Professional Documents
Culture Documents
Roughness Parameters: E.S. Gadelmawla, M.M. Koura, T.M.A. Maksoud, I.M. Elewa, H.H. Soliman
Roughness Parameters: E.S. Gadelmawla, M.M. Koura, T.M.A. Maksoud, I.M. Elewa, H.H. Soliman
Roughness parameters
E.S. Gadelmawlaa, M.M. Kourab, T.M.A. Maksoudc,*, I.M. Elewaa, H.H. Solimand
a
Production Engineering and Mechanical Design Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt
b
Design and Production Engineering Department, Faculty of Engineering, Ain Shams University, Cairo, Egypt
c
School of Design and Advanced Technology, University of Glamorgan, Wales, UK
d
Electronics and Communications Engineering Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt
Accepted 14 January 2002
Abstract
Surface roughness evaluation is very important for many fundamental problems such as friction, contact deformation, heat and electric
current conduction, tightness of contact joints and positional accuracy. For this reason surface roughness has been the subject of experimental
and theoretical investigations for many decades. The real surface geometry is so complicated that a ®nite number of parameters cannot provide
a full description. If the number of parameters used is increased, a more accurate description can be obtained. This is one of the reasons for
introducing new parameters for surface evaluation. Surface roughness parameters are normally categorised into three groups according to its
functionality. These groups are de®ned as amplitude parameters, spacing parameters, and hybrid parameters. This paper illustrates the
de®nitions and the mathematical formulae for about 59 of the roughness parameters. This collection of surface roughness parameter was used
in a new software computer vision package called Surf Vision developed by the authors. In the package, these de®nitions were extended to
calculate the 3D surface topography of different specimens. # 2002 Elsevier Science B.V. All rights reserved.
Roughness parameters can be calculated in either two- Amplitude parameters are the most important parameters
dimensional (2D) or three-dimensional (3D) forms. 2D to characterise surface topography. They are used to measure
pro®le analysis has been widely used in science and engi- the vertical characteristics of the surface deviations. The
neering for more than half a century. In recent years, there following sections give a brief description for each parameter.
was an increased need for 3D surface analysis. Recent
publications [1±4] emphasised the importance of 3D surface 2.1. Arithmetic average height (Ra)
topography in science and engineering applications.
3D roughness parameters are calculated for an area of the The arithmetic average height parameter, also known as
surface instead of a single line. Hence, in order to calculate the the centre line average (CLA), is the most universally used
3D roughness parameters, the SurfVision software considers an roughness parameter for general quality control. It is de®ned
area from the surface to be tested and divides it into a number as the average absolute deviation of the roughness irregula-
of sections. These sections represent a number of consequent rities from the mean line over one sampling length as shown
pro®les from the surface. The 2D roughness parameters then in Fig. 1. This parameter is easy to de®ne, easy to measure,
calculated for each section separately, and the average of each and gives a good general description of height variations. It
parameter is taken for all sections. This research presents all does not give any information about the wavelength and it is
roughness parameters and their calculation methods. not sensitive to small changes in pro®le. The mathematical
de®nition and the digital implementation of the arithmetic
Abbreviations: 2D, two-dimensional; 3D, three-dimensional; ADC, average height parameter are, respectively, as follows:
amplitude density curve; BAC, bearing area curve; BMP, type of graphics
Z
format stands for windows bitmap; CCS, Cartesian coordinate system; 1 l
CLA, centre line average; CPP, contact probe pro®lometry; EVC, Elf VGA Ra jy
xj dx
capture board; FFT, fast Fourier transformation; GIF, type of graphics l 0
format stands for graphics interchange format; h/v, horizontal/vertical 1X n
resolution; HSL, hue, saturation, lightness Ra jyi j
*
Corresponding author. Fax: 44-1443-48231. n i1
0924-0136/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 2 ) 0 0 0 6 0 - 2
134 E.S. Gadelmawla et al. / Journal of Materials Processing Technology 123 (2002) 133±145
shown in Fig. 4, as follows: 2.13. Mean of the third point height (R3z)
Rti Rpi Rvi
This parameter is the mean of the ®ve third point height
where i ranges from 1 to 5. From the ®gure, Rt1 Rp1 Rv1 , parameters (R3y1, R3y2, R3y3, R3y4, and R3y5). As shown in
Rt2 Rp2 Rv2 , etc. Fig. 5 R3z is equal to
R3y1 R3y2 R3y3 R3y4 R3y5 =5.
The mathematical de®nition of this parameter is as follows:
2.10. Mean of maximum peak to valley height (Rtm) !
1 X 5
Rtm is de®ned as the mean of all maximum peak to valley R3z R3yi
5 i1
heights obtained within the assessment length of the pro®le.
From Fig. 4, the mathematical de®nition of this parameter is
as follows: 2.14. Profile solidity factor (k)
1X n
Rtm Rti The pro®le solidity factor (k) is de®ned as the ratio
n i1
between the maximum depth of valleys and the maximum
where n is the number of samples along the assessment height of the pro®le. The mathematical de®nition of this
length of the pro®le. From the ®gure Rtm
Rt1 Rt2 parameter is as follows:
Rt3 Rt4 Rt5 =5.
Rv
k
2.11. Largest peak to valley height (Ry) Rmax
This parameter is de®ned as the largest value of the 2.15. Skewness (Rsk)
maximum peak to valley height parameters (Rti) along the
assessment length. From Fig. 4, Ry Rt3 . The skewness of a pro®le is the third central moment of
pro®le amplitude probability density function, measured
2.12. Third point height (R3y) over the assessment length. It is used to measure the
symmetry of the pro®le about the mean line. This parameter
To calculate this parameter, the distance between the third is sensitive to occasional deep valleys or high peaks. A
highest peak and the third lowest valley is calculated for symmetrical height distribution, i.e. with as many peaks as
each sampling length, then the largest distance is considered valleys, has zero skewness. Pro®les with peaks removed or
as the third point height (R3y). From Fig. 5 the third point deep scratches have negative skewness. Pro®les with valleys
height parameter (R3y) is the maximum value of the ®ve ®lled in or high peaks have positive skewness. This is shown
values of R3y1, R3y2, R3y3, R3y4, R3y5, that is R3y5. in Fig. 6. The skewness parameter can be used to distinguish
between two pro®les having the same Ra or Rq values but where Rq is the RMS roughness parameter and Yi the height
with different shapes. of the pro®le at point number i.
The value of skewness depends on whether the bulk of the The skewness parameter can be used to differentiate
material of the sample is above (negative skewed) or below between surfaces, which have different shapes and have
(positive skewed) the mean line as shown in Fig. 6. The the same value of Ra. In Fig. 6, although the two pro®les may
mathematical and the numerical formulas used to calculate have the same value of Ra, they have different shapes.
the skewness of a pro®le, which has number of points N, are
as follows: 2.16. Kurtosis (Rku)
Z
1 1 3
Rsk 3 y p
y dy Kurtosis coef®cient is the fourth central moment of pro®le
Rq 1 amplitude probability density function, measured over the
! assessment length. It describes the sharpness of the prob-
1 XN
3 ability density of the pro®le. If Rku < 3 the distribution
Rsk Yi
NR3q i1 curve is said to be platykurtoic and has relatively few high
peaks and low valleys. If Rku > 3 the distribution curve is 2.18. Auto correlation function (ACF)
said to be leptokurtoic and has relatively many high peaks
and low valleys. Fig. 7 shows these two types of kurtosis. The ACF describes the general dependence of the values
The mathematical and the numerical formula used to of the data at one position to their values at another position.
calculate the kurtosis of a pro®le with a number of points It is considered a very useful tool for processing signals
N are as follows: because it provides basic information about the relation
Z between the wavelength and the amplitude properties of
1 1 4 the surface. The ACF can be considered as a quantitative
Rku 4 y p
y dy
Rq 1 measure of the similarity between a laterally shifted and
! an unshifted version of the pro®le. The mathematical and
1 XN
Rku Yi 4 numerical representations of this function are as follows:
NR4q i1 Z
1 L
ACF
dx y
xy
x dx dx
where Rq is the RMS roughness parameter and Yi the height L 0
of the pro®le at point number i. X
N
1
The skewness parameter can also be used to differentiate ACF
dx yi yi1
between surfaces, which have different shapes and have the N 1 i1
same value of Ra. In Fig. 7, although the two pro®les may where dx is the shift distance and yi the height of the pro®le
have the same value of Ra, they have different shapes. at point number i.
The ACF can be normalised to have a value of unity at a
2.17. Amplitude density function (ADF) shift distance of zero. This suppresses any amplitude infor-
mation in the ACF but allows a better comparison of the
The term amplitude density corresponds exactly to the wavelength information in various pro®les.
term probability density in statistics. The ADF represents the
distribution histogram of the pro®le heights. It can be found 2.19. Correlation length (b)
by plotting the density of the pro®le heights on the horizontal
axis and the pro®le heights itself on the vertical axis as This parameter is used to describe the correlation char-
shown in Fig. 8. acteristics of the ACF. It is de®ned as the shortest distance in
To calculate the density of the pro®le heights, the ampli- which the value of the ACF drops to a certain fraction,
tude scale is divided into small parts dy. The measure of the usually 10% of the zero shift value. Points on the surface
amplitude values found within dy, can be made by calculat- pro®le that are separated by more than a correlation length
ing all amplitude values between y and dy relative to the may be considered as uncorrelated, i.e. portions of the
assessment length of the pro®le. The Amplitude density is surface represented by these points were produced by sepa-
hence de®ned by the following equation: rate surface forming events. Correlation lengths may range
from the in®nite correlation length for a perfectly periodic
P
y; y dy
p
y lim wavelength to zero for a completely random waveform.
dy !0 dy
For surfaces produced by a truly random process, the ADF 2.20. Power spectral density (PSD)
would be a Gaussian distribution of surface heights given by
the following equation: The PSD function is an important function for character-
q 2 ising both the asperity amplitudes and spacing. It is calcu-
y lated by Fourier decomposition of the surface pro®le into its
ADF
y 2pR2q exp
2Rq 2 sinusoidal component spatial frequency (f). For a 2D surface
pro®le it can be calculated from the following equation: mean line, per unit length along the assessment length. Fig. 9
Z 2 shows how to calculate the HSC parameter above a selected
1 L
PSD
f y
x exp
i2pfx dx level. The pro®le shown in the ®gure has eight HSC.
L 0
" #2 3.2. Peak count (Pc)
1 XN 1
j2pbi=N
PSD yi e
N 1 i0 The importance of the peak count parameter appears in
some manufacturing processes such as forming, painting, or
where b is the correlation length. coating surfaces. It is de®ned as the number of local peaks,
which is projected through a selectable band located above
3. The spacing parameters and below the mean line by the same distance. The number
of peak count is determined along the assessment length and
The spacing parameters are those which measure the the result is given in peaks per centimetre (or inch). If the
horizontal characteristics of the surface deviations. The assessment length is less than 1 cm, the results should be
spacing parameters are very important in some manufactur- multiplied by a factor to get the peak count per centimetre.
ing operations, such as pressing sheet steel. In such case, As shown in Fig. 10 the peak count is determined only for
evaluating the spacing parameters is necessary to obtain the closed areas of the pro®le, in which the pro®le intersects
consistent lubrication when pressing the sheets, to avoid each the upper and the lower bands in two points at least. The
scoring and to prevent the appearance of the surface texture pro®le shown in the ®gure has four peak counts.
on the ®nal product. One of the spacing parameter is the peak
spacing, which can be an important factor in the perfor- 3.3. Mean spacing of adjacent local peaks (S)
mance of friction surfaces such as brake drums. By con-
trolling the spacing parameters it is possible to obtain better This parameter is de®ned as the average spacing of
bounding of ®nishes, more uniform ®nish of plating and adjacent local peaks of the pro®le measured along the
painting. The SurfVision software calculates the most known assessment length. The local peak is de®ned as the highest
spacing parameters. The following sections give more infor- part of the pro®le measured between two adjacent minima
mation about the spacing parameters. and is only measured if the vertical distance between the
adjacent peaks is greater than or equal to 10% of the Rt of the
3.1. High spot count (HSC) pro®le. Fig. 11 shows how to measure this parameter. This
parameter can be calculated from the following equation:
The HSC parameter is de®ned as the number of high regions 1X n
S Si
of the pro®le above the mean line, or above a line parallel to the N i1
Fig. 10. Calculating the peak count (Pc) parameter within a selected band.
140 E.S. Gadelmawla et al. / Journal of Materials Processing Technology 123 (2002) 133±145
Fig. 11. Calculating the mean spacing of adjacent local peaks (S).
where N is the number of local peaks along the pro®le. intersections of the pro®le at the mean line can be calculated
from the following equation:
3.4. Mean spacing at mean line (Sm)
1X n
n
0 ci
L i1
This parameter is de®ned as the mean spacing between
pro®le peaks at the mean line and is denoted as (Sm). The where L is the pro®le length (in cm).
pro®le peak is the highest point of the pro®le between
upwards and downwards crossing the mean line. Fig. 12 3.6. Number of peaks in the profile (m)
shows how to measure the mean spacing at mean line
parameter. This parameter calculates the number of peaks of the
This parameter can be calculated from the following pro®le per unit length (centimetre or inch). Peaks are
equation: counted only when the distance between the current peak
1X n and the preceding one is greater that 10% of the maximum
Sm Si height of the pro®le (Rt). In Fig. 14 the three little peaks,
N i1
which follow the peaks m2, m3 and m4 are neglected because
where N is the number of pro®le peaks at the mean line. the distance between each peak and the preceding one in too
The difference between the two types of mean spacing small.
parameters, S and Sm, is that the ®rst parameter (S) is The number of peaks can be calculated from the following
measured at the highest peaks of the pro®le, whilst the equation:
second parameter (Sm) is measured at the intersection of 1X n
the pro®le with the mean line. m mi
L i1
3.5. Number of intersections of the profile where L is the pro®le length (in cm).
at the mean line (n (0))
3.7. Number of inflection points (g)
This parameter calculates the number of intersections of
the pro®le with the mean line measured for each centimetre This parameter calculates the number of in¯ection points
length of the pro®le. As shown in Fig. 13, the number of of the pro®le per unit length (centimetre or inch). An
Fig. 13. Calculating the number of intersections of the profile at mean line.
in¯ection point occurs when the pro®le changes its direction the assessment length. This parameter can be calculated by
at any point as shown in Fig. 15. This parameter can be calculating the radius of curvature for each peak along the
calculated from the following equation: pro®le, then calculating the average of these radii of cur-
vatures.
1X n
g gi The radius of curvature for a peak (rpi) can be calculated
L i1
from the following equation:
where L is the pro®le length (in cm). 2yi yi yi1
1
rpi
l2
3.8. Mean radius of asperities (rp)
where yi is the height of the peak at which the peak radius
The mean peak radius of curvature parameter is de®ned as of curvature (rpi) is to be calculated, yi 1 the height of
the average of the principle curvatures of the peaks within the preceding peak, and yi1 the height of the next peak.
Fig. 15. Calculating the number of inflection points along the profile.
142 E.S. Gadelmawla et al. / Journal of Materials Processing Technology 123 (2002) 133±145
The mean peak radius of curvature (r), then can be calcu- where n is the total number of intersections of the pro®le
lated from the following equation: with the mean line along the assessment length.
1 Xn 2
1 4.2. Mean slope of the profile (Da)
rp
n 2 i1 rpi
This parameter is de®ned as the mean absolute pro®le
slope over the assessment length. Many mechanical proper-
ties such as friction, elastic contact, re¯ectance, fatigue
4. The hybrid parameters crack initiation and hydrodynamic lubrication affect this
parameter. This parameter can be calculated by calculating
The hybrid property is a combination of amplitude and all slopes between each two successive points of the pro®le,
spacing. Any changes, which occur in either amplitude or then calculating the average of these slopes. As shown in
spacing, may have effects on the hybrid property. In tribol- Fig. 17, the mathematical and numerical formulas of calcu-
ogy analysis, surface slope, surface curvature and developed lating the mean slope parameter are as follows:
interfacial area are considered to be important factors, which Z
1 L dy
in¯uence the tribological properties of surfaces. The follow- Da dx dx;
ing sections describe the most common hybrid parameters. L 0
1 X
n 1
dy
4.1. Profile slope at mean line (g) Da i
n 1 i1
d xi
This parameter represents the pro®le slope at the mean
line. It can be calculated by calculating the individual 4.3. RMS slope of the profile (Dq)
slopes of the pro®le at each intersection with mean line,
then calculating the average of these slopes as shown in This parameter is the root mean square of the mean slope
Fig. 16. The numerical equation for calculating the pro®le of the pro®le. The mathematical and numerical formulas for
slope at the mean line is as follows: calculating this parameter are as follows:
s
1 X n 1 Z Z
1 dyi 1 L 1 L
g tan Dq
y
x y dx; y _ 2 _ y
x dx
n 1 i1 dxi L 0 L 0
1X n
4.4. Average wavelength (la) lo li
L i1
The average wavelength parameter is a measure of the
spacing between local peaks and valleys, taking into con- where li is the length of line number i in the pro®le, and it can
sideration their relative amplitudes and individual spatial be calculated from the following equation:
frequencies. This parameter can be calculated from the q
following equation: li
yi1 yi 2 dxi 2
2pRa
la where yi is the pro®le height at point number i, and dx the
Da
horizontal distance between each two successive points.
where Ra is the arithmetic average height and Da the mean
slope of the pro®le. 4.7. Bearing area length (tp) and bearing area curve
4.5. RMS wave length (lq) The bearing line length parameter is de®ned as the
percentage of solid material of the pro®le lying at a certain
The RMS wavelength parameter is similar to the average height. This parameter is a useful indicator of the effec-
wavelength (la) parameter. It is de®ned as the root mean of tive contact area as the surface wear. From Fig. 19, the
the measure of the spacing between local peaks and valleys, bearing area length can be calculated from the following
taking into consideration their relative amplitudes and indi- equation:
vidual spatial frequencies. It can be calculated from the
following equation: 1X n
tp li
2pRq L i1
lq
Dq
where L is the assessment length of the pro®le.
4.6. Relative length of the profile (lo) By calculating the bearing line length at different heights
of the pro®le, the bearing area curve (BAC) can be drawn, as
The relative length of the pro®le (lo) is estimated by shown in Fig. 20. The horizontal axis represents the bearing
calculating the lengths of the individual parts of the pro®le area lengths as a percent from the total assessment length of
Fig. 19. Calculating the bearing area length (tp) of the profile.
144 E.S. Gadelmawla et al. / Journal of Materials Processing Technology 123 (2002) 133±145
X1
1 NS
Hu S:D:
yiNPS1 ;yiNPS2 ;yiNPS3 ;...;yiNPSNPS
NS i0
The roughness height uniformity of a pro®le (Hu) is The proposed vision system SurfVision is divided into
de®ned as the standard deviations of the individual height two parts, hardware and software. The hardware includes
values of the pro®le constituting the arithmetic average an IBM compatible personal computer with Windows 95
height (Ra). To calculate this parameter the standard devia- operating system, frame grabber as a capturing board, charge
tion is calculated for the pro®le heights in each sampling coupled device (CCD) camera, and a microscope. The
E.S. Gadelmawla et al. / Journal of Materials Processing Technology 123 (2002) 133±145 145
software was written especially to perform different analysis After checking the accuracy of the system for calculating
on the captured images. The proposed software was written the Ra parameter, 59 roughness parameters were calculated
using Microsoft Visual C version 5.0 and it could run for the six sections using both the imperial and the metric
under Windows 95, Windows 98 or Windows NT operating units. The above table shows the symbols, the description
systems. and the value of the calculated roughness parameters for the
The software package was developed totally in-house 2 min. specimen.
such that it can be used independently without referring to
any other software. The package includes the unique feature
6. Conclusion
of containing a multitude of surface roughness parameters
that are not included in any other package hitherto. Also, the
Different manufacturing processes produce different sur-
software allows the building up of a data base information
face characteristics. Also, different applications require dif-
system during surface inspection. This database was made to
ferent surface properties. Surface parameters are therefore
allow the future inclusion of arti®cial intelligence module for
different and wide-ranging. Each of these parameters indi-
automated calibration of the system. The software is fully
cates a particular property of the surface and it could be the
integrated with AutoCAD and MS Word. The software has
most important for the particular application. This research
the professional look interface that is used by most Windows
presented the de®nitions and the mathematical formulae for
95 application.
about 59 of the surface roughness parameters. This collection
Standard surface roughness specimens were used to test
of surface roughness parameter was used in a new software
the proposed vision system. These specimens are the
computer vision package called SurfVision developed by the
RUBERT surface roughness scales no. 24 MK II, which
authors. In the package, these de®nitions were extended to
has 12 pieces with different values of Ra for different
calculate the 3D surface topography of different specimens.
machining operations. Three specimens with the same man-
ufacturing process and different values of Ra were selected
as shown in the table below. The values of Ra for the three References
specimens were given in min. The corresponding values in
mm were calculated by multiplying each value by (0.0254) as [1] U.B. Abou El-Atta, Surface roughness assessment in three-dimen-
shown in the table below. sional machined surfaces for some manufacturing operations, M.Sc.
Thesis, Industrial Production Engineering Department, University of
Mansoura, Egypt, 1991.
[2] E.C. Teague, F.E. Scire, S.M. Baker, S.W. Jensen, 3-Dimensional
Specimen Value of Calculated Manufacturing Accuracy stylus profilometry, Wear 83 (1) (1982) 1±12.
number Ra (min.) Ra (mm) process (%) [3] T. Pancewicz, I. Mruk, Holographic contouring for determination of
three-dimensional description of surface roughness, Wear 199 (1)
1 2 0.0508 Lapping 10 (1996) 127±131.
2 4 0.1016 Lapping 10 [4] B.G. Rosen, Representation of 3-dimensional surface topography in
3 8 0.2032 Lapping 10 CAD-systems and image processing, Int. J. Mach. Tools Manuf. 33 (3)
(1993) 307±320.