Int. J. Heat Mass Tran.,(fer. Vol. 10, pp. 37-59. Pergamon Press 1967.

Printed in Great Britain

STUDY ON FORCED CONVECTIVE HEAT TRANSFER

IN CURVED PIPES
(2NO REPORT, TURBULENT REGION)
YASUO MORI and WATARU NAKAYAMA
Department of Mechanical Engineering, Tokyo Institute of Technology

(Received 22 February 1966)

Al~traet--For this paper the effect of curvature on heat transfer for fully developed turbulent flow in
curved pipes on the condition of constant heat flux was studied theoretically and experimentally. In the
analysis, a boundary layer is considered to exist along the pipe wall. Local shear stress and local heat
flux at the wall are given on the basis of reducing the resistance (2s) and the Nusselt number (Nu,) formula
for straight pipes to the local relation of friction and heat transfer. When the formulae for straight pipes
are given as 2 s oc Re-1/m and Nu s oc Re I"- 1)/,,, it is shown that the dynamic similarity and also the simi-
larity for heat transfer in curved pipes depend upon Re(a/R)~/2.
The resistance coefficient and the Nusselt number for curved pipes are obtained by putting ra = 4 or
m=5.
The N u ~ l t numbers obtained from measurement of the velocity and temperature distributions in the
air flow through the curved pipes of R/a = 40 and 18.7 are in good agreement with the theoretical results.

NOMENCLATURE - [ 2 a Q w ~ / k ( T w - Tin)];
A, w 1 at the centre of a cross section n. constant giving # in the boundary
perpendicular to the pipe axis; layer [equation (81)];
a', 01 at the centre of a cross section n o, constant giving g in a straight pipe
perpendicular to the pipe axis; [equation (61)] ;
a, radius of the pipe; P. - (a21v2)(plp);
C. coefficient == - (aP/HaO); Pr. Prandtl number. - pcpv/k;
Cp, specific heat of fluid at constant p. pressure;
pressure; Q,. Q~. heat flux in the fluid;
D. dimensionless velocity of the second- Qw. heat flux at the wall to the fluid per
ary flow in the flow core; unit area and unit time ;
f Orl, -
=- (a2/v2)(zoCl/p);
q.,
q,,
- Q~/kt;
fo,. - Q,/kt;
G. - T~" T; qw. - Qw/kt;
g. - G/ta ; R. radius of curvature of the pipe axis;
gin, =- (T~ - Tm)/ta; Re, Reynolds number. - ( 2 a W ~ v ) ;
# in a straight pipe expressed by the r, co-ordinate in radial direction in
1~no power law; the cross section;
k~ heat conductivity of fluid; T, time-averaged temperature;
M, exponent of w16 [equation (18)] ; Tin, mixed mean fluid temperature;
m, constant giving exponent o f R e in the Tw, wall temperature;
formula for ;t, [equation (9)] or Nu~ U, radial component of time-averaged
[equation (68)] ; velocity;
Nu, Nusselt number, u, - Ua/v;
37
38 YASUO MORI and W A T A R U NAKAYAMA

V, circumferential component of time- z0,, r0o, shear stresses in the direction of

averaged velocity; pipe axis ;
V, -- Va/v; ~,, circumferential co-ordinate in the
U$ , circumferential component of dimen- cross section.
sionless friction velocity ;
W, axial component of velocity; Subscripts
W, =- Wa/v ; 1, value in the flow core region;
w., mean velocity; b, value at the boundary between buffer
W*~ axial component of dimensionless layer and turbulent core in a straight
friction velocity; pipe;
if, dimensionless velocity in a straight c, curved pipe ;
pipe or resultant velocity formed by f, value at the position where v in the
v and w; boundary layer becomes maximum;
dimensionless friction velocity, m, mean value around the periphery
=- (a/v) x/(rw/p); (~b = - ~ ,-~ n), (except gin, T,., W,.);
Y, coefficient given by equation (67). S, straight pipe;
6, value at ¢ = 6.
Greek symbols
c~, proportional coefficient in the form- INTRODUCTION
ula for 2s [equation (9)] ; IN A CURVED pipe, the centrifugal force of a
~c, proportional coefficient in the form- flowing fluid produces a pressure gradient in a
ula for 2c [equation (36)] ; cross section. This pressure gradient yields
fl, proportional coefficient in the form- a pair of secondary flows. The secondary flows
ula for Nus [equation (68)-] ; cause a larger amount of pressure drop or heat-
7o, constant in equation (54); transfer rate than those for a straight pipe. To
A, coefficient of correction term, evaluate the effect of curvature on the heat-
[Suffix (A', Or,, Nu, 2) denotes the transfer rate is of fundamental importance for
correction term for them] ; the design of heating or cooling coils and for
6, thickness of the boundary layer other industrial requirements.
divided by a ; In the previous paper [1], heat transfer in
6,., coefficient in equation (34) giving a fully developed laminar flow is discussed.
6,.; The velocity and temperature distributions in
~, eddy diffusivity; the curved pipe are measured. These profiles
H, =- R / a ; show that a thin boundary layer is formed along
rl, = r/a; the pipe wall, Theoretical analysis is done by
0, axial co-ordinate; dividing the flow and temperature field into
~c, exponent of Pr in the formula for a core region and the Boundary layer. The
Nu~ [equation (68)] ; formulae of Nusselt numbers applicable in a
2, resistance coefficient, fairly wide range of Dean numbers [K -= Re
=_ [(-OPIROO)(2a/½pW2)] ; x/(a/R)] are suggested.
v, kinematic viscosity; For a turbulent flow, the empirical formula of
~, =l-q; Jeschke [2] has been quoted in many books [3].
p, density ; Others can be found in the reports of Pratt [4],
z, temperature gradient along the pipe Seban and McLaughlin [5], Rogers and Mayhew
axis ; [6]. These results have been derived from
zw, frictional stress at the wall ; experimental data utilizing a few cases of curva-
FORCED CONVECTIVE H E A T T R A N S F E R IN C U R V E D P I P E S 39

ture to pipe radius ratio. The fluid is air or ments were made to investigate the velocity and
water. The heat-transfer rate is obtained by the temperature distributions in air flow. The
measurements of the bulk mean temperature Nusselt numbers were also obtained as the
at the fixed two stations and the total heat flux experimental results.
from the pipe wall to the fluid between the
stations. THEORETICAL ANALYSIS OF THE FLOW IN
This method is conventional for heat-transfer A CURVED PIPE
experiments in pipes or conduits, but provides In the present paper, the far side of the wall
little information about heat-transfer processes from the center of curvature is called the outer
inside the fluid. The effect of curvature is far wall, and the near side is called the inner wall.
less evident in a turbulent flow than in a laminar In a curved pipe, the fluid in the central part is
flow. Therefore, in order to establish a general driven toward the outer wall by centrifugal
correlation, an amount of data from different force. The fluid near the wall flows along the
cases ofradius ratio and a high degree of accuracy wall surface to the inner wall. Thus in the pipe
in experiments are required. It is thought that the regular secondary flow forms a pair of
there should be a shorter way to approach the vortices in a cross section. When the balance of
general formula of Nusselt numbers. The most forces in the direction of the pipe axis is con-
effective way is to examine the velocity and sidered, it is noticed that stress caused by the
temperature distributions inside a pipe theoretic- secondary flow shares with other kinds of
ally and experimentally, and show how the stress the role of counter force to the pressure
Nusselt number is affected by curvature through gradient.
the distortion of distribution profiles. Up to The system of co-ordinates is taken as shown
the present, theoretical study of the problem in Fig. 1. The shear stresses in the direction of
has scarcely been made.
The only theoretical analysis of the flow
resistance in curved pipes was done by Ito [7].
The boundary layer concept is employed. The
R
resistance formula is given in an asymptotic
form which gets nearer to the empirical cor-
relation in the region of large Reynolds numbers,
like Adler's analysis for laminar flow [8]. ,~neor stresses in the direction
of ~ exerted tAoon o small element
However, there is a difference between the of the fluid

theoretical results and the experimental results FlG. 1. System of co-ordinates.

over the practical range of Reynolds numbers.
The procedure of calculation is so complex that
it does not fit the purpose of applying the the axis (0) are %,, %,. All quantities are ex-
analysis of the flow field to the heat-transfer pressed in non-dimensional form, as follows:
problem.
In this report, variations of physical properties ~l = r/a, H = R/a, u = Ua/v,
with temperature are not taken into considera-
v = Va/v, w = Wa/v, P = a2 P
tion. The theoretical analysis for a turbulent
flow in curved pipes is made in a manner which
is rather simple and suitable for the discussion Y" = k--e)' ¢'*=
of heat transfer. Heat transfer in a fully developed
temperature field on the condition of constant When the flow is steady and fully developed, the
heat flux is discussed theoretically. The experi- pressure gradient along the pipe axis is constant.
40 Y A S U O M O R I and W A T A R U NAKAYAMA

The constant in a dimensionless form is written

as C =- -(OP/HOO). It is assumed that r / R ~ 1,
and all geometrical small terms are omitted.
Then, the force balance equation is expressed as
follows :

c3foq,
8 (tlfo,) + - C. (1)
o

The shear stresses are

FIG. 2. Secondary flow streamlines.

g3W
fo, = O--q- uw - U'W' }
thickness of the layer of diffusion is invariable
(2)
in 0-direction, and varies only in a cross section.
OW
fo~, - r l O O vw -
V'W'.
i. T h e velocity distributions in the core region
In this equation u, v, w are the time-averaged In this region the velocity components u, v,
mean-velocity components; u', v', w' are the w, are denoted by Ul, Vl, Wl respectively. The
components of turbulent fluctuation; and u'w', shear stresses are expressed by
v'w' are Reynolds stresses.
When the flow is turbulent, a mean velocity fo. = - u l w l , Jo• = - v l w l . (3)
is always large enough to produce the secondary The relations between the centrifugal force
flow which shows a remarkable effect in of fluid and the pressure gradient in a cross
determining the m a i n feature of the velocity section are
distribution. Only the shear stresses uw, vw are
2 8P 2 8P
supposed to be predominant over a cross wi sin qJ - (4)
H cos O = ~q, H r/t?0"
section of the pipe except a thin layer along the
wall. Hereafter, the region where the stresses Elimination of the pressure terms from these
caused by the secondary flow are predominant equations yields the following relation for wt :
is called a core region. The thin layer next to
the wall where all kinds of stresses cannot be 8wl 8wl
cos ~k~ + sin qJ ~ = 0. (5)
ignored is called a boundary layer. The region
of the boundary layer is represented by 6 which The equation of continuity is written as
is the dimensionless thickness divided by the
pipe radius a. 8 8vl
When the secondary flow is brought into r/O~ (r/u,) + q--~ = 0. (6)
consideration, its appearance is imagined as
The velocity components Ul, vx, w~ satisfying
follows. The fluid in the core region flows
equations (5) and (6) are put in the most simple
towards the outer wall, then enters the boundary
forms as follows :
layer. There is a return flow toward the inner
wall in the layer. The conceivable stream lines ul = Dcos~O ]
are shown in Fig. 2. In these circumstances,
vl = - D s i n ~ k I (7)
the work done by pressure in the direction of the
w 1 = A + (C/D) rl cos
pipe axis (which is called 0-direction) is lost by
turbulent and laminar diffusion mainly in the where A and D are constants. The secondary
boundary layer. When the pressure gradient flow in the core region is expressed by a uniform
in 0-direction is constant, 6 representing the flow toward the outer wall.
FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 41

2. The shear stress at the wall and the velocity

~': x ,lw:
distributions in the boundary layer
The distance from the wall is denoted by
= 1 - ~/. As shown in equation (7), the velocity
at the edge of the boundary layer (¢ = 5) varies
with the angle qJ in a cross section. The charac-
teristics of the flow in the boundary layer are
determined by the velocity at ~ = 5. Since the
local shear stress at the wall depends on the
velocity in the boundary layer, zw also varies
with ft. The shear stress % is expressed non- ¢1_ ¢: 4 v: ~--
dimensionally by the friction velocity # * =
Fro. 3. Velocitycomponentsand resultant velocitynear the
(a/v) x/(zw/p). pipe wall.
The circumferential component (~h-compon-
ant) of the velocity in the boundary layer v has
a distribution like that shown in Fig. 2. The after Blasius's resistance formula which is valid
distance from the wall where v becomes the for Reynolds numbers less than l0 s .
maximum value v: is denoted by ~:. When m is taken as 5, the value of a is 0"184.
The axial velocity component w at ~ = ~: is A form of the velocity distribution correspond-
written as w:. The components vy and w / f o r m ing to equation (9) may be deduced by means
a velocity fff as shown in Fig. 3. The relation of plausible assumptions [9]. As the result,
between if*, ey and i f / m i g h t be considered to the velocity distribution is expressed by the
be obtained from the law which describes the 1 / ( 2 m - 1) power law. The Reynolds number
relation in a straight pipe between the shear Re in equation (9) is calculated by using this
stress at the wall, the distance from the wall and velocity distribution. On the other hand, the
the velocity at the position. The law in regard total shear stress at the wall keeps balance with
to a straight pipe may be derived from the the pressure gradient. Then, from equation
resistance formula. The resistance coefficient (8) 2 is expressed by the stress at the wall zw and
2 is defined as Re. We can now deduce % from equation (9).
In a non-dimensional form this is written as
2= (-~.~ 2a (8)
1
k ~z/ ~.p W ,,2 if,2 (a~(zw~_ a
~,,V2) k P ) 23+(1]m'
where z is the co-ordinate taken along the pipe
axis. We use the suffix s to indicate that the case F!2m-
× [m-(4tn -- 1 ) . ] ~v~Zm-,),m
~;(,/,,) (10)
is one of a straight pipe. The formula for a
straight pipe is expressed in general terms as In order to make the analysis for a curved
2s = a Re -"/") (9) pipe on the basis of equation (10), the velocity
component w in the boundary layer is put in
where
the following form so as to satisfy the boundary
Re = 2aW,,,/v. conditions:
at
For the appropriate range of Reynolds num-
~=0, w=0,
bers, ~ and m are determined in order to approxi-
and at
mate the actual relation between 2s and Re.
The commonly used value of m is 4 or 5. = ~, w = w16.

When m is taken as 4, the value of ~t is 0.316 w = w ~ 6 ( U ~ ) 1 / ( ~ ' - ~) (11)

42 YASUO MORI and WATARU NAKAYAMA

The conditions for v are the boundary condi- in the circumferential direction 0p-direction)
tions and the condition of continuity. The is given by
boundary conditions are
I),2 Vf ¢v,2 ~m
at w: 21+('/")(2m- 1)
~=0, v=O, r(2m - 1)2](2m-1)/m
and at
~=3, V = V l. w(~-l)/"3-("+l)/"Dsin~O. (16)
In consideration of the flow field in a cross From equation (7)
section shown in Fig. 2, the flow rate of the
secondary flow through the plane B-O is to be wlo = A + (C/D)(1 - 3)cos ~b. (17)
equal to that through the plane A-B. The It is assumed that C/D is sufficiently small
condition of continuity is compared with A, and 3 ,~ 1. Let the power of
wl~ in equations (15) and (16) put in a general
v d~ = D(1 - 3) sin ~b. (12) symbol M and, the following expansion is
0
made in order to simplify the later calculation,
So as to satisfy these conditions and follow and only the first and second terms in the
the 1 / ( 2 m - 1) power law near the wall, v is bracket are taken into account.
written as
w~=aU[1 +M(C/AO)cos~k+...]. (18)
v = -Osin~, m- 1 - The results obtained finally show that the
ratio C/AD is, for example, about 0.4, when
+ 1 ( m 2 - 1) ( ~ ) ] (13) m = 4. In this case, M in equation (15) is ¼.
m-1 The third term in the bracket of equation (18)
becomes
From equation (13) v: and ~: are obtained.
Then, w: is obtained from equation (I I). ½ x ¼ x ~C/AD) 2 cos 2 ~ - 0"1 cos 2 ~k.
A friction velocity component in the direction
This is reduced to 0"05, when the qJ-averaged
of the pipe axis (0-direction) is
effect is considered. Moreover, ( 1 - 6) 2 is
w. 2 = w£ k,2 (14) multiplied to this term in the strict expansion.
Wf These reasons suggest that the elimination of
terms after the third is equivalent to the omission
where of only a magnitude of a few per cent.
,: = + w}). Since both the velocity distribution in the
boundary layer and that in the core region
Supposing that v ,~ w, we may write that are established, the mean velocity in the pipe
~ : ~ w:. With the aid of eq~uation (11), ~,: and may be calculated. The dimensionless mean
4: in equation (10) are replaced by w16 and 3. velocity is
Then, equation (14) becomes
;t 1--~
w, 2 ~ [ ( 2 m - 1)2] (2m-l)/" w,. = (Re~2) = (l/n) { S ~ w:/dr/d~9
--rt 0
= f i L m ( 4 m - 1)J ~t 6
+ ~ Sw(1 -- ~)d~d~0}. (19)
w(~ "- ')/" 3 -('"). (15) --~0

In like manner a friction velocity component The variation of 3 with ~0 is supposed to be

FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 43

very small. Hence, 6 is replaced by its mean. taken into account, and 6 is replaced by 6m.
value 6,.. Equation (24) becomes
Substitution of equations (7) and (11) into W. 2 = E + F cos ~, (25)
equation (19) yields the relation between A and where
8 m.
t 6m - 1 cos2
Re 1 E= 2 m + 1)(4m- 1)
A = 2 1 - (1/m)8,. + [l/(4m - 1)]62. (20)
4m(2m - 1) 1)sin2@1 C, (26)
Now we consider the region bounded by the + (2m + 1)(4m -
pipe wall and two cross sections, apart by a 6m - I
distance R dO. The equation describing balance F = (2m + 1)(4m - 1)AD. (27)
of forces exerted upon this portion of fluid is
The mean value of equation (25) around

ff
-nO
O-~ordrd@= %ad@.
-n
i
tReperiptlm'y(~ = - n ~ n) is C/2 and satisfies
(21")~ equation (22). The variation of E with ff is
small compared with F cos~, so that E is
Use of the dimensionless quantities gives the taken as constant and replaced by its mean
following relation between C and w.2 averaged value C/2.
around the periphery. On the other hand, w.2 is written in the
following form from equation (15)
C = 2w .2. (22) C ~ ( 2 m - 1) F(2m- 1)2] '2r'-l)/=
Equations (18) and (20) are substituted into w*2 = 2- + 23+---71/-~)m Lm(am - 1).J
equation (15). Then, substitution of equation
1
(15) into equation (22) gives x a (r'-''/m C ~ cos $ (28)
[ ( 2 m - 1)2] (2m-1)/m where the expansion (18) is used.
C = --~ L - ~ m -__-~] Ret2m-X)/m
Since w.2 of equation (25) should be identical
with that of equation (28), F must be equal to
X 8~-n(l/m) [1 + 2 mm2
-16m 1 . (23) the coefficient of cos ~ in equation (28). When
the largest components of A and C in equations
Here the terms having magnitudes of order
(20) and (23) are substituted, the following
less than 82 are neglected.
relation between D and 6= is found.
3. The boundary-layer momentum equations D26tm2/m) = ~t2(4m 2 - 1)(4m - 1)
The unknown quantities are now reduced 2Vm(6m- 1)
to D and 8. They are determined by solving F ( 2 m - 1)2][2(2m- 1)]/"
the momentum equations of the boundary X [~-~-m~__-~] R e I2(m-l)l/m. (29)
layer in the '-a~ial and circumferential directions.
The other relation between D and 6 is
The integrali~luation expressing the equilib-
obtained from the integral equation of momen-
rium of momentum in the direction of the pipe
tum in the circumferenfiat direction
axis (0) is
v* z - ~,~P~d~ sinH ~ / f w2d~
W* 2 = w l a ~ - ~ f v d { - f - - ~ f v w d , + C 8 . (24) o o
o o 6 6
0
Equations (7), (11) and (13) are substituted o@Ii 2 d -b v,;fod¢ (30)
into equation (24). Only the largest terms are o o
44 YASUO MORI and WATARU NAKAYAMA

and the one in the radial direction (q-direction). Equating these terms, we have
1-6
D6m[(2ra+ 1)/m] _ 2(2m - 1)
p~ = '~1 cos Ip dr# - CHO. (31) ctm(2m + 1)
H
0 x Pm(4m --- ')-I `2"~',/'' l
Equation (31) is substituted into equation (30). L(2m _ 1)eJ Re(,.+ 1)/,. --H" (32)
The pressure gradient in equation (8) that
defines the resistance coefficient 2 is expressed From equations (29) and (32), D and 6,. are
non-dimensionally as C given by equation (23). obtained as follows.
Therefore, for the purpose of obtaining 2 it D = D R e "1("+ i) H - [ l I 2 ( " + i)] (33)
is necessary to know only the mean value of
thickness 6. The details about the variation where
of 6 with ~b are not required, if 6,. can be de- 1
termined without integration of 6 with respect log/) - m + 1 [¼{(2m - 1)log(2m + 1)
to ~. Hence, a relation between D and 6,. is
+ (18m - 5) log(2m - l) - (10m - 1)logm
derived from equations (30) and (31) in the
following way. -(6m- 5)log(4m- 1)-(2m+ l)
Equations (7), (11), (13) and (16) are substituted x log (6m - 1) - (14m + 5) log 2}
into equations (30) and (31), where the friction + m log ~] (33')
velocity v.2 is written with the aid of equation
(18). When the integrations are carried out, r~
= ,$,,,Re [l/(m+l)lH"/[2(m+l)] (34)
both the right-hand side of equation (30) and
where
v.2 are expressed in the form E' sin qs + F'(cos qs)
sin qs, where E' is constant and F'(cos qs) is a
log ~,~ - m +
1 1 {¼(3mlog(2m + 1)
function of cos ~s. The terms reduced to the
symbol E' relate to the basic relation between
+ (15m - 8)log(2m - 1) - (7m - 4)
the wall shear stress and the driving force of
the secondary flow. The equation (30) is under- × [logm + log(4m - 1)] - mlog(6m - 1)
stood to consist of the terms E' and the terms - 9m log 2) + m log ct}. (34')
F'(cos qs) which represent the local deviation of
momentum flux from the mean value. The 4. The resistance coefficient
terms F'(cos ~s) are supposed to determine the The definition of the resistance coefficient
deviation of 6 from 6,.. In order to obtain 6,,, for a curved pipe is given by equation (8) when
the terms E are considered. They appear such as R d0 is put in place of c~Z. By using the non-
dimensional quantities we write
/9,2 ~ [ ( 2 m - 1)2] (2m-1)/"
--+ 4(2m - 1)Lm-~m 2_- i).] 16
2< = ~ C. (35)
x D6m [(m+ 1)/"1 Re(,. - l)lm
When equation (23), in which 6, is given by
6 equation (34), is substituted into equation (35),
r c~,. R e 2
the following formula for 2c is obtained.
0
6 2< 7Ia) = ~c 1/(,. + 1)
[Re (a/R) (m/z)]
sin~ ~ 2 2m- 1 Re 2
~- j w de ~ 4(2m + 1) 6,,, f t . A,~
0
X [1 q-[Re(a/R)(rn/2)]l/("+l t (36)
FORCED CONVECTIVE HEAT TRANSFER IN C U R V E D PIPES 45

where power law, and obtains the result

1 2, x/(R/a) = 0.29~[Re (a/R)2] ~.
log ~t, - m+l-- {¼( - 3 log (2m + 1)
This formula gives the value 7-8 per cent
+ (16m - 7)log (2m - 1) - (8m - 3) less than that from his empirical formula. The
x [log m + log (4m - 1)] + log (6m - 1) discrepancy is caused by lack of reasonable
+ 9 log 2) + m log ct} (37) consideration of velocity distribution in the
boundary layer in Ito's analysis. The latter
1 term in the bracket of equation (40) or (23) is
log Aa - - - {¼(3m log (2m + 1)
m+l due to the term of (6.,) in equation (20).
- (15m + 4)log m + (19m - 4)log (2m - 1) If only w x is used to calculate the mean
velocity, it is found that A = Re~2.
- (7m - 4) l o g ( 4 m - 1) - m log (6m - 1)
In that case, the velocity profile of the core
- 9m log 2 ) + m log ~}. (38) region is assumed to extend over the whole
The formulae for 2, derived from the com- cross section and the presence of the boundary
monly used formulae for 2s in the form of layer is ignored. In Ito's analysis [T], the
equation (9) are shown in the following para- relations between the unknown quantities are
graphs. not given in simple forms, and the complicated
According to the Blasius's formula 2., = procedure of calculation makes it difficult to
0.316 Re -~-, we put m = 4, and for this case we obtain 2c in the form of equation (40).
denote 2~ and 2, by 2~4 and 2~4 respectively. In order to obtain the formula for fairly
Since ~t = 0.316, equation (10) now becomes large Reynolds numbers we put m = 5. The
formula for a straight pipe is 2s = 0"184 Re -~.
if,2 = 0-0233 if} ~i ~. (39) We denote 2 s and 2, by 2s5 and 2,5 respectively.
The equation (36) becomes
It may be seen that the expression (39) is
in good agreement with experimental data, if 2c5 x/(a) - 0'192
the coefficient 0"0233 is modified as 0.0225. [Re (a/R)2"s] ~
Therefore, the value of ~ equivalent to the 0-068
modified coefficient, that is 0"305, had better x 1 + [Re(a/R)2.5]tj. (41)
be used in the calculation after equation (10).
Hence, in equations (37) and (38), by putting A further remark should be written about
at = 0-305. Equation (36) becomes equation (36). For laminar flow we put m = 1.
As reported in the previous paper [1], 2, x/(R/a)
2,4 /(J'~ _ 0"300 becomes a function of Dean number
[Re (a/R)2] t
[Re x/(a/R)].
0-112
x {1+ [Re(a/R)2]~,}. (40) We denote 2c by 2~x in the laminar region.
From equation (36) it is found that 2, is
This formula agrees well with Ito's empirical almost proportional to (a/R) ~/2~'+1). As the
formula value ofm increases, the difference in radius ratio
2, ~/(R/a) = 0"029 + 0.304 [Re (a/R)2] -°25 h a s still less influence on 2,. The tendency is
shown in Fig. 4. In the present discussion, as a
over the latter's application range 300 > matter of course, eases in which the radius
Re (a/R) 2 >0.034. ratio is so large that the state of flow is not so
Ito [7] also made the analysis using the -~ different as the one in a straight pipe and little
46 YASUO MORI and W A T A R U NAKAYAMA

importance in engineering applications is found, found that Re's- 5.0 x 104. Since Re'> Re~,
are excluded. equation (43) is not available for R/a larger than
The bend of a curve indicating the transition 170. For such large R/a, it is recommended to
from laminar to turbulent flow is written use Re~ as the lower limit for 2c5. However,

Xc= , R / o = 20

• ~Xcl ,R/o=I00
o,
Xc~ R / o = 2 0
k:5, R/o =20
X = 100
Xsl=64 Re-I"N,,j ~ ' ~
I
Recr
(=2300) Xs4= 0"316 Re -I/4
[Link]
Xc5=0.184 Re-I/5

io 3 tO4 10 5 106

Re
FIG. 4. 2-Re diagram.

according to the Ito's empirical formula for because of little difference between 2~4 and 2c5
critical Reynolds numbers [7]. in the slope of the curves against Re, the avail-
Re¢, = 2 x lO4 (a/R) °'32. (42) ability of 2¢4 would fail distinctly at Reynolds
numbers far larger than Re' given by equation
The Reynolds number at which 2¢4 comes to (43). Usually, it is difficult to make Reynolds
exceed 2¢1 increases, as R/a decreases. This fact number so large in curved pipes. Owing to
seems to imply the same inclination of Re¢, this reason the experimental data obtained up
as shown in equation (42). to the present by many investigators are in
Similar criterion are written for the Reynolds good agreement with 2¢4. Therefore, equation
number Re' which dermes the lower limit of (40) is convenient enough for practical use.
the application range of 2¢5. The real limit is To illustrate the increase of flow resistance
not obvious because of the continuous change in curved pipes resistance coefficient ratio,
of actual 2¢. Therefore, we define it roughly as 2J2s, is shown in Fig. 5, 2c4 and 254 being used.
the Reynolds number at which ;~c5 comes to
exceed 2c4. From equations (40) and (41) it is T U R B U L E N T H E A T T R A N S F E R IN A C U R V E D
found that PIPE
In the present analysis, the condition of
R e ' = 6"5 x 105 ~/(a/R). (43) uniform heat flux is defined as meaning that the
increasing or decreasing rate of total heat flux
If, for a straight pipe, the,Reynolds number through a cross section is constant along the
determined in a similar way is-taken to be the pipe axis. Since the temperature distribution
criticalpoint for the choice of 254 or 2~5, it is is not symmetrical like that in a straight pipe,
FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 47

1"3
/
I
I
f ~ I
,~ 1"2

I
f
H
/
I
I'0
4 6 8 I 2 4 6 s I0 2 • 4 6 8 I00 2
Re (o/,q )z
FIG. 5. 2j2,.

local heat flux from the wall to the fluid may be When r / R is so small as assumed in the flow
a function of ~k. This is the same as saying that field analysis, the energy equation becomes
the ~b-averaged heat flux at the wall is invariable
with the distance along the pipe axis (RO).
For fully developed flow under this condition
the time-averaged temperature T can be ex-
q•l (tlq") Oqq, = P r w. (46)

Heat fluxes are

pressed in the form
T = zRO - G(r, d/)
where z is a constant temperature gradient
(44) "
q" = - -Oq + P r (u9 + u'o')

dO
} (47)
along the pipe axis, and G(r, ~) is a function of q~' = - ,tg-~ + P r (vo + v'o')
r and qL
According to Seban and McLaughlin [5],
where u'o', v'q' are heat fluxes due to turbulent
there is a temperature variation around the
fluctuation.
periphery of a cross section. However, except
The temperature distribution in a pipe is
for fluids having Prandtl numbers considerably
considered to be determined mainly by the
less than unity, the Nusselt number for fully
effect of the secondary flow like the velocity
developed turbulent flow is hardly affected by
distribution. The contribution of the secondary
the wall temperature condition. This is ascer-
flow of heat transfer is supposed to be pre-
tained by the results for a straight pipe, for
dominant in the greater part of a cross section.
example the one given by Seban and Shimazaki
[10]. The present purpose is to obtain the Accordingly, only the terms uo, vo are taken
into account in the flow core region.
Nusselt number averaged around the periphery.
Therefore, in the following analysis the wall
1. T h e temperature f i e l d in the core region
temperature is assumed to be constant with
When g in the core region is denoted by g l,
respect to ~. Hence, we put
the heat fluxes are
T w = ~RO. (45)
qn = P r u191, q , = P r va01. (48)
Let Q,, Q, be a heat flux in the r-direction
and the ~k-direction respectively, and write Substituting these into equation (46), we find
Q, Q, 601
q~ = k z ' qq' = --~z ' u1 + vl q--~ = wa. (49)

G The dimensionless-temperature 9a is written

Ta in the following form so as to satisfy equation
48 YASUO MORt and W A T A R U N A K A Y A M A

(49) in which ul, vb and wl from equation (7) where e is the eddy diffusivity taken equally
are substituted. for heat and momentum.
91 = A' + (C/2D2)rl 2 + (A/D)rlcos~9 (50) The assumption is made that the ratio f / q
remains constant in the transverse direction 3.
where A' is a constant. This assumption is generally used in the
The constant A' is determined by considering elementary theories of turbulent convection.
total heat balance about a cylindrical slice of the Thus, from equations (51) and (52) the ratio of
fluid occupying a cross section of the pipe. It is qw to the friction velocity is written in the form
necessary to know beforehand the relation which
creates the local heat flux from the wall to the qw v + Pr e dg
fluid. ~,,z- v + e dr" (53)

Near the wall, the coefficient (v + Pr e)/(v + e).

2. The heat f l u x at the wall
may be expressed mainly as a function of
When the heat flux at the wall is denoted by
Prandtl numbers. We divide the flow field into
Qw, the non-dimensional heat flux is given by
three regions as usual. The regions where the
qw = Q~/kz. Interest is now concentrated on
effect of the Prandtl number is important are
the small local portion of the fluid adjacent to
the laminar sublayer and the buffer layer. In
the wall. The mechanism of heat transfer in
both regions, we reduce the coefficient to the
that place is supposed to be the same as that
following form
observed in a straight pipe. In other words,
when the velocity and the temperature at a v+ Pre
_ _ _ p r 1 - ~o (54)
given distance from the wall are specified, v-bE
Qw is determined by the law applicable to that
existing near the wall in a straight pipe. Thus The exponent (1 - 7o) would be chosen for
the relation between the non-dimensional vari- the appropriate range of Prandtl and Reynolds
ables, qw, 3, ff and g is derived from the Nusselt- numbers. Hence, equation (53) becomes
number formula for a straight pipe.
qw = prl_~o__dg (55)
The formula is usually based on the mixed ~,2 dr"
mean temperature. In order to obtain the
local relation, the mixed mean temperature In the turbulent core region, where the effect
should be disintegrated into the velocity and of turbulent fluctuation is predominant, equa-
the temperature. The expression for the tempera- tion (53) becomes
ture distribution in a straight pipe is found in qw - Pr do.
the theoretical paper by Martinelli [11]. How- (56)
~,2 dff
ever, it is not suitable for the present analysis
owing to the rather complicated structure. Suppose that the interface between the tur-
Hence, we begin with the discussion of a bulent core and the buffer layer exists at ~ = ~b,
straight pipe. and denote g, ~ at the interface by gb, ~'h re-
The Boussinesq expressions for the shearing spectively. The integration of equation (55)
stress and the heat flux are, in the dimensionless yields
forms,
qw = p r l - r o ~_. (57)
if,2 wl'
f = (1 + ev)dd~
~ (51)
From equation (56)
_ E d0
q= 1 + rr-l-z-; (52)
v/a¢ g--gb=~r -- .
FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 49

When the 1 / ( 2 m - l) power law is used, the The Nusselt number is defined by
velocity distribution is expressed by
Nu = 2aQw = 2qw (64)
(if/fib) = (~/~b) 1/t2m- 1) (59) k ( T w - Tin) g=
Equation (59) is substituted into equation where g= is the dimensionless difference between
(58) and elimination of (q~/w .2) from equations Tw and, the mixed mean temperature, Tin. When
(57) and (58) yields g in the turbulent core. the suffix s is used to denote the case of a
straight pipe, g,~ is given by
g = gb{Pr-r°[(~/~b) ~/~2=-~)- 1] + 1}. (60) 1
For the simplicity in calculation, equation g.,~ = (4~Re) S g~' r/d~/. (65)
0
(60) is approximated by
This is calculated by substitution of equation
= gb (~/¢b) (1/n°) (61) (61) and ff following the 1/(2m - 1) power law.
Heat flux qw from equation (62), in which
where
equations (10) and (63) are substituted, and
no = P d °(2m - 1). g=s yield the following Nusselt number for a
straight pipe.
Comparison between the equation (60) and
(61) is shown in Fig. 6 and the above approxima- -- 1)2 (2m + 1 ) p r , Re(m_l)/m Y(66)
tion is found to be reasonable. N u s = ct( (2m4m(4 m _ l) 2
If we put ~ = 6 in equations (59) and (61),
it is found that where
6m+ 1
~vb = Cv~(~d6) ~/~2m- ", ~ = ~ (~d6) "~° pr-~O)
4m(2m + 1) (1
where the suffix 3 denotes the value at ~ = 6. 1
+
Substituting these into equation (57), we find 4m(2m + 1) 'l
qw = Prl-~° if,2 (~/ff~)(~b/6)(l/no)-tl/(2ra- 1)]. (62) × (1 - Pr-r°)~ 6 -tl/t2m-l)l(l-e'-~ ) (67)
When the temperature distribution bears a In fluids having Prandtl numbers different
close resemblance to the velocity distribution, from unity, the strict analogy between heat and
n o is nearly equal to 2m - 1, that is, the exponent momentum transfer, in the sense usually em-
of (~b/6) in equation (62) is near zero. Thus qw ployed in the elementary theories of turbulent
depends on the Prandtl number through Pr ~-ro convection, is unlikely to exist. The exponent of
alone. However, this discussion may be allowed Re in the Stanton number formula which is
in the very limited range of Prandtl numbers formed by dividing N u with Re Pr is not always
about unity. In order to obtain the results identical with that in the resistance formula.
applicable in the relatively wide range of When the same amount of qw is supposed to be
Prandtl numbers, we put given to various fluids, difference in the exponent
p r 1- ~ o (~b/6)(1/no)-[1/(2ra- 1)1 = ( p r ~. (63) is caused by the effect of Prandtl numbers on
the mixed mean temperature. The temperature
The left-band side of equation (63) is a distribution is more flattened as the Prandtl
function of Pr, 6 and Re which determines ~b. number increases. In the present analysis, this
The coefficient ~ and the exponent x are con- contribution of Prandtl numbers is expressed
stants which should be chosen so as to approxi- by Y.
mate the true function in the appropriate range The behaviour of Y against Pr ~° is shown in
of these variables. Fig. 7. It is found that use of large m lessens the
50 YASI!O MORI and WATARU NAKAYAM,\

>~
105 for temperature and velocity is considered.
~ bO0 This equivalent distribution follows the l/
o,e.~ Pr=l and P:,~=20-- (2m - 1) power law, The index m is chosen with
I ~0 I00
the coefficient ct from the resistance formula,
so as to give the real Nusselt number from
equation (66) without Y.
Below the abscissa of Fig. 7 an estimation
of P r against Pr ~° is shown. The exponent 7o
is determined to give the agreement between
0 from equation (61) and the distribution
0.8 I 2 4 6 8 I0 20 obtained by Martinelli [11] as shown in Fig. 8.
pr~
The values of 7o are
FIG. 6. g/0.
Pr = 1 ,-, 1'5, 7o---2,

effect of Prandtl numbers. The values of Pr > 1'5, ";o = 0"7.

represent the order of magnitude of boundary- The previous steps relating to heat transfer
layer thickness expected in a curved pipe. The in a straight pipe are now summarized and
case of 6 = 0"1 is found when R e (a/R) 2 - 100. applied to the analysis for a curved pipe. The
This is in general the case when R e is very large. dimensionless temperature difference 06 and
For such a large mean velocity region, choice the velocity w6 are replaced by 016 and wl6
of large m makes the curve of Y come near the respectively, where 916 is the value of 91 at
line of unity. Moreover, in the region of high = 6. When the Nusselt number for a straight
Prandtl numbers, if a large value of m is pre- pipe is given in the form
ferred regardless of the magnitude of Re, the
deviation of Y from unity can be diminished. Nus = fl R e (m- I)/,. p r ~ (68)
For fluids of Prandtl numbers about unity, the local heat flux at the wall qw in a curved pipe
the value of m should be the same as that in is expressed by
the resistance formula. For other fluids, it is
assumed that m in the Nusselt number formula qw = OwPr~w(~-l)/m 6-(l/") 91~ (69)
implies the effect of Prandti numbers, and the where
value different from that in the resistance
formula can be chosen in the Reynolds number [(2m _ 1):l~,,- 1)/m
qw = 2 -[~"+ 1)/=1
region under consideration. With this under-
standing we put Y = 1. As a matter of facL 4m - 1
this assumption is equivalent to the following. × --/3. (70)
2m+ 1
Instead of the real temperature and velocity
distribution, the one supposed to be the same In order to ascertain the relation, Martinelli's
results [11] are calculated numerically, and
rearranged into the forms of equations (68)
I-2 1m=4/
•~.0- I and (69) independently. In the region where the
'-m -51
~m=4 1 Prandtl number is near unity and the Blasius's
1.0 8 ,T-O-~
resistance formula is applicable, Nu~ and qw are
I 2 4 6 8 I0 20
pr~ written as follows :
, ,.~ " ,'o ~o 20 ,6o Nu~ = 0-038 R e ~ Pr ~ (71)
Pr
FiG. 7. Y. qw = 0"023 P r ~ w ~ 6 - ~'-91 a. (72)
FORCED CONVECTIVE HEAT TRANSFER IN C U R V E D PIPES 51

Pr =100 for this control volume is

I.O
f /t ff a

[Link] w a dO = 0 cp-~-~z W T r dr d 0. (74)

0-6 This is expressed in the following dimension-

l!
less form :
".'.o., qw" -
Re Pr
4 (75)
Martinelli [ll]Ne =104
0.2 where q," is the mean value of qw around the
I
periphery of a cross section.
ol
--~ ~ . 0.2 0.4 0.6 0.8 I0
From equation (73) qw,, is obtained

Buffer layer r/o qw,. = q., P r~ 6~, (x/") A ("- 1)/m

(Re= IO4)
C m-1
FiG. 8. Comparison of ~ with Martinelli's results [11]. x A'+~+ 2~

1- 2 m ( 6 m - 1)
On the other hand, when we start with × 1 ( 4 m 2 - 1 ) ( 4 m - 1)1 A}. (76)
equation (68), we put fl = 0"038, m = 4 and
x = ½ according to the equation (71). It is When A, C, D and fir. given respectively by
found that qw from equations (69) and (70) is equations (20), (23), (33) and (34) are substituted
the same as that of equation (72). into equation (76), A' is determined from equa-
From equation (50), gx~ is obtained by tion (75) as follows :
putting q = 1 - 6. Equation (17) and gl~ from 2'
equation (50) are substituted into equation a ' = (Pr I -~ - Aa, ) 1 + [(m - 1)/m 2] fir.
(69). When the expansion (18) is used, qw
x R e x/("+ x) H1/2("+ l) (77)
becomes
qw = (lw P r~ 6 -(1/") A ( " - 1)/" where
C m-1 4 m ( 4 m - 1 ) ( 6 m - 1)fl
x A'+-~-~+--m AA, = (4m 2 -- 1)2 a (78)

x r 2m(6m-1)ll~Acos2~O 1
log A' - - - {¼[(4m + 7)log (2m + 1)
[(4m 2 - 1 ) ( 4 m - m+l
+ (4m - 7) log m - log (6m - 1)
+ [_~( 2m(6m-1) )
k(4m 2 - 1 ) ( 4 m - 1)} - 11 log(4m - 1) - (8m - 15)log(2m - 1)
- 9 log 2] + log a} - log ft. (79)

t 4. T h e energy integral equation o f the boundary

layer
3. Determination o f A ' In the boundary layer, heat transfer due to
The portion of fluid bounded by the pipe viscous and turbulent diffusion exists with that
wall and two cross sections apart by a distance caused by the secondary flow. The effect of the
R dO is considered. The heat balance equation former is implied in qw. The energy integral
52 Y,\SUO MORI and ~x,,XI"\R[; NAK,\Y\M\

equation is written where

,} h b
L = 4 m 2 ( 4 m - 1)(6m- 1 ) f l p r ~1-.~
Pr gl~ vd~- gvd~ ÷ wd~. (80) ( 2 m - 1)3(2m + lJ 2
0 0 0
p~! _ - ) + . . 4m(4m
. . . - 1)2
We put the dimensionless temperature dis- x L m (2m -- 1 t2 (2m + 1)
tribution in the boundary layer as
g = glo(~/6) ~/" (81) x fl P r - I1 Kt (84)
O~

where n is the unknown exponent to be deter- When m, ~, fl and • are chosen from a Nusselt
mined from equation (80). formula for a straight pipe as shown in the
Equations (11), (13) and (81) are substituted next section, n can be determined. When we
into equation (80). When only the largest put P r = 1,
terms are considered, it is found that the sum
of convection terms due to v is equal to qw/Pr. n = 2 m - 1. (85)
The contribution of w is neglected, because it Since the assumption which allows m to
has an order of magnitude less than that of the imply the effect of Prandtl numbers is intro-
other terms. Performing the integrations in duced in Section 2, the value of n for another
equation (80) yields P r seems to lack a reasonable ground. Rather,
equation (85) is interpreted as the exponent
qw= Pr os20 m- 1 which gives the equivalent distribution con-
sidered in Section 2. Hence, we use equation
(85) in calculation of the mixed mean tempera-
l/n) + [1/(2m - 1)] + 1 (l/n) + ture, T,,. It should be noted that rough estima-
tion of T,, is obtained by extending gl and Wl
× (cos2 @ - sin2 qJ + Pr D ' + over an entire cross section. When T,, is calcu-
lated more precisely, the distribution of g in
the boundary layer is used to modify gl near
),( -- __ the pipe wall. Thus, the exponent n forms a
m - i 1/n) + [1/(2m - 1)] correction term in the Nusselt number formula.
Consequently, the variation of n does not have
1 _']l cos ~. (82t a remarkable effect on Nusselt numbers.
(l/n) +
We find that the mean value of q,~ from equa- 5. N u s s e l t numbers
tion (82) satisfies the relation given by equation In equation (64), the dimensionless mean
(75). Similar to the analysis of flow field, the heat flux q,,. given by equation (75) is put in
terms in the first bracket of the right-hand side place of qw. Then, the Nusselt number in a
are replaced by the mean value of qw. Equating curved pipe N u c is written,
the coefficient of cos ~9 from equations (73) and 2qw,. Re Pr
(82), we find Nuc - - (86)
g,. 2g,.
1 3m- 1 where
n 2m- 1 n 1 6

[(3m-- lX] 2 4m LL]~ g" -- ~ R e g l w l ~/drl dO

+ L\2m - I] + 2rn 2 1 1 - (83)
FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 53

n 6 the applicability of equation (91) obtained by

putting m = 4 remains well except for extremely
large Reynolds number. Equation (91) seems to
be available for practical use.
Using all the knowledge obtained about the Most of the experimental data for water are
temperature and velocity field, we may calcu- arranged into the form Nuc Pr-0.4 [4-6]. This
late equation (87) by the order of magnitude 6. form is introduced from the following con-
Equation (86)becomes ventional formula for a straight pipe.
1 Pr Nu~ = 0"023 Re °'s Pr °'4. (92)
Nu~ = 24' Pr 1-~ - Ag,. Rem/~"+ U
If the derivation of Nuc is based on equation
× ¢) 1/I2~,,+1~1 {, 4 [Re (aiR)m/z] 1/~"+ 1,t,88, (92),
fl = 0.023, m = 5, tc = 0.4.
where A' is given by equation (79). and
When these values are used, Pr x -~/(Pr 1 -~ -
4 ( 6 m - 1 ) ( 4 m - 1 ) m ( m - 1)fl AO,,) in equation (88) comes near unity for large
A9,, = (2m + 1)2 ( 2 m - 1)3 (89)
Pr as shown in Fig. 10. Therefore, in the range
of large Prandtl numbers, it is convenient to
4m 2 - 2m - 1 ^
AN" = m2(2m + i) 6,.. (90) put
p r ~- ~
The coefficient 6,, is obtained from equation (34'). 1 (93)
prX-~ _ A9, .
If we take equation (71), we have
for the simplicity of calculation, and for direct
fl = 0.038, x = ½, m = 4.
comparison with the empirical formulae sug-
The equivalent resistance formula is 2~4 gested by some experimenters [4-6].
given by Blasius. The modified value of coef- It must be noted that this approximation
ficient, a = 0.305, is employed as in the deriva- more or less compensates for the over estima-
tion from the resistance formula to calculate tion caused by putting Y -- 1 in equation (66).
Nu¢. In this case equation (88) becomes In the derivation of Nus from qw, assumption
of Y = 1 causes underestimation of Nus. It is
Pr
conversely written that qw given by equation
Nuc = 26-2(Pr ~ - 0.074)
(69) may be larger than the value equivalent
(R)~ { - 0'098 ~ to Nus from equation (68), and gives a slightly
x Re~ 1 + [Re(a/R)Z]~j. (91) higher value of Nu~. Thus, we may write the
Nusselt number formula for liquids which have
This formula agrees well with the experi- in general high Prandtl numbers as follows:
mental results for air as shown in Fig. 9. In the
figure, the experimental data, the theoretical
curve for laminar air flow in the region of large N u ~ P r - ° ' 4 = 4 ~ . o R e ' ( R ) ~
Reynolds numbers, which is drawn from the
curves for first and second approximation [1], 1 0"061 _'~. (94)
and Nu~ given by equation (71) are also shown. x + [Re (a/R)25]tj
For gases, of which Prandtl numbers are in
the neighbourhood of unity, the conclusive Equation (94) is compared with the empirical
discussion of resistance coefficients may be formulae in Figs. 11 and 12. The empirical
applied to the Nusselt number formula. Namely, correlations are listed below.
54 YASUO MORI and W A T A R U NAKAY\MA

4 Experiments (air) _
0 R i o = 40

g6

_ ' ,(R~i40} ~J Startingpaint

I0-Straight ~ - '"

4102 2 4 6 8 I03 2 4 6 8 104 4 6 8 105 2

Re
FIG. 9. The theoretical curves for N u c and the experimental results.

Pratt [4]: Nuc to N G is expressed as

Nu~ P r f o.4 = 0.0225 (1 + 3"4 a/R) Re °'s. (95) Nu c Pr i - K
Seban and McLaughlin [5] : Nu, 2.~' fl (Pr* - ~ - dgra )

Nu~Pr-~ °'4 = 0"023 (a/R) °'x Re~ sS. (96) × [Re (R)'~/2] */t'tm+ l']
Rogers and Mayhew [6]:
Nuc Pr~ °4 = 0.021 (a/R) °'1 Re} "aS (97) x 1 + [Re(a/R)m/2]l/(m+l~j. (99)
Nu c Pr~-°'4 = 0.023 (a/R) °'1 Re °'ss. (98)
The values of m and e used up to now are
The suffixes b and f denote evaluation of given to equation (99). Figure 13 and Fig. 14
physical properties at bulk (mixed mean) and show the increase of Nusselt numbers for gases
film temperatures respectively. and liquids respectively. Though many formulae
The Reynolds number exponent of 0.85 for Nu, are suggested, Fig. 14 would be available
used in equations (96--98) is in the vicinity of for rough estimation of Nu~ for any kind of
the present result ~ = 0.833...). The difference liquid. However, the experimental data for
in the radius ratio exponent does not have a oil have not been fully established. The absence
significant effect on Nuc, because in any formula of data is due to difficulties in making viscous
it is very small. Bearing in mind the accuracy oil turbulent especially in a curved pipe where
of experiments, we may conclude that equation transition from laminar to turbulent occurs at
(94) agrees well with the experimental results. higher velocity than in a straight pipe.
From equations (68) and (88), the ratio of Equation (99) shows that the difference
FORCED CONVECTIVE HEAT TRANSFER IN C U R V E D PIPES 55

between Nuc and Nus decreases slightly as Pr and temperature distributions inside the pipe
becomes large. This behaviour is in contrast were measured by inserting probes through little
with the results for laminar flow [1]. The opposite access holes at the pipe wall far downstream of
inclination is caused by the difference in the inlet. One of the measured distributions is
mechanisms of heat transfer which can be shown in Fig. 15. The profiles show a steep
seen in the formula of Nus for laminar flow gradient of velocity and temperature near the
(Nus = ~,*s) and for turbulent flow (Nu~ oc Pr~). pipe wall. The presence of the boundary layer
Similar discussion to that about ,t.< can be is ascertained.
i-2

[ 4

I
I. 6 2 4 6 8 I0 20 40 6 0 80 I00
t 2
Pr
FIG. 10. P r I -~l(Pr i -'< - A(J.).
io=
8

2 - -
? /Theoretical =10"2 I04 2 4 6 8 105

Re

8
I I I (~) Theoretical curve
Theoreticall (~----- Rogers 81 Moyhew [6] (Nut, P/°~) _~R =10-8
curve ~-_17~-/70
\ ott [4] / -'° (~) ..... Rogers S Moyhew [6] (Nuc,PP°~

I I I (~) .... Seban Et McLouQhlin [5] [ =104

0 I04 7 6 B I0 ~
Re
FIG. 11. Comparison of the theoretical results with the FIG. 12. Comparison of the theoretical results with the
empirical formula [4-]. empirical formulae [-5, 6 ].

made from equation (88) or (99). As the value of The detailed description of the experimental
m to be applied becomes large, the difference apparatus is given in the previous paper [1].
in R/a has less influence on Nuc. The curved pipe of R/a = 40 is the same as
that used for laminar flow case [1].
EXPERIMENTS The pipe ofR/a = 18.7 is of steel and has the
Experimental investigation in air flow was following dimensions:
done on two cases of radius ratio, 40 and 18.7.
Inside diameter 2a = 53.6 mm
The pipes were heated by nichrom wires wound
Wall thickness 3"5 mm
around them. The heating wires were divided
Radius of curvature R = 500 mm
into several sections along the pipe axis. By
Angle of 0 between the inlet and the outlet
adjusting electrical power input to these sec-
300 degrees.
tions, a constant wall temperature gradient
along the pipe axis was maintained. The velocity A cross section of the pipe does not form a
56 YASUO MORI and W A T A R U NAKAYAMA

1'6
b5
1.4
,3

1
IOd.I 2 " 4 6 8 I 2 4 6 8 I0 2 4 6 8100 2
Re (o/R) 2

Fl(J. 13. N u j N u s for gases (Pr ~ 1).

complete circle because of distortion caused + 1.3Y'C at the inner side, - 1.15~C at the outer
by bending. Difference between the m a x i m u m side, +0.05°C at the top and - 0 . 2 5 ° C at the
outside diameter and the minimum diameter at bottom when the mean wall temperature =
a cross section is 2 ~ 3 mm. The inside diameter 44"15°C, T,, = 24.6°C, T = 6.2°C/m, Re = 3.6 x
is obtained by converting a cross sectional 104, R/a = 18.7. It was found that evaluation of
area, which was found by cutting a test piece from Tw at a fixed peripheral location introduces an
the pipe, to an area of a circle. This equivalent error of several per cent at worst in Nuc. How-
diameter differs about 1 per cent from the ever, this example is the worst one. The varia-
original diameter prior to bending. tions found throughout the experiments for
C u - C o thermocouples for measurements of R/a = 40 are far below the above example.
the wall temperature were buried in the wall F r o m the esperiments, Nusselt numbers are
to the depth of 2 m m from the surface. The obtained by the method shown in the previous
temperature drop due to conduction through paper [1]. The rough description of the method
the wall was taken into account, but the cor- is given here. Mean heat flux at the wall Qwm
rection gave little change on measured Tw. The is written from equation (74)
thermocouples are located at 9 stations at
Qw,, = ¼rk Re Pr. (100)
intervals of 30 degrees in 0. The velocity and
temperature distributions were measured at Therefore, the wall temperature gradient
0 = 150 ° and 210 ° (the inlet of the curved obtained from the measured wall temperature
pipe section is at 0 = 0°). The distribution distribution gives Qw,,. The velocity and tempera-
profiles are found to be similar to those for ture distributions on a horizontal line are
R/a = 40. assumed to have similar profiles to those
F o r both cases of radius ratio, wall tempera- obtained by transversing through the centre of
ture variation around the periphery of a cross a cross section. Both distributions extended
section was checked at two stations by attaching over a whole cross section are utilized to
four thermocouples at the top, bottom, inner multiply numerically temperature by velocity
and outer side of the wall. (The curved pipe is dividing the cross section into small parts.
placed so as to let the centre of curvature and Thus, the mixed mean temperature T,. is
the pipe axis be in a horizontal plane.) The wall computed. The physical properties are evalu-
temperature Tw used in plotting the temperature ated at T,.. The results are shown in Fig. 9.
distribution and calculating the Nusselt number The data for R/a = 40 agree well with the
is the mean value around the periphery. The theoretical curve of a fully turbulent region.
deviations from the mean value are, for example, The data for R/a = 18.7 at Re = 2.7 x 10 4
FORCED CONVECTIVE HEAT TRANSFER IN C U R V E D PIPES 57

F4
CONCLUSIONS
1"2 Turbulent flow and temperature field in a
curved pipe fully developed under the condition
4 6 8 I0 2 4 6 8100 2 4 of constant heat flux were analysed by theory
Re (o/R) z s and experiment, and the following conclusive
F=G. 14. N u c / N u ~ for liquids (Pr > 1). results were obtained.
(1) The theoretical analysis of a flow field
and 3"6 x 104 are found to be below the was done by assuming a thin boundary layer
analytical result. In these cases a little irregular along the pipe wall. Frictional stress at the wall
temperature rise along the pipe axis (0) caused was derived from the resistance formula for a
inaccuracy in determining Qw,,. The experi- straight pipe expressed by 2s oc Re -1/m. The
mental point at Re -- 1.9 x 104 is obtained by result shows that the resistance formula for
determination of z from fluid temperature rise a curved pipe is given in the form of 2c ~/(R/a)
which was measured by inserting the thermo- which is a function of Re (a/R) m/2. One of the
couple probe at several stations along the pipe resistance formulae is obtained by putting
axis. The same method for obtaining Nusselt m = 4 according to the Blasius's formula for
numbers experimentally was applied in the case 2s and taking into account not only the velocity
of the straight pipe. The straight pipe is of brass profile of a flow core occupying the greater
and 35.6 mm in diameter. The velocity and part of a cross section but also that of the
temperature distributions were measured in boundary layer. The formula agreed well in a
the region where both profiles were fully fairly wide range of Re (a/R) 2 with Ito's empirical
developed and the effect of buoyancy was formula [7] which summarizes the data of many
negligibly small. The results are shown in Fig. 9 investigators.
and are in good agreement with Nu~ given by (2) The analysis of heat transfer was done by
equation (71). deriving heat flux at the wall from the Nusselt

1.5 0.5

ot.O--o

I-0 I-0

J
A 4"
0,5 0"5

T~=52-6 *C
T =43-5 =C
T =,5.1 deg C/m

0 i i i r i i i r 0 i ~ i i ' l i ~ i
I'0 0 I'0 1.0 I.O
t7/o R/e
in * - h o r i z o n t a l traverse O --* o u t in +- horizontal traverse O - * o u t
bottom , - vertical traverse A - , top bottom , - vertical traverse A ~ top

FIo. 15. The velocity and temperature distributions (R/a = 40, R e = 2-5 x 104).
58 Y A S U O M O R I and W A T A R U NAKAYAM.X

n u m b e r f o r m u l a for a s t r a i g h t p i p e e x p r e s s e d by a p p r o x i m a t i o n was a s c e r t a i n e d by the m e a s u r e d

N u s oc R e " - ~ / " P r ~. T h e g e n e r a l e x p r e s s i o n of distributions. The Nusselt numbers obtained
the N u s s e l t n u m b e r for a c u r v e d p i p e Nu~ was e x p e r i m e n t a l l y a g r e e d well w i t h t h e t h e o r e t i c a l
o b t a i n e d . T h e r a t i o Nu~/Nu~ is e x p r e s s e d as a result.
f u n c t i o n o f R e (a/R) m/2 a n d Pr.
T h e f o r m u l a for gases (Pr ~ 1) g i v e n by REFERENCES
p u t t i n g m = 4 is in g o o d a g r e e m e n t w i t h the 1. Y. MORI and W. NAKAYAMA,Study on forced convec-
e x p e r i m e n t a l d a t a of air. T h e f o r m u l a for l i q u i d s tive heat transfer in curved pipes (lst report, laminar
region), Int. J. Heat Mass Transfer 8, 67-82 (1965).
(Pr > 1) e x p r e s s e d in t h e f o r m o f N u ~ P r - ° ' 4 by 2. D. JESCHKE, W~irmeiibergang und Druckverlust in
p u t t i n g m = 5 a g r e e d w i t h the e m p i r i c a l f o r m - Rohrschlagen, Z. Vet. Dt. Ing. 69, 1526 (1925); Z. V.D.L
u l a e s u g g e s t e d by s o m e e x p e r i m e n t e r s [4--6]. 24, 1 (1925).
3. W. H. McADAMS,Heat Transmission, p. 228. McGraw-
F i g u r e 13 a n d Fig. 14, s h o w i n g Nu~/Nu~, m i g h t Hill, New York (1954).
be a v a i l a b l e for r o u g h e s t i m a t i o n of Nu~ t h o u g h 4. N. H. PRATT, The heat transfer in a reaction tank
a different f o r m u l a o f Nu~ is to be a p p l i e d cooled by means of a coil, Trans. lnstn Chem. Engrs
25, 163-180 (1947).
a p p r o p r i a t e l y to v a r i o u s cases. 5. R. A. SEBANand E. F. McLAUGHLIN,Heat transfer in
(3) W h e n flow is t u r b u l e n t , 2~ a n d Nu~ d o n o t tube coils with laminar and turbulent flow, Int. J. Heat
s h o w s u c h a r e m a r k a b l e i n c r e a s e f r o m 2~ a n d Mass Transfer 6, 387-395 (1963).
6. G. F. C. ROGERSand Y. R. MAYHEW, Heat transfer
Nu.~ as f o u n d in l a m i n a r flow [1]. T h e v a l u e o f and pressure loss in helically coiled tubes with turbulent
m is to be c h o s e n in the a p p r o p r i a t e r a n g e of flow, Int. J. Heat Mass Transfer 7, 1207-1216 (1964).
R e a n d Pr. As R e o r P r increases, t h e v a l u e o f m 7. H. ITO, Friction factors for turbulent flow in curved
pipes, J. Bas. Engng 81, 123-134 (1959).
to be u s e d b e c o m e s larger, so t h a t the difference 8. M. ADLER, Str6mung in gekriimmten Rohren, Z.
in R / a has less influence on 2,. a n d Nu,.. Angew. Math. Mech. 14, 257-275 (1934).
(4) E x p e r i m e n t a l i n v e s t i g a t i o n in air flow 9. S. GOLDSTEIN,Modern Developments in Fluid Dynamics.
p. 339. University Press, Oxford (1957).
was d o n e in t w o cases o f R / a = 40 a n d 18.7. 10. R. A. SEBANand T. T. SHIMAZAKI,Heat Transfer to
T h e v e l o c i t y a n d t e m p e r a t u r e d i s t r i b u t i o n s in a fluid flowing turbulently in a smooth pipe with walls
the p i p e w e r e m e a s u r e d k e e p i n g the wall at constant temperature, Trans. Am. Soc. Mech. Engrs
73, 803-809 (1951).
t e m p e r a t u r e g r a d i e n t a l o n g t h e p i p e axis c o n - 11. R. C. MARTINELLI,Heat Transfer to Molten Metals,
stant. A p p l i c a b i l i t y of the b o u n d a r y l a y e r Trans. Am. Soc. Mech. Engrs 69, 947~-959 (1947).

R4sum6--L'effet de la courbure sur le transport de chaleur/l flux de chaleur constant pour un 6coulement
turbulent enti6rement 6tabli dans des tuyaux courbes a 6t6 6tudi6 th6oriquement et exp6rimentalement.
On suppose dans l'analyse qu'il existe une couche limite le long de la paroi du tuyau. La contrainte de
cisaillement locale et le flux de chaleur local/t la paroi sont donn~s dans but de r6duire les formules du
coefficient de perte de charge lin6ique (2s) et du nombre de Nusselt (Nu~) ~ la relation locale du frottement
et du transport de chaleur. Lorsque les formules pour les tuyaux rectilignes sont de la forme 2s ~c Re ~
et Nu s ~c Re Ira- 1)/,,, on montre que la similitude dynamique et la similitude pour le transport de chaleur
clans les tuyaux courbes d~pendent de Re (a/RW z.
Le coefficient de perte de charge lin6ique et le nombre de Nusselt pour les tuyaux courbes sont obtenus
e n p o s a n t m = 4 o u m = 5.
Les nombres de Nusselt obtenus/~ partir des mesures des distributions de vitesse et de temp6rature dans
un 6coulement d'air b, travers des tuyaux courbes tels que R/a = 40 et 18,7 sont en bon accord avec les
r6sultats th6oriques.

Zusalnmenfassnng--In dieser Arbeit wird der Einfluss der Kriimmung auf den W~irmeiibergang ftir
ausgebildete turbulente Str6mung in gekriimmten Rohren bei konstanter W~irmestromdichte theoretisch
und experimenteU untersucht. Der Analyse wird eine Grenzschicht entlang der Wand zugrundegelegt.
Die 6rtliche Schubspannung und die 6rtliche W~irmestromdichte werden angegeben um den Widerstand
(2s) und die Nusselt-Zahl (Nu s) for gerade Rohre auf die 6rtlichen Beziehungen fiir Reibung und War-
meiibergang zurtickzufiihren. Werden die Gleichungen ftir gerade Rohre mit 2~ ~ Re-t/,, und Nu,
Retm- l~/m angegeben, so kann gezeigt werden, dass die dynamische ~.hnlichkeit und die Ahnliehkeit des
FORCED CONVECTIVE HEAT TRANSFER IN CURVED PIPES 59

W~irmefibergangs in gekrfimmten Rohren abh~ingig ist von Re(a/R) m/2. Der Widcrstandskoeffizient und die
Nusseltzahl fiir gekrfimmte Rohre lassen sich erhalten wenn m a n m = 4 oder m = 5 setzt.
Die Nusselt-Zahlen, die durch Messungen der Gcschwindigkeitsund Temperaturverteilungen in einem
Luftstrom durch gekriimmte Rohre yon R/a = 40 und 18,7 erhalten wurden, stimmen gut mit theoretischen
Ergebnissen fiberein.

Am~oTa~Mi~[--B pa6oTe HpOBO~HTCH aHCl~epHMeHTaZtbHoe n TeopeTHqecHoe Hcc~eKOBaHHe

BJIHHHHH HpHBH3HIaI Ha TeHJIOO6MeH B HO,rIHOCTbIO pa3BHTOM Typ6y~ieHTHOM TeqeaHH B
HCHpHBJIeHHMX T p y 6 a x HpH HOCTOHHHOM TeIIJIOBOM IIOTOHe. YIpH aHa~iH3e IIpHHHMaeTcA
cyMeCTBOBaHHe HoFpaHHqHOFO CJIOH Ha CTeHHaX T p y f u . JIoHaJIbHHe H a n p A m e H . A TpeHHH H
~oHazIhH~i~ TenZIOaOfl novoR [Link] n a / ~ o p M y n conpoTna:IeHHA (A~) ~ q n c ~ a HyccezIbwa
(Nus) ~laA npAMUX T p y 6 , n p n a e ~ e H H u X H COOTHOmeHHAM n o H a : I ~ n o r o wpeHnA n wenno-
o6MeHa. IIoHaaaHo, qVO B TO BpeMA gaR (popMyJIU ~lnA npAMUX wpy5 ~Iamvca B nH;~e ,~socRe -~/'n
H g u s o c R e ~m-D/m, ~IHnaMtlqec~oe n venJxOBOe tio~oSYie HCHpHBJIeHHHX Tpy6 3aBHCHT OT
Re(a/R) 'n/2. IIoJIyqeHI~I I~O~{~(~HIII4eHTbI conpoTHB~eHI4A n qncJla HycceJlbTa KJIA n c ~ p n B ~ e H -
HUx~pySnpHm=4Hm=5.
q H c a a H y c c ~ : n , wa, n o : ~ y q c a H u e npH HaMepeHan npoqbn:tei~ cHopocTH n weMnepavyp~[ np~!
T e q e n n n B o a ~ y x a B H c ~ p . a : I e n n u x w p y 6 a x n p ~ R/a = 40 a 18,7, x o p o m o c o r a a c y m w c A c
weopevnqeC~nMn.