32 Int. J. Structural Engineering, Vol. 8, No.

1, 2017

Assessment of damper performance in controlling

cable vibrations using a reliability-based framework

Seyed Ali Mohammadi, Shaohong Cheng*

and Faouzi Ghrib
Department of Civil and Environmental Engineering,
University of Windsor,
401 Sunset Ave., Windsor,
Ontario, N9B 3P4, Canada
Email: mohamma3@[Link]
Email: shaohong@[Link]
Email: fghrib@[Link]
*Corresponding author

Abstract: Owing to their long flexible nature and low intrinsic damping, bridge
stay cables are prone to various types of wind-induced vibrations, among which
the rain-wind-induced vibration is most frequently observed on site. External
dampers are widely used to control such unfavourable cable oscillations and
their effectiveness in suppressing large-amplitude cable vibrations was
addressed in many studies using deterministic approaches. However, the
mechanical and/or physical properties of cables and the attached dampers could
not only deviate from their respective nominal design values at a given design
point, but also vary considerably during the lifetime of a cable-stayed bridge
and thus affect damper efficiency. Hence, for a realistic damper performance
assessment, these uncertainties should be taken into account. The objective of
this paper is to present a time-variant reliability-based framework model to
assess how uncertainties in the structural parameters of a cable-damper system
would influence the time specific reliability performance of an external damper
yielded from the current design practice.

Keywords: vibration control; stay cable; time-variant reliability analysis; limit

state function; rain-wind-induced vibration; damper.

Reference to this paper should be made as follows: Mohammadi, S.A.,

Cheng, S. and Ghrib, F. (2017) ‘Assessment of damper performance in
controlling cable vibrations using a reliability-based framework’, Int. J.
Structural Engineering, Vol. 8, No. 1, pp.32–59.

Biographical notes: Seyed Ali Mohammadi is a PhD candidate in Structural

Engineering at University of Windsor, Ontario, Canada. He received his
BSc and MSc in Civil Engineering from Amirkabir University (Tehran
Polytechnic), Iran. His research interests include structural analysis, design and
rehabilitation, wind engineering and aerodynamics, earthquake engineering,
and structural reliability analysis. He is a student member of American Society
of Civil Engineers.

Shaohong Cheng is an Associate Professor in the Department of Civil and

Environmental Engineering, University of Windsor, Windsor, Ontario, Canada.
She graduated in Structural Engineering from Tongji University, China. She
received her MASc in Structural Engineering from Tongji University, China
and her PhD in Civil Engineering from Carleton University, Canada. Her

Copyright © 2017 Inderscience Enterprises Ltd.

 Assessment of damper performance in controlling cable vibrations 33

research interests include structural dynamics, vibration and control,

aerodynamic response and wind-resistant design of bridges and buildings,
fluid-structure interaction, and strengthening of concrete structures. She is a
member of American Society of Civil Engineers.

Faouzi Ghrib is an Associate Professor in the Department of Civil and

Environmental Engineering, University of Windsor, Windsor, Ontario, Canada.
He graduated in Civil Engineering from the National Engineering School of
Tunis, Tunisia. He received his MASc in Applied Mechanics from the National
Engineering School of Tunis, Tunisia and his PhD in Civil Engineering from
École Polytechnique de Montréal, Canada. His research interests include
numerical simulation of piezoelectric materials, damage assessment of quasi-
brittle materials (ceramics, concrete, etc.), numerical simulation of fatigue
process under multi-axial loading stress, and seismic analysis of large
structures (dams, bridges, etc.). He is a member of the American Society of
Civil Engineers.

1 Introduction

Stay cables on cable-stayed bridges are sensitive to dynamic excitations induced by

various sources due to their low intrinsic damping and long flexible nature. In particular,
a study by the US Federal Highway Administration (Kumarasena et al., 2007) indicates
that majority of the large amplitude cable vibration field incidents are associated with
rain-wind-induced vibrations (RWIV). As revealed from numerous site reports and wind
tunnel studies (e.g. Hikami and Shiraishi, 1988; Matsumoto, 1998; Matsumoto et al.,
2003; Chen et al., 2004), for cables satisfying certain orientation and location conditions,
when mild wind is combined with moderate rain, water rivulet could form on cable
surface as a result of a sensitive equilibrium between gravity, capillary and aerodynamic
forces. Existing studies show that the presence of upper water rivulet would not only alter
geometric shape and aerodynamic feature of the cable cross-section, but its
circumferential oscillation could be aerodynamically coupled with flexural oscillation of
the cable and lead to negative aerodynamic damping (Yamaguchi, 1990). Over the last
two decades or so, besides an attempt to apply classical galloping theory to explain the
mechanism (Yamaguchi, 1990; Geurts et al., 1998; Geurts and Staalduinen, 1999), the
possible role of water rivulet thickness and its link with the rivulet motion speed
(Flamand, 2001), as well as the initial and instantaneous rivulet position (Gu and Lu,
2001) in exciting RWIV were investigated. Gu et al. (2009) carried out wind tunnel tests
to obtain aerodynamic forces acting on the cable and upper rivulet and developed a
theoretical model by considering the cable in-plane motion and rivulet tangential motion
in the analysis. Although encouraging progress has been made to better understand the
phenomenon, current knowledge is still inadequate to fully explain the underlying
physics associated with its excitation mechanism.
A practical criterion proposed by Irwin (1997) suggests that RWIV can be suppressed
to a harmless level if the Scruton number is greater than 10, i.e.
mζ
Sc = ≥ 10 (1)
ρD 2
34 S.A. Mohammadi et al.

where m is the cable mass per unit length, ζ is the damping ratio of the cable, ρ is the air
density, and D is the cable diameter. Consider an example of a typical bridge stay cable
having D = 20 cm and m =100 kg/m, this is equivalent to a structural damping
requirement of approximately 0.5%. The criterion is not only supported by experience in
Japan (Saito et al., 1994; Yamada, 1997) and France (Virlogeux, 1998), but also
confirmed by recent field monitoring program in China (Chen et al., 2004). In addition,
when revising design guideline of bridge stay cables, this criterion is recommended by
the Post-Tensioning Institute (PTI publication, 2001) to be used in practical design to
predicate the required damping in a stay cable to mitigate RWIV.
Besides modifying cable surface configuration and adding cross-ties, external
dampers are commonly used to suppress unfavourable cable motions on site. The
effectiveness of a damper design in controlling cable vibrations has been studied by many
researchers in terms of the structural modal damping level achieved by a stay cable when
attached with an external damper. Kovacs (1982) identified the existence of an optimal
damping in a taut cable-damper system, which was confirmed by a number of researchers
(Yoneda and Maeda, 1989; Uno et al., 1991; Pacheco et al., 1993; Krenk, 2000). In
particular, Pacheco et al. (1993) simplified the procedure of designing viscous dampers of
stay cables by deriving a universal damping estimation curve, which allowed relating the
modal damping level of a damped taut cable directly to the size and the location of a
damper. By solving the same problem using complex mode shapes, Krenk (2000)
developed an analytical form of this universal curve. Further, Mehrabi and Tabatabai
(1998), as well as Krenk and Nielsen (2002) presented a refined solution to a typical
cable-damper system by including the influence of cable sag and cable bending stiffness
in the formulation. The accurate asymptotic formula of the structural modal damping
ratio of a general cable-damper system was analytically derived by Fujino and Hoang
(2008). The study resulted in an explicit evaluation of reductions in the damper
effectiveness due to influential parameters such as the sag and the bending stiffness of a
cable and the stiffness of a damper support. Recently, Cheng et al. (2010) proposed an
energy-based method to evaluate the dynamic behaviour of a cable-damper system. A set
of damping estimation curves were developed for the practical parameter ranges of bridge
stay cables. These curves can be utilised to relate a specific damper design to the
corresponding equivalent structural modal damping of a damped stay cable.
At present, when an external damper is used to control RWIV of a stay cable, upon
considering the limitations of its installation location, the size of the damper is typically
selected using either the damper design curve or the design formula proposed in the
above mentioned studies so that the resulted equivalent structural damping ratio of the
damped cable would satisfy the damping requirement for suppressing RWIV proposed by
Irwin (1997). However, it is worth noting that most existing studies are based on the
assumption that the parameters of a cable-damper system are deterministic; whereas in
practice, it is expected that the system properties could not only deviate from their
respective nominal design values at a given design point, but also vary over time due to
changes in ambient conditions. For example, tension in a stay cable may increase or
decrease during its lifetime because of creep and shrinkage in concrete deck and/or pylon
and cable slacking (Au and Si, 2012). Thermal expansion or contraction of the fluid in a
viscous damper may result in excessive internal pressure in a damper or the formation of
a vacuum inside the damper, which would change the fluid properties or even cause
leakage and thus affect the damper capacity and degrade its efficiency. Therefore, the
assumption underlying deterministic approaches does not comply with the most common
 Assessment of damper performance in controlling cable vibrations 35

practical situations. For a rational assessment of the life-long performance of a cable-

damper system and for a better understanding of the impact of the system parameter
uncertainties on the vibration control efficiency of an external damper designed
according to the current practice, it is crucial to carry out a time-dependent probabilistic
analysis to consider these uncertainties. This can be achieved by combining the existing
time-invariant reliability analysis methods with the time varying characteristics of system
parameters. In the present paper, we propose to formulate the reliability of cable-damper
systems as a time-dependent up-crossing problem. The time dependency appears in the
form of propagation kinetics of structural properties decaying. Because only degradation
mechanisms are considered, the solution of the time-variant reliability problem can be
treated as a series of time specific reliability problem and solved using classical time
invariant reliability tools (Pagnini, 2010).
A number of time-invariant reliability analysis approaches are available in the
literature for the assessment of structural performance. These include the Monte Carlo
simulation (MCS) technique, the first and the second order reliability methods. Although
there might be many reasons for choosing one method over another, MCS is generally
selected for its simplicity in the formulation and its capability to handle problems with
complex limit state functions. However, the computational cost of this approach is
deemed very high when being used for parametric studies (Frangopol and Maute, 2003).
Mohammadi et al. (2011) studied the time specific performance of external dampers in
suppressing RWIV of bridge stay cables. The MCS was conducted by treating the cable
tension and the damper capacity as two random variables at a given design point. The
limit state function of a horizontal taut cable-damper system was expressed explicitly in
terms of an equivalent system modal damping ratio to restrain the occurrence of rain-
wind-induced cable vibrations. The high computational cost associated with MCS
motivated the authors of this paper to propose an alternative solution. In a subsequent
study, Mohammadi et al. (2013) proposed the use of the two-point adaptive nonlinear
approximations (TANA) method, which is a first-order-second-moment (FOSM) method,
to assess the time specific performance of an inclined sag cable equipped with a linear
viscous damper in resisting rain-wind-induced cable vibrations.
In this study, a time-variant reliability-based framework model will be proposed to
include the time-varying characteristics of system properties in the performance
assessment of a cable-damper system. The influence of structural parameter uncertainties
in a cable-damper system on the current deterministic-based damper design practice will
be evaluated. The problem will be formulated as a limit state up-crossing scenario to
determine the time specific reliability of a typical cable-damper system under rain-wind-
induced cable vibrations. The cable tension and the damper capacity are treated as two
independent time-varying random variables. The second-order-second-moment (SOSM)
method will be selected to solve this probabilistic problem. The accuracy of the applied
methods for analysing the reliable performance of a cable-damper system in resisting
RWIV will be investigated through an illustrative example. A set of sample reliability-
based damper design curves are proposed using a non-dimensional form of system
parameters. The relation between the time specific system reliability-index and various
cable-damper system parameter uncertainties is described. In addition, the potential
application of the proposed time-variant reliability-based framework model and results to
the development of maintenance strategy of a cable-damper system over its lifetime will
be demonstrated.
36 S.A. Mohammadi et al.

2 Formulation of time-variant reliability-based performance assessment

method

2.1 Control parameters of a cable-damper system

A schematic model of a typical inclined sag cable attached transversely with a linear
viscous damper is depicted in Figure 1. The coordinate system is defined such that the
X-axis is along the cable chord and the Y-axis is the perpendicular direction. The cable
has a length L, a mass per unit length m, a finite flexural rigidity EI, an inclination angle
θ, and carries a tension T. A damper is attached to the cable at a distance Ld from the
lower end support. The damping coefficient is denoted by c.

Figure 1 A schematic model of an inclined sag cable with a transverse linear viscous damper

The sagging effect in a real cable due to its self-weight is usually considered in terms of a
sag d at the cable mid-span. The sag is given by d = mgL2cosθ/(8H), where H is the cable
tension component along its chord direction, and g is the gravitational acceleration. A
non-dimensional sag parameter is defined by Irvine and Caughey (1974) as:
2
⎛ 8d ⎞ L
λ2 = ⎜ ⎟ (2)
⎝ L ⎠ [ e ( EA)]
HL /

where EA is the axial rigidity of the cable, and Le = L[1 + 8(d/L)2] is the static (stretched)
length of the cable. It should be noted that, as indicated by equation (2), the
non-dimensional sag parameter λ2 does not only include the effect of cable sag, but also
that of the cable axial stiffness and inclination. For simplicity, it is assumed that the
variation of λ2 is associated with the variation of the sag level. For example, λ2 = 0
represents physically the case of a taut cable, and larger λ2 values correspond to more
flexible cables. In practice, stay cables on cable-stayed bridges typically have λ2 values in
the order of ten or smaller (Johnson et al., 2002). This range will be considered in the
current study.
 Assessment of damper performance in controlling cable vibrations 37

For the cable-damper system shown in Figure 1, when damper location is selected,
the most influential factors that would affect the efficiency of a viscous damper include
the cable tension, T, and the damper capacity, c. As mentioned earlier, these two system
parameters could vary over the life time of a bridge due to change in the ambient
conditions such as creep and shrinkage in concrete deck and/or pylon, cable slacking, and
temperature-induced variation in damper fluid properties etc. Further, even at a given
time point, the actual cable tension and damper capacity in the system could deviate from
their respective nominal values. These facts lead us to consider cable tension and damper
capacity as time-dependent random variables in the current study, the uncertainties of
which are assumed to come mainly from the above two sources. The former type of
uncertainty can be included in the analysis by incorporating the time-varying
characteristics of cable tension and damper capacity over the bridge life time, which
typically can be collected from field monitoring program; whereas the latter type is
counted at a given point of time, by assuming that both the cable tension and the damper
capacity follow independent normal distributions, with mean values equal to their
respective nominal design values at the specific design point, and the standard deviations
defined by the specified coefficient of variations (COVs), as shown in Figure 2.

Figure 2 Schematic diagram showing uncertainty of cable tension and damper capacity

The representation of the dynamic response of a typical cable-damper system can be

simplified by introducing the non-dimensional form of cable-damper system parameters
similar to those defined by Mehrabi and Tabatabai (1998). They are defined as follows:
a Damper location parameter Γd = Ld/L. Due to the practical limitation on the damper
installation position, the maximum damper location parameter is restricted to 6%
(Γd ≤ 0.06).

b Cable bending stiffness parameter ξ = L H / EI , which reflects the flexibility level

of the cable. If we assume the flexural rigidity EI of a stay cable remains constant,
the cable bending stiffness parameter ξ can then be directly related to the cable
tension component, H, along its chord direction. Accordingly, for high ξ values, the
cable acts similar to a taut string.
38 S.A. Mohammadi et al.

c Damping parameter ψ = (πc)/(mLω1s), where ωls = (π / L) H / m is the first modal

frequency of a taut string equivalent to the cable. Hence, the definition of the
damping parameter can be simplified as ψ = c⏐ mH .

In the current study, the practical ranges of the non-dimensional parameters ξ and ψ are
selected based on the bridge stay cable database developed by Tabatabai et al. (1998).
The non-dimensional bending stiffness parameter ξ is limited to be no greater than 400,
whereas the damping parameter ψ is selected between 0 and 60.

2.2 Development of time-dependent limit state function (LSF)

If we assume that the vibration suppression effect provided by an external damper to the
attached cable can be defined as an equivalent structural damping within the cable, the
general form of the LSF for a cable-damper system can be proposed as the difference
between the available structural damping ratio in a cable-damper system, ζs, and the
damping demand of a damped cable, ζr, to avoid large amplitude vibrations. Since the
available structural damping ratio ζs in a cable-damper system does not only depend on
the location and the capacity of the external damper, but also relates to the amplitude of
cable vibration, the proposed form of the LSF is thus a function of the cable tension T
(in terms of its chord component H) and damper capacity c, or equivalently the
non-dimensional parameters ξ and ψ. In the current study, the cable tension and the
damper capacity are treated as two independent time-varying random variables; the LSF
thus also becomes a time-dependent function, i.e.

g ( H (t ), c(t ) ) = g ( ξ ( H (t ) ) , ψ ( c (t ), H (t ) ) ) = ζ s (t ) − ζ r (3)

where g(∙) is the time-dependent LSF of the cable-damper system. Because only the
structural properties are expected to decay and the stochastic nature of the RWIV is
simplified, the time-variant reliability problem associated with equation (3) can be solved
as a series of classical time-independent reliability problems. The underlying assumption
is to consider failure events at any time of the bridge lifetime to be independent.
To develop the full expression for ζs(t), we consider the in-plane transverse vibration
of the cable-damper system shown in Figure 1 at a specific design point (H, c), which can
be described by the following partial differential equation (Fujino and Hoang, 2008):
∂ 2 η( X , t ) ∂ 2 η( X , t ) d 2Y ∂ 4 η( X , t )
H −m +h − EI = f c δ ( X − Ld ) (4)
∂X 2
∂X 2
∂X 2
∂X 4
where η(X,t) is the in-plane transverse displacement of the cable motion in the
Y-direction; m is the cable mass per unit length; H is the cable tension component along
its chord direction at the given design point; h(t) is the additional tension in the cable due
to vibration; EI is the bending stiffness of the cable; fc(t) is the in-plane damping force of
the damper and is a function of damper capacity c, and δ(X-Ld) is the Dirac delta
function. Fujino and Hoang (2008) proposed an asymptotic solution to equation (4), of
which the equivalent nth modal damping ratio, ζn, of a damped cable is identified directly
from the imaginary part of the complex natural frequency solution:
 Assessment of damper performance in controlling cable vibrations 39

ζn η f ηsn ηn
= R f Rsn (5)
( η f ηsn ηn )
2
Ld / L

where ηn = nπ ( Ld / L)c / Hm is the dimensionless damper coefficient parameter of

mode n of a corresponding horizontal non-flexural taut cable-damper system (Krenk,
2000); Rf and Rsn are the reduction factors of the maximum modal damping ratio due to
respectively cable flexural rigidity and sag as defined by Krenk and Nielsen (2002); and
the modification factors for ηn due to the influence of cable flexural rigidity and sag are
represented by ηf and ηsn, respectively. A simplified form of ζn can be obtained when the
effects of sag and/or flexural rigidity are neglected. For example, by taking
Rsn = ηsn = Rf = ηf= 1, equation (5) can be reduced to ζ n = ηn / (l + ηn2 ) ⋅ ( Ld / L), which is
the equivalent nth modal damping ratio of a horizontal taut non-flexural cable
corresponding to a specific design point (H, c). By including the time-varying
characteristics of cable tension and damper capacity into formulation, the equivalent nth
modal damping ratio of the studied damped cable becomes time dependent, i.e.
ζ n (t ) = ηn (t ) / [l + ηn2 (t )] ⋅ ( Ld / L).
To complete the definition of the limit state function, the damping demand is
required. As reviewed earlier, although the mechanisms associated with RWIV have not
been fully understood yet, experimental results (Saito et al., 1994; Yamada, 1997) and
field experience (Virlogeux, 1998; Chen et al., 2004; Casas and Aparicio, 2010) show
that the minimum amount of required damping predicted by a Scruton number-based
empirical criterion (Irwin, 1997), as given in equation (1), is sufficient to effectively
mitigate RWIV. Considering the fact that this criterion is also recommended by PTI (PTI
Publication, 2007) when revising the design guideline of bridge stay cables, in the current
study, the damping demand for suppressing RWIV will be defined based on this criterion.
Establishing the onset of instability for RWIV as a function of a damping demand by
using the Scruton number limit, Sc,lim, equals to 10, the minimum required structural
damping ratio, ζr, of a damped cable can thus be expressed as:
Sc , lim ρD 2 10 ρD 2
ζγ = = (6)
m m
For any specific cable vibration mode n, by substituting equation (5) for ζs(t) and the
minimum amount of required damping defined in equation (6), the time-dependent LSF
in equation (3) can be written in the form:
⎛ η f ηsn ηn (t ) L ⎞ ⎛ 10 ρD 2 ⎞
g ( H (t ), c(t ) ) = ⎜ R f Rsn ⋅ d ⎟−⎜ ⎟ (7)
⎜ 1 + ( η f ηsn ηn (t ) ) L ⎟⎠ ⎝ m ⎠
2
⎝
The failure domain is defined by the region g(H(t), c(t)) < 0. Failure in this case refers
mainly to the loss of system aerodynamic stability exhibited through large amplitude
cable vibrations.
40 S.A. Mohammadi et al.

2.3 Structural reliability analysis methods

A number of reliability analysis methods, including the MCS, the FOSM, and the SOSM
methods are available to evaluate the reliability of structures. The MCS is commonly
used as a reference reliability analysis method (Frangopol et al., 2007). To reduce the
computational cost, the first and the second-order reliability methods (FORM/SORM) are
preferred to assess the probabilistic response and safety of structures by approximating
the limit state function (Choi et al., 2007). These methods require a search for the most
probable point of failure (MPP) on the failure surface in the standard normalised space.
FORM employs a linear approximation for the limit state function at the MPP, whereas
SORM features an improved accuracy by using a second order approximation.
In the current study, the two-point adaptive nonlinear approximation (TANA) method
is selected among various FORMs due to its established superiority. The method is based
on the definition of an improved linear approximation of LSF, g ( X ) by including a
nonlinear index, r, in the iteration process, i.e.
n
∂g ( X k )
g ( X ) = g ( X k ) +
1
r ∑x 1− r
i, k
∂xi
(x r
i − xir, k ) (8)
i =1

where X = {x1, x2,…,xn}T is the vector of the system design variables x1, x2,…,xn, Xk =
{x1,k, x2,k,…,xn,k}T is the vector X at the kth iteration, g(Xk) is the value of the nonlinear
LSF at Xk, xi,k is the ith element of X in the kth iteration, and n is the total number of
system random variables. In the current formulation, X = {x1 = H, x2 = c}T and n=2.
The solution uses an iterative process to minimise the discrepancy between the exact
nonlinear LSF and the approximate linearised LSF. The nonlinear index, r, can be
determined through a process called intervention method by using the values of LSF and
its gradients associated with the current and the last iterations as:
⎧⎪ n
∂g ( X k ) ⎫⎪
g ( X k −1 ) − ⎨ g ( X k ) +
1
r ∑x 1− r
i, k
∂xi
(x r
i , k −1 )
− xir, k ⎬ = 0 (9)
⎩⎪ i =1 ⎭⎪
However, the truncation error from the first-order Taylor series approximation used in
this method might be large. In particular, in the case of highly nonlinear LSFs associated
with large curvature failure surfaces, this approximation would reduce the accuracy of the
reliability assessment considerably. In addition, TANA method can only predict one MPP
in each iteration process. Hence, it is not capable of handling problems with multi-design
points in the LSF.
To overcome these limitations, second-order reliability methods (SORMs) are
selected. In these methods, when approximating the limit state function at MPP, the
original limit state surface is replaced by a second-order approximate one with the
addition of a second-order derivative term to the Taylor series expansion:
1
g (U ) ≈ g (U *) + ∇g (U *) (U − U *) + (U − U *) ∇ 2 g (U *)(U − U *)
T T
(10)
2
where the system variables are presented in a standard normalised U-space, U*is the
coordinate of the MPP, g (U ) is the approximate LSF, g(U*) is the LSF value at the
 Assessment of damper performance in controlling cable vibrations 41

MPP, ∇g(U*) is the gradient of the LSF at the MPP, and ∇2g(U*) represents the second-
order derivatives of the LSF at the MPP.
Wang and Grandhi (1995) simplified the calculations of the failure probability Pf by
proposing an adaptive approximation method for the SORM. In this method, the
nonlinear approximate LSF, g ( X ) developed in the TANA method is used to calculate
the second order derivatives of the LSF at the MPP, which considerably reduces the
computation time. In addition, it improves the accuracy of failure probability calculations
compared to the first-order reliability methods. The failure probability Pf can be
computed by either the Breitung’s formulation or the Tvdet’s formulation (Choi et al.,
2007).
The SORM with adaptive approximations method is applied to the studied cable-
damper system in the current work due to its formulation simplicity.

3 Reliability-based performance assessment of a cable-damper system

3.1 Selection of reliability analysis method

In this section, the reliability methods including TANA and SORM with adaptive
approximations are applied to a damped bridge stay cable to assess its time specific
capacity in resisting RWIV. Results from the MCS are used as a benchmark to compare
the efficiency of the two numerical methods in predicting the safe performance of the
system. Besides, how uncertain deviation of cable tension and damper size from their
respective nominal design values at a specific design point would affect the safe
performance of a damped cable will be investigated.
A typical bridge stay cable attached with an external linear viscous damper is
considered. The properties of the cable are assumed to be: length L = 150 m, unit mass
m = 72 kg/m, cable diameter D = 0.2 m and bending stiffness EI = 5420 kN∙m2 (including
grout and cover pipe). The damper is located at 6% L from the lower cable anchorage. At
any design point, the cable chord tension, H, and the damper damping coefficient, c, are
assumed to be random variables having independent normal distributions, with mean
values being their respective nominal design values. If the design point considered here
corresponds to a nominal cable chord tension H = 3700 kN and a nominal damper
damping coefficient c = 50 kN∙s/m, we propose to evaluate the time specific performance
of the system under RWIV at this design point by assuming the COV for H and c being
10% and 5%, respectively.
The non-dimensional parameters of the studied cable-damper system at the given
design point are as follow: the cable bending stiffness parameter, ξ = 123.94; the damping
parameter, ψ = 3.06; the damper location parameter, Γd = 0.06; and the sag parameter,
λ2 = 0.485.
The equivalent first modal damping ratio of the system, ζs, is calculated according to
equation (5). The non-dimensional damper coefficient parameter for the first mode of a
corresponding horizontal non-flexural taut cable-damper system is found to be
η1 = 0.577. By considering the effects of cable sag and flexural rigidity, the reduction
factors and modification factors in equation (5) are respectively ηsn = 1.020, Rsn = 0.948,
ηf = 0.798, and Rf = 0.938. Hence, the equivalent first modal damping ratio of the system
is ζs = ζ1 = 2.055%. On the other hand, if the air density is taken as ρ = 1.29 kg/m3, based
42 S.A. Mohammadi et al.

on equation (6), the minimum required structural damping ratio for such a cable-damper
system to prevent RWIV is ζr = 0.717%. The value of the LSF of the current cable-
damper system is determined according to equation (7) as:
g ( X ) = g ( x1 = H = 3, 700 kN, x2 = c = 50 kN ⋅ s/m ) = 1.339%

The TANA method is applied first to compute the reliability index of the studied damped
cable, with the iteration results summarised in Table 1. The convergence tolerance is set
to be 0.001. As can be seen from the table, the reliability index converges after four
iterations at point H4 = 4,410.8 kN, c4 = 171.3 kN∙s/m. This point is the MPP with
corresponding reliability-index equals to β4 = 11.79, and the associated probability of
failure equals to Pf = 5.6 × 10–16.
Table 1 Summary of reliability analysis results by TANA

Iteration no. 1 2 3 4
H (kN) 3,700.0 5,230.8 4,730.2 4,410.8
c (kN∙s/m) 50.0 111.8 182.4 171.3
g(Xk) 0.00134 0.0188 0.0081 0.00013
(∂g/∂H)|μH –2.12 × 10–7 –9.52 × 10–8 –1.32 × 10–7 –6.18 × 10–7
(∂g/∂H)|μc 2.62 × 10–5 8.32 × 10–6 4.11 × 10–5 7.17 × 10–5
β 10.83 11.47 11.77 11.79
ε – 0.1099 0.0260 0.0009

It is worth mentioning that although by including a non-linear index r in the formulation

and updating the approximate linear LSF during iterations, the TANA method has the
advantage over other FOSM approaches in evaluating the reliability of nonlinear systems,
the r value depends on the selected design point. For a cable-damper system, if the
performance is assessed at different design points, i.e. different cable tension and damper
capacity within the scope of the current study, the corresponding LSF would have
different curvatures. Thus, the associated nonlinear indices would be different. For
example, in the current example, the design point is H = 3,700 kN and c = 50 kN∙s/m, and
the corresponding nonlinearity index is r = 1.015. This suggests that a linear
approximation function would be able to describe the degree of nonlinearity for the LSF
at this point. While on the contrary, if the damper coefficient for the design point changes
to c = 200 kN∙s/m, the computation of the nonlinear index would yield r = 12.241, which
indicates a highly nonlinear LSF. Thus, if the linear approximation approach is used to
describe the LSF, the reliability assessment results could be inaccurate and misleading. In
addition, the selection of the initial trial point for the iteration will also affect the
convergence rate especially for a design point associated with large curvature LSF. This
is due to the fact that the nonlinear index r would remain the same for the rest of the
iteration process, which would affect the efficiency of calculations.
The accuracy of the TANA method can be improved by using the SORM with
adaptive approximations. The reliability assessment of the current cable-damper system
is continued by transforming the problem to a normalised space using an
orthogonalisation of the following matrix:
 Assessment of damper performance in controlling cable vibrations 43

⎡ ∂g (U *) / ∂H ∂g (U *) / ∂c ⎤
⎢ − ∇g (U *) − ∇g (U *) ⎥
⎢ ⎥
⎢⎣ 0 1 ⎥⎦

where ∂g(U*)/∂H and ∂g(U*)/∂c are the gradients of the limit state function g(∙)
evaluated at the MPP defined by H and c, and ⏐∇g(U*)⏐ is the magnitude of the gradient
function. The curvature corresponding to parameters H and c when the orthogonal
Y-space is used is calculated for the approximate LSF, g ( X ) at the MPP using the
method proposed by Wang and Grandhi (1995), which resulted in a reliability-index of
β = 11.79 and the corresponding probability of failure of Pf = 9.1 × 10–16.
To perform MCS, a set of 108 one-dimensional inputs are generated randomly for
each parameter and then the two variables H and c are combined together to generate a
total of 1016 runs. For each run, equation (5) is used to evaluate the equivalent first modal
damping ratio of the cable-damper system. The reliability-index is determined from the
ratio between the mean and the standard deviation of the limit state function, which is
β = 12.04. The corresponding probability of failure is Pf = 1.1×10–15. A comparison of
the reliability analysis results in terms of the probability of failure, Pf, shows that the
SORM with adaptive approximations method would yield more accurate reliability
evaluation results of the given cable-damper system than the TANA method. This is
mainly due to its capability of handling nonlinear LSF. In addition, it is computationally
more efficient than the MCS. Therefore, the SORM with adaptive approximations
method will be used in the rest of the paper to predict the performance of cable-damper
systems.

3.2 Impact of structural parameter uncertainty at a given design point

In this section, the proposed time-dependent reliability-based framework model will be

applied to the described cable-damper system in the previous section to assess the impact
of the structural parameter uncertainty at a given design point on the damper
performance. As mentioned earlier, in the current study it is assumed that cable tension
and damper size follow independent normal distributions at any design point, with the
mean being their respective nominal design value, and the standard deviation defined by
the COV. The design point considered here corresponds to a cable chord tension
H = 3,700 kN and a damper capacity c = 50 kN∙s/m. To isolate the respective uncertainty
effect of H and c, in the analysis, the response of the damped cable is evaluated by
varying COV of one variable while keeping that of the other at a constant level. A set of
sample results are presented in Figure 3. The solid line in the figure describes the
variation of system reliability index β when the damper size uncertainty, COV-c,
increases from 1% to 25% and COV-H remains at 10%. The dashed line portrays the
relation between β and cable chord tension uncertainty COV-H within the range of 1%
and 25% while COV-c is 10%.
44 S.A. Mohammadi et al.

Figure 3 Effect of uncertainty level of cable tension and damper size on the reliability response
of a cable-damper system

Notes: Design point, h = 3,700 kn, c =50 kn∙s/m

As can be seen from the figure, in both cases, the reliability index decreases
monotonically with increased level of uncertainty in cable tension and damper size. The
two curves intersect at a point representing the case of COV-H and COV-c both equal to
10%. The reliability index associated with this intersection point is β = 7.003. It is worth
mentioning that the deterministic scenario, which corresponds to COV-H and COV-c
both, being zero, would theoretically lead to a reliability index of infinity. In other words,
the presence of uncertainty in the structural parameters would have a sizable impact on
the safe performance of a cable-damper system and should not be ignored in the design.
Further, a comparison between the two curves in Figure 3 indicates that the system
performance safety is more sensitive to the uncertainty associated with the damper size.
For instance, by increasing the COV of the damper size from 0.01 to 0.1, the reliability
index of the system would drop from 59.4 to 7.0 by roughly 8.5 times, whereas the same
increase of uncertainty in the cable tension would decrease the reliability index by seven
times from 48.7 to 7.0.

4 Application of time-variant reliability-based framework model

The proposed time-variant reliability-based framework model of a cable-damper system

can be applied to refine the current deterministic-based damper design practice by taking
into account the possible deviation of structural parameters from their respective nominal
design values at a given design point and the impact of such uncertainty on the design
outcome. In addition, if the time-dependent variation of system properties is available
from site measurement, the influence of these uncertainties on the system performance
can be incorporated to assess the life-long performance of a damper and can be used as a
base for developing system maintenance strategies. These potential applications will be
illustrated in the current section through three numerical examples.
 Assessment of damper performance in controlling cable vibrations 45

Figure 4 Reliability-based damper design curves by assuming normally distributed system

parameters with σc = 0.1μc and σh = 0.1μh at each design point, (a) Γd = 0.02
(b) Γd = 0.04 (c) Γd = 0.06

(a)

(b)

(c)
46 S.A. Mohammadi et al.

4.1 Reliability-based damper design curves

It has been observed in prior section that the presence of uncertainty in system properties
at a design point would considerably affect the reliable performance of an external
damper. Therefore, the deterministic-based design approach might lead to too optimistic
damper design. To refine the current design approach, reliability-based damper design
curves can be developed using the proposed framework model. By setting different
uncertainty levels for system structural parameters, the relation between damper size and
system reliability index can be derived for a range of non-dimensional parameters
representing actual cable and damper properties on site collected from over 1,400 bridge
stay cables by Tabatabai et al. (1998).
Without losing generality, it is reasonable to assume that the random variables H and
c, are statistically independent. Since ξ and ψ are functions of cable chord tension H and
damper capacity c, they are also random variables. Although the non-dimensional
damping parameter is defined as ψ = c / mH , for a specific value of non-dimensional
bending stiffness parameter ξ, ψ is a function of damper capacity c solely. Therefore,
when assessing the performance of a cable-damper system for a particular level of ξ, it is
legitimate to uncouple the reliability analysis. Figure 4 shows a sample set of reliability-
based damper design curves at three different damper locations of Γd = 0.02, 0.04, and
0.06. In this figure, variability of the reliability-index with the damping parameter ψ is
plotted for different levels of non-dimensional bending stiffness parameter ξ. A range of
0 to 60 is considered for the non-dimensional damping parameter ψ. The cable bending
stiffness parameter ξ varies between 100 and 400. The damper location parameter Γd is
limited to a maximum of 0.06 due to practical installation limitation. The sag parameter is
considered to be less than ten, representing typical level of sag for stay cables on
cable-stayed bridges. The uncertainty level of cable tension and damper size, represented
by COV-H and COV-c, are both assumed to be 10% in this set of sample curves. Should
the coefficients of variation of cable tension and damper capacity be different from 0.1,
the associated reliability-based damper design curves can be developed similarly using
the same procedures as those for Figure 4.
It is noteworthy that mathematically, the reliability index β indicates how often the
standard deviation of the LSF, σg, may be placed between zero and its mean value, μg.
Therefore, the reliability index β computed at each design point depends on the level of
uncertainty of the system random variables. Referring to Figure 3, we can observe that
the β value would increase with smaller coefficients of variation of H and c, but decrease
if σH and σc become large.
Figure 4 shows that in general, the reliability index β increases with the non-
dimensional damping parameter ψ up to a maximum level, and then decreases.
Comparing the three subplots in Figure 4, it can be seen that at the same damper capacity
ψ, the further the damper is away from the cable end (larger Γd), the higher the reliability
index β for the cable-damper system would be, indicating that if achievable in practice,
the damper should be installed closer to the cable mid-span, which agrees with the
existing experience. The peak points in these curves represent the damper size ψr,max
associated with the maximum achievable reliability level at the studied damper position.
This finding suggests that for a specific damper location, there always exists a damper
size, ψr,max, which would yield the most reliable performance of a damped cable. This
phenomenon is consistent with earlier findings of which an optimum damper size, ψopt,
 Assessment of damper performance in controlling cable vibrations 47

has been observed for a given damper location to achieve the maximum damping ratio of
a damped cable (e.g. Kovacs, 1982; Pacheco et al., 1993; Krenk, 2000; Cheng et al.,
2010). In addition, the damper size ψr,max, which corresponds to the most reliable system
performance identified in the current study, is found to agree well with the optimum
damper size ψopt reported in the literature. Therefore, in the design of an external damper
to suppress cable vibrations, upon the determination of damper location, the objective is
then to select a damper size which would produce the optimum vibration control effect
with the most reliable performance. For instance, for a stay cable with a damper located
at 0.06L [Figure 4(c)], if ξ = 200, the optimum damper size which gives the maximum
achievable damping ratio of 3.11% would be ψopt = 6.0 (Cheng et al., 2010), whereas that
would yield the most reliable performance of the damped cable is ψr,max = 6.05, with the
associated reliability-index equals to β = 10.2.
It should be pointed out that in the case of cable-supported structures such as
cable-stayed bridges and suspension bridges, cables (cables and hangers for suspension
bridges) play an important role in resisting loads and transferring them from bridge deck
to tower and foundation. Therefore, they are designed with larger safety factor as
compared to other structural elements. A reliability-based performance assessment of a
suspension bridge in Japan (Imai and Frangopol, 2001) reported that the reliability-
indices of main cables, hanger ropes, and stiffening girders are very different. The β
values of the main cables are around 15, whereas those of the stiffening girders are within
the range of 3 to 6. Similarly, a reliability-based optimum design analysis of glulam
cable-stayed footbridges performed by Simões and Negrão (2005) showed necessity of
having reliability-index of the studied cable-stayed system within the range of 15 to 20 to
resist excessive cable deflection.
In addition, it is worth noting that the proposed reliability-based damper design
curves can be used to identify the practical range of damper size ψ to satisfy the required
reliability-index β for a particular cable with a bending stiffness parameter ξ. Due to the
dependence of the formulated limit state function in equation (3) on the structural
damping ζs, a practical range of damping parameter ψ should be determined to ensure a
safe design. The limit level of the damping parameter ψmin for a safe design is the locus of
the points where g(H, c) = 0. In the case of Γd = 0.02, when a very stiff cable is selected,
i.e. ξ = 100, the obtained structural damping ζs for the range of selected damper size
(ψ ≤ 60) is not adequate to resist RWIV. Furthermore, a comparison of the β-ψ curves in
Figure 4 suggests that the impact of cable bending stiffness on the damper size decreases
when the damper moves further towards the cable centre.

4.1.1 Example 1
For the same cable discussed earlier, which has L = 150 m, m = 72 kg/m, D = 0.2 m,
EI = 5,420 kN∙m2, and H = 3700 kN, an external damper is restricted to be installed at
Γd = 0.04. We propose to use the reliability-based damper design curves given in Fig. 4 to
determine the minimum required damper size in order to prevent rain-wind-induced
vibration with a reliability index β ≥ 4.7, and calculate the corresponding equivalent first
modal damping ratio of the damped cable.
Based on the given cable properties, the non-dimensional cable bending stiffness
parameter is ξ = 123.9. From Figure 4(b), when β = 4.7, the damping parameters
correspond to ξ = 100 and 200 are ψ = 6.6 and 3.7, respectively. Thus, that corresponds to
48 S.A. Mohammadi et al.

ξ = 123.9 can be obtained using linear interpolation, which gives ψ = 5.9. The damper
size is then determined as c = ψ Hm = 5.9 × 3, 700 × 103 × 72 = 96.3 kN s/m.
The equivalent structural damping ratio ζs for the first mode can be computed
according to equation (5). The non-dimensional damper coefficient parameter for the first
mode of a corresponding horizontal non-flexural taut cable-damper system is found to be:

ηn = η1 = nπ ( Ld / L ) c / Hm = 1× π × 0.04 × 96.3 ×103 / 3, 700 ×103 × 72 = 0.741

The reduction and modification factors are obtained from the numerical solution of the
formulation proposed by Fujino and Hoang (2008). They are Rs1 = 0.946, Rf = 0.913,
ηs1 = 1.020, and ηf = 0.700. Hence, the equivalent first modal damping ratio of the system
is ζs = ζ1 = 1.429%. On the other hand, the required damping ζr for such a cable-damper
system to prevent rain-wind-induced vibration is determined to be 0.717%.

4.1.2 Example 2
If the non-dimensional bending stiffness parameter of a stay cable is ξ = 250, and the
possible damper location is Γd = 0.028, we propose to determine the minimum required
damping parameter to satisfy the reliability index of the cable-damper system for
resisting rain-wind-induced vibration to be β ≥ 6.0.
Using Figure 4(a), along with linear interpolation for bending stiffness parameter
ξ = 200 and 300, the non-dimensional damping parameter corresponding to ξ = 250 at
Γd = 0.02 is 22.0. Similarly, ψ = 4.6 for ξ = 250 and Γd = 0.04 can be obtained from
Figure 4(b). Therefore, by assuming a linear interpolation between results of Γd = 0.02
and 0.04, for a cable with ξ = 250, if a damper is attached to it at Γd = 0.028, to ensure
β ≥ 6.0, the minimum required damping parameter would be ψmin = 15.4.

4.2 Reliability-based maintenance plan

For a typical stay cable, creep and/or shrinkage in concrete deck and/or pylon as well as
slacking in steel cable would cause change in cable tension and thus its bending stiffness.
Hence, it is beneficial to investigate the impact of cable bending stiffness variation on the
reliability index of a cable-damper system. Figure 5 shows the reliability index of a
cable-damper system as a function of cable bending stiffness parameter ξ for three
different levels of damping parameter, ψ = 0.8, 6.6 and 41.4, at damper location of 0.06L.
Refer to Figure 4(c), these three damping parameter levels correspond respectively to
β = 0, the maximum achievable β, and the upper limit of the studied damper size to
maintain β >0. It is intended to determine the required bending stiffness parameter
corresponding to each damper size to achieve a desired safety level. For example, in the
case of a stay cable with damping parameter ψ = 6.6, the required cable bending stiffness
parameter satisfying a reliability index of β ≥ 3 is ξ ≥ 44.7. It can be observed from
Figure 5 that at higher ξ values (ξ ≥ 250), the β-ξ curves would reach an asymptotic limit
to a certain reliability index level. In other words, for a cable that acts similar to a taut
string, a slight reduction in cable tension force would not have an appreciable impact on
its reliability response.
 Assessment of damper performance in controlling cable vibrations 49

Figure 5 Variation of reliability-index with bending stiffness parameter for three different
damper sizes at Γd = 0.06

To ensure the performance of an external damper sustains at the desired safety level,
maintenance should be carried out during the bridge service stage to minimise the impact
of cable tension and damper capacity variation on the response of a damped cable.

4.2.1 Example 3
In this example, we consider a damped cable on a concrete cable-stayed bridge, of which
the cable tension and damper property vary with time. The time-dependent effects on the
reliability of this cable-damper system are evaluated using the proposed time-variant
reliability analysis framework model. The properties of the cable used in this example
are taken from a study by Tabatabai and Mehrabi (2000): L = 93 m, D = 0.225 m,
EI = 3,049 kN∙m2, and m = 114.09 kg/m. The initial specified cable tension along
its chord direction is H = 5017 kN (ξ = 119.3). To suppress rain-wind-induced
cable vibrations, a linear viscous damper is attached 3.72 m from the cable lower end
(Γd = 0.04). In this example, two different damper design cases are considered. Case 1
represents an optimum damper design case with a copt = 271.1 kN∙s/m (ψopt = 11.3),
whereas Case 2 assumes a design case using a smaller damper size, c = 90.4 kN∙s/m
(ψ = 3.8). The cable tension loss is estimated based on the field monitoring data of a
cable-stayed bridge over a period of 300 days (Au and Si, 2012), whereas a linear
degradation from its original design is assumed for the damper capacity. In addition, it is
assumed that the cable chord tension and the damper capacity are random variables with
independent normal distributions, and the time variation of their respective mean values
over a period of 300 days are shown in Figure 6. We are interested in developing a
maintenance plan for the studied cable-damper system over a period of 300 days to
ensure its performance remain at a desired safety level during this time period.
50 S.A. Mohammadi et al.

Figure 6 Time variation of cable tension and damper capacity over a period of 300 days,
(a) cable tension along the chord (b) damping coefficient of damper

(a)

 Case1

(b)

Notes: Case 1: c0 = copt = 271.1 kN∙s/m; Case 2: c0 = 90.4 kN∙s/m

Let us define the reliability level of the cable-damper system corresponding to its initial
design point as the desired safety level over its lifetime and denote it as β0. In case 1, with
the initial damper design point being c0 = copt = 271.1 kN∙s/m and H0 = 5017 kN, β0
equals to 9.50; whereas for Case 2 (c0 = 90.4 kN∙s/m, H0 = 5,017 kN), the reliability
index is reduced to β0 = 2.24. This supports that a higher level of system reliability can be
achieved when a damper with a size closer to its optimum value is attached to the cable.
 Assessment of damper performance in controlling cable vibrations 51

The reliability response of the damped cable over this time period is portrayed in
Figure 7 as the time history of the relative reliability-index β(t)/β0 for cases 1 and 2. It
can be seen that as expected, the reliability index β(t) decreases with time, which
suggests a higher probability of failure associated with the loss in the cable tension and
damper capacity during the lifetime of the cable-damper system. Further, a higher
reduction of β value occurs within the first 50 days due to a rapid loss of cable tension
over this time period. As can be seen from Figure 6(a), within the first 50 days, the loss in
cable chord tension is 7.4%, whereas over the entire 300 days, the total loss in cable
chord tension is 12.2%. Based on Figure 7, the impact of cable tension and damper
capacity variation on the reliability of the system performance at any time within the
300 days can be obtained. Hence, the time-dependent reliability index, β(t), can be
expressed as a function of the damper coefficient and the cable chord tension, i.e., β(c(t),
H(t)).

Figure 7 Time variation of relative reliability index

 Case 1

Notes: Case 1: c0 = copt = 271.1 kN∙s/m, β0 = 9.50; Case 2: c0 = 90.4 kN∙s/m, β0 = 2.24
For a more convenient identification of the effect of cable tension loss and damper
capacity degradation on the system reliability, this set of results are also plotted in
Figure 8 in terms of the relative reliability-index β(t)/β0 versus the percentage loss of
cable chord tension [Figure 8(a)] and damper capacity [Figure 8(b)].Though the
presentation of Figure 8 isolates the dependence of the reliability index reduction on the
loss in cable chord tension and damper capacity, it should be pointed out that in the
analysis, the variation of H and c are considered simultaneously according to the patterns
shown in Figure 6. For example, at day 180, when the cable chord tension and the
damper capacity are reduced by 11.3% and 3%, respectively, the corresponding relative
reliability-index in case 1 is β (t = 180 day) / β0 = 0.980, i.e. the reliability-index drops
2% from its initial value.
52 S.A. Mohammadi et al.

Comparing the relative reliability index results of case 1 and case 2, it can be seen
that case 2 is more critical. This implies that when the damper capacity c is different from
the optimum value, the system performance would be more sensitive and have greater
degradation due to uncertainty in system parameters. This can also be verified by
comparing the slope of each reliability curve at different damping parameters in
Figure 4(b). The slope is less steep when ψ is closer to ψr,max.

Figure 8 Relative reliability-index estimation due to loss of (a) cable chord tension and
(b) damper capacity

(a)

(b)
 Assessment of damper performance in controlling cable vibrations 53

The integration of reliability analysis into the lifetime maintenance strategy for the
studied cable-damper system can be achieved by considering the time variation of the
system performance associated with the loss in cable chord tension and damper capacity.
To keep the cable-damper system operating at its designated performance, the lifetime
variation of the system reliability-index should remain zero, which means
∂β ( c(t ), H (t ) ) ∂β ∂c ∂β ∂H
= ⋅ + =0 (11)
∂t ∂c ∂t ∂H ∂t
in which the lifetime variation of the damper capacity and the cable chord tension are
considered by ∂c/∂t and ∂H/∂t, respectively. If we assume that the maintenance of the
studied cable-damper system can be achieved by adjusting cable chord tension to ensure
that the performance remains at the designed safety level, equation (11) can be rewritten
as:
∂H ∂β / ∂c ∂c
=− ⋅ (12)
∂t ∂β / ∂H ∂t

Thus, the required adjustment of cable tension at any time of interest can be obtained
from the tangents of the reliability index curves with respect to c and H, i.e.:
⎛ β (t ) ⎞ ⎛ β (t ) ⎞
∂⎜ ⎟ ⋅ β0 ∂⎜ ⎟
∂β ( c(t ), H (t ) ) ⎝ β0 ⎠ β ⎝ β0 ⎠
= = 0⋅ (13)
∂c ⎛ c − c(t ) ⎞ c0 ⎛ c − c(t ) ⎞
∂⎜ 0
c ⎟ ⋅ ( −c0 ) ∂⎜ 0 ⎟
⎝ 0 ⎠ ⎝ c0 ⎠

⎛ β (t ) ⎞ ⎛ β (t ) ⎞
∂β ( c(t ), H (t ) )
∂⎜ ⎟ ⋅ β0 ∂⎜ ⎟
= ⎝ β0 ⎠ β
= 0 ⋅ ⎝ β0 ⎠ (14)
∂H ⎛ H − H (t ) ⎞ H 0 ⎛ H 0 − H (t ) ⎞
∂⎜ 0 ⎟ ⋅ ( −H0 ) ∂⎜ ⎟
⎝ H0 ⎠ ⎝ H0 ⎠
where the lifetime variation of the relative reliability index β(t)/β0 versus the percentage
loss in cable chord tension and damper capacity can be computed from the results in
Figure 8.
In the current example, the initial design point in case 1 is H0 = 5,017 kN,
c0 = 271.1 kN·s/m and the initial reliability index equals to β0 = 9.50. For instance, at day
180, it can be deduced from equations (13), (14), and Figure 8 that ∂β/∂c = 3.7×10–5 and
∂β/∂H = 3.1 × 10–7. Hence, to ensure β180 = β0 = 9.50, the adjustment rate of the cable
tension, in terms of its chord component, can be obtained from equation (12) as

(∂H / ∂t ) Day180 = − ( 3.7 × 10−5 ) / ( 3.1× 10−7 ) ⋅ ( −0.056) = 6.73 kN/day

where the lifetime variation of damper capacity at day 180 can be obtained from Figure 6,
which is ∂c/∂t = –0.056 kN∙s/m per day.
By repeating the calculations for cable chord tension adjustment rate, ∂H/∂t, at
different time instants, a maintenance plan can be defined over the lifetime of the studied
cable-damper system. The required additional force after a certain time period can be
obtained by integrating the adjustment rates over the maintenance period, i.e.
54 S.A. Mohammadi et al.

∂H
t
Hm = ∫0 ∂t
dt (15)

where Hm is the amount of cable chord tension adjustment required at day t. Figure 9
shows the required Hm at different time to maintain the desired performance of the
studied cable-damper system. The results show that to keep the system perform at a level
corresponding to the designed reliability index β0 = 9.50, the required additional amount
of cable tension increases monotonically during the lifetime of the cable, which, at the
end of the studied 300-day period is 188.1 kN. This equals to 3.7% of the initial design
value.

Figure 9 Additional tension adjustment to maintain system reliable performance over the period
of 300 days

Case 2

Case 1

If conducting the same exercise for case 2, the required additional tension force over the
300-day period would be 11.2% of the initial design value. Comparison of the two curves
shown in Figure 9 indicates that more adjustment in cable tension is needed if the
selected damper size is different from the optimum one. Therefore, when an external
damper is applied to control cable vibrations, it is very important to select the optimum
size or a size as closer as possible to the optimum one. This is not only for a more
effective vibration control, but also to ensure a higher reliability of the system.

5 Summary

External dampers are commonly used on cable-stayed bridges to control various types of
large amplitude cable oscillations. In the case of RWIV, due to lack of thorough
understanding of its excitation mechanism, current prediction of the required damping to
effectively suppress it is mainly depending on a Scruton number-based empirical
criterion, and the damper is designed using tools derived by deterministic-based analysis
approaches. Therefore, the latent stochastic nature of the problem, including the
uncertainty of cable and damper properties with respect to their nominal design values
 Assessment of damper performance in controlling cable vibrations 55

and their potential time variations over the service life of a bridge, is generally excluded
in the formulation. Nevertheless, these uncertainties would render the actual system
properties deviate from their assumed nominal design value and could have a
considerable impact on the performance of a damped cable. The current study aims at
developing a time-variant reliability-based framework model to evaluate how these
uncertainties in the structural parameters of a cable-damper system would affect the
performance of an external damper designed according to the current practice.
The problem has been formulated as a limit state up-crossing scenario to determine
the time specific system reliability due to rain-wind-induced cable vibrations. The time-
variant reliability analysis is conducted by combing the existing time-invariant reliability
methods with the time-varying system parameters. The general form of the time-
dependent LSF was proposed as the difference between the available time-varying
equivalent structural damping ratio in a cable-damper system and the minimum required
structural damping ratio of a damped cable to avoid RWIV. Results showed that
compared to the first-order reliability method, the SORM with adaptive approximations
method could improve the accuracy of failure probability prediction while retaining
simplicity in the problem formulation and efficiency in computation. Various
applications of the proposed time-variant reliability-based framework model, including
the development of reliability-based damper design curves and long-term structural
maintenance plan, have been demonstrated through numerical examples. It has been
found that the presence of uncertainty in the structural properties of a cable-damper
system at a given design point could have a sizable effect on its reliability response and
thus should be considered in design. In particular, the performance of a cable-damper
system was observed to be more sensitive to the uncertainty associated with damper size.
Further, at a specific damper location, the existence of a damper size which would result
in the most reliable performance of a damped cable has been identified. Results showed
that it agreed well with the optimum damper size reported in literature which can attain
the maximum damping ratio at the same damper location.

References
Au, F.T. and Si, X.T. (2012) ‘Time-dependent effects on dynamic properties of cable-stayed
bridges’, Structural Engineering and Mechanics, Vol. 41, No. 1, pp.139–155.
Casas, J.R. and Aparicio, A.C. (2010) ‘Rain-wind-induced cable vibrations in the Alamillo cable-
stayed bridge (Sevilla, Spain). Assessment and remedial action’, Structure and Infrastructure
Engineering, Vol. 6, No. 5, pp.549–556.
Chen, Z.Q., Wang, X.Y., Ko, J.M., Ni, Y.Q., Yang, G. and Hu, J.H. (2004) ‘MR damping system
for mitigating wind-rain induced vibration on Dongting Lake cable-stayed bridge’, Wind and
Structures, Vol. 7, No. 5, pp.293–304.
Cheng, S., Darivandi, N. and Ghrib, F. (2010) ‘The design of an optimal viscous damper for a
bridge stay cable using energy-based approach’, Journal of Sound and Vibration, Vol. 329,
No. 22, pp.4689–4704.
Choi, S.K., Grandhi, R.V. and Canfield, R.A. (2007) Reliability-Based Structural Design, Springer,
London.
Flamand, O. (2001) ‘An explanation of the rain-wind induced vibration of inclined stays’,
Proceedings of the 4th International Symposium on Cable Dynamics, Montreal, Canada,
28–30 May, pp.69–76.
Frangopol, D.M. and Maute, K. (2003) ‘Life-cycle reliability-based optimization of civil and
aerospace structures’, Journal of Computers and Structures, Vol. 81, No. 7, pp.397–410.
56 S.A. Mohammadi et al.

Frangopol, D.M., Maute, K. and Liu, M. (2007) ‘Optimization of structural and mechanical
systems’, in Arora, J.S. and Arora, J.S. (Eds.): Design Optimization with Uncertainty, Life-
Cycle Performance and Cost Considerations, pp.291–329, World Scientific Publishing,
London, England.
Fujino, Y. and Hoang, N. (2008) ‘Design formulas for damping of a stay cable with a damper’,
Journal of Structural Engineering, Vol. 134, No. 2, pp.269–278.
Geurts, C. and Staalduinen, P. (1999) ‘Estimation of the effects of rain-wind induced vibration in
the design stage of inclined stay cables’, in Larsen, Larose and Livesey (Eds.): Wind
Engineering into the 21st Century, pp.885–892, Balkema, Rotterdam.
Geurts, C., Vrouwenvelder, T., Staalduinen, P. and Reusink, J. (1998) ‘Numerical modelling of
rain-wind-induced vibration: Erasmus Bridge, Rotterdam’, Structural Engineering
International, SEI, Vol. 8, No. 2, pp.129–135.
Gu, M. and Lu, Q. (2001) ‘Theoretical analysis of wind-rain induced vibration of cables of cable-
stayed bridges’, The 5th Asia-Pacific Conference on Wind Engineering, Kyoto, pp.125–128.
Gu, M., Du, X.Q. and Li, S.Y. (2009) ‘Experimental and theoretical simulations on wind-rain-
induced vibration of 3-D rigid stay cables’, Journal of Sound and Vibration, Vol. 320,
Nos. 1–2, pp.184–200.
Hikami, Y. and Shiraishi, N. (1988) ‘Rain-wind induced vibrations of cables in cable-stayed
bridges’, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 29, Nos. 1–3,
pp.409–418.
Imai, K. and Frangopol, D.M. (2001) ‘Reliability-based assessment of suspension bridges:
application to the Innoshima bridge’, Journal of Bridge Engineering, Vol. 6, No. 6,
pp.398–411.
Irvine, H.M. and Caughey, T.K. (1974) ‘The linear theory of free vibrations of a suspended cable’,
Proceedings of the Royal Society of London, Series A341, pp.299–315.
Irwin, P.A. (1997) ‘Wind vibrations of cables on cable-stayed bridges’, Building To Last:
Proceedings of Structural Congress XV, Portland, Oregon, 13–16 April, pp.383–387.
Johnson, E.A., Christenson, R.E. and Spencer, J.B. (2002) Flat-Sag Cables With Semiactive
Damping, Dept. of Civil and Env. Engrg. Los Angeles, University of Southern California.
Kovacs, I. (1982) ‘Zur frage der seilschwingungen und der seildämpfung’, Bautechnik,
pp.325–332, in German.
Krenk, S. (2000) ‘Vibrations of a taut cable with an external damper’, Journal of Applied
Mechanics, Vol. 67, No. 4, pp.772–776.
Krenk, S. and Nielsen, S. (2002) ‘Vibrations of a shallow cable with a viscous damper’,
Proceedings of the Royal Society of London, Series A458, pp.339–357.
Kumarasena, S., Jones, N.P., Irwin, P. and Taylor, P. (2007) Wind-Induced Vibration Of Stay
Cables, U.S. Department of Transportation, Federal Highway of Administration,
FHWA-HRT-05-083, Boston.
Matsumoto, M. (1998) ‘Observed behaviour of prototype cable vibration and its generation
mechanism’, in Larsen and Esdahl (Eds.): Bridge Aerodynamics, pp.189–211, Balkema,
Rotterdam.
Matsumoto, M., Shirato, H., Yagi, T., Goto, M., Sakai, S. and Ohya, J. (2003) ‘Field observation of
the full-scale wind induced cable vibration’, Journal of Wind Engineering and Industrial
Aerodynamics, Vol. 91, Nos. 1–2, pp.13–26.
Mehrabi, A. and Tabatabai, H. (1998) ‘Unified finite difference formulation for free vibration of
cables’, Journal of Structural Engineering, Vol. 124, No. 11, pp.1313–1322.
Mohammadi, S., Cheng, S. and Ghrib, F. (2011) ‘Reliability assessment for damped bridge stay
cables under dynamic excitation’, Proceedings of the 8th International Conference on
Structural Dynamics, Leuven, Belgium, pp.2993–2999.
Mohammadi, S., Cheng, S. and Ghrib, F. (2013) ‘Assessment of dynamic behavior of an inclined
sag cable with a transverse linear viscous damper using reliability-based approaches’, General
Conference. 336, Montreal, Canada, Canadian Society of Civil Engineering, pp.1–9.
 Assessment of damper performance in controlling cable vibrations 57

Pacheco, B.M., Fujino, Y. and Sulekh, A. (1993) ‘Estimation curve for modal damping in stay
cables with viscous damper’, Journal of Structural Engineering, Vol. 119, No. 6,
pp.1961–1979.
Pagnini, L. (2010) ‘Reliability analysis of wind-excited structures’, Journal of Wind Engineering
and Industrial Aerodynamics, Vol. 98, pp.1–9.
PTI Publication (2001) Recommendations For Stay-Cable Design, Testing and Installation, Fourth
Ed., Post-Tensioning Institute, USA.
Saito, T., Matsumoto, M. and Kitazawa, M. (1994) ‘Rain-wind excitation of cables on cable-stayed
Higashi-Kobe Bridge and cable vibration control’, Proceedings of the International
Conference on Cable-Stayed and Suspension Bridges (AFPC), 2, Deauville, France,
pp.507–514.
Simões, L.C. and Negrão, J.H. (2005) ‘Reliability-based optimum design of Glulam cable-stayed
footbridges’, Journal of Bridge Engineering, Vol. 10, No. 1, pp.39–44.
Tabatabai, H. and Mehrabi, A. (2000) ‘Design of mechanical viscous dampers for stay cables’,
Journal of Bridge Engineering, Vol. 5, No. 2, pp.114–123.
Tabatabai, H., Mehrabi, A.B., Morgan, B.J. and Lotfi, H.R. (1998) Non-Destructive Bridge
Evaluation Technology: Bridge Stay Cable Condition Assessment, Report prepared for Federal
Highway Admin. pSkokie, III: Construction Technology Laboratories.
Uno, K., Kitagawa, S., Tsutsumi, H., Inoue, A. and Nakaya, S. (1991) ‘A simple method of
designing cable vibration dampers of cable-stayed bridges’, Journal of Structural Engineering,
Japan Society of Civil Engineers, Vol. 6, No. 1, pp.167–178.
Virlogeux, M. (1998) ‘Cable vibrations in cable-stayed bridges’, in Larsen and Esdahl (Eds.):
Bridge Aerodynamics, pp.213–233, Balkema, Rotterdam.
Wang, L.P. and Grandhi, R.V. (1995) ‘Improved two-point function approximation for design
optimzation’, AIAA Journal, Vol. 32, No. 9, pp.1720–1727.
Yamada, H. (1997) ‘Control of wind-induced cable vibrations from a viewpoint of the wind
resistance design of cable-stayed bridges’, Proceedings of the 2nd International Seminar on
Cable Dynamics, Tokyo, pp.129–138.
Yamaguchi, H. (1990) ‘Analytical study on growth mechanism of rain vibration of cables’, Journal
of Wind Engineering and Industrial Aerodynamics, Vol. 33, pp.73–80.
Yoneda, M. and Maeda, K. (1989) ‘A study on practical estimation method for structural damping
of stay cable with damper’, Proceeding of Canada-Japan Workshop on Bridge Aerodynamics,
Ottawa, Canada, pp.119–128.

Notation

The following symbols are used in this paper:

A cross-sectional area of the cable
c viscous damper coefficient
c0 damper capacity at design point
d sag at mid-span of cable
D cable diameter
EI equivalent bending stiffness of the cable
fc in-plane transverse damping force by the damper
g gravitational constant
58 S.A. Mohammadi et al.

g(∙) limit state function

g(Xk) value of the nonlinear LSF at Xk
H cable tension component along its chord
H0 pretension force along the chord at design point
Hm amount of cable chord tension adjustment
h(t) additional tension in the cable due to vibration
L cable length
Ld distance from the damper to the nearest cable end
Le static length of cable
m mass per unit length of cable
Pf probability of failure
Rf reduction factor in maximum damping ratio due to cable flexural rigidity
Rsn reduction factor in maximum damping ratio due to cable sag
Sc Scruton number
T cable tension
T0 cable tension at design point
U standard normalised coordinate system
U* coordinate of the most probable failure point (MPP) in U-space
X coordinate along the cable chord
X vector containing system variables x1, x2,…,xn
xi,k ith element of X in the kth iteration
Xk value of X in the kth iteration
Y coordinate perpendicular to cable chord
β reliability-index
β0 reliability-index at design point
δ(∙) Dirac-delta function
ζn nth modal damping ratio of a damped cable
ζr required structural damping ratio
ζs available structural damping ratio
φ cumulative distribution function of the standard normally distributed variable
Γd non-dimensional damper location parameter
ξ non-dimensional bending stiffness parameter of a cable
 Assessment of damper performance in controlling cable vibrations 59

ηn dimensionless parameter of damper coefficient of mode n

ηf modification factor for ηn due to cable flexural rigidity
ηsn modification factor for ηn due to cable sag
ψ non-dimensional damping parameter of a cable
ψmin minimum damping parameter
ψopt optimum damping parameter
ψr,max damping parameter corresponding to the most reliable system performance
ρ air density
λ2 non-dimensional sag parameter
θ cable inclination angle
ω1s fundamental frequency of a taut string.