Table of Contents Author Guidelines Submit a Manuscript
Erratum

An erratum for this article has been published. To view the erratum, please click here.

Advances in Tribology
Volume 2008 (2008), Article ID 214894, 17 pages
http://dx.doi.org/10.1155/2008/214894
Research Article

A Model of the Transient Behavior of Tractive Rolling Contacts

Mechanical Engineering Department, Division Production Engineering, Machine design and Automation, Katholieke Universiteit Leuven, Celestijnenlaan 300B, B-3001 Heverlee, Belgium

Received 19 September 2007; Accepted 6 February 2008

Academic Editor: Mihai Arghir

Copyright © 2008 Farid Al-Bender and Kris De Moerlooze. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

When an elastic body of revolution rolls tractively over another, the period from commencement of rolling until gross rolling ensues is termed the prerolling regime. The resultant tractions in this regime are characterized by rate-independent hysteresis behavior with nonlocal memory in function of the traversed displacement. This paper is dedicated to the theoretical characterization of traction during prerolling. Firstly, a theory is developed to calculate the traction field during prerolling in function of the instantaneous rolling displacement, the imposed longitudinal, lateral and spin creepages, and the elastic contact parameters. Secondly, the theory is implemented in a numerical scheme to calculate the resulting traction forces and moments on the tractive rolling of a ball. Thirdly, the basic hysteresis characteristics are systematically established by means of influence-parameters simulations using dimensionless forms of the problem parameters. The results obtained are consistent with the limiting cases available in literature and they confirm experimental prerolling hysteresis observations. Furthermore, in a second paper, this theory is validated experimentally for the case of V-grooved track.

1. Introduction

When1 an elastic body of revolution rolls tractively over another, the traction field in the contact patch changes progressively with the distance traversed, from its initial distribution, until it reaches a certain constant distribution. This distribution, which is independent of the initial field prior to commencement of rolling, does not vary with further (steady-state) rolling. This eventual rolling regime is termed gross rolling; the period building up to it, from commencement of rolling, is termed the prerolling regime. The resultant traction in this regime is characterized by rate-independent hysteresis behavior with nonlocal memory in function of the traversed displacement [1]. Although steady-state gross rolling is fairly well understood and theoretically founded, the situation is different in regard to prerolling. This paper deals with the theoretical treatment of the prerolling period.

The research on tractive rolling contact phenomena dates back to 1875 when Reynolds [2] describes the phenomenon of creepage. He uses creepage measurements between a rubber cylinder and a metal plate to confirm his proposition that the contact region of a rolling contact is divided into stick zones and microslip zones, determined by frictional forces and elastic deformation in the contact. The findings of Hertz [3] in 1882 form the necessary basis for the beginning of research on rolling friction.

The treatment of rolling motions starts with Carter [4] in 1926. He considered the steady-state tractive rolling of an elastic cylinder, which transmits a tractive force at the plane on which it is rolling. Carter presented a solution to this problem in a two-dimensional form (plane strain). He defined the relation between creepage and creepage forces, applied on locomotive wheels, where high tangential forces are transmitted from the wheel to the rail during accelerating and braking the vehicle. He proved that from the moment that a braking or tractive couple is applied to the wheel, creepage occurs. This two-dimensional theory is extended to the three-dimensional case by Johnson [5, 6]. He considered two rolling balls, including the longitudinal and lateral creepages, however without spin creepage. Vermeulen and Johnson [7] subsequently extended this theory to arbitrary smooth half-space bodies.

Kalker [810] developed numerical methods which are able to deal with elastic rolling involving three-dimensional frictional contacts, with imposed creepage and spin, considered constant throughout the contact spot. His universal computer algorithm Contact [8] deals with all contact problems of half-space bodies. This algorithm, which is based on the exact theory of Kalker [10], is computationally intensive, thus not suitable for real-time applications. For this reason, Kalker developed the simplified theory, which is used in his algorithm Fastsim [9]. The reference work of Garg and Dukkipati [11] serves as a good summary of the preceding theoretical work.

Nielsen [12] considered corrugation by abrasive wear and the case of a velocity dependent friction coefficient in the contact of two-dimensional quasi-identical bodies. Two bodies are quasi-identical when they are geometrically and elastically symmetric, which means that their elastic constants are equal, considering the homogeneous isotropic case [10]. Li [13] investigated the wear of rolling contact in railway applications by developing a simulation tool to model the evolution of the contacting surfaces, with the objective to better predict the wear behavior and to optimize the profile geometry of the wheel and rail.

This short literature overview shows that the research on prerolling hysteresis was little in evidence despite its importance. The main scope of the research was up to now mainly situated in the field of locomotive and automotive design. In this paper, the hysteretic frictional behavior in the prerolling period is treated.

The objective of this study is to extend the understanding of the frictional behavior in the prerolling regime by developing a theoretical model for (pre-)rolling friction, partly based on existing theories [10]. In very precise positioning, this period, which occurs after every velocity reversal and extends for a distance on the order of magnitude of the contact patch radius, is of main importance [14]. It is mainly the hysteresis effect in this period that is responsible for the stiffness and damping characteristics of a rolling element guideway in the direction of rolling [15, 16]. As rolling element bearings are widely used in machine guidance, knowledge and theoretical quantification of this hysteretic behavior is important. Moreover, for the treatment of torsional or rolling vibrations in railway wheels or rolling elements, a theory of prerolling may be an important prerequisite. To this end, it is the intent of this paper to extend the simplified theory of rolling contacts, based on the work of Kalker [10], to the case of prerolling.

The paper is structured as follows. Section 2 gives an overview of the contact definition in a nonconforming Hertzian contact, the tractive rolling kinematics, and the rolling theory. A simplified traction-displacement relationship is formulated. Moreover, the initial traction field, formed when the bodies first come into contact, is determined as an initial value to the prerolling problem. Section 3 introduces the solution method developed in the scope of this research topic. Here, the implementation details are introduced and dimensional analysis is applied to the theory in order to yield tractable results. In Section 4, the model results are presented. The steady-state results as well as the transient results are discussed. Afterwards, the evolution of the tractive forces as function of the relative motion (creepage and/or spin) is given. Furthermore, a parameter study is applied to obtain better knowledge of the phenomenon and to make it possible to situate experimental results in a consistent framework. Finally, appropriate conclusions are drawn and future work is indicated.

2. Formulation

2.1. Contact Definition and Preliminary Assumptions

Let us consider the general elastic, nonconformal contact of two bodies of revolution. Depending on their elasticity moduli, the two bodies will deform to certain degrees (see Figure 1(a)). For the purpose of our study, we shall assume that the problem may be reduced to that of a single equivalent elastic body of revolution with a plain infinitely rigid body (see Figure 1(b)). This assumption is widely adopted in the theory and application of Hertzian contacts so that the conversion formulas of geometry and elasticity are well established [17, 18]. However, this assumption excludes certain phenomena such as Heathcote slip. Let us note, firstly, that when the externally imposed creepages result in appreciably larger microslip levels than Heathcote slip, the assumption will yield good approximations. Secondly, it would still be possible to adopt this assumption and account for Heathcote slip by calculating an equivalent creepage field that corresponds to that case.

fig1
Figure 1: Reduction of the elastic contact between two balls to that of one ball and a rigid flat surface.

When this assumption is used, the equivalent modulus of elasticity, 𝐸 (Hertzian modulus), and the equivalent radius of the resulting single body of revolution, 𝑅, are given by 1𝐸=1𝜈21𝐸1+1𝜈22𝐸2,1𝑅=1𝑅1+1𝑅2.(1) (NB. This formula is also applicable to each of the principle radii of an ellipsoid.)

In the rest of this article, we shall consider, without loss of generality, the contact of an elastic sphere with a smooth, rigid plane. The case of an elliptical contact may be dealt with in an analogous manner using appropriate conversion formulas.

2.2. Normal Stresses and Contact Patch

When an elastic sphere of radius 𝑅 and equivalent modulus of elasticity 𝐸𝑒=𝐸/(1𝜈2) is pressed with a load 𝑊 against a rigid plane surface, the following obtaining [17, 19].

The circular contact patch radius 𝑎 is given by𝑎=3𝑊𝑅4𝐸𝑒1/3.(2)The contact patch is defined as the region 𝐴 in the 𝑥𝑦-plane: 𝐴={(𝑥,𝑦)𝑥2+𝑦2𝑎2}.

The normal stress 𝑝𝑧 is given by𝑝𝑧(𝑥,𝑦)=𝑝01(𝑥/𝑎)2(𝑦/𝑎)2(3)with𝑝0=3𝑊2𝜋𝑎2.(4)

2.3. Tractive Rolling Kinematics

Figure 2 gives an overview of the different creepages which can occur in tractive rolling. Here, one considers a sphere of radius 𝑅 rolling in the 𝑥-direction such that its center is translating at a velocity 𝑉. In addition, the sphere is spinning around its 𝑥, 𝑦, and 𝑧 axes with angular speeds ̇𝜓, 𝜔, and ̇𝜙, respectively. The creepages, which characterize tractive rolling, are defined as follows:𝛿𝑉𝑥=𝜔𝑅+𝑉isthelongitudinalcreepage,𝛿𝑉𝑦̇=̇𝜓𝑅isthelateralcreepage,𝜙isthespincreepage.(5)In most cases of interest, the magnitudes of these creepages are proportional to the magnitude of the rolling velocity. In that case, they can be expressed in terms of displacements per traversed rolling distance (as will be apparent further below).

214894.fig.002
Figure 2: Overview of the rolling motion.

The problem of an elastic sphere tractively rolling on a rigid plane can be reduced to that in which the contact patch is stationary in space and time. This is accomplished by assuming the rigid plane upon which the sphere is rolling to be moving in the opposite direction with a velocity 𝐕 (see Figure 2). Looking through the stationary contact patch in a direction normal to it at points in the bodies that are sufficiently remote from the interface, one will then see two surfaces entering it: one moving with velocity 𝐕 and one with velocity 𝐕+𝐜, where𝐜=𝛿𝑉𝑥̇𝜙𝑦,𝛿𝑉𝑦+̇𝜙𝑥(6) is the creepage velocity vector, as depicted in Figure 3.

214894.fig.003
Figure 3: An overview of the relative motion between the two bodies.

In order to arrive at an equation to describe the kinematics of the surface points lying inside the contact patch 𝐴, one follows a point on the surface of the sphere, which enters the contact patch and mates at the entrance with a counterpoint on the rigid plane. Owing to the creepage, the point on the surface of the sphere will have to deform in the plane of the contact patch by an amount 𝐮=(𝑢,𝑣) being called the displacement, which is generally function of space and time. This situation is depicted in Figure 4, which illustrates, in the 2D case, the difference between pure rolling and tractive rolling. Now, defining the slip 𝐬 as the relative velocity between mating points in the interface, and assuming that (𝑢,𝑣) are small as compared to contact patch dimensions, one obtains the following differential equation for the plane deformations [10, 19, 20]: 𝐬=𝐜+𝐷𝐮𝐷𝑡.(7)Writing out the material derivative 𝐷/𝐷𝑡=𝜕/𝜕𝑡+̇𝑥𝜕/𝜕𝑥 and ̇𝑥=𝑉, one obtains𝐬=𝐜𝑉𝜕𝐮+𝜕𝑥𝜕𝐮𝜕𝑡.(8) Generally, 𝑉=𝑉(𝑡). If, as indicated earlier, slip and creepage are scaled with 𝑉, then one can rewrite the previous equation in terms of the traversed rolling distance 𝑞 rather than the time and rolling velocity by making use of the substitution𝑞=𝑡0𝑉(𝜏)𝑑𝜏.(9)Substituting this into (8) yields𝐬=𝐜𝑉𝜕𝐮𝜕𝑥+𝑉𝜕𝐮𝜕𝑞.(10)Finally, normalizing creepage and slip by the rolling speed, one obtains𝐒=𝐂𝜕𝐮+𝜕𝑥𝜕𝐮𝜕𝑞,(11)where 𝐒=𝐬/𝑉=(𝑆𝑥,𝑆𝑦) is the relative slip and 𝐂=𝐜/𝑉=(𝜉𝑥𝜙𝑦/𝑎,𝜉𝑦+𝜙𝑥/𝑎) is the relative creepage.

fig4
Figure 4: The case of rolling without slip (a) and rolling with slip (b).

(In this formulation, the creepages (𝜉𝑥=𝛿𝑉𝑥/𝑉,𝜉𝑦=𝛿𝑉𝑦̇/𝑉,𝜙=𝜙𝑎/𝑉) have units m/m.) Moreover, when the rolling speed is constant, then 𝐂 will be a constant vector. Otherwise, it will be function of the traversed distance. In the rest of the treatment, the form given by (11) will be adopted.

2.4. The Rolling Theory
2.4.1. Traction-Displacement Relationship: Simplified Theory

Following Kalker [10], we consider the simplified “Winkler bedding” model to determine the relationship between surface displacement (𝑢,𝑣) and the tangential traction field (𝑝𝑥,𝑝𝑦) of the rolling object for the case when no slip occurs. Allowance for slip is considered subsequently together with normal traction (Hertzian pressure) treatment.

In this simplified approach, the surface of the elastic object is considered to be covered by elastic “bristles”, normal to it, which have constant stiffness for tangential deformations, 𝐿, that is,𝑝(𝑢,𝑣)=𝐿𝑥,𝑝𝑦,(12)where (𝑝𝑥,𝑝𝑦) is the tangential traction field and 𝐿 is the tangential flexibility of the “bristle”.

The flexibility parameter 𝐿 depends not only on the elasticity characteristics and Poisson's ratio, but also on the aspect ratio of the contact ellipse and the magnitude of the creepages. The determination of an appropriate value for 𝐿, for a particular case, is carried out by formal comparison of analytical, no-slip solutions using the simplified theory with those using exact theory, see Kalker [10]:𝐿=𝐺,𝐶𝑖𝑗,𝜉𝑥,𝜉𝑦𝐿,𝜙,(13)𝐿=1|||𝜉𝑥|||+𝐿2|||𝜉𝑦|||+𝐿3||𝜙||𝜉2𝑥+𝜉2𝑦+𝜙2.(14)It is shown in Kalker [10] that this approximation yields solutions which are very close to those given by exact theory.

2.4.2. The Traction Bound and Slip Conditions

When a tangential displacement field (𝑢,𝑣) is given, (12) will yield the corresponding traction field𝑝𝑥,𝑝𝑦=(𝑢,𝑣)𝐿.(15)This traction field corresponds to the no-slip condition. In practice, the local friction coefficient and normal traction (Hertz pressure) will determine whether no-slip will hold. In other words, (15) will be valid only if𝑝𝑥,𝑝𝑦𝜇𝑝𝑧,(16)where, 𝑝𝑧 is given by (3), (4) and 𝜇 is the local coefficient of friction or adhesion. Generally, 𝜇 will depend on the contact conditions. However, in the present study, we consider only the case of constant 𝜇.

The condition (16) is known as the traction bound, which leads to the following relationship for determining the tangential tractions:𝑝𝑥,𝑝𝑦=(𝑢,𝑣)𝐿,𝑝𝑥,𝑝𝑦𝜇𝑝𝑧,𝜇𝑝𝑧(𝑢,𝑣)(𝑢,𝑣),otherwise.(17)

2.4.3. The Initial Traction Field: Hertzian Contacts with Friction

Before any prerolling is initiated, a traction field is already present in the contact due to the normal loading of the contact. The initial value of the traction field, which is needed to solve the prerolling problem, can be calculated, considering a Hertzian axisymmetric contact between a ball and a flat surface. Here, we consider the normal loading of the ball, including the frictional behavior between the ball and its contacting surface. The shear traction field is given by Hills et al. [19]:𝑝𝑥𝑦(𝑟)𝜇𝑝0=1𝑟2𝑟𝑟𝐻𝑠𝑟𝑟𝑠𝑟Ψ𝑡,𝑟𝑠𝑡21𝑡2𝑑𝑡,(18)where 0<𝑟<1 is the dimensionless radial coordinate, 𝜇 the local coefficient of friction, 𝑝0 the peak contact pressure, 𝐻() the Heaviside step function, 𝑟𝑠 denotes the dimensionless radius of the stick region, and Ψ(𝑟,𝑟𝑠) is given by the following integral expression:Ψ𝑟,𝑟𝑠=0,if𝑟𝑠2𝑟1,𝜋(𝑤𝑟0𝜋/2𝑑𝜃1(1𝑟2)sin2𝜃)1𝑟2𝑠sin2𝜃)𝑟𝑠𝐾𝑟𝑠𝑟𝑤𝑟𝑠𝑄0(𝑤)𝑄0(𝑟𝑠)if0𝑟𝑠,(19)where 𝐾(𝑟𝑠) is a complete elliptic integral of the first kind, 𝑟𝑠2=1𝑟2𝑠,𝑄01(𝑟)=2ln1+𝑟1𝑟,𝑤2=𝑟2𝑠𝑟21𝑟2.(20)In the case of elastic similar bodies, the entire contact zone sticks, thus 𝑟𝑠=1. This adhesive limit was studied by Goodman [21] and can be expressed in closed form by𝑝𝑥𝑦(𝑟)𝛽𝑝0=𝑟𝜋𝑟0ln𝑥𝑑𝑥1𝑥22,0<𝑟<1,(21)with 𝛽 is the Dundurs' constant [22, 23] given by𝛽=12𝜈1/2𝐺112𝜈2/2𝐺21𝜈1)/𝐺1+(1𝜈2)/𝐺2.(22)Figure 5 shows the radial shear traction distribution for different values of 𝑟𝑠. This field serves as an initial value for the prerolling model. Let us note however that once gross rolling is attained, the traction field corresponding to it will serve as the initial value for subsequent rolling.

214894.fig.005
Figure 5: The radial shear traction distribution for different partial slip cases.

3. Solution Procedure

Referring to the previous section, the problem to be solved may be stated as follows: solve (11) subject to conditions (12), (15), (16) to determine the tangential displacements and tractions. First, we show that (11) admits a closed form general solution for the case of zero slip. Secondly, we apply the stiffness and slip conditions (12), (15), (16) in a numerical implementation to determine the resulting traction field. Finally, the traction forces are determined.

3.1. Analytical, Zero-Slip Solution

First, we write the two members of (11) in the form𝜕𝑢+𝜕𝑥𝜕𝑢𝜕𝑞=𝑆𝑥𝜉𝑥+𝜙𝑦𝑎,(23)𝜕𝑣+𝜕𝑥𝜕𝑣𝜕𝑞=𝑆𝑦𝜉𝑦𝜙𝑥𝑎.(24)If the slip (𝑆𝑥,𝑆𝑦) is given, (23), (24) admit general solutions which are derived in detail in the appendix.

Since the slip field is not known beforehand, the solutions of interest for numerical implementation are those corresponding to the case of zero-slip. Thus, for (𝑆𝑥,𝑆𝑦)0, we have the following solutions (see the appendix):𝑢(𝑥,𝑦,𝑞)=𝑞𝜉𝑥+𝜙𝑦𝑎+𝑓(𝑥+𝑞,𝑦).(25)Similarly, the general solution for 𝑣 is𝑣(𝑥,𝑦,𝑞)=𝑞𝜉𝑦𝜙𝑎𝑞𝑥+2+𝑔(𝑥+𝑞,𝑦),(26)where 𝑓,𝑔 are any arbitrary functions. Note that outside the contact patch 𝐴, 𝑢𝑣0.

Let us remark first that the problem is not coupled in (𝑥,𝑦), that is, 𝑦 can be treated as a parameter in (25) and (26). Consequently, we can solve the problem on any line 𝑦=𝑦0 in the contact patch. Thus, referring to Figure 6, we consider the solution on the line segment 𝑦=𝑦0 which is bounded by the leading edge 𝑥=𝑎(𝑦0) and the trailing edge 𝑥=𝑎(𝑦0). Since the surface points enter the contact patch free of stress, we have𝑦𝑢=𝑣=0𝑞at𝑥=𝑎0,(27)which provides the necessary boundary condition for the solution.

214894.fig.006
Figure 6: The case of rolling without slip (left) and rolling with slip (right).

The initial surface displacement distribution at 𝑞=0 is assumed to be given, for example, from Section 2.4.3, as𝑞=0,𝑢=𝑢0𝑥,𝑦0,0,𝑣=𝑣0𝑥,𝑦0.(28)Substituting this into (25) and (26), we obtain𝑓(𝑥,𝑦0)=𝑢0𝑥,𝑦0,𝑔(𝑥,𝑦0)=𝑣0𝑥,𝑦0.(29)Equation (29) means that the initial (𝑢,𝑣) distribution determines the arbitrary functions 𝑓 and 𝑔 on the range of definition of (𝑢,𝑣). Thus, by replacing 𝑥 by 𝑥+𝑞, on 𝑎𝑥+𝑞𝑎, we have that𝑓𝑥+𝑞,𝑦0=𝑢0𝑥+𝑞,𝑦0,𝑔𝑥+𝑞,𝑦0=𝑣0𝑥+𝑞,𝑦0,𝑦𝑎0𝑦𝑥𝑎0𝑞.(30) In order to obtain the form of functions 𝑓 and 𝑔 in the interval 𝑎(𝑦0)𝑞𝑥𝑎(𝑦0), we make use of the initial condition at 𝑥=𝑎(𝑦0). Thus, substituting (27) into (25) and (26) yields𝑢𝑎𝑦0,𝑦0,𝑞=0=𝑞𝜉𝑥+𝜙𝑦0𝑎𝑎𝑦+𝑓0+𝑞,𝑦0,𝑣𝑎𝑦0,𝑦0,𝑞=0=𝑞𝜉𝑦𝜙𝑎𝑎𝑦0+𝑞2𝑎𝑦+𝑔0+𝑞,𝑦0.(31)Replacing 𝑞 by 𝑥+𝑞, on 𝑎(𝑦0)𝑞𝑥𝑎(𝑦0), and translating 𝑥 by 𝑎(𝑦0), we have that𝑓𝑥+𝑞,𝑦0𝑦=𝑥+𝑞𝑎0𝜉𝑥+𝜙𝑦0𝑎,𝑔𝑥+𝑞,𝑦0𝑦=𝑥+𝑞𝑎0𝜉𝑦𝜙𝑦2𝑎𝑥+𝑞+𝑎0,𝑎𝑦0𝑦𝑞𝑥𝑎0.(32)Finally, the analytic solution is obtained in closed form by substituting of 𝑓 and 𝑔 from (30) and (32) into (25) and (26):𝑢𝑥,𝑦0,𝑞=𝑞𝜉𝑥+𝜙𝑦𝑎+𝑦𝑥+𝑞𝑎0𝜉𝑥+𝜙𝑦0𝑎𝑦,𝑎0𝑦𝑞𝑥𝑎0,𝑢0𝑥+𝑞,𝑦0𝑦,𝑎0𝑦𝑥𝑎0𝑣𝑞,𝑥,𝑦0,𝑞=𝑞𝜉𝑦𝜙𝑎𝑞𝑥+2+𝑦𝑥+𝑞𝑎0𝜉𝑦𝜙𝑦2𝑎𝑥+𝑞+𝑎0,𝑎𝑦0𝑦𝑞𝑥𝑎0,𝑣0𝑥+𝑞,𝑦0𝑦,𝑎0𝑦𝑥𝑎0𝑞.(33)This solution shows that the initial displacement distribution passes as a wave, moving from right to left, through the contact patch. Gross rolling is thus achieved after a rolling distance equal to 2𝑎 is traversed. The no-slip displacement field corresponding to it is obtained from the top members of (33) as𝑢𝑥,𝑦0𝑦=𝑥𝑎0𝜉𝑥+𝜙𝑦0𝑎,𝑣𝑥,𝑦0𝑦=𝑥𝑎0𝜉𝑦𝜙𝑦2𝑎𝑥+𝑎0,(34)which is evidently independent of the initial distribution.

3.2. Numerical Implementation

In order to apply the procedure numerically to a given contact problem, the contact region is discretized into a set of nodes. In Figure 7, the grid is depicted, where we have chosen for programming convenience Δ𝑥=Δ𝑦. Starting from any initial distribution 𝑢0(𝑥,𝑦0),𝑣0(𝑥,𝑦0), (33) yields the zero-slip values at any desired value of 𝑞. In the particular case when the increments Δ𝑞=Δ𝑥, the analytical solution reduces to the following simple step-wise procedure. After a step Δ𝑞, the new (𝑢,𝑣) distribution is obtained by augmenting the previous value with Δ𝑞(𝜉𝑥+𝜙𝑦/𝑎,𝜉𝑦𝜙𝑥/𝑎), shifting each of the 𝑢 and 𝑣 vectors one position to the left, and padding them with the boundary condition 𝑢=𝑣=0 at the right.

214894.fig.007
Figure 7: The discretization of the contact patch.

Next, the traction bound needs to be verified at each step and the values of (𝑢,𝑣) accordingly corrected. This is carried out as follows.

With any new obtained value (𝑢,𝑣), we calculate𝑝𝑥,𝑝𝑦=𝐿(𝑢,𝑣).(35)If (𝑝𝑥,𝑝𝑦)𝜇𝑝𝑧, no slip occurs (𝑆𝑥=𝑆𝑦=0) and the values of (𝑢,𝑣) are retained. If, on the other hand, (𝑝𝑥,𝑝𝑦)>𝜇𝑝𝑧, then(𝑢,𝑣)new=𝜇𝑝𝑧𝐿(𝑢,𝑣)(𝑢,𝑣)old.(36) In Figure 8, the flowchart of the algorithm is presented.

214894.fig.008
Figure 8: The flowchart of the algorithm.
3.3. Calculation of the Tractions 𝐹𝑥,𝐹𝑦,𝑀𝑧

The total traction forces and moment are obtained from the following integrals:𝐹𝑥=𝑎𝑎𝑎(𝑦)𝑎(𝑦)𝑝𝑥𝐹(𝑥,𝑦)𝑑𝑥𝑑𝑦,(37)𝑦=𝑎𝑎𝑎(𝑦)𝑎(𝑦)𝑝𝑦𝑀(𝑥,𝑦)𝑑𝑥𝑑𝑦,(38)𝑧=𝑎𝑎𝑎(𝑦)𝑎(𝑦)𝑝(𝑥,𝑦)𝑟(𝑥,𝑦)𝑑𝑥𝑑𝑦(39)with 𝑟(𝑥,𝑦) the radial distance from the center of the contact patch to the location of the node.

For the no-slip, gross rolling solution, (37) and (38) yield (see also [10])𝐹𝑥=8𝑎3𝜉𝑥,𝐹3𝐿(40)𝑦=8𝑎3𝜉𝑦3𝐿𝜋𝑎3𝜙.4𝐿(41)Upon comparing these results formally with the results obtained from exact theory, the relations (12) and (13) are obtained.

3.4. Dimensional analysis

In order to generate, analyze, and present the results systematically, dimensional analysis is applied to the problem. This is achieved through the application of the Vaschy-Buckingham-Π theorem [24, 25] as follows. The rolling problem may be generally expressed as𝑓𝑊,𝐹,𝑀,𝑎,𝑝0,𝑝,𝐸,𝑅,𝑉,𝛿𝑉𝑥,𝛿𝑉𝑦,̇𝜙,𝜈,𝜇,𝑞,𝑥,𝑦=0.(42)Inspection shows that the problem posses three independent dimensions (length, force, and time). Choosing 𝑎, 𝑝0, and 𝑉 as the variables to be eliminated, we obtain the dimensionless form, in which the number of variables is reduced by three:𝑊𝑝0𝑎2,𝐹𝑝0𝑎2,𝑀𝑝0𝑎3,𝑎𝑎,𝑝0𝑝0,𝑝𝑝0,𝐸𝑝0,𝑅𝑎,𝑉𝑉,𝛿𝑉𝑥𝑉,𝛿𝑉𝑦𝑉,̇𝜙𝑎𝑉𝑞,𝜈,𝜇,𝑎,𝑥𝑎,𝑦𝑎𝑊=0,,𝐹,𝑀,𝑝,𝐸,𝑅,𝜉𝑥,𝜉𝑦,𝜙,𝜈,𝜇,𝑞,𝑥,𝑦=0.(43)Another possibility is to also eliminate the (dimensionless) friction coefficient 𝜇 since it only scales the traction forces.

We refer to the nomenclature for an overview of the used symbols. Since the model contains no system dynamics of the ball, the rolling velocity 𝑉 falls out of the equations. This dimensional analysis makes it easier to compare the influence of the different parameters of the problem in a consistent way. Therefore, in the next section, the results are also given in this dimensionless form. An arbitrary value of 0.5 has been used for the coefficient of sliding friction.

4. Results

4.1. Steady-State Rolling

Figure 9 depicts traction distributions pertaining to basic cases of steady-state rolling with longitudinal, lateral, and/or spin creepages of an elastic ball. While rolling with longitudinal and lateral creepage is of main interest in wheel-rail or tyre-road contact, rolling with spin is the dominant type in angular contact ball bearings and linear guideways (e.g., with V-grooved tracks). For the sake of better visibility, a coarse grid is used. Figure 9(a) gives the example of equal longitudinal and lateral creepage. The slip zone is symmetric w.r.t. the 𝑥-axis while the traction vectors are oriented at 45 degrees to it. Figures 9(b) and 9(c) represent the traction fields with spin creepage of different levels. With increasing spin, the slip zone becomes larger, until the total contact zone slips. Figure 9(d) shows how the traction field looks like when combining spin with creepage in both directions.

fig9
Figure 9: A few results of the steady-state rolling traction field, the rolling direction is from the left to the right.
4.2. Transient Rolling

In this section, the traction behavior during transition to rolling is analyzed. This behavior is marked by hysteresis of the traction forces in the rolling displacement. Again, we limit the presentation to the basic cases of (pre-)rolling with longitudinal, lateral, and spin creepage. The applied displacement trajectory is provided with reversal points, to ascertain the nonlocal memory character of the hysteresis curves, as discussed in [1, 26, 27]. The rolling trajectory is specified as 𝑞=[0𝑞1𝑞2]. For all cases considered, the initial traction field is a null field since the contacting materials are identical, which results in a Dundurs' constant 𝛽=0.

4.2.1. Pure Longitudinal Creepage

For this case, we chose 𝜉𝑥=0.015,𝜉𝑦=𝜙=0. The rolling trajectory is 𝑞=[02.12.12.1]. The longitudinal creepage gives rise only to a traction force 𝐹𝑥 in the rolling direction. Figure 10(c) plots 𝐹𝑥 against 𝑞 to show the resulting hysteresis loop. Figures 10(a), 10(b), 10(c), 10(d), and 10(e) show the characteristic traction fields at selected points during the motion. From these, we see also that all traction components lie in the rolling direction.

fig10
Figure 10: The results for transient (pre-)rolling with longitudinal creepage. In the middle, the hysteresis with around it, transient traction fields at selected points in the trajectory.
4.2.2. Pure Lateral Creepage

To illustrate this case, we put 𝜉𝑥=0,𝜉𝑦=0.015,𝜙=0. Rolling combined with a lateral creepage component gives rise only to a traction field in the 𝑦-direction, that is, perpendicular to the rolling direction. Consequently, the hysteresis loops in Figure 11(a) depict this traction force 𝐹𝑦 versus the rolling displacement (𝑞=[02.10.41.62.10.41.62.1]). This trajectory has been so constructed as to show inner hysteresis loops and thus the nonlocal memory character of hysteresis. The behavior is, otherwise, similar to that of pure longitudinal creepage. Figure 11(b) depicts the steady-state traction field.

fig11
Figure 11: The results for transient (pre-)rolling with lateral creepage. The left panel depicts the hysteresis curve while the right panel visualizes the steady-state rolling traction field.
4.2.3. Pure Spin

Here, we put 𝑣𝑥=𝑣𝑦=0,𝜙=0.0255. Rolling with pure spin results in a traction field having components in the 𝑥 and 𝑦 directions. The resulting traction force in the 𝑥 direction equals zero. The traction stress field results in a spin moment 𝑀𝑧. This moment is plotted as function of the rolling displacement to yield the hysteresis curve of Figure 12(c). The input is 𝑞=[02.040.461.542.040.461.542.04]. Note that the virgin curve, corresponding to start of motion until first gross slip, overshoots the subsequent hysteresis curves. The shape of the virgin curve and the amount of overshoot vary with the assumed initial traction distribution.

fig12
Figure 12: The results for transient (pre-)rolling with spin. In the middle, the hysteresis with around it, some transient traction fields at selected points in the trajectory.
4.3. Evolution of the Steady-State Tractions in Function of the Creepages

It is intuitively plausible to assume that the traction force (or moment in the case of spin) will increase with increasing creepage until gross slip is reached. In Figure 13, the evolution of the steady-state rolling traction force is depicted as a function of the relative creepage. For rolling with pure longitudinal creepage, an increasing trend is observed as shown in Figure 13(a). The same evolution is noticed for rolling with lateral creepage in Figure 13(b). Considering rolling with pure spin, the spinning moment 𝑀𝑧 increases towards a saturation value while the lateral force 𝐹𝑦 shows a local maximum. These trends are depicted in Figure 13(c). As already mentioned, the longitudinal force component 𝐹𝑥 equals zero for this case. These results agree with those obtained in [10].

fig13
Figure 13: The evolution of the steady-state traction as function of the creepage.
4.4. Parameter Analysis

In the previous section, the general behavior of the steady-state traction force in function of the creepage value is discussed. In this section, the influence of the other model parameters is discussed. For brevity, we confine the treatment to the case of pure spin creepage.

Consider the dimensionless model of (43). Because the model parameters are reduced by normalizing with respect to the radius of the contact patch 𝑎 and the maximum normal pressure 𝑝0, these parameters fall out of the model equations, so that (2) becomes1=3𝑊𝑅4𝐸1/3𝑅=4𝐸3𝑊.(44)In other words, 𝑅 increases if 𝐸 increases or 𝑊 decreases. In this way, 𝑅 contains (or coalesces) all geometry, elasticity, and load information.

To study the influence of 𝑅, this parameter is varied between reasonable bounds. To determine its influence on the prerolling behavior, three main parameters are chosen for analysis.

(1) The prerolling distance 𝑥pr.(2) The initial stiffness 𝑘𝑖 of the hysteresis system, that is, the initial slope of the hysteresis curve,𝑘𝑖=𝜕𝑀𝑧𝜕𝑞|||𝑞=0.(45)(3) The steady-state frictional moment 𝑀𝑧,ss.

These variables are depicted as functions of the dimensionless ball radius in Figure 14 for different values of the spin creepage 𝜙 and in Figure 15 for different values of the coefficient of friction 𝜇. One can notice from (44) that an increase in 𝑅 corresponds to a decrease in 𝑊 for a constant value of 𝐸. A certain limit of applicability, 𝑅lim applies, owing to geometrical limitations: 𝑎 cannot be larger than about 𝑅/4 without violating basic Hertzian assumptions. The plotted values in Figures 14 and 15 below this value of 𝑅lim (marked by the box “theoretical region”) are only given for the sake of mathematical completeness.

fig14
Figure 14: The behavior of the prerolling region on a variation of the ball radius for different spin creepages (𝜇 = 0.5).
fig15
Figure 15: The behavior of the prerolling region on a variation of the ball radius for different coefficients of friction 𝜇(𝜙=0.05).

From Figures 14(a) and 15(a), one can see that the prerolling distance stays constant for low and moderate values of 𝑅 (i.e., for highly loaded contacts), while for higher values (lightly loaded case), the prerolling distance decreases. Thus, the maximum possible value for the prerolling distance is twice the contact patch radius. The initial stiffness of the hysteresis at the beginning of the prerolling region is depicted in Figures 14(b) and 15(b). For low values of 𝑅, the stiffness is constant; for medium values, a maximal stiffness is observed, while for higher values of 𝑅, the stiffness decreases with 𝑅. In the fully elastic region, we have from the Hertz theory𝑎(𝑊𝑅)(1/3).(46)From Hills et al. [19], we have𝑀𝑞𝑊𝑅𝜃𝑅,𝑀(47)𝑞𝑊𝑅𝑅𝑎3𝑅,(48) since𝑀=𝑀𝑝0𝑎3.(49)This leads to𝑀𝑞𝑅𝑞𝑎.𝑎𝑅,𝑘(50)𝑖=𝜕𝑀𝜕𝑞1𝑅𝑞since𝑎𝑞,𝑅𝑎=𝑅.(51)The steady-state tractive moment is depicted in Figures 14(c) and 15(c). After a constant behavior for low values of 𝑅, a steep decrease is noticed to end with a quasizero value for very high values of 𝑅. To get a value for the coefficient of rolling friction 𝜆, this steady-state moment is divided by the dimensionless normal load 𝑊. These results are depicted in Figures 14(d) and 15(d). We notice a steep increase in the region of moderate values for 𝑅 to saturate towards a constant value for higher 𝑅, which results in a similar behavior as compared to “Amontons' law” [28] for sliding friction.

5. Discussion and Conclusions

In the foregoing, a theory is developed to characterize the traction behavior during the transition to rolling. The following remarks are in order. Firstly, although based on the “Winkler bedding” simplification, it is shown in [10] that this approximation, with appropriate choice of the stiffness parameter 𝐿, yields solutions which are very close to those given by exact theory. The advantage gained is the transparency and easy application of this theory. Secondly, although only the case of point contact has been treated, this theory can be directly extended to the general cases of elliptical and line contacts. Thirdly, the cases of variable (e.g., pressure dependent) local coefficient of friction in the contact patch, variable (e.g., time or position dependent) creepages, variable normal load, and variable rolling velocity can all be directly treated by this theory. In that way, dynamical contact phenomena, such as those obtaining during motion reversals, acceleration, deceleration, oscillation, and so forth, can be accommodated by this theory. Other cases not covered immediately by this theory are discussed subsequently.

Rough Contacts
The replacement of a smooth contact by a rough one generally requires a higher computational effort, which often mortgages the development of a simplified version suitable for real-time execution. Bucher et al. [29], for instance, outline that in the case of rough contact, the stresses and deformations can only be calculated using special boundary element methods, while transient three-dimensional rolling contacts comprising rough surfaces are not possible at present. However, this “bristle” model offers the advantage that it can be directly extended to rough contacts. The challenge is to translate an actual rough surface into an equivalent bristle set. Alternatively, recent numerical methods offer an affordable solution for calculating the rough surface contact by using dedicated contact solvers [30, 31].

Wear and Heathcote Slip
The influence of wear and running-in of the surfaces is obviously very important in many applications. In the former algorithm, the creepage was kept constant over the contact patch. Taking wear in consideration, the need for a variable creepage becomes important. When the flat surface in which the sphere is rolling, wears, a groove develops in which the contact between both bodies, is conformal. This situation is similar to that of a ball in a groove, for example, a deep groove ball bearing, where the contact is conformal. In that case, the traction field for pure rolling, that is, with zero relative creepage motion in a conformal groove, is described by Heathcote [32] and is qualitatively depicted in Figure 16. The contact area is no longer plane. When rolling freely, there is no net tangential force, which explains that the contact spot is subdivided into three zones: the central zone contains positive slip vectors, while in the outer zones, the slip is negative.

214894.fig.016
Figure 16: The traction distribution of pure rolling: Heathcote slip.

In order to be able to deal with this problem using the algorithm presented in this paper, the hitherto constant creepage over the whole contact patch should be replaced by a variable creepage field. The latter may be obtained from geometrical and kinematical considerations of the contact with the border line between positive and negative creepage fields being the only unknown.

In conclusion, this paper considers the hysteretic behavior of the prerolling friction between a ball and a flat surface. Extending the existing steady-state gross rolling model, due to J. J. Kalker, to a transient prerolling model, the evolution of the traction field in the rolling displacement is determined and hysteresis curves are generated. The traction field and its behavior in the presence of rolling with creepage and spin are systematically investigated. Finally, a parameter study is carried out to gain more insight into this phenomenon. In a forthcoming paper, an experimental validation of the model is carried out for the case of rolling with spin creepage, using a configuration consisting of two V-grooved tracks with 2 balls in between.

Notes

1The authors dedicate this paper to the memory of Prof. J. J. Kalker (1933–2006).

Appendix

Consider the p.d.e.𝜕𝑢+𝜕𝑥𝜕𝑢𝜕𝑞=(𝑥,𝑦,𝑞).(A.1) Equation A.1 corresponds to the Lagrangrian system [33]:𝑑𝑥=1𝑑𝑦0=𝑑𝑞1=𝑑𝑢(𝑥,𝑦,𝑞).(A.2) The integrals of the system A.2 are𝑥+𝑞=𝛼,𝑦=𝛽,𝑢𝒢(𝑥,𝑦,𝑞)=𝛾,(A.3)where 𝛼,𝛽,𝛾 are arbitrary constants and 𝒢 is determined by==𝒢(𝑥,𝑦,𝑞)=(𝑥,𝑦,𝑞)𝑑𝑞(𝛼𝑞,𝛽,𝑞)𝑑𝑞𝑑𝑔𝑑𝑞(𝛼,𝛽,𝑞)𝑑𝑞=𝑔(𝛼,𝛽,𝑞)=𝑔(𝑥+𝑞,𝑦,𝑞).(A.4) Note that the last integration step is carried out on the assumption that the function (𝛼𝑞,𝛽,𝑞) posses an antiderivative 𝑔. This is always true if can be expressed as a polynomial in 𝑞.

The general solution to A.1 is then [33](𝛼,𝛽,𝛾)=0(A.5)or𝑥+𝑞,𝑦,𝑢𝒢(𝑥,𝑦,𝑞)=0,(A.6)where is an arbitrary function. Assuming that (A.6) admits an explicit solution for 𝑢, this solution will have the form𝑢=𝒢(𝑥,𝑦,𝑞)𝑓(𝑥+𝑞,𝑦),(A.7)where 𝑓 is an arbitrary function. As an example, considering (𝑥,𝑦,𝑞)=𝜉𝑥+𝜙𝑦/𝑎 and assuming 𝜉𝑥 and 𝜙 to be constant, then𝑢(𝑥,𝑦,𝑞)=𝑞𝜉𝑥+𝜙𝑦𝑎+𝑓(𝑥+𝑞,𝑦).(A.8)

Nomenclature
𝑎:Footprint radius
𝐴:Contact patch
𝐜:creepage
𝐂:Relative creepage
𝐸:Hertzian modulus of elasticity
𝐹:Traction force
𝑔:Traction bound
𝐺:Modulus of rigidity
𝐻()Heaviside step function
𝑘𝑖:Initial stiffness
𝐿:Flexibility
𝑀𝑧:Spin moment
𝑝:Pressure
𝑞:Traversed rolling distance
𝑟:Distance from the center
𝑅:Contact radius
𝑅:Equivalent radius of contact
𝐬:Slip
𝐒:Relative slip
𝑡:Time
𝑢,𝑣:Particle displacement
𝑉:Rolling velocity
𝑊:Normal load
𝑥:Displacement
𝑥pr:Prerolling distance
𝛽:Dundurs' constant
𝛿:Normal elastic deformation
𝛿𝑉𝑥:Dimensional longitudinal creepage
𝛿𝑉𝑦:Dimensional lateral creepage
𝜆:Coefficient of rolling friction
𝜇:Coefficient of friction
𝜈:Poisson's coefficient
𝜉𝑥:Nondimensional longitudinal creepage
𝜉𝑦:Nondimensional lateral creepage
̇𝜙:Dimensional spin creepage
𝜙:Nondimensional spin creepage
𝜓:Angular lateral creepage
𝜔:Angular longitudinal creepage

Acknowledgements

This research is sponsored by the Fund for Scientific Research, Flanders (F.W.O.), under Grant no. FWO4283. The scientific responsibility is assumed by its authors.

References

  1. F. Al-Bender and W. Symens, “Characterisation of frictional hysteresis in ball-bearing guideways,” Wear, vol. 258, no. 11-12, pp. 1630–1642, 2005. View at Publisher · View at Google Scholar
  2. O. Reynolds, “On rolling friction,” Philosophical Transactions of the Royal Society of London, vol. 166, pp. 155–174, 1876. View at Publisher · View at Google Scholar
  3. H. Hertz, “Über die Berührung fester elastischer Körper,” Journal für die reine und angewandte Mathematik, vol. 92, pp. 156–171, 1882. View at Google Scholar
  4. F. W. Carter, “On the action of a locomotive driving wheel,” in Proceedings of the Royal Society of London, vol. 112, pp. 151–157, 1926.
  5. K. L. Johnson, “The effect of spin upon the rolling motion of an elastic sphere upon a plane,” Journal of Applied Mechanics, vol. 25, pp. 332–338, 1958. View at Google Scholar
  6. K. L. Johnson, “The effect of a tangential force upon the rolling motion of an elastic sphere upon a plane,” Journal of Applied Mechanics, vol. 25, pp. 339–346, 1958. View at Google Scholar
  7. P. J. Vermeulen and K. L. Johnson, “Contact of non-spherical bodies transmitting tangential forces,” Journal of Applied Mechanics, vol. 31, pp. 338–340, 1964. View at Google Scholar
  8. J. J. Kalker, On the rolling contact of two elastic bodies in the presence of dry friction, Ph.D. thesis, Delft University of Technology, Delft, The Netherlands, 1967.
  9. J. J. Kalker, “A fast algorithm for the simplified theory of rolling contact,” Vehicle System Dynamics, vol. 11, no. 1, pp. 1–13, 1982. View at Publisher · View at Google Scholar
  10. J. J. Kalker, Three-Dimensional Elastic Bodies in Rolling Contact, Kluwer Academic Publishers, Dordrecht, The Netherlands, 1990.
  11. V. K. Garg and R. V. Dukkipati, Dynamics of Railway Vehicle Systems, Academic Press, Waterloo, Ontario, Canada, 1984.
  12. J. B. Nielsen, New developments in the theory of wheel/rail contact mechanics, Ph.D. thesis, Technical University of Denmark, Lyngby, Denmark, 1998.
  13. Z. Li., Wheel-rail rolling contact and its applications to wear simulation, Ph.D. thesis, Delft University of Technology, Delft, The Netherlands, 2002.
  14. T. Prajogo, Experimental study of pre-rolling friction for motion-reversal error compensation on machine tool drive systems, Ph.D. thesis, Department of Werktuigkunde, Katholieke Universiteit Leuven, Leuven, Belgium, 1999.
  15. F. Al-Bender and W. Symens, “Dynamic characterisation of hysteresis elements in mechanical systems. I. Theoretical analysis,” Chaos, vol. 15, no. 1, Article ID 013105, p. 11 pages, 2005. View at Publisher · View at Google Scholar · View at PubMed
  16. F. Al-Bender and W. Symens, “Dynamic characterisation of hysteresis elements in mechanical systems. II. Experimental validation,” Chaos, vol. 15, no. 1, Article ID 013106, p. 9 pages, 2005. View at Publisher · View at Google Scholar · View at PubMed
  17. R. D. Arnell, P. B. Davies, J. Halling, and T. L. Whomes, Tribology Principles and Design Applications, MacMillan, New York, NY, USA, 1991.
  18. J. Halling, Principles of Tribology, MacMillan, New York, NY, USA, 1979.
  19. D. A. Hills, D. Nowell, and A. Sackfield, Mechanics of Elastic Contacts, Butterworth-Heinemann, Oxford, UK, 1993.
  20. K. L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, UK, 1985.
  21. L. E. Goodman, “Contact stress analysis of normally loaded rough spheres,” Journal of Applied Mechanics, vol. 29, no. 3, pp. 515–522, 1962. View at Google Scholar
  22. J. Dundurs and M. S. Lee, “Stress concentration at a sharp edge in contact problems,” Journal of Elasticity, vol. 2, no. 2, pp. 109–112, 1972. View at Publisher · View at Google Scholar
  23. D. A. Spence, “The hertz contact problem with finite friction,” Journal of Elasticity, vol. 5, no. 3-4, pp. 297–319, 1975. View at Publisher · View at Google Scholar
  24. E. Buckingham, “On physically similar systems; illustrations of the use of dimensional equations,” Physical Review, vol. 4, no. 4, pp. 345–376, 1914. View at Publisher · View at Google Scholar
  25. M. D. Hersey, Theory and Research in Lubrication: Foundations for Future Developments, John Wiley & Sons, New York, NY, USA, 1966.
  26. W. Symens, Motion and vibration control of mechatronic systems with variable configuration and local non-linearities, Ph.D. thesis, Katholieke Universiteit Leuven, Division of Production Engineering, Machine design and Automation, Leuven, Belgium, 2004.
  27. I. Mayergoyz, Mathematical Models of Hysteresis, Springer, New York, NY, USA, 1991.
  28. G. Amontons, “On the resistance originating in machines,” in Proceedings of the French Royal Academy of Sciences, pp. 206–222, 1699.
  29. F. Bucher, A. I. Dmitriev, M. Ertz et al., “Multiscale simulation of dry friction in wheel/rail contact,” Wear, vol. 261, no. 7-8, pp. 874–884, 2006. View at Publisher · View at Google Scholar
  30. I. A. Polonsky and L. M. Keer, “Fast methods for solving rough contact problems: a comparative study,” Journal of Tribology, vol. 122, no. 1, pp. 36–41, 2000. View at Publisher · View at Google Scholar
  31. I. A. Polonsky and L. M. Keer, “A numerical method for solving rough contact problems based on the multi-level multi-summation and conjugate gradient techniques,” Wear, vol. 231, no. 2, pp. 206–219, 1999. View at Publisher · View at Google Scholar
  32. H. L. Heathcote, “The ball bearing: in the making, under test and on service,” Proceedings of the Institute of Automotive Engineers, vol. 15, pp. 569–702, 1921. View at Google Scholar
  33. A. R. Forsyth, Theory of Differential Equations Part IV: Partial Differential Equations, vol. 5, University Press, Cambridge, UK, 1906.