Você está na página 1de 31

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling

Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

An Analytical Model of Four-Point Contact


Rolling Element Ball Bearings
Jacob D. Halpin1 and Anh N. Tran2

Abstract The purpose of this work is to establish an analytical model and standard way to predict the
performance characteristics of a four-point contact, or gothic arch type, rolling element ball bearing.
Classical rolling element bearing theory, as developed by A. B. Jones, has been extended to include the
unique aspects of the four-point contact bearing; thereby providing complete element-wise attitude and
internal load distributions of the bearing under operating load conditions. Independent geometry
definition for the inner and outer raceways, including their associated arching, is provided to allow
assessment of a broad range of manufacturing tolerance conditions. Standard performance parameters,
such as element contact stresses, contact angles, inner ring deflections, nonlinear stiffnesss, torque, and
L10 life are solved explicitly via standard Newton-Raphson techniques. Race control theory is replaced
with a minimum energy state theory that allows both spin and slip to occur at each contact.
The developed four-point model was programmed within the ORBIS software program. Various test
cases are analyzed and key analytical results are compared with the A. B. Jones Four Point Contact Ball
Bearing Analysis Program, the DG03 design guide and traditional two-point (angular contact) analysis
codes. Model results for the internal distribution of ball loads and contact angles match the Jones
program extremely well for all cases considered. Some differences are found with the DG03 analysis
methods and it is found that modelling a four-point contact bearing with two opposed angular contact
bearings positioned to a common center can result in gross errors.

Keywordsbearing analysis, four-point contact, gothic arch bearing, double arched bearing, ball bearings, split ring
bearing, rolling element bearings

1
J. D. Halpin is the owner of Halpin Engineering, LLC, located in Torrance, CA 90504 USA (phone: 310-650-8982; e-mail:
jake@HalpinEngineeringLLC.com).
2
Tran, A. N. is owner of ATEC Corporation, located in Cypress, CA 90630 USA (email: tran.atec@gmail.com)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

NOMENCLATURE

Symbol

1,2,2,2,3,4

G, H

M, N

Description
Ball spin axis angle
Axial distance from outer-right raceway curvature center to inner-left raceway curvature
center
Radial distance from outer-right raceway curvature center to inner-left raceway
curvature center
Contact angle
+ 1
L10 fatigue capacity of a single point contact
Simplifying expressions, see equations (46) through (51)
Ball diameter
Pitch diameter
Normal approach ball to raceway contact
Youngs modulus of elasticity
Curvature ratio of raceway
Elastic reaction force of bearing on shaft
Arching dimension
Simplifying expression, see equations (37) and (38)
Generic functions, see equations (34) and (35)
Stiffness, Hertzian contact
Generic functions, see equations (90) through (101)
Azimuth angle (ball station clocking angle)
Radial play of non-arched bearing
Enforced displacement of inner ring at its center
Normal ball load
Raceway radius
Radius from bearing spin axis to inner or outer raceway curvature center at initial free
contact
Radial play of an arched bearing
Axial play of an arched bearing
Radial height of ball center relative to outer raceway curvature centers
Axial distance to ball center from outer right raceway curvature center
Number of rolling elements (balls) in a bearing row
SUBSCRIPTS

1,2,3,4,5
rest
free

Inner raceway
Outer raceway
Left raceway
Right raceway
Coordinate directions (1 = x, 2 = y, 3 = z, 4 = yy, 5 = zz)
Resting contact angle
Free contact angle

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Introduction
Gothic arch, or four-point, bearings are a unique class of ball bearings in that they offer capability to
react combined axial, radial and moment loads in a compact single row configuration. This set of
capabilities is desirable for a large array of mechanical designs; perhaps most useful in large scale
applications, such as turrets for cranes or wind turbine pitch and yaw bearings, both due to the difficulties
in maintaining coaxial alignment requirements of double row angular contact bearings and the obvious
weight savings realized from use of a single row.
Despite size and weight advantages offered by gothic arch bearings their rolling kinematics can
result in undesirable performance. In particular, four-point contact bearings are known to exhibit large
frictional efficiency losses which, in turn, cause heat generation within the bearing and accelerate the
lubricant breakdown process. These efficiency losses prohibit four-point contact bearings from being
suitable for high speed applications.
With the aforementioned pros and cons it becomes necessary for the design engineer to have a
reliable tool to assess critical performance characteristics of a four-point contact bearing. This work
proposes such a model by solving the complete internal Hertzian contact distribution of a four-point
contact ball bearing. Model outputs include the following key performance characteristics:
Ball normal contact loads
Ball contact angles
Frictional losses due to contact spin and slip
Relative displacements of inner rings to outer rings due to external loading
Non-linear bearing stiffness
L10 Fatigue life
This model neglects effects from gyroscopic moments and only considers dynamic body forces from
centrifugal effects. A study is presented herein to show removal of gyroscopic moments has an
insignificant effect on accuracy for speeds extending well beyond those advisable for gothic arch type
bearings. Hamrock [1] also neglected gyroscopic moments in his model for an arched outer race bearing.
Nelias and Leblanc [2] provided a model to consider gyroscopic moments yet ultimately caution that
selection of a double arched bearing for high speed applications should be carefully considered.
The problem of ball spin axis determination, which A. B. Jones addressed with his race control
theory [3], is handled herein with a minimum energy theory. This method allows both spin and slip to
occur at a single contact while maintaining a minimum energy state of the power losses. This is
important for four-point contact bearings as it provides a means to model torque increases due to multiple
contact points. This phenomenon is sometimes referred to as wiping.
The developed model has been programmed into the ORBIS software program to aid in the
evaluation of model results. Multiple test cases are compared against the various references and
discussed in detail.
Geometry
Geometry of a gothic formed raceway is perhaps best described by splitting a conventional deep
groove bearing down the middle of the raceway, removing equal portions of material from the two halves,
and abutting them back together. FIG. 1 illustrates this concept and a bearing of this type is often referred
to as a split ring, or split raceway, bearing. Bearing manufacturers may also use an arched shape tool to
directly cut the gothic profile without requiring ring splitting. Both types are synonymous in the context
herein and no further distinction will be made.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

go

Pd/4

ro
ri

gi

CL

CL

FIG. 1 Conventional deep groove bearing with arching region identified and resulting double arched bearing

The removed material, denoted gi for inner race and go for outer race, is called the arching or shim
size. As arching is increased from zero the initial radial play (Pd) decreases and the ball is no longer able
to make contact at the arch point (zero contact angle). To maintain this key distinction between an arched
bearing and conventional bearing Sd shall be used to denote radial play of gothic formed raceways and Pd
will denote radial play of the original circular raceway profile.
Raceway arching results in two important initial contact angles: a resting angle and the conventional
free contact angle. The resting angle, as shown in FIG. 2, is defined as the contact angle made when the
ball is simply resting on the inner or outer raceway (initial unloaded radial contact position). The free
contact angle, which is the standard way to define a conventional angular contact bearing, pertains to the
angle made at the initial unloaded axial contact position. These two contact angles should not be mixed
or interchanged as they are often different. For all four-point bearing designs the free contact angles must
be greater than or equal to the resting angle. The two are equal when there is no internal free play in the
bearing.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

free

2rest, o

2rest, i

CL

CL

Sd

FIG. 2 Resting angle and free contact angle

The arching geometry necessitates definition of additional parameters compared to a conventional


deep groove or angular contact bearing. In particular, two of the following three parameters must be
known to define an arched bearing: arched internal clearance (Sd), resting angles (rest) or the arching
dimensions (gi and go). The most practical parameters to measure are the internal clearance and resting
angles. The relationship between resting angles and arch dimensions are described by
ri fi d

(1)

ro f o d

(2)

gi sin rest ,i (2ri d )

(3)

go sin rest ,o (2ro d )

(4)

Internal play is perhaps best visualized by considering a normalized circle of radius 2r-d. As shown
in FIG. 3 the relationships between internal clearances, contact angles and shim size become readily
apparent when shown on the normalized circle.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Pdi ,o

free
2ri ,o d

rest

Sdi ,o

g i ,o

Sei ,o

FIG. 3 Normalized internal clearance circle depicting geometrical relationship between internal play, resting and free
contact angles, and the shim size.

Jones [3] determined that for a conventional angular contact bearing the relationship between radial
play and the free contact angle is
Pd 2Bd (1 cos free )

(5)

However, as previously discussed, for a gothic formed raceway the radial play is reduced. To make
use of equation (5), the original radial play, prior to arching, must be established. Leveille [4] found the
reduction to internal play for each gothic formed raceway is expressed by
Pdi (2ri d )(1 cos rest ,i )

(6)

Pdo (2ro d )(1 cos rest ,o )

(7)

Hence, the non-arched radial play is simply the sum of Sd and the reductions from gothic forming of
the inner and outer raceways. With this relationship the remaining necessary geometry is developed as
follows.
Pd Sd Pd ,i Pd ,o

free cos 1 1

Pd
2 Bd

Se 2Bd sin free gi go

(8)
(9)
(10)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Ri 0.5dm ( fi 0.5)d cos free

(11)

Ro 0.5dm ( fo 0.5)d cos free

(12)

Analytical Model
Internal Load Distribution
The general modelling approach expresses each of the four raceway centers of curvature with a
circle. The model assumes inner and outer shim sizes be greater than zero (arching is implied) and thus
each center of curvature has a unique position. Outer raceways remain fixed and a moveable coordinate
frame is attached at the center of the two inner raceways. Displacements of the moveable coordinate are
analogous to shaft deflections at the center of the bearing, which result in change to the normal approach
between inner and outer raceway circles. The normal approach of each contact point is then solved at
each ball station, or azimuth angle, around the bearing as a function of inner ring displacements. The
location of the ball center, relative to the fixed outer raceways, is determined by ensuring the ball is in
quasi-static equilibrium.
Given the normal approach of the ball center to each raceway classical Hertzian contact analysis is
utilized to determine contact normal loads at each of the four possible points. The vector sum of all ball
loads then provides the resultant force reacting on the shaft for the given set of displacements to the inner
rings. Complete bearing equilibrium is achieved when shaft reaction forces are equal to and opposite of
externally applied loads.
FIG. 4 shows a four-point bearing along with a five degree-of-freedom standard right-handed
coordinate system. As shown, the x-axis is aligned with the bearing spin axis and the y-axis is upward in
the cross-section. The inner ring deflections, qs, correspond to this coordinate frame. Ball indexing
convention is depicted in FIG. 5; where the first ball is always placed along the y-axis and indexing
increases counter-clockwise in the YZ plane.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Y+
q4
q2

q3

q5

q1

X+

Z+

FIG. 4 Coordinate system showing relation to inner ring displacements

Y
n=1
n=z

n=2

n=3

FIG. 5 Ball position definition relative to global coordinates

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

FIG. 6 illustrates the internal kinematics of the ball center and each of the four raceway centers of
curvature for initial and deflected states of the inner rings. Implied here is that when inner ring
deflections are zero (i.e. q1,2,3,4,5 = 0) the bearing is in its free contact angle position. Note that this is not
necessarily the geometric center of the bearingonly in the case where there is zero internal clearance
will this correspond to the geometric center of the bearing.
Final inner-right raceway
curvature center

Final inner-left raceway


curvature center

gi

il

+
.5)d
(f i-0

(fi 0.5

)d+

il

ir

il

Ay
Final Ball Center

(f
o -0
.5)
d+
o

ol
.5)d+
(f o-0

or

Initial Ball Center

ol

W
Outer-right raceway
curvature center

go
Ax

Outer-left raceway
curvature center

FIG. 6 Position of ball center and raceway curvature centers at initial and final loaded positions

From FIG. 6, the following relationships are developed.


Ax fi 0.5 d sin free fo 0.5 d sin free q1 Ri [q4 sin q5 cos ]

(13)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

10

Ay fi 0.5 d cos free f o 0.5 d cos free q2 cos q3 sin

(14)

q2

q2
Ri 4 sin SGN (q4 ) 5 cos SGN (q5 )
2
2

sin il

Ax W
fi 0.5 d il

(15)

sin ir

gi cos(q5 cos q4 sin ) Ax W


fi 0.5 d ir

(16)

sin or

sin ol

cos il

(17)

fo 0.5 d or
go W
fo 0.5 d ol

(18)

Ay V

(19)

fi 0.5 d il
Ay V gi sin(q5 cos q4sin )

cos ir

(20)

fi 0.5 d ir

cos or

V
f

0.5
o
d or

(21)

cos ol

V
f

0.5
o
d ol

(22)

Out of plane inner ring rotations causes a tilting effect of the inner raceway curvature centers
(denoted by in FIG. 6). This tilting angle will be small and can be sufficiently expressed with a first
term Taylor Series expansion. From equations (13) through (22), and the Pythagorean Theorem, the
normal approach of each raceway contact is solved. Negative values of normal approach indicate loss of
contact at the raceway.
il

ir

V Ax W fi 0.5 d
2

(23)

V gi sin(q5 cos q4sin ) gi cos(q5 cos q4 sin ) Ax W fi 0.5 d


2

(24)

or V 2 W 2 fo 0.5 d

(25)

ol V 2 go W fo 0.5 d

(26)

Contact angles are then determined as a function of the normal approach and ball center location.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Ay V

11

fi 0.5 d il

(27)

Ay V gi sin(q5 cos q4sin )

fi 0.5 d ir

(28)

il cos 1

ir cos 1

V
f o 0.5 d or

(29)

(30)

or cos 1

V
f o 0.5 d ol

ol cos 1

The ball center coordinates, V and W, are now needed to determine both the normal approach and
contact angle at each raceway. Since this model is quasi-static each ball must be in a state of static
equilibrium. Equilibrium force equations in the axial and radial directions become
Qil sin il Qol sin ol Qir sin ir Qor sin or 0

(31)

Qil cos il Qir cos ir Qol cos ol Qor cos or Fc 0

(32)

The ball normal load can also be defined as


Q K 1.5

(33)

By substitution the equilibrium equations can be rewritten as follows. These equations are labeled G
and H for further use.
Kil il1.5 Ax W

K ol ol1.5 g o W

fi 0.5 d il fo 0.5 d ol

Kir ir1.5 gi cos(q5 cos q4 sin ) Ax W

fi 0.5 d ir

K or W

0 G (V , W )
fo 0.5 d or
1.5
or

Kil il1.5 Ay V

fi 0.5 d il

Kir ir1.5 Ay V gi sin(q5 cos q4 sin )

fi 0.5 d ir

K or or1.5V
fo 0.5 d or

K ol V

F 0 H (V ,W )
fo 0.5 d ol c
1.5
ol

(34)

(35)

Fc is the centrifugal force acting at the ball center and is defined by


Fc

M b d m e2
2

(36)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

12

Ball orbit, or separator, speed is taken from Jones [3]. Since these equations are based on a two
point contact bearing it becomes necessary to find the inner and outer contacts that provide the
dominating tractive forces. In general this can be determined by identification of the largest normal
approach for each set of inner and outer contacts per equations (23) through (26).
i

d cos i
dm

(37)

d cos o
dm

(38)

e i

1 i cos o
(for stationary outer ring)
1 o cos i 1 i cos o

(39)

e o

1 o cos i
(for stationary inner ring)
1 o cos i 1 i cos o

(40)

The normal contact stiffness, which is a function of body curvatures and material properties, of each
ball to race contact is thoroughly described per Jones [5] and will not be repeated herein. With the
equilibrium equations now defined the solution to ball center coordinates is developed. Per Hamrock and
Anderson [1] and Harris [6] the ball center coordinates can be iteratively solved via Newton-Raphson
method.
Vn 1 Vn V

(41)

Wn 1 Wn W

(42)

Assigning equilibrium equations (34) & (35) as functions G & H respectively, the operators from
equations (41) and (42) become
V

1 H
G
G
H

W
W

(43)

1 G
H
H
G

V
V

(44)

G H H G

V W V W

(45)

Partial derivatives of the equilibrium equations are needed to solve ball center coordinates. Using
the following simplifying constants the complete sets of partial derivatives are solved. Note that
technically the centrifugal force is also a function of the ball center; however, its rate of change is poorly
coupled with radial movements of the ball center and neglecting its derivatives has negligible impact to
the convergence rate of the Newton-Raphson method.
C1 Ay2 2 AyV V 2 Ax2 2 AxW W 2

(46)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

13

C2 A Ay V gi sin(q5 cos q4 sin )

(47)

C2 B gi cos(q5 cos q4 sin ) Ax W

(48)

C2 C22A C22B

(49)

C3 V 2 W 2

(50)

C4 V 2 go2 2 goW W 2

(51)

The partial derivatives of G(V,W) with respect to V become


G G1 G2 G3 G4

V V V
V
V

(52)

C1 fi 0.5 d Ax W V Ay
C1 f i 0.5 d Ax W V Ay
G1

1.5Kil
Kil
V
C1
C11.5
0.5

1.5

0.5

1.5

C2 fi 0.5 d C2 AC2 B
2 fi 0.5 d C2 AC2 B
G2

1.5Kir
Kir
V
C2
C21.5
0.5

(54)

1.5

C3 f o 0.5 d WV
3 f o 0.5 d WV
G3

1.5K or
Kor
V
C3
C31.5
C4 f o 0.5 d
G4

1.5K ol
V
C4

(53)

0.5

go W V

1.5

4 f o 0.5 d

K ol
C41.5

(55)

g o W V

(56)

The partial derivatives of G(V,W) with respect to W are


G G1 G2 G3 G4

W W W W W

(57)
0.5

1.5

C1 fi 0.5 d Ax W W Ax
C1 f i 0.5 d
G1

1.5Kil
Kil
W
C1
C1
1.5

C1 fi 0.5 d Ax W W Ax

Kil
C11.5
0.5

1.5

C2 fi 0.5 d C2 B 2
C2 f i 0.5 d
G2

1.5Kir
Kir
W
C2
C2
15

C2 fi 0.5 d C2 B 2

Kir
C21.5

(58)

(59)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

0.5

C3 f o 0.5 d W 2
C3 f o 0.5 d
G3

1.5Kor
Kor
W
C3
C3

0.5

14

1.5

C3 f o 0.5 d W 2

Kor
(60)
C31.5

0.5

1.5

C4 f o 0.5 d g o W W g o
C4 f o 0.5 d
G4

1.5K ol
K ol
W
C4
C4

(61)

1.5

C4 f o 0.5 d g o W W g o

K ol
C41.5

The partial derivatives of H(V,W) with respect to V are


H H1 H 2 H 3 H 4

V
V
V
V
V

(62)
0.5

C1 fi 0.5 d
H1

1.5Kil
V
C1

V V Ay

1.5

C1 f i 0.5 d

Kil
C1

(63)

C1 fi 0.5 d Ay V V Ay

Kil
C11.5
1.5

0.5

1.5

1.5

2
C2 fi 0.5 d
H 2 1.5Kir C2 fi 0.5 d C2 A

Kir
V
C2
C2

Kir C2 f i 0.5 d C22A


C21.5

(64)

0.5

1.5

C3 f o 0.5 d V 2

Kor
(65)
C31.5

0.5

1.5

C4 f o 0.5 d V 2

K ol
(66)
C41.5

C3 f o 0.5 d V 2
C3 f o 0.5 d
H 3

1.5Kor
Kor
V
C3
C3
C4 f o 0.5 d V 2
C4 f o 0.5 d
H 4

1.5K ol
K ol
V
C4
C4

1.5

1.5

Finally, the partial derivatives of H(V,W) with respect to W are


H H1 H 2 H 3 H 4

W W W W W

(67)

[ C1 ( fi 0.5)d ]0.5 ( Ay V )(W Ax )


[ C1 ( fi 0.5)d ]1.5 ( Ay V )(W Ax )
H1
1.5Kil
Kil
W
C1
C11.5

(68)

[ C2 ( fi 0.5)d ]0.5 C2 AC2 B


[ C2 ( fi 0.5)d ]1.5 C2 AC2 B
H 2
1.5Kir
Kir
W
C2
C21.5

(69)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

C3 f o 0.5 d
H3

1.5K or
W
C3

0.5

C4 f o 0.5 d
H 4

1.5K ol
W
C4

0.5

WV

1.5

C3 f o 0.5 d WV

K or
C31.5

V W go

15

(70)
1.5

C4 f o 0.5 d

K ol
C41.5

V W go

(71)

With the foregoing solution to the ball center coordinates established the internal load distribution at each
ball contact is obtained and standard Hertzian contact methods can be used to determine contact stresses
and sub-surface shear stresses.
Bearing Reaction Forces and Stiffness
Net reaction forces of the bearing on the shaft are obtained by summing the force components for
each inner ring contact. Referring to coordinate definitions shown in FIG. 4 and ball indexing definition
in FIG. 5, the five bearing reaction forces are defined.
z

F1 [Qil , n sin il , n Qir , n sin ir ,n ]

(72)

n 1

F2 [Qil , n cos il , n Qir ,n cos ir ,n ]cos

(73)

n 1

F3 [Qil , n cos il , n Qir , n cos ir , n ]sin

(74)

n 1

F4 Ri [Qil ,n sin il ,n Qir ,n sin ir ,n ]sin

(75)

n 1

F5 Ri [Qil , n sin il ,n Qir ,n sin ir ,n ]cos

(76)

n 1

Stiffness of a four-point contact bearing entails developing the Hertzian contact stiffness of all four
possible load paths. As illustrated in FIG. 7 this results in a parallel pair of springs for the inner-left and
inner-right races and another parallel pair of springs for the outer-left and outer-right races. However,
since both pairs of springs are tied together at the ball center, the two pairs are connected in series. Recall
from earlier that the outer raceway centers of curvature are fixed in space while the inner curvature
centers are allowed to move relative to the fixed outer races.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Kir

16

Kil

Kor

Kol

FIG. 7--Four-Point Contact Stiffness Diagram

The total stiffness of the inner ring pair relative to the fixed outer pair is then expressed by equation
(77); which represents the non-linear instantaneous stiffness of the bearing under a given relative
deflection of the inner rings.

Fj
1
1

Fj
Fj
Fj
q j Fj

q q

q j ,or
j ,ir
j ,ol
j ,il

(77)

Let the following notation be used to express the partial derivatives for each ring pair
Fj
qinner
Fj
qouter

Fj
qil
Fj
qol

Fj

(78)

qir
Fj

(79)

qor

Then the diagonal terms of the full Jacobian become


F1
Kil
q1,inner

M il ,sin
Qil il
M il ,sin Nil

Kir

q
q1
1
il

F1
Kol
q1,outer

M ir ,sin
Qir ir
M ir ,sin Nir

q
q1
1
ir

M ol ,sin
Qol ol
M ol ,sin Nol

Kor
q1
ol q1

M or ,sin
Qor or
M or ,sin Nor

q1
or q1

(80)

(81)

M il ,cos
M ir ,cos
Q
Qir ir
F2
Kil cos il il M il ,cos Nil ,cos
M ir ,cos Nir
Kir cos

q2,inner
q2
q2
il q2
ir q2

(82)

M ol ,cos
M or ,cos
Q
Qor or
F2
Kol cos ol ol M ol ,cos Nol
M or ,cos Nor
Kor cos

q2,outer
q2
q2
ol q2
or q2

(83)

M il ,cos
M ir ,cos
Q
Qir ir
F3
Kil sin il il M il ,cos Nil ,cos
M ir ,cos Nir ,cos
Kir sin

q3,inner
q3
q3
il q3
ir q3

(84)

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

17

M ol ,cos
M or ,cos
Q ol
Qor or
F3
Kol sin ol
M ol ,cos Nol
M or ,cos Nor
Kor sin

q3,outer
q3
q3
ol q3
or q3

(85)

M il ,sin
M ir ,sin
Q il
Qir ir
F4
Ri Kil sin il
M il ,sin Nil
M ir ,sin Nir
Ri Kir sin

q4,inner
q4
q4
il q4
ir q4

(86)

M ol ,sin
M or ,sin
Q ol
Qor or
F4
Ri Kol sin ol
M ol ,sin Nol
M or ,sin Nor
Ri Kor sin

q4, outer

q
q4
4
4
4
ol

or

(87)

M il ,sin
M ir ,sin
Q il
Qir ir
F5
Ri Kil cos il
M il ,sin Nil ,sin
M ir ,sin Nir ,sin
Ri Kir cos
(88)
q5,inner

q
q5
5
5
5
il

ir

M ol ,sin
M or ,sin
Q ol
Qor or
F5
Ri Kol cos ol
M ol ,sin Nol
M or ,sin Nor
Ri Kor cos

q5, outer
q5
q5
ol q5
or q5

(89)

Where the functions M and N in equations (80) through (89) are defined as
Nil
Nir

N ol
N or

il1.5
( fi 0.5)d il

ir1.5
( fi 0.5)d ir

ol1.5
( f o 0.5)d ol

or1.5
( f o 0.5)d or

(90)
(91)
(92)
(93)

M il ,sin Ax W

(94)

M ir ,sin gi cos(q5 cos q4 sin ) Ax W

(95)

M ol ,sin go W

(96)

M or ,sin V

(97)

M il ,cos Ay V

(98)

M ir ,cos Ay V gi sin(q5 cos q4 sin )

(99)

M ol ,cos V

(100)

M or ,cos V

(101)

Partial derivative of normal load with respect to the normal approach has a general form that is
independent of raceway. By consistent subscript assignment the following expression can be used.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

Q
1.5
1.5

( f 0.5)d [( f 0.5)d ]2

18

(102)

The normal approach equations and the functions expressed in equations (94) through (101) will
have two forms depending on whether the reference is an inner or outer raceway.
i i Ax i Ay

q Ax q Ay q

(103)

o o W Ax o V Ay

q W Ax q V Ay q

(104)

M i M i Ax M i Ay

q
Ax q Ay q

(105)

M o M o W Ax M o V Ay

q
W Ax q
V Ay q

(106)

il
Ax W

Ax il ( fi 0.5)d

(107)

Ay V
il

Ay il ( fi 0.5)d

(108)

ir
Ax gi W

Ax ir ( fi 0.5)d

(109)

ir Ay V gi sin q5 cos gi sin q4 sin

Ay
ir ( fi 0.5)d

(110)

ol
W go

W ( f o 0.5)d ol
ol
V

V ( f o 0.5)d ol

(111)
(112)

or
W

W ( f o 0.5)d or

(113)

or
V

V ( f o 0.5)d or

(114)

The remaining partial derivatives needed are shown in Table 1 and Table 2.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

19

Table 1Partial derivatives of Ax and Ay

Ax

Ay

q1
q2
q3

1
0
0

cos

q4
q5

Ri sin
Ri cos

Ri q4 sin * signum(q4 )
Ri q5 cos * signum(q5 )

0
sin

Table 2Miscellaneous partial derivatives

M il ,sin

M ir ,sin

M ol ,sin

M or ,sin

M il ,cos

M ir ,cos

M ol ,cos

M or ,cos

Ax

1
-

1
-

1
0
-

-1
0
-

-1
0

1
0

0
1
-

0
1
-

0
1

0
1

Ay
W
V

To make the foregoing analysis useful it is desired to compute the precise bearing deflections due to
a known externally applied load. This is achieved by numerical iteration, via the Newton-Raphson
technique, as follows.
1

qn 1

F
qn
Fapplied Freaction
qn

(115)

Frictional Effects
Bearing friction torque, or its internal resistance to rotation, is dependent upon a multitude of
physical phenomenasuch as lubricant shearing (viscous effects), retainer or cage drag forces, imperfect
geometry of rolling elements and associated raceways, contact surface asperities, material elasticity and
hysteresis, and rolling/spinning contact slip due to relative surface velocities. For bearings of precision
quality with well-designed mounting conditions, and under loaded conditions, the primary torque
components typically come from contact sliding and viscous effects. The focus herein will be to extend
the conventional torque equations for contact slip within an angular contact bearing to a gothic arch
bearing. A well-known understanding with balls containing four points of contact is the kinematic overconstraint prevents pure rolling motion at all contact points. This phenomenon is sometimes referred to
as wiping; which is a loose term that is not clearly defined in an analytical sense but shall be defined
here, qualitatively, to describe a situation where there is an appreciable angle between the surface
velocities of the ball and the raceway.
Jones [3] developed a general friction model for ball bearings that accounts for interfacial slip due to
rolling and spinning at the contact. His model includes a race control theory that requires one of the
contacts in a two point bearing to contain pure rolling while spin occurs at the other contact. Boness and
Chapman [7] later developed a novel test apparatus that provided contrary results to the race control
theory. Its worth mentioning that Jones race control theory does result in a very close approximation for
well-behaved operation of angular contact bearings. Chapman and Boness found that the ball spin axis

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

20

angle, which defines the roll-to-spin ratio at the contact, tends to take an angle that minimizes the energy
state of the frictional losses.
With gothic formed bearings there can be up to four contact points on a given ball, and race control
theory becomes untenable. The conclusion made by Chapman and Bonesswhere the ball spin axis
attitude resolves to the lowest energy stateshall be applied here with use of general rolling and sliding
friction models developed by Jones. Additionally, to further expand the torque model, rolling and
spinning coefficients of friction will be differentiated to allow separate assignments for each. Per Jones
[3] the torque required to rotate a ball bearing under complex loading is

63, 025H
N o Ni

(116)

Where torque has units of in-lbf, H is the horsepower generated from frictional losses, and Ni and No
are inner and outer race speeds in revolutions per minute. The frictional horsepower, which is now
extended to all four possible contact points, becomes
HP

n
1
( s ,ilq s ,ilq s ,irq s ,irq s ,orq s ,orq s ,olq s ,olq

6, 600 q 1

(117)

R ,ilq r ,ilq R ,irq r ,irq R ,orq r ,orq R ,olq r ,olq )

Where S is the spinning torque and R is rolling torque. Spinning torque at each contact point is
defined by
s

3spin aQ
8 E ( e)

(118)

Where E(e) is the complete elliptical integral of the second kind with modulus e (contact
ellipticity), spin is the coefficient of sliding friction, a is the semi-major dimension of the contact ellipse
and Q is the normal ball load. Rolling friction, or Heathcoate slip, is written as follows. Here the
tractive braking effects are assumed to be zero, which results in a pure rolling line at 34.37% of the semimajor ellipse dimension.
R

3roll PR
sin 41 2sin 4 2
(1 2 2 )
1
2sin 2 cos 2 sin 1 cos 1
2
2
4sin 1
16sin 1
4sin 1

(119)

(120)

2 sin 1

0.3437a

(121)

2 fd
2 f 1

(122)

1 sin 1
R

Ball spin velocity for a stationary inner ring is

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

d m o
(1 o )(1 i )
d (1 o ) cos i (1 i ) cos o

21

(123)

And for a stationary outer ring becomes


b

d m i
(1 o )(1 i )
d (1 o ) cos i (1 i ) cos o

(124)

Now, the relative surface velocities for spinning and rolling at each contact become
s ,il i sin il b sin(il )

(125)

s,ir i sin ir b sin(ir )

(126)

s,or o sin or b sin(or )

(127)

s,ol o sin ol b sin(ol )

(128)

R,il i cos il b cos(il )

(129)

R,ir i cos ir b cos(ir )

(130)

R,or o cos or b cos(or )

(131)

R,ol o cos ol b cos(ol )

(132)

The lowest frictional energy state is achieved when equation (117) is minimized as a function of the
ball spin axis angle (). Energy minimization can be achieved with a numerical search algorithm that
varies between zero and /2. While this method provides both spin and slip to occur on a given contact
the assumption for tractive braking effects to be zero at each contact is likely invalid for certain operating
conditions. Namely, in cases where a raceway pair (i.e. inner left and outer right) provide dominate
control of the ball spin axis any contact on the opposing raceways must be in a state of slip. Such
instances could be modelled by modifying equation (121) to adjust the no-slip lines.
Fatigue Life
The ANSI/AFBMA (now named ABMA) standards [8] were specifically developed to allow users to
derive bearing fatigue life estimates via hand calculations and lookup tables. According to Jones [3],
these simplifications required assumptions regarding the internal load distribution versus applied external
loading. In particular, the standards make the following assumptions: inner and outer rings remain
parallel (i.e. only axial and radial relative motion), the contact angle is only influenced by thrust loads,
body forces due to element motions are neglected, initial mounted conditions are neglected (i.e.
preloading), and combined axial and radial loading can be modelled with an equivalent radial load.
Harris [6] further noted the fatigue equations only predict raceway failure and that ball failure apparently
was not observed in the LundbergPalmgren test data. He further postulated that during the Lundberg
Palmgren era the ability to manufacture accurate geometry of balls with good metallurgical properties
exceeded that of the raceways, thus resulting in predominate raceway failure in the test data.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

22

Another implication, which has not been studied sufficiently to be quantifiable, is that sliding at the
contact is believed to accelerate metallic fatigue of the raceways. The ANSI standard is based on
extensive testing of deep groove radial bearings which can only contain two points of contact on the ball
and will therefore have minimal sliding at the contact. Since the L10 formulation is all that is available at
this time it will be simply extended to include all four contact points with no additional knockdown to
account for additional sliding in a four-point contact bearing.
In the case of a ball with four points of contact, versus only two contact points, it is clear the ball
stress cycles must increase and hence so will the cumulative fatigue damage; yet the following L10
calculations are unable to account for this. To better account for irregular internal load distributions
within a gothic formed raceway it is not recommended to use ANSI tables with an assumed double row
configuration to approximate a four-point contact bearing. Instead, capacities shall be derived at each
contact point. Per Harris [6], dynamic capacity of a point contact on a raceway is defined by equation
(133) for ball diameters up to 1.0 inch (25.4 mm) and equation (134) for larger ball sizes.
2f
C p Ap

2 f 1

0.41

2f
C p Ap

2 f 1

0.41

1
1

1.39

1
1

1.39
1

d

dm

0.3

d

dm

0.3

d 1.8
1
z3

(133)

d 1.4
1
z3

(134)

The upper sign in (133) and (134) pertains to inner ring contacts and the lower sign is used for
outer ring contacts. Material constant Ap tends to vary depending on the source, however Harris shows
this constant to be 7080 for Cp in units pounds (71.7 for Cp in Newtons). This constant is applicable to air
melted 52100 chromium steel. Note that equations (133) and (134) are the complete general forms for
dynamic capacity of a ball-to-raceway contact and are not restricted to a particular conformity, contact
angle or load distribution.
Life of each raceway, in total revolutions, is established by summing the ratio of each normal contact
force and the applicable capacity of the raceway contact. The exponent for this ratio has a slight
dependence on the relative fixity of external loading to the rotating ring. Also, since there will be load
cases where some normal contact forces can be zero it is numerically convenient to invert the standard
L10 equation such that the capacity term is in the denominator.
L10 life for a raceway that is rotating relative to the external load is
L10race, rotating

Q
z n 106
n 1
C p , n
z

(135)

L10 life for a raceway that is stationary relative to the external load is
L10race, stationary

Q
z n
n 1
C p , n
0.9

10

106

(136)

Each raceways life capacity is then combined to determine the total life capacity of the bearing. In
the case of a four-point contact bearing the final L10 life, expressed in bearing revolutions, becomes

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

L104 Pt

1
1
1
1

10
10
10
10
9
9
9
L10ir
L10or
L10ol 9
L10il

23

10

(137)

Discussion
Model Results
To enhance confidence in the model a set of analysis test cases are compared with various
references. The primary parameters needed to validate the core model pertain to the internal load
distributionwhich is comprised of ball normal loads and associated contact angles. Hertzian contact
stresses provide a convenient comparative metric but are essentially a secondary measure of normal loads
for a given set of contact conformities and material properties. Other performance parameters, namely
friction torque, stiffness and L10 life are essentially all driven by the internal load distributions and
therefore are not discussed. To simplify matters, and maintain focus on the core model, all boundary
conditions will be assumed to be rigid (a.k.a. fixed ring analysis).
Each of the analytical references available has certain limitations and therefore a consistent set of
references for each test case becomes impractical. The A. B. Jones High Speed Ball and Roller Analysis
Program3 and ORBIS bearing analysis program both provide reference for standard angular contact
bearing analysis capabilities. These two references, which are denoted herein as Jones-2PT and ORBIS2PT, will be utilized to compare various parameters applicable to two-point modelling. They will also be
used to demonstrate that modelling a four-point contact by placing two angular contact bearings with a
common center point and opposing contact angles is not always a valid technique. The only reference
available that models a true four-point contact bearing explicitly is A. B. Jones Four Point Ball Bearing
Analysis Program4, herein referred to as Jones-4PT. This program provides detailed output for all four
contact points. Its limitations, however, are that it only performs a static analysis and requires the inner
and outer curvatures be the same value. It therefore does not provide torque losses or accurate results for
moderate speed applications. Additionally, the Jones-4PT model is not documented and there appears to
be a discrepancy in the non-linear stiffness output and some minor numerical errors when resolving the
elliptical integrals needed for stress results. The Wind Turbine Design Guideline, DG03 [9], is another
reference and provides hand calculations for estimating maximum ball load, maximum stress, torque and
L10 life. However, it is limited to a fixed curvature assumption, two discrete contact angle options, and
zero internal play (resting angle and free contact angle are the same).
Three bearings, as defined in Table 3, will be used to compare the different model predictions.
Bearing A is a fairly large cross-section bearing with tight curvature and large internal play. Bearing
B has a large pitch diameter, thin cross-section, and open curvature. Bearing B-DG03 is a
modification to bearing B to accommodate the constraints of the DG03 analysis method.

3
A. B. JONES HIGH SPEED BALL AND ROLLER BEARING ANALYSIS PROGRAM, JONES ENGINEERING COMPANY AND
DAVID A. JONES, LICENSED FOR USE BY ATEC/ANH N.TRAN, CYPRESS CA, LICENSE 060907, VERSION 5.5Q, SERIAL NO.
06140755Q1385, ISSUE DATE JULY 3, 2007
4
A.B.JONES FOUR POINT CONTACT BALL BEARING ANALYSIS PROGRAM (4PT), JONES ENGINEERING COMPANY AND
DAVID A. JONES, LICENSED FOR USE BY ATEC, CYPRESS, CA, UNDER LICENSE NO. 060907, VERSION 2.0B, SERIAL NO.
06090720B0005, ISSUE DATE JUNE 16, 2007

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

24

Table 3Bearing Parameters


Parameter
Pitch Diameter, in. (mm)
Ball Diameter, in. (mm)
No. Balls
Radial Play, in. (mm)
rest
free
fi
fo
2
Young's Modulus, Balls & Rings, psi. (N/m )
Poisson's Ratio of Balls & Rings

Bearing 'A'
7.3838 (187.549)
0.875 (22.225)
24
0.0098 (0.249)
26.83
41.12
.52
.52
2.9E+07 (2.0E+11)
0.25

Bearing 'B'
30.5 (774.7)
0.25 (6.35)
189
0.0041 (0.104)
30
39.6
.543
.543
2.9E+07 (2.0E+11)
0.25

Bearing B-DG03
30.5 (774.7)
0.25 (6.35)
189
0.0 (0.0)
45
45
.53
.53
2.9E+07 (2.0E+11)
0.25

Static Load Capacity Static load capacity for conventional bearing steel (52100) is defined per
ISO 76 [10] to be the external load that results in 406 ksi (2.64 GPa) mean stress on one or more of the
ball contacts. Since stress is proportional to contact normal load, and normal loads are a function of
internal load distributions and element contact angles, the static load capacity is a convenient parameter
for comparative purposes.
Table 4 shows predicted thrust, radial and moment load capacities from the programmed model
(ORBIS-4PT) and reference programs for each of the three bearings listed in Table 3. ORBIS-4PT results
show strong correlation with Jones-4PT results for most load directions and bearing configurations. Since
the Jones-4PT model is expected to contain a rigorous four-point contact model, similar to the developed
model herein, good correlation between these two models is expected. With the exception of bearing A
thrust and moment cases the model results correlate with Jones-4PT to within about 1%. Further
investigation into the thrust and moment cases for bearing A, which consequently have about 6%
difference, showed both ORBIS-4PT and Jones-4PT will predict identical ball loads and contact angles
for a given input load but resolve the corresponding Hertzian contact stresses with some difference. Since
this study fixed the stress to 406 ksi, and stress is proportional to load to the third power, small
differences in resolved Hertzian contact stress account for the 6% differences in predicted static load
capacities. It was further found that these particular cases result in computation of elliptic integrals where
the modulus is very close to 1.0 and, knowing the integrals become tougher to resolve to high accuracy in
this vicinity, it is believed that numerical precision errors likely explain these differences.
The model also shows good correlation with the DG03 results for radial and moment loading.
However, the predicted thrust capacity from the DG03 model appears to disagree (~10% difference) with
all other models. This difference can be explained by the fact that the DG03 analysis only uses the initial
contact angle in the equations and provides no means to adjust it due to contact deflection. For this thrust
case the contact angle increases from 45 to 52.6; thus resulting in about 10% change to the normal ball
loads.
As expected, the 2PT results correlate well with thrust loading results from the 4PT models.
However, for bearings A and B, which contain internal clearances that produce differences between
the resting angle and free contact angle, the 2PT modelling technique consistently diverges from the 4PT
results for both radial and moment load cases. In the case of bearing B moment capacities the 2PT
model diverges by ~27%.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

25

Table 4Thrust, Radial and Moment Static Load Capacity Results


Bearing
A

B-DG03

Model

Thrust, lbf

Radial, lbf

Moment, ft-lbf

ORBIS-2PT

193,600

91,900

26,400

Jones-4PT

206,700

88,100

23,350

ORBIS-4PT

193,600

87,500

22,000

ORBIS-2PT

81,100

41,800

46,100

Jones-4PT

81,600

37,200

36,350

ORBIS-4PT

81,100

36,800

36,200

ORBIS-2PT

112,400

50,300

63,900

DG03

100,600

50,300

63,900

Jones-4PT

113,200

50,560

63,900

ORBIS-4PT

112,400

50,300

63,900

Four-Point Analysis Overview


Since most bearing performance parameterssuch as stress, stiffness, torque, and lifeare all
related to the internal load distribution it is important to compute distributions with good accuracy. FIG.
8 and FIG. 9 illustrate the effects to internal load distribution and friction torque when varying the
arching, or shim size, on a gothic arch bearing that is under pure radial loading. At low resting angles
(less arching) there are fewer balls available to react the applied load and thus the peak ball loads are
amplified. As arching, or resting angle, is increased the distribution spreads out the applied load over
more balls. For the example shown in FIG. 8 and FIG. 9 the free contact angle is 39.6 and once the
resting angle reaches this value all internal play within the bearing is taken up (Sd = 0). Further increases
to the resting angle result in internal preloading (interference fitting on the balls) and eventually the load
distribution transitions from a subset of balls carrying the load to all balls carrying the load. Once all
balls have been loaded the load distribution takes a fixed profile and any additional shimming shifts the
distribution curve up the ordinate axis (increase in ball loads). While achieving minimum peak ball loads
is generally desirable, as shown in FIG. 9, there is a rapid torque increase once all balls are engaged
(above 39.6 resting angle for the example shown).

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

26

250

Ball Load

200

150

100

50

0
0

20

10

40

60

20

80

30

100
Ball No.
39.6

120

140

41.63

160

180

200

43.6

FIG. 8Internal load distributions for a bearing under pure radial loading with various arching (expressed by resting
angle).

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

27

600

Roll & Spin Torque (in-lbf)

500

Free Contact Angle

400

300

200

100

0
0

10

20
30
Resting Angle (deg)

40

50

FIG. 9Torque due to slip and spin bearing B under pure radial loading with various arching (expressed by resting
angle)

Gyroscopic Moment AssessmentTo investigate any modelling errors attributed to omission of


gyroscopic moments a study of constant thrust loading versus speed is performed and compared against
the 2PT model (accounts for gyroscopic moment effects). FIG. 10 and FIG. 11 show this comparison for
bearing A and bearing B. Ball normal loads, contact angles and mean stresses for a constant thrust
load are plotted against shaft speed. As shown, the model shows negligible difference for all three
parameters at low speeds. Not until around 3,000 RPM is there a perceptible difference in results and
even up to 8,000 RPM these differences are not large. Maximum errors, at max speeds shown in the
figures, are: ~3% error for normal ball load, ~14% error for contact angles, and ~ 1% error for mean
Hertzian stresses. Bearing A exhibits a larger divergence than bearing B, which is expected since
bearing A will experience greater gyroscopic moments due to its larger ball diameter (maximum % error
in contact angle for bearing B is ~8% at 10,000 RPM compared with 14% error on bearing A at 8,000
RPM).

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

(a)

28

(b)

(c)
FIG. 10Comparison of 4PT model to 2PT analysis with gyroscopic moment effects, bearing A, 5,000 lbf thrust load:
(a) normal ball loads, (b) contact angles, and (c) mean Hertzian stress

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

(a)

29

(b)

(c)
FIG. 11Comparison of 4PT model to 2PT analysis with gyroscopic moment effects, bearing B, 10,000 lbf thrust load:
(a) normal ball loads, (b) contact angles, and (c) mean Hertzian stress

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

30

Conclusion
Classical rolling element theory, as originally developed by A. B. Jones, has been extended to
develop an analytical model of four-point contact ball bearings. The model includes independent
geometry definition for both inner and outer raceways, including their associated arching, thereby
allowing assessment of a broad range of manufactured bearing conditions. All necessary equations,
including detailed partial derivatives, are completely derived and provided. Key performance parameters,
such as element contact stresses, contact angles, deflections, nonlinear stiffnesss, torque and L10 life are
solved. The complete model has been programmed within the ORBIS software program and various test
cases covering all loading directions and different internal bearing geometry were analyzed. Model
results were verified with various references and thoroughly discussed. A study was presented to show
neglecting gyroscopic moments has negligible impact to accuracy for slow to moderate bearing spin
speeds that extend well beyond bearing manufacturer recommendations. The model has replaced Jones
race control theory with a minimum energy criterion; thereby allowing interfacial slip and spin to occur
simultaneously within a given contact.
Analytical modelling of a four-point contact bearing by overlay of two angular contact bearings, with
opposing contact angle orientation and common center location, is shown to produce gross internal load
distribution errors for radial and moment loading. For this reason use of a rigorous four-point model,
such as presented herein, is highly recommended.
Further Research Suggestions
The model provided herein establishes a core foundation for the analysis of gothic formed bearings.
However, it is expected the model will be expanded upon to include many of the common challenges with
real-world bearing applications. For instance, items such as press fitting, thermal expansions, and
clamping effects present real-world application challenges that can have significant influence on bearing
performance. Many techniques to model these conditions have developed and used on angular contact
bearings. Most of these techniques and are expected to be directly applicable, or useable, in this model by
simply adjusting the initial bearing geometry accordingly. One area that was not covered herein is
performance effects due to raceway runouts. This is likely of concern to large diameter, thin section,
four-point bearings. Higher aspect ratios (ratio of diameter to cross section) inherently lose their hoop
stiffness and pose challenges for manufacturers to produce precise circular geometry. Raceway runouts
will invariably cause ball contact load fluctuations that will result in increased rotational resistance of the
bearing. Such effects can be further magnified when internal clearances are small or zero. One method
for modelling this, which is showing promising results, is to superimpose a sinusoidal error function on
the circle describing the raceway center of curvature.
Another area requiring further study is the frictional losses due to wiping. The model provided
herein used assumptions for no-slip bands on sliding friction. It is highly likely these bands do not exist
on contacts where a pair of raceways defines the ball spin axis yet the opposing pair of raceways is in
contact and must be wiping (the entire elliptical contact is in a state of slip). Preliminary investigations
show dramatic torque increase can be predicted by extending the slip region to include the full major
dimension of the ellipse. A fairly simple test setup could be constructed to incrementally introduce
multiple ball contact points, perhaps by application of radial load increments, and would provide helpful
data for further model refinement.

2015 ASTM F34 International Symposium on Rolling Element Bearings An Analytical Model of Four-Point Contact Rolling
Element Ball Bearings by J. Halpin and A. Tran (submitted for review: 20-Apr-2015)

31

References

[1]

B. J. Hamrock and W. J. Anderson, "Arched-Outer-Race Ball-Bearing Analysis Considering


Centrifugal Forces," NASA, Cleveland, OH, 1972.
[2] D. Nelias and A. Leblanc, "Ball Motion and Sliding Friction in a Four-Contact-Point Ball Bearing,"
Journal of Tribology, ASME, vol. 129, no. Oct 2007, pp. 801-808, 2007.
[3] A. B. Jones, "The Mathematical Theory of Rolling-Element Bearings," in Mechanical Design and
Systems Handbook, New York, McGraw-Hill, 1964.
[4] A. R. Leveille, "The Non-Reversible Nature of Ball Bearing Internal Geometry," in REBG
International Bearing Symposium, Orlando, FL, 1997.
[5] A. B. Jones, Analysis of Stresses and Deflections, Vol I, Bristol, Conn: General Motors Corp, 1946.
[6] T. A. Harris, Rolling Bearing Analysis, 4th Ed., New York: John Wiley & Sons, 2001.
[7] J. J. Chapman and R. J. Boness, "The Measurement and Analysis of Ball Motion in High Speed
Deep Groove Ball Bearings," Journal of Lubrication Technology, ASME, no. July 1975, pp. 341-348,
1975.
[8] S.-9. ANSI, "Load Ratings and Fatigue Life for Ball Bearings," American National Standards
Institute, 1990.
[9] T. A. Harris, J. H. Rumbarger and C. P. Butterfield, "Wind Turbine Design Guideline, DG03: Yaw
and Pitch Rolling Bearing Life," National Renewable Energy Laboratory, Golden, CO, 2009.
[10] ISO-76, "Static Load Ratings, International Organization forr Standardization," Geneva,
Switzerland, 2006.
[11] ISO-281, "Rolling bearings - Dynamic load ratings and rating life," Technical Committee ISO/TC 4,
2007.

Você também pode gostar