Professional Documents
Culture Documents
www.elsevier.com/locate/jtri
PII:
DOI:
Reference:
S0301-679X(15)00460-0
http://dx.doi.org/10.1016/j.triboint.2015.10.011
JTRI3885
1. Introduction
After the disc machine has been used to study the elastohydrodynamic lubrication (EHL) since 1960s,
it is found that the lubricant films were much weaker than that the theories predict from the
Newtonian behavior [1]. It was realized that the lubricants behave in a highly non-Newtonian fashion
when they pass through an EHL contact region [2]. The conditions in the EHL contact region are
severe for extremely high pressure, usually above 0.5 GPa, and very high shear rate, typically
106-108 s-1 [3]. These conditions result in the lubricant experiencing a very large increase in the
lubricant viscosity and a very high shear stress which produce the non-Newtonian behavior of the
lubricant. The non-Newtonian behavior may exhibit shear-thinning/thickening, limiting shear stress,
viscoelastic, or Maxwell behavior [4-6]. And some widely accepted constitutive equations have been
proposed to describe the behavior [7-9].
With the development of the computer technology and numerical analysis, it is now possible to
numerically simulate a variety of lubrication phenomena with different constitutive equations that
have reasonable agreement with the experiments and the practice [10-12]. The main methods to solve
the non-Newtonian lubrication problem are: (1) to obtain the Reynolds equation[13,14] by
integrating their constitutive equations and other related equations directly [15], (2) to derive the
Reynolds equation from the simplified Navier-Stokes equation and the continuity equations [16], and
(3) to deduce the generalized Reynolds equation from a non-Newtonian constitutive equation [17].
The difficulty to theoretically obtain the modified Reynolds equation due to the complicated
relationship between the shear stress and shear rate greatly limits the application of the first method.
The Navier-Stokes equation is derived through the Cauchy equation and by specifying the stress
tensor in terms of the viscosity and fluid velocity through the constitutive equation. The solution of it
is flow velocities with the assumption of constant density and viscosity. The assumptions bring about
significant errors to the velocity so that limit its application, not to mention the difficulty of
calculation.
Therefore, the generalized Reynolds equation method maybe is the only suitable one to solve the
general non-Newtonian lubrication problem. This method is based on the generalized Newtonian
fluid model. And the generalized viscosity of it is simplified to a non-linear function of the shear rate
or shear stress , = / . Although it is applicable, it uses an average Newtonian constitutive
equation by integrating the generalized viscosity across the film thickness instead of the real
non-Newtonian one. Besides, the generalized Newtonian fluid model is useful for steady simple
shear flows, while the time dependent effects, the elongational effects, and the normal stress
differences are not considered [3]. And, this method is not valid to captures the mechanics of fluids
in all situations [18,19].
Many other available methods for non-Newtonian fluid lubrication problem have also been raised,
such as the Reynolds equation put forward by Najii et al. [20], the homotopy analysis method (HAM)
[21] and etc, but the complication and simplifications during the derivation greatly limit their
applications.
2
This paper presents a unified Reynolds equation of general non-Newtonian fluid. Similar to the
Navier-Stokes method, the flow velocity is focused on. The main idea is to take the viscous flow as
the sum of the Poiseuille flow and Couette flow. And based on their special characteristics, the
velocity is obtained easily. Then, the Reynolds equation can be deduced so that the other quantities,
such as the pressure, load, elastic deformation and the lubricant film thickness can be found as a
usual Newtonian lubrication problem. Finally, comparisons with the conventional methods are
carried out, which have established the validity of the present method.
2. Unified Reynolds equation
2.1. Velocity field equation
A general lubrication system in the Cartesian coordinate system shown in Fig. 1 is studied. The
assumptions during the following derivation are: the inertial and body forces of fluids are negligible,
the flow is laminar flow, the pressure of lubricant does not vary in the film thickness direction, and
no sliding occurs at the top and bottom surfaces. The assumptions are valid for most fluid lubrication
applications.
du
x =f
dz
dv
y =f
dz
(1)
where f() represents a non-linear function, is the shear stress, is the shear rate, and u and v are
the velocities of the fluid in the x and y direction.
The x direction is taken as an example to illustrate the method. Based on the balanced force of
arbitrary control volume in the x direction, the equilibrium equation is [22]
x p
z x
where p is the pressure.
Substitute Eq.(1) into Eq. (2) yields
3
(2)
p x du
=
f
x z z dz
(3)
u f 1 z C1 dz F z C1 C2
0
x
(4)
where C1 and C2 are the constants of integration, f -1() is the inverse function of f (), F() is the
integral function of f -1().
It is obvious that the analytical expression of Eq.(4) is not easy to get, and the analytical form of C1
and C2 are even harder to obtain with nonzero boundary velocities. Bird [23] found that the solution
of the Newtonian fluid can be seen to be a sum of the solutions of the two separate problems of the
Poiseuille flow and Couette flow. Therefore, we try to use it to the non-Newtonian fluid, and treat the
viscous flow as the sum of the Poiseuille flow and Couette flow. The results show great agreement
with the conventional methods which have established the feasibility of this treatment.
2.1.1. Velocity of the Couette flow
The Couette flow is the wall driven laminar flow, and the velocity distribution of it is shown in Fig. 2.
The Couette flow is driven by the viscous drag force acting on the fluid and the applied pressure
gradient parallel to the flow direction. Since the pressure vertical to the flow direction has no effect
on the Couette flow, then the equilibrium equation, Eq.(2), of it is
x p
=0
z x
(5)
Therefore, the shear stress here is a constant. By using Eq.(4), Eq.(5) together with the boundary
conditions of the velocity of the Couette flow, uC 0 U 0 , uC h U h , it gives:
uC
Uh U0
z U0
h
(6)
The Poiseuille flow is the pressure driven flow, and the velocity distribution of it is shown in Fig. 3.
Unlike the Couette flow, it is significantly affected by the rheological properties of the fluid. And the
two main features of the Poiseuille flow are:
4
duP
dz
z h /2
0.
C1
dz z h /2 2 x
Therefore, C1
(7)
h p
, and then rearrange Eq.(7) as:
2 x
duP
h p
p
f 1 z
dz
2 x
x
(8)
(9)
According to feature (b), we have C2=0. Then, the velocity of the Poiseuille flow in the x direction of
arbitrary non-Newtonian fluid can be expressed as
p
h
h
z 0, , uP F z
2
2
x
p
h
h
z , h , uP F1 z
2
2
x
(10)
(11)
h /2
qx = udz =2
h
h /2
q y = vdz =2
p
h
1
z dz + h U 0 U h
2
2
x
p
h
1
F z dz h V0 Vh
2
2
y
(12)
qx q y h
0
x
y
t
(13)
Then the general Reynolds equation of arbitrary non-Newtonian fluid is obtained based on Eq.(12)
and Eq.(13).
p
h2
p
h
h2
h
F
z
dz
2
0 F z dz
0
x
2
y
2
x
y
h
1
1
0
Uh
Vh
2 x
2 y
t
(14)
where U U h U 0 , V Vh V0 .
3. Results and discussion
In order to investigate the validity of the method, the isothermal EHL in the line contact, as shown in
Fig. 4, has been studied. The power law model is chosen since its simplicity and realism over a
range of shear rates [25]. Besides, it can be both shear thinning (n<1) and shear thickening (n>1) as
long as the Newtonian fluid (n=1).
(15)
where is a dimensional constant, whose dimension depends on the power law index n.
h n h n Uh U0
h
1 p n n
z 0, , u
z U0
z
2
h
2
x n 1
2
1
1
1
1
n
h n h n Uh U0
h
1 p
z , h , u
z U0
z
2
h
2
x n 1
2
1
n
(16)
And the velocity expression of the power-law model of the conventional method is
1
1
n 1 p n
1
u
z c1 n c2
n 1 x
(17)
where c1 and c2 are constants that can be obtained by iteration with the boundary condition of the
velocity, u(0)=U0,u(h)=Uh.
It should be noted that when the power law index n equals to 1, the power law model is identical
with the Newtonian fluid model. Then the velocity of Newtonian fluid is also studied, and the
analytical velocity of the Newtonian fluid is [22]
1 p 2
z
z zh U h U 0 U 0
2 x
h
(18)
d h
d 2 1n 1 dp n 2 n (2n 1)
h
U h U 0
dx
n
dx
dx
(19)
0 exp ln 0 9.67 1 1 p / p0
0.6 p
0 1
1 1.7 p
The following dimensionless parameters are introduced:
0.68
(21)
(22)
p
u
x
w
hR
z
*
, = , P=
, X= , H= 2 , W
,U 0 , Z
0
0
pH
b
ER
ER
b
h
where 0 and 0 are the ambient density and viscosity, pH is the maximum Hertzian contact pressure,
b is the semi-width of the Hertzian contact region, and w is the load capacity per unit length.
The calculation condition of the simulation is w=1105N, E=2.21011 Pa, R=0.05 m, 0=0.05 Pas,
7
U0=1 m/s, Uh=0 m/s. The solution domain for the line contact simulation ranges from X=-4 to X=1.5,
which is discretized using a uniform mesh of 130 nodes. The Reynolds equation is discretized using
a mixed second order central and first order forward differencing scheme and solved along with the
equilibrium condition with a relative accuracy of 110-5.
3.2. Numerical simulation
3.2.1. Velocity distribution
The velocity distribution shown in Fig. 5 is given at X=-1,0,1 under the calculation condition shown
in Table 1.
Fig. 5. Velocity distribution of different n (U1 is the velocity of the proposed method, U2 the
conventional one, U0 the Newtonian fluid)
8
As expected, the velocities are exactly the same when n=1, which demonstrate the feasibility of the
proposed method. However, the two velocities have some deviation when n1. According to Eq.(13),
the mass flow rate should be identical through the contact region. But due to the existence of
calculation error, they are not. Since the average mass flow rate is the same for both methods, it
makes the standard deviation a good way to study the calculation error. The data listed in Table 1
show that the standard deviations of the mass flow rate of the proposed method are all lower than the
conventional one. It is because the calculation of the constants c1 and c2 in Eq.(17) introduces more
error though its expression is accurate. It makes the proposed method a more accurate way to
calculate the velocity distribution.
Table 1 Calculation condition and the standard deviation of the dimensionless mass flow rate
n
0.8
1.1
1.2
1.0
-1
P
-2
-1
7.1110
2.2310
3.708341
1.031922
5.3410-2
9.9210-1
184.7567
1.107842
-2
-2
0.823779
1.002892
-1
5.054857
1.038969
-1
174.8376
1.106950
-2
-1
0.9
5.5610
-1
2.1410
-1
1.6710
-1
1.8510
2.7910
9.8010
1.5810
2.0910
1.135737
1.003250
-1
1.17100
4.1710-1
10.67764
1.055207
-1
184.6575
1.107833
-2
1.156975
1.003720
1.0010
-1
9.9110
9.1710
2.3910
-1
2.24100
5.0410-1
16.83787
1.064575
1.9410
1.1310
342.1098
1.117491
1.83100
2.6910-2
1.17793
1.004175
-1
5.4010-1
3.4410-1
7.217006
1.046836
4.0410
-1
-1
9.6410
162.7575
1.105786
4.0410-1
2.1610-2
1.140725
1.003361
(Q1)
(Q2)
(Q0)
3.8410-2
5.9910-2
--
1.5110-2
4.5410-2
--
3.6810-2
1.5910-1
--
8.6810-2
4.7010-1
--
3.1610-3
3.1610-3
3.1610-3
Fig. 6. Pressure and film thickness distribution with n=1 (H1 and P1 are the results of the present
method, H2 and P2 the generalized Reynolds equation, H0 and P0 the Reynolds equation of
Newtonian fluid (since it is identical with H1 or P1, its line is covered up and cannot be seen))
Table 2 The mean value and standard deviation of difference of pressure or film thickness with n=1
Mean value
P0-P1
P0-P2
H0-H1
H0-H2
-7
-3.9410
1.6810-6
2.6710-6
1.3510-1
Standard deviation
9.2510-7
3.8010-2
3.1310-6
1.5110-2
A constant load is given, and the difference of the pressure distributions is quite small so they are not
studied then. Fig. 7 gives the film thickness distribution with various n, and it tells that the smaller
the power law index n, the smaller the film thickness. It is in accordance with the generally known
theory that shear-thinning results in a diminished film thickness [3].
Fig. 7. Film thickness distribution (H1 is the result of the present method, H2 the generalized
Reynolds equation)
However, difference still exists between the results of the two methods. The mean value and standard
deviation of the difference of the film thickness between the two methods are given in Table 3.
Compared to the difference when n=1, the difference of the results is acceptable. Since the presence
of the calculation error for the generalized Reynolds equation, it is true that the present method is a
10
2.2810
6.1810-2
1.3510-1
2.3910-1
3.2910-1
4. Conclusion
A unified Reynolds equation of the general non-Newtonian fluid is put forward in the present paper
by treating the total flow as the sum of the Couette flow and Poiseuille flow. An extensive set of the
full EHL line contact simulations have been presented to investigate the method. The comparisons of
the velocities and film thickness between the present method and the conventional ones not only
establish the feasibility and validity of the method, but also illustrate its simplicity in calculation and
simulation.
Acknowledgment
This research was supported by grants from the National Natural Science Foundation of China
(Grant No. 51575190 and 51175182).
References
[1] Crook AW. The lubrication of rollers, IV. Measurements of friction and effective viscosity.
Philos Trans R Soc London, Ser A 1963; 255:281312.
[2] Spikes H A. Sixty years of EHL. Lubr Sci 2006; 18(4): 265-91.
[3] Bair S, Qureshi F. The Generalized Newtonian Fluid Model and Elastohydrodynamic Film
Thickness. J Tribol 2003; 125: 70-5.
[4] Aytouna M, Paredes J, Shahidzadeh-Bonn N, et al. Drop formation in non-Newtonian fluids.
Phys Rev Lett 2013; 110(3): 034501.
[5] Scaraggi M, Perssonb BNJ. Theory of viscoelastic lubrication. Tribol Int 2014; 72: 118-30.
[6] Myers TG. Application of non-Newtonian models to thin film flow. Phys Rev E 2005; 72(6):
066302.
[7] Hirst W, Moore AJ. Elastohydrodynamic lubrication at high pressures. II. Non-Newtonian
behaviour. Proc R Soc London, Ser A 1979; 365:53765.
[8] Marcinkowska-Gapiska A, Gapinski J, Elikowski W, et al. Comparison of three rheological
models of shear flow behavior studied on blood samples from post-infarction patients. Med Biol
Eng Comput 2007; 45(9): 837-44.
[9] Gner F S, Baranak M, Soyta S, et al. Flow behavior of oil-modified polymer solutions. Prog
11
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
12
Figure captions
Fig. 1. General lubrication system
Fig. 2. Velocity distribution of the Couette flow
Fig. 3. Velocity distribution of the Poiseuille flow
Fig. 4. Coordinate system of the line contact
Fig. 5. Velocity distribution of different n (U1 is the velocity of the proposed method, U2 the
conventional one, U0 the Newtonian fluid)
Fig. 6. Pressure and film thickness distribution with n=1 (H1 and P1 are the results of the present
method, H2 and P2 the generalized Reynolds equation, H0 and P0 the Reynolds equation of
Newtonian fluid (since it is identical with H1 or P1, its line is covered up and cannot be seen))
Fig. 7. Film thickness distribution (H1 is the result of the present method, H2 the generalized
Reynolds equation)
Highlights
13