Engineering Structures 30 (2008) 1160–1177

www.elsevier.com/locate/engstruct

Dynamic impact analysis of long span cable-stayed bridges

under moving loads
D. Bruno ∗ , F. Greco, P. Lonetti
Department of Structural Engineering, University of Calabria, 87030 - Rende (CS), Italy

Received 5 September 2006; received in revised form 19 June 2007; accepted 2 July 2007
Available online 21 August 2007

Abstract

The aim of this paper is to investigate the dynamic response of long span cable-stayed bridges subjected to moving loads. The analysis is
based on a continuum model of the bridge, in which the stay spacing is assumed to be small in comparison with the whole bridge length. As
a consequence, the interaction forces between the girder, towers and cable system are described by means of continuous distributed functions.
A direct integration method to solve the governing equilibrium equations has been utilized and numerical results, in the dimensionless context,
have been proposed to quantify the dynamic impact factors for displacement and stress variables. Moreover, in order to evaluate, numerically, the
influence of coupling effects between bridge deformations and moving loads, the analysis focuses attention on the usually neglected non-standard
terms related to both centripetal and Coriolis forces. Finally, results are presented with respect to eccentric loads, which introduce both flexural
and torsional deformation modes. Sensitivity analyses have been proposed in terms of dynamic impact factors, emphasizing the effects produced
by the external mass of the moving system and the influence of both “A” and “H” shaped tower typologies on the dynamic behaviour of the bridge.
c 2007 Elsevier Ltd. All rights reserved.

Keywords: Moving loads; Dynamic impact factors; Cable-stayed bridges; “A” and “H” shaped towers

1. Introduction The extension of the moving load problem to long

Cable-stayed systems have been employed, frequently, to
overcome long spans, because of their economic and structural
advantages. Moreover, improvements in the use of lightweight
and high strength materials have been proposed in different
applications, and, consequently, more slender girder cross
sections have been adopted. As a result, the external loads
have become comparable with those involved by the bridge
self-weight ones and an accurate description of the effects
of the moving loads is needed to properly evaluate dynamic
bridge behaviour. At the same time, new developments in
rapid transportation systems make it possible to increase the
allowable speed range and traffic load capacity; consequently,
the moving system can greatly influence the dynamic bridge
vibration, by means of non-standard excitation modes. To this
end, investigation is needed to quantify the effects produced by
the inertial forces of the moving system on the bridge vibration.
∗ Corresponding author. Tel.: +39 0984 496914; fax: +39 0984 494045.
E-mail address: d.bruno@unical.it (D. Bruno).
span cable-supported bridges requires a consistent approach,
appropriately formulated, in order to fully characterize the
bridge kinematics and train–girder interaction. In the literature,
several studies have been developed, which analyse dynamic
bridge behaviour with respect to different assumptions and
frameworks. In particular, Fryba and Timoshenko [1,2],
provided a comprehensive treatment concerning primarily
the dynamic response of simply supported girder structures
travelled by vehicles, and analytical as well as numerical
solutions for some specific problems have been presented.
During the last few decades, with advances in high performance
computers and computational technologies, more realistic
modelling of the dynamic interaction between a moving system
and bridge vibration has become feasible. In particular, Yang
et al. [3] presented a closed-form solution for the dynamic
response of simple beams subjected to a series of moving loads
at high speeds, in which the phenomena of resonance and
cancellation have been identified. Moreover, Lei and Noda [4]
proposed a dynamic computational model for the vehicle and
track coupling system including girder profile irregularity by

c 2007 Elsevier Ltd. All rights reserved.

0141-0296/$ - see front matter
doi:10.1016/j.engstruct.2007.07.001
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1161

surface roughness and long term deflection of the concrete

Nomenclature deck. An efficient numerical modelling has been developed
by Yang and Fonder [10] to analyse the dynamic behaviour
α Longitudinal stay geometric slope
of cable-stayed bridges subject to railway loads, taking into
α0 Longitudinal anchor stay geometric slope
account nonlinearities involved in the cable system. Dynamic
As Stay cross sectional area
interaction of cable-stayed bridges with reference to railway
As0 Anchor stay cross sectional area
loads has been investigated in [11], in which strategies to
b Half girder cross section width
reduce the multiple resonant peaks of cable-stayed bridges
β Transverse stay geometric slope
that may be excited by high-speed trains have been proposed
c Moving system speed
for a small length bridge structure. Finally, a computational
∆ Stay spacing step
model and a parametric study have been proposed in [12]
e Eccentricity of the moving loads with respect to
to investigate bridge vibration produced by vehicular traffic
the girder geometric axis
loads. The literature referred to above investigates dynamic
E Cable modulus of elasticity
bridge behaviour properly taking into account the effects
EI Flexural girder stiffness
of interaction between bridge vibration and the moving
EA Axial girder stiffness
system. However, only a few studies have concentrated on
E s∗ Stay Dischinger modulus
∗ the dynamic responses of long span bridges. This paper,
E s0 Anchor stay Dischinger modulus
therefore, focuses on the dynamic behaviour of long span cable-
g Girder self-weight per unit length
stayed bridges, evaluating the effects produced by the moving
γ Stay specific weight
system on the dynamic bridge behaviour. In particular, the
G Jt Torsional girder stiffness
main aims of this paper are to propose a parametric study
H Pylon height
p in a dimensionless context, which describes the relationship
I0 Pylon polar mass moment
between dynamic amplification factors and moving loads and
Kp Flexural top pylon stiffness
p bridge characteristics.
K0 Torsional top pylon stiffness
The structural model is based on a continuum approach,
l Lateral bridge span
which has been widely used in the literature to analyse
L Central bridge span
long span bridges [13–15]. In particular, Meisenholder and
Lp Total train length
Weidlinger [13] have schematized bridge structures as an elastic
λ Mass function of the moving system per unit
beam resting on an elastic foundation, whose stiffness is strictly
length
connected to the geometrical and stiffness properties of the
λ0 Polar mass moment of the moving system with
stays. Moreover, extended models which generalize the bridge
respect to girder geometric axis per unit length
kinematics have been proposed in [14,15], in which the stay
Mp Lumped top pylon equivalent mass
spacing is assumed to be small in comparison with the central
µ Girder mass per unit length
bridge span. As a result, the interaction forces between the cable
µ0 Polar inertial moment of the girder per unit length
system and the girder can be assumed as continuous functions
ω Girder torsional rotation
distributed over the whole girder length. The accuracy of the
p Live loads
continuum approach has been validated in previous works
σa Allowable stay stress
developed in both static and dynamic frameworks, through
σg Stay stress under self-weight loading
comparisons with numerical results obtained by using a finite
σg0 Anchor stay stress under self-weight loading
element model of the discrete cable system bridge [14–16].
ψ L(R) Left (L) and right (R) top pylon torsional
In the present paper, the bridge kinematics and the inertial
rotations
forces have been considered in a tridimensional context, in
u L(R) Left (L) and right (R) horizontal top pylon
which both in-plane and out-of-plane deformation modes have
displacement
been accounted for. Cable-stayed bridges based on both “H”
v Girder vertical displacement
and “A” shaped typologies with a double layer of stays
w Girder horizontal displacement
have been considered. However, cable-stayed bridges with
one central layer of stays, especially for eccentric railway
the finite element method, whereas additional references to bridges, are characterized by high deformability, and difficulties
the influence of AASHTO live-load deflection criteria on the verifying the design rules on maximum displacements occur
vibration in a railway track under moving vehicles can be found frequently. In particular, the girder torsional stiffness needs
in [5–7]. to be significantly improved with respect to those involved
With reference to cable-stayed bridges, in order to evaluate for “H” and “A” shaped typologies, because contributions
the amplification effects produced by the moving system, arising from the cable system are practically negligible. As
different investigations have been proposed. In particular, Au a matter of fact, torsional analysis carried out for typical
et al. [8,9] investigated the dynamic impact factors of cable- concrete or steel girder cross sections shows that in order to
stayed bridges under railway traffic using various vehicle limit torsional rotation to reasonable values (i.e. below 0.02),
models, evaluating the effects produced by random road the maximum allowable central length must be approximately
1162 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
equal to 400 m [16]. The equations of motion for the vehicle-
track-bridge element are derived by means of the Hamilton
principle. Subsequently, the boundary value problem, due to
the equilibrium equations, was solved, numerically, by means
of a finite difference scheme based on θ-family methods,
in which proper interpolation functions on both spatial and
time domains were adopted to obtain stable and accurate
results. A parametric study in a dimensionless context has been
analysed by means of numerical results, in terms of typical
kinematic and stress bridge variables for both in-plane and
eccentric loading conditions. In particular, results are proposed
to investigate the effects of moving the system description
with reference to non-standard forces, usually neglected in
conventional dynamic analyses, i.e. Coriolis and centripetal
accelerations. Finally, the influence on the dynamic bridge
behaviour of pylon typology with reference to both “A” and
“H” shapes has been analysed,and comparisons in terms of both
moving loads and tower characteristics have been proposed.
2. Cable-stayed bridge model
To begin with, the bridge geometry is presented with
respect to a fan-shaped self-anchored scheme and both flexural
and torsional deformation modes are evaluated for an “H”
shaped pylon typology. Subsequently, the formulation is
adapted for “A” shaped pylons. This can be easily derived,
as explained in the following sections, starting from the “H”
ones and introducing slight modifications to the main governing
equations. In both formulations, the cable system is arranged
symmetrically with respect to both zx and yz planes.
Long span bridges based on cable-stayed systems are
frequently analysed by means of a continuum approach, in
which the stays are assumed to be uniformly distributed along
the deck. In particular, the stay spacing is quite small in
comparison to the central bridge span (i.e. ∆/L  1). As
a result, the self-weight loads produce negligible bending
moments on the girder with respect to that raised by the moving
loads. The initial stress distribution, at the “zero configuration”,
is supposed to be produced by a correct erection process which
yields tension in the stays and compression in both the girder
and the pylons. Moreover, under dead loads only, the girder is
arranged with an initial straight profile, which is practically free
from bending moments for reduced values of the stay spacing
step. In particular, the erection procedure is based on the free
cantilevered method, which is able to control the initial tension
distribution in the cable system to a value practically constant
in each stay. This assumption has been verified for long span
bridges, in view of the prevailing truss behaviour of the cable-
supported structures [15–18]. Therefore, the moving loads
modify the initial configuration and, consequently, produce
additional stress and deformation states. It is worth noting that
for long span bridges, the initial stress state produced by the
dead loading needs to be accounted mainly in the cable-stayed
system, in which the initial tension strongly affects the stays’
behaviour due to Dischinger effects [17,18].
The geometry and stiffness characteristics of the bridges are
selected with respect to typical ranges suggested by practical
design rules [17,18]. In particular, the cross sectional stay areas
are designed so that the dead loads (g) produce constant stress
over all the distributed elements, which are assumed equal to
a fixed design value, namely σg . As a result, the geometric
area of the stays varies along the girder, but the safety factors
are practically constant for each element of the cable system.
Moreover, for the anchor stays, the cross sectional geometric
area, As0 , is designed in such a way that the allowable stress is
obtained in the static case, for live loads applied to the central
span only. Therefore, the geometric measurement for the cable
system can be expressed by the following equations:
g∆
As = ,
σg sin α
"  2 #1/2 " 2 # (1)
gl l L
As0 = 1+ −1 ,
2σg H 2l
where α is the slope of a generic stay element with respect
to the reference system, (L , l, H ) are representative geometric
lengths of the bridge structure, and ∆ is the stay spacing step
(for more details see Fig. 1). The bridge analysis is based on the
following assumptions:
(1) the stress increments in the stays are proportional to the live
loads, p;
(2) a long span fan shaped bridge is characterized by a
dominant truss behaviour.
In this framework, the tension σg and σg0 for distributed and
anchor stays, respectively, can be expressed by the following
relations:
g
σg = σa ,
g+p

 2 #−1 −1

" (2)
 p 2L
σg0 = σa 1 + 1− .
 g l 
It is worth noting that the allowable stay stress, σa ,
represents a known variable of the cable system in terms of
which the design tension under dead loading can be determined
by the use of Eq. (2). Since it is assumed that for dead loads only
the bridge structure remains in the undeformed configuration,
the application of moving loads leads to additional stress and
deformation increments with respect to the self-weight loading
condition. In particular, as reported in Fig. 1 with respect to
the reference system with the origin fixed at the midspan girder
cross section, the bridge kinematic, for the “H” shaped tower
typology, is described by following displacement variables:
• horizontal and vertical girder displacements [u(x, t),
v(x, t)],
• left and right horizontal pylon top displacements [u L (t),
u R (t)],
• girder torsional rotation [ω(x, t)],
• left and right pylon top torsional rotations [ψ L (t), ψ R (t)].
In particular, bridge deformations related to flexure and
torsion for the girder and pylons and axial deformations for
the girder and stays have been taken into account, whereas
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1163

Fig. 1. “H” shaped tower moving load problem: bridge kinematics and representative stiffness parameters.

pylon axial deformability has been neglected. Consistently that this influence is practically negligible (less than 3%),
with the bridge configuration reported in Fig. 1, the bridge whereas maximum relative percentage differences, less than
scheme is constrained with respect to both vertical and torsional 10%, are observed for speeds above 140 m/s. In addition, it has
displacements at boundary cross sections of the bridge and at been observed that the analysed maximum amplification factors
girder/pylon connections. occur when the moving system is basically applied on the
The stays are modelled as bar elements and the central bridge span. Therefore, the analysis has been developed
nonlinear behaviour is evaluated consistently with the by assuming the tangent modulus for the double layer of stays
Dischinger formulation [17], which takes into account acting on the lateral spans, whereas for the double layer of stays
geometric nonlinearities of the inclined stays introducing a acting on the central span and for the anchor stays the secant
fictitious elastic modulus for an equivalent straight member, in modulus has been employed.
this way: The axial deformation increments of a stay generic for the
left (L) and right (R) pylons produced by the moving system
E σ depend on both kinematic and geometric variables, as in the
E s∗ =   , with β =
σ0
, (3)
γ 2 l02 E 1+β following relationships:
1+
12σ03 2β 2
1
∆ε L = [(v ± ωb) · sin2 α
where E s∗ is known as the secant Dischinger modulus, E is the H
Young’s modulus of the cable material, γ the specific weight, − (u L ± ψ L b − u) · sin α cos α], (4)
l0 the horizontal projection of the stay length and σ0 and σ are 1
the initial and actual tension values of the stay, respectively, ∆ε R = [(v ± ωb) · sin2 α
H
i.e. σ0 = σg for the double layer of stays and σ0 = σg0 for − (u R ± ψ R b + u) · sin α cos α], (5)
the anchor stays. Moreover, the tangent value of the Dischinger
modulus can be obtained from Eq. (3) by putting β = 1. As where (+/−) refers, in Eqs. (4) and (5) and in the following
far as the secant modulus is concerned, its value depends on ones, to the right (+) and left (−) distributed stays with
cable stress states under self-weight and live loading conditions. respect to the longitudinal girder geometric axis. Similarly, for
Sufficient accuracy in the actual stress state might even be the left and right pylon anchor stays, the incremental axial
achieved by assuming β as proportional to the ratio between deformations are described as
live and self-weight loads (dominant truss behaviour) [16–18], 1
i.e. β = σσag ≈ p+g ∆ε L0 = [(u L ± ψ L b − u) · sin α0 cos α0 ] , (6)
g for the double layer of distributed stays H
(L/2l) 2
and β = σσag ≈ p+g g (L/2l)2 −1 for the anchor stays. On the
1
∆ε R0 = [(u R ± ψ R b + u) · sin α0 cos α0 ] , (7)
other hand, numerical investigations have been developed to H
analyse the influence of adopting the secant or the tangent where α0 is the girder/anchor stay orientation angle (Fig. 1).
equivalent moduli on the dynamic impact factors prediction. The external loads evolve at constant speed from left to right
The results, not presented here for the sake of brevity, show along the bridge development. A perfect connection between
1164 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177

the girder and the moving system is assumed. Interaction Hamilton principle is utilized to derive the dynamic equilibrium
forces produced by girder profile roughness and friction are equations. In particular, it is assumed that the damping energy
supposed to generate negligible effects with respect to the is practically negligible. This hypotheses is quite verified in
global bridge vibration. This assumption has been verified the context of long span bridges, where it has been proved
in the context of long span cable-supported bridge, where that the bridge damping effects tend to decrease as span
roughness effects have been considered as negligible [19]. As a length increases [20,21]. Detailed results about the influence of
result, the moving system has the same vertical displacements damping effects on the dynamic amplification factors (DAFs)
as the girder. Nevertheless, non-standard contributions arising have been presented in [9], from which it transpires that the
from Coriolis and centripetal inertial forces, produced by the assumption of an undamped bridge system leads to greater
coupling behaviour between the moving system and bridge DAFs. It is well known that the Hamilton principle can be
deformations, have been taken into account. With respect to a expressed for a conservative system and a generic time interval
fixed reference system, the velocity and acceleration functions as
of the moving system are evaluated consistently with a Eulerian Z t2
description of the moving loads as δ (T − V ) dt = 0 (13)
t1
∂v ∂v ∂ 2v ∂ 2v ∂ 2v
v̇ = + c, v̈ = 2 + 2c + 2 c2 , where T and V are the kinetic and the potential energy
∂t ∂x ∂t ∂ x∂t ∂x
of the whole dynamic system, respectively, and t1 and t2
∂x
with c = . (8) define the observation period. Therefore, with respect to these
∂t kinematic fields, the kinetic energy functional of the combined
The moving loads are consistent with a train system bridge–moving-load system may now be formulated as
typology, modelled by a sequence of lumped and distributed
1 L  2 1 L
Z Z
masses, representative of both bogie components and vehicle

T = µ v̇ + u̇ 2 dx + µ0 ω̇2 dx
bodies. However, for long span bridges, the internal bogie 2 −L 2 −L
spacing for an elementary vehicle is, usually, small in 1 p  1  
comparison with the whole bridge length. Moreover, within the + I0 ψ̇ L2 + ψ̇ D 2
+ M p u̇ 2L + u̇ 2D
2 2
same approximation level, the locomotive, even if it is much Z L 
1 L
Z
1 
heavier than the carriage, is distributed on a length, which is + ρ v̇ 2 + u̇ 2 dx + ρ0 ω̇2 dx,
assumed, in this context, to be smaller than the whole length 2 −L 2 −L
of the train. As a result, the moving system is supposed to be with L = l + L/2. (14)
described by equivalent uniformly distributed loads and masses
acting on the girder profile. However, improvements to the In particular, (µ, µ0 ) and (ρ, ρ0 ) are the mass functions and
moving load distribution can be easily provided just modifying polar inertial moments per unit length for both moving loads
Eqs. (9)–(12) and introducing a piecewise constant function to and girder, respectively. Moreover, M p is the equivalent lumped
describe carriages and locomotive loads. mass which refers to horizontal top pylon displacement, b is the
p
With respect to a moving reference system, x1 , from the bridge semi-width and I0 is the pylon polar mass moment for
p
left end of the bridge, the mass and loading functions during both left and right pylons, with I0 = b2 M p .
the external loading advance can be written by the following The total potential energy of the system, using a small
expressions, respectively: displacements formulation, can be written as

ρ = λH x1 + L p − ct H (ct − x1 ) , 1 L  1 L

Z Z
(9) 
V = E I v 002 + E Au 02 dx + G Jt ω02 dx
f = p H x1 + L p − ct H (ct − x1 ) ,


(10) 2 −L 2 −L
1 0 E s∗ As H 1 L E s∗ As H
Z Z
where x1 = x + (L/2 + l), (λ, p) are the vehicle body + 2
∆ε L dx + ∆ε2R dx
mass and loading forces per unit length and H (·) is the 2 −L ∆ sin α 2 0 ∆ sin α
Heaviside function. Moreover, the moving system is assumed 1 ∗  
+ E s0 As0 ∆ε 2L0 + ∆ε 2R0
to be eccentrically located with respect to the bridge half width, 2
and, consequently, distributed moment and rotatory inertial 1h p 2   i
+ K 0 ψL + ψ D 2
+ K p u 2L + u 2D
functions are introduced to properly describe the external loads 2
as Z L
− ( f · v + m · ω) dx
ρ0 = λ0 H x1 + L p − ct H (ct − x1 ) ,


(11) −L
L     
m = p · eH x1 + L p − ct H (ct − x1 ) ,
Z
1 L L


(12) + λv δ x + +δ x − v 2 dx (15)
2 −L 2 2
where λ0 represents the torsional distributed polar mass
moment produced by the external loading and e is the where (E I, E A, G Jt ) are the flexural, axial and torsional girder
stiffness and E s∗ As , E s0
∗ A


eccentricity of the moving loads with respect to the cross s0 are the axial stiffnesses of a
p
generic stay or the anchor cables. Moreover, K 0 , K p are the


sectional geometric axes. An energy approach based on the
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1165

torsional and flexural top pylon stiffness and δ (·) is the Dirac L
Z
qu R δu R + m ψ R δψ R dx


+
delta function. In particular, the last term on the right-hand side 0
of Eq. (15) denotes a penalty functional, with λv representing + K0
p
(ψ L δψ L + ψ R δψ R )
the penalty parameter, introduced to penalize girder vertical
+ K (u L δu L + u R δu R )
p
displacements and, consequently, to reproduce the connection  
between the girder and the pylons correctly. + SL0 δu − SL0 δu L + M L0 δψ L 

x=−L
By integrating by parts the first variation of the kinetic  
+ S R δu + S R δu R + M R δψ R 
0 0 0 
energy functional of the combined bridge/moving loads system x=L
and assuming that the virtual displacements vanish at both the Z L Z L
beginning and end of the actual varied path, the following − f · δvdx − m · δωdx
−L −L
expression is derived: )
Z L
Z t1 Z t2 Z L − [δ(x + L/2) + δ(x − L/2)]λv v · δvdx dt
δT dt = − [µ (v̈δv + üδu) + µ0 ω̈δω] dxdt −L
t1 t1 −L (17)
Z t2
p
ψ̈ L δψ L + ψ̈ R δψ R dt


− I0 where
t1
Z t2 E S∗ A S h i
v · sin3 α − u L(R) − (+)u · sin2 α cos α ,


− M (ü L δu L + ü R δu R ) dt
p qvL(R) =
H∆
t1
E ∗ AS h
Z t2 Z L qh L(R) = S (−)v · sin2 α cos α
− ρ̇0 (ω̇δω) dxdt H∆ i
− (−) u L(R) − u · cos2 α sin α ,


t1 −L
Z t2 Z L
− ρ0 ω̈δωdxdt E ∗ AS h i
t1 −L
qu L = S −v sin2 α cos α + (u L − u) · cos2 α sin α , (18)
H∆
t2 L E S∗ A S h
Z Z i
− ρ (v̈δv + üδu) dxdt qu R = −v sin2 α cos α + (u R + u) · cos2 α sin α ,
t1 −L H∆
E ∗ A S b2  
t2 L m ω L(R) = S ω sin3 α − ψ L(R) cos α sin2 α ,
Z Z
− ρ̇ (v̇δv + u̇δu) dxdt. (16) H∆
t1 −L E S∗ A S b2  
m ψ L(R) = −ω sin2 α cos α + ψ L(R) cos2 α sin α .
It is worth noting that in Eq. (16), the last two terms on H∆
the right-hand side denote the kinetic energy contributions
∗ A
E s0
produced by the external moving mass, in which the s0
0
u − (+)u L(R) cos2 α0 sin α0 ,


SL(R) =
time derivative for vertical displacement has been assumed H
∗ 2
(19)
consistently with Eq. (8)). As a consequence, non-standard E s0 As0 b
terms due to both Coriolis and centripetal forces are introduced
0
M L(R) = ψ L(R) cos α0 sin α0 .
2
H
in the kinetic functional, which are strictly connected with the
It is worth noting that Eqs. (18) and (19) correspond to
interaction behaviour between the bridge deformations and the
distributed internal forces due to cable system/girder interaction
moving system. Moreover, the time dependence of the moving
and concentrated forces and moments applied to the left
mass determines additional contributions to the inertial forces,
and right girder cross section ends due to the anchor stays,
which are, basically, produced by an unsteady distribution of
respectively (Fig. 2). Moreover, in Eq. (18), the angular
the train/bridge system mass. Taking into account the first
parameter α, as depicted in Fig. 1, depends on the longitudinal
variation of the total potential energy function, expressed by
coordinate x and the geometrical bridge lengths (l, H, L).
Eq. (15), the following expression is obtained:
Assuming the bridge scheme reported in Fig. 1, the boundary
Z t1 Z t1 (Z L h i conditions at the left and right girder cross section ends require
δV dt = E I v I V δv − E Au 00 δu − G Jt ω00 δω dx null values for vertical displacements, bending moments,
t1 t1 −L torsional rotations and specified horizontal axial forces, as in
+ T δv − Mδv 0 + N δu + Mt δω −L
 L the following equations:
ω = 0, v = 0, v 00 = 0


Z 0 at x = ±L;
+ (qvL δv + qh L δu + m ωL δω) dx
0
0
E Au 0 L − S 0R = 0.



−L E Au −L + SL = 0, (20)
Z L
+ (qv R δv + qh R δu + m ω R δω) dx The dynamic equilibrium equations can be obtained in
0 explicit form by means of the variation statement of the
Z 0 Hamilton principles. In particular, by substituting Eqs. (16)
qu L δu L + m ψ L δψ L dx


+ and (17) into Eq. (13) and taking into account the boundary
−L
1166 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177

Fig. 2. Dynamic interaction forces between bridge components.

conditions, the following dynamic equilibrium equations are the equilibrium conditions for both internal and external forces.
derived: In order to develop a generalized formulation, the dynamic
Girder equilibrium equations have been proposed in dimensionless
form, introducing the following parameters, related to both
µv̈ + E I v I V + H (x) qv R + H (−x) qvL + ρ̇ v̇ bridge and moving load characteristics:
 
+ ρ v̈ + 2cv̇ 0 + c2 v 00 Girder
− f − δ (x + L/2) + δ (x − L/2) λv v = 0,
 
(21) 4I σg 1/4 Ct σg 1/2
   
εF = , εω = ,
µü − E Au + H (x) qh R + H (−x) qh L + ρ̇ u̇ + ρ ü = 0, (22)
00 H 3g Eb2 H g
(28)
λ µ0
 
µ0 ω̈ − G Jt ω00 + H (x) m ω R + H (−x) m ωL + ρ̇0 ω̇ Aσg
εA = , ζ= , ζ0 = .
+ ρ0 ω̇ − m = 0. (23) Hg µ µb2

Left top pylon Pylon, cable system

!
K p σg γ 2H2E
Z 0   MP
0
SL − qu L dx − K P u L + M p ü L = 0, −L ≤ x ≤ 0, ζP = , ηp = , a= ,
−L µH Eg 12σg3
(29)
(24) 
1+β

Z 0 a=a .
p p
I0 ψ̈ L + K 0 ψ L + m ψ L dx + M L0 = 0, −L ≤ x ≤ 0. 2β 2
−L
Moving loads
(25)
1/2
µσg

Right top pylon pσg e
p= , e= , ϑ =c ,
Z L Eg b Eg H
0
  (30)
qu R dx + K P u R + M p ü R = 0,
 1/2
SR + 0≤x ≤L Eg
0 τ =t .
µH σg
(26)
p p
Z L Bridge kinematic
I0 ψ̈ R + K 0 ψ R + m ψ R dx + M R0 = 0, 0 ≤ x ≤ L.
0 v u uR uL
V = , U= , UR = , UL = . (31)
(27) H H H H
A synoptic representation of both internal and external In particular, (ε F , ε A , εω ) are the relative bending, axial
forces has been proposed in Fig. 2, and, as an alternative to and torsional stiffness ratios between the girder and the cable
the variational approach, the governing equations can be easily system, respectively. Moreover, (a) identifies a bridge size
derived consistently with a local approach taking into account parameter strictly connected with the deformability of the cable
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1167

system caused by the Dischinger effect and η p , ζ P define the

pylons, depend only on the time variables. Therefore, in order
inertial and stiffness properties for both right and left pylons. to utilize standard numerical methods to solve PDE system
The moving load characteristics are reported in Eq. (30), equations, the integro-differential equations, given by Eqs.
in which ( p, e) describe the dimensionless applied loads and (32)–(38), are converted into a purely differential form. The
eccentricity with respect to the girder cross sectional geometric equivalence between the original and the modified systems
axis, whereas (ζ, ζ0 ) correspond to the mass and polar mass is guaranteed by the use of the penalty method, by which
moment ratio between the external loads and the girder. Finally, it is possible to convert the equilibrium conditions, from
(ϑ, τ ) define the dimensionless moving load speed and the integral to local form. In particular, the pylon kinematics
time variables. Adopting the following notation for spatial and (namely U L , U R ψ R , ψ L ) are assumed, fictitiously, to depend on
time derivatives, f 0 = ∂∂ Xf , f˙ = ∂∂τf , with X = Hx , the both time and space variables. However, they are constrained
dynamic equilibrium equations are now presented in terms over the spatial domain by means of penalty
functionals,
of the normalized kinematic variables, dimensionless bridge involving two penalty parameters, i.e. K U , K ψ . Moreover,
and moving system parameters, by means of the following distributed and lumped forces arising from the cable system
expressions: and pylons have been taken into account, utilizing Heaviside
Girder and delta Dirac functions, respectively. As a result, the integro-
! differential dynamic equilibrium equations, i.e. Eqs. (35)–(38),
ε 4F are transformed into the following relationships:
V I V + V̈ + ϕV
4  
L
H 1 [−ϕ1 V − ϕ2 (U − U L )] + δ X +
− ϕ1 H 1 (U L − U ) + H 2 (U R + U ) − p f 1


2H
ζ  
× η p U L + ζ P Ü L + K U U L = 0,

00
(39)
+ f 2 V̇ + ζ f 1 V̈ + 2θ V˙ 0 + θ 2 V 00
θ 
L

H 1 (ϕ1 ϑ − ϕ2 ψ L ) + δ X + ζ p ψ̈ L + η p ψ L


+ ΛV δ (X + L/2H ) + δ (X − L/2H ) V = 0,
 
(32) 2H
ε A U 00 − Ü − H 1 (ϕ1 V + ϕ2 (U − U L ))

+ K ψ ψ L00 = 0, (40)
ζ 
L

+ H 2 (−ϕ1 V + ϕ2 (U + U R )) − f 2 U̇ − ζ f 1 Ü = 0, (33)

θ −H 2 [ϕ1 V − ϕ2 (U + U R )] + δ X −
2H
εω ω00 − ζ0 ω̈ − ϕω + ϕ1 H 1 ψ L + H 2 ψ R


× η p U R + ζ P Ü R + K U U R = 0,
00


(41)
H ζ0
f 2 ω̇ − ζ0 f 1 ω̈ = 0.
 
+ p f1e − (34) L
−H 2 (ϕ1 ϑ − ϕ2 ψ R ) + δ X − ζ p ψ̈ R − η p ψ R


b θ
2H
Left pylon
+ K ψ ψ 00 = 0. (42)
Z 0
[−ϕ1 V − ϕ2 (U − U L )] dX + η p U L + ζ P Ü L


By comparing the sets of Eqs. (39)–(42) and Eqs. (35)–(38),
−L/H related to the differential and the integro-differential forms,
− χ (U − U L ) = 0, (35) respectively, it is easy to recognize that the internal forces
Z 0 referred to the anchor stays have not yet been introduced. In
ζ p ψ̈ L + η p + χ ψ L + (ϕ1 ϑ − ϕ2 ψ L ) dX = 0.


(36) particular, the transformation of equilibrium equations related
−L/H to the pylons, from global into local form, leads to considering
Right pylon the concentrated forces and moments arising from anchor stay
Z L/H axial forces as boundary conditions, by means of the following
− [ϕ1 V − ϕ2 (U + U R )] dX + η p U R + ζ P Ü R

equations:
0
1
+ χ (U + U R ) = 0, (37) U L0 (−L/H, τ ) = − χ (U − U L ) , U L0 (L/H, τ ) = 0
KU
Z L/H
1
ζ p ψ̈ R + η p + χ ψ R + (ϕ1 ϑ − ϕ2 ψ R ) dX = 0, U R0 (L/H, τ ) = χ (U + U R ) , U R0 (−L/H, τ ) = 0


(38)
0 KU
1 (43)
E∗ A σ
with H 1 = H (−X ) , H 2 = H (X ), χ = S0E 0 g Hg sin α0 ψ L0 (−L/H, τ ) = χ, ψ L0 (L/H, τ ) = 0,
σ
Kψ
cos2 α0 , ΛV = λv Egg , whereas the functions ( f 1 , f 2 , ϕ, ϕ1 , ϕ2 ), 1
for the sake of brevity, are reported in Appendix A. Moreover, ψ R0 (L/H, τ ) = χ, ψ R0 (L/H, τ ) = 0.
Kψ
details of the derivation of dynamic equilibrium equations can
be found in Appendix B, in which, for conciseness, only Eq. Finally, the following relationships summarize the remain-
(32) has been discussed. ing boundary conditions, i.e. Eq. (20), which impose that the
The dynamic equilibrium equations introduce an integro- vertical and torsional displacements vanish at the left and right
differential boundary value system. Moreover, in Eqs. (35)– and bridge cross section ends, natural conditions for girder ax-
(38), the kinematic variables, related to both left and right ial displacements and homogeneous relationships with respect
1168 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177

Fig. 3. Moving load problem for “A” shaped tower typology: bridge kinematics and internal force distribution.

to the time variable: in Fig. 3, introducing the misalignment angles, i.e. α, β,

χ which describe the orientation of a generic stay element with
U 0 (−L/H, τ ) = (U − U L ) , respect to vertical and longitudinal directions, respectively,
εA
χ and transverse displacement w regarding the xy plane, the
U 0 (L/H, τ ) = − (U + U R ) , interaction forces between cable system and girder assume the
εA
following expressions:
V (−L/H, τ ) = 0, V (L/H, τ ) = 0,
V 00 (−L/H, τ ) = 0, V 00 (L/H, τ ) = 0, E S∗ A S h
qvL(R) = v · sin3 α
ω(−L/H, τ ) = 0, ω(L/H, τ ) = 0, (44) H∆ i
− u L(R) − (+)u · sin2 α cos α cos β ,


ω(X, 0) = 0, ω̇(X, 0) = 0,
U (X, 0) = 0, U̇ (X, 0) = 0, E ∗ AS h
qh L(R) = S (−)v · sin2 α cos α cos β
V (X, 0) = 0, V̇ (X, 0) = 0, H∆ i
− −u L(R) − u · cos2 α sin α cos2 β ,


ψi (X, 0) = 0, ψ̇i (X, 0) = 0, Ui (X, 0) = 0,
U̇i (X, 0) = 0 i = R, L . E ∗ AS h
qu L(R) = S −v sin2 α cos α cos β (45)
H∆
From a numerical point of view the penalty stiffness i
+ u L(R) − (+)u · cos2 α sin α cos2 β ,


parameters, ΛV , K U , K ψ , are assumed to be sufficiently high

to impose proper constraint conditions, but not, exceedingly, E S∗ A S b2


∂w
to introduce numerical instabilities in the computation. For a m ω L(R) = ϑ sin3 α + cos α sin2 α cos β
H∆ ∂x
numerical evaluation of the stiffness values, from the Author’s
w 2

experience of applying the model over different structures, a − sin α cos α sin β
range between [105 –106 ] is suggested. b
∗
E As0
Long span bridges are frequently designed with “A” shaped 0
= s0 u − (+)u L(R) cos2 α0 sin α0 cos β,


pylon typologies, which are able to efficiently reduce both SL(R)
H
torsional and flexural deformations produced by eccentric or
transverse loading. In order to analyse such a bridge typology,
E ∗ AS
the previous formulation still applies, but slight modifications qwL(R) = S sin α cos α sin β
to the main governing equations need to be performed. As H∆
∂w
 
a matter of fact, the dynamic equilibrium equations with × ωb · sin α − (+) b cos α cos φ − w cos α sin β ,
respect to flexural deformation in the plane xy are, basically, ∂x
∗ (46)
the same. In contrast, the top pylon kinematic is governed E AS b
m wL(R) = − (+) S sin α cos α cos β
by longitudinal displacements only, because both out-of-plane H∆
(xz) and torsional (with respect to y) kinematic parameters ∂w
 
are assumed to be meaningless. Moreover, the distributed × ωb · sin α − (+) b cos α cos β − (+)w cos α sin β ,
∂x
stays are now inclined with respect to the vertical direction,
i.e. the y axis, and determine a coupling behaviour between where (qi , m ω , S 0 ) with (i = v, h, u), are the internal forces
out-of-plane flexural and torsional deformations. As depicted produced by the cable system girder and pylons. Moreover,
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1169

Table 1
Comparison of maximum normalized torsional rotation in terms of the relative torsional stiffness (εω ) and bridge size parameter (a) between simplified (S) and
general (G) formulations

a Model εω = 0.05 εω = 0.1 εω = 0.15 εω = 0.2 εω = 0.25

0.05 G 115.39 93.21 74.85 61.69 53.08
S 115.89 93.57 75.06 62.05 53.50
0.10 G 184.65 139.79 107.17 86.44 77.08
S 183.78 137.35 105.84 86.82 77.22
0.15 G 237.77 174.45 133.39 111.17 237.77
S 237.84 174.37 133.29 111.14 237.84
0.20 G 272.12 209.57 164.16 135.73 272.12
S 271.76 210.71 164.20 135.97 271.76
0.25 G 338.72 250.96 195.65 160.61 338.72
S 332.65 246.57 191.63 157.65 332.65
0.30 G 381.34 283.29 218.71 176.48 160.61
S 377.06 280.07 217.45 176.61 151.20

Table 2
Comparisons of maximum normalized torsional rotation in terms of speed parameter (θ ) and relative torsional stiffness (εω ) between simplified (S) and general (G)
formulations
θ Model a = 0.1 a = 0.2
εω = 0.1 εω = 0.15 εω = 0.2 εω = 0.1 εω = 0.15 εω = 0.2
0.05 G 117.99 95.31 83.67 191.55 152.52 128.74
S 117.31 94.97 83.66 188.85 150.41 129.68

0.07 G 123.33 99.15 84.60 190.29 152.44 131.02

S 122.80 98.82 84.49 191.65 153.11 131.01

0.09 G 127.93 101.54 85.06 209.95 164.71 136.03

S 127.98 101.82 85.38 208.57 163.62 135.53

0.11 G 139.56 106.54 88.09 223.88 172.00 140.75

S 138.49 106.20 87.78 222.20 170.95 139.88

0.13 G 142.69 112.63 92.10 238.69 181.20 144.29

S 143.75 113.48 92.64 237.56 181.12 145.11

qwL(R) and m wL(R) represent distributed forces and moments
in the (xz) plane (Fig. 3), produced by the coupling behaviour
by torsional and flexural deformations, arising from the inclined
stays of the “A” shaped tower typology.
The dynamic equilibrium equations, for “A” shaped bridge
typology, can be easily derived, starting from Eqs. (21)–
(26) and making use of Eq. (45) to describe the interaction
forces between the cable system and the girder. Moreover, an
additional equilibrium equation is required due to the presence
of the transverse displacement, w. In particular, as reported in
Fig. 3, the following expression regarding flexural deformation
in the xy plane is introduced:
µẅ + E Is w I V + H (x) qs R + m 0s R


+ H (−x) qs L + m 0s L = 0.


(47)
Finally, the dynamic equilibrium equations for the “A”
shaped tower typology can be summarized by Eqs. (21)–(26)
and (47). However, in the typical range of both geometrical,
mechanical and moving load characteristics, investigations
have been shown that for in-plane loading conditions (loads
applied in the x y plane), the transverse deformations are
practically negligible and do not influence dynamic bridge
behaviour. To this end, sensitivity analyses have been proposed
in terms of maximum normalized torsional rotation during
moving load application, i.e. (ω/ pe). In particular, results
concerning the actual solution, namely the General Approach
(GA), derived from Eqs. (21)–(26) and (47), and a simplified
one, namely the Simplified Approach (SA), obtained assuming,
“a priori”, that the transverse displacements are negligible,
i.e. w(x, t) = 0, have been compared. The following bridge
and moving load parameters, typically utilized in practical
applications, have been assumed constant during the analyses:
e/b = 0.5, p/g = 1, Jw /J = 100, L p = 750 m,
ζ = 1, ζ0 = 0.5, η p = 0.085.
(48)
Variability with respect to dimensionless torsional girder
stiffness, εω , bridge size parameter, a, and moving system
speed, θ, have been investigated. The results reported in
Tables 1 and 2 denote that the dynamic bridge behaviour is
practically unaffected by the transverse flexural deformations
deriving from w displacements. The actual solution and the
simplified ones make the same prediction, with an error of less
than 2%.
1170 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
The dynamic equilibrium equations for “A” shaped towers
in the dimensionless context are not presented here for the sake
of brevity, but they can be easily derived by introducing the
dimensionless parameters previously defined in Eqs. (28)–(31)
in the governing equations.
3. Numerical procedure
The dynamic equilibrium equation system introduces a PDE
system from which it is quite difficult to derive an analytical
solution, because a large variable number and complexities are
involved in the main equations. The governing equations are
converted to an equivalent differential system of the first order.
In particular, for any dependent variables involved with an
order higher than the first one, additional functions representing
all lower order time derivatives are introduced, by means of
supplementary equations, which are appended to the main
system. As a result, the reformulated boundary value problem
assumes the following form:
ai j (X, τ ) y 0j + bi j (X, τ ) ẏ j = f i (y1 , y2 , y3 , . . . , y14 , X ) ,
with i, j = 1, 14, (49)
where y = (V, V 0 , V 00 , V 000 , V̇ , U, U 0 , U̇ , U R , U R0 , U̇ R , U L ,
U L0 , U̇ L ) is the vector

of unknown functions or primary
variables and ai j , bi j are constants depending on both the
moving loads and the bridge properties. Moreover, f i represent
proper transformation operators, which define the relationship
between primary variables and known quantities, in accordance
with the PDE system given by Eqs. (32)–(34) and (39)–(42).
In accordance with Eqs. (43) and (44), initial and boundary
conditions with respect to both space and time are introduced
by means of the following equations:
B1 yi X , 0 = e yi , B2 yi (0, τ ) = y i ,


with X = ±L/H, i = 1 . . . 14, (50)
where B1 , B2 are proper transformation matrices, which
guarantee the consistency of the boundary conditions with Eq.
yi , y i represent known quantities related to the


(49), and e
temporal and spatial variables, respectively.
A numerical integration scheme has been utilized by means
of a finite difference method, which uses a second-order centred
implicit scheme for both time and spatial derivatives [22].
The method has a truncation error of O((∆t)2 (∆x)2 ) and is
unconditionally stable for all time steps. In order to capture the
rapid changes of the solution during the time integration, the
whole domain has been discretized by means of an accurate
mesh point number. Moreover, requested values, those which
do not lie on a mesh point, have been computed using a
Lagrange interpolation that uses four space subdivisions and
three time points. The solution process is obtained consistently
with an error control procedure, which is able to integrate
the nonlinear equations with respect to upper bounds’ error
tolerances related to both time and spatial variables. In
particular, the spatial error estimate is obtained by comparing
the solution to another one computed on a coarser spatial mesh,
but assuming the same time step. This gives an estimate of the
O(h ∧ 2) local truncation error of second order in h (where h
is the mesh spacing for the spatial variable). Alternatively, the
time error estimate is obtained by comparing the solution to
another computed with a larger time step, but the same spatial
mesh. This gives an estimate of the O(k ∧ 2) local truncation
error of second order in k (where k is the time step). However,
these estimates are local, so they do not account for situations
where a small perturbation in the solution at a time t = t1 can
lead to a large change in the solution at a later time.
The numerical results are derived providing at first a
trial integration time step, which is subsequently reduced by
means of a proper adaptive procedure in order to satisfy the
convergence conditions. Contrarily, the spatial discretization
remains fixed during the analysis, and consequently, in order
to minimize the integration errors, a proper mesh point number
over the bridge structure has been adopted. In the following
results, the spatial domain is discretized utilizing more than
10 000 subdivisions over the whole bridge length. The initial
integration time step, which is automatically reduced due to the
time adaptation procedure, is assumed as at least 1/1000 of the
observation period defined as the time necessary for the moving
train to cross the bridge. On a Pentium IV processor at 3000 Mz
the CPU time required for performing the time history for each
case was approximately 3 min.
4. Numerical results and parametric study
The results define the relationship between the characteris-
tics of the bridge and applied moving loads, emphasizing the
effects produced by the external mass on the dynamic bridge
vibrations. In particular, a parametric study is proposed, which
describes cable-stayed bridge behaviour in terms of dimension-
less variables, strictly related to both the moving loads and the
bridge characteristics. Numerical results are presented in terms
of dynamic impact factors, in order to quantify the amplification
effects produced by the moving loads over the static solution
(i.e. st), by means of the following relationship:
max X t=0...T
ΦX = (51)
X st
where T is the observation period and X is the variable under
investigation. The parametric study has been developed to
investigate the following variables:
• φV dynamic amplification factor of the midspan vertical
displacement,
• φ M dynamic amplification factor of the midspan bending
moment,
• φσ0 dynamic amplification factor of the axial force in the
anchor stay,
• φσ dynamic amplification factor of the axial force in the
longest central span stay.
• φω dynamic amplification factor of the midspan girder
torsional rotation.
The bridge and moving load dimensioning is selected in
accordance with the values utilized in practical applications
and due mainly to both structural and economical factors.
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177 1171

Table 3
Percentage errors of midspan vertical displacement and bending moment
dynamic amplification factors (ΦV , Φ M ) between the moving force model
(MFM), standard acceleration (SA) and proposed results for different
normalized speed parameters (θ )

θ ΦV ΦM
Error % SA Error MFM % Error SA % Error MFM %
0.02 0.37 0.51 0.87 0.21
0.04 1.16 3.54 1.58 21.93
0.05 5.11 6.40 2.59 33.64
0.07 9.24 10.60 11.15 33.87
0.09 2.42 11.25 18.23 39.69
0.11 4.65 20.74 38.32 52.41
0.13 19.17 31.34 37.58 42.87

The dimensionless parameters related to aspect ratio, pylon

stiffness, allowable cable stress and moving load characteristics
are assumed as equal to the following representative values:

L/2H = 2.5, l/H = 5/3,

σa /E = 7200/2.1 × 106 , K p /g = 50, (52)
p/g = [0.5 − 1], ζ0 = [0 − 2].
Fig. 4. Midspan vertical displacement dynamic impact factor (ΦV ) vs normal-
In order to evaluate the influence of both mass schematiza- ized speed parameter (θ ).
tion and moving system speed, comparisons in terms of the
dynamic amplification factors (DAFs) have been proposed in
Figs. 4–7. In particular, the actual solution has been compared
with numerical results based on the following assumptions:
(1) Inertial description of the moving system completely
neglected, i.e. p 6= 0, ζ = 0, namely the moving force
model (MFM).
(2) Inertial description of the moving system neglected with
respect to non-standard inertial forces, i.e. p 6= 0, ζ 6= 0,
f 2 = 0, 2θ V˙ 0 = θ 2 V 00 = 0, namely Standard Analysis
(SA).
The proposed results do not agree with those arising from the
SA, especially at high speed of the moving system, where it
has been shown that non-standard terms in the acceleration
function provide notable amplifications in both kinematic
and stress variables. Moreover the comparisons between the
proposed formulation and those concerning the MFM are not
in agreement in wide ranges of the speed parameter. However,
for reduced values of moving system speed, i.e. θ → 0,
the results arising from the dynamic and static solutions are
practically coincident and, consequently, the influence of the
mass schematization becomes negligible. In order to quantify
numerically the influence of the inertial effects of the moving Fig. 5. Bending moment dynamic impact factor (Φ M ) vs normalized speed
system, the percentage errors between the SA and the MFM and parameter (θ).
the proposed formulation have been reported in Table 3. Finally,
in Figs. 4 and 5, dynamic amplification variability with respect structural and practical conditions, which guarantee stability
to the speed parameter for different intensity ratios between live of the anchor stays, avoiding excessive steel quantity amount
and self-weight loads, are proposed. in the cable system [16–18]. The DAFs for both bending
In Tables 4 and 5, the influence of the geometric ratios of moments and displacements generally grow for increasing
the bridge, i.e. L/l and l/H , on the DAFs is investigated at ratios between the central span and the height of the pylons,
fixed speed of the moving system and relative girder stiffness, because of the most greater deformability of the structural
(i.e. εω = 0.2, θ = 0.10. In particular, the bridge geometry is system. However, the impact factors based on the bending
assumed to verify well-known design rules derived from both moments are quite dependent from the geometric aspect
1172 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177

Table 4
Dynamic amplification factors for midspan vertical displacement (ΦV ) vs geometric bridge ratios L/l and H/L

L/l a = 0.1 a = 0.2

L/H = 4 L/H = 5 L/H = 6 L/H = 4 L/H = 5 L/H = 6
2.75 1.308 1.356 1.390 1.25 1.29 1.35
3.00 1.282 1.332 1.378 1.20 1.25 1.32

3.25 1.256 1.312 1.371 1.18 1.23 1.31

3.50 1.235 1.304 1.374 1.16 1.22 1.32

Table 5
Dynamic amplification factors for midspan bending moment (Φ M ) vs geometric bridge ratios L/l and H/L

L/l a = 0.1 a = 0.2

L/H = 4 L/H = 5 L/H = 6 L/H = 4 L/H = 5 L/H = 6
2.75 1.903 2.394 3.094 1.464 1.855 2.590
3.00 1.664 2.280 2.908 1.431 1.727 2.407

3.25 1.674 2.090 2.390 1.722 1.652 2.361

3.50 1.563 1.906 2.583 1.588 1.821 2.262

Fig. 6. Anchor stay dynamic impact factor (Φσ0 ) vs normalized speed Fig. 7. Longest centre span stay dynamic impact factor (Φσ ) vs normalized
parameter (θ). speed parameter (θ).

ratios than the corresponding ones for vertical displacements.

Moreover, at fixed L/H , the DAFs for vertical displacements
are quite unaffected by the ratio between the main and central
spans, because the cable system stiffness remains practically
constant during the analyses.
The relationship between the DAFs and bridge size is
investigated in Figs. 8 and 9. In typical allowable ranges
of the a parameter, the DAFs, related to cinematic and
stress bridge variables, are analysed for an external moving
system with a constant speed advance and different loading
lengths (namely c = 120 m/s, L p1,2 = 500, 1000 m,
M p σg /µK = 2.3). The comparisons are proposed to
investigate the effect of the external moving mass on the
dynamic bridge vibrations. In particular, the results show a
tendency to decrease with an oscillating behaviour and some
local peaks in curve development. The inertial effects produce
considerable amplifications in both the displacement and stress
variables, especially, for low values of the bridge size parameter
a. Moreover, results concerning the MFM determine notable
underestimates in both stress and displacement DAFs.
In Figs. 10 and 11, the dynamic behaviour of the bridge
is investigated with respect to the dimensionless parameter
ε F , which defines the normalized stiffness of the girder with
respect to the cable system. In particular, the analysis has
been developed at a constant speed of the moving system
and for different values of the bridge size parameter, namely
a = (0.1, 0.2), emphasizing the influence of the external
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
Fig. 8. Midspan displacement dynamic impact factor (ΦV ) vs bridge size
parameter (a).
Fig. 9. Bending moment dynamic impact factor (Φ M ) vs bridge size parameter
(a).
moving mass description on the dynamic behaviour of the
bridge. As a matter of fact, the actual solution is compared
to the case in which the inertial effects of the train loads
have not been accounted for. The dynamic bridge behaviour
appears to be quite sensitive to the external mass description,
and underestimates of the dynamic impact factors are noted
if the travelling mass has not been properly evaluated. The
major amplification effects are noted for low ranges of the
1173
Fig. 10. Midspan displacement dynamic impact factor (ΦV ) vs relative girder
stiffness parameter (ε F ).
Fig. 11. Bending moment dynamic impact factor (Φ M ) vs relative girder
stiffness parameter (ε F ).
ε F parameter, in which the bridge structure is basically more
flexible and, mainly, dominated by the cable-stayed system.
In contrast, for high values of ε F , corresponding to girder-
dominated bridge structures, the effects of the inertial forces
of the moving system are notably reduced.
The dynamic bridge behaviour is analysed with respect to
eccentric loads, which involve both flexural and torsional de-
formations. In particular, in order to evaluate the amplification
1174 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
Fig. 12. Midspan torsional rotation dynamic impact factors (Φω ) vs normalized
speed parameter (θ ) for HST.
Fig. 13. Midspan torsional rotation dynamic impact factors (Φω ) vs normalized
speed parameter (θ) for AST.
effects produced by moving loads for bridge structures based
on both “A” and “H” shaped towers (namely AST, HST), a sen-
sitivity analysis has been developed. The results are presented
in term of maximum normalized torsional rotation and rela-
tive DAF produced by the moving load application, at midspan
girder cross section, i.e. X = 0.
In Figs. 12 and 13, the effects of mass distribution of the
moving system for both AST and HST is investigated in terms
of midspan torsional rotation for different values of the loading
strip length L p . In particular, in Fig. 12, for the HST typology,
dynamic amplification displays a tendency to grow especially
for reduced ratios of loading application length and central
bridge span. Moreover, the inertial forces of the moving system
Fig. 14. Midspan torsional rotation dynamic impact factors (Φω ) vs relative
girder stiffness (εω ) parameter for AST-HST.
Fig. 15. Maximum normalized displacement (ω) vs relative girder stiffness
(εω ) parameter for AST-HST.
determine an intensification of the DAFs. In contrast, in Fig. 13,
the AST denotes a smaller dependence on the loading strip
length and the moving mass schematization. This behaviour
can be explained due to the fact that the AST typologies show,
generally, greater stiffness with respect to the HST ones, which
strongly reduce dynamic amplifications, and as a result the
effects of the moving mass becomes negligible.
In Figs. 14 and 15, sensitivity analyses of DAFs and
maximum normalized torsional rotation, i.e. Φω and ω = θ/ pe
respectively, with respect to girder torsional stiffness parameter,
εω , are proposed. The comparisons reported in Fig. 14, denote
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
Fig. 16. Midspan torsional rotation dynamic impact factors (Φω ) vs bridge size
parameter (a) for AST-HST.
Fig. 17. Maximum displacement (ω) vs bridge size parameter (a) for AST-
HST.
that DAFs for AST are generally greater than the corresponding
ones obtained for bridges based on HST. This behaviour can
be explained in view of the enhanced stiffness properties of
AST, which although reducing bridge deformations means an
intensification of the dynamic impact factors. Moreover, for
increasing values of εω , which basically correspond to girder-
dominated bridge structures, the dynamic impact factors are
practically unaffected by the shape of the towers.
In Fig. 15, comparisons in terms of normalized torsional
rotations for both “A” and “H” shaped tower typologies have
been reported. The results show a tendency to decrease for
increasing values of girder stiffness. The dynamic solution
1175
appears to be strictly dependent on the tower shape and
significant differences between the AST and HST typologies
are noted.
Similar results have been proposed in Figs. 16 and 17,
in which sensitivity analyses have been developed in terms
of the bridge size parameter a. In particular, in Fig. 16, the
effects of moving mass description have been investigated,
emphasizing the influence of the inertial forces of the moving
system on the DAFs. The comparisons denote that with respect
to the HST, the AST are subject to major amplification effects.
Moreover, DAFs are strictly dependent on inertial forces of
the moving system, because solutions arising from the MFM
greatly underestimate the dynamic response of the bridge for
both AST and HST. Finally, in Fig. 17, normalized torsional
rotations have been compared in terms of pylon typology. The
results show a tendency to increase for increasing values of the
bridge size parameter a, and as with previous results, the AST
with respect to the HST typology shows smaller displacements
in the whole investigation range.
5. Conclusions
Long span bridges under moving loads have been analysed
for both flexural and torsional deformation modes, in terms
of dynamic impact factors for typical kinematic and stress
variables of the bridge. The effects of the inertial description
of the moving system on the dynamic bridge behaviour have
been investigated, by means of a parametric study developed
in terms of both moving loads and bridge characteristics.
The influence of the inertial forces are considerable, while
those corresponding to non-standard contributions arising from
Coriolis and centripetal accelerations determine the major
amplifications, mainly at high speeds of the moving system.
The inertial effects of the moving system have been discussed
with respect to typical geometrical and stiffness parameters of
the bridge, emphasizing the amplification effects produced by
the inertial forces of the moving system. For eccentric loads,
sensitivity analyses have been developed in terms of dynamic
impact factors and maximum normalized displacements with
respect to both “A” and “H” shaped tower typologies. In
the framework of the “A” shaped tower typologies, the
coupling behaviour between torsional and transversal flexural
deformations has been discussed. In particular, the influence of
the transverse displacements has been investigated, by means
of sensitivity analyses. This establishes that, for an in-plane
loading condition, the effect of transverse deformability on
the dynamic behaviour of the bridge is practically negligible.
Moreover, numerical results have shown that, in comparisons
with the “H” shaped tower topology, the “A” shaped ones,
even if having greater dynamic amplification factors, are
characterized by enhanced stiffness properties, which are able
to efficiently reduce torsional bridge deformation.
The investigation is developed in terms of the main
dimensionless parameters related to both geometric and
stiffness properties of the bridge. As a result, a parametric
study may be useful in the design procedure since the dynamic
impact factors for typical deformation and stress variables can
be determined in advance.
1176 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177

Appendix A E S∗ A S  
qvL = v · sin3 α − (u L − u) · sin2 α cos α
H ∆
Eg
In order to simplify the presentation of the dynamic = (V · ϕ − (U L − U ) · ϕ1 ) (B.4)
equilibrium equations in dimensionless formulation, the σg
following relationships have been utilized: E ∗ AS  
qv R = S v · sin3 α − (u R + u) · sin2 α cos α

Lp
 H ∆
f 1 = H (τ θ − X 1 ) H X 1 + − τθ , (A.1) Eg
H = (V · ϕ − (U L + U ) · ϕ1 ) (B.5)
σg
  
Lp
ρ = λH x1 + L p − ct H (ct − x1 )


f 2 = δ (τ θ − X 1 ) H X 1 + − τθ
 H
 
Lp
= λH X 1 + − τ ϑ H (τ ϑ − X 1 )
 
Lp (B.6)
− δ X1 + − τ θ H (τ θ − X 1 ) , (A.2) H
H
Eg 1/2 λ
    
ϕ (X ) ρ̇ = δ X1 +
Lp
− τ ϑ H (τ ϑ − X 1 )
1 1 L µHσg ϑ  H 
* − ≤ X ≤ 0, Lp
1 + aX 1 + X2
2 H − H X1 + − τ ϑ δ (τ ϑ − X 1 ) . (B.7)
= 1 1 L (A.3) H
, 0 ≤ X ≤ ,
L
1 + a 2H

2
− X 1+ L
−X

2 H By substituting Eqs. (B.4)–(B.7) in Eq. (21), and taking into
2H account Eqs. (B.1)–(B.3), the following equation is obtained:
ϕ1 (X )
∂2V σg I I V
1 X L − − V − H (X ) (V · ϕ − (U L + U ) · ϕ1 )
*
2 2
, − ≤ X ≤ 0, ∂τ 2 gH3
1 + aX 1 + X H
= L
L (A.4) − H (−X ) (V · ϕ − (U L − U ) · ϕ1 )


1 2H − X

2
2 , 0 ≤ X ≤ , λ
 
Lp

L
1 + a 2H − X 1 + 2H L
−X H − δ X1 + − τ ϑ H (τ ϑ − X 1 )
µϑ H
ϕ2 (X )   
Lp
1 X2 L − H X1 + − τ ϑ δ (τ ϑ − X 1 ) V̇
* , − ≤ X ≤ 0, H
1 + aX 1 + X2
2

2 H λ

Lp

= L (A.5) − H X1 + − τ ϑ H (τ ϑ − X 1 )
1 2H − X L
µϑ

2
2 , 0 ≤ X ≤ , H
L
1 + a 2H − X 1 + 2HL
−X H
∂ V
 2 
× + 2ϑ 0
V̇ + ϑ 2 00
V + p f1
with X 1 = x1 /H . ∂τ 2
σg
 
+ δ (X + L/2H ) + δ (x − L/2H ) λv
 
· V = 0,
Appendix B Eg
(B.8)
Starting from Eqs. (28)–(31), the following expressions can and taking into account Eqs. (28)–(31) and Eqs. (A.1) and
be determined: (A.2), Eq. (32) is finally determined.
∂v ∂ V ∂τ Eg 1/2
   
=H = H V̇ , References
∂t ∂τ ∂t µH σg
(B.1) [1] Fryba L. Vibration of solids and structures under moving loads. London:
∂ 2v ∂2V
 
Eg Thomas Telford; 1999.
=H 2 , [2] Timoshenko SP, Young DH. Theory of structures. New York: McGraw-
∂t 2 ∂τ µH σg
Hill; 1965.
∂v ∂V ∂ X ∂ 2v 1 ∂2V [3] Yang YB, Liao SS, Lin BH. Impact formulas for vehicles moving over
 
1
=H = V 0, = = V 00 . . . simple and continuous beams. J Struct Eng 1995;121(11):1644–50.
∂x ∂ X ∂x ∂x 2 H ∂ X2 H [4] Lei X, Noda NA. Analyses of dynamic response of vehicle and track
(B.2)
∂ 4v 1 ∂4V 1 IV coupling system with random irregularity of track vertical profile. J Sound
= 3 = 3V Vibration 2002;258(1):147–65.
∂x4 H ∂ X4 H [5] Roeder CW, Barth KE, Bergman A. Effect of live-load deflections on steel
bridge performance. J Bridge Eng 2004;9(3):259–67.
∂ 2v Eg 1/2 Eg H 1/2
   
= V̇ 0 , c= ϑ. (B.3) [6] Warburton GB. The dynamical behavior of structures. Oxford: Pergamon;
∂t∂ x µH σg µσg 1976.
[7] Wiriyachai A, Chu KH, Garg VK. Bridge impact due to wheel and track
Moreover, in view of Eqs. (B.1) and (B.3) and Eqs. irregularities. J Eng Mech Div 1982;108:648–65.
(28)–(31), the interaction forces between the cable system [8] Au FTK, Wang JJ, Cheung YK. Impact study of cable-stayed bridge under
railway traffic using various models. J Sound Vibration 2001;240(3):
and the girder (qvL , qv R ) and the mass function of the
447–65.
moving system (ρ) can be expressed by the following [9] Au FTK, Wang JJ, Cheung YK. Impact study of cable-stayed railway
relationships: bridges with random rail irregularities. Eng Struct 2001;24(5):529–41.
 D. Bruno et al. / Engineering Structures 30 (2008) 1160–1177
[10] Yang F, Fonder GA. Dynamic response of cable-stayed bridges under
moving loads. J Eng Mech 1998;124(7):741–7.
[11] Yau JD, Yang YB. Vibration reduction for cable-stayed traveled by high- [17]
speed trains. Finite Element Anal Design 2004;40:341–59.
[12] Huang D, Wang TL. Impact analysis of cable stayed bridges. Comput [18]
Struct 1992;43(5):897–908.
[13] Meisenholder SG, Weidlinger P. Dynamic interaction aspects of cable- [19]
stayed guide ways for high speed ground Transportation. J Dyn Syst Meas
Control ASME 1974;74-Aut-R:180–92. [20]
[14] Chatterjee PK, Datta TK, Surana CS. Vibration of cable-
stayed bridges under moving vehicles. Struct Eng Int 1994;4(2):
116–121. [21]
[15] Bruno D, Leonardi A. Natural periods of long-span cable-stayed bridges.
J Bridge Eng 1997;2(3):105–15.
[16] De Miranda F, Grimaldi A, Maceri F, Como M. Basic problems in long [22]
1177
span cable stayed bridges. Internal Report no 25. Department of Structural
Engineering, University of Calabria; 1979.
Troitsky MS. Cable stayed bridges. London: Crosby Lockwood Staples;
1977.
Gimsing NJ. Cable supported bridges: Concepts and design. John Wiley
& Sons Ltd; 1997.
Xia H, Xu YL, Chan THT. Dynamic interaction of long suspension
bridges with running trains. J Sound Vibration 2000;237(2):263–80.
Kawashima K, Unjoh S, Tsunomoto M. Estimation of camping ratio of
cable-stayed bridges for seismic design. J Struct Eng, ASCE 1993;119(4):
1015–31.
Yamaguchi Hiroki, Ito Manabu. Mode-dependence of structural damping
in cable-stayed bridges. J Wind Eng Industrial Aerodynam 1997;72(1–3):
289–300.
MAPLE Maplesoft. Waterloo Maple Inc. 2006.