sensors

Article
Monitoring Bridge Dynamic Responses Using Fiber
Bragg Grating Tiltmeters
Feng Xiao 1 ID
, Gang S. Chen 2, * and J. Leroy Hulsey 1
1 Department of Civil and Environmental Engineering, University of Alaska Fairbanks,
Fairbanks, AK 99775, USA; xfeng2@alaska.edu (F.X.); jlhulsey@alaska.edu (J.L.H.)
2 College of Information Technology and Engineering, Marshall University, Huntington, WV 25755, USA
* Correspondence: chenga@marshall.edu

Received: 24 September 2017; Accepted: 18 October 2017; Published: 20 October 2017

Abstract: In bridge health monitoring, tiltmeters have been used for measuring rotation and curvature;
however, their application in dynamic parameter identification has been lacking. This study installed
fiber Bragg grating (FBG) tiltmeters on the bearings of a bridge and monitored the dynamic rotational
angle. The dynamic features, including natural frequencies and mode shapes, have been identified
successfully. The innovation presented in this paper is the first-time use of FBG tiltmeter readings to
identify the natural frequencies of a long-span steel girder bridge. The identified results have been
verified using a bridge finite element model. This paper introduces a new method for the dynamic
monitoring of a bridge using FBG tiltmeters. Limitations and future research directions are also
discussed in the conclusion.

Keywords: fiber Bragg grating (FBG) tiltmeter; vibration monitoring; bridge health monitoring;
dynamic identification; expansion bearing

1. Introduction
Dynamic feature identification is an important step for the bridge health monitoring. Dynamic
data, including natural frequencies, mode shapes, and damping factors, can provide meaningful results
if they are used to improve a bridge’s finite element model or conduct damage identification [1–4].
Traditionally, measurements of the dynamic parameters of structures are based on accelerometers.
Sensors, such as the fiber optic accelerometer [5,6], force balanced accelerometer [7,8], and wireless
accelerometer [9,10], are typically employed, as they can obtain acceleration directly.
With the development of sensor and measurement technology, other types of sensor, such as strain
sensors, GPS, version-based measurement systems, and tiltmeters, can also be used to identify the
dynamic features. Fiber optic strain sensors have been used to estimate structural vibration features
and have been tested for a beam [11] and model bridge [12] in the laboratory. Moschas and Stiros used
GPS to measure dynamic displacements of a bridge, and the modal frequencies were identified [13].
Lee et al. used a version-based dynamic rotational angle measurement system to monitor the dynamic
rotational angle of a five-story modal tower [14]. Fukuda et al. developed a vision-based displacement
sensor for monitoring the dynamic response [15]. Erol et al. used a tiltmeter to determine the structural
dynamic movements of a tall building, and identified the structure’s resonance [16]. This paper
includes a bridge vibration study based on the measurement of fiber Bragg grating (FBG) tiltmeters.
Tiltmeters are a measurement tool for monitoring the health of a bridge. Currently, commercial
tiltmeters can provide resolution up to 0.001◦ [14,17,18]. To monitor the bridge’s overall response,
it has been applied for different type of bridges. Zhou et al. installed tiltmeters on the top of a pylon
for a cable-stayed bridge to record the rotations of the pylon [19]. Caglayan et al. measured the slope
of a bridge deck to identify the vertical displacement of an arch bridge [8]. Helmi et al. measured the

Sensors 2017, 17, 2390; doi:10.3390/s17102390 www.mdpi.com/journal/sensors

Sensors 2017, 17, 2390 2 of 11

rotations and curvature of a box girder bridge to compute the deflection [20]. Norouz et al. installed
tiltmeters on the piers of a truss bridge to measure the top of the piers’ displacement [21]. Kim and
LamanSensors 2017, 17, a
placed 2390
tiltmeter on bridge abutments and monitored the abutment movements 2 of 11 [22].

In conclusion, these measurements were all performed for the purpose of bridge deflection
rotations and curvature of a box girder bridge to compute the deflection [20].
identification. However, monitoring the bridge’s dynamic features based on a tiltmeter’s measurements
Norouz et al. installed tiltmeters on the piers of a truss bridge to measure the top of the piers’
has been lacking.[21]. Kim and Laman placed a tiltmeter on bridge abutments and monitored the
displacement
In addition
abutment movements to monitoring
[22]. Inthe bridge’sthese
conclusion, longitudinal movement
measurements inperformed
were all response to
forathe
traffic loading
purpose of and
the environment
bridge deflection effects [23], this paper
identification. studied
However, the dynamic
monitoring rotational
the bridge’s signal
dynamic from the
features basedtiltmeter
on a and
identified the bridge’s
tiltmeter’s measurements dynamic features.
has been The innovation is the first-time use of FBG tiltmeter readings
lacking.
to identifyIn the
addition
naturalto monitoring
frequencies the
ofbridge’s longitudinal
a long-span movement
steel girder bridge.in response to a traffic loading
and the environment effects [23], this paper studied the dynamic rotational signal from the tiltmeter
2. FBG
and Tiltmeter
identified Technology
the bridge’s dynamic features. The innovation is the first-time use of FBG tiltmeter
readings to identify the natural frequencies of a long-span steel girder bridge.
A tiltmeter can not only record the rotation or deflection but can also provide the
dynamic
2. FBGfeatures
Tiltmeterbased on signal processing technology [16]. In this study, two FBG-TI-310
Technology
(FBG Korea, Inc., Gwangju, Korea) tiltmeters [24] were employed. The fiber optic sensor is a sensing
A tiltmeter can not only record the rotation or deflection but can also provide the dynamic
solution using
features theonfiber
based Bragg
signal grating
processing and measurement
technology [16]. In this system.
study, twoThe theories and
FBG-TI-310 (FBGcharacteristics
Korea, Inc., of
this Gwangju,
fiber opticKorea)
technology were
tiltmeters [24]presented in previous
were employed. The fiberresearch [25,26].
optic sensor is a sensing solution using the
The Bragg
fiber FBG sensor
gratingwith
and very high accuracy
measurement system. can
Theprovide
theoriesnon-stop and long-term
and characteristics angle
of this fibermonitoring.
optic
The technology
tiltmeter haswere a presented
measurement rangeresearch
in previous [25,26].and has a resolution of ±0.05% FS. A built-in
of 8 degrees
temperature The FBG sensor with
compensator very high
is used accuracy
to adjust caneffect
for the provide non-stop and on
of temperature long-term angle monitoring.
the measurement, besides the
The tiltmeter has a measurement range of 8 degrees and has a resolution
rotational angle measurement, the tiltmeter includes a second FBG which provides active of ±0.05% FS. A built-in
temperature
temperature[24].
compensation compensator is used to adjust for the effect of temperature on the measurement, besides
the rotational angle measurement, the tiltmeter includes a second FBG which provides active
The sampling rate of the FBG tiltmeter is determined by the optical sensing interrogator.
temperature compensation [24].
The optical sensing interrogator sends wide-spectrum light to optical sensors and indicates the change
The sampling rate of the FBG tiltmeter is determined by the optical sensing interrogator. The
in the reflected
optical wavelength
sensing interrogator from the
sends optical sensors.
wide-spectrum lightThe interrogator
to optical selected
sensors and for this
indicates study in
the change was the
Optical
the reflected wavelength from the optical sensors. The interrogator selected for this study was the1k Hz,
Sensing Interrogator sm130 (Micron Optics, Atlanta, GA, USA), the scan frequency was
and Optical
the sampling
Sensingrate for the FBG
Interrogator tiltmeter
sm130 was
(Micron 250 Hz.
Optics, Atlanta, GA, USA), the scan frequency was
1k Hz, and the sampling rate for the FBG tiltmeter was 250 Hz.
3. Bridge Description and Monitoring System
3. Bridge Description and Monitoring System
The 790-foot Chulitna River Bridge (Figure 1a) is a five-span steel girder plate bridge located in
The is
Alaska and 790-foot
part of Chulitna River
the Parks Bridge (Figure
Highway, which1a)links
is a five-span steelwith
Anchorage girderFairbanks.
plate bridgeHeavily
located in
loaded
Alaska and is part of the Parks Highway, which links Anchorage with Fairbanks. Heavily
vehicles of up to 410,000 pounds regularly travel the route. The bridge deck is 42 feet 2 inches wide loaded
withvehicles of up to 410,000 pounds regularly travel the route. The bridge deck is 42 feet 2 inches wide
two exterior steel plate girders and three sub-stringers. Three longitudinal steel trusses were
with two exterior steel plate girders and three sub-stringers. Three longitudinal steel trusses were
installed utilizing the stringers as top chords (Figure 1b). Figure 1c,d shows the bearing condition at
installed utilizing the stringers as top chords (Figure 1b). Figure 1c,d shows the bearing condition at
the bridge’s end.
the bridge’s end.

(a) (b)

Figure 1. Cont.
Sensors 2017, 17, 2390 3 of 11
Sensors 2017, 17, 2390 3 of 11

Sensors 2017, 17, 2390 3 of 11

(c) (d)
Figure 1. Photographs: (a) Chulitna River Bridge; (b) Bridge bottom view; (c) Expansion bearing at
Figure
the 1. Photographs:
bridge (a) Chulitna
end; (d) Expansion
(c) River Bridge; (b) Bridge bottom view; (c)(d)
bearing. Expansion bearing at the
bridge end; (d) Expansion bearing.
Figure 1. Photographs:
The structural (a) Chulitna system
health monitoring River Bridge; (b) Bridge
consisted bottomthe
of five parts: view; (c) Expansion
sensors, bearing the
the multiplexer, at
the bridgethe
interrogator, end;local
(d) Expansion
computer,bearing.
and the remote computer (Figure 2). The interrogator sends optical
The structural health monitoring system consisted of five parts: the sensors, the multiplexer,
signals (laser) to the multiplexer, and the multiplexer switches this laser to several lasers, and each
the interrogator,
laser The
goes intothe
structural onelocal
health computer,
optical monitoring andsystem
sensor array. the remote
There is an computer
consisted
optic of five
Bragg (Figure
parts: the
grating 2).
in The interrogator
sensors,
each thesensor
optical sends
whichoptical
multiplexer, the
signals
can (laser)
reflect ato
interrogator, the the
certain multiplexer,
local computer,of
wavelength and
andlightthe multiplexer
theback
remote
to the computerswitches
interrogator (Figure
andthis2).laser
allow Thethetoother
several
interrogator lasers,
sends and
wavelengths of each
optical
laser goesto(laser)
signals
light into
passone throughoptical
to the the sensor
sensor, array.
multiplexer, and the
which There
can is an for
multiplexer
be used optic Bragg
switches
the grating
this
next optical laser intoeach
sensor. The optical
several lasers,sensor
interrogator and which
caneach
canlaser
reflect
identify a
goesthe certain
into wavelength
one optical
reflected laser andsensorof
read light
array. back to the
Theresignal.
the optical interrogator
is an optic and
Bragg grating
The interrogator allow
willinstorethe
eachtheother
optical wavelengths
data insensor which of
the local
light computer.
can toreflect aThe
pass through localthe
certain computer
sensor,sends
wavelength thecan
of light
which real-time
backbetoused data
the tothe
the next
interrogator
for remote and computer
allow
optical through
the
sensor. other the internet. of
Thewavelengths
interrogator can
Figure
light
identify to
the 2 reflected
shows
pass throughtiltmeters
the and
laser 1 and
sensor, 2which
read installed caninbe
the optical one sensor
used
signal. for
The array.
the Figure
next 3 iswill
optical
interrogator a photograph
sensor.
storeThe of theinFBG
interrogator
the data thecan
local
tiltmeter
identify
computer. at reflected
the
The the
local bridge site.andsends
laser
computer read the theoptical
real-timesignal.
dataThe interrogator
to the will store the
remote computer data in
through theinternet.
the local
computer. The tiltmeters
The local were installed
computer sends onthe the expansion
real-time data bearings
to the (Figure
remote 3). Figure
computer 4a shows
through the the
internet.
Figure 2 shows tiltmeters 1 and 2 installed in one sensor array. Figure 3 is a photograph of the FBG
expansion
Figure 2 shows bearing location
tiltmeters at the abutment 2. Figure 4b shows the details at the abutment 2, and
tiltmeter at the bridge site. 1 and 2 installed in one sensor array. Figure 3 is a photograph of the FBG
tiltmeter
tiltmeter 1 and tiltmeter
at the bridge site. connected
2 are to the same fiber optical cable and that the cable is connected
The the
with tiltmeters
control were panel installed
on the bridgeon the site.expansion
The control bearings (Figure 3). Figure
panel communicated with the 4a interrogator
shows the expansion
and
bearing The tiltmeters
location at the were
abutment installed
2. on the
Figure 4b expansion
shows the bearings
details at (Figure
the 3). Figure
abutment 2, and4atiltmeter
shows 1the and
the local computer through the optical signal [23].
expansion bearing location at the abutment 2. Figure 4b shows the details at the abutment 2, and
tiltmeter The 2 areexpansion
connected to theissame
bearing a pinfiber
type optical
expansion cable and that
bearing the 4c),
(Figure cable is connected
which consisted with theatcontrol
of a pin
tiltmeter 1 and tiltmeter 2 are connected to the same fiber optical cable and that the cable is connected
panel the ontop
the that
bridge site. Therotation.
facilitates control panel communicated
The rotation with thecan
of the bearings interrogator
accommodate and the thelocal computer
bridge’s
with the control panel on the bridge site. The control panel communicated with the interrogator and
longitudinal
through the optical movement.
signal [23].Figure 4c shows the tiltmeter installed on the rocker assembly. The purpose
the local
isThe
to computer
monitor through the
the rotation optical signal [23].
expansion bearing of is the rocker
a pin typeassembly
expansion in response to the bridge’s
bearing (Figure longitudinal
4c), which consisted movement.
of a pin at the
The expansion bearing is a pin type expansion bearing (Figure 4c), which consisted of a pin at
top that facilitates rotation. The rotation of the bearings can accommodate the bridge’s longitudinal
the top that facilitates rotation. The rotation of the bearings can accommodate the bridge’s
movement.
longitudinal Figure 4c shows
movement. the tiltmeter
Figure 4c showsinstalled
the tiltmeteron the rockeron
installed assembly.
the rocker The purposeThe
assembly. is to monitor
purpose
theisrotation
to monitor of thetherocker
rotation assembly in response
of the rocker assembly toin the bridge’stolongitudinal
response movement. movement.
the bridge’s longitudinal

Figure 2. System configuration.

Figure2.2.System
Figure System configuration.
configuration.
Sensors 2017, 17, 2390 4 of 11
Sensors 2017, 17, 2390 4 of 11
Sensors 2017, 17, 2390 4 of 11

Figure
Figure 3.
3. FBG
FBG tiltmeter.
tiltmeter.
Figure 3. FBG tiltmeter.

(a)
(a)

(b)
(b)

(c)
(c)
Figure 4.
4. Tiltmeter
Figure 4. installation
installation location
Tiltmeter installation location (a)
(a) Bridge
Bridge plan
plan view;
view; (b)
(b) Abutment
Abutment 222 view;
view; (c) Tiltmeter
(c) Tiltmeter on
Tiltmeter on
on
Figure
expansion Tiltmeter
bearing. location (a) Bridge plan view; (b) Abutment view; (c)
expansion bearing.
expansion bearing.
4.
4. Dynamic
Dynamic TestTest and
and Dynamic
Dynamic Properties
Properties Identification
Identification
4. Dynamic Test and Dynamic Properties Identification
AA dynamic
dynamic loadload test
test was
was conducted
conducted on on thethe bridge
bridge using
using twotwo dump
dump trucks.
trucks. The
The gross
gross weight
weight of of
thoseA dynamic
two trucks load
are test was
weighted conducted
before the on
load the bridge
test whichusing
are two
80.3 dump
kips trucks.
and 82.1 The
kips. gross
The weight
two of
trucks
those two trucks are weighted before the load test which are 80.3 kips and 82.1 kips. The two trucks
those
were two trucks are weighted before the load 45 test which are 80.3 kips6 and 82.1 kips. The two trucks
were side
side by
by side,
side, heading
heading north
north at
at aa speed
speed of of 45 mph.
mph. Figures
Figures 55 and and 6 show
show the
the tiltmeter
tiltmeter reading
reading in in
were side
response by
to side,
the heading
truck north
loading. Theat a speed
readings of
can45 mph.
be Figures
separated 5
intoand 6
threeshow the
phases: tiltmeter
trucks reading in
approaching
response to the truck loading. The readings can be separated into three phases: trucks approaching
response
the to the truck loading. The readings can be separated into three In phases: trucks approaching the
the bridge,
bridge, trucks
trucks on on the
the bridge
bridge and
and trucks
trucks having
having left
left the
the bridge.
bridge. In Figures
Figures 55 and
and 6,
6, the
the trucks
trucks
arrived
arrived at
at the
the abutment
abutment 22 around
around at
at the
the 17
17 s,
s, and
and the
the bridge
bridge is
is in
in aa free-decay
free-decay response
response after
after the
the trucks
trucks
Sensors 2017, 17, 2390 5 of 11

Sensors 2017, 17, 2390 5 of 11

Sensors 2017, 17, 2390 5 of 11
bridge, trucks on the bridge and trucks having left the bridge. In Figures 5 and 6, the trucks arrived at
left the bridge,
the and this study usedthe
this free-decay response to identify the bridge’s natural
leftabutment 2 around
the bridge, at the
and this 17 s, used
study and thisbridge is in a free-decay
free-decay response toresponse
identifyafter
the the trucksnatural
bridge’s left the
frequencies.
bridge, and this study used this free-decay response to identify the bridge’s natural frequencies.
frequencies.

Figure5.5.Tiltmeter
Figure Tiltmeter 1 reading.
Figure 5. Tiltmeter 11 reading.
reading.

Figure 6.6.Tiltmeter
Figure6.
Figure Tiltmeter22 2reading.
Tiltmeter reading.
reading.

4.1.Spectrum
4.1. Spectrum Analysis
SpectrumAnalysis
Analysis
Figure
Figure777isis
Figure the
isthe vibration
thevibration signal
vibrationsignal after
signalafter the
afterthe trucks
trucksleft
thetrucks leftthe
thebridge.
bridge.TheThedynamic
The dynamiccharacteristics
dynamic characteristics
characteristicsof the
ofof
the
the
structure
structurewere
structure were
were extracted
extracted
extractedbyby
converting
by convertingthethe
converting time domain
thetime
time datadata
domain
domain to frequency
data domain
totofrequency
frequency data. data.
domain
domain Figure 8Figure
shows
Figure
data. 8 8
shows
the
shows the
natural natural
natural frequencies
the frequencies obtainedobtained
frequencies using the
obtained usingfast the
using thefast
fastFourier
Fourier Fouriertransform
transform (FFT)
(FFT)method.
(FFT) method.
transform Both Both
sensors
method. sensors
Both yielded
sensors
yielded
1.50 1.50
Hz, 4.47
yielded Hz,
1.50Hz, 4.47
Hz, and Hz,
4.474.90 and
Hz, Hz 4.90
andas Hz
theHz
4.90 as the
primary primary
as the natural natural
primaryfrequencies. frequencies. The identified
The identified
natural frequencies. frequencies
frequencies
The identified can be
frequencies
can
can be
used as used
be aused as
as aa characteristic
characteristic quantityquantity
characteristic or indexor
quantity for
or index for
damage
index fordamage identification.
identification.
damage identification.
Signal Corrupted
Signal Corrupted with Zero-Mean
Zero-Mean Random
Random Noise
Noise Signal Corrupted
Corrupted with
with Zero-Mean
Zero-Mean Random
Random Noise
Noise
0.15 Signal Corruptedwith
with Zero-Mean Random Noise 0.15
Signal
Signal Corrupted with Zero-Mean Random Noise
0.15
0.15 0.15
0.15

0.1
0.1 0.1
0.1
0.1 0.1

0.05
0.05 0.05
0.05
0.05 0.05

00 00
0 0
X(t)

X(t)
X(t)

X(t)
X(t)

X(t)

-0.05
-0.05 -0.05
-0.05
-0.05 -0.05

-0.1
-0.1 -0.1
-0.1
-0.1 -0.1

-0.15
-0.15 -0.15
-0.15
-0.15 -0.15

-0.2
-0.2 -0.2
-0.2
-0.200 22 44 66 88 10
10 12
12 14
14 16
16 18
18 00
-0.2 22 44 66 88 10
10 12
12 14
14 16
16 18
18
0 2 4 6 8
tt (seconds)
(seconds)10 12 14 16 18 0 2 4 6 8
tt (seconds)
(seconds) 10 12 14 16 18
t (seconds) t (seconds)

(a)
(a) (b)
(b)
Figure 7. Tiltmeter reading for trucks left bridge. (a) Tiltmeter 1 signal; (b) Tiltmeter 2 signal.
Figure7.7.Tiltmeter
Figure Tiltmeterreading
readingfor
fortrucks
trucksleft
leftbridge.
bridge.(a)
(a)Tiltmeter
Tiltmeter1 1signal;
signal;(b)
(b)Tiltmeter
Tiltmeter2 2signal.
signal.
Sensors 2017, 17, 2390 6 of 11
Sensors 2017, 17, 2390 6 of 11

Sensors 2017, 17, 2390 Single-Sided Amplitude Spectrum of X(t) Single-Sided Amplitude Spectrum of X(t) 6 of 11
0.012 0.012

Single-Sided Amplitude Spectrum of X(t) Single-Sided Amplitude Spectrum of X(t)

0.01 0.01
0.012 0.012

0.008 0.008
0.01 0.01

|P1(f)|

|P1(f)|
0.006 0.006
0.008 0.008
|P1(f)|

|P1(f)|
0.004 0.004
0.006 0.006

0.002 0.002
0.004 0.004

0 0
0.002 1 2 3 4 5 6 7 0.002 1 2 3 4 5 6 7
f (Hz) f (Hz)

0 0
1 2 3 (a)
4 5 6 7 1 2 3 (b)
4 5 6 7
f (Hz) f (Hz)

Figure 8. Fast Fourier transform

(a) (FFT) of tiltmeter reading. (a) Tiltmeter
(b)1 reading; (b) Tiltmeter 2
Figure 8. Fast Fourier transform (FFT) of tiltmeter reading. (a) Tiltmeter 1 reading; (b) Tiltmeter
reading.
Figure 8. Fast Fourier transform (FFT) of tiltmeter reading. (a) Tiltmeter 1 reading; (b) Tiltmeter 2
2 reading.
reading.
To characterize the bridge vibration in the time-domain, time-frequency analysis of the
Tospectrogram
characterize is used
the [27]. Figure
bridge 9 shows the
vibration in spectrogram of the measured
the time-domain, signal at the
time-frequency
To characterize the bridge vibration in the time-domain, time-frequency analysis of the
tiltmeterof1 the
analysis
and
spectrogram tiltmeter
is used2 during the
[27].[27]. free-decay
Figure period; Both Figure 9a,b shows two dominant components. The
spectrogram is used Figure9 9shows
shows the spectrogramofof
the spectrogram thethe measured
measured signal
signal at theattiltmeter
the tiltmeter
1
first
1 and and component’s
tiltmeter 2 during corresponding frequencies are 4.47 Hz and 4.90 Hz. The second component’s
tiltmeter 2 duringthe thefree-decay period;
free-decay period; Both
Both Figure
Figure 9a,b 9a,b
shows shows two dominant
two dominant components. components.
The
corresponding frequency is 1.50 Hz. From those results, we can see that all those kinds of spectrum
The first
firstcomponent’s
component’s corresponding frequencies
corresponding frequencies areare 4.47
4.47 HzHz
and and 4.90
4.90 Hz.Hz.
TheThe second
second component’s
component’s
have merged components, the next section will decompose the time-frequency signal.
corresponding
corresponding frequency
frequency is 1.50
is 1.50 Hz.Hz. Fromthose
From thoseresults,
results, we
wecan
cansee
seethat allall
that those kinds
those of spectrum
kinds of spectrum
have merged
have merged components,
components, thethe
nextnext sectionwill
section willdecompose
decompose thethetime-frequency
time-frequency signal.
signal.

(a) (b)
Figure 9. (a) (b) 2 reading.
Spectrogram. (a) Tiltmeter 1 reading; (b) Tiltmeter

Figure 9. Spectrogram. (a) Tiltmeter 1 reading; (b) Tiltmeter 2 reading.

4.2. Modified Empirical
Figure 9.Mode Decomposition
Spectrogram. (a) Tiltmeter 1 reading; (b) Tiltmeter 2 reading.
To decompose
4.2. Modified Empiricalthe signal
Mode as being characterized in the time-frequency expression of the signal,
Decomposition
4.2. Modified
we nextEmpirical
used the Mode Decomposition
modified empirical mode decomposition method [28] to decompose the tiltmeter 1
To decompose the signal as being characterized in the time-frequency expression of the signal,
reading (Figure 9a). This method was based on empirical mode decomposition (EMD) [29], which
To
wedecompose
next used the the signal empirical
modified as being mode
characterized
decompositionin themethod
time-frequency expression
[28] to decompose of the signal,
the tiltmeter 1
established a new framework and avoided the EMD drawbacks [28]. The method decomposed the
reading
we next used (Figure 9a). Thisempirical
the modified method was based
mode on empirical mode
decomposition methoddecomposition (EMD) [29],
[28] to decompose thewhich
tiltmeter
original signal into a set of elemental signals called “intrinsic mode functions (IMFs)”.
established
1 reading (Figure a new
9a).framework
Thissignalsand avoided
method the EMD on drawbacks [28]. Thedecomposition
method decomposed the [29],
The decomposed are was based
zero-mean empiricalwhich
components, mode must satisfy the following (EMD)
original signal
which established into a set of elemental signals called “intrinsic mode functions (IMFs)”.
requirements:a (1)
newTheframework and avoided
number of extreme and the the EMD of
number drawbacks [28].inThe
zero crossings the IMF method mustdecomposed
be equal
The decomposed signals are zero-mean components, which must satisfy the following
the original signal
or differ into
at most byaone;
set (2)
of elemental
at any point,signals
the mean called
value“intrinsic mode defined
of the envelopes functions by the (IMFs)”.
local maxima
requirements: (1) The number of extreme and the number of zero crossings in the IMF must be equal
anddecomposed
The local minima must be zero.
signals areThe signal is locally
zero-mean symmetric around
components, whichthemusttime axis,
satisfy and the
theoriginal
following
or differ at most by one; (2) at any point, the mean value of the envelopes defined by the local maxima
signal can(1)
requirements: nowThebe represented
number ofasextreme
the sum and
of n IMFs
the plus a residual
number of zeror (crossings
). in the IMF must be
and local minima must be zero. The signal is locally symmetric around the time axis, and the original
equal signal
or differ
canat
nowmost by one; (2)as
be represented atthe
anysum
point,
of n the
IMFsmean
plus avalue of the
residual r ( envelopes
). defined by the local
maxima and local minima must be zero. The ( )signal
= is(locally
) + r ( symmetric
) around the time axis, (1) and the
original signal can now be represented as the ( )= sum of( )+r ( )
n IMFs plus a residual r n ( t ) . (1)
Next, the Hilbert transform is applied to all IMFs, ( ) , in Equation (1) to derive model
parameters. n
(t) = ∑
Next, the Hilbert transform is xapplied to call IMFs, ( ) , in Equation (1) to derive model (1)
i ( t ) + r(n () t )
parameters. (i=
)1= (2)
( )
( )IMFs,
Next, the Hilbert transform is applied to all = (2)
c j (t), in Equation (1) to derive model parameters.

1 ∞ c j (t)
  Z
H c j (t) = dτ (2)
π −∞ t − τ
Sensors 2017, 17, 2390 7 of 11

 
The Hilbert transform, H c j (t) , and c j (t) form a complex signal, Zj (t), where

Zj (t) = c j (t) + iH c j (t) = a(t)eiϕ(t)

 
(3)

Then the envelop of each IMF can be given by

Sensors 2017, 17, 2390   7 of 11
q
2 H c j (t) 2
a j (t) = [c j (t)] + { H [c j (t)]} , ϕ j (t) = arctan{ } (4)
The Hilbert transform, ( ) , and ( ) form a complex signal, c(j (), t)where
( )+ ( ) = ( ) ( )
( )= (3)
in which a j (t), the instantaneous amplitude of x (t), reflects how the energy of x (t) varies with time.
The term Then
ϕ j (t) the envelop
is the of each IMF
instantaneous can be
phase ofgiven
x (t). by
The instantaneous frequency, ω (t), is defined as the
time derivative of the instantaneous phase ϕ(t) as follows ( )
( )= [ ( )] + [ ( )] , ( )= { } (4)
( )
dϕ(t)
in which ( ), the instantaneous amplitude ) = ( ), reflects how the energy of ( ) varies with
ω (tof (5)
dt
time. The term ( ) is the instantaneous phase of ( ). The instantaneous frequency, ( ), is
defined
Then theasoriginal
the time signal
derivative
x (t)ofcan
thebe
instantaneous phase ( ) as follows
expressed as
( )
n ( )= Z (5)

( ) can be∑
x (t) = a j (t) exp [i ω j (t)dt] (6)
Then the original signal expressed as
j =1
( )=∑ ( )exp ( ) (6)
The EEMD method is supposed to eliminate the mode mixing problem of EMD, and adds noise
The EEMD method is supposed to eliminate the mode mixing problem of EMD, and adds noise
to the signal x(t). The detailed algorithm has been included in Rato et al. [28].
to the signal x(t). The detailed algorithm has been included in Rato et al. [28].
Figure 10 shows the decomposed components of the measured tiltmeter 1 reading which included
Figure 10 shows the decomposed components of the measured tiltmeter 1 reading which
the tiltmeter
includedoutput and IMF
the tiltmeter 1 toand
output 6. The
IMFdamping
1 to 6. Thecoefficients can be identified
damping coefficients using the
can be identified slope
using theof the
amplitude–time plot [30–32]. plot [30–32].
slope of the amplitude–time
To consider the the
To consider complex
complex frequency
frequencyproperties, the Hilbert
properties, the Hilbertspectrum
spectrum is applied
is applied to re-treat
to re-treat the the
recorded signal.
recorded Figure
signal. Figure 1111shows
shows thetheHilbert
Hilbert spectrum
spectrum ofdecomposed
of the the decomposed signals
signals in Figure in
10. Figure
From 10.
FromFigure
Figure11,11,
we can
we see canthat
seethose respective
that components components
those respective are well separated,
are and
welltheseparated,
time-dependent
and the
properties of the
time-dependent frequencies
properties of are
theclearly shown. The
frequencies are dominant mode exhibits
clearly shown. The modulation centered
dominant mode at
exhibits
approximately 4.90 Hz in a red color.
modulation centered at approximately 4.90 Hz in a red color.
Tiltmeter Output

0.2
0
-0.2
0 1 2 3 4 5 6 7 8 9

0.01
IMF 1

0
-0.01
0 1 2 3 4 5 6 7 8 9
-3
x 10
5
IMF 2

0
-5
0 1 2 3 4 5 6 7 8 9

0.2
IMF 3

0
-0.2
0 1 2 3 4 5 6 7 8 9

0.05
IMF 4

0
-0.05
0 1 2 3 4 5 6 7 8 9

0.02
IMF 5

0
-0.02
0 1 2 3 4 5 6 7 8 9

0.01
IMF 6

0
-0.01
0 1 2 3 4 5 6 7 8 9
Time(s)

Figure 10. Decomposed

Figure components
10. Decomposed componentsofofthe
thetiltmeter
tiltmeter 11 reading
readingby
byusing
using a modified
a modified empirical
empirical mode
mode
decomposition (EMD)
decomposition method.
(EMD) method.
Sensors 2017, 17, 2390 8 of 11
Sensors 2017, 17, 2390 8 of 11

Sensors 2017, 17, 2390 8 of 11

Figure
Figure 11.Hilbert
11. Hilbert spectrum
spectrum of
ofintrinsic
intrinsicmode
modefunctions (IMFs).
functions (IMFs).
Figure 11. Hilbert spectrum of intrinsic mode functions (IMFs).
5. Discussion and Theoretical Consideration
5. Discussion and Theoretical Consideration
To verifyand
5. Discussion theTheoretical
dynamic characteristics
Considerationobtained from the tiltmeter readings, a bridge finite
To verify
element the dynamic
model was built characteristics
using SAP2000 obtained
according from to thethe tiltmeter
as-built plans. readings,
The bridge a bridge finite element
deck, stringers,
To verify the dynamic characteristics obtained from the tiltmeter readings, a bridge finite
modelandwas builtare
girders using SAP2000
modeled usingaccording to the
shell elements; as-built
truss membersplans.
are The
modeledbridge deck,
using stringers,
frame elements. and Thegirders
element model was built using SAP2000 according to the as-built plans. The bridge deck, stringers,
bridge hasusing
are modeled expansion
shell bearings
elements;at the abutment
truss members 1, pier
are 2,modeled
pier 3, pier 5, and
using abutment
frame 2; thereThe
elements. are bridge
fixed has
and girders are modeled using shell elements; truss members are modeled using frame elements. The
bearings
expansion at the pier
bearings at 4.
theThe bridge model’s
abutment piermaterial
1,abutment
2, pier 3,properties
pier are abutment
5, and based on the 2; construction
there2;are drawing.
fixed
bridge has expansion bearings at the 1, pier 2, pier 3, pier 5, and abutment there are bearings
fixed at
the pier Based
4. The on the
bridge modal
model’s participating
material mass
propertiesratios,arethe first
based three
on the longitudinal
construction modes
drawing. have been
bearings at the pier 4. The bridge model’s material properties are based on the construction drawing.
identified successfully through the numerical mode analysis. Figure 12a is the first mode of
Based on on
Based thethe
modal
modalparticipating
participating mass ratios,the
mass ratios, thefirst
first three
three longitudinal
longitudinal modes modes
have have
been been
longitudinal vibration, Figure 12b is the second mode of longitudinal vibration, and Figure 12c is the
identified successfully
identified through
successfully the numerical
through the numericalmode analysis.
mode Figure
analysis. 12a is12a
Figure theisfirst
themode
first of longitudinal
mode of
third mode of longitudinal vibration. According to the identified mode shapes, there exist large
longitudinal
vibration, Figure vibration,
12b is Figure
the 12b is
second the second
mode of mode of longitudinal
longitudinal vibration, vibration,
and Figureand12cFigure
is 12cthird
the is the mode
longitudinal displacements at the end bearings for each mode, and they will directly cause the
third mode of
of longitudinal longitudinal
vibration. vibration.toAccording to the mode
identified modethereshapes, there exist large
dynamic rotation at the According
rocker bearings theat identified
the abutment shapes,
2. The exist
bridge’s finite large
element longitudinal
model
longitudinal displacements at the end bearings for each mode, and they will directly cause the
displacements
theoreticallyatproved
the endthat bearings for each
monitoring the mode,
expansion andbearings
they willcan directly cause
identify the the dynamic
bridge’s rotation
natural
dynamic rotation at the rocker bearings at the abutment 2. The bridge’s finite element model
at the rocker bearings
frequencies at the abutment
in the longitudinal 2. The
direction. bridge’s finite
The longitudinal elementcan
frequencies model theoretically
be separated based proved
on the that
theoretically proved that monitoring the expansion bearings can identify the bridge’s natural
end bearings’
monitoring movement.
the expansion The comparison
bearings can identify between the experimental
the bridge’s and analytical
natural frequencies in theresults are
longitudinal
frequencies in the longitudinal direction. The longitudinal frequencies can be separated based on the
presented
direction. Thein longitudinal
the following section.
frequencies can be separated based on the end bearings’ movement.
end bearings’ movement. The comparison between the experimental and analytical results are
The presented
comparison between
in the followingthe section.
experimental and analytical results are presented in the following section.

(a)
(a)

(b)
(b)

Figure 12. Cont.

Sensors 2017, 17, 2390 9 of 11
Sensors 2017, 17, 2390 9 of 11

(c)
Figure 12. Longitudinal Mode Shapes from Finite Element Method (FEM) Analysis. (a) Mode 1 (1.584
FigureHz; Longitudinal
12.Modal ParticipatingMode
MassShapes from(b)
Ratios 0.10); Finite
ModeElement
2 (4.348 Method
Hz; Modal (FEM) Analysis.
Participating Mass(a) Mode 1
Ratios
0.17); (c) Mode 3 (5.060 Hz; Modal Participating Mass Ratios 0.18).
(1.584 Hz; Modal Participating Mass Ratios 0.10); (b) Mode 2 (4.348 Hz; Modal Participating Mass
Ratios 0.17); (c) Mode 3 (5.060 Hz; Modal Participating Mass Ratios 0.18).
To quantitatively determine the reliability of the natural frequencies obtained from the tiltmeter
reading, it was compared with the bridge’s numerical model, and the results are summarized in
To quantitatively
Table determine
1. The difference betweenthethereliability of the
field test and the natural
numerical frequencies
analysis wasobtained
5.30% forfrom themode,
the 1st tiltmeter
reading, it was compared with the bridge’s numerical model, and the results are summarized
2.81% for the 2nd mode, and 3.16% for the 3rd mode. The use of the FBG tiltmeter makes it possible in Table 1.
The difference
to acquirebetween the frequencies
the dynamic field test and
withthe numerical
high analysis was 5.30% for the 1st mode, 2.81% for
reliability.
the 2nd mode, and 3.16% for the 3rd mode. The use of the FBG tiltmeter makes it possible to acquire
Table
the dynamic frequencies with 1. Experimental
high reliability. and Analytical Natural Frequencies.
Number of Mode Analytical Model (Hz) Test Results (Hz) Error (%)
Table
1 1. Experimental1.584
and Analytical Natural1.50
Frequencies. 5.30
2 4.348 4.47 2.81
Number of3Mode Analytical5.060
Model (Hz) Test Results
4.90 (Hz) Error
3.16 (%)
1 1.584 1.50 5.30
6. Conclusions 2 4.348 4.47 2.81
3 5.060 4.90 3.16
The tiltmeter reading can extract the primary natural frequencies. This study introduced a new
approach to identify long-span steel girder bridge frequencies in the longitudinal direction based on
6. Conclusions
monitoring the dynamic rotational angle at the expansion bearing. The identified frequencies have
been compared with the numerical analysis, demonstrating the effectiveness of using a tiltmeter for
The tiltmeter reading can extract the primary natural frequencies. This study introduced a new
dynamic feature identification. The analytical and experimental results showed a very strong
approach to identify
agreement with a long-span steel girder
maximum difference bridge
of 5.3%. frequencies
Compared with thein accelerometer,
the longitudinal direction
the FBG tiltmeterbased
on monitoring the dynamic rotational angle at the expansion bearing. The
can not only capture the bridge natural frequencies, but also provide the rotational angle. identified frequencies
have beenThiscompared withinstalled
study only the numerical
the FBG analysis, demonstrating
tiltmeters on the rocker the effectiveness
bearings, of using
and it can monitora tiltmeter
the
for dynamic
swaying feature
movementidentification.
of the bottom The analytical
roller. and exists
There also experimental
a rotationresults
of the showed
bearing’s atop
very strong
plate
connected to the bridge girder. Future research should install additional tiltmeters
agreement with a maximum difference of 5.3%. Compared with the accelerometer, the FBG tiltmeter on the bridge deck
or only
can not girdercapture
to monitor
thethis rotation
bridge which
natural can help to identify
frequencies, but alsothe bridge’sthe
provide vertical dynamic
rotational movement,
angle.
This study only installed the FBG tiltmeters on the rocker bearings, and it can monitorthe
and provide accurate responses of the structure`s bearing. Installing the tiltmeters only on theroller
swaying
(swaying part of the motion) is not enough, and there should be sensors mounted for measuring the
movement of the bottom roller. There also exists a rotation of the bearing’s top plate connected to
rotation of the deck also. Additionally, the effect of input excitation on the identification results such
the bridge girder. Future research should install additional tiltmeters on the bridge deck or girder to
as road roughness, multiple truck interactions, etc. should also be studied in the next phase of the
monitor this rotation which can help to identify the bridge’s vertical dynamic movement, and provide
study.
accurate responses of the structure‘s bearing. Installing the tiltmeters only on the roller (swaying part
of theAcknowledgments:
motion) is not enough,
Support forand
this there should
research be sensors
by the Alaska mounted
University for measuring
Transportation Center Grantthe
No.rotation
510015 of
the deck also. Additionally, the effect of input excitation on the identification results such asofroad
is gratefully acknowledged. The writers also wish to acknowledge the support from the Alaska Department
Transportation and Public Facilities, University of Alaska Fairbanks, Chandler Monitoring System Inc., and
roughness, multiple truck interactions, etc. should also be studied in the next phase of the study.
Micron Optic, Inc.

Author Contributions:
Acknowledgments: SupportFeng Xiaoresearch
for this and J. Leroy Hulsey
by the conceived
Alaska and designed
University the experiments;
Transportation Feng No.
Center Grant Xiao 510015
and is
gratefully acknowledged.
J. Leroy Thethe
Hulsey performed writers also wish
experiments; Fengto acknowledge
Xiao and Gang S. the
Chensupport
analyzedfrom the Alaska
the data; J. Leroy Department
Hulsey,
of Transportation
Gang S. Chen and Public
and Feng XiaoFacilities,
contributedUniversity of Alaska
analysis tools; Fairbanks,
Feng Xiao Chandler
and Gang S. Chen wroteMonitoring
the paper. System Inc.,
and Micron Optic, Inc.
Conflicts of Interest: The authors declare no conflict of interest.
Author Contributions: Feng Xiao and J. Leroy Hulsey conceived and designed the experiments; Feng Xiao and
J. Leroy Hulsey performed the experiments; Feng Xiao and Gang S. Chen analyzed the data; J. Leroy Hulsey,
Gang S. Chen and Feng Xiao contributed analysis tools; Feng Xiao and Gang S. Chen wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2017, 17, 2390 10 of 11

References
1. Fan, W.; Qiao, P. Vibration-based damage identification methods: A review and comparative study.
Struct. Health Monit. 2011, 10, 83–111. [CrossRef]
2. Li, Y.; Cai, C.S.; Liu, Y.; Chen, Y.; Liu, J. Dynamic analysis of a large span specially shaped hybrid girder
bridge with concrete-filled steel tube arches. Eng. Struct. 2016, 106, 243–260. [CrossRef]
3. Bedon, C.; Morassi, A. Dynamic testing and parameter identification of a base-isolated bridge. Eng. Struct.
2014, 60, 85–99. [CrossRef]
4. Xing, S.; Halling, M.W.; Barr, P.J. Delamination detection of reinforced concrete decks using modal
identification. J. Sens. 2012, 2012, 156583. [CrossRef]
5. Talebinejad, I.; Fischer, C.; Ansari, F. Serially multiplexed FBG accelerometer for structural health monitoring
of bridges. Smart Struct. Syst. 2009, 5, 345–355. [CrossRef]
6. Lee, Y.G.; Kim, D.H.; Kim, C.G. Performance of a single reflective grating-based fiber optic accelerometer.
Meas. Sci. Technol. 2012, 23, 045101. [CrossRef]
7. Huffman, J.T.; Xiao, F.; Chen, G.; Hulsey, J.L. Detection of soil-abutment interaction by monitoring bridge
response using vehicle excitation. J. Civ. Struct. Health Monit. 2015, 5, 389–395. [CrossRef]
8. Caglayan, B.O.; Ozakgul, K.; Tezer, O. Assessment of a concrete arch bridge using static and dynamic load
tests. Struct. Eng. Mech. 2012, 41, 83–94. [CrossRef]
9. Xu, M.; Shen, Q.; Liu, Y.; Ding, H.; Yang, G.; Ma, Q. Large-scale bridge structural health monitoring
system based on smart sensor networks. In Proceedings of the ICTIS 2011: Multimodal Approach to
Sustained Transportation System Development: Information, Technology, Implementation, Wuhan, China,
30 June–2 July 2011; pp. 942–948.
10. Park, J.W.; Sim, S.H.; Jung, H.J. Development of a wireless displacement measurement system using
acceleration responses. Sensors 2013, 13, 8377–8392. [CrossRef] [PubMed]
11. Kang, L.H.; Kim, D.K.; Han, J.H. Estimation of dynamic structural displacements using fiber Bragg grating
strain sensors. J. Sound Vib. 2007, 305, 534–542. [CrossRef]
12. Wu, B.; Lu, H.; Chen, B.; Gao, Z. Study on Finite Element Model Updating in Highway Bridge Static Loading
Test Using Spatially-Distributed Optical Fiber Sensors. Sensors 2017, 7, 1657.
13. Moschas, F.; Stiros, S. Measurement of the dynamic displacements and of the modal frequencies of a
short-span pedestrian bridge using GPS and an accelerometer. Eng. Struct. 2011, 33, 10–17. [CrossRef]
14. Lee, J.J.; Ho, H.N.; Lee, J.H. A vision-based dynamic rotational angle measurement system for large civil
structures. Sensors 2012, 12, 7326–7336. [CrossRef] [PubMed]
15. Fukuda, Y.; Feng, M.Q.; Narita, Y.; Kaneko, S.I.; Tanaka, T. Vision-based displacement sensor for monitoring
dynamic response using robust object search algorithm. IEEE Sens. J. 2013, 13, 4725–4732. [CrossRef]
16. Erol, B. Evaluation of high-precision sensors in structural monitoring. Sensors 2010, 10, 10803–10827.
[CrossRef] [PubMed]
17. Park, Y.S.; Agbayani, J.A.; Lee, J.H.; Lee, J.J. Rotational angle measurement of bridge support using image
processing techniques. J. Sens. 2016, 2016, 1923934. [CrossRef]
18. Dong, Y.; Song, R.; Liu, H. Bridges Structural Health Monitoring and Deterioration Detection-Synthesis of
Knowledge and Technology; Final Report; Alaska University Transportation Center: Fairbanks, AK, USA, 2010.
19. Zhou, J.; Li, X.; Xia, R.; Yang, J.; Zhang, H. Health Monitoring and Evaluation of Long-Span Bridges Based
on Sensing and Data Analysis: A Survey. Sensors 2017, 17, 603. [CrossRef] [PubMed]
20. Helmi, K.; Taylor, T.; Zarafshan, A.; Ansari, F. Reference free method for real time monitoring of bridge
deflections. Eng. Struct. 2015, 103, 116–124. [CrossRef]
21. Norouzi, M.; Cormier, K.; Aydemir, M.; Hashtroodi, S.; Hunt, V.; Nims, D.; Helmicki, A. Measuring
Displacement of Tall Concrete Columns During Construction: Jeremiah Morrow Bridge, Interstate 71,
Lebanon, Ohio. Transp. Res. Rec. J. Transp. Res. Board 2014, 2408, 97–106. [CrossRef]
22. Kim, W.; Laman, J.A. Seven-year field monitoring of four integral abutment bridges. J. Perform. Constr. Facil.
2011, 26, 54–64. [CrossRef]
23. Xiao, F.; Hulsey, J.L.; Balasubramanian, R. Fiber optic health monitoring and temperature behavior of bridge
in cold region. Struct. Control Health Monit. 2017, 24, e2020. [CrossRef]
24. FBG KOREA, Inc. FBG Tiltmeter. Available online: http://www.fbg.co.kr/eng/bbs/board.php?bo_table=
fbgp04&wr_id=1 (accessed on 24 August 2017).
Sensors 2017, 17, 2390 11 of 11

25. Ansari, F. Fiber optic sensors for structural health monitoring of civil infrastructure systems. Struct. Health
Monit. Civ. Infrastruct. Syst. 2009, 4, 103–110.
26. Ye, X.W.; Su, Y.H.; Han, J.P. Structural health monitoring of civil infrastructure using optical fiber sensing
technology: A comprehensive review. Sci. World J. 2014, 2014, 652329. [CrossRef] [PubMed]
27. Alan, V.O.; Ronald, W.S.; John, R. Discrete-Time Signal Processing, 2nd ed.; Prentice Hall: Upper Saddle River,
NJ, USA, 1999.
28. Rato, R.T.; Ortigueira, M.D.; Batista, A.G. On the HHT, its problems, and some solutions. Mech. Syst.
Signal Process. 2008, 22, 1374–1394. [CrossRef]
29. Huang, N.E.; Shen, Z.; Long, S.R.; Wu, M.C.; Shih, H.H.; Zheng, Q.; Yen, N.C.; Tung, C.C.; Liu, H.H.
The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series
analysis. Proc. R. Soc. Lond. A Math. Phys. Eng. Sci. 1998, 454, 903–995. [CrossRef]
30. Shi, W.X.; Shang, J.Z.; Lu, X.L. Modal identification of Shanghai World Financial Center both from free and
ambient vibration response. Eng. Struct. 2012, 36, 14–26. [CrossRef]
31. Wang, Z.C.; Chen, G. Analytical mode decomposition with Hilbert transform for modal parameter
identification of buildings under ambient vibration. Eng. Struct. 2014, 59, 173–184. [CrossRef]
32. Chen, J.; Xu, Y.L.; Zhang, R.C. Modal parameter identification of Tsing Ma suspension bridge under Typhoon
Victor: EMD-HT method. J. Wind Eng. Ind. Aerodyn. 2004, 92, 805–827. [CrossRef]

© 2017 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/).