applied

sciences
Communication
Impedance-Based Non-Destructive Testing Method
Combined with Unmanned Aerial Vehicle for
Structural Health Monitoring of Civil Infrastructures
Wongi S Na 1, * and Jongdae Baek 2
1 Future Strategy & Convergence Research Institute,
Korea Institute of Civil Engineering & Building Technology, Gyeonggi-Do 10223, Korea
2 Highway & Transportation Research Institute, Korea Institute of Civil Engineering & Building Technology,
Gyeonggi-Do 10223, Korea; [email protected]
* Correspondence: [email protected]; Tel.: +82-31-910-0155

Academic Editors: Gangbing Song, Chuji Wang and Bo Wang

Received: 15 November 2016; Accepted: 20 December 2016; Published: 22 December 2016

Abstract: Unmanned aerial vehicles (UAVs), commonly known as drones, are a rising topic in
remote sensing technologies for structural health monitoring. With technology advancement in
cameras, the visual inspection method using drones is gaining much attention in the field of civil
engineering. However, although visual inspection methods are feasible for finding cracks in structures,
the limitations of image processing for finding internal damage or small defects cannot be ignored.
To overcome this problem, a possible application concept of UAV, combined with a vibration-based
non-destructive health monitoring method, is proposed. The idea is for the drone to temporarily
attach the piezoelectric transducer onto a specific region where excitation and data acquisition occurs
simultaneously. This eliminates the need for a structure to be covered with hundreds of sensors
for monitoring, as this concept uses a single piezoelectric transducer for monitoring a structure.
The proposed work offers new areas of research by converging UAV with a vibration-based method,
as associated problems are required to be solved.

Keywords: unmanned aerial vehicle (UAV); drone; structural health monitoring (SHM); piezoelectric
transducer; non-destructive testing

1. Introduction
Unmanned aerial vehicles (UAVs), commonly known as drones, are gaining a considerable
amount of attention in the field of structural health monitoring (SHM). The visual inspection method
with image processing is a common concept when using drones to find cracks, rusts, or other damage
types that can be visually identified. Due to the nature of the aforementioned method, damages is
usually identified when it has been noticeably progressed. For this reason, a new concept is required
to allow drones to identify damage at an earlier stage, or detect internal damage such as thickness
loss due to corrosion. To date, various authors have investigated structural health monitoring using
the electromechanical impedance (EMI) method, including monitoring of drones. However, there are
only a few studies available where drones are utilized to monitor structures which are focused on
image processing [1–3]. Ellenberg et al. performed an investigation on remote sensing capabilities of a
commercialized UAV (Parrot AR 2.0) for crack detection from various distances [1]. An algorithm was
developed in MATLAB for image post-processing where a field test was conducted on a pedestrian
bridge in order to evaluate the performance of the UAV. Sankarasrinivasan et al. introduced an
approach involving a combination of hat transform and HSV thresholding technique for detecting
cracks using UAV for a real time SHM [2]. Field testing was conducted in the outdoor environment at

Appl. Sci. 2017, 7, 15; doi:10.3390/app7010015 www.mdpi.com/journal/applsci

Appl. Sci. 2017, 7, 15 2 of 9

Vel Tech University in India to evaluate the performance for the proposed study, during which factors
such as wind and random image noises resulted in a few erroneous results. Eschmann et al. used an
octocopter (a UAV with eight blades weighing 2.5 kg) for a building inspection where photos were
taken at high speed and frequency [3]. During the study, more than 12,000 photos were taken over the
four day period for crack inspection of the target structure.
In this work, a new concept of converging a drone with a vibration-based non-destructive
evaluation (NDE) method is proposed. Such a concept will allow one to potentially identify damage
at an earlier stage, reducing the overall maintenance cost of a structure. With the advancement in
measurement technologies, the equipment used for conducting NDE methods are becoming smaller,
lighter, and cheaper. Thus, with the new concept of equipping a drone with a vibration-based NDE
method, possibilities are created for new research areas in the field of remote sensing-based structural
health monitoring.

2. Equipping a Vibration-Based NDE Method onto a Drone

The vibration-based NDE method introduced in this study uses a single piezoelectric (PZT)
material to act as both actuator and sensor. The method is known as the EMI technique, first proposed
by Liang et al. [4]. Currently, this technique requires one to permanently attach a PZT transducer
onto the surface of the target structure, usually in 10 mm square sizes for performing the NDE
method [5–17]. Electrical impedance of the attached PZT transducer is usually measured using
equipment such as 4294A (Agilent Technologies, Santa Clara, CA, USA) costing up to 50,000 USD,
or the AD5933 evaluation board (Analog Devices, Norwood, MA, USA) costing less than 100 USD,
which is light enough for a drone to carry. The attached PZT excites the structure (thickness and radial
vibration modes) within a selected frequency range (usually between 20 and 400 kHz), creating a
standing wave at each frequency, whilst also acquiring data. By observing the 1-D equation introduced
by Liang et al. [4], the measured electrical admittance (inverse of impedance) of the PZT, Y(ω) is
directly related to the mechanical impedance of the host structure, Zs (ω). Thus, any changes in the
structure can be detected by measuring the electrical impedance. The rest of the variables, including
T , δ, d2 and Y E
ω, a, ε33 3x xx represent the excitation frequency geometric constant, dielectric constant,
dielectric loss tangent, coupling constant, and the PZT Young’s modulus.
 
T Zs (ω) 2 E
Y (ω) = iωa ε33 (1 − iδ) − d Y . (1)
Zs (ω) + Za (ω) 3x xx

Data analysis of the EMI method is achieved by comparing two impedance signatures, before and
after damage measurements have been recorded. A statistical technique known as the root mean square
deviation (RMSD) is used to quantify the variations for the two impedance signatures into a single
number. The RMSD equation is shown in Equation (2), where Re( Zk )i is the real part of the reference
impedance signature, and Re( Zk ) j is the real part of the corresponding impedance signature for each
measurement time at the kth measurement step. In general, the real part of an impedance signature is
used as it has been experimentally proven by the authors of this study that it performs better than the
imaginary part for damage identification [18]. Here, large variations in the two impedance signatures
will result in a higher RMSD value.
 h i2 . 1/2
RMSD = ∑ N
k =1 Re( Zk ) j − Re( Zk )i ∑ N
k =1 [ Re ( Zk )i ]
2
. (2)

One of the advantages of the EMI method is the use of the high frequency range, allowing one to
detect small defects. However, the sensing area at this frequency range is relatively small and a large
number of PZT transducers are required for monitoring large infrastructures. The concept introduced
here, for using a drone to monitor structures, requires the modification of the PZT transducer so that
it is re-attachable. This means that it should be able to be temporarily attached to a structure, with
Appl. Sci. 2017, 7, 15 3 of 9

aAppl.
connection strong enough for a successful damage detection. To achieve this, the conventional
Sci. 2017, 7, 15 10.3390/app7010015 3 of 9
method of permanently attaching the PZT is modified so that it allows temporary attachment to a
host
where structure
the PZTusing (5A4E,magnetic
0.5 mmforce. This type
thickness concept is shown
purchased fromin the red box is
piezo.com) in attached
Figure 1,onwhere
top ofthea
PZT (5A4E, 0.5 mm thickness type purchased from piezo.com) is attached
magnet (size 25 mm × 15 mm × 3 mm). This allows the PZT device to be easily attached and detached on top of a magnet (size
25 mma ×
from 15 mm ×
structure, as 3long
mm).as This allowsstructure
the target the PZTisdevice to be easily
ferromagnetic. attachedifand
However, the detached from is
host structure a
structure,
non-metallic as long
(e.g.as the target
concrete, structure wooden
composite, is ferromagnetic.
structureHowever, if the host
etc.), a support magnetstructure
mustisbenon-metallic
attached to
(e.g.,
the host structure, as shown in Figure 1. The experiment setup is also shown in the figure, the
concrete, composite, wooden structure etc.), a support magnet must be attached to wherehosta
structure, as shown in
laptop is connected to Figure 1. Theevaluation
the AD5933 experiment setup
board is also
with the shown
positiveinand thenegative
figure, where
wires aconnected
laptop is
connected
to the PZTtodevice.
the AD5933 evaluation
The drone used board
for thewith theispositive
study and negative
the commercial wiresmodel
product connected
SYMA to the PZT
X5HW,
device. The drone used for the study is the commercial product model SYMA
costing 60 USD. The package provides an application for a smart phone for real time visualization X5HW, costing 60 USD.
The package
through provides
a Wi-Fi an application
connection. The wiresfor of
a smart
the PZTphone for real
device are time
attachedvisualization through
to the drone, wherea Wi-Fi
it is
connection. The
controlled by the user.wires of the PZT device are attached to the drone, where it is controlled by the user.

Figure 1.
Figure 1. Vibration-based
Vibration-based drone
drone structural
structural health
health monitoring
monitoring concept.
concept.

Although the PZT device is connected to the laptop through the AD5933 evaluation board, it is
Although the PZT device is connected to the laptop through the AD5933 evaluation board, it is
possible for one to convert the board into a wireless system, eliminating the USB data cable shown in
possible for one to convert the board into a wireless system, eliminating the USB data cable shown
the figure. Such a system has been investigated by several authors for attempting to create a wireless
in the figure. Such a system has been investigated by several authors for attempting to create a
impedance measuring device using the AD5933 chip. Some of the first wireless impedance
wireless impedance measuring device using the AD5933 chip. Some of the first wireless impedance
measuring devices using this chip were created by Mascarenas et al. (2007) and Overly et al. (2008).
measuring devices using this chip were created by Mascarenas et al. (2007) and Overly et al. (2008).
Mascarenas et al. presented a wireless system using the AD5933 impedance measuring chip with an
Mascarenas et al. presented a wireless system using the AD5933 impedance measuring chip with an
Xbee 2.4 GHz radio (MaxStream, Lindon, Utah, US) and an ATmega128L microcontroller (Atmel,
Xbee 2.4 GHz radio (MaxStream, Lindon, Utah, USA) and an ATmega128L microcontroller (Atmel,
San Jose, CA, US). The performance of the proposed system was demonstrated by comparing the
San Jose, CA, USA). The performance of the proposed system was demonstrated by comparing the
results with those from the data acquisition system (National Instruments, Austin, TX, US),
results with those from the data acquisition system (National Instruments, Austin, TX, USA), measuring
measuring the impedance up to 90 kHz [19]. Overly et al. created an extremely compact wireless
the impedance up to 90 kHz [19]. Overly et al. created an extremely compact wireless impedance
impedance measuring system with dimensions of 52 mm × 3.7 mm, using an ATmega128L
measuring system with dimensions of 52 mm × 3.7 mm, using an ATmega128L microcontroller. The
microcontroller. The performance of the system was compared to an Agilent 4294A impedance
performance of the system was compared to an Agilent 4294A impedance analyzer protecting against
analyzer protecting against corrosion, loosening and tightening bolts. Due to its small size, the
corrosion, loosening and tightening bolts. Due to its small size, the system has a very low power
system has a very low power consumption, which would be suitable for the drone monitoring
consumption, which would be suitable for the drone monitoring concept shown in this study [20].
concept shown in this study [20].
First, since the EMI method is a vibration-based technique, it is vital to investigate whether
First, since the EMI method is a vibration-based technique, it is vital to investigate whether
vibrations caused from the drone (wind or motor spinning) have any effect on the impedance signatures.
vibrations caused from the drone (wind or motor spinning) have any effect on the impedance
In order to check this, five impedance signatures were measured at hovering state in order to evaluate
signatures. In order to check this, five impedance signatures were measured at hovering state in
order to evaluate the repeatability performance of the drone equipped with the EMI method. Figure
2 shows the results of measuring the impedance signatures in the frequency range from 30 to 70
kHz, during which virtually no change in the signatures can be seen. This is an expected result as the
EMI method uses a high frequency range over 20 kHz and vibrations from the drone (e.g., motor
Appl. Sci. 2017, 7, 15 4 of 9

the repeatability performance of the drone equipped with the EMI method. Figure 2 shows the results
of measuring
Appl. the
Sci. 2017, 7, 15 impedance signatures in the frequency range from 30 to 70 kHz, during which
10.3390/app7010015 4 of 9
virtually no change in the signatures can be seen. This is an expected result as the EMI method uses a
high
blade, frequency
wind) arerange over 20
assumed kHz
to be and lower
much vibrations
thanfrom the drone
this value. (e.g.,
Here, themotor blade,
largest wind) between
difference are assumed
any
to be much lower than this value. Here, the largest difference between any two impedance
two impedance signatures did not exceed the value of 0.66%. Thus, this value can be defined as the signatures
did not exceed
threshold value thefor
value
the of 0.66%.
study Thus,
and anythis value
values can be0.66%
below defined as thebe
should threshold
ignoredvalue
whenfor the study
identifying
and any values below 0.66% should be ignored when identifying damage.
damage.

Figure 2.
Figure 2. Impedance
Impedance signatures measured at
signatures measured at drone
drone hovering
hovering state.
state.

Having established that the impedance measurements are barely affected by the vibrations
Having established that the impedance measurements are barely affected by the vibrations created
created from the drone, the next step was to evaluate the repeatability of attaching and reattaching
from the drone, the next step was to evaluate the repeatability of attaching and reattaching the PZT
the PZT device onto a structure. Since the impedance signatures are highly sensitive, reattaching the
device onto a structure. Since the impedance signatures are highly sensitive, reattaching the PZT device
PZT device can have a significant impact on the impedance signature. In order to investigate this
can have a significant impact on the impedance signature. In order to investigate this matter, the PZT
matter, the PZT device is detached and reattached to a metal plate with dimensions of 100 mm × 100
device is detached and reattached to a metal plate with dimensions of 100 mm × 100 mm × 0.15 mm,
mm × 0.15 mm, with impedance signatures being measured each time in the 55 kHz to 70 kHz range.
with impedance signatures being measured each time in the 55 kHz to 70 kHz range. For the purpose
For the purpose of the test, the reattachment was conducted in two ways. First, the reattachment was
of the test, the reattachment was conducted in two ways. First, the reattachment was achieved by
achieved by manually controlling the drone when placing the PZT device at the corner of the metal
manually controlling the drone when placing the PZT device at the corner of the metal plate, where the
plate, where the magnetic force of the device allowed it to be temporarily attach to the surface. This
magnetic force of the device allowed it to be temporarily attach to the surface. This process is shown in
process is shown in Figure 3a where the impedance signatures were measured 10 times for every
Figure 3a where the impedance signatures were measured 10 times for every reattachment of the PZT
reattachment of the PZT device. The second method of detaching and reattaching the PZT device
device. The second method of detaching and reattaching the PZT device was done manually, by using
was done manually, by using the hands to try and achieve exactly the same location placement of the
the hands to try and achieve exactly the same location placement of the device each time, at the corner
device each time, at the corner of the plate. Again, 10 impedance signatures were acquired for a
of the plate. Again, 10 impedance signatures were acquired for a comparison with the results of the
comparison with the results of the previous 10 signatures. Figure 3b shows the RMSD values
previous 10 signatures. Figure 3b shows the RMSD values calculated from these two sets of data,
calculated from these two sets of data, where the first measurements were used as the reference
where the first measurements were used as the reference signatures and the remaining signatures were
signatures and the remaining signatures were then compared to these values. The difference in the
then compared to these values. The difference in the RMSD values for the two sets of data is clearly
RMSD values for the two sets of data is clearly visible, with the drone reattachment method being
visible,
noticeablywithhigher
the drone reattachment
(between 4% and method beingwhen
7%) than noticeably higher
the PZT (between
device 4% and 7%)
is reattached by than when
the hands
the PZT device
(2.5%~3.5%). is reattached
This proves that byreplacement
the hands (2.5%~3.5%).
of the PZT This proves
device ontothat replacement
the same locationofisthe PZT device
a vital factor.
onto the same location is a vital factor. However, the RMSD values in the
However, the RMSD values in the range of 2.5% to 3.5% can still be considered as high, range of 2.5% to 3.5% and
can
still be considered as high, and additional research is required to minimize
additional research is required to minimize the impedance signature variations caused by the impedance signature
variations
replacement caused
of thebyPZTreplacement of the4PZT
device. Figure shows device. Figure 4 shows
the impedance the impedance
signatures signatures
used to obtain used
the RMSD
to obtain the RMSD values just discussed, in the frequency ranges from 55 to 70 kHz.
values just discussed, in the frequency ranges from 55 to 70 kHz. As expected, the reattachment of As expected,
the
the reattachment
PZT device usingof thethe
PZT device
drone using4a)
(Figure thehas
drone (Figure
higher 4a) has
signature higher signature
variations than thosevariations
in Figurethan4b.
those in Figure
Regardless, the 4b. Regardless,
general shape of thethe
general shape signatures
impedance of the impedance signatures
is maintained withisdownward
maintainedpeaks with
downward
being located peaks
at 63being
kHz.located at 63 kHz.
Appl. Sci. 2017, 7, 15 5 of 9
Appl.
Appl. Sci.
Sci. 2017,
2017, 7,
7, 15
15 10.3390/app7010015
10.3390/app7010015 55 of
of 99

Figure
Figure 3.
Figure 3.3. (a)
(a) Piezoelectric
Piezoelectric (PZT)
(PZT) device
device reattachment
reattachment test;
test; (b)
(b) root
root mean
mean square
square deviation
deviation (RMSD)
(RMSD)
results
results for
results for the
for the reattachment
the reattachment test.
reattachment test.
test.

Figure
Figure 4.
4. Reattachment
Reattachment of the PZT device: (a) by the drone; (b) manually.
Figure 4. Reattachment of
of the
the PZT
PZT device:
device: (a)
(a) by
by the
the drone;
drone; (b) manually.
(b) manually.

3.
3. Damage
Damage Detection
Detection Problems
Problems and and Possibilities
Possibilities
3. Damage Detection Problems and Possibilities
In
In the previous section, the reattachment of
the previous section, the reattachment of the
the PZT
PZT device
device caused
caused the the impedance
impedance signatures
signatures to to
vary, In the
which previous
can section,
result in the reattachment
difficulty when of the
identifying PZT device
damage. caused
The highestthe impedance
RMSD values signatures
produced to
vary, which can result in difficulty when identifying damage. The highest RMSD values produced
vary,
for whichdifferent
can result in difficultymethods when identifying damage. The highest RMSD values produced
for the
the twotwo different reattachment
reattachment methods were were approximately
approximately 7% 7% whenwhen usingusing the
the drone
drone and and 3.5%
3.5%
for
for the
manualtwo different
placement. reattachment
Thus, it methods
would be were
ideal forapproximately
a structure 7% when
experiencing using the
damage drone
to have and 3.5%
higher
for manual placement. Thus, it would be ideal for a structure experiencing damage to have higher
for manual
values placement. Thus, itbe would be ideal for a structure experiencing damage to have higher
values as as these
these two
two values
values can can be defined
defined as as the
the threshold
threshold values,
values, in in order
order to to differentiate
differentiate between
between
values
an as these two values can be defined as the threshold values, in order to differentiate between
an intact
intact casecase and
and aa damage
damage case. case. Keeping
Keeping this this inin mind,
mind, thethe purpose
purpose of of this
this section
section is is to
to conduct
conduct twotwo
an intact
experiments case and a damage case. Keeping this in mind, the purpose of this section is to conduct two
experiments involving a progressive damage case and a thickness loss damage case. The setup of the
involving a progressive damage case and a thickness loss damage case. The setup of the
experiments
test involving a progressive damage case and a thickness loss damage case. The metal
setup of the
test specimens
specimens is is shown
shown in in Figure
Figure 5. 5. Figure
Figure 5a 5a shows
shows the the PZT
PZT device
device attached
attached to to the
the metal plate,
plate,
test
where specimens is is shown in Figure 5. Figure 5a showseach the PZT device attached to the metal plate, where
where damagedamage is introduced
introduced by by cutting
cutting 11 mm mm each time time (up
(up to to 20 20 mm),
mm), with
with the the impedance
impedance
damage
signatures is introduced by cutting 1 mm eachIntime (up 5b,
to 20 mm), withare thestacked
impedance signatures being
signatures being measured at every step. In Figure 5b, five plates are stacked with the PZT device
being measured at every step. Figure five plates with the PZT device
measured
attached at every step. In Figure 5b, five plates are stacked with the PZT device attached at one end
attached at at one
one end
end ofof the
the stack.
stack. The
The stack
stack is is held
held together
together by by thethe magnetic
magnetic force force (without
(without any any
of the
additionalstack. The
magnet) stack
of is
the held
PZT together
device, by the
where magnetic
it is force
detached (without
and any
reattached additional
10 times magnet)
for of the
acquiring
additional magnet) of the PZT device, where it is detached and reattached 10 times for acquiring
PZT device, where it is After
impedance detached and reattached 10 times for acquiring impedance signatures. After
impedance signatures.
signatures. After measuring
measuring the the impedance
impedance signature,
signature, aa single
single layer
layer ofof the
the metal
metal plate
plate isis
measuring
removed the impedance signature, a single layer of the metal plate is removed from the stack and
removed from from the
the stack
stack andand thethe PZT
PZT device
device is is detached
detached and and reattached
reattached another another 10 10 times,
times, withwith the
the
the PZT device is detached
impedance and reattached anotherThis 10 times, with the impedance signatures being
impedance signatures
signatures being
being measured
measured at at each
each step.
step. This process
process isis repeated
repeated until until the
the last
last plate
plate of
of the
the
measured
stack is at each
remaining, step. This
totaling process
50 is
impedance repeated until
signatures. the last plate of the stack is remaining, totaling
stack is remaining, totaling 50 impedance signatures.
50 impedance signatures.
Appl. Sci. 2017, 7, 15 6 of 9
Appl. Sci. 2017, 7, 15 10.3390/app7010015 6 of 9
Appl. Sci. 2017, 7, 15 10.3390/app7010015 6 of 9

Figure 5. Experiment setup for: (a) the progressive damage case; (b) thickness reduction case.
Figure 5. Experiment
Figure 5. Experiment setup
setup for:
for: (a)
(a) the
the progressive damage case;
progressive damage case; (b)
(b) thickness
thickness reduction
reduction case.
case.

Figure 6a
Figure 6a shows
shows the the impedance
impedance signatures
signatures for for the
the experiment
experiment in in Figure
Figure 5a, 5a, where
where 21 21 impedance
impedance
Figure(including
signatures 6a shows the the impedance
reference signaturescan
signature) forbe the experiment
observed. in Figure
Although the 5a, where 21vary
signatures impedance
subject
signatures (including the reference signature) can be observed. Although the signatures vary subject
signatures
to progressive(including
damage, thethe reference
intensitysignature)
of can bevariations
signature observed.are Although
less the signatures
severe compared vary
to subject
Figure 4.
to progressive damage, the intensity of signature variations are less severe compared to Figure 4.
to progressive
Thus, this
this causesdamage,
causes difficulty the intensity
difficulty when of signature
when identifying
identifying damagevariations
damage as are less
as damaged severe
damaged structure compared
structure results to Figure
results in 4. Thus,
in smaller
smaller
Thus,
this causesvariations
signature difficultythan when identifying
variations damage
caused from asreattachment
damaged structureof the PZTresults in smaller
device. Figure signature
6b shows
signature variations than variations caused from reattachment of the PZT device. Figure 6b shows
variations
RMSD thanby
values variations
selecting caused
the fromcase
intact reattachment
as the of the PZT
reference device.with
signature, Figure all 6b shows
other RMSD values
signatures being
RMSD values by selecting the intact case as the reference signature, with all other signatures being
by selecting
compared to the
to this intact
this reference.case
reference. The as the reference
The first
first RMSD signature,
RMSD valuevalue which with
which is all other
is calculated signatures
calculated from from the being
the initialcompared
initial 11 mm
mm cut to
cut
compared
this
damagereference.
resultsThe first RMSD
in 1.01%
1.01% value which
(red square
square dot), andis calculated
and the remaining fromRMSD
remaining the initial 1 mm
values rangecut between
damage results
1.4% and in
and
damage results in (red dot), the RMSD values range between 1.4%
1.01% (red
1.8% (purple square
(purple circle dot),
circle dots). and
dots). One the remaining
One possible RMSD
possible explanation
explanation for values
for therange between
the relatively
relatively low 1.4%
low RMSD and
RMSD result 1.8% (purple
result (1.01%)
(1.01%) for circle
for the
the
1.8%
dots).
first One possible
RMSD value explanation
is that the for the relatively
damage was low RMSD
conducted result
near a (1.01%)
node (a for the first
point that RMSD
vibratesvaluewithis
first RMSD value is that the damage was conducted near a node (a point that vibrates with
that the damage
minimum was conducted
amplitude), causing near avariations
small node (a point in thatimpedance
the vibrates with minimum
signature. Usingamplitude),
a line ofcausing
best fit
minimum amplitude), causing small variations in the impedance signature. Using a line of best fit
small variations
(excluding the firstin the
first RMSD impedance signature.
value), shows
shows Using a line of best fit (excluding the first RMSDFor value),
(excluding the RMSD value), aa rising
rising trend
trend with
with an increase
an increase in damage
in damage intensity.
intensity. this
For this
shows a rising
progressive trend with
damage case, antheincrease
RMSD invalues
damage areintensity.
smaller For
thanthis
the progressive
values damage case,
calculated from the RMSD
manually
progressive damage case, the RMSD values are smaller than the values calculated from manually
values are smaller
reattaching the PZT than
PZT the values
device calculated
(2.5%~3.5%). fromamanually
Thus, different reattaching
approach other the PZT
other than device
using(2.5%~3.5%).
the RMSD
RMSD
reattaching the device (2.5%~3.5%). Thus, a different approach than using the
Thus, a different
equation, needs
needs to approach
to be
be researched other
researched in than using
in order
order to the RMSD
to differentiate equation,
differentiate between
between anneeds to
an intact be
intact caseresearched
case and in order
and aa progressive
progressive to
equation,
differentiate
damage casecasebetween
using the theanpresented
intact case
presented PZTanddevice.
a progressive damage case using the presented PZT device.
damage using PZT device.

Figure 6. Progressive damage case: (a)

Progressive damage Impedance signatures; (b) RMSD
RMSD results.
Figure 6.
Figure 6. Progressive damage case: (a) Impedance
case: (a) Impedance signatures;
signatures; (b)
(b) RMSD results.
results.

Figure 7a
Figure 7a shows
shows 5050 impedance
impedance signatures
signatures measured
measured for thethe experiment conducted
conducted in Figure
Figure 5b.
Figure 7a shows 50 impedance signatures measured for for the experiment
experiment conducted in in Figure 5b.
5b.
When observing
When observing this,
observing this, a trend
this, aa trend exists
trend exists for
exists for the
for the resonance
the resonance peaks
resonance peaks located
peaks located
located atat 35
at 35 kHz
35 kHz and
kHz and 63
and 63 kHz,
63 kHz, where
kHz, where
where
When
both peaks
both peaks show
show aa decrease
decrease in
in the
the amplitudes
amplitudes and
and a shift
shift to
to the
the left
left direction
direction as
as the
the thickness
thickness ofof the
the
both peaks show a decrease in the amplitudes and aa shift to the left direction as the thickness of the
metal stack layer is reduced. Figure 7b shows the RMSD values calculated using the first
metal stack layer is reduced. Figure 7b shows the RMSD values calculated using the first impedance impedance
signature, with
signature, with the
the remaining
remaining signatures
signatures being
being compared
compared to to this
this value.
value. The
The red
red dotted
dotted lines
lines in
in the
the
Appl. Sci. 2017, 7, 15 7 of 9

metal stack
Appl. Sci. 2017, layer is reduced. Figure 7b shows the RMSD values calculated using the first impedance
7, 15 10.3390/app7010015 7 of 9
signature, with the remaining signatures being compared to this value. The red dotted lines in the
figure separate
figure separatethe theRMSD
RMSDvaluesvaluesfrom
from a stack
a stack of of
fivefive metal
metal layers,
layers, to ato a single
single metalmetal
layerlayer
(first(first
nine nine
dots
dotsmeasured
are are measured from from
a stacka stack withlayers,
with five five layers,
the next the10next
dots10aredotsfromareafrom
stackawith
stackfour
with four and
layers, layers,
so
and so Here,
forth). forth).there
Here, is there
a clearisdifference
a clear difference
in the RMSD in the RMSD
values forvalues for thenumber
the different differentofnumber of layers,
layers, where five
where (0.75
layers five layers (0.75 mm)
mm) result result
in values ofin values four
2%~4%, of 2%~4%,
layers four layers three
in 4%~7%, in 4%~7%,
layersthree layers in
in 6%~10% 6%~10%
(except for
(except
the 23rdfor the 23rd measurement),
measurement), two layers in two layers inand
10%~14%, 10%~14%,
one layer and inone layer inThus,
12%~18%. 12%~18%. Thus,
for this metalforstack
this
metalone
case, stack cancase,
first one can the
acquire firstimpedance
acquire the impedance
signature signature
at five layers for at five
use aslayers for usesignature,
a reference as a referencethen
signature,
detach then device
the PZT detachfor the PZT
it to device
be used for in
again it the
to future.
be usedWhen againtheinstackthe experiences
future. When the stack
reduction in
experiences
thickness, thereduction
PZT deviceinshould thickness, the PZT
be attached device
to the sameshould
spot forbe attachedthe
measuring toimpedance
the same signature
spot for
measuring
again. If thethe impedance
stack thickness signature again.
has reduced to Ifonly
the two
stacklayers,
thicknessa highhaspossibility
reduced to onlyfor
exists twothelayers,
RMSDa
high possibility
value to be within exists for theofRMSD
the range valueIn
10%~14%. toaddition,
be withinone thecanrange of 10%~14%.
increase In addition,
the accuracy one can
of the result by
increasemeasuring
simply the accuracy of the resultsignature
the impedance by simply measuring
multiple timesthe impedance
before averaging signature
the RMSD multiple
values,times
as a
before measurement
single averaging thecan RMSDresultvalues, as a single RMSD
in an unexpected measurement can result
value, leading to a in an unexpected
wrong prediction. RMSDBased
value,
on thisleading
finding,toitabecomes
wrong prediction.
possible for Based
one toon identify
this finding, it becomes
thickness reductionpossible for one tosubjected
of structures identify
thickness
to corrosion reduction
such as of structures
bridges, subjected
buildings, and to corrosion
offshore such asbybridges,
structures, using abuildings,
drone equippedand offshore
with a
structures, byPZT
re-attachable using a drone equipped with a re-attachable PZT device.
device.

Figure
Figure 7.
7. Thickness
Thickness reduction
reduction case:
case: (a)
(a) Impedance
Impedance signatures;
signatures; (b) RMSD results.
(b) RMSD results.

4. Conclusions
4. Conclusions
This study proposes a new concept of employing UAV for structural health monitoring of civil
This study proposes a new concept of employing UAV for structural health monitoring of
infrastructures, using a vibration-based NDE method. The concept surpasses the current limitations
civil infrastructures, using a vibration-based NDE method. The concept surpasses the current
of using a drone in combination with a visual inspection method and image processing. The
limitations of using a drone in combination with a visual inspection method and image processing.
vibration-based method can detect small damages which are difficult to identify, and internal
The vibration-based method can detect small damages which are difficult to identify, and internal
damages such as thickness reduction. The key idea of the concept is to make a PZT transducer that
damages such as thickness reduction. The key idea of the concept is to make a PZT transducer that
can be temporarily attached and detached to a structure, which is a different way of conducting the
can be temporarily attached and detached to a structure, which is a different way of conducting
EMI method, in which the PZT transducer is permanently attached to the host structure for damage
the EMI method, in which the PZT transducer is permanently attached to the host structure for
identification.
damage identification.
The first part of the study investigates the effect of measured impedance signature while the
The first part of the study investigates the effect of measured impedance signature while the drone
drone is in a hovering state. Due to the high frequency usage of the NDE method, the effect on the
is in a hovering state. Due to the high frequency usage of the NDE method, the effect on the impedance
impedance signature can be ignored. The second part of the study evaluated the repeatability
signature can be ignored. The second part of the study evaluated the repeatability performance of the
performance of the PZT device which is attached, detached, and reattached, to a stack of metal
PZT device which is attached, detached, and reattached, to a stack of metal plates, where the RMSD
plates, where the RMSD values were calculated and analyzed. The last section of the study
values were calculated and analyzed. The last section of the study investigated the performance of the
investigated the performance of the PZT device against two different damage types. During the first,
PZT device against two different damage types. During the first, the PZT device was attached to a
the PZT device was attached to a metal plate with 1 mm progressive damage being introduced up to
metal plate with 1 mm progressive damage being introduced up to 20 mm, and during the second,
20 mm, and during the second, the PZT device was tested against a stack of metal plates subjected to
the PZT device was tested against a stack of metal plates subjected to a thickness loss. The RMSD
a thickness loss. The RMSD values for the progressive damage case were noticeably less than the
RMSD values calculated from the signatures acquired by reattaching the PZT device on an
undamaged plate. Thus, more research is needed to solve this problem. However, the case of
thickness reduction on a metal stack showed results which validates the drone SHM concept. For
each reduction in metal layers, the RMSD values were within a specific range for all five layers.
Appl. Sci. 2017, 7, 15 8 of 9

values for the progressive damage case were noticeably less than the RMSD values calculated from
the signatures acquired by reattaching the PZT device on an undamaged plate. Thus, more research
is needed to solve this problem. However, the case of thickness reduction on a metal stack showed
results which validates the drone SHM concept. For each reduction in metal layers, the RMSD values
were within a specific range for all five layers. These were 2%~4%, 4%~7%, 6%~10%, 10%~14% and
12%~18% for 5 layers, 4, 3, 2, and 1 layer, respectively. Therefore, through a consideration of the
concept of using a drone equipped with a vibration-based NDE method, new possibilities have been
realized and existing problems will need solutions for turning this new concept into reality.

Acknowledgments: The research was supported by a grant from “Next-generation high speed transfer system
based technology development (20160582-001)” funded by Korea Institute of Civil Engineering and Building
Technology (KICT), Korea.
Author Contributions: Wongi S Na is the principal author for the research involved in writing the manuscript,
performing experiments and analyzing data; Jongdae Baek contributed in conducting experiments and
analyzing data.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ellenberg, A.; Branco, L.; Krick, A.; Bartoli, I.; Kontsos, A. Use of Unmanned Aerial Vehicle for Quantitative
Infrastructure Evaluation. J. Infrastruct. Syst. 2014, 21, 04014054. [CrossRef]
2. Sankarasrinivasan, S.; Balasubramanian, E.; Karthik, K.; Chandrasekar, U.; Gupta, R. Health Monitoring of
Civil Structures with Integrated UAV and Image Processing System. Procedia Comput. Sci. 2015, 54, 508–515.
[CrossRef]
3. Eschmann, C.; Kuo, C.M.; Kuo, C.H.; Boller, C. Unmanned aircraft systems for remote building inspection
and monitoring. In Proceedings of the 6th European Workshop on Structural Health Monitoring, Dresden,
Germany, 3–6 July 2012.
4. Liang, C.; Sun, F.P.; Rogers, C.A. Coupled electromechanical analysis of adaptive material system
determination of the actuator power consumption and system energy transfer. J. Intell. Mater. Syst. Struct.
1994, 5, 2–20. [CrossRef]
5. Na, W.S. Progressive damage detection using the reusable electromechanical impedance method for metal
structures with a possibility of weight loss identification. Smart Mater. Struct. 2016, 25, 55039–55048.
[CrossRef]
6. Song, G.; Li, H.; Gajic, B.; Zhou, W.; Chen, P.; Gu, H. Wind turbine blade health monitoring with
piezoceramic-based wireless sensor network. Int. J. Smart Nano Mater. 2013, 4, 150–166. [CrossRef]
7. Yang, Y.; Lim, Y.Y.; Soh, C.K. Practical issues related to the application of the electromechanical impedance
technique in the structural health monitoring of civil structures: I. Experiment. Smart Mater. Struct. 2008, 17,
035008. [CrossRef]
8. Annamdas, V.G.; Radhika, M.A. Electromechanical impedance of piezoelectric transducers for monitoring
metallic and non-metallic structures: A review of wired, wireless and energy-harvesting methods. J. Intell.
Mater. Syst. Struct. 2013. [CrossRef]
9. Na, W.S. Distinguishing crack damage from debonding damage of glass fiber reinforced polymer plate
using a piezoelectric transducer based nondestructive testing method. Compos. Struct. 2017, 159, 517–527.
[CrossRef]
10. Song, G.; Mo, Y.L.; Otero, K.; Gu, H. Health monitoring and rehabilitation of a concrete structure using
intelligent materials. Smart Mater. Struct. 2006, 15, 309. [CrossRef]
11. Na, S.; Lee, H.K. Neural network approach for damaged area location prediction of a composite plate using
electromechanical impedance technique. Compos. Sci. Technol. 2013, 88, 62–68. [CrossRef]
12. Gulizzi, V.; Rizzo, P.; Milazzo, A. Electromechanical Impedance Method for the Health Monitoring of Bonded
Joints: Numerical Modeling and Experimental Validation. Struct. Durab. Health Monit. 2014, 10, 19–54.
13. Liang, Y.; Li, D.; Parvasi, S.M.; Kong, Q.; Song, G. Bond-slip detection of concrete-encased composite
structure using electro-mechanical impedance technique. Smart Mater. Struct. 2016, 25, 095003. [CrossRef]
Appl. Sci. 2017, 7, 15 9 of 9

14. Rugina, C.; Toader, A.; Giurgiutiu, V.; Ursu, I. The electromechanical impedance method for structural health
monitoring of thin circular plates. Proc. Romanian Acad. Ser. A Math. Phys. Tech. Sci. Inf. Sci. 2014, 15,
272–282.
15. Wang, D.; Song, H.; Zhu, H. Embedded 3D electromechanical impedance model for strength monitoring of
concrete using a PZT transducer. Smart Mater. Struct. 2014, 23, 115019. [CrossRef]
16. Na, S.; Lee, H.K. A multi-sensing electromechanical impedance method for non-destructive evaluation of
metallic structures. Smart Mater. Struct. 2013, 22, 095011. [CrossRef]
17. Na, W.S.; Lee, H. Experimental investigation for an isolation technique on conducting the electromechanical
impedance method in high-temperature pipeline facilities. J. Sound Vib. 2016, 383, 210–220. [CrossRef]
18. Sun, F.P.; Chaudhry, Z.; Liang, C.; Rogers, C.A. Truss structure integrity identification using PZT
sensor-actuator. J. Intell. Mater. Syst. Struct. 1995, 6, 134–139. [CrossRef]
19. Mascarenas, D.L.; Todd, M.D.; Park, G.; Farrar, C.R. Development of an impedance-based wireless sensor
node for structural health monitoring. Smart Mater. Struct. 2007, 16, 2137. [CrossRef]
20. Overly, T.G.; Park, G.; Farinholt, K.M.; Farrar, C.R. Development of an extremely compact impedance-based
wireless sensing device. Smart Mater. Struct. 2008, 17, 065011. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).