You are on page 1of 11

Computers and Structures 83 (2005) 2271–2281

www.elsevier.com/locate/compstruc

A substructure approach for the dynamic analysis


of train–track–bridge system
B. Biondi a, G. Muscolino b,*
, A. Sofi b

a
Dipartimento di Ingegneria Civile ed Ambientale, Università di Catania, v.le A. Doria 6, I-95100 Catania, Italy
b
Dipartimento di Ingegneria Civile, Università di Messina, Villaggio S. Agata, I-98166 Messina, Italy

Received 26 November 2003; accepted 17 March 2005


Available online 24 August 2005

Abstract

The dynamic interaction between a running train, the track structure and the supporting bridge is investigated
resorting to substructure technique. The train is idealised as a sequence of identical vehicles moving at constant speed.
Both the rails and the bridge are modelled as Bernoulli–Euler beams, while the ballast is characterised as a viscoelastic
foundation. A variant of the component-mode synthesis method is proposed to couple the continuous (rails and bridge)
and discrete (train) substructures. Numerical investigations have demonstrated the accuracy and efficiency of the pro-
cedure, which allows to compute simultaneously the dynamic responses of the train, rails and bridge.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Railway bridge; Running train; Railway track; Dynamic interaction; Substructures; Component-mode synthesis

1. Introduction derived closed-form solutions for the dynamic response


of simple beams under high-speed trains. To account
The dynamic behaviour of railway bridges has been a for the inertia of the moving vehicles as well, the moving
topic of great interest in the field of civil engineering mass model (see e.g., [5,6]) has also been extensively
since the nineteenth century. A comprehensive review used.
of the history and literature on this subject can be found Enhanced by the construction of an increasing num-
in Ref. [1]. In early studies, the bridge has been modelled ber of high-speed railway bridges, recent researches have
as a beam-like structure and the train as a series of mov- focused on the problem of bridge–vehicle interaction,
ing loads (see e.g., [2,3]). The main advantage of such a which deserves special attention by designers. Mean-
simplified model is that it allows to use analytical meth- while, the necessity of more sophisticated vehicle models
ods for analyzing bridge vibrations. By modelling the for assessing the riding comfort has been recognised.
train as the composition of two subsystems of wheel The analysis of bridge–vehicle interaction basically re-
loads of constant intervals, recently Yang et al. [4] quires to write two sets of motion equations, one for
the bridge and the other for each vehicle. Due to the
interaction forces existing at the contact points, the
*
Corresponding author. Tel.: +39 090 3977159; fax: +39 090 two sets of equations are coupled and characterised by
3977480. time-dependent coefficients. This interaction problem is
E-mail address: muscolin@ingegneria.unime.it (G. Musco- usually solved by adopting iterative procedures [7,8],
lino). whose efficiency, however, worsens when dealing with

0045-7949/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2005.03.036
2272 B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281

the more realistic case of bridges travelled by a large characterised as a couple of infinite rails supported by
number of vehicles. In the literature, several approaches a continuous and homogeneous viscoelastic foundation
based on the condensation method are also available. (rail bed). Both the rails and the bridge are modelled
Yang and Lin [9] proposed a substructuring procedure as linear elastic Bernoulli–Euler beams. A particular var-
based on the dynamic condensation method to eliminate iant of the traditional component-mode synthesis (CMS)
all the vehicle degrees-of-freedom (DOFs) on the ele- method [16–19] is proposed to couple the continuous
ment level. Such a procedure has also been used in (rails and bridge deck) and discrete (train) substructures.
Ref. [10] to investigate the impact effects caused by vehi- Such variant allows to condense the axle DOFs into
cles moving over simple and continuous beams. In those of the rails in contact, so that the motion of the
Ref. [11] an element for modelling the vehicle-bridge coupled system is ruled by a set of ordinary differential
interaction in the analysis of railway bridges carrying equations with time-dependent coefficients in the gener-
high-speed trains has been developed. Such interaction alised coordinates of the bridge and rails, and physical
element consists of a beam element and of the masses displacements of the railway vehicles. One of the main
and suspension units of the car-body directly acting on features of the present approach is the capability of anal-
it. The same element has been adopted in Ref. [12] ysing simultaneously the dynamic responses of the train,
for studying the vibration of simple and three-span track structure and bridge, taking into account the inter-
continuous beams travelled by trains moving at high action among the three subsystems.
speeds. The influence of rail irregularity, ballast stiffness, For comparison purpose, the single-span simply sup-
suspension stiffness and suspension damping on the rid- ported prestressed concrete bridge analysed in Ref. [13]
ing comfort of rail cars has been investigated. Both the is taken as case study. Numerical investigations have
formulations presented in Refs. [11,12] account for the demonstrated that the proposed substructure technique
ballast stiffness by introducing continuously distributed provides results in good agreement with those obtained
springs, while the flexural stiffness of the rails is ne- by a finite element model, despite the number of un-
glected. As a consequence, the interaction element knowns involved is substantially reduced.
used in these works does not allow to analyse simulta-
neously the dynamic response of the railway track,
which is of great theoretical and practical importance. 2. Modelling of the train–track–bridge system
In most of the studies dealing with this issue, the rail- by substructure approach
way track is usually modelled either as a beam on Win-
kler elastic foundation or a beam supported on a series Consider a single-span simply supported bridge trav-
of discrete springs and dampers. In Ref. [13] a new ele- elled by a train moving at constant speed v (Fig. 1).
ment, called bridge–track–vehicle element, has been pro- The running train is conceived as composed of a
posed for investigating the interactions among a moving number Nv of identical vehicles, each of which comprises
train and its supporting railway track structure and a car-body, two bogies and nðaÞ v axles, as shown in Fig. 1,
bridge structure. Such element consists of vehicles mod- where without loss of generality it is assumed nðaÞ v ¼ 4.
elled as mass-spring-damper systems, an upper beam ele- The connections between the car-body and a bogie, as
ment to model the rails and a lower beam element to well as between a bogie and a wheel-set are characterised
model the bridge deck. The two beam elements are inter- as linear springs and viscous dashpots, whose stiffness
connected by a series of springs and dampers to model and damping coefficients are kb, cb and kw, cw, respec-
the rail bed. The procedure has been extended in tively. The car-body, bogies and wheel-sets in each rail-
Ref. [14] to deal with the steady-state response of a way vehicle are regarded as rigid components. Under
train moving over a series of simply supported railway this assumption, the jth vehicle is characterised as a dis-
bridges. crete system with nv ¼ nðiÞ v þ nv
ðaÞ
DOFs, being nðiÞ v and
In spite of their high accuracy and versatility, the nðaÞ
v the internal and axle DOFs, respectively. The inter-
above described finite element models may be time con- nal DOFs are represented by the vertical and pitch dis-
suming due to the large number of unknowns involved. placements of the car-body (vj,1, uj,1) and bogies
In this paper, the problem of dynamic interaction be- (vj,2, uj,2; vj,3, uj,3), while the axle DOFs are vertical dis-
ðaÞ
tween the train and the railway bridge is tackled by an placements denoted by vj;k ðk ¼ 1; . . . ; nðaÞ ðaÞ
v ; nv ¼ 4Þ,
alternative approach, based on substructure technique. (see Fig. 1).
The proposed procedure may be regarded as an appro- The bridge structure is modelled as a uniform Ber-
priate extension of the formulation developed by the noulli–Euler beam of length lb resting on simple end sup-
authors [15] for railway suspension bridges, and basi- ports. The track system lying on the bridge is idealised as
cally it consists in treating the train, rails and bridge a couple of infinite rails supported by a continuous and
deck as three substructures. The running train is idea- homogeneous viscoelastic foundation (rail bed). Since a
lised as a sequence of identical vehicles moving at con- plane model is adopted in the present analysis, the two
stant speed. The track system lying on the bridge is rails are treated as one, which is in turn modelled as a
B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281 2273

j − th vehicle

ϕ j,1
vj,1
kb cb ϕ j,2 ϕ j,3
kw cw
(a)
v j,1 vj,2 v (a) (a)
v j,3 vj,3 v (a) rail
−∞ j,2 j,4 +∞

rail bed
bridge
x
y
lb
Fig. 1. A single-span simply supported bridge under a running train and railway vehicle model.

 
uniform Bernoulli–Euler beam of infinite length. Fur- our ðx; tÞ oub ðx; tÞ
rb ðx; tÞ ¼ k f ½ur ðx; tÞ  ub ðx; tÞ þ cf  ;
thermore, the influence of rail irregularities is neglected ot ot
and all wheels are assumed to contact rigidly and contin- ð2Þ
uously as they roll over.
Following an approach recently proposed by the where ur(x, t) denotes the vertical displacement of the
authors for analysing the dynamic response of railway rail, while kf and cf are the stiffness and damping coeffi-
suspension bridges [15], the running train, rails and cients of the rail bed, respectively.
bridge deck are herein regarded as three substructures. The infinite rail resting on a viscoelastic foundation
In order to couple the continuous (rails and bridge deck) can be conveniently treated as an equivalent finite beam
and discrete (train) subsystems, a variant of the tradi- of length lr = lb + 2l0 (see Fig. 2a), provided that l0 sat-
tional component-mode synthesis (CMS) method is pro- isfies the following condition [20]:
posed. Through an appropriate choice of the boundary sffiffiffiffiffiffiffiffiffiffiffi
4 4E r I r
interface conditions, the proposed CMS approach en- l0 P 2p ; ð3Þ
ables the condensation of the axle DOFs into those of kf
the rails in contact, properly accounting for the interac- where Er and Ir are the rail modulus of elasticity and
tion effects among the three subsystems. According to moment of inertia, respectively.
standard substructure technique, first the equations of Indeed, if Eq. (3) holds, then the rail elements of
motion of the substructures separately taken must be de- length l0 on the right- and left-hand sides of the bridge
rived, as outlined in the next section. behave like semi-infinite beams. Furthermore, it can be
verified that, when the train runs over the bridge, the rail
deflections and rotations at the abscissas x = l0 and
3. Equations of motion of the substructures x = lb + l0 are nearly zero. This implies that the finite
rail can be reasonably assumed clamped at both ends.
3.1. Rails and bridge Under the previous assumptions, the equation of
motion of the rail crossed by a train running at constant
Taking into account the above defined model, the speed v can be written as follows:
motion of the bridge structure is ruled by the following
fourth-order partial differential equation: o2 ur ðx; tÞ our ðx; tÞ o4 ur ðx; tÞ
qr Ar 2
þ cr Ar þ Er I r
o2 ub ðx; tÞ oub ðx; tÞ o4 ub ðx; tÞ ot ot ox4
qb Ab þ cb Ab þ Eb I b ¼ rb ðx; tÞ; ðaÞ
 
ot 2 ot ox4 Nv X
X nv
ðaÞ ðaÞ
ð1Þ ¼ rr ðx; tÞ þ fr;jk ðtÞvjk ðtÞd x  xr;jk ðtÞ ; ð4Þ
j¼1 k¼1
where t and x denote the time and the coordinate mea-
where qr, cr and Ar are the rail mass density, damping
sured along the axis of the bridge, respectively; ub(x, t) ðaÞ
is the vertical displacement of the bridge; qb, Eb, cb, coefficient and cross-sectional area, respectively; fr;jk ðtÞ
Ab and Ib are the bridge mass density, modulus of elas- is the force transmitted by the kth axle of the jth vehicle
ðaÞ
ticity, damping coefficient, cross-sectional area and at the position xr;jk ðtÞ, given by
moment of inertia, respectively; rb(x, t) is the reaction ðaÞ
exerted by the viscoelastic foundation, given by xr;jk ðtÞ ¼ vt  l0  d jk ð5Þ
2274 B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281

rail

rail bed bridge x


y
(a) l0 lb l0

(b)
+

(c) l0 lb l0

Fig. 2. Coupled track-bridge system (a); finite rail on viscoelastic foundation fixed at the interface with the bridge (b); single-span
bridge with free interface (c).

being djk the distance between the kth axle of the jth of the CMS method considers the simply supported
vehicle and the first (d11 = 0). According to the previous beam, modelling the bridge structure, as free (see
relationship, when t = 0 the first axle of the first vehicle Fig. 2c), while the track structure is supposed fixed at
ðaÞ
is located at the abscissa xr;11 ð0Þ ¼ l0 . Furthermore, in the interface with the bridge (see Fig. 2b).
Eq. (4) d(Æ) is the Dirac delta function; vjk(t) is the Under the previous assumptions, the bridge deflec-
so-called window function, defined as tion ub(x, t) can be approximated by the following series
( ðaÞ expansion:
1 for  l0 6 xr;jk ðtÞ 6 lb þ l0 ;
vjk ðtÞ ¼ ðaÞ ðaÞ
X
mb
0 for xr;jk ðtÞ < l0 or xr;jk ðtÞ > lb þ l0 ; ub ðx; tÞ ¼ /b;i ðxÞqb;i ðtÞ ¼ /Tb ðxÞqb ðtÞ; ð8Þ
i¼1
ð6Þ
where /b(x) and qb(t) are the vectors collecting the first
and rr(x, t) denotes the reaction transmitted by the visco-
mb eigenfunctions of the simply supported beam, /b,i
elastic foundation to the rail, given by
(x), and the associated generalised coordinates, qb,i(t),
rr ðx; tÞ ¼ k f ½ur ðx; tÞ  ub ðx; tÞðU ðxÞ  Uðx  lb ÞÞ respectively. As usual, the eigenfunctions /b,i(x) are nor-
  malised with respect to the mass of the bridge.
our ðx; tÞ oub ðx; tÞ
þ cf  ðU ðxÞ  U ðx  lb ÞÞ ; On the other hand, the rail deflection ur(x, t) can be
ot ot
expressed as follows:
ð7Þ
ur ðx; tÞ þ uðbÞ
ur ðx; tÞ ¼ ~ r ðx; tÞ; ð9Þ
where U(x) is the unit-step function. Notice that for
x < 0 and x > lb the reaction rr(x, t) is proportional to where ~ ur ðx; tÞ denotes the relative displacement of the
the displacement and velocity of the rail, ur(x, t) and rail with respect to the bridge, whereas uðbÞ r ðx; tÞ repre-
our(x,t)/ot, since the terms relating to the bridge are set sents the contribution due to the motion of the bridge
equal to zero. Indeed, the rail outside the bridge is char- itself. The relative displacement, ~ ur ðx; tÞ, is in turn
acterised as a beam on viscoelastic foundation. approximated by the following series expansion:
In order to apply the CMS method, each of the above X
mr

defined partial differential equations must be reduced to ur ðx; tÞ ¼


~ /r;i ðxÞqr;i ðtÞ ¼ /Tr ðxÞqr ðtÞ; ð10Þ
i¼1
a set of ordinary differential equations by approximating
the rail and bridge displacements in terms of generalised where /r(x) and qr(t) are the vectors collecting the first
coordinates and suitably selected eigenfunctions. In par- mr eigenfunctions of the rail, /r,i(x), and the associated
ticular, the choice of the eigenfunctions is related to the generalised coordinates, qr,i(t), respectively. Taking into
selection of the boundary interface conditions for each account the selected boundary interface conditions (see
subsystem. As outlined in a recent work [19], the interac- Fig. 2b), the rail eigenfunctions, /r,i(x), are obtained
tion effects between a discretized and a continuous sub- as solution of the eigenproblem deduced from Eq. (4)
structure cannot be accurately described by assuming by setting ub (x, t) = oub(x, t)/ot = 0 in the expression
free boundary interface conditions for both the subsys- of rr (x, t) (see Eq. (7)), and dropping the damping terms
tems. In view of this observation, the proposed variant and the forces transmitted by the axles, i.e.,
B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281 2275
  ðaÞ ðaÞ
Er I r /IV 2
r ðxÞ þ k f  xr qr Ar /r ðxÞ ¼ 0; ð11Þ The matrices Wr;j ðtÞ and Ur;j ðtÞ of order ðmr  nðaÞ v Þ, the
diagonal matrix Xj(t) of order nðaÞ v and the ðnðaÞ
v  1Þ-vec-
where xr denotes the generic natural frequency of the ðaÞ
tor f r;j ðtÞ, appearing on the right-hand side of Eqs. (15)
rail. The eigenfunctions /r,i(x) satisfy the following and (16), are given by
conditions:
Z lb þl0 h      iT
ðaÞ ðaÞ ðaÞ ðaÞ
Wr;j ðtÞ ¼ wr xr;j1 ðtÞ ; wr xr;j2 ðtÞ ; . . . ; wr x ðaÞ ðtÞ
qr Ar /r ðxÞ/Tr ðxÞ dx ¼ Imr ; r;jnv
l0 h      iT
Z lb þl0
ðaÞ
Ur;j ðtÞ
ðaÞ ðaÞ ðaÞ
¼ /r xr;j1 ðtÞ ; /r xr;j2 ðtÞ ; . . . ; /r x ðaÞ ðtÞ
T kf
Er I r /r ðxÞ/IV 2
r ðxÞ dx ¼ Xr  Im ; ð12Þ h i
r;jnv
l0 qr Ar r
Xj ðtÞ ¼ Diag vj1 ðtÞ; vj2 ðtÞ; . . . ; vjnðaÞ ðtÞ
v
where Imr is the identity matrix of order mr and Xr is a n oT
diagonal matrix listing the first mr natural frequencies ðaÞ ðaÞ ðaÞ
f r;j ðtÞ ¼ fr;j1 ðtÞ; fr;j2 ðtÞ; . . . ; f
ðaÞ
ðtÞ :
ðaÞ
r;jnv
of the rail.
The second contribution appearing in Eq. (9), ð17Þ
uðbÞ
r ðx; tÞ, is approximated as follows:
Notice that the diagonal matrix Xj(t) allows to take into
X
mb
account at each time instant only the contribution of the
uðbÞ
r ðx; tÞ ¼ wr;i ðxÞqb;i ðtÞ ¼ wTr ðxÞqb ðtÞ; ð13Þ
i¼1 wheel loads acting on the rail, according to the definition
of window function vjk(t) given in Eq. (6). The remaining
where wr,i(x) are appropriate shape functions of the rail, matrices introduced in Eqs. (15) and (16) are defined in
which may be regarded as the counterpart of the well- Appendix A.
known pseudo-static influence functions. Specifically, The sets of ordinary differential equations (15) and
the shape functions wr,i (x) are evaluated as solution of (16) can be rewritten in the following compact form:
the following fourth-order partial differential equations:
qB ðtÞ þ CB q_ B ðtÞ þ KB qB ðtÞ ¼ f B ðtÞ;
MB € ð18Þ
Er I r wIV
r;i ðxÞ þ k f wr;i ðxÞ ¼ k f /b;i ðxÞðU ðxÞ  U ðx  lb ÞÞ;

ði ¼ 1; 2; . . . ; mb Þ; ð14Þ where
   
which are derived from Eq. (4) by setting ur(x, t) = Imb þ DMb Mbr Nb þ DCb Cbr
MB ¼ ; CB ¼ ;
wr,i(x)qb,i(t) and ub(x, t) = /b,i(x)qb,i(t) (qb,j(t) = 0 for Mrb Im r Crb Nr
j 5 i), once the damping and contact forces have been " # 
2
Xb þ DKb 0 qb ðtÞ
dropped. KB ¼ ; qB ðtÞ ¼ ;
2
Let us substitute the approximate bridge and rail dis- 0 Xr qr ðtÞ
( ðaÞ )
placements (8) and (9) into the equations of motion (1) XNv
Wr;j ðtÞ ðaÞ
and (4). Pre-multiply both sides of Eq. (1) by /b(x) f B ðtÞ ¼ ðaÞ
Xj ðtÞf r;j ðtÞ. ð19Þ
and integrate over the span length lb of the bridge. Then, j¼1 Ur;j ðtÞ
pre-multiply both sides of Eq. (4) by wr(x) and integrate
over the rail length lr = lb + 2l0. By adding both sides of 3.2. Running train
the resulting equations, the following set of ordinary dif-
ferential equations is obtained: As outlined in Section 2, the running train model
ðImb þ DMb Þ€qb ðtÞ þ Mbr €qr ðtÞ þ ðNb þ DCb Þq_ b ðtÞ consists of a number Nv of nv ¼ nðiÞ ðaÞ
v þ nv -DOFs vehicles
with the same characteristics. By assuming that the
þ Cbr q_ r ðtÞ þ ðX2b þ DKb Þqb ðtÞ vibration amplitude of each component is small, the
XNv
ðaÞ ðaÞ
equations of motion of the jth railway vehicle can be
¼ Wr;j ðtÞXj ðtÞf r;j ðtÞ; ð15Þ written as follows:
j¼1

Mv€vv;j ðtÞ þ Cv v_ v;j ðtÞ þ Kv vv;j ðtÞ ¼ f v;j ðtÞ;


where the dot over a variable denotes total derivative
with respect to time t and Xb is a diagonal matrix listing ðj ¼ 1; 2; . . . ; N v Þ; ð20Þ
the first mb natural frequencies of the bridge. Further-
more, pre-multiplying both sides of Eq. (4) by /r(x) where Mv, Cv and Kv are the (nv · nv) mass, damping
and integrating over the rail length lr = lb + 2l0, the fol- and stiffness matrices of each vehicle; vv,j(t) and fv,j(t)
lowing set of ordinary differential equations is derived: are the (nv · 1)-vectors listing the displacements and
loads pertaining to the jth rail car.
qr ðtÞ þ Mrb €qb ðtÞ þ Nr q_ r ðtÞ þ Crb q_ b ðtÞ þ X2r qr ðtÞ
€ Collecting the displacements associated with the
ðiÞ
X
Nv internal and axle DOFs into the subvectors vv;j ðtÞ and
ðaÞ ðaÞ ðaÞ
¼ Ur;j ðtÞXj ðtÞf r;j ðtÞ: ð16Þ vv;j ðtÞ, respectively, Eq. (20) can be rewritten in the fol-
j¼1 lowing partitioned form:
2276 B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281

" #8 ðiÞ 9 2 38 ðiÞ 9


the equations of motion of the two substructures by
MðiÞ
v 0 < €vv;j ðtÞ = v CðiÞ
CðiaÞ
v
< v_ v;j ðtÞ =
þ4 5 applying the component-mode coordinate transforma-
0 0 : €vðaÞ
v;j ðtÞ
;
CðaiÞ
v CðaÞ
v
: ðaÞ ;
v_ v;j ðtÞ tion (which accounts for the compatibility conditions)
2 38 ðiÞ 9 8 ðiÞ 9 and imposing the equilibrium conditions at the common
KðiÞ
v KðiaÞ
v
< vv;j ðtÞ = < Mv sðiÞ v g
= interface. Relying on this approach, in the present paper
þ4 5 ¼ ; a variant of the CMS method is developed to analyse the
: ðaÞ ; : f ðaÞ ðtÞ ;
KðaiÞ
v KðaÞ
v vv;j ðtÞ v;j coupled train–track–bridge system, which is composed
ð21Þ of two continuous substructures (rails and bridge) and
a discrete subsystem (running train). For this purpose,
ðiÞ
where MðiÞ ðiÞ ðiÞ ðiÞ
v ; Cv and Kv are the ðnv  nv Þ mass, damp- in the previous section, the equations of motion of the
ing and stiffness submatrices associated with the internal rails and bridge have been written in a reduced general-
DOFs; CðaÞ v and KðaÞ
v are the ðnðaÞ ðaÞ
v  nv Þ damping and ised space by using suitable eigenfunctions and shape
stiffness submatrices of the axles, whose mass is assumed functions, defined according to the selected boundary
negligible; g denotes the acceleration of gravity; sðiÞ
v is a interface conditions. The next step of the proposed
ðnðiÞ
v  1Þ-vector whose entries are all zero except those CMS approach is to write the equations of motion of
corresponding to the vertical displacements of the car- the three substructures, (Eqs. (18) and (22)), in compact
ðaÞ
body and bogies which are 1; finally, f v;j ðtÞ denotes the form, as follows:
ðaÞ
ðnv  1Þ-vector listing the forces due to the contact be-
ðaÞ
tween wheels and rails, fv;jk ðtÞ. y ðtÞ þ C y_  ðtÞ þ K y ðtÞ ¼ f  ðtÞ;
M € ð24Þ
The set of ordinary differential equations ruling the where
motion of the running train can be easily obtained by
assembling the equations of motion (20) of all the Nv M ¼ Diag½MB ; MT ; C ¼ Diag½CB ; CT ;
railway vehicles K ¼ Diag½KB ; KT ;
MT€vT ðtÞ þ CT v_ T ðtÞ þ KT vT ðtÞ ¼ f T ðtÞ; ð22Þ y ðtÞ ¼ fqB ðtÞ; vT ðtÞgT ; f  ðtÞ ¼ ff B ðtÞ; f T ðtÞgT : ð25Þ

where The equations of motion of the coupled train–track–


T bridge system can be deduced from Eq. (24) by imposing
vT ðtÞ ¼ vv;1 ðtÞ; . . . ; vv;N v ðtÞ ; the following compatibility and equilibrium conditions
T
f T ðtÞ ¼ f v;1 ðtÞ; . . . ; f v;N v ðtÞ ; ð23Þ at the contact points between wheels and rails
h i
ðaÞ ðaÞT ðaÞT
while MT, CT and KT are diagonal super-matrices of order vv;j ðtÞ ¼ Xj ðtÞ Ur;j ðtÞqr ðtÞ þ Wr;j ðtÞqb ðtÞ ;
ðiÞ ðaÞ ðiÞ ðaÞ
nT ¼ nT þ nT ¼ nv  N v ; ðnT ¼ nðiÞ
v  Nv and nT ¼ ðaÞ ðaÞ
ðaÞ
nv  N v Þ, collecting the Nv mass, damping and stiffness f r;j ðtÞ þ f v;j ðtÞ ¼ 0; ðj ¼ 1; 2; . . . ; N v Þ: ð26a; bÞ
matrices, Mv, Cv and Kv, respectively. The compatibility conditions (26a) impose that the axle
ðaÞ
displacements of the jth vehicle, vv;j ðtÞ, coincide with the
ðaÞ
rail deflections, ur ðxr;jk ðtÞ; tÞ, (see Eqs. (9), (10) and (13))
4. Motion equations of the coupled train–track–bridge ðaÞ
at the abscissas xr;jk ðtÞ; ðk ¼ 1; . . . ; nðaÞ
v Þ, defining the
system by CMS method
instantaneous positions of the axles. The window func-
tions vjk(t), collected into the diagonal matrix Xj(t),
In this section, the equations of motion of the rails,
assure that the displacements of those axles which have
bridge and running train, (18) and (22), are coupled by
not yet entered the finite rail or have left it are automat-
applying the CMS method. This procedure has been
ically set equal to zero. Eq. (26b) expresses the equilib-
extensively used to reduce the number of generalised
rium of the contact forces between the Nv vehicles and
coordinates in the dynamic analysis of coupled struc-
the rails.
tural systems composed of two or more discretized sub-
Taking into account the compatibility conditions
structures [16–19]. Recently, an efficient variant of the
(26a), the following component-mode coordinate trans-
CMS method has been proposed to analyse structural
formation can be written
systems modelled as the combination of a continuous
substructure and a finite element (FE) discretized one y ðtÞ ¼ CðtÞqðtÞ; ð27Þ
[19]. Basically, this variant requires the following steps: ðiÞ
(i) to write the equations of motion of the discretized where q(t) is the vector of order ðmb þ mr þ  1Þ nT
and continuous substructures in the nodal and general- listing the generalised coordinates of the bridge and
ised space, respectively; (ii) to define the component- rails, and the internal displacements of the Nv railway
ðiÞ
mode coordinate transformation consistent with the vehicles, vv;j ðtÞ; ðj ¼ 1; 2; . . . ; N v Þ, collected into the
ðiÞ ðiÞ
selected boundary interface conditions; (iii) to couple ðnT  1Þ-vector vT ðtÞ
B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281 2277

8 9 8 ðiÞ 9
( ) > v ðtÞ > mark-b method with constant average acceleration, i.e.,
qB ðtÞ < qb ðtÞ >
> = >
< v;1 >
= with b = 1/4 and c = 1/2 [21].
qðtÞ ¼ ¼ qr ðtÞ ; ðiÞ
vT ðtÞ ¼ .. :
ðiÞ
vT ðtÞ >
: ðiÞ > ; >
> . > > It is worth noting that the solution of Eq. (30) may be
vT ðtÞ : ðiÞ ; time consuming when the running train consists of a
vv;N v ðtÞ
large number of vehicles. Nevertheless, the computa-
ð28Þ
tional efficiency of the present procedure can be greatly
The time-dependent transformation matrix C(t) in Eq. enhanced by taking into account that bridge vibrations
(27) is defined as follows: are practically unaffected by a train running over the
side rail elements of length l0. In view of this observa-
tion, the analysis can be carried out starting from the
time instant t0 = l0/v (when the first axle of the first vehi-
cle is located at the abscissa x = 0) and eliminating all
the items associated with those vehicles which have not
reached or have left the railway bridge. This implies
that, when the bridge response is of concern, a quite dif-
ferent definition of the window function given in Eq. (6)
can be conveniently adopted, i.e.,
( ðaÞ
1 for 0 6 xr;jk ðtÞ 6 lb ;
vjk ðtÞ ¼ ðaÞ ðaÞ
ð32Þ
0 for xr;jk ðtÞ < 0 or xr;jk ðtÞ > lb :
Since the span length of simply supported bridges is
ð29Þ usually small in comparison with the rail car length,
the previous relationship allows to reduce substantially
Substituting Eq. (27) into Eq. (24), pre-multiplying both the size of the matrices and vectors appearing in Eq.
sides by CT(t) and taking into account Eq. (17), the re- (24), and hence the order of the transformation matrix
duced equations of motion of the train–track–bridge C(t). Indeed, denoting by N v < N v the number of vehi-
system are obtained cles running over the railway bridge at the generic time
instant t, i.e., those for which Xj(t) 5 0, the number of
MðtÞ€qðtÞ þ CðtÞqðtÞ
_ þ KðtÞqðtÞ ¼ fðtÞ; ð30Þ
true unknowns in the equations of motion (30) is
ðiÞ ðiÞ ðiÞ
where mb þ mr þ  nT , where nT ¼ nðiÞ
v  N v < nT .

MðtÞ ¼ CT ðtÞM CðtÞ;


_
CðtÞ ¼ CT ðtÞ½C CðtÞ þ 2M CðtÞ; 5. Numerical application
_ €
KðtÞ ¼ C ðtÞ½K CðtÞ þ C CðtÞ þ M CðtÞ;
T 

fðtÞ ¼ CT ðtÞf  ðtÞ ð31Þ Consider a single-span simply supported prestressed


concrete railway bridge with span length lb = 20 m,
are the time-dependent mass, damping and stiffness Youngs modulus of concrete Eb = 29.43 GPa, mass
matrices, and the load vector of the coupled system. per unit length qbAb = 34,088 kg/m and moment of
Eq. (30) represents a set of ordinary differential equa- inertia Ib = 3.81 m4 [13]. Each rail is characterised by
tions with time-dependent coefficients, in terms of gener- the following parameters: flexural rigidity ErIr = 4.3 ·
alised coordinates of the bridge and rails, and internal 106 Nm2 and mass per unit length qrAr = 51.5 kg/m.
displacements of the railway vehicles. Notice that the The influence of damping is neglected for both the
proposed component-mode coordinate transformation bridge and rails. The stiffness and damping coefficients
(27) allows the condensation of the axle displacements of the rail bed for one rail are kf = 65,800 kN/m2 and
into those of the rails in contact, reducing the number cf = 32,100 Ns/m2, respectively. In the analysis, all the
ðiÞ
of unknowns from mb + mr + nT to mb þ mr þ nT ; rail properties have to be doubled to account for the
ðiÞ ðaÞ
ðnT ¼ nT  nT Þ. The present formulation thus enables presence of both rails. The length of the finite track
one to pursue simultaneously the dynamic responses of structure replacing the infinite one is assumed to be
the railway vehicles, rails and bridge deck by integrating lr = lb + 2l0 = 100 m, where l0 = 40 m is taken twice
the reduced equations of motion (30). As regard the ini- the span length lb (see Fig. 2a). The bridge is travelled
tial conditions, it is assumed that at the instant t = 0 by a train consisting of five identical vehicles (see
both the bridge and rails are at rest, and the first axle Fig. 3). The present problem was analysed by Cheng
ðaÞ
of the first vehicle is located at the abscissa xr;11 ð0Þ ¼ et al. [13] using a bridge–track–vehicle element to de-
l0 . The reduced equations of motion (30) can be solved scribe the dynamic interaction between the train, bridge
by using standard step-by-step algorithms, such as New- and rails. Specifically, they divided the bridge into 16
2278 B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281

(a) lv ld lv ld lv ld lv ld lv

mv v1
kv cv
(b)

mv,Jϕ ,v v1 ϕ 1
kv cv
(c)

Fig. 3. Moving train: general model (a); sprung mass model (b); 2-DOFs vehicle model (c).

bridge-track elements of equal length and adopted an 7.00


irregular mesh, consisting of 24 beam elements on visco- 6.00
elastic foundation, for each portion of track structure on Proposed
the two sides of the bridge. In order to compare the re- 5.00 Ref. [13]
sults obtained by the proposed substructure approach 4.00
with those reported in Ref. [13], the running train was Dub
first idealised as a series of sprung masses, mv = 3.00

22,000 kg, concentrated at the axle positions, with 2.00


suspension stiffness kv = 9.12 · 106 N/m and damping
1.00
coefficient of the viscous dashpot cv = 8.6 · 104 Ns/m,
as displayed in Fig. 3b. The arrangement of wheel 0.00
0.00 40.00 80.00 120.00 160.00 200.00 240.00
assemblies is shown in Fig. 3a, where ld = 6 m and (a) v [m/s]
lv = 18 m. The analysis was performed assuming a train
velocity v = 20 m/s and taking into account the first 7.00
eight eigenfunctions of both the bridge and rails, say set-
ting mb = mr = 8. The reduced equations of motion of 6.00
Proposed
the coupled system (30) were solved by Newmark-b 5.00 Ref. [13]
method (with b = 1/4 and c = 1/2) [21].
DMb

The dynamic effects induced by the moving train on 4.00

the railway bridge were investigated by computing the 3.00


dynamic magnification factor for the generic response
2.00
quantity R(x) at the abscissa x, defined as
Rd ðxÞ 1.00
DR ¼ ; ð33Þ
Rs ðxÞ 0.00
0.00 40.00 80.00 120.00 160.00 200.00 240.00
where Rd(x) and Rs(x) denote the maximum dynamic (b) v [m/s]
and static responses, respectively. In Fig. 4 the dynamic
magnification factors for mid-span displacement and Fig. 4. Dynamic magnification factors for bridge displacement
(a) and bending moment (b) at mid-span.
bending moment of the bridge deck, Dub and DM b ,
respectively, obtained by the present substructure model
are compared with those reported in Ref. [13]. Notice Appropriate comparisons with the results reported by
that, in spite of the lower computational burden the pro- Cheng et al. [13] further demonstrate the accuracy of
posed method provides results in good agreement with the present method.
those deduced by Cheng et al. [13]. In Fig. 5 the dynamic As outlined in Section 4, the vibration amplitude of
magnification factor for the displacement of the railway the track-bridge system is negligible as long as the train
track at bridge mid-span, Dur , against train speed is plot- has not entered the bridge. Hence, the time history of
ted. The results obtained neglecting the bridge structure each response quantity, measured within the span length
are also shown. It can be seen that the dynamic magni- lb, can be conveniently plotted starting from the time in-
fication factor for the coupled track-bridge system is stant in which the first vehicle reaches the bridge, say
much higher than that pertaining to the track structure assuming as time parameter t  t0 (t0 = l0/v). Fig. 6a dis-
modeled as a finite beam on viscoelastic foundation. plays the time histories of bridge and rail displacements
B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281 2279

3.00 4.00E-004
Bridge considered (proposed)
Bridge considered (Ref. [13])
2.40 Bridge ignored (proposed) 3.00E-004
Bridge ignored (Ref. [13])

u~r, MAX
2.00E-004
Dur

1.80

1.00E-004
1.20

0.00E+000
0.00E+000 1.00E+005 2.00E+005 3.00E+005 4.00E+005 5.00E+005
0.60
0.00 40.00 80.00 120.00 160.00 200.00 240.00 kf [kN/m2]
v [m/s]
Fig. 7. Maximum relative displacement of the rail at mid-span
Fig. 5. Dynamic magnification factor for the displacement of versus the stiffness coefficient of the rail bed kf.
the rail at mid-span considering and ignoring the bridge
structure.
cient of the viscoelastic foundation, kf, is plotted. It
can be observed that, as the ballast stiffness increases,
1.20E-003 the rail and bridge deck behave as a single beam and
consequently ~ur;MAX tends to zero. By contrast, both
Mid-span displacement [m]

Bridge
Rail the maximum displacement and bending moment of
8.00E-004
the bridge at mid-span, ub,MAX and Mb,MAX, increase
with the stiffness coefficient kf, approaching a constant
4.00E-004 value, as shown in Fig. 8. This may be explained taking
into account that a softer rail bed distributes the axle
0.00E+000 loads over a larger portion of the bridge deck, while
for higher values of kf the forces transmitted by the
-4.00E-004
0.00 2.00 4.00 6.00 8.00
(a) t-t0 [s] 9.00E-004

3.00E+003

8.90E-004
ub,MAX [m]
Mb(lb/2,t-t0) [kNm]

2.00E+003

8.80E-004
1.00E+003

0.00E+000 8.70E-004
0.00E+000 1.00E+005 2.00E+005 3.00E+005 4.00E+0055.00E+005
-1.00E+003 (a) kf [kN/m2]
0.00 2.00 4.00 6.00 8.00
(b) t-t0 [s]
2.46E+003

Fig. 6. Time history of bridge and rail displacements (a), and


2.44E+003
bridge bending moment (b) at mid-span.
Mb,MAX [kNm]

2.42E+003

at mid-span (v = 20 m/s). As expected, the track struc- 2.40E+003


ture vibrates together with the bridge, so that the rail
and bridge absolute displacements are nearly coincident. 2.38E+003
In Fig. 6b the time evolution of bending moment
Mb(x, t  t0) at mid-span of the bridge deck is plotted. 2.36E+003
0.00E+000 1.00E+005 2.00E+005 3.00E+005 4.00E+005 5.00E+005
The influence of rail bed stiffness on the dynamic re- kf [kN/m2]
(b)
sponses of the bridge, rails and vehicles was also investi-
gated. In Fig. 7 the maximum relative displacement of Fig. 8. Maximum bridge displacement (a) and bending moment
the rail at mid-span, ~ur;MAX , versus the stiffness coeffi- (b) at mid-span versus the stiffness coefficient of the rail bed kf.
2280 B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281

1.04E+000 1.20E-003
Train model in Fig. 3b
Train model in Fig. 3c
1.02E+000
8.00E-004

ub (lb/2, t-t0) [m]


v1,MAX [m/s2]

1.00E+000
4.00E-004
9.80E-001
..

9.60E-001 0.00E+000

9.40E-001 -4.00E-004
0.00E+000 1.00E+005 2.00E+005 3.00E+005 4.00E+005 5.00E+005 0.00 2.00 4.00 6.00 8.00
(a) kf [kN/m2] t-t0 [s]

1.20 Fig. 10. Time history of mid-span displacement of the bridge


Track structure ignored for two different train models.
1.00 3
kf = 5.0 ×10
v1,MAX [m/s2]

4
0.80 kf = 1.0 ×10
kf = 1.316 ×10
5 placement at mid-span is plotted. Fig. 11 displays the
0.60
time evolutions of the vertical displacement, v1(t), and
..

0.40 acceleration, €v1 ðtÞ, of the first rail car measured from
the static equilibrium position. For comparison purpose,
0.20
in both Figs. 10 and 11 the results obtained by adopting
0.00 the sprung mass model (Fig. 3b) are also plotted. It can
0.00 40.00 80.00 120.00 160.00 200.00 240.00 be seen that the bridge response can be accurately pre-
(b) v [m/s]
dicted by idealizing the running train as a series of
Fig. 9. Maximum acceleration of the first sprung mass versus lumped masses supported by spring-dashpot units. Con-
the stiffness coefficient of the rail bed kf (a) and train speed v (b). versely, such a model proves to be inadequate for assess-
ing the running safety of the train and riding comfort

railway vehicles become analogous to loads concen-


trated at the contact points between wheels and rails.
5.00E-003
With reference to the first sprung mass, Fig. 9a shows Train model in Fig. 3b
that a harder ballast reduces the maximum vertical 4.00E-003 Train model in Fig. 3c
acceleration of the rail car and, therefore, raises the rid-
3.00E-003
v1 (t-t0)[m]

ing comfort. In Fig. 9b the maximum vertical accelera-


tion of the first sprung mass, €v1;MAX , versus train speed 2.00E-003
is plotted for three different values of the stiffness coeffi-
1.00E-003
cient kf and for the case in which the track structure is
neglected. Owing to bridge–vehicle dynamic interaction, 0.00E+000
the train speed for which the largest resonant peak of
-1.00E-003

v1;MAX occurs is close to that observed in Figs. 4 and 5 0.00 0.50 1.00 1.50 2.00 2.50
for bridge and rail displacements, and bending moment (a) t-t0 [s]
of the bridge deck. Furthermore, it can be observed that,
according to the results displayed in Fig. 9a, a softer rail 2.00E-001

bed induces higher maximum accelerations of the vehicle


1.00E-001
and additionally it shifts the critical speed (the one asso-
v1 (t-t0)[m/s2]

ciated with the largest resonant peak) to smaller values.


0.00E+000
The present problem was reanalyzed by modelling
the running train as a series of 2-DOFs (vertical and -1.00E-001
pitch displacements) vehicles, each represented by a rigid
..

car-body supported by two spring-dashpot suspension -2.00E-001 Train model in Fig. 3b


systems (Fig. 3c). The parameters characterizing the Train model in Fig. 3c
wheel assemblies, ld and lv, as well as the spring and -3.00E-001
0.00 0.50 1.00 1.50 2.00 2.50
damping coefficients, kv and cv, are the same as those de- (b) t-t0 [s]
fined above. The mass and pitch mass moment of the
car-body are mv = 43,900 kg and Ju,v = 1690.2 · 103 kg Fig. 11. Time history of vertical displacement (a) and acceler-
m2, respectively. In Fig. 10 the time history of bridge dis- ation (b) of the first vehicle for two different train models.
B. Biondi et al. / Computers and Structures 83 (2005) 2271–2281 2281

since it underestimates the maximum vertical accelera- References


tion of the vehicle, as shown in Fig. 11b.
[1] Frýba L. Dynamics of railway bridges. London: Thomas
Telford; 1996.
6. Concluding remarks [2] Frýba L. Vibrations of solids and structures under moving
loads. London: Thomas Telford; 1999.
A substructure approach for analysing the vibration [3] Pesterev AV, Yang B, Bergman LA. Revisiting the moving
of railway bridges under moving trains taking into ac- force problem. J Sound Vibr 2001;261:75–91.
[4] Yang YB, Yau JD, Hsu LC. Vibration of simple beams due
count the track structure has been presented. The proce-
to trains moving at high speeds. Eng Struct
dure basically consists in treating the train, rails and
1997;19(11):936–44.
bridge deck as three substructures. The problem of vehi- [5] Sadiku S, Leipholz HHE. On the dynamic effects of elastic
cle-bridge dynamic interaction is solved by applying a systems with moving concentrated masses. Ingenieur
particular variant of the traditional component-mode Archiv 1987;57:223–42.
synthesis method. Such variant enables the condensation [6] Akin JE, Mofid M. Numerical solution for response of
of the axle degrees-of-freedom into those of the rails in beams with moving mass. J Struct Eng ASCE
contact, duly accounting for the interaction effects 1989;115(1):120–31.
among the three subsystems. One of the main features [7] Hwang ES, Nowak AS. Simulation of dynamic load for
of the present method is the capability of handling bridges. J Struct Eng ASCE 1991;117(5):1413–34.
[8] Yang F, Fonder GA. An iterative solution method for
simultaneously the dynamic responses of the vehicles,
dynamic response of bridge-vehicles systems. Earthquake
rails and bridge. Furthermore, it is particularly suitable
Eng Struct Dynam 1996;25:195–215.
for the analysis of bridges crossed by trains consisting of [9] Yang YB, Lin BH. Vehicle-bridge interaction analysis by
a large number of rail cars. The procedure proves to be dynamic condensation method. J Struct Eng 1995;121(11):
extremely versatile as it allows to deal with vehicle mod- 1636–43.
els of various complexities and different boundary condi- [10] Yang YB, Liao SS, Lin BH. Impact formulas for vehicles
tions for the bridge. moving over simple and continuous beams. J Struct Eng
The accuracy and efficiency of the present method 1995;121(11):1644–50.
have been checked through appropriate comparisons [11] Yang YB, Yau JD. Vehicle-bridge interaction element for
with a recently proposed finite element approach based dynamic analysis. J Struct Eng 1997;123(11):1512–8.
[12] Yau JD, Yang YB, Kuo SR. Impact response of high speed
on the use of a bridge–track–vehicle element. Numerical
rail bridges and riding comfort of rail cars. Eng Struct
results presented in the paper have demonstrated that
1999;21:836–44.
the substructure technique is an effective tool for reduc- [13] Cheng YS, Au FTK, Cheung YK. Vibration of railway
ing the number of unknowns involved in the dynamic bridges under a moving train by using bridge–track–vehicle
analysis of railway bridges. element. Eng Struct 2001;23:1597–606.
[14] Wu YS, Yang YB. Steady-state response and riding
comfort of trains moving over a series of simply supported
Appendix A bridges. Eng Struct 2003;25:251–65.
[15] Biondi B, Muscolino G, Sofi A. Analysis of dynamic
The matrices appearing in Eqs. (15) and (16) are de- interaction between suspension bridges and running trains.
In: Grundmann H, Schuëller GI, editors. Structural
fined as follows:
dynamics, EURODYN2002 2002. p. 1041–6.
[16] Hurty WC. Vibrations of structural systems by compo-
cb Ab þ cf cr Ar þ cf nent-mode synthesis. J Eng Mech ASCE 1960;86:51–69.
Nb ¼ I mb ; Nr ¼ I mr ; ðA:1Þ [17] Biondi B, Muscolino G. Component-mode synthesis
qb Ab qr Ar
Z lb method variants in the dynamics of coupled structures.
Mbr ¼ MTrb ¼ k f /b ðxÞ/Tr ðxÞ dxX2 r ; Meccanica 2000;35(1):17–38.
0
Z lb [18] Muscolino G. Dynamic analysis of structural systems using
Cbr ¼ CTrb ¼ /b ðxÞ/Tr ðxÞ dxðk f Nr X2 component mode synthesis. In: Topping BHV, Bittnar Z,
r  cf Imr Þ; ðA:2Þ
editors. Computational structure technology. Stirling,
Z 0 lb þl0
Scotland: Saxe-Coburg Publications; 2002. p. 255–82
DMb ¼ qr Ar wr ðxÞwTr ðxÞ dx; [Chapter 11].
l0
Z lb [19] Biondi B, Muscolino G. Component-mode synthesis
kf
DKb ¼ Im b  k f /b ðxÞwTr ðxÞ dx; ðA:3Þ method for coupled continuous and FE discretized sub-
qb Ab 0 structures. Eng Struct 2003;25:419–33.
Z lb þl0
[20] Den Hartog JP. Advanced strength of materials. New
DCb ¼ ðcr Ar þ cf Þ wr ðxÞwTr ðxÞ dx York: Dover; 1987.
l0
Z lb [21] Clough RW, Penzien J. Dynamics of structures. New
 2cf /b ðxÞwTr ðxÞ dx: ðA:4Þ York: McGraw-Hill, Inc.; 1993.
0

You might also like