Anda di halaman 1dari 11

Advances in Engineering Software 36 (2005) 827837 www.elsevier.

com/locate/advengsoft

Wheel-rail contact elements incorporating irregularities


C.J. Bowe*, T.P. Mullarkey
Department of Civil Engineering, National University of Ireland, Galway, Ireland Received 25 November 2003; accepted 17 March 2005 Available online 22 August 2005

Abstract The aim of this study is to simulate the dynamic vertical response of a vehicle traversing rigid rails and a railway bridge. This is achieved by using the authors wheel-rail contact element (WRC) to model the dynamic interaction that exists between a sprung wheel, using a Hertzian spring, and the rail. The objective in creating these elements was to model the rail and wheel irregularities, which was not a feature of the contact elements within the ANSYS nite element program. In this paper the numerical results generated using the authors WRC element are identical to the results generated using the commercial contact element of ANSYS for a smooth rail condition. In the case of irregular rails, the numerical results generated using the authors WRC elements compare very favourably with the results from the literature. q 2005 Civil-Comp Ltd and Elsevier Ltd. All rights reserved.
Keywords: Wheel-rail contact element; Hertzian spring; Rigid rails; Irregularities

1. Introduction In recent years, research on the dynamic response of trains and railway bridges has become an important topic in civil engineering. Engineers and researchers have a responsibility for ensuring the safe passage of trains traversing rails and railway bridges by carrying out extensive research on existing structures. Much of the dynamic response that the bridge and vehicle experience can be attributed to the contact that exists between the wheel and the rail. Therefore, by modelling the different wheel-rail conditions, one can better understand the dynamic response of vehicles and railway bridges. Many researchers such as Cheng et al. [1], Yang and Wu [2] and Yau et al. [3] have developed their own vehiclebridge elements, whereby the vehicle is the assembly of an unsprung mass for the wheel and sprung mass for the vehicle body. In these systems, the wheel is assumed to be in direct contact with the rail at all times; hence, the wheel and rail have the same deection and wheel-rail separation is not possible on a rigid rail. In doing so, these systems must also
* Corresponding author. Tel.: C353 91 524411x3086; fax: C353 91 750507. E-mail address: cathal.bowe@nuigalway.ie (C.J. Bowe).

include the additional convective accelerations due to the wheel as it travels along the beam. In the authors model, the wheel is represented by a sprung mass, thus a Hertzian spring exists between the wheel and the rail and is simulated by means of the WRC elements. In this case, the wheel and beam no longer have the same deection; hence the additional convective accelerations due to the wheel are not required. The authors technique involves modelling each wheel as a Hertzian spring perpendicular to the surface of the rail. Each WRC element consists of three stiffness matrices to simulate the action of the wheel on the exible rail, whereas only one stiffness matrix is required to represent the action of the wheel on the rigid rail. The appropriate non-zero values, inputted into the stiffness matrices, relate to the position of a wheel on a particular element using its shape functions. The WRC elements use the extension in the Hertzian spring at each time-step to calculate the contact force that exists between the wheel and the rail. Wheel-rail separation occurs when the extension in the spring becomes positive, thus all stiffness matrices related to that particular wheel are made equal to zero. Introducing irregularities into the model tends to affect the behaviour of the train as it travels along the rails. Authors such as Au et al. [4], Chu et al. [5] and Wiriyachai

0965-9978/$ - see front matter q 2005 Civil-Comp Ltd and Elsevier Ltd. All rights reserved. doi:10.1016/j.advengsoft.2005.03.026

828

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

Wheel Centre

Wheel Centre

Rail

Sprung mass
Fig. 1. Wheel modelled as a sprung mass.

et al. [6] have discussed the effects of irregularities on the rails. Certain irregularities can lead to wheel-rail separation and as the wheel regains contact with the rail, it can cause a sharp impact load to the structure. In the authors model the irregularities are created using the summation of a series of sine curves and are represented indirectly using forces and moments applied to the exible rail and wheel as well as modifying the equation for extension. Again the forces and moments used are related to the position of a wheel on a particular element using its shape functions. When the wheel travels on the rigid rail, only the force applied to the wheel is taken into account.

rotation about z-axis are denoted as U, V and q, respectively. The subscripts B and L denote beam and spring, respectively. The interaction between the spring and beam element over which it is travelling has to be in accordance with Newtons third law. The beam acts with a vertical force on the spring; the spring acts with an equal and opposite force on the beam. Point forces and moments on the nodes represent this force on the beam. From the free-body in Fig. 2, the nodal forces in the wheel-rail contact element can be expressed as follows 9 8 9 8 > ULx1 > > FLx1 > > > > > > > > " #> > > > > kL11 kL12 < ULy1 = < FLy1 = Z (1) kL21 kL22 > ULx2 > > FLx2 > > > > > > > > > > > > > ; : ; : ULy2 FLy2 where kLij (i, jZ1, 2) denotes a 2!2 stiffness matrix, ULxi, ULyi (i, jZ1, 2) are the nodal displacement vector and FLxi, FLyi (i, jZ1, 2) are the nodal force vector of the Hertzian spring (also referred to as a link). kL11 is a symmetrical matrix, kL22 is equal to kL11, kL12 is equal to kL21, the latter being equal to minus one times kL11. " # " # cos2 q sin q cos q 0 0 kL11 Z kH q Z 908 Z kH 2 0 1 sin q cos q sin q The vertical displacement at any point along the beam can be calculated using Eq. (2), where N1 and N2 are the transverse displacement shape functions and G1 and G2 are the rotational shape functions. VB x Z VB1 N1 xCqB1 G1 xCVB2 N2 xCqB2 G2 x where N1 x Z 1C2x=l3 K3x=l2 G1 x Z xf1K2x=lCx=l2 gN2 x Z 3x=l2 K2x=l3 G2 x Z xfx=l2 Kx=lg The length of the beam element is denoted by l and the distance travelled along the element is given by x, travelling from left to right. Because local node 1 of the spring is located at position x on the beam, we can state the following:
ULy2
2 2 2

2. Wheel-rail contact elements 2.1. Smooth rails The technique involves modelling a wheel as a Hertzian spring with one node at the centre of the wheel and the other node on the rail as illustrated in Fig. 1. The spring is always perpendicular to the surface. In Fig. 2, the reader can see a free-body diagram for the WRC element and one beam element representing part of the rail. In ANSYS, the user can create a stiffness matrix between any pair of nodes; hence, the authors are obliged to use this facility to input the three additional stiffness matrices, thereby simulating the vertical effects of the wheel. To simulate the spring, three additional stiffness matrices are created between local node 1 and local node 2 of the beam, between local node 1 of the beam and local node 2 of the spring and nally between local node 2 of the beam and local node 2 of the spring. Fig. 2 indicates that the coordinate system adopted in this study has x positive along the beam element, y positive upward and z positive outwards. The origin of the coordinate system is at local node 1 of the beam. For twodimensional problems, the deection in the x, y plane and

(2)

(3)

ULx2
3rd stiffness matrix

B1
UB1
0 1 1

2nd stiffness matrix

Hertzian Spring

1 2

UB2

1st stiffness matrix

VB1

VB2

B2

Fig. 2. Free-body diagram of the wheel-rails contact element.

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

829

ULy1 ZVB x

(4)

three stiffness matrices used by the WRC elements:


38 9 0 0 0 0 0 0 >UB1 > > > > 7> 6 > > 6 0 N1 N1 N1 G1 0 N1 N2 N1 G2 7> VB1 > > > > > 7> 6 > > > > 7> 6 > > >q > 6 0 G N G G 0 G N G G 7< = 6 1 1 1 1 1 2 1 27 B1 7 6 ZkH 6 FLy1 7> >0> > > > 60 0 0 0 0 0 7>UB2 > > > > > > > > 7> 6 > > > > > > > 7> 6 > > > > N2 > > 6 0 N2 N1 N2 G1 0 N2 N2 N2 G2 7> VB2 > > > > > > > > 5> 4 > > > > > > ; > ; : > : G2 qB2 0 G2 N1 G2 G1 0 G2 N2 G2 G2 9 9 38 38 2 2 0 0 0 0 0 0 > UB1 > 0 0 0 0 0 0 > UB2 > > > > > > > 7> 7> 6 6 > > > > 6 0 0 0 0 N1 0 7> V > 6 0 0 0 0 N2 07> V > > > > > > B1 > > B2 > 7> 7> 6 6 > > > > > > > > 7> 7> 6 6 > > > > 6 0 0 0 0 G 0 7> 6 0 0 0 0 G 07> = = <q > <q > 6 6 1 7 2 7 B1 B2 7 7 6 6 CKkH 6 CKkH 6 7> 7> > > > > 6 0 0 0 0 0 0 7>ULx2 > 6 0 0 0 0 0 07> ULx2 > > > > > 7> 7> 6 6 > > > > > > 7> 7> 6 6 > > > > 6 0 0 0 0 0 0 7>U > 6 0 0 0 0 0 07> U > > Ly2 > > Ly2 > > > 5> 5> 4 4 > > > > > > > > ; ; : : K K 0000 0 0 0000 0 0 8 9 >0> > > > > > > > > > N1 > > > > > > > > > > > >G > < 1= 2 (9)

The substitution of Eq. (4) into Eq. (1) gives the following: 9 8 9 8 > ULx2 > > FLx1 > > > > > > > > " #> > > > > kL11 kL12 < VB x = < FLy1 = Z (5) kL21 kL22 > ULx2 > > FLx2 > > > > > > > > > > > > > ; : ; : FLy2 ULy2 Since it is assumed that the Hertzian spring element remains perpendicular to the surface at all times, the horizontal displacement on both nodes are equal. The force imparted to the beam by the spring is equal to KFLy1 located at position x, which in turn is equal to minus one times the left hand side of Eq. (6). The right-hand side of Eq. (6) come from Eq. (5). This generalised force consists of vertical forces and moments applied to the nodes. 8 9 8 9 > 0 > > 0 > > > > > > > > > > > > > >N > >N > > 1> > 1> > > > > > > > > > > > > ( ) > > > > < G = < G = VB x 1 1 Z k KkH FLy1 (6) > 0 > > 0 > H ULy2 > > > > > > > > > > > > > > > > >N > >N > > 2> > 2> > > > > > > > > > > > > : ; : ; G2 G2 In Eq. (6), VB(x) is replaced by the right-hand side of Eq. (2), resulting in the following equation: 8 9 8 9 > 0 > > 0 > > > > > > > > > > > > > > > > > > N1 > > N1 > > > > > > > > > > > > > > > > > > > > > >G > >G > < 1 = < 1 = Z kH KkH FLy1 > 0 > > 0 > > > > > > > > > > > > > (7) > > > > > > > > > > > > > N2 > > N2 > > > > > > > > > > > > > > > > > : ; : ; G2 G2 ( ) VB1 N1 xCqB1 G1 xCVB2 N2 xCqB2 G2 x ! ULy2 Eq. (7) can also be rewritten as: 8 9 8 9 9 8 >0> >0> > > > > > > > > > > > > > VB1 > > > > > > > > > N1 > > N1 > > > > > > > > > > > > > > > > > >q > > > > > > > B1 > > > > > > " #> > >G > >G > > = < 1 = < 1 = N1 G1 N2 G2 0 < VB2 Z kH KkH FLy1 > >0> >0> > > > > > > 0 0 0 0 1 > > > > > > > > > > >q > > > > > > > B2 > > > > > > > > > > > > > > > N2 > > N2 > > > > > > > > ; > > > > : > > > > > > > > ULy2 : ; : ; G2 G2 (8) The right-hand side of Eq. (8) is developed further, giving rise to the following equation, which contain the

There are three matrix terms on the right-hand side of Eq. (9). The rst term represents the stiffness matrix along the beam element, the second term represents the stiffness matrix between local node 1 of the beam and local node 2 of the spring element and the third term represents the stiffness matrix between local node 2 of the beam and local node 2 of the spring element. The force imparted to the wheel axle by the spring is KFLy2, where FLy2 is derived from Eqs. (5) and (2) resulting in the following:
8 9 >0> > > > > > > > > >0> > > > > > > > > > > > > <0= 8 > > > > > < 8 9 > VB1 > > > > > > > >q > < B1 = 8 9 >0> > > > > 9> > > > > > >> 0 > > > >> > >> > >> > >> > =< 0 =

FLy2

CkH ULy2 Z KkH N1 G1 N2 G2 >V > > > > >> > > B2 > >0> > >> 0 > > > > > > >> > > > > > > >> > > > > > > >> > : > > : ; ;> > > > > > >1> >1> qB2 > > > > > > > > > > > > > > > > : ; : ; K K 2 2 38 3 9 0 0 0 0 0 0 > UB1 > 0 0 0 000 > > > 6 6 7> 7 > > > 6 0 0 0 0 0 0 7> V > 60 0 0 0 0 07 > B1 > > 6 6 7> 7 > > > > 6 6 7> 7 > > 6 0 0 0 0 0 0 7> 60 0 0 0 0 07 <q > = 6 6 7 7 B1 7 7 CKkH 6 ZKkH 6 6 6 7> 7 > 6 0 0 0 0 0 0 7> ULx2 > 60 0 0 0 0 07 > > > 6 6 7> 7 > > > 6 6 7> 7 > > 6 0 N1 G1 0 0 0 7> U > 6 0 N 2 G2 0 0 0 7 > Ly2 > > 4 4 5> 5 > > > > : ; K 0 0 0 000 0 0 0 000 2 38 9 9 8 0 0 0 0 0 0 > UB1 > > UB2 > > > > > > > > 6 7> > > > > > > 6 0 0 0 0 0 0 7> V > > > > > > VB2 > > B1 > 6 7> > > > > > > > > > > 6 7> > > > > > > > > 6 0 0 0 0 0 0 7< <q = 6 7 qB1 = B2 6 7 CkH 6 10 ! 7 > > > > > > 6 0 0 0 0 0 0 7> ULx2 > > > ULx2 > > > > > 6 7> > > > > > > > 6 7> > > >U > > 6 0 0 0 0 1 0 7> U > > > Ly2 > > > > > 4 5> Ly2 > > > > > > > > > ; ; : : K K 000000

830

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

The right-hand side of Eq. (10) contains three matrix terms, and the third term by itself represents the effects of the rigid rails. For programming purposes Eqs. (9) and (10) are combined, resulting in symmetrical stiffness matrices for the Hertzian spring, updated at each timestep. The wheel-rail contact force in the spring is evaluated by multiplying the extension by the spring stiffness. The Hertzian extension is given by Eq. (11), where q is equal to 908 for the spring and where Eq. (4) has been incorporated. 9 8 > ULx1 > > > > > > > > > < ULy1 = extension Z K q K q cos q sin q cos sin > ULx2 > > > > > > > > > ; : ULy2 9 8 > ULx1 > > > > > > > V x > = < B > Z 0 K 0 1 1 11 > ULx2 > > > > > > > > > ; : ULy2 In the model, the extension in the spring is calculated at each timestep to determine if the wheel is in contact with the rail. A negative extension (compression) indicates that contact exists between the wheel and the rail, while a positive extension (tension) means that there is no contact; thus all stiffness matrices related to that particular wheel are set equal to zero when the extension is positive. 2.2. Irregular rails As a wheel traverses a rail with irregularities, the vertical position of the wheel changes in accordance with the prole of the irregularity, thus inuencing the behaviour of the train. Certain irregularities can lead to wheel-rail separation and as the wheel regains contact with the rail, it can cause a sharp impact load to the structure. In this model the irregularity function, f, is the summation of several sine curves as given in Eq. (12) fZ
N X jZ1

ULy2

y
x
B1
UB1
1 0 1

2 2

ULx2

f (x )

Hertzian Spring
1

UB2
2

x
VB2

VB1

B2

Fig. 3. Free-body diagram of the WRC element with irregularities.

Modifying Eq. (6) to include the effects of irregularities on the rails yields the following equation: 8 9 >0> > > > > > > > > > N1 > > > > > > > > > > > >G > < =
1

FLy1

Z k KkH >0> >0> H ULy2 > > > > > > > > > > > > > > > > > > > > > > > > > N2 > > N2 > > > > > > > > > > > > > > > > > : ; : ; G2 G2 8 9 8 9 >0> >0> > > > > > > > > > > > > > > > > > N1 > > N1 > > > > > > > > > > > > > > > > > > > ( ) > > ( ) >G > >G > < 1 = < 1 = x VB x f  k KkH k KkH Z C > > H >0> H ULy2 >0> > > 0 > > > > > > > > > > > > > > > > > > > > > N2 > > N2 > > > > > > > > > > > > > > > > > : ; : ; G2 G2 (14)

8 9 >0> > > > > > > > > > N1 > > > > > > > > > > > >G > < =
1

( ) x VB xCf 

aj sin

 2px gj

where the irregularity function f  has been isolated on the x right-hand side. Once again, the force imparted to the beam by the spring is equal to minus one times the left hand side of Eq. (14). Modifying Eq. (10) to include irregularities yields the following: 9 8 > VB1 > > > > > > > >q > < B1 = >V > > B2 > > > > > > > ; : qB2 15

(12) FLy2 ZK H N1 k

where aj and gj are the amplitude and wavelength, respectively, of the jth wave and N is the number of sine curves. The horizontal distance from the left support of the  beam to the spring is denoted as x. A free-body diagram showing irregularities on the rail can be seen in Fig. 3. This gure shows the Hertzian spring following the prole of the irregularities on the rail. Both the  x and the irregularity function f are shown in this gure. An inspection of Fig. 3 tells us that Eq. (4) must be modied as follows:  ULy1 Z VB x C f x (13)

G1

N2

G2

 C kH ULy2 C kH f x

The Hertzian extension in the spring is now modied by the irregularity function f and takes the form of Eq. (16). This equation explains why the contact element of ANSYS is unsuitable for modelling irregularities. Its equation for extension cannot include function f. Eqs. (14)(16) are

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

831

of 80.77 m and a total mass of approximately 275 tonnes. All members in the bridge are modelled using three-dimensional elastic beam elements. Sleeper and ballast effects are ignored in the model. In our model the rail on the bridge is exible, but the rail to the left and right of the bridge is rigid.
Fig. 4. Three-dimensional model of the Boyne Viaduct railway bridge.

3.2. Train model updated at each timestep. extension Z 0 K 1 0 1 9 8 ULx1 > > > > > > > < VB x C f x >  = > > > > : ULx2 ULy2 > > > > ; The three-dimensional vehicle model considered in this paper comprises a six-axle locomotive and a single fouraxle railway carriage as illustrated in Fig. 5. Each vehicle consists of a vehicle body supported by a pair of bogies, with each bogie supported by axles and nally a pair of wheels supports each axle. The bogies are connected to the axles through primary suspensions and to the vehicle body through secondary suspensions, with each suspension consisting of a spring and dashpot. Fig. 6 presents a three-dimensional locomotive model. Lumped masses mw are used to model the wheel masses, while elastic beam elements are used to model the bogie and vehicle body components. The primary spring stiffness and damping is given by k1 and c1, while k2 and c2 denote the

(16)

3. Bridge and train model 3.1. Bridge model As a case study, one only considers the centre span of the Boyne Viaduct railway bridge located in Drogheda, Ireland. The three-dimensional model of the bridge comprises a simply supported truss as shown in Fig. 4. The bridge has a clear span

Fig. 5. Six-axle locomotive and four-axle railway carriage.

y z x

y x z

vehicle body

k2 k1 mw kH l2 c1 mw

c2

secondary suspension bogie primary suspension

k2

c2 k1
c1

mw
Hertian spring

mw

mw

mw kH

wheelset

l2 l3

l2

l2

l4

Fig. 6. Three-dimensional locomotive model.

832

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837


VERTICAL DISPLACEMENT (m)
0.0005

VERTICAL DISPLACEMENT (m)

secondary spring stiffness and damping. The Hertzian spring stiffness, also known as the wheel-rail contact stiffness, is given by kH. Horizontal links are required between the vehicle body and the bogie and also between the bogie and axles to prevent the structure from becoming a mechanism. These horizontal links do not interact with the suspension springs. The distance between the axles of a single bogie is l2, between centres of two bogies is l3 and between wheels of a single axle is l4. The three-dimensional railway carriage model is similar to the locomotive model expect that it has two axles instead of three per bogie as described in Esveld [7].

(a)
0.0000 0 0.0005 WRC ELEMENT 0.0010 0.0015 0.0020 0.0025 ANSYS CONTAC48 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

TIME (sec)
0.0005 0.0000 0 0.0005 0.0010 0.0015 0.0020 0.0025 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

(b)

ANALYTICAL (5 modes) WRC ELEMENT

4. Results 4.1. Model validation for a single mass In order to validate the system, the authors compare their use of the WRC element with their use of the commercial node-to-surface contact element in ANSYS (CONTAC48). They also compare their results with some simple analytical solutions developed by Biggs [8]. The bridge properties adopted in Sections 4.1.14.1.3 are similar to those of Yang and Wu [2], such that the bridge has a length lZ25 m, Youngs modulus of elasticity EZ2.87!106 kN/m2, moment of inertia IZ2.9 m4, mass per unit length mZ 2.303 t/m and a Poissons ratio nZ0.2. The gravitational and damping effects of the bridge are ignored. The wheel traverses the bridge at a constant speed cZ27.78 m/s. The Hertzian spring stiffness kHZ1595 kN/m. Each example uses the Newmark time integration method [9] with 200 equal time steps to solve the transient analysis. In Sections 4.1.14.1.2 time t is arranged in such a manner that the vehicle is at the left hand support at tZ0 s. 4.1.1. Wheel as a moving sprung load As shown in Fig. 7, a simply supported beam is subjected to a moving sprung load P travelling at speed c from left to right. In the authors model, the wheel is given a zero mass and the point force PZ56.4 kN is attached to the centre of the wheel i.e. local node 2 of the spring. The deection and

TIME (sec)

Fig. 8. Vertical displacement at mid-point of beam due to a moving sprung load. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing the WRC element with analytical solution [8].

the vertical acceleration at mid-point of the beam are plotted as a function of time in Figs. 8 and 9, respectively,. In each case, gure (a) compares the WRC element with ANSYS CONTAC48 element, while gure (b) compares the solution using the WRC elements with Biggss [8] analytical solution, where the rst 5 modes of vibration are used to determine the deection and acceleration of the beam.
VERTICAL ACCELERATION (m/s2)
0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

(a)

WRC ELEMENT ANSYS CONTAC48

TIME (sec)
0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

VERTICAL ACCELERATION (m/s2)

(b)

ANALYTICAL (5 modes) WRC ELEMENT

TIME (sec)

Fig. 7. Simply supported beam subjected to a moving sprung load. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing the WRC element with analytical solution [8].

Fig. 9. Vertical acceleration at mid-point of beam due to a moving sprung load. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) Comparing the WRC element with analytical solution [8].

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837


VERTICAL ACCELERATION (m/s2)

833

0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

(a)

WRC ELEMENT ANSYS CONTAC48

TIME (sec)
0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 .

VERTICAL ACCELERATION (m/s2)

Fig. 10. Simply supported beam subjected to a moving sprung mass. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing the WRC element with analytical solution [8].

(b)

ANALYTICAL (2 modes) WRC ELEMENT

One should notice from the plots a striking similarity between the solutions using the WRC element and the ANSYS CONTAC48 element; equally the numerical and analytical solutions are strikingly similar. 4.1.2. Wheel as a moving sprung mass Fig. 10 presents a simply supported beam subjected to a moving sprung mass MwZ5.75 t traversing the beam at a constant speed. In this example, the analytical solution [8] consists of the rst two modes of vibration. Again, one compares the deection and vertical acceleration at midpoint of the beam as a function of time in Figs. 11 and 12, respectively, while the vertical displacement and acceleration of the sprung mass as a function of time are shown in Figs. 13 and 14, respectively. In Fig. 15, a plot of the contact force between the wheel and the rail is presented. The contact force due to the motion of the wheel varies somewhat from the weight of the wheel.
VERTICAL DISPLACEMENT (m)
0.0005

TIME (sec)

Fig. 12. Vertical acceleration at mid-point of beam due to a sprung mass. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing WRC element with analytical solution [8].

From inspection, one can see that the results from the WRC element and ANSYS CONTAC48 element are very similar again, whereas there are some slight deviations between the analytical solution [8] and the solution using the WRC element. Two reasons for the inaccuracy in the analytical solution are as follows: (1) too few modes are used to describe the deection of the beam; and (2) the analytical solution [8] models the wheel as an unsprung
VERTICAL DISPLACEMENT (m)

(a)
0.0000 0 0.0005 WRC ELEMENT 0.0010 0.0015 0.0020 0.0025 ANSYS CONTAC48 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0.0005

(a)
0.0000 0 0.0005 WRC ELEMENT 0.0010 0.0015 0.0020 0.0025 0.0030 ANSYS CONTAC48 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 . 0.9 .

TIME (sec) VERTICAL DISPLACEMENT (m)


0.0005

TIME (sec) VERTICAL DISPLACEMENT (m)


0.0005 0.0000 0 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030 ANALYTICAL (2 modes) WRC ELEMENT 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

(b)
0.0000 0 0.0005 ANALYTICAL (2 modes) 0.0010 0.0015 0.0020 0.0025 WRC ELEMENT 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

(b)

TIME (sec)

TIME (sec)

Fig. 11. Vertical displacement at mid-point of beam due to a sprung mass. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing the WRC element with analytical solution [8].

Fig. 13. Vertical displacement of the wheel due to a sprung mass. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) comparing WRC element with analytical solution [8].

834
VERTICAL ACCELERATION (m/s2)
0.20 0.15 0.10 0.05 0.00 0.05 0.10 0.15 0.20 0 0.1 0.2 0.3

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

(a)

0.4

0.5

0.6

0.7

0.8

0.9

WRC ELEMENT ANSYS CONTAC48

TIME (sec)
0.20 0.15 0.10 0.05

VERTICAL ACCELERATION (m/s2)

(b)

Fig. 16. Rigid rail and simply supported beam subjected to a travelling bouncing wheel.
0.003 0.002 0.001 0 0.0 0.001 0.002 0.003 0.004 0.005 0.006 Whee on rigid rail Wheel on rigid rail WRC ELEMENTS ANSYS CONTAC48

0.00 0 0.05 0.10 0.15 0.20 ANALYTICAL (2 modes) WRC ELEMENT 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

VERTICAL DISPLACEMENT (m)

0.5

1.0

1.5

2.0

2.5

TIME (sec)

Fig. 14. Vertical acceleration of the wheel due to a sprung mass. (a) Comparing the WRC element with ANSYS CONTAC48 element; (b) Comparing WRC element with analytical solution [8].

TIME (sec)

mass, thus requiring the additional convective accelerations, which are ignored in the analytical solution. 4.1.3. A travelling bouncing wheel In the following example one examines the effects of wheel-rail separation, whereby the wheel is modelled as a travelling sprung mass with an initial positive extension resulting in a zero contact force. As mentioned earlier, when the extension in the Hertzian spring becomes positive, no contact exists between the wheel and the rail, the contact force is zero, and all stiffness matrices related to that particular wheel are put equal to zero. In this model, rigid rails are located on either side of a simply supported beam as shown in Fig. 16. The initial horizontal position of the wheel at time tZ0 s is located at a distance lrZ25 m from the left support of the beam, while the initial vertical extension is 0.01 m. The wheel traverses the simply supported beam between the times tZ0.91.8 s and at all other times the wheel travels on the rigid rail.
50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Fig. 17. Vertical displacement at mid-point of the beam due to the travelling bouncing wheel.

The vertical displacement of the beam (at mid-point) and the wheel as a function of time is plotted in Figs. 17 and 18, respectively, while the contact force between the wheel and rail is shown in Fig. 19. Again, excellent agreement can be seen between the solution using the WRC element and ANSYS CONTAC48 element results. One can also see from Fig. 19 that as the wheel regains contact with the rail it generates a large impact load, much greater than the weight of the wheel; thus the maximum deection of the beam shown in Fig. 17 is much greater than the corresponding value in Fig. 11. 4.2. Model validation for a pair of masses separated by a spring for a wide range of speeds The WRC element models the wheel of a vehicle as sprung mass involving a Hertzian spring. In this section,
VERTICAL DISPLACEMENT (m)
0.02 0.00 0.0 0.02 0.04 0.06 0.08 0.10 0.12 W1.0RC ELEMENTS ANSYS CONTAC48

CONTACT FORCE (kN)

0.5

1.5

2.0

2.5

52 54 56 58 Weight of the wheel 60 WRC ELEMENT ANSYS CONTAC48

TIME (sec)

TIME (sec)

Fig. 15. Wheel-rail contact force due to a sprung mass.

Fig. 18. Vertical displacement of the travelling bouncing wheel.

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837


0 0.0 20 40 60 80 100 120 140 160 WRC ELEMENTS ANSYS CONTAC48 Weight of the wheel 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1.0 0.0 OLSSON [10] UNSPRUNG MASS SOFT HERTZIAN SPRING HARD HERTZIAN SPRING 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

835

0.5

1.0

1.5

2.0

2.5

TIME (sec)

DYNAMIC / STATIC DEFLECTION

CONTACT FORCE (kN)

1.0

Fig. 19. The contact force between the rail and the travelling bouncing wheel.

VEHICLE SPEED / CRITICAL SPEED

Fig. 21. Simply supported beam subjected to a moving vehicle using different Hertzian spring stiffness on a smooth rail.

the authors compare their WRC element with the unsprung mass system of Olsson [10]; however, the authors use both hard and soft Hertzian springs for the WRC elements because Hertzian springs are absent from Olssons [10] model as shown in Fig. 20. The bridge properties of Section 4.1 are used in this example. The dimensionless parameters adopted are similar to those of Olsson [10], such that the vehicle to bridge mass ratio is 0.5; the unsprung wheel mass to sprung vehicle mass ratio is 0.25; the bridge to vehicle frequency ratio is 3; the vehicle damping ratio is 0.125. Other dimensionless parameters used are a speed ratio (that is the vehicle speed divided by the critical speed) and a dynamic amplication factor (ratio between the maximum dynamic deection and static deection of the mid-point of the bridge). The critical speed is the speed of the vehicle such that the vehicle travels a distance of twice the length of the bridge in a time equal to the natural period of the bridge. In Fig. 21, the authors compare their WRC element with the unsprung mass system of Olsson [10] under smooth rail conditions, whereby the hard to soft Hertzian spring stiffness has a ratio of 500. One can see from the results that the solution for the hard Hertzian spring stiffness is almost identical with the solution of Olsson [10] at all speeds, whereas the soft Hertzian spring stiffness compares better with Olsson [10] at lower speeds.

One now examines the dynamic effects of rail irregularities on the simply supported beam subjected to the same moving vehicle. Olsson [10] uses an unsprung mass system while the authors use their WRC element. Again, hard and soft Hertzian spring are used for comparison purposes. In the model, the ratio of the bridge length to the irregularity wavelength is equal to 10, while the ratio of the irregularity amplitude to the static deection is 0.05. In Fig. 22, one can see that the results for the hard and soft Hertzian springs vary somewhat from the results of Olsson [10]. 4.3. Boyne Viaduct subjected to a moving locomotive and carriage on smooth and irregular rails The authors will now use the WRC element to investigate the dynamic response of the Boyne Viaduct Railway Bridge subjected to a moving vehicle travelling at a constant speed. The bridge and vehicle properties used are dened in Table 1. This is a three-dimensional problem, unlike the examples of Sections 4.1 and 4.2, which are twodimensional. In this example, the authors use the Newmark time integration method [9] to solve the transient analysis with 500 equal time steps. The WRC element is used to model rail irregularities, which are not a feature of the contact element of ANSYS.
1.8

DYNAMIC / STATIC DEFLEVTION

1.7 1.6 1.5 1.4 1.3 1.2 1.1 1.0 0.0

OLSSON [10] UNSPRUNG MASS SOFT HERTZIAN SPRING HARD HERTZIAN SPRING

0.1

0.2

0.3

0.4

0.5

VEHICLE SPEED / CRITICAL SPEED

Fig. 20. Comparing Olsson [10] unsprung model with the author sprung mass system using WRC elements.

Fig. 22. Simply supported beam subjected to a moving vehicle using different Hertzian spring stiffness on a rail with irregularities.

836 Table 1 Model properties Bridge properties

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837

Overall length of bridge (m) Self-weight of bridge (t) First natural frequency of unloaded bridge (Hz) First natural frequency loaded bridge (Hz) Train properties Mass of car body (t) Mass of each bogie frame (t) Mass of each axle together with wheels, mw (t) Stiffness of spring in primary suspension, k1 (kN/m) Damping of spring in primary suspension, c1 (kN s/m) Stiffness of spring in secondary suspension, k2 (kN/m) Damping of spring in secondary suspension, c2 (kN s/m) Hertzian spring stiffness, kH (kN/m) Distance between two centre of axles, l2 (m) Distance between two centre of bogie frames, l3 (m) Overall length of vehicle (m) Number of vehicles General properties Youngs modulus of elasticity (kN/m2) Density of steel (t/m3) Poissons ratio

81.77 275 3.14 1.94 Locomotive 90.958 10.175 4.522 7000 58.8 4100 22 1.40!106 1.8 14 18 1 Carriage 33.7 3.15 1.5 700 5.88 410 Fig. 24. Vehicle traversing the Boyne Viaduct with rail irregularities. 2.2 1.40!106 2.56 14 18.78 1 2.05!108 7.850 0.3

In addition the WRC element is used as a lateral spring to prevent the wheels of the vehicle from sliding laterally off the rails, something not possible using the contact element of ANSYS. This example studies the rolling of a railway carriage about the long axis, induced by rails with irregularities out of phase. A plot of the two rails (A and B) with irregularities out of phase is presented in Fig. 23, with each irregularity having a wavelength of 40 m and amplitude of 0.01 m. Rails A and B are indicated on Fig. 24. The train model consists of a six-axle locomotive pulling a single four-axle railway carriage traversing the Boyne Viaduct railway bridge at (20 m/s) 72 km/h from left to right. Time t is arranged in such a manner that the front
IRREGULARITY FUNCTION (m) .
0.0100 Irregularity along rigid rail 0.0050 Irregularity along bridge

wheel of the train arrives at the left support at tZ0.5 s and the rear wheel of the train leaves the right support at tZ 6.67 s. The simulation assumes that the bridge is horizontal in the absent of gravity and bridge damping is neglected. In Fig. 24, the authors show an image, taken from an animation, of the train traversing the Boyne Viaduct, inuenced by the irregularities on the rail using their WRC element. One can see from the image that the irregularities cause the locomotive to rolls about its long axis as it traverses the bridge. It should be noted that the displacements have been increased by a factor of 10. In the results that follow the smooth rail is compared with the irregular rail. The vertical displacement at mid-point of the bridge (point T1 as shown in Fig. 4) and the vertical displacement of the front wheel of the train (point W1 in Fig. 5) on rail A are plotted as a function of time in Figs. 25 and 26, respectively. Fig. 27 presents the contact force that exists between the front wheel (W1) of the train and rail A, while the lateral displacement of the vehicle body (point V1 as shown in Fig. 5) can be seen in Fig. 28. Examining Figs. 23 and 26, one can see that the front wheel of the train follows the prole of the irregularity. From inspection of the graphs, one can also see that, for this particular speed and irregularity, the dynamic effects of the bridge are less signicant than those of the train.
VERTICAL DISPLACEMENT (m)
0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 Front wheel on left support Front wheel on right support 0 1 2 3 4 5 6 Rear wheel on right support 7 SMOOTH RAIL IRREGULARITIES

RAIL A RAIL B

0.0000

0.0050

0.0100 40

20

20

40

60

80

100

120

DISTANCE (m)

TIME (sec)

Fig. 23. Rail irregularities out of phase along bridge.

Fig. 25. Vertical displacement at mid-point of the Boyne Viaduct (T1).

C.J. Bowe, T.P. Mullarkey / Advances in Engineering Software 36 (2005) 827837


LATERAL DISPLACEMENT (m)
0.010 VERTICAL DISPLACEMENT (m) 0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070 0.080 0 1 2 3 4 5 6 7 SMOOTH RAIL IRREGULARITIES

837

0.02 IRREGULARITIES 0.01

0.00
0 1 2 3 4 5 6 7

0.01

TIME (sec)

0.02

TIME (sec)

Fig. 26. Vertical displacement of the rst wheel of the train on rail A (W1).

Fig. 28. Lateral displacement of the vehicle body (V1).

50 0 CONTACT FORCE (kN) 70 90 110 130 150 170 TIME (sec) 1 2 3 4 5 6 7

SMOOTH RAIL IRREGULARITIES

Fig. 27. Contact force between the rst wheel of the train and rail A (W1).

5. Conclusions In this paper, the authors develop their own wheel-rail contact element to simulate the dynamic interaction that exists between a sprung wheel and the rail. This system uses a Hertzian spring to represent the wheel, and is modelled using three stiffness matrices. Wheel-rail separation is determined by calculating the extension of the spring at each timestep. If the extension become positive, the contact force is put equal to zero and all stiffness matrices related to that wheel are set equal to zero. The authors compare the use of their WRC element with the use of the ANSYS CONTAC48 element. They also compare their results with some simple analytical solutions developed by Biggs [8]. In the case of the wheel considered as a moving sprung load, the wheel considered as a moving sprung mass, and the travelling bouncing wheel, the solutions using the WRC element are identical to the solutions using the ANSYS CONTAC48 element. For analytical solutions, the WRC element compares better with ve modes than with two modes. The model is next validated for a pair of moving masses separated by a spring for a wide range of speeds. The authors compare their WRC element with the unsprung mass system of Olsson [10]; however, the authors use both hard and soft Hertzian springs

for the WRC elements because Hertzian springs are absent from Olssons [10] model. The authors results compare very favourably with Olssons [10] results. Finally the authors subject the Boyne Viaduct to a moving locomotive and railway carriage on smooth and irregular rails. In this example, the WRC has a threedimensional formulation and it simulates the vertical and lateral interaction between the wheel and the rail, the rail being smooth or irregular. Rolling motion of the train is evident in the case of irregularities that are out of phase on the two rails. Based on the results, summarised above, the authors can claim that the WRC element is exceedingly robust.

References
[1] Cheng YS, Au FTK, Cheung YK, Zheng DY. On the separation between moving vehicles and bridges. Journal of Sound and Vibration 1999;222(5):781801. [2] Yang YB, Wu YS. A versatile element for analyzing vehicle-bridge interaction response. Engineering Structures 2001;23:45269. [3] Yau JD, Yang YB, Kuo SR. Impact response of high speed rail bridges and riding comfort of rails cars. Enginnering Structures 1999;21: 83644. [4] Au FTK, Wang JJ, Cheung YK. Impact study of cable-stayed railway bridges with random rails irregularities. Engineering Structures 2002; 24:52941. [5] Chu KH, Garg VK, Bhatti MH. Impact in truss bridge due to freight trains. Journal of Engineering Mechanics 1985;3:15973. [6] Wiriyachai A, Chu KH, Garg VK. Bridge impact due to wheel and track irregularities. Journal of the Engineering Mechanics Division 1982;108:64865. [7] Esveld C. Modern railway track, 1st Edition, MRT Production, 1989. [8] Biggs JM. Introduction to Structural Dynamics, 1st Edition, Englewood Cliffs, Prentice-Hall, 1964. [9] Bathe KJ. Finite element procedures, 1st Edition, Englewood Cliffs, Prentice-Hall, 1996. [10] Olsson M. Finite element, modal co-ordinate analysis of structures subjected to moving loads. Journal of Sound and Vibration 1985;99: 112.

Anda mungkin juga menyukai