Wear 311 (2014) 167179

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Study of contact of rough surfaces: Modeling and experiment

Stanislaw Kucharski n, Grzegorz Starzynski
Institute of Fundamental Technological Research, PAS, Warsaw, Poland

art ic l e i nf o

a b s t r a c t

Article history:
Received 6 September 2013
Received in revised form
13 January 2014
Accepted 20 January 2014
Available online 27 January 2014

In the paper a problem of contact of rough surface with rigid at plane is investigated experimentally and
numerically. Samples made of three different steels with roughness constituted in a sand-blasting
process were compressed in a special experimental setup. 3D surface topographies were measured in
initial and deformed state using scanning prolometry. An experimental procedure has been designed
that enables specifying load-approach and load-real contact area relations corresponding to plastic
deformation of roughness zone. These relations were also simulated using a simple model based on
statistical approach with special procedure proposed for a proper specication of sampling interval. The
experimental and numerical results have been compared.
& 2014 Elsevier B.V. All rights reserved.

Keywords:
Contact mechanics
Roughness
Real contact area
Asperity denition
Sampling interval

1. Introduction
The topographies of interacting surfaces can have a signicant
inuence on the global physical and mechanical behaviors of a
technical system. The evaluation of real contact area between two
rough surfaces is an important issue for understanding tribological
quantities and processes such as friction, wear, adhesion, lubrication, air or water leakage [1]. The relation between the thermal
and electromagnetic resistivity and real contact area between two
solids in contact is also important [2,3,4].
It is well known that roughness features can be dened in a
wide length scale ranging from the length of physical sample to
atomic scale. To study the mechanism of any contact problems, it
is necessary to characterize such multi-scale rough surface and to
know the structure at length scale to the examined phenomenon.
The modeling of the relevant contact between two rough surfaces
consists of two parts: the rst is geometrical the modeling of
topography of surface, and the second is mechanical modeling of
deformation of asperity. Combination of these two models can
give a general description of contact of two rough surfaces.
One of the most developed ideas of describing surface topography
are the methods of dening roughness using random process theory.
Many statistical parameters can be computed from mathematically
modeled surfaces. The distribution of surface summits (dened as the
point having a greater height than those of the four or eight neighbors)
is frequently assumed to be Gaussian.

Corresponding author: Tel.: 48 22 8261281 141; fax: 48 2282677380

E-mail address: skuchar@ippt.pan.pl (S. Kucharski).

0043-1648/$ - see front matter & 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.wear.2014.01.009

On the basis of probability theory, following the many surface

theoretical works, Greenwood and Williamson [5], Nayak [6,7],
Whitehouse and Archard [8], and Greenwood [9] have made an
important advancement in developing the model of the contact of
rough surface. When Greenwood and Williamson [5] formulated
their original description of elastic rough contact, they based it on
the assumption that only asperity height was a random variable,
and the radius of each spherical peak was constant, and they used
the Hertz solution of elastic deformation of sphere in rigid halfspace. Greenwood and Trip [10] expanded the model in 1970 to
the contact of two rough surfaces and concluded that the contact
between two rough surfaces is not signicantly different from the
contact between a rough surface and a at plane.
One of the drawbacks of this class of models, which rely on the
specication of a single radius of curvature, is the ambiguity of scale.
That is, the determination of the average radius of curvature of a
surface prole is sensitive to the scale of observation, or more
specically, to the lateral resolution used to measure the surface.
The GW theory assumes roughness on a single length scale, which
results in an area of real contact which depends (slightly) non-linearly
on the load even for very small loads. Bush et al. [11] have developed
a more general and accurate contact mechanics theory (BGT) where
roughness is assumed to occur on many different length scales. This
results in an area of real contact which is proportional to the
squeezing force for small squeezing forces. Important considerations
relating to the linearity of the real contact areaload have been
proposed in [12], considering not only the asymptotic BushGibson
Thomas (BGT) solution for very small loads, but also the full solution,
giving rise to a deviation from linearity in the intermediate and high
pressure regimes. In 2006 Greenwood additionally simplied the
(BGT) contact model but obtained very similar results [13]. All of these

168

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

models concerned the description of surface roughness described by

the single selected prole.
The statistics of Greenwood models [5,9] framework has been
preserved, but different models are implemented for the asperity
deformation. The GW model and generally their idea are widely
used and being modied up till now. In recent years, Greenwood
with other authors developed the GW theory by introducing the
interaction between asperities [14], and also evaluated the difference between the approaches of two and three dimensions. They
concluded that the mean real contact pressures calculated for two
dimensions will be much lower than in three dimensions, and will
depend strongly on the thickness of the slab used to represent
the elastic half-space [15].
Various extensions of the GW contact model have been developed to incorporate effects of adhesion and plastic deformation
[16,17]. With only a small fraction of the available area supporting
the load, the contacting asperities of the surfaces often carry very
high compressive stresses. These high stresses will often cause
yielding in the material and thus purely elastic contact models of
rough surfaces are not always adequate. Chang at al. [16] modied
the GW theory by introducing the plastic deformation of the most
highly loaded asperities. Buczkowski and Kleiber (1999) [18] proposed a random surface model of elasto-plastically yielding asperities with Gaussian height distribution combined with mechanical
description of a single peak based on the Hertz theory coupled with
the Mindlin friction theory. The stochastic model was included in a
incremental nite element procedure. A few years later authors
extended their model and presented the complete elasto-plastic
microcontact model of anisotropic rough surfaces [19].
Whitehouse and Archard [8] extended the random asperity
model to include random heights and curvatures. In the model the
assumption is used that any surface prole of random type can be
completely dened (in statistical sense) by two characteristics: the
height distribution and the autocorrelation function. In their theory
for the rst time the important problem appears asperity density
and curvatures and some others parameters resulting from random
theory are not intrinsic properties of the surface, and depend on the
correlation distance (and, indirectly, on sampling intervals). However,
Archard's contact model is based on hypothetical, idealized surfaces
and is difcult to apply to a real rough surface. Again, the rough
surface was described by a random prole, related only to the model
geometry, and did not include contact mechanics. This was in 1970
but the problem came back in the work of Greenwood (2001) [21],
where the authors criticize their own proposition and denition of
asperities. They claim that peaks or summits dened according to the
previous denition do not represent the asperities and correspond to
artifacts at the surface, especially when a small sampling interval is
used. It is not hard to imagine that there will be a lot of peaks that
will conform with the 3-point denition, but from the point of view
of the mechanical contact it will be completely irrelevant. Also
Thomas [20] and Thomas and Rosen [22] proposed the determination of the optimum sampling interval for rough contact mechanics
30 years after the paper of Whithouse and Archard, so, it is seen, the
problem is not trivial. They presented the relationship of relevant
roughness parameters to properties of the power spectral density
function (PSDF). Then they expressed the PSDF in terms of fractal
parameters, which are independent of condition of measurement.
The last part of the model is to combine the plasticity index based on
slope with second moment equation of PSDF and they obtain the
relationship between the critical wavelength, fractal dimension and
material properties. This approach is limited to the Gaussian distribution of heights of the surface and power spectra as a power law.
Recently some experimental works appeared showing the
inuence of both the denition of peak and sampling interval
[2326]. Different criteria that take into account the number of
required neighboring points on the prole (i.e., 3, 5 and 7 points),

the peak-threshold value (z-direction) and the effect of the data

resolution in the x-direction were applied in this study. The results
show the huge inuence of these pre-selected criteria for which
no veried guidelines exist.
In recent years there has been a return to the description of the
surface and contact mechanics based on the proles. In a large twopart work by Pugliese, Ciulli, and Ferreira (2008) [27], the authors
presented several ways to approximate the roughness prole through
a set of parabolas, based on the approach of Aramaki [28]. This is a
clear attempt to describe the asperities, avoiding the problem of
measuring resolution. The real prole is described by parabolas that
simulate it by maintaining the constancy of some specic characteristics (approach of same area, same Rq, least mean squares (LMS), etc.).
The contact mechanics models used include two different elastic ones
and two elasticplastic models (Chang et al. [16] and Zhao et al. [17]).
The combination of this approach with the contact mechanics model
including the elastoplastic transition developed by Zhao, Maietta and
Chang seems to guarantee the best results. However, it seems that the
authors have committed some inaccuracy in describing the surface
with a prole (2D) and used the mechanical solutions for threedimensional solids.
Some authors presented a completely different approach to
solving the problem of contact of rough surfaces. More recently,
Buchner et al. [29] developed a new concept based on a combination of the bearing area curve and a model asperity representing
the average asperity slope of the original surface prole (after
Hansen [30]). This paper presented a method for evaluating the
real contact area depending on the normal load that takes the
material properties and real asperity slopes into consideration,
and simplication is achieved by making use of the original
character of real surfaces. The deformation of the bearing area
curve and the Hansen prole were calculated by nite element
analysis. For a given remaining height the real contact area and
total normal force were determined. The new concept showed
very good correspondence with the data obtained by FEM simulating the compression of the original prole of tested sample.
Another new method to determined the contact between
rough deformable surface and rigid smooth plane has been
investigated by Belghith et al. in the work [31]. Roughness
parameters for the microscopic model were deduced using the
standard procedure for roughness and waviness motif parameters. The motif is dened as the part of the prole found
between two peaks. This study described asperity geometry by
Robbe-Valloire's approach [32], which assumes a perfect circular
shape of asperities radius with a lognormal distribution. The new
idea was to determine the mean radius of asperity from dimensional characteristics of each motif. Results of deterministic
microscopic model have been validated with an analytical study
and a good correlation is found.
In this paper the analysis of surface topography is presented in the
context of investigation of contact of rough surfaces. In particular the
problem of sampling interval in surface topography measurements is
considered. The inuence of sampling interval on some roughness
parameters that are important in contact process is studied. A simple
model of contact is proposed and veried experimentally.

2. Experiment
The experiment was consisted of 4 stages:
specication of stressstrain curves for selected steels,
measurement of surface topography of constituting rough
surfaces before contact loading,
contact loading and measurements contact compliance
pa, and

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

169

Table 1
Chemical composition of tested steels (%).

S235
45
40HM

Si

Mn

Cr

Mo

Ni

0.17
0.45
0.4

0.25
0.25

1.4
0.6
0.65

0.035
o 0.04
0.035

0.035
o0.04
0.035

o 0.03
1.05

o 0.01
0.2

o 0.03
0.3

1000
900
40HM

700

material ratio [%]

stress [MPa]

800
steel 45

600
500
400
300

S235

200
100
0
0.00

0.02

0.04
0.06
strain

0.08

100
90
80
70
60
50
40
30
20
10
0

As

0.55
o 0.03

o 0.08

45-Abbott 0MPa
40HM-Abbott 0MPa
S235-Abbott 0MPa

0.10

Cu

0.2
0.4
0.6
0.8
dimensionless height (h/Sz)

Fig. 2. The bearing curves of the sandblasted samples for three steels.

Fig. 1. Strength curves of tested steels.

measurement of surface topography of constituting rough

surfaces after contact loading.

S235-Distribution

3. Results
3.1. Inuence of the sampling interval of rough surface measurement
on geometric model input parameters
In many models describing contact between rough surfaces,
some statistical parameters of the surface topography are used.
These parameters are obtained from measurement of the surface
by the scanning prolometer which allow scanning of the surface
using a series of proles. Depending on the kind of prolometer
the distance between the proles (dy) and the value of the
sampling interval (resolution) in the direction of measuring
proles (dx) can be very different. The sampling interval selection
is important for rough contact mechanics. In many models a
number of calculated quantities (plasticity index, separation
between contacting rough surfaces and real contact area) change
their values when sampling interval changes [24,26].
The obtained result is mainly related to the denition of the
peak (2D) or summit (3D), which was proposed by Greenwood and

height distribution [%]

40HM-Distribution

Samples were made of three kinds of steels (Table 1) with

different stressstrain characteristics. The steels were selected
especially due to their various values of yield limit and ultimate
strength. Stressstrain curves, Fig. 1, were measured on standard
samples cut from pieces of steels which were then used to test the
contact (Fig. 1).
The samples for contact research were subjected to mechanical
surface treatments (sandblasting), resulting in surfaces topography
with similar characteristics on each kind of steel, (bearing curves,
height distribution, Figs. 2, 3 and parameters Table 2).
The roughnesses of surfaces of samples have been measured
before and after loading for a wide range of nominal pressures up
to 800 MPa. All surface topographies were measured on a scanning
prolometer Hommel Tester T8000 Nanoscan. All parameters and
analyses were determined using the Hommel Map Expert program
for surfaces (not proles) measurements.

45-Distribution

0
0

0.5
dimensionless height (h/Sz)

Fig. 3. The height distribution of the sandblasted samples for three steels.

Table 2
3D surface parameters of the sandblasted samples for steels S235, 45 and 40HM
(the equations used to compute the statistical parameters are presented in the
Appendix).

Sq
Ssk
Sku
Sp
Sv
Sz
Sa

S235

45

40HM

5.22
 0.509
4.02
14.9
20.1
35
3.98

4.64
 0.532
4.38
13.7
17.4
31.1
3.47

4.33
 0.462
4.11
12.5
15.5
27.9
3.26

mm

mm
mm
mm
mm

Root mean square height

skewness
kurtosis
Maximum peak height
Maximum pit height
Maximum height (range)
Arithmetic mean height

used in most computer programs that analyze prolometry

measurements. According to this denition each measuring point
where neighbors have lower values of the z-coordinate can be
considered as a peak. In the case of 2D, measurement (prole)
involves a three-point denition, and for 3D a ve or nine point
one [23]. In the program (Hommell Map Expert) used in our work,

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

2500
2

the nine-point denition was applied according to the EUR 15178N

standard.
It is therefore important to check how a change in the sampling
interval affects the values of relevant parameters for the models. In
some models the average radius or curvature of the asperities, the
average slope of asperities, and the density of peaks are used.
Tests on the inuence of sampling interval on surface parameters were carried out on the 45 steel sandblasted surface.
The minimum step in the x-direction was taken as dx 0.2 m,
and in the y-direction, for technical reasons, it could not be less
than dy 1 m. Measurement data was modied by changing the
sampling interval in the direction of x and y, (maximum
dx 100 m and dy 100 m). The inuences of sampling interval
on the following quantities (functions) were checked:

density of asperities [1/mm ]

170

2000
1500
1000
500
0
0

700

radius of asperity [m]

300
200
100
0
0

20

40
60
dx,dy [ m]

80

100

Fig. 4. Mean radius of asperity as a function of sampling interval changes, simultaneously dx and dy.

density [1/mm ]

600

400

60

80

100

Fig. 5. Density [1/mm2] of asperities as a function of sampling interval changes,

simultaneously dx and dy.

800

500

40

dx, dy [ m]

bearing (Abbott) curve,

height (amplitude) parameters,
radii of asperities,
density of asperities, and
distribution of asperities.

Changing the sampling interval has almost no relevance to the

bearing curve. It is clear that even a very big sampling interval
(dx 100 m and dy 100 m) does not change the bearing curve.
Changing the sampling interval has small effect (not more then
10%) also on the height parameters values.
The mean radius of asperities is one of the most important
parameters in statistical models based upon the GW model. The
fundamental work of Greenwood used the constant mean radius
of the peak, while the later work used different radii and their
distribution. Fig. 4 shows the sensitivity of this parameter to
changes of the sampling interval.
Changing the sampling interval in both directions results in the
increase of the mean asperity radius from 2 m to several hundred m
(Fig. 4). This indicates a very high sensitivity of this parameter on the
size of the sampling interval and also shows how ambiguous a result
of measurement can be without the proper interpretation.
The change in the measuring step has a signicant effect on the
density of asperities. Here, the density of asperities decreases from
a range of several hundred [1/mm2] to only several [1/mm2], when
simultaneously both sampling intervals are changed (Fig. 5), while
it also decreases, but much more slowly, to approximately 90
[1/mm2] when only the dy-step is changed. The summits and
curvatures are detected by local neighborhood with respect to
eight neighboring points, Hommell Map Expert EUR 15178N.
Another important quantity for the statistical description of
surface topography is the density distribution of asperities. In this

20

D-02x5
D-20x20
D-50x50
D-100x100
D-5x5

500
400
300
200
100
0
0.0

0.2
0.4
0.6
0.8
range of roughness (h/Sz)

1.0

Fig. 6. Comparison of the summits distribution as a function of sampling interval

dx and dy.

case, a signicant change in amplitude distribution is observed,

while the character of the distribution remains the same (Fig. 6).
Change in amplitude is obvious, since the integral of the distribution is the density of asperities and the reduced density results in
decrease of the distribution.
3.2. Measurement of contact loading
The experiment was carried out using a device shown in Fig. 7.
The device enables precise measurement of the approach as a
function of the contact pressure.
The contact loading is realized between the nominally at
rough surface of the upper specimen (A) and the bottom special
head (B) having a very smooth at surface (Sa 0.06 [m]) and
made from very hard steel (63 HRC). The test sample is in the form
of a 50 mm diameter by 30 mm long cylinder. The contact surface
of special head (B) is in the form of three uniformly placed
punches ring sectors (nominal area 194 mm2). The special head
B serves not only as a counter-sample but also as a base for
approach (displacement of upper specimen) measurement. The
quasistatic load is applied gradually by steps with a hydraulic
press until reaching nominal stresses 200, 300, 500 and 800 MPa.
The load is measured using a tensometric bridge and the resulting
approach of the upper specimen is registered by an inductive
sensor. Results, in the form of diagrams of approach versus contact
pressure, are produced on-line on the screen of a PC. An example
of the result is shown in Fig. 8.
The approachcontact stress curves have two branches that
correspond to loading and unloading processes (Fig. 8), and in the

-3

load, P

displacement [mm x10 ]

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

A- sample

contact
interfaces

point K

loading signal, P

171

14
12
10
8
S235
40H
s45

6
4
2
0

B-loading and
gauge head

computer
a

Fig. 7. Scheme of experimental set-up for measurement of separation a as a

function of load P.

40
displacement [m]

35
unloading

25
loading

15
10
5

after correction

0
0

100

200

400

600

Fig. 9. Plastic deformation of roughness zone of investigated samples.

displacement
signal, a

20

200

normal load [MPa]

displacement sensor

30

300

400

normal load [MPa]

Fig. 8. The dependence of displacement on normal load.

latter only an elastic deformation take place. The difference between

the curves is due to the fact that during loading the elastic and plastic
deformations occur, and while unloading only the elastic. After
complete unloading (0 MPa for the load) the value obtained on the
graph is the plastic deformation of the sample.
It should be noted that, due to a specic experimental device,
the displacement a is not measured directly at the mean planes of
contacting surfaces. The value of a (displacement measured at the
point K, Fig. 7) corresponds to deformation of the whole measuring system i.e. deformation of roughness zone and bulk material of
the sample and deformation of the head B, Fig. 7. Generally the
deformation of all these elements can be composed of elastic and
plastic parts, but plastic strain starts in the roughness zone; when
the load increases the plasticity spreads in the bulk material and
next eventually in the head B. However, in the experiments
performed in this work the level of load is restrained so that the
plastic area does not exceed the roughness zone. The initiation of
plasticity in bulk material was detected by means of observation of
proles crossing a boundary of deformed and non-deformed
roughness zones. To fulll the above condition the maximal
applied load should be different for different steels.
Thus, in the applied load range, the shape of loaddisplacement
curve in Fig. 8 corresponds to limited area of plastic deformation;
it occurs only in the roughness zone. It can be seen that the plastic
part of displacement i.e. its residual value after unloading, is much
smaller than the total displacement corresponding to maximal
load. Moreover, we have observed that the unloading curve is the
same as the reloading curve recorded by a repeated loading up to
the same value of contact pressure. It conrms the observation

that the plastic deformation of a thin surface layer (roughness

zone) and elastic deformation of the whole system (including the
layer) are measured in the experiment and the range of displacement represented by the unloading curve corresponds to a purely
elastic deformation of the system. As the effect of elastic strain in
the thin surface layer is much smaller than that in other components (bulk, head B, etc.), it would be difcult to distinguish elastic
displacement in the roughness zone from that resulting from
deformation of other components of the measuring system. Thus
to proceed further the whole elastic part of the measured
displacement is removed using the following operation. First the
unloading curve is shifted so that it starts from zero ordinate; the
shifted curve represents the value of elastic displacement of the
system versus load. Next the ordinates of the shifted unloading
curve are subtracted from the loading curve ordinates (representing elastic and plastic displacement) and a new curve (denoted as
after correction in Fig. 8) is generated, that with good accuracy
corresponds to plastic deformation of the roughness zone. As the
theoretical models generally refer to this zone, such an approach
enables a fair comparison of experimental data and theoretical
prediction. It is worth noting that for applied load ranges and for
engineering rough surfaces the elastic part of deformation of
roughness zone is much smaller than the plastic one and plastic
deformation starts practically at the beginning of load.
The results of compression experiments i.e. plastic deformation
of roughness for different steels are presented in Fig. 9. The highest
stiffness is exhibited by the 40H steel sample. The behavior of 45
and S235 steel samples is not evident. The whole tension curve for
45 steel lies above that for S235 steel; however the 45 steel sample
is more compliant in low loading range and stiffer for higher loads
than the S235 sample. This effect can be explained when one
compares asperities height distribution and bearing curves of
samples, Figs. 2 and 22.
In the initial stage of approach, the number of asperities is
higher and area of contact is greater for s235 steel than for 45 steel
and this can be a cause of higher stiffness of the S235 sample,
despite the fact that s235 steel is more compliant. On the other
hand, one can observe a clear difference between samples S235
and 40H. As for these samples the surface topography (bearing and
height distribution curves) is practically the same, and the material properties are responsible for different loaddisplacement
relations. Generally, one can conclude that in the performed
experiment a relatively ne difference between samples stiffness
can be detected.

3.3. An attempt to estimate the real contact area (RCA) after the load
on the basis of prolometric measurements
The next part of the experiment was to estimate the real
contact area (RCA). Since in our experimental setup it is not

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

100

50MPa-Distr

70

100MPa-Distr

6
5

200MPa-Distr

60

300MPa-Distr

400MPa-Distr

50

600MPa-Distr

40

700MPa-Distr

30

3
2

20

70

60

50
40

30
20

10
0
0.00

0MPa-Abbott
300MPa-Abbott
0MPa-Distrib
300MPa-Distr

80
material ratio [%]

80

90

height distribution [%]

0MPa-Distrib

material ratio [%]

100

7
0MPa-Abbott

90

height distribution [%]

172

10

0.20

0.40

0.60

0.80

0
1.00

ht /Sz

0
0.00

0.20

dimensionless height (h/Sz)

possible to evaluate the RCA under an applied load we estimate it

on the basis of topography measurement after unloading.
Sandblasted samples of tested steels were loaded from 50 MPa
(small loads, about 1/8 of yield limit) up to 700 MPa (the load nearly
twice above it). After each loading, the sample was measured by the
scanning prolometer (measured area 3.5 mm  3.5 mm).
In Fig. 10 a comparison of the ordinates distribution curves (the
right scale in the gure) corresponding to the different loads (50
600 MPa) is presented.
In the height distribution curve, the part before the load is very
close to the normal distribution (curve 0 MPa-Distr in Fig. 10); a
peak (bulge) appears corresponding to the increased number of zordinates resulting from the displacement of highest ordinates of
asperities, while the lower part of the roughness (valleys) remains
unchanged. This bulge varies with increasing load: it grows and
moves deeper into the roughness zone. It represents surface points
that were shifted as a consequence of contact with the rigid plane;
thus it corresponds to the real contact area, which increases with
increasing load. It is worth noting that the shapes of the distribution curves are very sensitive to the applied pressure. First time,
the deformed surface was measured after 50 MPa load, which
corresponded to only 1/8 of the yield limit of this steel, but the
peak on the distribution curve is very clear.
A more detailed comparison of two bearing curves Abbott's
(material ratio, left scale) and two distributions of ordinates (the
right scale on the gure) specied before and after loading with
nominal pressure of 300 MPa is presented in Fig. 11. In this gure,
it can be clearly seen that changes of surface topography due to
the load are observed in a high part of roughness zone. Almost all
of the bearing curve and distribution curve before and after
loading overlap; the main difference is the bulge formed on the
distribution curve of the deformed surface, which is a result of an
increased number of ordinates in a certain range of roughness
zone. It means that, although the surface of loading stamp is very
smooth, the range where the number of ordinates increases can be
estimated as a few micrometers. On the other hand, the deviation
of the bearing curve of the deformed surface from the initial shape
takes place in the range of heights where the bulge is formed on
the distribution curve.
In view of the above observations, we propose the following
algorithm to estimate the RCA. As in the undeformed state of
roughness, the ordinates exhibit a normal distribution; a value ht
of height should be found for which the actual distribution for
deformed surface starts to deviate from a normal one (or from the
initial distribution, specied before loading). This is marked with
the vertical line in Fig. 11. One assumes that the points having
ordinates higher than ht have changed their position (when
compared with the initial state), due to contact with rigid plane

0.60

0.80

0
1.00

dimensionless height (h/Sz)

Fig. 11. The idea of RCA assessment based on the material ratio and height
distribution.

40
35

RPS 40H

30

RPS S235
RPS 45

25

RCA [%]

Fig. 10. Comparison of the distribution curves for different loads of nominal surface
(50600 MPa) steel 45.

0.40

20
15
10
5
0

200

400

600

800

loading [MPa]

Fig. 12. Dependence of RCA on applied normal load.

in loaded state; hence these points constitute the area of contact

under loading. Next we consider the bearing curve, specied for
the same deformed surface. The RCA is equal to the bearing area
corresponding to the height ht on this curve, cf. Fig. 11, and can be
estimated as about 18%.
The next gure (Fig. 12) presents the change of the RCA,
determined as described above, as a function of normal load of
the sandblasted surfaces. This relationship, taking into account
measurement errors and the dispersion, is close to linear for a
large range of loads.

4. Modeling of contact
4.1. Basic assumptions
The proposed model of contact is based on statistical approach
presented by Grenwood et al. [5,9,10,13], i.e. the rough surface is
considered as a set of asperities that are deformed independently
in a contact process. If for a single asperity one denes a local
function that represents evolution of a specic quantity (contact
area, electrical resistance, loading force etc.) as a function of
displacement, for this asperity (Fig. 13), the evolution of this
quantity for the whole rough surface W(z) can be calculated using
a generalized formulae
Z 1Z 1
Wz
Pz; gf z  d; gdg dz
1
d

where f(z ,g) can be any function dened in terms of local

reference axis variable z0 on a single asperity having a radius g,
P(z,g) is a probability density of existence of asperity having radius

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

173

plane of contact

d (approach)
x

reference plane (mean

plane)

z i+1
z
g

zi

z'

d
z'
Fig. 13. Basic notations used in analysis of roughness.

g at the level z, and W(z) is a function that corresponds to the

rough surface, Fig. 13.
Assuming that a single asperity has a spherical shape the area
of section of single asperity at a distance z0 from its summit is
2z'g, and the bearing area of the surface at a level d with respect
to the reference (mean) plane can be calculated as
Z
Az

1
d

Pz; g2z  ddg dz

It should be noted that in formulae (1) and (2) the mean radius of
all asperities is not required and all asperities with their actual
radii are taken into account.
In many papers [59] the integrals in Eq. (1) are calculated on
the base of random surfaces theory; power spectral density
function is used and Gaussian distribution of asperities heights
and ordinates heights is assumed.
In the proposed approach the surface characteristic is based on
direct analysis of measured surface points, e.g. in analysis of
normal contact, for each level zi the points that are the summits
and that lie in the range zi, zi z are selected; next the radius of
each summit is calculated; then for any value of the approach the
number of asperities in contact and their deformation can be
estimated, Fig. 13. The contact forces generated at each asperity for
the assumed approach are summarized and the total contact load
is calculated. The integrals in formulae (1) are replaced by double
sums, (3).
In the case of contact problems, assuming that L(zi) denotes
number of summits between an ordinate zi and zi z, gli is the radius
of lth asperity in the interval (zi; zi z), and M is the number of
intervals on the z axis that corresponds to the applied approach d,
Fig. 13, for elastic contact, the contact load, contact area and contact
stiffness can be calculated using the Hertz solution.
For the elasticplastic contact, where the contact process is
governed by a function f(z,g) that can be specied numerically and
describes the loadinterference or loadcontact area relation of a
single asperity, the following formula is used to asses a deformation of rough surface:
M

Lzi

Fd f zi  d; g il ;
i1l1

for zi Z d

It is evident that the result of calculation depends on the

assumed value of z and values of gli that in turn, similar to L(zi),
depends on the sampling interval. Using the proposed approach, a
purely geometric analysis of the rough surface can also be
performed, e.g. area of section of the surface at the level d above

the mean (reference) plane (red dotted sections, Fig. 13) can be
predicted as
M

Lzi

Ag d 2zi  dg il ;
i1l1

and then, using different values of d, the whole bearing (Abott)

curve can be modeled.
On the other hand, a point of the bearing curve corresponding
to the section at the level d can be determined using its denition
i.e. as a product of total number of ordinates Nd having the value d
and elementary area dxdy. Comparison of Abott curves determined
using these two methods can be considered as a validation of the
proposed geometrical model of the rough surface.
In the geometrical model of the surface an important point is
the denition of asperity summit and as a consequence, the
method of calculation of asperity radius. Two approaches are
possible; in the rst one, presented in many papers concerning
random surface theory [114], the denitions can be explained on
the basis of 2D analysis; the point of prole is a summit (peak in
2D) when its height is greater than that of two closest points, and
a peak curvature at z0 is specied using the formula
k 1=g Z 1 2Z 0 Z  1 =dx2 ;

where dx is the sampling interval.

It should be noted that, as shown in Section 3, both matrix gij of
summits radii and summits distribution (summits density) L(zi)
depend on the sampling interval. Both theory and experiment
show that as the sampling interval is reduced, the peak density
increases indenitely; the rms prole curvature may vary by a
factor of 100 and the mean peak curvature varies in the same way,
[21]. Thus there are different sets of gij and L(zi) for different values
of sampling interval and an important question is which value
should be taken for proper contact analysis.
4.2. Actual shape of asperities that inuence contact process
The other approach has been proposed by Greenwood an Wu
(2001), [21], where the authors criticize the denition of peak
presented in their previous papers; they claim that peaks or
summits dened according to the above denition do not represent the asperities and correspond to artifacts at the surface,
especially when a small sampling interval is used, 3- point peaks
are just an artefact of the prole, not real physical features, c.f. [21].
The idea of specication of asperity radius presented in [21] can be
summarized as follows. The contact between a rough, deformable
surface and a plane, rigid surface is considered. The mechanical
response (deformation) of asperity above the line AB, cf. Figs. 13

174

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

and 14 will be close to that for a smoothly rounded asperity that

passes through A and B and has the same area Ac. Assuming a
parabolic shape of the smooth asperity between A and B, the
curvature of rough asperity is dened as
 
AB l;

1
Ac 12
l

12Ac =l

In [21] a single prole composed of 1200 points has been

analyzed and a relation between asperity lengths and asperity
curvatures is presented for asperities laying at all heights. It can be
observed that there is a variation in curvature for asperities of
particular size and larger asperities have smaller curvature. This
relation is similar to the variation of rms curvature for proles
with sampling interval assumed as l/2. In paper [21] the proposed
approach has been analyzed theoretically and has not been
applied to solve an actual contact problem.
It is evident that heights distribution of asperities and their
radii (mean radius) specied using this procedure depend on the
height where the rigid surface (line AB) is placed (level of section
of rough surface), denoted as d in Fig. 13.

Ac

l
Fig. 14. Denition of asperity, Greenwood 2001.

The approach proposed in [21] is analyzed in more detail in this

paper using a prole (or several proles) of sand blasted surface.
Special software has been developed in the framework of Mathematicas platform. The goal is to check if this procedure can
replace the well known three points peak denition and consequently if it can be applied to specify a shape and distribution of
asperities required in analysis of contact.
In Fig. 15 a mean value of asperities radius, Rm for different
(levels) heights of prole section is presented (level 0 corresponds
to the mean line of the prole).
One can observe that for the sections between 0.006 and
0.016 mm (maximal height) the variation of Rm is rather low, but
below the height of about 0.006 mm the mean radius increases
rapidly when the level of surface section is placed lower, as for
lower levels of surface section the asperities that lie above a single
segment of section join and then two or more asperities are
detected as one asperity, Fig. 16.
Thus, it is not evident which height of section should be
selected to obtain proper values of asperities radii and height
distribution in the whole roughness zone.
In Fig. 17, the total number of asperities Ng(Z), calculated as
number of separate segments of prole section, is presented.
These asperities lie above the level of section. The maximum
asperities number occurs for the section at the mean line level and
for successively lower levels the total number of peaks decreases.
This is again an effect of joining of asperities below the mean line,
Fig. 16. This result is opposite to that observed when asperities are
dened as three-point peaks, where the number of peaks above
any level can be calculated using summation (integration) of
Gaussian distribution of peak heights and therefore it monotonically increases with decreasing level of section, cf. Fig. 6.
The heights distribution of asperities (peaks) dened according
to the procedure presented in [21] is shown in Fig. 18a and b for
different heights of section, denoted as d_. Due to the specic
denition, only the peaks lying above the selected section height
are taken into account in the distribution diagram. It is evident
that the distribution depends on level of section; when the height
d of potential (ctitious) rigid surface (Fig. 13) decreases, the
number of peaks in each of intervals z above this surface

mean asperity
radius [mm]

0.2

roughness zone

0.15
0.1
0.05
0
-0.01
0
level of section [mm]

-0.02

-0.02

0.01

-0.01

0.01

Fig. 17. Number of all asperities dened according to [21] for different levels of
roughness section.

0.007

0.012

0.0065
0.13

0.14

0.02

level of section [mm]

Fig. 15. Mean asperity radius calculated using (6) for different levels of sections AB
of prole.

0.15

0.006

0.16

0.0055

0.008

0.005
0.0045

0.006
0.004

70
60
50
40
30
20
10
0

number of asperities
above section

0.3
0.25

1.17

1.18

1.19
0.0035

1.21

Fig. 16. Fragments of proles dened as one asperity for lower levels of roughness section.

1.22

d_0.0093
d_0.0073
d_0.0053
d_0.0033
d_0.0013

20

bearing area %

number of peaks

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

15
10
5
0

-0.02

-0.01

0.01

0.02

-0.005

number of peaks

z[mm]

d_0.0013

20

d_-0.0007

15

d_-0.0067

-0.01

model
experiment

0.005
h[mm]

0.01

0.015

4.4. Proposed method of specication of sampling interval

10
5
0

-0.02

80
70
60
50
40
30
20
10
0

Fig. 19. Comparison of experimental and modeled bearing curve.

d_-0.0027
d_-0.0047

175

0.01

0.02

z[mm]
Fig. 18. Asperities height distribution: (a) for section levels 0.00930.0013 mm and
(b) for section levels -0.0067-0.0013 mm.

changes; at rst it increases and then decreases. Thus the question

which level of section should be selected to specify asperity
heights distribution and mean asperity radius appropriate for
modeling of contact cannot be answered as yet.
One can observe that the distributions corresponding to different levels of section differ from the Gaussian distribution assumed
in random surface theory, especially for lower section levels.
4.3. Verication of the geometrical model of roughness zone
It should be noted that a verication of models of contact that
are based on the fundamental works of Greenwood, Nayak,
Archard (statistical models) can be performed in two stages. In
the rst stage, the purely geometrical analysis is performed and
Eq. (4) is applied to model the bearing curve of the surface. On the
other hand, the bearing curve can be specied using its denition
i.e. as an integral of ordinates distribution of the measured rough
surface and it is considered as an experimental bearing curve. In
the second stage the mechanical material properties are also
included and the real contact area, contact load and contact
stiffness as functions of approach d can be specied using Eq.
(3). These functions in turn can be specied from experiments. In
both stages the same asperity height distribution and asperity
radii distribution should be taken into account.
To check if the results presented in Figs. 15 and 18 (mean
radius, peak heights distribution) can be applied to model the
geometrical structure of sand blasted surface, formula (4) was
used to model a bearing curve (based on selected prole). The
section of prole at height d 0.003 mm has been chosen as a
representative for specication of asperities heights and radii
distribution. The simulated bearing curve is compared in Fig. 19
with bearing curve denoted as the experimental one, i.e. calculated
accordingly to its denition using prole ordinates.
One can observe that both curves match well only at height
0.0033 mm, that is at the section level d chosen for calculation of
mean radius and height distribution. In other regions of the
roughness zone the accuracy of the modeled bearing curve is not
satisfactory. Thus the approach presented in [21] is not precise
enough to model geometrical structure in the whole range of
roughness zone (its bearing curve), when formula (4) derived from
the statistical model is used.

Other possibility is to use a classical denition of peak to dene

asperities heights and radii distributions. It has been shown in
Section 3 that in this case the results strongly depend on the
sampling interval. As in commercial softwares the details of
calculation algorithms are generally not reported, in the framework of Mathematicas programming a special software has been
developed that enables calculation of some statistical parameters
(especially peak radii and heights distribution) in the case of both
2D (prole) and 3D (surface) analyses of roughness. As stated in
[21], for small sampling intervals the three points peak (or ve
points summit) does not correspond to actual asperity. On the
other hand, to fulll the basic assumptions of random surfaces
theory (e.g. Gaussian peak heights distribution) a small sampling
interval is required. The direct observation of proles measured
with different sampling intervals indicates that when the sampling
interval increases, the three-point peak is closer to the actual
shape of asperities that are important in the contact process,
Fig. 20b. However, when the sampling interval is further increased,
some important asperities are neglected in the model, Fig. 20c.
Thus the proper assumption of sampling interval is crucial in
modeling surface topography and contact problems when the
statistical model with three-point peak denition is taken into
account. In the proposed solution of contact problem the denitions applied in statistical models and formulae (3) and (4) are
used with appropriate selection of sampling intervals.
The usefulness of this approach in modeling the rough surface
has been checked applying the rst stage of verication described
above. For a selected prole the peak heights and peak radii
distributions have been specied using different sampling intervals and next used to model (predict) the bearing curves of the
prole accordingly to Eq. (4). Decrease of number of three-point
peaks is observed when the sampling interval increases. It should
be noted that the peak distribution is closer to the Gaussian
distribution for smaller rather than for bigger sampling intervals.
It has been observed that in a purely geometrical modeling of
rough surfaces (simulation of bearing curve) the sampling interval
does not inuence the results. This observation is similar to that
presented in Section 3, where the shape of experimental bearing
curves (specied according to their denition in 3D analysis) does
not depend on the sampling interval. Thus the condition of a
proper modeling of bearing curve cannot be used as a criterion of
choice of the sampling interval.
The proposed method of proper specication of sampling
interval is based on other geometrical characteristics of the rough
surface. As an example of the proposed idea an analysis of several
proles of a sand blasted surface will be used. To make the
calculations more representative, three random proles selected
from 3D surface measurement (about 400 proles) have been
joined to compose one long (about 10 mm) prole that is used in
subsequent considerations.

176

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

0.015

N g(Z), N(Z)

Z[mm]

dx=0.006mm

0.01
0.005
0.8

0.9
0.005

1.1

1.2

1.3

X [mm]

0.01

-0.01
Z[mm]

500
450
400
350
300
250
200
150
100
50
0

Greenwood 2001
dx=0.008
dx=0.012
dx=0.006

0.01

0.02

Z [mm]

0.015
0.01

Fig. 21. Total number of asperities on prole (calculated using different denitions)
above level Z of section.

dx=0.012mm

0.005
0.8

0.9
0.005

1.1

1.2

1.3

X [mm]

0.01

Z [mm]

dx=0.032mm

0.015
0.01
0.005
0.8

0.9
0.005

1.1

1.2

1.3

X [mm]

0.01

Fig. 20. Comparison of prole measured with high resolution (0.0002 mm, black
line) with proles measured using different sampling intervals: (a) dx 0.006 mm,
(b) dx0.012 mm, and (c) dx 0.032 mm.

Using the peak heights distributions, the total number of all

three-point peaks that lie above any level Z of horizontal section of
rough surface is calculated as an integral (sum) of peaks height
distribution between the level Z and the highest ordinate of the
surface:
M

NZ Lzi for
i1

zi Z Z

The functions N(Z) (results of summation), for sampling intervals

dx 0.006, 0.008, 0.012 mm, are presented in Fig. 21 and compared with Ng(Z) already shown in Fig. 18 specied using the
approach proposed in [21] and discussed in the previous section
(Figs. 1519). As stated previously, the functions N(Z) have a
strictly monotonic character, while Ng(Z) has a maximum at about
0.002 mm above the mean line.
The best coincidence of function N(Z) with Ng(Z) (its monotonically decreasing part) is observed for dxE 0.0080.012 mm;
thus it can be concluded that for this range of values of the
sampling interval, the three points peaks correspond to actual
asperities that should be taken into account in analysis of contact.
On the other hand, the curve N(Z) monotonically decreases in the
whole range of roughness zone; thus it corresponds to the actual
distribution of asperities heights. Therefore the asperities distribution that is calculated on the base of three-point peak has a
general character and does not depend on level of section as in the
case of approach proposed in [21]. The measurements corresponding to sampling intervals dx 0.012 mm and dx 0.0002 mm
(highest resolution) have been compared in Fig. 20b.
It is well known that statistical parameters specied on the
basis of proles are different from those calculated using surface
analysis [6]. In the proposed method the prole analysis is applied

to specify optimal sampling interval for surface measurement. It

can be justied as follows. In the case of isotropic surfaces, when
the prole is sufciently long, it can be regarded as a representative for this surface; therefore the sampling interval dx0 specied
using the above procedure should be the same for all proles
made in any direction. In particular the same sampling interval dx0
will be obtained in two perpendicular X and Y directions. Let us
consider a prole in the X direction; if there are n three-point
peaks on the prole, only a smaller number n1 of these n peaks
will constitute a summits detected in 3D surface analysis. The
point of A(xA, yA, and zA) of rough surface that constitutes a peak on
a prole in X direction is dened as a summit for surface when it
also constitutes a peak on a prole, that goes through point A in Y
direction. As for both X and Y directions the same sampling interval
dx0 is valid; it can be used in 3D surface analysis, where the summit
radius can be calculated as the mean of radii in X and Y directions.
The peak radius is dened in the proposed approach as the
radius of parabola determined by three points that dene the peak
for the selected sampling interval.
4.5. Characteristic of surface topography for investigated samples
Once the sampling interval is specied using analysis of a one
long prole (composed of several proles), the 3D analysis of
topography of each surface is performed. The ve-point summit
denition has been applied and a distribution of heights and radii
of the summits has been calculated for each sample assuming the
sampling interval specied previously i.e. dxdy 0.012 mm. The
software developed in Mathematicas framework has been used.
As an example, the summits (asperities) height distribution for
surfaces of 45, 40H and S235 steel samples for 3 mm  3 mm
measured area is presented in Fig. 22. One can observe that in the
rst phase of contact, for z/Sz greater than about 0.13, for steels
40H and s235, the increase of number of asperities (summits) with
growing approach is similar and is greater than in the case of steel
45. The maximal number of summits occurs for steel 40H and the
minimal for steel S235.
For the assumed sampling interval, the value of mean asperity
radius equals about 35 m and is similar for all considered surfaces.
4.6. Mechanical part of model deformation of single asperity
In statistical models the deformation mode of whole surface is
evaluated on the basis of deformation of single asperity and
statistical asperities distribution. In the fundamental models by
Greenwood et al. only an elastic deformation, according to the
Hertz theory, was assumed. In numerous subsequent papers
(Chang and Larsson, [16,25]) more advanced (elasticplastic,
plastic) material models have been included and simplied formulae have been derived to describe the behavior of a single asperity.

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

177

number o summits

250
45

200

H40

150

S235

100

50
0

-0.3

-0.1

0.1

0.3

0.5

dimensions less height z/Sz

Fig. 22. Heights distribution of ve-point summits for all samples, dx 0.012 mm.

similarly as in [34]. It can be seen that, in the normalized diagram,

the relation
n

F n h
n

h 3=2

Fig. 23. Deformation of single asperity, steel S235.

2.5
steel 40H

F[N]

2.0
1.5

steel 45

1.0
0.5

steel S235

0.0
0

10
-3

h[mm x 10 ]

25

a [mm x 10-3]

20
15
10

geometrical section
h40

S235

0
0

10
-3

h[mm x 10 ]

1.2
1

* 3/2

0.8
0.6

F /(h )

Larsson et al. [33] have proposed general formulae that describe

plastic deformation of single asperity (paraboloid) for exponential
plastic hardening of material. An exhaustive FEM analysis of hemispherical contact that can be used to model deformation of single
asperity has been presented in [34]. In the paper the elasticperfectly
plastic material model has been applied, and ve different material yield strengths were considered. The results were normalized
with values of critical load, approach and contact area that
correspond to the initial point of yielding.
In the proposed approach, the deformation of single asperity in
contact with rigid plane is simulated in the whole elasticplastic
range using the nite element ABAQUS code. A ne mesh
composed of 13,476 quadrilateral four nodes elements is used
and at the maximal load about 230 elements are in contact with
the rigid plane. The HMH plastic yield criterion and plastic ow J2
theory have been applied. The stressstrain curves presented in
Fig. 1 are assumed to be the material characteristic. The boundary
condition and an example of deformed asperity are presented in
Fig. 23. Similarly as in the elastic Hertz model, a deformation half
of sphere is taken into account; however in contrast to the Hertz
solution, the lower edge of asperity model is xed in both r and Z
directions. It enables accounting for a restriction of asperity
deformation in lateral direction, due to presence of other asperities in a neighborhood. Thus one assumes that for large loads the
deformation of a single asperity in lateral direction is not independent of other asperities. The same boundary conditions were
assumed in [34]. The FEM mesh for asperity model in deformed
state is presented in Fig. 23, where the contour lines for reduced
(Mises) stress are also shown. One can observe that while the base
of asperity does not increase its diameter, the dimension of the
deformed asperity increases in lateral direction above the base. It
can be considered as the mean inuence of neighboring asperities.
Two functions for single asperity are specied using FEM: the
loaddisplacement (Fh) and contact radiusdisplacement (ah)
curves that are shown in Fig. 24a and b for asperity radius
R 30 m and for all considered materials.
One can observed that the function Fh strongly depends on
the material properties of asperity; stiffer material yields a stiffer
behavior of asperity but relation ah is only slightly inuenced by
material properties and is close to the ah generated by a simple
geometrical section of non-deformed asperity.
The presented results are similar to those reported in [34].
It can be seen in Fig. 24c, where the F(h) relations for each steel are
scaled using critical values of displacement and load [34], that


CSy
3=2
hc
8
R; F c 4=3R1=2 E0 hc
2E0

steel s235
steel 45
steel H40

0.4
0.2
0
1

10

100

1000

10000

h*
Fig. 24. Functions F(h), a(h), and Fn(hn)/(hn)3/2 for single asperity.

4.7. Modeling results for rough surface comparison of theoretical

and experimental results

FRi

FRj R2i
R2j

hRi

hRj Ri
Rj

10

These functions were next applied in Eq. (3) to calculate the

nominal contact pressureapproach and nominal contact pressurereal contact area relations for the investigated rough surfaces. Comparison of these theoretical and the experimental
results specied in Section 3 is presented in Figs. 25 and 26.

10

model
model

8
6

experiment
experiment

4
2

steelH40
H40
40H
steel

-3
-3

500
normal load [MPa]

experiments

12
10
8
6

model

4
2

steel 45

14
12
10
8
6
4
2
0

200
400
normal load [MPa]

600

model

experiment

steel 235
0

200
400
normal load [MPa]

5. Discussion

600

Fig. 25. Loadapproach curves for the analyzed steels.

A r [%]

40
35
30
25
20
15
10
5
0

40H exp.
S235 exp.
h40 model
S235 model

200

400

600

800

nominal pressure [MPa]

35

A r [%]

It is evident that pd relations are predicted with better

accuracy than pAr relations for all considered materials. The best
accuracy of prediction has been achieved for steel s235 for pd
relation as well as for pAr relation. The worst accuracy is observed
in the case of samples made of steel 40H. This fact can be
explained as follows. In the proposed model a single asperity
characteristic is calculated using nominal material properties
presented in Fig. 1, specied in tension tests. However after the
sandblasting process the local hardening of material take place,
and the actual material of asperities is stiffer than that investigated
in tension test. The increase of stiffness due to sand blasting
treatment depends on initial material properties; a lower increase
can be expected in the case of steel s235, and a higher in the case
of steel 40H. Thus the discrepancy between simulated and
experimental pd, pAr relations can be a result of difference
between the actual material properties in roughness zone and the
nominal properties assumed in the model. The effect of a gradient
of material properties in the roughness zone on the loadreal
contact area relationship has been demonstrated also in [35].
It is worth noting that the proposed approach can be applied in
different scales, i.e. for more accurate analysis of very small loads
and small displacement of the rigid surface (initial stage of
contact) the coincidence of curves N(Z) with Ng(Z) is required only
in the zone of higher roughness e.g. in the range 0.010.02 mm,
Fig. 21. The best t of N(Z) to Ng(Z) in such a limited range will
yield a different value of dx than that specied above, as in this
range ner asperities are deformed.

1000

14

-3

Once F(h) and a(h) functions are specied for an asperity of

selected radius; they should be scaled to obtain characteristics for
asperities of any radius, as in Eq. (3) all asperities are taken into
account in the sum. Towards this aim one can use the solution
presented in [33], where it has been shown that for plastically
deformed asperity the relationship F=R2  h=R depends only on
material properties. Hence, if F(h) is known for RRj it can be
simply computed for R Ri (Ri aRj):

12

displacement [mm x 10 ]

where hn h/hc, Fn F/Fc is close to that presented in [34], however

small differences can be observed, that is for hn 100 (log scale) the
values of normalized F are slightly higher and a difference between the
analyzed materials is slightly greater than that shown in [34]. It can
result from the fact that in the present simulation a strain hardening of
material has been taken into account, while in [34] a perfectly plastic
material model has been assumed. It should be noted that according
to the authors experience the strain hardening inuence also characterizes the relation loadcontact area, namely the mode of deformation of hemisphere that exhibits an important strain hardening is
closer to that described as mostly elastic deformation than that
described as mostly plastic deformation in [34]. It should be noted
that for s235 steel and R30 m, hc equals 0.037 nm, this is not
measurable in our experimental setup.

displacement [mm x 10 ]

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

displacement [mm x10 ]

178

30

45 exp.

25

S45 model

20
15
10
5
0
0

200

400

600

800

nominal pressure [Mpa]

Fig. 26. Loadreal contact area curves for rough surfaces on analyzed samples:
(a) s235 and 40H and (b) steel 45.

S. Kucharski, G. Starzynski / Wear 311 (2014) 167179

6. Conclusions
In the paper the contact of stiff rigid plane with rough surfaces
manufactured using different steels has been investigated experimentally and theoretically. In experiments a special approach is
applied where the elastic deformation of samples and experimental setup is eliminated and both experimentally determined
relations: contact pressureapproach (d) and contact pressure
real contact area (Ar) correspond to purely plastic deformation of
rough surface. It should be noted that in the case of investigated
materials and applied loads, the elastic part of deformation of
roughness zone (but not elastic deformation of bulk material and
whole experimental setup) is small when compared with plastic
one and can be practically neglected. The real contact area is
estimated experimentally from comparison of ordinates distribution of deformed and non-deformed surfaces. The theoretical
prediction of the functions determined experimentally has been
based on FEM analysis of single asperity deformation and the use
of asperities radii and heights distribution. These functions in turn
were specied directly using surface points, measured with specially selected sampling intervals. The ve points summit denition
is applied. The procedure for estimation of proper value of
sampling intervals that enable an account of the asperities that
are important in contact process has been proposed. Similarly
as in experiments the model prediction corresponds to the plastic
deformation of the roughness zone, and the bulk material is not
taken into account.
The agreement of the results obtained by means of the
proposed simple model and experimental results is generally
satisfactory. As a better coincidence is observed in the case of
steel that exhibits lower plastic hardening one can conclude that
the model could be improved if the evolution of material properties (local hardening) due to surface nishing would be taken into
account. On the other hand an interaction of asperities in lateral
direction should be taken into consideration using more sophisticated boundary conditions in analysis of a single asperity. This
problem will be a subject of the subsequent papers.
The proposed approach can be applied in different scales of
roughness deformation.

Appendix A
Denitions of the statistical parameters describing surface:
Mean surface roughness
Sa

1 N 1 M 1
jzx;y j
NM x 0 y 0

Standard deviation of the height distribution or RMS surface

roughness
s
1 N 1 M1 2
Sq
z
NM x 0 y 0 x;y
Skewness of the height distribution. Third statistical moment,
qualifying the symmetry of the height distribution
"
#
1 1 N1 M 1 3
Ssk 3
zx;y
Sq NM x 0 y 0
Kurtosis of the height distribution. Fourth statistical moment,
qualifying the atness of the height distribution
"
#
1 1 N 1 M1 4
Sku 4
zx;y
Sq NM x 0 y 0

179

where N is the number of points in prole and M is the number of

proles.

References
[1] K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge,
1985.
[2] J.R. Barber, Bounds on the electrical resistance between contacting elastic
rough bodies, Proc. R. Soc. Lond. Ser. A 459 (2003) 5366.
[3] M. Paggi, J.R. Barber, Contact conductance of rough surfaces composed of
modied RMD patches, Int. J. Heat Mass Transf. 4 (2122) (2011) 46644672.
[4] M.G. Cooper, B.B. Mikic, M.M. Yovanovich, Thermal contact conductance, Int. J.
Heat Mass Transf. 12 (1968) 279300.
[5] J.A. Greenwood, J.B.P. Williamson, The contact of nominally at rough surfaces,
Proc. R. Soc. Lond. Ser. A 495 (1966) 300319.
[6] P.R. Nayak, Random process model of rough surfaces, J. Lubr. Tech.(ASME) 93
(1971) 398407.
[7] P.R. Nayak, Some aspects of surface roughness measurement, Wear 26 (1973)
165174.
[8] D.J. Whitehouse, J.F. Archard, The properties of random surfaces of signicance
in their contact, Proc. R. Soc. Lond. Ser. A 316 (1970) 97121.
[9] J.A. Greenwood, A unied theory of surface roughness, Proc. R. Soc. Lond. Ser. A
393 (1984) 133157.
[10] J.A. Greenwood, J.H. Tripp, The contact of two nominally at rough surfaces,
Proc. Inst. Mech. Eng. 185 (1970) 625633.
[11] A.W. Bush, R.D. Gibson, T.R. Thomas, The elastic contact of rough surfaces,
Wear 35 (1975) 87111.
[12] M. Paggi, M. Ciavarella, The coefcient of proportionality k between real
contact area and load, with new asperity models, Wear 268 (78) (2010)
10201029.
[13] J.A. Greenwood, A simplied elliptical model of rough surface contact, Wear
261 (2006) 191200.
[14] M. Ciavarella, J.A. Greenwood, M. Paggi, Inclusion of interaction in the
Greenwood and Williamson contact theory, Wear 265 (2008) 729734.
[15] J.A. Greenwood, C. Putignano, M.A. Ciavarella, Greenwood & Williamson
theory for line contact, Wear 270 (2011) 332334.
[16] W.R. Chang, I. Etsion, D.B. Bogy, An elasticplastic model for the contact of
rough surfaces, ASME J. Tribol. 109 (1987) 257263.
[17] Y. Zhao, D.M. Maietta, L. Chang, An asperity microcontact model incorporating
the transition from elastic deformation to fully plastic ow, ASME J. Tribol. 122
(2000) 8693.
[18] R. Buczkowski, M. Kleiber, A stochastic model of rough surfaces for nite element
contact analysis, Comput. Methods Appl. Mech. Eng. 169 (1999) 4359.
[19] R. Buczkowski, M. Kleiber, Elasto-plastic statistical model of strongly anisotropic rough surfaces for nite element 3D-contact analysis, Comput. Methods
Appl. Mech. Eng. 195 (2006) 51415161.
[20] T.R. Thomas (Ed.), Longman, London, 1982.
[21] J.A. Greenwood, J.J. Wu, Surface roughness and contact: an apology, Meccanica
36 (2001) 617630.
[22] T.R. Thomas, D.G. Rosen, Determination of the optimum sampling interval for
rough contact mechanics, Tribol. Int. 33 (2000) 601610.
[23] A. Pogacnik, M. Kalin, How to determine the number of asperity peaks, their
radii and their heights for engineering surfaces: a critical appraisal, Wear 300
(2013) 143154.
[24] P. Pawlus, W. Zelasko, The importance of sampling interval for rough contact
mechanics, Wear 276277 (2012) 121129.
[25] M. Scaraggi, C. Putignano, G. Carbone, Elastic contact of rough surfaces: a
simple criterion to make 2D isotropic roughness equivalent to 1D one, Wear
297 (2013) 811817.
[26] G. Zavarise, M. Borri-Brunetto, M. Paggi, On the resolution dependence of
micromechanical contact models, Wear 262 (12) (2007) 4254.
[27] E. Ciulli, L.A. Ferreira, G. Pugliese, S.M.O. Tavares, Rough contacts between
actual engineering surfaces. Part I, II, Wear 264 (2008) 11051128.
[28] H. Aramaki, H.S. Cheng, Y. Chung, The contact between rough surfaces with
longitudinal texture. Part I. Average contact pressure and real contact area,
ASME J. Tribol. 115 (1993) 419424.
[29] B. Buchner, M. Buchner, B. Buchmayr, Determination of the real area for
numerical simulation, Tribol. Int. 42 (2009) 897901.
[30] N. Hansen, Der Boschungswinkel von Rauheitsprolen, Maschinenmarkt 73
(1967) 17901791.
[31] S. Belghith, S. Mezlini, H. Belhadj Salah, J-L. Ligier, Modeling of contact between
rough surfaces using homogenization technique, C.R. Mec. 338 (2010) 4861.
[32] F. Robbe-Valloire, Statistical analysis of asperities on a rough surface, Wear
249 (2001) 401408.
[33] J. Larsson, S. Biwa, B. Storakers, Inelastic attening of rough surfaces, Mech.
Mater. 31 (1999) 2941.
[34] R.L. Jackson, I. Green, A nite element study of elasto-plastic hemispherical
contact against a rigid at, J. Tribol. 127 (2005) 343354.
[35] M. Paggi, G. Zavarise, Contact mechanics of microscopically rough surfaces
with graded elasticity, Eur. J. Mech./A Solids 30 (2011) 696704.