Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
冉 冊 冉 冊
vergence is further improved for cases under severe conditions.
Recently, Hu and Zhu 关16兴 and Zhu and Hu 关17兴 developed a x P y P 共¯H兲
+ = 共1兲
X X Y Y X
1
Corresponding author. MCC Mechanical Engineering, 2145 Sheridan Rd, Room where
冉 冊冉 冊
B224, Evanston, IL 60208.
Contributed by the Tribology Division of ASME for publication in the JOURNAL OF aph ¯H3 x
TRIBOLOGY. Manuscript received June 2, 2005; final manuscript received March 2, x = , y =
2006. Review conduct by Liming Chang. 12u0 ¯ K2e
Journal of Tribology Copyright © 2006 by ASME JULY 2006, Vol. 128 / 641
X
␣¯i,j 0 0 0
X
¯i,j 0 0 0
X
␥¯i,j 0 0 0
X ¯i,j − ¯i−1,j ¯i+1,j − ¯i−1,j
␦¯i,j 3¯i,j − 4¯i−1,j + ¯i−2,j
Hi.j Hi.j Hi.j
⌬X 2⌬X 2⌬X
⌷共⌬x2兲 = −
共⌬x兲2 3u
6 x3
冏 冏 i,j
+ ⌷关共⌬x兲4兴
Fig. 1 Profiles of dimensionless pressure, film thickness, and In the second-order backward scheme,
冏 冏
first-order film thickness derivative along the centerline in a
Hertzian contact u 3ui,j − 4ui−1,j + ui−2,j
= + ⌷关共⌬x兲2兴
x i,j 2⌬x
共9兲
The film thickness expression, including normal approach, ge-
ometry clearance, roughness, and elastic deformation, is as fol-
lows:
⌷共⌬x 兲 =
共⌬x兲2 3u
2
3 x3
冏 冏 i,j
+ ⌷关共⌬x兲 兴 3
H = H0 + BxX2 + ByY 2 + S共X,Y兲 + V共X,Y兲 共2兲 Comparison of the truncation errors of the three schemes draws
two conclusions. First, for a given mesh size, results obtained by
where S共X , Y兲 denotes roughness, and elastic deformation V共X , Y兲 applying the second-order central scheme most closely approach
is from the Boussinesq integration the exact solution. Solutions from the first-order backward scheme
may have the largest discretization errors. The solution from the
P共¯,¯兲
V共X,Y兲 = 共CE兲 /冑 ⍀ 共X − ¯兲2 + K2e 共Y − ¯兲2
d¯d¯ 共3兲 second-order backward scheme is between the two. Second, as the
mesh spacing decreases, the solution deviations reduce. However,
the superiority of the higher-order formulas relies on the smooth-
The Barus pressure-viscosity relationship is used for viscosity ness of the exact solution of the problem modeled. When speed is
variation in the present study low or load is high, the film thickness gradient of an EHL solution
¯ = exp共␣ P兲 共4兲 becomes sharp at the inlet area. The limit case is a Hertzian con-
tact as shown in Fig. 1. The inlet film thickness gradient under-
The Dowson-Higginson pressure-density relationship 关22兴 is goes a turning point, where the higher-order film thickness deriva-
used for density variation tives do not existed. When a sharp gradient occurs, the truncation
Ca P error of a higher-order discretization is much larger than that of a
¯ = 1 + 共5兲 low-order discretization. Consequently, on a given grid, the trun-
1 + Cb P cation errors of higher-order terms do not diminish at such a rapid
Load balance is analyzed through the following pressure inte- rate as when the underlying exact solution is smooth. Unless the
gration grid is made extremely fine, a higher order discretization cannot
show apparent superiority over a low-order discretization 关23兴. In
2
/ ⍀
PdXdY =
3
共6兲 later sections, one can find that in the thin film regime, the first-
order backward schemes seem to yield more accurate solutions
than the other second-order schemes.
3 Differential Schemes
All differential schemes are derived from the Taylor series ex-
pansion. The first-order backward, second-order central, and
4 Discrete Equation
second-order backward differentiations are often used for discreti- By using the differential schemes mentioned above, the Rey-
zating the Reynolds equation. Because downstream schemes may nolds equation in Eq. 共1兲 can be converted into a discrete differ-
encounter instability caused by convection, forward schemes are ential equation at each unknown pressure point
not preferred here. The truncation errors for the commonly used ␣i,j Pi−1,j + i,j Pi,j + ␥i,j Pi+1,j = ␦i,j 共10兲
three schemes are
in the first-order backward scheme, where ␣i,j, i,j, ␥i,j, and ␦i,j are all known numerical coefficients,
冏 冏 冏 冏
which will be further described later. The Eq. 共10兲 system is
u ui,j − ui−1,j ⌬x 2u solved by Jacobi line relaxation described in 关24兴.
= + ⌷共⌬x兲 ⌷共⌬x兲 = + ⌷关共⌬x兲2兴
x i,j ⌬x 2 x2 i,j
4.1 Second-Order Central Scheme for Poiseuille Flows.
共7兲 The Poiseuille flows are usually discretized by the Second-order
In the second-order central scheme central differential scheme
冏 冏
u
x i,j
=
ui+1,j − ui−1,j
2⌬x
+ ⌷共⌬x2兲 冋 冉 冊册
x P
X
X i.j
=
1
关x
⌬X2 i+1/2,j
Pi+1,j − 共i+1/2,j
x
+ i−1/2,j
x
兲Pi,j
共8兲 + i−1/2,j
x
Pi−1,j兴 + ⌷共⌬X2兲 共11兲
␣i,j
HX i,j
Di−1,j i−1,j
− Di−1,j i+1,j
Di−1,j i−1,j
− Di−1,j i−1,j
−2.0Di−1,j i−2,j
+ 0.5Di−1,j
−¯i,j −¯i,j −¯i,j
⌬X 2⌬X ⌬X
i,j
HX
−␣i,j
HX
0 i,j
1.5Di,j i−2,j
+ 0.5Di,j
−¯i,j
⌬X
␥i,j
HX i,j
Di+1,j i−1,j
− Di+1,j −␣i,j
HX i,j
1.5Di+1,j i−2,j
+ 0.5Di+1,j
−¯i,j −¯i,j
⌬X ⌬X
关 兴 关 兴 关 兴
␦i,j
HX
关Hi,j − 兴共Di−1,j
i,j
Pi−1,j兲 关Hi+1,j − 兴共Di−1,j
i+1,j
Pi−1,j兲 1.5关Hi,j − 共Di,j
i,j i,j
Pi,j + Di+1,j Pi+1,j兲兴
¯i,j +Di,j Pi,j + 关共Di+1,j Pi+1,j兲兴 ¯i,j +Di,j Pi,j关共 + Di+1,j Pi+1,j兲兴 ¯i,j −2.0关Hi−1,j − 共Di−1,j Pi−1,j兲兴
i,j i,j i+1,j i+1,j i−1,j
Hcen Hmin
共nm兲 共nm兲
冋 冉 冊册
y P
Y
Y i.j
=
1
关y
⌬Y 2 i,j+1/2
Pi,j+1 − 共i,j+1/2
y
+ i,j−1/2
y
兲Pi,j
+ i,j−1/2
y
Pi,j−1兴 + ⌷共⌬Y 2兲 共12兲
where Fig. 2 Present results obtained from the working conditions in
†21‡
1 x 1 x
i+1/2,j
x
= 共i,j + i+1,j
x
兲, i−1/2,j
x
= 共i−1,j + i,j
x
兲
2 2
i,j+1/2
y 1 y
= 共i,j
2
+ i,j+1
y
兲, i,j−1/2
y 1 y
= 共i,j−1
2
+ i,j
y
兲
冋 册
共¯H兲
X i.j
=
¯i,jHi,j − ¯i−1,jHi−1,j
⌬X
+ ⌷共⌬X兲 共15兲
冦 冧
␣i,j
p
=
i−1/2,j
x 2
Hi,j = H0 + BxXi,j 2
+ ByY i,j + 兺兺D i,j
k,l Pk,l 共16兲
⌬X2 k l
冦 冧
Pi,j−1
␦i,j
p
=− ¯i,jDi−1,j
i,j
− ¯i−1,jDi−1,j
i−1,j
⌬Y 2
␣i,j
W
=−
⌬X
4.2 First-Order Backward Scheme for Couette Flow. The ¯i,jDi,j
i,j
− ¯i−1,jDi,j
i−1,j
Couette flow has two forms, the combined form and the separated i,j
W
=−
form, corresponding to the left and the right sides of Eq. 共14兲 ⌬X
respectively. The separated form includes two terms, a density ¯i,jDi+1,j
i,j
− ¯i−1,jDi+1,j
i−1,j
冤 冥
共¯H兲
¯ H 1
= H + ¯ 共14兲 ␦i,j
W
= ¯i,j关Hi,j − 共Di−1,j
i,j i,j
Pi−1,j + Di,j i,j
Pi,j + Di+1,j Pi+1,j兲兴
X X X ⌬X
If the first-order backward scheme is used for the combined − ¯i−1,j关Hi−1,j − 共Di−1,j
i−1,j i−1,j
Pi−1,j + Di,j i−1,j
Pi,j + Di+1,j Pi+1,j兲兴
form, the expression for the Couette term becomes 共18兲
冦 冧
␣i,j = ␣i,j
P
+ ␣i,j
W scheme is used for the combined Couette flow, the expression of
the discrete term and the contributions to the coefficients in Eq.
i,j = i,j
P
+ i,j
W
共10兲 can be given, respectively, as follows:
冋 册
共19兲
␥i,j = ␥i,j
P
+ ␥i,j
W
共¯H兲 ¯i+1,jHi+1,j − ¯i−1,jHi−1,j
␦i,j = ␦i,j
P
+ ␦i,j
W = + ⌷共⌬X2兲 共20兲
X i.j 2⌬X
冦 冧
¯i+1,jDi−1,j
i+1,j
− ¯i−1,jDi−1,j
i−1,j
␣i,j
W
=−
2⌬X
¯i+1,jDi,j
i+1,j
− ¯i−1,jDi,j
i−1,j
i,j
W
=−
2⌬X
共21兲
¯i+1,jDi+1,j
i+1,j
− ¯i−1,jDi+1,j
i−1,j
␥i,j
W
=−
2⌬X
␦i,j
w
= 冋
1 ¯i+1,j关Hi+1,j − 共Di−1,j Pi−1,j + Di,j Pi,j + Di+1,j Pi+1,j兲兴
i+1,j
i−1,j
Pi−1,j + Di,j
i+1,j
i−1,j
Pi,j + Di+1,j Pi+1,j兲兴
册
As described in 关24兴, this scheme will be unstable for heavy loads. In this paper, oscillating results may be observed for the cases
studied using this scheme.
4.4 Second-Order Backward Scheme for Couette Flow. Another high-order differential scheme is the second-order backward
scheme. However, if one processes the wedge term contributions to the discrete Eq. 共10兲 using this scheme in the same way as shown
in Eqs. 共22兲 and 共23兲, divergence is encountered regardless of how the relaxation factor is chosen
冋 册
共¯H兲
X i.j
=
3¯i,jHi,j − 4¯i−1,jHi−1,j + ¯i−2,jHi−2,j
2⌬X
+ ⌷共⌬X2兲 共22兲
冦 冧
3¯i,jDi−1,j
i,j
− 4¯i−1,jDi−1,j
i−1,j
+ ¯i−2,jDi−1,j
i−2,j
␣i,j
W
=−
2⌬X
3¯i,jDi,j
i,j
− 4¯i−1,jDi,j
i−1,j
+ ¯i−2,jDi,j
i−2,j
i,j
W
=−
2⌬X
3¯i,jDi+1,j
i,j
− 4¯i−1,jDi+1,j
i−1,j
+ ¯i−2,jDi+1,j
i−2,j
共23兲
␥i,j
W
=−
2⌬X
冤 冥
3¯i,j共Hi,j − Di−1,j
i,j i,j
Pi−1,j − Di,j i,j
Pi,j − Di+1,j Pi+1,j兲
1
␦i,j
W
= − 4¯i−1,j共Hi−1,j − Di−1,j Pi−1,j − Di,j Pi,j − Di+1,j
i−1,j i−1,j i−1,j
Pi+1,j兲
2⌬X
+ ¯i−2,j共Hi−2,j − Di−1,j Pi−1,j − Di,j Pi,j − Di+1,j Pi+1,j兲
i−2,j i−2,j i−2,j
Venner and Lubrecht 关24兴 believe that the instability is due to the accumulation of changes in the summation in film thickness. They
proposed the Jacobi distributive line relaxation instead. However, the present authors have realized that if the contributions in Eq. 共23兲
are adjusted as follows, good convergence may be observed for the cases studied:
冦 冧
− 2.0¯i−1,jDi−1,j
i−1,j
+ 0.5¯i−2,jDi−1,j
i−2,j
␣i,j
W
=−
⌬X
1.5i,jDi,j + 0.5¯i−2,jDi,j
¯ i,j i−2,j
i,j
w
=−
⌬X
1.5¯i,jDi+1,j
i,j
+ 0.5¯i−2,jDi+1,j
i−2,j
共24兲
␥i,j
w
=−
⌬X
冤 冥
1.5¯i,j共Hi,j − Di,j
i,j i,j
Pi,j − Di+1,j Pi+1,j兲
1
␦i,j =
W
− 2.0¯i−1,j共Hi−1,j − Di−1,j Pi−1,j兲
i−1,j
⌬X
+ 0.5¯i−2,j共Hi−2,j − Di−1,j
i−2,j i−2,j
Pi−1,j − Di,j i−2,j
Pi,j − Di+1,j Pi+1,j兲
4.5 Mixed Scheme for Couette Flow. In Secs. 4.2–4.4, the under consideration.
first-order backward, second-order central and second-order back- First, applying the first-order backward, second-order central,
ward schemes are applied to the combined form of the Couette and second-order backward schemes to the density derivative term
flow. If these differential schemes are applied to the separated yields
form, shown in Eq. 共14兲, one can get various mixed discrete
schemes. One example is the first-order backward for the density
冋 册
derivative term and the second-order backward for the film thick-
ness derivative term. Actually, there could be a rather large num-
¯ ¯i,j − ¯i−1,j
H = Hi.j + ⌷共⌬X兲 共25兲
ber of possible scheme combinations when the mixed schemes are X i.j ⌬X
冋 册 H
¯
X i.j
= Hi.j
¯i+1,j − ¯i−1,j
2⌬X
+ ⌷共⌬X2兲 共26兲
listed in Table 1.
Second, the film thickness derivative term can be processed
similarly, if the first-order backward, second-order central, and
冋 册
H
¯
X i.j
= Hi.j
3¯i,j − 4¯i−1,j + ¯i−2,j
2⌬X
+ ⌷共⌬X2兲 共27兲
second-order backward schemes are used for this term
冋 册
¯
H
= ¯i.j
Hi,j − Hi−1,j
+ ⌷共⌬X兲 共28兲
The coefficient contributions to Eq. 共10兲 from Eqs. 共25兲–共27兲 are X i.j ⌬X
w 800 N
E⬘ 2.1978⫻ 1011 Pa
Rx 1.905⫻ 10−2 m
␣ 1.82⫻ 10−8 1 / Pa
0 0.096 Pa s
a 4.725⫻ 10−4 m
ph 1.711⫻ 109 Pa
冋 册 ¯
H
X i.j
= ¯i.j
Hi+1,j − Hi−1,j
2⌬X
+ ⌷共⌬X2兲 共29兲
冋 册¯
H
X i.j
= ¯i.j
3Hi,j − 4Hi−1,j + Hi−2,j
2⌬X
+ ⌷共⌬X2兲 共30兲
冦 冧
X
␣i,j = ␣i,j
P
+ ␣¯i,j + ␣i,j
HX
X
i,j = i,j
P
+ ¯i,j + i,j
HX
X
共31兲
␥i,j = ␥i,j
P
+ ␥¯i,j + ␥i,j
HX
X
␦i,j = ␦i,j
P
+ ␦¯i,j + ␦i,j
HX
Table 5 The corresponding applied schemes for seven differential schemes. 1B, 2C, and 2B
refer to the first-order backward scheme, second-order central scheme, and second-order
backward scheme, respectively.
共¯H兲 1B 2C 2B
X
¯ 1B 1B 1B 2B
H
X
H 1B 2C 2B 2B
¯
X
Fig. 6 Effect of differential scheme on the shapes of film thick- area. On the contrary, pressure is very high in the central area,
ness and pressure: film thickness and pressure profiles along which causes very high viscosity and density and large elastic
the centerline „a… in X direction and „b… in Y direction deformation. These features result in such small coefficients for
the Poiseuille flows, x and y, that Poiseuille flows can be ne-
glected. In this case, the Reynolds equation reduces to
⫻ 10−6, is selected as the criterion for contact determination. The
共¯H兲
use of a smaller would negligibly affect the calculated results. If =0 共34兲
H is larger than , the full Reynolds equation in Eq. 共1兲 is used to X
obtain the hydrodynamic pressure. Otherwise, the contact pressure Equation 共34兲 has a solution in the form
is obtained from the steady-state reduced Reynolds equation
¯H = const 共35兲
H
=0 共32兲 Because the change in the density is small, the film thickness
X along the X direction in the central area should be almost constant.
In the present algorithm, when the film thickness is smaller than , Here, the above statement is checked for validation since the
it is directly set to . In this case, the solution to Eq. 共32兲 is present method solves pressures in both areas from the same full
equivalent to H expressed in the following contact equation: Reynolds equation.
A numerical example for smooth surface EHL that supports the
H= 共33兲 above statement is illustrated in Fig. 3. The working conditions
The results obtained by using Eq. 共32兲 were compared to those are listed in Table 4. Contributions to the solutions are shown in
obtained from Eq. 共33兲. In both cases, Eq. 共1兲 was used for lubri- Figs. 3共b兲–3共d兲. All plots in these figures show the results along
cation pressure. It was found that there is no obvious difference the centerline. The vertical coordinates, A1, A2, A3, and b corre-
between these two treatments. However, the advantage of replac- spond to the coefficients ␣, , ␥, and ␦ in Eq. 共10兲. The dotted and
ing Eq. 共33兲 by Eq. 共32兲 is that both the reduced Reynolds equa- dashed lines are for the contributions from the Poiseuille flows
tion in Eq. 共32兲 and the full Reynolds equation in Eq. 共1兲 are all in and the Couette flow, while the solid line is for the summation of
the same equation system; therefore, Eq. 共32兲 can be easily pro- the contributions from both sides. The figures show the absolute
gramed. values of all contributions. Three regions can be identified. In the
The reduced Reynolds equation can also be used to explain the area far from the inlet border of the contact zone, the solid line
EHL behavior. Since pressure in the inlet area is low, viscosity strictly follows the dotted line, indicating that the Poiseuille flows
and density here are lower than those in the center zone, and the are dominant. In the area around the inlet, the solid line first
elastic deformation is also small. It is well known that film thick- follows the dotted line and then follows the dashed line. This
ness is formed mainly by the entraining action in this magic inlet means that both the Poiseuille flows and the Couette flow are
important. In the central area, the solid line follows the dashed
line, which means that the Couette flow is dominant. In general, corresponding to mesh sizes of ⌬X = ⌬Y = 0.02344, 0.01172,
Poiseuille flow and Couette flow have different weights of influ- 0.00586, and 0.00293, respectively. Third, solutions with seven
ence in different areas. differential schemes for the Couette term are compared. Details of
However, the above statement may only be valid for smooth theses schemes are listed in Table 5. For all the cases with contact,
surface contacts when the film is ultrathin. The results may be Hmin are identical, equal to the film thickness lower limit of
different when surface roughness is involved. A sample case with 0.47 nm, which corresponds to the dimensionless value of
a stationary transverse sinusoidal surface is shown in Fig. 4. The = 1.0⫻ 10−6.
working conditions are the same as those in 关25兴. The solution Figures 5 and 6 focus on comparisons of the film thickness and
indicates that, because of the effect of roughness, the solid line pressure profiles along the centerlines in the X and Y directions.
does not always follow the dashed line in the central high-pressure Figure 5 presents the effect of decreasing entrainment speeds. All
area, revealing that both Poiseuille flows and the Couette flow are results were obtained from scheme 1, the separated first-order
important, and completely using the reduced Reynolds equation in backward approach, on the finest mesh, 1024⫻ 1024. The results
Eq. 共34兲 in the central area for rough surfaces may result in some for U = 1250 mm/ s are not included in order to emphasize the
numerical error. This observation indicates that for rough-surface thin-film performance. It is clear that, as speed decreases, EHL
EHL cases, the reduced Reynolds equation should only be used pressure gradually approaches the Hertzian pressure. On the other
when condition h 艋 is satisfied in a point-by-point basis, as hand, the film thickness keeps its horseshoe shape and becomes
stated in 关16,17兴. flatter at lower speeds.
Figure 6 shows the effect of the differential schemes by using
an ultrathin-film case of U = 1 mm/ s on the finest mesh of 1024
7 Results
⫻ 1024. It is found that the first-order backward schemes, both the
Numerical analyses are conducted to investigate the influences separated 共scheme 1兲 and the combined 共scheme 5兲 yield the
of speed, differential scheme, and mesh size. The operating con- thickest film; the films from the second-order backward schemes
ditions and parameters are the same as those in Table 4. The 共scheme 3, 4, 7兲 are the thinnest; while those from the second-
solution domain is −1.9⬍ X ⬍ 1.1 and −1.5⬍ Y ⬍ 1.5. First, results order central schemes 共scheme 2 and 6兲 are in between. The os-
for seven entrainment speeds are selected to compare: cillating film is from scheme 6, the combined second-order central
1250 mm/ s, 312.5 mm/ s, 100 mm/ s, 30 mm/ s, 10 mm/ s, scheme.
3 mm/ s, and 1 mm/ s. These results correspond to the central film The following comparisons focus on the central film thickness
thicknesses from relatively thick 共⬃530 nm兲 to thin 共3 – 4 nm兲. and minimum film thickness. Figures 7 and 8 show the results for
Second, four computational meshes are chosen for the same solu- the high-speed cases of 1250 mm/ s and 312.5 mm/ s. The analy-
tion domain: 128⫻ 128, 256⫻ 256, 512⫻ 512, and 1024⫻ 1024, sis of both Hcen 共Figs. 7共a兲 and 8共a兲兲 and Hmin 共Figs. 7共b兲 and
8共b兲兲 indicate that as the mesh number increases, all solutions schemes, including schemes 1 and 5, get their maximum on mesh
from different schemes approach a constant value. The compari- 256⫻ 256, then slightly decrease toward the final solution, while
sons among different differential schemes indicate that for Hcen, in Fig. 8共b兲, the Hmin values from those increase toward the final
all the solutions from the first-order backward schemes, both sepa- solution. For other schemes, other than the first-order backward
rated and combined forms 共schemes 1 and 5兲, gradually decrease scheme, Hmin always increases toward the final solution. Simi-
toward the constant value, whereas all the solutions from the larly, the results from the first-order backward are the largest,
second-order backward schemes, both separated and combined whereas those from the second-order backward are the smallest.
forms 共schemes 3, 4, and 7兲, increase toward this value. Solutions The second-order central scheme yields results in between. In
obtained using the second-order central schemes 共schemes 2 and addition, the differences between Hcen and Hmin, which reflects the
6兲 are in between. For these high-speed cases, the second-order flatness of film thickness distribution in the nominal contact area,
central schemes yield solutions of the highest accuracy due to are 306 nm and 135 nm for U = 1250 mm/ s and 312.5 mm/ s, re-
their smallest truncation errors. By using a coarse mesh of 256 spectively.
⫻ 256, the solutions obtained by means of these schemes con- Figures 9 and 10 illustrate the results for the intermediate
verge to a value that can be considered sufficiently accurate. The speeds of 100 mm/ s and 30 mm/ s. The phenomena are similar to
second-order backward schemes also yield higher accuracy and those shown in the solutions for the high-speed cases. However, as
fast convergence without datum oscillation. It is fair to say that the speed and the film thickness decrease, the difference between
the second-order backward schemes are more suitable for thick Hcen and Hmin becomes smaller, approximately 67 nm for
film cases at high rolling speeds. 100 mm/ s and 30 nm for 30 mm/ s, respectively. This reduced
The solutions confirm that the effect of pressure on density is difference means that smaller film thickness corresponds to a flat-
limited. For mixed differential schemes, the final solution accu- ter film thickness distribution. Meanwhile, the pressure distribu-
racy is dominated by the film thickness derivative term. As a tion approaches the Hertzian pressure as the height of the pressure
result, although scheme 2 uses the first-order backward for the peak at the outlet reduces. Similar observations were found in
density derivative term, its solution behavior is similar to that of 关26兴. It is understood that, as the speed approaches zero, the film
scheme 5, which uses the full second-order central scheme. Simi- thickness approaches zero and the entire solution approaches that
larly, both schemes 3 and 4 use the second-order backward for the of the Hertzian dry contact.
film thickness derivative term. Although their schemes for the The results for the case of 30 mm/ s with a thinner film, as
density derivation term are different, their solutions show similar shown in Fig. 10共a兲, is different from that of the thicker film
behaviors. cases. Hcen from the first-order backward schemes 共1 and 5兲 on the
However, the tendency of Hmin variation is slightly different. In coarse mesh 128⫻ 128 becomes low compared to that on the fine
Fig. 7共b兲, all the Hmin values from the first-order backward meshes due to poor resolution in the inlet area. Similar to Hmin in