Professional Documents
Culture Documents
UN
NIV
VE
ER
RS
SIT
TA
AT
T PO
POLIT
T
C
CN
NIIC
CA
AD
DE
EV
VA
A
AL
L
N
NCIA
CA
Dep
De
pa
arrta
am
me
men
nto
od
de
e In
ng
ge
en
niie
era
a Me
Mecc
n
ni ca y d
de
eM
Ma
ate
e riale
es
s
TE
TES
SI
S SD
DO
OC
OCT
OR
RAL
RA
TO
MO
MO
OD
DEL
ELOS DIN
N
N
M
MIC
C
CO
OS
AV
VA
AN
N
NZ
ZADO
OS
OS A
DOS
DE
EJ
JE
ES
S FE
FERR
RRO
RO
OV
VIA
AR
AR
RIO
O
OS
S
DE E
P
Pre
esse
en
ntta
ad
da
a po
or:: D
D. Jo
Jos
s
Ma
Marrtn
ne
z Ca
Cassa
ass
ez
D
Dirrig
giid
da
ap
po
orr:
D
Drr. D
D. L
Lu
uis Ba
Bae
eza
a Go
Go nzz
le
ezz
D
Drr. D
D. F
Frra
an
nccis
scco
D.. D
De
en
nia
aG
Gu
uz
zm
mn
n
oD
m
V
Vale
en
ncia
a, No
Novvie
em
mb
de 2
20
01
13
3
brre de
TESIS DOCTORAL
RESUMEN
En esta Tesis se aborda el estudio de la interaccin dinmica de un
vehculo ferroviario con la va a travs de la incorporacin en el modelado
de la flexibilidad de los elementos que componen el sistema. La dinmica
acoplada del vehculo con la va determina algunos problemas importantes
de la ingeniera ferroviaria como son el ruido de rodadura y la generacin de
determinados desgastes anormales de la rueda y del carril, lo cual afecta a la
seguridad, al impacto ambiental y al coste de la explotacin.
Las principales contribuciones de esta Tesis se enmarcan en el
modelado del eje montado donde, adems de la flexibilidad, se consideran
los efectos inerciales asociados al giro. Los trabajos realizados abarcan
desde la profundizacin en el campo de la Mecnica de Rotores en el que se
fundamentan los modelos del eje montado, hasta la aplicacin a los casos de
inters industrial que las hiptesis adoptadas ms avanzadas hacen posible
plantear.
Para la obtencin de la ecuacin del movimiento del eje se propone
una metodologa para slidos de revolucin basada en el uso de un sistema
de coordenadas en el que, aprovechando la simetra axial, se utilizan las
geometras deformadas modales en ejes fijos como base para definir el
desplazamiento de una partcula en funcin de su posicin espacial en la
configuracin no deformada (planteamiento modal Euleriano). Al estar
formulado en coordenadas espaciales fijas, el modelo es eficiente en
aquellos casos en los que hay interaccin entre el slido rotatorio con una
estructura no rotatoria, tal como corresponde al caso de un eje ferroviario
con la va. Las hiptesis adoptadas permiten considerar tanto la dinmica
cuando el slido describe una trayectoria arbitraria aadida al movimiento
de rotacin como el efecto del amortiguamiento interno.
La adopcin de un sistema de coordenadas de trayectoria que se
mueve segn el eje de la va en combinacin con el planteamiento modal
Euleriano, han hecho posible obtener una formulacin del eje montado en
curva cuyo coste computacional asociado es muy bajo. Para ello se
desarrolla la ecuacin del movimiento adoptando la hiptesis de pequeos
desplazamientos con respecto al sistema de trayectoria, e incorporando un
modelo de contacto rueda-carril flexible basado en el mtodo de penalti, que
considera los perfiles reales de la rueda y del carril.
ii
ABSTRACT
In this Thesis the dynamic interaction of a railway vehicle with the
track is studied through the incorporation in the modelling of the flexibility
associated with the elements of the system. The coupled dynamics of the
vehicle with the track determines some important railway engineering
problems, such as rolling noise and generation of abnormal wear of wheel
and rail, which affects the safety, environmental impact and cost of
exploitation.
The main contributions of this Thesis are related to the modelling of
the wheelset, where in addition to flexibility, inertial effects associated with
the rotation are considered. The works carried out range from deepening in
the field of Rotors Dynamics in which wheelset models are based, to
application to cases of industrial interest where the more advanced
assumptions are fulfilled.
A methodology suitable for flexible rotating solids of revolution is
proposed to obtain the equation of motion of the wheelset. Taking
advantage of the axial symmetry, the model makes use of the vibration
mode shapes in a non-rotating frame as basis functions, in order to define
the displacements of a particle depending on its spatial position in the
undeformed configuration (Eulerian modal approach). Due to the fixed
coordinate system, the model is effective in those cases where there is an
interaction between the rotating solid and the non-rotating structure, as the
case of a wheelset with the track. The considered assumptions allow to
consider the dynamics when the solid describes an arbitrary trajectory added
to the rotation as well as the effect of internal damping.
Adopting a trajectory-based coordinate system moving along the
axis of the track, in combination with the Eulerian modal approach, it is
possible to obtain a formulation of the wheelset in curve, whose associated
computational cost is very low. This is achieved by combining the equation
of motion and the hypothesis of small displacements with respect to the
trajectory system. In addition, a wheel-rail flexible contact model is
incorporated, which is based on a penalty method that considers the actual
profiles of the wheel and rail.
The model of the overall system is performed by adopting a substructuring technique, where the different substructures are formulated
separately and then are coupled through the forces transmitted between
iii
them. Cyclic boundary conditions are taken in the model of the track that
reduce the edge effect associated with the finite length of the track. The
models are able to consider a straight track as well as a curved track of
constant radius.
The works show results associated with a number of practical
scenarios that give rise to the high frequency dynamics of the track unsprung masses system. The effects of irregularities of the wheel surface
and the rail head are addressed, such as wheel-flat, the rail corrugation (or
harmonic wear) and the presence of random defects of mid and short
wavelength (roughness). The simulations presented in this Thesis show the
calculation of stresses in the wheelset axle and the wheel-rail contact forces.
iv
RESUM
En esta Tesi s'aborda l'estudi de la interacci dinmica d'un vehicle
ferroviari amb la via a travs de la incorporaci en el modelatge de la
flexibilitat dels elements que componen el sistema. La dinmica acoblada
del vehicle amb la via determina alguns problemes importants de
l'enginyeria ferroviria com sn el soroll de rodament i la generaci de
determinats desgastos anormals de la roda i del carril, la qual cosa afecta la
seguretat, a l'impacte ambiental i al cost de l'explotaci.
Les principals contribucions d'esta Tesi s'emmarquen en el
modelatge de l'eix muntat on, a ms de la flexibilitat, es consideren els
efectes inerciales associats al gir. Els treballs realitzats comprenen des de
l'aprofundiment en el camp de la Mecnica de Rotors en qu es fonamenten
els models de l'eix muntat, fins a l'aplicaci als casos d'inters industrial que
les hiptesis adoptades ms avanades fan possible plantejar.
Per a l'obtenci de l'equaci del moviment de l'eix es proposa una
metodologia per a slids de revoluci basada en l's d'un sistema de
coordenades en qu, aprofitant la simetria axial, s'utilitzen les geometries
deformades modals en eixos fixos com a base per a definir el desplaament
d'una partcula en funci de la seua posici espacial en la configuraci no
deformada (plantejament modal Euleri). Degut a la formulaci en
coordenades espacials fixes, el model s eficient en aquells casos en qu hi
ha interacci entre el slid rotatori amb una estructura no rotatria, tal com
correspon al cas d'un eix ferroviari amb la via. Les hiptesis adoptades
permeten considerar la dinmica quan el slid descriu una trajectria
arbitrria afegida al moviment de rotaci i tamb l'efecte de l'amortiment
intern.
L'adopci d'un sistema de coordenades de trajectria que es mou
segons l'eix de la via en combinaci amb el plantejament modal Euleri, han
fet possible obtindre una formulaci de l'eix muntat en corba amb un cost
computacional associat molt redut. Per a aix es desenrotlla l'equaci del
moviment adoptant la hiptesi de xicotets desplaaments respecte al sistema
de trajectria, i incorporant un model de contacte roda-carril flexible basat
en un mtode de penal, que considera els perfils reals de la roda i del carril.
El modelatge general del sistema es du a terme adoptant tcniques de
subestructuraci en les quals els diferents subsistemes es formulen de
manera separada, acoblant-se a travs de les forces que es transmeten entre
vi
PREFACIO
Esta Tesis Doctoral se enmarca en el rea de Ingeniera Mecnica, y
pertenece a la lnea de investigacin sobre dinmica ferroviaria. Contiene
las investigaciones originales llevadas a cabo por el candidato, excepto
aquellas explcitamente referenciadas. La Tesis consiste en dos partes, la
primera de las cuales contiene una memoria que proporciona una
introduccin del trabajo y describe los diferentes modelos desarrollados y
las simulaciones dinmicas llevadas a cabo. Adems esta parte incluye
diversos resultados mostrados en los artculos adjuntos y las conclusiones
que se derivan. La segunda parte de la Tesis es una coleccin de los
siguientes artculos:
vii
viii
AGRADECIMIENTOS
En primer lugar, quiero expresar mi ms sincero agradecimiento a mis directores de
Tesis. Luis, qu decirte que no sepas. Ambos sabemos lo mucho que hemos luchado por
este trabajo desde el principio. Y as y slo as, es como se consiguen hacer bien las cosas,
en equipo. Sin duda alguna, el xito de este trabajo es en gran parte gracias a ti. Nunca
olvidar estos ltimos meses de angustia en los que has estado a mi lado codo con codo,
animndome y dndome energas para acabar. Acho, que esto sale, hazme caso, chale
horas y saldr. Maldito creepage, cuntas horas de sueo me ha quitado. En fin Luis,
muchsimas gracias por todo, por estar ah siempre que te necesitaba tanto a nivel
laboral como personal. Me tendrs para todo lo que me necesites, como compaero de
trabajo y sobre todo como amigo. Gracias RailMan.
A Paco, mi mentor. No tengo palabras de agradecimiento, de verdad. Decirte lo que
siempre grito a los cuatro vientos haya quien haya, eres mi ejemplo a seguir. Desde que
te conoc en clase, dije: quiero ser como l, y creme, lucho cada da porque as sea.
Descubrirte como profesor fue mi motivacin para estar aqu, para llegar tan lejos, y
ahora adems me siento afortunado de tenerte como amigo. Agradecerte enormemente
tus sabios consejos, tu grata compaa durante tanto tiempo, la preocupacin que
siempre has mostrado por m, y sobre todo, agradecerte que siempre me hayas tendido tu
mano para levantarme cuando me has visto cado. Eres mi ejemplo profesional y
personal. Gracias de corazn, amigo mo.
Desidero ringraziare Stefano Bruni per la possibilit che mi ha dato di lavorare a
Milano presso il suo gruppo e per tutto laiuto e il sostegno datomi da quando ci
conosciamo. stato un enorme piacere lavorare al tuo fianco, come ricercatore ho
appreso e sono cresciuto molto grazie a te. Sei un gran professore e una persona
meravigliosa. Ti considero come un ulteriore direttore della mia Tesi. Questo solo
linizio della nostra collaboraione e spero di continuare a lavorare al tuo fianco. Mille
grazie Stefano per tutto. Desidero anche ringraziare Laura Mazzola e Egidio per la
loro collaborazione in questa Tesi, grazie ragazzi.
Mis agradecimientos tambin para todos mis compaeros del departamento de
Ingeniera Mecnica y de Materiales. Paloma, Vicente, Paco V., Manolo, Pepe, Suer,
Juanjo, Jose Enrique, Juanvi, Toni, Ana V. y Ana P., Alex, Paco Rubio, muchas
gracias a todos por vuestro compaerismo y por hacer del lugar de trabajo un sitio tan
agradable. A Juan Fayos, con el que empec mi andadura en el mundo de los Rotores
pelendome por descifrar su gran Tesis. Eugenio, como no agradecerte esos raticos que
aunque cortos son tan especiales contigo, raticos en los que hablamos de todo, nos
remos y me aconsejas. Eres una persona excepcional. Gracias campen. Javi, axo hotia
ix
pijo, cuntos cafs y cigarrillos hemos compartido, cuntas conversaciones sobre mi tesis
hemos tenido, y cuntas ideas y nimos me has dado. Eres muy grande y siempre te
estar agradecido, Javi. Federico Manuel, sin ti se me hubiera colgado Matlab una y
otra vez, hubiera explotado algn servidor, se me hubiera formateado el ordenador y el
Intelaim perdiendo toda la Tesis, y un largo etctera. Gracias por todo el soporte
informtico que me has dado, y por esa sonrisa que me regalas todos los das pese al
folln que te doy. No puedo olvidarme de mis secretarias, Marga, ngels, Amparo y
Eva. Agradeceros la disposicin, preocupacin y apoyo que siempre habis mostrado
conmigo, y esa felicidad con la que me recibs siempre, sin olvidarme del folclore. Sois un
encanto. Gracias.
Javier Fuenmayor, decirte que te estar eternamente agradecido por haber apostado por
m, por haber luchado durante tanto tiempo para poder brindarme la oportunidad de
cumplir mi sueo, ser profesor. Gracias a ti empec esta andadura. Como hasta ahora,
me seguir esforzando al mximo en mi trabajo y en hacer que te sientas orgulloso de m.
Muchsimas gracias de corazn.
Este agradecimiento debo extenderlo a Alfonso. Has confiado en m y lo sigues
haciendo, te implicas y preocupas por todo lo que me rodea, mis renovaciones, la
evolucin de mi Tesis, mis hojas de ruta, me has dado muchsimos consejos y ratos de
desconexin y folclore tan necesarios. Por todo ello, gracias Alfonso.
A todos los becarios con los que empec esta andadura, muchas gracias chic@s. En
especial a Enrique, Eva, Onofre, Camila y Luca. Hemos compartido momentos
inolvidables, (vase fiestas de Planes). Os deseo lo mejor a todos en el futuro. Gracias. Y
Juan, ha sido un descubrimiento conocerte, primero como alumno, despus como
compaero y ahora como amigo. Quiero darte las gracias por todo el apoyo que me has
dado, y por compartir conmigo tantos eventos deportivos. Llegars donde te propongas,
eres un fiera aun siendo del Valencia, y tendrs mi ayuda siempre. nimo y gracias
chaval. A Antoine, Moha y Octavio, se os echa mucho de menos, dejasteis un gran vaco
cuando os fuisteis. Me alegro de que todo os vaya genial, y gracias por regalarme todo el
tiempo que compartimos. Rafa y Vir, gracias por vuestro apoyo an en la distancia que
nos separa, y por todos los momentos inolvidables que vivimos y nos quedan por vivir.
Mis amigos del pueblo, mis queridos amigos que tan abandonados he tenido estos
ltimos aos por razones obvias y que tanto he echado de menos. Habis sido muy
comprensivos conmigo, me habis apoyado y animado en todo, me habis demostrado
muchsimo cario pese a mi prolongada ausencia. Os pido perdn por la cantidad de
veces que no he podido compartir vivencias con vosotros, pero deciros que ya estoy de
vuelta, ahora s. Gracias por estar ah chic@s. Eloy, bendigo aquel da que nos
x
conocimos en Valencia, fue sin duda mi mejor poca aqu. Agradecerte todo el tiempo
que me has dedicado y todo tu apoyo. Ahora te puedo decir, kilo, que tiemble el
Crillas! Joven, que ganas tenas de que acabase. Cuantas veces me habrs dicho:
Muchacho, tengo una gana de que acabes y seas doctor de una vez. Pues se acab, y
ahora podr dedicarte el tiempo que te mereces. Gracias por pertenecer a mi vida Piwy!
Al grupico SuperGuay de Valencia, Mara, Rafa, Tere, Ana, Manu, Laura, Benja,
Juan, deciros que para m habis cado del cielo, conoceros ha sido un regalo y lo pienso
disfrutar de por vida. Os he conocido en el momento ms duro de mi carrera, y aun as
consegus que desconecte y que saque lo mejor de m. Felicidades por ser como sois, y
muchsimas gracias chic@s. Mara, Tere, Ana, os merecis mencin aparte. Gracias por
preocuparos, por preguntarme, por escucharme, por animarme en todo momento, por el
calor humano, por estar ah. Me habis demostrado muchsimo en muy poco tiempo. Os
quiero mucho nenas.
A mis padres, de los cuales me separ a los 17 aos siendo un cro para poder llegar
hasta aqu. An en la distancia que nos separa me habis transmitido una fuerza y
motivacin indispensable para afrontar esta etapa. Gracias por haberme dado la
oportunidad de cumplir un sueo, por confiar en m y ayudarme en todo lo que he
necesitado. Mi hermana Nata, gracias por esas frases llenas de cario y motivacin, me
han impulsado en este camino. A mi hermano Salvi, te agradezco todo el apoyo y los
momentos que has compartido conmigo, esos partidos de ftbol y descensos por el ro. A
todos, os quiero y os echo de menos.
Dedico esta Tesis a la persona ms importante de mi vida, a la que ms quiero, la que ha
estado a mi lado en todo momento, la que ha compartido todos mis estados de nimo,
ayudndome a levantarme en mis fracasos y disfrutando de mis xitos. Eva, va por ti, te
lo mereces. Este logro es gracias a ti, una Tesis no se acaba sin motivacin, sin cario y
sin apoyo, y t has sido mi principal fuente en todo. Te debo infinito tiempo, imposible
de recuperar, pero cada minuto a partir de ahora ser eterno. Te lo prometo. Gracias
cario, te quiero mucho.
Noviembre de 2013
xi
ndice general
Resumen
Abstract
iii
Resum
Prefacio
vii
Agradecimientos
ix
ndice general
xiii
PARTE I. MEMORIA
1 INTRODUCCIN ............................................................................................. 1
1.1
1.2
1.3
2.3
................................................................................................................. 14
3 MODELO DE VA FLEXIBLE..................................................................... 23
3.1
3.2
CONCLUSIONES......................................................................................... 57
DESARROLLOS FUTUROS........................................................................... 60
ARTCULO 1
ARTCULO 2
ARTCULO 3
.............................................................................................. 61
.............................................................................................. 62
.............................................................................................. 63
BIBLIOGRAFA .................................................................................................. 65
xiv
Parte I
Memoria
INTRODUCCIN
1 INTRODUCCIN
1.1 MOTIVACIN Y ANTECEDENTES
Esta Tesis Doctoral se enmarca en la lnea de investigacin de
dinmica ferroviaria desarrollada en el Departamento de Ingeniera
Mecnica y de Materiales de la Universitat Politcnica de Valncia. Dentro
de esta lnea se realizan modelos correspondientes a la dinmica acoplada
entre el vehculo ferroviario y la va, que afectan a problemas de
corrugacin de los carriles, ovalizacin y poligonalizacin de las ruedas,
planos de bloqueo en ruedas (desgastes producidos por abrasin del carril
sobre la rueda durante una prdida de adherencia) y otros tipos de desgastes
anormales que afectan a la emisin acstica, a la conservacin de la va y a
la fiabilidad de los rganos de rodadura de los vehculos. Estos fenmenos
se asocian a un rango de altas frecuencias y estn caracterizados por la
dinmica estructural de los elementos que conforman el sistema.
La importancia de este tema ha favorecido que numerosos
investigadores hayan propuesto metodologas para la simulacin dinmica
de la interaccin del vehculo con la va. Resulta destacable comprobar que
la prctica totalidad de los modelos publicados correspondientes a la
interaccin va-vehculo adoptan modelos del eje rgidos, con muy pocas
excepciones. Encontramos un reducido primer grupo en el cual se considera
la flexibilidad de un eje que no gira en las Referencias [1 - 4]. Con
excepcin del trabajo de Szolc [1] que desarrolla una metodologa basada en
parmetros concentrados, las referencias correspondientes a la Escuela de
Bilbao, KTH y Chalmers [2, 3, 4] incorporan un modelo de elementos
finitos del eje montado. Uno de los avances ms significativos es llevado a
cabo en la tesis de Thompson [5, 6] para abordar problemas de ruido de
rodadura. En su modelo no se consideran los efectos inerciales asociados al
giro de la rueda, pero s se incorpora el efecto correspondiente al
movimiento relativo de la fuerza de contacto rueda-carril con respecto a la
rueda. Esta misma hiptesis ha sido adoptada recientemente por Pieringer et
al. [7] a travs de una metodologa de clculo basada en el producto de
convolucin.
INTRODUCCIN
INTRODUCCIN
INTRODUCCIN
INTRODUCCIN
INTRODUCCIN
1.2 OBJETIVOS
El transporte ferroviario se encuentra inmerso en continua evolucin
y desarrollo a nivel mundial, orientado en los ltimos aos haca vehculos
de mayores velocidades y mayores cargas, acompaados con mejoras de
prestaciones en el servicio y de fiabilidad de componentes. Es por ello que
la mayora de investigaciones recientes estn enfocadas al desarrollo de
metodologas eficientes que permitan abordar el problema de interaccin
dinmica vehculo-va con mayor precisin, donde sin duda alguna el
modelo de eje ferroviario juega un papel de vital importancia.
Hasta la fecha, la mayora de trabajos han considerado en sus
modelos de interaccin el eje ferroviario como un slido rgido, limitando
as el rango de estudio al intervalo de bajas frecuencias. Por este motivo, los
modelos disponibles no han sido capaces de representar la dinmica de alta
frecuencia asociada a la flexibilidad del eje, excitada por irregularidades de
longitud de onda corta presentes en los elementos de rodadura.
El objetivo ltimo de la Tesis es el desarrollo de modelos dinmicos
del eje ferroviario que consideren la flexibilidad y efectos giroscpicos
asociados a la rotacin del mismo. De este modo, los modelos permitirn
incluir la dinmica de alta frecuencia en los fenmenos de interaccin del
vehculo con la va en condiciones reales de servicio, as como analizar su
influencia en la dinmica acoplada. Para la consecucin de este objetivo
final se plantea la obtencin de los siguientes objetivos parciales:
INTRODUCCIN
INTRODUCCIN
INTRODUCCIN
j z
1
j z
sen
,
mj
L
11
(2.1)
kj
mj
(2.2)
AL
2
I j 2
2L
E I j
.
2 L3
4
, kj
(2.3)
x( z , t )
y ( z , t )
j 1
q xj (t )
j
q y (t )
j ( z)
q
j
(2.4)
j 1
j 2 g j G c j I q j 2j I 2 c j G q j Q j ,
q
(2.5)
12
c j j j , g j
j 2 ,
2 j 2
0 1
.
G
1 0
(2.6)
X
X t
Y
Z
Y
Figura 2.1. Sistemas de referencia fijo XYZ y mvil XYZ para viga rotatoria y
cilindro biarticulado.
13
MO
OD
DE
ELO
O DIN
M
MIC
CO
O DE
DE S
S LID
DO
OS
SD
DE R
REV
VO
OL
LUC
CI
N
N FL
FLEX
XIB
BL
LES
S Y RO
OT
TAT
TO
OR
RIO
OS
S
22.2 MO
MOD
DE
EL
LO
O DI
DIN
N
M
M
MIIC
CO
OD
DE
ES
S
L
LIID
DO
O
OS
SD
DE
ER
RE
EV
VO
OL
LU
UC
CII
N
N
F
FL
LE
EX
XIB
X BL
BLE
ES
S Y RO
ROT
TA
ATO
AT
OR
RIIO
OS
En
En trrabbaajo
oss an
ntterrio
orress del
d l ggrru
upoo in
nvveesttiggaaddor reellaccioon
n laa
naaddos ccon
T
Teesiss [[27
7]] sse pplaan
nteeaa el
e des
d saarrrollo
o dee laa ddinm
miicaa dell ssliddoo ro
rotaato
orrio
o a parrtiir
ddee la
la ttraannsffo
orm
m cin
n dde un
un sis
s steem
maa de
d co
c orrd
den
naad
daas mo
moddall Lag
grran
nggiaan
noo a
mac
ootrro
o dde tippo
o Eu
Euleerriaanno. E
En
n est
e ta seec ciin
n se
se pro
p oppo
onee un
unaa exxp
pliiccaccinn
aaltternnaativaa en
n laa quuee se
se pr
presccinnddee ddee de
desarrroollo
os an
anteeriioorees ppro
poorccionnaad
do s
op
ppoor otr
o ro
os aau
utoorress y ssee paarrtee dir
d ecctaam
meen
d pprrinnccip
piooss geen
ntee de
neeraaless de laa
ddinnm
miicaa.
Co
Conn el
e fin
f n dee m
mo
oddellar laa ciineem
mtticca de
del s
lliddo
o ffleex
xibble,, ssee de
d efiin
nenn
ddoos co
configuurraccion
nees ddeel sllid
doo, aso
a occiaaddaas a la
l ggeeom
m
metrraa no
no defform
maad
daa y
ddeefo
orrm
maadaa, seeg
n see de
detaallla en
n la
l Fig
guuraa 22.2
2. La cco
onnfiigu
urraaciin no
no
g
ddeefo
orrm
maadaa esst assoociiaadaa co
onn el
el mo
movvim
mie
m en
o ddee ro
r otaaciin
n ddee cu
cuerrpo rg
giddo
o de
d el
nto
eejee (vveelo
occiddaad an
a guullarr cco
onnsttantte), m
miien
nttraas qqu
ue laa coonnffigguuraacci
n ddeefo
orrm
maadaa
n
ccoon
nsiideerra laa fle
f ex
biilidaadd y loos ppeq
qu
uee
oss ddeespllaazaam
miiennttos dee sliid
do rggiddo
o.
xib
F
Fiigu
urra 2..2.. Sis
S steem
maa de
d reefeereen
nciia fiijo
oX
XY
YZ
Z. Con
C nfi
figurraaciin
nn
no
od
deform
ma
da (e
en trra
azo
o
ad
diiscco nttin
nu
uo)) y d
deform
ma
ad
da..
14
4
r u w(u, t ) ,
(2.7)
r u (u) q(t ) .
(2.8)
0 0J.
0 0
(2.9)
~
T
~,
v v1 v2 v3 u J u u
(2.10)
15
Dr Du Dw
Dt Dt Dt
v u
Ju w
v u
i
w
.
u~i
u i
(2.11)
Dt q
E k
q
j i
Vol
2 T
ui
i 1,3
Vol
2 T
q
Vol
2
u~i u~ j
u i u j
u~ u d q
i
d q
(2.12)
d q 2 T E u d .
u i
Vol
~ ~
~
~
2 V q D 2 A C q 2 ~c Q ,
q
(2.13)
~
donde la matriz V representa las fuerzas de inercia debidas a la aceleracin
~
~
de Coriolis, A est relacionada con la aceleracin convectiva, C considera
c las
las fuerzas de inercia que surgen tras la deformacin del slido y ~
fuerzas de inercia constantes asociadas a la configuracin no deformada del
slido. El vector Q representa la fuerza generalizada debida a las cargas
~
externas aplicadas en el slido y la matriz D es la matriz de rigidez modal,
representada por una matriz diagonal que contiene el cuadrado de las
frecuencias naturales no amortiguadas del slido.
16
17
MO
OD
DE
ELO
O DIN
M
MIC
CO
O DE
DE S
S LID
DO
OS
SD
DE R
REV
VO
OL
LUC
CI
N
N FL
FLEX
XIB
BL
LES
S Y RO
OT
TAT
TO
OR
RIO
OS
S
22.3 MO
MOD
DE
EL
LO
O DI
DIN
N
M
M
MIIC
CO
OD
DE
ES
S
L
LIID
DO
O
OS
SD
DE
ER
RE
EV
VO
OL
LU
UC
CII
N
N
F
FL
LE
EX
XIB
X BL
BLE
ES
S Y RO
ROT
TA
ATO
AT
OR
RIIO
OS E
EN
NT
TR
RA
A
AY
YE
EC
CT
TO
OR
RIIA
A
GE
EN
N
R
RIIC
CA
A
Co
Conn el
e fiin
n dde oobbteen
nerr un
unaa for
f rm
mu
ul acci
nn deel ejjee ccu
uannddo
o ccirccuulaa po
p or una
u a
vva cu
curv
vaa, se va
va a ad
adop
pttarr uunaa m
meeto
oddo
oggaa bbasaadda eenn cooo
naadaass dee
olo
ordeen
ttraay
yeccttorriaa (addo
pcii
n ddee un
un sissteem
maa dee cooorrdennaaddass m
mv
vil)) que
d scurrree
op
q e dis
ssiggu
uieennd
do eel eejee de laa va
v a ((ssu po
posiici
nn es
es een
n tood
do iinsttannte con
c nooccid
daa).. L
Laa
pprrin
ncippaal hhiipttesiis ad
doopptaad
daa es
e que
q e lo
os ddesp
plaazzam
mieenntooss ddeel ejje con
c n reesspecctoo
aal sisteem
maa ddee ttraay
yeccttorriaa soonn pe
peq
queoos. P
Paaraa ell de
d esaarrroolloo ddeel mo
moddello sse han
h n
ccoon
nsiideerraddoo doss siisttem
m
mas dde reefferreenciia (veerr Fi
F guurra 22.3
3)). El
E pri
p m
meroo ess uunn
ssissteem
maa dee reefe
ferren
ncciaa fij
ijo
o X0Y0Z0, y ell sseegu
unnd
doo ees uunn sis
s steem
maa dee reefe
ferren
ncciaa
mv
vil XTYTZT qu
quee de
desccrrib
bee la trray
yeeccto
oriiaa co
occiddaa. U
Un
n vveecttoor ex
xppressaado
o enn
cono
eel siisttem
m
ma de re
f o y m
mvil see ddeennottar com
c mo
m a0 y a
refeerenncciaa fij
rresp
peecttiv
vaam
meenntee.
F
Figgu
ura
a 22.3. S
Sisteem
dee refferren
nccia
a y vvecctoorress d
dee pos
n
n. Co
Con
nfiigu
urraccin
maadaa
mass d
p sicci
n no
n defforrm
(een
n trraazo
od
disccon
ntin
nuoo) y d
defforrm
ma
ad a.
18
8
(2.14)
(2.15)
v u v u
i
~ (u + w ) + w
+Ju+
= p
w
,
u~i
u i
(2.16)
~ TT T
el tensor de velocidades angulares del sistema de
siendo
referencia mvil. Los dos primeros trminos de velocidad estn asociados al
movimiento de traslacin y rotacin del sistema de referencia mvil,
representa la velocidad del punto espacial
respectivamente. El trmino w
ocupado por la partcula debido a la deformacin del slido. El trmino
J u se debe a la velocidad de rotacin de slido rgido y el ltimo
19
2
q
vol
vol
dv 2
u~
vol
u~i u~ j
2
ui u j
~ dv q
dv
~
dv (2.17)
ui
dv 2 T
u~i
vol
vol
u i
i
i 1,3 u i
~ dv T
~T
~ dv q
T
vol
vol
~ J u dv T
~ u dv
2 T E u dv 2 T
2
vol
vol
vol
~T
~ u dv T dv T T p
0 Q .
T
vol
vol
~ ~
~ ~ ~ ~
~
~
+ 2 V 2 P q + 2 A - C + 2 S + R - B + D q
q
~ ~ ~ ~
= 2 ~c - 2 U - H + N - G + Q .
(2.18)
20
~
continuacin. La matriz P representa los efectos giroscpicos sobre el
~
slido debido a la rotacin del sistema de referencia mvil. La matriz S y el
~
vector U estn asociados a las fuerzas de inercia causadas por la
aceleracin de Coriolis, para la configuracin deformada y no deformada,
~
~
respectivamente. La matriz B y el vector N consideran las fuerzas
centrfugas debidas a la rotacin del sistema de referencia mvil, para la
~
~
configuracin deformada y no deformada. La matriz R y el vector H
determinan las fuerzas asociadas a la aceleracin angular del sistema de
referencia mvil, para la configuracin deformada y no deformada,
~
respectivamente. Por ltimo, la matriz G tiene en cuenta los efectos
centrfugos sobre el slido asociados a la traslacin del sistema mvil, y el
vector Q es el trmino de fuerzas generalizadas que actan sobre el slido.
Para ms detalles sobre el modelo dinmico de slidos rotatorios
flexibles que describen una trayectoria genrica, vase el Artculo 3. En
este artculo se ha adoptado dicho modelo para analizar el comportamiento
dinmico de un vehculo ferroviario en va curva.
21
MODELO DE VA FLEXIBLE
3 MODELO DE VA FLEXIBLE
La va es un sistema de grandes dimensiones que puede ser
considerado como una estructura infinita soportada por balasto y materiales
elastomricos no lineales. Debido a estas complejas caractersticas, en la
literatura se encuentran modelos lineales de vas infinitas [37 - 42] y
modelos no lineales de vas finitas [36, 43 - 46].
Los modelos de va de longitud infinita se basan en tcnicas de
propagacin de ondas y anlisis de Fourier que requieren adoptar hiptesis
lineales. Por otra parte, los modelos de va de longitud finita estn muy
condicionados por los fenmenos de reflexin de onda que ocurren en el
extremo del carril. Si la distancia del vehculo al extremo del carril no es lo
suficientemente grande, la onda generada en el carril por la interaccin con
el vehculo puede reflejarse en el extremo del carril e interactuar en la
dinmica del vehculo. Por lo tanto, es necesario modelar longitudes de va
considerables, aumentando el coste computacional. Adems, los intervalos
temporales de integracin deben ser considerablemente menores que el
tiempo necesario para recorrer la va.
Para evitar estos inconvenientes citados anteriormente, en la presente
Tesis se ha adoptado un modelo de va cclica desarrollada en [35]. Este
modelo cclico puede considerar elementos no lineales, evita fenmenos de
reflexin de ondas y la longitud de la va no depende del intervalo temporal
de integracin.
23
MODELO DE VA FLEXIBLE
L
x
24
M
MO
OD
DELO
O DE
DE V
VA
AF
FLE
EX
XIB
BLE
E
En
En esstee pla
p an
nteeam
am
mieennto
o see aado
op
ptta un
una tccn
niccaa de
de su
ubbeesttruucctu
urraccinn
ddoon
ndee lo
oss ccaarrriless y laas trraavvieesaass se
se mo
moddeelaann po
por seppaaraaddoo. Paarra looss ccaarrrillees
ssee co
c nssideerra uunaa reep
preesen
ntacci
nn en
n bbaasse a vig
v gaas dde T
Tim
mos
m shheenkko
o, inncclu
uyyeendoo
lla de
defo
forrm
maacin
n a fle
f exxi
nn laateerrall, ve
vertticcaal y ddeffoorm
maaccin
n a toorssinn. La
La
ddinnm
miicaa see ha anaalizzaddo
om
meediiaanttee un
un ppllan
ntteaam
miienttoo mo
moddaal a pa
l as
parttirr ddee la
pprrop
piieddaaddess m
mo
dalles dde lo
oss caarrrillees con
c n connd
dicciion
neess de
de co
onnto
od
orrnoo cccliicaass,
rresu
ulttaand
doo uun
n ssisttem
ma
m dde eccuuaaciio
oneess ddee uun
doo ddee llibbeerttadd deessacoop
plaaddo
o.
n gr
grad
Lo
Los soppo
d scrreetoos dde laa vaa sse han
m odelladdo
o coom
moo ellem
men
m nttos dee
ortees dis
h n mo
ppaarm
meetrroos vis
v sco--eelstticco
os cconcceen
ntrrad
do
oss, seeggn see obbsseervvaa en
en la
l F
Fig
gu
ura
ra 3.33.
L
Laas pllaacaass dde as
ntoo se
s reep
preessen
nttann co
mo
o ele
e em
meenntoos viiscoo-ellssticco s
asieen
com
ccoon
nceenntrraadooss a pa
parttirr ddee su
su rrig
giddeezz y aam
moorrtiiguuaam
miien
ntto, y gen
neeraan
n llass ffuuerzzaas
ddee in
inteeraacccci
nn en
entrree lo
l s carrriileess y laass ttraav
vieessass. La
Las tra
t avvieesaass see mo
m odellann
ccoom
mo
o mas
m saas con
balaassto
o cco
moo ellem
men
m ntto cconncceen
o a parrtiir
ncceenttraaddaas,, y ell ba
om
ntraddo
ddee su
su rrig
giddeez y am
amoorttig
guuaam
mieen
ntoo eq
quuiv
vaaleen
d spprecciaan
nddoo su din
d nm
miccaa.
ntee, de
Fig
3 3. D
Deeta
allle deel mo
modeelo
od
dee alg
noos dee lla va..
gurraa 3.3
a gunoos van
Al
Al ttraattarrsee de
d uun
na vvaa reecctaa, de
debid
doo a laa siim
meetrraa dee laa estrruucctu
uraa
rresp
peectto
o all ceen
es suufficcieen
ntee co
onn m
mood
deelaar uun
ntrroo dee laa vvaa, es
n soollo ccaarrrill,
rreddu
uccieen
nddo el
e co
com
mp
puutaaccioonnal. En
En eel Ar
ullo 2 ddee llaa pre
p essen
ntte Te
T esiis see
osstee co
Artcu
ccoon
nsiideerr eesttee mo
moddeelo
o dee vaa rrectta cc
c cliica pparraa eel eesttudiioo de
de la
l inntterraacccinn
ddinnm
miicaa coon
n uun
hccuulo
o ffeerrrovviiari o..
n ve
veh
25
25
M
MO
OD
DELO
O DE
DE V
VA
AF
FLE
EX
XIB
BLE
E
33.2 MO
MOD
DE
EL
LO
O DE
DE V
VA
AE
EN
NC
CU
UR
RV
VA
VA
Al
Al iigu
uaal qu
que parraa la connffig
guuraaccin
n re
rectta,, een
n vva ccurrv
vaa tam
mbbiin
n see haa
aaddopttaado
o eel plaanteeam
m en
mie
nto
o de tiip
po cc
c cliicco ddeesaarrro
olllaado
o ppor in
veesttig
gaadorre s
nv
ddeel grruup
po een eel quee en
nm
maarrcaa esstaa T
Tessiss [3
5],, ad
adappttaddoo all ccaasoo dee vaa ccu
urv
vaa.
35
E
Enn esstee caasso se han
onnsid
deeraaddo dif
d ferreent
ntees loon
ng
gittud
deess de
de vvaanoo parra el
e carrriil
h n co
iintteeriiorr y ex
xtterrio
orr, por
p r lo
o qu
que ca
cada ccarrrril tie
t enne lloong
giituud
d y pr
piieddaad
de s
prop
mod
daaleess ddiffereen
nciiaadaass.
El
E pplaannteeaam
mien
ntto de
de v
v a cc
c cliica mo
moddelaa uun
naa vva een
n ccu
vaa dee raad
dioo
urv
ccoon
nsttanntte.. P
Pu
ueed
om
moo un
naa v
va cir
c rcuullarr ppo
orr lla qque ciircu
ulaan
n un
un
de innterrppreetaarrsee cco
ccoonj
nju
unnto
o dee veh
dntticcoos a llaa m
missm
maa ve
veloocciddaad V y seep
paaraad
do s una
u a
hcuuloss id
ddisttan
ncciaa L (vveer F
Fiigu
urra 3.4
3 4)). La
Las hip
h pteessiss dee perriood
diccidaadd en
n va
v a cu
urrvaa
ppuuedeenn addoop
ptaarssee ssin
n eem
mb
barrg
goo aau
unn ccu
uannd
o llaa lon
l ng
giituud
d ddeel pperrm
m
mettro
o de
d laa
do
ccirrccun
nffereennciia no
n ees m
mlttip
plo dde laa llo
ongiitu
d ccaaraaccteersticcaa L
L. A
All iguuaal qque enn
ud
vva rec
r cta,, ddeebbid
o a la
l pe
periio
odiiccid
daad de
de laa esstrruuctuuraa y dee laass caarrgaass, eel
do
pplan
nteeaam
miienntto cc
c cliico pperrm
miitee rreedu
uccirr ell es
e tuud
dio
o ddee laa v
v a currv
un ttraam
moo
va a un
ccirrccullaar fiiniito
o dee lo
d L.
longiituud
El
E m
mo
oddelaadoo de lo
os ccarrriileess y de
de llos soop
poorttess ddiisccrrettos see lle
va a cab
c boo
l ev
ddee manneeraa an
nlo
gaa al
a dde laa vaa rec
r ctta,, de
desccrritoo en
n el appaarttaado
o an
orr. En
E eel
og
ntterrio
A
Arrttccu
ulo
o 3 de
d la ppreesennte T
Tesiis see co
on
nssid
mood
delloo de
de v
v a cccliicaa enn
deer esstee m
ccuurv
vaa ppaaraa eell eesttu
udiio
o ddee intterraaccciin
n din
d nm
m caa co
c n un
u vveh
hcuullo fer
ferroov
viaarrio
o.
mic
Fig
F gu
ura
a 3.4
3 4. M
Mo
odeeloo d
dee vva
a een cu
urrva
a.
a cc
c clicca
26
26
Subestructura
vehculo
Sistema completo
fp1
fc1
fc2
fc1
fp3
fp1
Subestructura traviesa 1
fp2
Subestructura traviesa 2
fp3
Subestructura traviesa 3
27
28
RESULTADOS
5 RESULTADOS
Los resultados que se muestran a continuacin corresponden en
primer lugar a los estudios que contribuyen a validar la metodologa
considerando amortiguamiento interno. A continuacin se muestran
resultados de la simulacin de la interaccin dinmica en recta, calculndose
las fuerzas en el contacto rueda carril y las tensiones en el cuerpo del eje.
Finalmente se obtiene la respuesta dinmica del eje en curva.
29
RE
ES
SUL
LT
TAD
DO
OS
S
T
Tablla 5..1.. P
Pro
op
pieed
dad
dees deel cilin
nd
droo.
L
Longgiitu
udd
R
Radiio
o
D
Deenssidaadd
M
M
ddu
o ddee Yo
You
unngg
ulo
T
Tassaa am
amoortig
guuaam
mien
ntoo in
nteerrno
C
Co
oefficciennte dee Pooiisssoonn
L 1.5 m
R 0.005
5m
78
7 800
0 kgg / m
E 2. 1 110
11
N /m
0.003
3
0. 3
Tab
a 5.2
5 2. Fre
F ecu
ueen
nciass n
natu
ura
d ciiliind
drro [H
Hzz]..
T bla
alees del
nnd
diccee de
de
m
mo
oddo
1
2
3
T
Tiipo
o de mo
moddo
1err m
mod
doo ddee
ffleex
xin
n
d
2do
m
modoo dee
ffleex
xin
n
3err m
mod
doo ddee
ffleex
xin
n
EF
EF
M
Modeelo viig
ga
ddee Ra
Rayyleeiggh
h
9 .114
90
9900.4
444
3 5..21
35
33660
0.2277
7 0..51
78
88005.1177
La
La F
Fiiggurraa 55.2 m
mu
ueestrra lo
oss au
uttovvaalo
orres pparraa llas trres for
form
mu
ulaaccio
onne s
ddeesccrrittass aannteerrio
m
men
ntte.. s
stooss ssee ha
han cal
c lccullaado
o coon
n uunn baarrriiddo en vel
v loociid
dadd
orm
ddee gi
giroo, aalccanzzaan
ndoo cassi el
e do
obble dde laa veelloccidaadd crrtticcaa ( crr 54
5 416 rppm
m) a
ppaarttirr ddee laa cuuaal eel mo
mod
doo ffo
orw
waarrd se
s vvu
i nesstaabblee. A lla izzqqu
uieerrdaa see
ueelvvee in
pprresseenttaa uunn diiaagrraam
maa ddee A
Arg
gaan
nd ddee lo
los aut
ovallorees,, y sse ppu
dee obbsservvaar
a to
ued
ccieerrtaa ddiisccrrep
paan
entrre mo
mod
dello
os. A la der
d reech
haa ssee m
muesstrra lla par
nciia en
p rtee reeaal dee
llos au
uttov
vaaloorress, do
dee tto
odoss lo
los mo
moddeloos des
n pprrctticcaam
meenttee la
l m
mism
maa
dond
d sccriib en
30
30
RESULTADOS
31
RESULTADOS
574
572
570
10
0
Re ()
568
Im ()
20
Modo Backward
Modo Forward
576
566
564
562
-10
Modo Forward
Modo Backward
-20
-30
560
-40
558
556
-50
-40
-30
-20
-10
Re ()
10
-50
0
20
2000
4000
6000
[rpm]
8000
10000
Figura 5.2. Izquierda: autovalores en diagrama de Argand. Derecha: parte real de los
autovalores.
,
, modos forward y backward, modelo de viga;
,
,
modos forward y backward, modelo analtico de slidos;
,
, modos forward
y backward, modelo numrico de slidos.
-5
10
90.8
90.6
[Hz]
90.2
Modo Forward
Modo Backward
90
89.8
Receptancia [m/N]
90.4
-6
10
89.6
89.4
89.2
0
-7
2000
4000
6000
[rpm]
8000
10
10000
85
87
89
[Hz]
91
93
95
32
RESULTADOS
90.8
10
90.6
[Hz]
90.2
Modo Forward
Modo Backward
90
89.8
Receptancia [m/N]
90.4
-6
10
89.6
89.4
89.2
0
-7
2000
4000
6000
[rpm]
8000
10
10000
85
87
89
[Hz]
91
93
95
33
RESULTADOS
34
RE
ES
SUL
LT
TAD
DO
OS
S
55.2 IIN
NT
TE
ER
R
RA
AC
CC
CI
N
ND
DIN
N
N
M
M
MICA
CA VE
VEH
HC
CU
U
UL
LO
O--V
VA
A EN
EN
R
RE
EC
CT
TA
A
En
En esstaa seecccci
nn see mu
muesstrraan div
d veerrsooss rreesu
ulltaadoss de la inntterraacccinn
ddinnm
miicaa vveh
hcu
vaa en tr
traam
mo
o reeccto
o co
nssid
deraan
ndoo dif
d feereenntees ffuen
nttess dee
ullo-v
on
eexxciitaacci
nn: co
corrru
ugacci
nn arm
a mnniicaa sim
s mpplle,, ru
uggoosiiddadd ppseu
uddo
d el
oalle attorriaa de
ccaarrrill, y un plaan
queeoo en
en lla bband
daa de
de rrodaaddu
uraa ddee llaa rueed
Ell m
mod
deeloo
no dde blooq
daa. E
ddee in
nteerracccci
nn din
miccaa vveh
hcu
ulo--vvaa see ddeefiinee m
meed
diaanntee laa ttccn
niccaa dee
d nm
ssuubesstrru
ucttuuraacci
n deeffinnid
daa en
n el
e aap
parrtaad
o 44.. E
El ej
ejee fllexxiblle ro
orrio
o see
n
do
rotaato
mod
deelaa m
mediiaantte la fo
orrm
mu
ulaaccinn de
de s
sliid
xiibllees dessaarrro llad
daa en
en 22.2,, y
doss flex
lla va
v a rrecctta fllexxiiblee co
con el
e pl
plan
ntteaam
miieenttoo cc
c cliico ddettaallaad
doo een
n 33.1.
El
E vveehcculoo feerrro
oviiaariio co
con siideerrad
o en
n eestte ccaasoo de
d est
e tuudio
o eess uunn
do
ccooch
hee de
d uun
naa un
uniddaad
d ET
ETR
R 50000 (ttreen
n ittalliaan
noo ddee aalta vveloocid
daad)). S
See haa
ccoon
nsiideerraddoo el tiipoo de
d v
v a r
g
gid
daa seg
s g
n lo
os ppaarm
meetrroos ddeesccrritoss en
n eel
n
pprroy
yeeccto
oE
EU
UR
R
RO
BA
AL
LT
T [[4
47].
OB
La
Las pro
p op
m daaless quuee co
con
uyeenn loos da
datoos ppaaraa eell m
mod
deeloo
pieeddaadeess mo
nsttittu
ddeel ejje se obttieen
d uun mo
mod
dello
o de
de ele
e em
meen
ntooss ffinnittoos cu
uyyaa ma
malllaa sse mu
m ueesttraa
nen de
eenn la
l F
Figuurra 55.5
onddee ad
deem
ms se
s ob
obseerrvaa laa seecciin ddee est
e tu
dioo essccog
giidaa
5, ddo
ud
ppaaraa ell c
clccu
uloo de
de llaas teen
nsiiooness aax
xiaalless ddeel ej
ejee. D
Dicchhaa sec
s ccci
nn dee esstu
uddio see
eenncu
ueenntrraa eentrre doss di
discco
os dde fr
freen
no,, aaleej ad
c am
mbbiooss ddee di
d m
o, yya que
q e
daa ddee ca
meettro
sto
os innttrooddu
uceen
ncenntraaci
nn de
de tten
nssioon
o ppueeddee co
consideerrarrsse conn
n co
con
nes quee no
pprreccisin
n coon
n lla ma
e em
meen
nto
os fin
finittoss co
nsiidderraada.
mallla dde ele
con
F
Fig
gu
uraa 5.5
5 5. Ma
ad
dee ele
e em
men
nttoss fin
f nittoss d
dell eje
f xiiblle y seecccin
nd
de esstuud
dio
o.
Malla
e e fle
35
35
RESULTADOS
324 kg
200 MN/m
Amortiguamiento
150 kNs/m
del balasto
Rigidez de la placa
1 GN/m
de asiento
Amortiguamiento de
50 kNs/m
la placa de asiento
Perfil del carril
UIC60
RESULTADOS
37
RESULTADOS
70
68
66
64
62
60
58
56
54
52
0
0.4
0.6
Distancia x [m]
0.8
1.2
110
100
90
80
70
60
50
40
30
20 Rugosidad del carril
10
0
0.2
0.4
0.6
Distancia x [m]
ISO 3095
0.8
1.2
10 x ISO 3095
Figura 5.6. Fuerza de contacto rueda-carril cuando el vehculo circula sobre una va
corrugada con longitud de onda 60 mm. Arriba: efecto de la velocidad y de la
flexibilidad del eje con amplitud de corrugacin segn ISO 3095. Abajo: efecto de la
amplitud de corrugacin para una velocidad de 300 km/h y eje flexible.
RESULTADOS
[MPa]
y
[MPa]
y
V = 300 km/h
10 x ISO 3095
Figura 5.7. Tensin axial en la seccin de estudio del eje cuando el vehculo circula
sobre una va corrugada con longitud de onda 60 mm. Izquierda: dos velocidades de
vehculo, amplitud de corrugacin segn lmite de ISO 3095. Derecha: dos amplitudes
de corrugacin (resultados indistinguibles).
39
RESULTADOS
kQ
Qd ,max
,
Qs
(5.1)
d ,max
,
s ,max
(5.2)
RESULTADOS
1.8
Sin corrugacin
Corru. excita modo pin-pin
1.8
1.7
1.6
1.6
1.5
1.5
1.4
1.4
1.3
1.3
1.2
1.2
1.1
1.1
1
0
1
50
100
150
200
250
300
0.9
0
350
50
100
150
200
250
300
350
41
RESULTADOS
42
RESULTADOS
100
90
80
70
60
50
40
30
20 Rugosidad del carril
10
0
0.2
0.4
0.6
0.8
1.2
Distancia x [m]
Eje flexible V = 300 km/h
1.4
1.6
1.8
250
250
200
150
Vista A
100
50
0.5
Distancia x [m]
200
Vista A aumentada
150
100
50
1.5
0.1
0.2
0.3
Distancia x [m]
0.4
0.5
43
RESULTADOS
[MPa]
y
[MPa]
y
Vista aumentada
Vista aumentada
Impacto
plano rueda
[MPa]
y
Impacto
plano rueda
Figura 5.11. Tensin axial en la seccin de estudio del eje cuando el vehculo circula a
300 km/h sobre una va corrugada aleatoriamente (izquierda), sobre un carril
perfectamente uniforme en presencia de un plano de rueda de 50 mm a 50 km/h
(centro) y 300 km/h (derecha).
44
RESULTADOS
4.5
1.5
Corrugacin aleatoria
Plano de rueda
1.45
1.4
3.5
1.35
1.3
3
2.5
1.25
1.2
1.15
1.1
1.5
1
0
1.05
50
100
150
200
250
300
1
0
350
50
100
150
200
250
300
350
45
RESULTADOS
46
RESULTADOS
47
RESULTADOS
Tabla 5.4. Fuerzas de contacto rueda-carril estacionarias para el planteamiento multicuerpo rgido del vehculo completo (software ADTreS [49]) y el modelo de eje
flexible.
Fuerza vertical
Fuerza lateral
Fuerza
longitudinal
Rueda exterior
Modelo
Modelo eje
rgido
flexible
multicuerpo
69.57 kN
70.01 kN
5.68 kN
5.91 kN
14.39 kN
15.16 kN
Rueda interior
Modelo
Modelo eje
rgido
flexible
multicuerpo
49.82 kN
49.71 kN
-3.75 kN
-3.99 kN
-14.39 kN
-15.16 kN
RESULTADOS
75
70
65
60
55
50
45
40
0
10
15
20
25
Tiempo t [ms]
30
35
40
45
Figura 5.13. Fuerza de contacto rueda-carril vertical cuando el vehculo circula a 150
km/h sobre una va curva corrugada con longitud de onda 60 mm. Amplitud de
corrugacin corresponde al lmite de ISO 3095.
RESULTADOS
-2
-4
Rugosidad carril interior
-6
0
10
15
20
25
Tiempo t [ms]
30
35
40
45
RESULTADOS
120
100
80
70
60
50
40
Rugosidad carril interior
20
0
0
10
Tiempo t [ms]
15
20
25
Figura 5.15. Fuerza de contacto rueda-carril vertical cuando el vehculo circula a 150
km/h sobre una va curva con rugosidad aleatoria en los carriles.
51
RESULTADOS
6
4
2
0
-2
-4
-6
10
Tiempo t [ms]
15
20
25
25
20
15
10
5
0
-5
-10
-15
-20
-25
Rugosidad carril exterior
10
Tiempo t [ms]
15
20
25
52
RESULTADOS
RESULTADOS
250
250
200
200
Vista A
150
100
100
50
50
Vista A aumentada
150
10
20
30
40
Tiempo t [ms]
50
60
Tiempo t [ms]
10
12
Figura 5.18. Fuerza de contacto rueda-carril vertical cuando el vehculo circula a 150
km/h sobre una va curva perfectamente uniforme en presencia de un plano de rueda
de 50 mm.
54
RESULTADOS
35
35
30
25
30
Vista A
20
15
10
5
0
Vista A aumentada
20
15
10
5
0
-5
-5
-10
25
10
20
30
40
Tiempo t [ms]
50
-10
60
Tiempo t [ms]
10
12
60
60
40
80
Vista A
20
-20
-40
10
20
30
40
Tiempo t [ms]
50
40
20
-20
-40
60
Vista A aumentada
Tiempo t [ms]
10
12
55
RESULTADOS
180
180
160
160
140
Vista A
120
100
80
60
40
20
0
0
140
Vista A aumentada
120
100
80
60
40
20
10
20
30
40
Tiempo t [ms]
50
60
Eje flexible
Tiempo t [ms]
10
12
Eje rgido
Figura 5.21. Fuerza de contacto rueda-carril vertical cuando el vehculo circula a 150
km/h sobre una va recta perfectamente uniforme en presencia de un plano de rueda
de 50 mm.
56
6 CONCLUSIONES Y DESARROLLOS
FUTUROS
En este apartado se recogen las principales conclusiones extradas de
la Tesis Doctoral aqu presentada y se enumeran posibles desarrollos futuros
que continen la lnea de investigacin en la que se enmarca.
6.1 CONCLUSIONES
Se ha desarrollado una metodologa completa para modelar la
dinmica de slidos elsticos rotatorios con geometra de revolucin, tanto
cuando describen una trayectoria recta como una trayectoria genrica. Este
modelo resulta adecuado para aquellos casos donde el inters yace en los
puntos espaciales del slido ms que en los puntos materiales. Por lo tanto,
es especialmente til para estructuras complejas donde hay interaccin entre
slidos no rotatorios y rotatorios, como el caso de la interaccin eje
ferroviario-va.
El mtodo est basado en un planteamiento modal Euleriano donde
la base modal del sistema est formada por las propiedades modales del
slido no rotatorio, aprovechando la simetra de revolucin. La formulacin
final consiste en un conjunto de ecuaciones diferenciales ordinarias y
lineales cuyos coeficientes son independientes del tiempo. En consecuencia,
las matrices se calculan nicamente al principio de la simulacin,
reducindose considerablemente el coste computacional. En el caso de
describir una trayectoria curva, se ha adoptado un sistema de coordenadas
de trayectoria que se mueve segn el eje de la va y permite modelar el
movimiento de arrastre. Aadiendo adems el movimiento vibratorio del eje
respecto a la va segn el planteamiento modal Euleriano, se ha obtenido la
ecuacin de movimiento del eje en curva con un coste computacional muy
reducido.
Inicialmente el modelo de slidos se ha aplicado para estudiar la
dinmica de un cilindro rotatorio biarticulado, geometra sencilla que
57
59
60
RESUMEN DE ARTCULOS
7 RESUMEN DE ARTCULOS
7.1 ARTCULO 1
Dynamics of damped rotating solids of revolution through an Eulerian
modal approach.
En este artculo se ha presentado una tcnica para modelar la
respuesta dinmica de slidos flexibles rotatorios con amortiguamiento
interno modal. El mtodo es aplicable a slidos de revolucin que giran a
velocidad angular constante sobre su eje de axisimetra. En el modelo se
considera un sistema de coordenadas modales Eulerianas, el cual adopta
como base los modos de vibracin del slido no rotatorio en un sistema de
referencia fijo. Este planteamiento parte de un modelo Lagrangiano
desarrollado previamente, por lo que se desarrollan las relaciones
matemticas entre ambos conjuntos de coordenadas. Considerando el
sistema de coordenadas utilizado, el modelo es muy adecuado para analizar
la interaccin dinmica entre slidos rotatorios y estructuras no rotatorias,
adems de permitir la obtencin de Funciones de Respuesta en Frecuencia.
La metodologa propuesta slo puede ser formulada analticamente en casos
sencillos, como modelos de vigas. Por ello se ha desarrollado una
metodologa numrica para analizar estructuras ms complejas con
geometra genrica de revolucin, siendo el Mtodo de Elementos Finitos la
tcnica numrica adoptada para obtener las matrices asociadas con la
ecuacin de movimiento del slido flexible rotatorio.
Con el modelo presentado se ha estudiado el comportamiento
dinmico de un cilindro rotatorio biarticulado y amortiguado internamente,
y se ha analizado su estabilidad, obtenido su receptancia y sus propiedades
modales. Esta geometra simple proporciona una solucin de referencia que
ha permitido comparar la formulacin analtica de la teora de viga rotatoria
de Rayleigh con los resultados del modelo propuesto.
61
RESUMEN DE ARTCULOS
7.2 ARTCULO 2
Numerical estimation of stresses in railway axles using a train-track
interaction model.
En este artculo se ha descrito una metodologa numrica para la
estimacin de tensiones en ejes ferroviarios originadas por la interaccin
dinmica entre el vehculo ferroviario y la va. Como modelo de interaccin
dinmica se ha adoptado una tcnica de subestructuracin, que divide el
sistema en tres tipos de subestructuras: el vehculo, los carriles y sus
soportes. Del vehculo ferroviario se han modelado las masas no
suspendidas, donde el eje es un slido flexible rotatorio modelado mediante
un planteamiento Euleriano. Los carriles se han modelado como vigas de
Timoshenko, considerando la dinmica vertical, lateral y torsional. Por
ltimo, los soportes (placas de asiento, traviesas y balasto) se han
representado como sistemas de parmetros concentrados, en los que las
traviesas reciben un tratamiento de masas concentradas, y las placas de
asiento y el balasto como elementos visco-elsticos.
Para llevar a cabo el clculo de fuerzas de contacto rueda-carril y de
tensiones dinmicas en el eje, se han considerado diversas fuentes de
excitacin: corrugacin armnica, rugosidad pseudoaleatoria del carril, y
plano de rueda. Adems, se han analizado diferentes longitudes de onda de
corrugacin que excitan resonancias del eje y de la va, niveles de rugosidad
(referenciadas al lmite de la UNE-EN ISO 3095) y velocidades del
vehculo. Para todos los escenarios de clculo se han obtenido las fuerzas de
contacto rueda-carril y las tensiones en el eje. Por ltimo, se han evaluado
los factores de amplificacin dinmica para las fuerzas de contacto y las
tensiones axiales, con el fin de cuantificar la importancia de los efectos
dinmicos en la vida a fatiga del eje ferroviario.
62
RESUMEN DE ARTCULOS
7.3 ARTCULO 3
A comprehensive model of the railway wheelset-track interaction in curves.
La interaccin vehculo-va ha sido ampliamente estudiada en los
ltimos 40 aos, dando lugar a planteamientos de modelado que pueden
describir satisfactoriamente muchos problemas dinmicos que surgen en la
interfaz rueda-carril. Sin embargo, los modelos disponibles generalmente no
consideran de manera especfica la dinmica del vehculo circulando sobre
una va curva. Dicha dinmica tiene asociada una serie de fenmenos de
interaccin vehculo-va caractersticos de la circulacin en curva que la
hacen merecedora de un tratamiento en profundidad.
El objetivo de este artculo es definir un modelo de eje flexible
circulando sobre una va curva flexible. La principal novedad de este trabajo
es combinar un sistema de coordenadas de trayectoria con coordenadas
modales Eulerianas; el sistema de trayectoria permite considerar la va
curva, y las coordenadas modales modelar los pequeos desplazamientos
relativos entre el sistema de trayectoria y el slido. El modelo de eje est
acoplado al modelo cclico de va curva de radio constante mediante un
modelo de contacto rueda-carril, que considera la geometra real de los
perfiles en contacto y la relacin no lineal entre las velocidades y fuerzas de
pseudodeslizamiento.
El mtodo propuesto puede ser usado para analizar una variedad de
problemas dinmicos de vehculos ferroviarios, incluyendo la dinmica
originada por la rugosidad del carril o bien un plano de rueda, la generacin
de ruido de rodadura y estimaciones de las cargas de servicio en ejes. En
este artculo se aplica el modelo propuesto para algunas condiciones de
circulacin en curva, y se presentan resultados de simulacin de casos
realistas de interaccin vehculo-va, sealando la importancia de los efectos
de circulacin en curva abordados especficamente en este trabajo.
63
BIBLIOGRAFA
BIBLIOGRAFA
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
K. Popp, H. Kruse, I. Kaiser, Vehicle-track dynamics in the midfrequency range, Vehicle System Dynamics 31 (1999) 423-464.
[9]
BIBLIOGRAFA
66
BIBLIOGRAFA
67
BIBLIOGRAFA
BIBLIOGRAFA
69
Parte II
Artculos
aArtculo 1a
n de Tecnologa de Vehculos, Universidad Politecnica de Valencia, Camino de Vera s/n, 46022 Valencia, Spain
Centro de Investigacio
nica de Valencia, Universidad Politecnica de Valencia, Camino de Vera s/n, 46022 Valencia, Spain
Instituto de Biomeca
article info
abstract
Article history:
Received 12 January 2011
Received in revised form
24 July 2011
Accepted 1 October 2011
Handling Editor: S. Ilanko
Available online 24 October 2011
This article presents a technique for modelling the dynamic response of rotating exible
solids with internal modal damping. The method is applicable to solids with geometry
of revolution that rotate around their main axis at constant spinning velocity. The
model makes use of an Eulerian modal coordinate system which adopts the vibration
modes in a non-rotating frame as basis functions. Due to the coordinate system, the
technique is particularly suitable for studying the dynamic interaction between rotating
solids and non-rotating structures and permits to obtain Frequency Response Functions.
The current investigation presents the development of the proposed technique from a
previous Lagrangian model, and consequently the mathematical relationships between
the two coordinate sets are found. The approach has been adopted to study the
dynamics of a simply supported cylinder including damping in order to obtain the
receptance function and the modal properties of the rotating solid.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
The dynamics of damped shafts was investigated in early works at the rst decades of the twentieth century [1,2]. The
unstabilising consequence of the internal damping is a well-known phenomenon that produces a critical angular velocity
above which the rotor becomes unstable (see monographs in Refs. [3,4]). The literature shows many articles that analyse
internal damping in Jeffcott rotors and rotating beams but few reported attempts have been found regarding the dynamic
models for generic geometries.
The model of exible rotating solids with generic geometry has to dene different shaft section properties (e.g. axles
with non-constant cross-sections) and sometimes it needs to be based on a three dimensional domain (e.g. railway
wheelsets; Jeffcott rotors and rotating beams are based on a zero- and one-dimensional domains, respectively). The Finite
Element (FE) Method is possibly the numerical approach that can be more easily adapted to a generic geometry. FE Method
has been adopted through beam elements [510], and 3D solid elements [1113]. The main inconveniences of this
technique are found when the solid interacts with a non-rotating structure. The force that the non-rotating system exerts
on the rotating one is applied at xed spatial points, and consequently there is a relative motion between the load and the
rotating solid (moving load problem). This fact requires the computation of the contribution of the external force to the
generalised force term in each integration step during the simulation. Furthermore the Frequency Response Function (FRF)
that correlates the steady response to harmonic excitation produced by a non-mobile force cannot be obtained directly. In
addition, a constant and non-mobile force applied on an external surface of a rotating FE mesh produces a parametric
n
Corresponding author.
E-mail address: baeza@mcm.upv.es (L. Baeza).
0022-460X/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsv.2011.10.003
869
excitation because the force has different effect if it is applied in a node or in the middle of a FE solid. The Jeffcott rotor
model and the rotating beam model solve this difculty because they are based on zero-dimensional and one-dimensional
domains, respectively. Moreover, the displacements and velocities of these models are obtained in rotating or non-rotating
coordinates indistinctly.
The work presented in Ref. [14] proposed a method that solved the above mentioned problems. The method is applied
to undamped solids with geometry of revolution; it is based on Eulerian modal coordinates and provides satisfactory
results. Lagrangian coordinates are the most frequently adopted reference in classic Mechanics; by means of this
coordinate system, the position vectors and other kinematic properties are associated with material points of the solid.
Eulerian coordinates are commonly used in Fluid Mechanics, and they relate spatial points through a xed coordinate
frame. These coordinates determine spatial points through which the material ows as time passes. The technique exploits
the properties of the solids of revolution whose mode shapes in a xed coordinate frame do not depend on the rotation of
the solid. Consequently the mode shapes of the solid are used as basis functions in order to dene a generic displacement
of the exible solid.
The present work develops a methodology based on Ref. [14] for modelling the dynamic behaviour of rotating solids of
revolution with internal modal damping. Results of the proposed model are presented for a simply supported rotating
cylinder. This simple geometry provides a reference solution which permits to compare the analytical formulation by
means of the rotating Rayleigh beam theory with the results from the proposed method. The formulation associated with
damped rotating beams which are needed in this article was developed in Ref. [3] and can be found in the Appendix. For
further details in the dynamic behaviour of rotating solids with complex geometry, readers are referred to Ref. [14].
Section 2 presents the development of the equation of motion associated with the proposed model. A modal approach is
carried out and consequently the modal properties of the non-rotating solid need to be obtained. Section 3 builds the
matrices of the equation of motion from the analytical modes of the simply supported Rayleigh beam. At this point, the
differences of the new formulation in comparison with the well-established methods (like spinning beam theories) can
be found.
The proposed methodology can only be formulated analytically in simple cases, e.g. beam models. In Section 4, a
numerical methodology is given to analyse more complex structures such as railway wheelsets, non-vented disc brakes or
other solids with generic geometry of revolution. Given that theoretical modal analysis of general structures is usually
made with the nite element method, this technique is adapted in this paper to obtain the matrices associated with the
equation of motion.
The results are presented in Section 5 of this article, where the FRF of the rotating cylinder is obtained and the stability
analysis is carried out. The main conclusions of this paper are remarked in Section 6.
2. Proposed model
The development of the proposed model is based on a previous article from Brown and Shabana [15]. This work models
rotating undamped solids and makes use of the oating frame of reference, which is a coordinate system described in
detail in Ref. [16]. The proposed method develops an Eulerian modal coordinate system, and the present section shows its
properties and a procedure to derive this coordinate system from the oating frame of reference. This mathematical
development leads the equation of motion from the model presented in Ref. [15] through a change of variables.
(1)
where u0 is the position vector of the particle in the undeformed conguration, the product U(u0 )p(t) corresponds to the
displacement of the particle due to the elastic deformation of the solid in the rotating frame, U(u0 ) contains the massnormalised mode shapes of the non-rotating solid, p(t) is the vector of modal coordinates and A is the rotation matrix,
870
X
X
t
Y
Z
Y
Z
Fig. 1. Coordinate frame systems.
dened by
cos y
B
A @ sin y
0
sin y
cos y
0
0
cos Ot
C B
0 A @ sin Ot
1
0
0
sin Ot
cos Ot
0
C
0 A,
(2)
(3)
r v Uvqt,
(4)
if the particle occupies the spatial position v in the undeformed conguration, where q(t) is dened the Eulerian-modal
coordinate vector.
From Eqs. (1) and (4), it follows:
(5)
Eq. (5) is multiplied by r U(v) , where r is the density, and an integration is then carried out over the volume of the
solid. Due to the orthogonality of the modes, the integral yields
ZZZ
qt
rUvT AUAT vdn pt:
(6)
T
Volume
Eq. [6] provides the coordinate transformation, which can be written as follows:
qt Btpt:
(7)
Volume
(9)
Volume
(10)
871
~ that is
the transformation matrix B(t) has no inuence if it is applied to the matrix M,
~
~
BtT MBt
M:
(11)
ZZZ
~
BtT
rUu0 T MUu0 du Bt
BtT MBt
Volume
ZZZ
rBtT Uu0 T MUu0 Btdu
ZZZVolume
rBtT UvT MUvBtdn
ZZZ Volume
rUu0 T AT MAUu0 du
ZZZ Volume
~
rUu0 T MUu0 du M:
(12)
Volume
UvBt AUAT v:
(13)
Each side of the last equation is multiplied by itself (transposed) and the density, giving
Volume
(14)
(15)
Considering now Eq. (15), the property obtained in Eq. (8), and the mode shapes orthogonality, the relation B(t)TB(t)I
is found and consequently B(t) is an orthogonal matrix.
2.1.5. Structure of the transformation matrix B(t)
Let us consider different nth and mth modes and the corresponding mode shapes U o n 4 and U o m 4 . Eq. (13) can be
written for the nth mode as follows:
(16)
om4
(17)
Volume
If the modes U/nS and U/mS are orthogonal, due to the geometry of revolution AUAT v/nS and U/mS are also
orthogonal, and consequently Bmn 0.
The rows and columns of matrix B(t) associated with modes with multiplicity 1 have null entries except the diagonal
entry, which is equal to 1. Due to the orthogonality property of B(t), the diagonal blocks associated with modes with
multiplicity 2 contain rotation matrices, that is
2
3
0
0
0 0
6
7
&
^
^
^
&
^
6
7
6
7
6
0
0
0 07
6
7
60 0
7 siny 0 0 7
cosy
6
7
B6
(18)
7:
6 0 0 8siny
cosy
0 07
6
7
60 0
7
0
0
6
7
6
7
4^ & ^
5
^
^
&
0
(19)
is veried if the pair of entries associated with each mode with multiplicity 2 are equal. The proof is found through the
structure of the matrix B(t) showed in the previous section.
872
(20)
(21)
where
J Ay AT AT Ay ,
Ay being the derivative of the matrix A with respect to y, that is
0
sin y cos y
B
Ay @ cos y sin y
0
The development of the Eq. (23) shows that
0
B
J@1
0
1
0
0
(23)
C
0 A:
(24)
1
0
C
0 A:
(25)
Some vibration modes produce deformed shapes which may permit centrifugal forces to produce work. This effect is
considered through the matrix E~ which is calculated as
ZZZ
rUT EUdu,
(26)
E~
Volume
where
E Ayy AT AT Ayy ,
Ayy being the second derivative of the matrix A with respect to y, that is
1
0
sin y
0
cos y
C
B
Ayy @ sin y cos y 0 A:
0
It can be easily obtained that
1
B
E@0
0
(27)
(28)
0
1
C
0 A:
(29)
If the external forces are applied in xed material points, the generalised force term is computed by
ZZZ
Uu0 T AtT f v u0 ,tdu,
Q p t
Volume
(31)
873
where fv(u0 ,t) are the external volume forces. However, if the external forces gv(v,t) are applied in xed spatial points, the
generalised force term can be calculated from the following expression
ZZZ
Q p t
UAtT vT AtT gv v,tdn:
(32)
Volume
(33)
Introducing Eq. (33) into Eq. (21), the equation of motion of the internally damped solid is
~ 2OJ
~ p_ K
~ O2 Ep
~ Q O2 L:
~
p D
p
(34)
~ T 2OBJB
~ T 2BB_ q_ BKB
~ T O2 BEB
~ T BB BD
~ B_ 2OBJ~ B_ q
q BDB
2 ~
BQ O BL:
T
(35)
Due to the properties stated in Sections 2.1.3 and 2.1.6, the following equations are found:
~ T J,
~
BJB
(36)
~ T E,
~
BEB
(37)
~
~
BKB
K,
(38)
~
~ T D,
BDB
(39)
~
BL~ L:
(40)
(41)
_
_ vT A UvT A:
B_ Uu0 T U
(42)
_ v
U
3
X
@Uv dvi
i1
@vi
dt
3
X
@Uv dAu0
i1
@vi
dt
3
X
@UvT
i1
@vi
3
X
@UvT
i1
@vi
3
X
@UvT
i1
@vi
3
X
@Uv
i1
@vi
OAy u0 i O
3
X
@Uv
i1
@vi
Jvi :
(43)
!
Jvi AUu0 BT OUvT Ay Uu0 BT :
(44)
!
Jvi UvBBT OUvT Ay AT UvBBT
!
Jvi Uv OUvT JUv:
(45)
Now Eq. (45) is multiplied by the density r and it is integrated on the volume of the solid. Applying the orthogonality
property, yields
!
ZZZ
3
X
@UvT
_ TO
~
Jvi Uvrdn OJ:
BB
(46)
@vi
Volume i 1
874
Considering now the rst property in Section 2.1.3, the integral in Eq. (46) does not depend on time. With the denition
!
ZZZ
3
X
@Uu0 T
Ju0 i Uu0 rdu,
(47)
G~
@ui
Volume i 1
Eq. (46) turns into
_ T OG~ J
~
BB
(48)
T
The calculation of B B is performed as follows. The product B_ BT is constant and consequently
d _ T T _ _T
BB BB B B :
dt
(49)
(50)
Further analysis shows that G~ is antisymmetric, and Eq. (50) can be written as
T B_ B_ T BB
_ T BB_ T O2 G~ J
~ G~ J:
~
BB
The generalised force of the Eq. (35) is, from Eq. (32)
ZZZ
Q q t BQ p t B
V olume
Volume
UvT gv v,tdn:
(51)
(52)
(53)
In standard cases, the vector associated with external volume forces can be written as
gv v,t Fv vgt,
and the generalised force expression is
ZZZ
Q q t
Volume
(54)
(55)
Now Eqs. (36)(40), (48), (51) and (55) are substituted into Eq. (35), yielding
~ q_ K
~ OD
~
~ G~ Jq
~
~
~ O2 C
q D2
F~ gt O2 L,
OG
(56)
~ J~ J
~ J~ G
~ E:
~
C~ G~ G~ G~ J
(57)
where
~ J,
~ F~ and L~ are constant matrices and they are
~ G,
~ K,
~ E,
~ C,
Attention must be drawn to the fact that the matrices D,
computed at the beginning of the simulation. An advantage of this Eulerian approach is that, if external forces do not rotate
with the solid, the generalised force vector in Eulerian coordinates Qq(t) is calculated by integrating the volume of the solid
independently of time. Therefore, it may be also calculated at the beginning of the simulation, unlike in the Lagrangian
approach where the generalised force term Qp(t) must be evaluated at each time instant by using a costly integration over
the volume. Moreover, the equation of motion (56) is linear and it can be adopted for obtaining the FRFs associated with
forces or displacements that are applied or measured at xed spatial points.
Finally, it is worth noting that Eq. (56) is expressed in a non-rotating reference frame whereas the Lagrangian methods
such as Finite Element Method or Eq. (34) are based in a oating frame.
3. Application of the method to a simply supported beam
The parameters of the beam that are considered in the model are the radius R, length L, Youngs modulus E, crosssectional area A and second moment of area I. In order to nd a parallelism between the proposed method and beam
theories, in the following the variables of the rotating Rayleigh beam formulation which are presented in the Appendix are
adopted.
If N bending modes of the Rayleigh beam are consider, the matrix of mode shapes U becomes
3
2
f1 z
0
fN z
0
6
0
f1 z
0
fN z 7
7,
(58)
U6
5
4
df1
d f1
d fN
fN
x dz y dz x dz y ddz
875
where fj is computed from Eq. (A.1). Taking into account the modal gyroscopic term gj dened in (A.8) and the modal
functions detailed in Eq. (58), the matrices that dene the equation of motion for rotating solids, Eq. (56), become
3
2
0 g 1 0
0
6g
0
0
0 7
7
6 1
7
6
~
^
^
& ^
^ 7,
(59)
G 6
7
6
7
6 0
0
0
g
4
N5
0
0
gN
0
2
0
6 1g
6
1
6
^
J~ 6
6
6 0
4
1g 1
0
^
&
0
^
1g N
0
2
6
6
6
E~ 6
6
6
4
7
7
7
7,
7
1g N 7
5
0
0
^
1g 1
1g 1
&
0
0
0
0
1g N
0
0
1g N
(60)
3
7
7
7
7,
7
7
5
(61)
L~ 0,
and
g 1
6 0
6
6
^
C~ 6
6
6 0
4
0
(62)
g 1
&
g N
0 7
7
7
^ 7:
7
0 7
5
g N
(63)
The matrices of Eqs. (59)(63) uncouple the inuence between the modes in Eq. (56). Therefore, Eq. (56) can be written
for each pair of modal coordinates associated with the jth mode qj, that is
j
j
q 2Og j G cj Iq_ o2j O2 g j I 2Ocj Gqj Q j ,
(64)
where all the terms are dened in the Appendix. Eq. (64) is identical to Eq. (A.6) with the exception of the term O gj that
accounts for the effect of the centrifugal forces when the section is deformed. The one-dimensional Rayleigh beam model
does not consider this effect and therefore, some differences are expected in the results. Further discussion will be
provided in Section 5.
The eigenvalues of the Eq. (64) are
q
lj,1 cj ig j O c2j o2j g j O2 g 2j O2 2iOcj g j 1,
q
lj,2 cj ig j O c2j o2j g j O2 g 2j O2 2iOcj g j 1,
2
lj,3 lnj,1 ,
lj,4 lnj,2 :
(65)
Instable behaviour happens when the rst eigenvalue in Eq. (65) has positive real part. It is then concluded that the
critical speed is
1
,
Ocr p
1g 1
(66)
such critical velocity being lower than that predicted by the one-dimensional beam model (A.11).
4. Computational method
In general, an analytical expression of the mode shapes cannot be obtained and consequently, numerical techniques
like the FE Method are used to compute the modal properties. This section presents a method for obtaining the matrices of
the equation of motion in Eq. (56) from the FE modal solution.
FE modal analysis provides the modal solution in the nodes of the solid mesh. The modal matrix UFE is obtained with
the FE model, whose columns contain the values given by U(u) on each node.
876
Uu Ne uUFE ,
(67)
where the matrix Ne(u) contains the shape functions (or basis) of the eth element.
Bearing in mind the nodal interpolation dened in Eq. (67), the matrices of the equation of motion in Eulerian
coordinates, Eq. (56), are shown below
!
ZZZ
ne ZZZ
ne ZZZ
X
X
J~
rUT JUdv
rUT JUdv UTFE
rNTe JNe dv UFE ,
(68)
Volume
Ve
e1
e1
ne ZZZ
X
E~ UTFE
Ve
e1
L~ UTFE
G~ UTFE
ne ZZZ
X
e1
Ve
Ve
rNTe Eudv ,
3
X
@Ne uT
i1
(69)
ne ZZZ
X
e1
Ve
@ui
(70)
!
Jui Ne dv UFE ,
(71)
where ne is the number of elements in the FE mesh, and Ve is the volume of the eth element.
5. Results
This section analyses the stability, modal properties and FRF of the simply supported rotating exible cylinder with
internal damping. The aim of the present section is to compare the results from three different approaches: (1) the
Rayleigh rotating beam model (in the following, the beam model) whose formulas can be found in the Appendix of the
present paper; (2) the proposed method in which the modes are computed analytically (analytical solid model); and (3)
the proposed method where the computations are performed numerically (numerical solid model).
The present study considers the rst bending mode, the forward mode being unstable (further details can be found in
Ref. [3] and will be also shown later). The properties of the simply supported cylinder are dened in Table 1. The rst three
natural frequencies associated with bending modes are given in Table 2. The cylinder mesh considered in the numerical
solid model is depicted in Fig. 2. The FE model implements three-dimensional linear elements.
Table 1
Cylinder properties.
Length
Radius
Density
Youngs modulus
Internal damping ratio
Poissons ratio
L 1.5 m
R 0.05 m
r 7800 kg/m3
E 2.1 1011 N/m2
x 0.03
n 0.3
Table 2
Natural frequencies of the cylinder (Hz).
Mode
number
Mode type
FE
Rayleigh
beam model
1
2
3
90.14
355.21
780.51
90.44
360.27
805.17
877
Fig. 3 plots the eigenvalues in the Argand diagram for the three formulations previously indicated. These have been
calculated with a spin speed sweep, reaching almost twice the critical speed. The differences between the models are analysed
in the following gures. Fig. 4a shows the real part of the eigenvalues. All the models describe virtually the same evolution of
the real part of the eigenvalues. In addition, it is clearly shown that the real part of the backward eigenvalue is always negative
and therefore stable, while the real part of the eigenvalue associated with the forward mode is positive beyond critical speed
and hence becomes unstable. For the forward mode, Fig. 4b presents the real part of the associated eigenvalue against the ratio
angular velocity/forward natural frequency. The instability of the forward mode takes place if the angular velocity is higher
than the forward mode frequency, in the absence of external damping as in the case under analysis.
The critical speed is now shown in Fig. 5 for several internal damping ratios x. The critical velocity does not depend on
the internal damping rate in the absence of external damping [17]. These results are in accordance with the conclusions of
rotating shaft literature [3].
576
Backward modes
Forward modes
574
572
570
Im ()
568
566
564
562
560
558
556
-50
Fig. 3. Real and imaginary part of the eigenvalues:
backward modes, analytical solid model;
,
-40
-30
-20
-10
Re ()
10
20
,
, forward and backward modes, beam model;
, forward and backward modes, numerical solid model.
, forward and
20
Stable
10
Unstable
15
Re ()
5
0
-5
-10
-15
-20
,
, forward and backward modes, beam model;
, forward and backward modes, numerical solid model.
878
, beam model;
The discrepancies between the models are associated with the imaginary part of the eigenvalues as can be seen in
Fig. 6. The evolution of the natural frequency with the angular velocity (Campbell diagram) for the three formulations is
shown. The analytical and numerical solid models predict virtually the same evolution, but differ in comparison with the
beam model. As can be seen, this discrepancy increases with spin speed. Although the difference between models is less
than 0.5%, the proposed solid model describes a very different evolution and trend compared with the beam model.
In the introduction section of this article, it was stated that one advantage of the proposed solid model is to allow the
calculation of FRF associated with forces or displacements that are applied or measured at xed spatial points. The FRF for
the simply supported cylinder is shown in Fig. 7 for different internal damping ratios. The force is applied transversely in
the central node of the cylinder, and the displacement is measured at the same point and same direction. It can be seen
that the backward bending mode is highly damped and its effect disappears for x 0.01 or higher. In this way, the steady
Fig. 7. Direct FRF (receptance) (numerical solid model, O 5000 rev/min) for different internal damping ratios:
, x 0.01;
, x 0.03.
, beam model;
879
, x 0;
, x 0.001;
response is only inuenced by the forward bending mode. The same case is calculated through the three formulations for
x 0.03 and the associated FRFs are shown in Fig. 8. As observed previously, the analytical and numerical solid models
predict the same behaviour with undistinguishable curves, but they differ from the beam model. It is worth noting that the
backward mode has disappeared, remaining only the forward bending mode.
It is important to emphasise the effect done by the centrifugal forces associated with the deformation, being the
fundamental difference between the proposed solid model and the one-dimensional beam model. This effect is
represented by the term O2 gj for the analytical solid model in Eq. (64) or its equivalent O2 C~ in the solid model from
Eq. (56), and it does not appear in the beam model. The proposed model by means of analytical data produces the same
equation of motion in Eq. (64) than the rotating beam model in Eq. (A.6) if the term O2 gj is neglected in the former model.
Therefore, the following calculations show the results from the rotating Rayleigh beam model and the proposed numerical
solid model where the term O2 C~ is omitted.
The solid model showed not only the discrepancies in the imaginary part but also a quite different tendency in
comparison with the beam model. Once neglected the effect done by the centrifugal forces associated with the solid
deformation, both models describe a very similar evolution of the natural frequency, as shown in Fig. 9.
Finally, Fig. 10 shows the FRF when the inuence of the centrifugal forces is neglected. As can be seen the predictions
associated with both models are almost undistinguishable, with overlapped FRF curves in the frequency range considered.
880
Fig. 10. Direct FRF (receptance) for O 5000 rev/min and x 0.03.
, beam model;
6. Conclusions
A complete methodology to model the dynamics of damped elastic solids of revolution rotating about their axis of
revolution has been developed in this article. This model is designed for those cases in which the interest lies in spatial
points of the solid rather than material points. Therefore, it is especially suitable for complex systems consisting of rotating
and non-rotating solids that are in mutual interaction.
The method is based on a modal approach where the modal properties of the non-rotating solid form the modal basis of
the system. The nal formulation consists of a set of linear ordinary differential equations where the coefcients are time
independent. Therefore the matrices are calculated only once at the beginning of the simulation, leading to a considerable
reduction of the computational cost.
The proposed method can be applied to the study of the stability, to obtain the equivalent modal properties of the
rotating solid and also to compute the FRFs of the solid. Two versions of the method have been presented: analytical
consideration of the modal properties of the solid (as a Rayleigh beam), and numerical implementation from FE data.
881
In addition, these two approaches are compared with a one-dimensional Rayleigh beam model. It has been shown that the
analytical and numerical solid models predict virtually the same dynamic behaviour of the case under study. However, the
proposed solid model shows discrepancies with the Rayleigh beam model. These discrepancies are due to the effect of the
centrifugal forces associated with the deformed shape of the solid, which is only considered in the proposed solid model,
represented by the terms O2 gj and O2 C~ in Eqs. (64) and (56), respectively.
Finally, in order to describe the dynamic behaviour of a beam by means of the proposed solid model, the terms O2 gj or
O2 C~ have been eliminated. In this case, the calculations show that the proposed solid model behaves like the Rayleigh
beam model (see Figs. 910).
Acknowledgement
The authors gratefully acknowledge the support for this work provided by the Project TRA2010-15669 (Ministerio de
Ciencia e Innovacion).
Appendix
In this appendix, the formulation associated with the one-dimensional simply supported Rayleigh beam model,
including rotation and internal modal viscous damping is summarised. Further details can be found in Ref. [3].
In accordance with the frame of reference in Fig. 1, the jth mass normalised bending mode calculated in the neutral axis
of Rayleigh beam is
1
jpz
fj z p sin
,
(A.1)
mj
L
and the jth natural frequency is
s
kj
,
oj
mj
(A.2)
mj
EIjp4
rAL
2
2L3
(A.3)
rIjp2
2L
(A.4)
The transverse displacements are computed from the xed frame through the following modal approach:
8
9
(
)
1
1
< qjx t = X
X
xz,t
fj z
fj qj ,
j
yz,t
: qy t ;
j1
(A.5)
j1
where qj is the two-dimensional vector with modal coordinates associated with the jth orthogonal bending mode shape.
The equation of motion for the damped simply supported rotating Rayleigh beam in modal coordinates is
j
j
q 2Og j G cj Iq_ o2j I 2Ocj Gqj Q j ,
(A.6)
Q being the generalised external force, I is the identity matrix 2 2, O is the beam spin speed and oj is the natural
frequency of the jth bending mode of the simply supported non-rotating Rayleigh beam [1820]. The term cj is computed
from the internal modal damping ratio of the jth bending mode xj as
j
c j oj x j :
(A.7)
jp2
z2 jp2
(A.8)
where gj is, by denition, bounded between 0 and 1, and z is the slenderness of the beam. The matrix G is anti-symmetric
and couples both orthogonal bending modes. It can be expressed as
0 1
G
:
(A.9)
1 0
The eigenvalues of the Eq. (A.6) are
lj,1 cj ig j O
q
c2j o2j g 2j O2 2iOcj g j 1,
882
(A.10)
where the notation z represents the complex conjugate of z, lj,1 and lj,3 are the eigenvalues for the forward mode, and lj,2
and lj,4 are associated with the backward mode. The forward modes reveal a critical speed beyond which the real part of
their eigenvalues is positive. The modes become then unstable (a conclusion also made in Ref. [21]) beyond the critical
angular velocity given by
1
:
Ocr p
12g 1
(A.11)
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
aArtculo 2a
Centro de Investigacin en Tecnologa de Vehculos, Universidad Politcnica de Valencia, Camino de Vera s/n, 46022 Valencia, Spain
Dipartimento di Meccanica, Politecnico di Milano, Via La Masa 1, 20156 Milano, Italy
a r t i c l e
i n f o
Article history:
Received 8 March 2012
Received in revised form 5 July 2012
Accepted 9 July 2012
Available online 28 July 2012
Keywords:
Railway vehicles
Wheelset design
Wheelset axle fatigue
Dynamic loads
Traintrack interaction
a b s t r a c t
The fatigue design of railway axles requires that the stresses arising in the axle in real service are accurately quantied. This paper describes a method to compute the dynamic stresses arising in railway axles
as the effect of traintrack interaction, based on the numerical simulation of the dynamic interaction
between a exible wheelset and a exible track. The wheelset is modelled as a exible rotating body
using an Eulerian approach, whereas track is regarded as an innite periodic system with the rail modelled as a Timoshenko beam resting on discrete elastic supports, considering the inertia associated with
the sleepers.
The paper presents an application of the proposed procedure to the calculation of the dynamic stresses
caused in the axle by different types of geometric imperfection occurring on the wheel and rail surfaces,
considering the cases of a single harmonic rail corrugation, random rail roughness and a wheelat.
2012 Elsevier Ltd. All rights reserved.
1. Introduction
Railway axles durability is a key issue in designing and correctly
maintaining railway vehicles, to ensure that the highest safety
standards are met and, at the same time, to optimise life-cycle
costs from a system point of view, i.e. considering not only the
vehicle but also the interacting infrastructure. From the single
point of view of fatigue resistance, the axle design should tend towards increasing the size to reduce stresses, but this component
also represents a signicant contribution to the wheelset unsprung mass, which, on the contrary, shall be minimised to reduce
the generation of dynamic force at wheelrail contact and hence
damage in the track and in the wheels, especially for high-speed
trains.
At present, wheelset axles are designed for innite fatigue life,
however a small number of axle failures due to fatigue continues
to be reported, sometimes with catastrophic consequences. One
of the reasons which have been proposed to explain these unforeseen failures is that the loads assumed by the standard for the purpose of fatigue design verication do not fully reect the actual
service loads of the vehicle, which are largely depending on the
vehicle design parameters (e.g. unsprung masses, suspension stiffness and damping, etc.) and on the vehicle service prole [1,2]. The
precise knowledge of service loads is also pivotal to the denition
of appropriate intervals for the non-destructive inspection of rail Corresponding author.
E-mail address: stefano.bruni@mecc.polimi.it (S. Bruni).
0142-1123/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijfatigue.2012.07.006
19
20
Fig. 3. Detail of the track model. Left: model of some sleeper bays. Right: sleeper and rail pad model.
damping are accounted for by means of lumped spring and dashpot elements.
2.3. The model of wheelrail contact forces
The track model is coupled with the exible wheelset via the
wheelrail contact forces, which are expressed as function of the
relative wheelrail displacement and velocity in the contact patch.
In the numerical model presented here, the theory of Hertz is
adopted to dene the normal contact force component and FASTSIM [17] is used to dene the tangential contact forces as function
of the normal contact force and of the creepage components.
At each time step, the displacement and velocity of the wheel in
the contact point are obtained using Eq. (1) evaluated at the contact point position.
3. The exible wheelset model
In order to model the kinematics of the exible wheelset, two
congurations (undeformed and deformed) are dened. The undeformed conguration is associated with the spinning velocity of
the wheelset (at constant angular velocity X). The deformed conguration considers the exibility and small rigid solid displacements. The displacement eld relates the deformed conguration
with the undeformed conguration as it will be shown in Eq. (1).
The coordinates that are implemented in the wheelset model do
not follow the material points of the solid which is the commonest
procedure in Mechanics, nevertheless they are associated with spatial points (Eulerian approach). Let u an Eulerian vector coordinate
in a xed coordinate frame. Any property of the solid u(u, t) corresponds to the material point of the solid whose undeformed conguration is in the spatial point u at instant t. Following this
criterion, the displacement eld is dened by means of the following formula:
r u wu; t;
1
0 0 1
B
C
J @ 0 0 0 A;
1 0 0
1
1 0 0
B
C
E @ 0 0 0 A:
0 0 1
1
0 0 X
B
~ @ 0 0 0C
X
A XJ;
X 0 0
~X
~ X2 E:
and X
~ XJu Xu
~;
v v 1 v 2 v 3 T Xu
X @w
Dr Du Dw X @u
v i w_ v i
Dt Dt
Dt
@u
@ui
i
i
i
X @w
_ X u
~i
XJu w
:
@ui
i
!
!
X @wT
X @w
DrT Dr
_ X2
_ Tw
~i
~i
u
u
X2 uT Eu w
@ui
Dt Dt
@ui
i
i
!
T
X
@w
_ T Ju 2X2
~i
2Xw
Ju
u
@ui
i
!
X @w
_T
~i
2Xw
:
u
@ui
i
wu; t Uuqt;
1
Dr Dr
q
dt
2 Vol Dt Dt
Z
Z
1
1
1
q_ T
qUT Udtq_ X2
quT Eudt X2 qT
2
2
2
Vol
Vol
!
!
Z
Z
X @UT
X @U
~i
~i
dtq Xq_ T
q
qUT Judt
u
u
@ui
@ui
Vol
Vol
i
i
!
Z
X @UT
~i
u
Judt Xq_ T
X2 qT
q
@ui
Vol
i
!
Z
X @U
~i
dtq:
q UT
u
@ui
Vol
i
Ek
Vol
!
T Z
Z
X @U
D @Ek
~i
dtq_ X2
q
UT Udtq_ X
q
UT
u
Dt @ q_
@ui
Vol
Vol
i
!
Z
Z
X @UT
~i
Udtq_
qUT Eudt X
q
u
@ui
Vol
Vol
i
!
Z
X @U
~i
X
dtq_ X2
u
q UT
@ui
Vol
i
!
Z
X @UT
~i
Judt X2
q
u
@ui
Vol
i
!
Z
X @U
dtq X2
qUT
ui
@ui
Vol
i1;3
!
!
Z
X @U
X @UT
~i
~i
dtq X2
u
u
q
@ui
@ui
Vol
i
i
!
Z
XX
@2U
~i u
~j
dtq;
qUT
u
@ui @uj
Vol
j
i
!
!
T
Z
X @U
X @UT
@Ek
~i
~i
dtq X2
X2
q
u
u
@q
@ui
@ui
Vol
i
i
!
Z
X @UT
~i
u
q
Judt
@ui
Vol
i
!
Z
X @UT
_
~i
Udtq:
X
q
u
@ui
Vol
i
dt
X Z
elements
Vole
dt:
13
2X
!
X @U
~i
dt 2XUeT
u
FE
@ui
i
!
Z
X @Ne
T
~i
q
Ne
u
dtUeFE ;
@ui
Vole
i
qUT
e
Vol
14
qNe
e
Ve
Vol
!
X @Ne
~i
dt:
u
@ui
i
15
Ae
Z
Ce
Vol
10
11
qNe
e
Z
ce
qNe
Vol
!
T T
Z
X @U
D @Ek
@Ek
~i
dtq_ X2
q_ 2X
q
UT
u
@q
Dt @ q_
@ui
Vol
i
!
Z
XX
@2U
~i u
~j
dtq
qUT
u
@ui @uj
Vol
j
i
!
Z
X @U
X2
dtq X2
qUT
ui
@ui
Vol
i1;3
Z
qUT Eudt:
21
Vole
!
XX
@ 2 Ne
~i u
~j
u
dt;
@ui @uj
j
i
X
ui
i1;3
!
@Ne
dt;
@ui
qNe Eudt:
T
16
17
18
19
Vol
Uu Ne uUeFE ;
12
being Ne(u) the basis function matrix of the eth element, and UeFE
the mode shapes computed in the nodes of the eth element through
the FE model.
being Q the generalised forces, and D a diagonal matrix that contains the square of the undamped natural frequencies of the freeboundary solid.
3.1. Calculation of stresses in the wheelset
The simulation of traintrack interaction is nalised in this paper towards the calculation of dynamic stresses arising in the
wheelset axle. To perform the stress calculation, rst of all some
sections of interest are identied along the axle. These sections
22
Fig. 5. Studied section of wheelset and spatial point where the stresses have been
calculated.
er speed, the result obtained for the case of a rigid wheelset (i.e.
excluding from the analysis all modes except the rigid ones) is also
shown, to evaluate the effect of wheelset exibility. Finally, the rail
roughness prole is also reported in the gure using an appropriate
scaling and offset to obtain a proper visualisation.
Two harmonic contents are observed in the vertical contact
force, the largest one having the same wavelength as the rail corrugation and a second one with wavelength equal to the sleeper
bay. This second harmonic component is due to the periodic variation of the rail stiffness seen by the wheelset as the consequence
of the discrete rail support. The amplitude of the dynamic contact
force component having the same wavelength as the corrugation is
highly inuenced by the wheelset speed, being approximately
three times greater at 300 km/h than at 100 km/h. The contact
force is also highly affected by wheelset exibility: the comparison
of the results obtained for the rigid and exible wheelset at the
speed of 300 km/h shows that neglecting wheelset exibility leads
to over-estimating the peak-to-peak amplitude of the dynamic
force by 25% approximately. This is due to the fact that at the considered excitation frequency of 1390 Hz approximately (corresponding to the ratio of the wheelset speed over the corrugation
wavelength) the wheelset mass participating to the vertical motion
is lower than the whole mass of the wheelset due to exibility effects, a mechanism which is not captured by the rigid wheelset
model.
The corrugation peakpeak amplitude of 4.84 lm is relatively
small compared to the typical amplitude of rail corrugation, which
is in the range of tens or even hundreds of lm. Indeed, ISO 3095
limit is dened for a random corrugated prole including several
wavelengths which altogether would lead to a greater peakpeak
amplitude. Therefore, in the lower subgure of Fig. 6 the vertical
contact force obtained for the exible wheelset at 300 km/h is
compared for two corrugation amplitudes, one corresponding to
the ISO 3095 limit (same as in the upper subgure) and the other
10 times greater; this comparison also allows to assess the importance of non-linear effects in the case considered. For the larger
corrugation amplitude, the peak-to peak amplitude of the vertical
force is approximately 93 kN, and leads to a maximum wheel overloading/unloading which is approximately 78% of the static load.
For the case considered here, non-linear effects appear to play a
relatively marginal role, since the maximum dynamic variation of
the vertical load for the larger corrugation amplitude is with good
approximation 10 times greater than for the small corrugation
amplitude. However, the force uctuation component having the
same wavelength as the sleeper bay is in this case less visible (note
that the amplitude of excitation associated with this effect does
not change with the corrugation amplitude).
Fig. 7 shows the y-axis normal stress in the studied section of
the axle (cfr. Fig. 5), plotted as function of the wheelset rotation
for the same cases considered in Fig. 6. On the left, the effect of
vehicle speed is analysed by comparing the results obtained at
100 and 300 km/h for the same corrugation amplitude corresponding to the ISO 3095 limit: the waveform of the stress is with good
approximation sinusoidal with 2p periodicity, and the peakpeak
23
longitudinal stiffness
lateral stiffness
vertical stiffness
vertical damping
longitudinal damping
lateral damping
Sleeper bay
Sleeper number
Sleeper mass
Track bed stiffness
Track bed damping
Rail pad stiffness
Rail pad damping
Rail section
0.6 m
70
324 kg
200 MN/m
150 kN s/m
1 GN/m
50 kN s/m
UIC60
70
68
66
64
62
60
58
56
54
Rail roughness
52
0.2
0.4
0.6
0.8
1.2
Distance x [m]
Flexible wheelset V = 100 km/h
110
100
90
80
70
60
50
40
30
20
10
Rail roughness
0
0.2
0.4
0.6
0.8
1.2
Distance x [m]
ISO 3095
10 x ISO 3095
Fig. 6. Wheelrail contact force when the vehicle circulates on a corrugated track with corrugation wavelength 60 mm. Above: effect of speed and of wheelset exibility for
corrugation amplitude corresponding to the ISO 3095 limit. Below: effect of corrugation amplitude for speed 300 km/h and exible wheelset.
24
kQ
Q d;max
;
Qs
kr
rd;max
;
rs;max
50
40
40
30
30
20
20
10
10
0
10
10
20
20
30
30
40
40
50
V = 100 km/h
V = 300 km/h
21
with rd,max maximum dynamic stress in the material point considered and rs,max the corresponding maximum stress under the action
of rotating bending produced by the static loads acting on the
wheelset.
In Fig. 10 the dynamic factors kQ and kr are reported as function
of the vehicle speed for the different types of rail corrugation presented above, considering a corrugation amplitude 10 times greater than the ISO 3095 limit; in this case, also the results for a
corrugation wavelength exciting the second forward bending
mode of the wheelset are presented. For the contact force dynamic
factor kQ, in the entire speed range considered, the largest values
are obtained for the corrugation wavelength exciting the rst forward bending mode of the wheelset, with a maximum value close
to 1.8. A local maximum appears for most of the corrugation cases
considered for speeds in the 125135 km/h range, this is due to a
50
20
with Qd,max the maximum value of the contact force and Qs the static wheel load. For the stress in the axle the dynamic factor kr is dened as:
y [MPa]
y [MPa]
50
ISO 3095
10 x ISO 3095
Fig. 7. y-Axis normal stress in the studied section of the axle (cfr. Fig. 5) when the vehicle circulates on a corrugated track with corrugation wavelength 60 mm. Left: two
vehicle speeds, corrugation amplitude corresponding to the ISO 3095 limit. Right: two corrugation amplitudes (the results are undistinguishable).
25
110
100
90
80
70
60
50
40
30
20
10
Rail roughness
0
0.2
0.4
0.6
0.8
1.2
Distance x [m]
ISO 3095
10 x ISO 3095
110
100
90
80
70
60
50
40
30
20
10
Rail roughness
0
0.6
1.2
1.8
2.4
3.6
4.2
4.8
5.4
Distance x [m]
ISO 3095
10 x ISO 3095
Fig. 8. Wheelrail contact force when the vehicle circulates on a corrugated track with different corrugation wavelengths. Two corrugation amplitudes are considered,
corresponding to the ISO 3095 limit and 10 times the ISO 3095 limit. Above: the corrugation wavelength excites the pinnedpinned mode of the rail. Below: the corrugation
wavelength excites the rst forward bending mode of the wheelset.
and the effect of the resonance associated with the sleeper passing
is less evident than in the trend of the kQ dynamic factor.
These results suggest that the bending stresses in the axle are
highly affected by the wavelength of rail corrugation, and that
combinations of train speed and corrugation wavelength leading
to a resonance of a bending mode may be especially relevant in
view of axle resistance to fatigue. Particularly signicant in view
of real applications is the case of rail corrugation, a form of irregular wear of the rail often appearing in railway systems and characterised quasi-harmonic wear patterns developing on the rail head
in longitudinal direction [18]. Corrugation wavelength can range
from 50 mm or less up to more than 1 m in the case of heavy haul
26
50
60
40
30
40
20
y [MPa]
y [MPa]
20
10
0
10
20
20
30
40
40
50
ISO 3095
60
10 x ISO 3095
ISO 3095
10 x ISO 3095
Fig. 9. y-Axis normal stress in the studied section of the axle (cfr. Fig. 5) when the vehicle circulates on a corrugated track at 300 km/h speed. Left: the corrugation wavelength
excites the pinnedpinned mode of the rail. Right: the corrugation wavelength excites the rst forward bending mode of the wheelset.
1.8
No roughness
Corrugation excites pinned-pinned frequency
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
50
100
150
200
250
300
350
No roughness
Corrugation excites pinned-pinned frequency
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
0.9
pean Standards EN13103 and EN13104 [19,20] assume a 1.25 dynamic amplication factor on the loads generated at the primary
suspension. Since the calculation of the bending stresses is then
based on static equilibrium, also the bending stresses are magnied by 25% with respect to the static case, whereas the results in
Fig. 10 suggest that larger dynamic factors might apply when a
particular combination of corrugation wavelength and train speed
excites one bending mode. For instance, a sinusoidal rail corrugation having approximately 350 mm wavelength and 0.15 mm
depth (i.e. 10 times the ISO3095 limit for the considered wavelength) would produce a dynamic stress amplication close to
40%. It shall be pointed out however that the equivalent static
loads prescribed by EN13103/104 also include lateral wheelrail
contact forces due to curving, which are not included in the analysis presented here.
50
100
150
200
250
300
350
corrugation, and typical wear depth values are in the range of some
tenths of millimetre.
To quantify the relevance to fatigue of rail corrugation, we note
that the design methods prescribed for railway axles by the Euro-
27
excitation produced by random rail corrugation (amplitude corresponding to the ISO 3095 limit). The results obtained considering a
rigid and exible wheelset model are compared. Important dynamic uctuations of the contact force are evidenced, with the
maximum dynamic value corresponding to approximately 1.6
times the static load. The differences between the results obtained
for the rigid and exible wheelset are relatively small in this case,
and do not affect signicantly the maximum contact force value.
Fig. 12 presents the vertical contact force vs. travelled distance
for a vehicle affected by a wheelat. The calculation has been carried out for 50 and 300 km/h by means of the rigid and exible
wheelset models. A severe dynamic effect is observed, consisting
of the complete loss of contact between the wheel and the rail, followed by an impact leading to a maximum value of the contact
force which is between 3 and 4 times the static load and nally
by a transient vibration. The differences between the results for
the rigid and exible wheelset model are small in terms of duration
of the contact loss and of maximum overloading, but the transient
following the impact is affected quite remarkably by wheelset exibility, as demonstrated by the different frequency contents of the
contact force signal, see particularly the zoomed view on the right
side of the gure.
The y-axis stresses for the random rail corrugation and wheelat excitation cases are shown in Fig. 13. In the random corrugation case, the dynamic uctuations of the stress component are
relatively low and lead to a maximum peakpeak stress amplitude
around 91.2 MPa, corresponding to a dynamic amplication with
respect to the maximum stress produced by the static load kr of
1.23 approximately. When the wheelat defect is considered, the
results are highly affected by the vehicle speed, with large dynamic
effects taking place at low speed (with a dynamic factor kr of 1.34
approximately) on account of the repeated loss of contact of the
wheel to the rail, and with a reduction of the dynamic stresses at
higher speed.
Finally, in Fig. 14 the dynamic factors kQ and kr are reported for
the random corrugation and wheelat as function of the vehicle
speed. The contact force dynamic factor kQ is nearly monotonically
increasing up to a maximum value close to 1.8 for random corru-
gation excitation, whereas in case of wheelat excitation the dynamic factor is much higher, in the range of 2.84.2, with larger
values occurring at lower speeds: this is because at low speed
the duration of the contact loss caused by the wheelat is longer
and hence the following impact is larger. The stress dynamic factor
kr for the random corrugation excitation case is almost monotonically increasing with speed, with a maximum value close to 1.3.
For the wheelat excitation case, a completely different trend is
observed, the stress dynamic factor kr being initially decreasing
with the speed and then increasing above 200 km/h. The maximum value of the stress dynamic factor for this excitation case is
obtained at the lowest speed considered in the analysis and is
slightly below 1.5.
As in the case of excitation generated by a single-harmonic rail
corrugation, it is interesting to observe that in some cases the dynamic factor kr exceeds the 1.25 value assumed in EN13103/104.
For wheelat excitation, this happens at low speeds (below
75 km/h) which are typical e.g. of freight application whereas for
random rail roughness the 1.25 value is exceeded only at very high
speeds above 300 km/h, which are only relevant to very high speed
trains. It shall be stressed however that the results shown in Fig. 14
depend on the amplitude of the defects being considered and, in
case of more severe irregularities, larger dynamic stresses shall
be expected.
It is also interesting to point out that by using a static calculation to derive the bending stresses, Standards EN13103/104 inherently imply a proportionality between the contact forces and the
stresses, whereas the results in Figs. 10 and 14 show that the stress
dynamic factor kr is generally lower (sometimes much lower) than
the dynamic factor for the vertical contact force kQ. This circumstance is due to the inertia forces generated in the wheels and to
the different magnication of wheelset exible modes produced
when resonance conditions occur and suggests that a method
based on a static calculation could be not fully suited to estimate
service stresses in the axle. Dynamic traintrack interaction models such as the one proposed in this paper could be envisaged as a
means to derive a more realistic estimate of service stresses in the
axle, but this requires the study to be extended to consider the ef-
100
90
80
70
60
50
40
30
20
10
Rail roughness
0.2
0.4
0.6
0.8
1.2
1.4
1.6
1.8
Distance x [m]
Fig. 11. Wheelrail contact force when the vehicle circulates at 100 and 300 km/h speeds on a randomly corrugated track.
250
250
200
200
150
28
View A
100
Zoomed view A
150
100
50
50
0.5
1.5
0.1
0.2
Distance x [m]
0.3
0.4
0.5
Distance x [m]
180
180
160
160
140
140
120
Zoomed view A
View A
100
80
60
120
100
80
60
40
40
20
20
0.5
1.5
Distance x [m]
0.1
0.2
0.3
0.4
0.5
0.6
Distance x [m]
Fig. 12. Wheelrail contact force when the vehicle circulates at 50 (above) and 300 km/h (below) speeds on a perfectly even rail in presence of a 50 mm wheelat.
5. Conclusions
Dynamic effects may be extremely important in determining
the fatigue resistance of railway axles, and need to be properly
considered in the axle design process. This paper has presented a
method for the numerical estimation of the dynamic stresses in
railway axles, based on the simulation of the interaction between
a exible wheelset and a exible track. The method is presently
limited to the case of tangent track running of the wheelset, but
the extension to the curving condition is envisaged as a next step
of the research, to consider the additional axle loading due to
wheelrail contact forces in a curve.
Results of the numerical procedures were presented in the paper, considering the dynamic excitation caused by rail corrugation
and by a wheelat for different wheelset speed values. The results
clearly show that dynamic effects may lead to a signicant dynamic amplication of the stresses in the axle, which is in some
cases close to 70%. However, the actual relevance of dynamic effects affecting axle stresses is strongly depending upon the type
of excitation and the vehicle speed.
When single harmonic corrugation is considered, the dynamic
amplication of axle stresses is relatively low except in the case
when the corrugation wavelength excites the bending modes of
the wheelset. In the case of random multi-harmonic rail corrugation, the dynamic amplication factors are relatively low because
the excitation is spread over a wide range of frequencies, thus
reducing the importance of resonance effects. In all rail corrugation
cases considered, regardless the waveform of the excitation, the
29
Wheelflat
impact
50
40
Wheelflat
impact
50
50
40
40
30
30
30
20
20
20
y [MPa]
10
y [MPa]
y [MPa]
10
10
0
10
10
Zoomed view
20
20
20
30
30
30
40
50
Zoomed view
10
40
40
50
Fig. 13. y-Axis normal stress in the studied section of the axle (cfr. Fig. 5) when the vehicle circulates at 300 km/h speed on a randomly corrugated track (left) and on a
perfectly even rail in presence of a 50 mm wheelat at 50 km/h (centre) and 300 km/h (right).
4.5
Random roughness
Wheelflat
3.5
3
2.5
2
1.5
1
50
100
150
200
250
300
350
Acknowledgements
1.45
Random roughness
1.4
Wheelflat
1.35
1.3
1.25
1.2
References
1.15
1.1
1.05
1
50
100
150
200
250
300
350
[1] Smith RA. Railway fatigue failures: an overview of a long standing problem.
Materialwiss Werkstofftech 2005;36(11):697705.
[2] Grubisic V, Fischer G. Procedure for reliable durability validation of train axles.
Materialwiss Werkstofftech 2006;37(12):97382.
[3] Beretta S, Carboni M. Variable amplitude fatigue crack growth in a mild steel
for railway axles: experiments and predictive models. Eng Fract Mech
2011;78:84862.
[4] Luke M, Varfolomeev I, Lthepol K, Esderts A. Fatigue crack growth in railway
axles: assessment concept and validation tests. Eng Fract Mech
2011;78:71430.
[5] Grubisic V, Fischer G. Bemessung von Radsatzwellen Einussgrssen und
Vorgehen bei der Auslegung. Bericht FB-226, LBF, Darmstadt; 2005.
[6] Al S, Braghin F, Bruni S. Numerical and experimental evaluation of extreme
loads for improved wheelset design. Vehicle Syst Dyn 2008;46S:43144.
[7] Watson AS, Timmis K. A method of estimating railway axle stress spectra. Eng
Fract Mech 2011;78:83647.
30
[8] Bruni S, Corradi R, Mazzola L. Wheel rail contact forces as input for optimal and
robust axle design. In: 8th international conference on contact mechanics and
wear of rail/wheel systems (CM2009), Firenze, Italy; September 1518, 2009.
[9] Thompson DJ, Jones Thompson CJC. A review of the modelling of wheel/rail
noise generation. J Sound Vib 2000;231:51936.
[10] Nielsen JCO, Igeland A. Vertical dynamic interaction between train and
trackinuence of wheel and track imperfections. J Sound Vib
1995;187(5):82539.
[11] Diana G, Cheli F, Bruni S, Collina A. Experimental and numerical investigation
on subway short pitch corrugation. Vehicle Syst Dyn 1998;29S:23445.
[12] Nielsen JCO, Oscarsson J. Simulation of dynamic train_track interaction with
state-dependent track properties. J Sound Vib 2004;275:51532.
[13] Fayos J, Baeza L, Tarancn JE, Denia FD. An Eulerian coordinate-based method
for analysing the structural vibrations of a solid of revolution rotating about its
main axis. J Sound Vib 2007;306:61835.
[14] Baeza L, Fayos J, Roda A, Insa R. High frequency railway vehicletrack
dynamics through exible rotating wheelsets. Vehicle Syst Dyn
2008;46:64759.
aArtculo 3
* Corresponding author
E-mail corresponding author: baeza@mcm.upv.es
Abstract
Train-track interaction has been extensively studied in the lat 40 years at least, leading to modelling
approaches that can deal satisfactorily with many dynamic problems arising at the wheel/rail
interface. However, the available models are usually not considering specifically the running
dynamics of the vehicle in a curve, whereas a number of train-track interaction phenomena are
specific to curve negotiation.
The aim of this paper is to define a model for a flexible wheelset running on a flexible curved track.
The main novelty of this work is to combine a trajectory coordinate set with Eulerian modal
coordinates; the former permits to consider curved tracks, and the latter models the small relative
displacements between the trajectory frame and the solid. The wheelset model is coupled to a cyclic
track model having constant curvature by means of a wheel/rail contact model which accounts for
the actual geometry of the contacting profiles and for the non-linear relationship between creepages
and creep forces.
The proposed model can be used to analyse a variety of dynamic problems for railway vehicles,
including rail corrugation and wheel polygonalisation, squeal noise, numerical estimation of the
wheelset service loads. In this paper, simulation results are presented for some selected running
conditions to exemplify the application of the model to the study of realistic train-track interaction
cases and to point out the importance of curve negotiation effects specifically addressed in the
work.
Keywords
Flexible wheelset, flexible curved track, train-track interaction
1. INTRODUCTION
Train-track interaction consists of the coupled vibration of a railway vehicle or train set and of a
flexible track, with coupling of the two sub-systems being provided by wheel-rail contact forces and
excitation arising mainly from surface imperfections in the rails and wheels, such as rail roughness
and wheel out-of-roundness. In some cases, large level of vibration and large dynamic fluctuations
of the contact forces may take place, leading to unwanted phenomena such as high levels of noise
and vibration [1], damage of the rolling surfaces in the form of corrugation [2] or rolling contact
fatigue [3]. Furthermore, train-track interaction also leads to dynamic stresses in the track
components and in the wheelsets, which need to be carefully considered in order to avoid failures
due to metal fatigue.
For these reasons, a large effort has been spent over the last 40 years to define suitable models for
studying train-track interaction. Early models were mostly based on the representation of the
vehicle as a system formed by rigid bodies, possibly simplified to considering only the wheelset as
a rigid mass resting on a Hertzian stiffness [4]. More recently, the need to widen the frequency
range of analysis led to the incorporation of wheelset flexibility in the models, leading to more
realistic representation of wheel-rail interaction effects at higher frequencies. For the study of rail
corrugation and wheel polygonalisation, generally a modal synthesis is introduced to reduce the size
of the problem [5 7], whereas in case the frequency range of interest up to 1kHz and above (such
as for the study of rolling noise) a Finite Element model of the wheel or wheelset is used without
condensation [8, 9]. Only very recently, a further model refinement was introduced to consider the
inertial effects due to wheelset rotation [10, 11].
It should be noted that many phenomena related with train-track interaction, particularly squeal
noise, short pitch rail corrugation and the largest stresses generated in the wheels and axle, are
mostly occurring when the rail vehicle negotiates a curved track, calling for a proper consideration
of the effects related with wheelset curving in train-track interaction models. When a rail vehicle
runs through a curve, two mutually influencing phenomena take place at wheel rail contact: on one
hand, contact parameters such as the contact point position, the normal and creep forces, creepages
are slowly evolving in response to curve negotiation. On the other hand, the same quantities are
subject to faster changes due to high-frequency interaction of the flexible bodies in contact. On
account of the non-linearity of the problem, the two effects cannot be superimposed and a more
comprehensive approach needs to be deployed which, to the Authors knowledge, has not been
presented yet.
The aim of this paper is therefore to propose a mathematical model for the dynamics of a flexible
wheelset running through a curve and interacting with a flexible railway track, considering in full
the non-linearities introduced by wheel-rail contact. In order to keep within reasonable limits the
computational complexity of the problem, a trajectory coordinates set is introduced that permits to
describe the large motion of the wheelset along the curved track, and the small relative movements
of the wheelset with respect to the trajectory frame are described by means of Eulerian modal
coordinates introduced with respect to a set of modal shapes obtained from a finite element model
of the wheelset. The first 100 modes of vibration of the flexible wheelset are considered, covering a
frequency range up to 2.7 kHz approximately. The wheelset model is coupled to a cyclic track
model having constant curvature by means of a wheel/rail contact model which accounts for the
actual geometry of the contacting profiles and for the non-linear relationship between creepages and
creep forces.
Results for the proposed modelling approach are presented for a selected vehicle type and curving
condition, and for different excitation sources including short wavelength geometric imperfections
in the rail profiles and singularities such as wheel flats. Results are also presented for the wheelset
running in tangent track, to point out the importance of curve negotiation effects specifically
addressed in this work.
The paper is organised as follows: in Section 2 the equations of motion for the flexible wheelset in a
curve are introduced. In Section 3 the model of a cyclic curved flexible track is presented. In
Section 4 the model of wheel-rail contact forces is introduced and the final equations of the traintrack interaction problem are derived. In Section 5 simulation results are presented for some
selected running conditions and finally in Section 6 conclusive remark are provided.
2. THE FLEXIBLE WHEELSET MODEL IN CURVED TRACK
The study of vehicle-track interaction is carried out in this paper considering one single wheelset,
given that in the frequency range of interest the dynamics of the sprung masses (bogie frame and
car body) are effectively isolated from the motion of the un-sprung masses (wheelsets and axle
boxes) on account of the mechanical filter introduced by the suspensions. The effect of low
frequency curving dynamics of the complete vehicle needs however to be included in the model, in
order to obtain suitable mean values for the creepages and contact forces, which affect the coupled
wheelset-track vibration also at higher frequency. This is accomplished by prescribing the forces
applied by the primary suspension to the wheelset in the vertical plane and the yaw displacement of
the bogie at the primary suspension, as explained in Section 2.2.
2.1. Equations of motion of the flexible wheelset
In order to model the flexible wheelset travelling on curved track, two frames of reference are
considered (see Fig. 1). The first one is an inertial frame X0Y0Z0 which is fixed in an arbitrary
point. The second is a trajectory coordinate frame XTYTZT that follows the motion of the wheelset.
The system XTYTZT is centred in the undeformed configuration of the wheelset, being the XT-axis
parallel to the forward speed, the YT-axis parallel to wheelset axis and the ZT-axis is vertically up.
A vector referred to the fixed and trajectory frame is denoted by a0 and a , respectively.
Figure 1: Frames of reference and position vectors. In dashed trace is shown the undeformed configuration of
the wheelset. In solid colours is sketched a generic position of the flexible wheelset.
The coordinates that are implemented in the wheelset model do not follow the material points of the
solid which is the commonest procedure in Mechanics, nonetheless they are associated with spatial
points (Eulerian approach). The position vector r0 of a material particle which is in the spatial
position u at instant t for the undeformed configuration, can be defined by means of the following
formula:
r0 = p 0 + T (u + w (u, t )) ,
(1)
where p0 is the position vector of the track frame; w corresponds to the displacement vector due to
the elastic deformation and small rigid body displacement of the solid; T is the rotation matrix that
relations the trajectory frame of the track with the fixed frame.
Considering that the coordinate frame is chosen so that the wheelset spin rotation is in the
~
second axle YT, the angular velocity tensor is defined as follows:
0 0
0 0 0
0 0
~
J ; and J
0 0 1
0 0 0 .
1 0 0
(2)
~,
Ju u
(3)
ww
~ (u + w ) + w
+ J u + u~i
= p
w
ui
i
(4)
~ TT T
the
being p the speed of the track frame centre (expressed in the trajectory frame) and
angular velocity matrix of the track frame. The two first velocity terms are associated to
translational and rotational movement of the track frame, respectively; w represents the velocity of
the spatial point due to the flexibility; the term J u is the velocity due to the rigid body spinning;
and the last term is the convective velocity associated with the Eulerian coordinate system.
In order to obtain the expression of the kinetic energy for the wheelset, the square of the particle
velocity is obtained, which reads:
Dr T Dr
Dt Dt
ww
~ u w 2 p T J u 2 p T w
2 p T
p T p 2 p T
u~i
w ui
i
~T
~ u w 2 u T w T
~T w
~T Ju
2 uT w T
uT w T
2
u~
w wT ~
u w w
Tw
2 w
T Ju 2 w
T
w ui
2 u T J T J u 2 2
ww
J u 2
u~i
w ui
w w
u~i
w ui
T
ww .
u~ w u
i
(5)
ww
u~ w u
i
Due to the geometry of revolution of the wheelset, the displacement vector w can be calculated
through superposition of mode shapes in the non-rotating trajectory frame XTYTZT:
w u, t = u qt ,
(6)
where (u) is the mode shape functions matrix of the free-boundary wheelset and q (t ) is the
Eulerian-modal coordinate vector. The small rigid body displacements of the solid are considered in
this approach through the rigid body modes of the wheelset. It must be pointed out that the mode
shape functions do not depend on time since the rotation of the solid does not change the mode
shapes functions in spatial coordinates, because of the axial symmetry of the wheelset. Once the
formula Eq. (6) is applied in Eq. (5), the kinetic energy results in the following expression:
1
Dr T Dr
U dv
2 vol Dt Dt
1 T
~ U u dv p T
~ U dv q p T J U u dv
p p mw p T
vol
vol
vol
2
1
w
~T
~ u dv
dv q U u T
p T U dv q p T U u~i
vol
vol
vol
u
2
w
i
i
EK =
~T
~ dv q U u T
~ T dv q
~T
~ dv q 1 q T U T
U uT
vol
vol
2 vol
~ T dv q U u T
~ T J u dv q T U T
~ T J u dv
qT U T
vol
vol
vol
w ~
w T ~
u dv q T U u~i
dv q
q T U u~i
vol
vol
w ui
w ui
i
i
T
1
w
dv q
q T q q T U T J u dv q T U T u~i
vol
vol
2
w ui
i
1 2
w T
J u dv
U u T E u dv 2 q T U u~i
vol
vol
2
w ui
i
1 2 T
w T ~ w
ui
dv q
q U u~i
vol
2
w ui i
w ui
i
(7)
Once the kinetic energy is known, the two terms of Lagranges equation are computed as follows:
w EK
wq
~ T p U u~ w dv p U T
~T
~ u dv
= U T dv
i w u
vol
vol
vol
i
~T
~ dv q U T
~ T dv q U T
~ T J u dv
U T
vol
vol
w
U u~i
vol
w ui
i
vol
w T ~
u dv U u~i
dv q
vol
w ui
T
~ T u~ w dv q U u~ w dv q
U T
i
i
vol i w ui
vol
w ui
i
w T
w T ~ w
J u dv 2 U u~i
ui
dv q
2 U u~i
vol
vol
w ui
w ui i
w ui
i
i
(8)
D w EK
Dt w q
~ u dv U T
~ dv q U T
~ dv q
U T
= U T dv p
vol
vol
vol
vol
w
w T
U T u~i
dv q U u~i
dv p
q
vol
vol
w ui
w ui
i
i
w T ~
~ J u dv
u dv U T
U u~i
vol
vol
u
w
i
i
w T ~
~ u~ w dv q
dv q U T
U u~i
i w u
vol
vol
w ui
i
i
i
w T
J u dv 2 U T E u dv
2 U u~i
vol
vol
w ui
i
w
w T ~ w
dv q
dv q 2 U T ui
ui
2 U u~i
vol
vol
w ui
w ui i
i
i 1,3 w ui
w2
2 U T u~i u~ j
i j
vol
w ui w u j
being E
dv q
(9)
1 0 0
0 0 0 .
0 0 1
J J
The equation of motion of the flexible and rotating wheelset is derived by means of Lagranges
~ is anti-symmetric, the
equation. Considering Eqs. (8) and (9) and taking into account that matrix
following equation is obtained:
2
q
vol
2
U T
2
~
U T
vol
vol
~ dv q
w 2
dv 2
i
j
u~ u~ w u w u
i
vol
w
u~ w u dv 2 U
U T
w
dv U
u~
w u
vol
vol
U T
w
u w u dv
i
1, 3
~ dv U T
~T
~ dv q .
vol
(10)
~ J u dv U T
~ u dv
2 U T E u dv 2 U T
vol
vol
vol
~T
~ u dv U T dv TT p
0 Q
U
vol
vol
The modal properties are computed from a finite element (FE) model, therefore it is adequate to use
the FE methodology for computing the equation of motion numerically. The mode shape functions
are obtained into the e-th element of the FE mesh as follows:
u N e u eFE ,
(11)
where N e u is the basis (or shape) function matrix of the e-th element, and eFE the mode shapes
computed in the nodes of the e-th element through the FE model.
This approach allows obtaining the matrices of the equation of motion by means of the matrices of
the elements. These matrices have to be assembled in global matrices by following the standard FE
assembling technique. The first matrix in Eq. (10) is obtained by means of the approach in Eq. (11)
as follows:
~
V
vol
ne
w
u~ w u dv
U T
T
FE
ne
U N
vol e
e 1
eT
vol e
e 1
wN
u~i
w ui
w Ne e
FE dv
w ui
u~
U eFE N e
T
dv FE
(12)
being defined V e as the corresponding matrix of the e-th element, that is:
Ve
vol
w Ne
~
dv , V
w ui
U N e u~i
TFE
ne
V
e
FE
(13)
e 1
where ne is the number of elements in the FE mesh, vol is the volume domain associated with the
undeformed solid, and vol e is the volume of the e-th element. Following the same procedure as in
Eq. (12), the remaining matrices of the equation of motion are obtained:
Pe
Ae
vol
T
U Ne
vol
1, 3
vol
vol e
He
w N e
~
dv , C TFE
w ui
w Ne
~
dv , S
u~i
w ui
ne
vol
vol
ne
vol
(15)
FE
(16)
(17)
e 1
ne
S
ne
FE
e 1
FE
(18)
e 1
ne
ne
T
U N e E u dv , ~c TFE
FE
(19)
e 1
c ,
(20)
e 1
T ~
~
U Ne
J u dv , U TFE
ne
U ,
FE
e 1
B
A
R
(14)
C
TFE
T ~
~
N e dv , R
U Ne
TFE
FE
e 1
T ~T ~
~
U Ne
N e dv , B TFE
ce
Ue
w 2 N e
~
u~i u~ j
dv , A TFE
w ui w u j
T ~
U Ne
Re
Be
T
U Ne
vol e
ne
P
Ce
Se
vol e
T ~
~
U Ne
N e dv , P TFE
T ~
~
u dv , H
U Ne
TFE
(21)
e 1
ne
H ,
e 1
(22)
Ne
Ge
vol
T ~T ~
~
U Ne
u dv , N TFE
ne
N ,
vol e
ne
(23)
e 1
G T
U N e dv , G TFE
0 ,
p
(24)
e 1
resulting the following equation of motion for the flexible wheelset running along a curved track:
~ ~
~ ~ ~ ~
~
~
~ ~ ~ ~
+ 2 V 2 P q + 2 A - C + 2 S + R - B + D q = 2 ~c - 2 U - H + N - G + Q c Q s .
(25)
q
~ ~ ~ ~ ~ ~
~
The matrices V , P , A , C , S , R and B account for inertial effects associated with the deformed
configuration originated by Coriolis, centrifugal and tangential acceleration of the wheelset, as
produced by the track frame motion and by the rotation of the wheelset around its axis. The vectors
~
~ ~ ~
~c , U
, H , N and G account for inertial effects not depending on wheelset deformation, which are
also originated by Coriolis, centrifugal and tangential acceleration experienced by the wheelset. The
~
diagonal matrix D is the modal stiffness matrix that contains the square of the undamped natural
frequencies of the free-boundary wheelset. Finally, Q c and Q s are the vectors of the generalized
forces acting on the flexible wheelset resulting respectively from wheel-rail contact forces, see
Section 4, and from the forces applied by the primary suspension, see Section 2.2.
All the above described boundary conditions are applied on the flexible wheelset model Eq. (25) by
appropriately setting the terms in vector Q s . These consist of the generalised forces associated with
the modal coordinates q of the concentrated forces applied at the axle-box seats, defined as
explained above in this section.
3. THE TRACK MODEL
The track model has been adapted from the one presented in reference [13] where cyclic boundary
conditions were adopted. With respect to this previous work, here different sleeper bay distances
have been considered in order to take into account the dynamics of a constant radius curved track.
The cyclic track approach that is used in the present paper models a circumferential constant radius
track negotiated by a set of identical vehicles, uniformly distributed in such a way that each vehicle
is set at a constant distance L apart from the adjacent ones and travel at the same velocity V, see
Figure 2. The constant distance L is set large enough to avoid the dynamic interaction between the
vehicles and cyclic boundary conditions are introduced at the ends of the model. Hence, due to the
periodicity of the structure and of the loading conditions, the study of the track is reduced to a
single section having finite length L.
The approach adopts a substructuring technique where rails and sleepers are treated separately. The
rails are modelled as Timoshenko beams, including bending deformations in vertical/lateral
directions, as well as torsional deformations. Rail vibration is introduced in terms of modal
superposition for the unconstrained rail with cyclic boundary conditions, hence resulting into a set
of de-coupled 1-d.o.f. equations.
The discrete rail supports are introduced in the form of lumped parameter systems. The rail pads are
modelled as lumped visco-elastic elements generating the interaction forces between the rails and
the sleepers, represented as lumped masses. Ballast dynamics is neglected here, being not relevant
for the dynamic behaviour of the wheelset, but the equivalent ballast stiffness and damping are
accounted for by means of lumped spring and dashpot elements.
The lateral and vertical displacements of the rail axis are [13]:
w y x, t
x qry t ,
(26)
x qrz t ,
(27)
w z x, t
and the torsion and the rotations of the rails cross-section are:
\ x x, t
<
<
<
x qrx t ,
(28)
x qrz t ,
(29)
x qry t ,
(30)
\ y x, t
\ z x, t
where Wr y x , Wrz x , < r x x , < r y x and < r z x are the r-th modal functions of the Timoshenko
periodic beam, and qrx t , qry t and qrz t are the modal coordinates associated with torsional,
lateral and vertical rail vibrations, respectively.
The resulting equations of motion for the cyclic track model in modal coordinates take the form:
qr 2 [ r Zr q r Zr2 qr
fr ,
(31)
being Zr the r-th undamped frequency and [ r is the modal damping. The modal forces f r are
computed from the wheel rail contact forces acting on the track.
In this way, the displacements of the rail in the present contact point can be evaluated from the
displacements and rotations of the rail axis as follows:
xr, j
E j w yj
w zj \ xj \ jy \ zj ,
(32)
where x r , j is the vector of contact point displacements in j-th rail, and the matrix E j relates the
displacements in rail axis and contact points.
4. THE MODEL OF THE WHEEL-RAIL CONTACT FORCES
Equations (25) and (31) are coupled by the wheel-rail contact forces, which can be defined as
function of the wheelset modal coordinates q and their time derivatives q and of the track
displacements x r together with their time derivatives x r . The calculation of the contact forces is
performed within the time step integration of the equations of motion for the wheelset and the track.
First the motion (position and speed) of the contact points on wheel and rail surfaces is determined,
then the normal and tangential wheel-rail contact forces are computed as a function of the relative
wheel-rail motion at the contact point, finally the generalised forces on the vehicle and track
coordinates are defined based on the principle of virtual work.
4.1 Contact kinematics
Using the modal superposition principle, the vectors rw, j of the wheel displacements at the contact
point (with j = 1,2 representing the left and right wheel) are computed as:
rw, j
)
u w, j qt rw(irr
, j t ,
(33)
)
with u w, j the position of the contact point on the wheel and rw(irr
, j a vector accounting for the effect
of wheel out-of-roundness, which is defined as a periodical function of time. In the same way, the
material velocity vectors rw, Mj of the wheel at the contact point, i.e. the velocity of the material
point on each wheel instantaneously in contact with the rail are defined as:
rw M, j
u w, j q t
w
u~ wu qt .
i
(34)
The rail displacements at the contact points on the left and right rails rr , j are obtained according to
the following expression:
rr , j
x r , j t rr(,irrj ) t ,
(35)
where rr(,irrj ) is a vector accounting for the effect of geometric imperfections in the track due to
irregularity and rail roughness, whereas the material velocity of the rails at the contact points is
computed as:
rr,Mj
x (rM, j )
E j w yj
w zj \ xj \ jy \ zj
V Ej
w
w jy
wx
w zj \ xj \ jy \ zj ,
(36)
(37)
where n i is the vector defining the normal direction for the i-th contact, G i , j is the undeformed
distance for the i-th contact of the j-th wheel-rail couple, and superscripts (t) and (t-'t) denote
quantities evaluated at the present and previous time step respectively.
The normal force N i,tj is computed as function of the elastic inter-penetration according to Hertzs
formulae using the profile curvatures retrieved from the contact table:
N i,t j 0
t
t
N i , j C i , j pi , j
3
2
if
pi,t j d 0
if
pi,t j ! 0
(38)
The creep forces are then computed as function of the creepages, according to the heuristic
formulae by Shen, Hedrick & Elkins [16]. The longitudinal and transversal creepages, H Li , j and H Ti , j
respectively, are computed as follows:
HL
HT
i, j
'R
i, j
R0
V
s
(1) j ,
R0
cosJ j ,
(39)
(40)
with l i and t i the vectors defining the longitudinal and transversal direction for the i-th contact, V
the speed of the wheelset, 'Ri , j the rolling radius variation for the contact point under
consideration, V j the angle of attack of the wheel, s half-distance between the wheel nominal
running circles and J j the contact angle of the wheel at the actual contact point.
Finally, the normal and creep forces obtained at each i-th wheel-rail contact point are projected
along the trajectory frame XTYTZT and summed over all active contacts occurring in the same
wheel-rail couple, and the components of the resulting contact forces along the modal coordinates
q are derived by standard application of the principle of virtual work, providing vector Q s in Eq.
(25).
5. RESULTS
In this section, results of wheelset-track interaction simulations are presented considering different
sources of excitation: a single harmonic rail corrugation, random rail roughness and a wheelflat.
Results are also presented for a rigid wheelset model, i.e. including in the modal synthesis only the
rigid modes of vibration of the unconstrained wheelset. In this way, the effect of wheelset flexibility
is pointed out. Furthermore, for the wheelflat excitation case, results are also presented for the
wheelset running at the same speed in tangent track, to assess the effect of curving on highfrequency wheelset-track interaction, which is the main innovation introduced in this paper.
The case study considered here refers to the trailed car of a concentrated power train for high-speed
passenger service. The vehicle is equipped with a solid axle wheelset with monobloc, light design
wheels. The track considered features UIC60 rails and track parameters are based on the
EUROBALT project [17], considering a stiff track. Table 1 summarises the input data used to setup the simulation model.
1375 kg
120 kN
7.5 MN/m
7.1 MN/m
200 MN/m
0.81 MN/m
150 kNs/m
100 kNs/m
1 GN/m
100 kNs/m
50 kNs/m
30 kNs/m
Rail section
UIC60
All simulation cases presented below consider the wheelset running at 150 km/h through a curve
with radius 1000 m with cant deficiency 150 mm approximately. For this running condition, Table
2 compares the steady-state values of the contact force components for the rigid body model of the
complete vehicle and for the model of the single flexible wheelset: a very good agreement of the
two series of data is observed, leading to the conclusion that, at least for the considered running
condition, the procedure introduce in Section 2.2 is able to correctly set-up the boundary conditions
for the flexible wheelset model. Note that in Table 2 and below in this section vertical, lateral
and longitudinal force respectively mean the component of the contact force along the ZT, YT and
XT axis of the track following reference.
Outer wheel
Inner wheel
Rigid
multi-body
model
Flexible
wheelset
model
Rigid
multi-body
model
Flexible
wheelset
model
Vertical force
69.57 kN
70.01 kN
49.82 kN
49.71 kN
Lateral force
5,68 kN;
5,91 kN
-3,75 kN;
-3,99 kN
Longitudinal force
14.39 kN;
15.16 kN
-14.39 kN;
-15.16 kN
Table 2: Steady-state wheel-rail contact forces for the rigid multi-body model of the entire vehicle and for the
single flexible wheelset model.
Two harmonic contents are observed in the vertical contact force, the largest one having the same
wavelength as the rail corrugation and a second one with wavelength equal to the sleeper bay. This
second harmonic component is due to the periodic variation of the rail stiffness seen by the wheelset
as the consequence of the discrete rail support. The contact force is also highly affected by wheelset
flexibility: considering a rigid wheelset leads to an over-estimation of the peak-to-peak dynamic
force by approximately 15% on the inner wheel and 40% on the outer wheel. This is due to the fact
that the mass participating in the high frequency vibration of the flexible wheelset is lower than the
whole mass of the wheelset, mainly due to bending deformations occurring in the axle, a
mechanism which is not captured by the rigid wheelset model.
75
70
65
60
55
50
45
40
0
10
15
20
25
Time t [ms]
30
35
40
45
Figure 3. Vertical wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a corrugated curved track
(1000 m curve radius) with corrugation wavelength 60 mm. Corrugation amplitude corresponding to the ISO 3095 limit
Figure 4 shows the lateral component of the contact force on the two wheels, for the same running
condition considered above. The steady state values of the contact forces (see also Table 2) reflect
the curving condition of the wheelset. Here, the leading wheelset in the front bogie of the vehicle is
considered, for which a small negative angle of attack takes place on account of curve negotiation,
causing a steady component of the transversal creep force which points towards the outside of the
curve. This is balanced by the lateral component of the flange force on the outer wheel, leading to
the typical condition of the two lateral forces pointing in opposite directions in a way that tends to
widen the track gauge. Like for the vertical contact force component, the dynamic component of the
lateral contact force shows two harmonic contents, one corresponding to the sleeper-passing
frequency, the other corresponding to the wavelength of the harmonic corrugation introduced as the
source of excitation. Also in this case, simplifying the problem to the case of a rigid wheelset leads
to an over-estimation of the dynamic contact force harmonics synchronous with the corrugation.
The longitudinal contact force component caused by a single harmonic rail corrugation is not shown
as the level of dynamic excitation for this case is very low.
-2
-4
Inner rail roughness
-6
0
10
15
20
25
Time t [ms]
30
35
40
45
Figure 4. Lateral wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a corrugated curved track
(1000 m curve radius) with corrugation wavelength 60 mm. Corrugation amplitude corresponding to the ISO 3095 limit
120
100
80
70
60
50
40
Inner rail roughness
20
0
0
10
Time t [ms]
15
20
25
Figure 5. Vertical wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a randomly corrugated curved
track (1000 m curve radius). Amplitudes corresponding to the ISO 3095 limit.
6
4
2
0
-2
-4
-6
10
Time t [ms]
15
20
25
Figure 6. Lateral wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a randomly corrugated curved
track (1000 m curve radius). Amplitudes corresponding to the ISO 3095 limit.
25
20
15
10
5
0
-5
-10
-15
-20
Inner rail roughness
-25
Outer rail roughness
10
Time t [ms]
15
20
25
Figure 7. Longitudinal wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a randomly corrugated
curved track (1000 m curve radius). Amplitudes corresponding to the ISO 3095 limit.
250
200
200
250
View A
150
100
100
50
50
Zoomed view A
150
10
20
30
Time t [ms]
40
50
60
Time t [ms]
10
12
Figure 8. Vertical wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a perfectly even curved track
(1000 m curve radius) in presence of a 50 mm wheelflat.
35
35
30
25
30
View A
20
15
10
5
0
Zoomed view A
20
15
10
5
0
-5
-5
-10
25
10
20
30
Time t [ms]
40
50
60
-10
Time t [ms]
10
12
Figure 9. Lateral wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a perfectly even curved track
(1000 m curve radius) in presence of a 50 mm wheelflat.
80
60
40
60
View A
20
Zoomed view A
20
-20
-20
-40
40
10
20
30
Time t [ms]
40
50
60
-40
Time t [ms]
10
12
Figure 10. Longitudinal wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a perfectly even curved
track (1000 m curve radius) in presence of a 50 mm wheelflat.
In Figure 11 results are presented for the same wheelset excitation considering the vehicle running
at the same speed (150 km/h) in tangent track. Because dynamic effects in lateral and longitudinal
direction are in this case relatively modest, only the vertical force component is shown.
Furthermore, because of the symmetry in the running condition considered, the forces on the two
wheels are the same and therefore results are shown for one single wheel. Apart from the difference
in the steady state value of the contact forces, the waveform of the contact force time history in
Figure 11 looks similar to the result obtained for curve negotiation in Figure 8. However, the peak
force is approximately 180 kN, compared to 250 kN for the outer wheel in the curving condition,
showing that wheelflat excitation in a curve may lead to considerable overloading of the wheelset
and rails and can therefore be expected to cause accelerated damage and degradation of the
contacting surfaces as well as increased noise and vibration.
180
160
160
140
180
View A
120
100
80
60
140
Zoomed view A
120
100
80
60
40
40
20
20
0
0
10
20
30
Time t [ms]
40
50
60
Flexible wheelset
Time t [ms]
10
12
Rigid wheelset
Figure 11. Vertical wheel-rail contact forces when the vehicle circulates at 150 km/h speed on a perfectly even tangent track
in presence of a 50 mm wheelflat.
6. CONCLUSIONS
This paper has presented a model for a flexible wheelset running on a flexible curved track. By
introducing a trajectory coordinates set describing the large motion of the wheelset along the curved
track and assuming small relative movements of the wheelset with respect to the trajectory frame,
the terms appearing in the wheelset equations of motion can be efficiently computed, keeping the
time required to carry the numerical simulation within acceptable limits. At the same time, the
model accurately describes the vibration of the flexible wheelset in the frequency range up to 2.7
kHz. The flexible wheelset is coupled to a periodic curved track model by a non-linear description
of wheel-rail contact forces.
Appropriate boundary conditions are prescribed at the wheelset axle-boxes in terms of vertical and
lateral forces and longitudinal displacements above the primary suspension, to correctly reproduce
the steady state contact forces and creepages as obtained from a simulation of the low frequency
running dynamics of the complete vehicle along the curve considered. The results reported in Table
2 show that this method succeeds in establishing the appropriate steady state contact condition on
both wheels, for the exemplary curving condition considered in this paper.
Results are presented for three types of excitation: single harmonic rail corrugation, randomly
corrugated track and excitation arising from a wheel flat when the wheelset runs over a perfectly
even track. One exemplary curving condition was considered, which consists of the wheelset
running along a 1000 m radius curve at 150 km/h and 150 mm cant deficiency. To point out the
implications of modelling wheelset flexibility, results are also presented for the case of a rigid
wheelset.
In all three excitation cases considered, the rigid wheelset model leads to an over-estimation of the
maximum contact forces, compared to the flexible wheelset model. For the vertical contact force
component, this effect is particularly significant with excitation coming from a corrugated rail
(either with single harmonic or random corrugation) whereas differences between the two models
are less important in the case of wheel flat excitation. However, in case of wheel flat excitation a
very large difference of the results obtained for the rigid and flexible wheelset model is observed for
the lateral component of the contact force.
Finally, for the case of wheel flat excitation results obtained for the wheelset running in curve
where compared to the case of the wheelset running at the same speed in tangent track: this
comparison shows that the peak load in the vertical force component is approximately 39% larger
when the wheelset runs in curve compared to the tangent track running case, leading to the
conclusion that wheelflat excitation in a curve may lead to considerable overloading of the wheelset
and rails and can therefore be expected to cause accelerated damage and degradation of the
contacting surfaces as well as increased noise and vibration.
It is worth noticing that the effects of wheel flat excitation in a curve have not been extensively
studied in the past, so this topic is suggested for further investigation in the future, looking e.g. at
the effect of parameters such as curve radius, cant deficiency, wheel profiles, bogie and suspension
design. Further use of the model presented here is also envisaged to investigate problems such as
rail corrugation, squeal noise, dynamic stresses in wheels, axles and rails.
ACKNOWLEDGEMENTS
The authors acknowledge the financial contribution of the Spanish Ministry of Economy and
Competitiveness through the project TRA2010-15669.
REFERENCES
[1]
[2]
[3]
A. Ekberg, Fatigue of railway wheels, In: The Wheel/Rail Interface Handbook Woodhead
Publishing in Mechanical Engineering, Pages: 211-244, 2009.
[4]
S.L. Grassie, R.W. Gregory, D. Harrison, K. L. Johnson, Dynamic response of railway track
to high frequency vertical excitation, Journal of Mechanical Engineering Science 24 (1982)
77-90
[5]
[6]
[7]
K. Popp and I. Kaiser, Interaction of elastic wheelsets and elastic rails: modelling and
simulation, Vehicle System Dynamics (2006) 44 (Suppl.), 932-939.
[8]
[9]
[10] L. Baeza, J. Fayos, A. Roda and R. Insa, High frequency railway vehicle-track dynamics
through flexible rotating wheelsets, Vehicle System Dynamics (2008) 46 (7) 647 662.
[11] P. T. Torstensson, J. C. O. Nielsen and L. Baeza, Dynamic train-track interaction at high
vehicle speeds-Modelling of wheelset dynamics and wheel rotation, Journal of Sound and
Vibration 330 (2011), 5309-5321.
[12] S. Bruni, A. Collina, G. Diana, P. Vanolo, Lateral dynamics of a railway vehicle in tangent
track and curve: tests and simulation, Vehicle System Dynamics (2000) 33 (Suppl.) 464-477.
[13] L. Baeza, H. Ouyang, A railway track dynamics model based on modal substructuring and a
cyclic boundary condition, Journal of Sound and Vibration 330 (2011) 7586.
[14] J. Piotrowski, H. Chollet, Wheel-rail contact models for vehicle system dynamics including
multi-point contact, Vehicle System Dynamics, 43 (6), 455-483. 2005.
[15] F. Braghin, S. Bruni, G. Diana, Experimental and numerical investigation on the derailment
of a railway wheelset with solid axle, Vehicle System Dynamics (2006) 44 (4), 305 325.
[16] Z. Y. Shen, J. K. Hedrick and J. A.Elkins, A comparison of alternative creep force models for
rail vehicle dynamic analysis, Vehicle System Dynamics, (1983) 12 (Suppl.) 79 82.
[17] L. Mazzola, Y. Bezin, S. Bruni, Vehicle Track interaction: MB simulation for track loading
limits and damage identification ECCOMAS Thematic Conference Multibody Dynamics
2011, July 4-7 2011, Bruxelles, Belgium.