You are on page 1of 4

Chemical Engineering Science, 1969, Vol. 24, pp. 103 I-1034. Pergamon Press. Printed in Great Britain.

On the solution of boundary-layer equations for a non-Newtonian


power-law fluid
TUNCER CEBECIP, J. M. CLAUSS and J. DYER
Mechanical Engineering Department, California State College at Long Beach, Calif. 90804, U.S.A.

(First received 6 August 1968; in revisedform 6 February 1%9)

Ah&r&-The accuracy of a formula for calculating velocity gradient at the wag for power-law fluids
is investigated by obtaining the exact numerical solution of the boundary-layer equations for Howarth’s
flow. In general, the formula is found to be quite accurate for the pseudo-plastic cases, n < 1.

INTRODUCTION and very useful formula for calculating skin fric-


IN RBFERENCE[~] Acrivos, Shah and Peterson tion for flows subject to Eq. (2). In reference [2]
have presented an asymptotic method for solv- Serth and Kiser have extended the Gortler-series
ing approximately the laminar boundary-layer method for the solution of two-dimensional
equations for a power-law non-Newtonian fluid boundary-layer equations for power-law fluids
under conditions where the flow external to the and have investigated the accuracy of Eq. (1).
boundary layer has a general form. They have They have shown that the approximate method
shown that for power-law fluids, for example, of Acrivos gives, in general, a good approxima-
T = K(du/dy)“, the velocity gradient at the wall tion to the velocity gradient at the wall, but that
may be approximated by the formula errors of 20 per cent or greater can be expected
when using this method in the vicinity of the
separation point for highly pseudo-plastic fluids.
Their study was mostly limited to cases for
which n > 1, since they could obtain only the

I(*-yl*myl)n+l
first four universal functions for values of
X
n = 0.8, 0.6, and only the first two universal
functions for values of n S O-5.
+ zmJn+

3n 1I
1) 213 3/2(n+l)
(1) The purpose of the present paper is to investi-
gate the accuracy of Eq. (1) by numerically solv-
ing the boundary-layer equations for power-law
for an external velocity distribution of the form
fluids and by comparing the results with those
given by Eq. (1) for the pseudo-plastic cases,
n,(x) = { jl ai.Pi}1’2 (2) n < 1. Such an investigation has not been made
in reference[2]. In addition, it is difficult to pre-
where Fi, which is a function of n, is the trans- dict the accuracy of Eq. (1) by the Gortler-series
formed velocity gradient for a flat plate at the method with the number of universal functions
wall. Serth and Kiser have calculated, since the
Equation (2) is indeed an important equation, higher order terms neglected in their calculations
since it avoids the complexity of solving the generally become quite important close to
boundary-layer equations and presents a simple separation.

Wenior Engineer, Dougias Aircraft Company, Long Beach, Calif. 90406, U.S.A.

1031

CES Vol. 24 No. 7 A


T. CEBECI, J. M. CLAUSS and J. DYER

ANALYSIS
The two-dimensional boundary-layer equations
for a non-Newtonian power-law fluid can be
-.f%4lf+~*JL+~3&2)1 (10)
written in the form[ l]
where, for simplicity, the subscript j is dropped.
Equation ( IO) is a nonlinear differential equation
that, with the boundary conditions given by
Eq. (9), is difficult to solve. For this reason, it
(4) is first linearized. The linearization scheme is
based on quasLlinearization[3]. If Eq. (10) is
and can be expressed in the form written in the form

f’“+ .(,: ,)f(f)*-“+pY”[1-


w>*1 d.fJ’,f”,f”, 7, $) = 0 (11)

and expanded by a Taylor series around a known


= y (fyn[j-r~-fz$j (5) solution designated by 0, we obtain a third order
equation of the form
by introducing the dimensionless quantities
E "'=G(,~,~,E,E',E") (12)
which we solve by an implicit finite-difference
21 = &lKn+l, t_
method[4]. Figure 1 shows the finite-difference
ae=z; (6)
hl'
scheme.

the definition of stream function +; and the


generalized Gortler transformation

I
(j,i)
I) ij-2.i) (j-1.i)
&i-l)
JI= cw1’(n+1)f(6, 7) (7) .b (j&-2)
where
uo*- nL” Y da,
Re=pK P=T-. (8)
ue4 t
Fig. 1. Finite-difference scheme.
The boundary conditions of Eq. (5) for no mass
transfer are RESULTS AND DISCUSSION
We consider the Howarth flow[5] with main-
f(4,O) =f’(5,0) = 0, f’(S, m) = 1. (9) stream velocity distribution of the form

Next, the streamwise derivatives in Eq. (5) i&= l--n (13)


are replaced by 3-point finite-difference formulas
at ,$= &, which gives and obtain solutions of (aii/ay) ar=o for values of
n= 1, O-5, O-3, and O-1. Generally, it is accepted
f’“+ .(,2+
*)foP-“+a(f”>‘-“[l- WY1
that
(for
this flow has a separation point at R = 0.120
IZ= l), which is usually obtained by extra-
1032
On the solution of boundary layer equations

polation. It also has quite accurate solutions for When the computed shear-stress values were
n = 1[6]. For this reason, at first the accuracy of compared with those given by Smith and Clutter
the present numerical method was checked. [6], excellent agreement was obtained. Then
a comparison of values of (&i/@)+,, by the
present method and values calculated from
Eq. (1) was made. Table 1 presents the results
in tabular form and Fig. 2 presents the results in
graphic form. Figure 3 shows the separation
point, GP., for various values of n. It can be

0 I I I I I
0 02 04 0% OS IO
n
Fig. 2. Comparison of exact and approximate velocity
gradient at the wall values, n = 1,05,0.3. Fig. 3. Effect of n on flow separation point.

Table 1. Comparison of exact and approximate velocity-gradient values at the wall for the flow P, = 1- f

n=l n = 0.5 n = 0.3 n=O.l

n Present Present Present Present


f\ method Eq. (1) method Eq. (1) method Eq. (1) method Fq. (1)

0.0125 2.73932603 2.75516301 5.68723661 5.63413149 IO.53290498 10.26583803 39.49134464 38.92608452


oG250 1.76453301 l-79523%3 3.25271040 3.23004830 5.58397436 5.43882209 18*%779418 18~518902%
0.0375 l-30853038 1.33980863 2.233%394 2.22607329 366435900 3.56777143 11.65631616 11.35586393
0*0500 1.01019333 lGW93081 164193366 164424691 260800940 2.53928521 790344888 7.68115687
0.0625 0.78933164 0.83888349 1.24298434 1.25370398 192838478 1.87878819 5.62453771 5.45 144588
0.0750 0*61083402 0.67253023 0.95018319 096860075 1*45061056 1.41555241 4.10320759 396459782
0.0875 0.45622633 0.53405774 0.72277781 0~74889408 lG9500025 1.07197368 3.02499527 2.91256332
o-09375 0.38361854 0.47209538
O*lOOO 0.3 1205884 0.41377323 0.53877598 0*57310050 0.81974252 0.80733483 2*22%7178 2.13835332
0.1075 0.22371871 0.34755278
0.1125 O-16287870 0*30514501 0.38476434 0.42853383 060058532 0.59809519 1.62672839 1.55339350
0.1250 0.25 155351 0.30715792 0.42238405 0.42968407 1.16130038 1~10393071
0.1375 O-13859608 0.20360135 O-274%123 0.29266338 0.79800992 O-75558199
0.1500 0.14802781 0.18081836 051350022 048559142
O-15125 0.13715382 0.17085131
0.1625 0287475 14 0.27879407
O-16875 0~19416700 Cb19593688
0.1750 0*01036690 0.12578948

1033
T. CEBECI, J. M. CLAUS!3 and J. DYER

fluids for the pseudo-plastic cases, n < 1.


Figure 4 shows a comparison of results obtained
by the present method, by the Gortler-series
-Pnsaltmsthod method of Serth and Kiser and with those given
-- ftemncdll
by Eq. (1). Naturally the two universal functions
--RofaUlwc21
given by Serth and Kiser for this case are not
sufficient, and the values computed with these
functions are quite inaccurate. On the other hand
the agreement between the present method and
the method of Acrivos is quite remarkable.

NOTATION
K,n parameters in the power-law model
L a characteristic length
Re a Reynolds number
4? edge velocity
UO reference velocity
0 02 0.4 06 0.8 I.2 14 u v
x- and y-components of velocity
,
P
X distance along the surface from the lead-
Fig. 4. Comparison of wall velocity gradient values obtained
by three methods for n = O-l. ing edge
Y distance normal to the surface
seen that as IZdecreases, flow separation occurs Primes denote differentiation with respect to r)
farther downstream.
It was pointed out in [l] that Eq. (1) had been Greek symbols
shown to give excellent results for Newtonian p density
boundary layers. The results obtained for n = 1 r shear stress
confirm this statement and further point out that E perturbation quantity, E = f-fo, where f.
Eq. (1) is even more accurate for non-Newtonian is a known solution

REFERENCES
[l] ACRIVOS A., SHAH M. J. and PETERSEN E. E,Ghem. Engng Sci. 1965 20.
[2] SERTH R. W. and KISER K. M., Chem. Engng Sci. 1967 22.
[3] SYLVESTER R. J. and MEYER F., SIAM Jnl1965 13 2.
141 SMITH M. 0. and CEBECI T., Douglas Aircrqft Company. No. DAC-33735 1967.
151 HOWARTH L., Proc. R. Sot. 1938 AU.
[6] SMITH A. M. 0. and CLUTTER D. W., AIAA Jnll963 19.

R&urn&On &die la precision dune formule permettant de calculer le gradient de velocitt, a la


paroi de fluides de loi de puissance, en obtenant la solution num&ique exacte des Cquations en
couche limlte pour l’&oulement de Howarth. En g6n&al, la formule est assez precise pour les cas
pseudoplastiques, n < 1.

Zusammenfasmmg-Die Genauigkeit einer Formel tiir die Berechnung des Geschwindigkeitsgra-


dienten an der Wand fur nicht-Newtonische Flilssigkeiten wird durch Ermittlung der genauen numeri-
schen Liisung der Grenzschichtgleichungen fur die Howartbs-Striimung untersucht. Im allgemeinen
wird festgestellt, dass die Gleichung fur die pseudo-plastischen Fie, n < 1, recht genau ist.

1034

You might also like