A Synthetic Phased Array Surface Acoustic Wave Sensor for Quantifying Bolt Tension

Abstract

In this paper, we report our findings on implementing a synthetic phased array surface acoustic wave sensor to quantify bolt tension. Maintaining proper bolt tension is important in many fields such as for ensuring safe operation of civil infrastructures. Significant advantages of this relatively simple methodology is its capability to assess bolt tension without any contact with the bolt, thus enabling measurement at inaccessible locations, multiple bolt measurement capability at a time, not requiring data collection during the installation and no calibration requirements. We performed detailed experiments on a custom-built flexible bench-top experimental setup consisting of 1018 steel plate of 12.7 mm (½ in) thickness, a 6.4 mm (¼ in) grade 8 bolt and a stainless steel washer with 19 mm (¾ in) of external diameter. Our results indicate that this method is not only capable of clearly distinguishing properly bolted joints from loosened joints but also capable of quantifying how loose the bolt actually is. We also conducted detailed signal-to-noise (SNR) analysis and showed that the SNR value for the entire bolt tension range was sufficient for image reconstruction.

Full text

Sensors 2012, 12, 12265-12278; doi:10.3390/s120912265
sensors
ISSN 1424-8220
www.mdpi.com/journal/sensors
Article
A Synthetic Phased Array Surface Acoustic Wave Sensor for
Quantifying Bolt Tension
Jairo Martinez †, Alper Sisman †,‡, Onursal Onen, Dean Velasquez and Rasim Guldiken *
Department of Mechanical Engineering, University of South Florida, 4202 E Fowler Ave., ENB 118,
Tampa, FL 33620, USA; E-Mails: jairo@mail.usf.edu (J.M.); alpersisman@mail.usf.edu (A.S.);
onursalonen@mail.usf.edu (O.O.); dvelasqu@mail.usf.edu (D.V.)
† These authors contributed equally to this work.
‡ Permanent address: Department of Electrical and Electronics Engineering, Marmara University,
Goztepe, Istanbul 34722, Turkey; E-Mail: alper.sisman@marmara.edu.tr.
* Author to whom correspondence should be addressed; E-Mail: guldiken@usf.edu;
Tel.: +1-813-974-5628; Fax: +1-813-974-3539.
Received: 10 July 2012; in revised form: 27 August 2012 / Accepted: 4 September 2012 /
Published: 7 September 2012
Abstract: In this paper, we report our findings on implementing a synthetic phased array
surface acoustic wave sensor to quantify bolt tension. Maintaining proper bolt tension
is important in many fields such as for ensuring safe operation of civil infrastructures.
Significant advantages of this relatively simple methodology is its capability to assess bolt
tension without any contact with the bolt, thus enabling measurement at inaccessible
locations, multiple bolt measurement capability at a time, not requiring data collection
during the installation and no calibration requirements. We performed detailed experiments
on a custom-built flexible bench-top experimental setup consisting of 1018 steel plate of
12.7 mm (½ in) thickness, a 6.4 mm (¼ in) grade 8 bolt and a stainless steel washer with
19 mm (¾ in) of external diameter. Our results indicate that this method is not only capable
of clearly distinguishing properly bolted joints from loosened joints but also capable of
quantifying how loose the bolt actually is. We also conducted detailed signal-to-noise
(SNR) analysis and showed that the SNR value for the entire bolt tension range was
sufficient for image reconstruction.
OPEN ACCESS

Sensors 2012, 12 12266
Keywords: bolt tension; real area of contact; non-destructive testing; surface acoustic
wave sensor; synthetic phase array; ultrasonic imaging
1. Introduction
Threaded fasteners are one of the most versatile methods for assembling structural components.
They are used in many types of physical assemblies, ranging from small toys to large-scale civil
infrastructures. Even in an assembly of a single structure, bolts with wide ranging sizes and material
properties are actively used. For example, for bridges, large bolts are used for fixing base columns and
small bolts are used for attaching maintenance access ladders. As expected, not all bolts are critical for
the safe operation of the structure. Relatively low load-bearing fasteners used in large quantities are
not monitored and maintained carefully due to cost and feasibility constraints. However, even
“non-critical” bolts can cause significant structural problems, even catastrophic failure, if many of
them fail.
Loosening of threaded fasteners is a common cause of structure failure. This phenomenon is
commonly referred as “slip” at the thread-plate and head-plate interface [1–7]. Loosened bolts are
problematic for the clamping force (CF) of the joint. CF is the parameter that defines the effectiveness
of a joint. Insufficient CF may lead to joint separation. In the case of separated bolted joints,
aggressive fatigue is induced on the remaining bolts due to increase in the tensional stress [2,5].
In addition, loosened bolts tend to loosen even faster than properly tensioned bolts leading to
accelerated bolt loosening as the structure ages [2]. Although CF is a critical parameter, direct
measurement of CF is very challenging. Instead, bolt tension is typically used to quantify the clamping
force in the joint. CF is proportional to the bolt tension by the combined stiffness of the clamped
members and the fasteners. Tension of all the fasteners in a joint builds up the common force that
ensures clamping. Hence, bolt tension of all the fasteners present in a joint is critical for CF [2,5].
Currently, there are four common methodologies used to control and quantify bolt tension: torque
control, turn-of-nut method, direct preload control and stretch control. Torque control is by far the
most frequently used method for bolt tension. In this methodology, manual, pneumatic or hydraulic
torque wrenches are used to apply wide range of torque to the bolt [2]. The inherent dependency of this
method on various variables such as friction factor, torsion, bending and plastic deformation of the
threads reduce the measurement accuracy of the applied tension to as low as 30% [2,8]. Turn-of-nut
control method has two stages. In the first stage, the bolt is tightened with a conventional torque
wrench until it reaches approximately 75% of the material ultimate strength [2]. The second stage
involves an additional turn of 180° after the initial tightening. Every 360° degree that the bolt rotates
increases the bolt length (and hence the tension) by the bolt pitch, so the final turn ensures a tension
that creates stresses larger than the bolt material yield point [2]. Tension accuracy of 5% is reported
with this methodology, however this method can only be used for bolts made out of ductile materials
with long and well-defined elastic deformation regions [2]. Direct preload control method uses direct
estimators of the tension such as strain, stress or deformation. Judiciously positioned strain gages can
measure the strain very precisely leading to tension estimation with an accuracy of as high as 1% [2].

Sensors 2012, 12 12267
Bolt stretch control is another way to quantify the bolt tension. This method employs the transit time of
ultrasound measurements along the bolt length to quantify the bolt tension. The main advantage of this
methodology is that measurement is independent of friction between the fastener and the clamped
plate. Avoiding the friction permits it to be almost as accurate as the instrument used to measure the
bolt length change [2]. Additionally, this approach makes it possible to monitor the fastener tension
levels only by comparing the new length of the bolt to the value recorded during installation [2].
However, as expected, many older structures with no recorded bolt lengths during installation cannot
be evaluated with this method. Also, length variations due to irregular surfaces, uneven machinated
processes, temperature changes, plastic deformations and bending displacements introduce significant
error to the length estimation.
The common issues with all these frequently used methodologies are their limitation to quantify the
bolt tension only during the assembly, measurement capability for a single bolt at a time, relatively
high cost and frequent calibration requirement. However, critical bolts should be monitored real-time
and single bolt measurement at a time is a significant limitation for large structures with high number
of bolts. Response analysis of induced vibrations may be used to characterize the general state of
bolted joints. For instance, statistical manipulation can be employed to obtain changes in the vibration
signals due to bolt preload variation [9,10]. Recently, signal processing algorithms based on empirical
mode decomposition (EMD) have been used to detect changes in vibration signals produced by impact
hammers [11]. This method was also validated empirically and by Finite Element Analysis (FEM).
Vibration can be used to monitor common problems such as stability or resistance in a structure, but
identifying the location of the problem is a fairly difficult task. Local approaches such as monitoring
the tension of individual fasteners can be employed [12–15]. The deformations on the fastener were
measured by using automatic digital image correlation (ADIC) [12]. Also, piezoelectric sensors were
installed in fasteners to sense changes in the electromechanical impedance of the bolts [15]. Guided
ultrasonic waves were also used to measure bolt tension. Modulation in Lamb waves due to loosened
bolts was used to calculate a joint damage index [13]. Transformation between the wave modes due to
stress was exploited by [14] with the purpose of calculating stress levels, tension and CF in the
threaded joints. These methodologies often require a direct contact with each bolt to be inspected.
Furthermore, majority of the methods require either disassembly of the components or data collected
during the installation in order to establish the tension levels.
In this study, we investigate the use of surface acoustic waves (SAW) to quantify bolt tension
without contacting the bolt. The advantages of SAW as compared to other acoustic modes are that they
propagate over long distances with little amplitude loss enabling large area inspection with minimal
number of sensors reducing the cost and complexity [16]. Secondly, SAWs are capable of propagating
through curves surfaces and to inaccessible locations, simplifying the design. Also, SAWs are sensitive
to many different kinds of flaws. Here, real contact area (RAC), which grows with increased bolt
tension, is quantified by its acoustic signature response to SAWs. RAC formed in-between the bolt
head and the clamped elements were imaged using a synthetic phase array. SAWs are generated and
sensed by off-the shelf bulk piezoelectric transducers connected to a custom-designed ultrasonic
wedge. In this paper, we first describe the methodology developed to quantify bolt tension. We, then,
discuss the design of experiments including the experimental setup. Finally, we present and discuss the
experimental results obtained with the synthetic phased array surface acoustic wave sensor.

Sensors 2012, 12 12268
2. Methodology
The interaction of various acoustic wave modes with the boundaries of rough surfaces has been
widely studied [17–22]. SAWs interact with solid elements in the surface of the propagation
medium [8]. The interaction (for instance with the bolted joint) takes place at the points where the
propagation surface and the external body are directly in contact with each other, referred as points of
real contact. As illustrated in the Figure 1(a), the points of real contact correspond to the tips of the
surface peaks in both materials making direct contact with each other and this represent only a small
portion of the entire apparent area of contact. Real area of contact (RAC) is the sum of all these
individual contact points [23–25].
Figure 1. (a) Schematic representation of the real contact area of a clamped plate and
washer; (b) schematic representation of change in the real area of contact as a function of
increasing bolt tension.
(a)
(b)
It has been reported that the number and size of points of contacts increase with increasing tension
in the bolt [24,25]. Figure 1(b) illustrates a schematic representation of the RAC growth due to
increase in bolt tension. As can be observed from this figure, the higher the tension (signified with
“arrow”), the larger the RAC between the plate and the washer (signified with the “donuts”). The RAC

Sensors 2012, 12 12269
is, in fact, different from the apparent area of contact, but at high-tension levels the difference between
those becomes minimized [24]. The saturation of the RAC at a specific preload level establishes the
maximum tension level that should be applied to the bolt for optimal and safe operation. The way in
which the RAC grows is of special interest for estimating the tension. As illustrated in Figure 1(b), the
RAC grows from the center of the washer towards the washer perimeter. This behavior can be
explained by analyzing the way the tension is applied to the bolt head. First, the preload transforms
into a pulling force exerted to the head center. The surface peaks closer to the center are then affected
more than the ones in the perimeter and therefore the expansion of the RAC can be predicted as
propagating radially outward.
Figure 2. (a) Schematic representation of the proposed methodology used to quantify
bolt tension; (b) representation of surface acoustic wave propagation and reflection from
four different boundaries. White arrows represent no load, green arrows low load, blue
arrows medium load and red arrows high load. The black arrows represent the incoming
acoustic wave.
(a) (b)
In this study, the interaction of the SAW with the bolt and RAC was investigated by a set-up
illustrated in Figure 2(a). Waves created by a SAW generator are directed towards a bolted joint. As
the tension at the bolt is increased, the position of the reflective boundary is expected to move from the
edge of the joint towards the outer washer perimeter, as presented in Figure 2(b). It is hypothesized
that the increase in the bolt tension results in a proportional increase in the RAC, represented by the
circles in Figure 2(b). The arrows represent the incident (black) and reflected (colored) waves. The
white, green, blue and red colored arrows represent the SAW reflections from the plate hole and the
RAC corresponding to tension states of low, medium and high intensity respectively. The method
employs a phased array to obtain the B-scan ultrasonic image of the bolted joint in order to quantify
the RAC, which is proportional to tension level in the bolted joint. Ultrasonic image of the bolt
assembly is obtained by generating SAW pulses and receiving the echoes using the array elements.
Each beam line in the image is constructed by summing the beamformed signals. The construction of
image is illustrated in Figure 3 where the τ1 to τN represent the proper delays and the output is the
beamformed signal corresponding to focal point. The delay profile for each focal point in the beam line
is calculated using the following equation:

Sensors 2012, 12 12270

󰇡󰇢 (1)
where c is the SAW velocity, (xi,0) is the coordinate of array element, (xp,zp) is the coordinates of the
focal point on the beam line and Rp is the distance between the array center and focal point. SAW
velocity (~2,900 m/s) was calculated by the Rayleigh wave equation [26], and verified by ultrasonic
delay time (pulse-receive) experiments performed on the 1018 steel plate by using two transducers.
The delay that is calculated for each transducer element basically eliminates the differences of flight
time that is the travel duration from focal point to the each array element. The sum of properly delayed
signals results in an in-phase sum called beamformed signal. The mathematical expression to
determine the beamformed signal is given in Equation (2). Here Ai is the aperture function applied to
each receive and transmit element, si,j is the received signal corresponding to ith transmit and jth
receive element, τi is refer to transmit and receive delays [27]. Dynamic focusing is employed in
receive mode, hence each beam line is obtained by summing the beamformed signals obtained by
focusing to each radial point [28]. The combined beamlines form the image of the interested area
(illustrated in Figure 3(b)).
(2)
The location of bolt boundary is obtained using the ultrasonic image by measuring the distance
between the boundary and the array center since it is proportional to the RAC. The distance is an
important parameter to determine the health of the bolted joint.
Figure 3. Scanning the interested area using transducer array. (a) Construction of a
beamformed signal; (b) Scanning by constructing successive beamlines.
(a) (b)
3. Experimental Design
The proposed concept was investigated by experimental studies. The experimental setup is
illustrated by the block diagram (Figure 4(a)). The tested joint was composed of a 1018 steel plate of
12.7 mm (½ in) thickness, a 6.4 mm (¼ in) grade 8 bolt and stainless steel washer with 19 mm (¾ in)
of external diameter, as illustrated in Figure 4(b). The joint was lubricated with general purpose oil in
order to facilitate the tightening process. The wideband excitation pulses were generated and the
()
,,, ,,
11
2
()
NN
p
tx i rx j i j tx i rx j
ij
R
Bt A A s t c
ττ
==
=−−+
∑∑

Sensors 2012, 12 12271
echoes were received and amplified by a pulsar-receiver (Olympus 5072PR, Olympus NDT Inc.,
Waltham, MA, USA). The received signals were displayed/digitized by a scope (Tektronix TDS
2024B, Textronix Inc., Beaverton, OR, USA). The digitized signals are stored in a computer and the
image is generated using a custom-designed MATLAB code.
We obtained the ultrasound images using simpler synthetic array imaging system instead of more
involved phased array system for the proof of the proposed concept studies. Synthetic array imaging is
chosen due to the simplified beamformer structure although it may degrade the image quality by
lowering the signal to noise ratio (SNR).
Figure 4. (a) The block diagram and (b) the photographic illustration of the
experimental setup.
(a) (b)
Synthetic phased array was formed by a single angle beam transducer actuated from judicious
several positions within a linear array. The wideband SAW pulse is transmitted and the echo is received
in each movement step (Figure 5(a)). Consequently N-element linear array is synthetically formed
using a single moving transducer. It should be noted that synthetic array imaging cannot achieve all
of the TX/RX combinations; the transmitted array element must receive the echo. The missing
transmit-receive (TX/RX) combinations resulting in decreased SNR as compared to phased imaging
modality. However, the major advantage of synthetic phased array modality is significantly simplified
beamformer since the total number of TX/RX combinations is decreased dramatically [27–31]:
(3)
In this study, we used a 5 MHz bulk piezoelectric transducer (C541-SM, Olympus NDT Inc.) with
circular aperture of 12.7 mm (½ in) attached to an ultrasonic wedge that is specifically designed for
converting bulk pressure wave (P-wave) into SAWs in steel (ABWML-5T, Olympus NDT Inc.).
Ultrasonic couplant (Soundsafe, Sonotech Inc., Glenview, IL, USA) was used at the surfaces between
the wedge and the transducer and between the wedge and the test piece to facilitate the transmission of
ultrasonic energy. The transducer-wedge system is linearly moved to realize synthetic array. The
position control of the angle beam transducer (transducer-wedge attachment) was achieved through a
manual translator stage (PT3, Thorlabs Inc., Newton, NJ, USA) consisting of micrometers with
resolution of 25 µm and maximum displacement range of 25.4 mm. The transducer and the wedge
were adapted to the translator stage by custom designed aluminum adaptors. In Figure 5(b), two
()
,,,
1
2
2
N
p
rx i tx i i i i
i
R
Bt A A s t c
τ
=
⎛⎞
=−+
⎜⎟
⎝⎠
∑

Sensors 2012, 12 12272
important features of the synthetic linear array are illustrated: array pitch (p) and total array length. In
this study, the array pitch was selected as 250 µm corresponding to less than half the wavelength. With
this selection criterion, sufficient lateral resolution was achieved and formation of side lobes in the
acoustic beam was avoided [30]. The number of elements in the array determines the array size which
is proportional to the lateral resolution of the system. In this study, an array with 50 elements was used
corresponding to the total array length of 12.5 mm that results in 1.25 mm beam diameter at 29 mm
away from the center of the array. Note that 29 mm corresponds to the distance between the transducer
array and the bolted joint. This 1.25 mm beam diameter provides sufficient resolution for imaging the
bolted joint.
Figure 5. (a) Illustration of the synthetic aperture imaging and (b) the experimental
implementation.
(a) (b)
We investigated the proposed methodology with several bolt tension values for comparison. During
the experiments, the torque applied to the bolt was precisely controlled as it is indirectly related to the
bolt tension. Torque and tension are related by the bolt diameter and a friction factor which is a
function of screw type and the materials used [5]. However, it is worth noting that for a particular bolt
installed in a specific plate, the torque and tension are directly proportional. Also as RAC is directly
proportional to tension, torque and RAC values are also proportional to each other:

 and    then    (4)
where T is the torque applied to the bolt, FT is the bolt tension. The theoretical value of maximum
torque that can be applied to the bolt is given in the Equation (5) below [5,30]:
   

 (5)
where K is the friction factor, At is the tensile stress area, Sy is the bolt yield strength, d is the bolt
nominal diameter and T is the torque applied. For the 6.3 mm (¼ in) grade 8 bolt used in experiments,
the values of the parameters K and Sy are 0.18 and 691 N/mm2 with the tensile stress area and nominal
diameter being 20.5 mm2 and 6.3 mm respectively [29]. By substituting these values into Equation (5),
the maximum allowable torque for the bolt can be obtained as 13.78 Nm. In this study, five
torque values equally spaced from 0 Nm to 13.78 Nm are selected for conducting the experiments. A
torque wrench with 29 Nm of maximum capacity and 0.11 Nm of resolution was used to apply the
required torque.

S
4
t
h
o
a
i
m
t
r
i
n
w
d
r
p
i
n
o
c
c
r
S
ensors 201
4
. Results
a
The resu
l
h
e experi
m
o
btained re
c
a
nd the vert
m
age was
0
r
ansducer
a
n
the shape
w
hen differ
e
d
isplaceme
n
r
eference c
o
p
ointed out
n
Figure 6
)
o
bserved.
A
c
ontact ho
w
c
an also ob
s
r
esponse is
g
Figur
e
from
a
for fi
v
2, 12
a
nd Discus
s
l
ts were o
b
m
ental set-
u
c
onstructed
ical extens
i
0
.92 cm to
a
rray is loc
a
of reconst
r
e
nt torque
v
n
t of the fi
r
o
ndition of
n
that even
w
)
, a noticea
b
A
s the appli
e
w
ever there
w
s
erve that
t
g
etting stro
n
e
6. Image
s
a
6.3 mm (
¼
v
e different
t
s
ion
b
tained for
f
u
p detailed
images wi
t
i
ons of eac
h
3.29 cm a
n
a
te
d
at the
o
r
ucted ima
g
v
alues are
a
r
st reflectio
n
n
o torque,
a
w
ith a very
b
le move
m
e
d torque is
w
as a simil
t
he brightn
e
n
ger.
s
reconstru
c
¼
in) bolt
w
t
orque val
u
f
ive differe
n
in the pr
e
t
h these di
f
h
image are
n
d the verti
c
o
rigin. As
c
g
es, especia
l
a
pplied. It
w
n
boundar
y
a
clear refl
e
loose tight
e
m
ent of the
r
increased,
ar trend of
e
ss of the i
m
c
ted with
d
w
ith a stai
n
u
es are illus
t
n
t torque l
e
e
vious sec
t
f
ferent torq
u
2.37 × 2.5
4
c
al extensi
o
c
an be obse
r
l
ly the loca
t
w
as observe
d
y
towards t
h
e
ction from
e
ning of th
e
r
eflective b
images be
c
movement
m
ages is i
n
d
ynamic ra
n
n
less steel
w
t
rated.
e
vels (with
t
ion. Figur
e
u
e levels. I
n
4
cm, resp
e
o
n was fro
m
r
ved from t
h
t
ion of the
f
d
that incre
h
e transdu
c
the plate h
o
e
bolt (corr
e
oundary to
w
c
ame noisie
of reflecti
v
n
creasing a
s
n
ge of 15
d
w
asher. Re
c
3.61 Nm t
o
e
6 illustr
a
n
this figur
e
e
ctively. Th
m
−1.27 c
m
h
is figure,
t
f
irst reflect
i
asing the a
p
c
er array. F
u
o
le was ob
s
e
sponding t
w
ards the
l
r due to in
c
v
e boundar
y
s
the torqu
e
d
B with th
e
c
onstructed
o
rque incre
m
a
tes the e
x
e
, the horiz
e range ext
e
m
to +1.27
c
t
here was a
i
on from th
e
p
plied torq
u
u
rthermore
,
s
erved. It s
h
o torque v
a
l
eft side of
c
rease in th
e
y
towards t
h
e
is decrea
s
e
SAW refl
ultrasonic
122
7
m
ents) usi
n
x
perimental
l
o
ntal (rang
e
e
nsion of t
h
c
m where t
h
clear chan
g
e
bolted joi
n
u
e resulted
i
,
even in t
h
h
ould also
b
a
lue 3.61 N
m
the image
e
real area
o
h
e array. O
n
s
ed since t
h
ections
images
7
3
n
g
l
y
e
)
h
e
h
e
g
e
n
t
i
n
h
e
b
e
m
is
o
f
n
e
h
e

Sensors 2012, 12 12274
Figure 7. (a) Averaged 1-D images of the first reflection boundary with a dynamic range
of 6 dB; (b) Experimentally obtained position of the first reflection boundary as a function
of the torque applied.
(a)
(b)
To quantify the displacement of the first reflection boundary from the joint center towards the
transducer array as a function of applied torque, the relative average signal intensity of the
reconstructed images as a function of distance from the transducer array is plotted in Figure 7(a).
The 1-D average plot, which is a simplification of the 2-D reconstructed image (Figure 6), is obtained
by taking the average of lateral dimension of the 2-D image. The 1-D plot allows establishing the
exact position of the maximum intensity point of the 2-D image. The position of this point is used to
2.64 2.66 2.70 2.72 2. 74 2.78 2.8 2.82
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
Distance (cm)
Relative intensity (dB)
0 Nm
3.61 Nm
7 Nm
10.39 Nm
13.78 Nm
0 2 3.61 6 7 10.39 13.78
2.68
2.7
2.72
2.74
2.76
2.78
2.8
2.82
Torque (Nm)
Distance (cm)

Sensors 2012, 12 12275
estimate the actual distance change of the reflective boundary. Figure 7(b) illustrates the position of the
first reflection boundary as a function of the applied torque. Firstly, one can clearly observe from
the maximum intensities of the reconstructed images that there is a clear trend of moving the
reflection boundary towards the transducer array as the torque is increased. It can also be observed
from Figure 7(b) that the distance between the position of first reflection boundary with no torque
applied and maximum allowable torque of 13.78 Nm is 0.1 cm. The first reflection boundary
corresponding to rest of the torque values are positioned between these two extreme values. Note that
the first reflection boundary position change does not have a linear trend with increasing torque. This
expected behavior can be explained by the fact that with this phased array image reconstruction
technique, we are quantifying the location of the first reflection boundary, not the real area of contact
(RAC). As a result, the boundary distance and the torque should not be expected to have a linear
relationship. It should be pointed out that this non-linear variation of the movement of the first
reflection of boundary as a function of applied torque can easily be used to estimate the torque (hence
the tightening of the bolt) by just obtaining the distance of the first reflection boundary from the
transducer array. Another important characteristic of the Figure 7(b) is the fact that RAC saturation
takes place around 10 Nm corresponding to first reflection boundary position of 2.7 cm.
4.1. Signal to Noise Ratio
Synthetic phased array modality was used in this study as opposed to phased array modality for
significantly simplified beamformer design. However, one drawback of synthetic phased array
modality is its lower signal to noise ratio (SNR). In order to ensure that SNR associated with the
generated images is sufficient, we conducted a detailed study measuring the SNR of all the received
signals (A-scans). In Figure 8, the SNRs of the signals reflected by the bolt are presented. This figure
illustrates the SNR levels for the 50 signals corresponding to each 50 positions of the transducer.
As expected, higher SNR values were obtained closer to the center position due to the reflections from
the washer (or the hole). It was also observed that average SNR decreases with the increasing torque
level. As expected, lower SNR values resulted in lower quality images for higher torque levels
compared to lower torque values. Lower image quality may potentially lead to less accurate estimation
and higher errors.
The minimum dynamic range that can be applied to the images was determined by the averaged
SNR. The reflective boundary position was calculated using the 1-D averaged plot based on the
location of the signal peaks. Consequently, the peak SNR value of Figure 8 at each torque level
corresponds to the peaks in Figure 7(a). The minimum peak SNR of the system was around 17 dB at
13.78 Nm, as shown in Figure 8. The averaged signal to noise ratio for each torque value is depicted in
Table 1 for reference. The decrease of SNR can clearly be seen as the applied torque is increased. The
average SNR value of the system is around 15 dB. This SNR level was good enough to differentiate
the signal from the ambient noise. The reduction of the SNR of the system around the bolt may be
attributed to some attenuation due to the bolt interference. This analysis allows identification of
sources of errors for further reducing the SNR of the system.

Sensors 2012, 12 12276
Figure 8. Signal to noise ratio of the bolt reflections based on original signals.
Table 1. Experimentally obtained signal to noise ratio (SNR) for different torque values
applied to the bolted joint.
Torque Level (Nm) 0 3.61 7 10.39 13.78
Average SNR (dB) 18 18.31 15.75 12 12.05
5. Conclusions
This study presents the first step towards the development of a bolt tension sensor based on surface
acoustic waves. The bolt tension was estimated using the reflection of SAWs created by the bolted
joint interference. Our results indicate that this method is not only capable of clearly distinguishing
properly bolted joints from loosened joints but also capable of quantifying how loose the bolt is.
Increase in the bolt tension resulted in an increase in the number of points of that are in-contact at the
bolt-plate interface (real area of contact). This increase in the real area of contact caused the first
reflection boundary position change measured from the bolted joint by using surface acoustic waves.
The boundary location change was quantified by using a 1-D and 2-D ultrasonic images obtained by
linear synthetic array image processing technique.
The 2-D ultrasound imaging allowed clear and detailed visualization of the position of the reflective
boundary when applied torque to the bolt was changed. For the entire range of torque values from zero
to maximum allowable, as hypothesized, increase in the torque value increased the real area of contact
and therefore moved the position of the reflective boundary towards the ultrasonic array. The boundary
position versus the torque applied to the bolt was also investigated. There were two distinct regions;
the first region presented a non-linear behavior, while the second one was identified as a saturation
region. The saturation region (presented at the highest torque values) was characterized by negligible
change in the boundary position due to further increase in the torque applied. The signal-to-noise
(SNR) analysis was also performed and the results indicate that the intensity of the signal was reduced
as the torque was increased. However, it should be noted that the SNR value even for the maximum
Element Number of the Synthetic Array

Sensors 2012, 12 12277
allowable torque value was sufficient for image reconstruction. The bolt tension sensor investigated
here has the potential to be used as a component in a non-destructive structural health monitoring
(SHM) system for local monitoring of civil infrastructures.
The proposed method has a significant potential to be used in real-life applications. However
there maybe limitations that needs to be addressed. Current method utilizes an off-the-shelf wedge
(8 cm × 3 cm × 5 cm in size) coupled to surface with an acoustic couplant. Although this fairly small
size is suitable for majority of applications, for those applications where the structure does not allow
direct access, sensor arrays can be directly integrated to structures. Wireless interrogation can then be
employed for data acquisition. Surface condition is also an important factor that may affect the
operation. Surface conditions at the plate-bolt and plate-wedge interface (such as contamination,
surface roughness) and along the propagation path (such as liquid loading) may possibly cause minor
problems such as wave attenuation, undesired scattering. However, it should be noted that these issues
can be overcome by using proper commercially available surface treatments and ultrasonic couplants.
References
1. Hess, D.P. Vibration and shock-induced loosening. In Handbook of Bolts and Bolted Joint;
Bickford, J., Ed.; ASM International: New York, NY, USA, 1988.
2. Bickford, J.H. Introduction to the Design and Behavior of Bolted Joints, 4th ed.; CRC Press:
Boca Raton, FL, USA, 2008.
3. Goodier, J.N. Loosening by vibration of threaded fastenings. Mech. Eng. 1945, 67, 798–802.
4. Sauer, J.A.; Lemmon, D.C.; Lynn, E.K. Bolts: How to prevent their loosening’s. Mach. Des.
1950, 22, 133–139.
5. Budynas, R.G.; Nisbett, K. Screw, fasteners, and the design of nonpermanent joints. In Shigley’s
Mechanical Engineering Design, 8th ed.; McGraw-Hill Companies, Inc.: Burlington, MA, USA,
2006; pp. 396–425.
6. Pai, N.G.; Hess, D.P. Experimental study of loosening of threaded fasteners due to dynamic shear
loads. J. Sound Vibrat. 2002, 253, 585–602.
7. Pai, N.G.; Hess, D.P. Influence of fastener placement on vibration-induced loosening. J. Sound
Vibrat. 2003, 268, 617–626.
8. Pedinoff, M.E.; Waldner, M.; Jones, W.R. Refraction and reflection of surface acoustic waves at
boundaries of layered anisotropic substrates: Gold on lithium niobate. J. Appl. Phys. 1971, 42,
3025–3034.
9. Huang, Y.H.; Liu, L.; Yeung, T.W.; Hung, Y.Y. Real-time monitoring of clamping force of a
bolted joint by use of automatic digital image correlation. Opt. Laser Tech. 2009, 41, 408–414.
10. Tanner, N.A.; Wait, J.R.; Farrar, C.R.; Sohn, H. Structural health monitoring using modular
wireless sensors. J. Intell. Mater. Syst. Struct. 2003, 14, 43–56.
11. Milanese, A.; Marzocca, P.; Nichols, J.M.; Seaver, M.; Trickey, S.T. Modeling and detection of
joint loosening using output-only broad-band vibration data. Struct. Health Monit. 2008, 7, 309–328.
12. Esmaeel, R.A.; Briand, J.; Taheri, F. Computational simulation and experimental verification
of a new vibration-based structural health monitoring approach using piezoelectric sensors.
Struct. Health Monit. 2011, 11, 237–250.

Sensors 2012, 12 12278
13. Amerini, F.; Meo, M. Structural health monitoring of bolted joints using linear and nonlinear
acoustic/ultrasound methods. Struct. Health Monit. 2011, 10, 659–672.
14. Kim, N.; Hong, M. Measurement of axial stress using mode-converted ultrasound. NDT E Int.
2009, 42, 164–169.
15. Mascarenas, D.L.; Park, G.; Farinholt, K.M.; Todd, M.D.; Farrar, C.R. A low-power wireless
sensing device for remote inspection of bolted joints. J. Aerosp. Eng. 2009, 223, 565–575.
16. Ballantine, D.S. Acoustic Wave Sensors: Theory, Design, and Physico-Chemical Applications;
Academic Press: San Diego, CA, USA, 1997.
17. Kendall, K.; Tabor, D. An ultrasonic study of the area of contact between stationary and sliding
surfaces. Proc. Roy. Soc. Lond. A Math. Phys. Eng. Sci. 1971, 323, 321–340.
18. Pau, M. Estimation of real contact area in a wheel-rail system by means of ultrasonic waves.
Tribol. Int. 2003, 36, 687–690.
19. Królikowski, J.; Szczepek, J. Prediction of contact parameters using ultrasonic method. Wear
1991, 148, 181–195.
20. Drinkwater, B.W.; Dwyer-Joyce, R.S.; Cawley, P. A study of the interaction between ultrasound
and a partially contacting solid-solid interface. Proc. Roy. Soc. Lond. A Math. Phys. Eng. Sci.
1996, 452, 2613–2628.
21. Królikowski, J.; Szczepek, J. Phase shift of the reflection coefficient of ultrasonic waves in the
study of the contact interface. Wear 1992, 157, 51–64.
22. Takeuchi, A. Establishment of lubrication diagnostic technique with ultrasonic method. J. Jpn.
Soc. Tribol. 2004, 49, 422–427.
23. Johnson, K.L. Contact Mechanics; Cambridge University Press: Cambridge, UK, 1985.
24. Czichos, H. Tribology: A Systems Approach to the Science and Technology of Friction,
Lubrication and Wear; Elsevier Scientific Publishing Company: New York, NY, USA, 1978.
25. Fischer-Cripps, A.C. Surfaces Forces, Adhesion and Friction; Springer-Verlag Inc.: New York,
NY, USA, 2000.
26. Auld, B.A. Acoustic Fields and Waves in Solids, 2nd ed.; Krieger Publishing Company: Malabar,
FL, USA, 1990.
27. Thomenius, K.E. Evolution of Ultrasound Beamformers. In Proceedings of IEEE Ultrasonics
Symposium, San Antonio, TX, USA, 3–6 November 1996; pp. 1615–1622.
28. Karaman, M.; Atalar, A.; Koymen, H. Optimization of dynamic receive focusing in ultrasound
imaging. Acoust. Imag. 1992, 19, 225–229.
29. Collins, J.A.; Staab, G.H.; Busby, H.R. Mechanical Design of Machine Elements and Machines;
John Wiley & Sons: Hoboken, NJ, USA, 2002.
30. Szabo, T.L. Diagnostic Ultrasound Imaging: Inside Out; Elsevier Academic Press: Burlington,
MA, USA, 2004.
31. Hoskins, P.R.; Martin, K.; Thrush, A. Diagnostic Ultrasound: Physics and Equipment, 2nd ed.;
Cambridge University Press: Cambridge, UK, 2010.
© 2012 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 license
(http://creativecommons.org/licenses/by/3.0/).