Professional Documents
Culture Documents
B S PETUKHOV
High Temperature Institrite. Acudcrr~yof Science of the USSR. Moscoqc.. U S S R
I . Introduction . . . . . . . . . . . . . . . . . . . . . .
I1 . Analytical Method . . . . . . . . . . . . . . . . . . .
A Basic Equations . . . . . . . . . . . . . . . . . . .
B . Eddy Diffusivities of Heat and Momentum . . . . . . .
C &4nalyticalExpressions for lemperature and Velocity Profiles,
Heat Transfer. and Skin Friction . . . . . . . . . . . .
111. Heat Transfer uith Constant Ph?sical Properties . . . . . . .
A . Analytical Results . . . . . . . . . . . . . . . . . .
B . Experimental Data . . . . . . . . . . . . . . . . . .
I\. Heat Transfer and Skin 1:riction for Liquids with Variablc
Viscosity . . . . . . . . . . . . . . . . . . . . . . . .
A . Theoretical Results . . . . . . . . . . . . . . . . . .
B . Experimental Data and Ihpirical Equations . . . . . . .
V . Heat Transfer and Skin Friction for Gases with Variable Physical
Properties . . . . . . . . . . . . . . . . . . . . . . .
A . Analytical Results . . . . . . . . . . . . . . . . . .
B . Experimental Data and Ihpirical Equations . . . . . . .
VI . Heat Transfer and Skin 1;riction for Single-Phase Fluids a t
Subcritical States . . . . . . . . . . . . . . . . . . . .
A . Analytical Results . . . . . . . . . . . . . . . . . .
B. Experimental Data and I<mpiricalEquations for Normal I I w t
Transfer Regimes . . . . . . . . . . . . . . . . . .
C . Experimental Data f o r Hcgimcs with Diminished and
Enhanced Heat Transfur . . . . . . . . . . . . . . .
VI1 . Conclusion . . . . . . . . . . . . . . . . . . . . . . .
.
.
Nomenclature . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . .
503
504
507
507
510
516
521
521
525
528
528
530
533
533
540
543
543
550
555
560
561
561
504
B. S. PETUKHOV
I. Introduction
Heat transfer in turbulent pipe flow has been investigated for almost
HEATTRANSFER
AND FRICTION
IN TURBULENT
PIPE FLOW 505
steady fluid flow with constant physical properties in circular tubes has
been rather fully investigated.
I n reality, fluid physical properties depend on temperature. That is
why heat transfer relations obtained with the assumption of constant
physical properties can only be used in practice either a t small
temperature differences in a flow or with physical properties changing
slightly in the temperature range considered. I n this case the effect of
changing physical properties can be approximately accounted for by
choosing the properties at a certain average fluid temperature.
I n heat transfer systems used in different fields of engineering, large
temperature drops and high heat fluxes are often realized. I n this case
the large temperature gradients occur in a fluid flow. For example, in
nuclear reactors the heat flux (E;cal./m2 hr) may be as high as several
millions. In cooling systems for jet propulsion engines it may rise to
several score of millions while a heat flux of hundreds of millions may
occur i n some special kinds of apparatus. Liquids and gases whose
physical properties are very responsive to temperature changes are
often used as heat transfer fluids. Gases flowing at large temperature differences or some liquids (single-phase fluid) at subcritical states
serve as examples. In these cases it is impossible to consider physical
properties constant, because great errors would otherwise result. Under
such conditions the analysis of the flow and heat transfer should include
the dependence of physical properties on temperature.
For various types of fluids and for a given fluid the variation of the
physical properties with temperature and pressure is not the same over
different ranges of the state parameters. For such a fluid under these
varying conditions it is presently impossible to describe the fluid flow
and heat transfer by a single relationship valid for all conditions. As a
consequence, the problem of a fluid flow and heat transfer with variable
physical properties divides into several problems, and each problem
corresponds to a certain type of dependence of physical properties on
temperature and pressure. Therefore, the analytical expressions for fluids
with constant physical properties are not universal in the case of variable
physical properties. Theoretical studies of flow and heat transfer
in fluids n i t h variable physical properties are hindered by different
mathematical and physical difficulties. T h e mathematical difficulties
can be explained by the fact that the momentum and energy equations
in the case of variable physical properties are coupled and nonlinear.
However, these difficulties can be overcome, e.g., by using numerical
methods and with the help of computers. T h e difficulties of a physical
nature are more serious. T h e y may be attributed to the inability to
prescribe analytical expressions for the turbulent diffusivities of heat
506
B. S. PETUKHOV
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 507
equations are the basis of this method. With the help of these expressions
we can calculate heat transfer and friction using the method of successive
approximations. Later on this method was used for calculation of heat
transfer and friction of some gases with and without dissociation, and
also for carbon dioxide at supercritical state parameters.
T h e fluid mechanics and heat transfer of turbulent pipe flow with
variable physical properties have been studied both theoretically and
experimentally. Published papers are available which contain many
experimental results. In particular they present data obtained on heat
transfer and skin friction for liquids under the conditions of substantial
changes in viscosity, for some gases at large temperature differences, and
for water, carbon dioxide, and some other substances at supercritical
states. Some of the papers contain empirical equations correlating the
experimental results.
I n this paper we shall consider heat transfer and skin friction in
turbulent pipe flow with variable physical properties. T h e constant
properties solution will be considered only so far as is necessary for the
flow and heat transfer analysis with variable physical properties.
A. BASICEQUATIONS
Turbulent flow is, of its nature, transient. Velocity, temperature, and
other properties change continuously in time at every point of a turbulent
flow. These changes are irregular fluctuations with respect to some
temporal mean. This behavior allows us to represent different turbulent
flow properties as the sum of the mean value, in time, and a pulsation of
this value. So we can describe the field of real (instantaneous) velocities
as a field of averaged (in time) velocities and the superimposed field of
velocity fluctuations. We can do the same with temperature, pressure,
and density fields and with other dependent variables. With this
approach, transfer processes in a turbulent flow are controlled by two
mechanisms: molecular and convective (turbulent). T h e first mechanism
results in the appearance of viscous stresses proportional to the gradients
of the averaged velocity and heat fluxes due to heat conduction which are
proportional to the averaged temperature gradients. T h e second
mechanism gives rise to turbulent stresses caused by momentum transfer
due to velocity fluctuations and turbulent heat fluxes caused by heat
transfer resulting from velocity and temperature fluctuations. This
approach suggested by Reynolds allows us to pass from energy,
momentum, and continuity equations for instantaneous values to the
508
B. S. PETUKHOV
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 509
where x and r are the axial and radial coordinates (x coincides with the
pipe axis); W, and W , are the averaged (in time) values of the axial and
radial components of the velocity vector, respectively; p, h, T , and P are
the averaged density, enthalpy, temperature, and pressure; WE,
W,,
h,
and T are the fluctuations of velocity (in the axial and radial directions),
enthalpy, and temperature, respectively; h and p are the thermal
conductivity and dynamic viscosity at temperature T and pressure P.
Since fully developed flow is being considered, it can be assumed that
the change in the axial component of the mass velocity along the axis x
is small, ix.,
tI(pwT)/aYw 0
(2;
.
p ? W )
w0
then it is seen from Eq. (3) that a/& = 0, i.e., pressure P is constant
over the cross section.
With these assumptions the set of Eqs. (1)-(4) is reduced to the
following two equations:
A(;iT/ar) - pWTh
(7)
T h e first term of this sum is the heat flux due to conductivity, while the
second item is accounted for by the eddy diffusivity of heat.
On the right-hand side of Eq. (6), the expression for shear stress is
given by
u =
ii W,/ar) - p W, W,.]
- [p(
(8)
Here the first term is the viscous stress and the second term is the
turbulent stress.
5 10
B. S. PETUKHOV
and
u =
uwR
(9)
and
u = uW
(10)
Here qw and ow are the heat flux and the shear stress at the pipe wall,
respectively; R = r/ro is the dimensionless radius; ro is the pipe radius.
We can see that the first assumption from (9) is fulfilled only in the
case of slug2 flow, and the second assumption only for fully developed
flow with constant physical properties. As for assumption (lo), it is not
fulfilled for pipe flow. However, in a number of cases assumption (9) and
even (10) do not introduce great errors into our analysis. This may be
attributed to the fact that in calculations of heat transfer and skin
friction a correct description of the flow near the wall is the most essential,
and in this region assumptions (9) and even (10) are fulfilled approximately. Quantitative error estimations are given in Section 111, p. 521
(see Table I, p. 522).
Assumptions (9) and (10) allow us to replace Eqs. ( 5 ) and (6) by
simpler ones which can be obtained from Eqs. (7) and (8) as a result of
the substitution of q and u from (9) and (10).
For calculations with the help of Eqs. (5) and (6) it is necessary to
express the turbulent heat flux pW,h and the turbulent shear stress
p WzW, as functions of the independent variables and the averaged flow
properties. Then it is convenient to introduce the coefficients of the
eddy diffusivities of heat and momentum.
B. EDDY DIFFUSIVITIES
OF HEATAND MOMENTUM
By the definition, the eddy diffusivity of heat is
~
Eq
W,h
ahiar
= ___
W,Wrt
a WJar
A slug flow is a flow ~ v i t l iil uniform (over the pipe cross section) velocity profile.
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 511
~-
the cross section), then the expressions for heat flux and shear stress
become
q = (A
+ pCDcO)aT/ar
(13)
= 71*r,/v.
B. S. PETUKHOV
512
functions of 7, and near the axis they are also functions of the radius R
(or of the Reynolds n ~ m b e r ) ~ :
v
EuiV
drl, R ) =
= (Eo/V)(%
d17,Re)
R) = (~U/V)(% Re)
( l / ~ In
) 7
+A
(16)
< <
E,/V
n2qy[l
(17)
exp(--n2~7)]
where K = 0.36
Deissler has noted that the velocity profile in a turbulent core
calculated from (18) for E<,/vis described by logarithmic relationships ( I 6)
when K = 0.36. Therefore we can approximate (,/v with the following
equation instead of (1 8):
C,/V
0.36(1
7/70) 7
(19)
zi*r,/v
$ Re([/8)'/*
where
HEATTRANSFER
.4ND F R I C T I O N
For 0
For q
> 50:
IN
TURBULENT
PIPE F L O W
513
where K
0.4 and qr1= 1 1 .
With large Prandtl numbers the main temperature change occurs
directly in the vicinity of the wall. In this case, for the calculation of heat
transfer, it is very important to describe turbulent transfer processes near
the wall correctly.
shall analyze Eqs. (17) and (20) from this point
of view. .At small q Eqs. (17) and (20) can be simplified by series
expansions of the exponential function and hyperbolic tangent, and by
considering only the first t\no terms of the series. I n addition, we can
take
71 in (17) which results in
1
c,,'u =
c174
and
E,,/U
= Cn'73
where
n4 = 2.365 )i
and f 2 = K/3qn2 = 1.102
lo--3.
T h u s , a t small 7 1 , according to Deissler co q4, and according to
Reichardt E ,
qB; also Eq. (17a) gives lower E, values than those
predicted by Eq. (20a).
I n the literature there is no consensus on the value of the exponent rn
in the equation for the region close to the wall i , = cq"' (here i , , = E , , / u ) .
We can only say from theoretical considerations that m :
: 3.4 Being
inaccurate, measurements near the wall do not produce a reliable value
for m. \Ye can indirectly infer m values from comparisons ofthe predicted
values of heat and mass transfer mith experimental data at large Pr or Sc.
Such a comparison was made in (25) on the basis of a statistical analysis
of the experimental data on heat and mass transfer which shows that m
ranges from 3 to 3.2. Apparently, this is close to reality.
Thus, Eq. (20) which gives m
3 is more likely to describe the
mechanisms of turbulent transfer near the wall than Eq. (17).
Equations (20) and (2 I ) have some other advantages over Eqs. ( 1 7)-( 19).
T h e relationship for c , : ! ~ , suggestccl b y Reichardt, has no discontinuities
in the range q -. 50 Lvhich is important for the calculation of heat
~
\I'
-~
B. S. PETUKHOV
5 14
transfer and skin friction. Besides, Eq. (21) takes into account changes in
EJv with R (and with Re) in the central part of the tube, and on the tube
axis it gives a nonzero value of E, which varies with Re that is in full
agreement with experimental data. This means that the velocity gradient
calculated from Eq. (21) is zero along the tube axis. Therefore, we shall
use mainly Eqs. (20) and (21). In Fig. 1, E,/V versus 7 is plotted for
FIG. 1.
c,/u
1s 7
HEATTRANSFER
A N D FRICTION
I N TURBULENT
PIPE F ~ o n 515
are not numerous. Resides, the accuracy of experimental values of eq is
rather poor (especially i n the wall region). T h e majority of measurements
was carried out with liquid nictals and seldom with air. In this paper we
shall not discuss the problems of eddy diffusivity and heat transfer for
liquid metals. T h e measurements with air (26-29) show that at some
small distance from the wall /3 :
1.2-1.5, but it decreases when the
distancc from the wall and Re increase. However, some other workers (30)
report that ,B is approximately unity near the wall and it increases with
the distance from the wall. Lack of experimental data has as yet prevented
us for determining the general relations for eq or p over a wide range of
Re and Pr.
Presently, in predicting heat transfer, ,B is usually taken as one ( a t least
for Pr 2 I ) due to the ambiguity of estimating 8. This is justified since
the results at constant properties predicted with such a n assumption are
in good agreement with experimental data.
If the physical properties change with temperature, i t becomes
necessary to take into account the influence of variable physical properties
on the turbulent diffusivity expressions. As is known, this problem has
not yet been systematically investigated. Therefore, a solution is usually
based on some assumptions, and their validity can be confirmed only
indirectly by comparing predicted values of heat transfer and skin
friction with experimental data.
Deissler (11,24) assumed that, to calculate the eddy diffusivity of
momentum with variable physical properties, relations (17) and ( 1 8) may
be used although they were obtained at constant physical properties. I n
these relations the kinematic viscosity is considered a variable. I n
accordance with such an assumption the equation for E,!V takes the form
where
e7.r
H'3!zw*,
7u
z'w*y;vu.,
Constants n and K have the samc values as in the case of constant physical
properties. Equation (22) is recommended when vw < 26, and Eq. (23)
is used when rlW .:' 26.
Goldmann (29) suggested a method of calculating E , , with variable
physical properties using the hypothesis that local turbulence characteristics a t a given point depcncl on physical properties a t that point and
do not depend on physical properties changing in the vicinity of that
point. In light of this hypothesis, Goldmann has come to the conclusion
516
B. S . PETUKHOV
C. ANALYTICAL
EXPRESSIONS
FOR TEMPERATURE
AND VELOCITY
PROFILES,
HEATTRANSFER,
AND SKINFRICTION
Considcr the problem of turbulent fully developed quasi-steady flow
and heat transfer in a circular tube, assuming that the fluid is
incompressible and its physical properties display some arbitrary
temperature dependence. T h e problem is analyzed for the case of
constant heat flux which is prescribed at the wall (qw = const) (14).
If the assumptions of Section 11, A are applied, this problem may be
described by the energy and momentum equations ( 5 ) and (6). The
left-hand side of Eq. (6) can be written as
As earlier, a@W,)/ax w 0 was assumed (see Section 11, A). Taking all
this into account and also using relationships (13) and (14) we obtain
the equations
1 a
-r -ar
( r!?)
pW2) = - l a
- - (ra)
ah
pw,- ax
a (P
ax
r ar
HEATTRANSFER
A N D FRIC.TION
I N TURBULENT
PIPEFLOW 517
where
f@)
(27)
= f&)
(28)
?k?1,Y
[aTjax = f(.)].
For liquids and gases flowing at sniall subsonic velocities, the pressure
gradients due to longitudinal density changes a/ax(pW.,.P)
are, by far,
smaller than the total pressure gradient dP/dx. Since P does not change
along the tube section assumption (28) is also well founded under these
conditions.
First, analytical expressions must be found for the enthalpy and
temperature fields and the Nusselt number.
Multiplying Eq. (25) by Y dr, taking into account assumption (27),
and integrating with respect to the radius from 0 to ro , we obtain
-
2qw/pW,ro
lll/t,V
gvz
(29)
where
is the (over the section) bulk velocity. Substituting (29)
into (25) and integrating from 0 to Y ~ v obtain
e
the expression for the heat
flux distribution along the radius
where R
r/r, .
Since the pressure P is uniform over the tube section, we have
dh/&
CDa T p r
(A/C',])(l
.,/a)
ah/&
B. S. PETUKHOV
518
or
9
where /3
= en/c0 .
(WXl + P Pr %/V)
Now we shall calculate the bulk enthalpy (A,,), or, to be more exact, the
enthalpy difference h, - h, , By definition,
Substituting h,
obtain
Now let us introduce the heat transfer coefficient from the definition
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 519
cp
Substituting h, - 12, from (37) into (39), we obtain the expression for
the Nusselt number
For constant physical properties, wc can reduce Eq. (41) to the wellknown Lyon integral (8)
o,R
pwx
1' pWrR dR
- 0
520
B. S. PETUKHOV
5fw = 8 ~ w P w / ( P w , ) 2
Substituting Fz
from (44) into (45),we obtain the expression
where
Rew = p,d/pW
where
Re
=P
d / p
- VW/VOW
where
TW
= .w*y/vw,
Tow =
~ W * ~ O / ~ W ,
v w * = (.w/f
w)12
HEATTRANSFER
A N D FRICTION
IN TURBULENT
PIPE FLOW 521
higher approximations. T h e calculation is performed till the difference
in temperature distribution of the ( n
1)th and nth approximations
becomes smaller than some prescribed value within the range of which
we can neglect the change of the physical properties.
4. E,/Y and the velocity distribution are calculated from the physical
properties distribution obtained in the last approximation. Then from
Eq. (41) we can find the value
Since q\,,d and T, are prescribed, then from the last relation we can
find the average bulk enthalpy h,) and the appropriate bulk temperature
T,, . After that the number Nu, = qwd/X(Tw
- Tb)is determined.
Using Eq. (46) we can calculate the friction factor
.
ew
A. ANALYTICAL
RESULTS
Consider the heat transfer solution for the case of a fully developed
flow with constant properties i n a circular tube with constant heat flux
at the wall. T h e calculation has been done by Petukhov and Popov (14) by
the method discussed in Section 11, C. T h e eddy diffusivity of momentum
was calculated from Keichardts Eqs. (20) and (21); j? = E , J ~ was
~
taken
to be one. T h e calculation was done for Re
104-5 x lo6 and
Pr
0-2000. T h e method of calculation takes into account the variation
of the heat flux q and shear stress u along the radius. I n order to estimate
the errors which could appear with the assumption of uniform q and u
along the radius, several values of Nu and E were calculated for q = qw
and u = u, . T h e ratios of thc corresponding values of N u and 5 for
varying q and u to these values for q = qw and u = a
, are tabulated
and presented in Table I . As can he seen from the table, the assumption
of uniform q and u produces noticeable errors in N u and [ values,
especially for low Re and Pr.
Nu vs Pr for various lie, according to the predicted values, is plotted
in Fig. 2 (for Pr 3, 0.5). T h e calculations over the range Re and
B. S. PETUKHOV
522
TABLE I
RATIOOF Nu AND $, CALCULATED
TAKING
INTO ACCOUNT
CHANGES OF q AND
RADIUSTO THEIR
VALUES WHEN Q = Qw AND O = O W .
105
o ALONG THE
106
N ~ / N ~ g = g ~ . u = o ~
1
10
100
lo00
1.14
1.09
1.08
1.07
1.11
1.08
1.06
1.06
1.10
1.07
1.05
1.04
l/lu=uw
1.17
1.12
1.10
FIG.2. Nu vs Re and Pr from predictions by Petukhov and Popov (solid lines) and
Deissler (dotted lines).
HEATTRANSFER
AND
TURBULENT
PIPE FLOW
F R I ( T I O N IN
523
Pr
104-5 x lo6 and 0.5-2000, respcctively, are described by the
interpolation equation
where
5
Kl([)= 1
(1.82 log Re
+ 3.45,
k,(Pr)
1.64)~
11.7
+ 1.8 Prr1I3
(49)
Nu
((it%) Re Pr
1.07
+ 12.7(t/8)1/2(Pr2I3
1)
524
B. S. PETUKHOV
where
E(=-
3 43
2n
= 0.0855
<
Ku
= KD(f/8)1/2
Re Pr1/4
(52)
where
K D = 2 42nI.rr
= 0.112
was obtained with the assumption that the eddy diffusivity of momentum
e o / v is described by Eq. (17a).
Unlike Eq. (51), the expression (52) gives a weaker dependence of
Nu on Pr. That is why at very large Prandtl numbers (or Sc) Nu (or Sh)
calculated from (52) appears to be lower than that calculated from
Eq. (51) (at Sc = lo5 it is lower by approximately a factor of two). Only
by comparing the predicted results with the experimental data can we
solve the problem to the extent that the predicted results correspond to
reality. Before the discussion of the experimental data we shall note a
very important fact.
HEATTRANSFER
AND F R I C T I O N
IN
TURBULENT
PIPE FLOW
525
B. EXPERIMENTAL
DATA
A great number of experimental papers on heat transfer in turbulent
pipe flow have been published. LJnfortunately, in many cases measurement accuracy was not high; therefore, heat transfer coefficients obtained
experimentally often contain substantial errors which are difficult to
estimate. Little experimental data of rather high accuracy have been
reported in recent years (34-37). Mainly heat transfer for air and water
flow has been measured, i.e., approximately over a range of 0.7-10 for
Prandtl numbers. Only a few authors have obtained heat transfer data
at Pr from 10 to 100-150 and a little higher. Heat transfer measurements
were not performed for Pr
1000 because of the great experimental
difficulties. Therefore mass transfer experimental data were used over
the range of Pr (Sc to be more exact) for which the exact analogy between
heat and mass transfer processes was valid.
For comparison with the predicted results the most reliable
experimental data have been chosen. T h e main characteristics of the
data are given in Table 11. T h e heat transfer experimental data were
extrapolated to the zero wall and flow temperature difference to avoid the
affect of the dependence of the fluid physical properties upon the
temperature. In some cases such extrapolation was performed rather
accurately, while in other cases only approximately. Naturally the mass
transfer experimental data need not be extrapolated because all the
measurements were performed under isothermal conditions.
I n Fig. 3 the predicted results described by Eq. (48) are compared with
All the listed results are for the case of fully developed Nu number. T h e difference
between Nu,, and Nu, can be larger at the thermal entrance region.
526
B. S. PETUKHOV
TABLE I1
THEMAIN CHARACTERISTICS
OF HEAT AND MASS TRANSFER
EXPERIMENTAL
DATA
Ref.
Symbols
Volkov and
Ivanova (38)
Petukhov and
Roizen (39)
Sukomel and
Tsvetkov
Allen and
Eckert (34)
Dipprey and
Sabersky (37)
Yakovlev (36)
Malina and
Sparrow (35)
Sterman and
Petukhov (40)
Hamilton ( 4 f ) b
Fluid
Re.
Pr or Sc
Io/d"
I/d"
Air
48-370
12.5-350
-0.7
Air
40
39
15-280
-0.7
Air. helium
50
80
9-40
0.67-0.71
Water
96
30
13-110
Water
48.5
46.2
150
1.2-5.9
Water
Water, oil
-4
96
70-80
30
19-140
12-100
2-12
3,48, 75
Monoisopropyldiphenyl
Water and
water solutions
of glycerine
and metaxyl
0, 30
89-125
22-260
12-35
75
13
10-100
430-105
lo/d and l/d are relative lengths of calming nonheated and heated sections.
In Hamilton ( 4 1 ) an experimental tube wall made of benzoic acid served as a solid
phase.
a
'.'
Nu0 1.0
( 0 )
ao
a 1 0
10'
4Re
HEATTRANSFER
A N D FRICTION
I N TURBULENT
PIPE F ~ o w 527
of the experimental data from predictions does not exceed 5-6(), (except
for a few points), this being ivithin the range of accuracy for both
predicted and experimental data.
I n Fig. 4 the predicted results are compared with the mass transfer
experimental data at large Sc. Ihe solid curve in the lower plot corresponds to Eq. (48) (when Sc . lo3) and to Eq. (51) (when Sc . > 10.
I n the upper plot the solid curve corresponds to Eq. (51). The dotted
lines show Deisslers calculations (when Sc is approximately lo3 and
higher, t h e curves being drakvn according to Eq. (52)). From this figure,
Eqs. (48) and (51) are in quite good agreement with the cxpcrimental
data over a wide range of Sc (up to 10. In comparison with the
experimental data Ileisslers calculations produce lower Sh numbers.
Therefore, when Sc =- lo3, the difference is approximately 20,,, and
when Sc = lo5, the differencc increases to 50%.
T h e analysis shows t h a t , for practical heat and mass transfer calculations over a range of Pr or Sc
0.5-IO3, Eq. (48) should be used, but for
calculations in the range of Pr or Sc
103-10, Eq. (51) is valid.
Equation (50) may also be used for Pr or Sc ranging from 0.5 to 200. As
mentioned above, these equations describe the range of Reynolds
numbers from lo4 to 5 x 106.7
With high Re a nd simultaneously hiah P r o r Sc, th e validity of Eqs. (48). (50). and (51)
has not yet been verified experimentally d u e to th e absence of experimental data. I-loweTw,
from the theoretical considerations \ve can assume that with high Re they +re also in
good agreement with expcrimental data.
B. S. PETUKHOV
528
= c Rem Prn
(53)
Comparing Eq. (53) with Eqs. (48) and (51) it is easy to see that with
Eq. (53) at constant c, m, and n it is impossible to describe to a reasonable
accuracy the change of Nu number with Re and Pr over a wide range of
these parameters. A direct comparison of Eq. (53) with experimental data
leads to the same conclusion. Allen and Eckert (34) have shown that
1.3 x 104-11 x lo4 and Pr = 8, Eq. (53) (when c = 0.023,
for Re
m = 0.8, and n = 0.4) produces an error of u p to 20%. An equation of
the type (53) can be used for N u = N u (Re, Pr) only assuming that
c, m, and n are functions of Re and Pr. For Re
104-5.106 and Pr
(or Sc) -0.5-105, m changes from 0.79 to 0.92, while n varies from
0.33 to 0.6.
A. THEORETICAL
RESULTS
For liquids (condensed medium) far from their critical point only
dynamic viscosity varies greatly with temperature; all the other physical
properties (p, C, , A ) depend on temperature rather weakly. Therefore
while investigating nonisothermal liquid flow, a model with variable
viscosity may be used as a good approximation, other physical properties
being assumed constant.
Deisslers paper ( I I ) should be noted as one of the papers devoted to
the analysis of flow and heat transfer for liquids with variable viscosity.
His analysis reduces to the simultaneous solution of Eqs. (13) and (14)
in dimensionless form by means of the successive approximations
method. As in the case of constant properties, q = qw and u = uw are
assumed. T h e eddy diffusivity of momentum is calculated from Eqs. (22)
and (23), and /3 is taken to be one. T h e variation of viscosity with
temperature is taken into account only near the wall (7 < 26). T h e
temperature dependence of viscosity is formulated as
where K is a constant; K varies from - 1 to -4. As for the turbulent core
viscosity, it is considered constant here.
Deissler has calculated heat transfer and skin friction over the range of
1-103 for Pr and 4 Y 103-2 x lo5 for Re. His results are given as
Nu = N u (Re,Pr,.) [where the subscript x means that the physical
HEATTRANSFER
A N D FRICTION
IN TURBULENT
PIPE FLOW 529
properties are evaluated at the characteristic temperature as defined
below] and 5 = ((Re,) for constant physical properties, if in calculating
Re, and Pr,r the values of dynamic viscosity are taken at the reference
temperature:
t,
= .l(tw
tb)
+ tb
(54)
I I I 1
I I l l
I I l l
x'
I I l l
/U
Pr
loo
/om
FIG.5 . Reference temperature for heat transfer (a) and friction (b) calculations by
Deissler's method for fluids with vari;lhlc viscosity. Key: solid line, heating of fluid;
dashed line, cooling of Ruid.
530
B. S. PETUKHOV
EQUATIONS
B. EXPERIMENTAL
DATA AND EMPIRICAL
Heat transfer and skin friction experimental data obtained under the
conditions of essentially varying viscosity are not numerous; with rare
exception they are not accurate and often do not agree well with each
other. Therefore we shall use only that small amount of data which may
be considered the most reliable.
Analyzing the experimental data we assume that the relation of N u
with Re and Pr, and the relation between 5 and Re at variable physical
properties (in this case at variable viscosity) is the same as in the case of
constant properties. This assumption is confirmed by the experimental
data for liquids with variable viscosity, gases with variable physical
properties, and certain other cases (see the following sections).
Of course, assuming a similar variation of N u with Re and Pr, and 8
with Re for both constant and variable physical properties is only
approximate. For example, Allen and Eckert discovered experimentally
(34) (see Fig. 6) that variable viscosity affected heat transfer to a greater
4.4
f.3
12
1.i
LO
1.r
1.2
f.3
1.4
16
f.8
2.0
2.5
pb'pw
FIG. 6 . Variable viscosity influence on the heat transfer in heated water: 0 ,0 ,
v , and are Allen and Ikkert's experimental data for Re = 13,000, 20,500, 35,500,
8. I, Deissler's predictions; 11, Nu/Nu, = (pb/pW)O.'*;
62,500, and 110,000 at I'r
A,
111, Nu/Nu,
(/Jb/pw)".".
3.0
HEATTRANSFER
AND
FliICTIoN IN
TURBULENT
PIPE F L O W 531
viscosity on heat transfer and skin friction can be estimated from the
following relationships:
Nu
= fNu(Pw/Pbh
6/60
(55)
f&w/Pb)
where N u and [ are the Nusselt number and the friction factor at variable
are the same numbers
viscosity obtained experimentally; Nu,, and
calculated by assuming constant physical properties with Re and Pr the
same as for the corresponding Nu and [; pw and p b are the dynamic
viscosities at T , and T,, , respectively. All the physical properties
(except pw) in the expressions for the dimensionless numbers are
calculated at the bulk temperature t,, for the given tube section.
I n Fig. 6, Allen and Eckerts experimental data (34) on heat transfer
for the case of water heating are compared with Deisslers predicted
results and some empirical relationships. As is readily seen, the predicted
results describe qualitatively the effect of variable viscosity on heat
transfer but produce quantitatively overestimated values.
T h e relative change of heat transfer due to viscosity dependence on
temperature can be expressed by the equation
so
WNu,
(56)
= (Pb/PWY
where Nu, is calculated from Eq. (48) or (50) and the n expression is
determined from the experimental data.
As we can see from Fig. 6, n = 0.14 as suggested by Sieder and
Tate (42) is overstated; n = 0.1 1 as suggested in ( 1 3 , 4 3 )corresponds to
experimental data for liquid heating better than n = 0.14.
To choose the correct value of n in Eq. (56) the heat transfer experimental data corresponding to heating and cooling for several liquids over
a wide range of values pw /pl, (the main characteristics of these data are
presented in Table 111) where treated. T h e results of the treatment are
TABLE 111
THEM A I N
CHARACTERISTICS OF
Ref.
Yakovlev (36)
Kreith and
Summerfield (45)
Petukhov ( I 3)
Petukhov (13)
HEATTItANSFEH EXPERIMENTAL
DATA
FOR
V A R I A R I . I IVISCOSITY
Fluid
Symbols
1
Water
I/d
70-80
Re.
19-123
Butyl alcohol
38
42-78
Iransformci- oil
88
88
5-44
5-14
0 1 1
MS
F L L I D S \VITH
Pr
Pw/W
2-12
0.19-0.77
23-30
0.08-0.45
39-61
I .2-8.6
1.6-38
134-140
532
B. S . PETUKHOV
FIG.7. Variable viscosity influence on the heat transfer in different fluids for heating
and cooling (for symbols see Table 111).
5/60
= (CLW/P~)O"~
(57)
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 533
p w pb
FIG. 8. Variable viscosity influence on friction in water for both heating and cooling:
13,000-110,000, P r = 8); 0 Rohonczys
experiments (Re = 33 . 103-225 . loJ, Pr = 1.3-5.8). I, Deisslers calculation (when P r = 8
for heating and P r = 2.3 for cooling); 11, f/[ ( f L ~ / p h ) ~ 111,
~ ; (/to= 1/6(7 - p h / p W ) ;
Iv, f / f o ( f L ~ / f L h ) ~ ~ ~ .
(0)Allen and Eckert experiments (Re
B. S. PETUKHOV
534
1.0
08
0.8
06
0.6
04
0.4
02
0.2
I 0
0.2
1.0
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 535
in the q/q, distribution. For cooling, when the mass velocitv profile is
more full, the pip, distribution becomes nearly linear. T h e property
variation also affects greatly the distribution (along the radius) of the
relative eddy diffusivity of monientum.
FIG. 10. Heat transfer versus tcmperaturc ratio parameter according to t h e analys~s.
,(
* From here on, th e suhscripts 1) and w mcan that t h e physical gas properties are
evaluated at temperatures Tt, and T,, \\lien calculating the corresponding dimensionless
numbers.
B. S. PETUKHOV
536
where
u = -(a log 8
8"
(59)
+ 0.36)
FIG. 11. Friction versus temperature ratio parameter (dots denote Petukhov and
Popov's predictions; for symbols see Fig. 10).
tb/t"b =
where
II
8"
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 537
for heating and
n
-0.6
+ 0.79(Rew*)-0.11
for cooling.
Equation (60) describes the calculated results within 2-3 010 over the
range 0.37-3.7 for 0 and 14 x 10"-1O6 for Re,*. As Re,* varies over the
indicated range, n goes from -0.44 to -0.58 for the case of heating and
from -0.32 to -0.42 for the case of cooling. If in Eq. (60) n is taken as
-0.52 for heating and -0.38 for cooling, this equation describes the
calculated data to within 700 accuracy in the first case and 4", in the
second.
Kutateladze and Leontiev (17) have obtained an analytical expression
for the functions describing the influence of variable physical properties
on heat transfer and skin friction in a turbulent gas flow when Re + 00.
By making some assuniptions (u and q, vary identically with respect to the
radius for both constant and variable properties, the velocity and
temperature fields are similar, C,, = const) they obtained the following
relationship:
Re;l4/31
(62)
T h e subscript x denotes that the physical properties in Xux and Re, are
evaluated at the temperature
I,.
r .
X(Tw
Tb)
+ Tb
TABLE IV
vl
00
OF HEATTRANSFER
EXPERIMENTAL
INVESTIGATIONSFOR GAS WITH VARIABLE PHYSICAL PROPERTIES
MAIN RESULTS
Ref.
Il'in (48)
Gas
Air
lld
59, 62
Reb .
7-60
Nu,,
0. j6-2.3
R c " . ~8"
0.5-0.9
0.9-1.2
1.2-2.3
0.0218
0.0212
0.0223
n
0
-0.27
-0.58
c
Air
30-1 20
7-300
0.46-3.5
> 60)
n = Oat0 < 1
n
Air
Hydrogen
29-72
124-435
1.1-1.73
1-4
-0.55
Hclium
60,92
Hydrogen
43, 67
Hydrogen,
helium
21-67
Nu
7-1500
1.5-2.8
5-1 500
1.5-9.9
Nu
0.021
0.045
Prt4
d
x
Pr;,' 0-".5
1.6-3.9
3.2-60
a t great I/d
cn
3:
at 0 > 1
Nub - 0.022
X 0 ' 1 5
0-O.'
Helium,
12.8
250
Nuf
hydrogen
Taylor (56)
Hydrogen,
0.021
Prf0.4
3:
m
77
1.5-5.6
160
1.1-2.5
Nuf
0.021
Pr;.*
helium
McEligot, Magee, and
Leppert (57)
Kirillov and Malugin (58)
Air, helium,
Nitrogen
138
7- 160
I . 1-2.3
Nu,,
Fedotov (59)
Air, carbon
77-206
14-600
1.1-2.7
Nu
=
b
3.
=!
0
z
30
at x/d
dioxide, argon
zU
nitrogen
> 50 and M
<1
Perkins and
Nitrogen
160
18-280
1.3-7.5
Nu
Air
48-370
14-400
1.1-2.1
m
c
r
at x / d > 40
Worsoe-Schmidt (60)
Volkov and Ivanov (38)
0.024 ReO.fl
b
Nitrogen
80-100
13-300
1-6
Nu
TZ =
'
fl
Prl 8 - O
r
m
r
55
-Q
100
+ 0.205)
>
:
80
<
ul
540
B. S. PETUKHOV
B. EXPERIMENTAL
DATAAND EMPIRICAL
EQUATIONS
A great number of experimental papers are devoted to the investigation
of the heat transfer between the tube wall and the gas flow at large
temperature differences when physical properties cannot be considered
constant. Table IV presents a schematic summary of the main results.
Heat transfer in the case of gas heating for constant heat flux at the wall
(and sometimes for variable qw) is the object of the majority of the papers.
In the case of gas cooling heat transfer, experimental data are neither
numerous nor complete. In the earlier papers investigators measured the
average (along tube) heat transfer coefficients. T h e data obtained from
these investigations, especially in short tubes, do not reveal the real
relation between the local Nu and 8. Therefore, in recent papers the local
Heat transfer
heat transfer is the prime subject of the investigati~n.~
measurements were performed mainly with diatomic gases (air, N, , H,)
and to a lesser degree with monatomic (He, Ar) and triatomic (CO,) gases.
T h e experiments cover the temperature ratio parameter range approximately from 0.4 to 4. In some papers the values of 8 approached 6 and
even 10. But such high values of 0 were as a rule found in the entrance
region of the tube.
T h e results of heat transfer measurements at large temperature
differences between the wall and the gas flow are usually presented as
Nu b
=c
Re:*
Prt4 8"
(63)
Table I V does not provide heat transfer data in relatively short tubes, as heat transfer
expressions in the entrance region have not been discussed in this paper.
HEAITRANSFER
A N D FRICTION
I N TURBULENT
PIPE FLOW 541
10
I2
14
16
18 2 0
25
30
3 5
40
B. S. PETUKHOV
542
< <
h'Ub/NU,b
1.27 - 0.278
(64)
< <
< <
HEATTRANSFER
AND FRlc'TION
IN
TURBULENT
P I P E I;I.OlV
543
= &0.264
(66)
, H as indicated
But, when the experimental data arc plotted as ( , , / f " ,vs
by (60), there is considerable sc'itter i n the results. According t o the data
of Lel'chuk and Ilyadyakin, the exponent in Eq. (66) should be -0.16
(instead of -0.264), while AlcEligot's work (57) indicates a value of
-0.1 for 6
1.0-2.4. 'Thus the evperiniental data from Rets. ( 5 7 , 6 0 ,62)
differ significantly. T h e general trend of these papers is to suggest a
weaker relationship between (,,/[,,,, and 6 than predicted analytically,
thus indicating the need for further investigation.
A. ANALYTICAL
RESULTS
By heat transfer in single-phase media at subcritical states (or,
shortly, in the supercritical range) is meant that the heat transfer takes
place at supercritical pressures and at subcritical or pseudocritical
temperatures (i.e., at temperatures corresponding to the maximum heat
capacity at constant pressure).
We can explain heat transfer anomalies in the supercritical region by
the fact that physical properties in this region change considerably and
in a special way.
Figure 13 shows the behavior of physical properties with temperature.
T h i s figure presents the data on carbon dioxide at P = 100 bar. T h e
majority of the physical properties do not change monotonically: heat
capacity a t constant pressure has characteristic maxima; heat conductivity
and viscosity coefficients usually pass through a minimum.
Density changes considerahly but the volumetric coefficient of thermal
expansion attains a maximum. 'I'herefore, even though the tcniperature
difference in the flow is sniall (T,\ - T,, rn lO-2Oc'C), the physical
properties change considerably across the tube. For instance, if
T,, < Till < T , (here T,,,is a pseudocritical temperature) as the distance
from the wall increases, C,, increases rapidly, goes through a maximum,
and then decreases. I n the case of heating density increases rapidly from
the wall to the tube asis. T h e variation of physical properties along the
radius is especially significant a t high heat fluxes and at large temperature
differences between the wall and the fluid, respectively.
If heat transfer takes place at rather small temperature differences then
B. S. PETUKHOV
544
100 bar.
HEATTRANSFER
A N D FRICTION
IN TURBULENT
PIPE FLOW545
FIG. 14. I-Ieat transfer coeficicnt for carhon dioxide in the supercritical region calculated \vith the assumption of coristant physical properties ( d = 6.7 m n i , G
100 kgihr).
Deissler predicted heat transfer and skin friction for the case of
heated water over a temperature range of 204-65OoC, pressure being
equal to 344 bar. T h e results tvere correlated by using the reference
temperature which proved to be a complicated and nonmonotonic
function of T,,./T,, and T,, . lhis function is plotted in Fig. 18. Goldmann
also carried out heat transfer and skin friction calculations for the case of
heated water at a pressure of 344 bar, but over a wider temperature
range of 260-840 C. Ihe analytical solution of the heat transfer is
presented by the equation
which is based on the assumption that with both constant and variable
546
B. S. PETUKHOV
FIG. 15. Goldmanns analytical results for water at P = 344 bar (solid lines) in
comparison with experimental data (dotted lines; circles represent averaged experimental data). Ranges of paratneters: d = 1.27-1.9 mm, qw = (0.32-9.5) . lo6 W/m*;
= (2-4) lo3 kg/sec-m2.
,,w
FW
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 547
fU
0
Q
U
(u
70
4
2
FIG. 16. 01 vs t b for carbon dioxide both heated and cooled ( P = 78.5 bar,
W/m2: ( 1 ) 0.70; (2) 1.4; (3) 2.8;
G = 3.9 x
kglsec, d = 10 nim) at qw x
(4) -0.70; ( 5 ) - 1.4; (6) -2.8.
and
0.97
The results are presented as interpolation equations for heat transfer and
skin friction.
Comparison of the analytical results with the experimental data shows
that the relation of the Nu number to Re and Pr and to 5 and Re at
variable properties is approximately the same as at constant physcial
properties. Then the ratios Nu,,/Nu,, and [ , , / f o b depend only on
the behavior of the physical properties with temperature. For the given
liquid NuJNu,, is a function of T,,, T,,,, and P or T,/T,, , TWIT,,,, and
P/P,,, which is more convenient for the comparison of the data obtained
for different pressures and liquids.
In Figs. 15, 17, and 18, some predicted values are compared with
experimental data and empirical equations. T h e behavior of Nu,,/Nu,,
with T w / T mis clearly seen in Fig. 18. When Tb/Tlrl< 1 heat transfer
1, it reaches the
increases with T w / T m increasing; when T w / T ,
maximum and then decreases. If T,,/Tnl3 1 heat transfer decreases when
T w / T , increases (see Fig. 17).
= 2150kg/m2-sec,
FIG. 17. cy vs Tw for water at P = 345 bar, T w / T b = 1.25,
and d = 9.4 mm: (1) Deisslers predictions; (2) Goldmanns predictions; (3) Svenson
et al.s empirical equation.
HEATTRANSFER
A N D FRICTION
I N TURBULENT
PIPE FLOW 549
It can be seen from Fig. 17, taken from (67), that Deisslers theoretical
results produce lower heat transfer values in comparison with
experimental data (approximately two times lower). Bringer and
Smith (68), who calculated the heat transfer for heated carbon dioxide
by Deisslers method, have found a disagreement of 2 5 O / , with their
measured values. Having calculated the heat transfer by using Deisslers
method but Prandtls expression for the mixing length in a turbulent
core and approximating the influence of the density fluctuations, MelikPashaev (69) have obtained better agreement between calculated and
experimental data. Goldmanns predicted results (see Figs. 15 and 17) are
in good agreement with the experimental data when T , - T,, 5 100C.
Disagreement increases with increasing T, - T,, , reaching 25 o/o when
T , - T,, = 250-300C. Povovs theoretical results (66) are in agreement
with the experimental data for carbon dioxide within 20,,. Figure 18
illustrates this comparison.
Hsu and Smith (70) made an attempt to take into account (approximately) the influence of density fluctuations on the turbulent momentum
diffusivity. T h e calculations for carbon dioxide have shown that this
correction improves the agreement between the predicted and experimental heat transfer data. However, the predictions for hydrogen (72)
with a correction for density fluctuations displayed less favorable
agreement with experimental data than the same predictions without the
correction. Thus, the problem of accounting for fluctuations due to
density and other physical properties has not as yet been satisfactorily
solved.
One might be tempted to expect a substantial influence of free
convection on heat transfer in a supercritical forced flow, because
density changes considerably with temperature. Hsu and Smith (70)
analyzed this problem theoretically for the case of a turbulent flow of
heated carbon dioxide in a vertical tube. As should be expected, the
effect of free convection on heat transfer becomes more pronounced
with higher G r and lower Re. For example, when G r = 108 and
Re = lo4 the heat transfer rate increases approximately two times due
to free convection; when G r = 108 and Re = lo5 free convection does
not affect the heat transfer.
T h e analysis of the predicted results reveals that when physical
properties do not change considerably over the flow cross section it is
possible to satisfactorily describe the heat transfer mechanism in the
supercritical region (but not for all possible regimes of flow and heat
transfer). T h e disagreement found between the predicted values and the
experimental data can probably be attributed to the inexact analytical
methods and to some uncertainity in the estimation of the physical
550
B. S. PETUKHOV
B. EXPERIMENTAL
DATA AND EMPIRICALEQUATIONS
FOR NORMAL
HEATTRANSFER
REGIMES
At present there is a considerable number of experimental works on
heat transfer in the supercritical region. In the vast majority of these
works investigators analyzed the heat transfer for turbulent flows of
water and carbon dioxide in circular tubes for heating when qw = const.
Table V gives an incomplete summary of these works. Some papers are
devoted to the heat transfer for heated hydrogen (71, SS), oxygen (89)
and Freon-12 (90). Experimental data on the heat transfer of cooled
fluids are not available.
Judging from the experimental data obtained for small temperature
differences in a flow when the physical properties do not change
considerably, heat transfer in the supercritical region displays no
striking differences and is described well by the known relationships
for constant properties.
For rather large temperature differences in a flow, i.e., under the
conditions when the physical properties change significantly, all the
possible flow and heat transfer regimes can be classified in the following
l1 For example, according to some data, thermal conductivity at the pseudocritical
temperature has a pronounced maximum; according to other data, it changes monotonically. In this paper the latter is adopted. But in Ref. (72) it is shown that the consideration
of the thermal conductivity maximum makes no difference for theoretical predictions.
TABLE V
EXPERIMENTAL
HEATTRLVSFER
DATAOF WATERAND CARBON
DIOXIDE
AT SUPERCRITICAL
PARAMETERS
Tb/Tm
. 10-6 W/mZ
P;Pcr
T /Tm
1.01-1.16
0.9 1 --I. 12
0.87-1.09
35- I80
0.89-25
1 . 1 1.2
0.6- I .4
0.5 1.05
18-800
0.9-8.7
1.02- I . 25
0.8 1-1.17
0.97-1.12
13.-570
0.84-55
0.23-2.5
1.02-1.11
1.11-1.38
1.04-1.57
0.8 1 - 1.44
0.85-1.15
0.74-1.30
21-101
0.8-14
0.27-1.1
0.08-0.13
1.09-1.4
I .56
1.03-1.87
0.69-1.35
0.74- 1.6
0.56-2.1
0.59- I . 22
0.74-1.25
0.54-1.9
58-860
64- 1020
27-680
0.75-16
0.82-3.3
0.8-9.4
0.87-1.7
3.1-9.5
0.2-1.8
1.03-1.25
1.11
0.89- 1.35
0.84-1.33
0.87-1.22
0.6-1.25
36-530
12-400
0.89-16
0.9- 15.4
0.32-3.6
0.12-0.93
lirasnoshcht.kov and
Protopopov (83 8.5)
I .Oh- I .4h
0.98-2.6
0.96-1.7
65-500
0.85-40
0.02-2.5
Illellh-Iilshac\ ( 6 Y )
Bringer atid Smith ( 6 8 )
1.1-5.7
1.13
1.OO- 1.03
1.004
I .4-2.9
0.99- 1-06
0.9 1-1 - 2
0.97- I .02
0.95-1.06
0.97
0.84-1.31
Author
Reb
prb
qw
( a ) Water
Arinand. T a r a s o v a , and
I i 0 I ~ k O V(73)
Vikhrev, U;irulin, and
Kinnkov (74)
Mirupolskl- and Shitsman
(75. 76)
Shitsman (77)
Aladev. Veltishchev, and
Kondratier (78)
Dickinson and Welch (79)
Chalfant and Randall (80, 81)
Swenson, Carver, and
Kakarala (67)
Bischop and Sandberg (67)
Yamagata, Nishikawa,
Hasegawa. and Fujii (82)
0.18--0.73
0.23-1.25
( b ) Carbon dioxode
1.51
0.97-1.05
0.84-1.32
150-650
38-270
30-300
910-950
52-400
I .(bKI
2.6-7.4
0.9-1 1
0.85-10.5
2.5-10.5
9.5
0.02--0.31
0.06-0.63
0.01-0.07
0.002-0.25
up to
z
24
552
B. S. PETUKHOV
way: (a) normal regimes, (b) regimes with diminished heat transfer,
(c) regimes with enhanced heat transfer.
Normal regimes are those for which the relations observed in
experiments can be explained and analyzed by the existing concepts of
turbulent flow and heat transfer formulations for variable physical
properties.
Regimes with diminished and enhanced heat transfer are those with
lower or higher heat transfer coefficients, respectively, in comparison
with those which should have been expected if the regime had been
normal. Such a classification of flow and heat transfer regimes is, of
course, rather conventional and most likely it is not a matter of principle
differences between these regimes but has appeared as a result of
insufficient knowledge of the actual mechanism of turbulent diffusivity
with variable physical properties. A rather elementary model of turbulent
diffusivity used at present allows us to describe normal regimes satisfactorily, but it appears insufficient and inadequate for description of
more complicated phenomena involved in regimes with diminished and
enhanced heat transfer.
In heat exchanger systems at supercritical states, normal regimes occur
mostly. Heat transfer formulations for these regimes are essentially in
good agreement with predicted results. Several empirical equations are
known for the calculation of heat transfer in the normal regime. One of
the first equations was suggested by Miropolskii and Shitsman (75, 9Z).
This equation has the form
Nul) = 0.023 Re:*
(68)
where Prmin is the smaller of the two Pr numbers calculated from the
bulk temperature (Pr,,) and the wall temperature (Pr,).
Krasnoshchekov and Protopopov (85), using their experimental data
for CO, and the experimental data of other workers for CO, and H,O,
suggested the equation
Nut)
NU~h(pw/pb)'.~
(cD/Cnb)R
(69)
HEATTRANSFER
AND FRICTION
IN TURBULENT
PIPE FLOW 553
FIG. 19.
17
vs TWITmand T b / T m .
plotted in Fig. 19. Equation (69) describes the experimental data with an
accuracy of & 15 O 0 and I S v d l i d for the following range of
1.01
I,Pqr
2 . lo4
0.02
Reb
c,/c,b
1.33,
0.6
8 . 10;
S 4.0;
Y I ) / T<~ ~1.2;
0.85
2.3
. lo4
Irb
~
qw
Swenson ~t al. (67) have correlated their data on the heat transfer for
water by the equation
B. S. PETUKHOV
554
KW
<b/'$ob
= (/%v/pb)0'22
(72)
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 555
0.6 x lo6-
zb
where
= 8'Tw/51,( p W ) * ; tot,
is calculated from Eq. (49) when Re =
Re, =
p,, ; jl,and F,, are averaged values of p,, and p,, over the
tube section under consideration (they are calculated from the prescribed
distribution of bulk enthalpy along the tube length); pw is the averaged
value of pw in this section.
iwd
C. EXPERIMENTAL
DATAFOR REGIMESWITH DIMINISHED
A N D ENHANCED HEAT TRANSFER
A number of experimental studies (74, 77, 82, 89, 95) have led to the
discovery of diminished heat transfer regimes. Such regimes are seen
very clearly in Shitsman's experiinents (77) with water at P -- 226-245
bar. For fluid heating a t T I , T,,, and a certain combination of mass
velocity and wall heat flux, heat transfer sharply decreases and the wall
temperature increases rapidly (in the case of pw = const).'" T h e heat
transfer and wall temperature found in this case are not described by the
relations valid for the normal heat transfer regime.
I n Fig. 22 the t, and t,, variations along the tube length are plotted for
upward liquid flow for several regimes of normal and diminished heat
transfer (77). When qw = 221 x lo3 and 281 x lo3 W/m2 (Curves 1
and 2), normal regimes are observed. But when qw = 300 . lo3 W/m2
(Curve 3 ) , the wall temperature in the tube section length 10d-15d
increases sharply and becomes approximately 100C higher than the
expected one. With further increase in heat flux (Curves 4 and 5 ) , the
l2
than
Usually (but not always) diminished regimes occur in the cases when T , is higher
Tm .
556
B. S. PETUKHOV
FIG.22. Distribution of ta (solid lines) and t b (dotted lines) along the tube length
for water at P = 226 bar, PW = 430 kg/m2-sec, and qw x lo-$ W/m2: 1 (w), 221;
2 (v), 281; 3 ( O ) , 300; 4 (o), 337; 5 ( 0 ) , 386.
HEATTRANSFER
AND FRICTION
IN TURBULENT
PIPE FLOW 557
FIG. 23. Limiting values of heat fluxes for the downward flow of water in a tube d
( d = 8 mm) with
measured in kg/m2 sec: (1) 430; (2) 700; (3) 950; (4) 1500. P i s
equal to: (0)
226, ( 0 ) 235, ( 0 ) 245, ( v ) 294, and ( 0 ) 343 bar.
r
w
the wall temperature begins increasing r a ~ i d 1 y . lA~ straight line corresponds to each value of bulk velocity; in other words, the relation of
qZm vs h, can approximately be considered linear. It is of interest to
note that qGn at prescribed values of h, and
does not depend on the
pressure, at least over the pressure range of 226-343 bar. T h e area below
the appropriate curves in Fig. 23 corresponds to normal heat transfer
regimes (from the start of heating to the section under consideration
l3
558
13.
S. PETUKHOV
which naturally can also be the outlet); the area above these curves
corresponds to decreased heat transfer regimes.
When decreased heat transfer regimes appear i n a certain range of
heat fluxes and other parameters, considerable pressure fluctuations a s
well as fluctuations in TrVand T , at the entrance are observed. For
example, according to (77), when P = 245 bar the amplitude of the
pressure fluctuations reached 25 bar.
T h e appearance of diminished heat transfer regimes can probably be
attributed to the change in flow mechanisms and, in particular, turbulent
diffusivity processes caused by large changes in the physical properties
over the flow cross section. Free convection appearing as a consequence
of large density gradients can affect considerably the flow regime. This
is confirmed by the fact that diminished heat transfer regimes are
observed in an upward flow in heated tubes and are not found in a
downward flow. In the latter case normal regimes occur under the same
conditions which may be apparently connected with niore intensive
fluid mixing when the directions of forced and free convection at the wall
are opposite. T h e small amount of available data ( 9 6 ) shows that, in the
case of a horizontal flow, diminished heat transfer regimes can occur too,
though not in a very pronounced manner.
Improved heat transfer regimes were discovered by Goldmann
( 6 4 ) and again later on in other works; for example, in (97). T h e
experiments described hy Goldmann were performed on a water flow
in an electrically heated tube 1.58 m m in diameter and 203 m m in
length. Figure 24 is a typical plot of liquid and wall temperature distribution along the tube length when P = 345 bar a n d g v
2 . lo3 kg/sec-m2.
FIG. 24. Distribution of 7; (solid lines) and T I ,(dotted lines) along the tube length
at the onset of the impnived heat transfer regime: ( I ) q w
3.54 x lo6 W / m 2 ; (2)
gw
3.67 x lo6\V/ni2. I<spcrinierits\vith water at P = 345 bar. PW
2 x lo3kg/sec-m2,
and d :
1.58 mni.
~
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 559
When q, = 3.54 x lo6 W;m2 (Curve l), the wall temperature increases
along the whole tube and is described by the relationships valid for the
normal heat transfer regime. But _with
_
a small increase in heat flux to
qw = 3.67 x lo6 W, m2(when Z, pW, and T,) are the same as at the
entrance) the wall temperature at a distance x c
z 125 m m from the
entrance (after a small local increase) decreases rapidly and over the rest
of the tube, which is 50d in length, has a constant value. I n this case the
constant value T , = 733K is a little bit higher than T,,, N 691K.
Heat transfer in the region with constant T , is much higher and T , is
much lower in comparison with the values which should have been
expected for the normal regime.
Improved heat transfer regimes are always accompanied by a loud
sound similar to a sharp whistle with a frequency of approximately
1400-2200 Hz.
Improved heat transfer regimes arise only with high heat fluxes
(usually higher than 3 :*: lo6 Wjm) and when T,, < T,, after the wall
temperature reaches a value somewhat lower or, more often, somewhat
higher than T, . Heat fluxes corresponding to the onset of improved
heat transfer, all other conditions being equal, increase with increasing
Pandand decrease when the fluid temperature at the entrance
increases. When improved heat transfer regimes occur, a further increase
of heat flux causes the wall temperature as a whole to increase in the
region adjacent to the exit, while for a given heat flux it remains constant
along the length of this region.
Apparently heat transfer improvement is connected with a rather
considerable change in physical properties (mainly p and C],)over the
flow cross section at high heat fluxes. Under the conditions when
improved heat transfer regimes are observed, fluid density at the wall is
several times smaller and heat capacity is several times higher than in the
core. Fluid particles coming from the core to the hot wall due to
turbulent diffusivity possess relatively high thermal conductivity and
small heat capacity. With high temperature differences between the
layer adjacent to the wall and the particles which come from the core, the
particles (64) are heated rapidly and explosively dilated. This process
yields intensive fluid mixing in the wall layer and, consequently, higher
heat transfer. T h e existing theory neglects the above-mentioned heat
transfer mechanisms with variable physical properties. Therefore,
the theory is not yet able to describe improved heat transfer regimes
satisfactorily.
For an interpretation of hcat transfer diminution and enhancement
some investigators use the analogy with surface liquid boiling. They
identify diminuation with the heat transfer crisis of subcooled liquid
560
B. S. PETUKHOV
VII. Conclusion
All that has been said regarding heat transfer in turbulent pipe flow
can be summarized in the following manner. T h e analytical methods
allow us to describe heat transfer mechanisms for constant liquid
properties quite satisfactorily and to take into account the influence of
the variation of physical properties with temperature versus heat
transfer and skin friction in a number of important cases. Disagreement
between theoretical and experimental results observed in other cases,
in particular, with a considerable change in physical properties over the
flow cross section, may be attributed to imperfect methods of estimating
the effect of the variation of physical properties on turbulent diffusivity.
Therefore, further refinement of the analytical methods demands
enhanced study of turbulent diffusivity with respect to variable physical
properties.
Important experimental material has been accumulated on heat
transfer and skin friction for variable physical properties. However,
certain portions of this material possess relatively low accuracy that
prevents its successful use. For a number of important cases there has
been no systematic data collection, or that which is available is scanty
and contradictory. Therefore, the need for further experimental
investigations, with a high degree of accuracy, into the fluid mechanics
and heat transfer for variable physical properties is quite urgent.
l4 T h e former takes place at relatively lower heat fluxes; the latter occurs at much
higher heat fluxes.
HEATTRANSFER
AND FRICTION
IN TURBULENT
PIPE FLOW 561
NOMENCLATURE
heat transfer coefficient
the ratio of t h e eddy
diffusivity of heat to th e eddy
diffusivity of momentum
C , specific heat at constant press ure
d=2r0, tube diameter
q, eddy diffusivity of heat
C~
eddy diffusivity of momentum
7 v * y / v , universal coordinate
Wz/u*, dimensionless velocity
p, 171 Goldmanns variables; see Eqs.
(24)
G mass flow rate
Gr Grashof num b er
h enthalpy
K
const
I tube length
h thermal conductivity
p dynamic viscosity
ISu Nusselt num b er
Y
kinematic viscosity
6 friction factor,
--I
80p/(p12,)
P r Prandtl num b er
P pressure
y heat flux
Re Reynolds number
I?= r i r , , dimensionless radial coordinate
r coordinate along radius
ti
,~=c,/c,,
density
mass velocity
s,average
given cross section
over a
S c Schmidt n u mb er
Sh Sherwood n u mb er
u shear stress
T temperature, K
t temperature, C
O= T w / T b ,temperature ratio parameter
w* friction velocity
W , , W , velocity vector components
x coordinate along tube axis
.%=(Tz- Tb)/(Tw- Tb), dimensionless characteristic temperature
y coordinate along the normal to
t h e wall measured from the
wall
SUBSCRIPTS
b
cr
m
0
w
x
REFERENCES
I . \V. Nusselt, Mitt. Forsch.-Arb. Ifig.-Hes. (VDZ-Forsch.-Heft) No. 89, 1-38 (1910).
2. 0. Reynolds, Scientific Papers. Vol. 1 . Cambridge Univ. Press, London an d New
York, 1901.
3. G. Y . Taylor, T ech . Rept. Adv. C o m m . Aer, Vol. XI, pp. 423-429. Rep. M e m . No.
272 (hlay 1916).
4. L. Prandtl, Phys. Zs. B d . 1 1 , 1072-1078 (1910).
5 . L. Prandtl, Phys. Z s , Bd. 29, 487-489 (1928).
6. T . Karman, Trans. A S M E 61, 705-710 (1939).
7. H. Reichardt, A r c h . Ges. ~+orrwtrchnik,6 7 , 129-143 (1951).
8. R. N. Lyon, Chrm. Eiig. P r o p . 47, No. 2 (1951).
9. R. A. Seban and T. T . Shimazaki, Trans. A S M E 7 3 . No. 6, 803-809 (1951).
10. R. G. Deissler, Trans. A S M E 7 3 . No. 2, 101-107 (1951).
562
B. S. PETUKHOV
11.
12.
13.
14.
HEATTRANSFER
AND FRICTION
I N TURBULENT
PIPE FLOW 563
44.
45.
46.
47.
564
B. S. PETUKHOV