Vous êtes sur la page 1sur 6

This article was downloaded by: [RMIT University]

On: 06 August 2013, At: 13:19


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

A S L E Transactions
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/utrb19

Surface Roughness Effect on Slider Bearing


Lubrication
a

S. T. Tzeng & Edward Saibel

Department of Mechanics, Rensselaer Polytechnic Institute, Troy, New York


Published online: 25 Mar 2008.

To cite this article: S. T. Tzeng & Edward Saibel (1967) Surface Roughness Effect on Slider Bearing Lubrication, A S L
E Transactions, 10:3, 334-348, DOI: 10.1080/05698196708972191
To link to this article: http://dx.doi.org/10.1080/05698196708972191

PLEASE SCROLL DOWN FOR ARTICLE


Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy, completeness, or suitability
for any purpose of the Content. Any opinions and views expressed in this publication are the opinions
and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of
the Content should not be relied upon and should be independently verified with primary sources of
information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or
indirectly in connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or
systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution
in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at
http://www.tandfonline.com/page/terms-and-conditions

" :'I.E TH.\ X".\CTIOX:'

10, 334- 338 096i)

Surface Roughness Effect on Slider Bearing Lubrication


By S. T. TZEKGI and EDWARD SAIBEU

Downloaded by [RMIT University] at 13:19 06 August 2013

Th e effe ct of roughn ess of surf aces of a slider bearing on the lotul-carruinu capacity and fri ction fo rce, is
studied. Th e distinction is made between wavin ess and roughn ess, the loiter being treated as a randoln
quantity whi ch may be characterized by a probability density j unction, determin ed ex peri mentally.
R esult s can be obtain ed in closed form sui table[or calculation ij a B eta distribution is a.s.~lIIned . A ctuollu,
thi s can approximate quit e closely a Gaus sian distribution, yi eldin(J quit e accurate result s. Ho wever, the
procedure deiclopeil can be carried out for an y desired di stribution ,

Introduct ion

Solution of the problem

IT has been conjectured for many years how surfa ce


roughness aff ect s the pressure d evelopment , loadcarrying ca pac ity, friction forc e, and oil flow in fu lly
lubrica t ed bearings. So far the on ly quantitative
results known to the authors have been obtained for
sinusoida l variations of film th ickness (1 )- (3). These,
how ever, are more readi ly associated with surface
waviness than with su rface roughness. Incid entally, it
would a lso be reali sti c and of interest to conside r the
combin ed effect of waviness and roughness.
Loca l roughness is in general much more com plica ted
than waviness and may produce significa nt ly different
effects on the paramet ers than waviness, particu larly
wh en th e d eviations in film thickness arc random and
may be about th e same ord er of magnitud e as the film
thi ckness it self.
It is this last property, that of the d eviations in film
th ickness b eing of t he sa m e ord er of magnitude as the
film thi ckness itself, that introduces difficulties in th e
treatment of this random process. Ordinarily in a
st ochas t ic process it may b e assumed that the random
deviation from an id eal paramet er is very small compared to th e paramet er it self and conseq uent ly a
perturbation type of process is justifi ed . In the case of
surface roughness in bearings, however , t his is no t th e
ease.
In th e presen t work, an analytic treatmen t based on
random surface roughness is presented . A numeri cal
exa m p le based on a roughness ch a rac te rized by a Beta
di stribution ha s b een includ ed to demonstrat e the ord er
of magnitud e of the effects produced but th e general
t reatmen t is not limit ed to thi s spec ific density fun ct ion .

Consider a wid e slide r bearing ill whi ch th e cha nge ill


oil pressure along the width, dp jdz, is negligible
(F ig. 1). Th e R eynolds equation is th en reduced to (.n

!i
(IPdP)
dx
dx

= AI dH

[see Append ix I]

d:c

[1]

wh er e p = p (x ) is the pressure,
H = H( x ) is the film thickness,
!J. is the viscosity of lubrica nt (cons ta nt ),
U is t he v elocity of th e slider,
a nd
A[ = G,u U .

A simp le int egrat ion of Eq. [1 ] yields


rip _ '\ 1
rlx - I/ ~

.\~

l~j

I p'

tl

791...----- - - X

1 Gra dua te Assist ant, D epartment of Mechuni cs, l len s-elaer


Polyt echn ic In sti tut e, T roy , New York . Presen t add ress: Westinghouse B ettis Atomi c Power Lab oratory, West Miffl in , P a ,
2 P rofessor of Xle ch uui cs, D epartment of Xlechuuics, Rensselaer
Poly techni c In stitut e, Troy, N ew York. Presen t address: Professor of Appli ed Me chani cs, Ca rneg ie-Me llon Univers it y. Pit t8burgh, Pa .

'1'"7''----,~...y.-.,r___'7''~.

I
FIn. I. Slider beuriug.

334

Su rf ace Roughness Effect on Slide r L ubrication

101 (Tl) d,

wh ere 1\2 is an integration cons ta n t. Let e denote the


random roughness d ev ia t ion from the mean b earing
sur fac es, and h ex ) th e nominal film thickness d efin ed
by th e m ea n su rfaces. Th en on e ca n writ e
H (x )

= hex )

Downloaded by [RMIT University] at 13:19 06 August 2013

and

p(l )

O.

rI(p )

[41

rlx

with corresp onding boundary cond itio ns


=

O.

(p (.r )\=/

()

[i)]

to be satisfied . Th e sy m bol ( ) represent s expected


valu es. Noticing that the m ember s on th e right-hand
side of Eq. [4] are expec te d values of functions of the
random variable f , it is seen t hat.

/ _1_" ",

""H n/

f_"" (h +

f (f )

df

(n = 2,3) .

f) "

[ 12)

(T2> d.

Cons ider a p lane-slider b earing of infinit e wid th. Let


t he ro ughness of the b earing su rfaces b e rep resen t ed by
a Beta d en sity function of the form
- c

<

e :::;

[3]

F or a fixed :1:, the above processes rep resen t t wo


random variables. Let t he random roughness variable f
b e cha racterized by it s probabi lity d en si ty Iunction j'(s).
Taking exp ected va lues of th ese va riables in Eq . [2], on e
finds

(p (.r ) )J:~ o

10

[ 11]

Example

f.

If one t reats th e right-hand side of Eq. [2] as an input


s to ch astic process, th e pressure di stribution p (x ) ca n be
regard ed as th e st ochastic output of a linea r sys te m .
Th e output p rocess must sa t isfy two boundary cond itions a t the ends, namely ,
p (O)

335

[ 13]

or in t erms of t he variance u 2 ,
1i)

f(f)

1G(7 1/ 2 ) U

(1
[141

O. else where .

Roughness is fr equently regard ed as bein g cha rac te rized by a Ga uss ia n di stribution although no publish ed
data exists so far to subs ta ntia te this notion in a n exac t
se nse . It h as recen t ly come t o the authors ' a ttention
that 'W illia mson (5) ha s verified th e G a uss ia n di stribution in a wid e vari et y of cases. Howev er, using th e B eta

[()]

Unit Normol Dist r ibut ion

j\

f=...L e - e I2
../2"";

Upon d et erminatio n of th e right-hand side of Eq . [4]


by Eq . [fiJ, the exp ected value of the p ressure as a
functio n of x is obtained hy solv ing t h is differential
eq uation. Sym bolica lly ,

Unit Beto Di str i bu tion


15

[7 ]

Th e integration cons ta nt A2 ca n now be d et ermined by


th e second cond it ion in [;''i].
I nt egra t in g th e fina l exp ression for (p(:t:) ov er t he
bea r ing su rface a rea , the expect ed va lu e of th e total
load-carry in e; capacity per unit wid t h becomes

fl
II" = l (p Cr

[8J

dx,

Ba sed OIl t he sa me assumptions and rea soning as


before, the sh ea r ing st resses are giv en by

cn

7~

I.d '

H rip

If

2 d.r

- IT - :2 dx
_ !J.e + !!. dp

on t he shoe.
on th e slider,

[!J ]
[10]

wh er e dp j dx ca n be rep la ced by t he exp ressi on [4]. Th e


ex pected value of th e tota l fri ctional force is lik ewise
ob tai ned a s

f = 16.rr

0 ,2

(I-

e /7 )

0 .1

-3

-2

- I

FH i .2 . Co m pa riso n of di st ribu tions .

di stribu t ion as an approximation will lea d to ac curat e


resu lt s since the analysis is no t sens it ive to an approximation of the d istribution fun ction . The unit B eta
d en sity function is compared with the un it Gaussian
distribution in Fig. 2.
Using Eq . [13] in Eq . [eiJ a nd th en Eq . [4], then '
resu lts

S.

336
15 { AI [
(p (r)
x -_ Hk"

+ 4C"/I]

- .\ "

1(j c3..r + 3m
8c (13
- 3"
t -

[~ (h" -

hoe) - st,

T . TZ E X G AX D EDWARD SAIBEL

h 03)

In t he case of smoot h surfa ces, using results from


Sh aw and Mack s (6), t he correspond ing values have
been obtained as follows:

41 3

+ 2czIo]} ,

[I G]

p(l ) = 7,900 psi;


FI

wh ere ho is t he minimum hex), m . t he in clinati on of the


shoe and

1
x

I ,,(x )

h In

G~ ~ ~) dx

(see Appendi x II) .

If' = 1,5,a {AI[_ ~ c3lz + ~

Downloaded by [RMIT University] at 13:19 06 August 2013

4./ 3

+ GJ

+ 4C

Z
'/ I ]

3 -

wh er e hI =

3m

(h

14

4m

h0

[17]

+ Az [_ ~ (h 3~ h
13

ho:J l)

021)

Jo]}'
h o + ml is t he ma ximum h (:r) and

2cZ

[1 8]

(see Appendix II ).

[19]

Sim ila r ca lculations for t he fri ctional for ces from


Eqs. [9] and [10] y ield

F\

~;~o ["~I if/I + ~2 if/2 }

[20]

F1 ~ - lu}'5.
[ 2A1 . /. + A2 . /. ]
e;) 3 '1'1
2 '1'2
-

[21 ]

'

wh er e

if/l = 14 (1) - 2c2 1z(1)

+ c410(l) + :~ (h

2
1 -

- ~ (h14

2m

h02)

h0 4)
.

[22]

+ 4c"II (l).

[23]

a nd
if/z = -

1()

c31

+ 8c
(h 3 m 1

1zo3) - 4 / 3(1 )

A set of cu rve s ca n be plo t t ed fr om t he for egoing


results wh en so d esired. For the presen t di scussion , it
is sufficient to com p ute resul t s for a sing le numeri cal
case. Co ns ide r a n exam ple with t he fo llowing data :
JJ.

R E F ERE N CES
1.

3.

and

4.

= 0.0003 in ..

T he res ults of t he n ume rica l com p uta t ion from Eqs,


[16], [18J, [20], and [21] are

(pO =

10,680 psi;
FI = 5.53 Ib ;

TV = 11,250 Ib :
8.28 lb.

r , ==

A com pa rison of the for egoing t wo se ts of results


reveals ra ther int eresting effects of surface roughness.
Both the load-carrying ca pac ity and the fri ctional
for ces are incr eased cons ide rably wh en su rface roughness is taken in to account. The in crease in the total
load-carrying ca pacit y is more significa n t than that of
frictional for ces, however. This leads to a lower average
coefficien t of fri ction (defined by the ratio F/ TV), in the
case of rough su rfaces than in the case of smoot h surfa ces. To be more exact, the average coefficients of
friction in the cas e of rough surfaces are 0.000675 on the
slide r and 0.000370 on the shoe, whereas those releva nt to smooth surfac es are 0.000735 and 0.000591,
respec t ive ly .
It must be pointed out her e that the stochastic
t reatment of the problem is by no means limited to the
application of a Beta di stribution. For other types of
di stributions, however, on e might find it difficult t o
ob tain solutions in closed form. In t hat event, numeri cal
tech niques would be justified. The Beta di stribution,
wh en applied to t his type of problem, offer s spec ia l
meri ts over other distributions. For instance, singula riti es in the in t egrand in Eq. 16] whi ch may exist for other
di stributions ca n be avoided if a B eta di stribution is
em ploye d . It is felt that t here are probably a great
many practi cal problem s in wh ich the su rface roughness
di stribut ion may be approximated by a Beta di stribution with reason able a ccuracy . This may be achi eved
eithe r by adjustin g t he power s of the d ensity functions
or t he va lue of t he maximum d evia tion c, or bo th . I t is
a lso possible to have asymmetrical d ensit y functi on s
wh en need ed .

2.

= 5 X 10- 6 R eyn. ;

U = - 320 in. / sec ;


m = 0.000145 (inc lination of t he shoe is 5 min) ;
l = 2 in.;
c = 0.000 2 in. ;

ho

6 .65 Ill;

Discussion and Conclusion

Th e t ota l load ca n now be ca lculated from Eq. [8] as


1lJc

IF =

14,GIO Ill/in. widt h ;


F2 = 9.84 lb .

5.
0.
7.

MI CH ~;LL , A. G . M. , "Lubrica tion ; Its P rinciples and


P ra cti ce," I3lackie, London , 1950.
CITRON, S. J., "S low Viscous F low Between R ota ti ng
Co ncentric I nfinite Cy linder with Axia l R ou ghness,"
J . A pp], M ech., Tran s. ASM E , \'0 1. 29, Ser, E (962 ) , 188192.
B UIlTON, R. A., "Effect of T wo-d im ensional , Sinusoidal
R oughn ess on the Load Support Characteristics of a
Lu brican t F ilm ," J . Basic En g., T rans. ASME, SeT. D
85, ( 1963), 258- 264.
BISSOI' , E. E ., and ANllEIlSON , W . J. , " Adva nce d Bearin g
T echnology ," NASA , Washington , D . C ., 1964.
\VILLIA.\1S0N, J . B. P ., (priva te com muni cation ).
S HAW, M . C ., and Mxc x s , E . F ., "Analysis and Lu brica ti on
of Bearings." M cGraw-Hill, Ne w York , 1949.
PAPOULIS, A., "P roba bi lity, R an dom Variables und Sto chastic Processes." M cGraw-Hili, N ew Yo rk, 1965.

337

Surface Roughn ess E ffect on Slider Lubri ca tion


Appendix I

./ o(x ) =

The gen eral equa t ion cont ains t he te rm 12j.L (aH /at )
on t h e righ t-hand side. This term contrib utes no thin g
howev er as may be seen from the following. Int egrating
with resp ect to x leads to a t erm a/at foX H dx. However ,
H

h ex )

I/21'a

2~ [/ (.1' ) -

Ilia -

!J4'l )

+ ~ (yl -

1
J 2(x ) = -3 [/3(X) - h3(0 )x In
In

Downloaded by [RMIT University] at 13:19 06 August 2013

under t he as sum ption of a stationa ry process. Furthermore, the remaining t erm a/at fo Xh(;r) dx is zero since h
is a function of :1; a lone .

+ Ya
(Y13 + Y2a -

X (Y2 4
X

Appendix"

!J1

Ya3

Y4

c
+ 3m
~c X(Ya2 + Y42) +
.ld )

In

+ c In (YY2Y3
IY4)

- 2mx } ];

and

+ c,
h (O ) + c,

/l i = h ex )

Y2 = h ex ) - c,

Ya

//4

+ 3cm X2] ;

~m

+ 3" (!Ja'l -

Y43) _

yl)

(~)

(~)
+ -j
';
Y4
X

X {h eX) In (Y IY2)
!J3Y4

Let

h(~)J}] ''

- 2[ (h (:r ) -

c2 / 0 (x) - W( D) - dx In

+ (j~n (.lh a + Yaa -

a nd

(~)
+ ms. {heX) In(.'.YI!J2)
Y4
YaY4

h (O):r In

+ c ln ('YIY4 )
./1(;1') =

+ t (x ,t)

~In [/I (X) -

./ 3(X)

= h (O ) - c.

Then the followin g exp ressions were obtained for t he


I n's and .I n'S:

i- [/4(X) ~nl

c4/ 0(x) - W (D) - c4)x In

(~)
Y4

+ 2~m 1/12 + y } - YI" - Y4 1 + ~ (Ya Y4


-1c
+ 3mc (/li + Y2 4 - Y34 - Y4 4) - 3"
x( y} + y4'l )
+ ~ (Y2: +
!J13 - !J43) + 3c2:r (Y32 - yn
5

l o(x )

In

[h (.T) In

a)
h (O) In ( !J
y,

(.If..!-)
Y2

1
[ha(x ) In (.If..!-)
m
Y2

= -3

+ ~ (Y2 a + Yaa -

2 2

Ia(;r)

1
-4
In

[h 4(X) In (~)
Y2

+ (Y24 + Ya

h(O ))

(~)
Y4

Yla - Y4 a)

+ 32c (Y12 + Y22 -

YI 4

+ 3c2(.lh2 + yl

!J44)

yl - yn
h 4(0) In

+ ca In (Y I.lh )J ;
YaY4

(lE) + c In ( lliY4
Y2Ya)
4

.1/4

+ ~c (!Jla + !J2'l -

- YI2 - Y42)

] .

DISCUSSION

Y42)

.,

//l- -

ha(O) In

y 3'l -

+ -1mC X

+ -1c(h (x ) 12 (x )

(~)
!J4

II (x ) = _1 [h 2 (X) In (YI) - h2 (0 ) In
2ln
Y2

!J IY4 + :21 (Y2-., + /N0' + c-., l n (/hYa)

+ c In (YIY2)]
;
YaY4

!J} - Y4a)

+ 8ca(h( x ) -

F. F . L I X G (Rensselaer Polylechnic I nsi.; T roy, New Yo rk ) :


The a u thors are to be complime nted for ha vi ng brough t for th
ran do m analysis to bearing problems. This is an approach wh ich
will receive a good deal of a ttention in t he fu ture as it ha s in
ot her field s (e.g., information).
This d iscusser woul d like to ha ve seen m ore parametric stud y
num erically . Also, an exam ple com pa ring the effect of roughness
a nd wa viness would hav e been illumina ting.
There a re two q ues tio ns whi ch t his di scusser wishes to ask :
(1) Wha t is t he justifica tion of the stateme nt that the der iva ti ve
wit h respect to an argume nt of the " expected value " of a fu nction
is th e same as th e " expec ted va lue " of t he derivative of the
fu nction ?
(2) Wha t. ha pp ens, p hy sically spea king, a t the wa ke of
per tuberances? Tha t is, classical lu bri ca ti on theor y has been used '
thro ughou t. But what is the pe rmi ssibility of this ap plication to
a more com plicated sit ua t ion where the re a re , so to spea k, these
alternatively convergen t and diver gen t " cha nnels"?

h (O) ] ;
A UT H ORS ' C L OSU R E :

~
[h (.I') In ( YI)
;) In
Y2
O

+ (h2(x)

- h"(O ) In

(~)
+c
Y

- h 2(0 ) ) W (;r )

In ( YI/h )

YaY4

+ h2(D) + 2c2)

The a u thors a re grateful to Professor F . F . Ling for his di scussion of ou r pap er . The a nswe r to his first qu estion can be found
from a mathematical point of vie w in a ny of a number of t rea tises
on stoc hastic p rocesses. F rom a ph ysical poin t of view , it m ay be
seen from a n exam ina t ion of F ig. 1 if th e hori zont al sea le is

338

S . T. TZ E"' G A "'D EDWARD S A IBEL

Downloaded by [RMIT University] at 13:19 06 August 2013

stretched a hun dre d 10 a th ou san d limes . The fo rm of t he film


thickne ss is a com plicated bu t di ffer en tia ble surface alm ost
everywhe re.
As for t he second question, t he re will und oubt edl y be spots
where r-nvitu t ion will take place, hut th ese will hav e littl e effec t

on t he ca lcula ted load ca rryi ng capacity. It should be noted that.


cavitat ion does not tak e place immedia tely on rea chin g th e
div ergen t portion of the film th ickness, This together with the
fact th a t th e mean film thi ckn ess is conve rge nt, mak es the effect
of ca vitation relat ively unimport an t.

Vous aimerez peut-être aussi