ht. .I. Heat Moss Transfer. Vol. 17, pp. 845-859. Pergamon Press 1974.

Printed in Great Britain

PREDICTION OF FULLY-DEVELOPED FLOW IN A TUBE

CONTAINING A TWISTED-TAPE

A. W. DATE*
Mechanical Engineering Department, Imperial College, London SW.7

(Received 26 November 1973)

Abstract-The problem of fully-developed, laminar and turbulent, uniform-property flow in a tube con-
taining a twisted-tape has been formulated in terms of partial differential equations of momentum and
heat transfer. These equations have been solved by adapting an existing numerical procedure for two-
dimensional elliptic equations to predict the friction and heat-transfer characteristics of the flow.
A set of laminar flow predictions are presented to demonstrate the influence of Reynolds number,
twist-ratio, Prandtl number and the fin-parameter on the flow characteristics.
The turbulent-viscosity, necessary to predict the turbulent flow characteristics, has been calculated by
solving differential equations for kinetic energy of turbulence and the energy dissipation rate. This
approach is, however, found to be inadequate for accurate quantitative predictions; a measure for
eliminating this inadequacy is suggested.

NOMENCLATURE Prandtl number;

a, bl, bz, coefficients in the general partial effective Prandtl number;
c, a, differential equation; heat fluxes, divided by C, in radial,
c,,G> coefficients in the general finite difference tangential and axial directions;
C,,cW* equation; heat input per unit length;
C P specific heat; radius of the tube;
Cfin, fin parameter; cylindrical polar coordinates;
C,, , C,, C,, constants in the k-6 turbulence model; coordinates of a rotating coordinate
Q diameter of the tube; system;
&, hydraulic diameter; Reynolds number based on hydraulic
constant in the log-law; diameter;
:: friction factor based on hydraulic Reynolds number based on internal
diameter; diameter;
fi> friction factor based on internal source term of finite-difference equation;
diameter; distance from the wall;
h, heat-transfer coefficient; temperature;
H, pitch for 180 rotation of twisted-tape; radial, tangential and axial time-mean
k, kinetic energy of turbulence; velocity;
K,, thermal conductivity of fluid; shear velocity;
Kl> thermal conductivity of tape metal; twist ratio.
1, length scale of turbulence;
n, distance between the wall grid node and
the adjacent interior node;
Nu, average Nusselt number based on thickness of the twisted tape;
hydraulic diameter; dissipation;
Nui, average Nusselt number based on dynamic viscosity;
internal diameter; effective viscosity;
P, pressure; density;
p, mean pressure; dependent variable of the partial differ-
P 01 variable pressure; ential equation;
_~ -_ stream function;
*Present address: Department of Mechanical Engineer- laminar diffusivity multiplied by p;
ing, Indian Institute of Technology, Bombay-400076. effective value of I-;
845
846 A. W. DATE

w. axial vorticity; firstly, to show how the necessary transport equations

& constant in the log-law. are derived, and solved numerically on a computer;
and secondly, to present predicted characteristics of the
Subscripts flow to demonstrate the scope of the numerical pro-
cedure described in the paper. Of course, the numerical
b. bulk value over the cross-section;
procedures do not eliminate the need for experiments.
mean value over the cross-section;
but they certainly diminish their extent and cost by
LIy wall grid node;
highlighting the most influential parameters governing
no, near-wall grid node;
the characteristics under given conditions.
turb, turbulent part of the effective value;
The paper is divided into further four sections:
eff, effective value.
Section 2 describes the derivation of the necessary
transport equations; Section 3 shows how an existing
1. INTRODUCTION numerical procedure is adapted to solve the equations:
TWISTED-TAPE inserts inside circular tubes provide a in Section 4, some predicted data are presented; and
means of increasing the surface heat-transfer coefficient finally, concluding remarks are made in Section 5.
within the tubes of tubular heat exchangers. Figure 1
shows the layout of a tube contai~ng a twisted-tape. 2. THE MATHEMATICAL PROBLEM
The tape consists of a long metal strip which has been 2.1. The domain of interest
twisted about its longitudinal axis, the width of the tape In the region of the fully-developed flow, the dis-
being equal to the internal diameter of the tube. tributions of the vefocities are independent of the axial

c3
Directio
of flow
LA T;be
Sect Ion AA

FIG. 1. Illustration of the twisted-tape inside a circular tube

The pressure-drop and heat-transfer characteristics distance along the tube; in other words, they are
of the fully-developed turbulent flow in tubes contain- identical at every cross-section of the tube. One such
ing twisted tapes have been investigated experimentally cross-section is shown in Fig. 1 (section A-A).
by many investigators (see, for example, [l-3]); and Contrary to the real situation, it is assumed here
empirical correlations to predict these characteristics that the width of the tape exactly equals the internal
have been proposed. A survey carried out by the present diameter of the tube. The ratio of the tape thickness 6
author (see Date [18]) showed that the available ex- to the tube diameter D is usually small for the free
perimental data are not extensive, and the correlations space to be regarded as two twisting ducts of semi-
to predict them are applicable only for a certain range circular cross-section which rotate through 180 degrees
of the values of the inde~ndent parameters. For in an axial distance H. Since the flow will be identical
example, no experimental data exist for laminar flows, in each of the two semi-circular sections, only one of
and for fluids with high Prandtl numbers. Similarly, these need be considered.
the proposed correlations do not demonstrate con-
2.2. The coordinate system
sistent agreement regarding, for example, the influence
The only suitable coordinate system is one in which
of ~uid-property variations or Prandtl number on
the angular coordinate is measured always from the
Nusselt numbers. In order to gain knowledge of the
surface of the twisted-tape. Such a rotating cylindrical
characteristics of the flow over a range of conditions,
polar coordinate system (r, z, 0) say (see Fig. 2) is
more experimental data are needed. However, the effort
then related to the customary stationary system (r, z, R)
involved will be prohibitively expensive.
by the foIlowing equations:
A less tim~and-cost-consuming approach is to
r = r,
derive and solve the differential equations which de-
scribe the transport of momentum and heat in the _ --2.
_ (1)
twisted tape flow. The purpose of the present paper is and 0 = t)+ x,-/H.
 Prediction of fully-developed flow in a tube containing a twisted-tape 84-I

The positive sign before nz/H implies anti-clockwise Radial momentum equation
rotation of the tape as z increases, when distances z
and z are measured in the direction of axial flow, and +$+(g++I:)g-q =-$
Oand 8 are measured in the clockwise direction.
As a consequence of equations (l), the following
1
1 aV7, ,I 1 ha 7e,8 n a7,,,
- _I
+rdB-T+Hx (7)
relations may be written: r dr I
d (! Temperature equation
dr W
a -a (2)
ii aw'
a a 71a
and z=aZ+Hj$
Since ap/laz and aT/dz are known constants,
These relations are now employed to transform the equations (4-8) are two-dimensional : the dependent
well-established (see Bird et al. [4]) transport equations variables, V,, &, V, and T, that were functions of
of momentum and heat in (r, z, 0) system to those in r, z and 0 are now functions of r and 0 alone. These
(I, z, @) system. equations are general, and applicable to both laminar
and turbulent flows if, in the latter case, the instan-
taneous velocities are replaced by their time-averaged
2.3. The transport equations in (r, z, 0) system values.
The starting point in the derivation of the necessary It is important, at this stage, to note the way in
transport equations is the equations in the (r, z, 0) which the pressure-terms in equations (4-7) are treated.
system. Before making use of these equations, the It is assumed that the pressure at any point in the flow
following mathematical condition of the fully- consists of two parts. Thus, P is expressed as:
developed flow must be noted. It is that, P(r, z, 0) = P(z) + P,(r, @) (9)
a
z-- 0,
where P(z) is the mean pressure at a cross-section
and Po(r, 0) is that component of the pressure which
varies over the cross-section.
for all dependent variables, except for the mean pressure
The terms on the left hand side of the equations
P, and temperature T, and ap/dz and aT/az are to
(5)-(S) represent the convective fluxes, whereas the
be treated as known constants.
terms containing 7s and qs represent the diffusive
It follows from equations (2) and (3) that the con-
fluxes of momentum and heat respectively. When in-
tinuity, momentum and heat transport equations in
(r, z, 0) system are as follows: formation regarding rs and qs is specified, the math-
ematical statement of the transport equations is
complete.
Continuity equation
2.4. The dz$iision fluxes
For laminar fluids, the stresses TSare related* to the
velocity gradients (or strain) via p. Similarly, qs are
related* to the temperature gradients via r; and p
Axial momentum equation and r are constants for uniform property fluids. For
turbulent flow one may postulate the existence of
effective exchange coefficients analogous to p and I-.
Thus, for turbulent flow,

and (10)

Tangential momentum equation The subscript turb signifies the turbulent part of
the effective values. Further, it may be postulated that:
,{,$+(;+?$!!!+!LJ!} = _fG!!$ r& = M/PGff,ry (11)

1 --+--
ah) 71 ah) *The relations for each stress and heat flux component in
V27,,0)
+7 ae
H a@
(6)
a rotating coordinate system are given in Date [19].
848 A. W. DATE
where Pr,fl.T if the effective Prandtl number. Experi- The appearance of PO from equation (5) for axial
mental evidence (see Spalding [5]) suggests that its momentum is still not removed, although it is easily
value is about 0.9. seen that the term aP,/iM can be found from equa-
Thus the TSand qs are replaced by one unknown ha; tion (6).
in Section 2.7, the method adopted for its determination Now the stream function II, is defined thus:
is discussed. For the time being it is assumed that pL,Ris
knowable. (14)
In the discussion to follow, use will be made of
and
equations (10) and (11) to represent the diffusion fluxes,
with a reminder that when laminar flow is considered, a* --$I/,. (15)
~~8 and pr&T are to be replaced by the respective
SF=
laminar values: the laminar viscosity p and the laminar
These definitions, along with the definition of vor-
Prandtl number Pr.
ticity can be manipulated so that, the statement of
mass conservation, as expressed by the continuity
2.5. Introduction of vorticity and stream function
equation (4), is implicitly satisfied. The result of this
Equations (5)-(g), after substituting expressions for
manipulation is generally termed as the stream function
7s and qs, form a set of equations which need to be
equation:
solved. From the point of view of numerical solutions,
the explicit appearance of pressure POin these equations The stream function equation
poses certain difficulties. To eliminate its appearance
from equations (6) and (7), the axial vorticity w is
introduced such that the information contained in the
two equations is completely described by a single
equation for w, where w is defined as:
= 0. (16)

(12)
Thus, by introducing w and $ the mathematical
The process of arriving at the vorticity equation is as statement of the problem has been redefined; but with
follows: differentiate equations (6) with respect to r, the following advantages:
and equation (7) with respect to Q, and subtract the 1. Since the stream function equation satisfies mass
first from the second. Using the definition (12), the conservation, no explicit recourse to the con-
resulting expression can be transformed into the re- tinuity equation is necessary.
quired vorticity equation. It is as follows : 2. The introduction of w has replaced equations
The vorticity equation (6) and (7) by a single equation for w; at the same
time, partial elimination of pressure PO has also
been achieved.
The equations, which now need to be solved are:
equation (16) for $, equation (13) for w, equation (5)
for V, and equation (8) for T.
Following Gosman et al. [6] these equations can be
represented in a general form.

where (13) 2.6. Generalized representation of the equations

The terms

and V,

appear as multipliers on the left-hand side of the

equations for w, V, and T. These terms can be replaced
by stream function via equations (14) and (15). This

+;{&(Lkfig)+g&(&*g)}. done, the equations for $, w, V, and T can be

represented by a general equation for any dependent-
 Prediction of fully-developed flow in a tube containing a twisted-tape 849

Table 1. Equations for $, w, V, and T

4 a tJ1 h1 C d

$ 0 Upr rlp 1

w 1 r

K 1 mr

1 7tr ?T
T 1 I-eff,T ;+s l- etT.T 1 -rV,-
( ) SZ

*The term d, is as follows:

variable 4. This general equation is as follows: Following the dimensional arguments of Prandtl [9]
and Kolmogorov [8] hUrb is written as:

+ convection terms + where k is the kinetic energy of turbulence, and E is

the energy dissipation rate; C, is a constant whose
value is to be specified. Now, the spatial distributions
of k and E are determined by solving, with appropriate
/- diffusion terms ____I boundary conditions, the partial differential equations
fd=O, (17) for the conservation of k and E. The complete descrip-
source terms tion of the derivation of these equations is beyond the
scope of this paper; here their final forms are presented
where expressions for a, bl, b2, c and d for each in the same manner as those for $, w, V, and T in
dependent variable 4 are given in Table 1. Table 1.
Such a generalized representation of the equations
Table 2. Equations fork and E
will allow concentration on equation (17), when the
numerical solution of the equations is discussed in 4 a bl h2 c d
Section 3. Before proceeding to Section 3, however,
here is .a reminder: A method for determining bff, r'(G- p&)
reff,kr 1
which is vital to the prediction of turbulent flow is still
not described. Determination of &f is the subject 1 &
6 1 Ta7.E -+jp reK,e
r 1 r CGiG-CDpe2/k
matter of the next sub-section. ( r ) (

2.7. Determination of ,uef for turbulent flow where

To determine ~a, one must determine the spatial
distribution ofkUrb over the semi-circular cross-section.
One may specify AU& as an external information to the
set of equations, or it may be related to the product of and preK,kand Prcf,E,like Pr&T are constants. C, and
(Prandtls) mixing length and the local velocity gradi- C, are other two constants; and G, which stands for
ent, where the mixing length is again specified by an the generation of kinetic energy, is given by:
algebraic expression. However, since the twisted-tape
flow is assymetric, the algebraic expressions, referred
to above, are difficult to invent, especially in the absence
of extensive experimental data for velocity profiles.
Here, a method is chosen which does not necessitate
specification of any algebraic expressions; but simply
requires specification of certain constants which may
be universal. The method is as follows:

HMT Vol. 17. No. 8-D

850 A. W. DATE

Jones and Launder [lo] have suggested the following There is one such equation for each variable 4 at
values for the various constants mentioned in the above every interior node ofthe grid; with conditions specified
discussion: on the boundaries (node referred to by 0 in Fig. 2),
c, = 0.09, there will be as many equations as there are unknowns.
c, = 1.55, The set of algebraic equations given by (22), therefore,
c, = 2.0, is a solvable set. The solution of this set, however,
(21)
will require an iterative procedure because the Cs and
Pr,,, = 1.0,
S depend on the 4s. Gosman et ul. adopt the Gauss-~
and Pr,,, = 1.3. Seidel iteration procedure which incorporates the
Note that Pr,qr equals 0.9 (see equation 11). successive substitution formula.
These values of the constants have proved to be valid E
in many flow situations. For example, Laufers [20]
experimentally measured axial velocity profiles for
fully-developed flow in a circular pipe can be predicted
within 1 per cent (see Date [19]); the performance of
two-dimensional film-cooling geometries was predicted
accurately by Mathews [ll].
The numerical procedure to be described in the next
section is applicable to the equations of k and E as Tape wall
well; here it is emphasized again that these equations
FIN;.2. Illustration of the grid-distribution,
have eliminated the need for algebraic specification of and a typical interior node P of the grid.
(t&. The extent of the success of this method will be Note the area called cell bounded by
demonstrated in Section 4, where some typical predic- dotted lines, and the wall node (0) and the
near-wall node (no).
tions are pre:ented.
For the twisted-tape flow equations described in
The set of differential equations specified above Section 2, this standard iteration procedure does not
cannot be solved unless the conditions of the variables 4 always produce convergence. For instance, in compu-
are specified on the boundaries. The boundaries of tations at very high Reynolds numbers and high twists
interest are: the tube wall surface and the twisted-tape of the tape, the source term of the vorticity equation
surface at any semi-circular cross-section. The dis- (see Table 1) becomes very large. Under these condi-
cussion of the boundary conditions is deferred to the tions, a given flow perturbation (i.e. a change in $
next section for convenience. and V,) produces a large change in (11~(see equation 22).
Thischangein(+,causes thestreamfunction to be altered
greatly, since (11~appears in the source term of the
3. THE NUMERICAL SOLUTION OF THE EQUATIONS
stream function equation. This greatly altered stream
3.1. The jnite-differencr equations and their solution
function distribution perturbs the flow further. Clearly
The finite-difference solution procedure adopted to
this computation scheme can become unstable, al-
solve the general partial differential equation (17) is
though the physical flow situation is a stable one.
described in detail by Gosman et a[. (IS]; therefore,
Gosman and Spalding [7] suggested an iterative pro-
only its main features are described here.
cedure to cure such an instability; the procedure was
Essentially, the partial differential equation (17) for
called the Multi-point circulation adjustment
each variable 4 (4 variables for laminar flow and 6
(MPCA) procedure. The application of the procedure
variables for turbulent flow) is converted into a set of
to the twisted-tape flow equations is described in detail
algebraic equations through a finite-difference algor-
in Date [19]; here, only its principal features are
ithm. These algebraic equations can be formulated as
summarized.
follows* :
The procedure simultaneously adjusts the value of
4,= Gh4 +G4 +wk+Gv4v+~, (22) vorticity wp and its neighbours wH, us, wE and ow,
where the cs are the coefficients which usually depend and subsequently the stream function $, in a single
on the variables 4s, and depending on the influence sweep of the field of computation. The constraints on
of convective and diffusive fluxes, determine how these adjustments are:
heavily the neighbouring values of Q, should weigh in
(a) that the imbalance in equation (22) for vorticity
the determinations of 4,. The effects of sources of 4
is reduced to zero at every grid node;
are contained in term S.
(b) that the circulation principle for the cells sur-
*See Fig. 2 for the meaning of the subscripts. rounding grid nodes P, N, S, E and W is satisfied.
 Prediction of fully-developed flow in a tube containing a twisted-tape 851

The circulation principle requires that: twisted-tape, and by solving a one-dimensional heat-
conduction equation along the width of the tape. Here
jAco.dA=~ V.&S (23) the result of the analysis which is given in Date [19]
is presented. The analysis gives rise to a quadratic
where A is area of the cell, V is the velocity along expression for the tape surface temperature To, say; this
the boundaries of the cell, and dS is the incremental expression is:
length along these boundaries.
aTi+bT,+c = 0, (25)
Besides MPCA, there are other conventional pro-
cedures such as the under-relaxation of the finite- where a, b and c are functions of the temperatures at
difference equations and the under-relaxation of source nodes adjacent to To, and Cfi,, where Cs, is the fin-
terms of the equations which are necessary to obtain parameter defined as :
convergence; the details of these are given in Date [ 191.
(26)
3.2. Boundary conditions
As mentioned in the previous section, boundary con- Turbulent pow. In the close vicinity of the wall,
ditions are necessary to close the numerical problem. dependent variables such as V,, k, o, T, etc., for
The equations for fully-developed flow are two-dimen- turbulent flow, vary rapidly with distance from the wall.
sional and elliptic, and conditions of the variables 4 Therefore, to preserve the accuracy of the finite differ-
must be specified on four boundaries of the domain ence procedure, very fine grids will be necessary in
of interest. These boundaries, with reference to Fig. 2, this region. Consequently, the computer time will
are: AB, BEC, CD, and DA. The boundary DA is increase substantially.
essentially the centre point of the twisted-tape; here it To obviate this shortcoming, implications of the
is treated as a boundary of zero radius. log-law of the wall suggested by Backshall and
In the subsequent discussion the boundary condi- Landis [17] are used to serve as the boundary condi-
tions for the laminar and turbulent flow are described tions. The suggested law pertains to the representation
separately. of the variation of the axial (V,) and the cross-stream*
Laminar jlow. Each boundary of the domain of (V,,) velocities in the near-wall region with the distance
interest represents a solid wall; the velocity V, is there- from the wall. The complete derivation of the boundary
fore zero on all boundaries, and the stream function ti conditions is given in Date [19], here the suggested
assumes a constant val.ue: IJ was assigned the value log-law and the boundary conditions are stated.
zero in the present calculations. The values of w and
T, however, require further consideration.
Vorticity is defined in terms of the gradients of the
cross-stream velocities but the distribution of these and
velocities in the near-wall region is unknown. The
boundary conditions for vorticity can however be
obtained from the no-slip condition, and from the
assumption regarding the profiles of the velocity com- where tl is the angle between the direction of the
ponents in the near-wall region. The complete deriva- absolute velocity and the cross-sectional plane and
tion of the boundary conditions can be found in V,,, and V,,, are the cross-stream and axial shear
Date [19]; here, it will suffice to note the general velocities respectively.
form of the boundary conditions appropriate to the With reference to Fig. 2, the boundary conditions
four boundaries. It is as follows: are as follows :

$0 = 0, (29)
w 0 = g,(~-~no)+g20J +g3v (24)
pn2 Ls*o
(30)
where subscripts 0 and no refer to the boundary node
and to the adjacent interior one respectively; n denotes
the distance between the two nodes (see Fig. 2 for
illustration); and gr, g2, g3 are dimensionless numbers.
In the present calculations it is assumed that the z& = c,- 12v,2, (32)
temperature on the tube wall BEC is maintained at a
uniform value. The temperature on the tape surface,
*V,, refers to Ve near the tube-wall; it refers to V,cos 0;
i.e. the boundaries AB, CD, and DA, is unknown. It near the tape-wall, 0; being the angle made by the first radial
can be determined bv, referring to the fin-actinn of the__ gnd-lme away trom the tape-wall with the tape-wall.
852 A. W. DATE

and c?T*Jaz is now a known constant since Rei and Pr

are known.
The predicted value of T*, and its mean value TX
where enables evaluation of the average Nusselt number Nui.
Thus :

and (40)

V-J
s( = tan- i ._._
r,
~~
\ v<mo/
1 (35) where,

h = Q/V,- TJW, (41)

In equation (35), V,,,,, is evaluated from the definition
and, therefore,
of stream function and the log-law of the wall: and
V,,,, and V,,, in equations (30) and (31) are calculated Nui = - llT,*/rc. (42)
by iterating equations (27) and (28) with V, = L&,
The evaluation of the mean values of I, and T* in
6.Y = V,, and s = n, all of which are known. Thus it
the above calculations is carried out by using Simpson
is seen that the variables w, k and E are specified at
rule.
the first grid-node away from the wall whereas $ and
The predictions to be presented were obtained with
V, are specified at the wall, V,,, being the slip velocity
15 x 15 uniformly-spaced grids. This grid size provided
deliberately introduced (see, Patankar and Spalding
sufficient numerical accuracy (see Date [ 191).
1261).
4.1.2. Friction characteristics. Figure 3 shows the
To account for the steep gradients of temperature
predicted variation off;: with Rei at different values of
near a wall, the value of I,, at the grid-node no is
y. The solid line for J = E corresponds to the flow in
calculated from the law suggested by Spalding and
a straight semi-circular duct for which an analytical
Jayatillake [25] ; the details of calculations can be found
solution is available (Weigand [12]). The present pre-
in Date [ 191.
diction agreed with this exact solution within 1 per cent.

4. PRESENTATION AND DISCUSSION OF RESULTS /o

4.1. Laminar flow 3 14
4.1.1. Method of prediction. To predict the friction
characteristics, the equations of $, (u, and V, were
solved for given input values of p, p, dP/rz and z/H.
The predicted profile of K enables the calculation of
Reynolds number Re, and the friction factor fi:. Thus:
i-

(36)
I0-l
and
Rr. = !!!%!!-!~ (37)
P
where V,,, is the mean value of Vz.
3xd
The predicted distribution of V, and $ is subsequently 4x IO' I03 2x103
used to solve the temperature equation. This requires Re,
additional inputs, which are: the Prandtl number Pr, FIG. 3. Predicted friction factor data for fully-developed
and the fin-parameter Cs,, and the source term aT/dz. laminar flow.
In solving the temperature equation, a dimensionless
temperature T* is used, where:
The predicted data which lie above this line cor-
T* = (T- LMQIkfL (38) respond to those cases in which the tape possesses a
finite twist. As would be expected. for the same flow
where Q is the uniform heat flux per unit axial
rate (or same Rri), the friction factor increases as y
distance. The dT*/az can then be determined from the
decreases. Also at lower Reynolds numbers the data
overall heat balance over a length AZ as:
converge on the line for y = sci, irrespective of the
aT* 4 value of y. Thus the friction factor data show plausible
-=--. (39)
azl xr,ReiPr trends.
 Prediction of fully-developed flow in a tube containing a twisted-tape 853

Figure 4 shows the effect of y on the predicted axial friction factor data, converge on the line for y = co.
velocity profifes as Reix 1200. The effect of twist is Efict of Prandtl number on Nusselt number at
evident from the assymetry exhibited by the profiles. C,, = W-Figure 6 shows the variation of Nui with
As the twist increases (or y decreases), the profiles Rei for various Prandtl numbers. The solid horizontal
disintegrate into a pattern which shows two peaks of line again corresponds to the case of y = cc ; as would
axial velocity. be expected, the predicted value of Nui is unaltered by
It should be noted here that unlike in the case of the value of Prandtl number.
y = co, secondary flow exists in cases for which y < co. The predicted curves which lie above the solid line
The magnitude of the secondary velocities increases represent predictions for y = 5.24, and they signify the
with increase in Rei and decrease in y. In the present fact that for y < CC, when secondary flow exists, the

y= 5.24 3 14 2.25

FIG. 4. Effect of twist-ratio on predicted profiles of axial velocity (V;/V ,) at

Rei z 1200 (arrows indicate rotation of tape in the direction of fl&

Predictions at Cfin=~,pr=l.~
Symbol Y

FIG. 5. Influence of twist-ratio on Nusselt number at C,, = T

predictions, the mean values of the secondary velocity Nussdt number is dependent on both Reynolds num-
ber and Prandtl number. The predictions are similar in
trend to those found in the case of flow in a curved
were found to be between 10 and 35 per cent of the pipe (see Akiyama and Cheng [ 141)where the secondary
mean axial velocity. flow induced by the curvature increases the convective
4.1.3. ~eut-transfer characteristics. EZect of twist- heat transfer.
ratio on Nusselt number at Cj, = co-Figure 5 shows The important difference between the results for
the variation of NUi with Rei for various values of y curved pipe and the present predictions is that the
for maximum fin-effect. The solid horizontal line cor- increase in the Nusselt number in the Prandtl number
responds to the analytical solution for y = co (Eckert range lo-100 is greater than that in the range l-10.
et al. [13]); and represents Nui = 5-4. This value, which Thisincreasein Nusselt number increases with increase
is independent of Reynolds number and Prandtl num- in the Reynolds number. This is because, added to the
ber is predicted accurately for three Reynolds numbers. convection due to secondary velocities (which are large
The data for other values of y demonstrate that at in the twisted tape flow), there is convection due to the
constant Reynolds number the Nusselt number in- component of the axial velocity [i.e. (n/H)f/l, see
creases as y decreases. This increase is particularly equation (8)] which is large at high Reynolds numbers.
appreciable at high Reynolds numbers. At low The substantial increases in Nusselt number obtained
Reynolds numbers, however, the NUi data, like the at high Prandtl numbers are therefore to be expected.
x54 A. W. DATE

Also within the range of Rei and Pr considered here equalled the average heat-transfer coefficient on the
at any fixed Reynolds number, the value of NuJPr tape surface when Cs, = co, then one could obtain that
decreases with increase in Pr; showing thus that Nusselt Nlri,~DINui,, = (~~+2~)/~~ = 1.64. In the present
number becomes bounded with increasing Prandtl computations the ratio is found to be 2.07 mainly
numbers. because the average heat-transfer coefficient on the tape
E&t of fin-parameter on Nusselt number-Figure 7 surface was greater than that on the tube wall for
shows the predicted values of Nui at 4 = 225 and co C,, = co ; the latter coefficient also being slightly larger
are shown for three values of Cfin: 0 (i.e. minims than the average coefficient for Cs, = 0.

4x102 Predict ions at C,,, =m, y=524

Symbol Pr
n 0. I
0 I ,o /
I
l 10-O /
IO2 * 100~0 9
/
---- Curves through /
/
: the dato 4

FIG. 6. Influence of Prandtl number on Nusselt number at Cfin = (~j.

2x10 Predictions at Pr-l,y:2.25

- 4-_cI-

___---
__--

FIG. 7. Influence of fin-parameter on Nusselt number at y = 2.25. Hori-

zontal lines correspond to y = a. Dashed lines are curves through the
predicted data.

fin-effect), and l-85 (i.e. an intermediate fin-effect), and The plausibility of the data for y = 2.25 is, of course,
XI (i.e. maximum fin-effect). The values of NUi for shown by their asymptotic behaviour at low Reynolds
y = co represented by the solid lines corresponding to numbers. However, from the designers point of view,
Cs, = 0, 1.85, and co are respectively 2.594, 3.801 and the important observation to be made concerns the
5.4. These values were found to be independent of Rei values of Nu~,,/N~Q,~. The value of this ratio is nearly
and Pr. 2.1 at low Reynolds number, whereas at Rq = lo, the
Note that the ratio Nui,,/Nul,, for 4 = co is 2.07. value of the ratio equals 262. Thus, the improvement
This can be explained as being primarily due to the in heat transfer available from the fin-effect depends
increase in the surface area available for heat transfer on the value of Reynolds number when y < co; in
when Cs, = co compared to when Cs, = 0. Thus, if the Date [19], it is shown that it also depends on the
average heat-transfer coefficients on the tube wall were value of the Prandtl number.
identical for Cs, = 0 and co, and if these coefficients In Date [19], where many more predicted data are
 Prediction of fully-developed flow in a tube containing a twisted-tape 855

presented, it is shown that for all practical purposes, a 4.2.2. Friction characteristics. A typical comparison
single carameter Re,/y can be used to correlate the Nui of the predicted friction factor with the available
data for a particular value of Pr and Ca, . experimental data is shown in Fig. 9 for y = 4.14. The
Figure 8 shows typical temperature profiles at experimental data are under-predicted by nearly 30
C,, = 0, 1.85 and cc. The figure shows that as Cs, per cent. In Date [19] where comparisons for other
increases, the difference between the value of the bulk values of y are presented, underprediction of the same
fluid temperature and the tube wall temperature de- order of magnitude is observed. While qualitative
creases. This supports the trend observed earlier that trends of the predictions is in agreement with the
the Nusselt number, which is inversely proportional to experimental data, their quantitative disagreement is
this temperature difference, must increase as the fin- certainly due to the inadequacy of the model of
parameter increases. Note also the manner in which turbulence used in the predictions. The main reason
the value of Cs, affects the gradient of temperature near for underprediction is that the value of bff, as pre-
the tape wall. dicted from k and E equations, is too small. Com-

FIG. 8. Influence of the fin-parameter on predicted profiles to T* at Re, = 810,

y = 5.24 and Pr = 1.0 (arrows indicate rotation of tape in the direction of flow).

zo-

----PredictIons [y--4.15]
00 o Seymour (1966)
0 0 0 Koch(l958)
rl
0 lo- o o~Thorseneta/.(1969)

- fo =0.046

FIG. 9. Comparison of predicted and experimental friction factors

This completes the presentation and the discussion putational experiments showed that the predictions
of the laminar flow characteristics. We now turn to the can be forced to match the experimental data by
turbulent flow characteristics. augmenting the effective viscosity (or C,) in the axial
velocity equation* by a factor greater than 1. Over the
4.2. Turbulent jiow range of Reynolds numbers considered, the value of
4.2.1. Method of presentation. The method employed this factor, however, depended on the value of y. For
for prediction of friction and heat-transfer charac- instance, at y = 11.0 the value of the augmentation
teristics ofthe turbulent flow is similar to that described factor was found to be 1.1; at y = 5.15 to be 1.2;
for laminar flow, except that the equations for k and E at y = 4.14 to be 1.35; and at y = 3.14 to be 1.6. This
are solved with values of C, , Cc, Cn, Pref,k, PrCREand means that the constant C,, is not universal, or simply
Pr,fl. r being specified as explained in Section 2. The that the turbulent viscosity is not isotropic. Lilley and
values off;:, Rei and Nui are calculated in the same man- Chigier [15] have provided experimental evidence to
ner asdescribed in Section 4.1, but here these values are the effect that, in axisymmetric swirling flows, turbulent
divided by the ratio of internal to hydraulic diameters viscosity must be non-isotropic. Roberts [16], who
(i.e. D/D,,). This is done to make a meaningful com- predicted a decaying swirling flow, could obtain reason-
parison of the predictions with the available experi-
mental data. The value of D/D,, used is 1.66. The friction
*The predictions can also be forced to match the experi-
factor, Reynolds number and Nusselt number thus mental data by augmenting the effective viscosity in the
calculated are referred to by f, Re and Nu respectively. vorticity equation.
856 A. W. DATE

able predictions only when the turbulent viscosities in longitudinal-strains. Account of this fact, the author
the axial and tangential directions were ascribed differ- believes, will bring about the further augmentation of
ent values. The turbulent viscosity has directional the implied turbulent viscosities to enable quantitat-
properties (or is non-isotropic) in these cases, because ively correct prediction of the friction factor. The com-
appreciable velocity gradients (or strains) exist in more putational effort involved, especially in terms of obtain-
than one direction. In the twisted-tape flow, which is ing convergent solutions to the equations cannot be
non-axisymmetric swirling-type, a similar situation undermined; and it may be more expedient to follow
exists. the concept of augmentation factor mentioned in the
If such flow situations are to be predicted without ad last paragraph.
hocmeans suchas the augmentation factor introduced Figure 10 shows the comparison of the predicted and
above, then one must do away with the turbulent experimental (Backshall [17]) profiles of axial velocity
viscosity concept. Instead, description of the turbulent at y = 3.08 and Rq = 5 x 104. The agreement between

FIG. 10. Comparison of predicted (---) and experimental I----, Backshall

[17]) axial velocity profiles ( Vz/Vz,,)at y = 3.08. (Arrows indicate rotation
of tape in the direction of flow.)

Kinetic energy Length scale

I
2; IO3
f0

FIG. 11. Predicted profiles of length scale and kinetic-energy at y = 4.15 and
Re, = 1.2 x 104. (Arrows indicate rotation of tape in the direction of flow.)

stresses must be obtained from the solution of Reynolds the two profiles is qualitatively acceptable, although
stress equations. The task of this approach is, however, the magnitude of the predicted axial velocity in the
quite formidable at the present time. In Date [19] centre of the cross-section exceeds the measured value
simplified forms of the Reynolds stress equations are by nearly 14 per cent.
used for prediction purposes. These predictions, al- Figure 11 shows the typical distributions of the pre-
though nearly 15 per cent better than the ones described dicted length scale and kinetic energy profiles. The
here, do not agree with data. The predicted stresses, kinetic energy profile is plausible; for near the wall,
however, are such that they indeed imply much larger where the shear stress is maximum, so is the value of
turbulent viscosities than the ones predicted by solving energy; near the centre of the cross-section the value of
k and E equations. energy diminishes since the shear stress diminishes.
Following Hanjalic and Launder [23] and Launder The profiles of the length scale clearly indicate the
and Ying [24], the simplified forms used in Date [19] usefulness of solving the dissipation (E) equation; for
can be further refined to account for the fact that the such a length scale distribution, if indeed true, would
stresses in the plane of cross-section are produced by be difficult to specify algebraically. The predicted dis-
 Prediction of fully-developed flow in a tube containing a twisted-tape 857

tribution, however, seems plausible; for, as would be Reynolds numbers of interest. The multi-point circu-
expected, the length scale increases as distance from lation adjustment procedure must be incorporated in
the wall. the procedure of dosman et al. to ensure convergence
The heat-transfer characteristics are examined next. of the solutions.
4.2.3. Heat-transfer characteristics. The comparison 2. The laminar flow predictions for tapes of finite
of the predicted and experimental Nusselt numbers is twist have agreed asymptotically with the analytical
shown in Fig. 12. The scatter in the experimental data solutions for the case of ffow in a straight semi-
is primarily due to the different fin-effects existing in circular duct. The predictions have shown that sig-
different experimental set-ups. The predictions were nificant augmentation in heat transfer can be obtained
therefore obtained for Ce, = 0 and co; the measured at high Reynolds numbers, high Prandtl numbers, low
Nusselt number data should lie between these two twist-ratios (i.e. high twists), and high fin-parameters.
extremes of the fin-effect. The exact values of the fin- 3. By solving the differential equations for kinetic
effect for the experimental data are unknown; the in- energy and its dissipation rate, the need for algebraic
vestigators expect the fin-effect to be not greater than specification of the effective viscosity or of the length-
10 per cent. This means that the measured Nusselt scale has been removed.
numbers cannot be more than 10 per cent greater than 4. The constants in the turbulence model, as
those for C,, = 0. suggested by Jones and Launder [IO], although quite
6~10~
F

--- Predictions[y=4.15]
0 Thorsen etUl.(l969)
0 Koch (1968)

Re
Fw. 12. Comparison of predicted and experimental Nusselt numbers.

That the experimental data lie much closer to the universal for flows in which velocity gradients are
line for C,, = co suggests that the Nusselt number appreciable only in one direction, are found to be
data, like the friction factor data, are under-predicted. inadequate for predicting the twisted-tape flow charac-
In Date L-193,it is shown that this under-prediction is teristics. The constants, for example, under-predict the
a direct consequence of the fact that the low magnitudes value of the friction factor (and consequently the
of the effective viscosity have been predicted. It is Nusselt number). Qualitatively, however, the constants
important to note, however, that the qualitative trend predict data which are in agreement with the experi-
of the predictions is in agreement with the experimental mental data.
data, and one can reasonably argue that if the friction 5. The quantitative disagreements, mentioned above,
factor data were predicted correctly, then so will be the expose the inadequacy of the effective-viscosity concept
Nusselt number data. Indeed when the augmented in modelling turbulent stresses in twisted-tape flow.
turbulent viscosities were used, the Nusselt number Future research must seek to obtain the information
predictions were more realistic. regarding the stresses from the Reynolds stress
equations.
5. CONCLUWONS Acknowledgements-The author is grateful to Professor
The following are the conclusions of the present D. B. Spalding for suggesting the rotating coordinate system
whlch forms the basis of the present work. The author is
paper : also grateful to Dr J. H. Whitefaw who read the first draft
1. The transport equations of the twisted-tape fIow of this paper and made suggestions for improvement.
The work reported in the paper was carried out as a
are strongly coupled; therefore, the iteration scheme part of the authors Ph.D. research programme; the financial
suggested by Gosman et al. [IS] cannot be employed assistance provided by the Science Research Council is
successfully over the entire range of twist-ratios and gratefully acknowledged.
858 A. W. DATE

REFERENCES 14 M. Akiyama and K. C. Cheng, Boundary vorticity

method for laminar flow forced convection heat transfer
1. E. Smithberg and F. Landis, Friction and forced con- in curved pipes, Int. J. Heat Mass Transfer 14, 165991675
vection heat-transfer characteristics in tubes with twisted (1971).
tape swirl generators, J. Heat Trunsfer 86, 39-49 (1964). 15 D. G. Lilley and N. A. Chigier. Non-isotropic turbulent
2. R. Thorsen and F. Landis. Friction and heat-transfer stress distribution in swirling flows from mean value
characteristics in turbulent swirl flow subjected to large distributions, Int. J. Heat Mass Transfer 14(4), 573
transverse temperature gradients, J. Heut Trawfer 90, (1971).
87-97 (1968). 16 L. W. Roberts, Turbulent swirling flows with recircula-
.1 R. F. Lopina and A. E. Bergles, Heat transfer and tion, Ph.D. Thesis, University of London (1972).
pressure-drop in tape generated swirl flow, M.I.T.. R. G. Backshall and F. Landis, The boundary layer
Dept. of Mech. Engng, Rep. No. DSR 7028 l-47 (1967). velocity distribution in turbulent swirling pipe flow,
4. R. B. Bird, W. E. Stewart and E. N. Lightfoot, Trunsport J. Bus. Engng 91,728-733 (1969).
Phenomena. John Wiley, New York (1960). A. W. Date, Review of single-phase forced convection
5. D. B. Spalding, Contribution to the theory of heat data for flow in tubes containing twisted-tapes, Heating
transfer across a turbulent boundary layer, Inc. J. Heat & Ventilating Engr J. Air Conditioning (November 1973).
Muss Transfer 7, 743- 761 (1964). A. W. Date, Prediction of friction and heat transfer
6. A. D. Gosman, W. M. Pun, A. K. Runchal, D. B. characteristics of a flow in a tube containing a twisted-
Spalding and M. Wolfstein, Hwt and Mass Transfer in tape, Rep. No. HTS/73!15, Dept. of Mech. Engng,
Recirculating Flows. Academic Press, London (1969). Imperial College (1972).
I. A. D. Gosman and D. B. Spalding, Computation of 20 J. Laufer. The structure of turbulence in fully-developed
laminar flow between shrouded rotating discs, Imperial pipe flow, NACA, Rep. No. 1174 (1954).
College, Mech. Eng. Dept., Rep. No. EF;TN;A/30 (1970). 21 E. V. Seymour, Fluid Row through tubes containing
8. A. N. Kolmogorov, Equations of turbulent motion in twisted tapes. The Engineer 222, No. 5779 (1966).
incompressible fluid, IX. -II&. iVauk. SSSR. Series Fiz. 22 R. Koch, Pressure loss and heat transfer for turbulent
6. 56 (1942). Also Imperial College. Mech. Eng. Dept., flow. ForschHft. Ver. Dt. Ing. 24. Series 13, No. 469.
Rep. No. N/6 (1968). 144 (1958). English translation AEC-tr-3875.
9. L. Prandtl, Uber ein neues Formelsystem der ausge- 23 K. Hanjalicand B. E. Launder, A Reynolds stress model
bildeten Turbulenz, Nachr. Akad. Wiss. Gottingen, of turbulence and its application to asymmetric bound-
Math-Phys. KL. p. 6 (1945). ary layers. Imperial College. Mech. Engng Dept.,
10. W. P. Jones and B. E. Launder. The prediction of TM/TN,A/R (1971).
laminarization with a two-equation model of turbulence, 24 B. E. Launder and W. M. Ying, Fully-developed tur-
Inc. J. Heat Muss Tran.sfer 15(2), 301-314 (1972). bulent flow in ducts of square cross-section, Imperial
II L. Mathews, Performance of some two-dimensional film College. Mech. Engng Dept.. Rep. No. TM/TN/A/11
cooling geometries. Ph.D. Thesis, University of London (1971).
(1972). 2s D. B. Spalding and C. L. V. Jayatillake, A survey of the
12 A. Weigand, The problem of torsion in prismatic mem- theoretical and experimental information on the resist-
bers of circular segmental cross-section, NACA Tech. ance of laminar sub-layer to heat and mass transfer.
Memo.. No. 1182, translation of the original German Proc. 2nd All Soviet Union Conference on Heat and
paper in Luftfithrt Forschung 20(12). 333-340 (1944). Mass Transfer, Minsk. U.S.S.R. (1966).
13 E. R. G. Eckert, T. F. Irvine. Jr. and J. T. Yen, Local 26 S. V. Patankar and D. B. Spalding, Heat and Mass
laminar heat transfer in wedge-shaped passages, Truns. Trtrnsfir in Bounduq Layers. MorganGrampian,
Am. SW. Mrch. Engrs 80. 143331438 (19.58). London (1967).

PREVISION DE LECOULEMENT ETABLI DANS UN

TUBE CONTENANT UN RUBAN HELICdiDAL

Resume-On formule par les equations aux derivees partielles du mouvement et du transfert thermique,
Itcoulement etabli laminaire ou turbulent dun fluide a proprittes constantes, dans un tube contenant
un ruban helicoidal. Ces equations sont resolues en adaptant une methode numerique deja existante
pour les equations elliptiques bidimensionnelles. On calcule les caracteristiques de frottement et de
transfert thermique.
On presente un ensemble de calculs pour Iecoulement laminaire de FaGon a montrer linfluence du
nombre de Reynolds, du pas de Thelice. du nombre de Prandtl et du parametre dailette.
Pour calculer la viscosite turbulente ntcessaire, on r&out les equations aux derivees partielles de
Ienergie cinetique turbulente et du taux dtnergie dissipee. Cette approche est neanmoins insuffisante pour
des previsions quantitative exactes; on suggere une modification pour eliminer cette imperfection.

BERECHNUNG EINER AUSGEBILDETEN STRBMUNG IN

EINEM ROHR MIT SPIRALEINBAUTEN

Zusammenfassung-Fur den Fall ausgebildeter laminarer und turbulenter Stromung mit einheitlichen
Stoffeigenschaften in Rohren mit eingebauten Spiralen wurden partielle Differentialgleichungen fur
Impuls- und Warmeaustausch aufgestellt.
 Prediction of fully-developed flow in a tube containing a twisted-tape 859

Urn Druckabfall und Warmetibergang der Stromung zu berechnen, wurden die Gleichungen durch
Anpassung einer bekannten numerischen Prozedur fur zweidimensionale elhptische Funktionen gel&t.
Die Ergebnisse fiir laminare Stromung sind wiedergegeben, urn den EinfluB der Re-Zahl, des Windungs-
verhlltnisses, der Pr-Zahl und des Rippenparameters auf die Striimungsform zu zeigen.Die scheinbare
Zahigkeit, die man braucht, urn die Eigenschaften der turbulenten Stromung zu bestimmen, wurde
berechnet aus der Differentialgleichung fur die kinetische Energie der Turbulenz und die Gri%e der
Dissipationsenergie. Es zeigte sich, daB diese Naherung fiir genaue quantitative Aussagen nicht ausreicht.
Es wird ein Verbesserungsvorschlag gemacht.

PAC4ET HO.JlHOCTbFO PA3BMTOFO HOTOKA B TPYBE CO IBHEKOM

~HHOT~~R-~3a~~~~O~O~HOCTbK,p~3BHTOM~~MHH~pHOM~Typ6yn~H~HOMO~HOpOLIHOM TeYeHWW
B Tpy6e CO IIIHeKOM C+OpMyJlkipOBW C rIOMOLLIbW ypilBHeHFi2irIep(?HOCaBMrIyJlbCa R TerlJIaB
YPCTHLIX lTpOrl3BOnHblX.3TH ypaBHeHHR PemeHbl ll)'TeMMOW$HKaUHli MeTOAB, npHMeHReMOI-0 B
,ilByMepHblX 3JlJIWlTki'feCKHX
ypaBHeHWlX, JlJlRpWieTa K03~~WUWeHTOB TpeHHIl H TtWIOO6MeHa
rIOTOK&
npHBeneHHblC paC'IeTHble WHHble &'ISlJlaMBHapHOrO TfFieHAR rIOKa3blBaIOT BJIURHWe 'IACJII
PeZLnonbnca, kosdxjmqeerrra 3axpy~nnamifl, qecna Hpaunrna u napaMerpa pe6pa ria xaparre-
pHCfUKllIIOTOKI.
Ko3+#wf~WeHT Typ6yJIeHTHOfi BI13KOCTH, HeO6XOnkiMbrk AJlflpaC'ieTa XapaKTepFiCTHK Typ6y-
neHTHOr0 TWeHHR, pac0i~TblBaJtCa rQTeMpemeHAR W.$~peHUHaJ-fbHblX ypaBHeH&iiiKWHeTWieCKOfi
3HeprWiTyp6yJleHTHOCTH HCKOpOCTU~HCCW~BUHH3HeprHH.~~H~KO,yCT~HOB~eHO,YTOT~KOiiMeTO~
He~pHrO~HblM~Jl~ TOYHbIX KOJIWYeCTBeHHblX paC'feTOB.npeJVGiraeTCR cnoco6 yCTpaHeHHR
OKil3FiJlCII
3TOrO HenOCTaTKa.