You are on page 1of 19

This article was downloaded by: [Anadolu University]

On: 25 December 2014, At: 14:36


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Vehicle System Dynamics: International


Journal of Vehicle Mechanics and
Mobility
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/nvsd20

Tangential problem solution for non-


elliptical contact areas with the
FastSim algorithm
a b
A. Alonso & J. G. Giménez
a
CEIT, Paseo M. Lardizabal, 13, 20018, San Sebastían (Guipúzcoa),
Spain
b
Construcciones y Auxiliar de Ferrocarriles Investigación y
Desarrollo , S.L. and TECNUN (University of Navarra) , CAF I+D.
J.M. Iturrioz, 26, 20200, Beasain (Guipúzcoa), Spain
Published online: 21 Mar 2007.

To cite this article: A. Alonso & J. G. Giménez (2007) Tangential problem solution for non-elliptical
contact areas with the FastSim algorithm, Vehicle System Dynamics: International Journal of
Vehicle Mechanics and Mobility, 45:4, 341-357, DOI: 10.1080/00423110600999763

To link to this article: http://dx.doi.org/10.1080/00423110600999763

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [Anadolu University] at 14:36 25 December 2014
Vehicle System Dynamics
Vol. 45, No. 4, April 2007, 341–357

Tangential problem solution for non-elliptical


contact areas with the FastSim algorithm
A. ALONSO*† and J. G. GIMÉNEZ‡
†CEIT, Paseo M. Lardizabal, 13, 20018 San Sebastían (Guipúzcoa), Spain
Downloaded by [Anadolu University] at 14:36 25 December 2014

‡Construcciones y Auxiliar de Ferrocarriles Investigación y Desarrollo,


S.L. and TECNUN (University of Navarra) CAF I+D. J.M. Iturrioz, 26,
20200 Beasain (Guipúzcoa), Spain

This article deals with the application of the FastSim algorithm to the solution of the tangential contact
problem for non-elliptical contact areas.
At first, the causes creating problems for the solution of non-hertzian contact areas with this
algorithm shall be analyzed. Then, different currently existing methods shall be studied, analyzing
their accuracy characteristics and computational cost to determine whether or not they are appropriate
to use in dynamic simulations. Finally, a new strategy shall be proposed that, in the opinion of the
authors, offers good characteristics of precision and computational cost.

Keywords: Tangential contact problem; Non-hertzian; Wheel–rail

1. Introduction

Over the last 40 years different methods have been developed to solve the wheel–rail contact
problem. The main differences between these methods regarding their possible implementation
in a railway dynamic simulation program are as follows.

1.1 Versatility

The various contact theories have been developed based on different initial hypotheses. Hence,
some are limited to the specific case where the contact areas have an elliptic or even circular
shape whereas others allow for arbitrary shapes.
Also, some algorithms allow for the use of variable friction coefficients dependent on the
local slip while others establish that the friction coefficient remains constant.
Other differences between the methods are the possibility, or not, of solving the problem
considering the transient effects of permitting, or not, general combinations between creepages
and spin, etc.

*Corresponding author. Email: aalonso@ceit.es

Vehicle System Dynamics


ISSN 0042-3114 print/ISSN 1744-5159 online © 2007 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/00423110600999763
342 A. Alonso and J. G. Giménez

1.2 Computational efficiency

In a railway dynamic simulation the contact problem must be solved many times and the use
of a method with an excessive computational cost may excessively slow down the simulations
until they are unfeasible.

1.3 Accuracy

This ranges from the achievement of merely qualitative results to the achievement of exact
values under certain conditions.
Among the various existing algorithms to solve the tangential problem of contact in dynamic
simulations the most used is FastSim [1]. This is because it has the best computational cost-
Downloaded by [Anadolu University] at 14:36 25 December 2014

accuracy ratio [2]: its maximum errors are around 10% with respect to the results offered by
the ‘exact’ methods and its computational cost is significantly less.
Regarding the field of application of FastSim (versatility) it can be said that this algorithm
allows arbitrary combinations between longitudinal, lateral and spin creepages. However,
during its initial development a series of hypotheses were made, limiting its field of application.
Among these, the following are worth pointing out: the friction coefficient is constant and
independent of the local slip value; the non-stationary effects do not affect the problem’s
solution; the contact area must have an elliptic shape.
The above hypotheses may be assumed for the solution of some railway problems. However,
in other cases (reduced radii curves negotiation, squeal noise generation and corrugation,
wheel-slide protection systems design, etc.) its use may lead to inaccurate results.
In order to overcome some of these limits, some modifications have been made to the
original algorithm.
Hence, Giménez et al. [3] have developed a formulation allowing the friction coefficient
to be introduced in FastSim according to the local slip at each point. Also, Shen and Li [4]
searched for a method of considering the non-stationary terms in the solution of the contact
problem.
With regards to the assumption that the contact areas have an elliptic shape, it can be said
that it is not true in a large number of cases (for example when the contact is produced in
the transition between the rolling surface and flange). As a result, recently various theories
have been developed to resolve both the normal contact problem and the tangential problem
(Kalker [5, 6], Alonso and Giménez [7], Knothe and Hung [8, 9], Kik and Piotrowski [10],
Ayasse and Chollet [11]) without being restricted to hertzian geometries.
Current work will focus on the solving of tangential problems with non-hertzian contact
areas by means of the FastSim algorithm. At first, the causes creating problems for the solution
of non-hertzian contact areas with this algorithm will be analyzed. Then, different currently
existing methods will be studied, analyzing their accuracy characteristics and computational
cost to determine whether or not they are appropriate to use in dynamic simulations. Finally,
a new strategy will be proposed that, in the opinion of the authors, offers good characteristics
of precision and computational cost.

2. Solution of the non-hertzian contact problem with FastSim

FastSim is an algorithm to solve the contact tangential problem. This algorithm supposes that
the normal and tangential problems can be solved separately. It is worth pointing out that this
assumption is true if both bodies are quasiidentical (material and geometrical symmetry).
Tangential problem solution with the FastSim algorithm 343

The most used theory to solve the normal problem (compatible with the original version
of FastSim) is the Hertz theory. Using this theory the contact area is elliptic in form and the
distribution of normal pressures, an ellipsoid. However, if the hypotheses of the Hertz theory
are not satisfied then it is necessary to resort to more complex methods. Most of these strategies
(such as the Kalker variational method [6]) have an excessive computational cost to be used
in a dynamic simulation program. Thus, over the last few years other less precise but more
computationally efficient methods have been developed. Among these it is worth pointing out
the works of Kalker [5], Alonso and Giménez [7, 12] Kik and Piotrowski [10] and Ayasse and
Chollet [11].
Once the normal contact problem has been solved the solution to the tangential problem
can be reached by means of the FastSim algorithm (using the contact area and the pressure
distribution obtained in the resolution of the normal contact problem as inputs).
One of the basic characteristics of the FastSim algorithm is that it is based on the simp-
Downloaded by [Anadolu University] at 14:36 25 December 2014

lified elasticity theory. This theory assumes a proportional ratio between stresses ‘τ ’ and
displacements ‘u’ according to the equation:

ui (x, y) = Li · τi (x, y) (1)

where Li is a flexibility coefficient, and (x, y) is a generic point of the contact area.
According to the equation (1), displacement in one point on the surface of a solid depends
only on the stress on this point, in opposition to the general elasticity theory where displacement
at one point depends not only on the stress produced as that point but also on the stresses
produced at all points of the contact area.
In order to apply the expression (1), it is necessary to assign a value to the flexibility
coefficient ‘L’. To this end, Kalker used as a basis the asymptotic theory for small slips
developed by himself [13]. This theory exactly solves the tangential problem where the contact
area is elliptic and the whole area is in adhesion conditions (i.e. in the case where the creepage
value is infinitesimal).
Equalizing the results obtained with the asymptotic theory of Kalker and FastSim
(maintaining the hypothesis in this algorithm that all of the contract area is in adhesion
conditions), the flexibility coefficients associated to each creepage component can be
calculated:

8a
L1 =
3C11 G
8a
L2 = (2)
3C22 G
πa 3/2
L3 = √
4 bC23 G

where: L1 , L2 and L3 are the flexibility coefficients associated with the longitudinal, transversal
and spin creepages respectively; a and b are the longitudinal and lateral semi-axes of the
contact ellipse; G is the combined shear strength module; C11 , C22 and C23 are the ‘Kalker
coefficients’.
The ‘Kalker coefficients’ are obtained directly from the application of the asymptotic theory
and have been tabulated according to the Poisson module and to the ratio between the semi-axis
of the contact ellipse [13].
As has been seen, the method used for the FastSim algorithm for the determination of the
flexibility coefficients requires the contact areas to be elliptic. Due to this, if the FastSim field
344 A. Alonso and J. G. Giménez

of application is to be extended, a way of calculating said coefficients for non-hertzian areas


must be found. To this end, two strategies can be followed:
• To search for a way of directly calculating the flexibility coefficients. Among the methods
developed following this strategy it is worth pointing out the one proposed by Kalker [5]
(In this work this method will be called ‘deformations levelling method (DLM)’).
• To follow a strategy similar to that of Kalker [1]; i.e. to develop an asymptotic theory for
small slips which are not restricted to elliptical areas and, subsequently, search for a method
that permits the calculation of the flexibility coefficients of the simplified theory from the
results of the asymptotic method. Among the methods developed to solve the tangential
problem for small slips and non-elliptical areas it is worth pointing out the Kalker variational
method [6] and the Knothe and Hung method [8, 9] (In this article this method will be called
‘asymptotic strips method (ASM)’).
Downloaded by [Anadolu University] at 14:36 25 December 2014

The following section will briefly include the theoretical bases of these methods and will
discuss the application possibilities according to their precision and computational cost.

3. Methods of determination of the flexibility coefficients

3.1 Analysis of the existing methods

As has been seen, the Kalker coefficients were initially defined for elliptic contact areas. Due
to this, and in order to be able to compare the precision of different methods the concept of the
Kalker coefficient will be generalized. To this end, we shall base ourselves on the fact known
according to which if the slips are infinitesimal, the forces transmitted through the contact
area respond to the following equations:
 3/2
A A A
Fx = −C̃11 G ξ − C̃12 G η − C̃13 G φ
π π π
 3/2 (3)
A A A
Fy = −C̃21 G ξ − C̃22 G η − C̃23 G φ
π π π

where: C̃ij are the Kalker coefficients for non-elliptical contact areas; A is the surface of the
contact area; Fx and Fy are the longitudinal and transversal forces transmitted through the
contact area.
In the case of elliptical contact areas, for symmetry reasons the coefficients C̃12 , C̃13 and
C̃21 are null. In another type of contact area it has been proven that the coefficients C̃12 and
C̃21 are very small. This does not happen, however, with the coefficient C̃13 that can reach
rather significant values in specific situations.

3.1.1 Kalker variational method. A first way of calculating the Kalker coefficients of
non-elliptical contact areas would be to use the Kalker variational method [6]. To this end, the
tangential problem must be solved, assuming that the friction coefficient is infinite.
It is worth remembering that this method converges towards exact results as the number of
elements the contact area is divided into increases.
The assumption that the whole contact area is in adhesion condition greatly simplifies the
algorithm, although the method can only be used in dynamic simulations if the number of
elements required to obtain sufficient accuracy is very low. This is because the computational
cost of the method increases quickly as the mesh refinement increases.
Tangential problem solution with the FastSim algorithm 345

In a first step, the convergence speed of the method to the exact result will be checked
in order to determine the number of elements required to achieve a specific precision level.
To this end, hertzian cases will be solved as the exact results are known (provided by the
Kalker asymptotic theory (AMK)). In the figures 1–3 below, the Kalker coefficients for a
circular contact area are represented depending on the number of elements used in both
directions.
As can be seen, the problem slowly converges (particularly in the case of coefficient C23 ).
With discretizations of approximately 20 × 20 elements the errors in the calculation of the
three coefficients are about 5%.
The reason for the slow convergence of the variational method (VM) is the assumption that
all the contact area is in adhesion condition. Under this hypothesis, on the trailing edge of the
contact area the tangential stresses tend to be infinite, leading to high errors on both the stress
calculation and its numerical integration to calculate the contact forces.
Downloaded by [Anadolu University] at 14:36 25 December 2014

From the above mentioned it is deduced that the Kalker variational method cannot be directly
used in a railway dynamic simulation, since in order to achieve the required precision, the
computational cost will be extremely high. Hence, its use in this work has been reduced to the
validation of the different methods considered.

Figure 1. Coefficient C11 according to the number of elements (a/b = 1).

Figure 2. Coefficient C22 according to the number of elements (a/b = 1).


346 A. Alonso and J. G. Giménez
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 3. Coefficient C23 according to the number of elements (a/b = 1).

3.1.2 Deformations levelling method. The idea of this method, proposed by Kalker [5],
consists of calculating the flexibility coefficient, making this minimize the difference between
the displacements calculated by numerical integration of the Cerruti formula and by the
simplified theory.
If the contact area is discretized in N elements and it is assumed that the tangential stresses
in each item are known ‘ταm ’, the distribution of displacements can be estimated by means of
the Cerruti equations:
uαm = Aαmn · ταn (4)

where Aαmn is the influence coefficients matrix.


However, if the simplified elasticity theory is used, the displacements must be calculated
using the following equation:
uαm = L · ταm + aα (5)

Applying the least squares method, the ‘L’ value can be obtained as that minimizing the
differences between the displacement fields calculated with both theories.
As can be seen in the equations (4) and (5), to perform the flexibility coefficient calculation
it is necessary to previously assume a distribution of tangential stresses. Normally (as a first
approximation) the stress distribution corresponding to the case, where all the contact area is
in slip conditions, is generally used. With this distribution a flexibility coefficient is calculated
whereby the tangential problem is solved again, in this way obtaining a new tangential stress
distribution. This distribution is used to calculate a new value for ‘L’. This process continues
until the convergence is achieved in the results.
Once the way to calculate the flexibility coefficient has been exposed, the precision of the
method shall be determined. To this end two checks are to be performed (both with hertzian
contact areas).
Firstly, the Kalker coefficients obtained via the DLM have been compared with those
obtained via the Kalker asymptotic method. As can be seen in table 1, the error made
in the calculation of coefficients C11 and C22 is approximately 15%. However, these errors
increase in the calculation of C23 reaching up to approximately 75%. It is important to
remember that using the flexibility coefficients calculated with the equation (2) exact results
are achieved in these calculations. Hence, the conclusion can be drawn that the use of this
method leads to a significant loss of precision in the Kalker coefficients calculation.
Tangential problem solution with the FastSim algorithm 347

Table 1. Kalker coefficients calculated by means of the DLM and by the TAK.

C11 C22 C23


a/b DLM TAK Error (%) DLM TAK Error (%) DLM TAK Error (%)

0.10 3.31 3.04 8.25 2.52 2.43 3.41 0.47 0.13 72.42
1.00 4.12 4.49 9.10 3.67 4.08 11.22 1.47 1.23 16.39
10.00 11.70 13.23 13.11 12.80 14.53 13.49 14.60 10.73 26.49
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 4. Longitudinal force vs. longitudinal creepage.

Secondly, the results provided by the VM have been compared with those obtained with
the FastSim algorithm using the new flexibility coefficients (FastSimM). The results obtained
with the original FastSim algorithm (FastSim) have also been included in the comparison.
As a result of this comparison it has been found that the precision of the new method is
similar to the original FastSim in certain cases. In other cases, however, there is an appreciable
loss of precision as can be seen in figures 4 and 5.
Finally and to complete the analysis of the method, it is worth pointing out that the new
method of calculating the flexibility coefficients significantly increases the computational cost

Figure 5. Lateral force vs. spin.


348 A. Alonso and J. G. Giménez

of the FastSim algorithm: in the best case, to solve the tangential problem it must be solved at
least twice (once for full slip and once with the second estimation of the flexibility coefficient);
also the least squares method must be used at least twice to calculate the value of ‘L’ which
best approximates the displacement distributions.

3.1.3 Asymptotic strip method. The basic idea of this method, developed by Knothe and
Hung [8, 9] consists of solving the contact equations (particularized for the case where the
whole contact area is in adhesion conditions) previously assuming a second-order polynomial
distribution for the tangential stresses.
To solve the rolling contact equations the contact area is divided into M strips parallel to
the rolling direction (‘y’ coordinate value constant). In each strip the real tangential stresses
are approximated by quadratic functions in the ‘x’ direction, which are cancelled on the
Downloaded by [Anadolu University] at 14:36 25 December 2014

leading edge. Hence, the distribution of pressures in each strip are defined by 4 coefficients
(4M coefficients in the whole contact area)
The 4M coefficients are determined by the collocation method imposing the equations be
fulfilled at two points per strip. The position of the collocation points is chosen such that the
errors are minimized in the case of elliptical contact areas.
Consequently, a system of 4M equations is obtained which is necessary to solve in order to
obtain the distribution of tangential stresses.
Once the calculation method has been presented the precision of the method will be
determined. For this, first the Kalker coefficients obtained for elliptical contact areas will
be compared with those provided by the AMK. The results of this comparison are included in
table 2.
As can be seen, the results are rather accurate (in all cases the errors are less than 10%
and in most cases less than 5%). To see this in a more graphic form, figure 6 shows the C23
coefficient as a function of the relation between the semi-axes of the contact ellipse.

Table 2. Comparative table between the ASM and the asymptotic method of Kalker (AMK) for v = 0.25.

ASM AMK
g C11 C22 C23 C11 C22 C23

g= b
a if b < a
0.1 3.34 2.53 0.55 3.31 2.52 0.47
0.2 3.38 2.61 0.65 3.37 2.63 0.60
0.3 3.44 2.71 0.74 3.44 2.75 0.71
0.4 3.51 2.81 0.83 3.53 2.88 0.82
0.5 3.59 2.92 0.92 3.62 3.01 0.93
0.6 3.65 3.02 1.01 3.72 3.14 1.03
0.7 3.74 3.13 1.11 3.81 3.28 1.14
0.8 3.82 3.24 1.20 3.91 3.41 1.15
0.9 3.91 3.35 1.30 4.01 3.54 1.36
g= a
b if b > a
1 4.00 3.46 1.40 4.12 3.67 1.47
0.9 4.09 3.57 1.51 4.22 3.81 1.59
0.8 4.21 3.73 1.65 4.36 3.99 1.75
0.7 4.35 3.90 1.83 4.54 4.21 1.95
0.6 4.56 4.16 2.09 4.78 4.50 2.23
0.5 4.83 4.49 2.44 5.10 4.90 2.62
0.4 5.25 4.99 3.01 5.57 5.48 3.24
0.3 5.93 5.81 4.01 6.34 6.40 4.32
0.2 7.21 7.39 6.18 7.78 8.14 6.63
0.1 10.76 11.92 14.13 11.70 12.80 14.60
Tangential problem solution with the FastSim algorithm 349
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 6. C23 according to the relation between the contact semi-axes calculated with the Kalker asymptotic method
and the Knothe and Hung Method.

Once the method accuracy has been determined for hertzian cases, this precision will be
checked in non-elliptical contact areas. To perform this comparison, the three contact areas
shown in figure 7 have been chosen. It is worth pointing out that in these areas, due to symmetry
reasons, the coefficients C̃13 and C̃21 are null in all of these cases (the coefficient C̃12 is also
null in the first two cases).

Figure 7. Non-hertzian contact areas.


350 A. Alonso and J. G. Giménez

Table 3. Comparison of the results between the ASM and the VM.

C̃11 C̃22 C̃23


Caso ASM VM Error (%) ASM VM Error (%) ASM VM Error (%)

Figure 4(a) 4.51 4.41 2.27 3.98 3.91 1.79 1.5572 1.462 6.51
Figure 4(b) 3.77 4.21 10.45 3.24 3.88 16.49 1.4874 1.5283 2.68
Figure 4(c) 4.27 4.25 0.47 3.73 3.78 1.32 1.5058 1.6866 10.72

As can be seen, the shapes of the contact areas tested are very different to those usually found
in the wheel–rail contact problems. However, they are useful to make simple comparisons with
the Kalker variational method.
Also, it is logical to assume that if good results are achieved with these areas then the
Downloaded by [Anadolu University] at 14:36 25 December 2014

precision using this method with more typical contact areas in the wheel–rail contact will be
similar or even better. This is because the collocation points have been determined, minimizing
the errors in hertzian cases, whereby it is logical to assume that the further the contact area is
from the elliptical shape the worse the results.
Table 3 shows the coefficients C̃11 , C̃22 and C̃23 for the asymptotic strips method (ASM)
and the VM.
As can be seen, the calculation of all the coefficients is performed in a rather precise manner
(although in some cases the error can reach around 20%).
However, using this method in dynamic simulations can also be advised against as, once pro-
grammed, it has been found that its computational cost is too high, and its use will excessively
slow down the dynamic simulations.

3.2 Method of assimilation of the contact area to two semi-ellipses (MAE)

The calculation of the Kalker coefficients by means of this method is performed by dividing
the real contact area in two parts and associating a semi-ellipse to each part of the contact area.
Finally it is assumed that the Kalker coefficients in each zone of the real area are the same as
those of the associated semi-ellipses.
The process is applied in five steps:

Step 1 Calculation of the centre of gravity and main axes of inertia


The centre of gravity of the real contact zone is calculated as well as the main axes of inertia.
In the case of the wheel–rail contact it can be considered, without committing a significant
error, that both axes of inertia coincide respectively with the vehicle’s travelling direction and
with the transversal direction. These axes are called axis X and axis Y respectively.

Step 2 Division of the contact area


The contact area is divided taking the X axis obtained as the separation line, creating two
contact sub-zones.

Step 3 Assimilation of each sub-area to a semi-ellipse


To liken each sub-zone to a semi-ellipse the average quadratic values associated to ‘x’ and
‘y’ coordinates of the edge of the real contact zone and of the related ellipse are made to
coincide.
Tangential problem solution with the FastSim algorithm 351

Hence, from the average quadratic values (VExi , VEyi ) of each part of the real contact area,
the values of the semi-axes of the related semi-ellipse can be calculated:

3
ai = VExi
2
 (6)
3
bi = VEyi
2

Step 4 Calculation of the Kalker coefficients for each contact sub-area


Once the size of the semi-axes is known, the TAK can be used for small slips to calculate
the Kalker coefficients.
It is important to mention that the usual Kalker coefficient tables only include the coefficients
Downloaded by [Anadolu University] at 14:36 25 December 2014

C11 , C22 and C23 for the whole contact ellipse. To solve the tangential problem using this
method, a table must therefore be created containing the coefficients Cijm , (i = 1, 2, 3, j =
1, 2, 3, the superindex ‘m’ states that the value only relates to half the contact area).
m m
As mentioned previously, the effects of the coefficients C12 and C21 can be neglected and
therefore there is no need for them to be included in the aforementioned table.

Step 5 Calculation of the forces transmitted


Finally, to calculate the force transmitted via the contact area, assuming small slips, the
equation (3) can be used where:
(Cij1 + Cij2 )
C̃ij = (7)
2

3.2.1 Method validation. To validate the method the C̃ij coefficients obtained with the
proposed method will be compared with those provided by the variational theory.
Before making the comparison it is important to point out that the results obtained with the
proposed method for hertzian cases are exact. Therefore, there is no loss of precision when
the new method is used in these cases.
Verification in the cases of non-elliptical contact areas is begun by solving the contact areas
shown in figure 7. The results obtained (included in table 4) show maximum differences,
compared to the VM, of around 15%.
Also comparisons have been made between the results obtained on solving the areas defined
by means of the following equations:
 2  2
x y
Si x < 0 −→ + =1
a1 b1
 2  2 (8)
x y
Si x > 0 −→ + =1
a2 b2

Table 4. Comparison of the results between the MAE and the VM.

C̃11 C̃22 C̃23


Caso MAE VM Error (%) MAE VM Error (%) MAE VM Error (%)

Figure 4(a) 4.12 4.41 6.58 3.67 3.91 6.15 1.47 1.46 0.55
Figure 4(b) 4.12 4.21 2.14 3.67 3.88 5.41 1.47 1.53 3.81
Figure 4(c) 4.10 4.25 3.53 3.65 3.78 3.44 1.45 1.69 14.03
352 A. Alonso and J. G. Giménez

Table 5. Values of C11 , C13 , C22 and C23 obtained with the MAE and the VM.

Caso C̃11 C̃13 C̃22 C̃23


g1 g2 VM MAE VM MAE VM MAE VM MAE

0.10 1.00 3.91 3.72 1.27 1.16 3.38 3.10 1.13 0.97
0.50 1.00 3.81 3.87 0.15 0.13 3.35 3.34 1.23 1.20
1.00 1.00 3.95 4.12 0.00 0.00 3.55 3.67 1.43 1.47
2.00 1.00 4.53 4.61 0.00 0.02 4.30 4.29 2.20 2.05
5.00 1.00 5.98 5.95 0.08 0.26 6.12 5.91 5.24 4.05

where the values of ai , bi are defined in the following manner:


1
ai · bi =
Downloaded by [Anadolu University] at 14:36 25 December 2014

π
(9)
ai
= gi
bi
Table 5 shows the results obtained for different combinations of g1 and g2 < 0.
As we can see, the results obtained are relatively good. In the case of coefficients C̃11 and
C̃22 the maximum differences between both methods are around 10%. In the case of coefficient
C̃23 these differences increase, reaching in one case, to around 25%.
In the case of C̃13 the results are very good except for the case of (g1 = 5, g2 = 1) where the
errors are greater. However, it is important to point out that except for the case of (g1 = 0.1,
g2 = 1) the value of coefficient C̃13 is lower than the other coefficients.
With regards to the computational cost of this method, it can be said that this is very low
since it only needs to calculate the centre of gravity and the average quadratic values of each
sub-zone of contact, and interpolate in the tables.
As a final comment it is worth mentioning the suitability of the assimilation method of the
contact areas to two semi-ellipses to be used in dynamic simulations. This is because said
method combines adequate precision with a low computational cost.

3.2.2 Flexibility coefficients in non-hertzian areas. Once the asymptotic problem has
been solved for small slips a method must be developed to obtain the flexibility coefficients.
For hertzian areas, Kalker proposed solving the tangential problem with FastSim assuming
adhesion conditions in all the contact areas and equalizing the results with those of the
asymptotic theory.
For non-hertzian areas the methodology to be used could be the same as that proposed by
Kalker. However, as the shape of the contact area generally does not respond to a geometrically
defined shape it is impossible to obtain an analytical equation of the flexibility coefficients,
with the subsequent increase in computational cost. As a result, a cost effective method of
calculating these flexibility coefficients is necessary.
The proposed solution is as follows: a semi-ellipse will be associated to each contact
sub-zone (used in step 2 of the Kalker coefficients calculation) and it is assumed that the
flexibility coefficients of each contact sub-zone are the same as those that shall correspond to
the associated semi-ellipse.
For the calculation of these semi-ellipses the following conditions are imposed:
• The ratio between the semi-axes of each semi-ellipse is the same as those of the semi-ellipses
calculated for the determination of the ‘Kalker coefficients’ (values calculated with the
equation (6)).
• The area of the associated semi-ellipses is the same as the area of each contact sub-zone.
Tangential problem solution with the FastSim algorithm 353

Once the semi-axes of the associated semi-ellipses have been obtained for these conditions
the equation (2) could be used directly to calculate the flexibility coefficients. However, using
this formula will only provide exact results for small slips in the calculation of the longitudinal
force generated by the longitudinal creepage and of the lateral force produced by the lateral
creepage and the spin. However, for non-elliptical contact areas both the lateral creepage and
the spin can also produce longitudinal forces. Moreover, the longitudinal creepage can also
produce lateral forces (see equation (3)).
Due to this, the present work proposes using four flexibility coefficients instead of the three
coefficients used originally in FastSim. In the asymptotic case, where the whole contact area
is in adhesion conditions, the null slip condition gives rise to the following equation:

⎡ ξ φy ⎤
  −
⎢ L1 L4 ⎥ ∂ τ
Downloaded by [Anadolu University] at 14:36 25 December 2014

0
=⎢

⎥+ (10)
0 η φx ⎦ ∂x
+
L2 L3

Integrating this equation in a semi-ellipse, the forces transmitted through this contact area can
be obtained:
⎡ 2 ⎤
4a b a 2 b2 ⎡ ⎤
  0 ± ξ
⎢ 3L 2L4 ⎥
Fx ⎢ 1 ⎥⎢ ⎥
= −⎢ ⎥⎣ ⎦
η (11)
Fy ⎣ 4a 2 b π a3b ⎦
0 φ
3L2 2L3
The sign of the item (1, 3) of the matrix relating forces and creepages depends on whether the
semi-ellipse is in the half plane y > 0 or y < 0.
As can be seen, in spite of the use of a fourth coefficient, L4 , in equation (11), the terms
(1, 2) and (2, 1) are null. However, in a general case, these coefficients are not zero. Hence,
it can be concluded that it is impossible to equalize the results obtained with those obtained
with the equation (11) and with the equation (3).
In spite of this, as we have mentioned, coefficients C̃12 and C̃21 are very small in the
practical cases of wheel–rail contact whereby its effect can be neglected without committing
a significant error.
If this assumption is made and the results obtained by the VM and FastSim are equalized
the following flexibility coefficients are obtained:

4a
L1 = m
3C̃11 G
4a
L2 = m
3C̃22 G
(12)
πa 3/2
L3 = √ m
8 bC̃23 G

ab
L4 = m
2C̃31 G

As can be seen the coefficients L1 , L2 and L3 are equal to those proposed by Kalker and L4
m
depends exclusively on the Kalker coefficient C̃31 .
354 A. Alonso and J. G. Giménez

3.3 Results of the FastSim algorithm for non-elliptical contact areas

This section analyses the accuracy of the FastSim algorithm (adapted according to the strategy
previously proposed) in the solution of the tangential problem with non-elliptical contact areas.
To this end the case shown in figure 8 shall be solved with different combination of creepages
and spin using the proposed method and the VM. The reason in order to choose such an area
are various: the contact area shape chosen is very different to that relating to a hertzian case,
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 8. Non-elliptical contact area.

Figure 9. Longitudinal force calculated with the VM and with FastSim according to the longitudinal creepage
(η = φ = 0).
Tangential problem solution with the FastSim algorithm 355
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 10. Transversal force calculated with the VM and with FastSim according to the longitudinal creepage
(ξ = φ = 0).

it is easy to make comparisons and finally, there is the opportunity to show the influence of
the parameter C13 (L4 ) in certain cases.
When making the comparison it shall be assumed that the results provided by the Kalker
variational theory are exact (although, as has been mentioned, these results are not completely
exact as the number of elements the contact area is divided into is not sufficient).
Figures 9–13 show the results of said comparison for different combinations of creepages.
As we can see, the differences between both methods are generally less than 10%. It can
also be stated that the order of the errors obtained is similar to that obtained for elliptical
contact areas.
Hence, the conclusion can be drawn that, using the strategy proposed in this chapter, the
FastSim algorithm can be used to calculate the non-hertzian contact areas.

Figure 11. Transversal force calculated with the VM and with FastSim according to the spin (ξ = η = 0).
356 A. Alonso and J. G. Giménez
Downloaded by [Anadolu University] at 14:36 25 December 2014

Figure 12. Longitudinal and transversal force calculated with the VM and with FastSim according to a combination
of longitudinal and lateral creepage (ξ = η = 0).

Figure 13. Longitudinal and transversal force calculated with the VM and with FastSim according to a general
combination of longitudinal and lateral creepages (ξ = h = f ).

4. Conclusions

To complete this work a summary of the main results and conclusions obtained is as follows:

• The application of the FastSim algorithm to non-hertzian geometry problems requires, as a


preliminary step, the determination of the flexibility coefficients.
• A new strategy has been proposed based on the assimilation of the contact zone with two
half ellipses which has been proven to be fast enough and precise enough to be used in
dynamic simulations.
• It has been stated that the precision of the FastSim algorithm in the solving of non-elliptical
contact areas using the method of assimilation of the contact area two semi-ellipses is very
similar to that obtained on solving hertzian cases.
Tangential problem solution with the FastSim algorithm 357

Acknowledgements

The authors wish to express their thanks to the financial support provided by the Ministry of
Foment of Spain (70028-T05) and by the Department of Education, Universities and Research
of the Basque Government (PI2003-2)

References
[1] Kalker, J.J., 1982, Fast algorithm for the simplified theory of rolling contact. Vehicle System Dynamics, 11, 1–13.
[2] Alonso, A., 2005, Nuevos desarrollos para la resolucíon de problema de contacto rueda-carril y su aplicacíon a
simulaciones dinámicas. Thesis University of Navarra.
[3] Giménez, J.G., Alonso, A. et al., 2005, Introduction of a friction coefficient dependent on the slip in the FastSim
algorithm. Vehicle System Dynamics, 43, 233–244.
[4] Shen, Z. and Li, Z., 1996, A fast non-steady state creep force model based on the simplified theory. Wear, 191,
242–244.
Downloaded by [Anadolu University] at 14:36 25 December 2014

[5] Kalker, J.J., 1983, A simplified theory for non-hertzian contact. Vehicle System Dynamics, 13, 43–45.
[6] Kalker, J.J., 1982, Two Algorithms for the Contact Problem in Elastostatics (University of Waterloo Press, Solid
Mechanics Division), 103–119.
[7] Alonso, A. and Giménez, J.G., 2005, A new method for the solution of the normal contact problem in the
dynamic simulation of railway vehicles. Vehicle System Dynamics, 43, 149–160.
[8] Knothe, K. and Le-The, H., 1985, A method for the analysis of the tangential stresses and the wear distribution
between two elastic bodies of revolution in rolling contact. International Journal of Solids and Structures, 21,
889–906.
[9] Knothe, K. and Le-The, H., 1986, Determination of the tangential stresses and the wear for the wheel–rail rolling
contact problem. Vehicle System Dynamics (Suppl), 264–277.
[10] Kik, W. and Piotrowski, J., 1996, A fast, approximate method to calculate normal load at contact between wheel
and rail and creep forces during rolling. 2nd Mini-Conference on Contact Mechanics and Wear of Rail/wheel
Systems, Budapest, Hungary, July 29–31.
[11] Ayesse, J.B. and Chollet, H., 2005, Determination of the wheel rail contact patch in semi-Hertzian conditions.
Vehicle System Dynamics, 43, 161–172.
[12] Alonso, A. and Giménez, J.G., 2005, Some new contributions to the resolution of the normal wheel rail contact
problem. Vehicle System Dynamics (Suppl), 44, 230–239.
[13] Kalker, J.J., 1967, On the rolling contact of two elastic bodies in the presence of dry friction. Thesis Delft.

You might also like