Vous êtes sur la page 1sur 5

Int. J. Mach. Tools Manufact, Vol. 38, Nos 5-6. pp. 567 571.

1998
1998 Elsevier Science Ltd. All rights reserved
Pergamon Printed in Great Britain
PII: S0890-6955(97)00103-X 0890-6955/98 $19,00 + 0.00

Fractal Dimension Measurement of Engineering Surfaces


J o h n C. Russ

Taylor Hobson Ltd, Leicester, England, and


Materials Science and Engineering Dept., North Carolina State Univ., Raleigh, NC USA

Summary

Many types of engineered surfaces have been observed to exhibit a fractal geometry. In some cases,
modeling the generation of the surfaces predicts this and provides correlation between the dimension
and history and properties of the surfaces. The complex process of machining precision surfaces
does not yet lend itself to such detailed modeling, but observations of the correlations between
production methods and surface dimension, independent of material type and hardness, offer
encouragement. ~ 1998 Elsevier Science Ltd

Introduction mean variance vs. sampled area, the Minkowski


method or range vs. neighborhood size, or a
Observations indicate that many surfaces are Richardson measurement of the outlines of slit
not Euclidean (defined by planes and analytical islands. Because in general different methods
functions) but instead have a fractal geometry. produce numeric values that cannot be
The reasons for this are not well understood, compared to each other, within each data set
but presumably arise from the operation of a only a single method appropriate to that
hierarchy of processes that dissipate energy particular application has been used.
chaotically at many dimensional scales. The
earliest papers in this field, only a few years Fracture and Wear
ago, were primarily discoveries of the existence
of fractal behavior in a wide variety of different One of the most widely studied classes of
circumstances. The next important step in fractal surface structures is that of deposited
uncovering an understanding of the causes and coatings. Agglomeration by diffusion, ballistic
importance of this behavior is to find impact, and chemical or electrochemical
functional dependencies and relationships that processes often results in self-similar structures.
correlate the fractal dimension of surfaces with Models explain these processes well and predict
their specific history and performance. quantitatively many properties of the resulting
surfaces, and they are not considered further
There are numerous procedures available for here. A comprehensive review can be found in
measurement of the fractal dimensions. Some (Vicsek, 1992).
are inappropriate for surfaces, which are
generally self-affine rather than self-similar. Another of the classes of surfaces known to
Others are unsuitable for surfaces that are exhibit fractal geometry is brittle fractures in
anisotropic. Most of the different techniques materials. As shown in Figure 1, fractures in
estimate bounds of the dimension, and the metals, glasses and ceramics create surfaces
methods have varying levels of numeric whose energy absorption is related to the
precision. It is not the purpose of this paper to dimension of the surface. The correlation
compare the various methods, which has been includes materials that are crystalline as well as
done elsewhere (Russ, 1994). All of the data amorphous, with atomic bonds that are
reported here used either the slope of the covalent, ionic and metallic, based on data
Fourier transform power spectrum, a plot of obtained by three different research groups.

567
568 J.C. Russ

wear processes as the tool tip (in these examples


5OO i • i • i • i • i • i • i • i • i
a diamond cutting tip) interacts with the
0 glass cerami~1
n alumina material. The tool also wears in the process,
450 vanadium-gol
dulling the cutting edge and m o d i f y i n g the tool
4oo flank which rubs against the surface created in
the machining process. A plot of the cutting
force and the thrust force as a function of the
g A
distance the tool has traveled on the part is
3.0O
shown in Figure 2 for two different materials.
The tool forces are observed to monotonically
,.,- 2 . s o A / A []
increase with tool wear.
2.00

1.50 ~-~-~ Cu - cutting force

-->-- Cu - thrust force


100 . . . . . . .
100 lOB 1.10 1.15 1.20 1.25 1.30 %35 1.40 ..... AI - cutting force
Fractal Dimension ~ 1 0 _

-~ AI - thrust force
Figure l. Correlation between material fracture
(
toughness and the fractal dimension of the brittle "d
fracture surface produced, for different materials.
5-

Fracture is not widely used at the present time


as a means of producing engineering surfaces,
of course. But it was at one time. Stone tools ~ . ~ ~ - - - ~ _
were shaped using fracture, and the surfaces of
the tools were then subjected to wear. Different
Distance(km)
types of tool use have long been known to
Figure 2. Cutting and thrust forces on single-point
produce signatures on the tool surface that can diamond tools cutting aluminum and copper as a
be recognized by experts. These surfaces are function of the distance the tool has traveled.
fractal, and the wear process reduces the fractal
dimension of the as-made surface producing
.--c~-- Cu - Ra
values that are specific to the type of use as
shown in Table I. ~ - Cu - Rq
30.
~ - - AI - Ra
~'* ___ _
Table 1. Fractal dimensions measured on modern stone g ~ AI- Rq
v /
tools made from chert and subjected to different usages. ~20- / /
Stone Tool Usage
Unworn Flint
Saw Bone
Scrape Hide
Scrape Oak Wood
Fractal Dimension
2.442
2.382
2.348
2.295
1o- ~

o
-

i J ,
1
Because of the variety of different materials
Distance(kna)
used to fabricate these tools, and the wide range Figure 3. Conventional statistical roughness measures
of uses that can modify the surface geometry, on the surfaces produced by the tools and materials in
there is no one-to-one match between Figure 2. The drop in amplitude at long distances is
dimension and tool use. That is, there are many presumably due to heating and ductile surface
combinations of material and use that can deformation by a worn tool.
produce a particular surface fractal dimension.
But there is consistency, in that each particular However, the surface finish that is produced
material and process produces surfaces with a undergoes a more complex evolution. Figure 3
very consistent dimensional value. shows the conventional statistical surface
roughness parameters Ra and Rq for the two
The precision machining of surfaces is materials, and Figures 4 and 5 show the
accomplished by a combination of fracture and variations in their fractal dimension. These are
Fractal dimension measurement 569

, . , , , , , , , .
2.30
2"301 ~ - - CU 0 Aluminum
[] Copper
AI "[] 2.27 []
2.25 ~
2.25 ~
D ,.:., / - 4 ~
[3-~- E::]
2.22 []

"~ 2 . 2 0

2.15 CIr.. 2.17 0 0

2.15
2.10
2.12 0
Oistanee(hn) D = 2 . 2 5 7 - . 0 0 3 ° Rq (nm); R 2 = . 6 4 5
Figure 4. Fractal dimension of surfaces produced by 2.10 , , , , ,
the tools and materials in Figures 2 and 3. Notice that 5 10 15 20 25 30 35

the two materials behave differently as tool wear Rq (nm)


increases. Figure 6. Partial correlation between Rq and surface
fractal dimension using the data in Figures 4 and 5. In
general, the smoothest surface in a classical sense has
[] Cu the highest (least correlated) fractal dimension.
-- -- AI []
/ Machined surfaces
1.12, / To consider the question of whether different
[] z . . . . 4::L~__ / surface preparation methods produce
1,10,
signatures that have unique fractal dimensions,
0 similar to the case of stone tool wear, a series of
1.08, aluminum, brass and stainless steel surfaces
produced by fly-cutting, grinding, shot blasting
1.06. and vapor polishing were examined. Range
O
images of 1 mm square areas on the surfaces
Dis~ce(km) were obtained using a scanned probe
Figure 5. Anisotropy of the FT power spectrum (ratio microscope (Talyscan).
of slopes in the cutting and transverse directions) for
the surfaces and materials in Figure 4.
This instrument uses a conventional diamond
stylus mounted on a cantilever arm whose
quite different for the two materials,-which vertical deflection is measured optically. The
machine in markedly different ways. low mass of the moving stylus allows rapid
A correlation plot (Figure 6) of the scanning (under 60 see for the full area) with a
conventional roughness vs. the fractal low applied force (<100 mgf). The large lateral
dimension shows that in general surfaces with and vertical range (200 ~ n ) permits useful
less relief tend to have higher fractal measurements to be obtained from engineering
dimensions, which is in qualitative agreement surfaces with a lateral sampling of 1 pan and
with the understanding that a surface with a vertical resolution of 3 nm.
dimension of 2.5 would correspond to the
random Brownian case with no lateral Figure 7 shows several representative images;
correlation, and might therefore be optimum as notice that some of the surfaces are isotropic
an engineering surface for some applications while others have either simple or complex
(electrical contacts, friction, etc.). Notice that anisotropy arising from the directionality of
this is a very different expectation than analysis tool motion. A total of 6 specimens of each
based on conventional Euclidean geometry material and process were produced, and
with ideally smooth surfaces would predict. multiple areas measured on each.
5"7[) J. ('. R u s s

a) fly cut aluminum c) vapor polished aluminum

b) ground stainless steel d) shot blasted stainless

Figure 7. Representative surface range images (1 mm square): a) fly cut aluminum; b) ground stainless steel;
c) vapor polished aluminum; d) shot blasted stainless steel.

Variance vs. area plots showed straight line Plotting the data for conventional surface
results (on log axes) that indicated fractal roughness (Sq, the area analog to Rq the rms
surface geometry. Although the appearances of roughness) and fractal dimension (Figure 8)
these surfaces varies widely, they are all capable shows no correlation of values, a substantial
of being generated using simple fractal scatter in the Sq data, and no grouping
generation routines, such as cross-extrusion of according to the material type.
profiles generated by fractal Brownian addition However, plotting the same mean values coded
or randomly generated Takagi pyramids (Russ, by the surface preparation method (Figure 9)
1974). indicates that the different machining
Fractal dimension measurement 571

2.45
2.45 Flycut []
Aluminum A
tA¢ + Ground +
T Bras s O
2.40 -
2.40 Shot Blasted X
4-
Stainless Steel o
.4- V a p o r Polished 0
rn
2.35
2.35
r~
1 lo D []
2.30
2.30 ,1
t-o,t 1 x

x
I,,- o ,,,,4 2.25
2.25
_ '~ ,.d _ ~ " e,i c,i
S q (laln)
sq (gin)
Figure 9. The same data as in Figure 7 but grouped
Figure 8. Scatterplot of conventional surface according to the surface preparation method rather than
roughness data (Sq) vs. fractal dimension (D) for a the material.
variety of machined surfaces produced in different
materials by fly-cutting, grinding, shot-blasting and
vapor-polishing.The scatter bands represent 2 standard
deviations, determined from multiple specimens. North Carolina State University. The data for
effect of tool wear on surface fmish were taken
by John Tyner as part of this thesis research at
techniques do produce consistent fractal North Carolina State University. The
dimension values, even though the conventional photographic images of stone tools were
roughness measures are not distinct. provided by J. Bueller from his thesis research
Work is continuing to determine the static and at the Hebrew University of Jerusalem. The
dynamic friction coefficients for these surfaces images of machined surfaces were provided by
and to analyze possible correlations with David Williamson of Taylor Hobson Ltd.
material hardness, conventional roughness and
surface fractal dimension. References

Acknowledgments J. C. Russ (1994) Fractal Surfaces, Plenum


Press, New York
The data on fracture surfaces were obtained by T. Vicsek (1992) Fractal Growth Phenomena,
Yusuf Fahmy as part of his thesis research at World Scientific Publ., Singapore

Vous aimerez peut-être aussi