Preload Monitoring in Bolted Connection Using Piezoelectric ... - MDPI

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; [email protected] (T.-C.H.); [email protected] (N.-L.D.) * Correspondence: [email protected]; 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 (mcan s, kbe s, and cs). When bolt preloads are changed, the(m contact parameters of bolt the the splice plate also modeled bythe the respectively structural parameters cs ). When the s , ks , and connection are altered stiffness leading to(e.g., the contact variation in the coupled preloads are changed, the (e.g., contactcontact parameters of thereduction), connection are altered stiffness reduction), responses of variation the system at resonance. By monitoring the impedance responses of thethe system in the leading to the in the coupled responses of the system at resonance. By monitoring impedance resonant band, is possible detect the bolt looseness preload that have occurred in the responses of the it system in the to resonant band, it is possible or to detect thechanges bolt looseness or preload changes connection. that have occurred in the connection.

(a) A prototype of PZT interface.

(b) Bolted connection equipped with a PZT interface.

Bolt Preload F

Impedance

PZT Interface PZT mi

ki

Z(ω) F F+δF

F

Frequency

ci

Splice Plate

(k s , cs , ms )

kc cc (c) PZT interface-bolted connection interacting system. Figure Figure 1. 1. Impedance 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 responses responses of of aa bolted bolted connection connection measured measured via via the the PZT PZT interface interface can can be be theoretically derived from a simplified impedance model. Based on the previous studies [10,30], theoretically derived from a simplified impedance model. Based on the previous studies [10,30], a 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 the bolt preload is proposed, as shown in Figure 2. In the model, one dof refers representing the bolt preload is proposed, as shown in Figure 2. In the model, one dof refers to to the the interface ), and and the interface (m (mii,, cci,, and and kki), the other other dof dof refers refers to to the the splice splice plate plate (m (mss,, ccss,, and and kkss)) with with the the contact contact parameters ). parameters (c (ccc and and kkcc).

Sensors 2018, 18, 2766 Sensors 2018, 18, x FOR PEER REVIEW

4 of 20 4 of 20

Z eq (ω )

Z a (ω )

Bolted Connection

Interface

I (ω ) V (ω )

~

PZT

f i (ω )

ms

mi

ki ci

ks cs

ui

us

kc

cc

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 . . mi ui + ci (ui −k uks ) + k i (ui − us ) = f i ccs ccc . .. . .  s + s c uus s−−cci (i (uuii−−uuss))++ ks ksc uc s − mm u sk−i (ukii (−uiu−s )u s=) =0 0 s ussu+ cs +cc k s +k c c s + cc

.

..

.

k s + kc

(1) (1)

..

where ui ,, ui ,, ui and and us ,, us ,, us are corresponding to to where are the the displacements, displacements, velocities, velocities, and and accelerations accelerations corresponding masses m and m , respectively. s i masses mi and ms, respectively. Under the 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 Under described by by the the following: following: described ui = Ui e jωt ωt (2) usi ==UUi es ejjωt u (2) U s e jω t s = dependent where Ui and Us are complex quantities thatu are on the excited frequency and structural parameters 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 where Ui and of the system are given as follows: parameters of the system. 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 2 s cc s kc −( jωci + k i )Ui + −ω 2 m jω ( c + c+ + i(− k i(+jωkksc+ ) Us s == F 0i −ω s +m i + ijω cics+ i k+c ki ) U ckci) U

(

)

(3)

 2 cs cc Zeq , is defined k k  the ratio between(3) The equivalent of the system, the ). + ( ki + s c )as ci + ki ) U i +impedance − ( jωmechanical  −ω ms + jω (ci + U s = 0 c c k k + + excitation force fi and the velocity at s cui , given as s follows: c   the PZT driving point The equivalent mechanical impedance off ithe system, Fi e jωt Zeq, is defined as the ratio between the Z = = (4) . eq excitation force fi and the velocity at the PZT driving point jωU e jωt, given as follows: u i

i

jωt

f Ui , is Fe i obtained. By substituting the obtained Ui (4) By solving Equation (3), the quantity, into Zeq = i = jωt Equation (4), the equivalent mechanical impedance the system Z is obtained as follows: ui jωUof e eq i 
 By solving Equation (3), the quantity, i, is obtained. By(ksubstituting thejωc 2 Ui into cs cc 1 2m U −ω 2 mi + jωc + k − ω + jω ( c + ) + + k ) − ( obtained s s 1+ ξ i i i i i + ki ) cs +cc EquationZ(4), the equivalent mechanical impedance of the system Z eq is obtained as follows:   (5) eq = 1 s cc jω −ω 2 ms + jω (ci + ccs + c c ) + ( k i + k s 1+ ξ )  cc 1  2 )  − ( jω ci + ki ) −ω 2 mi + jω ci + ki  −ω 2 ms + jω (ci + s c ) + (ki + k s + + 1 ξ c c where ξ = ks /kc is defined as the ratio s c plate’s stiffness and the contact stiffness.  between the splice Z eq = (5) ξ ≈ 0 indicates the infinitive value of contact stiffness (i.e., fixed boundary), c c 1  while ξ ≈ ∞ indicates ) jω  −ω 2 ms + jω (ci + s c ) + (ki + k s the unnoticeable value of the contact stiffness (i.e., cfree boundary).1If + ξthe splice plate’s stiffness, ks , s + cc  remains unchanged, the increment of the ratio ξ = ks /kc will be equivalent to the decrement of the where = ks/kc is defined ascan thebe ratio between as thethe splice and(i.e., thebolt contact stiffness. ξ ≈ contactξstiffness, kc , which interpreted bolt plate’s preloadstiffness reduction 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:  Z (ω ) =

 −1 wa la T Za ( ω ) 2 ˆ E tan( kl a ) 2 ˆE jω d Y ε 33 (1 − jδ) − d31 Y11 + ta Za (ω ) + Zeq (ω ) 31 11 kla

(6)

E = (1 + jη )Y E is the complex Young’s modulus of the PZT sensor (width w , length l , where Yˆ11 a a 11 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 E , where ρ is the mass density of the PZT. The mechanical impedance of the PZT is ρ/Y11 computed as Za = − jYˆ E wa t a /ωla .

as k = ω

11

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: E = 6.098 × 1010 N/m2 , εT = 1.505 × 10−8 Farads/m, d = −1.71 × 1010 m/V, δ = 0.015, ρ = 7750 kg/m3 , Y11 31 33 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 Sensors 2018, 18, x FOR PEER REVIEW Sensors 2018, 18, x FOR PEER REVIEW

Peak Peak 22 31.42 kHz 31.42 kHz

2500 2500 Real Part Real Part 2000 Imaginary Part 2000 Imaginary Part

Imaginary Part (Ohm) Imaginary Part (Ohm)

Peak Peak 11 18.96 kHz 18.96 kHz

Real Part (Ohm) Real Part (Ohm)

900 900 800 800 700 700 600 600 500 500 400 400 300 300 200 200 100 100 0 0 10 10

6 of 20 6 of 20 6 of 20

1500 1500 1000 1000 500 500 0

15 15

20 20

25 25

30 30

35 35

40 40

45 45

0

-500 -500 50 50

Frequency (kHz) Frequency (kHz)

Figure 3. Impedance responses the impedance model with infinitive contact stiffness (ξ = /k ksc/k=c 0). = 0). Figure ofof the impedance model with infinitive contact stiffness Figure 3. 3. Impedance Impedance responses responses of the impedance model with infinitive contact stiffness (ξ(ξ== kkss/k c = 0). 400 400

k /k k s /ksc =c k /k k s /ksc =c k /k k s /ksc =c k /k k s /ksc =c k /k k s /ksc =c k /k k s /ksc =c

350 350 Real Impedance (Ohm) Real Impedance (Ohm)

300 300 250 250 200 200 150 150

=0 0 = 0.05 0.05 = 0.1 0.1 = 0.15 0.15 = 0.2 0.2 = 0.25 0.25

100 100 50 50 0 0 10 10

15 15

20 20

25 30 35 25 30 35 Frequency (kHz) Frequency (kHz)

40 40

45 45

50 50

The frequency range 10–50 kHz (a)(a) The frequency range ofof 10–50 kHz

300 300

300 300 250 250

Contact Contact Stiffness-loss Stiffness-loss

Real Impedance (Ohm) Real Impedance (Ohm)

Real Impedance (Ohm) Real Impedance (Ohm)

400 400

200 200

200 200

150 150 100 100

100 100

0 0 12 12

Contact Contact Stiffness-loss Stiffness-loss

14 14

16 18 20 16 18 20 Frequency (kHz) Frequency (kHz)

The frequency range 12–22 kHz (b)(b) The frequency range ofof 12–22 kHz

22 22

50 50 0 0 27 27

29 29

31 33 35 31 33 35 Frequency (kHz) Frequency (kHz)

37 37

The frequency range 27–37 kHz (c)(c) The frequency range ofof 27–37 kHz

Figure 4. Changes in the impedance responses of the impedance model under contact stiffness-loss. Figure Changes the impedance responses the impedance model under contact stiffness-loss. Figure 4. 4. Changes inin the impedance responses ofof the impedance model under contact 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, 2018, 18, 18, 2766 x FOR PEER REVIEW Sensors

of 20 20 88 of

(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) Poisson’s ratio, υ Mass density, ρ (kg/m3 ) Damping loss factor, η

70 0.33 2700 0.02

200 0.3 7850 0.02

6 0.38 1700 0.02

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

Value 

Elastic compliance, E (m2 /N) sijkl

Dielectric coupling constant, dkij (C/N)

Permittivity, ε Tjk (Farad/m)

      

 −5.74 −7.22 0 0 0 16.4 −7.22 0 0 0   −7.22 18.8 0 0 0   × 10−12 0 0 47.5 0 0   0 0 0 47.5 0  0 0 0 44.3 0 0 0 −171  0 0 −171     0 0 374   × 10−12   0 584 0     584 0 0  0 0 0  1730 0 0
 0 1730 0  × 8.854 × 10−12 0 0 1700

16.4 −5.74 −7.22 0 0 0

Mass density, ρ (kg/m3 ) Damping loss factor, η Dielectric loss factor, δ

7750 0.0125 0.015

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

Damage Case

Description

Intact

0% contact stiffness-loss 12.5% contact stiffness-loss 25% contact stiffness-loss 37.5% contact stiffness-loss 50% contact stiffness-loss

D1 D2 D3 D4

Value of Contact Stiffness (N/m2 /m) kx = ky

kz

2.0 × 1011

4.0 × 1011

1.75 × 1011

3.5 × 1011

1.5 × 1011

3.0 × 1011

1.25 × 1011

2.5 × 1011

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 Sensors 2018, 18, x FOR PEER REVIEW Sensors 2018, 18, x FOR PEER REVIEW

10 of 20 10 of 20 10 of 20

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

Peak 2 Peak 2 34.05 kHz 34.05 kHz

Real Part (Ohm) Real Part (Ohm)

Imaginary Part (Ohm) Imaginary Part (Ohm)

2000 2000 1500 1500 1000 1000 500 500

0

0 10

3000 3000 Real Part 2500 Real Part 2500 Imaginary Part 2000 Imaginary Part 2000 1500 1500 1000 1000 500 500 0 0 -500 -500 -1000 -1000 -1500 -1500 40 45 50 40 45 50

10

15

15

20

20

25

25

30

30

35

35

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 (a) Longitudinal bending motion at 17.73 kHz

(b) Lateral bending motion at 33.73 kHz (b) Lateral bending motion at 33.73 kHz

Figure 7. 7. Two bending modes of of PZT interface corresponding to Peak 1 and Peak 2. 2. Figure Two bending modes PZT interface corresponding to Peak 1 and Peak Figure 7. Two bending modes of PZT interface corresponding to Peak 1 and Peak 2.

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

11 of 20 11 of 20

consistent with thethe previous observations the two-dof impedancefrequency model, and alsotodemonstrated demonstrated sensitivity of the PZTfrom interface’s predetermined range the bolt preload the change. sensitivity of the PZT interface’s predetermined frequency range to the bolt preload change. 1200 Intact D1 D2 D3 D4

Real Impedance (Ohm)

1000 800 600 400 200 0 10

15

20

25

30 Frequency (kHz)

35

40

45

50

(a) The frequency range of 10–50 kHz. 700

1000

Contact Stiffness-loss

500

Real Impedance (Ohm)

Real Impedance (Ohm)

600

400 300 200 100 0 12

14

16 18 Frequency (kHz)

20

22

(b) The frequency range of 12–22 kHz

Contact Stiffness-loss

800 600 400 200 0 27

29

31 33 Frequency (kHz)

35

37

(c) The frequency range of 27–37 kHz

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

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

Damage Case

Damage Case

Intact Intact D1 D1 D2D2 D3D3 D4D4

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

Δf2 (%) 0 −0.15 −0.29 −0.44 −0.59

4. Experimental Evaluation on Lab-Scaled Bolted Girder Connection 4. Experimental Evaluation on Lab-Scaled Bolted Girder Connection 4.1.4.1. Experimental Setup Experimental Setup 4.1.1. Test-Setup of Bolted Girder Connection 4.1.1. Test-Setup of Bolted Girder Connection An An experimental evaluation waswas conducted on aonlab-scaled bolted girder connection. Figure 9 9 experimental evaluation conducted a lab-scaled bolted girder connection. 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 length ofof4.14 supported by bysteel steelbars barsatat four locations, as shown in with a total length 4.14m, m,was was simply simply supported four locations, as shown 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) by Sensors 2018, 18, x FOR PEER REVIEW of 20 splice plates and bolts at two flanges, see Figure 9b. The connection splice plate (310 × 200 × 10 mm) clamped eight (20 mm-diameter) is schematized in Figure 9c.interface The PZTprototype interfacesketched prototype by eightbybolts (20bolts mm-diameter) is schematized in Figure 9c. The PZT sketched in Figure was fabricated and surface-mounted to the of middle of theplate. splice plate. Thebody whole in Figure 5b was 5b fabricated and surface-mounted to the middle the splice The whole of the interface, including thethe flexible andand side sections, was fabricated from an an aluminium plate body of the interface, including flexible side sections, was fabricated from aluminium plate using a precisioncutting cuttingmachine. machine. Loctite Loctite 401 thethe PZT to to thethe using a precision 401 instant instant adhesive adhesivewas wasused usedtotoattach attach PZT middle section andthe thebonded bondedsections sectionsto to the the host host structure. structure. middle section and 1

2

Simple support

Bolted Connection

Simple support

Splice Plate 310 x 200

Bolts φ 20

100

1300

Simple support

Simple support

H-200ⅹ180ⅹ8ⅹ10 1340

1

100

1300

2

(a) Steel girder

1-1

2-2

50

Splice Plate

10

180

70

70

50

50

10 8

70

PZT Interface

100

200

50 Bolt 1

200

50 100 (b) Cross-section

Bolt 2

50

Bolt 3 310

Bolt 4

(c) Bolted connection.

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

Thereal realsetups setups of of the the steel bolted connection are illustrated in Figure 10. As 10. The steel girder girderand andthe the bolted connection are illustrated in Figure designed, all of the bolts of the connection were fastened to the torque of 160 Nm. A torque wrench As designed, all of the bolts of the connection were fastened to the torque of 160 Nm. A torque (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 of (i.e., 160 Nm to the torque of The 110 girder Nm (i.e., 31% torque-loss), 60 Nmwhere (i.e., athe 62% torque-loss), 0 Nm a 100% torque-loss). wasa placed in the laboratory, torque-loss), Nm (i.e., anear 100% The girder in the laboratory, temperatureand was0 controlled 22torque-loss). °C by air-conditioners so was as toplaced 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 Loosened Bolt 1 Bolt Bolt 1

Bolt 2

Bolt 2

Bolt 3 Bolt 3

Bolt 4

Bolt 4

Variation of Torque Level (Nm) Bolt 1: 160 of → Torque 110 (−31%) 60 (−62%) → 0 (−100%); Variation Level→(Nm) all others: 160 Bolt 1: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%); Bolt 2: 160all →others: 110 (−31%) 160 → 60 (−62%) → 0 (−100%); all 1600 (−100%); Bolt 2: 160 → 110 (−31%) → 60 (others: −62%) → all others: 160 Bolt 3: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%); Bolt 3: 160 → 110 (−31%) → 60 −62%) → all(others: 1600 (−100%); all others: 160 Bolt 4: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%); Bolt 4: 160 → 110 (−31%) → 60 (−62%) → 0 (−100%); all others: 160 all others: 160

Sensors FOR PEER REVIEW Sensors 2018, 2018, 18, 18, x 2766

(a) Test structure

13 13 of of 20 20

(b) Bolted connection.

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

Impedance Measurement System 4.1.2. Impedance SSeL-I impedance measurement systemsystem developed by the research A low-cost low-costand andmulti-channel multi-channel SSeL-I impedance measurement developed by the group at group Pukyong National National University [22], was [22], used was to wirelessly acquire theacquire impedance data from research at Pukyong University used to wirelessly the impedance the PZT Figure 11aFigure shows11a a prototype of the wireless thatnode consists of three data frominterface. the PZT interface. shows a prototype of the SSeL-I wirelessnode SSeL-I that consists layers, SSeL-I board, an Imote2 platform, and a battery board. board. The schematic of the of threean layers, animpedance SSeL-I impedance board, an Imote2 platform, and a battery The schematic of the SSeL-I node is shown in Figure 11b. key Thecomponent key component ofSSeL-I the SSeL-I board is low-cost the lowSSeL-I sensorsensor node is shown in Figure 11b. The of the board is the cost impedance AD9533, which a capability to measure the impedance up100 to 100 impedance chipchip AD9533, which has ahas capability to measure the impedance up to kHzkHz withwith the to 16 PZT the resolution than The SSeL-I boardintegrates integratesa amultiplexer multiplexerfor formeasuring measuring up up to resolution lessless than 0.10.1 Hz.Hz. The SSeL-I board recording temperature patches and the SHT11 sensor for recording temperature and and humidity. humidity. The Imote2 toto control impedance measurements via the board. The Imote2platform platformisisused used control impedance measurements via impedance the impedance board. Imote2 has has a high-speed PXA27x processor (clock speed ofof 13-416 The Imote2 a high-speed PXA27x processor (clock speed 13-416MHz), MHz),SRAM SRAMofof256 256kB, kB,the the flash flash of 32 memory of 32 MB, MB, and andthe theSDRAM SDRAMofof32 32MB MB[34–36]. [34–36].This Thisplatform platformisisdesigned designed with a wireless radio with a wireless radio of of GHz Zigbee for data transmission to a distance 125an mexternal by an external The 2.42.4 GHz Zigbee for data transmission (up to(up a distance of 125 of m by antenna).antenna). The wireless wireless sensor unit is powered via the battery board V). Although the wireless impedance sensor unit is powered via the battery board (3.2 V).(3.2 Although the wireless impedance sensorsensor node node a limited measurable frequency (i.e., than less than 100 kHz), it costs 300 USD has a has limited measurable frequency rangerange (i.e., less 100 kHz), it costs onlyonly 300 USD and and has has multi-channels could enable the cost-effectiveness a health monitoring system of inmega situ multi-channels that that could enable the cost-effectiveness for a for health monitoring system of in situ mega structures. boltedbolted structures.

(a) A prototype of SSeL-I

(b) Schematic of SSeL-I.

Figure 11. 11. Wireless Wireless SSeL-I impedance sensor node. Figure

Sensors 2018, 18, 2766 Sensors 2018, 18, x FOR PEER REVIEW

14 of 20 14 of 20

4.2.Preload PreloadChange ChangeMonitoring MonitoringininBolted BoltedGirder GirderConnection Connection 4.2. 4.2.1. 4.2.1.Impedance ImpedanceMeasurement Measurementvia viaPZT PZTInterface Interface The Theimpedance impedancesignatures signatureswere weremeasured measuredininthe thefrequency frequencyrange rangeofof10–50 10–50kHz kHztotoidentify identifythe the sensitive Section 3. 3. The Theamplitude amplitudeof sensitiveimpedance impedancepeaks peaks(Peaks (Peaks 1–2), 1–2), as as numerically numerically pre-analyzed in Section ofthe theexcitation excitation voltage at and 1V, and the resolution thescanning PZT scanning frequency voltage waswas set set at 1V, the resolution of theofPZT frequency was 0.1 was kHz. 0.1 kHz. Four measurements repeated measurements were for conducted for bolt-loosening each of the bolt-loosening Four repeated were conducted each of the cases in Table cases 5. For in the Table 5. For theevaluation, performance impedance signatures by the wireless SSeL-I performance theevaluation, impedancethe signatures measured by measured the wireless SSeL-I system were system werewith compared with those using a wired high-performance analyzer HIOKI-3532. compared those using a wired high-performance impedanceimpedance analyzer HIOKI-3532. As shown As Figure theimaginary real and imaginary signaturesby measured bysystem the SSeL-I in shown Figure in 12a,b, the 12a,b, real and impedanceimpedance signatures measured the SSeL-I were system were well-matched with those by the wired HIOKI system for the same frequency range with well-matched with those by the wired HIOKI system for the same frequency range with identical identical patterns.patterns. 700

5000

Wired HIOKI Wireless SSeL-I

600

Wired HIOKI Wireless SSeL-I

4000 Imaginary Part (Ohm)

Real Part (Ohm)

500 400 300 200

3000

2000

1000 100 0 10

20

30 Frequency (kHz)

40

50

(a) Real impedance signatures

0 10

20

30 Frequency (kHz)

40

50

(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 of Bolt Bolt33were wereplotted plottedinin Figure Theimpedance impedanceresponses responsesunder under the the bolt-loosening bolt-loosening cases cases of Figure 13.13. As As expectedfrom from the analysis, twotwo resonant bandsbands (i.e., 12–22 andkHz 27–37and Hz) exist expected thenumerical numerical analysis, resonant (i.e.,kHz 12–22 27–37containing Hz) exist significant impedance peaks between 10–50 as zoomed Figure The comparison between containing significant impedance peaks kHz, between 10–50inkHz, as13b,c. zoomed in Figure 13b,c. The Figures 8 and 13 revealed certain gaps between the experimental measurement and the numerical analysis. comparison between Figure 13 and Figure 8 revealed certain gaps between the experimental These gaps couldand be caused by the differences in the structural parameters between experimental measurement the numerical analysis. These gaps could be caused bythe the differencesmodel in the and the FE model. For Peak 1, the numerical pre-analysis in Figure 8b predicted the peak frequencies structural parameters between the experimental model and the FE model. For Peak 1, the numerical around 18 kHz,inwhile the8b experimental measurement in Figure 13b showed thewhile peak frequencies near pre-analysis Figure predicted the peak frequencies around 18 kHz, the experimental 16measurement kHz (i.e., the prediction error of 11.1%). For Peak 2, the pre-analysis in Figure 8c estimated the peak in Figure 13b showed the peak frequencies near 16 kHz (i.e., the prediction error of frequency at about kHz, while the in experiment Figure 13c the peak frequency at about 11.1%). For Peak 2,34the pre-analysis Figure 8c in estimated themeasured peak frequency at about 34 kHz, while 30the kHz (i.e., the prediction of 11.8%). the peak frequency at about 30 kHz (i.e., the prediction error experiment in Figureerror 13c measured As observed in Figure 13, the impedance peaks tended to shift left as the torque was reduced. of 11.8%). As compared with the first resonant band (i.e., 12–22 thetosecond oneas(i.e., As observed in Figure 13, the impedance peakskHz), tended shift left the 27–37 torquekHz) was showed reduced. less the the torque-loss severity. The(i.e., changing thesecond two resonant bands was consistent Assensitivity comparedto with first resonant band 12–22 trend kHz),ofthe one (i.e., 27–37 kHz) showed with the previous numerical results. The first resonant band experienced both the frequency less sensitivity to the torque-loss severity. The changing trend of the two resonant bandsand was magnitude the second one showed slight changes in the peak frequency and almost consistent shifts, with while the previous numerical results. The first resonant band 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 FOR PEER REVIEW Sensors 2018, 2018, 18, 18, x2766

15 15 of of 20 20

600 Intact 31% Torque Loss 62% Torque Loss 100% Torque Loss

Real Impedance (Ohm)

500 400 300 200 100 0 10

15

20

25

30 35 Frequency (kHz)

40

45

50

(a) The frequency range of 10–50 kHz. 600

Real Impedance (Ohm)

Real Impedance (Ohm)

400

300

200

100

0 12

14

16 18 Frequency (kHz)

20

500 400 300 200 100 0 27

22

(b) The frequency range of 10–20 kHz

29

31 33 Frequency (kHz)

35

37

(c) The frequency range of 24–34 kHz.

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

4.2.2. 4.2.2. Detection Detection of of Preload Preload Change Change using using Impedance Impedance Response Response Statistical Quantification Quantification Method Statistical To detect detectthe thepreload preload change in bolted the bolted connection, two common damage-sensitive To change in the connection, two common damage-sensitive features features were extracted from the impedance data, the correlation coefficient(CCD) deviation (CCD) and were extracted from the impedance data, the correlation coefficient deviation and root-meanroot-mean-square These two impedance features quantify changesinin the the square deviation deviation (RMSD) (RMSD) indices. indices. These two impedance features quantify thethe changes impedance signatures with different manners. While the CCD index mainly quantifies the frequency impedance signatures with different manners. While the CCD index mainly quantifies the frequency shift of the RMSD index quantifies bothboth the frequency and magnitude shifts. shift the impedance impedancesignatures, signatures,the the RMSD index quantifies the frequency and magnitude According to [23], the CCD index can be computed using the below formula: shifts.

According to [23], the CCD index the below formula: o ncan be computed using  1 ∗ Z ) [ Re ( Z ∗ (ωi )) − Re( Z )] CCD = 1 − E Re ( Z ( ω )) − Re ( i 1σZ σZ∗ * *

{

}

(7)

E Re(Z (ωi )) − Re(Z )  Re(Z (ωi )) − Re(Z )  (7) * σ σ Z Z where E[·] is the expectation operation; Z (ω ) and Z ∗ (ω ) signify the impedance responses at the ith CCD = 1 −

i

i

∗

frequency and after a damage event, Z indicate the mean values ofat those where E[·] before is the expectation operation; ( respectively; ) and ∗ ( Z) and signify the impedance responses the ∗ impedance responses; and σ and σ are the corresponding standard deviations. * Z Z ith frequency before and after a damage event, respectively; Z and Z indicate the mean values of 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 u 2 t ∗ N ∑ [ Z ( ωi ) − Z ( ωi )] N RMSD = ∑ [Z(2 ωi )]2 2 * RMSD =  i = 1Z (ω i ) − Z (ω i )   [ Zi =(ω1i ) ] i =1

(8)

(8)

i =1

wherewhere N denotes the number of swept frequencies. N denotes the number of swept frequencies. For distinguishing the bolt-loosening state from state,an analarming alarming threshold known For distinguishing the bolt-loosening state fromthe thehealthy healthy state, threshold known as theas upper control limit (UCL) can be established using the values of the impedance features under the upper control limit (UCL) can be established using the values of the impedance features under the intact state state [25,37], as follows: the intact [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

60

60

50

50

40

40 RMSD(%)

RMSD(%)

The RMSD control chart waswas constructed dataininthe thepredetermined predetermined The RMSD control chart constructedusing using the the impedance impedance data frequency bandband of 10–50 kHz. showsthe theRMSD RMSD index plotted according frequency of 10–50 kHz.Figure Figure 14 14 shows index thatthat was was plotted according to the to the torque-loss forallallof of loosening of1–4. Bolts 1–4. As from observed 14a–d, torque-loss level level for thethe loosening casescases of Bolts As observed Figurefrom 14a–d,Figure the RMSD values were very the intact case, but case, became as Bolts 1–4as experienced 31%, 62%, a the RMSD values weresmall veryfor small for the intact butnoticeable became noticeable Bolts 1–4 aexperienced and 100% torque-loss. The UCL thresholds were computed to classify the bolt-loosening events. As 31%, 62%, and 100% torque-loss. The UCL thresholds were computed to classify the bolt-loosening plotted in Figure 14a–d, for all of the torque-loss events, the RMSD values were above the events. As plotted in Figure 14a–d, for all of the torque-loss events, the RMSD values weredefined above the thresholds, indicating the successful detection of the preload changes. defined thresholds, indicating the successful detection of the preload 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 of the PZT interfacebolts werenear quite to the(i.e., small preload occurred in the signatures bolted connection. The loosened thesensitive PZT interface Bolts 2 and 3)changes were detected in the with bolted connection. The loosened bolts near the PZT interface (i.e., Bolts 2 and 3) were detected higher severity estimations than those far from the interface (i.e., Bolts 1 and 4). These results with higher severity estimations than those far from the interface (i.e., Bolts 1 and 4). These results confirmed that the single PZT interface was able to monitor multiple loosened bolts on the splice confirmed plate.that the single PZT interface was able to monitor multiple loosened bolts on the splice plate. Forcomparison, the comparison, CCD control chartwas wasalso alsoconstructed constructed by impedance For the the the CCD control chart byusing usingthe thesame same impedance data, as shown in Figure 15a–d. As compared with the RMSD, the values of the CCD index were data, as shown in Figure 15a–d. As compared with the RMSD, the values of the CCD index were relatively smaller. It is noted from Figure 13 that the impedance signatures showed both frequency relatively smaller. It is noted from Figure 13 that the impedance signatures showed both frequency and and magnitude under the bolt-loosening events. Thus, RMSD approachconsidering considering both magnitude changes changes under the bolt-loosening events. Thus, the the RMSD approach both 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 in Figure Although the thresholds, values of the CCD index quitebolt-loosening small, those for forsketched the damage cases 15a–d. were above the UCL indicating the were successful the damage cases were above the UCL thresholds, indicating the successful bolt-loosening detection. detection.

30

UCL = 5.42 % 20 9.45

10

12.75

15.87

30 20

11.79

Intact

31% Loss 62% Loss Torque Loss in Bolt 1

12.33

10

1.77 0

23.11

UCL = 3.86 %

100% Loss

0

(a) Bolt 1-loosening cases

1.23 Intact

31% Loss 62% Loss Torque Loss in Bolt 2

100% Loss

(b) Bolt 2-loosening cases.

Figure 14. Cont.

Sensors 2018, 18, x FOR PEER REVIEW Sensors 2018, 18, 2766

17 of 20 17 of 20

60 18, x FOR PEER REVIEW Sensors 2018,

60 50.93

50 60

UCL = 6.66 % 21.54

20 30

24.42

UCL = 6.66 %

21.54

10 20

40 50 RMSD(%) RMSD(%)

RMSD(%) RMSD(%)

30 40

50 60

50.93

40 50

24.42

30 40

0

Intact 2.18 Intact

31% Loss

62% Loss

100

100% Loss

0

100% Loss

19.41 14.29

10 20

31% Loss 62% Loss Torque Loss in Bolt 3

UCL = 4.17 %

20 30

2.18 100

17 of 20

(c) BoltTorque 3-loosening Loss in Boltcases 3

UCL = 4.17 %

19.41

7.43 14.29

1.33 Intact 1.33 Intact

7.43 31% Loss 62% Loss Torque Loss in Bolt 4 31% Loss

100% Loss

62% Loss

100% Loss

(d) Bolt 4-loosening Torque Loss in Bolt cases. 4

Bolt 3-loosening cases (d)root-mean-square Bolt 4-loosening cases. Figure 14. (c) Preload change monitoring of bolted connection by deviation (RMSD) index. Figure change monitoring of bolted connection by root-mean-square deviation (RMSD)(RMSD) index. Figure14. 14.Preload Preload change monitoring of bolted connection by root-mean-square deviation 30

30

25 30

25 30

20 25

20 25 CCD(%)CCD(%)

CCD(%)CCD(%)

index.

15 20 10 15 105 50 0

UCL = 0.36 % UCL = 0.36 % 1.00

0.29 Intact 0.29 Intact

(a)

30

1.62

31% Loss 62% Loss Torque Bolt 1 1.00 Loss in1.62

15 20 10 15 105

2.40

50

100% 2.40Loss

0

31% Loss 62% Loss 100% Loss BoltTorque 1-loosening cases Loss in Bolt 1

UCL = 0.31 % 4.17

UCL = 0.31 % 1.41

0.27 Intact 0.27 Intact

1.54 4.17 100% Loss

31% Loss 62% Loss 1.41 Loss in1.54 Torque Bolt 2 31% Loss

62% Loss

100% Loss

(b) BoltTorque 2-loosening Loss in Boltcases 2

30

(a) Bolt 1-loosening cases

25 30

(b) Bolt 2-loosening cases

25 30 21.03 20 25 21.03

15 20 10 15 105 50 0

UCL = 0.41 % 4.50

5.47

0.30 0.30 Intact

(c)

5.47 4.50 31% Loss 62% Loss Torque Loss in Bolt 3

15 20 10 15

UCL = 0.32 %

105

UCL = 0.41 % Intact

CCD(%)CCD(%)

CCD(%)CCD(%)

20 25

100% Loss

31% Loss 62% Loss 100% Loss BoltTorque 3-loosening cases Loss in Bolt 3

50 0

UCL = 0.32 % 0.27 Intact 0.27 Intact

0.57

31% Loss 62% Loss 1.88 Torque 0.57 Loss in Bolt 4 31% Loss

3.43

1.88

62% Loss

3.43Loss 100% 100% Loss

(d) BoltTorque 4-loosening Loss in Boltcases 4

(d) Boltcoefficient 4-loosening cases (c) Boltchange 3-loosening casesof bolted Figure 15.15. Preload monitoring connection by correlation deviation (CCD) index. Figure Preload change monitoring of bolted connection by correlation coefficient 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, the piezoelectric-based interface technique was developed to acquire demonstrate the theoretical feasibility of thesmart proposed method, a simplified 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) (2)

(3)

The PZT interface’s sensitive frequency band, predetermined by the numerical simulation, was quite consistent with that measured from the experiment. 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. 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. 2. 3. 4. 5. 6. 7. 8. 9. 10.

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] Kim, N.; Hong, M. Measurement of axial stress using mode-converted ultrasound. NDT E Int. 2009, 42, 164–169. [CrossRef] Joshi, S.G.; Pathare, R.G. Ultrasonic instrument for measuring bolt stress. Ultrasonics 1984, 22, 261–269. [CrossRef] 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] Doyle, D.; Zagrai, A.; Arritt, B.; Çakan, H. Damage Detection in Bolted Space Structures. J. Intell. Mater. Syst. Struct. 2010, 21, 251–264. [CrossRef] 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] 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] 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] 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] 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

11. 12. 13. 14. 15. 16.

17. 18. 19. 20. 21. 22.

23. 24. 25. 26. 27.

28. 29. 30. 31.

32. 33. 34.

19 of 20

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] 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. 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] Wang, T.; Song, G.; Liu, S.; Li, Y.; Xiao, H. Review of Bolted Connection Monitoring. Int. J. Distrib. Sens. Netw. 2013, 9, 871213. [CrossRef] 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] 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] 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] 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] 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] 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] 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] 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. 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] 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] 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] 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] 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] Perera, R.; Pérez, A.; García-Diéguez, M.; Zapico-Valle, J. Active Wireless System for Structural Health Monitoring Applications. Sensors 2017, 17, 2880. [CrossRef] [PubMed] Johnson, K.L. Contact Mechanics; Cambridge University Press: Cambridge, UK, 1985. 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. 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] 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] 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] 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

35. 36. 37.

20 of 20

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] 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] 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/).