Anda di halaman 1dari 13

Fluid Dynamics Research 22 (1998) 345–357

An extended study of the hydrodynamics


of gravity-driven lm ow of power-law uids
Helge I. Andersson ∗, De-Yi Shang
Division of Applied Mechanics, Norwegian University of Science and Technology,
N-7034 Trondheim, Norway
Received 10 February 1997; revised 18 August 1997; accepted 28 September 1997

Abstract
A new similarity transformation has been devised for extensive studies of accelerating non-Newtonian lm ow. The
partial di erential equations governing the hydrodynamics of the ow of a power-law uid down along an inclined plane
surface are transformed into a set of two ordinary di erential equations by means of the dimensionless velocity component
approach. Although the analysis is applicable for any angle of inclination (06 ¡ =2), the resulting one-parameter
problem involves only the power-law index n. Nevertheless, physically essential quantities, like the velocity components
and the skin-friction coecient, do depend on and relevant relationships are deduced between the vertical and inclined
cases. Accurate numerical similarity solutions are provided for n in the range from 0.1 to 2.0. The present method enables
solutions to be obtained also for highly pseudo-plastic lms, i.e. for n below 0.5. The mass ow rate entrained into the
momentum boundary layer from the inviscid freestream is expressed in terms of a dimensionless mass ux parameter ,
which depends on the dimensionless boundary layer thickness and the velocity components at the edge of the viscous
boundary layer. , which is thus an integral part of the similarity solution, turns out to decrease monotonically with n.
This parameter is of particular relevance in the determination of the streamwise position at which the entire freestream
has been entrained and viscous stresses prevail all the way to the free surface of the lm. A short-cut method to facilitate
rapid and yet accurate estimates of the mass ux parameter is developed to this end. c 1998 The Japan Society of Fluid
Mechanics Incorporated and Elsevier Science B.V. All rights reserved.

1. Introduction

Ecient heating or cooling of liquids can be achieved by allowing the uid to ow in a thin lm
along a solid surface kept at a constant temperature. While the hydrodynamics of thin lm ow of
Newtonian liquids has been extensively studied for several decades, see e.g. Andersson (1987), only
modest attention has been devoted to gravity-driven lms of non-Newtonian liquids.
Fully developed laminar lm ow along a plane surface was considered by Astarita et al. (1964),
who measured the thickness of some non-Newtonian lms for various inclinations and ow rates.

Corresponding author.

0169-5983/98/$19.00 c 1998 The Japan Society of Fluid Mechanics Incorporated and


Elsevier Science B.V. All rights reserved.
PII S 0 1 6 9 - 5 9 8 3 ( 9 7 ) 0 0 0 4 5 - 2
346 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

Fig. 1. Schematic representation of accelerating lm ow.

In a similar investigation by Therien et al. (1970), experimental data for the lm thickness were
compared with an analytical expression for the thickness of fully developed lms of power-law uids.
Sylvester et al. (1973) compared predicted and experimental characteristics for wavy but otherwise
fully developed lms of power-law and Newtonian liquids.
The hydrodynamics of developing power-law lms has been studied by means of the integral
method approach and similarity analysis. Yang and Yarbrough (1973, 1980) and Narayana Murthy
and Sarma (1977) extended the conventional integral analysis for Newtonian lms to also include
power-law uids. Later, Narayana Murthy and Sarma (1978) included the e ect of interfacial drag
at the liquid–vapour interface in a similar analysis, while Tekic et al. (1986) presented results which
accounted for the streamwise pressure gradient and surface tension. More recently, Andersson and Ir-
gens (1990a) explored the in uence of the rheology of the lm on the hydrodynamic entrance length.
A di erent approach was adopted by Andersson and Irgens (1988, 1990b) namely to divide the ac-
celerating lm ow into a developing viscous boundary layer and an external inviscid freestream, as
depicted in Fig. 1. They furthermore demonstrated that a similarity transformation exists, such that
the boundary layer momentum equation for power-law uids is exactly transformed into a Falkner–
Skan type ordinary di erential equation. The resulting two-point boundary-value problem was solved
numerically with a standard shooting technique based on classical 4th-order Runge–Kutta intergra-
tion in combination with a Newton iteration procedure. Numerical results were obtained for values
of the power-law index n in the range 0:56n62:0. It was conjectured that converged results could
have been obtained also for highly pseudo-plastic uids, i.e. for n ¡ 0:5, by a di erent integration
technique, for instance a nite-di erence scheme.
More recently, Kafoussias and Williams (1993) pointed out some of the diculties associated
with the shooting method as applied to two-point boundary value problems and advocated the use
of nite-di erence techniques. Andersson and Toften (1989) successfully adopted the implicit Keller
Box scheme for a variety of power-law uid ows, including both similar and non-similar boundary
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 347

layer problems. This scheme is unconditionally stable, permits arbitrary grid spacing and has second-
order accuracy in all variables. The resulting non-linear system of algebraic equations was linearized
by Newton’s method. Due to the severe non-linearity introduced in the viscous term by the power-law
model, the ability of the scheme to converge depended crucially on whether the viscous term was
incompletely or fully linearized and whether some under-relaxation was allowed for in the iter-
ative Newton process. It is noteworthy that for the semi-in nite at plate, the fully linearized
nite-di erence scheme failed to converge for highly pseudo-plastic uids (n ¡ 0:3) while incomplete
linearization led to accurately converged solutions even for n = 0:05. It can therefore be concluded
that the non-linearity of the rheological model makes the numerical solution procedure sensitive to
non-physical parameters inherent in the adopted solution scheme.
In the present paper, a new similarity transformation is devised for non-Newtonian boundary layer
problems. This transformation is an extension of a transformation introduced by Shang and Wang
(1990) for Newtonian boundary layers and employed successfully to free convection by Shang and
Wang (1990) and Shang et al. (1993) and to lm condensation by Shang and Adamek (1994)
with the particular view to account for the variability of the thermophysical properties of the
uid. Now, the transformation is generalized to non-Newtonian uids which obey the non-linear
power-law equation-of-state and applied to gravity-driven lm ow. By means of a conventional
shooting method, which closely resembles that used by Andersson and Irgens (1988), numerical
results can now be obtained even for highly pseudo-plastic substances, i.e. n ¡ 0:5. Moreover, a
short-cut method for estimation of the mass ow rate within the lm will be devised. This approach
is based on correlation formulas derived from the accurately computed similarity solutions.

2. Physical model and governing partial di erential equations

Consider the accelerating laminar ow of a non-Newtonian liquid lm down along an inclined


plane surface, as shown schematically in Fig. 1. The incompressible and inelastic uid is assumed
to obey the Ostwald-de-Waele power-law model and the action of viscous stresses is con ned to the
developing momentum boundary layer adjacent to the solid surface. The basic conservation equations
for mass and momentum within the viscous boundary layer are:
@wx @wy
+ =0 (1)
@x @y
 (n−1)
@wx @wx K @wx @2 w x
wx + wy = g cos + n (2)
@x @y  @y @y2
with boundary conditions
y = 0: wx = 0; wy = 0; (3)
y → l : wx → wx;∞ (4)
where wx and wy are velocity components in the x and y directions, respectively, whereas g and
denote the gravitational acceleration and the angle of inclination of the plane wall. Here, it has been
anticipated that @wx =@y¿0 throughout the entire lm. The uid properties ; K and n, which are
348 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

assumed to be constant in the present analysis, are the density, the coecient of consistency and
the power-law index, respectively. The two-parameter rheological model represents pseudo-plastic or
shear-thinning uids if the power-law index n is smaller than unity and dilatant or shear-thickening
uids for n ¿ 1. The deviation of n from unity indicates the degree of deviation from Newtonian
rheology and the particular case n = 1 represents a Newtonian uid with dynamic coecient of
viscosity K.
No-slip and impermeability at the inclined wall y = 0 are expressed by the boundary conditions
given by Eq. (3), while the outer condition, Eq. (4), assures that the velocity component wx within
the boundary layer approaches the external velocity
p
wx;∞ = 2gx cos (5)

at the edge y = l of the boundary layer. Since the frictionless ow between the viscous boundary
layer and the free streamline bordering the constant-pressure atmosphere is quasi-one-dimensional,
the simple solution given by Eq. (5) is readily derived from Eq. (2) by assuming wx;∞ = 0 (and
in nite lm thickness) at the inlet x = 0, cf. Andersson and Irgens (1990b).
It may be worthwhile to recall that the boundary layer theory conventionally adopted in the analysis
of thin- lm ow may be inadequate if the Reynolds number is too low. Wu and Thompson (1996)
compared boundary layer theory predictions with solutions of the full Cauchy equation for ow of
a shear-thinning power-law uid past a at plate of length L. They found that the Reynolds number
Rex (with x = L), below which the boundary layer approximations become inaccurate, decreased
from 120 for a Newtonian uid (n = 1) to 45 for a highly pseudo-plastic uid (n = 0:1).

3. A new similarity transformation

Incidentally, as pointed out by Andersson and Irgens (1988), the external velocity, Eq. (5), belongs
to the Falkner–Skan class of freestreams wx;∞ ∝ xm which permits a similarity transformation of the
momentum boundary layer equation even for power-law uids. A generalized Falkner–Skan type
of transformation was therefore introduced by Andersson and Irgens (1988, 1990b). In the present
study, however, a new transformation is proposed, which unlike the Falkner–Skan type of approach
does not involve the stream function.
Let us rst introduce the dimensionless similarity variable

 = (y=x)(Rex )1=(n+1) (6)

where

Rex = xn (wx;∞ )2−n =K (7)

is a generalized local Reynolds number. Now, we follow Shang and Wang (1990) and de ne di-
mensionless velocity components
p
Wx = wx = 2gx cos ; (8)
p
Wy = (wy = 2gx cos )(Rex )1=(n+1) (9)
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 349

in the x and y directions, respectively. These dimensionless variables become analogous to the
similarity transformations used by Shang and Wang (1990) and Shang et al. (1993) for the particular
parameter value n = 1.
The partial di erential equations given by Eqs. (1) and (2) and their boundary conditions given
by Eqs. (3) and (4) are now transformed into the ordinary di erential equations
n dWx dWy
Wx −  +2 = 0; (10)
n + 1 d d
   (n−1)
−n dWx dWx dWx d 2 Wx
Wx  + Wx + 2Wy = 1 + 2n (11)
n + 1 d d d d2
subject to

=0: Wx = 0; Wy = 0; (12)

=∞: Wx = 1: (13)

Evidently, the power-law index n is the only explicit parameter in the transformed problem.
In order to demonstrate the linkage of the present approach with that of Andersson and Irgens
(1988, 1990b), the physical stream function (x; y) = xwx;∞ · Re−1=(n+1)
x f can be introduced. If
the velocity components are expressed in terms of the dimensionless stream function f(), the
mass continuity equation, Eq. (10), is automatically satis ed and the momentum equation, Eq. (11),
becomes an ODE for the unknown f, which is practically identical to the generalized Falkner–Skan
equation derived by Andersson and Irgens (1988).

4. Numerical solutions

The non-linear two-point boundary value problem de ned by Eqs. (10)–(13) was solved numeri-
cally for several values of the power-law index in the range 0:16n62:0. Here, the same shooting
method as that used earlier by Shang and Wang (1990) and Shang et al. (1993) was adopted. First,
Eqs. (10) and (11) were written as a system of three rst-order di erential equations, which was
solved by means of fth-order Runge–Kutta integration. Then, a Newton iteration procedure was
employed to satisfy the outer boundary condition, Eq. (13). As far as the numerical procedure is
concerned, the only essential di erence between the present approach and that of Andersson and
Irgens (1988) is that the fourth-order Runge–Kutta scheme used by Andersson and Irgens used
equidistant grid spacing, while the present fth-order scheme utilizes variable grid spacing.
Some of the computed velocity pro les Wx () are shown in Fig. 2. The power-law index appears
to have a substantial e ect on the velocity distribution within the boundary layer and, as observed
already by Andersson and Irgens (1988), the most striking feature being the monotonic thinning
of the boundary layer with increasing n-values. This is fully consistent with the ndings for other
two-dimensional plane ows, for instance the non-Newtonian analogue of the classical Blasius prob-
lem, i.e. ow past a semi-in nite at plate, which was rst solved by Acrivos et al. (1960) and
more recently by Andersson and Toften (1989).
350 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

Fig. 2. Numerical solutions for the streamwise velocity component Wx (). (Curves 1–9: n = 2:0; 1:5; 1:2; 1:0; 0:7; 0:5;
0:3; 0:2; 0:1:)

The gradient of the dimensionless velocity Wx () at the wall  = 0 is perhaps the single most
important characteristic of the solution. This is because the local skin-friction coecient Cf is a
dimensionless measure of the shear stress  = K(@wx =@y)n at the wall, i.e.
"  #n
w dWx
Cf ≡ = 2Re−1=(n+1)
x : (14)
1=2wx;∞
2 d =0

The numerical results for the dimensionless velocity gradient at the wall (dWx =d)=0 are given in
Table 1 and shown in Fig. 3, from which it is observed that the wall gradient gradually decreases
with increasing n. It is noteworthy, however, that since the local Reynolds number Rex , as de ned
in Eq. (7), varies as ∝ x(n+2)=2 , the streamwise variation of Cf becomes Cf ∝ x−(n+2)=2(n+1) , i.e.
the skin-friction coecient decreases in the streamwise direction, irrespective of the value of the
power-law index.
The non-linearity of the highest-order derivative in Eq. (11) increases with increasing devia-
tion of the power-law index n from unity. The thickening of the boundary layer with increasing
pseudo-plasticity 1 − n, in combination with the steeper slope of the dimensionless velocity pro le
Wx (), makes the mathematical problem de ned by Eqs. (10)–(13) increasingly sti . In fact, the
shooting method becomes gradually less attractive as the distance (in boundary layer coordinates )
from the wall to the outer edge of the calculation domain, at which the condition Wx = 1 should be
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 351

Table 1
Computed variation of (dWx =d)=0 with power-law index n
n Present Eq. (15) Eq. (16)

0.1 3.57308 — 3.6382


0.15 2.48411 — —
0.2 1.96020 — 2.0010
0.25 1.65736 — —
0.3 1.46275 — 1.4892
0.4 1.23218 — —
0.5 1.10437 1.1047 1.1234
0.6 1.02613 — —
0.7 0.97519 0.9753 —
1.0 0.89972 0.8997 0.9122
1.2 0.87902 0.8790 —
1.5 0.86592 0.8659 0.8749
2.0 0.86360 0.8636 0.8705

satis ed, increases. This is most likely the reason why Andersson and Irgens (1988) failed to obtain
converged solutions for highly pseudo-plastic uids with n ¡ 0:5. In the present study, however, this
diculty was remedied by using variable grid spacing.
In order to assess the accuracy of the present numerical results, comparisons are made with the
calculations by Andersson and Irgens (1988) according to the relationship
   1=(n+1)
dWx 3
= f00 (0) (15)
d =0 4

where f denotes the Falkner–Skan type stream function and the primes signify di erentiation with
respect to the similarity variable adopted in their analysis. Data obtained from the approximate
interpolation formula

  ( " #)1=(n+1)
dWx n + 1=2 (n + 1)2
= (f000 )n+1 + (16)
d =0 n+1 3n(n + 1=2)

derived by Acrivos et al. (1965) are also included in Table 1. Here, f000 denotes the dimensionless
wall shear stress for power-law boundary layer ow past a at plate, for which data are tabulated
by Acrivos et al. (1965).
The comparison in Table 1 shows that the present numerical solutions are practically indistin-
guishable from the similarity solutions of Andersson and Irgens (1988). The data derived from the
approximate formula given by Eq. (16) compares surprisingly well with the present similarity so-
lutions and the velocity gradient at the wall is overpredicted by no more than 2% throughout the
entire range of n-values considered. Here, it should be recalled that the similarity solutions can be
considered as exact in the sense that they do not involve other approximations than those inherent
in the boundary layer theory and the adoption of the power-law model.
352 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

Fig. 3. Numerical solutions for the dimensionless velocity gradient (dWx =d)=0 for di erent values of the power-law
index n.

5. Mass ow rate

Although the total mass ow rate within the lm is constant, the partition of the mass ow rate
between the viscous boundary layer and the external inviscid ow varies in the streamwise direction.
As the boundary layer thickens, uid is continuously being entrained from the freestream. Let Gx
denote the entire mass ux entering into the boundary layer from the freestream upstream of a
certain position x. The mass ux dGx entrained into an element of the boundary layer of streamwise
extent dx (and unit width) is
 
dl
dGx =  wx;l − wy;l d x (17)
dx
according to the principle of mass conservation. Since the boundary layer thickness is given as
l = l (Rex )−1=(n+1) x, Eq. (17) can be expressed in terms of dimensionless variables as
 
dGx p nl
=  2gx cos (Rex )−1=(n+1) Wx;l − Wy;l : (18)
dx 2(n + 1)
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 353

Fig. 4. Numerical results for l ; Wx;l ; Wy;l and  for di erent values of the power-law index n. (1. l ; 2. , 3. −Wy;l
and 4. Wx;l .)

The mass ux entering the boundary layer from the inlet x = 0 to a streamwise position x is
Z x
Gx = dGx (19)
0

which after integration can be expressed in dimensionless form as


 
Gx Re1=(n+1)
x 2(n + 1) nl
=≡ Wx;l − Wy;l : (20)
wx;∞ x 2n + 1 2(n + 1)

Here, the right-hand side de nes the mass ux parameter . Since the dimensionless boundary layer
thickness l and the velocity components Wx; l and Wy; l at the edge of the boundary layer may
depend on the power-law index n;  turns out to be a function of n alone. The most frequently
used de nition of l is the value of  for which the dimensionless velocity component Wx in Fig. 2
becomes equal to 0.99. Data for l ; Wx;l and Wy;l obtained from the numerical similarity solutions
presented in Section 4 are shown in Fig. 4. The resulting variation of the dimensionless mass ux
354 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

parameter  is also included and it is observed that  is a monotonically decreasing function of the
power-law exponent n.
To facilitate rapid estimates of the mass ux parameter  for any value of the power-law index
in the interval 0:26n62, accurate curve- t formulas for l and Wy;l
l = 4:9505 − 7:617(n − 0:54) + 11:214(n − 0:54)2 + 8:703(n − 0:54)3 − 0:37(n − 0:54)4
(0:26n61) (21)

l = 2:3201 − 1:0623(n − 1:425) + 0:9962(n − 1:425)2 − 0:7533(n − 1:425)3


(16n62) (22)

Wy;l = −1:8675 + 3:9616(n − 0:54) − 6:022(n − 0:54)2 − 3:22(n − 0:54)3 + 16:946(n − 0:54)4
(0:26n61) (23)

Wy;l = −0:53954 + 0:5002(n − 1:425) − 0:5078(n − 1:425)2 + 0:3946(n − 1:425)3


(16n62) (24)
can be used together with Wx; l = 0:99 in Eq. (20). This short-cut method turns out to be accurate
to within 0.01%.

6. Length of boundary layer region

Let us now denote the total mass ow rate within the lm Q, where Q is the volumetric ow
rate. Since the viscous boundary layer develops from x = 0, i.e. l (0) = 0, the entire mass ow is
initially carried by the freestream. At a certain streamwise position x = x0 , on the other hand, the
boundary layer extends all the way to the free surface of the lm and the total mass ux is within
the boundary layer, i.e.
Gx0 ≡ Gx (x0 ) = Q: (25)
This criterion, in combination with Eq. (20), can be rearranged to give the explicit relation
" n+1 #2=(2n+1)
Q (2g cos ) (1−2n)=2
x0 = (26)
 K=
for the particular streamwise position x0 . Since the lm inlet is at x = 0, cf. Fig. 1, the characteristic
coordinate value x0 de nes the streamwise length of the boundary layer region.

7. Critical lm thickness

When the boundary layer extends all the way to the free surface and the freestream disappears
at x = x0 , the lm, thickness equals the boundary layer thickness l (x0 ). The latter can be obtained
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 355

from the de nition, Eq. (6), of the similarity variable at the outer edge of the viscous boundary
layer, i.e. y = l (x0 ) for  = l and x = x0 , and expressed as
!−1=(n+1)
(2g cos ) (2−n)=2
l (x0 ) = l · x0n=2(n+1) : (27)
K=

It can therefore be concluded that both x0 and l (x0 ) are completely determined as long as the
problem characteristics n; K=; Q and g cos are known, along with the solution of the transformed
problem, Eqs. (10)–(13), which determines .
The lm thickness at the particular position x = x0 is a critical quantity in lm ow analysis
since the boundary layer concept is applicable only upstream of x0 where the local lm thickness
is everywhere greater than l (x0 ). Downstream of x0 , on the other hand, the ow is a ected by
viscous shear all the way to the free surface and the lm is everywhere thinner than the critical lm
thickness l (x0 ). Even further downstream, the action of the viscous shear stress  exactly balances
the gravitational acceleration g cos and a region of uniform lm thickness is established.

8. E ect of wall inclination

It is noteworthy that the angle of inclination does not appear in the transformed problem de ned
by Eqs. (10)–(13). Any solution Wx () and Wy () is accordingly independent of but, nevertheless,
valid for all inclinations 06 ¡ =2. Physically relevant quantities, on the other hand, do depend
on due to the similarity transformation, Eqs. (6)–(9). For a given quantity, say P, the relationship
 
Pi cos i
= = cos i (28)
Pv cos v
between the inclined and vertical cases, identi ed by subscripts i and v, respectively, holds. Here,
v denotes the angle of inclination in the vertical case, i.e. v = 0, and the exponent is provided
in Table 2 for some quantities of particular interest.

Table 2
Relationship between inclined and vertical lm ow
P wx wy Cf x0 l (x0 )
1 3 n−2 1 − 2n −1

2 2(n + 1) 2(n + 1) 2n + 1 2n + 1
The exponent in Eq. (28) depends on the physical quantity P under con-
sideration

9. Conclusions

A new similarity transformation has been used to study the gravity-driven ow of a non-Newtonian
liquid lm along an inclined surface. The partial di erential equations governing the hydrodynamics
356 H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357

of the power-law uid transform exactly into a set of two ordinary di erential equations, which can
be integrated numerically to an arbitrary degree of accuracy. The results are practically indistin-
guishable from those of Andersson and Irgens (1988) in the parameter range 0:56n62:0. With the
present approach, however, calculations could be accomplished also for highly pseudo-plastic liquids
and the numerical results compared accurately with results deduced from an approximate interpola-
tion formula due to Acrivos et al. (1965) throughout the entire parameter range of 0:16n62:0. The
non-linearity of the momentum boundary layer problem for a power-law uid increases with increas-
ing pseudo-plasticity 1 − n and the use of variable grid spacing is therefore increasingly important
for small n-values.
It is noteworthy that the resulting system of dimensionless ordinary di erential equations depends
only on the single parameter n. All other parameters, like the streamwise location x, the uid
properties K=, and the component of the gravitational acceleration along the surface gcos have
been combined into a generalized local Reynolds number Rex and dimensionless velocity components
Wx and Wy . Various ow characteristics can thus be expressed only in terms of n and Rex , except the
particular position x0 at which the entire freestream has been entrained into the momentum boundary
layer and viscous shear prevails throughout the lm. In order to determine x0 and the associated
critical lm thickness l (x0 ), knowledge about the total mass ow rate Q within the lm is also
required, together with the new dimensionless mass ux parameter . The latter quantity, which
depends on the dimensionless boundary layer thickness l and the velocity components Wx;l and
Wy;l at the edge of the boundary layer, is generally obtained as an integral part of the numerical
solution of the transformed problem and turned out to be a function only of the power-law index n.
However, to facilitate rapid and accurate estimates of , polynomial curve- t formulas have been
developed on the basis of the rigorous similarity solutions.
The new transformation is a generalization of a recent similarity transformation proposed by Shang
and Wang (1990) for boundary layer ow of Newtonian uids. The ability of the transformation to
handle also the accompanying heat transfer problem (even with variable thermophysical properties)
for Newtonian uids, as demonstrated by Shang and Adamek (1994), makes us believe that the
generalized transformation should be applicable to the analysis of heat transfer in power-law lms.

Acknowledgements

This study was partly supported by The Research Council of Norway under contract number
110657=410. The authors are grateful to the referees for their constructive criticism which helped to
improve the paper.

References

Acrivos, A., Shah, M.J., Petersen, E.E., 1960. Momentum and heat transfer in laminar boundary-layer ows of
non-Newtonian uids past external surfaces, AIChE J. 6, 312–317.
Acrivos, A., Shah, M.J., Petersen, E.E., 1965. On the solution of the two-dimensional boundary-layer ow equations for
a non-Newtonian power-law uid, Chem. Eng. Sci. 20, 101–105.
Andersson, H.I., 1987. The momentum integral approach to laminar thin- lm ow. In: Proc. ASME Symp. on Thin Fluid
Films, Cincinatti, OH, FED-48, pp. 7–13.
H.I. Andersson, D.-Y. Shang / Fluid Dynamics Research 22 (1998) 345–357 357

Andersson, H.I., Irgens, F., 1988. Gravity-driven laminar lm ow of power-law uids along vertical walls.
J. Non-Newtonian Fluid Mech. 27, 153–172.
Andersson, H.I., Irgens, F., 1990a. Hydrodynamic entrance length of non-Newtonian liquid lms. Chem. Eng. Sci. 45,
537–541.
Andersson, H.I., Irgens, F., 1990b. Film ow of power-law uids, in: Cheremisino , N.P. (Ed.), Encyclopedia of Fluid
Mechanics, vol. 9, Gulf Publishing, Houston, TX, pp. 617– 648.
Andersson, H.I., Toften, T.H., 1989. Numerical solution of the laminar boundary layer equations for power-law uids,
J. Non-Newtonian Fluid Mech. 32, 175–195.
Astarita, G., Marrucci, G., Palumbo, G., 1964. Non-Newtonian gravity ow along inclined plane surfaces, Ind. Eng. Chem.
Fundam. 3, 333–339.
Kafoussias, N.G., Williams, E.W., 1993 An improved approximation technique to obtain numerical solution of a class of
two-point boundary value similarity problems in uid mechanics, Int. J. Numer. Methods Fluids. 17, 145–162.
Narayana Murthy, V., Sarma, P.K., 1977. A note on hydrodynamic entrance lengths of non-Newtonian laminar falling
lms. Chem. Eng. Sci. 32, 566 –567.
Narayana Murthy, V., Sarma, P.K., 1978. Dynamics of developing laminar non-Newtonian falling liquid lms with free
surface. ASME J. Appl. Mech. 45, 19–24.
Shang, D.Y., Adamek, T., 1994 Study on laminar lm condensation of saturated steam on a vertical at plate for
consideration of various physical factors including variable thermophysical properties, Warme- und Sto ubertrag. 30,
89–100.
Shang, D.Y., Wang, B.X., 1990 E ect of variable thermophysical properties on laminar free convection of gas, Int. J.
Heat Mass Transfer 33, 1387–1395.
Shang, D.Y., Wang, B.X., Wang, Y., Quan, Y., 1993 Study on liquid laminar free convection with consideration of
variable thermophysical properties, Int. J. Heat Mass Transfer 36, 3411–3419.
Sylvester, N.D., Tyler, J.S., Skelland, A.H.P., 1973 Non-Newtonian thin lms: theory and experiment, Can. J. Chem. Eng.
51, 418– 429.
Tekic, M.N., Posarac, D., Petrovic, D., 1986 A note on the entrance region lengths of non-Newtonian laminar falling
lms, Chem. Eng. Sci. 41, 3230–3232.
Therien, N., Coupal, B., Corneille, J.L., 1970. Veri cation experimentale de l’epaisseur du lm pour des liquides
non-Newtoniens s’ecoulant par gravite sur un plan incline, Can. J. Chem. Eng. 48, 17–20.
Wu, J., Thompson, M.C., 1996. Non-Newtonian shear-thinning ows past a at plate, J. Non-Newtonian Fluid Mech. 66,
127–144.
Yang, T.M.T., Yarbrough, D.W., 1973. A numerical study of the laminar ow of non-Newtonian uids along a vertical
wall, ASME J. Appl. Mech. 40, 290–292.
Yang, T.M.T., Yarbrough, D.W., 1980. Laminar ow of non-Newtonian liquid lms inside a verticle pipe, Rheol. Acta
19, 432– 436.

Anda mungkin juga menyukai