Accepted Manuscript

Measurement of the Real-time Deflection of Cable-Stayed Bridge Based on

Cable Tension Variations

Yonghui Huang, Yang Wang, Jiyang Fu, Airong Liu, Wei Gao

PII: S0263-2241(18)30083-6
DOI: https://doi.org/10.1016/j.measurement.2018.01.070
Reference: MEASUR 5245

To appear in: Measurement

Received Date: 17 July 2017

Revised Date: 13 December 2017
Accepted Date: 31 January 2018

Please cite this article as: Y. Huang, Y. Wang, J. Fu, A. Liu, W. Gao, Measurement of the Real-time Deflection of
Cable-Stayed Bridge Based on Cable Tension Variations, Measurement (2018), doi: https://doi.org/10.1016/
j.measurement.2018.01.070

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Measurement of the Real-time Deflection of Cable-Stayed Bridge
Based on Cable Tension Variations
Yonghui Huang1, Yang Wang1, Jiyang Fu1*, Airong Liu1, Wei Gao2
1
Guangzhou University-Tamkang University Joint Research Center for Engineering Structure Disaster Prevention and
Control, Guangzhou University, Guangzhou 510006, China
2
Centre for Infrastructure Engineering and Safety (CIES), School of Civil and Environmental Engineering, The
University of New South Wales, Sydney, NSW 2052, Australia

Abstract: Measuring the real-time deflection of a bridge is vital to understand its

behavior and performance. Although automated structural health monitoring systems
have been established for many long span bridges, continuously measuring the real-time
deflection of a bridge is still cumbersome. According to the mechanical and structural
features of cable-stayed bridge, a novel method, for measuring the real-time deflection
of cable-stayed bridge based on variations of cable tension, is freshly proposed. The
relationship between the variation of cable tension and the deflection of the anchor point
between the cable and girder is formulized building upon the geometric relationship of
the deformation. Then, the real-time frequency of the cable is calculated from the
measured acceleration signals using a time-frequency analysis method referred to as the
Hilbert-Huang Transform. Consequently, the real-time cable tension is computed and
the deflections of the anchor points of the girder are obtained. In the practical structural
health monitoring of a cable stayed-bridge that realistically monitors the tension in
many cables, measurements of the kinetic deflection curve of the girder can be achieved
without requiring additional equipment. A numerical example is implemented to verify
the accuracy of the proposed method. In addition, through a real engineering application
of the He-dong Bridge, the feasibility and effectiveness of the proposed method are
demonstrated by a comparison with global positioning system (GPS) observation results.
The developed method can provide an alternative cost-effective option for the
real-time deflection measurement of cable-stayed bridges.

Keywords: Real-time deflection; Structural health monitoring; Cable-stayed bridge;

1
Cable tension variation; Hilbert-Huang Transform.

1. Introduction
Cable-stayed bridges are commonplace in engineering applications and have been
widely used as medium and long span bridges due to their great spanning capacity and
beautiful shape. However, large deflections usually occur in long-span bridges,
especially for bridges with steel box girders, because of their weak stiffness. The
consequences of large deformations of bridges are critical to their serviceability and
safety. Therefore, deformation measurement is essential for the health condition
assessment of long span cable-stayed bridges.
Over the past few decades, a great number of deformation measurement methods
have been developed. In the early stage, traditional surveying techniques were widely
used with instruments such as theodolite, level, rangefinder, total station [1], etc.
However, traditional measurement methods have the disadvantages of a heavy workload
and low efficiency. These methods are also easily influenced by climatic conditions, and
continuous and automatic measurements are difficult to implement because they require
inter-visibility between the monitoring point and base point. Later, the conventional
methods of ground-based measurements have continuously improved and developed.
Currently, the remarkable progress is the total station instrument which is widely used,
such as the Robotic Total Station (RTS) which is also known as Geo-robot sometimes.
This system has provided an excellent technical means for automatic deformation
measurements and has achieved automatic measurements under unattended, all-weather
and certain omnidirectional circumstances. However, using this method, one instrument
can only monitor one point and inter-visibility between the individual instrument and
monitoring point is needed [2]. These disadvantages significantly impair its
applicability in complex structures.
With the development of technology and improvements in deformation measuring
requirements, a number of advanced measurement methods have emerged, including the
inclinometer method [3], displacement sensor method [4], photogrammetric method [5],

2
laser measurements [2] and global positioning system (GPS) measurement technology
[6]. With servo-inclinometers the inclination (rotation) of the deflection shape relating
to the vertical axis can be measured. Then, a vertical deflection can be calculated as the
first-order integration of the inclination [3]. The displacement sensor method is a
high-precision method that can realize the consecutive collection of displacements, but
needs to be installed on a fixed basement, which limits its application in long span
bridges. The photogrammetric method is a new method that can realize the real-time
monitoring of structural deformations and has been used to monitor the structural
deformations of the second Bosporus Bridge of Istanbul [5]. However, a fixed pillar it
also needed to install the instrument.
Among the known deformation measuring methods, GPS has been proved to be a
state-of-the-art measuring tool because of its high-speed, all-climate working abilities
and dispensing with inter-visible measuring stations [7]. This system is commonly used
in large-scale structures, such as towers, suspension bridges, dams and high-rise
buildings [8, 9]. Real-time kinematic (RTK) global positioning system (GPS) [10, 11]
were designed and installed on the Humber Bridge in the UK [12] and the Akashi Kaiyo
Bridge in Japan [13], and obtained good results. However, the GPS measurement
method has two main shortcomings. One is that the price of a GPS receiver is very
expensive and one GPS receiver can only measure one point. The total cost may be
unaffordable if the deflection curve of the bridge’s girder is needed to be measured. The
other shortcoming is that the measurement precision of the vertical displacement of
GPS can only reach the millimeter level [14-16]. For example, in the Humen Bridge, the
tested results show that the RTK GPS system has only a potential resolution of ±2 mm
horizontally and ±5 mm vertically [9].
As aforementioned, direct measurement methods, including the most commonly
used GPS observation method, have the intrinsic limitations of arrangement and
precision. Therefore, to overcome these drawbacks, indirect measurement approaches
should be considered. For cable-stayed bridges, it is easy to install acceleration sensors
on the cables and convenient to realize the dynamic measuring of cable tensions by
using a vibration-based force identification method [9]. Building upon this achievable
3
realistic engineering condition, a new method for measuring the real-time deflection of
a cable-stayed bridge based on the cable tension variation is proposed in this paper. In
order to obtain the real-time deflection, four steps are devised in this method. First, the
relationships between the variation of the cable tension and the deflection of the bridge
girder are derived. Second, accelerometers are installed on the cables to acquire the
acceleration signals of the vibrations. Third, real-time frequencies of the cables are
identified by using the Hilbert-Huang Transform method, and then the real-time cable
tensions are calculated by using the vibration-based force identification method. Fourth,
the real-time deflection of the anchor points of the girder are calculated by substituting
the real-time cable tensions into the relationship equations derived in the first step.
Compared with the existing direct and indirect measurement methods, the proposed
method is obviously superior. First, this method can easily furnish automatic continuous
monitoring of the deflection of a bridge girder because that the real-time cable tension
can be easily and automatically measured. Second, this low-cost and high-efficiency
method can measure the deflection of all the anchor points between the girder and
cables simultaneously. Because cable tension monitoring is an essential component of
the structural health monitoring system of a cable-stayed bridge, no additional
instruments are needed to implement this method. Third, the proposed measurement
method is available under all-weather conditions. Its precision is only dependent on the
precision of the identified cable tension and the relationship between the cable tension
and girder’s deflection rather than the surrounding environment. If high-performance
methods are used to monitor the cable tension more accurately, this method can obtain
higher precision. The results of a numerical example and a case study show that the
accuracy of the proposed method can reach a sub-millimeter level and is accurate
enough for practical engineering applications.

2. Relationship between cable tension variation and girder’s

deflection
According to the theory of an elastic foundation beam, the main girder of a
cable-stayed bridge can be regarded as a continuous beam with elastic supports. When
4
the cable tension is changed, the structural deformation caused by cable tension is
shown in Fig. 1. If the bending stiffness of the pylon is assumed as infinity, the
deflection of the girder is only caused by the elongation of the cable, as shown in Fig.
1(a). If the axial stiffness of the cable is assumed as infinity, the deflection of the girder
is only caused by the bending of the pylon, as shown in Fig. 1(b). In fact, both the
bending stiffness of the pylon and the axial stiffness of the cable are not infinity, so the
deflection of the girder is composed of the superposition of the above two parts [17]. It
is worth pointing out that the effect of the longitudinal movement of the girder on the
deflection of the girder is ignored in the above assumption. The reason is that the
deflections δ1 and δ2 are very small compared with the length of the girder Lg, therefore,
the rotation angles β1 and β2 are very small, which lead to the deflections caused by the
longitudinal movement of the girder are higher order infinitesimal of δ1 and δ2 which
can be ignored without losing precision.
△p

δb
Lc Lc
Pylon Cable Before deformation Pylon Cable Before deformation
ht

ht

After deformation After deformation

x x
α

α
β1

β2
δ1

Girder Girder
δ2
δa δc
Lg Lg
y y

(a) (b)
Fig. 1. Deflection of the anchor point between the girder and cable
In Fig. 1(a), δ1 is the deflection of the anchor point between the stay cable and the
girder due to the elongation of the stay cable δa, and α is the angle between the cable
and the girder. According to the geometric relationships, the following formulation
stands:

a Lc
1   T (1)
sin  Ec Ac sin 

where T , Ec , Ac and Lc denote the variation of cable tension, Young’s modulus,

the cross-sectional area and the length of the cable, respectively.

5
 From Eq. (1), it can be seen that the deflection of the anchor point due to cable’s
elongation is linearly related to the variation of cable tension. This component of
deflection is easy to calculate.

In Fig. 1(b), δ2 is the deflection of the anchor point between the stay cable and

girder due to pylon’s bending, and is the longitudinal displacement of the anchor

point between the pylon and the stay cable. Because the cable has no elongation, there
is:

 c   b   p cos  (2)

and then

c  p cos a
2     p cot  (3)
sin  sin 

According to Eqs. (1) and (3), the expression of the relationship between the cable
tension variation (ΔT) and the deflection of girder (δ) can be expressed as:

Lc
  1   2  T   p  cot  (4)
Ec Ac sin 

A typical model of a cable-stayed bridge is shown in Fig. 2, there are n pairs of

cables anchored at each pylon. The corresponding mechanical model of the deformation
of a pylon is shown in Fig. 3.

Tn Tn
...

...

...
...

T n' T n'
Tj Tj
...
...

T j'
...

...

T j'
T 1' T1 T1 T 1'

Fig. 2. Mechanical model of a typical cable-stayed bridge

6
 △ pn △ T ncosαn
△ T n'cosαn'

...

...
△ T j'cosαj' △ T jcosαj Pj

...

...
E tI t
△ T 1'cosα1' △ T 1cosα1 hn
hj hj

h1
x

(a) (b)

Fig. 3. Mechanical model of the deformation of the pylon

When the cable tensions changed, the mechanical model of the pylon’s longitudinal
displacement calculation is shown in Fig. 3 (a). The pylon is seen as a cantilever beam
subjected to a series of concentrated loads. For a uniform section beam with
homogeneous material subjected to a concentrated load (Fig. 3 (b)), the function of the
deflection curve of the beam is:
 Pj x 2
  3h j  x  x  hj
 6 Et I t
y  x   3
(5)
 Pj h j Pj h 2j
 3E I  2 E I  x  h j  x  hj
 t t t t

According to the superposition principle, for a pylon subjected to a series of

concentrated loads (Fig. 3 (a)), there is:
i j

 h  T  cos   Ti '  cos  i' 

1
 Pj  i
3
i i
3Et I t i 1
i  j 1

 h h  hi   Ti  cos  i  Ti '  cos  i' 

1
 i
2
j (6)
2 Et I t i 1

1 i n 2
  h j  3hi  h j   Ti  cos i  Ti '  cos i' 
6 Et I t i  j 1
where  Pj is the longitudinal displacement of the anchor point between the jth

cable and the pylon. The parameters hi and h j are the heights of the anchor point of

the ith and jth cable, respectively. Ti is the tension variation of the ith cable. Et and

I t are the Young’s modulus and the second moment of area of the pylon, respectively.

There are three types of cable-stayed bridge systems: the fan system, semi-fan
system and harp system [18], and their configurations are shown in Fig. 4.
7
 For the fan system cable-stayed bridge, all of the cables are anchored at the same
position of the pylon. Therefore, Eq. (4) can be simplified as:
i n

 h  T  cos   Ti '  cos i' 

1
 Pj  t
3
i i (7)
3Et I t i 1

where ht is the height of the anchor point between cables and the pylon.

For the harp system cable-stayed bridge, parallel stay cables are used so that the
angles between the cables and the girder are the same. Eq. (4) can be simplified as:

cos   i j 3 i  j 1 i n 
 Pj    i i    i    i  j  i     i    h 2j  3hi  h j   Ti  Ti '  
' 2 '
2 h T T 3 h h h Ti T
6 Et I t  i 1 i 1 i  j 1 
(8)
where  is the common angle of the stay cables.
For the semi-fan system cable-stayed bridge, the original expression Eq. (6) is
used.

0.2-0.4 L L 0.2-0.4 L

(a) Fan System

0.2-0.4 L L 0.2-0.4 L

(b) Semi-Fan System

0.3-0.45 L L 0.3-0.45 L

(c) Harp System

Fig. 4. Cable-stayed bridge systems

8
3. Unified formula for cable tension calculation
It is required to know the variation of cable tension beforehand when using Eq. (4)
to calculate the deflection of the girder. Currently available techniques to identify the
cable tension include the static methods, which directly measure the tension by using a
load cell or a hydraulic jack [19], and the vibration-based method, which indirectly
estimates the tension from the measured natural frequencies [20]. If the direct
measuring method is used to measure the cable tension, then the real-time deflection of
the girder can be calculated directly by using Eq. (4) and the measured real-time cable
tension. However, for most of the in-service cable-stayed bridges, direct measuring
methods cannot be used because there are no pre-installed load cells or hydraulic jacks
in the anchorage of the cable. The vibration-based method is the most commonly used
method to identify the cable tension of the in-service cable-stayed bridges. In order to
obtain an accurate cable tension, one of the key problems for the vibration-based
method is to obtain an accurate relationship between the cable tension and natural
frequencies. In this paper, the unified practical formulas proposed by Huang et al. [21]
are used to calculate the cable tension from the measured frequencies. A brief
introduction about this method is described as follows.
The simplest relationship between the natural frequencies and the cable tension
calculated from the taut string theory is [22]
2
 f 
T  4ml  n 
2
(9)
 n 

where T is the cable tension, m and l are the mass per unit length and length of the
cable, respectively. fn is the frequency of the nth vibration mode and n is the frequency
order.
However, taut string theory assumes that the cable is an ideal string that may
introduce significant errors from ignoring the bending stiffness and the boundary of the
cables [23]. To overcome these disadvantages, a unified practical formula is proposed
[21]. Correction coefficients Kn are multiplied ahead the right side of Eq. (9) to consider
the effect of the bending stiffness and boundary conditions. The expression of Kn is as

9
follows
Kn   An n 2  Bn n  1 (10)


where n  EI 4m 2 f n 2l 4  and EI is the bending stiffness of the cable. An

and Bn are the boundary condition coefficients with the following expression

 98.2n 4  87.64n3  65.37 n 2 for fixed-fixed boundaries


An  97.51n 4  47.18n3  10.17n 2 for fixed-pinned boundaries (11-a)
 0 for pinned-pinned boundaries

9.31n  1.72 for fixed-fixed boundaries

Bn   4.78n for fixed-pinned boundaries (11-b)

 0 for pinned-pinned boundaries

Then, the expression of the unified practical formula is

2
 f 
T    An n  Bn n  1 4ml  n 
2 2
n  1, 2, 3 10 (12)
 n 

The cable tension T can be calculated from any of the first 10 order frequencies.
The accuracy of the unified practical formula has been validated against several
numerical examples as well as practical test cases [21].

4. Hilbert-Huang Transform

4.1. Brief introduction of Hilbert-Huang Transform

The Hilbert-Huang Transform (HHT) is a new method that has been introduced
recently to decompose nonlinear, non-stationary time series into a sum of different
modes, each one having a characteristic frequency. This method includes empirical
mode decomposition (EMD) and the Hilbert spectral analysis (HSA) [24]. The key part
of the HHT is the EMD method with which any complicated data set can be
decomposed into a finite and often small number of components, called intrinsic mode
functions (IMFs). An IMF is defined as any function that has the same (or differing at
most by one) number of zero-crossing and extrema, and symmetric envelopes defined
by the local maxima and minima. The definition of an IMF guarantees a well-behaved

10
Hilbert transform of the IMF. This decomposition method operating in the time domain
is adaptive, and, thus, highly efficient. With the Hilbert transform, the IMFs yield
instantaneous frequencies as a function of time, so the instantaneous frequency is also
called the real-time frequency. In this paper, the HHT method is used to analyse the
acceleration signals to obtain the real-time frequencies of the cable.

4.2. Example of an analogue sinusoidal signal

An analogue sinusoidal signal y (t ) is assumed, the expression of y (t ) is

y(t )  sin(4 t )  cos(24 t ) (13)

where t is the time. The sampling frequency of the signal is 100 Hz and analysis time is
50 s. The analogue signal has two natural frequencies: 2 Hz and 12 Hz. The waveform
of the analogue signal is shown in Fig. 5. Three IMF components decomposed from the
original signal by EMD are shown in Fig. 6. IMF1 and IMF2 are the two components
containing the second and the first natural frequency of the original signal, respectively.
IMF3 is the trend term. The instantaneous frequencies of IMF1 and IMF2 are obtained
using the Hilbert Transform, and the results are shown in Fig. 7 and Fig. 8.

Fig. 6. Decomposition of the analogue

Fig. 5. Analogue signal waveform
signals by EMD

11
 Fig. 7. Instantaneous frequency of
Fig. 8. Instantaneous frequency of IMF2
IMF1
From Figs. 7 and 8, the instantaneous frequencies of IMF1 and IMF2 are 12 Hz and
2 Hz respectively in a time range of 2-48 s with a very small amplitude oscillation. At
t=0-2 s and t=48-50 s the results oscillate significantly because of the end effect [25].
The results show that different components can be separated from the original signal by
EMD, and the instantaneous frequencies of the different IMFs obtained from the Hilbert
Transform agree well with the true values, except for the beginning and ending results.

4.3. Example of a cable with time-varying tension

In the example of section 4.2, the instantaneous frequencies of the analogue signal
are constant. However, in a real bridge, the tension and natural frequencies of the cable
change over time. Therefore, a simple supported cable shown in Fig. 9 with
time-varying tension is studied as a further verification example. The parameters of the
cable are as follows: L=150 m, A=0.026 m2, E=195 GPa, ρ=8300 kg/m3 and α=45˚.
T
B

α
A

Fig. 9. Simplified schematic of stay cable

12
 The time-varying tension of the cable is shown in Fig. 10. From t=0 s to t=50 s, the
cable tension increases proportionally from 600 kN to 1600 kN with a slope of 20 kN/s.
From t=50 s to t=100 s, the cable tension decreases proportionally from 1600 kN to
1000 kN with a slope of -12 kN/s. From t=100 s to t=150 s, the cable tension is equal to
1000 kN constantly. From t=150 s to t=200 s, the cable tension decreases proportionally
from 1000 kN to 600 kN with a slope of -8 kN/s. From t=200 s to t=250 s, the cable
tension is equal to 600 kN constantly.
The acceleration time history of the cable is obtained from finite element transient
analysis and is shown in Fig. 11. The first order real-time frequency is decomposed
using the HHT method and is shown in Fig. 12. By submitting the real-time frequency
into the unified formula of the cable tension calculation, the real-time cable tension can
be obtained. Comparisons of the calculated real-time cable tension and the true value as
well as the relative error are shown in Fig. 13. From these figures, it can be seen that the
cable tension calculated by the proposed method agrees well with the actual one, which
again verifies the accuracy of the proposed method. The trends in the tension data are
consistent with each other, and the errors are no more than 5%, except at the turning
points. The reason for the large relative errors near the turning points is that when the
cable tension has a sudden change (turning point), the real-time frequency oscillates
sharply which causes an unusual error (show in Figs. 12 and 13).

Fig. 11. Acceleration time history of the

Fig. 10. Time-varying cable tension
cable

13
 Fig. 12. The first order time-varying Fig. 13. Comparison of the calculated
frequency value and true value

5. Measuring method for real-time deflections

5.1. Procedures of the method

The proposed new method for measuring the real-time deflection of a cable-stayed
bridge based on the cable tension variation includes four steps:
(1) The relationship equations between the variation of cable tension and the
deflection of the bridge girder are derived following the procedures described in section
2.
(2) Accelerometers are installed on the cables, and the acceleration signals of
cable’s vibrations are acquired.
(3) Real-time frequencies of the cables are estimated by using the Hilbert-Huang
Transform method, and then real-time cable tensions are identified by using the
vibration-based force identification method shown in section 3.
(4) The real-time deflection of the bridge’s girder is calculated by substituting the
real-time cable tensions into the relationship equations derived in the first step.
(5) The kinematic deflection curve of the bridge’s girder is obtained after all the
real-time deflections of the anchor points between the stay cables and the girder are
calculated by the proposed method.
In summary, as long as the real-time cable tensions are measured, the proposed
method can be used to obtain the real-time deflection of a cable-stayed bridge.

14
5.2. Numerical verification

The elevation of an example cable-stayed bridge is shown in Fig. 14, which has a
main span of 70 m and a total length of 110 m. The girder is supported by a fixed
bearing in the left pier and by longitudinally movable bearings at the right pier. There
are no vertical supports between the girder and the pylons, and for such bridge it is also
called floating system cable-stayed bridge. From the left to right, the cables of the
bridge are numbered from LS1 to LS14, and the deflection measuring points 1 to 14 are
the anchor points between the cables and girder. A finite element model is established
by using ANSYS software [26] and is shown in Fig. 15. The pylons and girder are
simulated using spatial beam element Beam 188, and the cable is simulated using the
Link 180 element. The parameters of the bridge are listed in Table.1.
Table. 1. Parameters of the bridge
Young’s Density Cross sectional Inertia
Hight (m)
modulus(GPa) (kg/m3) area (m2) moment (m4)

Cable 200.0 7850 0.005 2×10-5 --

Girder 36.0 2600 10.0 0.83 1.0

Pylon 36.0 2600 12.0 16.0 24.0

20m 70m 20m

LS11-LS14
LS1-LS4 LS5-LS7 LS8-LS10
24m
20m

46°
16m

54°
35°
30°

46°

42°

Monitoring point1-4 Monitoring point 5-10 Monitoring point 11-14

Fig. 14. Elevation of the cable-stayed bridge example

15
 Fig. 15. Finite element model
The loads applied to the cable-stayed bridge model are as follows: First, an impact
excitation is applied at the mid-span of the bridge to enable the bridge to vibrate freely.
Then, a concentrated load, P=900 kN, is applied to simulate a vehicle load starting from
t=50 s. The vehicle moves from the left pier to the right pier at a speed of 18 km/h. After
the vehicle load leaves, the bridge freely vibrates again. The first 150 s of acceleration
signals of the cables are collected. Figs.16 (a) ~ (d) show the acceleration time history
of LS4, LS5, LS6 and LS7, respectively.

Fig. 16. Acceleration time histories (a) LS4, (b) LS5, (c) LS6, (d) LS7
The real-time first order frequency of the cable can be extracted by using the HHT
method. In this case, LS7 is used as an example. The real-time first order frequency of
LS7 is shown in Fig. 17. By submitting the real-time first order frequency of the cable
into the unified formula of the cable tension calculation, the real-time cable tension is
calculated and the result is shown in Fig. 18.

16
 Fig. 17. Real-time first-order frequency Fig. 18. Real-time cable tension of LS7
of LS7
Then, combining the real-time cable tensions with the relationship equations
shown in Eqs. (4) and (6), the real-time deflections of the anchor points between the
cables and girder are calculated. A comparison of the calculated real-time deflection
using the proposed method and the FEA results (directly obtained from the finite
element transient analysis) of measuring point 7, which is the anchor point between
cable LS7 and the girder, is shown in Fig. 19. From the figure, the calculated results
using the proposed method agree well with the FEA results. The maximum deflection of
the calculated result occurs at time t=59.6 s, and its value is 22.93 mm, while the
maximum deflection of the FEA result also occurs at time t=59.6 s, and its value is
24.34 mm; therefore, the relative error of the maximum deflection is 5.8%. The results
of the proposed method surrounding time t=54 s and t=68 s fluctuate significantly
because at these times, the load begins to act on or is then removed from the bridge, so
the bridge vibrates fiercely, and the cable tensions vary from the constant value.

Fig. 19. Comparison of the deflections calculated by the proposed method and FEA
The real-time deflection of other monitoring points can be obtained by the proposed

17
method with the same procedures. Then the deflection curve of the girder at a certain
time can be obtained. Fig. 20 shows the acting position of the moving load at different
times. Fig. 21 shows the kinematic deflection curve of bridge’s girder at various times.
In Fig. 20 and Fig. 21, the deflection at time t=0 s is the initial deflection. When t=52 s,
the vehicle load is acting on the middle of the left side span, so the maximum deflection
of the whole girder appears at the middle of the left side span. When t=61 s, the load is
acting on the middle of the main span, so the maximum deflection of the whole girder
appears at the middle of the main span. The rules of the calculated results agree well
with our theoretical judgment. Finally, the kinematic deflection curves of the girder are
drawn as shown in Fig. 21 because the deflection of each anchor point measured by the
proposed method are continuously dynamic and synchronous.

t =52s t =61s t =70s

t =50s t =54s t =72s
t =68s
Fig. 20. Position of moving load at different times

5 | | | | | | | | | |

0 __
Deflection (mm)

-5 __

-10 __

t=52s
t=54s
-15 __

t=61s
-20 t=68s __

t=70s
-25
0 10 20 30 40 50 60 70 80 90 100 110
Longitudinal distance (m)

Fig. 21. Kinematic deflection curves of the girder

5.3. Discussion and parametric analysis

As described in section 2, the deflection of the main girder is composed of two

parts δ1 related to the axial stiffness of the stay cable and δ2 related to the bending

18
stiffness of the pylon. In order to analyze the effect of stiffness ratio of the cable to the
pylon on the deflection of girder, a dimensionless parameter is defined as:

Ec Ac h 2
 (14)
Et I t

where h is the height of the pylon and γ denotes the stiffness ratio of the cable to the
pylon which may affect the ratio of δ1 to δ2. If γ=0, then the bending stiffness of the
pylon is infinity, which will lead to δ2=0 and δ=δ1. In addition, if γ=∞, then the axial
stiffness of the cable is infinity, which will lead to δ1=0 and δ=δ2.
Using the same example shown in section 5.2, a parametric analysis of the effect of
γ on the deflection of girder is studied. A concentrated load of P=900 kN is applied at
the middle of the main span. Only the value of the second moment of area of the pylon
is changed in the models. Fig. 22(a) shows the variations of deflection components of
monitoring point 7 versus γ. It can be seen that the deflection caused by the cable’s
elongation δ1 decreases as γ increases, while the deflection caused by the pylon’s
bending δ2 increases as γ increases. It can also be seen that the analytical solutions agree
with the FE results quite well for bridges with different stiffness ratio of the cable to the
pylon. Fig. 22(b) shows the variations of the ratio for different deflection components to
the total deflection versus γ. It can be seen clearly that the ratio of the deflection caused
by the cable’s elongation δ1 decreases from 100% to 51.5%, while the deflection caused
by the pylon’s bending δ2 increases from 0% to 48.5% as γ increases from 0 to 10. The
common range of stiffness ratios of the cable and pylon in cable-stayed bridges is from
1 to 3 [18], and in this range the deflection of the girder is dominated by the cable’s
elongation at a ratio of 80% to 65%.

19
 Fig. 22. Variations of deflection components of monitoring point 7 versus γ (a)
deflection, and (b) ratio

6. Case study
He-dong Bridge (shown in Fig .23) is a famous bridge in Guangzhou city that spans
the Pearl River. The bridge is a cable-stayed bridge with a steel-concrete composite
girder. As shown in Fig. 23, the middle span of the bridge is 360 m, and there are 72
pairs of cables spaced at a distance of 9.5 m on the bridge girder. In the health
monitoring system of the He-dong Bridge established by the authors, acceleration
sensors are installed on the cables to monitor the cable tension. In addition, three GPS
monitoring stations are installed on the top of the west and east pylons as well as the
middle of the main span (shown in Fig. 23). Leica GR10 receivers are used as the GPS
mobile stations. The synchronization of the acceleration signal and the GPS signal are
guaranteed by using NI data acquisition instruments to sample the GPS and acceleration
data with the same sampling frequency of 20 Hz.
Because of the large amount of GPS and cable tension monitoring data, only the
GPS data in the early morning of April 15, 2015, from 1:30 am to 2:00 am is selected
for a comparison analysis. During this period, the number of vehicles passing over the
bridge is less than at other times, and the temperature effect can also be ignored because
the temperature variation is very small. In this case study, the real-time deflection of the
middle span of the bridge is measured by the proposed method following the procedures
described in section 5.1. The parameters of the bridge are listed in Table. 2. Note that
the parameters of the composite girder are unified to a uniform material.
Table. 2. Parameters of cable C36, the pylon and girder of the He-dong Bridge

E(GPa) ρ(kg/m2) A (m2) I (m4) L(m) α (°)

Cable C36 185.0 7850 0.0081 5.37×10-6 191.66 28.5

Pylon 34.5 2600 27.34 220.627 121.27 /

Girder 206.0 14054 1.39 0.602 / /

20
 GPS GPS

West East
C19-C36 C37-C54
C1-C18 C55-C72
GPS

28.2m 36m 36m 43.8m 360m 43.8m 36m 36m 28.2m

Fig. 23. Elevation of He-dong bridge and the location of GPS stations

6.1. Acceleration signals of Cable C36

In the health monitoring system of the He-dong Bridge, an acceleration sensor is

installed on cable C36. The measured acceleration signals of cable C36 at April 15,
2015, from 1:30 am to 2:00 am are shown in Fig. 24.

Fig. 24. Acceleration signals of cable C36

6.2. Real-time cable tension of cable C36

Using the EMD method, the acceleration signals of cable C36 are decomposed into
a small number of IMF components. With the Hilbert transform, the instantaneous
frequency of each IMF component is yielded as a function of time. From the unified
practical formula of cable tension identification [21], it can be seen that, the cable
tension can be calculated from any order frequency. In this case study, the first order
frequency or so-called fundamental frequency was used. The first order real-time
frequency of cable C36 produced by the HHT method is shown in Fig. 25, and the
real-time tension of cable C36 is shown in Fig. 26.
21
Fig. 25. First order real-time frequency of Fig. 26. Real-time tension of cable C36
cable C36

6.3. Real-time deflection of the middle span

Because the He-dong bridge is a semi-fan system cable-stayed bridge, the tensions
of all cables are required to calculate the longitudinal displacement of the pylon using
Eq. (6). For the sake of simplicity, the longitudinal displacement of the pylon recorded
by the GPS at the top of the pylon is used, and combined with the real-time cable
tension of cable C36, the real-time deflection of the middle span of the bridge is
calculated. The deflection at time t =0 s is set as the original value, and the real-time
deflections are all relative to the original deflection. The measured real-time elevation
of GPS is shown in Fig. 27. A comparison between the girder’s deflection obtained by
the proposed method and GPS results is shown in Fig. 28.
From Figs. 27 and 28, it can be seen that during most of the testing time, the
deflection of the testing point swings within the scope of -10 mm to 10 mm, which
indicates that there is no heavy vehicle passing through the bridge. The peak points of
the curve indicate the times at which heavy vehicles passed through the mid-span of the
bridge. At the instant of t=1000 s, there is a maximum GPS measured deflection of 65.0
mm, while the calculated deflection is 59.8 mm, providing a relative error of 8.0%.
Though there are some errors between the proposed method and GPS method, their
trends are basically the same, which indicates that the proposed method can be used to
measure the real-time deflection of the girder.

22
Fig. 27. The measured real-time elevation Fig. 28. Comparison between the results
by using GPS observation of GPS and proposed method

6. Conclusions
Monitoring the real-time deflection of a bridge is of great significance for assessing
the bridge’s behavior and performance. In this study, a new method to measure the
real-time deflection of cable-stayed bridge based on the variations of cable tension is
presented. The new method includes four steps: First, the relationship between the cable
tension variation and the deflection of the anchor point is established based on the
theory of a beam with elastic foundation. Second, the acceleration signals of the cables
are acquired from accelerometers installed on the cables. Third, the time-varying cable
tension of the stay cable is obtained using the HHT method and the unified formula of
cable tension estimation. Fianlly, the variation of real-time deflection of the anchor
point between the cable and girder is calculated by using the equations derived. The
feasibility and effectiveness of the proposed method are verified by a numerical
example and a practical case study. The results of numerical example show that the
errors between the calculated deflection using the proposed method and the FEA result
are within 6%. Then, the girder’s deflection of the He-dong Bridge in Guangzhou which
has been established a structural health monitoring system is measured in real-time
using the proposed method. Through comparisons with the GPS measured results, the
results produced by the proposed method exhibit the same trends, with relative errors
less than 10%, which indicate that the proposed method can be used in practical
engineering. Compared with the traditional deformation measuring methods, the

23
developed method is a much easier and more economical method for real-time
deflection measurements because the tensions in many of the cables are already
auto-monitored and no additional equipment is needed in the structural health
monitoring system of the cable-stayed bridge. In addition, the proposed method is more
convenient because of its availability in all-weather conditions.

Acknowledgments
This research is financially supported by the Pearl River S&T Nova Program of
Guangzhou (Grant No. 201710010147), the Technology Planning Project of Guangdong
Province (Grant No. 2016B050501004) and Technology Planning Project of Guangzhou
Education Department (Grant No. 1201620446).

References
[1] E.W. Beshr, M.R. Kaloop, Monitoring bridge deformation using auto-correlation adjustment
technique for total station observations, Positioning. 4 (2013)1-7.
[2] D. Qiu, Y. Ding, Deformation monitoring of bridge approach using terrestrial laser scanning, in:
Facing the Challenges – Building the Capacity, FIG Congress, 2010, pp.11-16.
[3] Stöhr S, Link M, Rohrmann RG, Rücker W. Damage detection based on static measurements on
bridge structures. In: Conference & exposition on structural dynamics 2006 IMAC-XXIV, 30
January – 2 February 2006, St Louis, Missouri, USA.
[4] Nguyen V.H., Schommer S., Maas S., Zürbes A., Static load testing with temperature
compensation for structural health monitoring of bridges, Eng. Struct. 127 (2016) 700-718.
[5] Avsar O, Akca D, Altan O. Photogrammetric deformation monitoring of the second Bosphorus
Bridge in Istanbul. In: The international archives of the photogrammetry, remote sensing and
spatial information sciences, volume XL-5, 2014 ISPRS technical commission v symposium,
23 – 25 June 2014, Riva del Garda, Italy.
[6] O. Ogundipe, G.W. Roberts, C.J. Brown, GPS monitoring of a steel box girder viaduct. Struct.
Infrastruct. Eng. 10 (2014) 1-16.
[7] M.R. Kaloop, H. Li, Monitoring of bridge deformation using GPS technique, Ksce. J. Civ. Eng.
13 (2009)423-431.
[8] M. Figurski, M. Gałuszkiewicz, M. Wrona, A bridge deflection monitoring with GPS. Artificial
Satellites. 42 (2007)229-238.
[9] J. Guo, L. Xu, L. Dai, M. Mcdonald, J. Wu, Y. Li, Application of the real-time kinematic global
positioning system in bridge safety monitoring, J. Bridge. Eng. 10(2005)163-168.

24
[10] X. Meng, A.H. Dodson, G.W. Roberts, Detecting bridge dynamics with GPS and triaxial
accelerometers, Eng. Struc. 29 (2007)3178-3184.
[11] Z. Mat Amin. Experimental detection of the Penang bridge vibration with real time kinematic
gps, in: International Symposium & Exhibition on Geoinformation, 2003, pp.1-9
[12] V. Ashkenazi, G. W. Roberts, Experimental monitoring of the Humber bridge using GPS, Civ.
Eng. 120 (1997)177-182.
[13] S. Kashima, Y. Yanaka, S. Suzuki, et al, Monitoring the Akashi Kaikyo bridge: first experiences,
Struc. Eng. Int. (2)(2001)20-123.
[14] G. Roberts, C. Atkins, C.J. Brown, X. Meng, Further results from using GPS to monitor the
deflections of the forth road bridge, in: Strategic Integration of Surveying Services.
5(2007)1-11.
[15] M.R. Kaloop, H. Li, Sensitivity and analysis GPS signals based bridge damage using GPS
observations and wavelet transform, Measurement. 44 (2011)927-937.
[16] C.O. Yigit, M.Z. Coskun, H Yavasoglu, The potential of GPS precise point positioning method
for point displacement monitoring: a case study, Measurement. 91(2016)398-404.
[17] H.F. Xiang, Advanced theory of bridge structures (Second edition), China communications
press. 2013, Beijing, China. (In Chinese)
[18] N. J. Gimsing, C. T. Georgakis, Cable supported bridges: concept and design, third ed., A John
Wiley & Sons, Ltd., Publication, Chichester, 2012
[19] D.S. Jiang, S. Li, Design and research structural health monitoring system for an existing bridge
based on optical fiber sensing technology, in: The 2nd international conference on structural
condition assessment, monitoring and improvement, 2007, pp 959–965.
[20] B.H. Kim, T. Park, Estimation of cable tension force using the frequency-based system
identification method, J. Sound. Vib. 304 (2007)660-676.
[21] Y.H. Huang, J.Y. Fu, R.H. Wang, Q. Gan, A.R. Liu, Unified Practical Formulas for Vibration
Based Method of Cable Tension Estimation, Adv. Struct. Eng. 18 (2015) 405-422.
[22] Irvine, H. M. Cable structures, MIT Press, Cambridge, MA, 1981
[23] Casas, J.R. A combined method for measuring cable forces: the cable-stayed Alamillo bridge,
Spain, Struct. Eng. Int. 4 (1994) 235-240.
[24] Huang N E. Hilbert-Huang transform and its applications. Vol. 16. World Scientific, 2014.
[25] Huang N E, Wu Z. A review on Hilbert‐Huang transform: Method and its applications to
geophysical studies. Rev. of Geophys. 46 (2008) 1-23.
[26] ANSYS, Multiphysics. Release 14.5, Ansys Inc., Canonsburg, PA, USA, 2014

25
 Highlights

1. A new measurement method for real-time deflection of cable-stayed bridge is

proposed.

2. The proposed method can easily realize automatic and continuous measuring.

3. The proposed method is a low-cost and high-efficiency method.

4. The proposed method is more convenient for its availability in all-weather condition.

5. Feasibility and effectiveness of the method are validated by real engineering

application.

26