto, Rua Dr. Roberto Frias, s/n
structural performance. In Material Science and 4200–465, Porto, Portugal.
Engineering, this is referred to as a defect, and it Email: [email protected]
Tenreiro
2 et al. 229
Structural Health Monitoring 00(0)
data-acquisition equipment that stores and processes inexpensive, and can be implemented without nega-
the measured signals. In some cases, passive SHM sys- tively impacting the structure’s performance; (b) the
tems make use of only sensors, which measure ambient high frequencies used are immune to ambient noise and
signals to detect damage, meaning that there is no need vibrations commonly present in real-world applica-
for actuation elements and associated power driving tions; and (c) with a wide frequency spectrum, one can
systems. Nonetheless, both setups will always require a detect minute damage foci.4,13 In fact, EMIS is consid-
central unit to properly process the measured signals ered such a robust measuring method that it is also
for de-noising, compensate for environmental factors, starting to be used for industrial monitoring of system
and extract damage signature.5,7 In this last stage, dam- properties, such as fluid viscosity monitoring.22
age identification, localization, and characterization It must be noted that SHM systems for adhesively
algorithms, based on physical modeling or statistical bonded joints, where two mechanical components are
and machine learning techniques14 are required. joined together with a structural adhesive, are uncom-
A variety of NDT methods have been used in the mon and in an early stage of research. Furthermore,
scope of SHM to monitor and evaluate structural traditional NDT technologies used for adhesive con-
integrity, such as guided waves,5,15,16 acoustic emis- nections, such as the Ultrasound, Acoustic Emissions
sions,17,18 and infrared thermography,19 among others. and Infrared Thermography are time-consuming and
Particular attention was paid to several vibration-based incapable of detecting all types of damage.23–27 Among
SHM methods, where one can monitor changes in the the various defects present in adhesive joints, weak
system’s stiffness, mass, and energy dissipation adhesion, alternatively known as kissing-bonds or zero-
properties.20,21 volume unbound, is extremely difficult to detect with
One promising vibration-based SHM method is the conventional NDT methods.28,29 Weak adhesion is pre-
so-called electromechanical impedance spectroscopy sented as a defect in the interface between the adhesive
(EMIS). EMIS makes use of the coupled electromecha- and the adherend, where there may be physical contact
nical behavior of piezoelectric wafer active sensors between the two, but there is no bonding of the adhe-
(PWASs), which can work both as actuators or sensors. sive on the component to be connected.28,30 Therefore,
An electric AC voltage, generated by a function genera- the electromechanical impedance method has great
tor, is sent to the PWAS, acting as an actuator, and potential for damage detection and characterization for
converts the electric voltage signal into a mechanical adhesively bonded joints.
stress solicitation. At the same time, the PWAS acts as Although several reviews were published on this
a sensor, converting the structure’s mechanical response topic,4,31–33 the following article presents an updated lit-
to an electric signal, which is then sent to a data- erature review of the EMIS-based monitoring research,
acquisition unit, such as a computer or an impedance with focus on adhesively bonded joints, and is orga-
analyzer. This experiment is tested for a wide range of nized as follows. Section ‘‘Theoretical background of
frequencies, usually going from 1 kHz onward. A sche- EMIS’’ introduces a theoretical background regarding
matic representation of the following setup in adhesive the electromechanical coupling that occurs in the elec-
joints is presented in Figure 1. This online monitoring tromechanical impedance method. Section ‘‘Algorithms
technique seems advantageous over other methods for damage detection, localization, and quantification’’
since (a) the used transducers are light-weight, presents an overview of algorithms and indices used to
detect, locate, and quantify damage in a given structure.
Section ‘‘EMIS-based monitoring of adhesive joints’’
presents a review of the research of electromechanical
impedance monitoring applied to adhesively bonded
joints. Section ‘‘Hardware for impedance measure-
ment’’ gives an overview of the hardware equipment
used with EMIS measurements. Section ‘‘Challenges
for EMIS-based SHM’’ discusses the main challenges
and trends in EMIS research. Finally, the concluding
remarks are presented in section ‘‘Conclusion.’’
Theoretical background of EMIS

EMIS is an NDT method that takes advantage of the
Figure 1. Schematic representation of the electromechanical fact that PWAS elements can convert electrical signals
impedance setup in a bonded joint. into a mechanical actuation and vice versa. This is due
230
Tenreiro et al. Structural Health Monitoring 21(2)
3
Figure 3. Model of interaction between the piezoelectric

element and the monitored structure made by Xu and Liu.40
Figure 2. Model of interaction between the piezoelectric
element and the monitored structure made by Liang et al.35
excitation, F, and the velocity of a particle excited by
to electromechanical coupling property of piezoelectric the load, x_ , so that, Z = F=_x.
elements. In other words, this element can generate an This model was developed on the fact that the actua-
electric charge when subjected to a mechanical stress, or tor force is not constant under variable frequency excita-
can generate a mechanical strain when under the influ- tion, as first established by the static and early dynamic
ence of an electric field, as described by finite element approaches.36 This model formed the basis
of extensive research on structural monitoring.
S1 = �sE11 T1 + d31 E3 ð1Þ Giurgiutiu and Zagrai37 used the model presented in
�
D3 = d31 T1 + eT33 E3 ð2Þ equation (4) to detect axial and flexural resonance fre-
quencies. Sun et al.38 tested various algorithms based on
where S1 represents the mechanical strain, T1 is the this model to determine the system frequency behavior.
mechanical stress, E3 denotes the electric field, D3 is Zhou et al.39 extended the presented model and pro-
the electrical displacement (charge per unit area), �sE11 is posed a two-dimensional (2D) system, with the struc-
the mechanical compliance at zero electric field, d31 tural impedance depending on the position.
stands for the piezoelectric induced strain coefficient, Xu and Liu40 improved the 1D model in equation
�
and eT33 is the dielectric constant at zero stress.3,4,34 These (4), by adding a bonding layer between the PWAS and
relations constitute the backbone of all modeling of the monitored structure, as per Figure 3, obtaining
EMIS.

Liang et al.35 developed a one-dimensional (1D) wP lP T� ZP (v)d31 Y�11
E
tan (klP )
Y (v) = jv � E
e33 � d31 Y11 +
model for the electromechanical coupling, based on hP ZP (v) + jZs (v) klP
equations (1) and (2), where a piezoelectric element is
ð5Þ
connected to a structure of stiffness, k; mass, m; and
damping coefficient, c, as presented in Figure 2. The 1 1
researchers demonstrated that the electrical admittance, j= = ð6Þ
Ks �ms v2 + jvcs + ks
Y (v) (or the electric impedance, Z(v)), as a function of 1+ 1+
Kb �mb v2 + jvcb + kb
the angular frequency, v, is related to the structure’s
mechanical impedance, Zs , as per equation (4) where j is a parameter relating the structure’s dynamic
stiffness, Ks , with that of the bond layer, Kb , as
Z(v) = Y (v)�1 ð3Þ described by equation (6). The parameter j modifies
the dynamic behavior of the structure. If j = 1, then one
wP lP T� Zs (v)
Y (v) = jv e33 � d31 Y�11
E
ð4Þ obtains the model formulated in equation (4), which
hP Zs (v) + ZP (v)
means that Kb is much higher than Ks and perfect adhe-
The variables wP , lP , and hP are the width, length, sion of the PWAS is achieved. In practice, this may be
and thickness of the piezoelectric element, ZP corre- difficult to achieve, since bonding operations occur
sponds to the piezoelectric element’s mechanical impe- manually and are inherently susceptible of variations.
dance, and Y�11
E
is the complex Young’s Modulus E of the Therefore, the quality of the adhesive greatly influences
piezoelectric material, defined as Y�11 E
= Y11 (1 + jh), the obtained electrical admittance.
where h is the mechanical loss factor. The dielectric Zagrai and Giurgiutiu41,42 proposed a 2D model of
�
constant at zero stress, eT33 , is affected
by a dielectric an isotropic circular plate coupled with a PWAS, where
T�
loss factor, d, such that e33 = eT33 (1 � jd). It should be both the axial and flexural components of the natural
noted that both Zs and ZP are mechanical impedances vibrations are considered. The model was validated
that are defined as the ratio between the load with experimental results, and can be written as
Tenreiro
4 et al. 231
model using the concept of effective mechanical

impedance.
Pavelko48 developed an electromechanical impe-
dance model for PWAS–structure interaction that can
be applied for 1D, 2D, or 3D cases. However, such a
model becomes more complex, and, consequently, cum-
bersome for SHM applications. Furthermore, this
model only presents adequate results for large struc-
tures and looses accuracy for high frequencies.
Annamdas and Soh49–51 conceived a semi-analytic 3D
electromechanical model for a broad class of structures,
Figure 4. Shear lag effect phenomenon due to the adhesive with any spatial distribution of PWAS and with the
bonding the PWAS to the structure.43 sensor’s mass also in consideration. However, the
( " #)�1
2 kp 2 (1 + n)J1 (ua )
Z(v) = jvC(1 � kp ) 1 + 3 ð7Þ
1 � kp 2 ua J0 (u) � (1 � n)J1 (ua ) � x(v)(1 + n)J1 (ua )
where k represents the wave number, n is Poisson’s model is quite complex, requiring numerical data to
coefficient, kp is the planar coupling factor, and calculate model parameters.
ks (v) Several authors developed models of electromechani-
x(v) = denotes the dynamic stiffness ratio cal coupling between a specific structure and the PWAS,
kP
between the structure, ks (v), and the piezoelectric ele- such as for beams structures,52–55 axial rods,54,56 shell,57
ment, kP . This model was defined with Bessel functions, and plaque geometries.58 However, no analytical model
J0 (u) and J1 (u), of order zero and one, respectively, relating the mechanical impedance of adhesively bonded
vr structures with PWAS was made so far. While this
where u = , c being the sound speed in the PWAS
c approach provides a good estimation of the electrical
and r being the radius. The variable ua is defined for impedance measurements at high frequencies, one is lim-
the disk radius ra . ited to structures with simple geometries. Furthermore,
Bhalla et al.43 proposed a simplified 2D impedance in order to accurately predict the measured electrical
model considering shear lag effect, as presented in equa- impedance of a damaged structure, the model used must
tion (8). This effect occurs in the adhesive layer that include the influence of the damage.
connects the piezoelectric sensor to the monitored struc-
ture, and is defined as the lag (or difference) in displace- Algorithms for damage detection,
ment between the structure’s and the PWAS’s surfaces, localization, and quantification
as shown in Figure 4. The model states that the electri-
cal admittance, Y (v), is given by The objective of SHM is to detect the presence of damage
in a given structure, pinpoint its location, classify the type
� 2 E 2 E
� ��
lP2 T 2d31 Y11 2d31 Y11 tan (klP ) ZP, eff
Y (v) = 4jv e � + 3 ð8Þ
hP 33 1�n 1�n klP Zs, eff (v) + ZP, eff (v)
where ZP, eff and Zs, eff are the effective mechanical impe-
dance of the piezoelectric element and of the structure, of damage, evaluate its severity, and establish a prognosis
respectively, which combine the bidirectional impe- on the structure’s lifetime based on the data collected.
dance behavior. It must be mentioned that Bhalla and These tasks require advanced statistical and algorithmic
Soh44 previously developed a similar model, but their approaches that may not be able to fully accomplish all
approach differed from this one since they did not con- of these tasks. As such, most research on SHM is focused
sider an averaged shear strain along the structure’s on developing more robust and complex algorithms. This
length (x direction in Figure 4). Moharana and Bhulla45 section reviews statistical metrics and algorithms for
refined the model, similar to that of Equation (8), where EMIS-based structural monitoring for performing dam-
inertial and shear terms are included. Wang et al.46 and age detection, localization, and characterization.
Zuo et al.47 extended this approach and developed a However, no discussion on algorithms that estimate the
three-dimensional (3D) electromechanical impedance remaining useful life of the structure is presented.
232
5
Damage detection be null.60–62 Oliveira et al.63 adopted the Principal

When the properties of the monitored structure change Component Analysis to process time-domain readings.
due to the presence of some damage, so does the mea- Then, these results were used to calculate the RMSD. It
sured impedance signature. Therefore, once the impe- was verified that they were more sensitive than those
dance signatures are acquired, metrics and algorithms obtained directly with EMIS. The mean absolute per-
are required to determine whether there is damage in centage deviation, MAPD, evaluates the average of the
the structure. Most research makes use of damage absolute values of the deviations at each individual data
metrics/indices, which allow the comparison of impe- point of the signatures. If the deviations between spec-
dance (or admittance) readings of a pristine structure tra are small, then MAPD is nearly zero, thus reflecting
(as a baseline/reference) with the in situ impedance that there is no damage.62,64 The Covariance, Cov, eval-
readings. Various metrics have been proposed in the lit- uates the relationship between two impedance signa-
erature. The most commonly used are the root mean tures and determines whether these two impedance
square deviation, RMSD; the mean absolute percentage datasets resemble each other or not. If both signatures
deviation, MAPD; the Covariance, Cov; the Correlation peak at the same frequency, then the Covariance is pos-
Coefficient, CC; and the Correlation Coefficient itive. Conversely, if the peak of a dataset is placed
Deviation, CCD, defined as where the valley of the other dataset is, then the
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Covariance is negative. If both signatures are unrelated,
u vf then covariance tends to be null. The CC reflects the
uP 2
u ðRe(Zi ) � Re(Zi0 )Þ same information as the Covariance, but the value is
u i = vi
RMSD = u
u vf ð9Þ normalized so that, CC nears 1 when both signatures
t P 2
Re(Zi0 ) are similar, and is equal to �1 when they are opposite
i = vi
to each other. The correlation coefficient deviation,
vf � � CCD, also represents the same information as the corre-
1X �Re(Zi ) � Re(Zi0 )�
� �
MAPD = ð10Þ lation coefficient, CC. However, it better reflects how
N i = vi � Re(Zi0 ) �
much different the two impedance spectra are: if they
vf
�� are similar, CCD nears zero.61,62
1X �
Cov = Re(Zi ) � ZRe Re(Zi0 ) � ZRe
0
ð11Þ While these are the most commonly used damage
N i = vi indices, other metrics have been developed. Tinoco and
Cov Marulanda65,66 developed the ellipse damage index,
CC = EDI, which correlates the impedance readings from a
sZ s0Z
reference/pristine structure, Z 0 , and a damaged struc-
v
Pf � ��
Re(Zi ) � ZRe Re(Zi0 ) � ZRe 0 ture, Z. These impedance readings can be described by
1 i = vi two Gaussian distributions, one for the reference val-
= sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
N P vf � �2 P vf � �2 ues, N (mZ 0 , s2Z 0 ) and another for the measured readings
Re(Zi ) � ZRe Re(Zi0 ) � ZRe 0
N (mZ , s2Z ). From the covariance matrix that relates
i = vi i = vi
these two distributions, one can determine the eigenva-
CCD = 1 � CC ð13Þ lues, lmin and lmax , that describes an ellipsis, as repre-
sented in Figure 5. From these parameters, one can
where Zi0 is the baseline electrical impedance of the
define EDI as
pristine structure, Zi is the measured electrical impe-
dance, and N is the number of acquired samples. The plmin lmax
0 EDI = 2:4477 ð14Þ
variables ZRe and ZRe are the average of the measured
max (Z 0 ) max (Z)
and pristine impedance, while sZ and s0Z are the stan-
dard deviation of the measured and pristine impedance, Enciu et al.67 developed a normalized entropy index,
respectively. It is noted that, usually, these indices H, based on the statistical concept of entropy, which is
are calculated using only the real component of the the logarithm of a probability, and is defined as
electrical impedance, since it is more sensitive to dam-
age.3,33 However, research has shown that, after some vf
P
pi log2 pi
processing, the imaginary component can be used as i = vi
well.59 H= � ð15Þ
log2 n
The root mean square deviation, RMSD, which is
based on the euclidean norm, allows one to see, for a where n is the number of datapoints in the impedance sig-
given frequency range, if the two impedance spectra are nature, and pi is a probability associated with the electri-
similar or not. If they are similar, then RMSD tends to cal impedance measurement at frequency vi , defined as
Tenreiro
6 et al. 233
where a0 denotes the incident wave/vibration amplitude, at

is the received wave/vibration amplitude after propagation
from the actuator has occurred, and m(v) is the acoustic
attenuation coefficient. Since this method requires spatial
discretization, a meshing of the component being moni-
tored is necessary, and, therefore, the variable xi, j trans-
lates the distance between the transducer at position i and
the center of the mesh j. The value of acoustic attenuation
Ai, j is used to calculate a position dependent damage
index, DIj . The following damage metrics were proposed
" �#�1
N �
X RMSDi
DIj(RMSD) = ð18Þ
i=1
Ai, j
" �#�1
N �
X Df i
DIj(Df ) = ð19Þ
i=1
Ai, j
Figure 5. Ellipse obtained from the Covariance matrix and
respective eigenvalues.65 where RMSDi is the calculated root mean square devia-
tion for the signatures obtained with sensor i, Df i is the
Re(Zi )
pi = vf ð16Þ averaged shift in frequency peaks for obtained signa-
P
Re(Zi ) tures from sensor i, and N stands for the total number
i = vi of piezoelectric sensors. These damage indices were
used to detect small impact damage in a composite I-
The difference between a baseline entropy index, H 0 , beam (1D geometry) and a clamped composite plate
and the entropy index of the damaged specimen impe- (2D geometry). It was verified that the algorithm was
dance readings allowed to determine the presence of able to detect and locate the damage foci.
damage in a circular aluminum plate. Kralovec et al.72 developed a triangulation algo-
Other approaches based on statistical and machine rithm for damage localization in an aluminum plate,
learning algorithms have been proposed. Based on the based on both experimental and finite elements method
work of Sohn and Farrar,68 Park et al.69 developed an (FEM) numerical results. Magnets were used to simu-
unsupervised learning auto-regressive model with exo- late damage. These experiments determined the impe-
genous inputs in the frequency domain for damage dance spectrum for baseline and damaged plates (21
identification in structures. de Oliveira et al.70 devel- damages were considered). Then the RMSD and MAPD
oped a convolutional neural network (CNN) that was values were calculated. These indices are function of
able to detect the presence of one of three possible dam- the distance between PWAS and damage location. As
age foci in an aluminum plate. Their results provided such, it was found that the results could be fitted using
100% accurate results in the Testing phase when using either an exponential law or a two-term Hankel func-
three epochs, but also require a longer training time tion of the first kind. These empirical relations were
period than other methods. able to detect other damage foci.
Dragan et al.73 combined the reconstruction algo-
Damage localization rithm for probabilistic identification of damage
(RAPID) imaging algorithm with the EMIS transfer
For a more robust damage algorithm, detection may impedance approach to locate damage. In EMIS, trans-
not be sufficient to characterize the defect, especially fer impedance is an experimental procedure where one
for large structures. It is also of importance to locate PWAS acts as a vibration actuator, and another
the focus of damage. Hence, damage localization algo- PWAS acts as vibration sensor. The impedance transfer
rithms are of interest. function, T (v), is defined as
Cherrier et al.71 developed a damage detection and
localization algorithm, where acoustic attenuation, Ai, j , Us (v) jUs j ju(v)
was used as a tool for damage localization, and is T (v) = = e = jT (v)jeju(v) ð20Þ
Ua (v) jUa j
defined as
where Us and Ua are the measured voltage signals from
at the PWAS sensor and the PWAS actuator, respectively,
Ai, j = = e�m(v)xi, j ð17Þ
a0 and u(v) is the phase shift between the two signals. The
234
7
RAPID algorithm, which is extensively used for Lamb damages. However, its performance was unsatisfactory
wave–based SHM,74,75 calculates a damage intensity for 225 mm2, since the experimental data used for this
probability, I, which can be generically calculated as case had damaged surfaces with a larger area.
Loendersloot et al.75 used a combination of EMIS
X
Na X
Ns and Lamb wave to detect and localize damage caused
I= DI(a, s)Ra, s ð21Þ by fatigue loading in glass-fiber reinforced polymer. In
a=1 s=1 a first stage, damage detection was performed using
EMIS readings from PWAS. It was found that both
where Na and Ns are the total number of actuators and
the RMSD and MAPD indices allowed the detection of
sensors, respectively, DI is the obtained damage index
damage and index values increased with the number of
for the combination of actuator a and sensor s, and Ra, s
load cycles. Afterward, Lamb waves were generated
is a function that represents the range between actuator
and the RAPID algorithm was used to locate the dam-
a and sensor s. The impedance transfer function, T (v),
age with accuracy.
is used as the damage index, DI(a, s), to calculate the
Rabiei et al.78 developed an EMIS-based algorithm for
damage intensity probability.
damage detection and localization on a replica of an alu-
Giurgiutiu and Kropas-Hughes76 performed the train-
minum mobile bridge structure. In a first step, the dam-
ing and validation of a probabilistic neural network
age indices RMSD, CC, and CCD were used to detect
(PNN) to detect damage in a 2024T3 aluminum plate of
damage. Afterward, the damage location map was made
100 mm diameter, where the sensor was placed at the
using a mixture of Gaussian kernels, where impedance
center. Two high-frequency bands were considered: 10–
spectra for the various PWASs were inputted. Gonsalez
40 kHz and 10–150 kHz. In this approach, the research-
et al.79 performed a variability analysis using an ANOVA
ers defined five groups of damage: pristine (no crack),
approach and Tukey’s multiple comparison tests. RMSD
slight damage (crack at 40 mm from PWAS), intermedi-
values were obtained from impedance signatures of an
ate damage I (crack at 25 mm from PWAS), intermedi-
aluminium aeronautical panel instrumented with eight
ate damage II (crack at 10 mm from PWAS), and strong
PWAS. Additionally, six structural scenarios were consid-
damage (crack at 3 mm from PWAS). First, a feature
ered: pristine condition, three cases with a loosened bolt
extraction algorithm was developed that extracted the
in different locations, one consisting of three bolts simul-
resonant peak frequencies and amplitudes, as well as the
taneously loosened, and one where all the bolts were re-
corresponding log-amplitudes. However, only the reso-
tightened. The ANOVA approach proved to be effective
nant peak frequencies were considered as input for the
in detecting and locating the damage. However, this
PNN. This was essential, since it allowed for dimension-
method requires the determination of the frequency band-
ality reduction of the problem at hand, and, conse-
width over which the impedance signatures are obtained.
quently, a more efficient and less time-consuming PNN
training procedure. In the 10–40 kHz frequency range, 4
frequency features (major peaks only) were not enough Damage quantification
for accurate classification of the damage location, while Given the high and ever-increasing complex computing
6 and 11 features provided good results. For the 10–150 advancements, algorithms have been developed that
kHz range, 22 frequencies were required in order to accu- can detect, locate, and characterize damage. These will
rately classify the damage in the aluminum disk. be presented here.
Selva et al.77 also developed a PNN for damage loca- Khodaei and Aliabadi80 developed a three-layered
lization on a carbon fiber–reinforced polymer (CFRP) algorithm to detect, localize, and quantify, to some
instrumented with three PWASs. FEM simulation extent, the damage in a curved composite fuselage panel
results obtained from various models were used, with with composite stiffeners. Their approach made use of
damage being placed at various locations. In their work, guided waves and the electromechanical impedance
the simulated damage corresponds to an impact damage method, and a network of 16 PWAS elements. In the
caused by a drop-weight impact tower with a mass of first level of the diagnosis algorithm, the existence of
2 kg and energy of 20 J. Various metrics were consid- damage was evaluated by processing Lamb wave mea-
ered for the input of the PNN, such as correlation coef- surements from the sensor network, using the maxi-
ficient of Re(Z), area subtraction of Re(Z) and Im(Z), mum value of the energy envelops difference, E(Rij ),
RMS or quadratic mean of Re(Z) and Im(Z), and root between the pristine and the current states, given by
mean square deviation using Re(Z) and Im(Z).
� �
Afterward, the neural network was tested with experi- E(Rij ) = �Rij � jH(Rij )� ð22Þ
mental data results to see if it was properly trained. The
PNN provided good results for locating damage with where Rij (t) is the residual that is calculated as the
600 mm2 of area, and placed correctly three of five energy envelope difference between the pristine state
Tenreiro
8 et al. 235
and the measured state, and H(Rij ) is the Hilbert trans- aircraft panel with and without window, with defects such
form of the residual signal. Various thresholds were as cracks and rivet loss as present in the structures in
defined to determine whether damage was present and question. Concurrently, the authors also developed a
which path between sensors best detects the damage. The Gustafson-Kessel fuzzy cluster algorithm to classify the
second level of the algorithm considered the EMIS signa- type of damage. Both algorithms were able to detect and
tures to determine an approximate localization of the classify the damage with minimal error if the sensors are
damage. To easily process the signatures, the correlation placed near the damage, hence also presenting an approxi-
coefficient, CC, was used to determine which sensors mated form of damage localization. However, no accurate
were the most sensitive to the presence of damage. damage localization is performed. De Moura and
Afterward, in the third level of the algorithm, an accurate Steffen84 developed a Statistical Meta-modeling technique
damage localization and characterization was performed. of surface responses involving Correlation Coefficients,
The two most sensitive sensors were selected to execute CC, of impedance measurements from an aluminium air-
triangulation of the damage. In order to choose a sensor plane fin. These results were used to characterize damage,
for accurate localization, various guided wave paths were such as the radial distance from the fault to the sensor,
analyzed using the energy envelops difference, as in Level and the severity of damage. Their experimental proce-
1, where the previously chosen sensors from Level 2 dures were also used to train a PNN to detect damage
acted as actuators, and the others were considered as sen- without the need for determining the impedance signal
sors. With the third sensor chosen, the weighted energy peaks in the monitored frequency band.
arrival method (WEAM) was used to accurately detect Zahedi and Huang85 developed an interesting
the location and size of the damage, and is defined as approach based on the combination of EMIS and Lamb
waves. The excited waves are generated by a Hanning
1X N X N
windowed tone-burst signal, like traditional Lamb waves
DI(x, y) = E(Rij (tij (x, y)))wij (s, n) ð23Þ
N i=1 j=1 SHM method, but the same piezoelectric elements that
excite the structure also measure the response. Afterward,
where E(Rij (tij (x, y))) is the energy envelope determined both the electric admittance, Y (v), and the current of the
in equation (22), for coordinates x and y and time t, and generated signal, I(v), are used to calculate the frequency
wij (s, n) denotes a window centered at the first peak of domain response of the PWAS, O(v), such that
the residual signal to reduce the effect of reflections and
mode superposition. The presented algorithm led to O(v) = Y (v)I(v) ð24Þ
accurate results both in location of damage and in esti-
which is then converted to the time-domain by using
mation of damage size.
the inverse Fast Fourier Transform (FFT). Finally, the
Fan et al.81 developed a damage identification, loca-
time-domain signal envelope, O(t, f ), is calculated with
lization, and quantification methodology using sparse
the Hilbert–Huang transform. With this procedure, the
information from resonant frequency shifts of impe-
amplitude can be represented as a function of frequency
dance spectra. Given that the number of measured fre-
and time. While this does not enable the detection or
quency shifts is smaller than the number of structural
direct characterization of damage, the transient EMIS
parameters, l1 regularization method is used to solve
time–frequency analysis allows the detection of damage
the underdetermined inverse problem. This method
and, in an indirect manner, the sensitivity of the PWAS
allowed for the detection and localization of damage,
to damage by measuring echoes.
and determined the severity of the damage, but its accu-
As a final remark, one can see that damage metrics are
racy is limited to small structures. For larger structures
commonly used in EMIS for damage detection, of which,
or structures with multiple damage foci, this technique
the most commonly used are the RMSD; the MAPD; the
underestimates the severity of the damage.
Covariance, Cov; the CC; and CCD. These can also be used
He et al.82 trained and tested an artificial neural net-
for more complex algorithms that are able to determine
work (ANN) to detect, localize, and quantify damage
damage localization or characterization. However, in this
in a truss structure. Damage was simulated by loosen-
stage, researchers tend to focus more on either complex
ing the screws to various levels, and thereby creating
physics-based model-driven approaches, or ANN to per-
various damage intensities. The measured impedance
form damage localization and/or quantification.
spectra were divided into various sub-bands, and RMSD
values were calculated for each portion of the spectrum.
EMIS-based monitoring of adhesive joints
These values are then inputted into the ANN, and three
output neurons indicated the damage location (i.e. the EMIS is an ever-growing field of research since the first
rod and joint where the screw is loosened) and severity. articles were published in the mid 1990s. However,
Palomino et al.83 trained a PNN for detection, localiza- research on its application to adhesively bonded joints
tion, and classification of damage in both an aluminum has only started in this last decade. This section
236
9
presents a review of the most relevant work performed where G0 (v) is the reference conductance, Gn (v) is
so far by the scientific community. conductance after the n th loading cycle, t is the fre-
quency lag, and � is the convolution operator. From
Experimentation this, the authors used An (v)=A0 (0) as a damage metric.
A normalized resonant frequency shift, DfR =fR , was
The backbone of any progress or innovative approach also used as a damage metric. Using FEM numerical
in SHM, in general, and in EMIS-based SHM, in par- simulations, the authors observed that both indices
ticular, has been the use of experimental testbeds for augmented slightly when the lack of adhesion origi-
concept validation and discovery. This subsection dis- nated in the extremities of the interface layer between
cusses the experimental work where EMIS is used to the adhesive and the adherend. Then the indices
monitor adhesive joints. reached a local minimum when the damage neared
In one of the earliest works, Na et al.86,87 monitored the PWAS, and, finally, the indices augmented expo-
adhesively bonded joints immersed in corrosive envir- nentially until failure occurred. However, this was
onments, such as water and acetone, using reusable only experimentally verified for DfR =fR . Despite not
PWAS sensors, where the piezoelectric element is reporting the difference of numerical and experimen-
attached to a neodymium magnet. Their findings tal behavior of the cross-correlation function ratio,
showed that this adapted sensor was able to detect An (v)=A0 (0), the researchers deduced that: (a) the
debonds occurring in the adhesive due to the nature of adhesive’s loss factor, hb , might influence conduc-
the environment. A comparative analysis between the tance signals due to environmental perturbations and
performance of these sensors and the traditional PWAS (b) the local minimum might occur near the SLJ stress
noted that, despite still being able to detect damage in failure. Either way, this damage metric is unreliable
the adhesive, sensitivity to damage was sacrificed at the for damage detection.
cost of measurability simplicity. Zhuang et al.90 studied the EMIS performance of a
Lim et al.88 performed a more comprehensive and modified epoxy-based adhesive SLJ with distinct levels
thorough monitoring research of epoxy-based adhesive of bad adhesion, and having piezoelectric sensors
cure for 21 days. First, an analysis of the PWAS reso- embedded inside the adhesive layer. SLJ specimen were
nances was made. For the first 6 h, as the adhesive cyclically loaded with increasing load steps of 111 N,
became more viscous, a dampening effect occurred and every time load ceased to be applied, impedance
reducing the PWAS resonance magnitude and frequency. measurements were performed. Damage index, RMSD,
Afterward, between 6 and 24 h, as the adhesive became of the measured impedance did not increase signifi-
progressively more solid, the frequency peak shape gra- cantly until the stress values were roughly at 80% to 90
dually altered, and several smaller peaks appeared due to % of the failure stress (with a threshold value of
the adhesive stiffness. At the same time, a rightward shift RMSD = 0:2). The parasitic effect of having an
of the piezoelectric resonance frequency was observed. embedded sensor in the adhesive was studied, conclud-
From 1 to 7 days, there is still some hardening of the ing that, given the small size of the sensor, no significant
adhesive, and a slight rightward movement of the PWAS effect on the mechanical performance was observed.
resonant frequency was observed. Finally, from 7 to This seems coherent with research on mechanical beha-
21 days, the hardening of the adhesive was almost indis- vior of adhesive joints with artificially placed voids91,92
cernible. One drawback of this approach was the fact which concluded that, for small defects, failure is con-
that, for high frequencies, the EMIS sensing zone was trolled by the adherends’ plasticity. Zhuang et al.93 also
highly localized and, therefore, the repeatability between performed fatigue testing of the SLJ specimen, and peri-
different sensors was less satisfactory. A second analysis odically stopped the test to perform EMIS monitoring.
approach using the structural resonance peaks as indica- They found that only at 75% of its full fatigue life (i.e.
tors yielded similar conclusions. These results were corro- at 36,000 of 48,000 load cycles), some change of the
borated with Lamb wave monitoring results. measured impedance spectrum occurred, and only at
Dugnani et al.89 studied the influence of bad adhe- 88% of its life (i.e. at 42,000 of 48,000 load cycles) did
sion, more commonly known as kissing-bonds, in the RMSD increase significantly.
epoxy-based adhesive single lap joints (SLJs) under sta- Malinowski et al.94,95 compared the electromechanical
tic testing conditions, where the SLJ was cyclically impedance readings of pristine joints with various defec-
loaded with incremental load differences, until failure. tive adhesive joints in a high-frequency range (3–5 MHz)
Using the cross-correlation function, An (v), defined as and used the RMSD damage index and the frequency shift
to measure the bond degradation. Three different defect
Z‘
causes were studied: (a) contamination with a silicon
An (v) = G0 (v) � Gn (v + 2pt)dv ð25Þ release agent, (b) moisture contamination, and (c) lower
�‘ adhesive curing temperature. They concluded that bad
Tenreiro
10 et al. 237
curing caused greater decrease in RMSD values and fre- and quantify both the debond in the structural adhesive
quency shifts. Meanwhile, weak adhesion caused by either layer (SD for structural debond), as well as the debond
release agents or moisture presence resulted in a big varia- in the adhesive layer connecting the PWAS to the adhe-
tions of RMSD and a small variation of frequency peaks. sive joint (TD for transducer debond). Impedance spec-
In a more recent study, Malinowski et al.96 per- tra from cases where TD, SD, and a combination of
formed mechanical, ultrasound, and EMIS testing on TD and SD, where the debonded lengths varied, were
adhesive joints with CFRP. In their experiments, the used to train the ANN. The neural network obtained
joint specimen were subject to (a) varying of thermal interesting results, especially in estimating the length of
treatment temperature, (b) contamination with potas- TD. However, in order to have better results, appar-
sium formate–based de-icing fluid, and (c) combined ently more data should have been used to train the
modification due to thermal treatment and de-icer con- ANN, in order to better estimate the length of SD.
tamination. Ultrasonic testing was unable to detect It must be noted, once again, that only in recent years
damage or bond weakening, but was able to detect has EMIS been used to study adhesive joints. While
delamination in the CFRP due to excessive temperature some research has proven that this method can detect
for thermal treatment. However, the EMIS monitoring issues like weak adhesion and debond, no systematic
in a high-frequency range (4.25–4.7 MHz) could unra- analysis of the effects in impedance signatures caused by
vel information regarding the bond quality. Indeed, the various defects in adhesive joints has been made.
reduction in RMSD values for each contaminated speci-
men could be correlated with a reduction in both mode
I and mode II fracture-toughness results. Simulation
Gulizzi et al.97 first studied the applicability of moni- Often, the EMIS behavior of a given structure is vali-
toring the adhesive’s curing process at ambient tem- dated or predicted using the finite elements approach.
perature by means of EMIS, using a simple data This allows to have a basic understanding of the struc-
acquisition setup, with a chirp signal exciting the tural response to EMIS monitoring and to optimize
PWAS, as proposed by Baptista and Filho.98 Their SHM parameters.103 This subsection, reviews the use of
findings revealed that, as the cure progresses, the adhe- FEM as a tool to study the EMIS behavior in adhe-
sive gains stiffness and the impedance amplitudes are sively bonded joints.
increased. In the first 6 h of curing, significant changes Zhuang et al.90 performed a direct steady-state linear
in the impedance spectrum were observed, and reso- dynamic type FEA simulation using ABAQUS to mini-
nant peaks started to appear. Afterward, changes in the mize the parasitic effect on the mechanical behavior of
impedance spectrum were less significant. In a second a bonded joint with an embedded sensor, and to opti-
study,99,100 the same authors analyzed the repeatability mize its ability to detect damage. Therefore, a para-
of monitoring the cure of adhesives, using the same metric analysis of the dimensions of the sensor was
setup. They concluded that no significant changes in made. The researchers concluded that reducing the
the impedance readings took place. They also found thickness of the sensor would optimize its sensitivity to
that the ratio of resin/hardener had a significant influ- damage, while an optimal diameter of 3 mm would
ence on the curing time and on the resultant stiffness. maximize its ability to detect bad adhesion.
As the percentage of resin in the mixture increased, Roth and Giurgiutiu104,105 proposed an FEM meth-
resulting in a higher ratio, the increment of RMSD with odology with two types of numerical simulations to
time became less significant. Furthermore, one could determine the best frequency range and location of the
observe when RMSD became constant and, conse- sensors on an adhesively bounded doubler, using
quently, when the curing stopped. ANSYS. In a first stage, a modal analysis determined
Cuc and Giurgiutiu101 studied the applicability of the best mode shapes to evaluate the presence of an
using EMIS for monitoring adhesively bonded titanium artificial debond in the adhesive. In order to avoid
C-sections for helicopter rotor blades. A comparison interpretation errors, a numerical algorithm was devel-
between pristine and debonded rotor blades revealed oped by the authors. A mode-matching index between
that significant changes in the frequency spectrum were the pristine and damaged simulation models was calcu-
observed in the lower frequencies (below 200 kHz). In lated, where the dot product of the normalized nodal
this region of the frequency spectrum, the resonant displacements for the pristine and debond model was
peaks shifted leftward to lower frequencies. calculated and then the mean of all dot products for
Rautela and Bijudas102 developed an EMIS-based each node is determined. To choose the pristine mode,
SHM system to detect damage in adhesively bonded which best matches a given debond mode, the maxi-
joints. Two ANN, one for FEM simulation and mum of the mode-matching index for the nth mode was
another for experimental results, are trained to detect determined. Afterward, local debond displacements
238
11
were compared through the root mean square devia-

tion, RMSD. The obtained results were then ranked in
descending order, thus easily detecting the mode shapes
of largest value. Once having the frequency range deter-
mined, a finite element sensitivity study both on the
pristine and debonded models estimated the best sensor
placement. The simulations results were finally verified
experimentally, both using the EMIS method as well as
with laser Doppler velocimetry.
Rautela and Bijudas102 compared the impedance
spectra obtained from a 1D analytic model, FEM
Figure 6. Voltage divider used by Peairs et al.106
simulations, and experimental results. Both numerical
simulations and experimental procedures of the bonded
Given that impedance analyzers are expensive,
structure were made, where two steel adherends were
bulky, and may be over-weighted for some applications,
bonded with an epoxy-based adhesive. Conversely, the
Peairs et al.106 proposed an alternative composed of a
analytical model considered only a medium composed
simple and low-cost voltage divider in conjunction with
of steel, and, therefore, damping effects caused by the
an FFT analyzer. By using this approach, the measured
adhesive were not included in the formulation.
electrical impedance can be approximated by dividing
Resonant peaks in experimental and FEM results were
the input voltage, Vi with the current passing through
closely matched. However, the peak amplitudes of the
the sensing resistor, I = Vo =Rs , as presented in Figure 6,
experiments are smaller than the ones from the analyti-
where Vo is the output voltage and Rs is the sensing
cal or numerical results. This discrepancy may be due
resistor. Panigrahi et al.107 improved this experimental
to factors like thickness of the adhesive layer, or its var-
setup by using a function generator as pure sine signal
iation, temperature dependence of piezoelectric para-
generator and an oscilloscope to measure Vo . Bhalla
meters, among others.
et al.108 implemented a simple impedance measurement
FEM numerical simulation is commonly used by
setup using piezoelectric elements, a function generator,
researchers for (a) determining the placement of sensors
and a digital multimeter. This setup proved to be effec-
for optimal performance in damage detection in the
tive in detecting changes in structural behavior due to
monitored structure, (b) estimate the behavior of the
damage, but the measured results are quite different
electrical impedance signatures in a wide frequency
from those obtained with conventional equipment.
range, and, therefore, determine the optimal frequency
Xu and Giurgiutiu109 have also developed a low-cost
range for structural monitoring using EMIS, or (c)
setup that included a function generator with a linear
gather data from numerical models to train advanced
chirp signal source and a frequency-swept signal source,
damage algorithms like, for example, ANN. This is also
while the measurements were acquired by a data-
the case for bonded joints. Nonetheless, given that only
acquisition (DAQ) board and a computer. Baptista and
recently has EMIS been used for SHM of adhesive
Filho98 used a chirp signal to excite the piezoelectric
joints, few cases of FEM simulation has been found in
actuator, and used a DAQ board connected to a com-
the literature.
puter with LabView, as well as an auxiliary circuit for the
EMIS readings. The reported results show that measure-
ments acquired by this system are quite similar to the ones
Hardware for impedance measurement
obtained by an impedance analyzer (model HP4219A).
An EMIS system designed for the continuous surveil- This setup was then used by Gulizzi et al.97,99,100
lance of structural health is composed of one or more Wang and You110 developed an impedance measur-
PWAS, as described in section ‘‘Introduction,’’ and an ing system based on a Wheatstone bridge, where one of
impedance analyzer that derives the measured electrical the impedances is the piezoelectric sensor, ZPWAS , and
impedance, or admittance, in a frequency spectrum. an adjustable digital resistor, Zp , is placed parallel to
However, while this is the conventional equipment the PWAS, to balance the bridge, as presented in
setup in EMIS-based SHM, various alternatives have Figure 7. Kim and Wang111 used an array of bi-stable
been put forth. It must be noted that these alternatives circuits integrated with the structure and piezoelectric
have been made for monitoring of any structure, and, transducers to obtain measurements less sensitive to
as such, some of these techniques have also been noise and to allow for better localization of the damage.
employed to monitor adhesive joints. However, no Maruo et al.112 developed a hardware and firmware
hardware setup has been developed for exclusive moni- system based on a microcontroller for measuring the
toring of bonded structures. electromechanical impedance of multiplexed bonded
Tenreiro
12 et al. 239
piezocomposite is comparatively stiffer and has more

flexibility, allowing it to be mounted in curve geometries,
and, as such, researchers have used it in experimental set-
ups.120–123 However, it is noted that, with high tempera-
tures, post-curing of the matrix may occur, leading to
unwanted changes in impedance readings.124
EMIS is an effective local damage detection
method, however it has the disadvantage of requiring
the implementation of a network of sensors for health
monitoring of large structures, and therefore, of
establishing a complex hardware architecture for sig-
Figure 7. Wheatstone bridge proposed by Wang and You.110 nal generation and data acquisition, with a multitude
of electric cables. An alternative to this is to use wire-
PWAS, without making use of expensive FFT analy- less sensors in SHM.125,126 This strategy was put forth
zers or algorithms and without need for high-speed as an alternative to EMIS-based SHM.123,127–129 This
data-acquisition equipment. The results obtained in this is done so with the help of the AD5933 impedance
case were almost identical to those obtained by an measurement evaluation board (manufactured by
HP4194A impedance analyzer. Analog Devices, Inc.), and allows for impedance mea-
In the meantime, new piezoelectric sensors have also surements for frequencies up to 100 kHz.130 It is
been developed by several researchers. Based on a reported that the use of this equipment is in EMIS-
Lamb wave mode decomposition sensor setup to iden- based research.123,127,129,131
tify individual modes,113 Song et al.114 applied a dual
piezoelectric transducer composed of a piezoelectric
ring and a piezoelectric disk contained inside the ring, Challenges for EMIS-based SHM
where the top and bottom surfaces are coated with
nickel. These elements can work both as actuators and SHM is a field where intensive research has been made
sensors simultaneously, or can act separately as a sen- in the last two decades, but real-world implementation
sor and as an actuator. Experimental results revealed of such systems are rare, only being present in niche
that the measured amplitudes are higher than those applications. Among these are NASA’s Shuttle Modal
achieved with conventional PWAS. Furthermore, the Inspection System for subsystem’s inspection,132 and
imaginary component of the measured impedance is the Israeli’s Air Force unmanned aerial vehicle (UAV)
only sensitive to the structural impedance, and is not planes for damage detection during flight operations.133

influenced by dielectric constant, eT33 . Recently, the first real-world application of EMIS-
Piezoelectric diaphragms, also known as ‘‘buzzers,’’ based SHM was documented in Singapore, where mon-
are a sound component normally used for producing itoring of a mass rapid transit (MRT) heavy metro
alarms and sound elements in electronics, but have also station’s structure is performed during construction
recently been candidate for SHM sensors.115 This com- operations.134 However, despite these efforts, SHM sys-
ponent, which is a circular piezoelectric ceramic tems are far from commercial use. Therefore, in this
bounded to a circular metallic plate, seems an interest- section, current EMIS-based SHM research, develop-
ing alternative sensor for the electromechanical impe- ment and implementation issues, and challenges are
dance monitoring of structures, since it acts similar to a presented.
conventional PWAS.116,117 de Freitas et al.118 devel-
oped an equivalent circuit model to physically describe
the electromechanical coupling in a 1D medium, and
Temperature effects and compensation algorithms
experimentally evaluated its sensitivity. One of the main problems faced by EMIS-based moni-
PWAS are basically ceramic components with toring systems is that the measured signals are suscepti-
mechanically brittle behavior and are susceptible to frac- ble to unwanted variations with temperature changes.

ture in real-world applications, where one cannot control This happens because the dielectric constant, eT33 (see
the environment. Furthermore, these elements cannot be section ‘‘Theoretical background of EMIS’’), varies sig-
fixed to non-planar surfaces, given their lack of mechani- nificantly with temperature, and consequently, changes
cal flexibility. For this reason, NASA researchers devel- the capacitive component of the measured electrical
oped a macro-fiber composite (MFC) sensor, which is a impedance.135,136 However, the piezoelectric-induced
piezocomposite composed of a sheet of aligned piezo- strain coefficient, d31 , of certain piezoelectric materials
ceramic fiber, a couple of thin polymer layers with con- is also temperature sensitive.135 Furthermore, material
ductive electrode patterns, and a matrix material.119 This properties of the monitored structure may also vary
240
13
significantly with temperature. As such, the structural measurements, while changes in the measured impe-
mechanical impedance may vary.61 dance due to damage are local to the nearest sensors.
Several studies show that the measured impedance Therefore, comparing statistics, such as the mean
amplitude decreases with increasing temperature, when square root deviation, RMSD, between two different
plotted versus the excitation frequency, while the phase PWAS readings, would indicate that, if both values are
remains unaffected.135,136 Baptista et al. corroborated identical, then impedance changes are due to tempera-
these findings and verified that the measured resonance ture changes. Otherwise, values differ if changes are
peaks reveal a shift toward lower frequencies. This shift due to the detection of damage. However, one must
is a function of both the temperature and the baseline note that this method is only effective when various
frequency of the resonance peak.61 The frequency shifts sensors are used.
and variations of the impedance amplitude may be sig- Standard EMIS-based SHM systems may require the
nificant enough to jeopardize the detection algorithms, detection and compensation of temperature variations
leading to false positive detection. in impedance readings, however, some applications
Several algorithms have been proposed to fix this should include the temperature effect. Kundu et al.55
issue. Krishnamurthy et al.135 proposed the formulation examined the electromechanical impedance behavior of
of a frequency-independent temperature coefficient, dT , continuous welded rails with axial loads and tempera-
that can be used to establish a linear relationship in ture variations, which may cause thermal stresses, and,
order to determine a compensated impedance, Zcomp , at consequently, may affect impedance readings.
a given reference temperature, Tref Given the research conducted so far, advances are
needed to improve the compensation of temperature
Zcomp = Zm 1 + dT (T � Tref ) ð26Þ influence upon impedance spectroscopy measurements.
The approach of Bastani et al., based on the compari-
where Zm is the measured impedance at temperature T . son between readings of different piezoelectric sensors,
However, this type of approach can only be approxi- allow to see whether significant impedance changes are
mately defined for a given temperature range, since the either due to temperature or due to damage. However,
piezoelectric parameters do not vary linearly with tem- such procedure needs to be extended by including other
perature. Park et al.136 followed a similar approach to statistical indices for comparison. Another approach is
shift the measured frequency readings until it matched to develop temperature compensation algorithms with-
the baseline measurements. Baptista et al.61 developed out potentially losing damage information, such as the
an automatic method, using the linear shifting compen- resonant peak frequency shift.
sation method, where the measured impedance was Research on the temperature effects on EMIS read-
incrementally compensated in a loop until the correla- ings of adhesively bonded joints have been performed.
tion coefficient, CC, between the measured and the ref- Temperature can have a much stronger impact on the
erence impedance was maximized (nearly unit, CC’1). impedance readings, since the adhesive’s mechanical
However, such a procedure was only feasible without behavior can be temperature dependent, and can change
errors for readings of small frequency spectra. Koo drastically if the ambient temperature is superior to the
et al.137 demonstrated that CC values of measured and adhesive’s glass transition temperature, Tg .140,141 As
frequency-shifted impedance signals have fluctuations, such, electrical impedance measurements are highly sus-
which may result in false damage characterization, and ceptible to variations due to both variations of piezo-
propose the use of an outlier analysis with the frequency electric properties, as well as variations of adhesive
shifted measurements. Siebel and Lilov138 have shown properties, as demonstrated by Bilgunde and Bond.142
that temperature compensated impedance measure-
ments using the CC is effective for excitation frequencies
below 30 kHz, but may not be sensitive to detection of
Issues for real-world applications
impact damage with temperature fluctuations. The temperature perturbation on electromechanical
Sun et al.139 used a probability correlation function impedance signatures is recognized as the most influen-
to perform temperature compensation. This approach tial variable that may cause false positives. Nonetheless,
also enables damage detection, but it can only be per- other factors may also influence the measured signals.
formed by measuring the differences in the resonant As previously discussed, piezoelectric elements need
peak amplitudes, since the measurement of frequency to be connected to the structure using adhesives.
shift is no longer possible. However, the bonding procedure is manually made,
Bastani et al.131 conjectured that, if the monitored and, consequently, may lead to imperfect adhesion of
structure has more than one PWAS, then the changes the PWAS. This is of great importance, since a badly
in the electric impedance measurements due to tem- connected sensor may lead to either the false detection
perature would be reflected in all PWAS of damage, or the inability of detecting real damage.
Tenreiro
14 et al. 241
Knowing this problem, procedures need to be designed that can connect to the PWAS and acquire the impe-
in order to evaluate the quality of the bond. Giurgiutiu dance spectrum.33 Loading conditions may also affect
et al.143 found that, when the sensor is detached from the impedance measurements, since they have influence
the structure, the sensor’s resonant frequencies can be on the mechanical properties of the structure being
easily detected from the electrical reactance, Im(Z(v)), examined. Ding et al.151 verified that a downwards and
measurements. Park et al.144 further analyzed the feasi- rightwards shift of impedance peaks were observed for
bility of using the imaginary component of the electric concrete structures. Neuschwander et al.152 developed a
admittance (the inverse of the impedance) to determine load compensation algorithm that related the strain
the sensor’s integrity to accurately conduct structural suffered by a CFRP structure with the impedance sig-
monitoring. They concluded that breakage of PWAS natures. Furthermore, this strategy allowed one to
would result in a downward shift of the slope of the ima- determine when the structure became damaged or
ginary part of the admittance, while, on the other hand, reached failure. However, loading conditions need to
bonding defects would cause an upward shift in the be further studied, particularly when knowing that
slope. Park et al.145 also compared the performance of external loading may have similar effects to those
piezoceramic and MFC sensors and concluded the com- resulting from the influence of temperature.
posite sensors could still perform monitoring, since they
are more robust and reliable than PWAS. As previously New algorithms and metrics for SHM
mentioned, Xu and Liu40 developed a mathematical for-
mulation of the electromechanical impedance behavior, SHM is a field where data are a key component and,
where the adhesive bond between the structure and the therefore, data processing and interpretation tools are
sensor could be adjusted. Mulligan et al.146 studied the of the utmost importance. Unlike traditional NDT
performance of PWAS in structural excitation and sen- methods, SHM may yield huge amounts of data in a
sing, and found that both non-uniform adhesion and short period of time. Therefore, it is vital to develop
degradation of the adhesive properties influence the data compression tools and efficient damage detection
PWAS behavior, both as an actuator and sensor. Ai algorithms for future commercial use.153
et al.147 developed an EMIS-based damage detection Commonly, EMIS-based health monitoring systems
SHM system that was able to detect if there was dete- use statistical metrics for damage detection, such as the
rioration in the bond layer or in the sensor, as well as mean square root deviation, RMSD, or the correlation
breakage. Following these ideas, new advances should coefficient, CC. These indices use the real part of the
focus on developing an automated method for control- measured electric impedance, which is sensitive to dam-
ling the bonding of PWAS to any given structure. age. Bhalla et al.59 demonstrated that the measured
It was also reported that piezoelectric ceramics are electric admittance can be divided in passive, YP (v),
sensitive to humidity, thus affecting the measured impe- and active components, YA (v), respectively as
dance spectra. Bhalla et al.148 determined that protect-
ing the sensor with silica gel could mitigate this wP lP T
YP (v) = jv e ð27Þ
problem, but it would render the piezoelectric sensor hP 33
less sensitive to damage. Furthermore, humidity also w P lP Zs (v)
affects the adhesives and composites and it has been YA (v) = jv d31 Y11
E
ð28Þ
hP Zs (v) + ZP (v)
shown that water uptake will have a negative effect on
the mechanical behavior of adhesive joints.149,150 To where YP (v) and YA (v) were deduced from the model
the authors’ best knowledge, the monitoring of adhe- proposed by Liang et al.35 It can be noted that the passive
sive joints under the effect of hygrothermal aging was component does not contribute to the detection of dam-
not yet tackled and recognize this as an interesting age, since it is not dependent on the structural mechanical
topic, especially when considering the effects of water impedance, Zs . Therefore, the passive component from
on the piezoelectric properties of PWAS. experimental measurements was not considered and
Another important issue with EMIS-based SHM merely the imaginary part of the measured impedance
systems is the existence of wire resistance, which, if was used for structural damage monitoring. However,
long, may negatively impact the measurements of the since their works publication, there has been very little
electrical impedance, Z(v). Annamdas and Yang134 follow-up research that the authors have seen. As such,
viewed that the addition of the wire resistance caused a the authors find that potential application of these data
shift of resonance peaks to lower frequencies. As such, can be useful for damage detection, localization, and
wire length is limited. Alternatively, to avoid this prob- quantification, either by the development of new algo-
lem, one can use wireless transmission of data as pre- rithms or by the conception of new statistical metrics.
sented in section ‘‘Hardware for impedance Various algorithms have been developed in order to
measurement.’’ Another alternative is to use drones detect, localize, and characterize damage. Most either
242
15
Figure 8. Schematic representation of the guided wave setup in a bonded joint.
use a model-based approach, meaning that their devel- Rayleigh waves, which are also known as surface acous-
opment is based on physical phenomena, or take tic waves or surface-guided waves, are a type of elastic
advantage of simple damage metrics. In recent years, guided wave that travels close to the surface with almost
some researchers developed ANN to process electrome- no penetration through the solid’s depth. Shear hori-
chanical impedance signatures, with varying degrees of zontal waves are horizontally polarized waves, meaning
success. However, further research is still needed, since that the particle motion is parallel to the surface and
most algorithms have only been tested for simple struc- perpendicular to the wave’s propagation direction.
tures in a laboratory environment. Furthermore, other Lamb waves are more complex in nature and can be
data-driven and machine learning approaches, besides subdivided as (a) symmetric lamb waves, which are
CNN and PNN, can be used.14,154 As an example, Gu quasi-axial waves and (b) antisymmetric lamb waves,
et al.155 developed a fuzzy-logic system that could corre- which behave as quasi-flexural waves. For a given value
late EMIS measurements with the compressive strength of the quantity frequency–thickness product, that is, of
of concrete structures during the curing process. vh, a combination of symmetric (S0, S1, S2, ...) and
Another unexplored alternative in the study and antisymmetric (A0, A1, A2, ...) waves may coexist.
modeling of EMIS signatures is using fractional calcu- However, given the complex nature of these waves,
lus, which is a generalization of the classical differential research tends to use lower frequencies, obtaining there-
calculus. This mathematical tool has been used to fore only the fundamental modes (S0 and A0).3,5
model complex and nonlinear systems and compo- For the creation and measurement of guided waves,
nents,156,157 and recently has been used in conjunction one can use various types of actuators and sensors,
with electrical impedance spectroscopy to characterize among which, one of the most popular is the PWAS.
the electrical behavior of an epoxy adhesive in SLJs.158 The EMIS- and guided wave–based SHM systems can
coexist, as discussed in a number of studies.75,88,143,160–162
In fact, one can use EMIS health monitoring to monitor
EMIS-based SHM with other systems near-sensor incipient damage, and, at the same time, use
Piezoelectric elements are an essential part of EMIS- guided wave–based SHM for monitoring of distant dam-
based monitoring of structures. However, these sensors age. However, one must note that extra equipment is
can also be used in other SHM systems, such as guided required for the generation of guided waves, such as a
waves (Lamb, Shear Horizontal, etc.), and Acoustic voltage amplifier and a dedicated DAQ system. A sche-
Emissions,3,143,159 or even in structural control of vibra- matic diagram of the guided wave SHM setup for adhe-
tions.119,160 These systems will be discussed herein. sive joints is presented in Figure 8.
Guided waves are a family of ultrasonic elastic waves Acoustic emission is an SHM and NDT method where
that propagate through a bounded medium, which may elastic waves are generated by small impacts and instanta-
result in multi-modal, dispersive waves. These waves neous solicitations on the monitored structure.17,18
have a special relevance in SHM, since they can propa- Therefore, unlike guided wave SHM systems, structural
gate over large distances and inside curved walls. integrity monitoring using acoustic emissions does not
Tenreiro
16 et al. 243
specialized algorithms to detect, locate, and character-

ize damage in a given structure.
Another important aspect of this technique is the
development of special hardware and signal processing
tools for implementation of an EMIS-based SHM sys-
tem. Here, the main focus has been on developing
cheaper alternatives, and more robust systems allowing
for (a) the design of signal processing with lower com-
plexity, (b) the implementation of a wide sensor net-
work, and (c) the transmission of data by means of
Figure 9. Block diagram of an active vibration control system.
wireless communication systems.
However, despite such progress, further ground-
require signal generation hardware. It is also noted that
breaking research still needs to occur in order to reach
PWAS sensors can be used in acoustic emission–based
a broad implementation of EMIS as a feasible and reli-
SHM systems. Therefore, since acoustic emission and
able structural monitoring tool. Among these issues,
EMIS systems are similar in what concerns the hardware
robust compensation methods for ambient perturba-
implementation, they can work concurrently to synergisti-
tions on electromechanical impedance readings need to
cally detect damage in structures.115,121,163
be developed, as well as the development of novel and
Piezoelectric elements have also been used in the
innovative damage detection, localization, and charac-
control of vibrations in structures, or, in other words,
terization algorithms.
in the design of control systems to monitor and reduce/
nullify vibration loads in structures.119,160,164,165 There
Declaration of conflicting interests
are two types of vibration control systems: passive and
active control systems. Active control systems use a The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
controller and a power source to control the actuators
article.
that apply a force to the monitored structure, as illu-
strated in Figure 9. Conversely, passive control systems
do not require a power system and are activated by the Funding
resultant motion from the vibration.160 The author(s) received no financial support for the research,
authorship, and/or publication of this article.
Conclusion
ORCID iD
This article reviewed EMIS as a vibration-based SHM
A Francisco G Tenreiro https://orcid.org/0000-0002-6410-
method, with emphasis in adhesively bonded joints. 0196
SHM is seen as the successor of NDT to mechanical
structures, and aims to perform the tasks of detecting, References
localizing, and characterizing damage on a structure as
1. Farrar CR and Worden K. An introduction to structural
well as determining the severity of the damage and the
health monitoring. Philos Trans Royal Soc A: Math Phys
remaining useful life of its lifecycle. Up to the present Eng Sci 2007; 365(1851): 303–315.
day, no EMIS-based SHM system has been able to per- 2. Worden K, Farrar CR, Manson G, et al. The fundamen-
form all of these tasks. This is particularly true for tal axioms of structural health monitoring. Proc Math
adhesively bonded structures, where research with Phys Eng Sci 2007; 463(2082): 1639–1664.
EMIS-monitoring only began to take place in this last 3. Giurgiutiu V. Structural health monitoring with piezoelec-
decade, focusing on the adhesive curing process and tric wafer active sensors. 2nd ed. Oxford: Academic Press,
the monitoring of the joint’s mechanical behavior. 2014.
Meanwhile the detection of damage and defects in 4. Yan W and Chen WQ. Structural health monitoring
bonded joints still requires much attention. using high-frequency electromechanical impedance signa-
The EMIS method has been an ever-growing field of tures. Adv Civil Eng 2010; 2010: 429148.
5. Mitra M and Gopalakrishnan S. Guided wave based
research in SHM, and several approaches have been
structural health monitoring: a review. Smart Mater
made to study structures. One of the first methods is Struct 2016; 25(5): 053001.
the mathematical modeling of the PWAS interaction 6. Güemes A, Fernandez-Lopez A, Pozo AR, et al. Struc-
with the structure and, consequently, of the connection tural health monitoring for advanced composite struc-
between the electrical behavior of the piezoelectric ele- tures: a review. J Compos Sci 2020; 4(1): 13.
ment and the mechanical impedance of the structure. 7. Abbas S, Li F and Qiu J. A review on SHM techniques
Another possible approach is to use damage metrics or and current challenges for characteristic investigation of
244
17
damage in composite material components of aviation structures, part III: investigation of mechanical strength.
industry. Mater Perform Character 2018; 7(1): 224–258. J Adhesion 2002; 78(12): 1027–1038.
8. Giurgiutiu V. Structural health monitoring of aerospace 28. Brotherhood C, Drinkwater B and Dixon S. The detect-
composites. 1st ed. London: Elsevier, 2016. ability of kissing bonds in adhesive joints using ultrasonic
9. Staszewski WJ, Boller C and Tomlinson GR. Health mon- techniques. Ultrasonics 2003; 41(7): 521–529.
itoring of aerospace structures -smart sensor technologies 29. Vijaya Kumar R, Bhat M and Murthy C. Evaluation of
and signal processing. 1st ed. Chichester: John Wiley & kissing bond in composite adhesive lap joints using digi-
Sons, 2004. tal image correlation: preliminary studies. Int J Adhesion
10. Soliman M, Barone G and Frangopol DM. Fatigue relia- Adhesiv 2013; 42: 60–68.
bility and service life prediction of aluminum naval ship 30. Jeenjitkaew C and Guild F. The analysis of kissing bonds in
details based on monitoring data. Struct Health Monit adhesive joints. Int J Adhesion Adhesiv 2017; 75: 101–107.
2015; 14(1): 3–19. 31. Park G, Sohn H, Farrar CR, et al. Overview of piezoelec-
11. Garevski M. Earthquakes and health monitoring of civil tric impedance-based health monitoring and path for-
structures. Dordrecht: Springer, 2013. ward. Shock Vib Digest 2003; 35(6): 451–463.
12. Chang PC, Flatau A and Liu SC. Review paper: health 32. Annamdas VGM and Soh CK. Application of electrome-
monitoring of civil infrastructure. Struct Health Monit chanical impedance technique for engineering structures:
2003; 2(3): 257–267. review and future issues. J Intell Mater Syst Struct 2010;
13. Park G, Cudney HH and Inman DJ. Impedance-based 21(1): 41–59.
health monitoring of civil structural components. J Infra- 33. Na W and Baek J. A review of the piezoelectric electrome-
struct Syst 2000; 6(4): 153–160. chanical impedance based structural health monitoring tech-
14. Farrar CR and Worden K. Structural health monitoring: nique for engineering structures. Sensors 2018; 18(5): 1307.
a machine learning perspective. 1st ed. Chichester: John 34. ANSI/IEEE 176-1987 Standard on Piezoelectricity.
Wiley & Sons, 2013. 35. Liang C, Sun FP and Rogers CA. Coupled electro-
15. Su Z, Ye L and Lu Y. Guided Lamb waves for identifica- mechanical analysis of adaptive material systems-
tion of damage in composite structures: a review. J Sound determination of the actuator power consumption and
Vib 2006; 295(3–5): 753–780. system energy transfer. J Intel Mater Syst Struct 1997;
16. Lammering R, Gabbert U, Sinapius M, et al. Lamb-wave 8(4): 335–343.
based structural health monitoring in polymer composites. 36. Liang C, Sun FP and Rogers CA. An impedance method
1st ed. Cham: Springer, 2018. for dynamic analysis of active material systems. La Jolla,
17. Ono K. Structural health monitoring of large structures CA: American Institute of Aeronautics and Astronautics,
using acoustic emission–case histories. Appl Sci 2019; pp. 3587–3599.
9(21): 4602. 37. Giurgiutiu V and Zagrai AN. Embedded self-sensing
18. Ono K. Review on structural health evaluation with piezoelectric active sensors for on-line structural identifi-
acoustic emission. Appl Sci 2018; 8(6): 958. cation. J Vib Acoust 2002; 124(1): 116–125.
19. Bagavathiappan S, Lahiri B, Saravanan T, et al. Infrared 38. Sun FF, Liang C and Rogers CA. Structural modal anal-
thermography for condition monitoring—a review. Infra- ysis using collocated piezoelectric actuator/sensors: an
red Phys Tech 2013; 60: 35–55. electromechanical approach. In: Proceedings smart struc-
20. Doebling SW, Farrar CR and Prime MB. A summary tures and materials 1994: smart structures and intelligent
review of vibration-based damage identification methods. systems, Orlando, FL, 14–16 February, 1994, vol. 2190,
Shock Vib Digest 1998; 30(2): 91–105. pp. 158–161. New York: SPIE.
21. Fan WF and Qiao P. Vibration-based damage identifica- 39. Zhou SW, Liang C and Rogers CA. An impedance-based
tion methods: a review and comparative study. Struct system modeling approach for induced strain actuator-
Health Monit 2011; 10(1): 83–111. driven structures. J Vib Acoust 1996; 118(3): 323–331.
22. Wang G and Li F. An online viscosity measurement 40. Xu YG and Liu GR. A modified electro-mechanical
method based on the electromechanical impedance of a impedance model of piezoelectric actuator-sensors for
piezoelectric torsional transducer. IEEE Sens J 2018; debonding detection of composite patches. J Intel Mater
18(21): 8781–8788. Syst Struct 2002; 13(6): 389–396.
23. Guyott CCH, Cawley P and Adams RD. The non- 41. Zagrai AN and Giurgiutiu V. Health monitoring of aging
destructive testing of adhesively bonded structure: a aerospace structures using the electromechanical impe-
review. J Adhesion 1986; 20(2): 129–159. dance method. In: Proceedings, smart nondestructive eva-
24. Adams RD and Drinkwater BW. Nondestructive testing of luation for health monitoring of structural and biological
adhesively-bonded joints. NDT&E Int 1997; 30(2): 93–98. systems, San Diego, CA, 11 June 2002, vol. 4702, pp.
25. da Silva LFM, Öchsner A and Adams RD. Handbook of 289–300. New York: SPIE.
adhesion technology. 2nd ed. New York: Springer, 2018. 42. Zagrai AN and Giurgiutiu V. Electro-mechanical impe-
26. Yang X, Sun L, Huang B, et al. Tuning the properties of dance method for crack detection in thin plates. J Intel
functional adhesives with hybrid nanofillers for structural Mater Syst Struct 2001; 12(10): 709–718.
health monitoring. J Adhesion 2019; 97: 101–116. 43. Bhalla S, Kumar P, Gupta A, et al. Simplified impedance
27. Banks WM, Boinard P and Pethrick RA. Use of dielec- model for adhesively bonded piezo-impedance transdu-
tric spectroscopy to assess adhesively bonded composite cers. J Aerospace Eng 2009; 22(4): 373–382.
Tenreiro
18 et al. 245
44. Bhalla S and Soh CK. Electromechanical impedance aerospace systems. San Diego, CA, 8–12 March 2014, vol.
modeling for adhesively bonded piezo-transducers. J Intel 8345, p. 83451Z. New York: SPIE.
Mater Syst Struct 2004; 15(12): 955–972. 59. Bhalla S, Surendra A and Soh CK. Influence of
45. Moharana S and Bhalla S. Modelling of shear lag effect structure-actuator interactions and temperature on piezo-
for piezo-elastodynamic structure for electro-mechanical electric mechatronic signatures for NDE. In: Proceedings,
impedance technique. In: Proceedings, sensors and smart smart materials, structures, and systems, Bangalore, India,
structures technologies for civil, mechanical, and aerospace 12–14 December 2002, vol. 5062. New York: SPIE.
systems, San Diego, CA, 9–12 March 2015, vol. 9435, p. 60. Giurgiutiu V and Rogers CA. Recent advancements in
94352J. New York: SPIE. the electro-mechanical (ElM) impedance method for
46. Wang D, Song H and Zhu H. Embedded 3D electrome- structural health monitoring and NDE. In: Proceedings
chanical impedance model for strength monitoring of SPIE 3329, smart structures and materials 1998: smart
concrete using a PZT transducer. Smart Mater Struct structures and integrated systems, San Diego, CA, 27 July
2014; 23(11): 115019. 1998, vol. 3329.
47. Zuo C, Feng X and Zhou J. A three-dimensional model 61. Baptista F, Budoya D, Almeida V, et al. An experimental
of the effective electromechanical impedance for an study on the effect of temperature on piezoelectric sensors
embedded PZT transducer. Math Probl Eng 2013; 2013: for impedance-based structural health monitoring. Sen-
218026. sors 2014; 14(1): 1208–1227.
48. Pavelko V. New applications of a model of electromecha- 62. Tseng KKH and Naidu ASK. Non-parametric damage
nical impedance for SHM. In: Proceedings, health moni- detection and characterization using smart piezoceramic
toring of structural and biological systems, San Diego, material. Smart Mater Struct 2002; 11(3): 317–329.
CA, 10–13 March 2014, vol. 9064, p. 90640Y. New York: 63. Oliveira MA, Filho JV, Lopes V, et al. Damage detection
SPIE. based on electromechanical impedance principle and prin-
49. Annamdas VGM and Soh CK. Three-dimensional electro- cipal components. In Allemang R, Clerck J De, Niezrecki
mechanical impedance model. I: formulation of directional C and Wicks A (eds) Topics in modal analysis vol 7. New
sum impedance. J Aerospace Eng 2007; 20(1): 53–62. York: Springer, 2014. pp. 307–315. DOI: 10.1007/978-1-
50. Annamdas VGM and Soh CK. Three-dimensional elec- 4614-6585-0 28.
tromechanical impedance Model. II: damage analysis 64. Lecce L, Viscardi M and Zumpano G. Multifunctional
and PZT characterization. J Aerospace Eng 2007; 20(1): system for active noise control and damage detection on
63–71. a typical aeronautical structure. In: Proceedings, smart
51. Annamdas VGM and Soh CK. Three-dimensional elec- structures and materials 2001: smart structures and inte-
tromechanical impedance model for multiple piezocera- grated systems, Newport Beach, CA, 2001, vol. 4327.
mic transducers—structure interaction. J Aerospace Eng New York: SPIE, pp. 201–212.
2008; 21(1): 35–44. 65. Tinoco HA and Marulanda DJ. Damage identification in
52. Ritdumrongkul S, Abe M, Fujino Y, et al. Quantitative active plates with indices based on Gaussian confidence
health monitoring of bolted joints using a piezoceramic ellipses obtained of the electromechanical admittance. J
actuator–sensor. Smart Mater Struct 2004; 13(1): 20–29. Nondestruct Eval 2015; 34(3): 28.
53. Kuang YD, Li GQ and Chen CY. An admittance func- 66. Tinoco HA and Marulanda DJ. A new index for damage
tion of active piezoelectric elements bonded on a cracked identification in active beams with electromechanical
beam. J Sound Vib 2006; 298(1-2): 393–403. impedance technique (EMI) Approach to SHM, Bogotá,
54. Tseng KK and Wang L. Impedance-based method for Colombia, 2014, pp. 1–6, https://www.researchgate.net/
nondestructive damage identification. J Eng Mech 2005; publication/267331933_A_New_Index_For_Damage_Iden
131(1): 58–64. tification_In_Active_Beams_with_Electromechanical_Impe
55. Zhu X, Lanza Di Scalea F and Fateh M. Temperature dance_Technique_EMI_Approach_to_SHM
and axial stress effects in electromechanical-impedance- 67. Enciu D, Ursu I and Toader A. New results concerning
method-based structural health monitoring. In: Proceed- structural health monitoring technology qualification for
ings, health monitoring of structural and biological systems, transfer to space vehicles. Struct Control Health Monit
San Diego, CA, 10–13 March 2014, vol. 9064, p. 90641O. 2017; 24(10): e1992.
New York: SPIE. 68. Sohn H and Farrar CR. Damage diagnosis using time
56. Park G, Cudney HH and Inman DJ. An integrated health series analysis of vibration signals. Smart Mater Struct
monitoring technique using structural impedance sensors. 2001; 10(3): 446–451.
J Intel Mater Syst Struct 2000; 11(6): 448–455. 69. Park G, Rutherford AC, Sohn H, et al. An outlier analy-
57. Yang Y and Hu Y. Electromechanical impedance model- sis framework for impedance-based structural health
ing of PZT transducers for health monitoring of cylindrical monitoring. J Sound Vib 2005; 286(1-2): 229–250.
shell structures. Smart Mater Struct 2008; 17(1): 015005. 70. de Oliveira M, Monteiro A and Vieira Filho J. A new struc-
58. Annamdas VGM, Yang Y and Park S. Application of tural health monitoring strategy based on PZT sensors and
three dimensional electromechanical impedance model convolutional neural network. Sensors 2018; 18(9): 2955.
for damage assessment of plate. In: Proceedings, sensors 71. Cherrier O, Selva P, Pommier-Budinger V, et al. Damage
and smart structures technologies for civil, mechanical, and localization map using electromechanical impedance
246
19
spectrums and inverse distance weighting interpolation: Statistical Meta-modeling. J Intel Mat Syst Str 2006;
experimental validation on thin composite structures. 17(11): 1023–1036. DOI:10.1177/1045389X06063087.
Struct Health Monit 2013; 12(4): 311–324. 85. Zahedi F and Huang H. Time–frequency analysis of elec-
72. Kralovec C, Schager M and Mayr M. Localization of tromechanical impedance (EMI) signature for physics-based
damages by model-based evaluation of electro- damage detections using piezoelectric wafer active sensor
mechanical impedance measurements. In: Proceedings of (PWAS). Smart Mater Struct 2017; 26(5): 055010.
the 9th European workshop on structural health monitor- 86. Na S, Tawie R and Lee HK. Impedance-based nondes-
ing, Manchester, 10–13 July 2018. tructive evaluation of the FRP adhesive joints in corro-
73. Dragan K, Dziendzikowski M and Kurnyta A. Damage sive environment with re-usable technique. In:
detection and localization with use of PZT sensors and Proceedings, sensors and smart structures technologies for
transfer impedance approach. J KONES Powertrain civil, mechanical, and aerospace systems, San Diego, CA,
Transp 2016; 23(4): 643–648. 7–10 March 2011, vol. 7981, p. 79811B. New York:
74. Moll J, Kathol J, Fritzen CP, et al. Open Guided Waves: SPIE.
online platform for ultrasonic guided wave measure- 87. Na S, Tawie R and Lee HK. Electromechanical
ments. Struct Health Monit 2019; 18(5–6): 1903–1914. impedance method of fiber-reinforced plastic adhesive
75. Loendersloot R, Ehsani M and Shamshirsaz M. Fatigue joints in corrosive environment using a reusable piezo-
damage identification and remaining useful life estima- electric device. J Intel Mater Syst Struct 2012; 23(7):
tion of composite structures using piezo Wafer Active 737–747.
Transducers. In: Proceedings of the 32nd international con- 88. Lim YY, Tang ZS and Smith ST. Piezoelectric-based
gress and exhibition on condition monitoring and diagnostic monitoring of the curing of structural adhesives: a novel
engineering management, Huddersfield, 26–28 May 2020. experimental study. Smart Mater Struct 2019; 28(1):
76. Giurgiutiu V and Kropas-Hughes CV. Comparative study 015016.
of neural network damage detection from a statistical set 89. Dugnani R, Zhuang Y, Kopsaftopoulos F, et al. Adhe-
of electro-mechanical impedance spectra. In: Proceedings, sive bond-line degradation detection via a cross-
smart nondestructive evaluation and health monitoring of correlation electromechanical impedance–based
structural and biological systems II, vol. 5047, San Diego, approach. Struct Health Monit 2016; 15(6): 650–667.
CA, 10–14 March 2013, p. 108. New York: SPIE. 90. Zhuang Y, Kopsaftopoulos F, Dugnani R, et al. Integrity
77. Selva P, Cherrier O, Budinger V, et al. Smart monitoring monitoring of adhesively bonded joints via an electrome-
of aeronautical composites plates based on electromecha- chanical impedance-based approach. Struct Health Monit
nical impedance measurements and artificial neural net- 2018; 17(5): 1031–1045.
works. Eng Struct 2013; 56: 794–804. 91. Karachalios E, Adams R and da Silva LF. Strength of
78. Rabiei M, Sheldon J and Palmer C. An impedance-based single lap joints with artificial defects. Int J Adhes Adhesiv
approach for detection and quantification of damage in 2013; 45: 69–76.
cracked plates and loose bolts in bridge structures. In: 92. Ribeiro F, Campilho R, Carbas R, et al. Strength and
Proceedings health monitoring of structural and biological damage growth in composite bonded joints with defects.
systems, vol. 8348, San Diego, CA, 12–15 March 2012, Compos Part B: Eng 2016; 100: 91–100.
vol. 8348, p. 834828. New York: SPIE. 93. Zhuang Y, Li YH, Kopsaftopoulos F, et al. A self-
79. Gonsalez CG, da Silva S, Brennan MJ, et al. Structural diagnostic adhesive for monitoring bonded joints in aero-
damage detection in an aeronautical panel using analysis of space structures. In: Proceedings, sensors and smart struc-
variance. Mech Syst Signal Process 2015; (52–53): 206–216. tures technologies for civil, mechanical, and aerospace
DOI:10.1016/j.ymssp.2014.04.015. systems, Las Vegas, NV, 2016, vol. 9803, p. 98030I. New
80. Khodaei ZS and Aliabadi M. A multi-level decision York: SPIE.
fusion strategy for condition based maintenance of com- 94. Malinowski PH, Wandowski T and Ostachowicz WM.
posite structures. Materials 2016; 9(9): 790. Characterisation of CFRP adhesive bonds by electrome-
81. Fan X, Li J and Hao H. Impedance resonant frequency chanical impedance. In: Proceedings, health monitoring of
sensitivity based structural damage identification with structural and biological systems, San Diego, CA, 10–13
sparse regularization: experimental studies. Smart Mater March 2014, vol. 9064. New York: SPIE.
Struct 2019; 28(1): 015003. 95. Malinowski PH, Wandowski T and Ostachowicz W. The
82. He C, Yang S, Liu Z, et al. Damage localization and use of electromechanical impedance conductance signa-
quantification of truss structure based on electromechani- tures for detection of weak adhesive bonds of carbon
cal impedance technique and neural network. Shock Vib fibre–reinforced polymer. Struct Health Monit 2015;
2014; 2014: 727404. 14(4): 332–344.
83. Palomino LV, Steffen V and Finzi Neto RM. Probabilis- 96. Malinowski PH, Tserpes K, Ecault R, et al. Mechanical
tic neural network and fuzzy cluster analysis methods and non-destructive study of CFRP adhesive bonds sub-
applied to impedance-based SHM for damage classifica- jected to pre-bond thermal treatment and de-icing fluid
tion. Shock Vibration 2014; 2014: 401942. contamination. Aerospace 2018; 5(2): 36.
84. De Moura JDRV and Steffen V. Impedance-based 97. Gulizzi V, Rizzo P and Milazzo A. On the use of the
Health Monitoring for Aeronautic Structures using EMI for the health monitoring of bonded elements. In:
Tenreiro
20 et al. 247
Proceedings, sensors and smart structures technologies for 111. Kim J and Wang KW. Electromechanical impedance-
civil, mechanical, and aerospace systems, San Diego, CA, based damage identification enhancement using bistable
10–13 March 2014, vol. 9091, p. 906104. New York: and adaptive piezoelectric circuitry. Struct Health Monit
SPIE. 2019; 18(4): 1268–1281.
98. Baptista F and Filho J. A new impedance measurement 112. Maruo IIC, Giachero GDF, Steffen Junior V, et al.
system for PZT-based structural health monitoring. Electromechanical impedance -based structural health
IEEE Trans Instrum Meas 2009; 58(10): 3602–3608. monitoring instrumentation system applied to aircraft
99. Gulizzi V, Rizzo P and Milazzo A. On the repeatability structures and employing a multiplexed sensor array. J
of the EMI for the monitoring of bonded joints. In: Pro- Aerosp Tech Manag 2015; 7(3): 294–306.
ceedings, health monitoring of structural and biological 113. Yeum CM, Sohn H and Ihn JB. Lamb wave mode
systems, San Diego, CA, 8–12 March 2015, vol. 9438, p. decomposition using concentric ring and circular piezo-
9438. New York: SPIE. electric transducers. Wave Motion 2011; 48(4): 358–370.
100. Gulizzi V, Rizzo P and Milazzo A. On the repeatability 114. Song H, Lim HJ and Sohn H. Electromechanical impe-
of electromechanical impedance for monitoring of dance measurement from large structures using a dual
bonded joints. AIAA J 2015; 53(11): 3479–3483. piezoelectric transducer. J Sound Vib 2013; 332:
101. Cuc A and Giurgiutiu V. Disbond detection in adhe- 6580–6595.
sively bonded structures using piezoelectric wafer active 115. Castro B, Clerice G, Ramos C, et al. Partial discharge
sensors. In: Proceedings, health monitoring and smart monitoring in power transformers using low-cost piezo-
nondestructive evaluation of structural and biological sys- electric sensors. Sensors 2016; 16(8): 1266.
tems III, San Diego, CA, 21 July 2004, vol. 5394, p. 66. 116. de Freitas ES and Baptista FG. Experimental analysis
New York: SPIE. of the feasibility of low-cost piezoelectric diaphragms in
102. Rautela M and Bijudas C. Electromechanical admit- impedance-based SHM applications. Sens Actuat A:
tance based integrated health monitoring of adhesive Phys 2016; 238: 220–228.
bonded beams using surface bonded piezoelectric trans- 117. Budoya D, Castro BD and Campeiro L. Analysis of
ducers. Int J Adhesion Adhesiv 2019; 94: 84–98. piezoelectric diaphragms in impedance-based damage
103. Liu W and Giurgiutiu V. Finite element modeling and detection in large structures. Proceedings 2017; 2(3):
simulation of piezoelectric wafer active sensors interac- 131.
tion with the host structure for structural health moni- 118. de Freitas ES, Baptista FG, Budoya DE, et al. Equiva-
toring. In: Proceedings, smart structures and materials lent circuit of piezoelectric diaphragms for impedance-
2006: sensors and smart structures technologies for civil, based structural health monitoring applications. IEEE
mechanical, and aerospace systems, San Diego, CA, 11 Sens J 2017; 17(17): 5537–5546.
April 2006, vol. 6174. New York: SPIE. 119. Wilkie WK, Bryant RG, High JW, et al. Low-cost piezo-
104. Roth W and Giurgiutiu V. Structural health monitoring of composite actuator for structural control applications.
an adhesive disbond through electromechanical impedance In: Proceedings, smart structures and materials 2000:
spectroscopy. Int J Adhesion Adhesiv 2017; 73: 109–117. industrial and commercial applications of smart structures
105. Roth W and Giurgiutiu V. Adhesive disbond detection technologies, Newport Beach, CA, 12 June 2000, vol.
using piezoelectric wafer active sensors. In: Proceedings, 3991, pp. 323–334. New York: SPIE.
structural health monitoring and inspection of advanced mate- 120. MacLennan D, Elgoyhen J, Button T, et al. Properties
rials, aerospace, and civil infrastructure, San Diego, CA, 9– and application-oriented performance of high frequency
12 March 2015, vol. 9437. New York: SPIE, p. 94370S. piezocomposite ultrasonic transducers. In: IEEE ultra-
106. Peairs DM, Park G and Inman DJ. Improving accessi- sonics symposium proceedings, New York, 28–31 Octo-
bility of the impedance-based structural health monitor- ber 2007, pp. 100–103. New York: IEEE.
ing method. J Intel Mater Syst Struct 2004; 15(2): 121. Brunner AJ, Birchmeier M, Melnykowycz MM, et al.
129–139. Piezoelectric fiber composites as sensor elements for
107. Panigrahi R, Bhalla S and Gupta A. A low-cost variant structural health monitoring and adaptive material sys-
of electro-mechanical impedance (EMI) technique for tems. J Intel Mater Syst Struct 2009; 20(9): 1045–1055.
structural health monitoring. Exp Tech 2010; 34(2): 122. Rybyanets AN, Domashenkina TV and Rybyanets AA.
25–29. Complex material constants for PZT/PZT ceramic com-
108. Bhalla S, Gupta A, Bansal S, et al. Ultra low-cost adap- posites. In: 2010 IEEE international symposium on the
tations of electro-mechanical impedance technique for applications of ferroelectrics (ISAF), Edinburgh, 9–12
structural health monitoring. J Intel Mater Syst Struct August 2010, pp. 1–4. New York: IEEE.
2009; 20(8): 991–999. 123. Park S, Yun CB and Inman DJ. Structural health moni-
109. Xu B and Giurgiutiu V. A low-cost and field portable toring using electro-mechanical impedance sensors. Fati-
electromechanical (E/M) impedance analyzer for active gue Fract Eng Mater Struct 2008; 31(8): 714–724.
structural health monitoring. Stanford, CA: University 124. Faria CT, Owen RB and Inman DJ. Macro-fiber compo-
of South Carolina, 2005, p. 13. sites performance under thermal cycling for impedance-
110. Wang S and You C. A circuit design for impedance- based SHM applications. In: Proceedings, active and passive
based structural health monitoring. J Intel Mater Syst smart structures and integrated systems, San Diego, CA,
Struct 2008; 19(9): 1029–1040. 10–13 March 2014, vol. 9057, p. 9057. New York: SPIE.
248
21
125. Spencer BF, Ruiz-Sandoval ME and Kurata N. Smart effective frequency shift for compensating temperature
sensing technology: opportunities and challenges. Struct effects. J Intel Mater Syst Struct 2009; 20(4): 367–377.
Control Health Monit 2004; 11(4): 349–368. 138. Siebel T and Lilov M. Experimental investigation on
126. Sarwar MZ, Saleem MR, Park JW, et al. Multimetric improving electromechanical impedance based damage
event-driven system for long-term wireless sensor opera- detection by temperature compensation. Key Eng Mater
tion for SHM applications. IEEE Sens J 2020; 20(10): 2013; 569–570: 1132–1139.
5350–5359. 139. Sun FP, Chaudhry Z, Rogers CA, et al. Automated
127. Taylor SG, Farinholt KM, Park G, et al. Wireless impe- real-time structure health monitoring via signature pat-
dance device for electromechanical impedance sensing tern recognition. In: Proceedings, smart structures and
and low-frequency vibration data acquisition. In: Pro- materials 1995: smart structures and integrated systems,
ceedings, sensors and smart structures technologies for San Diego, CA, 27 February–1 March 1995, vol. 2443,
civil, mechanical, and aerospace systems, San Diego, CA, pp. 236–247. New York: SPIE.
9–12 March 2009, vol. 7292, p. 729228. New York: 140. Machado J, Marques E and da Silva LF. Influence of
SPIE. low and high temperature on mixed adhesive joints
128. Cortez NE, Filho JV and Baptista FG. Design and under quasi-static and impact conditions. Compos Struct
implementation of wireless sensor networks for 2018; 194: 68–79.
impedance-based structural health monitoring using 141. Avendaño R, Carbas R, Marques E, et al. Effect of tem-
ZigBee and global system for mobile communications. J perature and strain rate on single lap joints with dissimi-
Intel Mater Syst Struct 2015; 26(10): 1207–1218. lar lightweight adherends bonded with an acrylic
129. Feng Y, Tang X, Shao B, et al. Self-powered wireless adhesive. Compos Struct 2016; 152: 34–44.
impedance sensing for structural health monitoring. In: 142. Bilgunde P and Bond LJ. Temperature dependence of
Proceedings of 2019 IEEE international conference on electromechanical impedance based bond-line integrity
unmanned systems (ICUS), Beijing, China, 17–19 Octo- monitoring. In: Proceedings, Health monitoring of struc-
ber 2019, pp. 51–55. New York: IEEE. tural and biological systems, Portland, OR, 26–29 March
130. Analog Devices, Inc, AD5933 1 MSPS, 12-Bit Impedance 2017, vol. 10170, p. 101702Y. New York: SPIE.
Converter, Network Analyzer, 2020, www.analog.com. 143. Giurgiutiu V, Zagrai A and Jing Bao J. Piezoelectric wafer
131. Bastani A, Amindavar H, Shamshirsaz M, et al. Identi- embedded active sensors for aging aircraft structural health
fication of temperature variation and vibration distur- monitoring. Struct Health Monit 2002; 1(1): 41–61.
bance in impedance-based structural health monitoring 144. Park G, Farrar CR, di Scalea FL, et al. Performance
using piezoelectric sensor array method. Struct Health assessment and validation of piezoelectric active-sensors
Monit 2012; 11(3): 305–314. in structural health monitoring. Smart Mater Struct
132. Hunt DL, Weiss SP, West WM, et al. Development and 2006; 15(6): 1673–1683. DOI:10.1088/0964-1726/15/6/
implementation of a shuttle modal inspection system. 020.
In: Proceedings of the 8th international modal analysis 145. Park G, Farrar CR, Rutherford AC, et al. Piezoelectric
conference, Kissimmee, FL, 29 January–1 February active sensor self-diagnostics using electrical admittance
1990, vol. 2, pp. 919–925. Bethel, CT: Society for Experi- measurements. J Vib Acoust 2006; 128(4): 469–476.
mental Mechanics. DOI:10.1115/1.2202157.
133. Shapira O, Ben-Simon U, Bergman A, et al. Structural 146. Mulligan KR, Quaegebeur N, Ostiguy PC, et al. Com-
health monitoring of a UAV fleet using fiber optic dis- parison of metrics to monitor and compensate for piezo-
tributed strain sensing. In: Proceedings of the 10th inter- ceramic debonding in structural health monitoring.
national workshop on structural health monitoring, Struct Health Monit 2013; 12(2): 153–168.
Stanford, CA, 10–12 September 2015, vol. 2. Hong 147. Ai D, Luo H and Zhu H. Diagnosis and validation of dam-
Kong, China: Destech Publications. aged piezoelectric sensor in electromechanical impedance
134. Annamdas VGM and Yang Y. Practical implementa- technique. J Intel Mater Syst Struct 2017; 28(7): 837–850.
tion of piezo-impedance sensors in monitoring of exca- 148. Bhalla S, Naidu ASK, Ong CW, et al. Practical issues in
vation support structures. Struct Control Health Monit the implementation of electro-mechanical impedance
2012; 19(2): 231–245. technique for NDE. In: Proceedings, smart structures,
135. Krishnamurthy K, Lalande F and Rogers CA. Effects devices, and systems, Melbourne, VIC, Australia, 1 Jan-
of temperature on the electrical impedance of piezoelec- uary 2002, vol. 4935, p. 484. New York: SPIE.
tric sensors. In: Proceedings, smart structures and materi- 149. Machado J, Marques E, Barbosa A, et al. Effect of
als 1996: smart Structures and Integrated Systems, San hygrothermal aging on the quasi-static behaviour of
Diego, CA, 26–29 February 1996, vol. 2717, pp. CFRP joints varying the overlap length. Compos Struct
302–310. New York: SPIE. 2019; 214: 451–462.
136. Park G, Kabeya K, Cudney HH, et al. Impedance-based 150. Machado JJM, Marques EAS, Barbosa AQ, et al. Influ-
structural health monitoring for temperature varying ence of hygrothermal aging on the quasi-static and
applications. JSME Int J Series A Solid Mech Mater impact behavior of single lap joints using CFRP and
Eng 1999; 42(2): 249–258. aluminum substrates. Mech Adv Mater Struct. Epub
137. Koo KY, Park S, Lee JJ, et al. Automated impedance- ahead of print 31 January 2019. DOI: 10.1080/
based structural health monitoring incorporating 15376494.2019.1675104.
Tenreiro
22 et al. 249
151. Ding J, Kang J, Zhu J, et al. Damage detection of epox- 159. Giurgiutiu V and Cuc A. Embedded non-destructive
ied joint model in precast concrete segmental bridges evaluation for structural health monitoring, damage
using PZT technology. J Aerosp Eng 2020; 33(3): detection, and failure prevention. Shock Vib Digest
04020012. DOI: 10.1061/(ASCE)AS.1943-5525.0001123. 2005; 37(2): 83–105.
152. Neuschwander K, Moll J, Memmolo V, et al. Simulta- 160. Amezquita-Sanchez JP, Dominguez-Gonzalez A, Seda-
neous load and structural monitoring of a carbon fiber ghati R, et al. Vibration control on smart civil structures:
rudder stock: experimental results from a quasi-static ten- a review. Mech Adv Mater Struct 2014; 21(1): 23–38.
sile test. J Intell Mater Syst Struct 2019; 30(2): 272–282. 161. Lim SI, Liu Y and Soh CK. Comparative study of elec-
DOI:10.1177/1045389X18806392. tromechanical impedance and Lamb wave techniques
153. Cawley P. Structural health monitoring: closing the gap for fatigue crack detection and monitoring in metallic
between research and industrial deployment. Struct structures. In: Proceedings, sensors and smart structures
Health Monit 2018; 17(5): 1225–1244. technologies for civil, mechanical, and aerospace systems,
154. Worden K, Staszewski WJ and Hensman JJ. Natural San Diego, CA, 27 August 2012, vol. 8345, p. 83450W.
computing for mechanical systems research: a tutorial New York: SPIE.
overview. Mech Syst Signal Pr 2011; 25(1): 4–111. 162. Sepehry N, Asadi S, Shamshirsaz M, et al. A new model
155. Gu H, Song G, Dhonde H, et al. Concrete early-age order reduction method based on global kernel k -means
strength monitoring using embedded piezoelectric trans- clustering: application in health monitoring of plate
ducers. Smart Mater Struct 2006; 15(6): 1837–1845. using Lamb wave propagation and impedance method.
156. Tenreiro Machado JA and Lopes AM. Analysis of natu- Struct Control Health Monit 2018; 25(9): 2211.
ral and artificial phenomena using signal processing and 163. Nasrollahi A, Deng W, Ma Z, et al. Multimodal struc-
fractional calculus. Fract Calculus Appl Anal 2015; 18(2): tural health monitoring based on active and passive sen-
459–478. sing. Struct Health Monit 2018; 17(2): 395–409.
157. Tenreiro Machado JA and Lopes AM. Fractional-order 164. Lallart M, Lefeuvre E, Richard C, et al. Self-powered cir-
modeling of a diode. Commun Nonlin Sci Numer Simul cuit for broadband, multimodal piezoelectric vibration
2019; 70: 343–353. control. Sensor Actuat A: Phys 2008; 143(2): 377–382.
158. Tenreiro Machado JA, Lopes AM and de Camposinhos 165. Gaudenzi P, Carbonaro R and Benzi E. Control of beam
R. Fractional-order modelling of epoxy resin. Philos Trans vibrations by means of piezoelectric devices: theory and
Royal Soc A: Math Phys Eng Sci 2020; 378(2172): 0292. experiments. Compos Struct 2000; 50(4): 373–379.

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Copyright:

Available Formats

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Copyright:

Available Formats

Review

Structural Health Monitoring

A review of structural health The Author(s) 2021

monitoring of bonded structures using sagepub.com/journals-permissions

A Francisco G Tenreiro1 , António M Lopes2 and Lucas FM da Silva2

Introduction appears in a progressive or sudden manner, such as

Theoretical background of EMIS

Figure 3. Model of interaction between the piezoelectric

model using the concept of effective mechanical

Damage detection be null.60–62 Oliveira et al.63 adopted the Principal

where a0 denotes the incident wave/vibration amplitude, at

were compared through the root mean square devia-

piezocomposite is comparatively stiffer and has more

Figure 8. Schematic representation of the guided wave setup in a bonded joint.

specialized algorithms to detect, locate, and character-

You might also like

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Copyright:

Available Formats

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tenreiro Et Al 2021 A Review of Structural Health Monitoring of Bonded Structures Using Electromechanical Impedance

Uploaded by

Copyright:

Available Formats

Review

Structural Health Monitoring

A review of structural health  The Author(s) 2021

monitoring of bonded structures using sagepub.com/journals-permissions

A Francisco G Tenreiro1 , António M Lopes2 and Lucas FM da Silva2

Introduction appears in a progressive or sudden manner, such as

Theoretical background of EMIS

Figure 3. Model of interaction between the piezoelectric

model using the concept of effective mechanical

Damage detection be null.60–62 Oliveira et al.63 adopted the Principal

where a0 denotes the incident wave/vibration amplitude, at

were compared through the root mean square devia-

piezocomposite is comparatively stiffer and has more

Figure 8. Schematic representation of the guided wave setup in a bonded joint.

specialized algorithms to detect, locate, and character-

You might also like

A review of structural health The Author(s) 2021