Professional Documents
Culture Documents
Communication
Detailed surface roughness characterization of engineering surfaces
undergoing tribological testing leading to scuffing
Allison Y. Suh a , Andreas A. Polycarpou a,∗ , Thomas F. Conry b
a Department of Mechanical and Industrial Engineering, 1206 West Green Street, University of Illinois at Urbana, Champaign, Urbana, IL 61801, USA
b Department of General Engineering, 104 South Mathews Avenue, University of Illinois at Urbana, Champaign, Urbana, IL 61801, USA
Abstract
Controlled tribological scuffing experiments were conducted on typical engineering surfaces (Al390-T6 discs and 52100 steel pins)
under starved lubrication conditions (a mixture of R410A refrigerant combined with a polyolester (POE) lubricant) to simulate the contact
conditions in an automotive air conditioning compressor. The time-to-scuff was established and was repeatable using a test protocol with
increasing load increments. Test were then conducted for periods of 25, 50, and 75% of the time-to-scuff to investigate the progression of
surface change preparatory to scuffing. Detailed studies of the surface topography were conducted using line and areal (usually referred as
one dimensional (1D) and two dimensional (2D), respectively) analyses. Simple amplitude roughness parameters as well as more detailed
spatial, hybrid, and functional parameters were calculated and used to track detailed roughness changes as the Al390-T6 samples undergo
progressive wear until scuffing occurs. One-dimensional amplitude descriptors, such as the root-mean-square value, were not reliable in
tracking surface topographic changes. However, 2D functional parameters, such as the surface bearing index and the fluid retention index,
clearly showed progressive changes as the surfaces wear and reach scuffing.
© 2003 Published by Elsevier Science B.V.
Keywords: Surface roughness; Engineering rough surfaces; Al390-T6; Scuffing; Compressor surfaces
2. Background
Fig. 2. The SEM surface images of Al390-T6 disc sample showing primary wear track magnification: (a) 20×; and (b) 50×.
instruments, such as contact profilometers [12]. The ver- observed, designated as Ra and Rq in Fig. 3. When compar-
tical scales of optical profile systems are based on the ing a virgin part to mildly-worn part at (3/4)Tscuff as shown
reflective coefficients over the sample surface; depending in Fig. 3(a) and (b), the surface is smoothened out signif-
on the reflectivity of the region, the vertical and lateral res- icantly from the previous stages. Because the disc at this
olution may also suffer due to optical noise (i.e. artifacts) stage is near scuffing in which destroyed protective surface
originating from wave interference. Despite these potential films can no longer “protect” the surfaces from damage, the
drawbacks, in this work the optical profiler has proven to induced surface damage is expected to be much greater than
be a sufficiently accurate method when the roughness data just change in vertical distribution of the asperities. Other
are correlated with data previously obtained using a line topographic changes, however, cannot be inferred from vi-
contact profilometer [11]. The details on the sampling con- sual analyses based on simple roughness profiling.
ditions of the obtained optical data are given in Table 1. When the sample is scuffed, different regions on the
Refer to Fig. 3 for the coordinate orientation of the data. disc exhibit different topographical features, as expected.
Fig. 3 displays typical roughness images of the virgin, Fig. 3(c) shows one of the most severely scuffed regions,
(3/4)Tscuff , and Tscuff samples in both top and isometric referred asprimary wear scar, which was also shown in the
views. As seen in all the images, both the virgin and worn SEM image of Fig. 2(b). In this region, faint scuffing tracks
samples seem to be highly anisotropic with traces of the in the sliding direction (perpendicular to the machining
original texture from machining. Notice how the roughness marks) are visible. At Tscuff , the scuffing is presented as
image of the fully-scuffed region (Fig. 3(c)) displays mild material transfer in small, defined regions between the steel
wear tracks due to scuffing damage. For all the samples, the pin and aluminum disc sample along the “wear track,” in
sliding direction is in the direction perpendicular to the ma- which the surface is extremely smooth. In essence, the disc
chining marks, as indicated by arrows. sample has already failed at this stage, thus comprehensive
According to the roughness images obtained in this study, analysis at this stage will probably not provide any insight
all the virgin discs have a predominant directionality in their into the surface changes leading up to scuffing. However,
topography due to the original machining process. These Fig. 3(c) shows the importance of 2D (versus 1D profiling)
discs were prepared from the identical blanks and machined roughness analysis since measuring in different areas will
using the same methods as the actual discs in swash-plate give different results.
automotive air-conditioning compressors. Although the ma- From observation of the optical images of the discs at
chining textures on the discs after (1/4)Tscuff and (1/2)Tscuff various wear stages, the lack of gross visual topographical
have become less dominant, no significant visual change changes seems to reinforce the hypothesis that mild progres-
other than small decrease in the CLA and rms values are sive wear may accompany other geometric and functional
topographical changes that become apparent only upon de-
Table 1 tailed surface characterization.
Sampling conditions of the optical profiler
x-direction y-direction 3. Areal roughness characterization using the
Number of sampling points 368 240 Birmingham-14 parameters
Image aspect ratio 1 1.167
Image size/length (mm) 1.1964 0.9093 As a standard procedure prior to areal roughness analy-
Sampling interval (m) 3.260 3.804
sis, all the topographical images are filtered under exactly
560 A.Y. Suh et al. / Wear 255 (2003) 556–568
Fig. 3. 2D surface roughness images of Al390-T6 discs: (a) virgin; (b) (3/4)Tscuff ; and (c) scuffed.
the same filtering conditions. Specifically, low-pass filtering surface roughness parameters may be used as the indicator
with cutoff wavelength of 8.5 m is applied to remove the for quantifying the physical changes occurring at the surface.
high frequency noise. Due to the inherent method of opti- With the filtered roughness data, all parameters specified in
cal measurements, there is no need for high-pass filtering to the 2D Birmingham-14 [16] are used to describe the distin-
remove low frequency components (waviness). By keeping guishing aspects of the surface roughness of the Al390-T6
the same filtering conditions in all the topographical data, it discs undergoing mild progressive wear. Specifically, the to-
enables fair and consistent “roughness” comparison, avoid- pographic evolution during wear leading to scuffing can be
ing the issues and complexities associated with filtering and described through the characterization of amplitude, spatial,
its effect on roughness parameters [19]. and hybrid properties of the disc surfaces. More importantly,
In order to characterize the geometric properties of a sur- the evolution of functional properties, i.e. bearing, sealing,
face at various stages prior to scuffing, pre-established 2D lubrication retention capabilities at each tribological stage
A.Y. Suh et al. / Wear 255 (2003) 556–568 561
Table 2
Amplitude parameters as samples undergo tribological scuffing testing
Amplitude property (1/4)Tscuff (1/2)Tscuff (3/4)Tscuff Tscuff
height variation (Sz ) of the surface is plotted in Fig. 4(b). 3.2. Areal spatial parameters
Interestingly, the disc-to-disc variation in the virgin discs is
smaller in the Sz analysis compared to the amplitude pa- In addition to the amplitude parameters examined in the
rameter (Sq ). Within a sample, the mildly-worn Sz is al- previous section, spatial parameters are then utilized to char-
ways smaller than the virgin Sz as with CLA and rms. How- acterize roughness information in the planar directions. Spa-
ever, the amount of the drop between the virgin Sz and the tial parameters include the density of summits (Sds ), measure
mildly-worn Sz in each stage is nearly identical for all the of isotropy–anisotropy of the surface (Str ), and directionality
wear stages. Due to its insensitive disc-to-disc variation in of surface texture (Std ). The areal autocorrelation function
the virgin states, 10-point height variation can be a more ap- (AACF), areal power spectrum density (APSD), and angu-
propriate amplitude parameter for determining the surface lar spectrum of the APSD are used to obtain the anisotropy
wear evolution, resulting in more steady, progressive wear of the surfaces and direction of the lay [16,23,24], also sum-
behavior. marized as Eqs. (A.1), (A.3) and (A.4), respectively, in the
Referring to Table 2 and Fig. 4(c), the skewness value Appendix A. Eqs. (A.2) and (A.5) in the Appendix A sum-
(Ssk ) at each tribological stage behaves relatively well. Not marize the mathematical definitions for calculating Str and
only is disc-to-disc variation in the initial virgin states neg- Std , respectively, in continuous integral forms. Table 3 sum-
ligible, but the skewness value also progressively decrease marizes the areal spatial parameters calculated in this study
from −0.25 at the virgin stages to −1.7 at (3/4)Tscuff . The for the samples undergoing tribological testing.
skewness decreases between virgin and mildly-worn states Using the “eight nearest neighbor summit–valley” defi-
for all the samples. The highly negative skewness values in nition, the density of summits (Sds ) is calculated for all the
the mildly-worn states support the theory that mildly-worn disc samples. As shown in Table 3, despite minor variation
asperities result in the surface being more negatively from disc-to-disc in their initial virgin states, it is clear
skewed by revealing more pits and valleys than peaks. The that the density of summits in the mildly-worn states is
most change is observed when moving from (1/2)Tscuff to always higher than the virgin states, albeit not by much.
(3/4)Tscuff . At this worn stage just before full scuffing, con- Most change occurs at (3/4)Tscuff , which explains that as
dition of the surface seems to be characterized by losing the the surface smoothens out, it reveals more minor asperities
spikiness of the top asperities by significant amounts. Not in the lower zones. The increase in the number of summits
only does the skewness show a clear trend with progressive in the “valley” in turn causes the total density of summits.
tribological testing but the lack of sample-to-sample vari- Other than a general increase from virgin to mildly-worn
ability in the virgin state seems to indicate a potential for states, no conclusive trend can be seen from the Sds
its use as precursor to scuffing. analysis.
The kurtosis (Sku ) values for all the virgin states are near A useful spatial parameter is the texture aspect ratio (Str ),
Gaussian, varying between 3 and 4. After the progressive sometimes referred to as the isotropy index, which describes
tribological testing, the kurtosis values seem to increase up the degree of isotropy of the surface roughness. The texture
to around 5 at all stages except (1/2)Tscuff . Unlike in the aspect ratio is based on the calculation of the areal autocor-
skewness analysis, no conclusive trend can be deduced from relation function (AACF) over the surface of interest. For
the kurtosis analysis other than its generally increasing ten- strongly anisotropic surfaces exhibiting a lay direction, the
dency towards being centrally distributed (Sku > 3). Str for a surface is typically <0.3. Indeed, this seems to be the
In the amplitude roughness analysis of Al390-T6 discs, case in Table 3 for all of the virgin and mildly-worn disc sam-
the CLA and rms parameters that are used extensively to ples, mainly due to the highly directional originally present
quantify surface roughness and changes do not show any machining marks. In a traditional wear test, the initial ma-
systematic trends that can be used as precursors to scuffing. chining marks are obliterated significantly, thereby chang-
On the other hand, the skewness seems to present clearly ing the surface from highly anisotropic to isotropic. In this
a well-behaved progressive trend with tribological testing, case, however, Str does not change much, indicating that the
suggesting its use as a precursor to scuffing. “wear” does not significantly affect the original machining
Table 3
Spatial parameters as samples undergo tribological scuffing testing
Spatial property (1/4)Tscuff (1/2)Tscuff (3/4)Tscuff Tscuff
Density of summits (mm−2 ) Sds V 5.089 × 103 4.777 × 103 4.690 × 103 5.587 × 103
W 5.657 × 103 4.969 × 103 5.769 × 103 5.157 × 103
Texture aspect ratio (0–1) Str V 0.28 0.094 0.28 0.036
W 0.12 0.041 0.28 0.070
Texture direction (◦ ) Std V −88.32 −89.16 88.32 −88.05
W −85.01 83.15 89.37 −89.27
A.Y. Suh et al. / Wear 255 (2003) 556–568 563
marks of the top surface, which was highly anisotropic be- ing on the relative influence of either the amplitude or
fore the tribological scuffing testing. the spacing or both, hybrid parameters will represent the
In order to determine the texture direction of the surface, combined aspects of roughness. Table 4 summarizes the hy-
first the areal power spectrum density (APSD) analysis, im- brid parameters, extracted from the virgin and mildly-worn
mediately followed by the angular spectrum analysis, must samples.
be performed [16]. The lay directions are found to be the an- The hybrid slope S q is the scalar magnitude of the
gle at which the maximum peak occurs in the angular spec- continuous surface slope at any point. Between virgin and
trum matrix, as illustrated in Fig. 5. Due to the pronounced mildly-worn states of each tribological testing stage, the
directionality in the Al390-T6 discs, each disc, whether vir- S q value of the mildly-worn state is always smaller as
gin or worn, has a distinctive maximum peak in its respec- summarized in Table 4. The reason for this reduction may
tive angular spectrum plot (i.e. anisotropic). be that the top asperities are being polished with tribological
From the spatial parameter characterization of the testing. Despite the large virgin sample variability, the dif-
Al390-T6, it is found that the Sds does not yield a signifi- ference between worn S q and virgin S q seems to increase
cant trend with tribological scuffing testing. Similarly, not for more severely worn stages, especially at (3/4)Tscuff .
much change is observed in Str and Std either, though they This indicates that the S q decreases steadily as the surface
are unique 2D spatial parameters not found in 1D analysis. approaches scuffing. This result seems to agree with the
From a practical point of view, knowledge of the degree of previously regarded observation on the reduction of asper-
anisotropy and its direction are very important, especially ity heights, in which the slope of a summit is expected to
in cases where there is no severe wear to completely alter reduce when the top asperity is worn off. However, because
the surface roughness and lay direction. the sample-to-sample variability is relatively large in the
S q , wear evolution through different stages of tribological
3.3. Areal hybrid parameters testing is difficult to substantiate.
Table 5
Functional parameters as samples undergo tribological scuffing testing
Functional propertya (1/4)Tscuff (1/2)Tscuff (3/4)Tscuff Tscuff Gaussian reference
was performed by Mr. J.J. Patel in the Tribology Labora- A.2. Hybrid parameters
tory of the University of Illinois. The surface roughness
measurements were performed at Professor C. Liu’s Micro • Mean summit curvature, Ssc
Actuators, Sensors, and Systems Group at the University
t
of Illinois. The scanning electron microscope images were 11 ∂z2 (x, y) ∂z2 (x, y)
Ssc = − +
taken at the Center for Microanalysis of Materials at the 2t ∂x2 ∂y2 for any summit
k=1
University of Illinois, which is supported by the US De-
(A.6a)
partment of Energy under Grant DEFG02-96-ER45439.
where t is number of summits and
1
Appendix A. Formulae for calculating the 2D R= (A.6b)
Ssc
roughness parameters
• Developed area ratio, Sdr
A.1. Spatial parameters
total interfacial area − sampling area
Sdr (%) = × 100
sampling area
• Areal autocorrelation function, r(τ x , τ y )
(A.7)
r(τx , τy ) = E[z(x, y) z(x + τx , y + τy )]
Ly Lx or
1
= lim z(x, y) total interfacial area − (M − 1)(N − 1) x y
Lx ,Ly →∞ 4Lx Ly −Ly −Lx Sdr (%) =
(M − 1)(N − 1) x y
× z(x + τx , y + τy ) dx dy (A.1) × 100
• Texture aspect ratio, Str See [16,22] for calculating the “total interfacial area”.
• Root-mean-square slope, S q
min τi 2 + τ j 2
0 < Str =
≤ 1, Lx Ly
1 ∂z(x, y) 2
max τi 2 + τj 2 S q=
Lx Ly 0 0 ∂x
r(τi , τj ) 1/2
where ≤ 0.2 (A.2) ∂z(x, y) 2
r(0, 0) + ∂x ∂y (A.8)
∂y
where i = 0, 1, ..., m < M; j = 0, 1, ..., n < N; τi =
i x; τj = j y
• Discrete Fourier transform, F(ωx , ωy ) A.3. Functional parameters
∞ ∞
F(ωx , ωy ) = z(x, y) • Surface bearing index, Sbi
−∞ −∞
× exp[−j2π(ωx x + ωy y)] dx dy (A.3) Sq 1
Sbi = = (A.9)
(z)Stp =5% h0.05
• Areal power spectrum density, G(ωx , ωy ), and angular
spectrum Ga (θ) • Core fluid retention index, Sci
G(ωx , ωy ) = F(ωx , ωy ) F ∗ (ωx , ωy ) (A.4a) ((Vv (h0.05 )−Vv (h0.8 ))/((M−1)(N−1) x y))
Sci =
Sq
rr(θ) (A.10)
Ga (θ) = G(θ, r) dr
0 • Valley fluid retention index, Svi
2 −1/2
rr(θ) = 1
2 [( x cosθ) + ( y sinθ) ]
2
(Vv (h0.8 ))/((M − 1)(N − 1) x y)
◦ Svi = (A.11)
where 0 ≤ θ ≤ 179 . (A.4b) Sq
• Texture direction, Std
π
−θ, θ≤ References
2
Std = (A.5)
π − θ, π
≤θ≤π [1] T.K. Sheiretov, Scuffing of Aluminum–Steel Contact Under Dry
2 Sliding Conditions, Ph.D. Thesis, University of Illinois at Urbana,
Champaign, 1997.
where θ is the position where the maximum value of the [2] J.F. Archard, Contact of rubbing surfaces and flat surfaces, J. Appl.
angular spectrum or Ga (θ) appears. Phys. 24 (1953) 981–988.
568 A.Y. Suh et al. / Wear 255 (2003) 556–568
[3] J.T. Burwell, Survey of possible wear mechanisms, Wear 1 (1957) [16] K.J. Stout, P.J. Sullivan, W.P. Dong, E. Mainsah, N. Luo, T. Mathia,
119–141. H. Zahouani, The Development of Methods for the Characterization
[4] R.D. Arnell (Ed.), Tribology, Principles and Design Applications, of Roughness in Three Dimensions, Commission of the European
Springer–Verlag, New York, 1991. Communities, Brussels, Luxembourg, 1993.
[5] M.P. Cavatorta, Running-In and Scuffing in Aluminum–Steel Sliding [17] H. Yoon, T. Sheiretov, C. Cusano, Tribological evaluation of some
Contacts, Ph.D. Thesis, Politecnico di Torino, 1998. aluminum-based materials in lubricant–refrigerant mixtures, Wear
[6] H. Yoon, T. Sheiretov, C. Cusano, Scuffing behavior of 390 aluminum 218 (1998) 54–65.
against steel under starved lubrication conditions, Wear 237 (2000) [18] T.Y. Lin, L. Blunt, K.J. Stout, Determination of proper frequency
163–175. bandwidth for 3D topography measurement using spectral analysis.
[7] M.P. Cavatorta, C. Cusano, Running-in of aluminum–steel contacts Part I. Isotropic surfaces, Wear 166 (1993) 221–232.
under starved lubrication. Part I. Surface modifications, Wear 242 [19] C.Y. Poon, B. Bhushan, Comparison of surface roughness
(2000) 123–132. measurements by stylus profiler, AFM and non-contact optical
[8] M.P. Cavatorta, C. Cusano, Running-in of aluminum–steel contacts profiler, Wear 190 (1995) 76–88.
under starved lubrication. Part II. Effects on scuffing, Wear 242 [20] T.R. Thomas, Characterization of surface roughness, Prec. Eng. 3 (2)
(2000) 133–139. (1981) 97–104.
[9] K.C. Ludema, A review of scuffing and running-in of lubricated [21] W.P. Dong, P.J. Sullivan, K.J. Stout, Comprehensive study of
surfaces, with asperities and oxides in perspective, Wear 100 (1984) parameters for characterizing three-dimensional surface topography.
315–331. Part III. Parameters for characterizing amplitude and some functional
[10] H.K. Yoon, Scuffing Under Starved Lubrication Conditions, Ph.D. properties, Wear 178 (1994) 29–43.
Thesis, University of Illinois at Urbana, Champaign, 1999. [22] W.P. Dong, P.J. Sullivan, K.J. Stout, Comprehensive study of
[11] J.J. Patel, Investigation of the Scuffing Mechanism Under Starved parameters for characterizing three-dimensional surface topography.
Lubrication Conditions Using Macro, Meso, Micro, and Nano Part IV. Parameters for characterizing spatial and hybrid properties,
Analytical Techniques, M.S. Thesis, University of Illinois, IL, 2001. Wear 178 (1994) 45–60.
[12] ASME B46.1-1995, Surface Texture (Surface Roughness, Waviness, [23] H.C. Xie, D.R. Chen, X.M. Kong, An analysis of the
and Lay): An American National Standard, New York, 1996. three-dimensional surface topography of textured cold-rolled steel
[13] ISO 4287:1997(E/F), Geometrical Product Specifications (GPS)— sheets, Tribol. Int. 32 (1999) 83–87.
Surface Texture: Profile Method—Terms, Definitions, and [24] I. Sherrington, G.W. Howarth, Approximate numerical models of
Surface Texture Parameters, The International Organization for 3D surface topography generated using sparse frequency domain
Standardization, 1997. descriptions, Int. J. Mach. Tools Manufact. 38 (5–6) (1998) 599–606.
[14] DIN 4776, Measurement of Surface Roughness Parameters Rk , Rpk , [25] J.I. McCool, Comparison of models for the contact of rough surfaces,
Rvk , Mr1 , Mr2 for the Description of the Material Portion in the Wear 107 (1986) 37–60.
Roughness, German Standard, 1990. [26] E.J. Abbott, F.A. Firestone, Specifying surface quality: a method
[15] BSI 1134, Assessment of Surface Texture. Part 1. Methods and based on accurate measurement and comparison, Mech. Eng. 55
Instrumentation, British Standard, 1988. (1933) 569–572.