NASA

/ TM--2000-210368

Effect Roller
Joseph J.V. Poplawski Erwin Glenn Steven

of Roller Bearing
and Associates,

Profile

on Cylindrical

Life Prediction
Bethlehem, Pennsylvania

V. Poplawski

V. Zaretsky Research Center, M. Peters and

Cleveland,

Ohio

J.V. Poplawski

Associates,

Bethlehem,

Pennsylvania

Prepared for the 2000 Annual Meeting sponsored Nashville, by the Society of Tribologists Tennessee, May 7-11, 2000 and Lubrication Engineers

National Space

Aeronautics Administration

and

Glenn

Research

Center

August

2000

Available NASA Center for Aerospace 7121 Standard Drive Hanover, MD 21076 Price Code: A03 Information

from National Technical Information Service

5285 Port Royal Road Springfield, VA 22100 Price Code: A03 Available electronically at http: / / gltrs.grc.nasa.gov / G LTRS

Effect

of Roller

Profile on Cylindrical Life Prediction

Joseph Bethlehem, Erwin National V. Poplawski & Associates Pennsylvania V. Zaretsky J.V. Poplawski

Roller

Bearing

Aeronautics and Space Administration Glenn Research Center Cleveland, Steven J.V. Poplawski Bethlehem, Ohio

M. Peters & Associates Pennsylvania

Four sohttion

roller and

prq[Hes finite andJhlly prqfile The

used element

in cylindrical analysis loaded were loading profile

roller (FEA] against

beatqng for a fiat stress

design and Lundberg predicted lives

and li['e. Four

mam_/'acture The roller rolling-element Palmgren, not significant Weibull in

were pt_)files

analyzed analyzed and

using were Harris, than

both fiat, were and

a closed.form tapered chosen.lot" Zarets_'. end. this The as

aerospace, analysis .fiat roller 98 percent. The Jidly and for equal

ccw,'ned These edge

raceway.

bearing loannides edge loading

life models can reduce condition upon

and compared. without profile used the

those

of Weibull,

and but For

has the longest produced lives. slope lives exponent

l!fe. However, predicted

life by as much not the only Hertz depends stresses exponent

end tapered

the highest

O, di]ferent stress

the aerospace

prq/He.

ctwvned

produces but also accepted

the lowest on the Weibull predict lower load-l(fe

The resultant assumed. than q/ 10/3

l!fe at each slopes a standards.

on the l!/e equation laonnides-Harris line contact,

q[" 1.5 and 2, both Lundberg-Palmgren Based Hertz maximum stress-li/'e

equations

the ANS1/ABMA/ISO results in the [ield.

to 6. 6. This value

is" inconsistent

with that experienced

SYMBOLS A C
e

material-life dynamic critical roller Weibull probability probability exponent life, number adjusted adjusted 10-percent characteristic roller

factor load capacity, N (lbD exponent

shear diameter, slope

stress-life m (in.)

d
e

F Ax) h L La Lm Llo L_

of failure, of survival

fraction function

or percent

of stress

cycles

or hr cycles or hr of stress cycles or hr survives, fails, number number of stress of stress cycles cycles or hr

life, number life based

of stresses

on fatigue

limit,

number

life or life at which

90 percent 63.2

of a population

life or life at which m (in.) m (in.)

percent

of a population

flat length,

It.

total length

of raceway,

NASA/TM

2000-210368

[r

effective total roller

roller

length,

m (in.)

I, N
I1

length,

m (in.) cycles exponent or number of components, elemental volumes

life, number maximum normal load-life comer

of stress Hertz

stress-life radial

P P

or equivalent exponent radius, radius, m (in.) m (in.)

load, N. (lbf)

1", S

crown

probability maximum residual

of survival, Hertz stress, volume, stress.

fraction GPa (ksi)

or percent

GPa (ksi) m _ (in. _)

V
X

stressed exponent load, depth stress

X Z

time,

or stress critical GPa (ksi) GPa (ksi) GPa (ksi) GPa (ksi) GPa (ksi) GPa (ksi) shear stress, m (in.)

to maximum or strength, parameter, stress, shear

or,,

location vonMises

critical maximum

stress.

shear shear

stress, stress.

orthogonal r,, fatigue

limit,

GPa (ksi)

Subscripts i
tl

i'h component number reference system related

or stressed

volume or elemental volume, volumes

of components point, stress,

ref sys
V

or life of survival or life

or component to stressed

probability volume

designates

characteristic

life or stress

N ASA,'TM--2000-210368

INTRODUCTION Thisbasis fortheANSI/AFBMAndISOlife predictions a (1-3)forcylindrical rollerbearings thetheory G. is of Lundberg A. Palmgren Theirlifetheory and (4,5). isbasedpon u theworkof Weibull (6-8). ubsequently, have S others published modifications ofLundberg Palmgren Among arethetheories Ioannides and (4,5). these of andHarris9)and ( Zaretsky (10-12). Zaretsky, Poplawski, Peters and (13)comparing resultsfthe the o different lifetheories discussing and theirimplication in thedesignnd a analysis ofballbearings presented a critical analysis. Foraninverse ninth-power relationetween b life and maximumertz H stress for"point ontact" onraceway), c (ball theLundberg-Palmgren qualitatively theory predicts life best. However, tbraninverse2th 1 power relationetween b lifeand maximumertztress, Zaretsky modified theory is H s the
best. Using a "fatigue-limiting stress" such as proposed by loannides and Harris to predicting failed with (9) without modifying factors components He also significantly and, noted more as the that had over predicts the life of ball bearings (Zaretsky, et al. (13)). A. Palmgren (14,15) in 1924 suggested a probabilistic approach specifically, time there shorter at which was rolling-elements 10 percent an apparent size bearings. of a population effect on life. bearings. (4), the L,) life of a bearing can be determined from the equation: On the basis of bearings that will have bearings the lives of machine that" an acceptable will have survived. load equivalent

of his test results,

he suggested or 90 percent the same

life is defined bearings

is, larger

as smaller

lives than the smaller

From Lundberg

and Palmgren

LI 0 = [C/p]

1'

(1)

where load

Lt0 is the bearing that will radial produce the lives

life in millions

of race

revolutions, race revolutions

C is the dynamic with

load capacity probability This

of the bearing of survival,

or the theoretical P is the applied is a have a

a life of one-million of roller of these bearings bearings.

a 90 percent bearings.

equivalent Predicting variable

load and p is the load-life

exponent. is more Because complex than that of ball effects is because to edge the roller geometry of the deleterious on life due loading, the rollers

in the design

full crown or a partial crown rather than a flat roller profile. As a result the Hertz contact in most roller bearings is a hybrid between "line contact" (flat roller profile on a plain) and "point contact." Lundberg and Palmgren (5) state that with line contact p equal contact same They become exponent. between assumption Tests that roller range and exponent between 3. They at some both rings the exponent p = 4. They further state that with point contact between both rings the exponent

observe that, as a rule, the contacts between the rollers load. Accordingly, the load-life exponent p varies from In this regard, that it be applied value between used Lundberg and Palmgren roller stress suggest bearings that to all cylindrical Hertz

and the raceways transform 3 to 4 for different loading a suitable point value of load-life and line contact.

from a point to line intervals within the exponent This p is 10/3. ofp has value

roller further

bearing. suggest

for mixed (1-3).

the accepted For line maximum is made, analysis of 4, which From

in the ANSl/ABMA

and ISO Standards

The relationship

load and maximum

is Sm,_ _ P' where

Sma,, is the maximum

Hertz

stress

and x is an relation on what

contact x = 2 and for point contact x = 3 (16). Hence, for line contact the theoretical resultant Hertz stress and life is L - S'_ma_. Based upon a load-life exponent p of 10/3 and, depending the resulting by Rumbarger for those Parker that with and stress-life and Jones reasonable Zaretsky (18) bearings exponent (17) n can either for oscillatory a Hertz stress-life the correct contact, be 6 2/3 straight value values exponent or 10 for line needle roller or point bearings there contact, resulted exponent exponent

respectively. in a load-life data for cylindrical experimentally in his computer the segment as power load-life bearing life in life For a

established

of 8. However, stress-life stress-life

is no controlled

is published bearings. from (19)

to establish

certainty for point of stress

of the Hertz of the Hertz contact

8.4 to 12.4. recognized defining the state in a roller-race is difficult. As a result, by treating uses a 4th the roller into thin slices. He calculated the Hertz stress in each segment Jones (19) does not relate the Hertz stresses in the contact to life but life is not reflective et al. (13), a 20-percent an exponent a similar result is based varying variation of 6 2/3 variation of the actual the Hertz or in Hertz stresses stress-life stress selected between Hertz under in the bearing. exponent n can significantly a two affect can result in a nearly to one difference value. Lundberg by Zaretsky bearings, on whether roller

Jones

program he segmented a thin roller. However, exponent. As was predictions. prediction relation given Palmgren load, Hence, discussed For roller depending is used (4,5) different

the predictive

10 was

in the calculations. the actual stresses predict and thus

Conversely, life. Both life.

if a load-life and

independent and Jones (19),

of stress, geometries which

can occur

life and the predicted roller bearing

in significantly upon Lundberg

different and Palmgren

NASA/TM--2000-210368

Inviewofthe aforementioned discussion,objectivesthe the of

(1) (2) (3) Determine the three dimensional flat raceway using finite element Evaluate Determine and compare what effect the various volumetric analysis. life theories stress field lbr cylindrical

work between

reported

herein

were

the following: different roller crown profiles and a

a square bearings

roller with roller

with

roller

different bearing

geometry.

the presumption

of a fatigue

limit has on cylindrical

life prediction.

LIFE

THEORIES

Weibull

Equation

I')'acture

Strength

in 1939

W. Weibull

(6,7)

published

two papers in material

that strength

describe

a statistical

approach group

to determine of test

the

strength of solids. could be expressed

Weibull postulated that the dispersion according to the following relation: in in(l/S)=e

for a homogeneous

specimens

ln[X / X,t3]

(2)

where

X = G and A'_ = o'# (see Appendix (2) relates specimen slope survival (and

A). S to the fracture fatigue) data modulus (or rupture) designated a cumulative of exponential, slope, is also strength o. When In In(l/S) is used itself as the ordinate of this line is referred from (statistical) distributions, slope, the to an

Equation

and In o" as the abscissa

and fracture

are assumed usually

to plot as a straight by the letter distribution

line.

The slope

(tangent)

is referred to as the Weibull as a Weibull plot. By using extremely distribution. respectively The larger scatter a Weibull small sample Weibull (8). plot, size. slopes

or Weibull

e or m. The plot of an infinite data and and normal the value variation

it becomes The

possible slope

to estimate is indicative

population

Weibull

of the dispersion

of the

its density (Gaussian)

of 1, 2, and 3.57 proportional versa. strength (20).

are indicative

Rayleigh,

in the data making related

is inversely and vice

to the Weibull slope The smaller

that is, the lower liable to statistical size the greater subjected of survival

of the Weibull depending variation solid

the scatter (6,7)

in the data

The Weibull to the volume

on the sample in the slope. the solid to be by

size (database) Weibull

up the distribution the material

the sample

the statistical to stress. for the entire

of the material

[f we imagine

divided in an arbitrary manner into n volume elements, the probability multiplying the individual survivabilities together as follows: S = S1.$2.$3...S,_ where the probability of t'ailure F is F = 1- S

can be obtained

(3)

(4)

Weibull

(6,7)

further relation:

related

the probability

of survival

S, the material

strength

_J, and the stressed

volume

V, according

to the following

lnl S
where

. f(X)dV

(5)

f(X) For a given probability of survival S,

= ge

(6)

(7)

NASA/TM

2000-210368

From Eq.(7)['orthesamerobability p ofsurvival thecomponents withthelarger stressed volume ill have w thelower strength (orshorter life).
Fatigue Li[e In conversations A. Palmgren with E.V. Zaretsky on January 22, 1964, Sweden W. Weibull related how he had suggested bearing (fatigue) to his life contemporaries where and G. Lundberg in Gothenburg, to use his equation to predicl

f(X)

= rCN e

(8)

and where in the publication Ref (13)

r is the critical past we have

shear credited

stress this

and N is the number relation of the to WeibuIl. application

of stress

cycles

to failure. there is no Weibull documentation in the open of the above nor any in

However, of Eq. (8)

to the

authors'

knowledge

by

literature.

However,

we did apply

Eq. (8) to Eq. (5) where

(9)

The parameter scatter From Hertz

c/e is the stress-life theory

exponent.

This

implies

that the inverse

relation

of life with

stress

is a function

of the life

or data dispersion. l' and r can be expressed as a function of S,1,_ (13) and substituting L for N

,
L = ALr j LVJ
Sm_x

"

(10)

From

(13),

solving

tbr the value

of the exponent

n for line contact

(roller

on raceway)

from

Eq. (10) gives

c+l n e

( 11 a)

For point

contact

(ball

on raceway) c+2 n = -e

(1 lb)

It should Lundberg order these values

be noted and

that before were ofp

the Lundberg-P_rlmgren empirical used and made

life theory to conform

(4) was published, p. The values values ofp slope

Palmgren previously e must

(21 ) had already c and e published

published selected be 9.3. data by In If 11.1 varies

Eq. (1) relating to retain

bearing Palmgren the value

life to the inverse

of load P to an exponent the values

for the exponents

to the

by Palmgren. 10.2 and

by Palmgren, (Experience first used

for the Weibull then from the Weibull bearings that

be 1.11 and c/e must bearing fatigue (9),

from

Lundberg contact,

and Palmgren respectively.

for c and e are retained, has shown

Eqs. (11 a) and (11 b), n equals slope e for most

for line and point from 1 to 2.) Using life of individual

a finite-element stressed

analysis volumes

(FEA)

for rolling-element as follows:

by Ioannides

and Harris

the computed

can be integrated

1_

Nef

In 2

Jv rcdV

(12)

Equation

(12) can be rewritten

to represent

each

individual

stressed

volume

and associate

stresses

as follows:

LTJJ"ac,,'e rL-_, nl/e J

5

(13)

NASA/TM

2000-210368

where Eq. 10),he from ( t material lhctor

a=L"eiir"eS] l%J
From Lundberg and Pahngren (4) (see Appendix survival are summarized as follows: B) the lives
r 1C/Cr I ll/C

(14)

of the individual

stressed

volumes

at a given

probability

of

(15)

By replacing

X in Eq. (2) with L, the probability

of survival

S and the life L can be related

to Sr_t-and Lr_t as tbllows:

S = Sire '_f''LI" L

(16)

Lundberg-Palmgren

Equation

in 1947, G. Lundberg and A. Palmgren (4) applied Weibull analysis life. The Lundbcrg-Palmgren theory expressed/(X) in Eq. (5) as

to the prediction

of rolling-element

bearing

fatigue

,lffCN

f(X)--

--7

117)

where shear

r is the critical stress

shear

stress,

N is the number contact. From

of stress

cycles

to failure,

and Z

is the depth

to the maximum

critical

in a concentrated

(Hertzian)

Eqs. (5) and (12)

N_p1 L_J I"'F LT_I

From Hertz theory ( 16), I; r, and Z can be expressed as a function of S .... and substituting L for N

(18)

c:
Substituting these values

Ay"'1 ' tTJ tfrj

lbr the exponent c+l-h n -

1
S_a..< n for line contact gives

(19)

into Eq. (12) and solving

(20a)

and Ibr point

contact c+2-h n (20b)

From For the

Lundberg

and

Pahngren

(4),

using using

the

values

of

c and analysis

e previously (FEA),

discussed of

and

h =

2.33, stressed

then

from

Eqs. (2a) and (20b), can be computed

n equals

8.1 and 9 for line and point contact, theory, a finite-element

respectively. the lives the individual volumes

Lundberg-Palmgren as follows:

NASAffM

- 2000-210368

In I

(21)

As was follows:

done

for

Eq. (12).

Eq. (21)

was

rewritten

to rcpresent

each

individual

stressed

volume

and associate

stresses

as

Li

--[ Tref [

L r,. j
where from Eq. (19), the material factor

/ _ef [ L I"i J

t zi [ LZref J

(22)

Lref[12reflc'e[ j

, 1l''e = r LI'refi LZref 1h/e i

(23)

Using

Eq. (15),

the lives of the individual Equation

stressed

volumes

are summarized

to obtain

the component

life L.

loannides-Harris loannides where from and

Harris

(9),

using

Weibull

46,7)

and

Lundberg

and

Palmgren

(4,5}

introduced

a fatigue-limiting

stress

Eq. (5)

f(X)

= (r - r,,)CN Zh

(24)

The equation stress where

is identical

to that

of Lundberg

and

Palmgren
\C'f'

(Eq.
l "e

(18))

except

for the

introduction

of a fatigue-limiting

N-

IZ]

(25)

Equation

(25)

can be expressed

a function

of Sma_ where

A(

1 c/e

1 t'e

t.'e

(26)

S;;ax
ioannides exponent loannides predict and Harris of and (r (9) use the same - r,,). (9) using equation finite relates element each analysis elemental (FEA) volume integrated as follows: the computed life of elemental stress volumes to values of Lundberg and Palmgren and Palmgren (Eqs. (20a) for e, c, and h. If r, equal and (20b)). However, 0, then the values of r,, of the

n are identical Harris life. Their

to those of Lundberg

for values

> O, n is

also a function bearing

lnl_

Nel,

(r-

Zh

r,,)CdV
volume and associated stresses as follows:

(27)

Equation

(27) can be rewritten

to represent

each

individual

stressed

r'-,l:r(+-,),o,+',, +t iv+_.,//+__, / L ;J
/Lref J kZref 3

NASA/TM--2000-210368

If welettrot=(r
used to calculate are summarized Zaretsky to obtain

r,,)_f thevalue f A then o

lives for each life L. the component

for Eq. (26) volume.

is the same Using

as Eq.

(23).

The

value

of A from

Eq. (23) stressed

is then volumes

the individual

stressed

Eq. (15),

the lives of the individual

Equation

Both

the Weibull

and

Lundberg-Palmgren cA" thus exponent becomes, depends

equations

above

relate

the critical

shear

stress-life

exponent

c to the Weibull

slope e. The critical shear wide variety exponent equation

parameter stress-life of materials

in essence, the effective critical shear stress-life exponent, implying that the on bearing life scatter or dispersion of the data. A search of the literature tbr a fatigue or dispersion the exponent reveals that most stress-life Zaretsky slope exponents (12) has vary from 6 to 12. The the Weibull in the data. c independent Hence, rewritten

and tbr nonrolling-element of scatter by making

appears to reflect

to be independent that obse_'ation

of the Weibull

e, where

f(X) From Eqs. (5) and (29)

= rCeN e

(29)

(30)

Equation Zaretsky that once the critical

(30) assumes initiated, shear

differs based stress fatigue

from upon

the Weibull experience takes a crack

Eq. (9) and the Lundberg-Pahngren that the value of the stress-exponent to the surface bearing The fatigue and fonn to propagate

Eq. (4) in the exponent c = 9. Lundberg a fatigue spall propagation propagation relation Zaretsky (1947) (13) Lundberg

of the critical

stress

r. to

and Palmgren is a function time dependent.

(4) assumed of the depth However, small time is true. their

the time

Z. Hence, life can

by implication, be categorized life

life is crack Crack

rolling-element fraction Palmgren theoD', To decouple

as _'high-cycle and cycle" crack fatigue

fatigue." rate,

is an extremely implies dispensed with

for the total life or running the dependence ofL of"high relation the concepts

time ot" the bearing. (It should

Lundberg-Palmgren that at the time

that the opposite published

of bearing cycle" as

propagation were

the Lundberg-

- Z _'' in Eq. (30).

be noted

and Palmgren

and "lot,

only then beginning

to be formulated.)

Equation

(30) can be written

(31)

From Ref. (13),

solving

for the value

of the exponent

n, for line contact c+l n e

from

Eq. (31) gives

(32a)

and for point

contact c+2 n e

(32b)

where

c = 9 and e = 1.11, n = 9.9 for line contact (10) that as well uses as ioannides and Harris a discrete-stressed-volume

and n = 10.8 lbr point contact. (9) proposed approach. a generalized August and Weibull-based Zaretsky (11) that is based on stressed volumes methodology extended this finite element as follows: for structural methodology stress analysis. life by

Zaretsky prediction

developing a technique Zaretsky, like loannides

tbr predicting component life and survivability and Harris, integrates the complete life of elemental

NA SA/TM--2000-210368

And.aswith loannides
follows:

and

Harris,

an elemental

reference

volume

and

stress

is required.

Equation

(33)

is rewritten

as

IL,
where from Eq. (31). the material factor A= Lref [_'ref ]c [I'ref ] lie

(34)

(35)

The

lives of the individual Zaretsky (12)

stressed does

elements

are summarized a fatigue-limiting

in accordance stress,

with Eq. (15). not exclude that concept either. However, his stress, the be affected

Although

not propose

he does

approach is entirely different life for the elemental stressed where L = 1/V/''.

from that of loannides and Harris volume is assumed to be infinite. a reduction in stressed volume

(9). For critical stresses less than the fatigue-limiting Thus, the stressed volume of the component would of 50 percent results in an increase

As an example,

in life by a factor of 1.9.

ROLLER

TYPEAND

PROCEDURE

Roller

Geometr3' of a nonlocating ring have than greater cylindrical axial ball roller surface bearings. bearing area with than is shown in contact roller thrust in Fig. with the 1. This inner bearing and outer type allows build races, primarily of their sensitive axial movement Because support radial no of

A schematic the inner roller greater loads. apparent bearings loads If properly

or outer

to accommodate sized they

thermal

expansion Cylindrical nominal ball

of the shaft bearings loads roller

and tolerance are designed

up in an assembly. they generally to carry radial

rolling-element

comparably designed, bearings

heavy

can be operated greater loads

of up to 5 percent bearings are more

load with

degradation roller

of performance. support bearings, rollers fatigue. shaft can to misalignment effect of edge loading on "'straight" bearing life due to rolling-element stress distribution rings. Moment loading on the rollers. loading on bearing on the on load or stress profile Angular misalignment of the bearings misalign the as shown also is shown in Fig. 2. The higher between the shaft and housing is another with roller reason for the In order to minimize

Although

and/or edge loading. The stresses result in reduced also effect partial The cage general provides considered roller causes nonuniform inner misaligned

Poor

alignment

on the shaft bearing.

and outer

of misalignment

and edge

life, the rollers

are profiled

in Fig. 3, usually precision d, referred grade,

a full or guidance, For roller, is the

crown. The effect of a partial crown on load or stress profile is shown in Fig. 2. limiting speed of a cylindrical roller depends on roller length-to-diameter ratio, type and material, the best balance equal type of lubrication, having and speed series shaft and housing roller The presumed accuracy, speed and heat to roller limitation use, roller dimensions an effective length/,.equal diameter of a roller with

dissipation bearing

of the overall having

mounting. rollers

to as a _'square'" "square"

of load

capacities.

to that of a comparable effective length can be written

ball bearing. to be in contact the races under loading. Generally,

In Fig. 3, the roller effective length

l,.is the length as tbllows:

l,. = It - 217

(36)

where

r, is the roller

comer

radius

or the grinding roller profiles

undercut, shown

whichever

is larger. roller bearing life prediction, loaded model were roller we selected a 12.7-mm is shown studied. having (0.5-in.) in Fig. 4. were a taper

To compare

the effect

of various

in Fig. 3 on cylindrical

simple roller-race geometry model diameter rollers running in a linear, The effective roller length l,. is equal roller (a) flat (straight) cylindrical

for evaluation. The model assumes a plurality of normally raceway having a length IL. A schematic of the roller-race to the roller with diameter, edge 12.7 mm (0.5 in.). loading; (b) partially Four roller profiles and without (end) tapered

These

profile

NASA/TM--2000-210368

angle 0.20 witha flatlength f 8 mm(0.314 oi" o in.);(c)aerospace (partiallyrowned) witha flatlength t"8mm c roller o (0.314 in.)and 965-mm a (38-in.)adius; r and(d)fullycrowned rollerhaving 965-mm a (38-in.) rown c radius. Three maxmmmertz H stresses chosen were forcomparisons witheach rollergeometry. were These nominally.4,1.9, nd 1 a 2.4GPa (200, 75, nd350ksi).The 2 a normal loadsoproduce stresses different t these were foreach roller profile. he T loads, stresses, anddimensions foreach used roller profile Fig.3 intheroller-race ofFig.4 are of model summarized in Table 1.
Finite Element Stress Analysis and Life Prediction (FEA) advantage for the geometry of the symmetric of the roller-race nature model used in the studies (Fig. 4)

A three-dimensional, is shown significant in Fig. 5. The surface shear

finite-element model stresses geometry

analysis takes

of the Hertzian

contact

for the case

of no

or misalignment. area face was divided into -162 elements. Element 1.0. Beyond size ranged stress that depth from level. 0.0991x0.0330 Element thickness mm was stress. of with The turn (0.00503x0.00125 in.) until a depth -3500 stress to 5900 The in.) depending z/b of about solid on the Hertzian thickness

The quarter (0.0039x0.0013 in the depth gradually The model freedom We

section direction

of the contact was 0.0254 A typical (350-ksi) constraint model nonhal

in.) to 0.1278x0.03175

mm (0.0010 contained

the element

increased. after

model

isoparametric was performed classical

elements nodes,

depending giving about

on the Hertz computer theory over (13).

for 2.4-GPa applying the FEA

maximum boundary results stresses available

Hertz

had -5800 analysis values shear

elements

and 7000

18 000 degrees

conditions. calculated

on a 450-MHz Hertz contact

personal stress of theory

the COSMOS/M checked

commercially

FEA software. against by stress using

FEA-predicted stress range

principal studied. distributions three

and the in-plane

z'45 agreed

within

3 percent

the Hertz

Three fatigue shear

stress used

that stresses

have were

been

discussed

over

the years shear stress

as being within

the "critical this paper.

stress" They

in determining were or equivalent

bearing stress ksi) in at using

life. These stress

examined

as the "stress (b) in-plane

of choice"

(a) the orthogonal

by Lundberg

and Palmgren:

l"45; and the Von Mises

effective

field. Figure 6(a) shows the three-dimensional orthogonal shear stress field for a aerospace roller with edge loading. Figure 6(b) shows the corresponding Von Mises stress distribution. A maximum stress of about 0.84 GPa occurred The evaluating 0.114 results mm (0.005 of the FEA in.) below runs examined the surface three for a 1.4-GPa stresses For purposes (200-ksi) at each maximum roller roller profile only geometry were Hertz were stress. saved as databases will no edge at the Hertz

(122

to be used be considered and

the life theories

in this paper.

of analysis,

the life of the race assuming using

each load condition. The Li0 life at 2.4 ANSI/ABMA/ISO standards is normalized The analysis. a reference Pahngren, By using element model can component These selected respectively, percent have component life L_f, life and survivability a unit or gage factor a material

GPa (350 ksi) for a flat and assumed to be 1. for each of the of the

loading

life equations to the gage Eq. (35). stress individual for L and of 1.0 results Hence, life equations

predicted Zrcrand

results

of the finite-element stress r_t. all related Lundbergthe finiteanalysis to Eq. (14); from

By establishing

volume A for each

Vrer, a depth and Zaretsky, shear the values values a relative

volume

a reference for Weibull, elemental of survival

can be calculated and respective the probability

Eq. (23): loannides-Harris, the appropriate analysis, L and S values

Eq. (23); tbr each Eq.

life equations

and critical element

volumes

are computed.

lbr the entire

be obtained by using can be detennined. provide element.

(3) to multiply or normalized reference imply values

survivabilities. S in relation 0.9 are

By using to reference

Eq. (15), values

the Lic_ life of the chosen S_f from the of 90 we of and 1.11 The and is

equations reference

relative These

Generally,

and

assigned can

to the

Lre_ and at random.

variables,

in the equations.

or normalized element

life of unity

and a probability be chosen

of survival However,

for the reference primarily c and used

element

or volume with

l_cr. A reference resultant 9 in.3). The specific critical ksi); and

or volume

the element 5.32x10

the highest parameters reference stress,

stress

at a reference value material.

depth For the

below

the surface,

Zrer. The

value

l'_r selected exponent exponent evaluated. values presented were Ioannides-Harris

by us was

14 m3 (3.244x10 slope e are Three shear 0.64

corresponding to the stresses

of Z_r is 210x10 Weibull, slope to V_. were and Von (A primer Mises

6 m (8.25x10 Lundberg-Pahngren, to be used stress this

3 in.). The

the Weibull equations,

c = 10.3.

For the Zaretsky

equation,

c = 9. The

Weibull

e is assumed

h is assumed These 0.82 in Ref were GPa (22).)

to be 2.33. the maximum (119 ksi);

r_r corresponding shear (228.1 stress; ksi).

in the analysis whose methodology

r45; the orthogonal 1.57 GPa

respective

GPa (93.3

detailing

NASA/TM

2000-210368

10

RESULTS

AND

DISCUSSION

Life Theory Using roller These

Comparison tbrm Hertzian solution (16) and assuming no edge stresses, normal loads were calculated for the flat

a closed

geometry thai would produce maxinaum Hertz stresses of nominally loads and stresses are summarized in Table 1, Based on these normal Hertz that stresses lbr diameter radius roller 2).--For are the end and profile r, was tapered, length zero. were The aerospace, both N (953 contact, 2. The contact and width lb) normal the maximum equations crowned (0.5 in.). load the roller 12.7 mm

1.4, 1.9, and 2.4 GPa (200, 275, and 350 ksi). loads and the lamina method of Jones (19), the roller It was geometries assumed profiles were that was determined. For these length the the rollers equal length had no corner

maximum calculations, radius,

is, the corner

for all the roller

to the roller

except for the crowned roller width. Weibull Weibull, modulus. (3.33) should while tbr Slope These be 4 that (Table results results

at the 4239

that produced stress-life standards

a contact relationship of the use

of 90 percent

line and point Ioannides-Harris, in Table contact. Hertz but limited

Hertz

was determined Weibull slope exponent This

for the p of value 10/3 of n

Lundberg-Palmgren, line contact

and Zaretsky From stress-life Lundberg exponent 18).

as a function

or Weibull contact

summarized

ANSI/ABMA/ISO and Palmgren Based upon

a load-life exponent (see

and 3 for point in a maximum with available results

(4) the load-life the Hertz with stresses

p for line

n of 8.1 for line contact is inconsistent

Eq. (20a)). for line contact database "mixed be true and tapered point

low is consistent p of 3.33,

data (13,

and the load-

life exponent

in a value

of n equal

to 6.6. This

the available can experience (5). This roller : may

and can account contact," in limited roller that bearings is,

in part for the lower life predictions than that experienced in the field. Lundberg and Palmgren's justification for a p of 10/3 was that a roller one raceway bearing designed Referring The value For mixed resultant Accordingly, slope is 1.11. that the value As currently Weibull resulted shown reflect Zaretsky equations slope the critical exponent for Weibull 12.7 mm Hertz highest loading, calculated. ANS1/ABMA/ISO The assumed. equation Weibull resultant As (Eq. slope reasonably Applying slope in their in Table a strong (18) then, shear slopes (0.5 can experience but "line contact" not and the other with the book raceway vast (2l), point "point majority designs back contact life of the This ofp it is certainly 1945 appears edition normal The that consistent

bearing contact"

roller

of cylindrical ofp

and used today. to the ofA. load, the Palmgren's from he uses a value database contact will case where 1.11. have will 3 for both in Palmgren's life, which having of -3.3 However, on both upon Palmgren and line contact. paper (14,15). the contact. will where dominate of p = 3.33 to come initially bearing load-life an unreported discussed 1924

for a given bearing. verifies should

the race having life will exponent

the lowest have a value contacts.

resultant

be less than p in that bearings to be and Palmgren

the life of the raceway for mixed

the point

we have

calculated

the Weibull

the recommendation and as discussed above,

of Lundberg roller both

it is our opinion raceways. the value this of the as and the of assumption However, by factored Parker into and Harris

be not less than 4 for cylindrical bearings

line contact relations Lundberg

occurs and bearing

practiced which analysis

the load-life

and stress-life For

are based

for rolling-element matching on in Table no longer relation slope. 3).--It The

is assumed relations The

preexisting the Weibull of

life equations slope.

(21 ) and their of Weibull, existing some cases

nonpublished

life database. and loannides reported were with

2, both the load-life dependence reflect slopes as shown the relation stress-life between (Table

and stress-life

Lundberg higher

and Palmgren, fatigue data If the slope decreases

rolling-element or lower. significantly a slight

in the range 2, the

1 to 2 and reality.

stress-life

(load-life) Accordingly, slope between

exponent only

increases Hertz

in Weibull stress-life contact of

whereby

matches value

the Zaretsky

equation variation

that decouples of the maximum

the dependence

and the Weibull of n varies

shows

n and Weibull

9.5 and 9.9 tbr line contact roller load bearings capacity Also, plastic GPa the These with lives,

and 10 and 10.8 for point rollers bearings Hertz of 4.14 having produced stress GPa of 4.14 (600

2 and 1.11. was calculated of 12.7 mm 3.96 (0.5 by us for cylindrical in.) that the dynamic (575 to 625 ksi). contact stress normalized of 2.4 GPa (350 strongly depends in predicted predicts the for the dynamic steel roller-race Hertz comprising of these the stress deformation geometry ksi) roller results used slope tbr (600 flat a maximum a diameter maximum GPa can ksi) is the

Li['e Prediction stresses stress

in.) and a length of approximately that can be placed

to 4.31 GPa stress

Accordingly, without of 4.14 to ksi). life highest

be chosen the four The

as a representative life equations lives life

load capacity. standards

on a hardened normalized results

of the contact. and assuming each roller based in Table with lives. are 3. slope for the by the the Except Zaretsky no edge was the upon geometry

and the ANSI/ABMA/ISO to a maximum were stress subsequently Hertz stress variation condition

to the flat roller

the theoretical relative

geometry are shown but also comes lives

standards predicted we previously (31)). of

and a maximum life at each discussed, the

on the equation Weibull result the standards

the Weibull

the least calculated, Weibull

At all conditions 1.11 at which

the ANSI/ABMA/ISO equation

in the lowest highest

predicted

NASA/TM--2000-210368

11

Zaretsky equation. ForWeibull slopesf 1.5and2,boththeLundberg-Palmgren o andloannides-Harris r,,= 0) (where equations lowerives the predict l than ANSI/ABMA/ISO standards.
Fatigue Limit (Fable 4). As we previously where
(','o

discussed,

Ioannides

and Harris

advocates

the use of a fatigue

limit

r,, in the

Lundberg-Pahngren

equation

(37)

and where equal Table data limit The Table solution

Lm is the life with assuming maxinmm to support

the fatigue no edge _GPa Hertz

limit

r,,, L is the life without and a Weibull 60 ksi). The exponent, of a fatigue induced using lives failure a finite increased What those Hertz-stress slope n was limit of results

a fatigue 1.11, lives

limit were

r,, and r,, is the critical calculated in Table be noted bearing of most for assumed 4(a). that steels. For each there

shearing values value

stress. of r,, of r,, of

For a fiat roller, 4, a resultant

loading stress-life

to 138, 276, and 276x10

(20, 40 and

are summarized It should hardened range

calculated. for through

are no definitive if a fatigue would in form Hertz

in the literature were to exist,

the existence of fatigue matter. was the Also, repeated predicted

However,

the probability above results, amount.

in the operating element with

rolling-element results are

bearings summarized the closed maximum

virtually

not exist as a practical analysis 4(b). but For described the FEA to a lesser

analysis increases

(FEA). were

These limit

in fatigue higher

as with same

the resultant

life exponents

tbr the

stress and assumed fatigue limits with the FEA analysis. with and without the assumption of a fatigue limit exceeds We have a fatigue concluded In 1965 that loannides Zaretsky, and Harris (9) have limit. et al. (23) published

is important to note is that in all cases the values predicted using the ANSI/ABMA/ISO standards. the existence relation of compressive residual stresses

predicted for that of

confused

the following

(1Jmax)r

= -Tmax

- 1 (__S r

(38)

where residual

r,,,,

is the maximum the positive would to the ninth

shear

stress,

(r,,,,,),. sign

is the maximum a tensile stress

shear

stress

modified residual

by the residual stress,

stress,

and S,. is the Accordingly, relation a of

stress,

or negative reduce power.

indicating shear

or compressive be

respectively.

compressive life and stress

stress

the maximum The modified

and increase life L4 would

the fatigue

life according

to the inverse

or adjusted

=+m4I.ax '--', l1
2 )J If in Eq. (37) stress-life 276xl0 Roller Roller finite we let r,, equal 1/2 S, and 138x10 c/e equal 6 GPa 9, the (20 ksi) two equations in Table become identical. The resultant with exponents 6 GPa (40 ksi). Profile Prolile Comparison (TablesL analysis --Four (FEA) analysis of this, considers only and roller profiles and previously life. The analysis described loannides-Harris is similar Hertz stress to that and the and the Zaretsky were analyzed analysis equations stressed of Zaretsky using without if the were volume both a closed exponents as defined form limit are by n for a r,, of 4 are certainly consistent a residual

(39)

maximum

Hertz stress of

solution chosen

and to The and to be

element

for stress

a fatigue used

is identical

Lundberg-Palmgren identical. Because closed form solution Palmgren where

the Weibull only the

the Lundberg-Palmgren maximum

for this comparison. Lundberg

V:

1 / _{ ,..I L . Z)

(40)

It does

not consider

the effects

of stress

concentrations

and the entire

subsurface

stressed

volume.

NASA/TM

2000-210368

12

The theoretical foreach lives roller geometry calculated normalized were and toamaximumertz H stressf4.14 o GPa
(600 The exponents 1.11 from Lundberg With predicted crowned aerospace The stress that the ksi). results The relative standard for the closed and form produces these consider and edge geometry method life results form were subsequently Hertz stress without summarized The lives profile, While edge normalized loading in Table values those or stress are under four are to the flat roller in Table by Zaretsky's the crowned the the geometry 5(a). The based Hertz exceed upon the ANSI/ABMA/ISO and a maximum solution those predict and and lowest entire from of 2.4 GPa (350 ksi). summarized calculated concentrations, profile and the stress-life slope those of for 2 and calculated of the standard. flat roller profile, end stress profile tapered has the profile longest The and full the respectively. tbr line contact and a Weibull method

n of 8.1 and 9.9 were Lundberg Palmgren and Palmgren. the closed life profile profile, followed

Zaretsky.

of life

Both methods solution the the

exceeding edge the there

not considering lives. volume These effective those lives

by the end-tapered differences

aerospace

life differences the Hertzian in Table analyzed. solution. from should

between contact 5(b). Except As with that be those by the and with life. and This

may not be significant. stressed results of the with distribution would form strongly profile, solution, including suggest the lives for the end are summarized profiles form the closed significantly these can or profiled 2 that for relative roller stresses the element analysis

FEA

results

concentrations flat roller with profile of

loading. is least exceed

for the

flat roller

predicted tapered critical Effect roller edge smooth Table (350 The However, edge profile lacking stress ksi)

the FEA produced where Loading. edge

the closed

the highest The roller with

but not

different two profiles rollers reduce roller from as actual

of the aerospace of preference. fact that In order Iife was end

profile.

Certainly,

applications Edge have will loading stress 5(b). The

life and reliability as illustrated contact no end

are factors, in Fig.

use of blended a finite loading 6. The flat

is dictated roller profile (FEA)

the ends conducted

or edges

of a flat of a in As and used. with is

stresses flat

bearing of stress and one

to evaluate as

the effcct considering summarized

on the results

analysis

distribution were

a flat lives profile, The

loading Hertz

normalized in Table on lower the

to the ANSI/ABMA/ISO results in a higher

standards life prediction. expected,

at a nominal

maximum

stress

of 2.4 GPa purposes. as 98

and are summarized discussed of the method edge loading and that appears ksi))

the standard is (350

are presented to reduce calculated and As with exponent

for comparison life by as much

previously 82 percent

of Zaretsky load,

effect

at the higher except will exceed

respectively. with

percentage 1.9 GPa standard. stress-life companies,

depends

on the

analysis life even

for the values

at the higher predicted to reflect and, perhaps for which

of 2.4 and ANSI/ABMA/ISO Hertz

275 ksi),

the predicted

loading,

the previous database

analysis accepted. at lower

for roller There Hertzian

the FEA analysis in the open (less than GPa (300

a higher in the

maximum files analyses

n then normally

literature

of the bearing

a definitive

any of these

can be benchmarked.

GENERAL

COMMENTS

The

basis

for

the

ANSI/ABMA published

and

ISO in 1947

life and

prediction 1952 (4,5).

for

cylindrical upon

roller

bearings

is

the

life

theory

of

G. Lundberg (14) assumed In their 1952 point contact states,

and A. Palmgren

Based

an unpublished proportional for roller edition

database,

Palmgren

in 1924

roller bearing life based on a modified publication Lundberg and Palmgi'en and the other raceway has line contact.

line contact is inversely calculate a 10/3 exponent Palmgren, in the third

to radial load to the 10/3 power. bearings where one raceway has published in 1959 (21)

of his book,

"Pure line contact occurs only in certain exceptional cases. In many types of roller bearings, at least one track is slightly crowned, so that in the case of zero load there is point contact, which, as the load increases, becomes line contact. The exponent values p = 3 and p = 4 are therefore the limit values for roller bearings. As it is desirable to have a uniform method of calculation for all designs of roller bearings under all conditions, it is of advantage to introduce a mean value of the exponent for all types, namely p = 10/3. The basic dynamic load rating (capacity) of the roller bearings must then be adapted so that the error is small in the most common range, L - 100 millions to L - 10,000 millions of (race) revolutions." The bearing incorrect standards consideration 10/3 exponent assumption company and roller under be given has been him, incorporated and line contact it is no longer and bearing revised this value life. to reflect to p = 5. Experience roller be into the ANSI/ABMA/ISO correct roller for most standards types suggests bearings bearings herein roller first of roller that published bearings today the that the in then 1953. While certainly is with

Palmgren's for cylindrical

of point bearings. predicts bearings

may have been the case the analysis Accordingly,

for many

in use by the exponent ABMA/ISO p = 4

employing

manufactured

and most ANSI/

presented

10/3-power exponent

it is our

recommendation

for roller

for cylindrical

a load-life

to increasing

NASA/TM--2000-210368

13

In 1985 S.loannides T.A.Harris9)published they and ( what claimed as w a"'new life"theory incorporating a fatigue limit.The concept fatigue ofa limittbrrolling-element bearings firstintroduced was byPahngren (14)in 1924 ndthen a abandoned byhimbythetirnehewrote thefirstedition hisbook of (21). undberg Paimgren do not consider the L and
concept of a fatigue limit in their 1947 life theory 1965 relation if any, (4). What of Zaretsky limit that exists loannides a fatigue and Harris limit to account and To assume (9) do is to tack in our paper for the effect is nothing else bearing Qualitatively, closed form the entire analysis anything onto the 1947 above. than result life theory However, residual the over in an of Lundberg this relationship stress value The from paper provides Lundberg critical contact loading closed exponent and Pahngren is the same (4) a relationship as the fatigue residual fatigue life. stress, life. to use finite of their of the as does that for each assumes FEA, part element "new life analysis life" theory. equations form stress for rolling-element The FEA solution. volume stresses analysis The discussed above. life prediction by us both the solution FEA first value comes in this by analysis used of the to edge was applied Hence, incorporating the fatigue as discussed Harris herein more will el al. (23) of loannides in the steel. of compressive

on rolling-element of a compressive of rolling-element suggestion in our the and shearing same and Harris previous ordering first and

one-half

prediction

and methodology paper (13)

loannides

(9) in the application of life prediction using

the closed the defined

Pahngren stress.

by implication Whereas,

incorporates throughout the FEA

the maximum

the distribution in a higher

of shearing life prediction.

subsurface is sensitive

Hertzian

is considered

resulting

for the most

Because

and stress concentrations, form solution regardless n ranged a valid FEA from -14 database analysis

it is our opinion of the life theory to are 18 and not verified in exist does

that it may provide a more accurate used. However, in some cases the case, validate n = 29. These or field values data, or invalidate

quantitative life prediction than the resultant maximum Hertz stress-life were the more higher it is our than we anticipated. that be values recommendation life

a single to either either

Untbrtunately, until the various relied upon.

the analysis,

with

experimental

conservative

SUMMARY

OF RESULTS

Four solution and were fatigue exponents maximum were were maximum

roller

profiles

used The

in cylindrical analysis (FEA) profiles

roller normally

bearing used

design vary with

and

manufacture profiles

were analyzed Four The analysis at the

analyzed were

using

both bearing

a closed end,

tbrm

and finite crowned. The

element

for stress These on roller roller

and life. The roller were those

flat, tapered loannides

aerospace life models and Harris, without nominal lives The lives has the ksi). a if the

fully

roller

manufacturer. evaluated.

rolling-element and Palmgren, loannides-Harris is similar loads The ksi). and

chosen limit

_br this analysis effect is identical arc Hertz Hertz the chosen stresses stress-life The closed predicted The stress of profile

and compared. limit The

of Weibull, life was were

Lundberg

and Zaretsky.

of a fatigue to the Lundberg

bearing analysis

analysis that produced (350

and Palmgren

and the Weibull evaluated (200, tbr cylindrical roller

to that

of Zaretsky relative individual profile crowned between solution

to be identical.

geometries

normal

on a flat raceway exponents results solution followed form life were were and by

of 1.4, 1.9, and 2.4GPa life prediction determined obtained: not the considering end-tapered these produces edge standards

275, and 350 roller profiles

theoretical at 2.4 GPa

compared compared. 1. With longest tapered considered from 2. 3. The

to the ANSI/ABMA/ISO following

bearings

for the individual or stress the may profile is lives.

the resultant the flat roller and For the the

concentrations, aerospace there highest life profile

profile,

profile, the end which different

respectively.

full crowned concentrations profile

profile

the lowest differences profile profile

While not the

are life differences FEA lives at the but not higher

and the aerospace

profile, followed the flat

be significant.

the end tapered on roller

produced

significantly load by but as

that of the aerospace effect edge

by the crowned

and the flat roller

profile,

respectively. much on as the

loading

to reduce

98 and 82 percent at the lower load. The actual percentage calculated depends on the analysis used. The resultant predicted life at each stress condition not only depends on the life equation used Weibull Except predicted Harris slope assumed. The slope least variation in predicted slope of the Weibull slopes life with equation Weibull predicts slope comes standard the highest standard. in a maximum Lundberg that not consistent with At all conditions calculated for a Weibull equation. 1.t 1, the ANSI/ABMA/ISO of 1.5 and 2, both result lives,

also lowest

the Zaretsky in the the highest

equation. lives. are lives

for the Weibull by the Zaretsky r,, equal the Hertz n equal for a p of "line contact" of cylindrical (where upon

of 1.11 at which For Weibull predict lower

the Lundberg-Palmgren

and Ioannides-

0) equations stresses to 6.6.

lives than the ANSI/ABMA/ISO the load-life bearing can exponent that experience contact." This

4.

Based

for line contact, This value that other a roller roller

p of 10/3, results in the "mixed today. field. contact,"

Hertz

stresscan

life exponent justification experience majority

is inconsistent raceway "point

experienced

and

Palmgren's the vast

10/3 was and the roller

is, one raceway with

is certainly

and tapered

bearings

designed

and used

NASA/TM

2000-210368

14

Appendix

A--Derivation

of Weibull

Distribution

Function

As presented

in Melis

et al. (24) and according

to Weibull

(6) any distribution

function

can be written

as (A1)

F(X)

= 1 - exp -[f(X)]

where occurring

F(X)

is the probability

of an event as

(failure)

occurring.

Conversely,

from

the above

the probability

of an event

not

(survival)

can be written

1-

F(X)=

exp - [f(X)]

(A2a)

or

1 - F = exp -[f(X)]

(A2b)

where

F = F(X)

and (1

F) = S, the probability components, each not occurring

of survival. with a probability total of the event (failure) not occurring from being (1 as F), the of all components can be expressed Eq. (A2b)

If we have probability

n independent

of the event

in the combined

(1-F")=exp-[nf(X)]

(A3)

Equation generally, consisting "single" link has

(A3) for the

gives size

the appropriate effect on failures links. Also, to find chain

mathematical in solids.

expression

for the principle

of the weakest of Eq. of failure of n links, fails,

link (A3),

in a chain we assume assume of the

or, more a chain to a if one has that

As an example

of the application consisting is equal

of several link. failed

we assume the probability fails. In other

that by testing of failure words, if any

we find the probability F, of a chain single part (1 - F,,),

F at any load X applied we must the whole component simultaneous

If we want the whole

of a component

failed. Accordingly, the probability nonfailure of all the links. Thus,

of nonfailure

of the chain

to the probability

1 - F,, = (1 - F)"

(A4a)

or

S n = S ll

(A4b)

Or, where

the probabilities as

of failure

(or survival)

of each

link are not necessarily

equal

(i.e.,

$1 ::$2 _: $3 ...), Eq. (A4b)

can be expressed

S,, = S 1 .S 2 .S 3 ....

(A4c)

This is the same From

as Eq. (2) of the main

text. of stresses throughout a volume V (A5a)

Eq. (A3) for a uniform

distribution

F v = 1 - exp - [Vf(o)]

or

S = 1 - F v = exp - [Vf(o)]

(A5b)

NASA/TM---2000-210368

15

Equation can (A5b) beexpressed asfollows:

in

lnEll =

In fic,+lnV

(A6)

It follows variation function.

that

if In In(l/S)

is plotted

as an ordinate will imply

and In/'(_) only

as an abscissa displacement

in a system but

of rectangular

coordinates,

of volume l" of the test specimen Weibull (6) assumed the form

a parallel

no deformation

of the distribution

f(_)

(A7)

and Eq. (A6)

becomes

In

FiT
'nk_j = e In(c-_u)-e In ell3 +lnV (AS)

If r_,,, which written as

is the location

parameter,

is assumed

to be zero

and

V is normalized

whereby

InV is zero,

Eq. (A8)

can bc

,n : l ,nf
Equation The assumed (A9) form of is identical Eq. (A9) to Eq. (2) of the main where _,, is assumed text. to be zero is referred to as "two-parameter Weibull." Weibull.'" Where G, is not to be zero, the form of the equation is referred to as "three-parameter

NASA/TM

-2000-210368

16

Appendix

B--Derivation

of System

Life Equation

As discussed element following Front bearing

and

presented is based

in (24), first

G. Lundberg derived

and A. Palmgren between as

(4) in 1947, individual

using

the Weibull lives

equation and system

for rollinglife. The

life analysis, A, Eq. (A9),

the relationship equation

component

derivation Appendix

on but is not identical the Weibull

to the Lundberg-Palmgren can be written

(4) analysis.

(B1)

where

N is the number

of cycles ofa

to failure. plot in Fig. 7, the slope e can be defined as follows:

Referring

to the sketch

Weibull

xnloFxl_,ndt
L&y_J
e_

LSref J

(B2a)

In N - In Nre f

or

,nVI N c ,nIL1
Ls_fJ From Eqs. (B l) and (B2b)

(B2b)

,nI--'
and

Ls_.,,_J s,._fj L

IN]e

(B3)

Referring reliability to Fig. S. From 8, for a given time Eq. (A4c) tbr a series reliability Combining Eqs. (B4) and (B5) gives

Ssy s = exp-

(B4)

or life N, each system

component

or stressed

volume

in a system

will

have

a different

Ssy s =S I.S 2.S 3....

(B5)

_exp-F--] x_x_- x...

exp_ =exp_ LNI_2J

(B6a)

NASA/TM--2000-210368

17

or

exp- :ex_-_ +[N_ _ I +i_--7'J +"

It is assumed that the Weibull slope e is the same for all components. From Eq. (B6b)

r ,ct

_"_'

(B7a) [L N[31J L N_ 2 J L N_ 3 ]

Factoring

out N from

Eq. (B7a)

gives

From

Eq. (B3)

the characteristic component

lives N_,

N_,

N_), etc., probability

can be replaced of survival

with

the respective

lives

NI,

N_. F,5, etc.,

at S_r

(or the lives of each

that have the same

S_r) as follows:

r,n' 7_:r,oll[_]_ rlo;lr_ it, [ ,][,]c

L MjLUrefJ
where, in general, from Eq. (B3) ---

L srefj _

ln-_

L s_.fJLN3J

+...

(B8)

I,]e:r, llIll e
_ L s_JLN_,j
and tile 1][1] = I Ins-_e f _ to,etc. Factoring out In (1/S,_r) from Eq. (B8) gives

IB9a)

(B9b)

t
or rewriting Eq. (B 10) results in
I_ 1 to

Equation

(BI

1) is identical

to Eq. (21) of the text.

NASA/TM

2000-210368

18

REFERENCES

I. Anon., 2. Anon.,

"Load "Load "Rolling

Rating Association, Ratings Association, Bearings 1990.

and and

Fatigue Fatigue

Life DC, Life DC, Load

for 1990. for

Ball Roller

Bearings," Bearings,"

ANSI/ABMA ANSI/ABMA

9-1990, 11-1990,

The The

American American

Bearing Bearing for

Manufacturers Manufacturers 3. Anon., 4. 5. 6. 7. Standardization, Lundberg,

Washington, Washington, Dynamic

1990. Ratings and Rating of Rolling of Roller 1952. Ingeniors Vetenskaps Akademien, No. 153, 1939. Applicability," Journal of Applied Mechanics, Life Distribution of Rolling Bearings," Journal 18, Rolling 3, 1951, Life," ISO 281:1990(E), Acta International Organization Engineering The Royal

G. and Palmgren,

A. "Dynamic

Capacity Capacity Stockhohn, in Solids," of Wide

Bearings," Bearings,"

Polytechnica

Mechanical No. 210,

Series, 1,3, Stockholm, 1947. Lundberg, G. and Palmgren, A., "Dynamic Swedish Weibull, Weibull, Academy of Engineering Sciences, of Rupture Distribution tbr W., "The Phenomenon W., "A Statistical 297. W., "Efficient

Handingar

Proceedings,

pp. 292 8. Weibull, 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 24. 22. 23. 25.

Methods

Estimating

Fatigue

Contact 107, 3,

Phenomenon, edited by J.B. Bidwelt, Elsevier, New York, Ioannides, E. and Harris, T.A., "A New' Fatigue Life July 1985, pp. 367 E.V.. 378. Criterion E.V., for Life, J.V., 1996, Life to System "Incorporating 1993, pp. 706 and Peters, Plan for Death," pp. 237 Design, Finite 710. Machine Zaretsky, Jan=Feb. August, Zaretsky, Zaretsky, STLE pp. 339 Zaretsky, Palmgren, "Fatigue

1962, pp. 252 265. Model for Rolling Life and Reliability," into

Bearings," Journal Component 1994,

of Tribology, and Power,

of Propulsion Life and pp. 55 for 59.

3, 1, Journal

1987, pp. 76,83. R., and Zaretsky, Design, "Design 39, 2, April E.V., E.V.,

Element

Analysis

Reliability,"

of Mechanical

115, 4, Oct.

Design,

66, 15, Aug. of Life des

Poplawski, Service Palmgren

S.M.,

"Comparison 1996. Zectsckrift

Theories

Rolling-Element lngenieure, Engineering, 68,

Bearings," 14, 1924, 54,

Trans., 341. E.V.,

248, 501-503, Bearings," Basis

A., "The "A.

of Ball

Vereines

Deutscher Lubrication 1946. Bearings,"

Revisited-A

for Bearing Vols. Capacity

Life Prediction,"

J. STLE,

2, Feb. 1998, pp. 18 24. Jones, A.B., Analysis of Stresses Rumbarger, Technology, Parker, R.J., J.H., 90, and l, Jan. Jones, 1998, pp. E.V., Theory

and Deflections. A.B., 130 "Dynamic 138, 1968.

1 and 2. New Departure, of Oscillating Life Ball Relation Rolling

Journal

of

Lubrication NASA Load TN and

and Zaretsky,

"Reevaluation for Elastically

of the Stress Constrained

in Rolling-Element Roller Bearings 1964.

Bearings," Under Arbitrary

D-6745, 1972. Jones, A.B., "A General

and Radial

Speed Conditions," Journal of Basic Engineering, Johnson, L.G., The Statistical Treatment o/Fatigue Palmgren, Industries, Zaretsky, Zaretsky, Residual Melis, Reliability," A., Ball and Roller 1959. Criterion R.J., E.V., to System Fatigue," and August, Trans. Bearing Philadelphia, E.V., "Fatigue E.V., Stress M.E., Parker, Zaretsky, J. Propulsion

82, 2, April 1960, pp. 695-703. Experiments, Elsevier, New York, First Ed., SKF Industries, Primer,

En'gineering, Design, W.J., R., AIAA, and NASA

Philadelphia, NASA TM-

1945. 106022,

Third 1992.

Ed..

SKF

Life and Reliability-A Miller, S.T., "Effect 1965. Analysis 1999, TN D 2664, 15, 5, Sept.-Oct.

Anderson,

of Component of Aircraft 666. Gas

Differential Turbine

Hardness Disk Life

on and

and Rolling-Contact and Power,

"Probabilistic

pp. 658

NASA/TM

2000-210368

19

TABLE [Roller

I.--MAXIMUM HERTZ STRESS AS FUNCTION OF NORMAL LOAD AND ROLLER PROFILE Dia., 12.7 mm (0.5 in. ): roller length, 12.7 mm (0.5 in.).] la) Constant normal load I

Nomlal load, P, N. llbs) 4239 (953) 8016 (1802) 12980 t2918)

Flad

[ End taperedZ [ Aerospace-" [ Crowned Maximum Hertz stress, GPa, (ksi) 1.84 (267)
_._9

1.38 1.53 1.57 (200) (222) (227) 1.90 2.02 2.05 (275) (293) (298) 2.41 2.52 2.54 (350) (365) (369) (b) Constant Hertz stress load. P, N, (lbs)

(333) 2.76 (400)

Maximum Normal Hertz stress, GPa, lksb 1.38 4239 3327 (200) (953) (748) 1.9 8016 6993 (275) (1802) (t572) 2.4 12980 11824 (350) (2918) (2658) Based on closed form solution. -'Based on laminated roller analysis.

3158 (710) 6699 (1506) 11521 (2590)

1557 (350) 3955 (889) 8154 (1833)

TABLE 2.--MAXIMUM HERTZ STRESS-LIFE EXPONENT FUNCTION OF WEIBULL SLOPE FOR FOUR LIFE EQUATIONS Equation Weibull slope Stress-life n Line contact 6.6 10,2 7.5 5,7 8.1 6.0 4.5 8.1 6.0 4.5 9.9 9.7 9.5 , equal 0. Point contact 9 11.1 8.2 6.2 9 6.7 5.0 9 6.7 5.0 10.8 10.3 10.0 Line contact 3.33 5.1 3.8 2.9 4.1 3 2.3 4.1 3 2.3 5 4.9 4.8 exponent, Load-life

AS

ANSI/ ABMA/ISO Weibull eq. (10) LundbergPalmgren, eq. (19) loannidesHarris, eq. (26) Zaretsky, eq. t31) LNo fatigue

1.11 1.11 1.5 2.0 1.11 1.5 2.0 1.11 1.5 2.0 1.11 1.5 2.0 limit assumed,

exponent, P Point contact 3 3.7 2.7 2.1 3 2.2 1.7 3 2.2 1.7 3.6 2.2 3.3

NASA/TM--2000-210368

20

TABLE 3.--COMPARISON LIFE OF RELATIVEFROM FOUR EQUATIONSLATOLLER LIFE FOR R F WITH NO LOADING EDGE t
Maximum Hertz stress, GPa (ksil-" ANSI/ ABMA/ Weibull Relative theoretical life Lundbergloannides] Zaretsky Palmgren eq. (19) slope, 1.11 16 2.2 slope, 1.5 Harris eq. (26)_ ] eq. (31)

ISO Standard 3 eq. (10) Weibull 4.9 1 79 6.4 Weibull

1.9 (2751 2.4 (350)

16 2.2

65 5.9

1.9 (275t 2.4 (3501

4.9 1

,o 12, 2,o i.l

9.9 1.6 3.1 0.7 3.1 0.7 4 2.4 0.3 551 5.3 971 47 4.8 1.1 l. ,, equal 0.

Weibull slope, 2.0 1.4 (200) 40 15 4 1.9 (2751 4.9 2.4 2.4 2.4 (350) l 0.6 0.3 )mmlized to 4.14 GPa (600 ksi). :Refer to Table 1 for values of load. _Based on Weibull _No fatigue slope equal limit assumed,

TABLE

4.--EFFECT

OF FATIGUE LIMIT ON RELATIVE LIFE OF FLAT ROLLER LOADING USING IOANNIDES-HARRIS EQUATION t [Weibull Slope, 1.11.] Relative theoretical life, L, and resultant stress-life exponent, Closed form soluuon Fatigue 0 138 10_'(20) limit .... 276 GPa (ksi) 10-(40)

WITH

NO EDGE

Maximum Hertz stress, GPa (ksi)Z

ANSI/ABMA/ISO standard

414

10-_(60)

L
1.4 (200) 1.9 (275) 2.4 (350) 40 4.9 1

'n
6.6 6.6 --

Ll n
209 16 2.2 8.1 8.1 --

Ll'n
9074 209 16 11.4 10.7 - 5.7

L
eq. (261 18.4 15.7 10 _ 7545 190 analysis 20.5 27 -limit. 9.7 2.8

L
% l0 n 4e 25.6 5815

(a) loannides-Harris,

(bt Finite element 1.4 (200) 1.9 (275) 4.0 49 6.6 6.6 661 37 12.2 14.1 8267 215 16 19.1 1.5 l0 e

106

27.8 38.5 - - -

"_'_ _,_4 10 a fatigue

47165 22

2.2 -4.7 - 2.4 (350) 1 -tNormalized to maximum Hertz stress of 4,14 GPa (600 ksi) without -'Refer to Table 1 for values of load. 3Normalized to maximum Hertz stress of 2.4 GPa (350 ksit. 41nfinite life,

TABLE

5.--EFFECT [Weibull

OF ROLLER slope,

PROHLE

ON RELATIVE _,l

LIFE I

1.11, critical

shear stress,

Maximum Hertz stress, GPa, (ksi)" 1.4 (200) 1.9 (275l 2.4 (350)

ANSI/ ABMABSO standard Life 'n 40 6.6 4.9 1 6.6 --

(a) Closed

form solution

without Hat Life 1509

edge loading Zaretsk_, (eq. (31)} End ta _ered Aeros _ace Life 3n Life _n 537 9.9 431 9.9 34 3.9 9.9 -29 3.5 9.9 -Crowned Life 3n 86 9.9 9.7 1.6 9.9 --

Lundber_-Palmgren (eq. (19)) Hat End tapered Aerospace Crowned Life _n Life _n Lile 3n Life _n 209 8.1 90 8.1 75 8.1 20 8.1 16 2.2 8.1 - 9.5 8.1 8.3 8,1 1.6 -1.5 - (hi Finite Element Anal;'sis

_n 9.9

3.4 8.1 65 9.9 0.8 -5.9 -(FEAI with edge loading

1.4(2001 144iO 16.6 71 15.51630 33 13.9 528 1.9(275} 6,6 1.2 1.5,9 17.9 33 2.4 (350} -0.05 0.93 --3.5 _Normalized to maximum Hertz stress of 4.14 GPa (600 ksi). :Table 1 for values of load. 3Normalized to maximum Hertz stress of 2.4 GPa (350 ksi).

10.'7 11.2 ---

96 13 0.65

10.7 15,2 ---

296 4,9 0.19

15.7 16.3 ---

824112.6[756 43 14.9 2.2 ---

35 5.4

10.6 9.3 ---

1211128.6 12 8.7 2.2

NASA/TM--2000-210368

21

TABLE 6.--EFFECT OF EDGE STRESSES ON RELATIVE LIFE OF FLAT ROLLER BASED ON FINITE ELEMENT ANALYSIS j Maximum Hertz stress, GPa (ksir" [Weibull Slope, 1.11 : critical shear stress, 45.1 ANSI/ABMA/ISO ILundberg-Palmgren Zaretsky standard (eq. (19)) (eq. (31 )) Relative theoretical life, L, and resultant stress-life exponent, n No edge stresses 1.41200) 24 50) 1.4i200) 1.91275) t 49 40 i 40 6.6 -6.6 6.6 37 667 22 71 12.2 --14.1 I 16_228 I 33 I I 14.1 13 15.7 16.3

Edge stresses 15.5 I 296

1.9 (275} 4.9 6.6 1.2 15.9 4.9 2.4(350) 1 -0.05 --0.19 Normalized to maximum Hertz stress of 4.14 GPa (600 ksi}. :Refer to Table 1 for values of load. 'Normalized to maximum Hertz stress of 2.4 GPa (350 ksil.

Outer ring Shoulder --_/_

_ _.r_

_t" . . _

Flange Roller

I_
l_

Total length, It

LI

Flat length,

If

I/] ff-_ jr/i, ,//__

-"-- Inner ring uore. _ Raceway

F
Crown drop I
/

. _

radius Crown

lil,

Blend point ---"

Diameter

/-- No / Contact stress Figure 1.---Cylindrical roller bearing with nonlocating inner raceway. Bearing accommodates axial movement by not restraining rollers axially on inner raceway. Similar beadng with flanged inner ring allows axial roller movement on outer raceway. Figure 2.mRoller // / crown /

,-- Crown

I
Length profile influence on stress pattern.

NASA/TMI2000-210368

22

It

_-_

p_l-_lf_

"-- Eclge loading

/f -_

14

,-- Contact

geometry

.-- Contact

geometry

Figure 3.--Roller

profile types and Hertzian crown

contact

geometry.

(a) Flat roller profile. (b) Tapered

crown roller

profile. (c) Aerospace

roller profile. (d) Full crown roller profile.

NASA/TM--2000-210368

23

Contact ellipse

Figure 4.--Schematic

of loaded

crowned

roller on race.

Roller--,
\

Figure 5._Quarter

section finite-element

model of roller-race

contact.

NASA/TM--2000-210368

24

GPa (ksi) 0.43 (62)--0.32 (46) .... 0.1 (1)---

o.21 (31)--0.19 (27) -0.13 (19) _0.10 (15) --

0.08 (11) - _. 0.05(7) --_ 0.02 (3)_(a) _

0.57 (82) -0.71 (103) ..... 0.85 (123) 1.1 3 (164) _ _ _ _ _ 0.99 (143)_ 0.77 (112) -

-_

(b) Figure 6.--FEA stress profile of quarter section of raceway for roller with aerospace shear

(partial) crown. Maximum Hertz stress, stress. (b) Von Mises stress.

1.9 (275) GPa (ksi). (a) Orthogonal

NASA/TM--2000-210368

25

i- N -]e Ssys = exp - I N_ I Ssys / /

/
.- lu)
rE

/
Sp _arc Sref Itar e

Nref

Np In N

Figure 7.--Sketch or tangent of 36.8 percent

of Weibull

plot where (Weibull) slope of survival

of line is e. SI_ is probability

at which N = N_ or NINI3 = 1.

Ssys

$1
v-103

/
/

/ /

-= c

$2

$3 Sref

///
Fig. 7)_-_//I//_ Nre f N 1

/
Ssys=S 1 .S 2-S 3....

Ili N2 N 3 N In N plots where each numbered of each component distribution in system of plots slope e.) combined

Figure 8.---Sketch plot represents 1,2,

of multiple cumulative

Weibull

distribution

and system Weibull

plot represents

3, etc. (All plots are assumed

to have same Weibull

NASA/TM--2000-210368

26

REPORT

DOCUMENTATION

PAGE

FormApproved
OMB No. 0704-0188

Public reporting burden for this collection of information is estimatedto average1 hour per response,including the time for reviewing instructions, searching existingdata sources, gathering and maintaining the data needed, and completing and reviewing the collection of information, Send comments regarding thisburden estimate or any other aspect of this collection of information,including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204. Arlington. VA 22202-4302, and to the Office of Managemenl and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 13, REPORT TYPE AND DATES COVERED

August 4. TITLE AND SUI_TITLE of Roller Profile on Cylindrical Roller

2000

Technical

Memorandum NUMBERS

5. FUNDING Bearing Life Prediction

Effect

6.

AUTHOR(S)

WU-253-02-98-00

Joseph

V. Poplawski,

Erwin

V. Zaretsky,

and

Steven

M.

Peters

7.

PERFORMING

ORGANIZATION

NAME(S)

AND

ADDRESS(ES)

8.

PERFORMING REPORT NUMBER

ORGANIZATION

National John H.

Aeronautics Glenn Ohio Research

and

Space Center

Administration at Lewis Field E-12037

Cleveland,

44135-3191

9.

SPONSORING/MONITORING

AGENCY

NAME(S)

AND

ADDRESS(ES)

10.

SPONSORING/MONITORING AGENCY REPORT NUMBER

National Washington,

Aeronautics DC

and 20546-0001

Space

Administration NASA TM--2000-210368

11. SUPPLEMENTARY Prepared Nashville, 528 Erwin

12a.

NOTES the 2000 Annual May Bethlehem, organization Meeting 7-11, 2000. sponsored Joseph by the Society and of Tribologists Steven NASA M. and Peters, Lubrication J.V. Poplawski Center. Engineers, and Associates, person,

for

Tennessee, New Street,

V. Poplawski Erwin

N.

Pennsylvania: code
STATEMENT

V. Zaretsky, I.

Glenn

Research

Responsible

V. Zaretsky,

5900,

(216)

433-324

DISIHiBUTION/AVAILABILITY

12b. DISTRIBUTION

CODE

Unclassified Subject

- Unlimited Categories: 37 and 39 Distribution: Nonstandard

This
13.

publication

is available
200

from
words)

the NASA

Center

for AeroSpace

Information,

(301)

621--0390.

ABSTRACT

(Maximum

Four solution

roller and

profiles finite and and fully

used element

in cylindrical analysis loaded These edge tapered

roller (FEA) against were loading profile those for

bearing stress a flat

design and raceway. life.

and The Four Lundberg

manufacture roller profiles

were

analyzed

using were life flat, models and

both

a closed end, chosen and

form

analyzed bearing Ioannides edge

tapered were

aerospace, this The much profile. depends Palmgren stresses 6.6. This for analysis flat

crowned

rolling-element and life. lives resultant assumed. but Palmgren, However, not

for

compared. without The crowned life equation end

of Weibull, the longest the lives.

Harris, can reduce than

Zaretsky. life by as

roller

profile percent. fully the

has

predicted highest The

loading

as 98 The on

produced the on lowest

significantly life

different at each of stress

the

aerospace not only

profile used

produces but equations also

predicted For Weibull

condition 2, both upon

the Weibull lower exponent

slope lives of

slopes

1.5 and Based

Lundbergthe Hertz to

and line

Iaonnides-Harris contact, the

predict load-life

than 10/3

the results

ANSI/ABMA/ISO in a maximum

standards. Hertz stress-life

accepted with that

exponent

equal

value

is inconsistent

experienced

in the

field.

14. SUBJECT

TERMS bearings: Life prediction methods: Rolling-element fatigue:

15. NUMBER
16. PRICE

OFPAGES 32

Rolling-element Stress 17. analysis

CODE

A03 SECURITY CLASSIFICATION OF REPORT Unclassified NSN 7540-01-280-5500 18. SECURITY CLASSIFICATION OF THIS PAGE Unclassified 19. SECURITY CLASSIFICATION OF ABSTRACT Unclassified
Standard Prescribed 298-102 Form by ANSI 298 Std. (Rev. Z39-18 2-89)

20.

LIMITATION

OF ABSTRACT

ERRATA NASA/TM--2000-210368 EFFECT OF ROLLER PROFILE ON CYLINDRICAL ROLLER BEARING LIFE PREDICTION

Joseph V. Poplawski,

Erwin V. Zaretsky, August 2000

and Steven M. Peters

1
Page 7, equation (26): Replace with -Snax

1
--maxSn( ru )

c+l
Page 8, equation (32a): Replace with c +-e

1
e

c+2
Page 8, equation (32b): Replace e

with c + -e

Page 20, table 2: In the Load-life the 2.2 with 3.4

exponent,

p, Point contact column

and the Zaretsky,

eq. (31) row, replace