See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334284887

Optimization of concrete cable-stayed bridges under seismic action

Article  in  Computers & Structures · July 2019

DOI: 10.1016/j.compstruc.2019.06.008

CITATIONS READS

2 38

3 authors:

Alberto Martins Luis M. C. Simões

University of Coimbra University of Coimbra
25 PUBLICATIONS   97 CITATIONS    150 PUBLICATIONS   688 CITATIONS   

SEE PROFILE SEE PROFILE

J.H.J.O. Negrao
University of Coimbra
133 PUBLICATIONS   632 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sustainable Optimum Design of Structures View project

Sustainable Optimum Design of Bridges View project

All content following this page was uploaded by Alberto Martins on 20 June 2020.

The user has requested enhancement of the downloaded file.

Preprint version of article

Optimization of concrete cable-stayed bridges under seismic action

Alberto M. B. Martins, Luís M. C. Simões and João H. J. O. Negrão

Published journal article version information

Publisher: Elsevier
Journal: Computers & Structures
ISSN: 0045-7949
Journal homepage: https://www.sciencedirect.com/journal/computers-and-structures
DOI: https://doi.org/10.1016/j.compstruc.2019.06.008
Received: 01 November 2018
Accepted: 28 June 2019
Available online: 06 July 2019

Please cite this article as:

Martins, A. M. B., Simões, L. M. C., & Negrão, J. H. J. O. (2019). Optimization of concrete
cable-stayed bridges under seismic action. Computers & Structures, 222, 36–47.
https://doi.org/10.1016/j.compstruc.2019.06.008
 Optimization of concrete cable-stayed bridges under seismic
action

Alberto M. B. Martins*, Luís M. C. Simões and João H. J. O. Negrão

Department of Civil Engineering, Faculty of Sciences and Technology, University of Coimbra, 3030-788
Coimbra, Portugal

Abstract

The design of cable-stayed bridges built in earthquake prone areas is governed by their response
to the seismic loading. These structures have low damping and given their flexibility exhibit
long vibration periods. Moreover, the dynamic response of this type of structure is highly
dependent on the stiffness and mass distribution. This paper concerns an optimization-based
computational tool to help in the design of concrete cable-stayed bridges under seismic action.
The finite element method is used for the three-dimensional analysis of the structure under dead
and live loads considering time-dependent effects and geometrical nonlinearities. The dynamic
analysis is solved by the modal/spectral approach. The design of concrete cable-stayed bridges
under seismic loading is posed as a multi-objective optimization problem. The cable forces and
the cross-sectional dimensions of cables, deck and towers are the design variables. Design
objectives of minimum cost, deflections, natural frequencies and stresses considering both,
serviceability and ultimate limit states are considered. The discrete direct method is used for
sensitivity analysis. An entropy-based algorithm finds economically and structurally efficient
solutions by rearranging the stiffness and mass distribution to enhance the response of concrete
cable-stayed bridges. A numerical example is presented to illustrate the features of the proposed
computational tool.
Keywords: cable-stayed bridges; optimization; seismic action; concrete.

1. Introduction

Cable-stayed bridges have been constructed all over the world, are mainly used for medium-to-
long spans and are part of important transportation networks. Besides their structurally

*
Corresponding author
Email addresses: alberto@dec.uc.pt (A.M.B. Martins), lcsimoes@dec.uc.pt (L.M.C. Simões), jhnegrao@dec.uc.pt
(J.H. Negrão)
efficiency they owe their popularity due to an elegant and transparent appearance. The bridges
constructed in earthquake prone areas, must be designed to withstand the seismic action.
Cable-stayed bridges, due to the long spans and their flexibility, typically present long vibration
periods which theoretically makes them not sensitive to dynamic excitation (Walther, 1999;
Svensson, 2012). However, their inherent low damping and its dynamic behaviour is highly
dependent on the stiffness and mass distribution. Any an attempt to minimize the inertia forces
and to maximize the resistance leads to an undesired decrease in the vibration periods and
consequently to higher seismic forces. Moreover, for concrete bridges, although the higher
damping, they are heavier than steel or composite bridges which implies higher inertia forces.
The dynamic behaviour of cable-stayed bridges has been extensively studied by several authors
(Abdel-Ghaffar and Khaliffa, 1991; Abdel-Ghaffar and Nazmy, 1991; Morgenthal, 1999;
Soneji and Jangid, 2008, Caetano et al., 2008; Camara, 2011; Camara and Efthymiou, 2016).
They focused in topics, such as: numerical modelling and dynamic properties, spatial variability
of the seismic action, soil-structure interaction, cable vibrations, devices for energy dissipation
and isolation, the deck-tower interaction. Concerning the optimization of cable-stayed bridges
under seismic action only few studies have been reported (Negrão and Simões, 1999; Ferreira
and Simões, 2011 and 2012). To the best of the author’s knowledge, the optimization of
concrete cable-stayed bridges considering the seismic effects has not yet been reported.
Cable-stayed bridge design is a daunting task which involves some complex problems, such as:
defining the structural system, finding the members’ cross-sections, the calculation of the cable
forces distribution, the construction stages and geometrical non-linear effects. For concrete
bridges the time-dependent effects must also be considered. The seismic action adds more
complexity to find an adequate mass and stiffness distribution that optimizes the bridge
dynamic response. Therefore, optimization algorithms are particularly suited to handle the large
amount of information involved and thus obtaining economical and structurally efficient
solutions under both, static and dynamic loading.
Previous works concerning the optimization of cable-stayed bridges studied the cable force
calculation in steel (Negrão and Simões 1997a; Sung et al., 2006; Baldomir et al., 2010),
composite (Hassan et al., 2012 and Hassan 2013) and concrete bridges (Martins et al., 2015a
and 2015b). The use of geometric and cross-sectional design variables was also reported in the
optimization of steel and composite steel-concrete bridges (Torii and Ikeda, 1987; Ohkubo and
Taniwaki, 1991; Long et al., 1999; Simões and Negrão 1994; Negrão and Simões, 1997b;
Simões and Negrão, 2000; Hassan et al., 2013 and 2015) subjected to static loading. The main
objective was to the minimize the structure cost while ensuring that the stresses and
displacements throughout the structure remain within allowable limits.
The current research work is a development of previous research works by the authors
concerning the optimization of concrete cable-stayed bridges subjected to static loading
(Martins et al., 2015a; 2015b and 2016). The main goal is to develop an optimization-based
computational tool to assist in the design of concrete cable-stayed bridges under seismic action.
It is considered relevant to have, in the preliminary design stage of large and complex structures
like cable-stayed bridges, a tool that provides solutions with an adequate behaviour to face not
only static, but also seismic loading. The latter being of major concern for the design of bridges
in earthquake prone areas.
A direct search algorithm was selected to handle efficiently a large optimization problem, not
only in terms of the number of design variables and design objectives, but also with some added
complexity due to several load cases, the consideration of geometrical nonlinearities and the
dynamic analysis to access the structural response under seismic action. A concrete bridge
needs including of the time-dependent effects and poses additional difficulties to the
optimization problem when formulating the sensitivities for the design objectives. This is due
to the fact that the resistance of each concrete member depends on the correspondent cross-
sectional design variables.
Therefore, a computer program previously developed in MATLAB environment was improved
to allow three-dimensional modelling, static and dynamic analysis. This program includes two
modules: a structural analysis module and a sensitivity analysis and optimization module. The
first module is based in the finite element method and takes into account static loading
(permanent load and road traffic live load), the time-dependent effects and the geometrical
nonlinearities. A modal/spectral approach is used to obtain the structural response to the seismic
action defined according to Eurocode 8 (EN 1998-1-1, 2010).
In the second module the design of concrete cable-stayed bridges under seismic action is
formulated as a multi-objective optimization problem. This is a minimax problem which is
solved indirectly, using an entropy-based approach, through the minimization of a convex scalar
function. The design variables considered are the cable-stays areas and prestressing forces, the
deck and towers cross-sections. The design objectives are the cost, deflections, natural
frequencies and stresses considering both, serviceability and ultimate limit states. The stress
objectives for the concrete members were established according to the Eurocode 2 (EN 1992-
1-1, 2010) provisions. For a gradient-based optimization algorithm the sensitivity analysis is of
major relevance to access the structural response to changes in the design variables. Considering
the accuracy, the computational efficiency and given that the number of design objectives is
much larger than the design variables, the analytical discrete direct method was selected for the
sensitivity analysis.
The features and applicability of the procedure are demonstrated by numerical examples
concerning the optimization of a real-sized concrete cable-stayed bridge.

2. Structural analysis

2.1. Structural concrete modelling

The analysis of concrete structures requires the consideration of the instantaneous and the time-
dependent behaviour of concrete. In this work, the structural concrete was modelled as a linear
viscoelastic material and the time-dependent effects of creep, shrinkage and aging of concrete
were computed considering the Eurocode 2 (EN 1992-1-1, 2010) formulation. The deformation
of a concrete member loaded with a stress σc at an age t0 is time-dependent and can be expressed
as the sum of stress dependent, ε cσ (t ,t 0 ) , and stress independent, ε cn (t ) , contributions as
follows
ε c (t ) = ε cσ (t , t 0 ) + ε cn (t ) = J (t , t 0 ) ⋅ σ c (t 0 ) + ε cn (t ) (1)
where J (t , t 0 ) is the creep function. If the acting stresses are less than 45% of the characteristic
value of concrete compressive strength (fck), the principle of superposition is valid and the creep
strain varies linearly with the applied stress.
For the general case of varying stress, Equation 1 can be rewritten as
t
∂σ c (τ )
ε c (t ) = J (t , t 0 ) ⋅ σ c (t 0 ) +  J (t ,τ ) dτ + ε cn (t ) (2)
t0
∂τ

that requires solving the superposition integral. This can be done with different approaches:
simplified methods, step-by-step numerical integration and approximation of the creep
function. In this paper, the latter approach is used and the creep function is approximated by a
Dirichlet series (Bazant, 1988) leading to:

 a (t )(1 − e )
1 1 n
J (t , t 0 ) ≅
α ( − t −t 0 )
+ j
(3)
Ec (t 0 ) Ecm
j 0
i =1

where n is the number of terms of the Dirichlet series and aj are coefficients obtained from a
curve fitting using the least squares method. The coefficients 1 / α j are called retardation times

and are chosen to cover the range of time values for the creep coefficients calculation.
The Eurocode 2 (EN 1992-1-1, 2010) model for concrete shrinkage considers that the total
shrinkage strain at an age t, ε cs (t ) , is given by

ε cs (t ) = ε ca (t ) + ε cd (t ) (3)
where ε ca and ε cd are the autogenous and the drying shrinkage, respectively. The shrinkage
deformation depends on the member notional size and the age of concrete at the beginning of
the drying shrinkage, the environmental relative humidity, the cement type and the concrete
compressive strength.
The concrete aging, which translates to an increase in concrete strength and modulus of
elasticity with time as result of curing can be expressed as
0, 3
    28 1 / 2   
Ecm (t ) = exps 1 −      Ecm
 (4)
    t    

where Ecm is the mean modulus of elasticity of concrete at an age of 28 days, t is the age of the
concrete in days and s is a coefficient depending on the cement type.
The time-dependent effects were simulated in the finite element model by equivalent nodal
forces computed from the creep and shrinkage deformations for a given time interval. These
forces produce the same displacement field as the time-dependent effects and from which the
actual deformation state is calculated. The stresses are assessed using only the elastic
constitutive relationship between stresses and mechanical origin deformations.
For detailed considerations concerning the modeling of the time-dependent effects please refer
to a recent research work by the authors (Martins et al., 2015a).

2.2. Geometrical nonlinearities

In cable stayed-bridges there are three main sources of geometric nonlinearities that arises from
the nonlinear axial force-elongation relationship for the inclined cable stays due to the sag
caused by their own weight; the nonlinear axial force and bending moment-deformation
relationships for the towers and the deck under combined bending and axial forces; and the
geometry change caused by large displacements (Nazmy and Abdel-Ghaffar, 1990; Karoumi,
1999; Freire et al., 2006). To consider the geometric nonlinear effects, the structural analysis
was carried out in an iterative manner to perform a second-order elastic analysis.
To account for the cable sag, the cables were modelled as truss element with an equivalent
modulus of elasticity given by Ernst formulation (Ernst, 1965)
 E
E eq =
1+
(γ ⋅ L cos α )2 E (5)
12σ 3

where Eeq is the equivalent cable modulus of elasticity, E is the effective cable material modulus
of elasticity, γ is the specific weight of the cable material, L is the length of the chord, α is the
angle between the cable chord and the horizontal direction and σ is the tension stress in the
cable.
The towers and the deck longitudinal beams were modelled with 2-node and 12-degrees of
freedom Euler-Bernoulli beam elements represented in Figure 1.

Figure 1 – Degrees of freedom of the beam element in local coordinates

To include the geometrical nonlinearities in these elements, the stiffness matrix is obtained
considering the elastic (KE) and geometric (KG) contributions given by
 EA EA 
 L 0 0 0 0 0 − 0 0 0 0 0 
L
 12 EI z 6 EI z 12 EI z 6 EI z 
 0 0 0 0 − 0 0 0 
 L3 L2 L3 L2 
 12 EI y 6 EI y 12 EI y 6 EI y 
 3
0 − 2
0 0 0 − 3
0 − 0 
L L L L2
 GI t GI 
 0 0 0 0 0 − t 0 0 
 L L 
 4 EI y 6 EI y 2 EI y 
 0 0 0 2
0 0 
 L L L
4 EI z 6 EI 2 EI z 
 0 − 2z 0 0 0 
KE =  L L L  (6)
 EA 
 0 0 0 0 0 
L
 12 EI z 6 EI 
 0 0 0 − 2z
 L3 L 
 12 EI y 6 EI y 
 0 0 
L3 L2
 GI t 
 0 0 
 L 
 4 EI y 
 0 
L
 4 EI z 
 
 L 
where E, G, A, Iz, Iy and It are the modulus of elasticity, the shear modulus, the cross sectional
area, the moment of inertia about z-axis, the moment of inertia about y-axis and the torsional
constant, respectively.
0 0 0 0 0 0 0 0 0 0 0 0 
 6 L 6 L 
 0 0 0 0 − 0 0 0
5 10 5 10 
 6 L 6 L 
 0 − 0 0 0 − 0 − 0 
 5 10 5 10 
 0 0 0 0 0 0 0 0 0 
 2 L2 L L2 
 0 0 0 0 0 
 15 10 30 
 2 L2 L L2
0 − 0 0 0 − 
P 15 10 30 
KG =  0 0 0 0 0 0  (7)
L
6 L
 0 0 0 − 
 5 10 
 6
0 
L
 0
5 10
 0 0 0 
 
 2 L2
0 
 15 
 2 L2 
 
 15 
where P is the axial force in the element and L is its length.
In previous works by the authors (Martins et al., 2015a and 2015b) the equivalent lateral force
method was used to consider the second-order effects in the deck and towers. In the present
work the option was to include the geometrical nonlinearities by using the geometric stiffness
matrix. This is option was selected because a modal analysis is required for the structural
analysis under seismic action, and as stated by Abdel‐Ghaffar and Nazmy (1991) and Karoumi
(1999), for this purpose the stiffness matrix must be obtained for the structure in the dead-load
permanent state. Therefore, the geometrical nonlinear effects were considered within the
stiffness matrix instead of using equivalent lateral forces.

2.3. Structural analysis including seismic action

The analysis of any structure under seismic action involves solving the dynamic equilibrium
equation given by:
M ⋅ u&&(t ) + C ⋅ u& (t ) + K ⋅ u(t ) = M ⋅ r ⋅ u&&g (t ) (8)

where u&&(t ) , u& (t ) and u (t ) are, respectively, the vectors of acceleration, velocity and
displacement of the structure; M, C and K are, respectively, the mass, the damping and the
stiffness matrices of the structure; r is an influence matrix relating the degrees of freedom of

{ }
the structure and the ground acceleration components u&&g (t ) = u&&gX , u&&Yg , u&&gZ .
T

The structural response due to the seismic action can be obtained by step-by-step integration of
the dynamic equilibrium equation or by a modal/spectral approach. In this work the latter
approach was adopted because it is particularly suited for an integrated analysis-and-
optimization process (Negrão and Simões, 1999).
For the modal analysis a lumped mass matrix formulation was considered for the truss and beam
elements used in modelling the bridge. As previously referred, the stiffness matrix of the
structure was computed in the dead-load permanent state and includes the elastic and geometric
contributions. The eigenvalue and eigenvector problem for obtaining the natural vibration
frequencies and the corresponding mode shapes can be expressed as
[K − λ ⋅ M ]⋅ φ = 0 (9)

where λ = ω 2 are the eigenvalues or characteristic values indicating the square of the free
vibration frequencies ( ω ) and φ represents the eigenvectors or mode shapes of the vibrating

system (Clough and Penzien, 2003). This problem was solved using Lanczos method (Lanczos,
1950).
Assuming an elastic behaviour, a uniform support excitation and that no control devices are
present, the structural response due to the seismic action was evaluated with a modal/spectral
approach. This gives a set of pseudo-static forces leading to an envelope of the critical structural
responses throughout the vibration process.
The seismic action was quantified according to the Eurocode 8 (EN 1998-1-1, 2010) elastic
response spectra from which the maximum spectral accelerations, S ai (ξ i , Ti ) , are obtained

according to the damping ratio, ξ i , and vibration period, Ti , of each mode. A constant damping

ratio of ξ = 2% was considered (Abdel‐Ghaffar and Nazmy, 1991; Morgenthal, 1999; Camara
and Astiz, 2014).
The maximum modal force vector for any of the p vibration modes under consideration is
obtained from the expression
Γi
f i ,max = M ⋅ φ i S ai (ξ , Ti ) (10)
Mi

( )( )
where φ i is the i-th eigenvector, Γi = φ i ⋅ M ⋅ r / φ i ⋅ M ⋅ φ i is the modal participation factor
T T

and M i = φ i ⋅ M ⋅ φ i is the modal mass of the corresponding mode.

T
The complete quadratic combination (CQC) (Wilson et al., 1981) was used for the combination
of the maximum contribution of each mode, because as stated by Walker and Stafford (2010)
is the most adequate modal combination option to evaluate the response of cable-stayed bridges.
This is due to the modal coupling that is present in the dynamic response of these bridges. Using
the CQC, the forces for each of the k degree of freedom are combined applying the expression
p p
fk =  f
i =1 j =1
ki ⋅ ρ ij ⋅ f kj (11)

Assuming a constant modal damping ratio, the cross-modal coefficients, ρij, are given by
8ξ 2 (1 + r ) ⋅ r 3 / 2
ρ ij = (12)
(1 + r )
2 2
+ 4ξ 2 r ⋅ (1 + r )
2

( ) (
where r = min ωi , ω j / max ωi , ω j ).

3. Optimization strategy

3.1. Objective function

The design of concrete cable-stayed bridges under seismic action is formulated as a multi-
objective optimization problem leading to an optimum solution in the Pareto sense. This can be
expressed as a minimax problem, which is discontinuous and non-differentiable and therefore
poses difficulties to its numerical solution
Minx Max j (g1 , g 2 ,..., g j ) (13)
However, using the maximum entropy principle the solution of this problem can be obtained
indirectly by means of a constraint aggregation approach and thus through the minimization of
an unconstrained convex scalar function (Simões and Templeman, 1989) given by

1  M ρ ( g j ( x )) 
min F ( x ) = min ln  e  (14)
ρ  j =1 
where x = {x1 , x 2 , ..., x N }T is the vector of N design variables and ρ is a parameter which must
not be decreased through the analysis and optimization process.
Since the design objectives, g j (x ) , do not have an explicit algebraic form, the problem is

solved by an explicit approximation given by taking Taylor series expansions of all the
objectives, around the current design variable vector, truncated after the linear term
 N dg ( x )
 
Δxi  

1  M ρ g j ( x )+ 
 j

min F ( x ) = min ln  e

dxi 
i =1
(15)
ρ  j =1 
 
where M is the number of design objectives and dg j ( x ) / dx i represents the sensitivity of the

j-th design objective with respect to i-th design variable. A detailed description of the
sensitivities calculation is described is section 3.4. The MATLAB function fmincon, which
minimizes a scalar function of several variables subjected to bound constraints using a sequence
of quadratic problems, was used to minimize the objective function expressed by Equation 15.

3.2. Design variables

Two sets of design variables were considered: sizing and mechanical. The first set concerns the
cross-sectional dimensions of towers, deck and cable-stays, which have direct impact in the
mass, stiffness and cost of the structure. The second set represents the cable-stays prestressing
forces which do not have direct impact in cost but play a central role in the behaviour of cable-
stayed bridges. Figure 2 presents the types of design variables considered.
Cable-stay element Cables cross section

Force
Area

Force

Rectangular hollow section (towers) Beam-and-slab cross section (deck)

tf 0.20 m

tw h
h
b b
17.00 m
b

Figure 2 – Design variables

3.3. Design objectives

The structural design, besides satisfying safety and service criteria, should also seek economic
solutions. Considering this the first objective is related to the minimization of the structure cost
and can be stated as

g1 ( x ) =
C
−1 ≤ 0 (16)
C0
where C is the current cost of the structure and C0 is a reference cost, that represents the initial
cost of each analysis and optimization cycle. This approach makes the cost always one of the
primary objectives for the optimization algorithm. The cost of the structure was expressed in
terms of cost of the volume of materials (structural concrete and prestressing steel for the cable-
stays). The materials’ unit prices were obtained from consulting Portuguese bridge designers.
A second set of objectives is defined to limit the deck vertical displacements and the towers
horizontal displacements to attain the desired final deck profile and to minimize the tower
bending deflections
δ
g 2 (x ) = −1 ≤ 0 (17)
δ0
where δ and δ0 are, respectively, the displacement value and the limit value for the displacement
under control.
The stress objectives for the concrete members were established according to the Eurocode 2
(EN 1992-1-1, 2010) provisions. Regarding the service conditions, the acting stress, σ c , in the
concrete members must lie within 45% of the characteristic value of the concrete compressive
strength (fck) and the mean value of axial tensile strength of concrete (fctm). The former limit
ensures that the concrete remains within the range of linear creep and also the longitudinal
cracking is avoided. The latter limit is imposed to prevent cracking and, thus, ensuring
durability. These goals can be expressed by
σc
g 3 (x ) = −1 ≤ 0 (18)
0.45 f ck

σc
g 4 (x ) = −1 ≤ 0 (19)
f ctm
where the tensile and compressive stresses are used considering the respective signals.
An essential aspect when considering the seismic action is the strength verification of the
different structural members, which involves some complexity for structural concrete. For
members under bending and axial force the corresponding objective can be formulated as
ω req
g 5 (x ) = −1 ≤ 0 (20)
ω exist
where ω req and ω exist are the required and existent mechanical reinforcement ratios,

respectively. The existent mechanical reinforcement ratio is given by

As f yd
ω exist = (21)
Ac f cd
where As and Ac are the reinforcing steel and concrete cross-sectional areas, respectively; fyd
and fcd are the design values of the yield strength of reinforcement and the concrete compressive
strength, respectively. The required mechanical reinforcement ratio is dependent upon the
cross-sectional dimensions, the concrete grade and the acting axial force and bending moment.
This parameter is computed according to the formulas that describe the non-dimensional
bending moment - axial force interaction diagram (Arga e Lima et al., 1999).
For members under biaxial bending and axial force the corresponding objective can be
expressed as
a
  µ Edy
a
µ 
g 6 ( x ) =  Edz  +   −1 ≤ 0 (22)
 
 µ Rdz   µ Rdy 
where µ Ed and µ Rd are the normalized values of the acting and resistant bending moments,

respectively and a is parameter that depends from the shape of the cross-section and from the
ratio of the acting axial force (NEd) with respect to the resistant axial force (NRd) in pure
compression. The normalized values of the acting bending moments are given by
M Edz M Edy
µ Edz = , µ Edy = (23)
Ac ⋅ hz ⋅ f cd Ac ⋅ h y ⋅ f cd

Where MEdz and MEdy are the acting bending moments; hz and hy are the height of the cross-
section for the direction considered. The normalized values of the resistant bending moments
are calculated with the same interaction formulas referred for bending and axial force (Arga e
Lima et al., 1999). The resistant bending moments are dependent from the acting axial force,
the cross-sectional dimensions, the amount of reinforcing steel and the concrete grade.
The shear force resistance of the concrete members was also considered. This goal can be
expressed as
V Ed
g 7 (x ) = −1 ≤ 0 (24)
V Rd

where VEd and VRd are the acting and resistant values of the shear force.

The most difficult part of handling the objectives given by Equations 20, 22 and 24 is that,
besides the acting internal forces, the values of the resistance are also dependent on the design
variables, namely the cross-sectional dimensions. This poses an additional effort computing the
required sensitivities for the optimization algorithm.
Another set of objectives is related to the stresses in the cable-stays which can be written as
 g 8 (x ) =
σ
−1 ≤ 0 (25)
k ⋅ f pk

g 9 (x ) = 1 −
σ
≤0 (26)
0.1 f pk

where σ and fpk are the acting stress in the stays and the characteristic value of the prestressing
steel tensile strength, respectively. The value of k in Equation 26 is equal to 0.50 for service
conditions and 0.74 for strength verification (EN 1993-1-11 2006).
A last set of goals is related to the dynamic properties of the structure, namely the natural
frequencies and can be posed as
λ
g10 ( x ) = −1 ≤ 0 (27)
λ0
where λ 0 is a reference value for the vibration frequency under consideration.

3.4. Sensitivity analysis

The sensitivity analysis is a fundamental feature when using direct search optimization
algorithms from which depends the evolution of the analysis and optimization process and its
accuracy. From the several methods available to compute sensitivities, the analytical discrete
direct method was chosen. This is justified by accuracy purposes, by the availability of the
source code and given the large number of objectives when compared to the number of design
variables.
The sensitivities of displacements, u, with respect to the design variables are obtained by
differentiating the equilibrium equations
K ⋅u = F (28)
leading to

du d F d K du −1  d F dK  du −1
K = − u = K ⋅  − u   = K ⋅ Q vi (29)
dxi dxi dxi dxi  dxi dx i  dx i

where F is the global force vector of the and Q vi is the virtual pseudo-load vector of the system

with respect to the i-th design variable.

The stress sensitivities, which are computed from the chain derivation of the finite element
stress-displacement relation, are given by

dσ
e
=
d D⋅B( e
)
e du
⋅u + D ⋅ B ⋅
e
e

(30)
dxi dxi dxi
The sensitivities of the internal forces (Xe) are computed from the relation between internal
forces and stresses, which can be written as
e e
dX dσ e dCSP
= ⋅ CSP + σ ⋅ (31)
dxi dxi dxi
where CSP represents cross-sectional properties according to the internal force under
consideration.
The sensitivities of the natural vibration frequencies are given by the derivation of Equation 9
and can be expressed as
dλ j T dK dM 
= φ i  −λj  ⋅ φ i (32)
dxi  dxi dxi 
Concerning the sensitivities of the objectives related to the strength verification of concrete
members (Equations 20, 22 and 24) and since the resistance depends from the design variables,
the corresponding sensitivities are given by
dg 5 ( x ) d  ω req  1  dω req dω 
=   =  ⋅ ω exist − ω req ⋅ exist  (33)
dxi dxi  ω exist  ω exist
2
 dxi dxi 

dg 6 ( x )
a
d  µ Ed   dµ Ed a dµ Rd 
a
1  
=   = ⋅ µ − µ ⋅ (34)
dxi  µ Rd µ Rd 2 a  dx 
Rd Ed
dxi   i dxi 

dg 7 ( x ) d  V Ed  1  dV Ed dV 
=   =  ⋅ V Rd − V Ed ⋅ Rd  (35)
dx i dxi  V Rd  V Rd
2
 dxi dxi 
To ensure the accuracy of a linear sensitivity analysis procedure within a non-linear structural
analysis, upper and lower bound constraints with move limits between 1.5% and 5% of the
values of the design variables, were imposed.

4. Numerical examples and results

4.1. Description of the numerical model

The numerical model comprises of a symmetrical concrete cable-stayed bridge with a total
length of 312 m and a main span of 160 m. The total height of the towers is 64 m, with the deck
placed 20 m above the foundation. A semi-harp cable arrangement with lateral suspension and
a total of 80 cables was considered. The cable spacing is 8 m on the deck and 1.5 m on the
towers.
The deck is simply supported at the abutments and continuously supported along the bridge
length by the inclined cable-stays. The deck is not supported at the towers thus a solution with
full suspension was adopted. “H”-shaped towers with fixed supports were considered, since the
soil-structure interaction was disregarded. A beam-and-slab cross-section was considered for
the deck and was modelled with longitudinal and transversal beams. Rectangular hollow
sections were adopted for the towers. The finite element model of the bridge has a total of 218
nodes and 375 finite elements. Table 1 presents the properties of the materials used and Figure
3 shows the finite element mesh of the bridge example.

Table 1 – Material properties

Concrete C40/50 (Deck and Towers)
E = 35 GPa; γ = 25 kN/m3
fck = 40 MPa; fctm = 3.5 MPa; fcd = 26.7 MPa
Mean Relative Humidity = 70%
Type N cement
Cost: 450 €/m3 (including reinforcement)
Prestressing steel – Cable-stays
E = 195 GPa; γ = 77 kN/m3
fpk = 1860 MPa; fp0,1k = 1770 MPa
Cost: 15.000 €/m3 + 18.500 €/cable
Reinforcing steel – A500 NR
E = 200 GPa; fyk = 500 MPa; fyd = 435 MPa
εyk = 2.5×10-3; εyd = 2.174×10-3

Figure 3 – Finite element mesh of the bridge example

Six load cases were defined. The first one represents the end of construction and the bridge is
considered under self-weight and an additional permanent load of 2.5 kN/m2 (flooring,
walkways, safety barriers and guardrails). The second one refers to the long term analysis
(10000 days) of the bridge under permanent load. Three additional load cases for the bridge
under permanent load plus a live load of 4 kN/m2, corresponding to the road traffic, were
defined. The live load was applied on the whole deck or only on central or side spans. The last
load case concerns the bridge under permanent load and subjected to the seismic action.
As previously referred, the seismic action was computed according to the Eurocode 8 (EN
1998-1-1, 2010) elastic response spectra which was defined considering a type A ground (rocky
soil) and type 1 spectrum (more dangerous because presents higher spectral accelerations for
the long period area that characterizes of these structures). The value of the design ground
acceleration, a g = 0.5g , was defined according to the Portuguese National Annex for a seismic

zone 1 and an importance class IV. Based on the same document the design ground acceleration
in the vertical direction was taken as avg = 0.75a g . Since cable-stayed bridges present long

vibration period, the requirements about spectra for periods longer than 4 s have been followed.
Figure 4 presents the response spectra used to characterize the seismic action.

2.0
1.8
Spectral acceleration, Sa [g]

1.6
1.4 Horizontal
1.2 Vertical
1.0
0.8
0.6
0.4
0.2
0.0
0 1 2 3 4 5 6
Period, T [s]

Figure 4 – Response spectra used to characterize the seismic action

According to Eurocode 8 (NP EN1998-1-1) the 30% rule was used to combine the seismic
effects in the horizontal and vertical directions.
The construction stages were not directly considered in this numerical example which focuses
on the seismic behaviour of the entire bridge. The problem of optimization of concrete cable-
stayed bridges considering construction stages was solved in previous papers by the authors
(Martins et al., 2015a and 2016).
A total of 50 design variables were considered and are presented in Table 2. The cables were
numbered from the abutment to the central span. The reinforcing steel area was not considered
as a design variable and was assumed a constant design parameter expressed as As = 0.02 ⋅ Ac .
The design objectives, approximately 1200 for the six load cases, were defined taking into
account the formulation presented in section 3.3 and the different load cases considered.

4.2. Optimization results

Based on the bridge numerical model described in section 4.1 two examples were analysed. In
Example 1 the bridge was optimized considering the five static load cases. Example 2 concerns
the optimization of the bridge under static loading and seismic action.
The evolution of the bridge cost throughout the optimization process is presented in Figure 5.
It is worth noting that the initial solutions for the two examples are different since the structural
behaviour is strongly influenced by the seismic action and a feasible solution for static loading
is unfeasible when considering seismic loading. Therefore, in Figure 5 is represented the
normalized cost with respect to the initial cost. The optimum solution for Example 1 presents a
cost reduction of 26.8% with respect to the initial solution. In Example 3 despite an initial cost
increase it can be noticed a cost reduction of 17.7% with respect to the initial solution. The
optimum solution for Example 2 presents a cost 2 times higher than the cost obtained for
Example 1. This is due to the greater demands that the seismic action poses in the strength
verification of the structural members mainly the towers. This emphasizes the relevance of the
seismic action in the design of these structures.
1,20
1,15 Example 1
1,10
Example 2
Normalized bridge cost

1,05
1,00
0,95
0,90
0,85
0,80
0,75
0,70
0 5 10 15 20 25 30 35 40
Number of iterations

Figure 5 – Normalized bridge cost vs. number of iterations

Table 2 presents the description, the initial and final values of the design variables. Comparing
the optimum solutions obtained in Examples 1 and 2 it can be observed an increase in the cross-
sectional dimensions of the towers and the cables for Example 2. Also the height of the deck
beams is increased. This translates to the cost increase already stated. The solution for Example
2 presents, in average, higher cable forces. The highest forces occur in the backstays in order
to control de tower bending deflections and stresses. The optimization algorithm finds an
optimum solution rearranging the stiffness and mass distribution throughout the structure to
improve its behaviour under seismic action.

Table 2 – Design variables (initial and final values)

Example 1 Example 2
Number Design variable
Initial Final Initial Final
1 Cable force [kN] - stay 1 2500 1492 2500 2092
2 Cable force [kN] - stay 2 2000 1464 2000 1558
3 Cable force [kN] - stay 3 2000 1963 2000 1562
4 Cable force [kN] - stay 4 2000 1912 2000 1552
5 Cable force [kN] - stay 5 2000 1805 2000 1817
6 Cable force [kN] - stay 6 2000 1652 2000 1894
7 Cable force [kN] - stay 7 2000 1744 2000 1862
8 Cable force [kN] - stay 8 2000 1978 2000 1934
9 Cable force [kN] - stay 9 2000 1946 2000 1988
10 Cable force [kN] - stay 10 2000 1955 2000 1998
11 Cable force [kN] - stay 11 2500 1481 2500 1980
12 Cable force [kN] - stay 12 2000 1391 2000 1946
13 Cable force [kN] - stay 13 2000 1373 2000 1893
14 Cable force [kN] - stay 14 2000 1380 2000 1769
15 Cable force [kN] - stay 15 2000 1304 2000 1720
16 Cable force [kN] - stay 16 2000 1450 2000 1625
17 Cable force [kN] - stay 17 2000 1456 2000 1671
18 Cable force [kN] - stay 18 2000 1628 2000 1673
19 Cable force [kN] - stay 19 2000 1792 2000 1630
20 Cable force [kN] - stay 20 2000 2051 2000 1807
21 Cable area [m2] - stay 1 3.00×10 -3
2.18×10-3 3.00×10 -3
3.20×10-3
22 Cable area [m2] - stay 2 2.25×10 -3
1.60×10-3 2.25×10 -3
2.59×10-3
23 Cable area [m2] - stay 3 2.25×10-3 1.57×10-3 2.25×10-3 2.43×10-3
24 Cable area [m2] - stay 4 2.25×10-3 1.50×10-3 2.25×10-3 2.36×10-3
25 Cable area [m2] - stay 5 2.25×10-3 1.77×10-3 2.25×10-3 2.03×10-3
26 Cable area [m2] - stay 6 2.25×10 -3
1.91×10-3 2.25×10-3 1.99×10-3
27 Cable area [m2] - stay 7 2.25×10-3 1.83×10-3 2.25×10-3 2.02×10-3
28 Cable area [m2] - stay 8 2.25×10-3 1.70×10-3 2.25×10-3 2.03×10-3
29 Cable area [m2] - stay 9 2.25×10-3 1.77×10-3 2.25×10-3 2.20×10-3
30 Cable area [m2] - stay 10 2.25×10-3 1.77×10-3 2.25×10-3 2.25×10-3
31 Cable area [m2] - stay 11 2.25×10 -3
1.67×10-3 2.25×10-3 2.12×10-3
32 Cable area [m2] - stay 12 2.25×10-3 1.70×10-3 2.25×10-3 2.15×10-3
33 Cable area [m2] - stay 13 2.25×10-3 1.79×10-3 2.25×10-3 2.46×10-3
34 Cable area [m2] - stay 14 2.25×10-3 1.83×10-3 2.25×10-3 2.66×10-3
35 Cable area [m2] - stay 15 2.25×10 -3
1.94×10-3 2.25×10-3 2.78×10-3
36 Cable area [m2] - stay 16 2.25×10-3 1.97×10-3 2.25×10-3 2.92×10-3
37 Cable area [m2] - stay 17 2.25×10-3 2.04×10-3 2.25×10-3 2.89×10-3
 38 Cable area [m2] - stay 18 2.25×10-3 1.97×10-3 2.25×10-3 2.87×10-3
39 Cable area [m2] - stay 19 3.00×10-3 1.88×10-3 3.00×10-3 2.97×10-3
40 Cable area [m2] - stay 20 2.25×10-3 1.78×10-3 2.25×10-3 2.65×10-3
41 h [m] of the deck beams 2.00 1.75 1.50 2.28
42 b [m] of the deck beams 1.00 0.47 0.60 0.50
43 h [m] of the towers under the deck 6.00 3.41 6.00 5.44
44 b [m] of the towers under the deck 5.00 2.44 5.00 3.77
45 tw [m] of the towers under the deck 3.00 1.85 6.00 5.03
46 tf [m] of the towers under the deck 3.00 1.89 5.00 4.36
47 h [m] of the towers above the deck 0.30 0.15 2.00 1.78
48 b [m] of the towers above the deck 0.30 0.15 2.00 2.05
49 tw [m] of the towers above the deck 0.30 0.16 2.00 1.79
50 tf [m] of the towers above the deck 0.30 0.15 2.00 1.80

The deformed configuration of the bridge under permanent load is presented in Figure 6 where
can be noticed the small values of the deck vertical displacements (maximum value of 10 cm)
and towers horizontal displacements (maximum value of 4 cm). Figure 7 presents the deformed
configuration of the bridge for the seismic load case were can be noticed the maximum deck
vertical displacements of approximately 37 cm and the large bending deflections of the towers
(maximum value of 42 cm). This highlights the high seismic demands for the towers of cable-
stayed bridges.

Figure 6 – Deformed configuration of the bridge under permanent load (Example 2 – optimum solution)
 Figure 7 – Deformed configuration of the bridge for the seismic load case (Example 2 – optimum solution)

The values of the design objectives concerning the strength verification of the towers are
presented in Figure 8. It can be stated that in all load cases the design objectives are satisfied
and that the parts of the tower under the deck and above until the first cables are the most
stressed. Concerning the results obtained for the modal analysis of the optimum solution
obtained for Example 2 present the following first ten vibration frequencies (0.42, 0.55, 0.79,
1.01, 1.16, 1.75, 2.06, 2.40, 3.12 and 4.76 Hz)

50,0
Service conditions
45,0
Dead Load+Live Load
40,0
Seismic Load
Towers height [m]

35,0
30,0
25,0
20,0
15,0
10,0
5,0
0,0
-1,00 -0,90 -0,80 -0,70 -0,60 -0,50 -0,40 -0,30 -0,20 -0,10 0,00
Normalized strength design objectives

Figure 8 – Values of the design objectives along the towers height

5. Conclusions

The following conclusions can be drawn:

• The design of concrete cable-stayed bridges under seismic loading can be formulated and
solved as a multi-objective optimization problem.
• The optimization algorithm finds economical and structurally efficient solutions with an
adequate mass and stiffness distribution throughout the structure that enhances its
behaviour under seismic action.
• To obtain an efficient and accurate algorithm the discrete direct method was selected for
sensitivity analysis. This required a large amount of time but proved to be adequate in
predicting the bridge structural behaviour under both, static and dynamic loading.
• The resistance of concrete members depends on the correspondent cross-sectional design
variables. This poses additional difficulties when formulating the sensitivities for the
respective design objectives. An efficient way to handle the strength design objectives of
concrete members was developed and implemented.
• The design algorithm is efficient and robust. It considers all relevant actions and effects
(time-dependent effects and geometrical nonlinearities) and allows checking the service
and strength criteria for the complete bridge under dead load, traffic load and seismic
action. Even dealing with a large number of objectives and design variables, the solutions
are obtained after a relatively small number of iterations.
• Important aspects of the seismic analysis of cable-stayed bridges, like the use of passive
and active control devices, the spatial variability of the seismic ground motion and the soil-
structure interaction, were not addressed in this article, however will they be considered in
future developments of the current research work.

References

Abdel‐Ghaffar, A. M., & Khalifa, M. A. (1991). Importance of Cable Vibration in Dynamics

of Cable‐Stayed Bridges. Journal of Engineering Mechanics, 117(11), 2571–2589.
https://doi.org/10.1061/(ASCE)0733-9399(1991)117:11(2571)
Abdel‐Ghaffar, A. M., & Nazmy, A. S. (1991). 3‐D Nonlinear Seismic Behavior of Cable‐
Stayed Bridges. Journal of Structural Engineering, 117(11), 3456–3476.
https://doi.org/10.1061/(ASCE)0733-9445(1991)117:11(3456)
Arga e Lima, J., Monteiro, V., Mun, M. (1999). Betão Armado – Esforços normais e flexão
(REBAP-83). LNEC – Laboratório Nacional de Engenharia Civil, Lisboa (in Portuguese).
Baldomir, A., & Hernández, S. (2009). Cable optimization of a long span cable stayed bridge
in La Coruña (Spain) (pp. 107–119). Apresentado na OPTI 2009, Algarve, Portugal.
https://doi.org/10.2495/OP090101
Bazant Z. P. (1988). “Material models for structural creep analysis.” In Mathematical
modelling of creep and shrinkage of concrete, edited by Z. P. Bazant, 99-215. John Wiley and
Sons, Ltd.
Caetano, E., Cunha, A., Gattulli, V., & Lepidi, M. (2008). Cable–deck dynamic interactions at
the International Guadiana Bridge: On-site measurements and finite element modelling.
Structural Control and Health Monitoring, 15(3), 237–264. https://doi.org/10.1002/stc.241
Cámara, A., & Astiz, M. Á. (2014). Aplicabilidad de las diversas estrategias de análisis sísmico
en puentes atirantados en rango elástico. Revista Internacional de Métodos Numéricos para
Cálculo y Diseño en Ingeniería, 30(1), 42–50. https://doi.org/10.1016/j.rimni.2012.10.001
Camara, A., & Efthymiou, E. (2016). Deck–tower interaction in the transverse seismic response
of cable-stayed bridges and optimum configurations. Engineering Structures, 124, 494–506.
https://doi.org/10.1016/j.engstruct.2016.06.017
Clough, R.W. and Penzien, J. (2003). Dynamics of Structures, Third edition, Computers and
Structures, Inc., USA.
EN 1992-1-1 (2010). NP EN 1992-1-1 Eurocódigo 2 – Projecto de estruturas de betão, Parte
1-1: Regras gerais e regras para edifícios. IPQ – Instituto Português da Qualidade.
EN 1993-1-11 (2006). EN 1993-1-11 Eurocode 3 – Design of steel structures, Part 1-11:
Design of structures with tension components. CEN – Comité Européen de Normalisation.
EN 1998-1-1 (2010). NP EN 1998-1-1 Eurocódigo 8 – Projecto de estruturas para resistência
aos sismos, Parte 1: Regras gerais, acções sísmicas e regras para edifícios. IPQ – Instituto
Português da Qualidade.
Ernst, J.H. (1965). Der E-Modul von Seilen unter berucksichtigung des Durchhanges. Der
Bauingenieur, 40(2), 52–55.
Ferreira, F. L. S., & Simoes, L. M. C. (2011). Optimum design of a controlled cable stayed
bridge subject to earthquakes. Structural and Multidisciplinary Optimization, 44(4), 517–528.
https://doi.org/10.1007/s00158-011-0628-9
Ferreira, F., & Simões, L. (2012). Optimum cost design of controlled cable stayed footbridges.
Computers & Structures, 106–107, 135–143. https://doi.org/10.1016/j.compstruc.2012.04.013
Freire, A. M. S., Negrão, J. H. O., & Lopes, A. V. (2006). Geometrical nonlinearities on the
static analysis of highly flexible steel cable-stayed bridges. Computers & Structures, 84(31–
32), 2128–2140. https://doi.org/10.1016/j.compstruc.2006.08.047
Hassan, M.M. (2013). Optimization of stay cables in cable-stayed bridges using finite element,
genetic algorithm, and B-spline combined technique. Engineering Structures, 49, 643–654.
https://doi.org/10.1016/j.engstruct.2012.11.036
Hassan, M.M., Nassef, A. O., & El Damatty, A. A. (2012). Determination of optimum post-
tensioning cable forces of cable-stayed bridges. Engineering Structures, 44, 248–259.
https://doi.org/10.1016/j.engstruct.2012.06.009
Hassan, M.M., Nassef, A. O., & El Damatty, A. A. (2013). Optimal design of semi-fan cable-
stayed bridges. Canadian Journal of Civil Engineering, 40(3), 285–297.
https://doi.org/10.1139/cjce-2012-0032
Hassan, Mahmoud M., El Damatty, A. A., & Nassef, A. O. (2015). Database for the optimum
design of semi-fan composite cable-stayed bridges based on genetic algorithms. Structure and
Infrastructure Engineering, 11(8), 1054–1068. https://doi.org/10.1080/15732479.2014.931976
Karoumi, R. (1999). Some modeling aspects in the nonlinear finite element analysis of cable
supported bridges. Computers & Structures, 71(4), 397–412. https://doi.org/10.1016/S0045-
7949(98)00244-2
Lanczos, C. (1950). An iteration method for the solution of the eigenvalue problem of linear
differential and integral operators. Journal of Research of the National Bureau of Standards,
45(4), 255–282.
Long, W., Troitsky, M. S., & Zielinski, Z. A. (1999). Optimum design of cable-stayed bridges.
Structural Engineering and Mechanics, 7(3), 241–257.
https://doi.org/10.12989/sem.1999.7.3.241
Martins, A.M.B., Simões, L. M. C., & Negrão, J. H. J. O. (2015a). Optimization of cable forces
on concrete cable-stayed bridges including geometrical nonlinearities. Computers & Structures,
155, 18–27. https://doi.org/10.1016/j.compstruc.2015.02.032
Martins, Alberto M. B., Simões, L. M. C., & Negrão, J. H. J. O. (2015b). Cable stretching force
optimization of concrete cable-stayed bridges including construction stages and time-dependent
effects. Structural and Multidisciplinary Optimization, 51(3), 757–772.
https://doi.org/10.1007/s00158-014-1153-4
Martins, Alberto M.B., Simões, L. M. C., & Negrão, J. H. J. O. (2016). Optimum design of
concrete cable-stayed bridges. Engineering Optimization, 48(5), 772–791.
https://doi.org/10.1080/0305215X.2015.1057057
Morgenthal, Guido (1999). Cable-stayed bridges – Earthquake response and passive control.
MSc Dissertation, Imperial College of Science, Technology and Medicine, London.
Nazmy, A. S., & Abdel-Ghaffar, A. M. (1990). Three-dimensional nonlinear static analysis of
cable-stayed bridges. Computers & Structures, 34(2), 257–271. https://doi.org/10.1016/0045-
7949(90)90369-D

Negrão, J. H. J. O., and L. M. C. Simões (1997a). “Cable stretching force optimization in cable-
stayed bridges.” Paper presented at the WCSMO-2 The Second World Congress of Structural
and Multidisciplinary Optimization, IFTR: 983-987.
Negrão, J. H. O., & Simões, L. M. C. (1997b). Optimization of cable-stayed bridges with three-
dimensional modelling. Computers & Structures, 64(1–4), 741–758.
https://doi.org/10.1016/S0045-7949(96)00166-6

Ohkubo S., and K. Taniwaki (1991). “Shape and sizing optimization of steel cable-stayed
bridges.” In Proceedings of OPTI 91 – Optimization of Structural Systems and Industrial
Applications, edited by S. Hernandez and C. A. Brebbia, Elsevier Applied Sciences,
Cambridge, MA, USA.
Simões L. M. C., & Templeman, A. B. (1989). ENTROPY-BASED SYNTHESIS OF
PRETENSIONED CABLE NET STRUCTURES. Engineering Optimization, 15(2), 121–140.
https://doi.org/10.1080/03052158908941147
Simões, L. M. C., & Negrão, J. H. J. O. (1999). Optimization of cable-stayed bridges subjected
to earthquakes with non-linear behaviour. Engineering Optimization, 31(4), 457–478.
https://doi.org/10.1080/03052159908941382
Simões, L. M. C., & Negrão, J. H. O. (1994). Sizing and geometry optimization of cable-stayed
bridges. Computers & Structures, 52(2), 309–321. https://doi.org/10.1016/0045-
7949(94)90283-6
Simões, L. M., & Negrão, J. H. J. (2000). Optimization of cable-stayed bridges with box-girder
decks. Advances in Engineering Software, 31(6), 417–423. https://doi.org/10.1016/S0965-
9978(00)00003-X
Soneji, B. B., & Jangid, R. S. (2008). Influence of soil–structure interaction on the response
of seismically isolated cable-stayed bridge. Soil Dynamics and Earthquake Engineering,
28(4), 245–257. https://doi.org/10.1016/j.soildyn.2007.06.005
Sung, Y.-C., Chang, D.-W., & Teo, E.-H. (2006). Optimum post-tensioning cable forces of
Mau-Lo Hsi cable-stayed bridge. Engineering Structures, 28(10), 1407–1417.
https://doi.org/10.1016/j.engstruct.2006.01.009
Svensson, H. (2012). Cable-stayed bridges: 40 years of experience worldwide (1. ed). Berlin:
Ernst, Wiley-Blackwell.
 Torii K, and K. Ikeda (1987). “A study of the optimum design method for cable-stayed bridges.”
Paper presented at the International Conference on Cable-Stayed Bridges, Bangkok, Thailand.
Walker, C., & Stafford, P. J. (2010). The use of modal-combination rules with cable-stayed
bridges. Proceedings of the Institution of Civil Engineers - Bridge Engineering, 163(4), 225–
240. https://doi.org/10.1680/bren.2010.4.225
Walther, R. (Ed.). (1999). Cable stayed bridges (2. ed). London: Telford.
Wilson, E. L., Der Kiureghian, A., & Bayo, E. P. (1981). A replacement for the srss method in
seismic analysis. Earthquake Engineering & Structural Dynamics, 9(2), 187–192.
https://doi.org/10.1002/eqe.4290090207

View publication stats