Fully Developed Viscous Flow in Coiled

Circular Pipes
The Navier-Stokes equations in stream-function/vorticity form were
solved numerically by over-relaxation for the case of steady state, fully de-
veloped, isothermal, incompressible viscous Newtonian flow within a rigor-
ously treated toroidal geometry. Solutions were obtained for curvature ratios LARRY R. AUSTIN
ranging from 5 to 100 and for Dean numbers as low as 1 and as high as and J.D. SEADER
1000. The Dean number was demonstrated to be the principal parameter to
characterize toroidal flow; however, a second-order dependence upon the Department of Chemicul Engineering
curvature ratio above that expressed in the Dean number was observed. The University of Utah
Comparisons of the numerically computed axial-velocity profiles were made Salt Lake City, Utah 841 12
with experimental data. The cross-sectional pressure distribution was cal-
culated, and a correlation is presented for a diametral pressure drop in terms
of the Dean number.

SCOPE
Straight circular tubes have found widespread use in as compared with simple straight-tube parabolic flow,
operations involving fluid dynamics, heat transfer, mass causes a higher axial-pressure gradient, a higher critical
transfer, and chemical-reaction kinetics. For laminar flow, Reynolds number for transition to turbulent Row, a di-
a large number of analytical investigations and supporting ametral-pressure gradient, a fluid-element residence-time
experimental studies have been reported that permit, in distribution that more closely approximates plug flow, rela-
many cases, quite accurate predictions of transport phe- tively high average heat-transfer and mass-transfer rates
nomena, including transport with simultaneous chemical per unit axial pressure drop, especially for high-handtl-
reaction. number and high-Schmidt-number fluids, and significant
Curved configurations of circular tubes-such as partial peripheral distributions of the transport rates. The latter
coils, single coils, helical coils, and spiral coils-have re- effect can be utilized to advantage in applications where
ceived far less attention in the literature despite their fre- the peripheral boundary conditions are asymmetrical.
quent use in heat exchangers, chemical reactors, rocket en- Of fundamental interest to the development of a com-
gines, and other apparatus, equipment, or devices. In some plete understanding of viscous-flow phenomena in curved
cases, the use of curved tubes is necessitated because of or coiled tubes is the nature of the velocity and pressure
geometrical restrictions. However, it is also becoming in- distributions in the fully-developed flow region. In studies
creasingly apparent that the nature of the complex primary by previous investigators, these profiles have been found
(axial direction) and secondary (normal to primary) flow to depend strongly on the Dean number N D , where N D , =
patterns in curved tubes makes possible some definite ad- NRe(R/Rc)'h,R is the pipe inside radius, and Rc is the coil
vantages of this configuration over straight tubes for a radius of curvature. At low Dean numbers, open-form
number of situations. In fully developed curved-tube vis-
analytical solutions have been achieved; numerical meth-
cous flow, the primary-flow-velocity profile is distorted
ods have been successful to moderate Dean numbers of
from its parabolic straight-tube-flow counterpart, a sec-
ondary flow is established that consists of two vortices, and approximately 300; and an approximate combined bound-
the resultant combined primary and secondary flow pat- ary-layer/potential-flow core theory has been applied at
terns cause a fluid element to have a screw-like motion. high Dean numbers. The objective of this study was to
At one instant, a fluid element may be traveling near the develop a rapidly converging and reasonably accurate nu-
very center of the tube cross section. After a short period merical solution to the equations of fluid motion so that
of time and a short axial distance downstream, the same velocity and pressure profiles could be computed over a
fluid element may be found very near the outside wall of much wider range of Dean numbers and curvature ratios
the tube. The nature of curved-tube viscous-fluid motion, R J R than previously.

CONCLUSIONS AND SIGNIFICANCE

Solutions of the equations of fluid motion were achieved was not investigated. The computations covered a Dean-
for the case of steady state, fully-developed, isothermal, number range of 1 to 1000, a Reynolds-number range of
incompressible, viscous Newtonian flow within a toroidal- 10 to 4000, and a curvature-ratio range of 5 to 100. The
type coiled-tube geometry. Thus, the effect of coil pitch numerical mathematical technique that was employed in-
volved the expression of the equations of motion in terms
of the axial-velocity component, the stream function of
Correspondence concerning this paper should be addressed to J. D. the secondary flow,'and tge axial componentof the vofii~-
Seader. L. R. Austin is now with E. I. du Pant de Nemours and Com-
pany, Chattanooga, Tennessee. ity. The resulting partial-differential equations were ex-

AlChE Journal (Vol. 19, No. 1) January, 1973 Page 85

pressed as finite-difference equations by the use of first Pressure distributions for the same locations in the sec-
central-difference operators and then solved by a sequen- ondary-flow plane as those in Figure 6 are shown in Figure
tial application of a successive over-relaxation technique. 8. In general, the pressure distribution is dependent
Several methods were used to check convergence. mainly on the Dean number; and the pressure increases
As shown in Figure 4, the low-Dean number secondary- smoothly from the inside to the outside of the pipe with
flow streamlines are essentially symmetrical and in agree- respect to the coil axis. The computed dimensionless
ment with the previous analytical solution of Dean, the diametral-pressure-drop data are correlated in Figure 9
pioneering theoretician for curved-tube viscous flow. The and are in reasonably close proximity to the plot of Lans-
secondary flow is directed outward (with respect to the coil fords equation (1935), which was determined under the
axis) along a path near the pipe diameter and then back assumption of an ideal, uniform, primary-flow velocity
along a path near the pipe wall. As the Dean number is distribution. As shown in Table 2, at moderate-to-high
increased, causing an intensification of the secondary flow, Dean numbers, the dimensionless average diametral-pres-
nonconvex regions are developed by the formation of sure gradient in the secondary-flow plane is greater than
dimples in the streamlines of the secondary flow. These the negative axial-pressure gradient.
streamlines depart from their symmetrical nature in the The axial-pressure gradient for curved-tube flow is gen-
half plane, and the vortex center shifts toward the wall erally correlated in terms of a friction factor that is nor-
and toward the axis of the coil. As shown in Figure 4, at malized with respect to the straight-tube-flow friction
a Dean number of approximately 400, little difference can factor for the same Reynolds number. The computed re-
be distinguished in the secondary-flow streamlines over a sults shown in Figure 10 are well grouped around Whites
wide range of curvature ratio. Thus, the effect of the experimental correlation and exhibit a slight dependence
curvature ratio is accounted for mainly in the Dean num- on the curvature ratio, as predicted previously by
ber. The secondary-flow streamlines for the approximate Topakoglu (1967), at low Dean numbers.
combined boundary-layer/potential-flow core theory at a The results presented here or calculated at other con-
Dean number of 400, as shown in Figure 5, differ signifi- ditions by the computer program cited can be utilized to
cantly from the numerical solution. compute pressure drops, residence-time distributions, and
At low Dean numbers where the secondary flow is heat- and/or mass-transfer rates (both axial and peripheral
small, the axial-velocity profile is essentially parabolic and variations). In an extension of the work presented here,
unaltered from straight-tube laminar flow. As shown in Kalb and Seader (1972) have utilized the velocity profiles
Figure 6 for a Dean number of approximately 400, the
to compute heat transfer to steady viscous flow in a toroid
maximum velocity at the diameter of full-circle symmetry
and that for four adjacent parallel chords spanning the for fully developed velocity and temperature fields under
half-circle cross section are skewed. For the same condi- the thermal boundary condition of axially uniform wall
tions of Dean number and curvature ratio, Figure 7 shows heat flux with peripherally uniform wall temperature for
that the central portion of the velocity surface is somewhat Dean numbers from 1 to 1,200 and Prandtl numbers from
dished in the direction parallel to the axis of the coil. This 0.005 to 1,600. Order-of-magnitude increases in the aver-
dished effect becomes slightly more pronounced at higher age heat-transfer coefficient for curved tube flow over that
Dean numbers. The numerically computed profiles were for straight-tube flow are predicted, particularly at high
in excellent agreement with experimental data. Dean numbers and for high Prandtl number fluids.

Fluid flow within helically coiled tubes is common in White ( 1929). His law has been verified many times since.
many industrial operations, particularly those involving He introduced a dimensionless parameter, now known as
heat transfer. Toroidal flow is the limiting case of helical the Dean number, which has proved to be the most im-
flow with zero pitch. The mode of fluid flow in coils is portant single parameter with which helical flow can be
characterized by a secondary flow field, which is superim- characterized. Dye-injection experiments, performed first
posed upon the axial-velocity flow field. The secondary by Eustice (1911), and later by Taylor (1929), were
flow field divides itself along the diameter of the toroid fundamental in demonstrating the secondary flow, which is
into two mirror images. Flow is directed outward along the characteristic of all helical flow. Taylors experiments also
diameter, which then returns to the center by following showed that, depending on the ratio of the radius of curva-
the tube wall. This secondary-flow action causes the diame- ture to the tube radius (curvature ratio), the critical Reyn-
tral axial-velocity profile to be skewed toward the outside olds number was increased to values beyond the value of
of the coil. Not only are velocity profiles altered as com- 2,100 usually accepted for straight-tube flow. For example,
pared to straight-pipe flow, but also the axial-pressure drop at a curvature ratio of 18.7, a critical Reynolds number
is increased and a diametral pressure difference is estab- of 5830 was observed.
lished. Fundamental information on velocity and pressure Several theoretical attempts have been made by Dean
profiles is useful in establishing heat- and mass-transfer ( 1927, 1928), Topakoglu ( 1967), McConalogue and
rates and homogeneous chemical-kinetic phenomena for Srivastava (1968), Truesdell and Adler (1970), Larrain
flow in curved tubes. and Bonilla (1970), and Akiyama and Cheng (1971) to
Active experimental work with helical-flow systems solve the equations of fluid motion for the case of toroidal
dates back to the beginning of the twentieth century. Con- flow. The first classical work is attributed to Dean (1927,
siderable experimental work has been done by Grindley 1928). His original solution was valid for creeping flow
(1908), Eustice (1910), White (1929), Keulegan and only ( N D , < 1) and failed to show any increased flow
Beij (1937), %ban and McLaughlin (1963), Kubair and resistance over straight-pipe flow but did predict the form
Kuloor ( 1963), Schmidt ( 1967), Ito ( 1969), and Larrain of the secondary-flow field, as previously observed by
and Bonilla ( 1970) to establish a pressure-drop relation- Eustice ( 1911) in his dye-injection experiments. Truesdell
ship. The first reliable resistance law was determined by and Adler (1970) produced a numerical solution that was

Page 86 January, 1973 AlChE Journal (Vol. 19, No. 1)

completely rigorous for a toroidal system; however, their of the coil tube. The toroidal coordinate system is shown
method of solution proved to be unstable at Dean numbers in Figure 1 where the position variables are $, 8, and T. A
much larger than 200. The work described here is an out- radius-of-curvature parameter R, is also required to com-
growth of the work by Truesdell and Adler. Basic differ- plete the geometric specification. Here, both T and R, are
ences, as discussed in detail by Austin (1971), are: understood to be normalized with respect to the character-
1. A toroidal coordinate system is used here, while istic length R which was chosen to be the tube radius.
Truesdell and Adler employed a rectangular coordinate Equation ( 6 ) was applied to this toroidal coordinate sys-
system. tem by means of dimensionless scale factors, which are
2. The solution algorithm in this work is based on the readily shown by Austin (1971) to be
vorticity field, while Truesdell and Adler solved a fourth-
order partial differential equation in terms of the stream & = R, + T sin e (7)
function.
Akiyama and Cheng (1971) also utilized the vorticity and
concept but restricted their solution to very large ciirvature
ratios and could not obtain solutions beyond a Dean num-
ber of 300. Fully developed flow was assumed.
By expanding Equation ( 6 ) into its three components,
then differentiating the T and 6 components with regard
THE EQUATIONS OF FLUID MOTION to 0 and r, respectively, and combining the resulting ex-
A vector representation of the steady state, constant- pressions, the unknown pressure is eliminated. The un-
property Navier-Stokes equation can be expressed ( Bird knowns W, and we are eliminated by introducing V@through
et al., 1960) as its relationship to the definition of vorticity in Equation
1 (5)
MV(V*U)-VX (VxU)=--vVp+vV2V, (1)
P
This equation may be rendered dimensionless by introduc- and
ing the following dimensionless variables, where R is a
characteristic length:
"
Rv
V=-
V The W# component of Equation ( 5 ) is retained.
The unknowns V, and Ve may be eliminated by intro-
p = - R2 P (3) ducing the definition of a stream function for the secondary
PG flow in the cross section such that the equation of con-
tinuity for the fully developed toroidal flow is satisfied.
and by replacing the customary dimensional-vector differ- These defining equations are
ential operator V with a dimensionless form RV. The re-
sulting dimensionless equivalent of Equation ( 1 ) is 1 as
vr -- - - -
%V(V .V) - v x ( V x V) = - VP + V V . ( 4 ) r h@ ae
It is convenient to introduce the vorticity w as discussed and
by Brodkey (1967), into Equation ( 4 ) , where
o=vxv. (5)
Accordingly, an alternate form of Equation ( 4 ) is The final result is three elliptic second-order partial dif-
ferential equations in three dependent variables-@, Va,
Y2V ( V * V) - V x w = - VP - V X W. (6) and S-and two parameters-R, and aP/a$-namely:
Equation (6) was solved numerically in terms of a 1. The defining equation for the axial component of the
toroidal geometry, that is, with the assumption that the vorticity
coil has no pitch. The pitch effect might become important
as the radius of curvature for the coil approaches the radius

t' /
2. The axial component of the Navier-Stokes equation

where

Fig. 1. The toroidal geometry.

AlChE Journal (Vol. 19, No. 1) January, 1973 Page 87

and The boundary conditions were deduced in a logical man-
ner from the equations of motion by Austin (1971), based
on two main assumptions:
1. That the no-slip condition exists at solid boundaries,
3. The combined equations for the T and 0 components that is, the velocity at the wall is zero, and
of the Navier-Stokes equation 2. That a symmetrical condition exists across the centrnl
diameter.
ao,
ao, + C - a%, 1 aZo+
A, q,+ B - - -- --= D, The resulting boundary conditions for each of the four
ae ar arz r2 ad2 boundary sections are summarized in Table 1.
(19)
where FINITE-DIFFERENCE EQUATIONS
Each of the governing relationships given by Equations
(14)) (15), and (19) were expressed as finite-difference
and equations by use of first central-difference operators. As
described by Austin (1971)) special one-sided operators
after the method of Vrentas et at. (1966) were used at the
Although Equations (15) and (19) are nonlinear, the pipe wall to improve solution stability. Corresponding
iterative-solution procedure described below involved lin- terms were then combined such that each equation could
earized forms of these equations, wherein the dependent- be expressed as coefficients multiplied by the dependent
variable terms in the defining relationships for A,, A,, 3, variable at the mesh point in question and its four nearest
C , and D, were based on the solution of previously calcu- neighbors. These equations, in the computational molecule
lated values. form of Lapidus (1962), are

BOUNDARY CONDITIONS 1. The axial component of the vorticity

Because of the symmetry of the flow field, only one half
of the cross-sectional region required consideration. The 1 21
11
regional boundary was divided into four distinct sections
which are shown in Figure 2 where I and J are radial and
peripheral grid-point locations, respectively, that were used
I1 zz
Z3 2,

1
S =r2(Ar)hao, (22)

in the numerical calculations described later. These sec- where

sine 1
tions are
1. The curved portion of the outer wall ( I = 1, 1 6
4 21)
21 = T2 [-g+ y- 2h,
21)
2. The inside half of the diameter ( 1 4 1 20, J = 1 )
3. The outside half of the diameter ( 1 I

4. The singular point ( I = 21, 1 J A 21).

20, J = z2=- AT
ae rose
-+'I 2h, Ae

J=II

and
25 = TZ [ --+2~ -+ -
2ha AT

2. The axial component of the Navier-Stokes equation

c x1 1

Inner
wal I I Outer where
c x5 1
waI I
r2C P
Fig. 2. The computational grid region.
xl=---
2Ar AT)^

FOR THE THREE

TABLE1. BOUNDARYCONDITIONS VARIABLES

Variable Section 1 Section 2 Section 3 Section 4

and

Page 88 January, 1973 AlChE Journal (Vol. 19, No. 1)

 3. The combined r and 0 components of the Navier- values were set equal to the final results obtained from
Stokes equation computations for the next lowest Dean number. The first
case, for a low axial-pressure gradient corresponding to a
r 1 Dean number of approximately one, was initiated under
the assumption of straight-pipe values; that is, S = W* = 0
and V* based on a parabolic profile.
As shown in Figure 3, the method of solution was, first,
to iterate Equation (24) one time, updating the vorticity
where at each interior point, then to solve Equation (22) for the
Yk = Xk stream function. Normally, this equation was iterated a
k = 1,2,4,5, number of times (5 to 15) to obtain a rather smooth solu-
and tion surface, which greatly facilitated computational sta-
bility since the boundary condition for the vorticity de-
pends upon the derivative of the stream function. Next,
Equation (23) was solved for the axial component of the
The above three finite-diff erence equations were solved velocity; this nonlinear equation was also iterated to ac-
by a standard successive over-relaxation technique, as de- celerate overall convergence. These three steps constituted
scribed by Greenspan (1965). The order in which the one complete major iteration.
equations were solved was critical for stable convergence Generally, a number of major iterations were performed
and is discussed next. in order to obtain a converged solution. As discussed in
more detail below, satisfactory convergence of the solution
COMPUTATIONAL ALGORITHM
was tested by comparing successive results as the spatial
grid size was reduced, comparing friction factors computed
A flow diagram of the major calculational steps involved for a given run by two alternate expressions, and compar-
in solving the governing equations appears as Figure 3. ing the numerical results with experimental data.
The starting data included boundary conditions as well as The spatial-convergence check was performed by ob-
the assumed initial values of the three dependent variables taining solutions with an 11 x 11 grid-point mesh and a
at all other grid points. In general, in computations for a 21 x 21 grid-point mesh for Dean numbers as high as 400.
particular curvature ratio and axial-pressure gradient cor- Agreement between the two results was excellent. As a
responding to a certain Dean number, the assumed initial further check on spatial convergence, dimensionless sec-
ondary velocities at the singular point I = 21 were com-
puted for all values of the peripheral grid-point location J .
START Excellent agreement among the values was obtained. For
example, at a Dean number of 603 and a curvature ratio
4 READ s t a r t i n g data
of 9.06, the 21 values covered a very narrow range of
only 21.88 to 21.90.
In practice, it was found possible to obtain a satisfactor-
I+
I ily converged solution by terminating a run when suc-
Prepare f o r i t e r a t i o n

-L
I cessive values of the three dependent variables obtained
during major iterations changed by less than 0.001%.
t This normally required on the order of 100 major itera-
Boundary c o n d i t i o n f o r
tions. Convergence was aided by proper selection of re-
laxation factors that were assigned in the manner of
One i t e r a t i o n on
Greenspan (1965). The best relaxation factor for the
stream function was one. This is in agreement with Aki-
yama and Cheng (1q71). Factors for the vorticity and
I t e r a t e on S once
axial velocity were required to decrease from one with in-
creasing Dean number. Once a completely converged solu-
tion was obtained, values of the Reynolds number, the
for error Dean number, and the friction factor were computed.

RESULTS AND DISCUSSION

Update non-linear coeff'c'ents As discussed above, the three variables actually obtained
from the previously described numerical computations
I t e r a t e on V, once
were the axial component of the vorticity, the stream func-
tion, and the axial component of the velocity. Primitive
variables were computed numerically as follows: The
Test radial and angular components of the secondary velocity
were obtained by solution of Equations (12) and ( 13),
respectively, with fourth-order difference operators. The

.
Check Converqence pressure distribution for the cross section was calculated
by integration of the radial component of Equation ( 6 ) ,
starting with a reference value of zero for the pressure at

0 A listing of the Fortran computer program has been deposited as

Document No. 01994 with the National Auxiliary Publications Service
(NAPS), c/o Microfiche Publications, 305 East 46 Street, New York
10017 and may be obtained for $2.00 for microfiche or $5.00 for
Fig. 3. Flow chart for computer program. photocopies.

AlChE Journal (Vol. 19, No. 1) January, 1973 Page 89

 FOR NUMERICAL CALCULATIONS
TABLE2. RUNCONDITIONS

Run apm fcM -A P D / A P ~

1 -4,000 100.0 9.999 0.999 1.om 0.0105

2 -20,000 100.0 49.98 4.998 1.001 0.053
3 -50,000 100.0 123.9 12.39 1.009 0.1295
4 -90,000 100.0 213.6 21.36 1.053 0.215
5 -600,000 100.0 1,003.0 100.3 1.495 0.595
6 - 1,020,000 100.0 1,499.0 149.9 1.701 0.765
7 - 1,510,000 100.0 2,001.0 200.1 1.886 0.91
8 -2,650,000 100.0 2,997.0 299.7 2.210 1.14
9 -4,000,000 100.0 4,018.0 401.8 2.489 1.325
10 -421,000 29.1 1,620.0 300.4 2.232 2.065
11 -570,000 29.1 2,014.0 373.4 2.431 -
12 - 148,000 14.4 1,138.0 300.0 2.257 2.835
13 -75,000 9.06 904.4 300.5 2.288 3.45
14 -201,000 9.06 1,816.0 603.3 3.054 5.425
15 -309,000 9.06 2,437.0 809.6 3.499 5.765
16 -321,000 9.06 2,500.0 830.6 3.543 -
17 -430,000 9.06 3,034.0 1,008.0 3.911 6.5
18 -50,300 6.94 786.0 298.4 2.305 3.825
19 -200 5.0 9.980 4.463 1.002 0.219
20 -450 5.0 22.17 9.917 1.015 0.465
21 -950 5.0 44.19 19.76 1.075 0.845
22 -18,000 5.0 448.7 200.7 2.006 3.50
23 -31,500 5.0 670.6 299.9 2.349 4.33
24 -47,400 5.0 897.3 401.3 2.641 5.05
25' - 1,910,000 40.0 3.956.7 625.6 3.017 -
Close to the conditions for the experimental velocity profile reported by Mori and Nakayama (1965) in Figure 14 of their article.

the center point. Center of circulation

A total of 25 computational cases were run on a Univac
1108 digital computer. The run conditions and computed NDe =
parameters are listed in Table 2. Curvature ratios were
varied from 5 to 100, and specified dimensionless axial-
Rc = 1
pressure gradients resulted in a Dean-number range of
approximately 1 to 1000 and a Reynolds-number range of
approximately 10 to 4000. Runs at conditions other than
those listed in Table 2, for Dean numbers as high as 1000,
can be made rapidly with the aforementioned computer
program.* Although runs for Dean numbers greater than
500 are shown for a curvature ratio of 9.06 only, calcula-
tions at other curvature ratios can be made with the com-
puter program.
Velocity Profiles
The velocity distribution in the secondary-flow plane is
best displayed by means of the stream function. Lines of
constant stream-function values represent the projected
secondary-motion streamlines that are superimposed on the
axial-velocity flow field. Typical contours of this function
are shown in Figure 4 for several conditions. At a Dean
number of 1.0 and a curvature ratio of 100, the streamline
curves bound convex regions that are essentially symmetri-
cal about a radius at 0 = 0" in the half circle. The center
of the secondary vortex lies somewhat closer to the diame-
ter of full-circle symmetry than to the upper wall. The
low-Dean-number streamlines are in agreement with the
analytical solution obtained by Dean (1927). However,
as the Dean number is increased causing an intensification
of the secondary flow, nonconvex regions are developed
by the formation of dimples in the streamlines, which also
depart from their symmetrical nature in the half plane. In
addition, the vortex center shifts toward the upper wall
Inner
I-\ Outer
and toward the axis of the toroid. Such alterations in the
nature of the streamlines are noticeable above Dean num- wall wa II
bers of approximately 100, as observed by Truesdell and Fig. 4. Computed stream-function contours.

Page 90 January, 1973 AlChE Journal (Vol. 19, No. 1)

 ND, = 400 to be due to a combination of factors, including experi-
Center of circulation mental error, a pipe-coil cross section that was slightly
R, >>I flattened during bending to form the coil, and the coil
pitch.
A comparison was also made with the experimental
axial-velocity-profile data of Mori and Nakayama ( 1965)
for a Reynolds number of 4000, a curvature ratio of 40,
and a corresponding Dean number of 632.5. This is in-
cluded as Run 25 in Table 2. The agreement was excellent
for the velocity profile in the horizontal plane. For the
vertical plane, the data of Mori and Nakayama scattered
somewhat. Our numerical results for the vertical plane
were within their data scatter.
Pressure Distribution in the Secondary-flow Plane
Inner Outer
wal I wal I A typical plot of the dimensionless pressure [see Equa-
tion ( 3 ) ] distribution in the secondary-flow plane for the
Fig. 5. Stream-function contour for boundary-layer, potential-flow diameter of full-circle symmetry, as well as for four ad-
model.

Adler (1970). Truesdell (1963) shows the secondary-flow

pattern for a Dean number of 122.9 and a curvature ratio 1 ,N, = 401.3
of 100. Secondary-flow patterns similar to those of Figure
4 are shown by Akiyama and Cheng (1971) for very large
curvature ratios. Figure 4 includes the results of our com-
putations for a Dean number of ,401.8 and a curvature ratio 0
9
of 100. If this Dean number is held essentially constant 0

while the curvature ratio is reduced drastically to a value

of 5.0, the nature of the secondary flow is altered only
slightly, as can be observed in Figure 4. Thus the effect
7
of the curvature ratio is accounted for mainly in the Dean
number, but it does have an additional, although slight,
effect on the secondary flow.
Results of the application of the approximate boundary-
layer, potential-flow model of Mori and Nakayama (1965)
to the calculation of streamlines for a Dean number of 400
and a large curvature ratio are shown in Figure 5. Their
assumption of uniform secondary flow outward in the cen-
tral-core region and their calculation of the location of the
center of the secondary-flow circulation are seen to differ
significantly from the results of the numerical solution,
shown in Figure 4 for the same Dean number. This differ-
ence was very marked below a Dean number of 200.
At a Dean number of 1.0, the axial-velocity profile for
R, = 100 is essentially parabolic and unaltered from fully
developed, straight-pipe laminar flow. However, at Dean 9
numbers greater than approximately 10, the maximum
velocity at the diameter of full-circle symmetry tends to O 1.0 - .5 5 .o .5 1.0
shift outward, away from the axis of the toroid.* Dimen- RAOILIS
sionless axial-velocity [see Equation ( 2 )] profiles for the Fig. 6. Axial-velocity profiles.
diameter of full-circle symmetry, as well as for four ad-
jacent parallel chords spanning the half-circle cross sec-
tion, are shown in Figure 6 for a Dean number of 401.3
and a curvature ratio of 5.0. The entire axial-velocity sur-
,N
, = 401.3
face for the same conditions is depicted in Figure 7. The
central portion of the velocity surface is somewhat dished
in the direction parallel to the axis of the toroid. This
dished effect becomes slightly more pronounced at higher
Dean numbers.
Comparisons of the numerically computed axial-velocity
profiles were made with experimental data obtained by
Austin (1971) with air flow at Dean numbers of 372 and
832 for curvature ratios of 29.1 and 9.06, respectively. The
agreement was excellent, with the small deviations believed

*Lorrain and Bonilla (1970) show that, for very low Dean numbers
and low curvature ratios, the maximum velocity can occur slightly closer
to the inner wall of the pipe. This effect was observed also in the nu-
merical results of this study. Fig. 7. Axial-velocity surface.

AlChE Journal (Vol. 19, No. 1) January, 1973 Page 91

jacent parallel chords spanning the half-circle cross sec- curved pipe to that for a straight pipe is equal to the in-
tion, is shown in Figure 8 for a Dean number of 401.3 and verse of the ratio of the volumetric flux for a curved pipe
a curvature ratio of 5.0. In general, the pressure increases to that for a straight pipe. The latter is given by the well-
smoothly from the inside to the outside of the toroid with known Hagen-Poiseuille relationship. The volumetric flux
respect to the axis of the toroid. The secondary pressure for the toroid was calculated by integration of the com-
distribution was dependent mainly on the Dean number puted axial-velocity profile over the flow cross section. The
and only slightly on the curvature ratio as an additional results are included in Table 2 and plotted in Figures 10
independent parameter. Of particular interest is the di- and 11. In Figure 10, the values are well grouped around
mensionless diametral pressure difference,* which is cor- Whites experimental correlation ( 1929), and a slight
related empirically with the Dean number, as shown in
Figure 9. An approximate solution for the diametral pres-
6
sure drop was obtained by Lansford (1935) by making a
force balance with the assumption of an ideal, uniform,
primary-flow velocity distribution. The dimensionless form ?ol
of Lansfords resulting equation is A P D = ( N D e ) / 2 . As
shown in Figure 9, this equation is in reasonably good
agreement with the computed points. The experimental
data of Kubair and Kuloor (1963) and Tsuji and Kawa-
shima as shown by Ito (1969) were found to be in excel-
lent agreement with the empirical correlation of Figure 9.
The ratio of the computed average diametral-pressure
gradient to the negative axial-pressure gradient -AP,/AP,
depends upon both the Dean number and the curvature
ratio. Computed values, included in Table 2, ranged from
0.0105 to 6.5. At high Dean numbers and small curvature Ap, =
ratios, the diametral-pressure gradient can be quite sig-
nificant.
Friction Factor
If the friction factor for curved-tube flow is defined in
a manner similar to that for straight-pipe flow, it can be
readily shown that the ratio of the friction factor for a

LANSFORD ( 1935

1
10
1 10 100 1000
ND!
Fig. 9. Correlation of diametral-pressure difference.

-This work.
Expel

3-

fc fs

2-

Fig. 8. Pressure distribution in secondary-flow plane. .

30
* The dimensionless form of the diametral pressure drop is based on
the tube inside diameter. Therefore, APD may also be referred t o as an
average diametral pressure gradient.

Page 92 January, 1973 AlChE Journal (Vol. 19, No. 1 )

 1.41 I I I I I I I R, = radius of curvature
RC S = stream function, defined by Equations ( 1 2 ) and
a 5.0, 100.0 This work.
White (1929) (13)
-
- - - - - - - Topakoglu ( 1967) -u = velocity vector
I I u = average velocity
1.2 , I
V = dimensionless velocity, defined by Equation (2)
Greek Letters
AP, = dimensionless average diametral-pressure gradient
in the secondary-flow plane
-APD/AP@ = ratio of diametral- to axial-pressure gradients
AP@ = dimensionless axial-pressure gradient = (aP/
0.91 5 I
10
I
15
I
20
I
25
I
30 35
19
a$) /&
= angular coordinate in secondary-flow plane in
ND.3 Figure 1
Fig. 11. Effect of curvature ratio on friction in curved tubes. p = fluid viscosity
u = fluid kinematic viscosity
p = fluid density
dependence on the curvature ratio can be noted. The ana- 4 = angular coordinate in axial direction in Figure 1
lytical work of Topakoglu (1967) for low Dean numbers w = vorticity
supports the validity of the curvature-ratio dependence,
as shown in Figure 11. Topakoglu, extending Deans origi- Subrcripts
nal work but without simplifying the equations of motion, r = radial direction
demonstrated a resistance law for toroidal flow that de- 0 = angular direction in secondary-flow plane
pended upon both the Dean number and the curvature 4 = angular direction in axial direction
ratio, As with Deans, Topakoglus solution was limited
to low Dean numbers because of the difficulty of calculat- LITERATURE CITED
ing additional coefficients in his expanded-series solution.
The agreement of Topakoglus solution with this work Akiyama, M., and K. C. Cheng, Boundary Vorticity Method
for Laminar Forced Convection Heat Transfer in Curved
below a Dean number of approximately ten is noted in Pipes, Intern. J. Heat Mass Transfer, 14, 1659 (1971).
Figure 11. Austin, L. R., The Development of Viscous Flow Within
A second method was used to compute the friction fac- Helical Coils, Ph.D. thesis, Univ. Utah ( 1971 ).
tor for curved-tube flow as a further check on the conver- Bird, R. B., W. E. Stewart, and E. N. Lightfoot, Transport
gence of the numerical procedure. This method involved Phenomena, p. 119, Wiley, New York (1960).
the computation of the peripheral distribution of point fric- Brodkey, R. S., The Phenomena of Fluid Motions, p. 60,
tion factors from the gradients at the wall of the axial com- Addison-Wesley, Reading, Mass. ( 1967).
Dean, W. R., Note on the Motion of Fluid in a Curved Pipe,
ponents of the velocity. The velocity gradients were ob- Phil. Mag., 4, 208 (1927).
tained by use of second- or third-order forward finite- Dean, W. R., The Stream-Line Motion of Fluid in a Curved
difference expressions. Typical distributions are shown by Pipe, ibid., 5, 673 ( 1928).
Austin (1971). At high Dean numbers, the point friction Eustice, J., Flow of Water in Curved Pipes, PTOC.Roy SOC.,
factor increased dramatically in going from the inside of A84,107 (1910).
the coil to the outside of the coil. For example, at a Dean Eustice, J., Experiments of Stream-Line Motion in Curved
Pipes, ibid., A85, 119 (1911).
number of 400 and a curvature ratio of five, the overall Greenspan, D., Introductory Numerical Analysis of Elliptic
increase corresponded to a factor of almost four. The Boundary Value Problems, Harper and Row, New York
peripheral distribution of the point friction factors was (1965).
integrated to determine an average curved-tube friction Grindley, J. H., and A. H. Gibson, On the Frictional Resistance
factor. I n general, friction factors computed in the second to the Flow of Air Through a Pipe, Proc. Roy SOC., A80,
manner agreed with those computed from the first method 114 (1908).
Ito, H., Laminar Flow in Curved Pipes, ZAMM, 49, 653
to within one percent. (1969).
Kalb, C. E., and J. D. Seader, Heat and Mass Transfer Phe-
ACKNOWLEDGMENT nomena for Viscous Flow in Curved Circular Tubes, Intern.
J. Heat Mass Transfer, 15, 801 ( 1972).
This research was supported primarily through a National Keulegan, G. H., and K. H. Beij, Pressure Losses for Fluid
Defense Education Act Fellowship in the Chemical Engineering Flow in Curved Pipes, J. Res. Nut. BUT. Stand., 18, 89
Department at the University of Utah. Charles E. Kalb pro- ( 1937).
vided assistance in certain computations. Kubair, V., and N. R. Kuloor, Secondary Flow in Helical
Coils, Indian J. Technol., 1, 333 (1963).
NOTATION Larrain, J., and C. F. Bonilla, Theoretical Analysis of Pres-
sure Drop in the Laminar Flow of Fluid in a Coiled Pipe,
fc = friction factor for curved tube, dimensionless Trans. SOC.Rheol., 14, 135 (1970).
fs = friction factor for straight tube, dimensionless Lansford, W. M., discussion in Trans. ASCE, 100, 1036 (1935).
h = scale factor, dimensionless Lapidus, L., Digital Computation for Chemical Engineers,
N D , = Dean number = N R e ( R / R c ) h McGraw-Hill, New York ( 1962).
NRe = Reynolds- number = 2Rp G@/p McConalogue, D. J., and R. S. Srivastava, Motion of a Fluid
= combined static pressure and gravitational force in a Curved Tube, Proc. Roy. SOC., A307, 37 ( 1968).
p Mori, Y., and W. Nakayama, Study on Forced Convection
in the form of a scalar potential, as defined in Heat Transfer in Curved Pipes (1st Report, Laminar Re-
Bird et al. (19601, pp. 43-44 gion), Intern. J. Heat Mass Transfer, 8, 67 (1965).
P = dimensionless pressure, defined by Equation ( 3 ) Seban, R. A., and E. F. McLaughlin, Heat Transfer in Tube
r = radial coordinate in Figure 1 Coils with Laminar and Turbulent Flow, ibid., 6, 387
R = inside tube radius (1963).

AlChE Journal (Vol. 19, No. 1) January, 1973 Page 93

Schmidt, E. F., Warmeiibergang und Druckverlust in Rohr- Through Helical Conduits, Ph.D. thesis, Case Inst. Techno].,
schlangen, Chem. Ing. Tech., 39, 781 ( 1967). Cleveland, Ohio (1963).
Taylor, G. I., The Criterion for Turbulence in Curved Pipes, Vrentas, J. S., J. L. Duda, and K. G. Bargeron, Effect of
Proc. Roy. Soc. London, A124,243 ( 1929). Axial Diffusion of Vorticity on Flow Development in Circular
Topakoglu, H. C., Steady Laminar Flows of an Incompressible Conduits: Part I. Numerical Solutions, AZChE J., 12, 837
Viscous Fluid in Curved Pipes, J. Math. Mech., 16, 1321 ( 1966 ).
(1967). White, C. M., Stream-Line Flow Through Curved Pipes,
Truesdell, L. C., Jr., and R. J. Adler, Numerical Treatment of Pmc. Roy. Soc., A123, 645 (1929).
Fully Developed Laminar Flow in Helically Coiled Tubes,
AZChE J., 16, 1010 (1970). Manuscript 7eceiz;ed January 5, 1972; reeisiun receioed August 2,
Truesdell, L. C., Jr., Numerical Treatment of Laminar Flow 1972; paper accepted August 7, 1972.

Effect of Mixing on Nonuniformly Initiated

Polymerization by Radiation
An analytical study was carried out on the effects of mixing on radia-
tion-induced polymerization in which the species initiating the polymeriza- W A l CH IRO KAWAKAMI
tion were not uniformly produced because of a nonuniform dose rate distri-
and S U E 0 MACHI
bution. The analysis is based on the periodic irradiation of fluid elements
circulating in a stirred-tank reactor having a high dose rate region and a very Takasaki Radiation Chemistry
low dose rate region. It is shown that the rate of polymerization increases Research Establishment
with agitation speed and that the bimodal molecular weight distribution of Japan Atomic Energy Research Institute
the polymer formed at low agitation speed moves to a unimodal distribution Takasaki 370-12, Japan
as the agitation speed increases.

SCOPE
In this paper we present a simplified analytical method uniform dose rate distribution gives a higher overall reac-
to predict the effect of mixing on radiation-induced poly- tion rate for the same absorbed energy. If one can make
merization in which the species initiating the polymeriza- the distribution of the active species more uniform by
tion were not uniformly produced becawe of a nonuniform agitation before the deactivation, an increase in the re-
dose rate distribution. Under the influence of the nonuni- action rate may be expected.
form dose rate distribution, the rate of active species for- Up to this time, two different calculational methods
mation is naturally nonuniform, and the efficiency of active have been reported. One is based on a diffusion model in
species utilization in propagation reactions is low because which the radicals produced in the higher dose rate area
of the high probability of their mutual deactivation in the
are assumed to be dispersed throughout the reactor by
high concentration.
The rate of radiation-induced chemical reaction usually diffusion (Noyes, 1959; Hill and Reiss, 1968). The other
has a dose rate exponent ranging from 0.5 to 1.0, depend- is based on a circulation flow model, in which fluid ele-
ing on the deactivation mechanism of active species. In ments are assumed to be irradiated periodically by going
the case of reaction with dose rate exponent less than around in the reactor (Fredrickson et al., 1961; Kawakami
unity, it has been established that irradiation with a more and Isbin, 1970).

CONCLUSIONS AND SIGNIFICANCE

The analysis is based on the periodoc irradiation of fluid tion speed to attain values obtained under uniform irradia-
elements circulating in a stirred tank reactor having a tion by the average dose rate in the reactor. Also, the
high dose rate region and very low dose rate region as bimodal molecular weight distribution of the polymer
shown in Figure 2. The practical consequences of the formed at low agitation speed moves to a unimodal distri-
calculations are shown for the rate of polymerization and bution as the agitation speed increases. If the energy ab-
molecular weight diqtribution of polymer based on the sorbed in the reaction system is constant, there is a non-
polymerization of acrylamide in aqueous solution by an uniformity of dose rate distribution which gives a mini-
electron beam. It is shown that the rate of polymerization mum polymerization rate.
and number average molecular weight increase with agita- The agitation speed required to attain perfect mixing

Page 94 January, 1973 AlChE Journal (Vol. 19, No. 1)