sensors

Article
Preload Monitoring in Bolted Connection Using
Piezoelectric-Based Smart Interface
Thanh-Canh Huynh ID
, Ngoc-Loi Dang and Jeong-Tae Kim *
Department of Ocean Engineering, Pukyong National University, Busan 48723, Korea;
ce.huynh@gmail.com (T.-C.H.); loi.ngocdang@gmail.com (N.-L.D.)
* Correspondence: idis@pknu.ac.kr; Tel: +82-51-629-6585




Received: 2 August 2018; Accepted: 21 August 2018; Published: 22 August 2018


Abstract: In this study, a preload monitoring method using impedance signatures obtained from a
piezoelectric-based smart interface is presented for bolted girder connections. Firstly, the background
theory of the piezoelectric-based smart interface and its implementation into the health monitoring of
bolted connections are outlined. A simplified electro-mechanical (EM) impedance model of a smart
interface-embedded bolted connection system is formulated to interpret a mechanistic understanding
of the EM impedance signatures under the effect of bolt preload. Secondly, finite element modeling
of a bolted connection is carried out to show the numerical feasibility of the presented method,
and to predetermine the sensitive frequency band of the impedance signatures. Finally, impedance
measurements are conducted on a lab-scaled bolted girder connection, to verify the predetermined
sensitive frequency range and to assess the bolt preload changes in the test structure.

Keywords: preload monitoring; bolted connection; bolt-loosening; piezoelectric sensor; impedance

response; smart interface

1. Introduction
Bolting is a widely-accepted method for making connections in steel structures in the field.
Bolts are torqued to a high tensile stress, developing clamping pressures at the interfaces of the
structural members to hold them in position. Followed by the use of high-strength bolts, this fastening
method has enabled the advantages of easy installation, time efficiency, and a high strength for field
connections. After a long-term service life, however, the bolted connections could experience a loss
of preloads (i.e., self-loosening) due to repetitive external forces and vibrations, which threaten their
functionality. Therefore, bolt preload monitoring is essential and recently gained growing interest
in efforts to ensure the safety of bolted joints, and to prevent the catastrophic failures of the entire
structures [1–6].
To assess the structural integrity of the local critical members in the mechanical and civil systems,
there have been many research attempts on the impedance-based method [7–13]. The fundamental
part of the method is to utilize electromechanical (EM) impedance responses as local dynamic features
for assessing the structural damage. The frequency band used in the impedance-based method is often
in the ultrasonic range, hence the method is able to effectively capture incipient damages. Owing to
the advantage associated with the use of high-frequency responses, the impedance-based method has
been applied for the health assessment of bolted joints [8,14–18].
Bolt-loosening in a bolted joint can be monitored via its EM impedance responses, measured
by piezoelectric sensors or piezoelectric washers [15,18–22]. Because the EM impedance correlates
with the structural properties of a bolted connection, any damage occurrence could be detected
via observing the changes in measured impedance data. From the previous research attempts on
the impedance-based bolt-loosening monitoring, an important question has been raised on how to

Sensors 2018, 18, 2766; doi:10.3390/s18092766 www.mdpi.com/journal/sensors

Sensors 2018, 18, 2766 2 of 20

identify an effective frequency band of impedance signatures that is sensitive to the preload change or
bolt-loosening. In real situations, the effective frequency band is often determined by trial-and-error,
because it is dependent on the local dynamic characteristics of a monitored structure.
The mountable interface technique can be a potential solution to cope with the above-mentioned
problem [23]. This technique uses an interfacial structure equipped with a piezoelectric sensor (e.g., PZT
(lead zirconate titanate)) to indirectly acquire the sensitive impedance data from a target structure.
The geometry and material properties of the interface should be appropriately designed so that the
sensitive impedance response is occurred within a pre-defined frequency band. For the damage
monitoring of bolted joints, the mountable interface technique could offer unique advantages in
comparison with the piezoelectric washer technique [20–22]. Firstly, the mountable interface can be
post-installed into an existing connection, whereas the piezoelectric washer requires pre-installation
during the construction. Secondly, a single mountable interface can be used to monitor multiple
bolts in a connection, meanwhile, a single piezoelectric washer is particularly fit with a single
bolt. Thus, the use of the mountable interface technique could reduce the number of sensing
channels for impedance monitoring of a large bolted connection. However, the previous studies
have mainly focused on developing the mountable interface technique for the health monitoring of
tendon-anchorage systems [23–26]. So far, the effectiveness of the mountable interface technique for
bolt-loosening detection problems has not been evaluated. Additionally, the mechanistic understanding
of the impedance response under bolt-loosening has not been sufficiently explained via a mathematical
model that considers the effect of a bolted connection. Also, there is a need to identify the sensitive
frequency band for the impedance monitoring of a bolted connection by using finite element modeling.
In this study, a PZT interface-based impedance monitoring method is developed to detect the
bolt-loosening events in a bolted connection. To demonstrate the theoretical feasibility of the presented
method, a simplified impedance model is newly designed with the consideration of the bolt preload
effect. Next, the sensitive frequency band of the EM impedance responses is numerically predetermined
for a bolted connection example embedded with a PZT interface. Finally, impedance measurements
are conducted on a lab-scaled bolted girder connection to verify the pre-analyzed sensitive frequency
band and to evaluate the effectiveness of the proposed method for bolt-loosening detection.

2. Piezoelectric-Based Smart Interface for Bolted Connection

2.1. Piezoelectric-Based Smart Interface Technique

An impedance monitoring method using the PZT interface technique is designed in order to
acquire the impedance data with predetermined sensitive frequency bands from bolted connections.
As shown in Figure 1a, the PZT interface prototype is a plate-like structure, having two outside bonded
sections and a middle flexible section, that is embedded with a PZT sensor. The flexible section is
intentionally made to provide free vibrations during the PZT’s excitation. The bonded sections allow
the PZT interface prototype to be mountable and easily reconfigured if needed.
To monitor multiple bolts in a connection, the PZT interface should be mounted to the splice plate
connection, which is clamped by the bolt preloads, as shown in Figure 1b. Under the PZT’s excitation,
there are coupled interactions between the PZT and the interface, and then between the PZT interface
and the connection splice plate. The coupling between the PZT interface and the splice plate opens
a potentiality to assess multiple loosened bolts on the splice plate. The flexible section of the PZT
interface allows for predetermining the sensitive frequency band of impedance signals below 100 kHz,
and thus enabling the use of a low-cost wireless impedance measurement system [22,27,28].
In equilibrium, the bolt preloads can be transformed into contact pressures and bearing stresses at the
contact between the main structure and the splice plate, as seen in Figure 1b. According to the previous
studies [17,29], the contact parameters of the bolted connection can be represented by a system of a spring
and dashpot (kc , cc ), whose values represent the amount of bolt preloads, see Figure 1c. At the PZT driving
point, the interface can be modeled with the mass, stiffness, and damping parameters (mi , ki , and ci ) and
Sensors 2018, 18, x FOR PEER REVIEW 3 of 20

Sensors 2018, 18, 2766 3 of 20

1c. At the PZT driving point, the interface can be modeled with the mass, stiffness, and damping
parameters (mi, ki, and ci) and the splice plate can be also modeled by the respectively structural
parameters
the splice plate(mcan
s, kbe
s, and cs). When
also modeled bythe bolt preloads
the respectively are changed,
structural parametersthe(mcontact parameters
s , ks , and cs ). When theof bolt
the
connection
preloads are altered
are changed, the (e.g.,
contactcontact stiffness
parameters of thereduction), leading
connection are to(e.g.,
altered the contact
variation in the
stiffness coupled
reduction),
responses
leading of variation
to the the system at resonance.
in the By monitoring
coupled responses the impedance
of the system at resonance. responses
By monitoring of thethesystem in the
impedance
resonant band,
responses of the it is possible
system in the to detect the
resonant band,bolt looseness
it is possible or preload
to detect thechanges that have
bolt looseness occurred
or preload in the
changes
connection.
that have occurred in the connection.

(a) A prototype of PZT interface.

(b) Bolted connection equipped with a PZT interface.

PZT Interface Z(ω)
Impedance

Bolt Preload F PZT F F+δF

F
mi Frequency

ki ci Splice Plate
(k s , cs , ms )

kc cc
(c) PZT interface-bolted connection interacting system.

Figure 1. Impedance
Figure 1. Impedance monitoring
monitoring method
method for
for bolted
bolted connection
connection via
via PZT
PZT interface.
interface.

2.2.
2.2. Analytical Modeling
Modeling of
of Piezoelectric-Based
Piezoelectric-Based Smart
Smart Interface
Interface

2.2.1.
2.2.1. Impedance
Impedance Response
Response of
of Bolted
Bolted Connection
Connection
The
The impedance
impedance responsesresponses of of aa bolted
bolted connection
connection measured
measured via via the the PZTPZT interface
interface can
can bebe
theoretically derived from a simplified impedance model. Based on the
theoretically derived from a simplified impedance model. Based on the previous studies [10,30], a previous studies [10,30],
atwo-dof
two-dof(degree
(degreeofoffreedom)
freedom)impedance
impedancemodel
model with
with the
the consideration
consideration of of the
the contact
contact parameters
parameters
representing
representing the bolt preload is proposed, as shown in Figure 2. In the model, one dof refers to
the bolt preload is proposed, as shown in Figure 2. In the model, one dof refers to the
the
interface (mii,, cci,, and
interface (m and kki),
), and
and the
the other
other dof
dof refers
refers to
to the
the splice
splice plate (mss,, ccss,, and
plate (m and kkss)) with
with the
the contact
contact
parameters
parameters (c and kkcc).
(ccc and ).
Sensors 2018, 18, 2766 4 of 20
Sensors 2018, 18, x FOR PEER REVIEW 4 of 20

Z eq (ω )
Z a (ω )
Interface Bolted Connection
I (ω )
mi ms kc
f i (ω )
V (ω ) ~ PZT
ki ks
ci cc
cs
ui us
Figure
Figure 2.
2. Impedance
Impedance model
model of
of the
the PZT
PZT interface-bolted
interface-bolted connection
connection system.
system.

As illustrated in Figure 2, when the PZT sensor is excited by a harmonic voltage, V(ω), with a
As illustrated in Figure 2, when the PZT sensor is excited by a harmonic voltage, V(ω), with a
current, I(ω), a harmonic force is introduced into the system at the PZT driving point. The equation
current, I(ω), a harmonic force is introduced into the system at the PZT driving point. The equation of
of motion under the PZT’s harmonic force = jωt can be given as follows:
motion under the PZT’s harmonic force f i = Fi e can be given as follows:
m u + ci (ui − u s ) + ki (ui − u s ) = f i
.. i i . .
ccs ccc . mi ui + ci (ui −k uks ) + k i (ui − us ) = f i (1)
(1)
s ussu+
mm  s + s c uus s−−cci (i (uuii−−uuss))++ ks ksc uc s −
.. . .
u sk−i (ukii (−uiu−s )u s=) =0 0
c s + cc
cs +cc k s + kc
k s +k c

. .. . ..
where ui ,, ui ,, ui and
where and us ,, us ,, us are
are the
the displacements,
displacements, velocities,
velocities, and
and accelerations corresponding to
accelerations corresponding to
masses m and m , respectively.
masses mi and ms, respectively.
i s
Under the
Under the harmonic
harmonic excitation
excitation force,
force, the
the steady
steady states
states of
of the
the interface
interface and
and the
the splice
splice plate
plate can
can be
be
described by
described by the
the following:
following:
ui = Ui e jωt
ωt (2)
usi ==UUi es ejjωt
u
(2)
where Ui and Us are complex quantities thatu are U s e jω t
s = dependent on the excited frequency and structural
parameters
where Ui and of U the system. By substituting Equation (2) into Equation (1), equations to obtain Ui and Us
s are complex quantities that are dependent on the excited frequency and structural
of the system are
parameters of the system.given as follows:
By substituting Equation (2) into Equation (1), equations to obtain Ui and
Us of the system are given as follows: 2


 −ω mi + jωci + k i Ui − ( jωci + k i )Us = Fi
−ω
−( jωci + k i )Ui + −ω 2 m 2
s +mjω ( c + c+ (
i + ijω cics+
s cc
ckci) U k i(+jωkksc+
+ i(− s kc
)
i k+c ki ) U
) Us s == F
0i
(3)

 2 k k  the ratio between(3)

cs cc Zeq , is defined
− ( jωmechanical
The equivalent ci + ki ) U i +impedance
 −ω ms + jω (ci +
of the system, ). + ( ki + s c )as U s = 0
the
excitation force fi and the velocity at c
 the PZT driving point
s + c k
cui , given as + k c 
s follows:

The equivalent mechanical impedance off ithe system,Fi e jωt Zeq, is defined as the ratio between the
Z eq = . =
excitation force fi and the velocity at the PZT driving point (4)
u jωU e jωt, given as follows:
i i
jωt
By solving Equation (3), the quantity, f Ui , is
Fei obtained. By substituting the obtained Ui (4)
into
Zeq = i = jωt
ui jωUof
Equation (4), the equivalent mechanical impedance i e the system Zeq is obtained as follows:

By solving Equation (3), the quantity,

2m U i, is obtained. By(ksubstituting thejωc
( obtained 2 Ui into

 cs cc 1

−ω 2 mi + jωc i + k i − ω s + jω ( c i + cs +cc ) + i + k s 1+ ξ ) − i + ki )
EquationZ(4),
eq = the equivalent mechanical
 impedance of the system Z eq is obtained
 as follows: (5)
jω −ω 2 ms + jω (ci + ccs + s cc 1
c c ) + ( k i + k s 1+ ξ )
 cc 1 
( )
−ω 2 mi + jω ci + ki  −ω 2 ms + jω (ci + s c ) + (ki + k s
c + c 1 + ξ
)  − ( jω ci + ki )
2

where ξ = ks /kc is defined as the ratio

Z eq =  between the splice s c plate’s stiffness and the contact stiffness.
(5)
ξ ≈ 0 indicates the infinitive value of contact stiffness c c
(i.e., 1  while ξ ≈ ∞ indicates
fixed boundary),
jω  −ω 2 ms + jω (ci + s c ) + (ki + k s )
the unnoticeable value of the contact stiffness (i.e., cfree
 s + cc
boundary).1If + ξthe splice plate’s stiffness, ks ,
remains unchanged, the increment of the ratio ξ = ks /kc will be equivalent to the decrement of the
contactξstiffness,
where = ks/kc is defined ascan
kc , which theberatio between as
interpreted thethesplice
bolt plate’s
preloadstiffness
reduction and(i.e.,
thebolt
contact stiffness. ξ ≈
looseness).
0 indicates the infinitive value of contact stiffness (i.e., fixed boundary), while ξ ≈ ∞ indicates the
unnoticeable value of the contact stiffness (i.e., free boundary). If the splice plate’s stiffness, ks,
Sensors 2018, 18, 2766 5 of 20

The EM impedance, Z(ω), of the bolted connection measured via the PZT interface is a combined
function of the equivalent mechanical impedance of the interface-bolted connection system, Zeq , and the
mechanical impedance of the PZT sensor, Za , given by [30,31] the following:
  −1
wa la T 2 E Za ( ω ) 2 E tan( kl a )
Z (ω ) = jω ε 33 (1 − jδ) − d31 Ŷ11 + d Ŷ (6)
ta Za (ω ) + Zeq (ω ) 31 11 kla

E = (1 + jη )Y E is the complex Young’s modulus of the PZT sensor (width w , length l ,

where Ŷ11 11 a a
and thickness t a ) at the zero electric field; ε T33 is the dielectric constant at the zero stress; d31 is the
piezoelectric coupling constant in the 1-direction at the zero stress; the terms η and δ are the structural
damping q loss factor and the dielectric loss factor of the PZT. The wave number of the PZT is given
as k = ω E , where ρ is the mass density of the PZT. The mechanical impedance of the PZT is
ρ/Y11
computed as Za = − jŶ11 E w t /ωl .
a a a
From Equations (5) and (6), it has been shown that the impedance response, Z(ω), measured
via the PZT interface, would contain the structural parameters of the interface and the connection.
Thus, any damage that occurred in the bolted connection (e.g., preload change) can be diagnosed by
tracking the variation in the impedance response, Z(ω).

2.2.2. Impedance Response versus Preload Change

To demonstrate the theoretical feasibility of the PZT interface technique for the impedance monitoring
of a bolted connection, an example of the simplified two-dof model was investigated. The PZT sensor
has the following dimensions: wa = 15 mm, la = 15 mm, and ta = 0.51 mm; and the following properties:
ρ = 7750 kg/m3 , Y11E = 6.098 × 1010 N/m2 , εT = 1.505 × 10−8 Farads/m, d = −1.71 × 1010 m/V, δ = 0.015,
33 31
and η = 0.0125. The interface has the following structural properties: mi = 0.1 kg, ki = 2 × 109 N/m,
ci = 200 N/ms−1 ; and the connection splice plate has the following structural properties: ms = 1 kg, ks = 2 ×
1010 N/m, cs = 200 N/ms−1 . The contact damping is assumed as cc = 500 N/ms−1 . Assuming that the
splice plate was undamaged (i.e., ks remained constant), the bolt preload reduction can be simulated by
increasing the stiffness ratio ξ = ks /kc (i.e., kc was reduced with respect to ks ).
Figure 3 shows the real and imaginary impedance responses, Z(ω), of the two-dof model when
the contact stiffness was infinitive (ξ = ks /kc = ks /∞ = 0). Two resonant peaks (i.e., Peak 1 at 18.96 kHz
and Peak 2 at 31.42 kHz) can be clearly observed from the impedance responses, representing the two
coupling responses of the system. It is noted that the resonant impedance peaks represent the significant
contributions of the equivalent structural impedance, Zeq , to the total impedance Z(ω) (see Equation (6)).
The effect of the bolt preload reduction on the impedance responses was investigated for the
different stiffness ratio, ks /kc , in the range of 0–0.25, as plotted in Figure 4a. The changes in the real
impedance values of the two impedance peaks were zoomed in Figure 4b,c, respectively. From the
figures, it can be seen that as the ratio, ks /kc , was increased from 0 to 0.25 (i.e., the contact stiffness, kc ˆ,
was reduced), the two resonant peaks clearly shifted to the left side, indicating the reduction in the
resonant frequencies of the system.
When the ratio, ks /kc , was varied from 0 to 0.25, Peak 10 s frequency was varied from 18.96 kHz
to 17.06 kHz (i.e., 10.02% variation) and Peak 20 s frequency was shifted from 31.42 kHz to 31.25 kHz
(i.e., 0.54% variation). The results suggested that the impedance peak at a lower frequency exhibited
a larger frequency shift than that at a higher frequency under the same bolt preload change.
Importantly, the results evidenced the theoretical feasibility of the PZT interface technique for the
bolt-loosening monitoring of a bolted connection.
Sensors 2018, 18, 2766 6 of 20
Sensors 2018, 18, x FOR PEER REVIEW 6 of 20
Sensors 2018, 18, x FOR PEER REVIEW 6 of 20

Peak
11 Peak
Peak Peak 22
18.96 kHz
18.96 kHz 31.42 kHz
900 31.42 kHz 2500
900 2500
Real Part
800 Real Part
800 2000
Imaginary Part 2000
700 Imaginary Part
700

Imaginary Part (Ohm)

Imaginary Part (Ohm)
Real Part (Ohm)

600 1500
Real Part (Ohm)

600 1500
500
500 1000
400 1000
400
300 500
300 500
200
200 0
100 0
100
0 -500
0 -500
10 15 20 25 30 35 40 45 50
10 15 20 25 30 35 40 45 50
Frequency
Frequency (kHz)
(kHz)
Figure 3. Impedance responses
ofof the impedance model with infinitive contact stiffness (ξ = /k ksc/k=c 0).
= 0).
Figure
Figure 3.
3. Impedance
Impedance responses
responses of the
the impedance
impedance model
model with
with infinitive
infinitive contact
contact stiffness
stiffness (ξ(ξ== kkss/kc = 0).

400
400
k /k =0
350 k s /ksc =c 0
350
k /k = 0.05
k s /ksc =c 0.05
300
Real Impedance (Ohm)

300
Real Impedance (Ohm)

k /k = 0.1
k s /ksc =c 0.1
250 k /k = 0.15
250 k s /ksc =c 0.15
200 k /k = 0.2
200 k s /ksc =c 0.2
150 k /k = 0.25
150 k s /ksc =c 0.25

100
100
50
50
0
0 10 15 20 25 30 35 40 45 50
10 15 20 25 30 35 40 45 50
Frequency (kHz)
Frequency (kHz)

(a)(a)
TheThe frequency
frequency range
range ofof 10–50
10–50 kHz
kHz
400 300
400 300

Contact 250 Contact

Contact 250 Contact
Real Impedance (Ohm)

Real Impedance (Ohm)

300 Stiffness-loss Stiffness-loss

Real Impedance (Ohm)

Real Impedance (Ohm)

300 Stiffness-loss Stiffness-loss

200
200

200 150
200 150

100
100
100
100
50
50

0 0
0 12 14 16 18 20 22 0 27 29 31 33 35 37
12 14 16 18 20 22 27 29 31 33 35 37
Frequency (kHz) Frequency (kHz)
Frequency (kHz) Frequency (kHz)

(b)(b) The
The frequency
frequency range
range ofof 12–22
12–22 kHz
kHz (c)(c)
TheThe frequency
frequency range
range ofof 27–37
27–37 kHz
kHz

Figure 4. Changes
Figure in the impedance responses of the impedance model under contact stiffness-loss.
Figure 4. 4. Changes
Changes inin
thethe impedance
impedance responses
responses ofof the
the impedance
impedance model
model under
under contact
contact stiffness-loss.
stiffness-loss.
Sensors 2018, 18, 2766 7 of 20

3. Predetermination of Sensitive Frequency Band for Impedance Response

3.1. Finite Element Model of PZT Interface-Bolted Connection

3.1.1. Finite Element Modeling

The EM impedance’s sensitive frequency band was predetermined for a bolted connection by
using the PZT interface technique. The finite element (FE) model of a bolted connection example was
established by using COMSOL multiphysics. As shown in Figure 5a, the connection example is a steel
bolted joint that was used to connect two H-beam segments. The splice plate (310 × 200 × 10 mm3 )
was clamped by the eight bolts (20 mm diameter) at each flange of the H-beam. To monitor the bolt
preload, the connection was equipped with a PZT interface at the middle of the splice plate. The effect
of the bolt preload was simulated by the equivalent contact spring (kx , ky , and kz ) and the damper
systems (cx , cy , and cz ), as shown in Figure 5b [17,30]. The main concern of the FE study was to
numerically examine the effect of the contact parameters on the impedance responses of the PZT
interface. So, the effect of the H-beam segments in Figure 5a was neglected for the simplification.
As detailed in Figure 5b, the PZT interface has two bonded sections (33 × 35 × 5 mm3 ) and
a flexible section (33 × 30 × 4 mm3 ) embedded with a PZT-5A patch (15 × 15 × 0.51 mm3 , Piezo
Systems Inc). The interface body is made of aluminium. The PZT patch was attached to the interface
by the bonding layer of 0.1 mm. The PZT interface was also mounted to the splice plate by the
0.1 mm bonding layer. The PZT patch was simulated using the piezoelectric elements that have both
mechanical and electrical properties. The FE model was discretized by three-dimensional (3D) solid
elements, as shown in Figure 5c. A complete mesh of the FE model consists of 4155 elements.
The structural properties of the splice plate, the interface, and the bonding layers are listed in
Table 1. It is noted that the similar structural parameters of the bonding layers were recommended
in the previous studies [32,33]. The piezoelectric properties of the PZT-5A are listed in Table 2 [26].
The thickness frequency of the PZT patch is about 4 MHz. For acquiring the EM impedance from the
PZT interface, the harmonic excitation voltage with 1 V amplitude (V = 1e jωt ) was applied to the top
surface of the PZT sensor, while the bottom surface was set as the ground.

3.1.2. Simulation of Bolt Preload Change

As explained previously, the bolt preload change can be represented by the variation of the
contact parameters. The contact stiffness was assumed to be uniform over the contact area of the splice
plate. For the intact state, the contact stiffness was set as kz = 4.0 × 1011 N/m/m2 and kx = ky = 0.5 kz .
The contact damping loss factor was assumed to be η x = η y = η z = 0.02. As given in Table 3, four
damage cases of the contact stiffness (D1–D4) were investigated. The contact stiffness was reduced by
12.5%, 25%, 37.5%, and 50% in the cases D1, D2, D3, and D4, respectively. It is noted that the contact
stiffness-loss of 12.5% could be interpreted as the equivalent damage severity of a completely loosened
bolt in the eight-bolt connection.
Sensors 2018,
Sensors 2018, 18,
18, 2766
x FOR PEER REVIEW 88 of
of 20
20

(a) A bolted connection example

(b) An equivalent FE model of bolted connection (unit: mm)

(c) Meshing

Figure 5. A finite element (FE) modeling of a bolted

bolted connection
connection example
example with
with aa PZT
PZT interface.
interface.
Sensors 2018, 18, 2766 9 of 20

Table 1. Material properties of the splice plate, the PZT interface, and the bonding layer.

Parameters PZT Interface Splice Plate Bonding Layer

Young’s modulus, E (GPa) 70 200 6
Poisson’s ratio, υ 0.33 0.3 0.38
Mass density, ρ (kg/m3 ) 2700 7850 1700
Damping loss factor, η 0.02 0.02 0.02

Table 2. Properties of the PZT-5A patch.

Parameters Value
 
16.4 −5.74 −7.22 0 0 0

 −5.74 16.4 −7.22 0 0 0  
Elastic compliance,  −7.22 −7.22 18.8 0 0 0   × 10−12
E (m2 /N)
sijkl


 0 0 0 47.5 0 0  
 0 0 0 0 47.5 0 
0 0 0 0 0 44.3
0 0 −171
 0
 0 −171 
Dielectric coupling constant,  0
 0 374  × 10−12
dkij (C/N)  0
 584 0  
 584 0 0 
 0 0 0
1730 0 0
Permittivity,
0  × 8.854 × 10−12


 0 1730
ε Tjk (Farad/m)
0 0 1700
Mass density,
7750
ρ (kg/m3 )
Damping loss factor,
0.0125
η
Dielectric loss factor,
0.015
δ

Table 3. Damage cases of the finite element (FE) model.

Value of Contact Stiffness (N/m2 /m)

Damage Case Description
kx = ky kz
Intact 0% contact stiffness-loss 2.0 × 1011 4.0 × 1011
12.5% contact
D1 1.75 × 1011 3.5 × 1011
stiffness-loss
D2 25% contact stiffness-loss 1.5 × 1011 3.0 × 1011
37.5% contact
D3 1.25 × 1011 2.5 × 1011
stiffness-loss
D4 50% contact stiffness-loss 1.0 × 1011 2.0 × 1011

3.2. Predetermination of Sensitive Frequency Band for Bolted Connection

Figure 6 shows the EM impedance of the PZT interface-bolted connection system, including
the real and imaginary parts in the frequency range of 10–50 kHz with the resolution of 0.05 kHz.
Within the examined range, there were both resonant and non-resonant regions of the impedance
signatures. Two resonant bands containing two significant peaks (i.e., Peak 1 at 18.05 kHz and Peak 2 at
34.05 kHz) were observed in the figure. In the resonant bands, the aspect of the real impedance values
becomes significant as that of the imaginary impedance values. Because the impedance signatures
of Peaks 1–2 would be sensitive to structural damage, it is necessary to predetermine the frequency
ranges containing these peaks.
Sensors 2018, 18, 2766 10 of 20
Sensors 2018, 18, x FOR PEER REVIEW 10 of 20
Sensors 2018, 18, x FOR PEER REVIEW 10 of 20

ToTo
To
identify the
identify
identify the the modal
modal responses
modal responses
corresponding to
responses corresponding
corresponding to to the
the two
the two
impedance peaks,
two impedance
impedance peaks, peaks, the
the Eigenvalue
the Eigenvalue
Eigenvalue
analysis
analysis of of
thethe
PZT PZT interface
interface waswas performed.
performed. The The interface
interface waswas fixed
fixed at at
thethe bottom
bottom surfaces
surfaces of of
two two
analysis of the PZT interface was performed. The interface was fixed at the bottom surfaces of two
bonded sections.
bonded sections.
sections. As As
Asshown shown in Figure
shownininFigure 7a,b,
Figure7a,b,
7a,b,thethe bending
thebending modes
bendingmodes
modes of the PZT interface corresponding to
bonded ofof
thethe PZT
PZT interface
interface corresponding
corresponding to
Peak
to Peak 1 and
1 and Peak
Peak 2 were
2 were found
found at the
at the frequencies
frequencies of 17.73
of 17.73kHz (i.e., longitudinal bending motion) and
Peak 1 and Peak 2 were found at the frequencies of 17.73 kHzkHz
(i.e.,(i.e., longitudinal
longitudinal bendingbending motion)
motion) and
and33.73
33.73kHzkHz (i.e.,
(i.e.,lateral
lateral bending
bending motion),
motion), respectively. The
respectively. The frequency
frequencydifferences
differencesbetween
betweenthethe
33.73 kHz (i.e., lateral bending motion), respectively. The frequency differences between the
impedance
impedance analysis and
analysis the modal analysistheof isolated
the isolated PZT interface were onlyfor5.4% for1Peak 1
impedance analysisand andthethemodal
modalanalysis
analysisofof the isolatedPZTPZTinterface
interface werewereonly 5.4%
only 5.4% Peak
for Peakand 1
andfor0.9%
0.9% for The
Peak 2. Thesuggested
results suggested that thefrequency
sensitive bands
frequency bands of thesignatures
impedance
and 0.9%Peak
for 2.Peak 2.results that thethat
The results suggested sensitive
the sensitive frequency of the
bandsimpedance
of the impedance
signatures
can be easily can be easily
predetermined predetermined
by the numerical by the numerical modal analysis of the isolated PZT interface.
signatures can be easily predetermined by themodal
numericalanalysis of the
modal isolated
analysis PZT
of the interface.
isolated PZTThe results
interface.
The
also results
revealed also
that atrevealed
least two that at least
significant two
peaks significant
(Peaks 1–2) peaks
can be (Peaks
expected 1–2)
in can
the be expected
frequency band inofthe
The results also revealed that at least two significant peaks (Peaks 1–2) can be expected in the
frequency
10–50 forband
kHz band the of 10–50 kHz for the impedance measurement via the PZT interface.
frequency ofimpedance
10–50 kHz measurement
for the impedance via the PZT interface.
measurement via the PZT interface.
Peak 1 Peak 2
Peak 1 kHz
18.05 Peak 2
34.05 kHz
18.05 kHz 34.05 kHz 3000
3000
Real Part
Real Part 2500
2000 2500
2000 Imaginary Part
Imaginary Part 2000

Imaginary Part (Ohm)

2000

Imaginary Part (Ohm)

Real Part (Ohm)

1500
Real Part (Ohm)

1500 1500
1500 1000
1000
500
1000 500
1000
0
0
500 -500
500 -500
-1000
-1000
0 -1500
0 -1500
10 15 20 25 30 35 40 45 50
10 15 20 25 30 35 40 45 50
Frequency (kHz)
Frequency (kHz)
Figure 6. Numerical impedance signatures of PZT interface-bolted connection system.
Figure
Figure 6.
6. Numerical
Numerical impedance
impedance signatures
signatures of
of PZT
PZT interface-bolted
interface-bolted connection
connection system.
system.

(a) Longitudinal bending motion at 17.73 kHz (b) Lateral bending motion at 33.73 kHz
(a) Longitudinal bending motion at 17.73 kHz (b) Lateral bending motion at 33.73 kHz
Figure Two
7. 7.
Figure bending
Two modes
bending of of
modes PZT interface
PZT corresponding
interface to Peak
corresponding 1 and
to Peak Peak
1 and 2. 2.
Peak
Figure 7. Two bending modes of PZT interface corresponding to Peak 1 and Peak 2.

3.3.
3.3. Evaluation
Evaluation of of Predetermined
Predetermined Frequency
Frequency Band
Band
3.3. Evaluation of Predetermined Frequency Band
The The sensitivity
sensitivity of of
thethe predetermined
predetermined frequency
frequency band
band to to
thethe bolt
bolt preload
preload change
change was
was numerically
numerically
The sensitivity of the predetermined frequency band to the bolt preload change was numerically
evaluated.
evaluated. TheThe impedance
impedance signatures
signatures in in 10–50
10–50 kHzkHz were
were numerically
numerically analyzed
analyzed forfor
thethe four
four damage
damage
evaluated. The impedance signatures in 10–50 kHz were numerically analyzed for the four damage
cases, as plotted in Figure 8a. Two resonant bands containing Peak 1 and Peak 2 were zoomed in in
cases, as plotted in Figure 8a. Two resonant bands containing Peak 1 and Peak 2 were zoomed
cases, as plotted in Figure 8a. Two resonant bands containing Peak 1 and Peak 2 were zoomed in
Figure
Figure 8b,c,8b,c, respectively.
respectively. TheseThese impedance
impedance peaks sensitively
peaks sensitively shifted along
shifted leftward leftward
withalong with the
the reduction
Figure 8b,c, respectively. These impedance peaks sensitively shifted leftward along with the
in reduction
the contactinstiffness.
the contactWhilestiffness. While
Peak 1 in 12–22Peak
kHz1(see
in 12–22
FigurekHz (see Figure 8b)
8b) experienced bothexperienced
the frequency both
andthe
reduction in the contact stiffness. While Peak 1 in 12–22 kHz (see Figure 8b) experienced both the
frequencyshifts,
magnitude and magnitude
Peak 2 in 27–37 shifts,
kHzPeak
(see2Figure
in 27–37
8c) kHz
showed (seeonlyFigure
the 8c) showedvariation.
frequency only the frequency
frequency and magnitude shifts, Peak 2 in 27–37 kHz (see Figure 8c) showed only the frequency
variation.
It is shown that Peak 1 was more sensitive to the bolt looseness than Peak 2. As listed in Table 4,
variation.
when the It iscontact
shown that Peakwas
stiffness 1 was more sensitive
reduced by 50%,toPeak 10 s looseness
the bolt frequencythan Peak1.25
shifted 2. As
kHzlisted in 6.93%
(i.e., Table 4,
It is shown that Peak 0
1 was more sensitive to the bolt looseness than Peak 2. As listed in Table 4,
when the
variation), contact
while Peakstiffness was reduced
2 s frequency by 50%,
shifted only 0.2 kHzPeak
(i.e.,1′s frequency
0.59% shifted
variation). These1.25 kHz were
results (i.e., well
6.93%
when the contact stiffness was reduced by 50%, Peak 1′s frequency shifted 1.25 kHz (i.e., 6.93%
variation), while Peak 2′s frequency shifted only 0.2 kHz (i.e., 0.59% variation). These results were
variation), while Peak 2′s frequency shifted only 0.2 kHz (i.e., 0.59% variation). These results were
well consistent with the previous observations from the two-dof impedance model, and also
well consistent with the previous observations from the two-dof impedance model, and also
Sensors 2018, 18, 2766 11 of 20
Sensors 2018, 18, x FOR PEER REVIEW 11 of 20

consistent with thethe

demonstrated previous observations
sensitivity of the PZTfrom the two-dof
interface’s impedancefrequency
predetermined model, and alsotodemonstrated
range the bolt preload
the change.
sensitivity of the PZT interface’s predetermined frequency range to the bolt preload change.

1200
Intact
1000 D1
D2
Real Impedance (Ohm)

D3
800
D4

600

400

200

0
10 15 20 25 30 35 40 45 50
Frequency (kHz)

(a) The frequency range of 10–50 kHz.

700
1000
600
Contact Contact
Real Impedance (Ohm)

Real Impedance (Ohm)

Stiffness-loss 800 Stiffness-loss

500

400 600

300
400
200
200
100

0 0
12 14 16 18 20 22 27 29 31 33 35 37
Frequency (kHz) Frequency (kHz)

(b) The frequency range of 12–22 kHz (c) The frequency range of 27–37 kHz

Figure
Figure 8. Numerical
8. Numerical impedance
impedance signatures
signatures of model
of FE FE model under
under contact
contact stiffness-loss.
stiffness-loss.

Table
Table 4. Change
4. Change in peak
in the the peak frequencies
frequencies duedue to contact
to contact stiffness-loss.
stiffness-loss.

Peak Frequency (kHz)

Damage Case Peak Frequency (kHz)
Damage Case f1 Δf1 (%) f2 Δf2 (%)
f1 ∆f 1 (%) f2 ∆f 2 (%)
Intact 18.05 0 34.05 0
Intact 18.05 0
D1 17.80 −1.3934.05 34.000
−0.15
D1 17.80 −1.39 34.00 −0.15
D2D2 17.50 −3.05 −3.0533.95
17.50 33.95
−0.29 −0.29
D3D3 17.20 −4.71 −4.7133.90
17.20 33.90
−0.44 −0.44
D4D4 16.80 −6.93 −6.9333.85
16.80 −0.59
33.85 −0.59

4. Experimental
4. Experimental Evaluation
Evaluation on Lab-Scaled
on Lab-Scaled Bolted
Bolted Girder
Girder Connection
Connection

4.1.4.1.
Experimental Setup
Experimental Setup

4.1.1. Test-Setup
4.1.1. of Bolted
Test-Setup Girder
of Bolted Connection
Girder Connection
An An
experimental evaluation
experimental waswas
evaluation conducted on aonlab-scaled
conducted bolted
a lab-scaled girder
bolted connection.
girder Figure
connection. 9 9
Figure
shows the schematic of a three-span steel girder with a bolted connection at the middle. The girder,
shows the schematic of a three-span steel girder with a bolted connection at the middle. The girder,
with a total
with length
a total ofof4.14
length 4.14m,
m,was
was simply supported by
simply supported bysteel
steelbars
barsatat four
four locations,
locations, as shown
as shown in
in Figure
9a. The girder was assembled from two single H-shaped beams (H-200 × 180 × 8 × 10 mm) by splice
plates and bolts at two flanges, see Figure 9b. The connection splice plate (310 × 200 × 10 mm) clamped
Sensors 2018, 18, 2766 12 of 20

Figure 9a. The girder was assembled from two single H-shaped beams (H-200 × 180 × 8 × 1012mm)
Sensors 2018, 18, x FOR PEER REVIEW of 20
by
splice plates and bolts at two flanges, see Figure 9b. The connection splice plate (310 × 200 × 10 mm)
clamped
by eightbybolts
eight(20bolts (20 mm-diameter)
mm-diameter) is schematized
is schematized in Figure
in Figure 9c. The PZT 9c.interface
The PZTprototype
interfacesketched
prototype
sketched
in Figurein Figure
5b was 5b was fabricated
fabricated and surface-mounted
and surface-mounted to the of
to the middle middle of theplate.
the splice splice plate.
The Thebody
whole whole
of the
body interface,
of the interface,including thethe
including flexible andand
flexible side sections,
side was
sections, fabricated
was fabricatedfrom
froman an
aluminium
aluminium plate
plate
using
using a precisioncutting
a precision cuttingmachine.
machine. Loctite
Loctite 401
401 instant
instant adhesive
adhesivewaswasused
usedtotoattach
attachthethe
PZT
PZTto to
thethe
middle
middle section
section andandthe
thebonded
bondedsections
sectionstoto the
the host
host structure.
structure.

1 2
Bolted Splice Plate
Simple Simple Simple Simple
Connection 310 x 200 support
support support support

Bolts
φ 20
H-200ⅹ180ⅹ8ⅹ10

100 1300 1340 1300 100

1 2

(a) Steel girder

1-1 2-2 50 70 70 70 50

Splice Plate
10 50

10 PZT
100 Interface 200
180 8

50
Bolt 1 Bolt 2 Bolt 3 Bolt 4
200 50 100 50 310

(b) Cross-section (c) Bolted connection.

Figure9.9.Schematic
Figure Schematic of
of the
the bolted girder
girder connection
connection(unit:
(unit:mm).
mm).

Thereal
The realsetups
setups of
of the
the steel
steel girder
girderand
andthe
the bolted connection
bolted are illustrated
connection in Figure
are illustrated 10. As 10.
in Figure
designed, all of the bolts of the connection were fastened to the torque of 160 Nm.
As designed, all of the bolts of the connection were fastened to the torque of 160 Nm. A torqueA torque wrench
(TOHNICHI QL280N) was used to fasten the bolts and to control the bolt torque. Four of the eight
wrench (TOHNICHI QL280N) was used to fasten the bolts and to control the bolt torque. Four of the
bolts in the connection (Bolts 1–4) were selected to simulate the loosening events, as indicated in
eight bolts in the connection (Bolts 1–4) were selected to simulate the loosening events, as indicated in
Figure 10b. Among the four bolts, Bolts 1 and 3 are close to the PZT interface, while Bolts 1 and 4 are
Figure 10b. Among the four bolts, Bolts 1 and 3 are close to the PZT interface, while Bolts 1 and 4 are
more distant. Table 5 describes the loosening cases of Bolt 1–4. Each of the four bolts was loosened
more distant. Table 5 describes the loosening cases of Bolt 1–4. Each of the four bolts was loosened
from the initial torque of 160 Nm to the torque of 110 Nm (i.e., a 31% torque-loss), 60 Nm (i.e., a 62%
from the initialand
torque-loss), torque of (i.e.,
0 Nm 160 Nm to the
a 100% torque of The
torque-loss). 110 girder
Nm (i.e.,
wasa placed
31% torque-loss), 60 Nmwhere
in the laboratory, (i.e., athe
62%
torque-loss),
temperatureand was0 controlled
Nm (i.e., anear
100%22torque-loss). The girder
°C by air-conditioners so was
as toplaced in the laboratory,
avoid temperature effects.where the
temperature was controlled near 22 ◦ C by air-conditioners so as to avoid temperature effects.
Table 5. Preload change cases of bolted girder connection.
Table 5. Preload change cases of bolted girder connection.
Loosened Bolt Variation of Torque Level (Nm)
Loosened Bolt → Torque
1: 160 of Level→(Nm)
110 (−31%) 60 (−62%) → 0 (−100%);
Bolt 1 Bolt Variation
all others: 160
Bolt 1: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%);
Bolt 1 →others:
Bolt 2
Bolt 2: 160all 160 → 60 (−62%) → 0 (−100%);
110 (−31%)
Bolt 2: 160 → 110 (−31%) → 60 (others:
all −62%) → 1600 (−100%);
Bolt 2
Bolt 3: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%);
all others: 160
Bolt 3 Bolt 3: 160 → 110 (−31%) → 60 −62%) →
Bolt 3 all(others: 1600 (−100%);
all others: 160
Bolt 4: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%);
Bolt 4 Bolt 4: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%);
Bolt 4 all others: 160
all others: 160
Sensors
Sensors 2018,
2018, 18,
18, x FOR PEER REVIEW
2766 13
13 of
of 20
20

(a) Test structure (b) Bolted connection.

Figure 10.
Figure 10. Experimental setup of bolted girder connection.

4.1.2. Impedance
Impedance Measurement System
A low-cost
low-costand andmulti-channel
multi-channel SSeL-I
SSeL-I impedance
impedance measurement
measurement systemsystem developed
developed by the
by the research
research
group at group
Pukyong at Pukyong
National National
University University
[22], was [22],
used was used to wirelessly
to wirelessly acquire theacquire
impedancethe impedance
data from
data frominterface.
the PZT the PZT interface.
Figure 11aFigure
shows11a shows a prototype
a prototype of the SSeL-I
of the wireless wirelessnode
SSeL-I
thatnode that consists
consists of three
of threean
layers, layers,
SSeL-I animpedance
SSeL-I impedance
board, anboard,
Imote2an Imote2
platform, platform, and a battery
and a battery board. board. The schematic
The schematic of the
of the SSeL-I
SSeL-I sensorsensor
node is node is shown
shown in Figure
in Figure 11b. key
11b. The Thecomponent
key component of theofSSeL-I
the SSeL-I
board board is low-cost
is the the low-
cost impedance
impedance chipchip AD9533,
AD9533, whichwhich
has ahas a capability
capability to measure
to measure the impedance
the impedance up100
up to to 100
kHzkHzwithwiththe
the resolution
resolution lessless
thanthan
0.10.1
Hz.Hz.TheThe SSeL-I
SSeL-I boardintegrates
board integratesa amultiplexer
multiplexerfor formeasuring
measuring up to 16 PZT
up to
patches and the SHT11 sensor for recording recording temperature
temperature and and humidity.
humidity.
The Imote2
Imote2platform
platformisisused
used toto
control
controlimpedance
impedance measurements
measurements via the
via impedance
the impedance board. The
board.
Imote2 has has
The Imote2 a high-speed
a high-speedPXA27x
PXA27x processor
processor (clock
(clockspeed
speed ofof
13-416
13-416MHz),
MHz),SRAMSRAMofof256 256kB,
kB,the
the flash
flash
memory of of 32
32 MB,
MB, and
andthe
theSDRAM
SDRAMofof32 32MBMB[34–36].
[34–36].This Thisplatform
platformisisdesigned
designed with
with a wireless
a wireless radio
radio of
of
2.42.4
GHz GHz Zigbee
Zigbee for data
for data transmission
transmission (up to(up to a distance
a distance of 125 of 125an
m by mexternal
by an external
antenna).antenna).
The wirelessThe
wireless
sensor unitsensor unit is powered
is powered via the battery
via the battery board (3.2 boardV).(3.2 V). Although
Although the wireless
the wireless impedance
impedance sensorsensor
node
node
has a has a limited
limited measurable
measurable frequency
frequency rangerange (i.e., than
(i.e., less less than 100 kHz),
100 kHz), it costs
it costs onlyonly 300 USD
300 USD and andhas
has multi-channels
multi-channels that that
couldcould
enableenable the cost-effectiveness
the cost-effectiveness for a for a health
health monitoring
monitoring system system of inmega
of in situ situ
mega
boltedbolted structures.
structures.

(a) A prototype of SSeL-I (b) Schematic of SSeL-I.

Figure 11.
Figure Wireless SSeL-I impedance sensor node.
11. Wireless
Sensors 2018, 18, 2766 14 of 20
Sensors 2018, 18, x FOR PEER REVIEW 14 of 20

4.2.Preload
4.2. PreloadChange
ChangeMonitoring
MonitoringininBolted
BoltedGirder
GirderConnection
Connection

4.2.1.
4.2.1.Impedance
ImpedanceMeasurement
Measurementvia
viaPZT
PZTInterface
Interface
The
Theimpedance
impedancesignatures
signatureswere
weremeasured
measuredininthethefrequency
frequencyrange
rangeofof10–50
10–50kHz
kHztotoidentify
identifythe
the
sensitive
sensitiveimpedance
impedancepeaks
peaks(Peaks
(Peaks 1–2),
1–2), as
as numerically
numerically pre-analyzed in Section
Section 3.
3. The
Theamplitude
amplitudeof
ofthe
theexcitation
excitation voltage
voltage waswasset set at and
at 1V, 1V, and the resolution
the resolution of theofPZT
thescanning
PZT scanning frequency
frequency was 0.1 was
kHz.
0.1 kHz.
Four Four measurements
repeated repeated measurements were for
were conducted conducted for bolt-loosening
each of the each of the bolt-loosening
cases in Table cases
5. For in
the
Table 5. For theevaluation,
performance performance theevaluation,
impedancethe impedance
signatures signatures
measured by measured by the
the wireless wireless
SSeL-I systemSSeL-I
were
system
comparedwerewith
compared with those
those using using
a wired a wired high-performance
high-performance impedanceimpedance analyzer HIOKI-3532.
analyzer HIOKI-3532. As shown
Asin shown
Figure in Figure
12a,b, the 12a,b, theimaginary
real and real and imaginary
impedanceimpedance signaturesby
signatures measured measured
the SSeL-Ibysystem
the SSeL-I
were
system were well-matched with those by the wired HIOKI system for the same frequency
well-matched with those by the wired HIOKI system for the same frequency range with identical range with
identical
patterns.patterns.

700 5000
Wired HIOKI
Wired HIOKI
600 Wireless SSeL-I
Wireless SSeL-I
4000
500 Imaginary Part (Ohm)
Real Part (Ohm)

3000
400

300
2000

200
1000
100

0 0
10 20 30 40 50 10 20 30 40 50
Frequency (kHz) Frequency (kHz)

(a) Real impedance signatures (b) Imaginary impedance signatures

Figure 12. Impedance signatures in 10–50 kHz: wired versus wireless measurements.
Figure 12. Impedance signatures in 10–50 kHz: wired versus wireless measurements.

The
Theimpedance
impedanceresponses
responsesunder
under the
the bolt-loosening
bolt-loosening cases cases of of Bolt
Bolt33were
wereplotted
plottedinin Figure
Figure 13.13.
As
As expectedfrom
expected from the
thenumerical
numerical analysis, twotwo
analysis, resonant bandsbands
resonant (i.e., 12–22
(i.e.,kHz andkHz
12–22 27–37andHz) exist
27–37containing
Hz) exist
significant
containing impedance
significantpeaks between 10–50
impedance peaks kHz, as zoomed
between 10–50inkHz,Figure as13b,c.
zoomedThe comparison between
in Figure 13b,c. The
Figures 8 and 13 revealed certain gaps between the experimental measurement
comparison between Figure 13 and Figure 8 revealed certain gaps between the experimental and the numerical analysis.
These gaps couldand
measurement be caused by the differences
the numerical in the
analysis. structural
These parameters
gaps could between
be caused bythetheexperimental
differencesmodel
in the
and the FE model. For Peak 1, the numerical pre-analysis in Figure
structural parameters between the experimental model and the FE model. For Peak 8b predicted the peak frequencies
1, the numerical
around 18 kHz,inwhile
pre-analysis Figurethe8bexperimental
predicted themeasurement in Figure
peak frequencies 13b showed
around 18 kHz,thewhile
peak frequencies near
the experimental
16measurement
kHz (i.e., the prediction error of 11.1%). For Peak 2, the pre-analysis in Figure 8c
in Figure 13b showed the peak frequencies near 16 kHz (i.e., the prediction error of estimated the peak
frequency
11.1%). Forat about
Peak 2,34the
kHz, while the in
pre-analysis experiment
Figure 8c in Figure 13c
estimated themeasured the peak
peak frequency frequency
at about at about
34 kHz, while
30the
kHz (i.e., the prediction
experiment in Figureerror of 11.8%). the peak frequency at about 30 kHz (i.e., the prediction error
13c measured
As observed in Figure 13, the impedance peaks tended to shift left as the torque was reduced.
of 11.8%).
As compared
As observedwith the first resonant
in Figure band (i.e., 12–22
13, the impedance peakskHz),
tended thetosecond oneas(i.e.,
shift left the 27–37
torquekHz)
was showed
reduced.
less
Assensitivity
comparedto the the
with torque-loss severity.
first resonant The(i.e.,
band changing
12–22 trend
kHz),ofthethesecond
two resonant bands
one (i.e., 27–37was
kHz)consistent
showed
with
less sensitivity to the torque-loss severity. The changing trend of the two resonant bandsand
the previous numerical results. The first resonant band experienced both the frequency was
magnitude
consistent shifts,
with while the second
the previous one showed
numerical slight
results. The changes in the peak
first resonant frequency
band and almost
experienced both no the
noticeable changes in the magnitude.
frequency and magnitude shifts, while the second one showed slight changes in the peak frequency
and almost no noticeable changes in the magnitude.
Sensors
Sensors 2018,
2018, 18,
18, x2766
FOR PEER REVIEW 15
15 of
of 20
20

600
Intact
500 31% Torque Loss
62% Torque Loss
Real Impedance (Ohm)

100% Torque Loss

400

300

200

100

0
10 15 20 25 30 35 40 45 50
Frequency (kHz)

(a) The frequency range of 10–50 kHz.

400 600

500
Real Impedance (Ohm)
Real Impedance (Ohm)

300
400

200 300

200
100
100

0 0
12 14 16 18 20 22 27 29 31 33 35 37
Frequency (kHz) Frequency (kHz)

(b) The frequency range of 10–20 kHz (c) The frequency range of 24–34 kHz.

Figure 13.
Figure Experimental impedance
13. Experimental impedance signatures
signatures under
under bolt-loosening
bolt-loosening cases
cases of
of Bolt
Bolt 3.
3.

4.2.2.
4.2.2. Detection
Detection of
of Preload
Preload Change
Change using
using Impedance
Impedance Response
Response

Statistical Quantification
Statistical Quantification Method
To detect
To detectthe
thepreload
preload change
change in bolted
in the the bolted connection,
connection, two common
two common damage-sensitive
damage-sensitive features
features
were were extracted
extracted from the impedance
from the impedance data, the correlation
data, the correlation coefficient(CCD)
coefficient deviation deviation (CCD) and
and root-mean-
root-mean-square
square deviation deviation
(RMSD) (RMSD)
indices. indices. These
These two two impedance
impedance features
features quantify
quantify thethe changesinin the
changes the
impedance signatures with different manners. While the CCD index mainly quantifies
impedance signatures with different manners. While the CCD index mainly quantifies the frequency the frequency
shift of the
shift the impedance
impedancesignatures,
signatures,the RMSD
the RMSD index quantifies
index bothboth
quantifies the frequency and magnitude
the frequency shifts.
and magnitude
shifts.According to [23], the CCD index can be computed using the below formula:
According to [23], the CCD index ncan be computed using the below formula: o
1 ∗
( Z ∗ (ωi )) − Re( Z )]

CCD = 1 − E Re ( Z ( ω )) − Re ( Z ) [ Re (7)
1σZ σZ∗
{ }
i
E Re(Z (ωi )) − Re(Z )  Re(Z (ωi )) − Re(Z ) 
CCD = 1 − * *
(7)
σ σ *
where E[·] is the expectation operation; Z (ω ) and Z ∗ (ω ) signify the impedance responses at the ith
Z Z i i
∗
frequency
where E[·] before and after a damage
is the expectation event,
operation; ) and ∗ ( Z) and
( respectively; Z indicate
signify the mean
the impedance values ofat
responses those
the
impedance responses; and σ and σ ∗ are the corresponding standard deviations.
*
ith frequency before and after a damage event, respectively; Z and Z indicate the mean values of
Z Z
those impedance responses; and  and ∗ are the corresponding standard deviations.
As another damage-sensitive feature, the RMSD index can be obtained by [23] the following:
Sensors 2018, 18, 2766 16 of 20

As another damage-sensitive feature, the RMSD index can be obtained by [23] the following:
Sensors 2018, 18, x FOR PEER REVIEW 16 of 20
v ,
u N N
N ∑ [ Z ( ωi ) − Z ( ωi )] N ∑ [Z(2 ωi )]2
2
u
RMSD = t ∗ (8)
RMSD =  1Z (ω i ) − Z (ω i )   [ Zi =(ω1i ) ]
2
i =
* (8)
i =1 i =1

N denotes
wherewhere the number
N denotes the numberof swept frequencies.
of swept frequencies.
For distinguishing the bolt-loosening
For distinguishing the bolt-loosening state from
state fromthe
thehealthy state,an
healthy state, analarming
alarming threshold
threshold known
known
as theas
upper control limit (UCL) can be established using the values of the impedance features
the upper control limit (UCL) can be established using the values of the impedance features under under
the intact state state
the intact [25,37], as follows:
[25,37], as follows:
UCL = µ + 3σ (9)
UCL = μ + 3σ (9)
where µ and σ are the mean and the standard deviation of the impedance feature values, respectively.
where μ and σ are the mean and the standard deviation of the impedance feature values, respectively.
The UCL determined by three standard deviations of the mean has the confidence level of 99.7%.
The UCL determined by three standard deviations of the mean has the confidence level of 99.7%.

Preload Change Detection Results

Preload Change Detection Results
The RMSD
The RMSD control chart
control waswas
chart constructed
constructedusing
using thethe impedance
impedance data dataininthe
thepredetermined
predetermined
frequency bandband
frequency of 10–50
of 10–50 kHz.kHz.Figure
Figure 14 showsthe
14 shows theRMSDRMSD index
index thatthat
was was plotted
plotted according
according to the to
the torque-loss
torque-loss level forallallof of
level for thethe loosening
loosening casescases
of Boltsof1–4.
Bolts 1–4. As from
As observed observed
Figurefrom
14a–d,Figure
the RMSD14a–d,
values
the RMSD were very
values weresmall
veryfor the intact
small for thecase, but case,
intact became butnoticeable as Bolts 1–4as
became noticeable experienced
Bolts 1–4 aexperienced
31%, 62%, a
and 100%
31%, 62%, torque-loss.
and 100% The UCL
torque-loss. Thethresholds were computed
UCL thresholds to classify the
were computed bolt-loosening
to classify events. As
the bolt-loosening
events. As plotted in Figure 14a–d, for all of the torque-loss events, the RMSD values weredefined
plotted in Figure 14a–d, for all of the torque-loss events, the RMSD values were above the above the
thresholds, indicating the successful detection of the preload
defined thresholds, indicating the successful detection of the preload changes. changes.
It should be noted that the 31% torque-loss of a single bolt is equivalently corresponding to the
It should be noted that the 31% torque-loss of a single bolt is equivalently corresponding to
3.8% preload reduction in the test connection, which consists of eight bolts. This means that the
the 3.8% preload reduction in the test connection, which consists of eight bolts. This means that the
impedance signatures of the PZT interface were quite sensitive to the small preload changes occurred
impedance
in the signatures of the PZT
bolted connection. interfacebolts
The loosened werenear
quite
thesensitive to the(i.e.,
PZT interface small preload
Bolts 2 and 3)changes occurred
were detected
in the with
bolted connection. The loosened bolts near the PZT interface (i.e., Bolts 2 and
higher severity estimations than those far from the interface (i.e., Bolts 1 and 4). These results 3) were detected
with higher severity estimations than those far from the interface (i.e., Bolts
confirmed that the single PZT interface was able to monitor multiple loosened bolts on the splice 1 and 4). These results
confirmed
plate.that the single PZT interface was able to monitor multiple loosened bolts on the splice plate.
For theForcomparison,
the comparison, the the
CCD CCD control
control chartwas
chart wasalso
alsoconstructed
constructed by byusing
usingthe
thesame
same impedance
impedance
data, as shown in Figure 15a–d. As compared with the RMSD,
data, as shown in Figure 15a–d. As compared with the RMSD, the values of the CCD index the values of the CCD index were
were
relatively smaller. It is noted from Figure 13 that the impedance signatures
relatively smaller. It is noted from Figure 13 that the impedance signatures showed both frequency and showed both frequency
and magnitude
magnitude changes changes
under the under the bolt-loosening
bolt-loosening events.events.
Thus, Thus,
the the RMSD
RMSD approachconsidering
approach considering bothboth of
of the magnitude and frequency shifts is expected to result in higher severity estimations than the
the magnitude and frequency shifts is expected to result in higher severity estimations than the CCD
CCD approach quantifying only the frequency shift. The thresholds of the CCD index were computed
approach quantifying only the frequency shift. The thresholds of the CCD index were computed and
and are also sketched in Figure 15a–d. Although the values of the CCD index were quite small, those
are also
forsketched
the damage in Figure
cases 15a–d.
were above Although the thresholds,
the UCL values of the CCD index
indicating the were quitebolt-loosening
successful small, those for
the damage cases were above the UCL thresholds, indicating the successful bolt-loosening detection.
detection.

60 60

50 50

40 40
RMSD(%)

RMSD(%)

30 30

UCL = 5.42 % 23.11

UCL = 3.86 %
20 20
15.87
12.75 11.79 12.33
10 9.45 10

1.77 1.23
0 0
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 100% Loss
Torque Loss in Bolt 1 Torque Loss in Bolt 2

(a) Bolt 1-loosening cases (b) Bolt 2-loosening cases.

Figure 14. Cont.

 Sensors 2018, 18, x FOR PEER REVIEW 17 of 20
Sensors 2018, 18, 2766 17 of 20
60 18, x FOR PEER REVIEW
Sensors 2018, 60 17 of 20
50.93
50
60 50
60

50.93
RMSD(%) 40
50 40
50

RMSD(%)
30
40 UCL = 6.66 % 30
40
24.42
RMSD(%)

RMSD(%)
21.54 UCL = 4.17 %
20 UCL = 6.66 % 20 19.41
30 30
24.42 14.29
21.54 UCL = 4.17 %
10
20 10
20 19.41
7.43
14.29
2.18 1.33
100 100 7.43
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 100% Loss
2.18 Torque Loss in Bolt 3 1.33 Torque Loss in Bolt 4
0 0
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 100% Loss
(c) BoltTorque
3-loosening
Loss in Boltcases
3 (d) Bolt 4-loosening
Torque Loss in Bolt cases.
4

Figure 14. (c) Bolt 3-loosening

Preload cases
change monitoring of bolted connection by(d)root-mean-square
Bolt 4-loosening cases.
deviation (RMSD)
index.
Figure 14.Preload
Figure14. Preloadchange monitoring
change of bolted
monitoring connection
of bolted by root-mean-square
connection deviation
by root-mean-square (RMSD)(RMSD)
deviation index.
index.
30 30

25
30 25
30

20
25 20
25
CCD(%)CCD(%)

CCD(%)CCD(%)

15
20 15
20

10
15 UCL = 0.36 % 10
15 UCL = 0.31 %

105 UCL = 0.36 % 105 UCL = 0.31 % 4.17

2.40
1.00 1.62 1.41 1.54
0.29 0.27
50 50 4.17
Intact 31% Loss 62% Loss 100%
2.40Loss Intact 31% Loss 62% Loss 100% Loss
0.29 1.00 Loss in1.62
Torque Bolt 1 0.27
1.41 Loss in1.54
Torque Bolt 2
0 0
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 100% Loss
(a) BoltTorque
1-loosening cases
Loss in Bolt 1 (b) BoltTorque
2-loosening
Loss in Boltcases
2
30 30
(a) Bolt 1-loosening cases (b) Bolt 2-loosening cases
25
30 25
30
21.03
20
25 20
25

21.03
CCD(%)CCD(%)

CCD(%)CCD(%)

15
20 15
20

10
15 UCL = 0.41 % 10
15 UCL = 0.32 %
5.47
105 4.50 105
UCL = 0.41 % UCL = 0.32 % 3.43
1.88
0.30 5.47 0.27 0.57
50 4.50 50
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 3.43Loss
100%
1.88
0.30 Torque Loss in Bolt 3 0.27 Torque
0.57 Loss in Bolt 4
0 0
Intact 31% Loss 62% Loss 100% Loss Intact 31% Loss 62% Loss 100% Loss
(c) BoltTorque
3-loosening cases
Loss in Bolt 3 (d) BoltTorque
4-loosening
Loss in Boltcases
4

Figure (c)
15.15.
Figure Boltchange
Preload
Preload3-loosening
change casesof bolted
monitoring
monitoring connection
of bolted connection (d)
by Boltcoefficient
by correlation 4-loosening
correlation cases
deviation
coefficient (CCD) index.
deviation (CCD)
index.
Figure 15. Preload change monitoring of bolted connection by correlation coefficient deviation (CCD)
index.
5. Conclusions
In this study, the piezoelectric-based smart interface technique was developed to acquire
5. Conclusions
sensitive impedance signatures from a bolted connection for bolt-loosening detection. To
In this study,
demonstrate the piezoelectric-based
the theoretical feasibility of thesmart interface
proposed technique
method, was developed
a simplified to acquire
EM impedance model
sensitive impedance signatures from a bolted connection for bolt-loosening detection. To
demonstrate the theoretical feasibility of the proposed method, a simplified EM impedance model
Sensors 2018, 18, 2766 18 of 20

5. Conclusions
In this study, the piezoelectric-based smart interface technique was developed to acquire
sensitive impedance signatures from a bolted connection for bolt-loosening detection. To demonstrate
the theoretical feasibility of the proposed method, a simplified EM impedance model was newly
designed with the consideration of the bolt preload effect. Secondly, the EM impedance’s sensitive
frequency band was numerically pre-analyzed for a bolted connection via the PZT interface technique.
Finally, the impedance measurements were conducted on a lab-scaled bolted girder connection to verify
the pre-analyzed sensitive frequency range and to assess the preload change in the test connection.
From the numerical and experimental observations, the following concluding remarks can be made:
(1) The PZT interface’s sensitive frequency band, predetermined by the numerical simulation,
was quite consistent with that measured from the experiment.
(2) The impedance signatures obtained from the PZT interface were sensitive to the minor preload
change in the bolted connection. For the tested eight-bolt connection, a 31% torque-loss of a
single bolt can be detected using the PZT interface technique.
(3) A single PZT interface was able to monitor multiple loosened bolts in a connection, thus reducing
the number of sensing channels for the impedance monitoring of a large bolted connection.
Future works will need to optimize the geometric size of the PZT interface so as to enhance the
sensitivity of the impedance signatures and to quantitatively estimate the sensing area of the PZT
interface technique. Also, there is a need to evaluate the presented method for the simultaneous
loosening of multiple bolts.

Author Contributions: T.-C.H. conceived and designed the methodology. T.-C.H. performed all the experiments
with assistance from N.-L.D., J.-T.K supervised the whole work. T.-C.H prepared the manuscript. J.-T.K. revised
the manuscript. All authors read and approved the final manuscript.
Funding: This research was supported by a grant (18CTAP-C142999-01) from the Technology Advancement
Research Program (TARP), funded by the Ministry of Land, Infrastructure, and Transport of the Korean
government. T.C. Huynh and N.L. Dang were partially supported by the Brain Korea 21 Plus program of
the Korean government.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Chaki, S.; Corneloup, G.; Lillamand, I.; Walaszek, H. Combination of Longitudinal and Transverse Ultrasonic
Waves for In Situ Control. of the Tightening of Bolts. J. Press. Vessel Technol. 2006, 129, 383–390. [CrossRef]
2. Kim, N.; Hong, M. Measurement of axial stress using mode-converted ultrasound. NDT E Int. 2009, 42,
164–169. [CrossRef]
3. Joshi, S.G.; Pathare, R.G. Ultrasonic instrument for measuring bolt stress. Ultrasonics 1984, 22, 261–269.
[CrossRef]
4. Wang, T.; Song, G.; Wang, Z.; Li, Y. Proof-of-concept study of monitoring bolt connection status using a
piezoelectric based active sensing method. Smart Mater. Struct. 2013, 22, 087001. [CrossRef]
5. Doyle, D.; Zagrai, A.; Arritt, B.; Çakan, H. Damage Detection in Bolted Space Structures. J. Intell. Mater.
Syst. Struct. 2010, 21, 251–264. [CrossRef]
6. Kong, Q.; Zhu, J.; Ho, S.C.M.; Song, G. Tapping and listening: A new approach to bolt looseness monitoring.
Smart Mater. Struct. 2018, 27, 07LT02. [CrossRef]
7. Park, G.; Sohn, H.; Farrar, C.R.; Inman, D.J. Overview of piezoelectric impedance-based health monitoring
and path forward. Shock Vib. Dig. 2003, 35, 451–464. [CrossRef]
8. 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]
9. Chaudhry, Z.A.; Joseph, T.; Sun, F.P.; Rogers, C.A. Local-area health monitoring of aircraft via piezoelectric
actuator/sensor patches. Proc. SPIE 1995, 2443. [CrossRef]
10. Huynh, T.-C.; Kim, J.-T. Quantification of temperature effect on impedance monitoring via PZT interface for
prestressed tendon anchorage. Smart Mater. Struct. 2017, 26, 125004. [CrossRef]
Sensors 2018, 18, 2766 19 of 20

11. Lim, Y.Y.; Soh, C.K. Towards more accurate numerical modeling of impedance based high frequency
harmonic vibration. Smart Mater. Struct. 2014, 23, 035017. [CrossRef]
12. Huynh, T.C.; Dang, N.L.; Kim, J.T. Advances and challenges in impedance-based structural health monitoring.
Struct. Monit. Maint. 2017, 4, 301–329.
13. Huynh, T.C.; Kim, J.T. RBFN-based temperature compensation method for impedance monitoring in
prestressed tendon anchorage. Struct. Control Health Monit. 2018, 25, e2173. [CrossRef]
14. Wang, T.; Song, G.; Liu, S.; Li, Y.; Xiao, H. Review of Bolted Connection Monitoring. Int. J. Distrib. Sens. Netw.
2013, 9, 871213. [CrossRef]
15. Park, G.; Cudney, H.H.; Inman, D.J. Feasibility of using impedance-based damage assessment for pipeline
structures. Earthq. Eng. Struct. Dyn. 2001, 30, 1463–1474. [CrossRef]
16. Hong, D.-S.; Nguyen, K.-D.; Lee, I.-C.; Kim, J.-T. Temperature-Compensated Damage Monitoring by Using
Wireless Acceleration-Impedance Sensor Nodes in Steel Girder Connection. Int. J. Distrib. Sens. Netw. 2012,
8, 167120. [CrossRef]
17. Ritdumrongkul, S.; Abe, M.; Fujino, Y.; Miyashita, T. Quantitative health monitoring of bolted joints using a
piezoceramic actuator–sensor. Smart Mater. Struct. 2004, 13, 20. [CrossRef]
18. Min, J.; Park, S.; Yun, C.-B. Impedance-based structural health monitoring using neural networks for
autonomous frequency range selection. Smart Mater. Struct. 2010, 19, 125011. [CrossRef]
19. Nguyen, T.-C.; Huynh, T.-C.; Yi, J.-H.; Kim, J.-T. Hybrid. bolt-loosening detection in wind turbine tower
structures by vibration and impedance responses. Wind Struct. 2017, 24, 385–403. [CrossRef]
20. Wang, B.; Huo, L.; Chen, D.; Li, W.; Song, G. Impedance-Based Pre-Stress Monitoring of Rock Bolts Using a
Piezoceramic-Based Smart Washer—A Feasibility Study. Sensors 2017, 17, 250. [CrossRef] [PubMed]
21. Kim, J.-T.; Nguyen, K.-D.; Park, J.-H. Wireless impedance sensor node and interface washer for damage
monitoring in structural connections. Adv. Struct. Eng. 2012, 15, 871–885. [CrossRef]
22. Nguyen, K.-D.; Lee, S.-Y.; Lee, P.-Y.; Kim, J.-T. Wireless SHM for bolted connections via multiple
PZT-interfaces and Imote2-platformed impedance sensor node. In Proceedings of the 6th International
Workshop on Advanced Smart Materials and Smart Structures Technology (ANCRiSST2011), Dalian, China,
25–26 July 2011; pp. 25–26.
23. Huynh, T.-C.; Kim, J.-T. Impedance-Based Cable Force Monitoring in Tendon-Anchorage Using Portable
PZT-Interface Technique. Math. Probl. Eng. 2014, 2014, 11. [CrossRef]
24. Huynh, T.-C.; Lee, K.-S.; Kim, J.-T. Local dynamic characteristics of PZT impedance interface on tendon
anchorage under prestress force variation. Smart Mater. Struct. 2015, 15, 375–393. [CrossRef]
25. Huynh, T.-C.; Kim, J.-T. Compensation of temperature effect on impedance responses of PZT interface for
prestress-loss monitoring in PSC girders. Smart Mater. Struct. 2016, 17, 881–901. [CrossRef]
26. Huynh, T.-C.; Park, Y.-H.; Park, J.-H.; Kim, J.-T. Feasibility Verification of Mountable PZT-Interface for
Impedance Monitoring in Tendon-Anchorage. Shock Vib. 2015, 2015, 11. [CrossRef]
27. Park, J.-H.; Kim, J.-T.; Hong, D.-S.; Mascarenas, D.; Lynch, J.P. Autonomous smart sensor nodes for global and
local damage detection of prestressed concrete bridges based on accelerations and impedance measurements.
Smart Mater. Struct. 2010, 6, 711–730. [CrossRef]
28. Perera, R.; Pérez, A.; García-Diéguez, M.; Zapico-Valle, J. Active Wireless System for Structural Health
Monitoring Applications. Sensors 2017, 17, 2880. [CrossRef] [PubMed]
29. Johnson, K.L. Contact Mechanics; Cambridge University Press: Cambridge, UK, 1985.
30. Huynh, T.-C.; Kim, J.-T. Quantitative damage identification in tendon anchorage via PZT interface-based
impedance monitoring technique. Smart Mater. Struct. 2017, 20, 181–195.
31. Liang, C.; Sun, F.P.; Rogers, C.A. Coupled Electro.-Mechanical Analysis of Adaptive Material
Systems-Determination of the Actuator Power Consumption and System Energy Transfer. J. Intell. Mater.
Syst. Struct. 1994, 5, 12–20. [CrossRef]
32. Gresil, M.; Yu, L.; Giurgiutiu, V.; Sutton, M. Predictive modeling of electromechanical impedance
spectroscopy for composite materials. Struct. Health Monit. 2012, 11, 671–683. [CrossRef]
33. Ong, C.W.; Yang, Y.; Wong, Y.T.; Bhalla, S.; Lu, Y.; Soh, C.K. Effects of adhesive on the electromechanical
response of a piezoceramic-transducer-coupled smart system. Proc. SPIE 2003, 5062. [CrossRef]
34. Kim, J.-T.; Huynh, T.-C.; Lee, S.-Y. Wireless structural health monitoring of stay cables under two consecutive
typhoons. Struct. Monit. Maint. 2014, 1, 47–67. [CrossRef]
Sensors 2018, 18, 2766 20 of 20

35. Kim, J.-T.; Nguyen, K.-D.; Huynh, T.-C. Wireless health monitoring of stay cable using piezoelectric strain
response and smart skin technique. Smart Struct. Syst. 2013, 12, 381–397. [CrossRef]
36. Huynh, T.-C.; Park, J.-H.; Kim, J.-T. Structural identification of cable-stayed bridge under back-to-back
typhoons by wireless vibration monitoring. Measurement 2016, 88, 385–401. [CrossRef]
37. Kim, J.-T.; Park, J.-H.; Hong, D.-S.; Ho, D.-D. Hybrid acceleration-impedance sensor nodes on
Imote2-platform for damage monitoring in steel girder connections. Smart Struct. Syst. 2011, 7, 393–416.
[CrossRef]

© 2018 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/).