Professional Documents
Culture Documents
Rough Surface
555
z
L
1 4
2
i i −1 i i +1
3
i +1
x
l
There exist other methods to calculate the peak radius of curvature. For example, rpi may be
calculated from formula:
where Dli is the length of the section of the asperity by the line equidistant to the mean line and
spaced from the peak to a distance hi. Here it is assumed that hi is approximately equal to 0.3 Ra.
This list of roughness parameters currently in use is far from complete. The majority of the
above parameters are included (or at least had been included) into national standards in various
combinations (see Table A1.1 which includes the ISO standard for comparison).
Foreign standards should be used with specific caution because:
– the same parameter may be designated differently in different standards (these differences
are specifically marked in Table A1.1, e.g., km instead of Ra in the Spanish standard);
– the same designation has different or differently assigned parameters. For example, unlike
the above definition of Rz corresponding to Russian or British standard, the German standard
determines this parameter as the average of the maximum peak-to-valley heights in five
consecutive sampling lengths;
– different standards recommend different numbers of sample lengths to be used as the
assessment length of one and the same parameter.
Probabilistic approach based on the probability theory and the random process theory is often
used under the assumption that rough surfaces are Gaussian (normal).
TABLE A1.1
Some National Roughness Standards
Ra Rz Rmax Rp tp S Sm Country
C C UK
C C C Germany
km H Spain
C C C C C C Russia
C USA
C C C C Sweden
C R C C (Tp)C AR France
C C C Japan
C C C ISO
Skew
0
Kurtosis
FIGURE A1.2 Skewness Sk and kurtosis Ks of real profile. (according to Thomas, T. R., Precision Eng., 3, 97, 1971.)
Skewness Sk and kurtosis Ks are the parameters serving to assess the deviation of the real surface
from the surface with symmetric distribution. The skewness Sk is a measure of deviation of
the probability density from the symmetric shape (Figure A1.2). It is defined in the following
manner:
ð
N
1
Sk Z z3 f ðzÞdz (A1.2)
s3
KN
Symmetric distributions (for example, the Gaussian distribution) naturally possess zero skew-
ness. The skewness is negative when profiles have worn peaks and deep scratches, while the
skewness is positive for a surface with high peaks and/or shallow valleys filled up with debris.
The excess Ks characterizes the peakedness of the distribution and is defined as (Figure A1.2):
ð
N
1
Ks Z 4 z4 f ðzÞdz: (A1.3)
s
KN
The symmetric Gaussian distribution has the kurtosis equal to three. If the profile has compara-
tively few high peaks and deep valleys, e.g., as a result of surface turning with a sharp pointed tool,
then Ks !3 and the distribution is termed platikurtic. The distribution is termed leptokurtic, Ks O3,
in the opposite case when there is a large number of high peaks and deep valleys.
The probalistic approach is generalized by the approach based on the theory of random fields
(processes).
It has been noted above that the surface roughness ordinate z is measured, say, from the mean
plane. The coordinate is treated as a random function of two variables z Z z(x, y). According to the
theory of random fields, in case the field z is homogenous and normal, as it is commonly assumed
when analyzing rough surfaces, it is fully determined by its correlating function:
Lð1 Lð2
1
Kðx; yÞ Z lim zðx1 ; y1 Þzðx1 C x; y1 C yÞdx1 dy1 :
L1 ;L2/N 4L1 L2
KL1 KL2
Practically any problem of determination of field characteristics (some of them are listed below)
can be solved. Still, it is frequently convenient to apply Fourier transformation to the correlation
function:
ð
N ð
1
Fðkx ; ky Þ Z 2 Kðx; yÞexp½Kiðxkx C yky Þdxdy;
4p
KN
which is termed as the spectral density. Here kx, ky are the components of the wave vector k.
The moments of spectral density are intensively used in the subsequent analysis:
ð
N ð
mpq Z Fðkx ; ky Þkxp kyq dkx dky
KN
where m0, m2, m4 are the moments of the spectral density of the profile.
The moments are practically determined from the surface profilograms using the Longuet-
Higgins formulas for the densities of zeroes Dz,0 and extrema (maxima and minima) De,0:
1 m2 1=2
Dz;0 Z ; (A1.5)
p m0
1=2
1 m4
De;0 Z : (A1.6)
p m2
m2 Z p2 s2 D2z;0 ; (A1.7)
m4 Z p4 s2 D2z;0 D2e;0 ; (A1.8)
where the standard deviation s equal to (m0)1/2 is also determined from the profilogram. Available
modern measuring instruments allow us to determine the densities of zeroes and extrema
automatically.
In order to determine a number of characteristics of a rough surface it may be enough to use the
joint probability density of the random function and its first two derivatives:
vzðx; yÞ vzðx; yÞ
x1 Z zðx; yÞ; x2 Z ; x3 Z ;
vx vy
v2 zðx; yÞ v2 zðx; yÞ v2 zðx; yÞ
x4 Z ; x5 Z ; x6 Z :
vx2 vxvy vy2
It is assumed that these random variables satisfy the general conditions of the central limit
theorem, so their joint probability density is written as
1 1 ðT ð
pðx1 ; x2 ; .; x6 Þ Z 3 1=2 exp K x M x ; (A1.9)
8p D 2
where xð is the column vector of values xj (j Z1, 2,., 6); T, the symbol of transposition; M, the
quadratic matrix 6!6 inverse with the matrix N:
0 2 1
x1 x1 x2 x1 x3 . x1 x6
B C
B C
Bx x x22 x2 x3 . x2 x6 C
N ZB 2 1 C;
B , , , , , C
@ A
x6 x1 x6 x2 x6 x3 . x26
Now it is the matter of techniques to determine various useful characteristics of the rough
surface. For example, again leaving the details apart, let consider the distribution of the peaks. The
surface has a summit (i.e., local maximum) in the point (x, y), providing the following conditions
are satisfied:
x2 Z x3 Z 0; x4 ! 0; x6 ! 0; x4 x6 Kx25 R 0:
If the point is within the elementary region dA above which the summit is disposed, the
variables x2, x3 vary within the region
vðx ; x Þ
dx2 dx3 Z 2 3 dA Z jx4 x6 Kx25 jdA:
vðx; yÞ
Then the density of the summits with the ordinate x1 is determined by integration:
ððð
Psum Z pðx1 ; 0; 0; x4 ; x5 ; x6 Þjx4 x6 Kx25 jdx4 dx5 dx6 ;
V
x4 ! 0; x6 ! 0; x4 x6 Kx52 R 0:
With some experience and taste for non-trivial transformations it is easy to demonstrate that the
last integral is taken in the finite form, after it is normalized for the density of the summits (the
number of summits per square unit of area):
1 m4
DZ pffiffiffi : (A1.11)
6p 3 m2
It is determined after transformations that the probability density of the summits is:
8
pffiffiffi < 2 31=2 pffiffiffiffiffiffi
* 3 * 4 3ð2aK3Þ 5 *2 3 2p * 2
pðx1 Þ Z x expðKC1 x1 Þ C ðx1 K1Þ
2p : 1 a2 2a
2 0 0 11=2 13 0 1 2 31=2
3 A A5exp@K1 x*1 2 A C 4 2pa 5
!41 C erf @x1* @ ; (A1.12)
2ð2aK3Þ 2 3ðaK1Þ
2 0 0 11=2 13 9
*2 =
a ax
!41 C erf @x1* @ A A5exp 1
K
2ðaK1Þð2aK3Þ 2ðaK1Þ ;
Here by a peak is meant a maximum in the profilometer trace (profile) which does not need to be
coincident with a maximum in the surface named a “summit”.
A comparison of probability densities of summits and peaks (Figure A1.3) shows how the
profile distors the surface. With the same parameter a the profile gives smaller amounts of high
peaks and underestimates their height than there exist on surface. The reason is that the chance that
stylus of profile-measuring instrument will travel over the summit is negligible. That is, the stylus,
0.7
0.6
a = 1.5
0.5 a=5
0.4
a=∞
p (ξ1*)
0.3
0.2
0.1
0.0
−2 −1 0 1 2 3
ξ 1*
FIGURE A1.3 Probability densities for peak (thin lines) and summit (thick lines) heights.
as a rule, travels over the shoulder of an asperity. The distortion decreases when the parameter a
varies from 1.5 (the greatest distortion) to N (zero distortion).
Let us consider some other measures of surface roughness.
The surface gradient is defined by the variables (x2, x3), the joint probability density of which is
found by the equation
2
1 x2 C x23
pðx2 ; x3 Þ Z exp K : (A1.13)
2pm2 2m2
In deducing Equation A1.14 we suggested the axis X1 coincides with the major direction, so that
m11 Z 0 and m20 R m02. In the case of the isotropic roughness the gradient modulus has the
Rayleigh distribution
z
pðzÞ Z expðKz2 =2m2 Þ (A1.15)
m2
ð
N
1=2
z Z zpðzÞdz Z pm2 (A1.16)
2
0
Surface curvature at summit is defined as follows. There are two major directions and the
corresponding them principal curvatures K1 and K2 at each surface point. Their half-sum is the
mean curvature Km Z (K1 CK2)/2 Z K(x4Cx6)/2. The joint probability density for summit
heights x1 at the summit t1 is
1=2
3C1 1
pðx1 ; tÞ Z ½t12 K2 C 2 2
expðKt1=2 ÞexpðKC1 x2
1 Þexp ðC t2 C C2 t1 x1 Þ (A1.17)
2p 2 11
m 1=2 I ðx ÞK2I ðx Þ C 2I ðx Þ
K m ðx1 Þ Z 4 3 1 1 1 5 1
; (A1.18)
3 I2 ðx1 ÞK2I0 ðx1 Þ C 2I4 ðx1 Þ
where
1=2
p 1
I0 Z exp K x1 ð1 C erf b1 x1 Þ;
2C1 2
1 2 1 2
I1 Z expðKC1 x1 Þ C p b1 x1 exp K x1 ð1 C erf b1 x1 Þ ;
1=2
C 2
1
pffiffiffi 3=2 1 1 2
I2 Z 2C1 b1 x1 expðKC1 x2 1 Þ C p 1=2
b x
1 1
2
C exp K x ð1 C erf b1 1 ;
x
Þ
2 2 1
2 3 1 2
I3 Z 2 ð1 C b21 x2 1 ÞexpðK C x
1 1
2
Þ C p 1=2
b x
1 1
3
C b x
exp K x ð1 C erf b1 x1 Þ;
C1 2 1 1 2 1
1=2
p ax2
I4 Z ð1 C erf b2 x1 Þexp K 1
;
2ðC1 C 1Þ 2ðaK1Þ
1 ax2
I5 Z expðKC1 x2 1 Þ C p 1=2
b x
2 1
ð1 C erf b x
1 1
Þexp K 1
;
C1 C 1 2ðaK1Þ
1=2 1=2
3 a
b1 Z ; b2 Z :
2ð2aK3Þ 2ð2aK1Þð2aK3Þ
Analysis of dependence Equation A1.18 shows (Figure A1.4) the curvature radius at the
summit is decreased with increase of the peak height (higher peaks is “more acute”), i.e., the
curvature is increased.
1=2 a: as a / 1.5 the mean curvature is directly
Its increase depends on bandwidth parameter
proportional to the peak height K m ZK 2m4 =3 x1 ; as a / 1.5 it becomes independent of the
peak height K m Z ð8=3Þðm4 =pÞ1=2 .
2.5
2.0
a=∞
Km /(m4)1/2
1.5
1.0 a=6
a = 1.5
0.5
a=2
0.0
−2 −1 0 1 2 3
ξ1*
It is pertinent to note that the dimensionless peak curvature is calculated by the formula:
pffiffiffi
K peak ðx1 Þ pffiffiffi xb C p expðx2 b2 Þð1 C erf xbÞðx2 b2 C 0:5Þ
pffiffiffiffiffiffi Z b a pffiffiffi ;
m4 1 C xb p expðx2 b2 Þð1 C erf xbÞ
where bZ 1=½2ðaK1Þ1=2 .
Comparison of peak and summit curvatures presented in Figure A1.5 shows that here, too,
there is a distortion of the surface by the profile. This distortion is small for a O2 within the range,
0% x1 % 3 which is of interest to tribological calculation. It is notable that the profile gives both
overestimated (a !2.5) and underestimated (a O2.5) values of peak curvature when compared to
the summit curvature.
Average roughness R’a is defined as mean value of absolute surface deviations from mean plane:
ðð
1
R’a Z jzðx; yÞdxdyj
A
A
3.0
2.5 a = 1.5
Km /(m4)1/2
2.0
a = 2.5
1.5
a=4
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
ξ1 *
FIGURE A1.5 Comparison of peak (thin lines) and summit (thick lines) curvatures for aZ 1.5, 2.5 and 4.
ð Lð0 ðð
4s2 X
N
1 ð2kK1Þð0Þ
2
DR’a Z 4 F K 2k ðrÞdx1 dx2 dy1 dy2
L0 kZ1 ð2kÞ!
0
Here F(k)(X) is the derivative of the Laplace function; K(r) is the correlation function of
the field.
Bearing area ratio is defined in the following manner. Let the surface zZ z (x, y) is intersected
by horizontal plane parallel to the mean plane and far from it on some distance u. In this horizontal
plane regions are cut out by surface z. Their summary area referred to unit of the nominal area is
named the bearing area at level u and denoted by hu. The mean value of the bearing area is
hu Z 1KFðu=sÞ; (A1.20)
where
ð
N
1
FðxÞ Z expðKt2 =2Þdt:
ð2pÞ1=2
KN
ð Lð0 ðð
1 XN
1 h ðkÞ
u i1=2
Dhu Z 4 F K k ðx2 Kx1 ; y2 Ky1 Þdx1 dx2 dy1 dy2
L0 kZ1 k! s
0
The distribution of value hu may be assumed nearly normal with the mathematical expectation
h u and variance Dhu.