Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

MECHANICS RESEARCH COMMUNICATIONS Vol. 18 (4), 199-206, 1991. Printed in the USA.

0093-6413/91 $3.00 + .00 Copyright (c) 1991 Pergamon Press plc

FLOW THROUGH P I P E S WITH SUDDEN ENLARGEMENT

R.Srlkanth" and E.Rathakrishnan °"


Department of Aerospace Engineering
Indian Institute of Technology
K&npur, I n d i a

(Received 10 September 1990; accepted for print 27 November 1990)

Introduction
The problem of flow in pipe with abrupt expansion, having
wade practical application, has been a subject of research over a
long period. Many theoretical and experimental investigations
have been reported An l i t e r a t u r e on t h e above problem. Although
attempts have been made t o describe the base pressure as function
of other parameters of the process viz..~ach number, Area ratio -
defined &s ratio of enlargement to i n l e t pipe area. etc. they are
all based on ~ssumptions which severely restrict the application
of the result. Some of the best known theories are by Chapman
et.al~, Korst 4 and Tanner~ However almost all the investigators
have used either a convergent or a c o n v e r g e n t - d i v e r g e n t nozzle to
generate the required Hath number at the entry to the
enlargement. But In practical piping systems one mostly
encounters flow £n s t r a i g h t pipes with abrupt change in area.
Th£s practical aspect has been considered by R&thakrishnan and

m Or<~kA~te Student
o~ Aee~gtGr~t P r o r ~ r

199
200 R. SRIKANTH and E. RATHAKRISHNAN

Sreekanth 6 in t h e i r experimental investigation. In t h e i r study,


a straisht circular pipe has been used as the inlet passase.
They have emphaslsed that it is the overall pressure ratio rather
than the Nach number at the entry to enlargement which is o~
importance in p r a c t i c a l applications. The pressure ratio covered
(stasnatlon to b a c k pressure) in t h e i r InvestiSation varied from
3.0 to 1.1. This corresponds to flow conditions from chokln~ to
low incompressible Mach number at the exit o£ the inlet pipe.
Based on their experimental study they concluded that £o~- a ~ i v e n
pressure carlo and area ratio it is poss4ble to Identi~y an
optimum enlarged pipe len@rh for any flxed inlet pipe re~ultln~
An m a x i m u m suction at the base with minimum pressure loss 4n the

pipe. Osin 6 the above experimental data It is attempted ir~ the

present study to develop an empirical relation £o~ the base


pressure as a £unction o£ p r e s s u r e ratio, area ratio and the
len6th to d i a m e t e r ratio of the enlarged pipe.

No~r.cl slurs

P - Base Pressure
b

P - Ambient Pressure

Po - Stagnation Pressure

d - Inlet Pipe Diameter

d - Enlarged Pipe Diameter

L - Enlarged Pipe Length

D~velopmont of the ompirlcal relation


~ithin the frame 0£ the parameters of the present study, the
~eneral form of behaviour of t h e s t a t i c pressure ratio betweet~
the base and the external atmosphere may be d e s c r i b e d by

P P d L

P
[o P d
. dll .............. (la)
PIPES WITH SUDDEN ENLARGEMENTS 201

L~I: E P d L
- - : ¥ , - - = ~r , ~ = ~, , - - = X
P P d d
..t • 2

then ¥ = f ( x , ~ , - ) .............. C1b)


A mathemat1"c~l model of form (lb), whlch represents
experLmental data, a s~ple of whlch Is shown Ln £ 1 G . 2 and f~6.3,
has ~o be developed. Since ve have a dlstr/butlon of data for the
four va~ables In (Ib), whlch Is nonllnea~, nonllnea~
srarlsrlcal resressLon technLques have ~o b e used. The ewe ~¥pes
(Re£.3) vlz.. Inrrlnslcally linear, nonllnear models (type I) a n d
£ntr~nslcall¥ nonlinear, nonl~near models (~ype 2) have Eo be
fLrted. A model c o r r e s p o n d l n 8 to t y p e i may be expressed as

y = a e (b~ + cx + d.) ............ (Za)


which can b e made linear by t~cln8 natural lo6 on either aide,
whZch @Ires,

in(~) = InCa) + l~bt + CX, + d/z, ............ C2b)

where ~ denotes the data po£nt considered. Applying linea~ least


square technlque 61vex,

y = 0.941 • (0"022~ + 0.055x - 0.Z44-) ........ (Zc)

To study the 8oodness of EL~ ( R e £ . 3 ) o£ s u c h model~ we use


the defined sum of the square Of the errors (SSE), betveen the
data polnt and the model predicted value and the aveca~ error
(At) as

ss~ :E C T-y,)" and AE =


rl

where N Is the total number of data points considered. Uhen


applied on model (2c) It resulted In
202 R. SRIKANTH and E. RATHAKRISHNAN

SSE = 7.1977 ; AE = 16.6~


In c a ~ c u l a t l o n oE the ~ o o d n e s s o£ ilt £actors certlan data
points In t h e range 1.0 < x < 2.0 h a v e been excluded as they are
not In the practical ~n~e. Thls is d o n e on every model that ~s
evaluated in t h i s study. One way to s e a r c h £or better models that
mlnlmlse SSE and AE Is to study plots whlch show as to how
parameters depend on each other separately. Such a study on
experimental data (2e£.6) su88ests the Eollowln8 Eunctlons

¥~ e-x . . . . . . . . . . (3a)

A combination oE (3a), with the parametric constants placed


in linear positions, 81YES

¥ = a e Cb l n C t a n h C ~ ) ) + cx z + d lnC-)) ..... (3b)

Any model Eormed by combinations oE functions oE (3a) should


retain the identity oE t h e Individual proportlonallt~es. This Is
achieved to some extent In C3b) by using natural loa on tanh(.~)
and ~ as they are combined in the exponentlal form. Usina linear
least square technique on (3b) we 8 s t

¥ = 4.414 • C81"3 lnCtamh(@)) - 0.016x ~ - 1.8 ln(~))

For this model, the resultZn8 8oodnesa oE Elt Eactors are


SSE - 81.7532 ~d AEsS6t . Thoush the results are ver¥ poor the
lncreaslni and decreaain8 behavlour with respect to each vaclable
Is obeyed. Howeve~ (2c) dld not obey le..wlth respect to x It
increased when It should have decreased. The reason Eor the poor
perEormance oi (3b) c~n be attributed to the posltlonin8 oi the
pArametrlc constants. Another comblnatlon of C3a) which was trled
Is
PIPES WITH SUDDEN ENLARGEMENTS 203

y = a tanh(b~) + c • dxz + f. + h . . . . . . . . . . . . (4a)


Since it is a model of type(2) an iterative nonlinear least
square technique of the £ollowln8 type was 1:tied. Grouplnkl (4a)
as

G ( ¥~,&,b,~,c,d,x ,f,n~,h ) = 0
where k denotes the iteration ztep, then at the k th s1:ep in the
i1:era1:ion process we have

~G = Ok~ - Gk

= G .As + G~b + G &c + G~d + G ~f + ~,L~h

where parameter subscrLpts denote partial derivatives vlth


respect to the par&meters at the da~a point L. SLnce Gk~* should
L

finally @o t o zero in the least square process we s h a l l set it to


zero (Re/.1) and proceed. This @enerates the matrix equatlon

[-o: ] - [,o.l ]
[mxl]-[mx6] [6xl]
The error matrix ls e v a l u a t e d in e a c h Itematlon, after which
each parameter Is corrected by At - zo that it can be substituted
in PDH, the partial derivative matrix and in the Gk matrix to
carry out the next iteration and so on - u n t l l the error matrix
ls sufficiently zmall when It c o n v e r 6 e z . However, conver@ence can
be ensured only for suitable In4tlal @uezs values foc the
parameters. The model (4a) Is so d e l i c a t e that it dlveDsed in
the very second step. To art closer Eo a good guess another
technique was trled. At first (4a) was expanded ~n the power
serlee form and then approximated to the ilcsr term. Since it is
llneac in t h i s form, linear leazt square technlque was applled to
@st the valuer of the parametrlc constants. There valuer were
then tried OUt as the initial guess fOr the nonlinear lease
square algorithm. It improved tO the extent that divergence was
204 R. SRIKANTH and E. RATHAKRISHNAN

seen only after sixty steps. Howeve~ owin@ ~o severe dlvecsenee


problem in Ehls complicated four variable, six parameter system,
nonlinear leas~ squares was abandoned a~ter it was ~ound that the
power serles expanslon to h i ~ h e £ terms g=ve excellent results.
Expanslon tO £1~sr two si~n1~Icant terms ~ave ~he £ollowln~
model.

Y = a~ + a~
= = * b,x = * bF
= * ~- + d (5)

Applyln~ linear least square technique 8ave ~he following


results

• 5 0 " 3 3 1 5 ' == = - O . O O Z S , b = - O . O 1 2 . ~ = O . O O O Z 3 , ~ 5 - O . 0 7 5 6 , d =-0.016

The 8oodness ol flt result were SSE = 5.2895 and AE ; 14.26%


Uhen the app~oxlmatlon was up to three te~ms the results
Improved drastleall¥.

¥ = =., + ==,'+ %,', b2. ÷ d .... (6>

After applyln~ linear least technique the parameters were


found to have the followln8 values

a = 0.589 , a 5-0.0178 , a = 0.000169

b = -0.0411 , ~5 0.00192 , ~5 -0.000027

c = -0,174 , d =-0.0000433

SSE = 0.8717 • AE = 5.8%

Cot== 1 u s i o n
From the practical viewpoint the results obtained for model
(6) can be considered mood e n o u s h . Purther It can be proved that
hlsher approximations ot (4a), Improves the 8oodness of fit
resultln 8 In hlsher accuracy. However the equations become
cumbersome with. l&rse number of terms. F18(5 ) shows a comparison
be=ween the sample data and the model (6) predicted values. It
can be seen from the error plot (f18.6) that there Is some
oscillation In the error In the ranae 2.0 < x < 6.0. This can
PIPES WITH SUDDEN ENLARGEMENTS 205

be e x p e c t e d out of polynom£al flts llke (5) and (6). It can also


be s h o w n that as we take hi@her term approximations while
expandln@ (4a) the osclllatlons damp out. A l s o r l t h m s (Ref.l,5) in
l~terature w~th suitable mod~£~cat&ons ~o su~t a sudden
enlargement ~low problem are be~n~ trled now.

Ref oro~o~

"Barb°m, R.H. and Dr°he, U., "An alaorlthm for least square
estlmat~on of nonl~near parameters when some are linear,"
Technometrlcs 14, 1972, pp. 757-766.

2Chapman,D.R., U1mbrow,U.R. and Kester,R.H., "Experimental


I n v e s t 1 6 a t l o n s o f base pressures on b l u n t - trallin G edse wlnss
at s u p e r s o n l c velocltles," NACA Technlcal Note, 2611, Jan. 1952.

aD~ape~, N.R. and H.Smith.3r, AppZ~ Re~r~eee~o~ Analys~e,1966,


Ulley,New York.

4Korst, 'A theory for Base Pressures In Transonlc and Supersonic


Flow," Journal o~ applled Hechanlcs, Vol.23, No.4, Dec. 1956, pp.
593-600.

5 M a r q u a r d t , D.U., "An also for least square estimation of


n o n l i n e a r parameters," SIAM J.Appl.Math., 1963, 11, pp. 431-441.

~ R a t h a k r £ s h n a n , E. and Sreekanth, A.K., "Flow in Pipes with


S u d d e n Enlaraement," Proc. of 14 tn Int. Symp. on Space Technology
and Sclence, 1984, pp. 491-496.
?
Tanner,N., "Steady Base Plows," Pro~. Aerospace Science, Vol.
21, 1984.

1.00

0.80 0
o O 0
Pressure +
÷
CL 0.60 t
• " ot

n t t *
I I vol,, I w°, stotic Q_ 0.40
Bese Pressure .
Gate Tcppmgs Toppmgs (d2/dl} 2 = 178 ~ f ~
0.20
t = 2.6
• = 3.0
High Pressure Air 0.00
1.00 2.00 3.00 4.00 5.00 6.00 7.00
L2/d2
~'IG.I 5CHE~,~AT:C ~I;~..CFL~ C~~- ~'HE EXPERIMENTAL
ss TL~ (Ret :- :' Fig.2 Experimentol data poinls (somple)
206 R. SRIKANTH and E. RATHAKRISHNAN

1.00

o o o
0.80 , o
t
=
= •

0.60 ~k . • • O
Q_
\
.O r. ' I, ,~o~o,,.
,°°9 dos,od :~ ,~o~$2o
o.. 0.40 n

(¢~dl) ] - 4.1S
0.20 • =22
• -26
• = 5.0
0.00
,.oo '~.~o '3.oo '4.6o ' 5.60 '6.6o 7.oo
L2/c12
r;g.3 Experimentol doto points (somple)
Fig.4 Trends in somple doto

1.0
0.9 ,"- ,,,
0,8
0.7
00.6
EL
~'~-0.5
EL 0..4
(d2/dl) 2 = 4..19
0.3
doshed : eq.6
0.2 Po/Pa • = 2.0 -93,
solid : eq.6
0.1 Po/Po o = 3.0 i
0.0 l' I l I I ~ I i I i -15 ,~-'-~----i - , , , ] , i ,
.0 2.0 3.0 4.0 5.0 6.0 7.0 2 3 4 5 6
L2/d2 L2/d2
Fig.5 Somple experimental doto and fit[ed curve Fig.6 Error plot for fig.5

You might also like