You are on page 1of 18

Computers and Structures 85 (2007) 458–475

www.elsevier.com/locate/compstruc

Longitudinal forces in continuously welded rails on bridgedecks


due to nonlinear track–bridge interaction
P. Ruge, C. Birk *

Lehrstuhl Dynamik der Tragwerke, Fakultät Bauingenieurwesen, Technische Universität Dresden, D-01062 Dresden, Germany

Received 14 December 2005; accepted 5 September 2006


Available online 29 November 2006

Abstract

In this contribution longitudinal forces in continuously welded rails on railway bridges are addressed. Here, the interaction between
track and structure cannot be neglected. European codes recommend an independent treatment of different longitudinal loading cases
taking a nonlinear stiffness law into account. The final rail stresses are obtained by a summation of the results of the above independent
calculations. The influence of the load history is completely neglected in this conventional approach. Based on an explanation of the
track–bridge interaction problem a truly nonlinear description is proposed in this paper. Here, the deformation history is of essential
importance. Exact stiffness formulations are derived for the loading cases ‘temperature change’, ‘sudden change of ballast stiffness when
the train reaches the bridge’, ‘bending of the supporting structure’ and ‘braking’. Numerical results are presented for a typical track–
bridge system and different combinations of the above loading cases. Extensive parameter studies enable a general comparison of the
rail stresses obtained using the conventional approach to the results of the proposed truly nonlinear analysis based on a correct combi-
nation of loading cases.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Track-bridge interaction; Railway engineering; Longitudinal rail forces; Deformation history; Nonlinear stiffness law; Expansion device

1. Introduction welded. In this case the track–bridge interaction cannot


be neglected.
The track forms the basis of railway traffic. An overview In this context, longitudinal loads are of special impor-
of different track structures and models can be found in tance. These are caused by braking, uniform temperature
Ref. [1]. In general, this major part of the railway construc- change of the supporting structure or bending of the bridge
tion rests on the subgrade. The support of the track by an deck. In Refs. [2–4] it has been demonstrated that espe-
engineering structure constitutes a special situation. Here, cially the latter two loading cases can lead to a remarkable
rail, fastening, sleeper and ballast or rail, fastening and increase of the longitudinal rail stresses.
concrete strip in case of a ballasted or rigid track, respec- Apart from the above situations, an additional loading
tively, rest directly on the bridge deck. In former times a case which is commonly neglected is identified in this
separate treatment of the track and the supporting struc- paper. It is shown that longitudinal forces are caused by
ture in a design analysis was the norm. However, this the sudden change of ballast stiffness when the train
approach cannot be used if the rails are continuously reaches the bridge. A detailed discussion and mathematical
description of this loading case is given in Section 3.2. The
importance of this effect is substantiated by the numerical
*
results presented in Section 5.
Corresponding author. Tel.: +49 351 4633 5325; fax: +49 351 4633
4096.
The investigation of longitudinal loads and their influ-
E-mail addresses: Peter.Ruge@tu-dresden.de (P. Ruge), Carolin. ence on the forces in continuously welded rails on bridge-
Birk@tu-dresden.de (C. Birk). decks has been much discussed for the last 10 years. As a

0045-7949/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2006.09.008
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 459

result the UIC recommendations [5] and the European on the subsoil. It should be noted that in a practical anal-
norm EN 1991-2 [6] provide information on loads on ysis it is necessary to include a part of the rail resting on the
bridges, design methods and approaches today. Together subsoil in the numerical model, since the above springs of
with Appendix A2 of the European norm EN 1990 [7] a stiffness k0/k1 can only be used in regions where an elastic
common set of rules is available which will replace national stiffness law is valid.
norms throughout Europe. Nevertheless, these approaches Bridge and rail are coupled through the ballast or fas-
have to be verified from time to time, since both vehicles tening system in case of a ballasted or rigid track, respec-
and track constructions are subject to a permanent tively. This is modelled   as a continuous elastic bond of
development. distributed stiffness c mN2 . The value of the latter parameter
In Germany the design of railway tracks on bridges is cur- depends on whether the track is loaded or not:
rently controlled by the DIN Fachbericht 101 [8]. According
cu ½N=m2  for unloaded track;
to Ref. [8] different longitudinal loading cases are analysed ð1Þ
2
separately considering a nonlinear stiffness law of the ballast. cl ½N=m  for loaded track:
Such separate treatments of the loading cases braking and The above elastic coupling between bridge and rail can be
temperature change can be found for example in Refs. visualized as a coupling element subject to a displacement
[9,10]. A comprehensive description of the latter loading case difference uD as shown in Fig. 1.
is also contained in the book [11], chapter 14. The maximum
longitudinal stresses follow from a subsequent summation of uD ¼ uR  uB for juR  uB j < u~;
ð2Þ
the results corresponding to different loading cases. This uD ¼ signðuR  uB Þ~u for juR  uB j P ~u;
approximate approach completely neglects the influence of
preceding events on the current load process. In fact, a more where uR longitudinal displacement of the rail, uB longitu-
realistic combination of loading cases is required to obtain dinal displacement of the upper surface of the bridge.
correct results. The sequence of events is of importance, One important characteristic of the track–bridge system
the history of a load process has to be taken into account. is that the above deformation of the coupling element is
For this purpose, a truly nonlinear track–bridge interaction limited to a critical value ~u. Below this limit value, a linear
model resulting from a correct combination of loading cases elastic relationship between the displacement difference uD
is developed in this paper. Here, the corresponding differen- and the longitudinal restoring force q shown in Fig. 1 is
tial equation is transformed into an exact stiffness formula- valid
tion. The latter is used for the computation of detailed unloaded track: q ¼ cu uD ;
numerical results including parameter studies and a compar- ð3Þ
loaded track: q ¼ cl uD :
ison of approaches in Section 5.
Here, a positive force +q is assumed to act on the rail if the
displacement of the former is smaller than that of the
2. Track–bridge interaction bridge.
Above ~u the rail is slipping relative to the ballast or
In the following a track–bridge system as shown in concrete strip. The corresponding nonlinear stiffness law
Fig. 1 is considered. A rail of longitudinal stiffness EA is for the ballasted and rigid track is shown in Fig. 2. The lat-
resting on a bridge structure of length L. On the left-hand ter illustration indicates that a situation where the rail slips
side abutment (point A) the horizontal bridge movement is relative to the coupling element corresponds to a constant
restrained by
 an elastic support modelled as a spring with longitudinal restoring force ~q with
stiffness k A N . On the right-hand side abutment (point
m q ¼ ~q ¼ c  signðuR  uB Þ~u for juR  uB j P ~u: ð4Þ
B) the horizontal displacement of the structure is enabled
by a sliding
pffiffiffiffiffiffiffiffiffiffibearing.
ffi  Two springs with a stiffness However, Eq. (4) is valid for a virgin track–bridge system,
k 0 ¼ k 1 ¼ EAcu N m
represent that part of the rail resting which has not undergone any previous deformation only.

Fig. 1. Longitudinal track–bridge interaction: (a) system model, (b) coupling element uB > uR.
460 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

respectively. The above available capacity can be calcu-


lated using Eqs. (5) and (6)
jr ¼ ~u  uD ; ð5Þ
jl ¼ ~u  uD : ð6Þ

The following limiting relationships apply:


2~u 6 jl 6 0; 0 6 jr 6 2~u: ð7Þ

At the same time, the sum of the absolute values of the


positive and negative capacity is constant.
jjl j þ jr ¼ 2~u: ð8Þ

In Fig. 4 it can be seen that the negative and positive avail-


able deformation capacity enclose a band of width 2~ u. A
subsequent loading case proceeds elastically if the addi-
tional relative deformation zD of the coupling element,

Fig. 2. Nonlinear stiffness law for rails UIC 60 and sleepers B 70 W.


zD ¼ zR  zB ; ð9Þ

zR additional longitudinal displacement of the rail, zB


additional longitudinal displacement of the upper surface
For a sequence of loading cases the situation is more com- of the bridge, lies within this band. Otherwise, the rail slips
plicated, as will be explained in the following. relative to the coupling element

2.1. Influence of the load history jl 6 zD 6 jr ! elastic


ð10Þ
zD < jl ; zD > jr ! plastic
Consider a track–bridge system which has experienced
one single longitudinal loading process, such as tempera- If a subsequent loading case proceeds plastically, a lon-
ture change. The displacement difference (uR  uB) gitudinal restoring force q as given in Eqs. (11) and (12)
obtained for this first loading case is shown in Fig. 4. acts on the rail
The relative deformation uD follows immediately. Based q ¼ jr  c for zR  zB P 0; ð11Þ
on this result the rail can be divided into regions where
q ¼ jjl j  c for zR  zB < 0: ð12Þ
the elastic coupling to the bridge is retained and where
the rail slips relative to the coupling element. In the follow- In case of a positive additional deformation of the coupling
ing, these two situations will be referred to as ‘elastic’ and element, the positive capacity jr is used to determine q,
‘plastic’, respectively. otherwise jl is relevant. Using Eqs. (5) and (6), formula-
Depending on the state of deformation of the coupling tions (11) and (12) can be written as:
element there is a remaining deformation capacity j avail-
able for a subsequent loading case to proceed elastically. q ¼ ðuD  ~uÞc for zR  zB P 0; ð13Þ
This is illustrated in Fig. 3 for the case of a positive relative q ¼ ðuD þ ~uÞc for zR  zB < 0: ð14Þ
deformation 0 < uD < þ~ u. Here, the terms jr and jl denote
the capacity corresponding to a possible positive and neg- Eqs. (11)–(14) will be used for a derivation of stiffness for-
ative additional deformation zD of the coupling element, mulations in the following sections.

3. Exact stiffness formulations for longitudinal loads

The loading cases

• uniform temperature change of the supporting structure,


• sudden change of ballast stiffness when the train reaches
the bridge,
• bending of the supporting structure,
• braking

are described mathematically in the following. In any case,


the actual relative deformation uD of the coupling element
Fig. 3. Definition of deformation capacity. between rail and bridge due to an arbitrary number of
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 461

Fig. 4. Deformation situation after a preceding loading case. (a) Displacement difference (uR  uB), (b) deformation uD of the coupling element and (c)
corresponding available capacity for a subsequent elastic deformation.

preceding load processes is assumed to be known. The spe- d d2 Du2 d3 Du3


cial case of a virgin structure is included in the following fNL½ugju1 Du þ 2 fNL½ugju1 þ 3 fNL½ugju1 ¼ f2 :
du du 2 du 6
formulation with uD = 0. ð18Þ
At this point it should be recalled that the response utotal
of a nonlinear system to a sum of loads is not equal to the
sum of the individual responses to parts of the total Evaluation of Eq. (18) yields a third-order problem similar
excitation. to Eq. (17).
utotal ¼ uðf1 þ f2 Þ 6¼ uðf1 Þ þ uðf2 Þ ¼ u1 þ u2 : ð15Þ
ðC 1 þ 2C 2 u1 þ 3C 3 u21 ÞDu þ ðC 2 þ 3C 3 u1 ÞDu2 þ C 3 Du3 ¼ f2 :
Nevertheless, the total displacement can be expressed as
the sum of the displacement due to a first load u1 and an ð19Þ
additional deformation due to a subsequent load f2
Thus, an equivalent nonlinear problem with different coef-
utotal ¼ u1 þ Du: ð16Þ ficients results. The latter depend on the response u1 to the
This is illustrated considering the following scalar nonlin- first load f1. These considerations can be transferred to the
ear problem: nonlinear track–bridge system. As explained in Section 2.1,
the response of the latter to a subsequent load strongly
NL½u ¼ C 1 u þ C 2 u2 þ C 3 u3 ¼ f ; NL½u1  ¼ f1 : ð17Þ
depends on the state of deformation due to a preceding
The additional displacement Du due to a subsequent load f2 sequence of loading cases. Consequently, the total displace-
follows from a Taylor-series expansion: ment of the rail is the sum of the displacement due to
462 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

preceding loading cases uR and the additional deformation zR ðxÞ ¼ pT hðxÞ;


zR due to the current load 2 3
z0
uR;total ¼ uR þ zR : ð20Þ 6z 7
6 17
p ¼ 6 7;
4 zA 5
In order to derive a stiffness formulation,
1
Kz ¼ r ð21Þ 2 3
sinh cðL  xÞ
6 sinh cx 7 1
for the additional displacements z a typical continuous 6 7
hðxÞ ¼ 6 7 :
track–bridge system as shown in Fig. 1 is analysed. Here, 4  sinh cðL  xÞ  sinh cx þ sinh cL 5 sinh cL
it is distinguished between an elastic coupling of rail and ~aL sinh cx þ ~ax sinh cL
bridge and a situation where the rail slips relative to the
ð29Þ
structure.
The longitudinal forces L0 and L1 at the end points follow:
3.1. Uniform temperature change of the supporting 2 3
 coth cL
structure
6 7
pffiffiffiffiffiffiffiffiffi6 sinh1 cL 7
This loading case is characterized by an imposed defor- L0 ¼ EAz0R ðx ¼ 0Þ ¼ pT cEA6 6 tanh cL 7;
7 ð30Þ
4 2 5
mation of the structure, ~
aL
 sinh cL
þ ~ac
zB ¼ zA þ aDTx ¼ zA þ ~
ax; a ¼ aDT ;
~ ð22Þ 2 3
 sinh1 cL
6 7
where a denotes the thermal expansion coefficient of the pffiffiffiffiffiffiffiffiffi6 coth cL 7
L1 ¼ EAz0R ðx ¼ LÞ ¼ pT cEA6 6 cL
7;
7 ð31Þ
bridge. 4  tanh 2 5
~aL coth cL þ ~ac
3.1.1. Elastic coupling
Provided that the elastic coupling is retained, the addi- Evaluating the equilibrium of forces at the points 0, 1, A
tional deformation of the rail is described by the differential and the overall system,
equation (23),
x ¼ 0: L 0 ¼ k 0 z0 ;
EAz00R  cðzR  zB Þ ¼ 0; ð23Þ x ¼ L: L1 ¼ k 1 z1 ;
ð32Þ
with zB given in Eq. (22). An analytical solution (24), A: L A ¼ k A zA ;
c overall: LA ¼ L1  L0
zR ðxÞ ¼ Aecx þ Becx þ zA þ ~
ax; c2 ¼ ; ð24Þ
EA and eliminating the longitudinal forces L0, L1 and LA in
contains two integration constants A and B. These Eqs. (30)–(32),
2 3 2 3 2 3
coefficients can be replaced using the end displacements L0 k0 0 0 z0
z0 and z1. 6 7 6 7 6 7
ð32Þ: 4 L1 5 ¼ Kk z; Kk ¼ 4 0 k 1 0 5; z ¼ 4 z1 5;
z0 ¼ zR ðx ¼ 0Þ ¼ A þ B þ zA ; ð25Þ LA 0 0 kA zA
2 3
cL
z1 ¼ zR ðx ¼ LÞ ¼ Ae þ Be cL
þ zA þ ~
aL: ð26Þ L0
6 7
ð30Þ–ð31Þ: 4 L1 5 ¼ Ksc z þ r;
Using Eqs. (25) and (26), the coefficients A and B can be
LA ¼ L1  L0
formulated in terms of the unknown displacements z0, z1
and zA.
a stiffness formulation (33) is obtained to describe the elas-
2 3 2 3
A z0 ecL  z1 þ zA ð1  ecL Þ þ ~
aL tic behaviour of a continuous track–bridge system due to
4 5¼ 1 4 5: uniform temperature change of the structure
ecL  ecL z ecL þ z þ z ð1 þ ecL Þ  ~ a L
B 0 1 A
ðKsc þ Kk Þz ¼ r; ð33Þ
ð27Þ 2 3
coth cL 1
sinh cL
 tanh cL2
pffiffiffiffiffiffiffiffiffi6 7
Using this result and the definition (28) of the hyperbolic Ksc ¼ cEA6 4  coth cL  tanh cL2 7
5 ¼ KTsc ; ð34Þ
sine function,
  2 tanh cL2
1 2 1 3
sinh cL ¼ ðecL  ecL Þ; ð28Þ cL
 sinh1 cL
2 pffiffiffiffiffiffiffiffiffi6 1 7
r ¼ aDTL cEA6 7
4  cL þ coth cL 5:
the analytical solution (24) can be rewritten in a normalized
form:  tanh cL2
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 463

pffiffiffiffiffiffiffiffiffih i
The above derivation is valid if the complete track–bridge Lj ¼ cEA  coth clj sinh1 clj tanh 2j zj
cl

system behaves elastically and can thus be described by  


pffiffiffiffiffiffiffiffiffi 1 1 pffiffiffiffiffiffiffiffiffi clj
one single element of length L. However, plastic and elastic þ aDTlj cEA  þ aDTaj cEA tanh ;
clj sinh clj 2
regions will alternate along the structure in practice. In a ð38Þ
general analysis the rail will therefore be divided into sev- pffiffiffiffiffiffiffiffiffih 1 cl
i
Ljþ1 ¼ cEA sinh clj coth clj  tanh 2j zj
eral elements of length lj as illustrated in Fig. 5. In a corre-  
pffiffiffiffiffiffiffiffiffi 1 pffiffiffiffiffiffiffiffiffi clj
sponding local description the prescribed displacement of þ aDTlj cEA  þ coth clj þ aDTaj cEA tanh :
the structure due to temperature change is characterized clj 2
by an additional term ~ aaj ð39Þ

zB ðxj Þ ¼ ðzA þ ~
aaj Þ þ ~
ax j : ð35Þ
3.1.2. Plastic situation (slip)
If the rail slips relative to the coupling element, the addi-
Consequently, an element stiffness formulation for an
tional deformation zR is described by the differential equa-
embedded part of the rail is obtained replacing zA by
tion (40).
zA þ ~aaj in formulation (33). That means, the third column
of the stiffness matrix (34) is multiplied by ~ aaj and moved EAz00R ¼ qðxÞ: ð40Þ
to the right-hand side. The resulting element stiffness Here, the elastic coupling between rail and bridge is ineffec-
matrix Kj and element load vector rj corresponding to the tive, the rail is loaded by the longitudinal restoring force q
degrees of freedom zj are given in Eqs. (36) and (37) defined in Eqs. (11) and (12). In regions where the previous
2 1 cl 3 loading case proceeded elastically, this longitudinal force q
coth clj sinh clj
 tanh 2j
follows an unknown continuous curve with
pffiffiffiffiffiffiffiffiffi6
6
7
clj 7 T
Kj ¼ cEA6  coth cl j  tanh 7 ¼ Kj ; 2~u  c 6 qðxÞ 6 2~u  c for juD j 6 ~u: ð41Þ
4 2 5
clj
  2 tanh 2
It is constant in regions where the rail slipped relative to the
2 3 coupling element due to the previous loading case
zj 8
6 7 < 2~u  c
>
zj ¼ 6 7
4 zjþ1 5; ð36Þ
q¼ 0 for juD j > ~u:
zA >
:
þ2~u  c
2 1
3
clj
 sinh1 clj Since the exact function of q(x) in previously elastic regions
pffiffiffiffiffiffiffiffiffi6
6
7
7 is a result of a complete load process and not known in
rj ¼ aDTlj cEA6  cl1j þ coth clj 7
4 5 closed form, we suggest to use a linear interpolation of
clj the nodal values qj and qj+1 in the following. This assump-
 tanh 2
2 cl
3 tion is correct for regions where the preceding loading case
tanh 2j proceeded plastically. The error introduced in previously
pffiffiffiffiffiffiffiffiffi6
6
7
7 elastic regions is minimized using a sufficiently fine discret-
þ aDTaj cEA6 tanh cl2j 7: ð37Þ
4 5 ization of the rail. Moreover, it should be recalled that the
clj exact function of q(x) is limited according to Eq. (41).
2 tanh 2
Therefore, no sharp peaks occur in the course of q(x),
The longitudinal forces at the element nodes follow from a which is beneficial for a linear approximation of the latter.
subsequent calculation For
qjþ1  qj
qðxj Þ ¼ x j þ qj ; ð42Þ
lj
an analytical solution of the differential equation (40) is
given in Eq. (43)
qj  qjþ1 3 qj 2
zR ¼ Ax þ B þ x  x: ð43Þ
6EAlj 2EA
The coefficients A and B are determined using the nodal dis-
placements zj and zj+1
" # " #   
A  l1j l1j zj 2qj þ qjþ1 lj
¼ þ : ð44Þ
B 1 0 zjþ1 6EA 0

Fig. 5. Embedded rail element of length lj described by the local The normalized solution (45) and the nodal forces Lj, Lj+1
coordinate xj. follow:
464 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

zR ¼ pT hðxj Þ; ð45Þ The right-hand side restoring force is given in Eq. (48).
2 3 In the special case of one single preceding loading case,
2 3 1  ljj
x
zj 6 7 namely a uniform temperature change of the structure
6 7 6 xj 7 which proceeded elastically, Dq is known in closed form.
p ¼ 4 zjþ1 5; h ¼ 6 lj 7;
4 5 In this situation a hyperbolic right-hand side of Eq. (50)
qj qjþ1 3 qj 2 2qj þqjþ1
1 x  x þ l x
6EAlj j 2EA j 6EA j j is obtained using Eqs. (22) and (29) to describe the relative
" # " #" # deformation of the coupling element. It is possible to derive
Lj EA 1 1 zj
¼ a corresponding closed-form solution of Eq. (50) which
Ljþ1 l j 1 1 zjþ1 leads to an exact stiffness formulation for the sequence of
" # " #
q j lj 2 qjþ1 lj 1 loading cases ‘DT (elastic) ! Dc (elastic)’. For a detailed
þ þ : ð46Þ mathematical description of this special case the reader is
6 1 6 2 referred to Ref. [2].
The element stiffness matrix Kj and right-hand side element However, in this paper the more general situation of an
vector rj corresponding to the degrees of freedom zj are ob- arbitrary sequence of loading cases is addressed. As
tained in a completely analogous manner to the preceding explained in the preceding section, the exact function
paragraph. Dq(x) is not known in this case. Therefore, a linear interpo-
2 3 2 3 lation of nodal values of the restoring force,
1 1 0 zj
EA 6 7 6 7 Dqjþ1  Dqj
Kj ¼ 4 1 1 0 5; zj ¼ 4 zjþ1 5; Dqðxj Þ ¼ xj þ Dqj ; ð51Þ
lj lj
0 0 0 zA
2 3 2 3
2 1 is used as right-hand side of Eq. (50) in the following. The
q j lj 6 7 qjþ1 lj 6 7
rj ¼ 4 1 5 þ 4 2 5: ð47Þ corresponding analytical solution of Eq. (50) is given below
6 6
3 3 Dqjþ1  Dqj Dqj cl
zR ¼ Aecl x þ Becl x þ xþ þ zA ; c2l ¼ :
lj c l cl EA
3.2. Sudden change of ballast stiffness ð52Þ

According to Section 2 the stiffness of the coupling ele- The coefficients A and B and the normalized solution
ment between rail and bridge depends on whether the track obtained using the end displacements zj, zj+1 are given in
is loaded or not. Consequently, the value of the parameter Eqs. (53) and (54), respectively
c changes suddenly when a train reaches the bridge. This " # 2 Dqj c lj Dqjþ1
3
c lj c lj
leads to an additional restoring force, A 1 4 zj e l  zjþ1 þ zA ð1  e l Þ  cl e l þ cl 5
¼ ;
B 2sinhcl lj zj ecl lj þ zjþ1  zA ð1  ecl lj Þ þ Dqj ecl lj  Dqjþ1
Dq ¼ Dc  uD ; Dc ¼ cl  cu ; ð48Þ cl cl

acting on both the rail and the structure. In Eq. (48) uD is ð53Þ
the relative deformation of the coupling element due to all Dqjþ1  Dqj Dqj
zR ¼ pT hðxj Þ þ xj þ ; ð54Þ
preceding loading cases. It is defined in Eq. (2). A limita- lj c l cl
2 3 2 3
tion of Dq according to Eq. (49) follows: zj sinhcl ðlj  xj Þ
6z 7 6 sinhcl xj 7
Dc  ~
u 6 Dq 6 Dc  ~
u: ð49Þ 6 jþ1 7 1 6 6
7
7:
p¼6 7; h ¼
This additional restoring force should be analysed as a sep- 4 zA 5 sinhcl lj 6
4 sinhcl xj  sinhcl ðl j  x j Þ þ sinhc l
l 5j
7
Dqjþ1 Dqj
arate loading case in connection with bending of the bridge 1  cl
sinhcl xj  cl
sinhcl ðlj  xj Þ
deck due to live load and braking. Consequently, a reliev-
ing load Dq should be considered prior to a subsequent The longitudinal forces at the element nodes follow
temperature change, if it is assumed that the train has left 2 32 z 3
  1
pffiffiffiffiffiffiffiffiffiffi cothcl lj  sinh cl lj  tanh 2 6 j 7
cl lj
the bridge again. Lj
¼  cl EA4 cl
54 zjþ1 5
Ljþ1  sinh1cl lj coth cl lj  tanh l2 j
3.2.1. Elastic coupling zA
2 Dq Dqjþ1 Dqjþ1 Dqj
3
Provided that the elastic coupling is effective, the addi- j 1
pffiffiffiffiffiffiffiffiffiffi cl cothcl lj  cl sinhcl lj þ lj cl cl
tional displacement of the rail is described by the differen- þ cl EA4 Dq Dq Dqjþ1 Dqj
5:
tial equation (50).  cl j sinh1cl lj þ cjþ1
l
coth c l l j  l j c l cl

EAz00R  cl ðzR  zA Þ ¼ Dq: ð50Þ ð55Þ


It should be noted that the stiffness of the loaded ballast, cl, The resulting element stiffness matrix is identical to Kj
is used in Eq. (50). The additional displacement of the given in Eq. (36) for the loading case ‘DT (elastic)’. The
bridge due to the sudden change of ballast stiffness is con- corresponding right-hand side vector can be found in
stant, zB = zA. Eq. (56).
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 465

2 Dqj Dqjþ1 Dqjþ1 Dqj


3
coth cl lj  1
þ This situation is illustrated in Fig. 6. Here and in the fol-
cl cl sinh cl lj lj cl cl
pffiffiffiffiffiffiffiffiffiffi6
6 Dq Dqjþ1 Dqjþ1 Dqj 7
7 lowing the symbols zB, zN and zA are used to denote the
1
rj ¼ þ cl EA6  c j þ coth c l
l j  7: longitudinal displacement of the upper surface, the neutral
4 l sinh cl lj cl l j c l cl 5
Dq þDq cl axis and the left-hand side elastic support of the bridge,
 j cl jþ1 tanh l2 j
respectively. The deformation zB is defined by the displace-
ð56Þ ment of the neutral axis and the slope of the bending line
w 0 (x).
3.2.2. Plastic situation (slip) zB ðxÞ ¼ zN þ w0 ðxÞh0 : ð60Þ
Here, the additional displacement of the rail due to the
sudden change of ballast stiffness is described by the differ- The rail stress due to bending is influenced by the relative
ential equation (57). location of the neutral axis g and the relative maximum
vertical displacement d
EAz00R ¼ DqðxÞ  qðxÞ: ð57Þ  
h0 w^ L
The elastic coupling between rail and bridge is ineffective, g¼ ; d¼ ; w ^¼w x¼ : ð61Þ
hu L 2
two types of loading are acting on the former. As in the
preceding subsection, Dq is the restoring force given in For a single-span beam of bending stiffness EI under con-
Eq. (48) due to the sudden change of ballast stiffness. Since stant live load q0 the deflection of the neutral axis is de-
the rail slips relative to the coupling element, an additional scribed by Eq. (62)
longitudinal restoring force q defined in Eqs. (11) and (12)  x
3 x
4 
q0 L4 U x
is present. As explained in Section 2.1, the latter depends wðxÞ ¼ 2 þ : ð62Þ
24EI L L L
on the available deformation capacity jr (5)/jl (6) resulting
from a sequence of loading cases. Assuming a linear varia- Here, U is a dynamic magnification factor which can be
tion of both Dq and q, chosen according to national design codes. Using Eq.
Dqjþ1  Dqj qjþ1  qj (62) the maximum deflection at the centre of the span
Dq þ q ¼ xj þ Dqj þ þ qj ; ð58Þ can be evaluated:
lj lj  
L 5 q0 L4 U 5 q0 L3 U
the solution of Eq. (57) can be found in a completely anal- ^ ¼w x¼
w ¼ ; !d¼ : ð63Þ
2 16 24EI 384 EI
ogous manner to the situation where the rail slips relative
to the coupling element due to temperature change of the The relative maximum deflection follows. Using Eq. (63),
structure. The resulting element stiffness matrix is identical the slope of the bending line can be formulated in terms
to Kj given in Eq. (47). The corresponding right-hand side of d
vector is obtained replacing q by (Dq + q).  x
2 x
3 
2 3 2 3 0 16
2 1 w ðxÞ ¼ d 1  6 þ4 : ð64Þ
ðDqj þ qj Þlj 6 ðDq þ q Þl 5 L L
7 jþ1 jþ1 j 6 7
rj ¼ 4 1 5þ 4 2 5: ð59Þ
6 6 Eq. (64) is rewritten in terms of a local element coordinate
3 3 xj as defined in Fig. 5.
 a þ x
2 a þ x
3 
0 16 j j j j
3.3. Bending of the supporting structure w ðxj Þ ¼ d 1  6 þ4 ; 0 6 x j 6 lj :
5 L L
Since rail and bridgedeck are coupled through the bal- ð65Þ
last or fastening system, longitudinal rail forces are caused
by bending of the structure due to the applied live load. 3.3.1. Elastic coupling
Provided that the elastic coupling between rail and
bridge is retained, the additional displacement of the rail
due to bending of the structure is described by the differen-
tial equation (66)
EAz00R  cl ðzR  zB Þ ¼ 0: ð66Þ
It should be noted that the stiffness cl is used since the track
is loaded by the live load q0. Using Eqs. (60) and (65) to de-
scribe the additional displacement of the upper face of the
bridge, Eq. (66) is specified as:
 a þ x
2 a þ x
3 
16 j j j j
EAz00R  cl zR ¼ cl zN  cl h0 d 1  6 þ4 :
5 L L
Fig. 6. Bending of the supporting structure. ð67Þ
466 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

The corresponding analytical solution is given in Eq. (68). The longitudinal forces at the element nodes follow:
" # 2 c lj
3
Lj pffiffiffiffiffiffiffiffiffiffi coth cl lj  sinh1cl lj  tanh l2
zR ¼ A cosh cl xj þ B sinhcl xj þ zN ¼  cl EA4 5z
 a þ x
3  Ljþ1  sinh1cl lj coth cl lj  tanh l2 j
cl
16 12 24 aj þ xj
a þ x
2
j j j j
þ h0 d 1  2 2 þ 2 2 6 þ4 : 2 3
5 cl L cl L L L L c lj
pffiffiffiffiffiffiffiffiffiffi coth cl lj  sinh1cl lj  tanh l2
ð68Þ þ cl EA4 cl
5~h
 sinh1cl lj coth cl lj  tanh l2 j
2 3
The integration constants A and B are determined using 24
 L122 aj þ L123 a2j
c2l L3
16 6
the nodal displacements zj = zR(xj = 0) and zj+1 = þ EAh0 d4

2 7 5:
zR(xj = lj), 5 24
 c2 L3 þ L 12 aj þlj
L 12 aj þlj
L L
l
2 3 ð74Þ
" # " # zj
A 1 sinh cl lj 0  sinh cl lj 6 7
¼ 6 zjþ1 7 The corresponding element stiffness relationship is given in
sinh cl lj  cosh c lj 4 5
B l 1 ðcosh cl lj  1Þ Eq. (75).
zN 2 3 2 3
Lj zj
2 3 6 7 6 7
~ 4 Ljþ1 5 ¼ Kj 4 zjþ1 5 þ Kj ~h þ rjj ; rj ¼ Kj ~h þ rjj ;
" # h0
1  sinh cl lj  sinh cl lj 0 66 ~ 7
7
þ 6 hj 7; Ljþ1  Lj zN
sinh cl lj ðcosh c lj  1Þ
l cosh cl lj 1 4 5
ð75Þ
~
hjþ1 2 3
cl
coth cl lj  sinh1cl lj  tanh l2 j
ð69Þ pffiffiffiffiffiffiffiffiffiffi6 7
Kj ¼ cl EA6 1
4  sinh cl lj coth cl lj  tanh
cl l j
2
7;
5
with  tanh
cl l j
 tanh
cl l j
2 tanh
cl lj
2 2 2
  cl
16 12 c2l ¼ ;
~
h0 ¼ h0 d 1  2 2 ; ð70Þ EA
2 a aj 3
5 cl L 2
1  Lj
c2l L2 L


2
3  6
7
EA 6 aj þlj 7
~h ¼ 16 h0 w
^
~hj ¼ 16 h0 d 24 aj  6 aj þ 4 aj 2 aj þlj
; ð71Þ ~
rjj ¼ 12h 6  þ 1  7; :
L 6 7
2
c L 2 L L
5 c2l L2 L L L 4 l 5 5 L
2
ðaj þlj Þ 2
aj
3 !
lj
   2  L2
 L2
 L
~ 16 24 aj þ lj aj þ lj aj þ lj
hjþ1 ¼ h0 d 2 2 6 þ4 : It can be seen that the element stiffness matrix for the load-
5 cl L L L L
ing case ‘bending (elastic)’ is identical to Kj given in Eq.
ð72Þ (36) for the loading case ‘DT (elastic)’.
The contribution of the element j to a global stiffness
Substituting Eq. (69) into (68) the normalized solution (73) formulation is obtained evaluating the equilibrium of
results forces at the nodes j, j + 1 and the overall system
x ¼ aj: þLj þ additional terms ¼ 0;
zR ¼ zT hðxj Þ  ~
hT hðxj Þ þ zN þ ~
h0 x ¼ aj þ lj: Ljþ1 þ additional terms ¼ 0;
 a þ x
3  ð76Þ
16 24 aj þ xj
a þ x
2
j j j j A: LA ¼ k A zA ¼ k A ðzN  hu w0 ð0ÞÞ;
þ h0 d 2 2 6 þ4 ;
5 cl L L L L overall: LA ¼ Ljþ1  Lj :
ð73Þ Summarizing:
2 3 2 32 3 2 3
2 3 2 3 Lj 0 0 0 zj 0
zj ~
hj 6 7 6 76 7 6 7
6 7 4 Ljþ1 5 ¼ 40 0 0 54 zjþ1 5  4 0 5:
6 7 6 7
z¼6 7
4 zjþ1 5;
~
h ¼ 6~hjþ1 7; LA ¼ Ljþ1  Lj 0 0 kA zN 16 w
k A hu 5 L^
4 5
zN ~
 h0 ð77Þ
2 3 It should be noted that the reaction LA is formulated in
cosh cl xj  coth cl lj sinh cl xj
6 7 terms of the slope of the bending line at x = 0,
6 sinh cl xj 7 w0 ð0Þ ¼ 165 d, in Eq. (76). This corresponds to the situation
hðxj Þ ¼ 6
6
7:
7
4 sinh cl lj 5 where the elastic support is situated at the left-hand side
cl l j
of the bridge. The resulting global stiffness formulation
 cosh cl xj þ tanh 2
sinh cl xj for a system with (N  2) rail elements has the form:
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 467
X X
Kztot ¼ r; K¼ Kj þ Kdiscrete ; r¼ rj þ rA ; ð78Þ

with
zTtot ¼ ½ z1 z2   zN ; Kdiscrete ¼ diagf k 0 0   0 k 1 k A g ;
16 w^
rTA ¼ ½ 0  0 þ1 k A hu ; for k A at x ¼ 0:
5 L
A slightly different right-hand side vector rA results, if the
Fig. 7. Typical track–bridge system with constant braking force p.
spring kA is situated at the right-hand side of the structure.
16 w
^
rTA ¼ ½ 0    0 1 k A hu ; for k A at x ¼ L: The longitudinal forces at the bridge ends follow:
5 L 2 3
  " # z
L0 pffiffiffiffiffiffiffiffiffiffi coth cl L  sinh1 cl L  tanh c2l L 6 0 7
3.3.2. Plastic situation (slip) ¼  cl EA 4 z1 5
L1  sinh1 cl L coth cl L  tanh c2l L
The additional displacement of the rail is described by zA
the differential equation (79). " #
pffiffiffiffiffiffiffiffiffiffi p tanh c2l L
þ cl EA : ð82Þ
EAz00R ¼ qðxÞ: ð79Þ cl tanh c2l L
The coupling between rail and structure is not effective any-
Evaluation of the equilibrium of forces (83) using Eq. (82),
more. A longitudinal restoring force q(x) acts on the rail
which depends on the available elastic deformation capac- x ¼ 0: L 0 ¼ k 0 z0
ity due to the complete preceding loading process. The sit- x ¼ L: L1 ¼ k 1 z1 ;
uation ‘bending (plastic)’ is thus completely analogous to ð83Þ
A: L A ¼ k A zA ;
the loading case ‘DT (plastic)’. Assuming a linear variation
of the longitudinal resistance q(x) between nodal values, overall: LA ¼ L1  L0 þ pL
the element stiffness matrix Kj and right-hand side vector yields the same stiffness matrix K as given in Eqs. (33) and
rj given in Eq. (47) can be used to describe this situation. (34), but a different right-hand side vector r:
However, the stiffness cl of the loaded coupling element 2 3
should be used when evaluating the nodal values of the lon- tanh c2l L
p6 7
gitudinal resistance according to Eqs. (5)/(6) and (11)/(12). r¼ 4 tanh c2l L 5: ð84Þ
cl
cl L
2 tanh 2 þ cl L
3.4. Braking

In the following a constant braking force p as shown in 3.4.2. Plastic situation (slip)
Fig. 7 is considered. Since the track is loaded, the coupling This situation is described by the differential equation
element is characterized by the stiffness cl. (85).
EAz00R ¼ qðxÞ  p: ð85Þ
3.4.1. Elastic coupling
Provided that the elastic coupling is retained, the addi- In addition to the longitudinal restoring force q(x) the
tional displacement of the rail due to the braking force p braking force p acts on the rail. The constant value of
is described by the differential equation (80). the latter can be seen as a special case of the above assumed
EAz00R  cl ðzR  zB Þ ¼ p: ð80Þ linear variation of the axial load. As all other plastic load-
ing cases, the situation ‘braking (plastic)’ is thus described
The normalized analytical solution of Eq. (80) for a contin- by the element stiffness matrix Kj given in Eq. (47). The
uous track–bridge system of length L is given below right-hand side vector r defined in Eq. (86) results for
pj = pj+1 = p
zR ðxÞ ¼ pT hðxÞ; ð81Þ 2 3 2 3 2 3
2 3 2 1 1
z0 q j lj 6
6z 7 7 qjþ1 lj 6 7 plj 6 7
6 17 rj ¼ 4 1 5þ 4 2 5þ 4 1 5: ð86Þ
p ¼ 6 7; 6 6 2
4 zA 5 3 3 2
1
2 3
sinh cl ðL  xÞ 4. Numerical realization
6 sinh cl x 7 1
6 7
h¼6 7 : The theoretical concepts derived above have been imple-
4  sinh cl ðL  xÞ  sinh cl x þ sinh cl L 5 sinh cl L
p mented in a FORTRAN programme. The latter can be used
cl
ð sinh cl ðL  xÞ  sinh cl x þ sinh cl LÞ to model an arbitrary sequence of the loading cases
468 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

the commonly neglected loading case ‘sudden change of


ballast stiffness’ is addressed in Section 5.4.

5.1. Comparison of the conventional approach and a truly


nonlinear analysis

Consider a track–bridge system described by the data


Fig. 8. Typical result of a first loading case. Temperature change DT.
given in Table 1. The stiffness of the coupling element cor-
responds to the situation ‘rigid track’. For the system
temperature change, sudden change of ballast stiffness, described in Table 1 the sequence of loading cases ‘warm-
bending and braking. ing up by DT = +30 K ! ‘bending’ ! ‘braking from
It should be recalled that the stiffness formulations pre- support A to support B with p ¼ þ20; 000 N m
’ leads to the
sented in Sections 3.1, 3.3 and 3.4 are exact if a loading case maximum rail stresses. However, in general the interaction
proceeds elastically or if slipping of the rail occurs in a pre- of different loading cases is not easy to foresee owing to the
viously plastic region. On the other hand, approximate big number of parameters involved. In a specific applica-
results are obtained if a plastic situation follows an elastic tion it should be examined which of the combinations of
one. Consequently, the numerical model of the first loading ±DT and ±p is most critical. It should be noted that the
case is exact throughout the track–bridge system. In fact, loading case ‘sudden change of ballast stiffness’ is neglected
only four elements were necessary to accurately describe in this example for clarity. Consequently, the loaded stiff-
the example situation shown in Fig. 8. However, subse- ness cl is used for the coupling element between rail and
quent loading cases lead to a different distribution of the bridge throughout the complete analysis. The unloaded
‘elastic’ and ‘plastic’ regions. For example the second part stiffness cu is used to describe the elastic support of the rail
BC of the rail shown in Fig. 8 which corresponds to an on the subsoil.
elastic situation due to DT could be split into ‘elastic’ and The maximum rail stresses obtained analysing each lon-
‘plastic’ regions due to a subsequent change of ballast stiff- gitudinal loading case separately and adding the results are
 N
ness. The associated increased number of necessary degrees listed in Table 2. Here and in the following the unit mm 2

of freedom complicates the conception of a transparent has been used for stresses, since this is the preferred unit
computer programme. of the German DIN Fachbericht [8]. For comparison, the
For this reason, the FORTRAN code developed by the maximum rail stresses resulting from a correct combination
authors is based on the use of a large number of elements of loading cases in a truly nonlinear analysis according to
of small length, typically 10 cm. One element is treated as
either completely elastic or plastic, depending on whether Table 1
the absolute value of the difference in additional displace- Example track–bridge system (geometry, material and loading data)
ments zR  zB is smaller than the corresponding capacity Length of the bridge L = 60 m
l
for more than 2j or not. Starting from an assumed configu- Stiffness of the elastic support (spring) k A ¼ 6  108 N=m
kA
ration the additional deformation of the rail is calculated in Parameter K ¼ L10 6 K = 10
a first step. The resulting elastic/plastic regions are com- Longitudinal rail stiffness EA = 3.23 · 109 N
pared to the initial guess and used as an improved assump- Distributed stiffness of unloaded track cu ¼ 6:0  107 N=m2
tion in a repeated calculation. This iterative solution Distributed stiffness of loaded track cl ¼ 12:0  107 N=m2
process is continued until the assumed distribution of ‘elas- Critical elastic relative deformation Du = uR  uB ~u ¼ 0:0005 m
Distance from neutral axis to upper surface h0 = 1.21 m
tic’ and ‘plastic’ regions is reproduced.
Distance from neutral axis to lower surface hu = 4.79 m
Ratio g = h0/hu g = 0.2526
^
w 1
Maximum relative deflection L ¼ 2500
5. Numerical results Temperature DT = +30 K
Thermal expansion coefficient of the bridge a = 1.0 · 105
The developed computer programme has been used in a Braking p = +20,000 N/m
number of parameter studies to demonstrate the influence
of the load history on the longitudinal rail stresses. Selected
results are summarized in the following. First, a detailed
comparison of the stresses obtained using the conventional Table 2
Maximum longitudinal rail stresses due to temperature change, bending
approach and a truly nonlinear analysis is given in Section and braking obtained analysing each loading case separately
5.1 for an example system of length L = 60 m. The influ-  N   N 
Load rmin mm 2 rmax mm 2
ence of the load history is further substantiated considering
a sequence of two opposite braking forces in Section 5.2. (1) DT = +30 K 66.506 58.624
(2) p = +20,000 N/m 19.276 19.276
An attempt to draw general conclusions is made in Section (3) Bending 34.051 49.914
5.3 based on a parameter study for the loading process
(1) + (2) + (3) 119.833 76.383
‘DT ! bending ! braking’. Finally, the importance of
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 469

Table 3 ping of the rail relative to the coupling element, as can be


Maximum longitudinal rail stresses due to a correct combination of the seen in Fig. 11. Here, the plastic region extends over a con-
loading cases temperature change, bending and braking in a nonlinear
analysis
siderably larger area of the track–bridge system in case of a
 N   N  correct combination of successive loading cases. It should
Load rmin mm rmax mm
2 2
be noted that the preceding loading case temperature
(1) DT = +30 K 66.506 58.624 change proceeded plastically in major parts of this region
(2) Bending after (1) 15.322 53.174
(3) p = +20,000 N/m after (1) + (3) 7.514 9.527
already. There, the elastic deformation capacity of the cou-
pling element is completely exhausted. Thus, a zero restor-
(1) + (2) + (3) 87.472 69.895
ing force q according to Eq. (11) acts on the rail in these
regions of the system. This leads to the almost constant
stress curve visible in Fig. 10. On the contrary, a constant
Sections 2 and 3 are given in Table 3. In general, the restoring force of absolute value jqj = c Æ 2~u is assumed to
extreme sum is not identical to the sum of the extreme val- act in plastic regions if the influence of the preceding load
ues of the loading cases because the latter do not necessar- case is neglected. The resulting stress distribution varies lin-
ily occur at the same points. However, the minimum values early as illustrated in Fig. 9.
ð1Þ ð2Þ ð3Þ
rmin , rmin , rmin obtained from an independent treatment of Analogous conclusions can be drawn for the final load-
loading cases are deduced at the same point x = 60 m, such ing case ‘braking’, although the differences are less pro-
ð1Þ ð2Þ ð3Þ
that rmin ¼ rmin þ rmin þ rmin is true for Table 2. The nounced here. The symmetry of the stress distribution
results of the independent and combined analysis are visu- obtained from an independent analysis of this situation is
alized in Figs. 9 and 10, respectively. lost when the deformation history is taken into account.
It can be seen that the biggest portion of stress is caused Summarizing it can be said that in the example consid-
by the temperature change DT. The corresponding results ered herein the decreased maximum stresses visible in Table
in Tables 2 and 3 are identical, since this loading case acts 3 and Fig. 10 can mainly be attributed to the reduced stress
on the virgin structure. The biggest tensile and compressive due to bending. In general, the stress distributions obtained
stress occur at approximately x ¼ L3 and the sliding support analysing each loading case separately and using a correct
(B), respectively. A completely unsymmetrical distribution combination differ widely, as can be seen in Fig. 12. Here,
of stress is obtained for both temperature change and bend- the maximum stresses obtained from a truly nonlinear
ing. The latter loading case leads to a maximum tensile and analysis are considerably smaller than the results of the
compressive stress at the supports (A) and (B), respectively, conventional approach. According to the DIN Fachbericht
and thus amplifies the effect of warming up. N
101 [8] a maximum compressive stress rsafe ¼ 92 mm 2 is

However, strong differences in the distribution of stress admissible. Thus, the maximum compressive stress
N
due to bending can be noticed comparing the respective r ¼ 120 mm 2 obtained in the above example using the

curves shown in Figs. 9 and 10. This discrepancy is espe- conventional approach would demand for an expansion
cially pronounced in the vicinity of the sliding support device. On the contrary, the result of a truly nonlinear anal-
N N
(B). In this region the loading case ‘bending’ leads to slip- ysis, r ¼ 87:5 mm 2 > rsafe ¼ 92 mm2 , does not require

Temperature ΔT = +30 K
Bending wmax/l = 1/2500
Braking p = +20000 N/m
50 Superposition
Longitudinal rail stress σ [N/mm ]
2

-50

-100

-20 0 20 40 60 80 100
Coordinate x [m]

Fig. 9. Longitudinal rail stresses due to warming up DT = +30 K, bending and braking p = +20,000 N
m
obtained analysing each loading case separately.
470 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

(1) Temperature ΔT = +30 K


(2) Bending wmax/l = 1/2500 after (1)
Braking p = +20000 N/m after (1) and (2)
50 Temperature -> Bending -> Braking

Longitudinal rail stress σ [N/mm ]


2
0

-50

-100

-20 0 20 40 60 80 100 120


Coordinate x [m]

Fig. 10. Longitudinal rail stresses due to a correct combination of warming up DT = +30 K followed by bending and braking p = +20,000 N
m
in a nonlinear
analysis.

depends on the state of deformation resulting from the


three preceding loading cases. Again, the maximum rail
stress due to the sequence of loading cases is not equal to
the sum of the maximum values corresponding to the single
load steps.
In a conventional independent analysis of the different
situations the additional stresses due to the opposite brak-
ing forces p and p would cancel each other. This is clearly
not the case if the load history is taken into account, as is
impressively shown in Fig. 13. Here, the strong dependancy
on the actual state of deformation is expressed by the com-
pletely different distribution of additional stresses due to p
Fig. 11. Elastic and plastic regions due to warming up DT = +30 K, and p. In contrast to that, an independent treatment of
1
bending wL^ ¼ 2500 and braking p = +20,000 N
m
. (a) Independent analysis, (b) the loading cases would lead to a dashed line in Fig. 13
correct combination. which is a mirror image of the solid line. Fig. 14 shows a
comparison of the total stresses obtained after the
such an expansive system and thus leads to less mainte- sequences
nance effort and cost. However, a local stress increase in
the rails is possible in isolated parts of the track–bridge sys- (a) ‘DT ! bending’,
tem and on the embankment. (b) ‘DT ! bending ! +p’,
(c) ‘DT ! bending ! +p ! p’.

5.2. Relief by opposite braking force p In case of an independent treatment of loading cases, the
solid line in Fig. 14 corresponding to the situation (a)
In this section the action of an opposite braking force would be identical to the dotted line corresponding to the
p following the loading sequence ‘DT ¼ þ30 K ! situation (c). However, the most remarkable result of this
bending ! þp ¼ 20; 000 N m
’ on the example system example based on a correct combination of loading cases
described above is analysed. This is done in order to dem- certainly is the fact that after the opposite action of the
onstrate that two subsequent loading cases of equal magni- two braking forces the total stress is reduced below the
tude but opposite sign do not cancel each other in case of a value obtained due to temperature change and bending.
truly nonlinear analysis. The resulting maximum rail stres- This decrease is observed throughout the complete track–
ses given in row (2) of Table 4 are identical to the values bridge system. However, the maximum compressive stress
given in Table 3, row (3). The additional stress due to the is most affected. Here, a decrease by approximately
relieving braking force p given in row (3) of Table 4 27.5% is obtained.
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 471

Separate analysis of load cases - Summation


Correct combination
50

Longitudinal stress σ [N/mm ]


2 0

-50

-100

-20 0 20 40 60 80 100
Coordinate x [m]

Fig. 12. Longitudinal rail stresses due to warming up DT = +30 K, bending and braking p = +20,000 N
m
. Comparison of approaches.

Table 4 5.3. Parameter study for the sequence ‘DT ! wL^ ! p’


Maximum longitudinal rail stresses due to a correct combination of the
loading cases DT, bending, braking +p and opposite braking force p In Section 5.1 it has been demonstrated that the maximum
 N   N 
Load rmin mm 2 rmax mm 2 longitudinal rail stresses occurring in a selected system differ
(1) DT = +30 K ! Bending 79.972 60.365 considerably, depending on whether a truly nonlinear or a
(2) + p = +20,000 N/m after (1) 7.514 9.527 conventional approach is used for the computation of the
(3)  p = 20,000 N/m after (1) and (2) 16.448 23.97 latter. In the following an attempt is made to generalize the
(1) + (2) + (3) 58.275 45.752 above result by means of a more detailed parameter study.
However, the large number of parameters complicates the
Besides ‘braking’ and ‘inverse braking’ the situations latter. Therefore, a given sequence of loading cases,
‘bending’ and ‘inverse bending’ will be relevant in reality.
These pairs occur in a rather short time. Thus, mass accel- (1) Temperature change DT = +30 K.
1
eration effects should be considered, too. This will be the (2) Bending wL^ ¼ 2500 .
subject of future research. (3) Braking p ¼ þ20; 000 N
m

30
σ due to +p (sequence: ΔT -> bending -> +p)
σ due to -p (sequence: ΔT -> bending -> +p -> -p)
25
Additional longitudinal rail stress σ [N/mm ]
2

20

15

10

-5

-10

-15

-20
-40 -20 0 20 40 60 80 100 120
Coordinate x [m]

Fig. 13. Additional longitudinal rail stresses due to braking in opposite directions taking the load history into account.
472 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

80
Temperature ΔT = +30 K -> Bending
Temperature -> Bending -> Braking +p
60 Temperature -> Bending -> Braking +p -> -p

40

Longitudinal rail stress σ [N/mm ]


2 20

-20

-40

-60

-80

-100
-40 -20 0 20 40 60 80 100 120
Coordinate x [m]

Fig. 14. Total longitudinal rail stresses due to warming up DT = +30 K, bending, braking p = +20,000 N
m
and opposite braking force p.

~x of the box girder. This leads to a fourth order expression


for the unknown parameter ~x. In the parameter study con-
ducted herein, ~x has been computed iteratively. A summary
of results is given in Table 5.
The maximum compressive and tensile stresses present
in track–bridge systems of different length are shown in
Fig. 16. It should be noted that these extreme values do
occur at different locations. In general, smaller compressive
stresses are obtained for smaller spring stiffnesses kA of the
Fig. 15. Assumed box girder cross-section of the bridge deck. left-hand side elastic support. For short- and medium-span
bridges, a smaller support stiffness leads to smaller tensile
stresses also. However, with respect to tension the influence
is considered in the following. The parameters EA, cu, cl, ~u of the stiffness parameter K is negligible for very long
and a are chosen according to Table 1. The length L of the structures.
bridge and the stiffness kA of the elastic support are varied. The maximum compressive and tensile stresses follow-
In this connection, the choice of one and the same pre- ing from a truly nonlinear analysis of the sequence of load-
scribed cross-section for all bridges is not sensible from a ing cases are compared to those obtained using the
constructional point of view. Instead, the height of the conventional approach for an example stiffness K = 5 in
bridge deck and the parameters h0, hu are calculated such Fig. 17. Here, for all bridge lengths smaller compressive
that the chosen maximum deflection occurs due to a given
live load q ¼ 80 kN m
. For this purpose an idealized box gir-
der of variable height as shown in Fig. 15 is assumed. The Table 5
Height of the box girder for bridges of different length L
width of the bridge deck as well as the thicknesses of the
slabs and webs are chosen according to structural Length L [m] ~x [m] h0 [m] hu [m] g ¼ hhu0 h [m]
requirements. 10 0.550 0.467 0.633 0.738 1.100
For given values of the maximum relative displacement 20 0.695 0.527 0.718 0.734 1.245
30 1.425 0.841 1.134 0.742 1.975
^
w
L
, the live load q, the modulus of elasticity E ¼ 34; 000 MN
m2 40 2.204 1.195 1.559 0.767 2.754
and the dynamic magnification factor U = 1.05 the 50 3.035 1.583 2.002 0.791 3.585
moment of inertia I follows from Eq. (63). 60 3.875 1.983 2.442 0.812 4.425
70 4.735 2.398 2.887 0.831 5.285
L 1 L3 80 5.615 2.826 3.339 0.846 6.165
I ¼ Uq    : ð87Þ 90 6.495 3.256 3.789 0.859 7.045
^ 76:8 E
w
100 7.385 3.693 4.242 0.871 7.935
110 8.285 4.136 4.699 0.880 8.835
On the other hand, the moment of inertia I and the dis-
1
tances h0, hu can be formulated as a function of the height
^
w
L ¼ 2500, q = 80 kN MN
m , U = 1.05, E = 34,000 m2 .
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 473

100
K=2
K=5
K = 10
K = 20

Maximum longitudinal rail stress σ [N/mm ]


2
50
Tension

Compression
-50

-100

-150
0 20 40 60 80 100 120
Bridge length L [m]
^
w 1
Fig. 16. Maximum longitudinal compressive and tensile stresses due to a correct combination of warming up DT = +30 K followed by bending L
¼ 2500
and braking p = +20,000 N
m
in a nonlinear analysis.

100
Correct combination
Separate analysis of loading cases
Maximum longitudinal rail stress σ [N/mm ]
2

Tension
50

-50
Compression

-100

-150
0 20 40 60 80 100 120
Bridge length L [m]

Fig. 17. Maximum longitudinal compressive and tensile stresses due to warming up DT = +30 K, bending and braking p = +20,000 N
m
. Comparison of
approaches for K = 5.

stresses are obtained by means of a correct combination of 5.4. Importance of the loading case ‘sudden change of ballast
loading cases. These differences are especially pronounced stiffness’
for long bridges. For example, the maximum compressive
N N
stress is decreased from 123:5 mm 2 to 100 mm 2 for In order to assess the need of a detailed consideration of
L = 110 m, K = 5. This difference corresponds to 23.5% the loading case ‘sudden change of ballast stiffness’ in a
of the stress obtained using a correct combination of load- track–bridge interaction analysis a parameter study for
ing cases. This reduction of stress is of great technological the sequence ‘temperature DT followed by braking p’ has
and economical importance as has been explained in Sec- been evaluated. Here, three different approaches have been
tion 5.1. used:
474 P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475

Table 6 of ballast stiffness from cu to cl’ and braking assuming


Material and loading data used in a parameter study to evaluate the cl (loaded track).
importance of the loading case ‘sudden change of ballast stiffness’
     
DT [K] p N m cl mN2 cu mN2 u~ [m] It should be recalled that only the static case was
+30 +20,000 1 · 107 3 · 107 0.002 addressed in each of the above three analyses. The assumed
material and loading data are summarized in Table 6. The
bridge length L was increased in steps of 10 m. The stiffness
(a) conventional, separate treatment of the loading cases of the elastic support spring kA was varied such that
kA
DT and p assuming a constant distributed ballast stiff- parameters K = 2; 5; 10; 20 ðK ¼ L10 6 Þ were considered.

ness cl (loaded track) for both cases; addition of The obtained maximum compressive rail stresses at the
results, sliding support B are given in Table 7 for K = 10. The
(b) correct combination of two loading cases: DT assum- results corresponding to K = 2; 5; 20 can be found in
ing a distributed ballast stiffness cl (loaded track) on Ref. [2]. The separate nonlinear treatment of the loading
the bridge followed by braking p, cases DT and braking and the subsequent addition of
(c) correct combination of three loading cases: DT results leads to greater rail stresses than a correct combina-
assuming a distributed ballast stiffness cu (unloaded tion of the loading cases, regardless of whether the sudden
track) followed by the loading cases ‘sudden change change of ballast stiffness is taken into account or not.

Table 7  N 
Maximum compressive rail stresses r mm2 at the sliding support B (K = 10)
L [m] (a) Conventional (b) Combination (c) Combination
LCl = rT,l + rp DT ! p with cl DT with cu ! change cu/cl ! p with cl
rT,l rp LCl
10 3.96 5.29 9.25 9.25 9.24
20 14.15 8.74 22.89 22.87 20.57
30 26.17 10.95 36.67 35.57 29.88
40 37.28 12.44 49.72 46.09 36.77
50 47.13 13.50 60.63 55.35 42.24
60 56.00 14.28 70.28 63.69 47.02
70 64.06 14.88 78.94 71.30 51.47
80 71.47 15.35 86.82 78.32 55.72
90 78.32 15.74 94.06 84.83 59.82
100 84.71 16.07 100.8 90.94 63.82
110 90.73 16.34 107.1 96.69 67.72

100
K=2
K=5
90
K = 10
Maximum compressive rail stress σrail [N/mm2]

K = 20 Approach (b) cl
80

70

60
Approach (c) cu -> cl
50

40

30

20

10

0
0 20 40 60 80 100 120
Bridge length L [m]

Fig. 18. Maximum longitudinal compressive rail stresses due to DT = +30 K and braking p = +20,000 N
m
. Influence of the loading case ‘sudden change of
ballast stiffness’.
P. Ruge, C. Birk / Computers and Structures 85 (2007) 458–475 475

Thus, the conventional approach (a) can be seen as a safe Summarizing it can be said that the maximum longitudi-
estimate of the real situation. However, a considerable nal rail stresses calculated using a correct combination of
reduction of maximum rail stresses can be obtained using loading cases are smaller than the corresponding results
the more realistic and thus preferable correct combination of a separate treatment of loading cases. In particular,
of loading cases. Fig. 18 shows a comparison of the maxi- the truly nonlinear simulation proposed in this paper leads
mum compressive rail stresses calculated using approaches to a considerable reduction of the maximum compressive
(b) and (c). Here, safe results are obtained assuming a con- rail stress for medium and large span bridges. This is of
stant stiffness cl of the loaded track for both loading cases great technological and economic importance, since expan-
DT and braking. The numerical treatment of the loading sion devices are not required in certain situations.
sequence ‘DT ! sudden change from cu to cl ! braking’ The results presented in this paper for single-track
is significantly more expensive from a computational point bridges can be applied to double-track systems also. The
of view. However, it can be seen that the proposed static proposed numerical single-track model can be used for
modelling of this phenomenon leads to a considerable the analysis of a double-track bridge with an abutment
reduction of the maximum compressive stress for bridges stiffness kA2 using k A ¼ k2A2 in the calculation.
of medium and great length. Nevertheless, the influence The importance of the phenomenon ‘sudden change of
of the mass acceleration should be considered before finally ballast stiffness when the train reaches the bridge’ has yet
assessing the importance of the loading case ‘sudden to be clarified. In this paper, an exact stiffness formulation
change of ballast stiffness’. This is the subject of current for this loading case has been derived which does not con-
and future research. tain any mass accelerations. This static formulation shows
a considerable reduction of the maximum compressive
6. Conclusions stress for bridges of medium and large span. The dynamic
analysis of the vibrations which are caused by the impulse
For the computation of longitudinal forces in continu- nature of the sudden change of ballast stiffness are the sub-
ously welded rails on railway bridges a coupled track– ject of current research.
bridge analysis is required, since track and structure
interact with each other. In this context, the coupling ele-
References
ment ballast or fastening system in case of a ballasted or
rigid track, respectively, is of great significance. Both situ- [1] Popp K, Kruse H, Kaiser I. Vehicle-track dynamics in the mid-
ations are described by a nonlinear stiffness law which frequency range. Vehicle Syst Dynam 1999;31:423–64.
allows the rail to slip relative to the structure if the relative [2] Ruge P, Trinks C, Muncke M, Schmälzlin G. Längskraftbeanspru-
displacement juR  uBj exceeds a critical value ~ u. chung von durchgehend geschweißten Schienen auf Brücken für
European codes recommend an independent treatment Lastkombinationen. Bautechnik 2004;81:537–48.
[3] Ruge P, Schmälzlin G, Trinks C. Schienenlängskräfte auf Brücken
of the loading cases temperature change, bending of the infolge Biegung. Bautechnik 2005;82:69–80.
supporting structure and braking taking into account the [4] Ruge P, Birk C, Muncke M, Schmälzlin G. Schienenlängskräfte auf
above nonlinear stiffness law and a subsequent summation Brücken bei nichtlinearer Überlagerung der Lastfälle Temperatur,
of the results. Following this conventional approach it is Tragwerksbiegung, Bremsen. Bautechnik 2005;82:818–25.
[5] Union Internationale des Chemins de fer (UIC), 774-3E. Track/
assumed that the coupling element has not experienced
bridge interaction. Recommendations for calculations, Paris,
any deformation prior to the considered loading case. 2001.
Thus, the influence of the loading history is completely [6] European Committee for Standardization (CEN), Brussels, Eurocode
neglected. 1, Part 2 (EN 1991-2). Actions on Structures; Traffic loads on bridges,
In this paper it has been shown that the actual state of 2003.
[7] European Committee for Standardization (CEN), Brussels, EN 1990-
deformation of the coupling element is of crucial impor-
Eurocode. Basis of Structural Design; Appendix 2: Application to
tance for a realistic simulation of longitudinal loading cases bridge design, 2001.
on railway bridges. Depending on the loading history of [8] DIN, DIN-Fachbericht 101. Einwirkungen auf Brücken, Berlin, 2003.
the ballast there is a certain deformation capacity available [9] Müller G, Jovanovic D, Haas P. Tracks–gravel–bridge interaction.
for a subsequent loading case to proceed elastically. Under Comput Struct 1981;13:607–11.
certain circumstances it is possible that no restoring force [10] Ruge P, Pahnke U, Toth J. Dynamic longitudinal behaviour of
railway bridges. In: Structural dynamics – Eurodyn 1999, Prague,
at all is transferred by the coupling element. In this case June 7–10, 1999. p. 657–62.
a remarkable reduction of the longitudinal rail forces in [11] Fryba L. Dynamics of railway bridges. London: Thomas Telford;
comparison with the conventional approach can result. 1996.

You might also like