Acta Mech

https://doi.org/10.1007/s00707-020-02802-8

O R I G I NA L PA P E R

Yunyue Cong · Houjun Kang · Guirong Yan · Tieding Guo

Modeling, dynamics, and parametric studies of a

multi-cable-stayed beam model

Received: 7 May 2020 / Revised: 26 July 2020 / Accepted: 13 August 2020

© Springer-Verlag GmbH Austria, part of Springer Nature 2020

Abstract To study the dynamic behavior of cable-stayed bridges, a linear multi-cable-stayed beam model is
developed to investigate its in-plane and out-of-plane transverse vibrations. From the full-bridge perspective,
an in-plane nonlinear single-mode discrete model is established. First, the in-plane and out-of-plane motion
equations of the system and their boundary conditions are derived. Second, by employing the separation of
variables method, the linear eigenvalue problems are solved. The influences of the mass ratio, stiffness ratio,
and cable sag on the occurrence of global, local, and coupled vibration modes are studied. Third, frequency
response, amplitude response, phase diagram, time history, and power spectrum are extracted to investigate
the system’s nonlinear dynamic behaviors. The obtained results demonstrate that for the in-plane motion, the
occurrence of global and local modes of the system depends on the mass and stiffness ratios between cable
and beam significantly; for the out-of-plane motion, without the elastic support of the cable, the global modes
occur, which can be suppressed by adjusting the mass and stiffness ratios between cable and beam but may in
turn induce the cable’s local vibration modes; for the nonlinear analysis, the single-degree-of-freedom system
behaves like a hardening spring. Its lower branch behaves more complicated than the higher one and has a
double-periodic steady-state solution. The system with large damping ratio behaves shows weak hardening
spring property.

1 Introduction

Among the infrastructure systems, cable-stayed bridges have been popular due to their large spanning capac-
ity, rational mechanical properties, reasonable cost, and mature construction technologies. In the past several
decades, the main span has been continuously increasing and currently the span has reached more than one
kilometer. The larger span bridges bring new challenges for design and construction. Due to the long spanning
of cable-stayed bridges, their dynamic behaviors become more complicated and more sensitive to small distur-
bances due to coupling interactions, making structural analysis more complicated. Therefore, it is imperative
to extensively investigate the dynamic behaviors and coupling properties from different perspectives.
Thus far, researchers have carried out a lot of work on the family of cable-beam models, consisting of a
cable and a beam with different type of excitations, in order to explore their interaction mechanism. In 1993,
Fujino et al. [1] first proposed a 3-DOF analytical model of a cable-stayed beam based on the local–global mode
approach and studied the auto-parametric resonance of the model under harmonic excitation experimentally and

Y. Cong · H. Kang (B) · T. Guo

College of Civil Engineering and Architecture, Guangxi University, Guangxi 530004, China
E-mail: [email protected]

Y. Cong · G. Yan
Department of Civil, Architectural and Environmental Engineering, Missouri University of Science and Technology, Rolla, MO
65401, USA
 Y. Cong et al.

analytically. Then, Warnitchai et al. [2] generalized the 3-DOF model and established a cable model with small
sag where small support motions as well as geometrical nonlinearities were included, which then was applied
to constitute the cable-structure system. Xia et al. [3] studied the auto-parametric vibration of a cable-stayed
beam structure under a random excitation with emphasis on the cable’s local vibration. Gattulli [4] developed
a cable-stayed beam model to understand the complex interaction between a single cable and a beam. Later
on, he and his coauthors [5] performed a parametric investigation of linear and nonlinear behaviors of the
same cable-stayed beam model. Fung et al. [6] studied the nonlinear oscillation of the flexible cable-stayed
beam with the cable having time-varying length and tension. Lenci et al. [7] analyzed the nonlinear dynamics
of a Helmholtz-Duffing oscillator derived by the Galerkin single-mode discretization of a cable-stayed beam.
Zhu et al. [8] established a continuous model of an inclined cable and the in-plane single-degree-of-freedom
nonlinear cable-stayed beam model with cable damage. Wei et al. [9] investigated the bifurcation and chaostic
between of a cable-beam coupled system under simultaneous internal and external resonances and three
separate loading cases acting on the beam and the cable. Then they [10] investigated the nonlinear responses
of the cable-beam system under the combined effects of 1:2 internal and 1:1 external resonances by utilizing
Galerkin’s method and the method of multiple scales. Thereafter, the nonlinear dynamics of the cable-stayed
beam driven by the subharmonic resonance of the beam and the principal parametric resonance of the cable
were further studied by Wei et al. [11]. Wang et al. [12] obtained mode shapes of the cable-stayed beam by
using a piecewise function and investigated nonlinear modal properties of the system with a approach and
a discretization approach. Furthermore, they [13] extensively studied nonlinear normal modes of the cable-
stayed beam based on energy-based formulation and investigated the influence of nonlinear coupling terms on
the modal properties of the system. Peng et al. [14] established a cable-stayed beam with time delay feedback,
aiming to study the nonlinear primary resonance in the vibration control. Wang [15] investigated the dynamic
behavior of a cable-stayed beam with the beam nonlinearity by using the Lagrangian multiplier method and
conducted a parametric analysis on the in-plane and out-of-plane eigen-problems and experimental analysis.
Liu et al. [16] studied the stability and dynamic analysis of in-plane parametric vibration of a cable in a
cable-stayed bridge with superlong spans subjected to axial excitation by using the multiscale method and
also investigated the influence of the cables’ damping ratio, excitation frequency, and amplitude on the cable
vibration. Liu and Xiang [17] conducted the parametric study of the CFRP cable regarding the global modal
frequency of the overall cable-stayed bridge by using the transfer matrix method; the results showed that the
global modal frequency is sensitive to the stiffness of the cables. Zhang et al. [18] studied the static performance
of a long-span concrete cable-stayed bridge subjected to multiple cable loss during construction by applying
the field inspection and finite element analysis. However, the work on the cable-beam or cable system can only
reflect the mechanical properties and dynamic behavior of cable-stayed bridge in the construction stage. There
are still many dynamic behaviors worth to be explored through a more complicated model.
From the perspective of a full-bridge model, there are few theoretical studies to explore the interaction
among its different components, namely, cables and bridge deck. Cao et al. [19] established an in-plane model
of a cable-stayed bridge that consists of a simply-supported four cable-stayed deck beam and two rigid towers.
Then, Song et al. [20] further studied the dynamic behavior of the model subjected to a moving vehicle load
with constant velocity using the Galerkin method in conjunction with the Runge–Kutta–Fehlherg method in
MATLAB. Kang et al. [21] established a novel nonlinear dynamic double-cable-stayed shallow arch model
to model a cable-stayed bridge, where the in-plane 1:1:1 internal resonance among the first three modes of a
shallow, arch and two cables under external primary and subharmonic resonances were investigated. Based on
this model, Cong et al. [22] studied planar 1:2:2 internal resonances among the first three modes of cables and the
shallow arch with external harmonic excitation applied on the shallow arch, and obtained exact eigensolutions
of the linearized model simultaneously. Then, they [23] studied the nonlinear dynamic behavior of the system
when cables are subjected to harmonic excitations and the 1:1:1 internal resonance.
Indeed, the internal and external resonances of different models of the cable-stayed bridge have been
extensively investigated by discretizing it into a multiple degree of freedom system. Despite previous research,
taking the cable-stayed bridge as a united system has never been studied from the overall perspective. This
research is badly needed, as it can provide the overall structural responses, which are used for active control
of this type of bridge [14]. To address this research gap, in this study, a linear multi-cable-stayed beam
model including the in-plane and out-of-plane transverse vibrations is developed to investigate the system’s
local, global, and coupling modes. From the full-bridge perspective, a united in-plane nonlinear single-mode
dynamic model is developed. Previous research was mainly focused on dynamic responses under a single
periodic excitation, which ignored environmental loads (they can be modeled by a finite sum of harmonic
terms). To model the actions induced by winds, rain, and traffic loads, multi-frequency excitations should
Modeling, dynamics, and parametric studies

be considered. In this study, a representative case, a two-frequency excitation, is applied and the induced
dynamic responses are studied. The remainder of this paper is organized as follows. In Sect. 2, the multi-cable-
stayed beam model is established with the geometrical and mechanical boundary conditions. Meanwhile, the
linearized equations for governing in-plane and out-of-plane transverse vibrations are derived, and the in-plane
single-degree-of-freedom (SDOF) discrete nonlinear model of the system is developed. In Sect. 3, the first 10
natural frequencies of the system are extracted from modal analysis, where the linear theoretical validity is
verified by a finite element model with the same parameters. In Sect. 4, parametric studies of the system based
on the linearized model are conducted, where the relations of natural frequencies between the subsystems
(the beam and the cable) and the overall system are plotted to explore modal properties of the system. Next,
parametric analysis of the nonlinear SDOF system regarding single-frequency and two-frequency excitations
are conducted based on the frequency response, amplitude response, phase diagram, time history and power
spectrum in Sect. 5. In Sect. 6, conclusions and some interesting results are obtained.

2 Problem formulation

A cable-stayed beam model with multiple cables and two rigid towers is considered, as shown in Fig. 1. The
beam is pinned at its both ends, and each cable is anchored to the rigid tower at its upper end and connected
with the beam at its lower end, designated as Junctions s j . All the junctions divide the beam into a number of
segments and each segment is designated as Beam i. The static equilibrium configuration of this beam model
is described by the cable displacement components Ucj , Vcj , Wcj ( j = 1, 2, . . . , n) and by the beam transverse
displacements Vbi (i = 1, 2, . . . , n + 1), as shown in Fig. 1. It is assumed that the deformation of each cable
follows the parabolic curve Ycj = 4Dcj [X cj /lcj − (X cj /lcj )2 ], due to the small ratio of sag (Dcj ) to length
(lcj ), which is less than 1/10.
The towers are assumed to be rigid, because vibrations on towers are minimal, which has been demonstrated
by experimental measurements and finite element analysis of a real cable-stayed bridge [1]. The axial motion
of the beam is negligible, because the axial stiffness of the beam is much larger than the bending stiffness in
transverse vibration, namely, E b Ab /lb >> 48E b Ib /lb3 . In addition, the influence of the horizontal component
of the cable tension on the mechanical properties of the entire system can be neglected, since it is very weak,
which had been verified in [5,12]. Additionally, the following assumptions [5] are introduced:
(a) The homogenous one-dimensional elastic continuum obeying a linear stress–strain relation of cables and
beam is considered;
(b) The Lagrangian strain of the centerline is used to describe the axial extensions of the cables;
(c) The longitudinal inertia force, flexural, torsional and shear rigidities of cables are neglected;
(d) For the beam, the axial, torsional, shear strain, and the geometric nonlinearities are neglected.
Based on the above assumptions and simplification, by using the classical Hamilton principle and the
standard condensation procedure [24], the following in-plane and out-of-plane motion equations governing
the transverse vibrations of the cable and the sub-beam are obtained:
 
m cj V̈cj + ζcj V̇cj − [Hcj Vcj + E cj Acj Ycj + Vcj ecj ] = Pvcj , (1)
m cj Ẅcj + ζcj Ẇcj − [Hcj Wcj + E cj Acj Wcj ecj ] = Pwcj , (2)

Fig. 1 Configuration of the multi-cable-stayed beam model

 Y. Cong et al.

Fig. 2 Diagram of in-plane static equilibrium at Junction s j

m bi V̈bi + ζbi V̇bi + E bi Ivbi Vbiiv = Pvbi , (3)

m bi Ẅbi + ζbi Ẇbi + E bi Iwbi Wbiiv = Pwbi , (4)

where m bi , ζbi , E bi and Ivbi (Iwbi ) are the mass per unit length, damping factor, Young’s modulus and the
in-plane (out-of-plane) second moment inertia of the cross section of the i-th sub-beam, respectively. m cj ,
ζcj , Hcj , E cj and Acj are the mass per unit length, damping factor, the initial tension, Young’s modulus and
cross-section area of the j-th cable, respectively. Pvbi (Pwbi ) and Pvcj (Pwcj ) are the in-plane (out-of-plane)
external excitations that are distributed along the cables and beams, respectively. ecj represents the uniform
dynamic elongation of the j-th cable, which is expressed as
 lcj  
1 2 1 2
ecj = Vcj (X cj , t) tan θ j + Ycj V  cj + V  cj + W  cj dXcj . (5)
0 2 2

For the above partial differential equations of the multi-cable-stayed beam model, Eqs. (1) to (4), the following
geometric boundary and compliance conditions should be satisfied:
⎧
⎪ Vb1 (0, t) = Vb(n+1) (lb , t) = Wb1 (0, t) = Wb(n+1) (lb , t) = 0,
⎪
⎪
⎪
⎪  
⎪ Vbi (S j , t) = Vb(i+1) (S j , t), Vbi (S j , t) = Vb(i+1) (S j , t),
⎪
⎪
⎨ W (S , t) = W  
bi j b(i+1) (S j , t), Wbi (S j , t) = Wb(i+1) (S j , t),
(6)
⎪ Ucj (lcj , t) sin θ j + Vcj (lcj , t) cos θ j = Vbi (S j , t),
⎪
⎪
⎪
⎪
⎪ Ucj (lcj , t) cos θ j − Vcj (lcj , t) sin θ j = 0,
⎪
⎪
⎩
Wcj (lcj , t) = Wbi (S j , t), Vcj (0, t) = Wcj (0, t) = 0,

and the following relevant mechanical continuous conditions at each junction, as given in Eq. (7), should be
satisfied. This can be obtained from the variational process. Figure 2 presents the diagram of in-plane static
equilibrium at each junction.
In Fig. 2, Q i and Mi are the shear forces and bending moments acting on the left side of the junction,
and Q i+1 and Mi+1 are for the right side of the junction; Tx j = E cj Acj ecj and Ty j = Hcj Vcj ’(lcj , t) +
E cj Acj ecj (Vcj ’(lcj , t) + Ycj ’(lcj )) are the longitudinal and transverse components of the tension in the j-th
cable, respectively [19]. Imposing the force balance at each junction, the mechanical condition can be obtained:
⎧
⎪ E bi Ibi Vbi (S j , t) − E b(i+1) Ib(i+1) Vb(i+1)
 (S j , t) = E cj Acj ecj sin θ j
⎪
⎪
⎪
⎪
⎪
⎪ +[Hcj Vcj (lcj , t) + E cj Acj ecj (Vcj (lcj , t) + Ycj (lcj ))] cos θ j ,
 
⎪
⎨
E bi Ibi Wbi (S j , t) − E b(i+1) Ib(i+1) Wb(i+1)
 (S j , t) = Hcj Wcj (lcj , t) + E cj Acj ecj Wcj (lcj , t), (7)
⎪
⎪
⎪
⎪ Vbi (S j , t) = Vb(i+1)
 (S j , t), Wbi (S j , t) = Wb(i+1)
 (S j , t),
⎪
⎪
⎪
⎪  (0, t) = V   
⎩ Vb1 b(n+1) (lb , t) = Wb1 (0, t) = Wb(n+1) (lb , t) = 0.

Previous research showed that that an asymmetrical bridge may be easily damaged since the vibration energy
may be concentrated on a fewer number of cables in certain situations where the excitation frequencies are in the
vicinity of the natural frequencies corresponding to the localized mode shapes [19]. Hence, a symmetrical model
is introduced in this work and the assumptions that m bi = m b , m cj = m c , E cj Acj = E c Ac , E bi Ivbi = E b Ivb
Modeling, dynamics, and parametric studies

and E bi Iwbi = E b Iwb are considered. A non-dimensional form of Eqs. (1) to (4), and all the relevant boundary
conditions, Eqs. (6) and (7), can be obtained by introducing the following variables:
⎧
⎨ xcj = X cj , τ = ω0 t, ycj = Ycj , u cj = Ucj , vcj = Vcj , wcj = Wcj , dcj = Dcj ,
lcj lcj lcj lcj lcj lcj
(8)
⎩ xbi = X bi , vbi = Vbi , wbi = Wbi , s j = S j , ξbi = ζbi , ξcj = ζcj ,
lb lb lb lb m bi ω0 m cj ω0

where ω0 is a given frequency, such as the first in-plane natural frequency of the system, and it can be 1.
Moreover, the following dimensionless quantities are introduced:
mc E c Ac E b Ivb lb Ivb
ρ= , μcj = , χ= 2 , γcj = , ι= . (9)
mb Hcj lb E c Ac lcj Iwb
Additionally, the following dimensionless spatial frequencies of the cable and the sub-beam are introduced:
2
m b lb4 m c lcj
βbi
2
= ωk , βcj = ωk , (k = v, w), (10)
E b Ivb Hcj
where ωk represents the dimensional temporal frequencies of the in-plane and out-of-plane modal motion
of the system, respectively. The relation of spatial frequencies between the cable and the sub-beam can be
obtained by solving the expression in Eq. (10) for ωk , and it is expressed as:
αj αj
βvcj = βvb
2
, βwcj = βwb
2
√ , (11)
γcj ιγcj
√
where α j is defined as α j = μcj ρχ .

2.1 In-plane eigenvalue problem

Expanding the dimensionless equation of the cable and neglecting the nonlinear and structural damping terms,
the linearized self-adjoint boundary value problem of the multi-cable-stayed beam model can be obtained by
employing the separation of variables method and the one-dimensional coordinate [4] xcj = xbi = x:

−βvb
4
φvbi + φ   vbi = 0, (12)

βvcj
2
φvcj +φ vcj = 8μcj dcj êcj . (13)
In addition, the linearized geometrical and mechanical boundary conditions can be obtained at the same time,
and they are expressed as
⎧
⎪
⎪ φvcj (0) = φvb1 (0) = φ  vb1 (0) = φvb(n+1) (1) = φ  vb(n+1) (1) = 0,
⎪
⎪
⎨ φvcj (1) = γcj φvbi (s j ) cos θ j ,
φvbi (s j ) = φvb(i+1) (s j ), φ  vbi (s j ) = φ  vb(i+1) (s j ), φ  vbi (s j ) = φ  vbi (s j ), (14)
⎪
⎪
⎪
⎪   cos θ j 
⎩ χ φ  (s j ) − φ  (s j ) − sin θ j + y  (1) cos θ j êcj − φ (1) = 0,
vbi vb(i+1) cj μcj vcj

where φvbi and φvcj are modal eigenfunctions. êcj is the linearized dynamic elongation of the cable, and it is
expressed as
 1
 
êcj = φcj (1) tan θ j + ycj φcj dxcj . (15)
0
It is well known that the exact closed-form solutions of Eqs. (12) and (13) have the following form:
φvbi (x) = cbi cos βvb x + dbi sin βvb x + ebi cosh βvb x + f bi sinh βvb x, (16)
φvcj (x) = acj sin βvcj x + bcj cos βvcj x + cj , (17)
where
8μcj dcj êcj
cj = . (18)
βvcj
2
 Y. Cong et al.

By manipulating Eqs. (16) and (17) with the relevant geometrical and mechanical compliance conditions (14),
the following characteristic equation can be obtained:

F(ωv ) = 0, (19)

where F(ωv ) is the coefficient matrix.  is the vector consisting of coefficients in Eqs. (16) and (17), which is
defined as
 = [· · · , acj , bcj , · · · , cbi , dbi , ebi , f bi , · · · ]T . (20)
The frequency ωv can be determined by making the determinant of the coefficient matrix F(ω) in Eq.
(19) equal zero. Once the frequency ωv is obtained, the coefficient vector  can be easily determined by
combining with the boundary conditions, and then the mode functions of each order can be obtained, where
the normalization can be performed with max {φvcj , φvbi } = 1.

2.2 Out-of-plane eigenvalue problem

Similarly, the out-of-plane eigenvalue problem can be obtained by using the separation of variables method,
which are given as

βwcj
2
φwcj + φ  wcj = 0 (21)
−βwb
4
φwbi + φ  wbi =0 (22)

and the following geometrical boundary and compliance conditions

⎧
⎪
⎪ φwcj (0) = φwb(n+1) (1) = φ  wb(n+1) (1) = φwb1 (0) = φ  wb1 (0) = 0,
⎪
⎪
⎨ φwcj (1) = γcj φwbi (s j ),
φwbi (s j ) = φwb(i+1) (s j ), φwbi  (s ) = φ    (23)
⎪
⎪ j wb(i+1) (s j ), φ wbi (s j ) = φ wb(i+1) (s j ),
⎪
⎪
⎩ χ φ  wbi (s j ) − φ  wb(i+1) (s j ) − 1 φ  (1) = 0.
ι μcj wcj

Without the influence of initial configurations of the cable, the out-of-plane eigenfunction equations of the
cable become homogeneous equations. In this case, solutions of Eqs. (21) and (22) are obtained:

φwcj (x) = acj sin βwcj x + bcj cos βwcj x, (24)

φwbi (x) = cbi cos βwb x + dbi sin βwb x + ebi cosh βwb x + f bi sinh βwb x. (25)

With the same approach, the characteristic equations can be deduced by applying Eqs. (23) to (25) simul-
taneously, and they are expressed as
F(ωw ) = 0, (26)
where F(ωw ) is the coefficient matrix. The solvability condition of Eq. (26) is that the determinant of coefficient
matrix F(ωw ) equals zero. Similarly, the normalization can be performed with max {φwcj , φwbi } = 1.

2.3 Nonlinear discrete model for planar oscillations

Imposing the stationary functional method, a single-mode discrete nonlinear model of the multi-cable-stayed
beam model is derived in order to explore the influence of some mechanical parameters on the dynamic
behavior of the nonlinear system. For achieving the single-degree-of-freedom nonlinear model, the system
displacement can be expressed in the form
 
vcj (xcj , τ )  T
= φvc1 (xc1 ) · · · φvcj (xcj ) φvb1 (xb1 ) · · · φvbi (xbi ) q(τ ), (27)
vbi (xbi , τ )

where { }T denotes the matrix transposition. φvcj and φvbi are the modal eigenfunctions of the j-th cable and
the i-th sub-beam, respectively, which are furnished by Eqs. (16) and (17). q is the generalized coordinate. By
Modeling, dynamics, and parametric studies

Table 1 The first 10 natural frequencies of the double-cable-stayed beam model

Natural frequency 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
In-plane
Analytical 0.1355 0.2307 0.4354 0.7848 1.2162 1.3503 1.3503 1.7417 2.3740 2.6854
FEM 0.1360 0.2307 0.4349 0.7840 1.2147 1.3436 1.3437 1.7391 2.3696 2.6890
Error (%) −0.3676 0.0000 0.1150 0.1020 0.1235 0.4987 0.4912 0.1495 0.1857 −0.1339
Out-of-plane
Analytical 0.0489 0.1936 0.4354 0.7741 1.2093 1.3428 1.3429 1.7417 2.3706 2.6856
FEM 0.0487 0.1937 0.4352 0.7735 1.2080 1.3433 1.3435 1.7390 2.3660 2.6870
Error (%) 0.4107 −0.0516 0.0460 0.0776 0.1076 −0.0372 −0.0447 0.1553 0.1944 −0.0521

applying the stationary Hamiltonian functional method, the following nonlinear ordinary differential equation
(ODE) governing the planar oscillation can be obtained:
q̈ + q + c1 q̇ + c2 q 2 + c3 q 3 = F1 cos 1 τ + F2 cos 2 τ, (28)
where c1 , c2 and c3 are polynomial coefficients; and Fn and Ωn are the amplitude and frequency of external
excitation, respectively. Their exact expressions are given in Appendix. It should be noted that Eq. (28) has
the typical form of a nonlinear oscillator with quadratic and cubic nonlinearities. In spite of this, due to the
different expressions of coefficients, new and significant behaviors can be found in the following analysis.

3 Verification of the Linearized dynamic model

To validated the linearized dynamic model developed in this study, a symmetric cable-stayed beam model
with two identical cables is studied. Junctions among the cable and the deck beam divide the beam into three
identical sub-beams. The following properties are applied on the deck beam: Young’s modulus E b = 34.5
GPa, moment inertia of the cross-section Ib = 9.8 m4 , cross-sectional area Ab = 16.3 m2 , mass per unit length
m b = 4.4 × 104 kg/m, and length of the overall beam lb = 300 m. The following properties are applied on
the j-th cable: Young’s modulus E cj = 210 GPa, cross-sectional area Acj = 6.273 × 10−3 m2 , mass per unit
length m cj = 10.4 kg/m, initial cable force Hcj = 1 × 106 N, inclined angle θ j = π/6 rad, length of the cable
lcj = lb /3/cosθ j m, and sag of the cable Dcj = m cj glc2 cosθ j /(8Hcj ). In this study, the gravity acceleration is
chosen as 9.8 m/s2 . Based on the developed characteristic Eqs. (19) and (26), the first 10 natural frequencies
associated with the in-plane and out-of-plane motion are extracted, as listed in Table 1. For comparison, a
finite element model (FEM) of this cable-stayed beam model is developed and modal analysis is conducted to
obtain the first 10 natural frequencies corresponding to the in-plane and out-of-plane motions. It shows that the
results obtained from the proposed theoretical approach in this study match those obtained from finite element
analysis very well, with the maximum error of 0.5%, verifying the theory developed in this study.

4 Parametric analysis of the linearized model

To further validate the proposed theory, systematic parametric analyses are conducted on a symmetrical model
to explore modal properties and nonlinear dynamic behaviors of the system. In this model, the cables with
length lcj = lc and inclined angle θ j = θ are further considered, and the coefficients ci and Fi in Eq. (28) are
given in the Appendix. In particular, the influence of some key parameters on the eigenvalue problem of the
linearized dynamic model is investigated through the eigenfunctions. The parametric study on the nonlinear
dynamic behavior of the system will be presented in Sect. 5. In order to describe the inherent characteristics
of the double-cable-stayed beam model effectively, two kinds of beam model are introduced as the reference
cases. The first model is a hinged-hinged beam, in which the cable is negligible (BM1 in Fig. 3a); and the
second model is a hinged–hinged beam with two vertical supports in the span, which is to simulate the case
where the axial stiffness of the cable is much higher than flexural stiffness of the beam (BM2 in Fig. 3b). The
first four modal solutions of the two reference beam models are presented in Fig. 4.
The parametric analysis about the modal behaviors of the double-cable-stayed beam model is conducted,
starting from a cable-beam system without cable sag, which can be treated as a string-beam system. Then,
the sag influence on the modal behaviors follows. In the following analysis, the non-dimensional parameters
 Y. Cong et al.

Fig. 3 Two kinds of single beam models

Fig. 4 Mode shapes for the references: BM1 and BM2

d = 0, θ = π/6 and μ = 486 [5] are used to study the modal properties of the system with different (ρ,
χ )-pairs. The values βvb and βvc associated with the first four modes of the system are calculated according to
the characteristic equation (19) and the relation (11). Moreover, it should be noted that the horizontal straight
lines in the following figures denote the first four eigenvalues of the BM1 (I1, II1, III1, IV1), the BM2 (I2, II2,
III2, IV2) and the string (π, 2π, 3π, 4π).

4.1 In-plane motion

Figures 5a and b present the first modal property of the system under different (ρ, χ )-pairs. In Fig. 5a, it can
be seen that when the parameter χ varies within a small range, the system behaves a global mode with a first
mode shape (I2) of the model BM2 (βb = 9.4247) for all values of parameter ρ while the value of the cable βc
does not correspond to a specific mode. This is obviously different from that in the cable-stayed beam system
[5], where the parameter range that exhibits the global mode is extremely small. Actually, it is worth noting
that the parameter χ represents the ratio of flexural rigidity of the deck beam to axial rigidity of cables from
the former expression in Eq. (9). Hence, the smaller value the parameter χ appears, the larger supporting effect
of the cable on the deck beam exerts. One of the limit states is that the double-cable-stayed beam system is
equivalent to a hinged-hinged beam with two vertical rigid supports at its trisection points, namely, the model
BM2 as shown in Fig. 3, and its first four modes is given in Fig. 4. In Figs. 5a and b, with further increasing
of the parameter χ , a local mode of the system emerges in the form of a first mode of the cable when ρ = 0.1
and χ > 0.002, in which the eigenvalue βc of cables reaches to the first frequency π. But the value of βb , in
this range, does not represent a specific modal shape of BM1 or BM2. Here, the deck beam can be regarded
as a continuous beam with two certain elastic supports.
Figures 5c and d present the second modal properties of the system with different (ρ, χ )-pairs. The behavior
is similar to those of the first mode of the system in Figs. 5a and b. The difference is that when the parameter χ
varies in the range from 0 to 0.001, the system hardly exhibits the global mode. The first mode of the cable may
Modeling, dynamics, and parametric studies

Fig. 5 In-plane eigenvalue (ρ, χ)-dependence: a, b 1st mode; c, d 2nd mode; e, f 3rd mode; g, h 4th mode

occur but act as the second mode of the system with parameters ρ = 0.1 and χ > 0.00026. For the third modal
property of the system in Figs. 5e and f, it is noted that increasing χ causes the transition of modal eigenvalue
from III2 of the model BM2 to III1 of the model BM1 with a certain level of stiffness concentrated at the
mid-two supports. Then, a global mode appears in the form of the third mode (III1) of the model BM1 when
ρ < 0.1 and χ > 0.002. When the higher mode (mode IV) is considered, increasing χ causes the transition
of modal eigenvalues from IV2 of the model BM2 to IV1 of the model BM1. And the first mode of the cable
appears as a local mode of the system when ρ = 0.1 and 0.00002 < χ < 0.00015, ρ = 0.01 and χ > 0.0005.
In addition, it can be observed that in the range of all the considered parameters, the possible local mode of
the system from mode I to mode IV, as shown in Figs. 5b, d, f and h, is not the high modes of the cable but
the first symmetric one (βc = π). However, the second mode of the cable, the antisymmetric mode shape,
was found in certain parameter range in Reference [5]. Furthermore, it can be observed from Fig. 5 that when
the parameter χ is larger, the local mode of the system, i.e. the large vibration of the cable, can be induced by
adjusting the parameter ρ. However, this is not allowed in engineering practice and should be avoided.
In addition, the mode shapes of the system associated with the four chosen points (A, B, C and D in Fig. 5)
are presented in Fig. 6 to demonstrate the modal variation, in which points A and B are related to the local
modes (only the cable vibrates; the deck beam hardly vibrates) and points C and D are related to the global
modes (only the deck beam vibrates; the cable hardly vibrates). In other cases, the system exhibits the coupled
modes (both the beam and the cable are dynamically involved in the motion).

4.2 Out-of-plane motion

When the out-of-plane motion of the double-cable-stayed beam model is considered, it should be noted that
there only exists the geometric stiffness of the cable in the out-of-plane motion, while geometric and elastic
stiffness simultaneously exist in the in-plane motion [5]. In this case, the same parameters of the model
are introduced as in the previous in-plane case. In addition, the identical bending stiffness of in-plane and
out-of-plane motion of beam is considered.
Figure 7 presents the out-of-plane eigenvalue (ρ, χ )-dependence of the first four eigenvalues of the double-
cable-stayed model. As shown in Figs. 7a and c, it can be seen that the influence of mass ratio ρ on the lower
modes (the first two orders) of the deck beam is very weak when their stiffness ratio χ varies within a small
range. Without the elastic support of the cable, the out-of-plane stiffness of the system is obviously less than
the in-plane one. Hence, the system mainly exhibits the global mode, as shown in Figs. 7a, c, e and g, which
 Y. Cong et al.

Fig. 6 Global, local mode shapes of system for different (ρ, χ)-pairs: A and B (0.1, 0.0005); C (0.0001, 0.0005); D (0.0001,
0.001)

Fig. 7 Out-of-plane eigenvalue (ρ, χ)-dependence: a, b 1st mode; c, d 2nd mode; e, f 3rd mode; g, h 4th mode

are associated with the first four modes (I1, II1, III1, IV1) of the model BM1, namely, the hinged beam at
two ends. Especially, it is noted that the difference is more evident by comparing modes in Figs. 5c and 7c.
In the higher modes of the system, (e.g., mode III and mode IV, as shown in Figs. 7e to h, it is noted that the
out-of-plane vibration of the deck beam can be suppressed by adjusting the stiffness ratio χ among cables and
the deck beam when their mass ratio ρ is given a certain value, such as 0.01 and 0.1. However, at the same
time, the system may exhibit the local mode, namely, the cable’s large vibration. It is well-known that the large
vibration of the cable, no matter the in-plane one or the out-of-plane one, can cause fatigue damage, threatening
Modeling, dynamics, and parametric studies

the safety of bridge. Hence, the mass ratio ρ and stiffness ratio χ should be studied comprehensively. The
simply local reinforcement through adjusting specific parameters of the bridge is undesirable, which can cause
damage of its local components.

4.3 Influence of cable sag

For the cable-stayed system that consists of long cables, the influence of the cables’ static configuration on
the mechanical properties of the system cannot be ignored. In previous research, the sag d was introduced to
explore its influence on the inherent properties of the system. Gattulli [5] and Irvine [25] investigated the sag
influence in which they had found that only the in-plane problem is affected by sag. In the following analysis,
two different values of sag d are considered, where d = 1/142 represents a beam supported by two taut
cables and d = 1/45 represents a beam supported by two shallow cables, while the other parameters remain
unchanged.
Figure 8 presents the first four eigenvalues of the system with different (ρ, χ )-pairs and the sag d = 1/142.
The modal behaviors are similar to the former string-beam system shown in Fig. 5. Here, the main difference
is that the local mode of the system is not easy to be evoked, especially at the first two modes of the system
from the comparison between Figs. 5b, d and 8b, d. A local mode at the fourth mode of the system occurs in
the form of the first symmetric mode of the cable with parameters ρ = 0.01 and χ > 0.00064. In addition,
the first frequency of the cable is replaced by the value βc = 3.334 because of the sag influence. As the sag
increases to the value d = 1/45, the sag influence is more evident, as shown in Fig. 9. It is observed that local
modes do not occur for all given parameters and should be noted that the first frequency of the cable changes
to the value βc = 4.694 in this case, which is never reached at the first two modes of the system, as shown in
Figs. 9a to d. This phenomenon is consistent with that obtained by Gattulli [5]. This phenomenon occurs due
to the fact that the increase in cable’ sag causes a relatively small decrease in its stiffness; at the same time, the
support stiffness of the cable on the beam decreases, which causes a decrease in global stiffness of the system
while causing an increase in the cable’s relative stiffness. Hence, the system mainly exhibits a global mode
acting as the beam’s lower modes. Under this circumstance, the first two modes of the double-cable-stayed
beam system always exhibits global modes, ranging from eigenvalues of the BM2 (I2, II2) to the BM1 (I1,
II1), respectively. Additionally, the coupled modes (modes where both the beam and cables are dynamically
involved in the motion) can also be found when the parameter χ is greater than a certain value (ρ = 0.1) in
higher modes (mode III and mode IV) in Figs. 8 and 9.

5 Parametric analysis of the nonlinear dynamic model

In this section, the nonlinear dynamic behaviors of the SDOF system (28) are investigated. Two cases are
studied: (1) the beam is subjected to single-frequency excitations, and (2) the beam is subjected to two-
frequency harmonic excitations. Here, it should be noted that the possible interaction between the beam and
the cable, bifurcations and internal resonance are not considered. Wei [10,11] and Kang et al. [21–23] used
the Galerkin discretization method to discretize the cable-beam coupled system into a multiple degrees of
freedom model, in which the cable and the beam are coupled by their continuous conditions so as to study the
interaction between them. However, the discrete model cannot allow one to analyze the parametric influence on
the dynamic response of the overall system. To this end, the cable-beam coupled system is treated as a united
system. Gattulli [5], Lenci [7] and Peng et al. [14] investigated the influence of parameters and time delay
feedback of a united SDOF model of cable-beam coupled system. For simplicity, the out-of-plane motion of
the system is also excluded so that the possible interaction between in-plane and out-of-plane vibrations of the
system is not considered in this study.

5.1 The system to primary resonance: single-frequency excitation

5.1.1 Perturbation analysis

A bookkeeping parameter ε is introduced to reorder the ODE (28) so that the damping term, nonlinear term and
external excitation term can appear at the same order for obtaining the uniformly valid approximate solution,
 Y. Cong et al.

Fig. 8 Sag influence on in-plane eigenvalue (ρ, χ)-dependence: d = 1/142

Fig. 9 Influence of sag on in-plane eigenvalue (ρ, χ)-dependence for d = 1/45

as follows:
q̈ + q + ε2 c1 q̇ + εc2 q 2 + ε2 c3 q 3 = ε2 F1 cos 1 τ. (29)

Expanding the solution in the form of different time scales according to the method of multiple scales [26]
yields
q(τ ) = ε0 q1 (T0 , T2 ) + εq2 (T0 , T2 ) + ε2 q3 (T0 , T2 ), (30)
Modeling, dynamics, and parametric studies

where T0 = τ is the fast time scale and T2 = ε2 τ is the slow time scale. By substituting the expression (30)
into Eq. (29) and equating the coefficients at the same order at both sides, one obtains

ε0 : D02 q1 + q1 = 0, (31a)
ε :
1
D02 q2 + q2 = −c2 q12 , (31b)
ε :
2
D02 q3 + q3 = −c1 D01 q1 − 2D01 D21 q1 − 2c2 q1 q2 − c3 q13 − F1 cos 1 T0 , (31c)
where Dmn = ∂ /∂ Tn , m = 1, 2 and n = 0, 2. The external detuning parameter σ is introduced to indicate the
m
nearness of frequencies between external excitation and the system

1 = 1 + ε2 σ. (32)
The general solution at order ε0 is taken in the following form

q1 (T0 , T2 ) = A(T2 )ei T0 + Ā(T2 )e−i T0 , (33)

where Ā is the complex conjugate term of A. After a recursive process according to the method of multiple
scales, and eliminating the terms in Eq. (31c) that produce secular terms, one can obtain the solvability condition
 
1 i T2 σ 10 2
e F1 − ic1 A + c2 − 3c3 A2 Ā − 2i Ȧ = 0. (34)
2 3

Then, letting A(T2 ) = 1/2a(T2 )eiψ(T2 ) , and separating real and imaginary parts, one can obtain the
modulation equations governing the amplitude and phase:
1 1
a  = − c1 a + F1 sin α, (35a)
2  2 
 5 2 3 1
aα = σ a + c2 − c3 a 3 + F1 cos α, (35b)
12 8 2
where α(T2 ) = T2 σ − ψ(T2 ). The steady-state solutions of Eqs. (35a) and (35b) occur when a = α = 0. The
a − σ frequency-response equation can also be obtained by squaring and adding Eqs. (35a) and (35b) and
imposing the identity sin(α)2 + cos(α)2 = 1, as follows:
 
3 5 2 2 1 F12
σ = c3 − c2 a ± − c12 . (36)
8 12 2 a2

5.1.2 Case analysis

In this case, the frequency responses of the single-mode nonlinear model of the cable-stayed beam with
double cable (DCS) and a single cable (SCS) are investigated, respectively. Two different cable sags, namely,
d = 1/142 and d = 1/45, are considered to explore the sag effect on the nonlinear response of the system.
Figure 10 presents the frequency response of the nonlinear model for different (ρ, χ )-pairs with the cable
sag d = 1/142 when two cables (DCS) are considered. It is observed that when the mass ratio is a very
small value (ρ < 0.01), the frequency response relations is of linear type, as shown in Fig. 10b–d. In [5], it
was pointed out that such behavior was a consequence of a global mode type involved in the first primary
resonance, where the nonlinearity caused by the cable vibration is very weak. On the other hand, the parameter
ρ represents the mass ratio of cable to beam; the smaller the parameter, the lighter the cable. Meanwhile, the
influence of the nonlinearity caused by the cable vibration may be smaller. From the perspective of energy
transfer, more energy is concentrated on the beam, which dominates the vibration of the overall system. With
an increase in the parameter ρ, the modal component of the cable becomes larger and the nonlinearity caused
by the cable’s sag appears. The system’s response gradually appears as a local mode, which causes hardening
characteristics, as shown in Fig. 10a. This conclusion is consistent with the previous results that a local mode
of the system can be induced by adjusting the mass ratio ρ, as presented in Sect. 4.1.
When the sag d is increased gradually, the quadratic coefficient c2 in Eq. (28) becomes larger. From the
expressions of c2 and c3 given in the Appendix, it can be seen that although the cable sag only affects the
quadratic nonlinearity, the quadratic nonlinearity not only depends on the cable sag, but also depends on the
 Y. Cong et al.

Fig. 10 Frequency response curves of the first in-plane mode for different (ρ, χ) values with d = 1/142 (DCS)

Fig. 11 Frequency response curves for the first in-plane mode for different (ρ, χ) values with d = 1/45 (DCS)

values of βc , ρ and the modal expressions φcj and φbi ; and the cubic nonlinearity depends on the values
of βc , ρ and the modal expressions φcj and φbi . Figure 11 presents the frequency response of the nonlinear
model for the first in-plane mode for different (ρ, χ )-pairs with the cable sag d = 1/45. Comparing with that
in Fig. 10, the major difference is that the frequency response of the nonlinear model exhibits a hardening
behavior all the time with the same parameters. This behavior is a consequence of the larger positive value of
the cubic coefficient induced by the larger sag, as shown in Table 2, which confirms its effect on the system
with cubic nonlinearities [26]. From the results in Sect. 4, a large ρ can have a significant influence on the
Modeling, dynamics, and parametric studies

Table 2 Quadratic and cubic coefficients of nonlinear model (Eq. 28) with different parameters (DCS)

d = 1/142 d = 1/45
ρ χ c2 c3 ρ χ c2 c3
0.1 0.0005 −2.05809 42.7098 0.1 0.0005 0.245542 40.9245
0.0008 −3.48731 185.445 0.0008 0.61212 125.64
0.001 −4.31991 358.056 0.001 0.901035 207.756
0.01 0.0005 −0.332808 0.142199 0.01 0.0005 −0.758608 10.8494
0.0008 −0.646561 0.528767 0.0008 −1.28541 33.0258
0.001 −0.863397 0.941268 0.001 −1.66427 53.7251
0.001 0.0005 −0.273178 0.100179 0.001 0.0005 −0.766312 9.4348
0.0008 −0.527663 0.370254 0.0008 −1.45538 28.3779
0.001 −0.701969 0.656027 0.001 −1.69944 46.5452
0.0001 0.0005 −0.267823 0.095528 0.0001 0.0005 −0.505328 9.68799
0.0008 −0.524023 0.387457 0.0008 −1.91134 26.7866
0.001 −0.729204 0.704683 0.001 −1.62193 46.2782

Table 3 The first 10 frequencies of the in-plane single-cable-stayed beam model (SCS)

Frequency 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
Analytical 0.0960 0.1932 0.4434 0.7733 0.8952 1.2111 1.7392 1.7906 2.3678 2.6865
FEM 0.0957 0.1935 0.4438 0.7741 0.9064 1.2125 1.7417 1.7903 2.3722 2.6859
Error (%) 0.3135 −0.1550 −0.0901 −0.1033 −1.2357 −0.1155 −0.1435 0.0168 −0.1855 0.0223

system response, hence c2 associated with ρ = 0.1 and d = 1/45 is different and has a positive value. c3
remains hardly changed for ρ = 0.1 due to the fact that, different from the results for d = 1/142, the system
does not behave as a local mode in this case. In spite of this, the cubic nonlinearity is still high and plays an
important role. In addition, the parameter ρ simultaneously affects the jump phenomenon of the system in the
transformation of its frequency response from a hardening characteristic to a linear type.
In the following analysis, a single-cable-stayed beam system (SCS) is developed. The cable is anchored at
the midpoint of the beam and the beam is hinged at its two ends [7]. In this case, it should be noted that the
nonlinear dynamic equation has the identical expression, as shown in Eq. (28), but with different coefficients.
Similarly, theoretical derivation has been verified with a finite element model with the same parameters. The
results reach a good consistency, as seen from the first 10 natural frequencies as given in Table 3. The largest
relative error is only 1.24%. For comparability, the identical parameters θ = π/6 and μ = 486 are used and
the influence of cable sag is also studied in this case.
Figure 12 presents frequency response of the first in-plane mode of the single-cable-stayed beam system
with different (ρ, χ ) values with the cable sag d = 1/142. It shows that compared with the frequency responses
in the former DCS case in Fig. 10, the single-cable-stayed beam system shows a stronger hardening property
within the whole parameter range. Especially, it shows a slight hardening behavior when ρ < 0.01, which
indicates the cable’s nonlinear influence. Likewise, the stiffness ratio χ affects the system’s nonlinear spring
property and the jump phenomenon. With increasing parameter χ , the system’s frequency response has a
stronger hardening property [8] and further shows a jump phenomenon.
Figure 13 presents the frequency response of the first in-plane mode of the single-cable-stayed beam system
with different (ρ, χ ) values with the cable sag d = 1/45. Similar to the former DCS case, when the cable
sag increases to a larger value, the SCS also presents a hardening behavior within the range of all parameters,
where the cable’s nonlinearity dominates the system’s frequency response.
As can be seen from Figs. 10, 11, 12, and 13, it should be noted that the cable mainly causes the quadratic
and cubic nonlinearities in the DCS and SCS systems, hence the hardening response obtained with large ρ
and sag d in all considered cases (DCS and SCS) is clearly linked with the major role played by the cable’s
nonlinearity.
Herein the proposed theory for the single-mode dynamic model of the complex multi-cable-stayed beam
system reproduces the typical frequency response of a nonlinear oscillator with quadratic and cubic nonlin-
earities describing structures with initial curvature ...[27–29], which is interesting. Moreover, the parametric
studies on the dynamic response of the DCS and SCS systems will lay the foundation for further investigation
of more complex models.
 Y. Cong et al.

Fig. 12 Frequency-response of the first in-plane mode for different (ρ, χ) values with d = 1/142 (SCS)

Fig. 13 Frequency-response for the first in-plane mode for different (ρ, χ) values with d = 1/45 (SCS)

5.2 The system to simultaneous resonance: two-frequency excitation

5.2.1 Perturbation analysis

As can be seen in Eq. (28), there exist the quadratic and cubic nonlinear terms due to the consideration of cable’s
sag effect and the interaction between the cable and the beam. Hence, there are many types of simultaneous
resonance between two external excitations, such as Ωn = 1, 2Ωn = 1, 3Ωn = 1, Ωn = 2, Ωn = 3, where
n = 1, 2. In the study, the case of Ω1 = 1 and Ω2 = 3 is studied.
Modeling, dynamics, and parametric studies

In this case, the system is subjected to the primary resonance and 1/3 order sub-harmonic simultaneous
resonance simultaneously. Similarly, a bookkeeping parameter ε is introduced to reorder the ODE (28) so that
the damping term nonlinear term can appear at the same order for obtaining the uniformly valid approximate
solution. Moreover, it should be noted that primary excitation is reordered so as to avoid the small divisor [30]:

q̈ + q + ε2 c1 q̇ + εc2 q 2 + ε2 c3 q 3 = ε2 F1 cos 1 τ + F2 cos 2 τ. (37)

The solution is assumed to have the polar expansion, according to the method of multiple scales [26]

q(τ ) = ε0 q1 (T0 , T2 ) + εq2 (T0 , T2 ) + ε2 q3 (T0 , T2 ), (38)

where T0 = τ is the fast time scale and T2 = ε2 τ is the slow time scale. By substituting Eq. (38) into Eq. (37)
and equating the coefficients at the same order on both sides, one obtains

ε0 : D02 q1 + q1 = 2F2 cos(2 T0 ), (39a)

ε :
1
D02 q2 + q2 = −c2 q12 , (39b)
ε :
2
D02 q3 + q3 = 2F1 cos 1 T0 − c1 D01 q1 − 2D01 D21 q1 − 2c2 q1 q2 − c3 q13 , (39c)

where Dnm = ∂ m /∂ Tn , m = 1, 2 and n = 0, 2. The external detuning parameters σ1 and σ2 are introduced to
indicate the nearness of frequencies between external excitation and the system,

1 = 1 + ε2 σ1 and 2 = 3 + ε2 σ2 . (40)

The general solution at order ε0 is taken in the following form:

q1 (T0 , T2 ) = A(T2 )ei T0 + E 2n ei T0 2 + cc, (41)

where
F2
E 2n = . (42)
1 − 22
After a recursive process according to the method of multiple scales [26], by eliminating the terms in Eqs.
(39b) and (39c) that produce secular terms, one can obtain the solvability condition

ei T0 1 F1 − 2iei T0 A − ic1 ei T0 A + 1 ei T0 A + 2 ei T0 (2 −2) B 2 + 3 ei T0 A2 B = 0, (43)

(T )
where 1 , 2 and 3 are given in the Appendix. Then, the amplitude is assumed as A(T2 ) = 1/2a(T2 )eiβ 2 ,
where a(T2 ) and β(T 2) denote the amplitude and phase, respectively. By substituting A(T2 ) into the solvability
condition, separating real and imaginary parts and considering the following relationships:

γ1 (T2 ) = T2 σ1 − β(T2 ), (44a)

γ2 (T2 ) = T2 σ2 − 3β(T2 ). (44b)

One can obtain the autonomous modulation equations. The steady-state solutions occur when a = γ1 = γ2 =0.
Hence, it leads to the requirement that σ2 = 3σ1 and Ω2 = 3Ω1 , due to the 1/3 order sub-harmonic resonance.
Then the following expressions are defined:
σ2
σ = σ1 = and γ (T2 ) = T2 σ − β(T2 ). (45)
3
Finally, the autonomous modulation equations are presented as

1 1
a  = − c1 a + F1 sin(γ ) + 2 a 2 sin(3γ ), (46)
2 4
8aγ  = 8aσ + 41 a + 3 a 3 + 8F1 cos(γ ) + 22 a 2 cos(3γ ). (47)
 Y. Cong et al.

Fig. 14 Frequency response with f 2 = 0.01 when: a f 1 = 0.001, b f 1 = 0.001, f 1 = 0.004, f 1 = 0.008

5.2.2 Case analysis

In the analysis, the dynamic behaviors of the lowest modes of the cable and the beam are studied. For the cable,
λ2 = 64μd 2 is the Irvine parameter, and the sag to span ratio d is defined as 0.001. The sine function is regarded
as the trial function in the Galerkin integration only when λ2 < 4π 2 , where the first symmetric vertical mode
shape has no internal nodes along the span [31]. As illustrated in previous description, different excitations are
rescaled to different orders to avoid the small divisor. The amplitudes of primary and sub-harmonic excitations
are chosen as f 1 = 0.001 and f 2 = 0.01. The system’s damping ratio is 0.003. The solid line represents the
stable solution and the dot line represents the unstable solution. Importantly, the numerical results are obtained
to verify the correction and a good agreement between the analytical and numerical results is reached.
Figure 14a presents the frequency response of the system when f 1 =0.001 and f 2 =0.01. The frequency
response bends to the right, which illustrates the hardening spring property. Lenci [7], Peng [14] and Zhao [30]
also observed the hardening spring property in their SDOF system with quadratic and cubic nonlinearities. It is
a consequence of the prevailing of the cubic nonlinearity in the equation of motion, as previously mentioned in
Sect. 5.2. With the variation of the excitation frequency, the system has the jump phenomenon occurring at the
saddle-node bifurcation (SN). The phase diagrams and time history before and after the jump point are given
in Fig. 15. A substantial difference is observed, especially in their phase diagrams. Zhao [30] also obtained
phase diagrams before and after the jump point in different cases of cable sags in the model of a single cable
subjected to planar simultaneous primary and sub-harmonic resonances. The results obtained from this study
are similar to the results in [28] in terms of the jump phenomenon, while they are different in terms of the
oscillation properties at different branches, as shown in Fig. 17.
Figure 14b presents the influence of different amplitudes of primary excitation on the system’s frequency
response when f 1 = 0.001, f 1 = 0.004 and f 1 = 0.008, respectively. Since the excitation amplitude f 1 comes
from the primary excitation, it has a significant influence on the frequency response of the system. At the same
time, the region of solutions enlarges. However, their backbones are unchanged and always turns to the right,
which exhibits a hardening spring property.
Figure 16 presents the amplitude response of the system as σ = 0.1 and σ = −0.1 with the amplitude
change of primary ( f 1 ) and sub-harmonic ( f 2 ) excitations, respectively. It can be clearly seen that the responses
in Fig. 16b with the variation of sub-harmonic excitation’s amplitude is much simpler than that of responses in
Fig. 16a with the variation of primary excitation’s amplitude. Moreover, quantitatively, the amplitude change of
sub-harmonic excitation only has a small effect on the system’s amplitude response at each branch solution. In
Fig. 16a, there are four saddle-node bifurcation points, which cause very complex dynamic response including
the jump phenomena and multi-solution. In a certain domain, as many as five solutions for a given detuning
parameter σ can be found where three are stable solutions and two are unstable solutions. This phenomenon
occurs in a very small region, and it is different from the amplitude response obtained in the SDOF system
subjected to single-frequency excitation [14,32]. Moreover, the system’s response decreases with the increasing
in the sub-harmonic excitation amplitude f 2 when σ = −0.1, as given in Fig. 16b, which confirms that the
addition of sub-harmonic excitation with certain parameters will decrease the dynamic response of the system
Modeling, dynamics, and parametric studies

Fig. 15 Phase portraits and time history when σ = 0.1: a upper branch, b lower branch

Fig. 16 Amplitude response: a with the excitation amplitude f 1 , b with the excitation amplitude f 2

to primary resonance [30]. Based on this principle, engineers can suppress the vibration of the system by
adding an extra excitation, which is well known in both Bridge Engineering and Mechanical Engineering and
called the active control technique.
Figure 17 presents the phase diagrams of the overall system at different excitation frequencies and different
branches in the frequency responses in Fig. 14a. As can be seen from Fig. 17, the lower responses of the system
have more complex vibration properties. When the detuning parameter σ = −0.4 (shown in Fig. 17a) and
σ = 0.4 (shown the lower branch in Fig. 17d), these phase planes are described by three circles, while the
phase planes in Figs. 17b and c are described by a single circle. The single circle in Figs. 17b and c suggests
that the system has period-1 steady-state solution, which can be confirmed in their power spectrums, as shown
in Figs. 18b and c. The three circles shown in Figs. 17a and d suggest that the system has period-2 solutions.
Their power spectrums are given in Figs. 18a and d, in which they have two peaks. Zhao [30] studied the
vibration properties at different branches of the frequency response of a single cable subjected to primary
excitation and 1/3 sub-harmonic excitations from the perspective of the Poincaré section. In his research, the
period-1 solutions were obtained. The difference is intriguing. In the complex system, the coupling among
different components may cause more complicated response. For a specific system, it is necessary to find the
response property in order to further control its vibration. Phase plane, Poincaré map and Power spectrum are
very useful indicators to characterize the system’s oscillation properties.
To investigate the influence of damping ratio on the system’s dynamic response, four cases with different
damping ratios c1 = 0.001, c1 = 0.003, c1 = 0.008 and c1 = 0.012 are considered. When the damping ratio
 Y. Cong et al.

Fig. 17 Phase portraits at different excitation frequencies and different branches: a σ = −0.4, b σ = 0, c σ = 0.4, d σ = 0.4

is small, as shown in Figs. 19a and b, the influence is very weak. With the increase in damping ratio, as given
in Figs. 19c and d, the system behaves a weak hardening spring property. Moreover, finite peak amplitudes
occur, their locus decreases as damping increases [27]. More interestingly, multi-peak amplitudes appear [30].
This will cause multiple solutions and jump phenomena with a specific value of detuning parameter σ , in spite
of the very small interval. In this case, there are six solutions, as shown in Fig. 19c, in which three are stable
solutions and three are unstable solutions. In the frequency response, more saddle-node bifurcations appear
in the same parameter domain. At each saddle-node bifurcation, the response curve will change its direction,
where a jump happens.

6 Conclusions

A multi-cable-stayed beam model for cable-stayed bridges and the related in-plane and out-of-plane vibration
equations are developed with their boundary and compliance conditions. The natural frequencies and mode
shapes are determined by using the separation-of-variables method. Then, the parametric study on eigen
problem is conducted by taking two single beam models as the reference cases. Meanwhile, a simplified single-
degree-of-freedom nonlinear model of the system is derived by applying the stationary Hamiltonian functional
method. The nonlinear dynamic responses to single-frequency and two-frequency excitations are investigated
through frequency response, amplitude response, phase diagram, time history of amplitude responses. It should
be noted that the nonlinear analysis is based on the simplified single-degree-of-freedom model, and hence the
results can only help to understand the parametric influence on the dynamic responses of the overall system
qualitatively. Rigorous analysis may be needed for quantitative analysis. The following conclusions are drawn
from these obtained results.
Modeling, dynamics, and parametric studies

Fig. 18 Power spectrum and time history at different excitation frequencies and different branches: a σ = −0.4, b σ = 0, c
σ = 0.4, d σ = 0.4

1. The occurrence of global, local and coupled modes of the system depends on the mass and stiffness ratios
between the cable and the deck beam. For the lowest in-plane mode, when the stiffness ratio χ varies within
a small range, the support stiffness of the cable on the beam is very large. The system mainly exhibits a
global mode. The influence of the mass ratio ρ on the modal the characteristics is small. When the stiffness
ratio becomes larger, the beam is supported with a certain level of stiffness. Global and local modes of the
system may occur. The system’s local mode can be affected by adjusting the mass ratio ρ. Without elastic
support in the out of plane motion, the system mainly exhibits the global mode. The vibration of the deck
beam can be suppressed by adjusting the stiffness ratio of the cable to the deck beam with a certain value
range of mass ratio, but may induce a local mode of the system.
2. When cable sag is considered, the local mode of the system is not easily evoked, especially at the first two
modes of the system. An increase of the cables’ sag causes a relatively small decrease of its stiffness; at the
same time, the support stiffness of the cable on the beam decreases, which causes a decrease in the global
stiffness of the system while the cable’s relative stiffness increases. Hence, the system mainly exhibits a
 Y. Cong et al.

Fig. 19 Effect of damping ratios on nonlinear response: a c1 = 0.001, b c1 = 0.003, c c1 = 0.008, d c1 = 0.012

global mode acting as the beam’s lower mode. Moreover, the cable only exhibits the lowest mode as the
local mode of the system with the given parametric combination in the first four modes of the system,
either in the string beam system or in the cable beam system with consideration of the cable sag.
3. The double-cable-stayed beam system can easily produce a global mode without the elastic support of
cables in out-of-plane motion. It can be inferred that the performance will become more prominent with
increasing beam span.
4. The single-degree-of-freedom nonlinear system subjected to single-frequency excitation shows a stronger
hardening spring property when adjusting the cable’s sag, and mass and stiffness ratios between the cable
and the beam, where the cable nonlinearity plays the major role.
5. Before and after the jump point occurs at saddle-node bifurcation, the system’s response to two-frequency
excitation has a substantial difference in its phase plane diagram and time history. Its lower branch behaves
more complicated than the higher one, and it has a double periodic steady-state solution. The damping
ratio can have a significant effect on the system’s softening/hardening spring property. The system with a
large damping ratio exhibits a weak hardening spring property. Moreover, in the amplitude and frequency
responses, multiple saddle-node bifurcation points appear, which will result in more complex behavior.

Acknowledgements This study is financially supported by National Science Foundation of China under Grant Nos. 11572117,
11502076, 11872176 and 11972151, and China Scholarship Council.

Compliance with ethical standards

Conflict of interest The authors declare that they have no conflict of interest.
Modeling, dynamics, and parametric studies

Appendix

For compact expression, the following integrals are introduced to define the coefficients ci and the excitation
amplitude Fi in Eq. (28).
 1  1  1  1
2
φ  c2 d x,
2
c1 = φc1 d x, c2 =
2
φc2 d x, c3 =
2
φ c1 d x, c4 =
0 0 0 0
 1  1  1  1
   
c5 = φc1 φ c1 d x, c6 = φc2 φ c2 d x, c7 = y  c1 φc1 d x, c8 = y  c2 φc2 d x,
0 0 0 0
 s1  s2  1
b1 = φb1
2
dx + φb2
2
dx + φb3
2
d x,
0 s1 s2
 s1  s2  1
b2 = f 1 φb1 d x + f 1 φb2 d x + f 1 φb3 d x,
0 s1 s2
 s1  s2  1
b3 = f 2 φb1 d x + f 2 φb2 d x + f 2 φb3 d x,
0 s1 s2
 
γ3
mv = βc2 c1 + c2 + b1 , (A.1)
ρ
where f i are the amplitudes of external harmonic excitation applied to the components of the beam. Therefore,
the coefficients ci and Fi are expressed as
 
βc2 γ3
c1 = ξc (c1 + c2 ) + ξb b1 ,
mv ρ
 
μ 1 1
c2 = − c3 c7 + c4 c8 + ε̄1 c5 + ε̄2 c6 ,
mv 2 2
μ
c3 = − (c3 c5 + c4 c6 ) ,
2m v
β2 γ 3
F1 = c b2 ,
mv ρ
β2 γ 3
F2 = c b3 . (A.2)
mv ρ
The coefficients in Eq. (43) are given as

6c3 F22 4c22 F22 8c22 F22

1 = − + − ,
(1 − 22 )2 (22 − 1)2 (22 − 4)(22 − 1)2
2c22 F2 3c3 F2 4c22 F2
2 = − − + ,
3(1 − 22 ) 1 − 22 (2 − 2)2 (22 − 1)
10 2
3 = c − 3c3 . (A.3)
3 2

References

1. Fujino, Y., Warnitchai, P., Pacheco, B.M.: An experimental and analytical study of autoparametric resonance in a 3DOF
model of cable-stayed-beam. Nonlinear Dynam. 4, 111–138 (1993)
2. Warnitchai, P., Fujino, Y., Susumpow, T.: A nonlinear dynamic model for cables and its application to a cable-structure
system. J. Sound Vib. 187(4), 695–712 (1995)
3. Xia, Y., Fujino, Y.: Auto-parametric vibration of a cable-stayed beam structure under random excitation. J. Eng. Mech.
132(3), 279–286 (2006)
4. Gattulli, V.A.P.: Planar motion of a cable-supported beam with feedback controlled actions. J. Intel. Mat. Syst. Str. 8, 767–774
(1997)
 Y. Cong et al.

5. Gattulli, V., Paolone, A.: A parametric analytical model for non-linear dynamics in cable-stayed beam. Earthq. Eng. Struct.
D 31, 1281–1300 (2002)
6. Fung, R.F., Lu, L.Y., Huang, S.C.: Dynamic modelling and vibration analysis of a flexible cable-stayed beam structure. J.
Sound Vib. 254(4), 717–726 (2002)
7. Lenci, S., Ruzziconi, L.: Nonlinear phnomena in the single-mode dynamics of a cable-supported beam. Int. J. Bifurcat. Chaos
19(3), 923–945 (2009)
8. Zhu, J., Ye, G.R., Xiang, Y.Q., Chen, W.Q.: Dynamic behavior of cable-stayed beam with localized damage. J. Vib. Control
17(7), 1080–1089 (2010)
9. Wei, M.H., Xiao, Y.Q., Liu, H.T.: Bifurcation and chaos of a cable-beam coupled system under simultaneous internal and
external resonances. Nonlinear Dynam. 67(3), 1969–1984 (2011)
10. Wei, M.H., Xiao, Y.Q., Liu, H.T., Lin, K.: Nonlinear responses of a cable-beam coupled system under parametric and external
excitations. Arch. Appl. Mech. 84(2), 173–185 (2013)
11. Wei, M.H., Lin, K., Jin, L., Zou, D.J.: Nonlinear dynamics of a cable-stayed beam driven by sub-harmonic and principal
parametric resonance. Int. J. Mech. Sci. 110, 78–93 (2016)
12. Wang, Z., Sun, C., Zhao, Y., Yi, Z.: Modeling and nonlinear modal characteristics of the cable-stayed beam. Eur. J. Mech.
A-Solid. 47, 58–69 (2014)
13. Wang, Z., Yi, Z., Luo, Y.: Energy-based formulation for nonlinear normal modes in cable-stayed beam. Appl. Math. Comput.
265, 176–186 (2015)
14. Peng, J., Xiang, M., Wang, L., Xie, X., Sun, H., Yu, J.: Nonlinear primary resonance in vibration control of cable-stayed
beam with time delay feedback. Mech. Syst. Signal Pr. 137(106488), 1–16 (2019)
15. Wang, Z.: Modelling with Lagrange’s method and experimental analysis in cable-stayed beam. Int. J. Mech. Sci. 176(105518),
1–13 (2020)
16. Liu, M., Zheng, L., Zhou, P., Xiao, H.: Stability and dynamics analysis of in-plane parametric vibration of stay cables in a
cable-stayed bridge with superlong spans subjected to axial excitation. J. Aerospace Eng. 33(1), 04019106 (2020)
17. Haibo, L., Jianjun, X.: Analysis on sensitivity of global mode of floating cablestayed bridge to CFRP cable. J. Dynam.
Control 16(3), 250–257 (2018). (In Chinese)
18. Zhang, Y., Fang, Z., Jiang, R., Xiang, Y., Long, H., Lu, J.: Static performance of a long-span concrete cable-stayed bridge
subjected to multiple-cable loss during construction. J. Bridge Eng. 25(3), 04020002 (2020)
19. Cao, D.Q., Song, M.T., Zhu, W.D., Tucker, R.W., Wang, C.H.T.: Modeling and analysis of the in-plane vibration of a complex
cable-stayed bridge. J. Sound Vib. 331(26), 5685–5714 (2012)
20. Song, M.T., Cao, D.Q., Zhu, W.D., Bi, Q.S.: Dynamic response of a cable-stayed bridge subjected to a moving vehicle load.
Acta Mech. 227(10), 2925–2945 (2016)
21. Kang, H.J., Guo, T.D., Zhao, Y.Y., Fu, W.B., Wang, L.H.: Dynamic modeling and in-plane 1:1:1 internal resonance analysis
of cable-stayed bridge. Eur. J. Mech. A-Solid. 62, 94–109 (2017)
22. Cong, Y., Kang, H., Guo, T.: Planar multimodal 1:2:2 internal resonance analysis of cable-stayed bridge. Mech. Syst. Signal
Pr. 120, 505–523 (2019)
23. Cong, Y., Kang, H., Guo, T.: Analysis of in-plane 1:1:1 internal resonance of a double cable-stayed shallow arch model with
cables’ external excitations. Appl. Math. Mech. Engl. 40(7), 977–1000 (2019)
24. Luongo, A., Rega, G., Vestroni, F.: Planar non-linear free vibrations of an elastic cable. Int. J. Nonlin. Mech. 19(1), 39–52
(1984)
25. Irvine, H.M.: Cable Structures. MIT Press, Cambridge (1981)
26. Nayfeh, A.H.: Nonlinear Oscillations. Wiley, New York (1979)
27. Benedettini, F., Rega, G.: Non-linear dynamics of an elastic cable under planar excitation. Int. J. Nonlin. Mech. 22(6),
497–509 (1987)
28. Kamel, M.M., Hamed, Y.S.: Nonlinear analysis of an elastic cable under harmonic excitation. Acta Mech. 214(3), 315–325
(2010)
29. Nayfeh, A.H.: The response of non-linear single-degree-of-freedom systems to multifrequency excitations. J. Sound Vib.
102(3), 403–414 (1985)
30. Zhao, Y., Guo, Z., Huang, C., Chen, L., Li, S.: Analytical solutions for planar simultaneous resonances of suspended cables
involving two external periodic excitations. Acta Mech. 229(11), 4393–4411 (2018)
31. Irvine, H.M., Caughey, T.K.: The linear theory of free vibrations of a suspended cable. P. Roy. Soc. A Math. Phy. 341(1626),
299–315 (1974)
32. Wang, L., Zhao, Y.: Nonlinear interactions and chaotic dynamics of suspended cables with three-to-one internal resonances.
Int. J. Solids Struct. 43(25–26), 7800–7819 (2006)

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional
affiliations.