New Dynamic Analysis Methods For Railway Bridges in Codes Iapf and Eurocode 1

RALWAY BRIDGES
Design, Construction and Maintenance

Spanish group of IABSE
Madrid, 12–14 june 2002
NEW DYNAMIC ANALYSIS METHODS FOR RAILWAY BRIDGES IN

CODES IAPF AND EUROCODE 1
J.M. Goicolea∗ , J. Domínguez∗ , J.A. Navarro∗ and F. Gabaldón∗
∗
Computational Mechanics Group,
Escuela Técnica Superior de Ingenieros de Caminos (ETSICCP),
Universidad Politécnica de Madrid,
Ciudad Universitaria, 28040 Madrid, España
e-mail: [email protected], web: http://w3.mecanica.upm.es
Keywords: Structural dynamics, high speed railways.
Abstract.
Bridges and viaducts for high speed trains are subject to demanding dynamic loads, as
to the classical effect of the moving (single) load the dynamic resonance appears for speeds
above 220 km/h. The classical methods for evaluation of dynamic impact factors available in
engineering, reflected in the codes of practise existing until recently, do not cover this possibility
of resonance. The design of such structures requires dynamic calculations which are the object
of this paper. In it we cover a general revision of available methods for calculation, as well as
a description of the provisions in the new (draft) codes IAPF [13] and Eurocode 1 for actions
on bridges [22].
Additionally, some recent research results obtained by our group are presented for high speed
traffic loads on bridges. The object of these studies is diverse: sensitivity to integration time-
step in modal analysis, simplified torsion analysis, evaluation of the bridge-vehicle interaction
in isostatic bridges and a proposal of a simplified method for dynamic analysis of portal frames.
All of these topics originate from issues in the application of the new regulations for high speed
lines, and are oriented toward being of practical use to designers of railway bridges.
/ J.M. Goicolea , J. Domínguez , J.A. Navarro and F. Gabaldón .
1 INTRODUCTION AND MOTIVATION

The construction of new transport infrastructure has experienced a boom in the last few years
in Spain and in other European countries. The main part of the investment in Spain has been
dedicated to high speed railway lines. This chapter has also been dominant in other neigh-
bouring countries such as France. These new railway lines are a very competitive alternative
for transport between cities at intermediate distances. Currently in Spain there is one line in
operation between Madrid and Seville, in early 2003 the inauguration of the section Madrid-
Lérida is foreseen within the line Madrid-Barcelona-French border. Several other lines have
been decided between Cordoba-Málaga, Madrid-Segovia-Valladolid, the new access Madrid-
Valencia-Murcia and Madrid-Toledo, these being in different states of development (award of
contracts, project or construction).
This important engineering activity highlights one of the main structural aspects associated
specifically to the design of bridges and structures in railway lines: the dynamic effects due
to moving loads from train traffic. The relevance of the dynamic response has been known
since the early stages of railways, having been considered as one of the design requirements for
the structures. This phenomenon propitiated the study of the basic phenomenon of a moving
load on a simply supported beam, whose classical solutions were developed (between others)
by Timoshenko [14]. More recently, the works of Fryba [15, 16] have gathered very diverse
models and features of the dynamics of railway bridges. Finally, one must cite the notable
contributions performed in Spain by Alarcón [1, 2].
The design codes existing up to now [18, 21, 19] for design of railway bridges consider
the dynamic response through an impact factor, which represents the increase in the dynamic
response with respect to the static one for a single moving load.
According to this coefficient, the dynamic increment [18] reaches a maximum value of
0
ϕ = 1.32, for an ideal straight track (without irregularities). The impact factor will be finally
obtained as the envelope
Φ = max(1 + ϕ0 + ϕ00 ),
where this last factor (ϕ00 ) stems from the effect of track irregularities.
As a representative example, consider the case of a point load of 195 kN, corresponding to
an axle of the engine of the high speed train ICE2, crossing a simply supported bridge with
span L = 15 m at a constant speed. The remaining mechanical parameters of the bridge are
the mass per unit length m = 15 t/m, flexural stiffness EI = 7694081 kN/m2 , fundamental
frequency (first mode of vibration) f0 = 5 Hz and damping rate ζ = 2%. This bridge belongs to
a catalogue of isostatic bridges employed by ERRI in [7] for dynamic calculations. The result of
the dynamic analysis at speed v = 220 km/h is shown in figure 1, where the maximum dynamic
deflection is 2.80 mm. Varying the velocity of the load one may perform a sweep in velocities,
for which figure 2 shows the maximum dynamic deflection for each velocity. The absolute
maximum within this sweep is δmax = 3.02 mm for v = 330 km/h. Taking into account that the
static deflection is δest = P L3 /(48EI) = 1.78 mm, this yields a dynamic factor of Φreal = 1.69.
This dynamic factor is covered by the design value prescribed in [18], which for this case results
2
in ΦUIC = 1 + ϕ0 = 2.16 (not considering the effect of track irregularities).

from the previous result one may conclude that consideration of the impact factor Φ is suffi-
cient for taking into account the dynamic effect of a single moving load. Let us consider now an
(ideal) load train, consisting of 10 axles of equal load to that considered above, with a uniform
separation between them of D = 16 m. The (dynamic) response obtained for two velocities of
circulation (v = 288 km/h and v = 360 km/h) is shown in figure 3. Notice that the response
is much higher for the lesser of the above two speeds, indicating that in this case a resonance
phenomenon occurs, which does not increase with the train velocity but rather appears at certain
critical velocities. In figure 4 the result of the sweep in velocities for this case of the load train
is shown, where the maximum corresponds to resonance at a critical velocity of 288 km/h.
The interpretation of this resonant phenomenon is simple: the frequency of application of
the cyclic loads due to the axles for v = 288 km/h, taking into account their uniform spacing,
is fP = v/D = 5 Hz. The coincidence of this excitation frequency with that of the fundamental
vibration mode of the bridge (fP = f0 = 5 Hz) determines the appearance of resonance.
Another (equivalent) manner to interpret resonance is through the so called wavelength of
excitation,
v
λ= .
f0
Resonance occurs when the characteristic length Dk of separation between axles coincides with
a multiple of the said wavelength:
Dk
λ= , i = 1, 2, 3, 4. ⇒ resonance
i
In our case, λ = 16 m, hence the previous condition is clearly fulfilled as it coincides with the
regular distance between axles.
Hence it must be clear that the impact factor Φ does not take into account the possible
resonance that would occur as a result of the cyclic repetition of loads. However, it must be
mentioned that for the bridge frequencies in practise and for the regular distances between
axles of real trains resonance has not occurred in practise. . . until the appearance of high speed
trains!
Indeed, for speeds above 200 or 220 km/h, for the regular distances between axles of current
railway cars (between 13 and 20 m)—which for a given velocity are the factors which determ-
ine the cyclic frequency of the applied loads—resonant phenomena may start to occur. As a
real life example, in figure 5 the measured resonant response obtained in the Tajo viaduct is
shown, for a speed of 219 km/h [3]. In [4] this case is analysed in greater detail, performing the
dynamic calculation with a simple model which nevertheless coincides very well with experi-
mental measurements (figure 6). These resonant measurements would have resulted even more
pronounced for a longer (double composition) train, with a greater number of axles.
The Tajo viaduct in the AVE HS line consists of isostatic simply supported spans of L =
38 m, whose fundamental frequency of f0 = 3.31 Hz. Hence, for the speed given the excitation
wavelength is λ = 18.4 m, very close to the characteristic regular distance between bogies in
the AVE, Dk = 18.7 m.
3
Figure 1. Dynamic response of a simply supported ERRI bridge L = 15 m under a single moving load, P = 195
kN, at a speed of 220 km/h.
Figure 2. Maximum deflection at centre of span as a function of load speed. Single moving load, simply supported
beam, L = 15 m, ζ = 2%.
4
Figure 3. Maximum displacement at centre of span as a function of time for train speeds of v = 288 km/h and
v = 360 km/h.
Figure 4. Maximum displacement at centre of span as a function of train speed for the case of single moving load
and load train.
5
Figure 5. Deflections measured in Tajo viaduct (HS line Madrid-Seville) crossed by AVE train (simple composition)
with v = 219km/h [3].
Figure 6. Displacements obtained with a dynamic model with direct time integration and moving loads for the Tajo
viaduct, for a speed of AVE HS train of v = 219 km/h (simple composition).
6
In summary, resonance in railway bridges, in spite of being a well known feature within
dynamic response of systems, has not been included prior to now in engineering design codes.
As a consequence, it has not been considered in project, except for the margin provided by the
safety factors.
The technical problem posed in structural engineering has far-reaching implications: society
demands the realisation of a great number of structures and bridges for the new high speed lines,
but analysis methods and design provisions which are reliable, practical and sufficiently valid-
ated are not available in the engineering design codes. Some new design codes have recently
started to remedy this lack, between these one may point out the Italian code [20], the new draft
of Eurocode 1 [22] and the new Spanish draft [13]. Eurocode 1
However, in spite of these new proposals which do consider resonant phenomena, there is
still a lack of practical knowledge of the implications of dynamic effects on numerous struc-
tural systems in railway bridges, for which reason in our view an important research effort is
essential. On the other hand, in spite that from a technical viewpoint a number of calculation
models are available, based on linear structural dynamics, it can be said that these methods are
not known well enough at this moment by engineers in charge of design and project.
The simplest models with regard to their application are those based on the decomposition as
a harmonic series of the dynamic response, and the consequent establishment of upper bounds
of these sums through analytic procedures [7, 4]. The drawback associated to these models is
that their application is restricted to isostatic structures (e.g. simply supported decks in which
dynamic response may be adequately defined by one mode of vibration). They cannot be applied
(directly at least) to statically redundant structures. This type of models is described briefly in
section 2.4. In the report [12], described in section 7, an analogy is proposed which allows
to extend the above models to certain redundant structures such as portal frames in railway
underpasses.
The next type of models available for analysis are those based on direct dynamic calculation,
with time integration of the response for each degree of freedom, for a set of moving loads
representative of the axles of the train [4]. These procedures may be implemented within finite
element codes, for which the data preparation (preprocessing) stage for the moving loads may
be the the most cumbersome. These models are discussed in section 2.2.1, and with them the
dynamic analysis of arbitrary structures may be performed. In certain cases of simple structures
(isostatic beams, or continuous beams of two or three vanes) it is possible to apply this procedure
also through an analytical exact definition of the modes.
Finally the most complete models are those which consider jointly the vibration of the struc-
ture as well as that of the vehicle. The latter is taken into account through springs and dampers
for the suspensions and the masses and connections provided by the vehicle boxes. Some of
these models with vehicle–structure interaction are described in section 2.3, having been dis-
cussed in greater detail in [4] and references there cited. Obviously, these models are also apt to
analyse arbitrary structures, provided the dynamic characteristics of the real trains are known,
something which unfortunately is not always the case. This possibility has the drawback of
greater calculation costs and an increased complexity in the definition of the model. From the
7
point of view of research they are interesting or even often indispensable, however it is not
reasonable to apply these methods for standard design calculations.
The different engineering design codes prescribe the obligation of specific dynamic analyses
when the simplified methodologies are not applicable, a case which is increasingly frequent
in structures within high speed lines. Thus, with the view put in obtaining practical and suf-
ficiently validated methods for calculation which facilitate the work of engineers designing
railway bridges as well as those responsible for maintenance, several calculation procedures,
which complement existing methods, have been proposed recently. These have been employed
mostly in the academic research community, the mathematical models involved often exhibit an
excessive complexity and calculation cost for project stages. The need for obtaining simplified
methods has hence been a goal of engineering research groups, in order to provide calculation
tools which are sufficiently precise and contribute to a better comprehension of the dynamic
behaviour of structures. The practical studies and applications developed by the authors which
are presented in the final sections of this paper follow this line of motivation. These studies
focus on the sensitivity to time-step in modal analysis, consideration of torsion, evaluation of
vehicle-structure interaction in isostatic bridges, and finally a proposal for a simplified dynamic
calculation method for portal frame underpasses.
The consideration of the vehicle-structure interaction models discussed in section 2.3 pro-
duces a reduction of the effects due to the existence of mechanisms which permit energy dissip-
ation (dampers) or systems which interchange energy between structure and vehicle (suspension
springs). For non resonant situations or statically redundant bridges, the interaction effects are
not usually relevant in the calculation, being sufficient to consider constant load models. How-
ever, for isostatic decks with short spans (10 m - 30 m), significant resonant effects appear with
high accelerations, and often these constant load models yield results above the design limits.
With the vehicle–structure interaction models an effective reduction of these results may
be obtained. The problem with these models is that they are often excessively complex for
their application in project stages. The work described in section 6 quantifies numerically the
reduction obtained in the dynamic response of isostatic bridges as a result of the application of
the models explained in section 2.3.
The calculation procedures included in the more recent engineering design codes are simple
methods applicable for isostatic structures, such as simply supported bridges, for which in prac-
tise a single mode of vibration may be considered in the response. For these cases a direct
dynamic calculation may be avoided, through the use of analytic envelopes for such effects
(section 2.4). However, for statically redundant structures, such as continuous deck viaducts, a
direct dynamic analysis is needed, as the response includes a contribution from several vibration
modes.
The railway underpasses (portal frames, vaults) are also in this latter category, as in general
the deck has a statically redundant support, hence the more simplified methods for evaluation
of dynamic effects are not applicable. Other aspect which complicate a correct model of the
dynamic response are the possible earth cover between the deck and the ballast bed, the vibra-
tion transmitted to the earth fill in contact with the lateral walls or piers, etc. All the above
8
contributes to an undesirable paradox: the simplest structures, for which in practise significant
resonance effects have not been observed, are the ones which a priori require a greater effort
for calculation and correct evaluation of dynamic effects. The analysis of the response of portal
frame underpasses is described in section 7, proposing a calculation method which adequately
covers the observed dynamic response and may be evaluated through the simplified procedures
for isostatic beams.
A code of good practise before performing a final time integration is to perform a sensitivity
analysis with regard to the time-step. Some such considerations are discussed in general terms
in section 8, regarding the recommendations which appear both in engineering codes and in the
technical literature.
A simplified method for evaluating torsion effects in dynamic studies of railway bridges is
proposed in [8]. According to this method, a conservative envelope results from linear super-
position of the effects associated to bending and torsion, each of them analysed separately. In
section 9 we present an evaluation of the fit of this simplified method.
In the first part of this paper we present a description of the basic features of the calculation
methods available for dynamic analysis of railway bridges subject to traffic loads. Following a
summary of the methods prescribed in the new drafts of codes IAPF and Eurocode 1 is done.
Finally some research results for specific problems obtained by our group are presented.
2 ANALYSIS METHODS
2.1 Impact factor Φ
The basic method followed up to now in the existing engineering codes for design of railway
bridges has been that of the impact factor, generally called Φ. As has been previously discussed
in section 1, such coefficient represents the dynamic effect of (single) moving loads, but not
resonance.
The general expressions contained in the codes attempt to offer simple formulae which serve
as an envelope for the range of train speeds, vibration frequencies of structures and train types.
2.1.1 Code IAPF-75

The existing Spanish code IAPF-75 [19] defines a “dynamic increment”, expressed as a per-
centage (%), for which it prescribes the following values, for simply supported decks:

0.33v,
 span L < 6 m;
I= √
114 L

 √ , span L ≥ 6 m.
3.10 − 1.76 L + L
In this expression v is the train speed in km/h. The dynamic increment so defined gives rise to
an impact factor Φ = 1 + I/100 which is applied to the static effects of certain nominal train
types, called “train type A” and “train type B”, which include point loads of 30 t and distributed
loads of up to 12 t/m.
9
For continuous beams the proposed expression is

µ vT
I = 65 2
; µ= , (1)
1−µ+µ 2L
where T is the fundamental period of vibration. In this formula the train speed v must be
expressed in m/s (there is an typographical error in the published code, where it says erroneously
km/h).
The range of speeds for application of this coefficient is v ≤ 200 km/h.
2.1.2 Code UIC-776-1R

In this code [18], published after the Spanish proposal of IAPF-75, an impact factor is proposed
associated to the static effects of a nominal train type (UIC71), which has been adopted by the
majority of codes in the different countries, as well as in the Eurocode currently in effect [21].
The train type (UIC71) considers concentrated loads of 250 kN, as well as distributed loads of
80 kN/m. Note that the values of this train type are smaller than the one defined by the Spanish
code IAPF-75 [19], detailed in the previous subsection.
The impact factor, for well maintained tracks, and for applying to bending moments, is
1.44
Φ2 = √ + 0.82; Φ2 ≥ 1. (2)
LΦ − 0.2
The equivalent span LΦ coincides with the real one for a simply supported isostatic element,
and an equivalence table is provided for other structural types.
This impact factor is applied to the static values of the effects obtained from train UIC71:
ΦSest,tipo ≥ Sdin,real , (3)
where Sest,tipo and Sdin,real are, respectively, the effects corresponding to the nominal train type
under static conditions and to each real train under dynamic conditions.
The above value of Φ results from obtaining a dynamic envelope for all the real trains. The
interpretation of this envelope is as follows:º
ΦSest,tipo ≥ (1 + ϕ0 + ϕ00 )Sest,real . (4)
It’s important to remark that impact factor Φ is applied to the effects obtained for the nominal
train type, whereas the factors (1 + ϕ0 + ϕ00 ) are applied to the real trains, normally much lighter
than the nominal train type.
The value of ϕ0 corresponds to the dynamic increment itself, for the train on an ideal track
without irregularities, and is expressed as:
K v
ϕ0 = 4
; K= . (5)
1−K +K 2LΦ f0
(Note that in this formula K coincides with parameter µ of expression (1) within code IAPF-
75, and that both expressions are very similar, with the only difference of the last term in the
denominator, in one case a fourth power and a second power in the other).
10
The value of ϕ00 arises from the effect of track irregularities, with the following value:
2 L 2
f0 LΦ

L
00 − 10Φ − Φ
ϕ = a 0.56e + 0.50 − 1 e 20 (6)
80
v

where a = min 22 , 1 , with velocity v expressed in m/s.
The application of impact factor Φ is subject to some conditions, which ensure that it cor-
responds to the real scenarios of bridges and trains for which it was formulated and its validity
checked. Specifically, the fundamental frequency of vibration of the bridge must be within the
limits of a band, defined with respect to the span of the bridge in figure 7.
2
10
Frecuencia natural [Hz]
1
10
0
10
1 2
10 10
Luz [m]
Figure 7. Bounds for the natural frequency f0 in Hz, in relation to the span of the element, for application of
impact coefficient Φ
The coefficient Φ so defined does not take into account resonant effects. With the object of
avoiding this possibility, most engineering design codes that include it (e.g. [21]) limit its use
to velocities v ≤ 200 km/h.
2.2 Dynamic analysis with moving loads

As has been said above, the method of the impact factor, as a counterpart to its simplicity,
has a number of limitations. The main one is that, not taking into account resonance, it is not
applicable for high speeds (v > 200 km/h). In these cases one may perform a dynamic analysis
with moving loads.
These methods are based on time integration of the dynamic equations for the structure,
when subject to a series of moving loads of fixed values, representative of each axle of a given
train. The model for the structure may be analysed either through an integration of the com-
plete system with N degrees of freedom, or through a reduction of degrees of freedom from a
modal analysis which reduces substantially the number of equations to integrate. This modal
11
reduction may be performed through an approximate numerical procedure to obtain the eigen-
modes of vibration, a capability which is provided by the majority of finite element programs.
Alternatively this may be achieved through an analytical (closed form) calculation, for certain
cases of simple structures.
2.2.1 Analytical methods

The classical problem of a simply supported isostatic bridge may be treated through the exact
eigen modes of vibration, which correspond to the hypotheses of the Bernoulli beam [9], for
whichp the modal shapes are φn (x) = sin(nπx/l) and the associated eigenfrequencies ωn =
2
(nπ) EI/(ml4 ). Figure 8 shows the first three vibration modes for this case. Generally,
x
r
φ1 (x) = sin(πx/L) ω1 = π 2 EI
ρL4
M1 = 12 ρL
r
φ2 (x) = sin(2πx/L) ω2 = 4π 2 EI
ρL4
M2 = 12 ρL
r
φ3 (x) = sin(3πx/L) ω3 = 9π 2 EI
ρL4
M3 = 12 ρL
Figure 8. First three vibration modes for an isostatic simply supported beam
for an isostatic case, it is enough to consider a single vibration mode; this way the problem is
reduced to a dynamic equation with one degree of freedom, whose solution and interpretation
is much simpler than other cases with multiple degrees of freedom.
For more complex statically redundant structures it is not possible in general to perform an
analytical extraction of vibration modes and frequencies. Nevertheless, analytical closed-form
solutions may be obtained for some specific cases, such as (intraslational) portal frames and
continuous beams with two or three spans [15]. For rectangular portal frames the procedure
detailed in [12] is slightly more complex than for the simply supported bridge. For example,
the two first vibration modes are shown in figure 9. The expression for the frequency associated
12
to the first mode is defined through a parameter b with the equation:

2 r
b EI
ω1 = (7)
ld md
where ld is the span of the deck, Ed Id its bending stiffness and md the mass per unit length.
Parameter b is obtained as the solution of the following nonlinear equation:
kp (1 − cosh(ki b) cos(ki b)) 1 − cosh b cos b
+ =0 (8)
cosh(ki b) sin(ki b) − sinh(ki b) cos(ki b) (cosh b + 1) sin b − (cos b + 1) sinh b
being: s r
4 Id3 md lh 4 Id mh
kp = ki = (9)
Ih3 mh ld Ih md
In this equation, subindex h refers to the piers or vertical walls of the portal frame.
b=3.2491 b=4.3363
Figure 9. Two first vibration modes of a portal frame corresponding to an underpass of a new high speed railway
line, with the value of parameter b for calculation of eigenfrequencies through equation (7)
Once the vibration modes are known, it is necessary to integrate the dynamic equations.
For this, the basic solution is the response of the structure to a single moving load (figure 10).
Consider a continuous beam of length l, being φi (x), Mi and ωi respectively the modal shape,
the modal mass and the eigen frequency of the i-th mode. The differential equation for a point
load F crossing the beam at a constant speed v is:
Mi ÿi + 2ζi ωi Mi ẏi + ωi2 Mi yi = F hφi (vt)i (10)
where yi is the modal amplitude of the i-th mode, ζi the damping fraction with respect to the
critical value, and hφ(•)i represents a bracket notation with the following meaning:
(
φ(x) if 0 < x < l
hφ(x)i = (11)
0 otherwise.
After obtaining the response for a single moving load, the response for a load train may
be assembled as the superposition of the responses for the point loads Fk (figure 11). The
13
v
F
vt y(t)
L
Figure 10. Response for a single moving load
differential equation corresponding to mode i is in this case:

nejes
X
Mi ÿi + 2ζi ωi Mi ẏi + ωi2 Mi yi = Fk hφi (vt − dk )i. (12)
k=1
Fk Fk−1 F4 F3 F2 F1
d1
dk−1
L
Figure 11. Response for a load train
2.2.2 Finite element methods

Dynamic analysis of railway bridges based on moving load models may also be performed
through finite element methods. These methods are applicable generally to arbitrary structures,
and may include if necessary nonlinear effects.
A spatial discretisation of the structure is performed into subdomains called finite elements,
obtaining an approximate model with a discrete number of degrees of freedom N , followed
by a time discretisation in time-steps. The analysis may then be carried out by direct time
integration of the complete model, or alternatively through modal reduction. In both cases the
basic problem to be solved is the system of differential equations:
Md̈ + Cḋ + Kd = f , (13)
where M is the mass matrix, C the damping matrix, K the stiffness matrix, f the external load
vector, and d the vector of (unknown) nodal displacements.
By means of the direct integration of the model, the complete system (13) of N degrees of
freedom would be solved for each time step; the equations are generally coupled, and there-
fore must be solved simultaneously. This procedure is also valid when nonlinear effects are
14
to be included in the response; in this case the elastic internal forces and viscous damping
forces from the previous expression should be replaced by a general term (nonlinear) of the
type Fint (d, ḋ, . . .).
If the structural behaviour is linear, a modal analysis may be performed, with a remarkable
reduction of degrees of freedom. In a first stage, the eigenvalue problem is solved, obtaining
numerically the more significant n eigenfrequencies and associated normal vibration modes
(generally n N ). Afterwards, these vibration modes are integrated in time. This way the
equations become uncoupled, and the modal response of each mode is reduced to the dynamic
equation of a system with a single degree of freedom [9].
The simplest procedure to model load trains is applying load histories in each node. At
a certain time-step, a load is assigned to each node if the load axis is above an element that
contains the node. The magnitude of the nodal load depends on the distance from the axis to
the node. This procedure is outlined in the figure 12 for a generic node A.
F F F
v v v
A−1 A A+1 A−1 A A+1 A−1 A A+1

t = t1 t = t2 t = t3
fA
F
t1 t2 t3 t
Figure 12. Nodal force definition for node A for a single moving load F .
This scheme is applied to the real trains defined in code [13], and has been implemented in
finite element program FEAP [10]. The results described in the report [12] have been obtained
with this methodology and time integration of the vibration modes.
2.3 Dynamic analysis with vehicle-structure interaction

The dynamic analysis with vehicle-structure interaction consists, like the analysis with moving
loads, of a direct time integration of the dynamic equations of the structure jointly with the
vehicle vibration due to the its own suspension; thus, the axis loads do not have in fact a fixed
value during the crossing of the bridge.
This type of model represents, in the most general case (figure 13), the primary suspension
of each axle with the values of stiffness and damping (Kp , cp ), the secondary suspension, with
the corresponding values of bogie stiffness and damping, (Ks , cs ), the non-suspended mass,
corresponding to the nominal mass of the axis of the wheel (mw ), the length, mass and moment
15
of inertia of the bogie (LB , mb , jb ), the suspended mass and moment of inertia that corresponds
to the box of the vehicle (M, j) and the geometry of the vehicle: total length (L), distance
between the centre of gravity of the box of the vehicle and the front and rear axis (dBd , dbt ), and
the distance between the axis of a bogie (deB ). For those vehicles whose guidance system is not
dtd dBd
M, J

MB , JB

MB , JB

LB deB
L
Figure 13. Complete vehicle–structure interaction model.
accomplished by bogies, the previous scheme should be adapted to the particular configuration
of the axes and the suspension system, with equivalent level of detail.
The level of detail in the interaction models described above is not always necessary. Simpli-
fied models of vehicle–structure interaction may also be employed, where each axis suspension
is modelled independently from the others, without considering the effect of connection with
the vehicle box. This way, they consider (figure 14) primary axis suspension with its corres-
ponding values of stiffness and damping (Kp , cp ), non-suspended mass corresponding to the
nominal mass of axis wheel plus the proportional part of totally suspended mass (vehicle box)
(mns ) 1 , and the suspended mass; in this case, its value is equivalent to the proportional part
of the bogie mass (ms ). Another variant equivalent to this model is proposed in the new UIC
776-2 leaflet[8]; this model is represented in the figure 14.
It is important to mention that in simplified interaction models each axis is independent
from the rest —this means that there is no interaction between the axes of a same vehicle—,
whereas in the complete models there exists certain interaction among them, because the model
represents the complete vehicle box.
Proposed interaction model.— This model has been implemented in [5], a computer applic-
ation which has been used for the research work reported in this article. It considers a train of
k loads, representing each axle according to a simplified vehicle–structure interaction model
(figure 15).
1
Remark that even though mns (simplified model) and mw (complete model) are referred by the same words
16
ms
masa suspendida F fuerza aplicada
ms masa suspendida
Kp Cp
Kp Cp

mmasa
no suspendida

ns

Figure 14. Simplified vehicle–structure interaction model (left). Variant to the proposed model in the UIC-776-2
code [8] (right)
y k (t) y 2 (t) y 1 (t)

Pn
w(x, t) = i=1 qi (t) · φi (x)
mja
kj cj
&'

$ ! "# $%

mjs dk d1 = 0
x
mj = mja + mjs d2
(a) (b)
Figure 15. Crossing of a train of loads, according to the vehicle–structure interaction simplified model: a) inter-
action element; b) geometric definition of variables
17
When analysing for a load train, the number of differential equations to solve is increased; for
a single axle, the number of equations considered are vibration modes n plus the corresponding
one to the mechanical system of the simplified interaction element, altogether n + 1. Supposing
a group of k loads, a system of n + k differential equations is obtained.
The equations corresponding to the vibration modes of the bridge differ only in the term
which represents of the modal load for each time step, for which it will be necessary to calculate
the axles which are on the bridge and the value of the amplitude corresponding to the position
of these.
For a general case, the following equations are obtained for the model:
• For each vibration mode (i = 1 . . . n):
k
X
hφi (djrel )i g mj + mja ÿ j

Mi q̈i + Ci q̇i + Ki qi = (14)
j=1
• For each interaction element (j = 1 . . . k):

n
" n n
#
X j
X j
X j
mja ÿ j + k j [y j − qi hφi (drel )i] + cj ẏ j − q̇i hφi (drel )i − qi vhφ0i (drel )i = 0
i=1 i=1 i=1
(15)
In equations (14) and (15) the notation hφ(•)i has been employed, defined previously in
equation (11). Additionally, djrel represents the relative position on the bridge for each element
j. Considering the initial time t = 0 when the head of the train enters the bridge (x = 0), djrel is
obtained as:
djrel = vt − dj (16)
Considering the nature of the resulting equations, (system of linear second order ordinary
differential equations), the trapezoidal rule is recommended for its integration. The trapezoidal
rule is a member of the β-Newmark family, defined by β = 1/4 and γ = 1/2. In [4] this and
some other aspects of the implementation of integration models in dynamic analysis of railroad
bridges are discussed.
2.4 Models based on harmonic series

This type of models avoid a direct dynamic analysis by time integration. In counterpart its
application is limited to simply supported bridges, which can be represented dynamically by
means of a single harmonic vibration mode.
The different models of this type all develop the response as a combination of harmonic
series, and establish an upper bound of this sum. The different available models of this type also
introduce another concept, of special relevance in the intuitive interpretation of the response:
dynamic signature associated to a given train of loads.
—non-suspended mass—, their values are obtained from different concepts.
18
As its name indicates, the dynamic signature of a train may be understood as a function
which characterises its aggressiveness in relation to the dynamic effects produced in a railroad
bridge. The models of this type proposed are:
• DER: Based en the Decomposition of the Resonance Excitation.
• LIR: Simplified method based on the Residual Influence Line.
• IDP: Simplified method based on the Proportional Dynamic signature.
The methods DER and LIR originate within the group of experts of the European Railway
Research Institute (ERRI), in the D214 committee on railroad bridges for speeds greater than
200 km/h [7]. The method IDP has been developed in [4] and [23].
All these models have their application limited to isostatic bridges; it is considered that the
dynamic response of this kind of bridges may be represented by only the first bending vibration
mode of the structure. In addition, the first mode shape is an harmonic function, a fact which
facilitates the analytical development of the series.
The DER method originates from the decomposition of the dynamic response of the bridge
in Fourier series, and focuses on the study of the term that corresponds to the resonance in
frequencies. This way, it obtains an upper bound of the maximum acceleration in the centre of
the beam as a product of two functions: the first one characterises the response of the bridge
and the second one is the so called dynamic signature of the train.
The mathematical development of LIR method is based on the analysis of the free vibrations
produced after each individual single load crosses a simply supported bridge, according to the
dynamic analysis of a beam put under the action of successive single loads.
Proposal IDP is centred in the study of residual maximum acceleration of each single load,
obtaining a slightly improved interpretation of the dynamic signature than LIR method. In
particular, it considers the damping that takes place since an axis enters the bridge until it leaves
it with a state of residual vibration.
All these methods end up limiting analytically the maximum dynamic response, in terms
of acceleration or displacement at a given point, as a product of three terms. In this triple
product the contribution from the structure and from the aggressiveness of the circulating train
are clearly differentiated .
Take as an example the LIR method proposed for the maximum acceleration. This value at
the centre of the span, Γ, is obtained as a product of the following factors:
Γ = Cacel · A(K) · G(λ), (17)
where Cacel = 1/M is a constant (the inverse of the total mass of the isostatic bridge),
λ = v/f0 (wavelength), with v the circulation speed and f0 the eigen frequency (Hertz) of the
first vibration mode, and K = λ/(2l), being l the span of the simply supported bridge. The
19
other terms have the following definition:

r
K π
−2ζ K
π π
A(K) = 2
e + 1 + 2 cos e−ζ K (18)
1−K K
v" #2 " x #2
u xi i
N t X
u X
G(λ) = max Fi cos (2πδi ) e−2πζδi + Fi sin (2πδi ) e−2πζδi (19)
i=1
x1 x1
In these expressions ζ is the damping rate, xi are the distances of each one of the N load axes
Fi to the first axis of the composition, and δi = (xi − x1 )/λ.
The term G(λ) (equation (19)) is the previously dynamic signature referred to above. It
depends only on the distribution of the train axles and the damping rates. Each load train has its
own dynamic signature, which is independent of the mechanical characteristics of the bridges.
The figure 16 represents the dynamic signature of train ICE2, for different values of damping.
Figure 16. Dynamic signature of train ICE2 with different damping rates.
A(K) is a function determined for each particular bridge; it depends on the span of the bridge
l, its natural frequency (f0 ), damping (ζ) and the range of speeds of circulation (v) under study.
This function of parameter r is called the bridge dynamic influence line.
Taking the three parameters considered, neither Cacel nor A(K) depend on the characteristics
of the train. Separating the contributions from the bridge and those from the train (G(λ), dy-
namic signature), it is possible to determine easily the critical parameters of span and wavelength
20
(proportional to the train circulation speed, v) that maximise the response of the bridge.
3 HIGH-SPEED REAL TRAINS AND INTEROPERABILITY

A characteristic of resonance is that its occurrence for a certain bridge depends mainly on the
circulation speed (critical resonance speeds) and the type of train.
With relation to the circulation speeds, dynamic analysis sweeps should be undertaken as
a general rule with speed increments of 2, 5 or 5 km/h (never greater than 10 km/h). For the
frequent speeds of circulation, this sweep should be refined.
With respect to the train types, although initially a certain rail line may be operated by one
or more given high speed train types, it seems clear that the trains which can circulate should
not be restricted. This would allow to modify the operation conditions and the high speed trains
that may circulate; furthermore, it opens the possibility of interconnection with other European
high speed lines, permitting all the European trains to circulate along the lines that have been
developed with interoperability criteria.
In order to guarantee this capability, the dynamic analysis of a certain bridge or viaduct
should be performed made for all the current and future types of real trains.
Another possibility that could be more advantageous is to establish interoperability condi-
tions which cover the characteristics of the existing real trains and the anticipated evolutions;
thus, these criteria could be imposed to the new trains to be developed from now on. Given
these conditions, it should be able to establish universal trains which are dynamic envelopes of
the effects of the possible real trains.
3.1 Real trains

European high speed real trains are of three different types (figure 17):
1. Conventional trains: each passenger car has two bogies, with two axes each one. Of this
type are the trains Ice2, Etr-y, Virgin.
2. Articulated trains: the passengers cars have one bogie of two axles in each end, shared
with the adjacent car and centred under the joint between them. This type includes the
Thalys and Eurostar. The Ave train is similar to the Thalys.
3. Regular trains: the passenger cars are also articulated, but this joint is not supported on a
shared bogie, but on a single axis between them. It is the case of Talgo AV.
The new code IAPF [13] details the compositions of each one of these types of trains. Know-
ing these, it is possible to obtain the dynamic signature of each one of the real trains (figure 18),
and from these, to obtain a signature envelope (figure 19). This envelope could be used in the
dynamic analysis, with a clear benefit as to simplification of calculations.
3.2 High-speed interoperability load model (HSLM)

The following objections may be raised with regards to the previous description of a reference
envelope study (figure 19):
21
Figure 17. Different types of high-speed trains, according to Eurocode 1 [22]
22
Figure 18. Dynamic signature (undamped) for European high-speed trains.
Figure 19. Envelope of the dynamic signatures of the European high-speed trains; structural damping rate ζ = 0%
.
23
• Generality: The obtained envelope is not general enough as to be proposed to rule out
designs in future trains: a small variation of parameters of a present train — say the length
of the vehicles, the nominal value of the load by axis or the distance of connections— could
modify its original signature, and it would subsequently not be covered by the envelope.
Therefore, it is possible to conclude that adopting the reference envelope thus defined is
too restrictive;
• Analysis procedure: The envelope of the dynamic signature of the real trains defines a
reference aggressiveness for the construction of new trains, but it does not propose a spe-
cific method of calculation associated to this envelope. This fact represents a drawback
in the envelope proposal, since it does not solve one of the principles of the rail lines in-
teroperability: to provide a simplified methodology of analysis associated to the reference
dynamic envelope.
In order to allow a more general analysis method, not subjected to these disadvantages,
committee ERRI D214 [26] has defined analytically a family of universal trains whose dynamic
effects on the structures covers any real train that circulates at the present (and future) time.
This family —called UNIV-A— has the characteristics gathered in the table 1.
UNIV-A
Type articulated
Total length ' 400 m
Cars lengths D from 18 to 27 m
Axis load 170 kN
Distance between axes of the same bogie 2.5 m
Head and tail locomotives yes
Table 1. Characteristics of UNIV-A universal trains. Font: Technical report of ERRI D214 Committee [26]
From its definition, the universal trains constitute a family that depends on the vehicle length
D. That is, each value of D from 1 which is within the range proposed in the table corresponds
to a member of this family. The set of trains that are generated by variation of parameter D
constitutes the envelope of the universal trains UNIV-A.
The envelope of the UNIV-A trains is shown in figure 20, based on the adopted discretisation
of D. Further, figure 21 represents, for each value of the wavelength λ, the length of the car
D that corresponds to the critical signature in that point, the one that defines the value of the
envelope.
3.2.1 UNIV-A and the Talgo AV

In order to obtain an envelope dynamic signatures which would cover the dynamic effects of all
the real trains, [26] proposed the envelope of the UNIV-A trains; however, figure 22 shows that
this signature does not cover the regular train Talgo AV for wavelengths smaller than 15 meters
24
Figure 20. Dynamic signature of the UNIV-A trains: obtaining the envelope for different vehicle lengths D.
Figure 21. Relation λ − D for the UNIV-A trains envelope. Damping rate ζ = 0%.
25
(λ < 15). A similar result would be obtained for other real regular trains, since Talgo AV is a
representative example of such trains.
Figure 22. Dynamic signatures of the UNIV-A and the Talgo AV envelope.
In order to complete the envelope with regard to regular trains, a study was carried out which
concluded with a proposal to introduce some modifications to the definition of the UNIV-A
family, so that for short wavelengths certain additional trains would be considered to cover
the envelope of such a regular trains like the Talgo AV [27]. These proposals were taken into
account and have given rise to the final definition of the High Speed Load Model (HSLM),
family of trains contained in the last and recent version (final draft) of Eurocode 1 of actions in
railroad bridges [22].
4 DYNAMIC ANALYSIS IN THE SPANISH CODE FOR ACTIONS ON RAILWAY

BRIDGES (INSTRUCCIÓN DE ACCIONES EN PUENTES DE FERROCARRIL,
IAPF)
Taking into account the above discussion the basic provisions for considering the dynamic re-
sponse of bridges in the new draft code on actions in railroad bridges (IAPF) are defined next.
[13].
4.1 Vertical traffic load model

The load model adopted for vertical actions of railway traffic is the UIC71 —also named LM71
—, which constitutes a standard in practically all the countries of our surroundings and in
26
Eurocode 1 [21, 22]. Further, the use of a classification factor α = 1.21 for international and
RENFE rail lines is prescribed. For metric (narrow gauge) rail lines, α = 0.91 has been adopted.
This factor α in fact increases the loads of model LM71, the concentrated loads of 250 kN
become 302 kN, and the distributed loads of 80 kn/m become 96.8 kn/m. These values are
similar to those considered until now in the existing Spanish code for design of railroad bridges
IAPF-75 [19] (see also section 2.1.1).
The main reasons that have led the technical commission to prescribe the factor α = 1.21
are the following:
• it has been verified that it leads to broadly equivalent effects as the existing load models
from the IAPF-75 [19];
• it preserves the capacity of the railway network; otherwise, using the new model LM71
without α the capacity would have been reduced and the network performance decreased;
• the cost of introducing this factor α = 1.21 has been evaluated, resulting in extremely
small increases in relation to the complete structure.
• With this increase it suffices consider only one load model, additional trains SW/0 (for
continuous bridges), SW/2 (heavy traffic) [22] are not necessary;
• it allows to extend the validity of the impact factor Φ up to 220 km/h, as the margin obtained
by multiplying the static loads of the load model, with respect to the static loads of the real
trains is greater and can accommodate moderate dynamic increases.
4.2 Envelope dynamic factor Φ

The basic result of the evaluation of the dynamic effects is the impact factor Φ. It is defined as
the envelope of the effect under consideration:
Sdin,real
Φ = max
Sest,tipo
The relevance of this definition is that the same concept of envelope impact factor is applicable
also to the cases where a dynamic analysis is carried out (here Sdin,real will be calculated), whose
result will also be synthesised as an impact factor.
The direct formula of the impact factor is the same one included in Eurocode which origin-
ates from [18],
1.44
Φ= √ + 0.82,
LΦ − 0.2
as was described in section 2.1.2 (equation (2) for tracks with good maintenance). The limita-
tions for using this factor are:
1. v ≤ 220 km/h;
2. conventional structure (that is, included in the corresponding tables for equivalent span
LΦ );
3. fundamental frequency of the bridge f0 within the limits of figure 7.
27
In any other case, when one or more of the previous conditions are not fulfilled, it will be ne-
cessary to perform a more detailed dynamic analysis; the methodology is described in appendix
B of [13]. These procedures are summarised in the following subsections.
4.3 Dynamic analysis procedures

For the cases when it is not possible to use the envelope factor Φ, several dynamic analysis
procedures are established. These procedures (except for the one mentioned in first place, which
is only applicable for v ≤ f or220, km/h) will be applied either for high speed real trains
(the axle loads are defined in the code), or for the universal load model HSLM which is also
included; this HSLM model allows to guarantee interoperability of the rail lines. It consists of
a family of 10 HSLM-A trains, except for short spans L < 7 m, for which a special family
HSLM-B needs to be applied.
As the critical resonance speed is unknown beforehand, the calculations need to be carried
out as sweep, analysing all the cases between v = 220 km/h and the maximum velocity of the
line, increased by factor 1.2 as a safety margin. A large number of dynamic analyses may hence
be necessary, specially if it is desired to evaluate different structural hypotheses; for this reason
this dynamic analysis work should be planned carefully.
1. Real impact factor.
This procedure boils down to evaluating more precisely the individual dynamic factors ϕ0
and ϕ00 , according to the formulae (4), (5) and (6).
This method may be used for rail lines with circulation speed v ≤ 220 km/h, when some
of the other specified requirements for the factor Φ are not fulfilled, for example when the
structural fundamental frequency is not within the limits of figure 7. For the analysis of
factor ϕ0 a set of trains defined specifically for speeds v ≤ 220 km/h should be used.
2. Dynamic signature (DER, LIR)
This procedure allows to evaluate the dynamic effects without making a dynamic analysis
with time integration, as has been described previously in section 2.4. It suffices to evalu-
ate the expressions indicated in the formulae (17), (18), (19); even though at first sight the
expressions seem somewhat complex, they can be programmed very easily in any com-
puter spreadsheet. The analysis method is considerably simpler than a direct time dynamic
analysis.
This method has the disadvantage that it is only applicable for isostatic bridges; this is an
important limitation for many viaducts and real structures that do not fulfil this condition.
3. Dynamic analysis by direct time integration with moving loads
For the general cases of bridges which are not simply supported beams, a dynamic analysis
with moving loads and the consideration of the real structural model (or modes of vibra-
tion) must be carried out. as has been already described above in section 2.2. It is possible
to use either direct analytical procedures for the extraction of the vibration modes (in the
simplest cases) or more general methods by means of finite elements, that allow to analyse
any type of structure.
28
4. Dynamic analysis by time integration with vehicle-structure interaction.

These are the last type of models mentioned in the section 2.3. The level of effort and
complexity of these models may result in an excessive complexity for ordinary project
analysis. Nevertheless, they can be useful to improve the analysis either in some special
project situation or as a part of a research work. It should be considered that with the
interaction models the the bridge dynamic effects result generally in smaller, more realistic
values, comparing to simpler moving load models, specially in short span bridges (see
[11], described also in section 6; reductions of dynamic effects of up o 45% are obtained
in resonance).
The additional complexity makes the work more difficult and risky for the analyst who
defines the model and interprets the results, which is why these analysis must be approved
by the competent authority.
In any of the previous cases, the result of the dynamic analysis may be interpreted to yield a
dynamic factor (without considering track irregularities):
Sdin,real
dynamic analysis → (1 + ϕ0 ) = (20)
Sest,real
Once this is calculated, the effect of the track irregularities may be added by means of factor
ϕ00 , which should be applied to the previous factor (20) according to:
(1 + ϕ0 )(1 + ϕ00 /2). (21)
The maximum of all these values for all the possible trains and circulation speeds will constitute
the impact factor Φ.
It is also convenient to remember that the dynamic analysis results must be applied to verify
not only the ultimate limit states (ULS), but also the service limit states (SLS) related to de-
formations and maximum accelerations; this is the object of another paper in this congress [24].
In particular, one of the conditions that might be difficult to satisfy for certain low spans bridges
is the limit of maximum accelerations (0, 35g for bridges with ballast bed, 0, 50g for the rest).
Finally, it is noticed that for bridges with two or more tracks, the static loads α × LM71 must
be simultaneously applied in the two tracks. However, the absolute maximum dynamic effects
should not be added for the two tracks, rather they should be combined by means of the rule of
the square root of the sum of the squares. These dynamic effects are the ones included within
coefficient ϕ0 in the expression (20).
5 DYNAMIC ANALYSIS IN EUROCODE 1 (PREN 1991-2)

The last Eurocode 1 of structural actions, section 2 (actions on bridges), is the last proposal of
a revision period; the last studies and analysis results arrived at by ERRI have been included,
in particular those from the ERRI D214 committee for structural effects of high speed rail lines
[8]. It is a very recent report, dated 10/01/2002. The more remarkable aspects follow in the next
sections.
29
5.1 Vertical traffic load model

Load model LM71 (UIC71) is adopted, allowing also the possibility to establish by the na-
tional administration a classification factor ×α. Additionally, trains of loads SW/0 (continuous
bridges), SW/2 (heavy trains), and the train without load should be used in certain specific
situations.
5.2 Conditions for dynamic analysis

These conditions are established by the Eurocode by means of a flow chart where the different
possibilities are included (figure 23).
We indicate in summary the main aspects deduced from this chart:
• Situations where only a static analysis is required, with impact factor Φ:
– v ≤ 200 km/h and continuous bridge;
– v ≤ 200 km/h and f0 within the limits of figure 7;
– v > 200 km/h and isostatic bridge with span L ≥ 40 m, and f0 within the limits of
figure 7;
– v > 200 km/h and isostatic bridge, ftorsion > 1.2f0 , + use of tables F1/F2 tables to
verify accelerations (ELS);
• Situations where a direct dynamic analysis is necessary (rest of cases):
– v > 200 km/h and non isostatic bridge (always)
– v > 200 km/h and isostatic bridge (where L < 40 m, or ftorsion ≤ 1.2f0 , or (v/f0 )lim
do not fulfil the requirements of tables F1/F2.
As a result of the dynamic analysis, the following value is obtained:

0
ydin
ϕdin = max − 1. (22)
yest
This factor will be calculated for all real trains (RT) or for HSLM trains. The maximum dynamic
effect will be finally calculated as:
(1 + ϕ0din + ϕ00 /2) × (HSLM ó RT), ó Φ × (LM71 + SW/0)
Finally, it should be mentioned that for the case of bridges with two tracks, the Eurocode
prescribes the dynamic analysis only in one of them.
6 EVALUATION OF THE DYNAMIC VEHICLE–STRUCTURE INTERACTION IN

SIMPLY SUPPORTED BRIDGES
6.1 Scope of the study
The object of this application is to evaluate the effective reduction which is obtained, with
respect to the dynamic analysis made without considering the vehicle-structure interaction, such
30
Figure 23. Flow chart of the new Eurocode 1 [22] for determining whether a dynamic analysis is required
31
as the models based on series of harmonics or models of single moving loads, more common in
engineering practise.
The calculations are based on a modal analysis considering only first mode of vibration,
without shear deformation. A model of moving loads is compared to the model with interaction
proposed in the section 14. Time integration has been carried out using the trapezoidal rule.
Isostatic bridges of spans (l) between 10 and 40 m have been considered, with characteristics
according to the catalogue of isostatic bridges from [7]. The velocity sweep is (120 − 420)
km/h, with δv = 2.5 km/h. The trains employed are the Ice2, Eurostar and Talgo AV, defined in
[13], with damping rates ζ = 0.5%, ζ = 1%, ζ = 1.5% and ζ = 2.0%. The calculations have
been performed with the computer program [5].
The analysis results, as was predictable, show a significant reduction of the maximum dis-
placements and accelerations for models with interaction. Some of the results obtained are
included in table 2.
6.2 Results
In view of the results shown, one may conclude in first place that the moving load models clearly
overestimate, in general terms, the response in accelerations and displacements of an isostatic
structure; in comparative terms, the interaction models can reduce the maximum acceleration
values in isostatic bridges up to 45% respect to acceleration obtained with single loads models.
Additionally, the dynamic response reduction, for the same hypothesis of span and damping,
is greater for accelerations than for displacements, and the reduction increases as the line design
speed is increased. Finally, it is also observed that the reduction of the response decreases when
we increase the damping rate or the bridge span.
7 DYNAMIC RESPONSE OF PORTAL FRAMES FOR UNDERPASSES

7.1 Model used for calculations
For isostatic beams with load trains, the first vibration mode is the one that has a preponder-
ant importance. For a beam with a span with an general support conditions, the fundamental
eigenfrequency of the first vibration mode can be expressed as:
2 r
π/β1 EI
ω= (rad/s) (23)
l m
with β1 = 1 for isostatic beams and β1 = 0.6642 for fully restrained beams. The deck of a portal
frame underpass is somewhere in between both cases, as it can be assumed to a certain extent
to behave like a beam with flexible restraints in its ends. This way, the parameter obtained from
the equation (7) for the first vibration mode of the portal frame will tend to π or to π/0.6642, for
the more flexible or stiffer piers, respectively. The relevant parameters of the deck are, in a first
approach, its length l, bending stiffness EI and linear mass m. Considering a certain length of
an equivalent isostatic beam, and fitting the rest of parameters, this beam could provide identical
fundamental eigenfrequency as the frame, and therefore would have similar dynamic response.
32
Table 2. Reduction obtained of maximum acceleration and displacement with interaction model with respect to the
line
moving load model. Vmax = V0 =220, 250, 270, 300, 350 and 375 km/h
33
With this philosophy, the analogy could be made in several different ways, either conserving
the deck length l, m, EI or its total mass M = ml. This last option would ensure preserving
the kinetic energy of vibration in the equivalent beam.
The differences between the dynamic behaviour of the frame and its equivalent beam could
be improved fitting the assigned m and EI proportionally, so that the ratio k̂ = EI/m between
them remains unchanged. According to [4], this proportional variation would allow to maintain
the first vibration frequency invariable, without modifying the critical speeds for resonance,
that is, without modifying the form of the envelopes, decreasing or increasing the maximum
displacements and accelerations. Therefore, the objective will be to first obtain a beam whose
envelopes are as similar as possible to those of the frame, and then fitting (if necessary) m and
EI conserving their ratio (factor k̂), to fit the maximum displacements and accelerations.
7.2 Verification of simplified model

In order to establish the most adequate model for an equivalent beam, four beams have been
defined for each frame of given span l. The length of these equivalent beams (leq ) is respectively
l, 0.95l, 0.90l, and 0.85l. The mass of the equivalent beams is calculated so as to conserve the
total mass of the deck, mviga = mdintel l/leq . To obtain the equivalent beam bending stiffness EI,
its first mode vibration frequency is adjusted to the frequency of the portal frame:
2 4
ωmarco mviga leq
(EI)eq = (24)
π4
The same damping rate has been considered within the frame and the beam, and neither shear
deformation nor the contribution of the earth cover have been considered. This last aspect
provides a more conservative evaluation, as the additional masses would decrease the maximum
accelerations and displacements. Only the first vibration mode has been taken into account, and
neither the effects of vehicle–structure interaction nor track irregularities have been considered.
The structures selected for the analysis have been 4 representative underpasses of high speed
railway line between Córdoba and Málaga, with spans of 8.5, 8.7, 9.8 and 15 m. Thus, a total
of 20 beam or frame structures have been studied. The frames have been analysed with the
finite element program [10], and the equivalent beams with the program [5]. For each structure,
the envelopes of accelerations, displacements and impact factors Φ have been calculated with
the 7 real trains specified in [13]. The calculations are performed for a range of velocities of
(120 − 420) km/h, with δv = 5 km/h. This makes a total of 8540 dynamic analyses. Some of the
results are included in figures 24, 25 and 26, showing the envelope of maxima for each velocity
for the frame of length 8, 5m, with a representative train of each one of the three types of real
trains mentioned in figure 17.
7.3 Discussion of results

The criterion for selection of the most appropriate equivalent beam model was the greatest
similarity between the accelerations envelope with the frame. since this is the critical aspect
in these structures. With this criterion, the equivalent beam of length l was selected. Hence
34
3 0.45
acel. marco 1 desp. marco 1
acel. viga equiv. l=L desp. viga equiv l=L
acel. viga equiv. l=0.95*L 0.4 desp. viga equiv l=0.95*L
2.5
0.35
Desplazamiento (mm)
2 0.3
Aceleracion (m/s2)
0.25
1.5
0.2
1 0.15
0.1
0.5
0.05
0 0
150 200 250 300 350 400 150 200 250 300 350 400
Velocidad del tren (km/h) Velocidad del tren (km/h)
Figure 24. Envelopes of amax , dmax from portal frame 1 and their equivalent beams of lengths l and 0.95l for Ave
train
3 0.45
marco 1 marco 1
viga equivalente l=L viga equiv l=L
viga equivalente l=0.95*L 0.4 viga equiv l=0.95*L
2.5
0.35
Desplazamiento (mm)
2 0.3
Aceleracion (m/s2)
0.25
1.5
0.2
1 0.15
0.1
0.5
0.05
0 0
150 200 250 300 350 400 150 200 250 300 350 400
Figure 25. Envelopes of amax , dmax from portal frame 1 and their equivalent beams of lengths l and 0.95l of train
talgo AV
35
the equivalent isostatic beam will have the same span l, linear mass m and damping rate ζ as
the frame deck. The bending stiffness EI is obtained according to equation (24). In summary,
other conclusions of this study are:
• It is possible to define an equivalent isostatic beam for the dynamic analysis of the usual
frames in railroad underpasses, which conserve the form of envelopes of accelerations (a),
displacements (δ) and impact factors (Φ).
• The isostatic equivalent beam yields, in practical terms, a conservative evaluation in crit-
ical design points, which are the circulation speeds where the maximum of the envelopes
appears, for a, δ and Φ. In addition, the adjustment mentioned in the section 7.1 of the
quotient k̂ = EIm
is not necessary.
• For non critical speeds, i.e. speeds at which the value of the response is not the maximum
value, the results obtained for the equivalent beam are almost always greater than the res-
ults for the portal frame (figures 24, 25 and 26). However, it cannot be stated with absolute
generality that the equivalent isostatic beam presents greater effects than those obtained
for the frame for any non critical speed. Nevertheless, this aspect lacks relevance for the
design of the structure.
• The frames which were studied are acceptable from a design point of view, their dynamic
behaviour yields values within the acceptable limits: always Φ ≤ 1, and amax,cdv =
1.69 m/s2 ≤ 0.35 g, limit established in [13, 22].
8 SENSITIVITY TO TIME-STEP IN TIME INTEGRATION OF MODES

Before making a modal analysis, it is convenient to estimate the most adequate time-step for the
time integration. In order to illustrate the importance of this point, the maximum accelerations
produced by a train Talgo AV in a isostatic bridge of 10 m span are studied (mechanical char-
acteristics taken from the catalogue of bridges of [7]). Figure 27 compares the results analysed
with a moving load model and modal analysis for different time-steps. The structural damping
rate is ζ = 0.5 %. The integration algorithm used is the trapezoidal rule, with fixed steps of
δt = 0.0001 s, δt = 0.001 s, δt = 0.01 s, δth = 0.1 s and δt = 1 s.
It is possible to observe that, as the integration step δt is reduced, the results obtained con-
verge to the solution which could be considered exact. In fact, the curves corresponding to steps
δt = 0.0001 s and δt = 0.001 s are indistinguishable. However, for large values of δt (in this
case: δt = 1 s and δt = 0.1 s) the curve of maximum accelerations differs radically from this
exact solution. This fact shows that with large time integration steps taking into account the
structural vibration frequency (f0 = 8 Hertz), the dynamic behaviour of the bridge cannot be
characterised adequately.
In [7] and [21] the criterion for performing a dynamic analysis is to consider only the modes
with frequencies f < 20 Hertz. In the last draft of the Eurocode-1 [22], this frequency limit has
been increased to f < 30 Hertz.
Following are summarised the results obtained in a study of the different recommendations
that appear in the codes and in technical literature for the integration step in dynamic analysis
36
3 0.45
acel. marco 1 desp. marco 1
acel. viga equiv. l=L desp. viga equiv l=L
acel. viga equiv. l=0.95*L 0.4 desp. viga equiv l=0.95*L
2.5
0.35
Desplazamiento (mm)
2 0.3
Aceleracion (m/s2)
0.25
1.5
0.2
1 0.15
0.1
0.5
0.05
0 0
150 200 250 300 350 400 150 200 250 300 350 400
Figure 26. Envelopes of amax , dmax from portal frame 1 and their equivalent beams of lengths l and 0.95l of train
Ice2
Figure 27. Maximum acceleration at centre of span, function of the speed of train Talgo AV, for different integration
time-steps h. Simply supported beam with 10 m span from the catalogue of [7]. ζ = 0.5 %.
37
of railroad bridges. According to [7], [6] and [25], the following integration steps h could be
used:
• Determination of the integration step h based on the higher vibration frequency of the
considered structure:
1
h1 = (25)
8fmax
• Determination of the integration step h based on the minimum number of time intervals
(in this case, two hundred) existing during the transit of an axis by the shortest span of the
structure:
Lmin
h2 = (26)
200 v
• Determination of the integration step h based on the number n of vibration modes con-
sidered and the length of the shortest span of the structure:
Lmin
h3 = (27)
4nv
• Time-step h independent from other parameters:
h4 = 0.001 s (28)
• Fixed integration step h that acts as filter of frequencies higher than 50 Hertz (models of
direct time integration of the structure):
h5 = 0.002 s (29)
Figure 28 shows, for a reference case, the maximum acceleration results obtained2 according
to the different alternatives for selection of time-step outlined above.
The results obtained for the steps h2 , h4 and h5 are similar. The determination of the optimal
one between these three steps will depend on other factors like, for example, the convenience
of using the same integration step in a speed sweep —in this case we recommend the use of h2
with v = vmax — or the optimisation of the total time for integration.
9 SIMPLIFIED EVALUATION OF TORSION IN REAL VIADUCTS

A simplified method to evaluate the torsion effects in railroad bridges is proposed in [8]; As a
conservative design envelope, this committee considers the linear superposition of the associ-
ated dynamic effects of bending and torsion, studying separately both effects.
This way, two independent analyses would be performed (bending and torsion), determ-
ining the maximum responses in absolute value (Rf and Rt , respectively) for the variables
under study, usually displacements or accelerations. The simplest method of combination is
proposed, obtaining the total response Rtotal as the direct absolute sum of the individual re-
sponses: Rtotal = Rf + Rt . It should also be possible to consider –although this possibility
2
The results presented in this section have been obtained with a moving load model, with integration of the
resulting equations using the trapezoidal rule.
38
Figure 28. Maximum acceleration at centre of span, function of the train speed of a Talgo AV, for different in-
tegration time-steps h. Simply supported beam from the catalogue of bridges from ERRI D214 with 10 m span.
ζ = 0.5 %.
is not mentioned in [8]— the application of the square root of the sum of squares (SRRS) for
the combination of these two actions, whenever the frequencies of both vibrations are suffi-
ciently separate (ωi − ωj ≥ 20%ωi ). We remark that the use of this hypothesis allows the use
of simplified methods of analysis.
In this section we report an evaluation of the adjustment of this simplified method, applying
it to two bridges of the future high speed line between the cities of Córdoba and Málaga. For the
analysis, six high speed European trains have been considered (Virgin, Ave, Ice2, Etr-y, Eurostar
and Talgo AV). The structural sections corresponding to these bridges are representative of those
being built in modern railroad bridges for HS lines: hollow slab and box section. The bridges
were considered as simply supported in bending, but with a full restraint for torsion in the
supports.
Bridge Box section Hollow slab section

span [m] 46 23.5
mc [*] [kg/m3 ] 3804 3840
Area [m2 ] 10.50 10.22
Ix [m4 ] 21.03 3.35
Iy [m4 ] 120.06 99.66
Iω [m4 ] 27.03 7.64
E [MPa] 36149.6 36149.6
G [MPa] 15062.3 15062.3
ζ 2% 2%
[*] The material density mc has been modified to include the effect of non-structural dead load.
Table 3. Characteristics of bridges analysed for bending and torsion.
39
The computer program employed for these dynamic analyses is [10]. In order to compare
the proposal of simplified evaluation of [8] with the application of a complete three-dimensional
model, the following process has been followed:
• Dynamic analysis to evaluate the effects due to longitudinal bending.
• Dynamic analysis to evaluate the effects due to transversal torsion (only torsion). These
results have been obtained by two different ways:
1. From a bending analysis, applying the existing proportionality between the maximum
accelerations curves associated to torsion and bending.
2. With the results obtained directly from the finite elements model used in torsion.
• Complete dynamic analysis combining bending and torsion.
The comparison criterion that has been established is to relate the results obtained with the
simplified method (direct sum of the absolute maximum of response obtained from the only
bending analysis plus the maximum obtained from the only torsion analysis) with the maximum
corresponding to the complete model (bending and torsion). In addition, the possibility of
combining the hypotheses of only bending and only torsion with method SRSS (square root of
the sum of squares) has been evaluated.
The table 4 shows the maximum values of the resulting accelerations in the centre of the span
for the all the different hypotheses. It also includes the existing deviation between the simplified
model and the complete one considering simultaneous combination of bending and torsion.
Bridge Box section Hollow slab section

Max. acceleration only bending [m/s2 ] 1.27 3.32
Max acceleration only torsion [m/s2 ] 0.05 0.28
Max. acceleration combined model [m/s2 ] 1.29 3.41
Max. acceleration simplified model [m/s2 ] 1.32 3.61
Deviation simplify model vs. combined ' +2.2% ' +5.72%
Max acceleration max. SRSS [m/s2 ] 1.27 3.33
Table 4. Summary of analysis results obtained for different alternatives for considering torsion, as only bending,
only torsion and as a combined model.
From the study of the results obtained, the following conclusions can be extracted:
• For the sections studied, the simplified method proposed in [8] for the analysis of the
phenomena of combination of bending and torsion is valid.
The simplified method modified with the use of SRSS combination, is not on the safety
side. However, this fact is due to the reduced value of torsion accelerations, and the devi-
ation is small.
40
• For structural sections with large torsional stiffness GIω , the deviation of results obtained
with the simplified method and those obtained with the combined model of bending–
torsion is almost insignificant; thus, this simplified method is recommended.
• For sections with reduced torsional stiffness (for example, hollow slabs, open sections),
the existing deviation of results between the simplified models and those of interaction
bending–torsion is more significant (in the case studied, near 5%). In case of being near
to the limits of accelerations or total displacements established by the code, it is recom-
mended to obtain more precise results using combined models which include bending and
torsion.
This study has been limited to bridges with closed boxed sections and hollow slabs. There
are sections, of common use in other countries, where the torsion effects are more important. In
[17] the analysis of a composite steel-concrete bridge with open boxed section and with cross
lower bracing are detailed; in those bridges the torsion accelerations are more relevant as design
limitations.
10 CONCLUSIONS
Following we summarise some final remarks:
• The design of high speed railroad bridges, because of the real possibility of resonance, re-
quire consideration of the dynamic vibration under moving loads. For this purpose several
models of analysis are described in this article, of smaller or greater complexity.
• It is very important to apply these dynamic analysis methods in applied research to improve
our knowledge about the most determinant factors for the dynamic response of the bridges
from the project point of view, as well as to be able to develop engineering design methods
and codes which are sufficiently practical, secure, and simple to use.
• The new draft of IAPF [13] and the final draft of Eurocode 1 of actions in bridges [22]
cover adequately this necessity of dynamic analysis for the high speed lines.
• The dynamic analysis methods have been applied to some representative design problems,
obtaining as a result the validation of simplified methodologies which may help in a more
reliable evaluation of engineering designs.
REFERENCIAS
[1] E. Alarcón; El coeficiente de impacto en puentes de ferrocarril; Revista de Obras públicas;
september 1971.
[2] E. Alarcón, R. Álvarez, M. Doblaré, J. Molina; Efectos dinámicos en puentes de ferrocarril;
Hormigón y acero 155, pp 173–186; 1985.
[3] Ministerio de Fomento, Dir. Gral. de ferrocarriles y transporte por carretera; Viaducto sobre
el río Tajo, Línea AVE Madrid–Sevilla, ensayos dinámicos; developed by IIC S.A.; 1996.
[4] J. Domínguez; Dinámica de puentes de ferrocarril para alta velocidad: métodos de cálculo
y estudio de la resonancia; doctoral thesis; Escuela Técnica Superior de Ingenieros de
41
Caminos, Canales y Puertos de Madrid (UPM); published by the Asociación Nacional de

Constructores Independientes (ANCI); 2001.
[5] Caldintav v.2.0; Cálculo Dinámico para trenes de alta velocidad; Grupo de Mecánica Com-
putacional; 2002.
[6] ERRI D214 comittee; Ponts-Rails pour vitesses > 200 km/h; Calcul des ponts a structure
complexe au passage d’un convoi. Logiciels de calcul dynamique; European Rail Research
Institute (ERRI); 1999.
[7] ERRI D214 comittee; Design of Railmode Bridges for Speed up to 350 km/h; Dynamic
loading effects including resonance; Final report. Draft C; European Rail Research Insti-
tute (ERRI); 1998.
[8] ERRI D214 comittee; Ponts-Rails pour vitesses > 200 km/h; Final report. Part B. Propos-
ition de fiche UIC 776-2R; European Rail Research Institute (ERRI); 1999.
[9] Clough, R. y Penzien, J.; Dynamics of Structures; Mc Graw-Hill; 1993.
[10] Taylor, R.L.; FEAP. A Finite Element Analysis Program; User’s Manual; http:www.
ce.berkeley.edu/~rlt
[11] Goicolea, J.M., Domínguez, J., Gabaldón, F. y Navarro, J.A.; Estudio de fenómenos
resonantes en puentes de ferrocarril: I. Valoración de la interacción dinámica vehículo-
estructura; technical report; december, 2001.
[12] Goicolea, J.M., Domínguez, J., Gabaldón, F. y Navarro, J.A.; Estudio de fenómenos reson-
antes en puentes de ferrocarril: II. Cálculo de pasos inferiores; technical report; december,
2001.
[13] Ministerio de Fomento; Instrucción de acciones a considerar en el proyecto de puentes de
ferrocarril; Draft, not published yet; 2002.
[14] Timoshenko, S.P. y Young, D.H.; Vibration problems in engineering; Van Nostrand, NY;
1955 (3"a ed.).
[15] Fryba, L.; Vibration of solids and structures under moving loads; Academia; Prague,
Noordhoff; 1972.
[16] Fryba, L.; Dynamics of railways bridges; Thomas Telford; 1996.
[17] Ville de Goyet, V. de and Cremer, J.M. and Lothaire, A. and Del Forno, J.Y.; Viaduc de
l’Arc, viaducs de Mornas et de Mondragon, viaduc de Donzère: 4 ouvrages, 3 conceptions;
Bulletin Ponts métalliques (19); 1999.
[18] Union Internationale des Chemins de Fer (UIC); UIC Code 776-1 R: Charges a prendre
en consideration dans le calcul des ponts-rails; 1979.
[19] Ministerio de Obras Públicas y Urbanismo; Instrucción ralativa a las acciones a consid-
erar en el proyecto de puentes de ferrocarril; 1975.
[20] Ferrovie dello Stato; Sovraccarichi per il calcolo dei ponti ferroviari; 1997.
[21] Comité Europeo de Normalización; Eurocódigo 1: Bases de proyecto y acciones en es-
tructuras; parte 3, acciones de tráfico en puentes; translated and published by AENOR;
1998.
[22] European Committee for Standardization; prEN 1991-2: EUROCODE 1 - Actions on
structures, Part 2: Traffic loads on bridges; 2002.
42
[23] Jorge Nasarre y de Goicoechea; particular report; Subdirección General de Planes y

Proyectos de Infraestructuras Ferroviarias del Ministerio de Fomento; 2000.
[24] Jorge Nasarre y de Goicoechea; Estados límite de servicio en relación con la vía en
puentes de ferrocarril; railroad bridges congress from the Spanish group of IABSE; Madrid,
june 12–14, 2002.
[25] Museros, P. y Vivero, G. y Alarcón, E.; Propuesta de articulado a la IAPF: Efectos
dinámicos y Apéndice de Cálculo dinámico; technical report for the developing comission
IAPF 2000.
[26] Tartary, J.P. y Robert, N.; Recherche du train universel pour calculs dinamiques; technical
report to ERRI D214 comittee; A.V.L.S.; 2000.
[27] Domínguez, J. y Goicolea, J.M.; Dynamic effects of regular trains: proposal for inclusion
within universal train envelope; technical report from Grupo de Mecánica Computacional
for the ERRI D214 comittee; 2000.
43

New Dynamic Analysis Methods For Railway Bridges in Codes Iapf and Eurocode 1

Uploaded by

Copyright:

Available Formats

New Dynamic Analysis Methods For Railway Bridges in Codes Iapf and Eurocode 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

New Dynamic Analysis Methods For Railway Bridges in Codes Iapf and Eurocode 1

Uploaded by

Copyright:

Available Formats

RALWAY BRIDGES

Design, Construction and Maintenance

NEW DYNAMIC ANALYSIS METHODS FOR RAILWAY BRIDGES IN

Keywords: Structural dynamics, high speed railways.

1 INTRODUCTION AND MOTIVATION

in ΦUIC = 1 + ϕ0 = 2.16 (not considering the effect of track irregularities).

2.1.1 Code IAPF-75

For continuous beams the proposed expression is

2.1.2 Code UIC-776-1R

2.2 Dynamic analysis with moving loads

2.2.1 Analytical methods

to the first mode is defined through a parameter b with the equation:

Figure 10. Response for a single moving load

differential equation corresponding to mode i is in this case:

Figure 11. Response for a load train

2.2.2 Finite element methods

A−1 A A+1 A−1 A A+1 A−1 A A+1

2.3 Dynamic analysis with vehicle-structure interaction

       

Figure 13. Complete vehicle–structure interaction model.

y k (t) y 2 (t) y 1 (t)

• For each interaction element (j = 1 . . . k):

2.4 Models based on harmonic series

other terms have the following definition:

3 HIGH-SPEED REAL TRAINS AND INTEROPERABILITY

3.1 Real trains

3.2 High-speed interoperability load model (HSLM)

Figure 17. Different types of high-speed trains, according to Eurocode 1 [22]

Figure 18. Dynamic signature (undamped) for European high-speed trains.

3.2.1 UNIV-A and the Talgo AV

4 DYNAMIC ANALYSIS IN THE SPANISH CODE FOR ACTIONS ON RAILWAY

4.1 Vertical traffic load model

4.2 Envelope dynamic factor Φ

4.3 Dynamic analysis procedures

4. Dynamic analysis by time integration with vehicle-structure interaction.

5 DYNAMIC ANALYSIS IN EUROCODE 1 (PREN 1991-2)

5.1 Vertical traffic load model

5.2 Conditions for dynamic analysis

6 EVALUATION OF THE DYNAMIC VEHICLE–STRUCTURE INTERACTION IN

7 DYNAMIC RESPONSE OF PORTAL FRAMES FOR UNDERPASSES

7.2 Verification of simplified model

7.3 Discussion of results

8 SENSITIVITY TO TIME-STEP IN TIME INTEGRATION OF MODES

9 SIMPLIFIED EVALUATION OF TORSION IN REAL VIADUCTS

Bridge Box section Hollow slab section

Bridge Box section Hollow slab section

Caminos, Canales y Puertos de Madrid (UPM); published by the Asociación Nacional de

[23] Jorge Nasarre y de Goicoechea; particular report; Subdirección General de Planes y

You might also like