Fan-Chao1965 Article UnsteadyLaminarIncompressibleFrect

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Vol.

16, 1965 351

Unsteady, Laminar, Incompressible Flow Through


Rectangular Ducts 1)
By CttlESI FAN, School of Engineering Science, Florida State University, Tallahassee,
Florida and BEI-TsE CHAO, Dpt. of Mech. and Ind. Eng., University of Illinois, Urbana,
Illinois, USA

I. Introduction
The steady, laminar, fully developed flow of viscous, incompressible fluids between
two parallel plates has long been studied. The velocity distribution is parabolic and
the pressure gradient along the direction of the flow is constant. LANDAU and LIF-
SrlITZ [2] 3) have reported the results of a purely harmonic flow. For non-steady flow
in rectangular conduits there is a conspicuous lack of information in the literature.
In the following we present some results of a study of parallel, viscous, incompres-
sible flow through long rectangular ducts when the axial pressure gradient is an
arbitrary function of time. The solution for an impulsive pressure gradient is first
sought by transforming the governing equation of motion to a homogeneous one along
with appropriate changes in initial conditions. Solutions for other time dependent
pressure gradients are subsequently obtained through appropriate integration of the
impulse response.

2. Governing Equations
For parallel flow of an incompressible Newtonian fluid of constant viscosity in a
horizontal rectangular duct of uniform cross-section, the axial velocity w satisfies
the following equations of motion and continuity

0w 0p t 02~
Q o r - = -- a T + # ~-J2i- +
02w~
Oy~ ] ' (1)

o~_ = 0 (2)
0z
where x , y , z are the three space coordinates as shown in Figure 1, t the time variable,
Q the density, p the pressure, and # the dynamic viscosity of the fluid.

~ z 2b 0 x
Y _

Figure 1
Coordinate System.
1) This paper summarizes the second portion of the Ph.D. Thesis [1]~) by C. FAN.
~) Numbers in brackets refer to References, page 360.
352 CtIIEN FAN a n d BEI-TsE CHAO ZAMP

I t is seen that w is only a function of x, y and t. Since p is independent of x and is


hydro-statically distributed in the y direction, the axial pressure gradient Op/Oz is
spatially uniform but could be time dependent. By letting Op/Oz = ~ f ( t ) , Equation (1)
becomes
[ 02w c)2w ~ Ow
v ~ - 3 ~ + Oy" ! Ot - - f ( t ) ' (3)

where v is the kinematic viscosity of the fluid. We shall consider that initially the
fluid is at rest, i.e., w ( x , y , 0 ) = 0. The appropriate boundary conditions are
w ( • a, y, t) = 0 and w(x, 4- b, t) = 0. A second physical requirement is that w must
be finite.
3. F l o w U n d e r an I m p u l s i v e P r e s s u r e G r a d i e n t

We shall first seek a solution when f ( t ) = - - A 6(t), where A is a positive constant


and ~(t) is the Dirac delta function. Physically this represents the case when an
impulsive pressure gradient of magnitude ~A is suddenly impressed on the fluid at
t : 0 +. Here the problem is completely prescribed b y

V \ Ox ~ + Oy ~1 Ot -- '
(4)
w ( x , y, O) = O , w ( 4 - a, y, t) = O , w ( x , 4- b, t) = O .

Without losing any generality, we have chosen A = 1 for convenience.


The solution of the foregoing set of equations can he shown mathematically to be
identical to that of the following
( O~w O~w ] Ow _ 0
v \ Ox 2 + Oy 2 ] Ot (5)
w(x, y, O+) = l , w ( 4 - a, y, t) = O , w ( x , 4- b, t) = O .

We note that the differential equation in (5) is now homogeneous but the initial
condition no longer is. A physical explanation for the equivalence of (4) and (5) m a y
be given" Immediately after the impulse is applied, only the inertia force would be
acting and thus the equation of motion m a y be represented b y d w / d t = d(t) which,
upon integration, gives
w=/d(t) dt=l, e - + O +.
,J
0

This means that o p / O z = 0 for t > 0 + and at t = 0% w ( x , y , 0 +) = 1.


The solution of (5) is found to be

16 ~ ~ ( - 1),o+,, /2 m + 1t ~ x (2 n + 1 ) ~ y
w(x, y, t) = ~ m-O~ .:0~'~ (2 ~ q-~) ( 2 ~ + 1) COS 2~ COS 2b
~' r(2~+i), (2,+i),i (6)

which is uniformly convergent. It represents the unit impulse response of the system
under consideration. It also represents the motion of fluid inside a rectangular duct
which moves at constant unit velocity for t ~ O, and is suddenly stopped at t = 0+.
Vol. 16, 1965 Unsteady, Laminar, Incompressible Flow T h r o u g h Rectangular Ducts 353

Once the impulse response is obtained, the solution of Equation (3) for any
arbitrary f ( t ) is given by either of the following convolution integrals
t
w(x, y, t) = - f f(2) w, (x, y, t - - 2) d2, (7a)
0
t
= -ff
o
(t -- 2) wi(x, y, 2) d 2 . (7b)

The subscript i refers to the impulse response as given b y (6).

4. F l o w U n d e r a C o n s t a n t P r e s s u r e G r a d i e n t

Here,
f(t) = - - K o H(t) (8)

where K 0 is a positive constant and H(t) is the Heaviside unit step function.
Upon replacing t by t -- 2 in (6), followed by its substitution into (7a), one obtains,
after integration and some re-arrangement, the solution in dimensionless form

Z~(ff, g , Z ) = __ (1//*) (Op/Oz) 4 a b = ~-m=o n=o


( - 1),.+., ~
(9)
X ( 2 m + 1 ) a ( 2 n + 1) + ( 2 m + 1) ( 2 n + 1)aq)e

• ( 2 m 2+ 1) ~_ ~ecos (2 n +2 1)~ $ l - - e - ~ [ ~ - - + ( 2 . + 1 ) ~ ] ,

where ~ = x/a, ~ = y / b , T = v t/a b, and 9 = a/b which is the aspect ratio of the duct
cross-sectional area. Figure 2 shows the transient velocity distributions in the central
plane of a square duct at several times after the application of a constant pressure

-- l i I ~ I i /4/

-/.0
Figure 2
Transient velocity distributions for flow under constant pressure gradient.

ZAMP 16/23
354 Cu,rN FAN and BEI-TsE CHAO ZAMP

gradient. As expected, at small times the viscous effect is confined to a narrow region
adjacent to the wall. In general, the manner of transition to steady motion is similar
to that in pipe flow [3].
Solution (9) is the step response. The solution of (3) for any arbitraryf(t) m a y also
be derived from either of the following Duhamel's integrals,
t
w(x, y, t) = --f(O) w,(x, y, t) - - / w ~ ( x , y, t -- 2) f'(~) d~, (lOa)
0

t
= -ff(2) ows(x, ot
y' t - 2) d2 (10b)
0

where w~ is the unit step response and is given b y w / K o of (9).


Once the velocity distribution is known, other quantities of interest such as the
mass flow rate, sectional average velocity, shear stress, etc. can all be obtained.
These are given in Reference [1I.
The friction factor m a y be defined as

Drag/Wetted area D/(a + b) (11)


f= 4 (1/2) e(Wa,,oo) 2 = (1/2) e(Wa~,oo) 2

where D is the drag force per unit length of the duct and way, oo is the steady state
cross-sectional average velocity. Both of which can be derived from (9). Thus one
obtains

.~=o,~=o ( 2 m + 1 ) 2 ( 2 n + 1) 2 e ~ --+(2n+1)~*]* /
1 V4
f = R e 8 hal2 (~' + 1)-
22
,~o,=0 (2m+1)4(2n+l)2+(2m+1)2(2n+l)4~0~
(12)
where R e is the Reynolds number based on equivalent diameter of the duct, i.e.,
R e = de w . . . . Q/# and d e = 4 1/a b/~. We note that d e is not the conventional equiva-
lent diameter based on the hydraulic radius.
B y letting ~ --~ ~ , the steady state friction factor results, and, for a square duct,
it reduces to
64 (12 a)
f~176 Re

which is exactly the same as that of the Darcy-Weisbach friction factor for pipe flow.
This means that the frictional force per unit length of perimeter per unit mass ilow
rate is the same for either a circular pipe or a square duct of identical cross-sectional
area. Since the square duct has 13 per cent longer perimeter than the pipe, the wall
frictional drag is larger by the same amount.
Figure 3 shows the variation of the product f. Re with ~ for several values of the
aspect ratio ~0. In the lower right-hand corner, the steady state value is shown plotted
against ~v. It is seen that the narrower the duct (smaller q~) the more friction it develops.
Vol, 16, 1965 Unsteady, Laminar, Incompressible Flow Through Rectangular Ducts 355

T h e m i n i m u m friction f a c t o r occurs w h e n 99 = 1. I t is clear t h a t , in F i g u r e 3, all t h e


c u r v e s r e m a i n u n a l t e r e d if 9 is r e p l a c e d b y its reciprocal.

120 @=0.2

/.Re

80

z5

_ t ~ 0.1, 02 ~ ~g4 a.6 0810 ,__

0.2 ~, a.6 a.8 ZOr

Figure 3
Variation of transient friction factors with ~b based on same mass flow rate.

T h e r a t e of e n e r g y d i s s i p a t i o n p e r u n i t l e n g t h of t h e d u c t is [4]
a b
u,=4Z[ f ri
l\~-! +\Oyll
axay (13)
0 0

w h i c h , for t h e s t e a d y s t a t e , gives

E i, co _ ~o
(14)
#(wavoo) 2
,
64 m=0 n=0
(2m+ 1) 4 ( 2 n + 1) 2 + ( 2 m + 1) 2 ( 2 n + 1) 4cp2

E q u a t i o n (14) e x h i b i t s a m i n i m u m w h e n 9 = 1.

5. F l o w U n d e r a Harmonically Oscillating Pressure Gradient

H e r e , we c o n s i d e r
f (t) = - - K cosro t . (15)

A p p l i c a t i o n of t h e C o n v o l u t i o n i n t e g r a l (7) or t h e DUHAMEL'S i n t e g r a l (10) l e a d s t o

w ( x , y, t) -- K 16 ~ 2 (-- 1)~+~ (2 m + 1) x * cos ( 2 n + l ) xy


60 ~2 (2 m ~ - 1 i ( 2 ~ - ~ - 1) COS 2~ 2 b
(16)
co v (~2/4) [m, n] cos co t + co~ sin co t co v [m, n] e-(~,'/i)Ira,-> t /
X 6o~ + v2 (~t/16) [m, ~]2 -- ~ ~ 2 ( ~ 4 ~ i ~ ) ~ , n]~ J,
356 CHIEN FAN and BEI-TsE CHAO ZAMP

in which [m, n] is an abbreviation for,


[ ( 2 m + 1/2 1t2]
a2 + b2 . (16a)

E q u a t i o n (16) is uniformly convergent and satisfies the initial and b o u n d a r y condi-


tions. As t -> o~, we arrive at the purely periodic solution which, in dimensionless
form, becomes

t)
w 6 (- 1/'o+. I2 m + ,1
K/co :~2 ~ ~ ( 2 m + 1) ( 2 n + 1) COS
,~=o n=o 2 (17}
xcos ( 2 n + 1 )
2 r {m, n}
wheIe {m, n} is an abbreviation for
{~ (~2/4) [(2 m + 1)2/9 + (2 n + 1) 2 ~0] cosw_t + ~ 2 sinw t[ (17a}
yJ~ + ~ ' / l g [(2 m § 1))~v + (2 n + 1)2 ~o]2 J"

In the latter ~ = a b (o/v, a dimensionless frequency parameter.


If we formally let a -> ~ while keeping b fixed, we obtain the one-dimensional
solution

w($'t)-- K z~
42 2;+i cos(2n+l)~ { 02(~2/2)(2n+
04 zd/16(2n
n~O 2 4 + + 1) 4 '
(18}
where $ = y/b and 0 = b I/o)/2v.
LANDAU and LIFSHITZ[2] gave the periodic solution of oscillating flow between
two parallel plates in complex form as follows

[ cosl/: (i ~o/v) y ] (19)

Numerical calculations have shown t h a t (18) and the real part of (19) are identical.
Evaluation of (17) was performed on an IBM 7090 digital computer and extensive
numerical results have been obtained. The velocity profiles at various instants of a
cycle are shown plotted in Figures 4 and 5 at various $'s for ~v = 1 (a square duct) and
~v = 10 (a narrow rectangular duct) and for $ = 0 (at duct center) and ~ = 0.8 (near
the vertical wall). I n each case, two extreme values of ~v are chosen : ~0 = 0.05 and 50'
for the square duct, and ~ = 0.5 and 500 for the narrow duct. Small values of ~'
correspond to highly viscous fluid oscillating slowly in a small duct, while large values
of ~o represent fast oscillations of low viscosity fluid in a large duct. I t is seen t h a t at
small values of ~o, the velocity is nearly in phase with the exciting force; while at large
values of ~o, there is almost a 90 ~ phase shift between the velocity near the duct axis
and the pressure force. Moreover, for fast oscillation, there is a flattening of the velocity
profile in the core region and the m a x i m u m velocity does not occur on the axis but
near the wall. Similar characteristics have been experimentally observed b y RI-
CHARDSON and TYLER [ 5 ] and theoretically investigated b y SEXL [6] for flow t h r o u g h
circular pipes.
The above phenomena of phase shift for the case of very small and very large
values of ~o can also be deduced from strict mathematical analysis [1].
adol. 16, 1965 Unsteady, Laminar, Incompressible Flow Through Rectangular Ducts 357

, 7t-..... ~-1.o . ~

-o o b~ ~ .oo

-)
t
= 50
o4 o15 7.'o -o.o
3 <---_ 3

0 Q.OI ) y = 0.05
-o.e

/,'0

Figure 4
l =50
=o.8

Velocity distributions of pulsating flow at different instants of a period.

F r o m (13) and (17) the energy dissipation per cycle of motion per unit length of
duct m a y be derived. I t is,
2 ~[~o

0 m~0 n=0 (20)


X
[ ~oz + + ]
(~4/16) [(2 m + 1)2/q9 + (2 n + 1) 3 ~]2 ,
which is a positive definite quantity.

6. A P u r e l y H a r m o n i c M o t i o n S u p e r i m p o s e d on a S t e a d y F l o w
We consider a pressure gradient of the form

f(t) = -- K o- K cos~o t , (21)


358 CHIEN FAN a n d BEI-TsE CHAO ZAMP

J=500
=0.0

~rO.8

~ = 500

_3 , , 5
Figure 5
Velocity distributions of pulsating flow at different instants of a period.

where K 0, K and eo are all positive constants; K 0 = - (l/Q) (Op/Oz) and (OplOz) is
the temporal mean of the pressure gradient, K is the amplitude of the oscillating
pressure gradient.
Since the problem under consideration is linear, the velocity is superposable.
Other quantities which are linearly dependent on velocity, such as mass flow rate,
shear stress, frictional drag, etc., are also superposable. The velocity distribution in
this case is simply given by the sum of Equations (9) and (16).
Under the condition of periodic motion, the amount of energy dissipated during
one cycle per unit length is also found to be the sum of the energy dissipated during
the time 2 ~/a~ for the steady motion and the energy dissipated per cycle of the
harmonic motion. We have,

-- K2 e(~)], (22)
Vol. 16, 1965 Unsteady, Laminar, Incompressible Flow Through Rectangular Ducts 359

where

9 =~ ~ ~ (2 m + 1)~ (2 n + 1)~ + (2 m + 1)~ (2 n + 1 7 ~o~


and, (22a)
~ 2 [ ~o 1 ] I
=, m=O n=O
(2 m + 1) 3 + (2 n + 1) 3 ~ 93 + (~'/16) [(2 m + 1)3/~ + (2 ~ + 1)8 ~]3

m=0 n=O ( 2 m + 1) 4 ( 2 n + 1) 8 + ( 2 m + 1) 3 ( 2 n + 1) 498


(22b)
The function G(~p) is called the coefficient of excess work. I t represents the extra
energy dissipated due to the pulsating motion when K / K o = 1.
An examination of (22b) shows that when the oscillation is extremely slow, i.e.,
--~ 0, G(~) + 1/2; and for extremely fast oscillation, G(~) --~ 0. These results are
identical to the conclusions arrived at b y UCHIDA [71 in his analysis of flow in circular
pipes.
The coefficient of excess work is graphically represented in Figure 6. The dotted
curve is reproduced from Reference [71 ; it corresponds to flow in a pipe with the same
cross-sectional area as the duct, thus ~v = (zr r~/4) (co]v). I t is observed that, for fixed
values of % G(~v) is larger for a square duct than a circular pipe with the same cross-
sectional area; and the narrower the rectangular duct, the more extra energy is
dissipated. On the other hand, whatever the aspect ratio of the conduit m a y be, G(~)
decreases as ~ increases, which indicates that the faster the oscillation, the less extra

05
~
.
energy per cycle is dissipated.
*gloZed againstf ro2~-
§

o.~

~N

o /o 2o ao ~ so 6o 7o 8o so loo},
Figure 6
Coefficient of excess work for a harmonic motion superimposed on a steady flow.

In conclusion, we remark that our results in general, exhibit similar characteristics


as those of the corresponding flow through circular pipes. The latter was first in-
vestigated b y SEXL [6] and then b y UCI-IIDA[7] for periodic motion. Their method of
solution is not applicable to transient problems.
Finally, we note that the results reported herein can be directly used for analogous
problems of heat conduction in rectangular bars with internal heat generation.
360 S.I. PAI and C. K. TsAo ZAMP

REFERENCES
[1] FAN, C., Non-steady, Viscous, Incompressible ]:;low in Cylindrical and Rectangular
Conduits (With Emphasis on Periodically Oscillating Flow) (Ph.D. Thesis, University
of Illinois, 1964).
[2] LANDAU, L. D. and LIFSHITZ, E. M., Fluid Mechanics, Translated by J. B. SYKES and
W. H. REID (Pergamon Press, 1959).
[31 SZYMANSKI,P., Quelques solutions des dquations de l'hydrodynamique de [luides visqueuz
dam le cas f u n tube cylindrique, J. de math. pures et appliqu6es, Series 9, 71, 67 (1932).
[4] MILNE-THoMSON,L. M., Theoretical Hydrodynamics (MacMillan Co., 1960).
[5] RICHARDSON,E. G. and TYLER, E., The Transverse Velocity Gradient Near the Mouth of
Pipes in Which an Alternating or Continuous Flow o/A ir is Established, Proc. Phys. Soc.
London d2, 1 (1929).
[6] SEXL, T., [?ber den yon E. G. Richardson entdeckten ,AnnulareMekt~>, Z. Physik 61, 349
(1930).
[7] UCHIDA, S., The Pulsating Viscous Flow Superimposed on the Steady Laminar Motion
o] Incompressible Fluid in a Circular Pipe, ZAMP 7, 403-422 (1956),

Rdsumd
L'6tude porte sur la th6orie de l'6coulement non permanent d ' u n fluide visqueux
incompressible dans de canaux rectangulaires d'allongement divers, sous l'influence d'un
gradient de pression arbitraire, d6pendant du temps. Des solutions ont ~t6 obtenues dans
4 cas particuliers: 1. gradient de pression impulsif, 2. gradient de pression constant et
6tabli brusquement, 3. gradient de pression en fonction harmonique du temps, 4. gradient
de pressjon ~ une composante constante et une composante harmonique. On donne les
r@artitions de vitesse, les coefficients de frottement et la dissipation d'6nergie par unit6
de longueur.
(Received: September 1, 1964.)

Nozzle Flow of a Fully Ionized Plasma Based


on Two Fluid Theory 1)
By S. I. PAI, Institute for Fluid Dynamics and App1. Mathematics, and C. K. TsAo,
Aeronautical Eng. Dpt., University of Maryland, College Park, Maryland, USA

I. I n t r o d u c t i o n

Most of the i n v e s t i g a t i o n of m a g n e t o - g a s d y n a m i c c h a n n e l flow is based on 'classical'


single fluid t h e o r y in which the generalized OHM'S law is used i n s t e a d of the exact
differential e q u a t i o n of electrical c u r r e n t d e n s i t y a n d the difference of t e m p e r a t u r e s
of ions a n d electrons is ignored. One w a y to i m p r o v e the results of classical m a g n e t o -
g a s d y n a m i c s is to use m u l t i - f l u i d theory in which the effect of various forces - b o t h
g a s d y n a m i c a n d electromagnetic - on the electric c u r r e n t d e n s i t y has been t r e a t e d
exactly from the macroscopic p o i n t of view a n d the behaviors of electrons a n d ions

1) This research was supported in part by the Air Force Office of Scientific Research under the grant
No, AFOSR 141-64. A part of this paper was submitted to University of Maryland as the M. S. thesis of the
Junior author.

You might also like