19690021129

TECH LIBRARY F ,N
1 0069025
A1taft 1- A r -;)fj6
LIQUID METALS
By P. L. Kirillov, V. I. Subbotin, and P. A. Ushakov, Editors
Translation of "Zhidkiye Metally."

Atomizdat, Moscow, 1967
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
For sale by the Clearinghouse for Federal Scientific and Technical Information
Springfield, Virginia 22151 CFSTI price $3.00
Page intentionally left blank
TABLE OF CONTENTS
Page
List of Symbols, Standard Throughout the Book vii
Foreword viii
Part One 1
Heat Transfer in Round, Annular and Square Channels
Study of Heat Transfer in Liquid Metals in Round
Pipes (V. D. Talanov and P.A. Ushakov) 1
Heat Transfer with Liquid Metal in a Vertical Pipe
with Low Peclet Numbers (N. A. Ampleyev, P. L. Kirillov,
V. I. Subbotin and M. Ya. Suvorov) 8
Heat Transfer of Sodium and a Sodium-Potassium Alloy
in Short Vertical Pipes by Combined Convection
(L. G. Volchkov, M. K. Gorchakov, P. L. Kirillov and
F. A. Kozlov) 28
Taking into Account Longitudinal Heat Flows when Measuring
the Average Temperature of Fluids in Pipes and Heat
Exchanges for Low Peclet Numbers (M. K. Gorchakov) 40
Statistical Characteristics of Turbulent Temperature
Fluctuations in Fluid Flows (V. P. Bobkov, M. Kh. Ibragimov
and V.I. Subbotin) 53
Statistical Characteristics of Temperature Fluctuations in
Heat Exchanger Walls Attendant to High Thermal Fluxes
(M. Kh. Ibragimov, V. I. Merkulov and V. I. Subbotin) 71
Heat Transfer in Turbulent Flow of Fluids in Square Ducts
(M. Kh. Ibragimov, Yu. M. Rabotyashnik, G. I. Sabelev,
V. 1. Sidirov and V. I. Subbotin) 82
Fluid Dynamics in an Annular Space with Longitudinal Ribs
(Yu. D. Levchenko, V. I. Subbotin, P. A. Ushakov and
A. V. Sheynina) 100
Effect of Eccentricity on Heat Transfer to Liquid Metals
in Annular Spaces (V. I. Subbotin, V. D. Talanov and
P.A. Ushakov) 108
Transfer of Heat in Turbulent Flows of Fluids in Annular
and Plane Spaces (N. I. Buleyev, V. A. Mosolova and
L. D. Yel'tsova) 120
iii
F
Experimental Study of Heat Transfer From a Flat

Horizontal Surface to Sodium by Free Convection
(A. P. Kudryatsev, D. M. Ovechkin, D. N. Sorokin,
V. I. Subbotin and A.A. Tsyganok) 130
Part Two
Heat Transfer and Fluid Dynamics of Intricately-Shaped Passages
Approximate Thermal Simulation of Cylindrical Fuel-Element
Rods (P. A. Ushakov) 136
Heat Transfer From Loosely-Spaced [Fuel] Rod Clusters to
Liquid Metal Flowing in the Axial Direction (A. V. Zhukov,
V.I. Subbotin and P.A. Ushakov) 149
Experimental Study of Temperature Fields of Fuel Elements,
Using Models (A. V. Zhukov, L. K. Kudryatseva, Ye. Ya. Sviri-
denko, V. I. Subbotin, V. D. Talanov and P. A. Ushakov) 170
Heat Transfer in the Intertubular Space of Liquid-Metal Heat
Exchanges (P. L. Kirillov and M. Ya. Suvorov) 190
Analysis of the Applicability of Average Heat-Transfer
Coefficients in Heat-Exchanger Design (P. A. Ushakov) 200
Resistance to Axial Flow of Fluids Past Rod Clusters
(A. V. Sheynina) 207
Velocity Distribution in a Cell Formed by a Densely-
Packed Cluster of Rods (Yu. D. Levchenko, V. I. Subbotin,
P. A. Ushakov and A. V. Sheynina) 220
Calculation and Experimental Study of Velocity Fields in
Intricately-Shaped Passages (M. Kh. Ibragimov, I. A. Isupov
Part Three
Boiling and Condensation of Metals
Electronically-Heated Setup for the Study of Heat Transfer
From Boiling Metals by Free Convection (A. A. Ivashkevich,
A, P. Kudryavtsev, D. M. Ovechkin, D. N. Sorokin,
V. I. Subbotin and A. A. Tsyganok) 248
Transfer of Heat From Sodium Boiling in a Large Vessel
(A. P. Kudryavtsev, D. M. Ovechkin, D. N. Sorokin,
Experimental Study of Critical Thermal Fluxes in Sodium
Boiling in a Large Vessel (A. P. Kudryavtsev, D. M. Ovechkin,
D. N. Sorokin, V. I. Subbotin and A. A. Tsyganok) 281
iv
Mean Temperature Difference in Low-Pressure Vapor
Generators and Condensers (P. L. Kirillov) 287
Transfer of Heat From Condensing Sodium Vapor
(N. B. Bakulin, M. N. Ivanovskiy, V. P. Sorokin,
V. I. Subbotin and B. A. Chulkov) 294
Part Four
Techniques of Work With Liquid Metals
Corrosion Resistance of Construction Materials to
Sodium (V. V. Zotov, B. A. Nevzorov and Ye. V. Umnyashkin) 300
Determination of the Oxygen Content of Sodium and of the
Sodium- Potassium Alloy by Vacuum Distillation (I. B. Dmitryeva,
F. A. Kozlov and E. K. Kuznetsov) 318
The Cock-Type Indicator as an Instrument for Detecting
Admixtures in Sodium (F. A. Kozlov, E. K. Kuznetsov and
V. I. Subbotin) 324
Effect of Admixtures on the Electric Resistance of Molten
Lithium (M. N. Arnol'dov, M. N. Ivanovskiy, V.I. Subbotin
and B. A. Shmatko) 336
The Use of Hot Traps for Removing Oxygen From Sodium
(F. A. Kozlov and E. K. Kuznetsov) 340
Maximum Permissible Leakage in a Sodium-to-Water Steam
Generator (N. N. Ivanovskiy and F. A. Kozlov) 357
Some Problems of Operation and Design of Liquid-Metal
Loops (A. V. Drobyshev, V. A. Kurov and I. T. Filipov) 360
Heat Exchangers of Experimental Liquid-Metal Loops
(V. A. Kurov and V. F. Men'shikov) 368
Study of Velocity Fields by the Electromagnetic Method
(N. N. Turchin and R. V. Shumskiy) 377
Electromagnetic A-C Pump for Liquid Metals Rated at
120 m3/hour (Ye. M. Avilova, T. V. Doktorova, N. I. Marin
and V.A. Povsten') 385
Electromagnetic Screw Pump for Pumping of Liquid Metals
(Ye. M. Avilova, Ye. L. Ivanter, N. I. Marin and V. A. Povsten') 389
Hydrostatic Bearings in Sodium Pumps (N. N. Aristarkhov,
Yu. Ye. Bagdasarov, Yu. L. Kulikov, Ye. N. Larin,
M. S. Pinkhasik, A. A. Rineyskiy and B. S. Tymosh) 392
v
3
Flowmeter for Measuring Large Liquid-Metal Flows

(N. L. Loginov) 403
Time Needed for Melting Cylindrical [Metal] Volumes
(V. M. Selivanov, N. M. Trebina and V. I. Sharypin) 406
Frozen Sodium Seals (A. V. Drobyshev, Yu. 1. Likhachev
and I. M. Saratov) 412
Appendices
I. Calculations of Heat Transfer in Liquid Metals 437
II. Corrosion Resistance and Temperature Limit of Utilization
of Certain Construction Materials in a Sodium Flow 446
Abstracts 448
vi
Nu = adA Nusselt number
Pe = wd/a Peclet number
Pr = v/a Prandtl number
Gz = Pe d/1 Gratz number
Re = wd/v Reynolds number
Gr = a13Atg/v2 Grashof number
Bi = adAw Biot number
q, qv, Specific heat load.

Ra = Gr Pr Rayleigh number
a Heat-transfer coefficient
v Thermal conductivity coefficient
Cp Specific heat
v Specific weight
a Coefficient of thermal diffusivity
v Kinematic viscosity
Dynamic viscosity
a Coefficient of volumetric expansion
g Acceleration of gravity
d Tube diameter
F Heat transfer area
t Temperature, C
At Mean temperature difference, C
vii
FOREWORD
The production of atomic [nuclear] power may become an important branch

of the national economy and satisfy to a substantial degree the requirements for
electric power and heat, provided that economic nuclear power generation sta-
tions are constructed, in particular those using fast-neutron reactors. Such
reactors, because of their physical characteristics, make it possible to utilize
almost all the uranium, as compared with the negligible part of this element which
is utilized by thermal neutron reactors; in addition, thorium can be used in
fast-neutron reactors. These remarkable properties increase the supplies of
utilizable nuclear fuel to such an extent that it becomes entirely possible to set
up a large atomic [nuclear] power producing system on a "country-wide" scale,
and provided with fuel for long periods of time. Hence, the development of
fast-neutron reactors is one of the primary goals of the nuclear power technology.
The optimal heat transfer agent in these reactors is liquid sodium, due to
the favorable combination of properties such as the nuclear-physical, thermo-
physical and corrosion, as well as because of its comparatively low cost. While
the use of liquid sodium has little effect on the nuclear properties of reactors, it
permits the removal of large specific quantities of heat (500-1000 kw per liter
in the active zone of the reactor).
One can expect that the sodium temperature at the reactor outlet will be
500-650 C, whichwill make it possible to obtain steam of sufficiently high pres-
sures to operate the power station at high efficiency.
Studies in the field of thermal physics and technology of sodium, study of
thermal processes in active zones, in reactor reflectors, heat exchangers, steam
generators and other equipment are tasks of paramount importance, Successful
solution of these tasks is closely tied to the development of fast-neutron reactor
technology.
In the Physics and Power Institute, where the development of fast-neutron
reactors is regarded as a major task, a large group of investigators has been
working for several years on the thermal physics and technology of liquid metal
heat transfer agents. Some results of this work are published in the present col-
lection.
Studies in this field already permit the practical utilization of liquid
metal heat transfer agents on an appreciable scale. This is attested to by the
successful operation for many years of the BR-5 experimental fast-neutron reac-
tor. The content of the present collection reflects the level of research work in
this field.
The problems of thermal physics of real active zones and reflectors of re-
actors as well as of impurity removal from the liquid metals, development of
Viii
of standard technologies and corrosion studies of structural materials are being
successfully solved. They serve as a basis for optimism as to the feasibility of
developing efficient electric power stations employing fast-neutron reactors
within the next few years.
A. I. Leypunskiy
ix
i
I
I
I
WI
JI
x
Î
FBI
uZI
O
QI
F
Î
U
I
FOLD LINE
-- -- -------------
I NATIONAL AERONAUTICS AND SPACE ADMINISTRATION POSTAGE AND FEES PAID
WASHINGTON, D.C. 20546 NATIONAL AERONAUTICS & SPACE ADMINISTRATION
OFFICIAL BUSINESS
I
I
I
I
i
I
I
I NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
I CODE USS-T
I WASHINGTON, D.C. 20546
I
I
I
I
NASA TTF No.
522
I- - - - - - -- - - FOLD LINE
I
NATIONAL, AERONAUTICS AND SPACE ADMINISTRATION Budget Bureau No. 104 -8037
TF HN'rAL TRANSLATInN EVALUATIO N I Approval Expires: Sept. 30, 1969
TO:
I
NASA TTF NO.
THE USERS OF THIS TRANSLATION I
522
I
MAINTAINING THE QUALITY OF NASA TRANSLATIONS REQUIRES A CONTINUOUS EVALUATION PROGRAM. PLEASE IW
COMPLETE AND MAIL THIS FORM TO AID IN THE EVALUATION OF THE USEFULNESS AND QUALITY OF THE I?
TRANSLATING SERVICE. I^
H
1=
THIS PUBLICATION (Check one or more)
a FURNISHED VALUABLE NEW DATA OR A NEW APPROACH TO RESEARCH. 10
VERIFIED INFORMATION AVAILABLE FROM OTHER SOURCES. la

H
i^
U
I
FURNISHED INTERESTING BACKGROUND INFORMATION.
I
I
OTHER (Explain): I
I
I
FOLD LINE FOLD LINE 1
I
TRANSLATION TEXT (Check one) I
I
IS TECHNICALLY ACCURATE. I
I
IS SUFFICIENTLY ACCURATE FOR OUR PURPOSE. I
I
U IS SATISFACTORY, BUT CONTAINS MINOR ERRORS. I
L^ IS UNSATISFACTORY BECAUSE OF (Check one or more): i
POOR TERMINOLOGY. 0 NUMERICAL INACCURACIES. I

1
INCOMPLETE TRANSLATION. E- 1 ILLEGIBLE SYMBOLS, TABULATIONS,
OR CURVES. 1
I
OTHER (Explain):
I
I
I
I
I
I
I
I
I
I
FOLD LINE FOLD LINE
I
REMARKS
I
i
W
Iz
I -j
1=
I~
is
IJ
I^
IU
I
FROM DATE
I
I
NOTE: REMOVE THIS SHEET FROM THE PUBLICATION, FOLD AS INDICATED, STAPLE OR TAPE, ANDMAIL. I
NO POSTAGE NECESSARY.
I
NASA Form 1332 AU G 66 GP 0 9T5-777
I
NASA TT F-522
ANNOTATION
This collection presents results of original studies performed during the

last few years at the Physics and Power Institute of the State Committee for
Utilization of Atomic Energy.
The collection contains theoretical and experimental solutions of problems
of convective heat transfer, pertaining to the flow of liquid metals in ducts of
various shapes. Of particular importance are ducts formed by the cluster of
round rods, since this is the most prevalent form of fuel elements in active
zones of nuclear reactors.
This book considers problems of heat transfer and fluid flow attendant to
fluid flow past clusters of round rods in the longitudinal direction, as well as
problems of boiling and condensation of metals. It presents techniques of work-
ing with liquid metals, methods for removing impurities from them, methods of
impurity control, problems of corrosion and the experience in operating various
kinds of equipment used in experimental facilities of the Physics and Power In-
stitute.
This collection illuminates,' problems of design of facilities employing liquid
metals. It presents recommendations on calculating the heat transfer in liquid
metals and a summary of physical properties, which were used in analyzingthe
experimental results.
The materials in this collection can be used extensively for the development
of facilities in which liquid metals are used as heat transfer agents.
PART ONE
HEAT TRANSFER IN ROUND, ANNULAR AND SQUARE CHANNELS
STUDY OF HEAT TRANSFER IN LIQUID METALS IN ROUND PIPES
V. D. Talanov and P. A. Ushakov
Investigations have been carried out at the Physics and Power Institute con- /9
cerned with the heat transfer to mercury flowing inside tubes made of nickel and
stainless steel, as well as with heat transfer to a sodium-potassium alloy flowing
Numbers in the margin indicate pagination in the foreign text.
inside a copper tube [1]. The experimentally obtained data for different conditions
were in good agreement with each other and are described by the following [empi-
rical] formula
Nu = 4.3 + 0.025Pe O.8 . (1)
Using data for sodium, whose Prandtl number is much lower than that correspond-
ing to mercury or to the sodium-potassium alloy, it was recommended in [1] to
calculate the heat transfer to liquid metals in tubes by using the formula
Nu = 5 4 . 0.025Pe O.8 . (2)
The sodium-potassium alloy wets the copper surface well, while the sur-
face of stainless steel is not wetted by this low-temperature alloy. This was
experimentally confirmed.
To obtain data on heat transfer between the stainless steel wall and the so-
dium-potassium alloy two series of tests were performed. In them use was made
of an eutectic alloy(22%Na +78%K) at a temperature not above 100C [2]. Before
pouring the alloy into the circuit, it was passed through a wire-gauze filter and the
oxides were continuously removed by a cold trap. The oxygen content of the alloy,
which was determined by a distillate sampling device, varied from 3 . 10 -4 to 10-3%
by weight. The design of the testing tube is shown in Fig. 1. The calibrated tube
from 1Kh18N9T stainless steel, 19.5 mm in diameter and with a wall thickness of
0.4 mm, was plated by a copper layer of about 7.8 mm by the zone melting
method. The entrance flow stabilization segment comprised 20d, the heating
length was about 940 mm (about 50d). The pipe was provided at its ends with
mixing chambers in which the mixed mean temperature of the sodium-potassium
coolant was measured. The outlet mixing chamber is removed from the end of the
heated segment by approximately 10d.
To measure the pipe wall temperature six thermocouples, two of which could /10
be moved along the entire heated segment, were built into the copper shell. The
movable thermocouples were placed in copper capillaries 1.5/0.6 mm in diameter,
inserted at a distance of 2.4mm from the heat transfer surface and placed in lon-
gitudinal slots in diametrically opposite sides of the pipe. The slots and the cop-
per plates by which the capillaries were calked were rounded off along the radii of
the latter, which improved the thermal contact between the capillaries and the cop-
per shell. In addition, when inserted into the pipe wall, the capillaries were sol-
dered with tin.
The movable thermocouples were made from copper and constantan wires
0.09 in diameter, pulled through a capillary from 1Kh18N9T steel 0.5/0.3 mm in
diameter, and welded to it at one end. The thermocouples were moved auto-
matically by means of a device consisting of an electric motor, speed reducer,
cam, Maltese-cross type gear and a drum on which the driving cable was wound.
On each revolution of the gear the thermocouples, moved through 20 mm. The time
2
elapsed between the thermocouple motions ( ~30 secs) was sufficient to stabilize
their readings and for making the necessary measurements.
Stationary Movable
Thermocouple Thermocouple
^T 60 0.5/0,3
.n N N
95
S Seco 6 "
1A -A
s
^1
Figure 1. Schematic Diagram of the Test Pipe. 1) Thermo-

couple Sleeve; 2) End Chamber; 3) Stainless Steel Pipe;
4) Copper Shell; 5) Movable Thermocouples; 6) Stationary
Thermocouples
In order to check the readings of the movable thermocouples, use was

made of four [additional] thermocouples, which were built into the wall at dis-
tances of 440, 540, 640 and 740mm from the beginning of the heated segment.
The distance from the junctions of these thermocouples to the heat-transfer sur-
face was 2.4 mm. The copper capillary served as one of the. thermocouple elec-
trodes, while the other was Kopel wire in alundum insulation. Preliminary tests
produced good agreement in the readings of the movable and the stationary ther-
mocouples, hence the main measurements were made by using the former.
The test pipe was heated by a band-type nichrome heater with a cross sec- /11
tion of 6 x 0. 25 mm insulated by an organic silicon compound and micanite. Heat
losses to the surrounding medium were reduced to a minimum by the thermo-in-
sulation and by a compensating heater, which was automatically adjusted by a
system of thermometers and devices. The power of the heater was adjusted by
the AOSK-25/0.5 transformer and was measured by class 0.5 astatic instruments.
The thermocouple emfs were measured by R-2/1 semiautomatic potentiome-
ters. The flow of rate of the sodium-potassium alloy was measured by magnetic
flowmeters and on the basis of the heat content of the alloy flowing in the test
pipe (calorimetric method) .
The double-wall pipe was made as follows. A calibrated pipe from the 1Kh18N9T
stainless steel was placed coaxially into a thin-walled cylindrical shell to which it
was then welded at the bottom, and [then] copper rings were inserted into the annual
space [thus formed]. The entire structure was suspended. from a molybdenum wire
pulled. through an auxiliary copper tube, the upper end of which was rigidly connect-
ed to the structure, while the lower end was placed on a ,washer fastened to the molyb-
denum wire.
3
The structure was subjected to zonal melting in a small shaft-type furnace.
As the copper rings, and after them the auxiliary copper tube, melted, the struc-
ture lowered itself under its own weight. The heating zone thus moved along the
length of the pipe in . the upward direction and after leaving the furnace it was cooled
by natural convection of the surrounding air. After the section was copper plated in
this manner the external steel shell was removed on a lathe.
The experimental specimens exhibited good diffusion bonding of the copper
with the stainless steel. Penetration of stainless steel components into the cop-
per and vice versa was observed at the boundary between the two metals, i. e.,
a transition layer was formed several tens of microns thick.
In order to determine the magnitude of the effective temperature correction
for the .depth of the thermocouple insertion experiments were performed using wa-
ter in a range of Reynolds numbers from 18; 000 to 600, 000. Comparison of the
data thus obtained with the results of experiments conducted with a copper tube in
which the correction for the depth of thermocouple insertion is small showed that
the formation of the transition layer between the steel and the copper attendant to
the melting of the latter is equivalent to reducing the thickness of the steel wall
by 67 microns. If we assume that the effective wall thickness of the inter-
nal steel pipe is 0.333 mm, thenthe data for water obtained with this two-metal pipe
and with copper tube segments are in agreement to within t 6%.
Moving the thermocouples along the test section made it possible to deter-
mine the temperature distribution at the heat transfer surface along the length of
the passage. Characteristic temperature distributions for two Peclet numbers are
shown in Fig. 2. The temperatures of the sodium-potassium alloy in any cross
section were found by linear interpolation of the heat-transfer agent's temperature /12
at the pipe inlet and outlet.
The results of experiments concerned with determining the stabilized values
of the Nusselt numbers as a function of the Peclet numbers for the two series of
tests are shown in Tables 1 and 2, as well as in Fig. 3. The same figure shows
the previously obtained [1] data, results obtained in [3] and curves obtained by us-
ing formulas
Nu = 7 + 0,025Pe O8 ; (3)
Nu = 5.2+ 0.025Pe- 8 ; (4)
Nu=5+0.014Pe O.S (5)
recommended in [4] and [5], as well as those obtained by using Eq. (1).
The good agreement between experimental data for mercury and for the so-
dium-potassium alloy, which were passed through pipes of various materials, show
that the effect of the thermal contact resistance is of no importance under the condi-
tions considered. The results we have obtained again confirmed the hypothesis on
4
a)
60
SO
40
30,
60
b) 000 0
4000
50
00
000 o
o^
000
0
30
0 0p00
o
o
100
200 400 600 800 1, nn
Figure 2, Temperature Distribution Along

the Length of the Pipe for Pe = 796, q =109
10 3 kcal/m 2 -hour (a) and Pe=163, q=
= 39 . 10 3 kcal/m 2 -hour (b), The Circles
Show the Temperature of the Heat Trans-
fer Surface, the Line Denotes the Temper-
ature of the Sodium-Potassium Alloy.
TABLE 1. Heat Transfer Experiments Using a Sodium-Po-

tassium Alloy in a Stainless Steel Pipe
q X 10- 91 ^('wt f)> Re X Pe I Nn Re X 10-3 Pe

Pe ( Su
:call /m2-hr, 10-3I ( kcal/ m 23h (tW
17,12 2.9 3.22 80 5.7 62,3 6.0 35.4 797 10,0

21,8 3.5 3.69 93 6.0 46,7 4.5 29.1 8.30 10,1
15.6 2.3 4.23 95 6.4 101.2 8,6 41.1 945 11,2
15,6 2.8 4.53' 104 5.4 101.2 7.85 57,2 1217 12.3
19.45 2,9 5,46 122 6.4 101.2 7.3 68.1 1770 13.2
31,18 4,7 5.25 133 6.3 101.2 6.0 102 2250 16.2
20,2 3,1 6.43 144 6.2 101.2 5.3 121 2800 18.2
38.9 5:8 6,46 170 6;4 101,2 5.4 137 3080 18.0
54.5 8,4 8.33 216 6.2 101,2 4.9 149,6 3290 19.8
49.8 7,0 8.28 208 6.8 101.2 4.5 161 3700 21.7
54.5 7,4 9.55 266 7,2 101,2 3.7 209.5 4610 25.8
49,8 6.5 17.0 408 7.3 101.2 3,4 244.5 5470 28,5
49.8 5,6 25,2 605 8.5 101.2 2.9 278 6490 33.1
10910 10.2 29.3 796 10,3
I
5
M
4.
10 20 40 60 80 100 200 400 600 800 1000 2000 4000 NO 8000 Pe
Figure 3. Data on Heat Transfer in Liquid Metals Flowing in Round. Pipes for 0, 02::5 Pr <_
<_ 0.030 1, 2, 3, 4 Are Curves Obtained by Eqs. (3) and (4), (1) and (5), Respectively;
the White and. Black Triangles Pertain to Tests Using Mercury in Nickel and.Stainless Steel
Pipes, Respectively [1]; the Black Circles are for Tests With the Na-K Alloy in a Copper
Pipe [1]; the Half-Shaded. Circles Show Experimental Data for Mercury [3]; the White Circles
Show the Results of the Two Series of Tests With the Na-K Alloy in a Stainless Steel Pipe.
TABLE 2, Results of Control Tests Using the Sodium-Po-
tassium Alloy for a Stainless Steel Pipe
Pe I Nu II Pe Nu II Pe Nu II. Pe I Nu
I I
134 6.2 692 8.9 1290 11.7 1995 14,5

172 6.2 700 9,4 1290 12.6 2065 14,2
309 7.2 733 9,9 1445 12,9 2320 17,0
462 8.3 881 11,1 1480 12.6 2630 19,1
537 8,3 891 10,4 1570 13,5 2920 17,0
550 8,7 1000 10,2 1585 12,3 5290 27,6
631 8.9 1000 11.9 1780 14,5 6160 31,6
669 9.0 1100 11,0 1820 14,8
the negligible effect of the thermal contact resistance on heat transfer provided the /13
oxides of the liquid metal are in the dissolved state.
The experimental data for 0. 02 < Pr c 0. 03 are situated somewhat lower than
the curve obtained in Eq. (2) and, together with data of [3], can be approximated
by Eq. (1). As the Prandtl numbers are made smaller, the Nusselt numbers increase
with the Peclet numbers held constant, which follows, for example, from calculations
made by Buleyev, which are quoted in [1]. Hence, heat transfer in liquid metals (as
well as in sodium, Pr ^ 0.005) flowing through pipes should be calculated by Eq. (2).
The loss of accuracy when using this equation will not exceed 10% even for low (^-100)
Peclet numbers.
REFERENCES
1. Subbotin, V.I., et al.: Inzhenerno-fizicheskiy zhurnal, Vol. VI, No. 4, 16 (1963) /15
2. Subbotin, V.I., et al.: Vliyaniye ekstsentrisiteta na teploobmen zhidkikhmetal-
lov v kol'tsevom zazore [Effect of Eccentricity on the Heat Transfer in Liquid
Metals in an Annual Space], Present collection, p. 108.
3. Petukhov, B. S. and Yushin, A. Ya.: Doklady AN SSSR, 136, 6 (1960).
4, Lyon, R.: Chem, Eng. Progr. , 47, 2, 75 (1951).
5, Kutateladze, S. S. , et al.: Zhidkometallicheskiye teplonositeli [Liquid-Metal
Heat Transfer Agents]. Moscow, Atomizdat, 1958.
2
HEAT TRANSFER WITH LIQUID METAL IN A VERTICAL PIPE
AT LOW PECLET NUMBERS
N. A. Ampleyev, P. L. Kirillov, V. I. Subbotin

and M. Ya. Suvorov
Heat transfer with liquid metal in pipes has been thoroughly studied during
the last fifteen years. At present a large amount of experimental data has been
accumulated and the main governing relationships needed for practical calcula-
tions have been, for the most part, clarified. Still remaining unclear are the
more complex cases of heat transfer (combined forced and free convection) and
the physicochemical processes at the liquid-solid metal interface. The most
general solution of heat transfer in a round pipe with q = const has been given by
Lyon [1]
g 2
(, ug d^)
1 o
Nu
_2 d^.
S (1)
0 i+s v`Pr
In order to evaluate this integral, in addition to the velocity distributionuQ),

one must also know the quantity `=away -1, i. e. , the expression which is cal-
culated by differentiating the velocity distribution, and E, the ratio of coefficients of
turbulent transfer. The work of E. V.. Nomofilov was directed at investigating the
relationship
s e = f (Re, y). (2)
Other theoretical solutions reduce, in one way or another, to the evaluation of

Eq. (1), upon making use of Eq. (2).
The first experiments on the heat transfer in liquid metals [2] were more
or less roughly qualitative in character, and their accuracy was not too high.
The large number of experiments [performed since], [3,4] have resulted in the
following [empirical] formulas
Nu= 3+0.014Pe O.8 . (3)
Nu = 5 + 0.0021 Pe. (4)

When performing these experiments the metal was not specially purified, and its /16
purity was not checked. In experiments with pure metals, higher heat transfer
coefficients were obtained, and these are given by the expression
Nu -- 5 +0.014PeO.8 . (5)
The difference in the Nusselt numbers obtained by the above expression and
by Eq. (3) was attributed to the so-called thermal contact resistance at the bound-
ary between the wall and the liquid metal. The nature of this resistance, despite
experiments performed to clarify it, remained unclear for a long time. The ma-
jority of hypotheses have not been verified experimentally or the results of sub-
sequent experiments were contradictory (effect of wettability, gas, etc.). In ad-
dition, some investigators attributed to contact resistance any data having values
below those which at that time were considered as standard, though these values
were quite possible either results of experimental inaccuracies or pro-
cedural errors.
The thermal contact resistance to an alkaline metal flowing in a pipe, was
recently investigated as well as the conditions leading to its appearance on the
heat transfering surface [6]. One can conclude from these experiments that the
Nusselt numbers calculated by Eq. (3) apparently, corresponds to the maximum
value of the thermal contact resistance produced by impurities suspended in the
heated coolant. The maximum value of this resistance is determined by the thick-
ness and structure impurities which precipitate on the wall, and depends on the
rate of flow of the liquid metal but does not depend on the impurity content in the
flowing metal.
In order to eliminate the effect of the thermal contact resistance on the re-
sults of heat-transfer experiments, a method was developed which makes it pos-
sible to measure the temperature fields in the liquid stream. The results of ex-
periments performed by using this method have yielded higher heat transfer co-
efficients than those given by Eqs. (3) and (4) and were in close agreement with
Lyon's formula
Nu = 7 +0.025Pe- 8 , (6)
which is obtained by evaluating the integral in Eq. (1) with attendant use of the
Nikuradze velocity distribution on the assumption that e = 1. It is .possible to
give a definitive physical meaning to the formula
Nu=Nuo+AN O.8 (7)
if we regard the expression Nu0 (Re) as the value of the Nusselt number when Pr =
= 0. In this case the values of Nu0 which characterize the heat transferred
by conduction only will not be constant, but will depend on the velocity
distribution, i.e., on the Reynolds number. The second term in Eq. (7) will re-
flect the contribution made by turbulent heat transfer.
Various authors have obtained close results for Nu 0 [7]; it may be as- /17
sumed approximately, with an error of 10%, that
Nuo=8 w (8)
wmax
It is known, that for laminar flow w min/wmax = 0.5 and this expression thus
yields Nu = 4. As is known, the theoretical value is 4. 36, as this was substan-
tiated experimentally in [8]. On the other hand, other experiments have yielded
higher (Nu = 7) as well as lower (Nu = 0.6) values [9-12].
The opinions to the effect that the high heat coefficients obtained in papers [9]
and [13] were due to the factthatthe heat transfer was measured in a section with
unsteady heat transfer . (Z = 5d) is incorrect. In this case, if there is no experimen-
tal error, Nu/Nu .= 1.15-1.2 and then Nu should be not more than about 5, but not
7 or 8 as was obtained in [9) and [13]. Consequently, the 50% difference remains un-
explained.
The opinion that the low heat transfer coefficients in the region of low Peclet
numbers are due to improper determination of the average temperature of the liq-
uid, which is caused by not taking into account longitudinal, heat leakages along the
wall and the heat transfer agent [14] and [15], is only partially correct. The ex-
periments performed by Johnson are in doubt from the point of view of methodol-
ogy due to the distortion of the temperature field and the heat fluxes by transverse
slots in the test section. However, the same results were obtained in a copper
pipe [10], which did not have slots and which was cooled by a sodium-potassium
alloy. Consideration of longitudinal heat fluxes in the fluid does not make a large
contribution, since the Nusselt numbers still differ by a factor of two when Pe = 25.
Analysis of data on measured temperature fields expressed in the dimen-
sionless form [11] has made it possible to approximately establish the experi-
mental temperature distribution a stream of liquid metal and to use this distri-
bution to set up the interpolational formula
Nu= 5 +0.025PeO.8
or
Nu = 4.36 + 0.343Pe0 4 , where Pe* = U*v*d .
The theoretical results obtained in [16] are practically identical withthe above results.
10
Reference [17] describes approximate calculations of heat transfer to the in-
let part of a pipe during flow of a liquid metal. It has been determined that the
length of the thermal stabilization segment becomes larger with an increase in
the Reynolds number. The values of the Nusselt number were estimated in a
wide range of Reynolds and Prandtl numbers. Deissler [18] has analyzed the ef-
fect (on the heattransfer) of the length of thermal stabilization, of Re and Pr, the
shape of the channel, and the initial velocity distribution. Asa result it was shown
that at Pr = 0.01 and a given l/d, the ratio Nuloc /NuOD for parallel plates /18
is somewhat lower than for pipes. At high Reynolds numbers and a given
l/d, the ratio Nuloc/NuCD becomes smaller with an increase in the Prandtl number.
For low Reynolds numbers the changes are more complex. Ratio Nuloc/NuOD first
decreases and then, with an increase in the Prandtl number, increases somewhat.
Results of calculations performed in [17] show that fully developed heat transfer
starts as early as at a distance of less than 10d.
N. S. Kondrat'yev has studied heat transfer (l/d = 17.4, 8.4) attendant to the
flow of the lead-bismuth eutectic alloy [12]. The results of his experiments can
be approximated by the expression
Nu
N U-1 si = 4.6 C d / 1-0.4, d 30.
oc
Kondrat'yev has not detected. the effect of the Reynolds number on the length of
the thermal section. By comparing the experimental data obtained by various
authors it follows from [4] that, when 1/d ^! 30, the effect of the inlet section has
little influence on the average heat transfer, while for short pipes it is recom-
mended to introduce the coefficient s l which. takes this effect into account:
\ 0.i6
ei =1,72 Cd
l
It was discovered in [19] by measuring the temperature distribution in the

inlet section of the pipe that stabilization of local values of the heat transfer co-
efficient takes place at a distance of less than 10d when the flow is hydordynam-
ically stabilized at Pe = 50-1500. On the other hand, Isakoff and Drew [20]
found that the heat transfer coefficient is not stabilized even at a length of 58d
and it is by 12% higher than the values obtained at a distance of 138d, from the
pipe inlet. Similarly, Subbotin et al, [21] have determined the local coefficients
of heat transfer for Pr - 0.02 in the thermal stabilization section (1/d = 5, 10
and 15) atPe=104-2.105
Stabilization of heat transfer in the range of Re under study took place prac-
tically at l/d 15. At distance 1./d = 5 the heat transfer coefficient was by 20-
40% higher than the stabilized values. The disagreement between the experiment-
ally obtained values of E I and those obtained theoretically in [17] is as high as 30 %.
11
The length of the thermal stabilization section and E I = Nuloc/NuOD increase, ac-
cordingto [17], with an increase in the Reynolds number. The heat transfer coef-
ficients obtained by measuring the wall temperature are lower than those obtained
on the basis of the measured temperature distribution. The length of the thermal
stabilization section and the magnitude of E l in this case are smaller; the coeffi-
cient is equal to ten and does not depend on the rate of flow of the metal. This
difference is attributed to the effect of the thermal contact resistance.
The local Nusselt numbers for Re = 7. P 103 ,3.4 . 104, 1.6-10 5 and 7.2- 10 5
in a round pipe were determined numerically in [22]. It follows from the results /19
of calculations that the length of the thermal stabilization section (Pr = 0.025)
increases as the Peclet number is increased from 10 to 3000 and decreases when
the latter number is increased further.
In the present paper an attempt is made to clarify the cause of the disagree-
ment between results obtained experimentally at low Peclet numbers and the the-
oretical value of Nu = 4.36. We have set out to determine the length of the initial
thermal [stabilization] section as a function of the Peclet number for turbulentflow.
Figure 1 shows the schematic diagram of the circulation circuit made from
1Kh18N9T stainless steel pipes with an a-c electromagnetic volute pump, provid-
ing a delivery rate of up to 10 m3/hour. The metal in the circuit is circulated
through the test section and an alloy-water heat exchanger. The alloy is pressurized
from tank 1 into tank 4 and into the circuit by means of an inert gas and flows
through a mesh filter made of several layers of nickel wire mesh. Measuring
tank 7 was provided for calibrating the flowmeters. The oxides are continuous-
ly removed from the alloy by cold traps. The oxide content is monitored by a
float indicator and a distillation sampler [23]. The flow rate of the sodium-po-
tassium alloy is measured by a magnetic flowmeter combined with the R2/1 po-
tentiometer and is adjusted by valves.
The test section (Fig. 2) consists of a 2700 mm long copper pipe with an / 21
inside diameter of 50 mm and wall thickness of 5 mm. The pipe is connected to
mixing chambers by flanges employing a ball-and-socket joint. Themechanismfor
moving the thermocouple and a rubber packing gland, which is fastened by a con-
nector nut, is placed on the top chamber cover. Longitudinal grooves 3 mm deep
are milled in the pipe wall and chromel-alumel thermocouples are placed in them.
The hot junctions of the thermocouples are calked by copper wedges, and. the elec-
trodes are provided with alundum insulation. The copper pipe is insulated by mi-
canite 0.2-0.3 mm thick, on which is placed a nichrome heater from wire 3.5 mm
in diameter with a pitch of 4mm; then asbestos insulation, thermometers and com-
pensating heaters were placed over the pipe, in this order (see Fig. 2).
When measuring heat transfer (Pe < 100) we have discovered appreciable
heat losses at the section from the end of the heater to the mixing chamber, for
which reason we have placed thermometers there. The heater was made in two
sections, each 1 meter long. The electric power used by the heaters was meas-
sured by a set of class 0.5 instruments. The thermocouple emf was measured
by the PPTN-1 potentiometer, together with the M 21/4 mirror galvanometer.
12
Figure 1. Schematic of Experimental Setup. 1) Filling Tank;
2) Level Meter; 3) Electromagnetic Pump; 4) Pump Tank;
5) Toluene-Cooled Cold Trap; 6) Water-Cooled Cold Trap;
7) Tank for Calibrating the Electromagnetic Flowmeter;
8) Float Indicator; 9) Sample Taker; 10) Heat Exchanger;
11) Test Section; 12) Valve; 13) VN-2 Vacuum Pump;
14) Electromagnetic Flowmeter.
The temperature distribution in the metal flow was measured by a movable /22
thermocouple (Fig. 3). The frame of the mechanism is pipe 11, which has been
lead out from the metal flow through a packing gland, and the thermocouple is
moved in the longitudinal direction by a jack-type mechanism. The main compo-
nents of the mechanism used for moving the thermocouple in the lateral direction
are a rod with an inclined slit (wedge) and a probe with a shank and axle. When
the rod is moved in the longitudinal direction, the shank which is inserted into
the inclined slit moves across the pipe. The thermocouple which is fastened to
the other side of the axis, then also moves across the flow. Pipe 12 is moved
relative to pipe 11 by another jack, which imparts motion to the rod. The shank
and the entire inside cavity of pipe 11 are protected. by a bellows from penetration
by the metal. The chromel-alumel wires are welded hermetically to the end of the
probe, pass inside of flexible insulation through the probe, shank and pipe 12, and
are lead out to the outside. The movement of the end of the probe (of the thermo-
couple) in the transverse direction is recorded by an indicator, which bears on the
pullbar, which is connected to pipe 12. The movement along the pipe was measured
by a ruler (with scale division of 1 mm). The thermocouple was calibrated by com-
paring the displacement of its end part with the indicator readings. The thermo-
couple junction thickness is 0. 7mm. The probe was centered in the pipe by means
of a guiding bushing, into which three guiding blades have been inserted.
13
11
12
Sec. A-A 11
Figure 2. Test Section.
' 1) Indicator; 2) Pipe 202 mm
of
18 in Diameter; 3) Plate;
B
NI o 4) Graduated Circle;
o
5) and 9) Connector Nuts;
6) Support; 7) Triple-
Thread Nut; 8) Movable
Thermocouple; 10) Pack-
ing Gland; 11) Pipe Union;
12) Mixing Chamber;
13) Copper Pipe; 14) Guide
Bushing; 15) Asbestos
Insulation; 16) Compen-
sating Heater; 17) Ther-
mometer; 18) Insulation
(Mica).
14
Methodology Used in the Experiments and in Analyzing the Experimental Data
Before starting the experiments the sodium-potassium alloy was filtered
through the cold trap. When doing this the sampling device was periodically
started up. After an oxygen concentration of 3 . 10-3 -5 . 10 -4% by weight was
reached, the given flow rate of the alloy through the test section was established
and the heaters were started up. After the wall temperature was measured, the
temperature in the alloy flow (28 points) was taken. The density of measurements
was the highest in the region closest to the wall. The measurements were made
with the thermocouple moving from the center to the wall. Here, for more reli-
able temperature readings at the wall,
the thermocouple was moved until two
consecutive temperature measurements
were identical. The bending of the ther-
mocouple (0.5-1 mm) which is inevitable
[under these circumstances] did not go
past the limits of the device's flexibility.
The theoretical power was calculated
as an average
Q =Qel+
_Z Qth
11
12
The disagreements in the thermal balance

13
magnitude did not basically exceed 3-7%.
The heat transfer coefficient was deter-
mined by two methods: 1) from the tem-
perature distribution of the wall along the
15 Section A-A length of the section and of the temperature
9
0
0
ti 16
77 Cross Section
of Probe Figure 3. Movable Thermocouple.
18 1) Indicator; 2) Plate; 3) Pullbar;
13 4) Bolt; 5) Graduated Circular;
6) Bushing; 7) Nut; 8) Support;
20
9) Spline; 10) Washer; 11) Pipe
With Diameter of 20 x 2 mm;
21 ^^.
12) Pipe With Diameter of 16 x
x 1mm; 13) Reducing Adapter;
14) Wedge; 15) Shank; 16) Rim;
17) Bellows; 18) Axle; 19) Ring;
20) Pipe With Diameter of 10 x
0.3 x 2 mm; 21) Pipe With Diameter
of 6 x 1 mm; 22) End Pieces.
15

of the liquid in the mixing chambers and 2) from the temperature distribution in
the alloy flow. The heat transfer coefficient was calculated from the expression:
_ g
a
(tWtf) (9) /23
where
Q Q
Odin
l.
here (tw - tf) is the difference of temperatures at the inside surface of the wall
and the average calorimetric temperature of the liquid, averaged over the length.
This difference was determined on the basis of the temperature distribution of
the wall along the length of the pipe and the readings of the thermocouples used
to measure the alloy's temperature at the inlet and exit from the heated section.
The temperature drop in the wall was determined by the ordinary formulas
Odin In dext
_ din
At
2k,0 (10)
The temperature of the alloy at the inlet and outlet of the heated section was de-
termined by taking into account the heat lost in the flow stabilization section and
from the end of the heater to the mixing chamber
t nl tmch. ate t out = tmch.+ tlos
The heat losses were determined from the formula
XasF103
s 540.10-3 AE.
Q =
The average liquid temperature t was determined upon taking into account the
flow of heat in the liquid and the wall t =T mch. + 6t
1
SL = Cn l ^f X. n/`- / 71
[V l dt _ 4qq
(11)
dx cpyd
_ in
16
According to the second method the wall temperature was found by extrapo-
lating the temperature distribution to the wall's surface. Here the convergence
was checked with the boundary condition
dt 4
dr Ir=ro
jr=roXf .
The average temperature of the liquid metal could have been found from the ex-
pression
i
W'
t ! =d-
tt =
0
w
where Z=^2 T`Y .
ro J
w;
_ -d::
0 7O (12)
In calculating tf the integration was replaced by summing. We shall now consi- /24
der in more detail the selection of the velocity distribution for calculating this
temperature. As was shown in [18], there is no difference in t f and, consequently,
also in the Nusselt numbers, if it is assumed that the turbulent flow is governed
by Karman's logarithmic velocity distribution
0< y+<5 u+=y+

5<y+ <30 u+=5Iny+-3,05
30 < y + u+= 2.5y + + 5.5 (13)
or by the power law
(2 I a)21+Q, 1_ rj
ro )a ; a= f (Re).
:o (14)
Expressions (13) and (14) are suitable for developed turbulent flow. For laminar
flow with Re < 2300 it would have been natural to use the parabolic expression for
the velocities
mt 2 P
00, Y 1 - (15)
17

However, Eqs. (13-15) are suitable only for an isothermal flow. When the liq-
uid is heated or cooled the velocity distribution changes. In addition, turbulent
flow ensues with lower Reynolds numbers as a result of acceleration or turbuli-
zation of the flow, which is brought about by natural convection. The velocity
distribution in a nonisothermal laminar flow has been analyzed in [24]. The pa-
rameter which characterizes the natural convection intensities is the quantity
f g V2 r dt
The curve of Re cr = f(Z), according to data of [24] is shown in Fig. 4.
R
11P
10^
8
6
4 Turbulent Flow
2
102
8
6
4 Laminar Flow
2
10
G' Z 4 6 9 10 Z
Figure 4. The Critical Reynolds Number as a

Function of Z.
In our experiments Z varied from 5 to 10. Figure 4 shows that under the /25
circumstances we always had turbulent flow. Hence, strictly speaking, the use
of the expression for the velocity distribution given in [24) is also incorrect
w i _ Z r ber (Z') i bei (Z) ber (Z) bei (Z') (^7)

W 2 L bei' (Z) bei (Z)+ber' (Z) ber (Z) '
,gpPr at Re denotes the
Z = r0 V V2 TX (Z) = Re [Jo (Z ^)^
real part;
bei (Z) _ Jm [Jo (Z Im denotes the
imaginary part;
Z' =
r, , jgA rr
V at ber' (Z) = Re [ ^J, (Z Yi )]
J0J1 is a Bessel
bei' (Z) = Jm [r1 J, (Z Yi)1 1 function.
18
The velocity distribution obtained by Deissler [25, 26] for heat transfer in
a round pipe was calculated by considering the viscosity only along the pipe cross
section, and hence is suitable only for gases. Ber [27] has calculated. the velocity
field in a round pipe with heat transfer in a flow of fluid. To determine the ve-
locity for Re < 2300 Gorchakov [28] has suggested to use the energy equation
F1 dt
W !t

ri Tr z (e
q +a) Trdt/dz
dt ] 1 '
( 18 )
It may be assumed in the first approximation that 1) E q = EM and 2) that the tur-
bulent viscosity in the flow
EM
V = 0.4 $
pro ^ ( 1 ^) [26],
Exn= 89 = 0.4v* rod ( 1 s),

Eebg(1^)
Differentiating Eq. (18) we get
wt C 1 (a + 2by 3b;=) x.
` '4 o
X +(a+b^+b^2)df } ' (19)
where
dt
A = dx = coast.
The first and second derivatives of the temperature were determined graph- / 26
ically from the experimental temperature curves in the sodium-potassium flow.
It was thus found that W = fQ). In the region next to the wall W was found by in-
terpolation. After the velocity distribution has been interpolated, the value thus
obtained was computed with the measured value of W and the difference did not
exceed 10%. In this manner we have found the approximate distribution of Wi/_
= fQ) for Re < 2300 in a nonisothermal flow. Using this distribution we have cal-
culated the Nusselt number. Figure 5 shows the velocity distribution obtained. by
differentiation for Z = 5 and 9 and the theoretically calculated curves from [24],
19
Results
The results of studies, expressed in terms of similarity criteria, are shown
in Figs. 6 and 7. The experimental data for laminar flow gave values of Nu > 4.36.
This divergence from the theoretical value
is due to the effect of natural convection.
Consequently, also in previously perform-
Wi ed work this difference, whenever observed, is
w attributable to precisely this circumstance,
0
I
I
The measured temperature distri-
bution in the flow of the sodium-potassium
1 alloy has shown that the actual tempera-
ture is higher than the mixed mean,
1 I which is determined in the mixing cham-
:5
^ 4 ber. Figure 8 shows the temperature
distribution in the wall and in the liquid
along the length where heat transfer takes
place (Pe = 10, Re = 400). As can be seen
from the figure, the actual temperature of
1.0
the alloy is higher than the temperature
given by the straight line drawn through
5\ - I 1 points obtained in the mixing chamber
/ / 1
tmch.' which is usually assumed in cal-
v
culating heat transfer coefficients for
q = const. This temperature increase is
satisfactorily taken into account by the
formula presented in [14].
The length of the initial thermal /28
0.5 [stabilization] section for Pe < 100 is
about l. 0/d. 1. The sharp drop in the
Figure 5. Velocity Distribution
According to Data of [24] (Dashed heat transfer (in the Nusselt numbers) in
the above range of Peclet numbers ob-
Line) and That Found Graphically tained in [10] is due to the fact that in an-
(Solid. Line). alyzing the results no consideration was
given to heat losses from the end. of the
heated section at the exit to the mixing
chamber. This is attested to by a recalculation of one of the regimes. The lower-
than-average Nusselt numbers obtained in [15] have been produced, as was shown
in [8], by distortion of the thermal flux which was brought about by the transverse
grooves in the test sections.
The heat transfer, which was determined by two methods yielded close re-
sults for Pe < 100, which points to the absence under the conditions at hand (ox-
-4
ygen content in the sodium-potassium alloy of 3 . 10 -3 -5 . 10 % by weight) of ap-
preciable thermal contact resistances.
The local value of the Nusselt number at different distances from the inlet
is shown in Fig. 9. Using these experimental points we have found the averaging
20
u^
4W
2 4 6 8 1p 2 2 4 6 8 1p3 2
Figure 6. Experimental Data Using the Coordinates

lg Nu = f(lg Pe). 1) Nu = 7 + 0. 025 Pe 0.8 ; 2) Nu =
= 5 + 0. 025 Pe 0.8 ; the White Circles Denote
the Temperature Distribution in the Flow of the
Sodium-Potassium Alloy; the Black Circles Denote
the Same in the Wall.
10
,YL
to
1 1
i 1 ^ 1 { 1
n
S
^^
8
^ ^ ^ t l ' , I I i t
L
e ^
I( I 1 ^ I
C R m 2 4 9 R en2 7 4 6 R mJ 7 4 4 R Aa
Figure 7. Results of Heat Transfer Experiments in a Pipe.

0.8
1) Nu = 7 + 0.025 Pe 08 ; 2) Nu = 5 + 0.025 Pe ; the Black
Circles Denote Curves of Temperature Distribution in the
Wall; the White Circles Denote the Same in the Sodium-Po-
tassium Alloy; the Triangles Are for Data of [8], the Half-
Blackened Circles are for Data of [10] and the Crossed
Circles Are for Data of [15]
curves and calculated the coefficients s l = Nuloc/NuOD, which characterize the

increase in the heat transfer at the inlet section of the pipe. Figure 10 compares
the experimental results with those calculated theoretically in [22]. The length
of the initial thermal [stabilization] section was here determined from the
21
t, c
tf tf
70 dtlOssl'- --
/ I
so
dta /
tmch. y
to
50
40
Ii
30 1
/
!I
n... I
-4 0 4 8 12 16 20 24 28 LID
Figure 8. Characteristic Temperature Distributions Along

the Length of the Test Section. t" is the Temperature, Cal-
culated by Taking Into Account Heat Losses, but Without Ac-
counting for Heat Flow in the Longitudinal Direction; w Is
the Wall Temperature; tmch is the Temperature Without
y
Consideration of Heat Losses and the Longitudinal Flow of
Heat; tf is the Temperature Calculated by Considering the
Longitudinal Flux, But Without Taking Into Account the Heat
Losses; tf Is the Temperature Calculated by Taking Into Ac-
count the Longitudinal Flux and the Heat Losses; the Cross-
ed Circles Denote the Wall Temperature Measured by Thermo-
couples Inserted in the Wall; the Squares Denote the Tem-
perature in the Mixing Chamber; the Black Circles Show the
Liquid Temperature Measured by the Movable Thermocouple;
the White Circles Denote the Wall Temperature.
distribution of At along the length of the pipe. It follows from Figs. 10 and 11
that this length and the coefficient s Z first increase with an increase in the veloc-
ity, and then decrease. Qualitatively this phenomenon is in agreement with re-
sults obtained in [18] and [22]. Stabilization of heat transfer in the range of Pe-
clet numbers under study takes place at l/d <_ 15.
The experimental data (Fig. 12) lie below the theoretical curve obtained in
[25]. The lower values of Nusselt numbers, as compared with those calculated /30
theoretically, are, in the opinion of the authors, due to an increase in parameter Z.
22
A
o
All
a
Figure 9. Local
^ ^ m
Nusselt Numbers
for Cross Sections
Situated at Differ-
1i ent Distance l/d.
1 3 6 10 15 16.5
0O0O
2 4 8 10 3 2 oe
l Id Figure 10. Length of the

14 0 Initial Thermal [Stabiliza-
0 0 tion] Section as a Function
10
of the PecletNumbers, The
White Circles Denote Data
0 0 Obtained by the Present Au-
s 1 1
thors; the Black Circles Are

0 for Experimental Data Ob-
tained in [21]; the Calcula-
70Z 2 3 4 5 6 7 E310 2 3 4 5 Pe tions Are From [22].
Nun
SUM
1,7
i6
1,5
1. 3
e=900
5G0 2000 200
1, 2
1,1
0
0 1 2 3 4 5 6 7 8 9 10 11 Z/d
Figure 11 Change in the Relative Heat Transfer at the In-

let Section of the Pipe. The Black Circles Denote Data Ob-
tained by the Present Authors,
23
u m
40
0
10 +
++
8
0
6 0
7 2 4 6 8 10 20 1
Figure 12. Comparison of Experimental Data

With the Theoretical Curve of [24]. The White
Circles Denote Results Obtained by the Pres-
ent Authors; the Black Circles Are for Data
Due to Gorchakov; the Crosses Denote Results
Obtained in [28].
As was shown, when natural convection arises, the flow becomes turbulent even
with Re = 400. Hence, the turbulent Z will be somewhat lower (v becomes great-
er) and the experimental points will be displaced to the left. A more detailed
study of this problem is presented in [28].
NOTATION
I length of heated section
r0 pipe radius
y distance from wall
Q heater rating
E thermocouple emf
G flow rate
m number of junctions in thermometer [thermocouple]
S thermometer [thermocouple] thickness
e
turbulent heat transfer coefficient
24
e turbulent momentum transfer coefficient
M
e = e q/e M ratio of turbulent transfer coefficients

$ Z = Nuloc/Nu. ratio of the local Nusselt number and of that stabilized along
the length
$ Z = Nu/Nu. average ratio
Dimensionless quantities
y = y v*/v distance from wall

= r/r0 distance from center
Subscripts /31
in internal side
ext external side
f fluid
W wall
loc. i local value
-(bar) average value
el electrical
th thermal, pertaining to the thermometer [thermocouple]
Go at infinity from the inlet
los pertaining to losses
inl pertaining to the inlet
out pertaining to the outlet
as asbestos
max maximum value
mch mixing chamber
25
fl pertaining to the flow stabilization section
cr critical, i.e., corresponding to transition from laminar to
turbulent flow
REFERENCES
1. Lyon, R. N.: Chem. Eng. Progr. 47, 75 (1951).

2. Styrikovich, M. A. and Semenovker, I. Ye.: Zhurnal ekspe rimental I noy
i teoreticheskoy fiziki, 10, 1324 (1940).
3. Mikheyev, M. A., et al.: In the book Reaktorostroyeniye i teoriya reak-
torov [Reactor Building and Theory] . page 139. Moscow, Publishedby Acad-
emy of Sciences of the USSR (Report of the Soviet Delegation to the Inter-
nal Conference on the Peaceful Uses of Atomic Energy, Geneva, 1955).
4. Kutateladze, S. S., et al.: Zhidkometallicheskiye teplonositeli [Liquid-
Metal Heat Transfer Agents]. Atomizdat, Moscow, 1958.
5. Dwyer, O. E.: A. I. Ch. E. Journal, 9 (2), 261 (1963); Nucl. Sci. Eng.,
15, 336 (1963).
6. Ivanovskiy, M.N., et al.: Kontaktnoye termicheskoye soprotivleniye pri
turbulentnom techenii zhidkikh metallov [Thermal Contact Resistance in
the Turbulent Flow of Liquid Metals]. In print.
7. Kirillov, P. L.: Teplofizika vysokikh temperatur, 1 (1), 102 (1963).
8. Petukhov, B.S. and Yushin, A. Ya.: Doklady AN SSSR. 136 (6), page 132,
1961.
9. Pirogov, M. S.: Teploenergetika, No. 3, 62 (1961).
10. Subbotin, V.I. , et al.: Issledovaniye teploodachi pri techenii zhidkikh
metallov v trubakh [Study of Heat Transfer in Liquid Metal Flows in Pipes].
Inzhenerno-fizicheskiy zhurnal, 6, No. 4, 16 (1963).
11. Kirillov, P. L.: Atomnaya energiya, 13, No. 5, 481 (1962).
12. Kondrat'yev, N. S.: In the collection, Voprosy teploobmena [Problems of
Heat Transfer] 96. Published by the Academy of Sciences of the USSR.
Moscow, 1959.
13. Pirogov, M. S.: Atomnaya energiya, 8, No. 4, 367 (1960).
14. Trefethen, L.M.: Trans. ASME, 78 (6), 1207 (1956).
15. Johnson, H.: Trans. ASME, 75 (D, 1191 (1953).
16. Buleyev, N.I.: Teploperedacha, Published by the Academy of Sciences of
the USSR, 64, 1962.
17. Poppendiek, H. F.: Nucl. Sci. Eng, 5 (6), 390 (1959).
18. Deissler, R. G.: Trans. ASME, 77, No. 8 (1955).
19. Kirillov, P. L., et al.: Atomnaya energiya, 6, No. 4, 382 (1959).
20. Isakoff, S. E. and Drew, T. B.: General Discussion on Heat Transfer,
405 (1951).
21. Subbotin, V.I. , et al.: Atomnaya energiya, 13, No. 2, 155 (1962).
22. Buleyev, N. I., et al.: Teplofizikavysokikhtemperatur, 4, No. 4, page 540, 1966.
23. Dmitriyeva, I. B., et al.: Opredeleniye soderzhaniya kisloroda v natrii i
splave natriy-kaliy metodom vakuumnoy distillatsii [Determination of the
Oxygen Content of Sodium and of a Sodium-Potassium Alloy by Vacuum
Distillation].
26
24. Brown, W. G.: VDI-Forschungsheft, 480 (1960).
25. Deissler, R. G.: Trans. ASME, 76, (1), 73 (1954).
26. Deissler, R. G.: NACA TN 2242, 3145 (1954).
27. Ber, L. E.: Izvestiya AN SSSR, OTN, No. 11, page 77, 1957. /32
28. Gorchakov, M. K., et al.: Teplootdacha natriya i splava natriy-kaliy v
korotkikh vertikal'nykh trubakh pri smeshannoy konvektsii [Heat Transfer
of Sodium and of a Sodium-Potassium Alloy in Short Vertical Pipes by
Combined Convection], Article following this one in the present collection.
27
HEAT TRANSFER OF SODIUM AND A SODIUM-POTASSIUM ALLOY
IN SHORT VERTICAL PIPES BY COMBINED CONVECTION
L. G. Volchkov, M. K. Gorchakov, P. L. Kirillov
and F. A. Kozlov
Heat transfer.with liquid metal by combined convection (forced and natural) /32
is studied because of the need to design heat exchanger apparatus, particularly,
cold traps for oxides of alkaline metals. Data on this topic are not available,
with the exception of individual remarks [1] to the effect that on parallel convec-
tion (when forced and natural convection proceed in the same direction) the heat
transfer coefficient increases substantially. The published results of theoretical
[2-5] and experimental [6-10] studies using liquids with Pr > 1 do not satisfac-
torily clarify the problem of heat transfer with liquid metals by combined convec-
tion. The qualitative pattern of the fluid flow and heat transfer for combined con-
vection can be observed in [2], which solves the problem for laminar stabilized
flow with qw = const. Subsequently the results of these calculations will be com-
pared with experimentally obtained data.
The heat transfer was studied at test sections - recuperators* - with the
following basic parameters:
Section number I 1 i 2 I 3
Central tube diameter, 102 I 68 I 135

d, mm
Length of heat transfer I 5 ( 5 ( 5
section, 1h. t/d
Central tube length, 1/d ( 8 8 ` 8
Wall material Copper I1Eh18N9Tj Copper

( steel
( 6.0 1.2 5.0
Wall thickness, mm I
Width of annular space 5.0 6.0 7.5

6, mm
Heat transfer agent Na and NaK Na Na
4
* [Translator's Note: Also called regenerators].

28
Section number 1 2 I 3
Range of Peclet num- 7.5-130 (Na) 3.1-150
hers under study 3.6-350 (NaR) 4.8-1550
Range of Z under 15.8-21.5 (Na) I
study 19-29 (NaK) 10.5-17 19-26
At test section No. 1 (external surface of the annular space) we have in- /33
stalled thermocouples, the design of which (similar to that described. in [61)made
it possible to measure the temperature along the pipe cross section at the entire
heat-transfer section. At section No. 2 (see Fig. 1) the movable thermocouple
could be used to measure the temperature distribution only at the center of the
heat transfer section. The thermocouples at section No. 3 were placed both on
the heat transfer and on the external wall. Except for this the sections differ
from one another only by inconsequential details. This design of the test sections
makes it possible to safely reach a thermal flux density of 3 . 105 kcal/m2-hour
using a low-capacity adjustable cooler on the . line connecting the outlet of the cen-
ter pipe with the inlet into the annular space. In addition, this simplifies the nec-
essary measurements.
The experimental sections were placed in sequence on a multipurpose test
stand, provided with a cold trap for oxides, oxide content indicator and a measur-
ing tank for calibration of the flowmeters. The manner in which the test section
with its main elements was connected to the circuit is shown in Fig. 2.
Experience of our work has shown that, in spite of the presence of a gas
separator, gas accumulated at the inlet part of the center pipe, decreasing the
mean heat transfer (a part of the surface was not wetted by the metal). An
additional blow-down line was constructed for continuous removing this gas. The
experiments were performed using pure metal with a continuously operating oxide
trap. The oxygen content of the metal was determined only before performing a
series of experiments; it comprised 5 . 10-40/o by weight.
When steady flow was established we have recorded at least three times the
following data: metal flow rate, temperature at the inlets and outlets of the center
pipe and of the jacket, readings of thermocouples which covered the entire surface
of the test section. The heat losses through the surface of the [test] section because
appreciable' for moderate metal flow rates and at high temperatures, hence the real
temperature was always calculated for each point.
To calculate the heat transfer coefficient we have used the difference 6t be- /35
tween the metal's temperature at the inlet of the center pipe and at its exit, as
well as the mean temperature difference Attheo between the pipe's outlet and the
inlet into the annular space, with a correction for heat lost through the outside
surface of the annular space. The theoretical heat transfer coefficient Ktheo
29
3
.9
Figure 1. Design of No. 2 Test

Section. 1) Inlet Connecting
Stub; 2) Pocket (Capillary);
3) Recoiling Mushroom; 4) Wire
Mesh; 5) Jacket; 6) Heat-Trans-
fer Wall; 7) Movable Thermo-
couple; 8) Pocket With Thermo-
couple; 9) Mechanism for Mov-
ing the Thermocouple Probe.
(without giving consideration to heat flow along the axis) was found from the
expression
Ktheo n d1 h,
h. t .theo (1)
30
Figure 2, Schematic Diagram Showing How
the Test Section Was,Connected to the Main
Circuit. 1) Cooler; 2) Test Section; 3) Blow-
Down Line; 4) Regulating Valve; 5) Gas Trap;
6) Heating Section; 7) Electromagnetic Pump,
Partially Adjusting the Flow Rate Through
the Section.
The actual heat transfer coefficient for the center pipe was found from the re-
lationship
i
a ^+
1 d
d Indi
I^ Aw d d1ag
(2)
The relationship between Ktheo and K was found by a method presented in

[11] (Fig. 3, a and b). The error of the correction increases highly with a reduc- /36
tion in the Peclet number, Hence K and a were not calculated for very low Pe.
The pattern of the experimental points for
Ktheo points to the appreciable role
31
p
m4 a)
W v eT '
IV
M
U
.1C
f 10 20 50 100 200 Pe
2 S 10 20 50 100 200 Pe
Figure 3. Values of K (a) and KtheO (b), the Heat

Transfer Coefficients, at the Test Sections. a) A -
Section No, 1 (Sodium); :Y - the Same (Sodium-
Potassium Alloy); - Section No. 2 (Sodium);
- Section No, 3 (Sodium); b) o; v. o, v - Corre-
sponding Values of Ktheo
of longitudinal heat fluxes in the greater part of Peclet numbers under study, in
particular in the sodium-potassium alloy. Calculations of K made with moder-
ate Peclet numbers yielded the expected result, i, e. , heat transfer by combined
convection improves with a reduction in Pe.
We wish to note two factors bearing on heat transfer with very low Peclet /37
numbers. On the one hand, the experimental study under these circumstances
is made difficult by the fact that it is necessary to take into account longitudinal
heat fluxes and due to technical difficulties (maintaining the flow rate, taking the
heat losses into account). On the other hand, the smaller Pe, the lower the ex-
perimental accuracy which is acceptable for making design calculations. In fact,
32
for specified inlet and outlet passage temperature and for a given fluid flow rate
the actual thermal flux through the wall will be expressed as [12]
_ Nu
NUfl eo 4Nu
1 {- Pee C1 I ^P)
With a reduction in the Peclet number the Nutheo becomes increasingly less de-
pendent on Nu, and when
2
Pea (1 + W) > 1 Nu o 4(11e-T) .
Since Nusselt numbers attendant to parallel combined convection are high, the
above factors have the most direct bearing on the results obtained.
The data on heat transfer in the annular gap for very low Peclet numbers,
needed for calculating a have not been published. However, extrapolation of the
available [13], [14] data to the needed range of Pe, is apparently permitted, just
as in the case of a round pipe. Reference [15] presents a study of the heat trans-
fer by mercury flowing in a pipe up to the laminar regime and it shows that the
heat transfer is practically unchanged for Pe < 100. The extrapolated data can
be only used when one is assured of the absence of natural convection in the an-
nual space. (Estimates which were made show that this condition is satisfied.)
According to [13], [14] the heat transfer in an annular space heated from one side
[only] is described by the expression
Nu .- 4.9 + 0.0175PO-18.
The thermocouples built into the wall of section No. 3 could not as yet be used to
find the heat transfer coefficient in the annular space due to low-frequency tem-
perature pulsations.
The data on heat transfer obtained for p arallel combined convection were
analyzed in coordinates Nu(Z), where Z = 2 'V4/va ax is a parameter which fully
describes the fluid flow and heat transfer in the steady-state laminar flow [2].
The use of Z for analyzing heat transfer in the nonstabilized region is not en-
tirely correct. But since the accuracy of experimental data is such that it is
[anyway] impossible to take into account the effect of the Reynolds and Prandtl
numbers, as well as other factors, Z was found to be the most suitable parameter for
purposes of analysis. Figure 4 shows Nu(Z) for stabilized laminar flow (solid
line) [2]. The results could also have been analyzed in the coordinates Nu(Cr/ /38
/Re), but this would not have reduced the scatter of points and would have intro- "
duced additional inconvenience in using the data. To find Z it is necessary
33
to know the liquid's proper-
ties and the temperature gra-
dient along the length, [while]
for calculating Cr and Re one
must know the rate of flow of
VV
the fluid and the mean temper-
ature difference. In the pres-
ence of longitudinal heat fluxes
the calculation of t requires
several. approximations.
It can be seen from Fig. 4
that there exists a relation-
ship for Nu(Z) which is simi-
lar to that taking place in a
laminar flow with parallel
combined convection, i. e. ,
Nu increases with an increase
in Z, remaining, however, be-
low the laminar curve.
3 4 if 10 15 20 Z
. The unusual location of
Figure 4. Nusselt Numbers for Parallel Com- points may be explained as fol-
bined Convection. o - Section No. 1 (Sodium); lows. In a purely laminar flow
v - Same as Above (Sodium-Potassium Alloy); the effect of the natural on the
o - Section No. 2 (Sodium); q - Section No. 3 forced convection is expressed
Sodium. as follows: the fluid layers
closest to the wall of the pipe
are imparted high velocities,
while in the center the veloc-
ity is reduced. Hence, the mean temperature difference between the wall and the
fluid drops, while Nu increase. Turbulence makes the velocity distributions more
flat, the velocity maximum near the wall becomes "washed out" to a certain de- /39
gree. In other . words, the increase in turbulence acts in the same direction as a
reduction in Z. If we analyze the structure of parameter Z derived_ for laminar
flow, then it becomes possible to qualitatively estimate the effect of turbulence
on the relationship between Nu and Z. As is known, in turbulent flow
vh. t
vh. t (Re) and ah. t = ah. t (Re); here always vh. t > v and ah. t > a. It is seen
from this that Z for turbulent flow calculated on the basis of the molecular values
of v and a is always higher than some actual Z h. t which is a function of Re and
Pr. Thus, relating the experimentally obtained values of Nu from a real turbu-
lent flow to parameter Z calculated. on the basis of properties of laminar flow one
should expect that the Nusselt numbers will lie below the line corresponding to
heat transfer in laminar flow.
The accuracy of the data obtained (A K theo Ktheo-116.5%) does not allow one to
follow quantitatively the effect of the magnitudes of Pr, Re and t on the Nusselt
numbers. Qualitatively it is possible to observe the heat transfer as a function
34
of Re (or Pe): at each section the heat transfer coefficient depends on the Rey-
nolds numbers for the same Z. The higher Re, the lower the Nusselt number.
The contribution made by the pipe's diameter is also noticeable, i. e. , for so-
dium the heat transfer coefficient in the region at hand depends weakly on d and
we get that Nu ^ d.
Points obtained in the sodium-potassium alloy are situated somewhatlower.
This apparently is due to the fact that the turbulent component of thermal conduc-
tivity of the alloy is higher than that of sodium. Hence, when using the coordi-
nates Nu(Z) the points for the alloy fall below those for sodium, although the val-
ues of the Nusselt numbers are situated at the same level. We can use the fol-
lowing relationship to describe the experimental points obtained for sodium
Nu = 0.3471.4.
The Nusselt numbers for the alloy are situated by 40% lower.
At section No. 2 we have obtained data on the heat transfer attendant to op-
positely directed [counterflow] combined convection (Fig. 5). In laminar flow de-
scribed in [2] the Nusselt numbers become lower in this case for Z = 0-3, while
when Z > 3 the flow is unstable. In our experiments we have not noticed a de-
pendence on Z, since the turbulence of the flow was quite appreciable. The ef-
fect of natural convection is obvious, i. e. , the lower the Peclet number, the low-
er the heat transfer rate. The increase in Nu with increasing Peclet numbers to
values above Nu = 5 + 0. 025 Pe 0. 8 is due to the fact that the section is short
(1 h. t/d = 5) and a turbulizing mesh has been installed at the inlet.
Figures 6 and 7a and b show the temperature distributions recorded under

different conditions. The not too high smoothness of the distributions is due to
the low-period temperature pulsations in the flow. For accurate reading of a
single point of the distribution with small Peclet numbers it is necessary to av-
erage these pulsations over a large time interval (up to 10 minutes). The most
minor variations in the flow conditions distort the results of temperature read- /41
ings. Hence,the temperature distributions are only qualitatively correct. We
note that, unlike the ordinary pattern, the temperature distributions at the be-
ginning of the heat transfer section are here always greater than in the following
sections (see Fig. 6). It is seen from Fig. 7b that, for the same average ve-
locity, the shape of the temperature distribution is highly affected by the thermal
flux at the wall. At the same time the heat transfer coefficients does not vary ap-
preciably (Nu jlSt). When the average velocity of the liquid is reduced, the
drop in the temperature distribution becomes larger, it widens. At very low ve-
locities the distribution becomes flat. Qualitatively this is similar to the behav-
ior of the temperature distribution for laminar combined convection [2].
The thermocouples installed on the jacket of section No. 1 and on the heat
transferring wall of section No. 3 made it possible to determine the trend in the
average temperature along the length of the section. The temperature variation
in the jacket wall along the length remains linear except for very small Peclet
35
yu num bers, when the he at
lost through the lateral
: surface of the jacket re-
sults in an appreciable
. increase in the mean
temperature difference
s between the passages.
Linearity in the temper-
2 t = ature trend of the jacket
to such an extent shows
that the heat transfer co-
' 10 40 100 200 600 Pe efficient and
( consequently
q y
also the heat transfer in
Figure 5. Nusselt Numbers for Oppositely Directed the passages) does not
[Counterflow] Combined Convection, o - Section No, 2 vary too much along the
(Sodium); Solid Line - Curve of length of the section.
This fact justifies the
Nu = 5 + 0.025 Pe08, applicability of the meth-
od for taking into account
longitudinal heat fluxes to
the analysis of preliminary results, which is presented in [11].
The only insignificant variation in K along the; length maybe attributed to the
species of the fluid flow with combined. convection (return flow of fluid. in the cen-
ter of the pipe),
Figure 6. Tempera-
ture Distributions in
Section No. 1 at Differ-
ent Distances From the
1.0 0,5 0 0.5 1,0 1,0 0.5 0 0.5 r/R Inlet (Sodium).
36
t-tW,
3 71_
0 F-
0 017 0.4 0.6 0.8 Ro_r 0 0,2 0,4 0,6 0.8 K.-r
a) Ro b) Ro
Figure 7. The Temperature Distribution in Section No. 2 as a Function of:

a) -Average Velocity of the Liquid b) - the Heating in the Section
Metal; (Re = 7 103):
w, mm/sec Re St, C
82
Q -- 55 A - 27 15 000; 4700; O -224,. V - 99;
CO 0-9.1 9700; 1500; it - 180; - 80,
.1 - 36 0 - 4,0 6300; 740 A - 1414;

Results
1. In the region of Peclet numbers and Z under study the heat transfer of
sodium and of the sodium-potassium alloy on parallel combined convection is ap-
preciably higher (and on counterflow it may be lower) than the heat transfer with
forced convection only.
2. The heat transfer coefficient (and consequently also the heat transfer)
in short ducts of experimental sections varies little along the length, which is
shown by the linear temperature variation of the section's surface,
3. The study of heat transfer in the region of low Peclet numbers is dif-
ficult. At the same time, it is possible to use for practical purposes data the
less accurate the lower Pe and the higher Nu and (p.
4. The qualitative picture of the behavior of the temperature fields in the
sections is in agreement with theoretical predictions made on the basis of the so-
lution to the problem of stabilized laminar flow [2].
5. Despite the small variation in the heat transfer coefficient in short ducts
its dependence on the duct length remains unclear.
NOTATION

t he t length of the heat transfer section /43
I total length of center pipe
x longitudinal distance
t average temperature of liquid in the center pipe, C
St difference between the liquid temperature at the inlet and outlet
of the center pipe, 'C
At theoretical value of the liquid-to-liquid mean temperature dif-
theo ference in the heat exchanger, 'C
G flow rate of liquid through the test section
cp ratio of the longitudinal conductivities of the pipe wall and of the
liquid. in the pipe
38
REFERENCES
1. Subbotin, V.I. , et al.: Teplofizika vysokikh temperatur, 1, No. 3 (1963).

2. Brown, W. G. and Grassman, P.: Forsch. Geb. Ingenieurwesens, 25,
No. 3, pages 69-78 (1959).
3. Ber, L. E.: Izvestiya AN SSSR, OTN, No. 6, page 25, 1962.
4. Ostroumov, G. A.: Svobodnaya konvektsiya v usloviyakh vnutrenney zadachi
[Free Convection Under Conditions of the Internal Problem]. Gostekhizdat,
Moscow-Leningrad, 1952.
5. Hanratty, T. J. , Rosen, E. M. and Kabel, R. L.: Industr. and Engng Chem.,
No. 5., 1958.
6. Brown, W.G.: VDI-Forschungsheft, 480, 1960.
7. Watzinger, A. and Johnson, D. G.: Forsch. Geb. Ingenieurwesens, 10,
No. 4, pages 182-196 (1939).
8. Wetjen K. A.: Chem. -Ingr. -Techn. , 26, No. 8/9 (1954).
9. Alad'yev, I. T., et al.: Izvestiya AN 9SSR, OTN, 1951.
10. Petukhov, B. S. and Nol'de, L. D.: Teploenergetika, No. 1, 72 (1959).
11. Gorchakov, M. K.: Uchet prodol'nykh peretechek tepla pri izmerenii sredney
temperatury zhidkosti v trubakh i teploobmennikakh pri malykh chislakh
Pekle [Taking Into Account Longitudinal Heat Fluxes When Measuring the
Average Temperature of Fluids in Pipes and Heat Exchangers for Low
Peclet Numbers]. Article immediately following this in the present
collection.
12. Trefethen, L.: Trans. ASME, 78, No. 6, pages 1207-1212 (1956).
13. Subbotin, V. I. , et al.: Atomnaya energiya, 9, No. 6, 461 (196 0).
14. Harrison, W. B. and Menke, J. R.: Trans. ASME, 71, No. 7, pages 797-
804 (1949).
15. Yushin, A. Ya, et al.: Izvestiya vuzov. Energetika, No. 7, pages 79-85 (1961).
39
TAKING INTO ACCOUNT LONGITUDINAL HEAT FLOWS WHEN MEASURING
THE AVERAGE TEMPERATURE OF FLUIDS IN PIPES AND
HEAT EXCHANGES. FOR LOW PECLET NUMBERS
M. K. Gorchakov
It is assumed in the study of heat transfer with qw = const that the real mean /43
temperature of the liquid in the duct varies linearly from t ( X _ 0 = t I X _ - 00 to
t X = L = tIX _ -, i. e., t = ttheo (X = 0 and X = L pertain to the start and end of
I
the heated section). However, for low Peclet numbers the axial heat fluxes re-
sult in the fact that t differs from ttheo by some value St. If this difference is not
taken into account in analyzing the experimental data and in calculating heat ex- /44
change devices, this can produce erroneous results. The authors of [1-5] had
as their purpose finding the average temperature of the liquid in the presence of
longitudinal flows of heat. Expressions were found for 6t for heat transfer in
hydrodynamically stabilized laminar flow and stabilized heat transfer in any flow.
It turns out that these data are insufficient for analyzing experimental data ob-
tained in the study of heat transfer in the recuperating part of cold traps of alka-
line metal oxides The heat transfer conditions are quite varied: a short double-
tube heat exchanger (see figure), heat transferring wall with an appreciable longi-
tudinal conductivity, and the velocity and temperature fields are interrelated, and only
their qualitative picture is clear; heat transfer coefficients in both passages are
significant.
Below we attempt to find a relationship between the actual mean tempera- /45
ture of the fluid t and its theoretical value ttheo in tubes with qw = const
and in heat exchangers with identical flow rates in both passages.
The effect of longitudinal flows of heat on the mean temperatures of the fluids
in passages of a real heat exchanger is a complex problem, although in principle
solvable. By considering separately the following principal problems and then com-
bining the results thus obtained. it is practically possible to achieve the same results:
1) behavior of the mean temperature of the liquid in the general case of nonstabilized
heat transfer in the presence of longitudinal heat fluxes; 2) dependence of 6t on the
shape of the inlet and outlet parts of the passage; 3) the value of 6t h e in two-dimen-
sional flows in the passages of a counterflow heat exchanger without taking into ac-
count the conductivity of the heat transferring wall; 4) taking into account the latter
conductivity; 5) the dependence of 6th, e in the heat exchanger on the changes in the
heat transfer coefficient and of the mean temperature difference along the length
of the passages.
40

As a result of consideration of these problems it is possible to obtain ameth-

od for introducing corrections for longitudinal fluxes of heat for calculations of the
mean temperature of the fluid in passages with q w = const and in heat exchangers.
1. The temperature distribution in a nonstabilized flow in a pipe is given by

the expression
1 2
Yotheo [w (x, r) ae { v (x, r) Tr
_ = ar
T {T[^
AT ( x , r)) as I T
TX (1)
if it is assumed that the physical properties of the liquid are constant, while the
turbulent component of the thermal conductivity is small in comparison with the
molecular component. Since we
are interested only in the behav-
ior of the mean temperature of
the liquid in the presence of long-
itudinal heat fluxes, it is possible
to obtain the required result by
averaging Eq. (1) over the pipe
cross section. Multiplying it by
27rrdr and integrating both parts
from r = 0 to 0rwe get
b) I I 11 I 111
-
- YotheoFwOX =F a 2nr o q , (2)
I [
where q = - X(000/i9r)r = r is
0
the thermal flux at the wall It
is clear from this that the char-
acter of the heat transfer does
not affect the distribution of the
mean fluid temperature along the
length of the duct. In any case
this distribution can be found
from the simple equation of bal-
Change in Mean Fluid Temperatures in Heat ance (2).
Transfer Passages a) Heat Exchanger De- Consideration of a specific
sign [1) Mixing Chamber; 2) Heat Transfer- problem with qw = const, with-
ring Wall; 3) Annular Space; 4) Center Pas-
sage; 5) Inlet Screen]; b) Schematic Replacing out taking into account the wall
the Actual Heat Exchanger When Finding conductivity, [4], yields the
at h. ea - t + atsp (I, H and III Are Heat Transfer expression
Zones); c) Curve Showing Variations in t and Pe(L -s))

= 4[1e-
St (3)
ttheo in [Simplified] Circuit b,
41
when 0:5 X C L
r_
St= L 6tdX= AP Pe L)1
e 3 Cl--P' (1--e JJ (4)
If Pe . L > 6, then /46
6t_

._ APe C 1 PeL (5)
2. When working with short passages without hydrodynamic stabilization

sections it is important to know the effect exerted by inlet and outlet devices on
the magnitude of St for small Peclet numbers. The heat transfer coefficient at
the initial section definitely depends on the shape of the inlet, and in some cases
also on the shape of the outlet (in the case of intensive effect of natural convection)
[sic]. This dependence on the shapes of the inlet and outlet canbe clarified by solv-
ing the following problem.
Let the passage in the zone 0:5 X:5 L in which heat transfer takes place
have a cross section F II, cross section F for X < 0 (narrowing down the fluid-
passing cross section by means of a grid with low-thermal conductivity in which
case F < FII, or its expansion in the form of a chamber in which case F > FII),
and cross section F III when X > L. In this case we shall write Eq. (2) for zones
I, II and III
anal
BFI axT wlFlc the0T 191,_
ax ^'
a
"pll a ^ WIIFIICthe'V aaxl + 27troq == 0, (6)
AFIII as l WIIIFIII0 the0Y aaxl 0'
Boundary conditions
X =-00, tI= 0;
X = 0, tI = tII, XFI
axI = XFII a" ;
X= L , t II = t III, a'FII aaz = ^FIII allll ' (7)
X = 00, t III = Const.
42

Reducing Eqs. (6) to the dimensionless form and making use of the fact that
wI F I =W 11 F Ii = w III FII19 we can get expressions for the mean temperature by
zones:
FT Pe X
Pe L) a Fl (8)
tl XPe2(Ia
tII XPe X+ XPe2

(1 e" Pe (L-X)l+ (9)
till = IN
qd
L. (10)
It can be seen from Eq. (9) that the shapes of the inlet and outlet parts of
the passage do not affect the mean temperature in the heat transfer zone. In the
section before heat transfer starts the reduction in 6t as X -- -.0 depends on the
relationship between F 1 and F I ; this factor should be taken into account when
placing the pickup measuring the temperature of the fluid at the inlet. It is thus
clear that d t = t ttheo in the heat exchanger zone can be calculated without con- /47
sideration of the internal structure of the flow and without taking into account the
specific end devices.
3. The above simplifies the analysis of heat exchangers to the extreme. For
finding the distribution of mean temperatures and of 8ttheo the actual heat exchang-
er can be replaced by a simplified arrangement (see Fig. 1b and c), since only the
heat transferring zone is of interest. Assuming that K, the heat transfer coeffi-
cient, is constant along the length and treating the wall through which heat trans-
fer takes place simply as a thermal resistance, it is possible to write a system
of equations describing the trend of the mean temperature in both passages in all
the zones. For this it is necessary to replace q = const in Eq. (2) by the expres-
sion K(t t sp). In zones I and II K = 0, in zone II
K=1 d j d d1
T +a s d i
+2K Ind
(11)
The subscript II is dropped for convenience when writing equations in the dimen-
sionless form.
Zone I
a2tSpi _ Pe 8tspi 0,
8X 2 V. OX

(12)
43
a 2 tl at.I.
ax2 + Pe ax
= 0. (13)
Zone II
a2tTp Pe atyp 4Nuk

x ax x (t tsp) = 0.,
ax e (14)
a 2 +Pe aX-4Nu k (t tS,) =0.

X (15)
Zone III
a'III
^p Pe al
-y-- SPI I I= 0
aX2 % ax (16)
aaX, + Pe
aax l = 0. (17)
Boundary conditions
X = CO tI = Const, tspl = 0; 1
X = 0 tI atI _ at atAp l _ at sp.
t t '_' ax ax' ax ax'
}
at atIII at sp_a t III
X L t = tIII, tsp = tIII, ax aX ' ax ax (18)
X = 00 tIII = t o, tsplII = COIlst. )
Solving simultaneously Eqs. (12)-(17) with the above boundary conditions

yields a system of algebraic equations with 12 unknown constants of integration.
Since the arrangement assumed to represent the heat exchanger yields the cor-
rect trend of mean temperatures only in the heat transfer zone, it makes sense /48
only to find the four unknowns, which enter into the description of t and t g , Due
to the cumbersomeness of intermediate calculations only the expressions for the
heat transfer zone are presented here
top= CI eRIx +. C2 p eR2x +. CSX + Cv,

â ( 19 )
le
t Ci C^ +P / eslx+C' ^Ra +Q / es2x -f Cs C'4Nuk +`YJ +C4i
i ( 20 )
44

= Pe 12x xf ^C12
01,2 Pe x 4N+ul,
/2
P=
4Nu k CxYi
X Pe
Q 4Nu k \xP2
Ci = [(1 + Q^s) e SaL ]. (21)
The value of H is not presented, since it is very cumbersome and it is not needed
for calculating the theoretical expressions.
C2 = 8 [esiL(1+PNi)l,
C3
= H [ (1 + PPi) ep2L + ( I + QPz) esiL]
\
C4 = Ci
C Pe ri Yi l
+ C2 C P p 2 2 / +C3 Pe '
At theo = (t ts)I X= - w
Ci
CP + Pe pi /+ C2 (Q 1 + C3
+ Pe P2 / C 4Nu h + Pe ) '
K _ p
_at 1+ St _1+wL,h.e.
Kp (itsP Oi (22)
(3 L (3 L
If e 1 >> 1 and e 2 1, we have
(Q+- e ('+QP;) C 4Nuk + Pe

^2 / P2
L [(1
1-Qî) PQ1 =-( 1 } Q^2) 4-N, (23)
When L -
y2
WL, h.e.It -,m = W T = Pea
4Nukx Y2 (Y2 - 1 )- 1 (24)
where y2 is found from the expression
yi, z =1 PPex = 0,5 (1 ; Y)1 ^1 } 4Nu" 1+x ' (25)
45

Calculations using Eq. (24) yield the expression /49
= p ,, (1 + x) ( 26 )
Substituting the above expression into Eq. ( 23), we will get a simpler formula for
a heat exchanger of arbitrary length
1+(D
1 + u)L _ Y!
CO
1 PjL r YA L (27)
The above expression can be used to calculate K/Ktheo for a heat exchanger in
which the longitudinal thermal flux along the intermediate wall is negligible as
compared with the flux along the fluid in both passages.
4. If the wall is conducting in the longitudinal direction, the situation is
much more complicated. The problem of the behavior of mean temperatures in
the passages of heat exchangers cannot as yet be solved exactly. However for
moderate co and for heat exchangers of moderate length it is apparently possible
to use results obtained on the following assumptions: a) the longitudinal conduc-
tivity of the wall is regarded as an additional conductivity of the fluid in the pas-
sages; b) in the statement of the problem the heat exchanger ' s passages do not
in anyway differ from one another, except for the difference in the area of the
flow passing cross sections. Hence the conductivity of the wall is divided be-
tween the passages in proportion to their cross sections, i, e., co = co 1 + c02,
x 1
^1 = 1+x rP, rP2 = q
On these assumptions we write Eqs. (14) and (15) in the form
r7=r Pe 8t 4Nuk
OX 2 x+(Pi O+ x+(P1 (ttsp)=^, (28)
Pe 81 _ 4Nuk
(t tsp)= 0.
aX
a22 +1+T2 aX 1 +(P2 (29)
46

Similar changes will take place also in the expressions for zones I and III; the
boundary conditions retain their form. The form of the solution of the new system
of equations remains the same
K 1+(Oh e. (P ,
1 + WL h.e. Ir _
Ktheo 1 _ y î h.e.. 1 + h.e.
PWiL y 2 p QS L (30)
1-}- x x
y^ 1 , 2 C 1 +
14wh . e.^(i { x) 2 ^ (31)
2=
1+x
Apt, z = (1 y(pi, z) (I +X + (F) r. Pe, (32)
4Nuk
Wh.e.m=^L.h.e.ÎL-,^= p e2 (1 -r r. c^). {33)

The use of the above set of equations is convenient when calculating Ktheo on the /50
basis of a known K.
When analyzing experimental data the calculations proceed in the opposite
direction, K is unknown, it can be found by successive approximations. However,
it is more convenient to use the relationships obtained by solving the problem of
variation in the mean temperatures in the reduced representation of a heat ex-
changer in which a constant thermal flux through the wall is assumed. This prob-
lem, together with the assumptions made for finding Eqs. (30-33), makes it pos-
sible to consider each passage separately with a part of the wall, i. e. , the prob-
lem of a heat exchanger with qw = const consists of two problems about passages
with qw = const, similar to expression (6), where XF H should be replaced by
XF II + X w F w ( 1/1 +:x) for the center passage and by XF II + X w Fw (x/1+x) for the an-
nular space. The expressions for St and 8t sp found under these circumstances make
it possible to find the relationship between the actual and theoretical heat trans-
fer coefficients.
atheo
Q = (t+x)PeL
4otheo W 1 -} %+q) (1 e i+x+m )]1
1 XPe2 ( + f+x L 1(1 { x) Pe L (34)
^sp theo
Qa = (1-^-x)PeL
4a 1 (P 1 _x (1 } x I q)) (1 _e ('+x+(P)x)1 (35)
1 %Pe2 x C1+ 1+X) (1+X) PeL J
47

Substituting the two above expressions into Eq. ( 11), we get
_ Ktheo
_
Kq
_ 4theod'
1 j,P_O ('+ X. 'P) tq (36)
where
(f+x)PeL _ (i+x)PeL
(11 x)
PeL
[1e^ 1+x +(p .+ x2 (1 a O+x+Ox )), (37)
37
or, in other words (calculating Kq, theo on the basis of K)
K
Kq, theo = 1 (38)
(O T,
In principle Kq, theo obtained from the above expression and Ktheo obtained from
Eq. (30) have different values, which difference disappears with an increase in
L and Pe.
Practically for calculating K on the basis of Ktheo it is sufficiently conven-
ient to use the expression
K_ Ktheo
1 ` 4Khheo d ( 1 + x To ^q
%Pe 2 jp (39)
where /51
^h.e.ro s't

W L, h.e. tp (40)
The magnitude of T can be calculated for a specific heat exchanger by assum-

ing values of Nuk. Numerical calculations show that for a specific heat exchanger
T depends little on Nuk, varying appreciably (by several percents) with changes
in the Peclet number.
5. Since all of the preceding analysis was performed on the assumption that
the heat transfer coefficients remain unchanged along the length of the passages,
it is necessary to examine the behavior of St and S'th e for the case when they are
48
variable. In a duct with a thin wall, when q w, = const, St obviously is independent
of changes in a along the length. Approximately (without taking into account the
effect of heat leakages in the wall at the ends of the heat supply section) the same
may be claimed also for a wall with an appreciable longitudinal thermal conduc-
tivity.
_ The b_eh_avior of mean temperatures in heat exchanger passages, as well as

Sttheo and K/Ktheo are affected by changes in the heat transfer coefficient along
the length, as well as by the heat losses through the outside surface, since qw(X)
= K(X)-At(X). Also here it is possible to analyze both passages of the heat ex-
changer separately [system of Eqs. (6)], with q W = q(X).
Since in order to estimate the effect of q(X) on St it is necessary to know the

specific relationship governing the former quantity, it therefore makes sense to
consider only one passage without the conducting wall. Solving Eqs. (6) with
boundary conditions (7) for a variable qW in the heat transfer zone we get
% L
t$ -
TPe [ 4
(X) dX + ePex J 4 (X) e Pex dX
o x JJ (41)
Lp
$tg = 74d , pex \ Q (X) e--Pex dX.
X (42)
If information on q(X) is available it is possible to calculate 6t X and S for

both passages making up the heat exchanger and, upon using Eq. (11), to find
X^ --X, theo' However, this is apparently not needed since the difference between
KK^^
6t and 6t found from Eq. (3), with q = q(X^ is found to be quite moderate. In fact,
let q(X) = q 0 + bX, q(X) = q 0 + 0.5bL, then
PeL+1
6tX= XPe 2 1( t b Pe )(1e Pe([. x) )b(L-1^)1
(43)
ePei)1 2 (44)
8tx = APe2
I ( 4 + b 1'e-Pt l 1. 1 P ,L ( i
49
The difference between Rx and 8t can be seen from the following expression /52
(for e- Pe- L 1)
_ bL 1 _ 2
8tx _ 1 = 2 PeL Pe2L2
St
(40 2 ) C 1 PeL
For q0 = 0 we have
_
bt 1 _ PeL
2
(45)
at _ i PeL-1
Thus, even when the thermal flux varies appreciably along the length of the
passage, the St (and consequently also K/Ktheo) calculated on the basis of the av-
erage magnitude of the flux does not produce appreciable errors if the product Pe.
L is not very small. In the presence of a conducting wall the following correction
should be introduced: Pe. L in Eq. (45) should be replaced by Pe- L/ 1 + co.
Results
1. In the case of qw = const the displacement of the actual mean tempera-

ture relative to its theoretical value does not depend on the shape of the velocity
distribution and on velocity variations along the length of the passage.
2. The shape of the passage ahead of and past the heat transfer section does
not affect St
3. Finding the distribution of mean temperatures in the heat exchanger pas-
sages in the case of a low-conductivity wall and with a constant K yields a method
for finding K/Ktheo.
4. Upon making certain assumptions it is possible to find a method for tak-

ing into account the longitudinal conductivity of the heat conducting wall when cal-
culating K/Ktheo'
5. The use of the average thermal flux for finding K/Ktheo does not yield,
under certain conditions, appreciable errors even when the thermal flux varies
highly along the length of the passage.
50
6. The use of the above results in analyzing experimental data of [6] for not
too low Peclet numbers has yielded expected results.
NOTATION
O(x, r) temperature of a point, C

6t difference between the actual mean temperature in a
passage and the theoretical value brought about by
longitudinal heat fluxes, 8th, e = 6t + 8tsp, C
At actual mean temperature difference, At = t t
for a heat exchanger, C sp
W relative difference in the fluid's temperature due to

longitudinal heat fluxes, averaged over the length
x longitudinal coordinate /53
r current radius
r0 radius of pipe, d = 2r0
d1 internal diameter of the annular space
dsp [representative] flow diameter of the annular space
1 length of the heat conducting part of passage or heat
exchanger
x = Fsp/F ratio of cross sectional areas of the heat exchanger
passages
rp = F 'w/F-' ratio of longitudinal conductivities of the heat conduct-
ing wall and of the fluid in the main passage
v(x, r), w(x, r) transverse and longitudinal components of the flow
velocity in a point
Nuk = Kd/X Nusselt number calculated for the heat transfer coef-
ficient
Dimensionless quantities
X = x/d, R = r/r 011 L = 1/d are the longitudinal coordinate, radius and
length of the heat conducting section, respectively.
51

Subscripts
theo theoretical value

w wall
sp space
L arbitrary length of the heat conducting section
h.e heat exchanger
q constant thermal flux on the wall
x variable thermal flux on the wall
REFERENCES
1. Labuntsov, D. A.: Doklady AN SSSR, 118, No. 6, 1118 (1958).

2. Genin, L. G.: Teplofizika vysokikh temperatur, 1, No. 2, 247-253 (1963).
3. Pirogov, M. S.: Zhidkiye metally [Liquid Metals], page 94, Gosatomizdat,
Moscow, 1963.
4. Trefethen, L.: Trans. ASME, 78,, No. 6, 1207 (1956).
5. Petukhov, B. S. and Yushin, A. Ya.: Doklady AN SSSR, 136, No. 6, 1321 (1961).
6. Volchkov, L. G., et al.: Teplootdacha natriya i splava natriy-kaliy v korotkikh
vertikal'nykh trubakh pri smeshannoy konvektsii [Heat Transfer of a Sodium
and a Sodium-Potassium Alloy in Short Vertical Pipes by Combined Convec-
tion]. Article in present collection immediately preceding the above.
52
STATISTICAL CHARACTERISTICS OF TURBULENT TEMPERATURE
FLUCTUATIONS IN FLUID FLOWS
V. P. Bobkov, M. Eh. Ibragimov and V. I. Subbotin
A characteristic feature of turbulent flow is the disorderliness, randomness /53

of motion which results in random variations in the velocity, pressure, tempera-
ture, etc. , in time and space. Statistical averaging of instantaneous values makes /54
it possible to separate the steady part of the velocity and temperature fields. The
irregular part of turbulent flow can be described by statistical methods. Total sta-
tistical description of turbulent motion is sufficiently complex and requires exten-
sive information on the values of the correlation and spectral tensors, on the prob-
ability distribution, etc. New experimental data in the field will make it possible to
more fully represent the mechanism of turbulence processes and to create a phys-
ically substantiated statistical theory of fluid flow and heat transfer, free of arti-
fical prerequisite conditions and assumptions.
At present many works have been published on the statistical nature of tur-
bulent flow [1], [2]. All publications support the idea that it is necessary to
measure experimentally the following specific quantities:
1) Averaged velocity and temperature fields, their dependence on the physi-
cal parameters of the fluid and on the Reynolds number. Experimental work which
was performed makes it possible to qualitatively estimate the magnitude of the tur-
bulent transfer of momentum and heat for different fluids in different flows; 2) root
mean square values of the velocity and temperature fluctuations, reflecting the in-
tensity of the process; 3) spectral densities of the fluctuations; 4) space-time cor-
relation coefficients in one and more points of the flow; 5) probability distribution
at various time instants and in different points of the flow.
The existing statistical theories make use of average statistical character-
istics. Only time averaging is practically possible in experiments. The relation-
ship between the average statistical values and values averaged with respect to time
is established by the ergodic theorem. In hydrodynamics this theorem is not theo-
retically proved, however, experiment data show that it can be used in the study of
hydrodynamic phenomena [2]. The validity of the ergodic theorem for temperature
fluctuations is proven in [3].
Under conditions of constant heat transfer, the velocity and temperature fluc-
uations are steady-state random functions. Such a function can be represented as a
system of random quantities, i.e., of ordinates of a random function during speci-
fied, sufficiently small time intervals. Consequently, the study of statistical char-
acteristics of stationary random functions can be reduced, to the study of a system
of random quantities making use in the process of the probability theory theorem.
53
This circumstance will be used in the present article for obtaining the distribution
function and other statistical characteristics.
The empirical information on the statistical characteristics of turbulentflow
in ducts is not too extensive. The available experimental data pertain mainly to /55
turbulent fluctuations of velocities in the flow of air or water [4-9]. No syste-
matic information is available on the statistical characteristics of turbulent tem-
perature fluctuations, although, obviously, only the study of temperature fluctua-
tions and of the statistical relationship between them and the velocity fluctuations
will make it possible to clarify the mechanism of turbulent heat transfer.
The present article presents results of measurements of certain statistical
characteristics of temperature fluctuations in turbulent flow of various fluids with
highly different thermal conductivities (water, mercury, lead).
Methods of Measurements, Apparatus
The measurements were performed in a round pipe 52.2 mm in diameter at

a distance of 30 diameters from the start of the heated section and of 40 diameters
from the pipe inlet. The temperature in the flow were measured by chrome-alumel
thermocouples with an external diameter of 0.3 and 0.5 mm. They were fastened
in a special probe. Reference [3 and 41 depict the layouts of the section, probes and
microthermocouples. The test pipe was held vertically. The thermal flux was pro-
duced. by a band-type electric heater, which was double wound. on the pipe to reduce
electromagnetic interference in the thermocouple circuit. The fluid was lead into
the pipe past the elbow in the loop. A part of experiments with mercury was per-
formed with a grid, consisting of a steel plate 3 mm thick with holes 2.5 mm in
diameter, forming a square lattice with a hole-to-hole distance of 4 mm at the pipe
inlet. These experiments have shown that the measurements were performed un-
der conditions of hydrodynamic and thermal stabilization of the flow.
To measure and automatically calculate the root mean square values of the
temperature fluctuations and the correlation coefficients we have used the 46K1
correlometer (analog computer) and special amplifiers with a transmission band
of from 0.15 to 200 cps. The oscillograms of turbulent temperature fluctuations
were recorded on 35-mm film by means of the MPO-2 oscillograph. This infor-
mation was used for determining the probability density distributions and other
statistical characteristics.
In measuring turbulent temperature fluctuations one most know how the in-
ertia, relationship between the specific heats of the thermocouple and the turbu-
lent flow and distortions of the structure of the turbulent flow by the thermocouple
in the point under measurement effect the experimental results. The effect ,of the
inertia on the measurement of fluctuations can be estimated by considering the ef-
fect of the latter on the root mean square values of the temperature fluctuations,
if the spectral function of fluctuations and the time constant of the temperature
under the specified conditions of heat transfer between the thermocouple junction
and the medium are known.
54

If we replace the stationary random function by a stationary linear system,

each of the ordinates of its spectrum will be multiplied by the square of the modu- /56
lus of the transient function of the system for the corresponding frequency [11], [12].
S i ( o)) = I (D ( i(o ) 1 2 S o (co). (1)
The transient function for a thermocouple can, in the first approximation, be rep-
resented as a purely aperiodic component [13], i.e.,
io) _ 1 8 i arctgTTw
( ) V(TT(0 )2 + 1 (2)
Multiplying both sides of Eq. (1) by dw and using Eq. (2) and integrating, we will
get the dispersion of the real (undistorted) signal
D = 2 S Si (co) [(TT(0)2+1[ dco. (3)

0
Thus, in order to find the real value of the temperature fluctuation intensity, it is
necessary to know the measured spectral density and T th, the thermocouple's time
constant which, in general, depends not only on the dimensions of the junction and
its thermophysical properties, but also on the conditions of heat transfer between
the fluid and the thermocouple junction.
The inertia was studied. separately. Numerical values of the time constant
were obtained (experimentally) for different thermocouples under unequal condi-
tions of heat transfer between the thermocouple junction and the surrounding med-
ium. It was found that for the thermocouple 0.5 mm in diameter the time constant
does not exceed 0.04 secs (when wetted with water flowing at 0.2 m/sec); for a
thermocouple 0.3 mm in diameter it was not more than 0.01 secs (when wetted
with water flowing at lm/sec). It is natural that when wetted with mercury and
lead the heat transfer in the junction is very high and the time constant will be
even smaller [14]. Preliminary results [3-11] have shown that S 1w, the spec-
tral function of temperature fluctuations has a value close to unity as early as at
a frequency of 10 cps, which attests to the negligible energy which is carried by
the higher frequencies (above 10 cps). Calculations using Eq. (3) have shown that
for frequencies from 0 to 10 cps the error produced by the inertia in measuring
the temperature fluctuation intensity does not exceed 1% for thermocouples 0.5
and 0.3 mm in diameter. The measured fluctuation intensities obtained by differ-
ent thermocouples do not differ. Equation (3) is suitable for estimating the cor-
rection only in the case when Tth, the time constant, does not vary with time, i.e.,
when the coefficient of heat transfer between the junction and the fluid is constant.
An estimate of the effect of unsteadiness on the heat transfer has shown that the
55
heat transfer coefficient under conditions of our experiment differs little from the
steady-state coefficient [14].
The effect of the distortions introduced by the thermocouples was estimated
indirectly: the intensities of fluctuations in the given point of the flow (center of
pipe) were measured with the thermocouples situated at different distances from
one another (Fig. 1); it was found. that Qt , does not change. This, apparently, can /57
be explained as follows: the temperature fluctuations in the given point are deter-
mined mainly by the turbulent field in the preceding, rather than in the given sec-
tion of the pipe. The second thermocouple also had no effect on the probability
distribution function.
It is known that the accuracy in determining the turbulent characteristics de-
pends on the averaging time T. Reference 15 recommends formulas for selecting
the time interval over which T is to be integrated for a given experimental accuracy.
An estimate made using these formulas has shown that in the case of T = 5 min (max-
imum integration time of the . correlometer) the error in determining second-order
moment is not more than 20/6. The oscillograms which were obtained are recordings
over time intervals from 15 secs to 2 mins. When T = 15 secs the maximum error
due to the fact that the integration time is finite should comprise not more than 45%.
Actually it turned out to be much lower. Comparison of results obtained by the cor-
relometer and of those produced. by analysis of the oscillograms shows that they agree
satisfactorily. It is natural that the error in measuring moments of higher order for
the same integration time.
Intensity of Temperature Fluctuations
The intensity of temperature fluctuations was measured for the flow of water / 58
(Re = 16, 000), mercury (Re from 5, 000 to 125, 000) and lead (Re from 30, 000 to
, 216, 000). Experimental data for water (thermal flux density 2.3-10 4 watts/m2)
and mercury (for fluxes from 1. 16 . 104 up to 2.3 . 104 watts/m2 ) were obtained by
using the correlometer; for lead (for fluxes from 1.8 . 104 to 3.1 . 104 watts/m2)
they were obtained by analysis of the temperature fluctuation oscillograms. Here
the time pace was 0.01 secs, , which enabled us to satisfactorily reproduce the fluc-
tuation oscillograms up to frequencies of the order of 10 cps [16]. The measurements
were made over the entire pipe cross section (from the pipe axis to contact with the
wall).
To determine the manner in which the intensity of temperature fluctuations

depends on the thermal flux density, we have performed experiments (for mercury)
with Re = 16, 000 and thermal fluxes of 1.2 10 4 , 1.8-10 4 and 2.3 . 104 watts/m2.
It was found that the intensity o vt, is proportional to the thermal flux density.
The manner in which the dimensionless fluctuation intensity varies for mer-
cury and water for Re = 16, 000 over the pipe cross section is shown in Fig. 2.
56
The fluctuation intensity
curves have a maximum for
water near the very wall, when
while those for mercury and
lead are at maximum at [some]
distance from it. This is due
to the fact that the tempera-
ture distribution in the main
body of the flow for a fluid
with large Reynolds numbers
is simplified and changes high-
ly near the wall, while for
liquid metals the mean tem-
perature distribution is rep-
resented by a flat-curve over
the entire flow cross section.
The maximum relative
fluctuation intensity for mer-
cury changes in a complex
manner with an increase in
Re (Fig. 3), i. e., first it in-
creases rapidly, reaching a
maximum when Re zz^ 20, 000,
and then decreases smoothly.
The fluctuation intensity curve /59
Figure 1. Location of Thermocouples in the tends to zero as Re approaches
Pipe. No. 1 and No. 2 are Thermocouples. the critical value, since the
temperature fluctuations pro-
duced by natural convection
are of such a low frequency
that they are not within the
transmission band of the am-
plifier. Experimental data
for lead were obtained with
Re > 20, 000. Here we were
0
able to observe the section
in which the fluctuation rate
drops with an increase in the
X Reynolds number.
Cd
5w
The above behavior of
\0. 5
the fluctuation intensity with
o o an increase in the Reynolds
number is due to the interac-
tion of two factors acting in
0 0,2 0.4 0,6 a8 1-^ opposite directions. First,
the turbulence is still devel-
Figure 2. Intensity of Temperature Fluctua- oping in the transition region
tions for Re =16, 000 Where the White Circles of the flow, the velocity fluc-
Denote Water, While the Black Circles are tuations increase with an in-
for Mercury. crease in Re, and consequently,
57
3

0 i
a
2
j
if
0 4 8 12 is Kei10-^
Figure 3. Maximum Intensity of Temperature Fluc-

tuations as a Function of Re. The Squares Denote
Lead, the Circles are for Mercury.
the temperature fluctuations increase also. Secondly, the maximum mean tem-
perature difference in the fluid. drops with an increase in the Reynolds number.
This should result in reducing the temperature fluctuation intensity. The latter
quantity thus changes depending on the relative strength of the two aforementioned.
factors. In the transition region the effect of the first factor predominates with
an increase in Re, while when Re is higher than 20, 000 the second. factor predom-
inates.
The above behavior of the fluctuation intensity with changes in Reynolds
numbers was observed almost over the entire cross sectional area of the pipe.
The fluctuation intensity increases continuously with an increase in Re to 70, 000,
then some drop in the fluctuation intensity is observed. For lead and mercury
the distance between the point of maximum fluctuation intensity and the wall first
increases rapidly with an increase in Re, and then this increase slows down. The
fluctuation intensity curve becomes more flat with an increase in the Reynolds
number.
Analyzing the experimental data for the three above heat transfer agents
we have found general quantitative laws governing the behavior of the tempera-
ture fluctuations. The first governing relationship is a satisfactory generaliza-
tion of experimental data for the maximum value of the fluctuation intensity and
the distance between this maximum point and the wall. Figure 4 shows the
curve of ovt /,max as a function of y/y max for all the heat transfer agents (the
1 1
Prandtl numbers differ by a factor of 300) obtained for the Reynolds numbers /61
and thermal fluxes at hand. The observed scattering of points is within the lim-
its of experimental error. The generalization attest to the similarity between
temperature fluctuation intensity curves as a function of ymax. The second re-
lationship pertains to the generalization of the dependence of the maximum fluc-
tuation intensity and of the distance between this maximum and the wall for dif-
ferent Reynolds and Prandtl numbers (Fig. 5). The experimental points for mer-
cury and lead can be generalized satisfactorily in logarithmic coordinates.
58
ô o ^ o
QS 0
^y p 0
0 p
0
0 1 2 3 4 S Y/Ymax
Figure 4. Fluctuation Intensity Curves.
O Re = 16,000 for water); Re = 30,000; Re = 64,000;
n Re = 216,000 for mercury, Re = t0,000; Cl Re = 161000;
Re = 28,000; Re = 55 , 00; Re = 70,000; q Re = 125,000
(for mercury);
N
x
X
ca
cd
200 400 600 1000 2000 4000 Pe
Figure 5. Curves Showing Y max Pr (1) and vt, /T+

(2) as a Function of the Peclet Number, o and are
for Mercury; o and are for Lead.
The above generalization shows that the observed. inhomogeneity of the tem-
perature fluctuation intensities over the pipe's radius disappears gradually with an
increase in the average velocity, and the fluctuation intensity maximum becomes
degenerate. This means that the turbulent flow tends to an isotropic flow with an
increase in the Reynolds number,
59
Spatial Correlation Coefficients and Scales of Temperature Disturbances
Spatial correlation coefficients R1.2 are major statistical characteristics of

turbulent flow. Study of these coefficients between velocity fluctuations in the flow
gives an idea of the scales of velocity disturbances, making it possible to estimate
the structure of the eddies [2]. Similarly, R1.2 between temperature fluctuations
in two points of the flow determine the dimensions of the temperature disturbances;
changes in these dimensions with a change in the character of the flow for fluids
with different Prandtl numbers provide us with some idea on the character of the
velocity fluctuations.
It is known D ] that R1, 2 between random quantities of the same kind of two
points of the turbulent flow has the following properties. It varies within the lim-
its of t 1. When the distance between the points is reduced, R 1.2 goes to unity,
which denotes a tendency to functional relationship between the fluctuations in the
two points. When the distance between the points is increased, R 1 , 2 goes to zero.
For independent random stationary processes R 1 , 2 is equal to zero, which is known
from the probability theory [171. For random stationary processes with normal
distributions the converse is also true. The fact that the correlation coefficient
is zero does not mean that there is no statistical relationship (in general).
When the distance between two points is reduced, the first derivative of the
correlation coefficient (with respect to space) tends to zero, while the second de-
rivative has a finite value when S = 0. The correlation curves for sufficiently
small 6 are parabolic [1]. This property is due to the fact that R1.2 varies sym-
metrically with respect to 6 near S = 0. Upon expanding the correlation coefficient
in a Taylor series, we will obtain the following relationship for sufficiently small S:
3
R1.2= (4)
0
where X 0 is a microscale, characterizing the average size of the dissipation eddies. /62
Except for these general properties it is impossible to analytically determine

the shape of the curve for anisotropic flow. In each particular case the character
of the curve is determined by the specific conditions: geometry of the passage, rate
of flow, properties of the fluid, degree of stabilization, the fluctuating quantities
which are measured (velocity, temperature, concentration, pressure, etc.).
We have studied the correlation coefficient for the flow of water (Re = 16, 000)
and mercury (Re = 10, 000-125, 000). Themovement of thermocouples duringmeasure-
ments is shown diagramatically in Fig. 1. Thermocouple No. 1 was placed at
a given point in the flow, while thermocouple No. 2 was moved relative to the first
60

along the pipe's diameter in the direction

of the axis of the flow. The distance be-
aa tween the thermocouples was varied from
a6 0.5 mm (contact) to 20 mm. These exper-
iments were performed without a screen
at the inlet.
- For each flow mode curves of R1.2
1 were obtained for different positions of
thermocouple No. 1 [from the center of
the pipe up to contact with the wall (Figs. 6
and 7)]. For mercury the correlation co-
ati efficient varies appreciably as thermocouple
No. 1 is moved along the pipe's radius and
it depends on the Reynolds number. In the
a2 a,b center of the pipe, for low Reynolds num-
y=Smm bers, the correlation coefficient goes more
} rapidly to zero with an increase in the dis-
tance between the two thermocouples, than
Q8 for high Re. Consequently, as the Reynolds
number in the center of the pipe is increased
b the size of the disturbances increases. For
the case when one of the thermocouples is
a4 in contact with the wall curves of the cor-
y= 415,mm relation coefficient change little with an
a increase in Re, while as the thermocouple
is moved. away from the wall, the change
0 Q2 0,4 0.6 d in R1, 2 with changes in the Reynolds num-
0 5 10 15 d mm
ber becomes increasingly more perceptible.
Figure 6. Correlation Coef- This shows that the correlation relationships
ficients for Mercury for Dif- in the region close to the wall do not change
ferent Reynolds Numbers. when Re > 10, 000 which, apparently, is due /63
a) Re = 10, 000; b) Re = to the limiting effect of the wall, while the
= 28, 000; c) Re = 125, 000. center of the flow the character of the cor-
relation coefficient, and consequently also,
the structure of the eddies, changes with:
an increase in Re. Self-similarity ensures
only for Reynolds numbers in excess of
30, 000. The curves of the correlation co-
efficient are sufficiently smooth (without sharp bends), butwith a sharp drop near 6 = 0,
which points to the presence of a continuous wide spectrum of eddies in the flow. For
mercury at high Reynolds numbers we have noticed the existence of regions with
negative correlation coefficients (see Fig. 6). For lower values of Re negative
regions were not observed in the explicit form (with 6 varied from 0.5 to 20 mm),
although the curves do have a tendency to such values.
The fact that velocity fluctuations must have negative correlation coefficients
was proven in [17]. The existence of negative regions of correlation coefficients
61
also for temperature fluctuations can be proven similarly. From the fact that the
flow and the heat transfer take place under steady-state conditions we have
ro ro
(t -^ t') r dr = const = tr dr,
o (5)
consequently, at each time instant
ro
t'rdr=0.
o (6)
Multiplying both parts of the above expression ty, we get
ro
t^ t'r dr = 0,
S
(7)
where t' is the magnitude of the temperature fluctuations in the center of the pipe,
which is independent of radius r. Thus, putting t' under the integral sign and av-
eraging it with respect to time, we get
rp
r dr = 0;
(8)
since tct' = at, vt' R't' t" it follows that

C c
ro
vt^
S
vt R,^,' r dr = 0
o (9)
or
ro
vt ,Ri^,,r dr = 0.
(10) /64
62
The validity of the above expression is
substantiated by experimental data on the
correlation coefficient and by the root-
mean-square value of temperatures [18].
Figure 8 shows Rt, t,vt ,r as a function of
c
the dimensionless radius of the pipe for
Re = 125, 000. The areas of the positive
and negative regions coincide satisfac-
torily with one another. Similarly this
kind. of relationship can also be observed.
2 4 6 8 10 6,nrr for other, lower Reynolds numbers (in
particular, for water). However, for
Figure 7. Correlation Coef- lower Reynolds numbers in order to sat-
ficient for Water with ^ = isfy Eq. (10) very small negative Rt , t,
= 0-0.99, Re = 16, 000. c
are required (less than 0.04), since a
large difference in the fluctuation
intensity is observed in the center
of the pipe and at its wall. Hence
zt^t'
these values cannot always be ob-
dt%'l
served due to measurement errors.
0.4 V
:t' o
0., -- 7.6
Satisfaction of Eq. (10) means
,P that the measurements of a t , and
0.:
g ,. Rt, t , in the experiments were of
Q; ).2 c
sufficient accuracy. The physical
0 7 meaning of the negative correlation
8
coefficients consists in the presence
Rtct - __-- -0.2
of regions with reversed flows, hence
^t r
1^ negative correlation coefficients are
-0.4
a produced by the main pulsating mo-
-0, 6 tion and do not describe the scales
0.8
of the disturbances.
0 0.2 0.4 0.6 1-f

For water with Re = 16, 000 / 65
08
the character of the curves for cor-
Figure 8. Experimental Substantia- relation coefficients differ little over
the pipe's cross section, with the ex-
tion of Eq. (10).
ception of a region near the wall where
a weaker variation in Rrwas observed
than in the center of the flow.
We have not discovered parabolic segments in the correlation coefficient
curves near d = 0. Apparently, our thermocouples are not capable to directly de-
termine the microscale of the temperature disturbances.
63
If the ratio of the macro-
and microscales for the temper-
ature fluctuations has the same
order as that for the velocity
fluctuations [1]
L arô
Ro v ' (11)
then l 0 for our conditions should

be about 0.1 mm. A quantity of
this magnitude can obviously not
be detected by the microthermo-
couples we have used.
0 0.2 A4 96 0.8 U 1.2 7.4 1.6 1-; For all the Reynolds num-
bers we have noted that the cor-
Figure 9. Correlation Coefficients for Dif- relation coefficients are not the
ferent Directions of Motion of same when 6 varies in the posi-
Thermocouple No. 2. tive and negative directions. The
coefficients are symmetrical only
along the pipe's axis. The lack
of symmetry becomes less pro-
nounced with an increase in the
Reynolds number (Fig. 9).
In determining the scales of the disturbances the problem arises of analyz-
ing the experimental data on the basis of correlation coefficients. Strictly speak-
ing, a calculation using the expression /66
LP R (y; y+ 6) d6,
n (12)
is valid [only] for a normal joint distribution of random functions, which describe
the temperature fluctuations in two points of the flow. Our preliminary calcula-
tions have shown that this distribution is close to normal in the center of the flow.
A certain deviation from normal distribution exists near the wall, in particular
of low Re. No recommendations exist on calculating the scale of disturbance for
a non-normal joint distribution. However, it may be assumed that disregarding
higher-order moments which are brought about by a not too large deviation from
a normal joint distribution will have little effect on the scales of disturbances.
On the other hand, the use of Eq. (12) is apparently not entirely logical, since
it does not take into account features produced by the nearness of the wall and by
the correlation curves. In addition, this expression assumes that, within the
limits of the transverse dimensions of the passage and at the wall, R1.2 reaches
a zero value (for 6 0). It also takes into account negative regions of the correlation
64
coefficient, which unjustifiably reduces the scales of disturbances. In another meth-
od for determining the scales of disturbances [5], using the expression
a
Lr^ R ( y -2; y -I 2)d6
0 (13)
it is assumed that the correlation coefficient varies symmetrically to both sides of

the given point of the flow along the radius of the pipe. Allowing for the fact that
R1.2 does not vary symmetrically, we propose another method for obtaining L2,
using the formula
Lr= Lr++Lr-
2
(14)
where
60 60
Lr+= ^R(y; Y+6)d6--dlr-= ``R(y; y-6)dS:

o, b
Integration with respect to 6 0 means that the region with negative correlation coef-
ficients R is disregarded. For comparison we have calculated the microscales
of temperature fluctuations using Eqs. (12) and (14). The region of negative cor-
relation coefficients was not taken into account in these calculations; it was as-
sumed that R1@ 2 at the wall becomes zero, since there is ho turbulent transfer
at this point, although naturally, temperature fluctuations produced by thermal
conductivity do penetrate it. Results of calculations for mercury and water for
Re = 16, 000 and 125, 000 are presented in Fig. 10. It can be seen from the fig-
ure that L near the wall depends appreciably on the selected method of calcula- /67
tions. The first method gives a maximum value of L r, since in this case we do
not use the condition for which R1, 2 at the wall is zero (integration takes place
into the flow). The second method yields the lowest values of L due to the fact
that the effect, of the zero condition at the wall is extended to the center of the flow.
The third method yields intermediate values. In the center of the flow all the three
methods give the same size of temperature disturbances, since the structure of the
flow is close to homogeneous.
Figure 12 also shows values of L obtained for water flowing with Re =
= 16, 000. Data for water have been calculated from Eq. (13). In the center of the
pipe and near the wall the scales of temperature disturbances are the same, butin
65
the region where L has its maximum, for the case of mercury they turned out to
be somewhat higher than for water. It is possible that this is a result of the effect of ther-
mophysical properties of the [different] liquids. The scales of temperature dis-
turbances for mercury as a function of Re, obtained from Eq. (14), are shown in
Fig. 11. The greater the distance from the wall, the more does the scale vary as /68
a function of the Reynolds number, The scale reaches its full size for large Re in
the center of the flow, but not at the wall. In the immediate vicinity of the wall the
scale does not change with an increase in the Reynolds number (for Re > 10, 000).
This is due to the fact that near the wall the scale of turbulent disturbances is
limited by the presence of the latter.
Probability Distribution for Temperature Fluctuation Amplitudes
We have studied the one-dimensional probability distributions for water,

mercury and lead. This study consisted in checking the hypothesis on the normal
distribution of the temperature fluctuations in a given point in the flow, the devi-
ation from which was checked by constructing rectified diagrams [17]. Figure 12
shows the rectified [linearized] diagrams for mercury at Re = 28, 000 for different
distance from the wall. It was not-
ed that the experimental points agree
03 with the straight line describing the
1 2 normal distribution only in the cen-
ter of the flow. The closer to the
oz - ! wall the larger the deviation from
3 the normal distribution; in addition,
i the character of this deviation
changes. Qualitative study of this
0.1
4
deviation has shown that the dis-
Re=16000 tributions under study are some-
what asymmetrical and have an
k:1 0 excess, which is characterized
IN
J
b ythird-and fourth-order mo-
ments, respectively. Deviations
from the normal distribution are
not too large and they are or-
dered. It is known that these
distributions can be smoothed
out according to the Charlier
distribution [19] (a particular
case of which is the normal
distribution), which is express-
ed as
0. 1 0,3 0.5 0.7 0,91 -f
N
Figure 10. Scales of Temperature Dis- at,
turbances. 1) Equation (12); 2) Equation
(13); 3) Equation (14) (Mercury); 4) Equa-
tion (13) (water). [14 6 (t'3-3t')+f 4 3 (t'' - 6t'a+3)] .
66
1
0
42
2 Figure 11. Scales of Tempera-
,8
3
ture Disturbances as a Function
6 of the Reynolds Number. 1) y =
J =1mm;2)y=5mm;3)y=r0=
a^
0 4 8 Rex10'f = 26.1 mm.
9,
3 -2 ^
.e
$
0,3
0,2
0,1

-3
3 1 1 0 2 tl6t
Figure 12. Probability Distribution Density, Linearized

Diagrams and Parameters of Temperature Fluctuation
Distributions for the Flow of Mercury,
Re
Y. mm I 2.8!04 I 1.25.105
0,25
4 1 O
Q f
Then we have determined the parameters of the above distribution and we

have followed their variations along the pipe's cross section. Figure 12 shows
67
variations of parameters A and E over the pipe's cross section for two different
Reynolds numbers for mercury. As the pipe wall is approached, the excess as
well as the asymmetry change. In conclusion it may be stated that near the wall
the probability of the appearance of large positive temperature fluctuations de-
creases, while that of small fluctuations increases. Scatter of points in the di-
rection of cooling (the main body of the flow cooler than the wall) have a distri- /70
bution close to normal (see Fig. 14). The coefficients of excess and of as-
symmetry also change
NOTATION
r0 pipe radius
y distance from wall
6 distance between thermocouples
y position of the maximum temperature fluctuation
max intensity
t' temperature fluctuation, C
t' temperature fluctuation in the center of the pipe, C
Qt, temperature fluctuation intensity, C
vu ' fluctuation intensity of the velocity in the transverse
direction
u+ rate of friction
D dispersion
R1.2 correlation coefficient
R2 spatial correlation coefficient in the radial direction
L microscale of temperature fluctuations over the pipe's
radius
co angular frequency
S0 (w), S1 (w)real and measured densities of the spectral function,
respectively
m(iw) transient function
Tth thermocouple time constant
68
T time
frequency of appearance of fluctuations in the given

n range of values;
0
E
i=1
d probability density of the normal distribution
A coefficient of asymmetry
E coefficient of excess
Dimensionless Quantities
A = 61r distance between thermocouples

t
yt = y u /v distance from the wall
^ = r/r0 radius
m atl.X/g0r0 temperature fluctuation intensity
REFERENCES
1. Khintse, 1. 0.: Turbulentnost' [Turbulence]. Fizmatgiz, Moscow, 1963.

2. Townsend, A. A.: Struktura turbulentnogo potoka s poperechnym sdvigom
[Structure of Turbulent Flow with Shear], Izdatel'stvo inostrannoy litera-
tury [translated from English], 1959.
3. Subbotin, V.I. , et al.: Teplofizika vysokikh temperatur, No. 1 (1964).
4. Velikanov, M.A.: Dinamika ruslovykh potokov [Dynamics of Channel Flows],
1, Moscow, 1954.
5. Minskiy, Ye. M.: Turbulentnost' ruslovogo potoka [Turbulence of Channel /71
Flow]. Gostekhizdat, Moscow, ,1953.
6. Taylor G. J.: Proc. Roy. Soc. A., 135 (1932).
7. Fage A.: Philos. Mag. , 7, 21 (1936).
8. Laufer, J.: NACA, Report-1053 (1951).,
9. Laufer, J.: NACA, Report-1174 (1953)-.
10. Subbotin, V.I. , et al.: Teploenergetika, No. 3 (1962).
11. Venttsel', Ye. S.: Teoriya veroyatnosti [Theory of Probability]. Fizmatgiz,
Moscow, 1962.
12. Pugachev, V. S.: Teoriya sluchaynykh funktsii [Theory of Random Functions].
Fizmatgiz, Moscow, 1957.
69
13. Solodovnikov, V. V.: Osnovy avtomaticheskogo regulirovaniya [Fundamentals
of Automatic Control]. Gostekhizdat, Moscow, 1954.
14. Bobkov, V. P., et al.: Teploenergetika, No. 5 (1965).
15. Opredeliniye parametrov sluchaynykh protsessov [Determination of the Pa-
rameters of Random Processes]. Collection of articles under the editor-
ship of V.I. Chaykovskiy. Gostekhizdat, Kiev, 1962.
16. Harley, N. and Buch, K.: Introduction to the Theory of Probability and to
Mathematical Statistics [translated from English], 1951.
17. Taylor, G. J.: Proc., Roy. Soc. A., 537, (1936).
18. Bobkov, V. P., et al.: Teplofizika vysokikh temperatur, No. 5 (1965).
19. Romanovskiy, L. Ya.: Primeneniye matematicheskoy statistiki v opytnom
dele (Utilization of Mathematical Statistics in the Performance of Experi-
ments]. Gostekhizdat, Moscow, 1947.
70
STATISTICAL CHARACTERISTICS OF TEMPERATURE FLUCTUATIONS IN
HEAT EXCHANGER WALLS ATTENDANT TO HIGH THERMAL FLUXES
M. Kh. Ibragimov, V. I. Merkulov and V. I. Subbotin
Under conditions of steady-state turbulent heat transfer temperature fluc- /71

tuations arise in heat conducting walls [1], [2]. The cause of these fluctuations
is the disorderly, random motion of a part of the fluid (eddies) in the turbulent
flow. In a nonisothermal flow of a fluid, when the temperature has a gradient
over the flow cross section, the motion of an eddy from one layer to another re-
sults in temperature fluctuations. When moving, the more energetic eddies pen-
etrate up to the wall of the duct, producing rapid temperature changes in it. The
majority of the eddies does not reach the wall, it breaks up in the region close to
the wall, producing in it temperature fluctuations which by thermal diffusivity are
transported to the wall of the duct. It is obvious that the maximum temperature
fluctuation at the heat conducting surface of the wall may be equal to Ata. How-
ever, the probability of the fact that eddies from the center part of the flow will
reach the wall is very small, since the integral scales of the temperature dis-
turbances in turbulent flow is smaller than the transverse dimensions of the duct
[3]. As a result of this, the maximum amplitude of temperature fluctuations in
the wall should be lower than the mean temperature difference.
The amplitude of the temperature fluctuations in the layer near the wall /72
and in the wall proper is damped exponentially; here the degree of damping de-
pends on the frequency of fluctuations and on the thermal. diffusivity of the fluid
and the wall
a
A,
A
A
=e x a
(1)
m
where Am and Ax are the amplitudes of the temperature fluctuations at the point
of origin and at a distance x from that point, and n is the frequency of fluctua-
tions, cps. Consequently, the fluctuation intensity and the spectral composition
of the temperature fluctuations will vary depending on the thickness of the wall
and will be determined by the behavior of the fluctuations in the fluid region ad-
joining the wall.
Heat exchange apparatus operate at very high mean temperature differ-
ences. Hence, studies of temperature fluctuations in the wall are performed with
attendant high heat fluxes, since in this case even very high frequency fluctua-
tions have large amplitudes and may penetrate the channel's wall still retaining
a sufficiently high intensity. It can be expected that not only the fluctuation in-
tensity increase with an increase in the thermal flux, but also the contribution
71
of high-frequency fluctuations, which are more dangerous from the point of view
of thermal fatigue stresses.
As a result of temperature fluctuations thermal stresses alternating at
high rates appear in the wall of the channel. Their effect on the fitness of de-
vices at high thermal fluxes is apparently substantial and should be taken into
account in strength calculations. These additional thermal stresses are partic-
ularly dangerous when the heat conducting wall contains microcracks, deep,
slots, etc., which produce stress concentrations [2]. The magnitude of the
thermal stresses is determined by the magnitude of temperature disturbances
in the wall of the channel, i.e., by the values of the spatial correlation coeffi-
cients of temperature fluctuations in two different points of the channel's wall.
This article presents some statistical characteristics (the intensity and
the cross-correlation function) of temperature fluctuations in the wall of a dou-
ble-tube heat exchanger. The same heat transfer agent (liquid metal) was pump-
ed through both heat exchanger passages, at a different temperature.
Experimental Setups and Methods of Measurement
The heat exchanger was made from 1Kh18N9T steel with inner tube diam-
eters of 25 x 2.5 mm and outer tube diameters of 42 x 3 mm. The manner in
which the heat exchanger was connected to the circulation loop, when the heat
transfer agent flow rate in both parts of the heat exchanger was the same, made
it possible to perform the experiments under conditions close to that of constant
thermal flux, passing through the heat conducting wall and directed into the /73
annular space [space between the tubes]. The heat transfer took place over an 1
length 1 h. e = 900 mm [1 h. e /dl = 43, lh. e / 2) = 82].
(Dl - d
Nine chromel-alumel microthermocouples were inserted into the inner
tube wall at a given cross section along the heat exchanger length i. e., at a
distance of 185 mm from the heat transfer agent inlet into the intertubular space;
the diameter of the outer hood of each thermocouple being 0.5 mm. The thermo-
couples were metallized in slots 0., 6 mm deep, milled in the tube wall. The cen-
ters of thermocouple junctions were situated at a depth of 0.35 mm from the out-
er surface of the tube at specified distances along the circumference (Fig. 1).
The heat exchanger was operated on the principle of counterflow (see Fig.
1). In order to reduce the effect of inlet conditions and to improve the mixing
of the liquid metal in the heat exchanger, provisions were made for three mix-
ing chambers, two of which were placed at the points where the liquid. metal en-
ters and. leaves the intertubular space, and one at the point where the metal leaves
the inner tube. Bushings for measuring the average temperatures of the liquid
metals were inserted in the supply and discharge pipes of the heat exchanger.
Temperature expansions in the heat exchanger were compensated for by a dish-
type compensator, made from 1Kh18N9T stainless sheet steel 1.5 mm thick. In
order to provide for a large temperature difference between the liquid metal flow- /74
ing in the inner and outer passages of the heat exchanger, a heating section and
72
s a cooling device were installed in the cir-
culation loop. The liquid metal was heat-
ed. by passing electric current through the
heating section. The heat generated by
the cooling device was absorbed by water.
The flow rate in the loop was measured
by a magnetic flowmeter and was checked
inlet on the basis of the heat balance at the
heating section. For this purpose appro-
priate thermocouples in hoods were plac-
ed at the heating section and in the loop.
Section A-A The temperature fluctuations in the
heat-conducting wall of the heat exchanger
were recorded by the EPP-09 high-speed
automatic potentiometer with an upper
scale limit of 1, 2.5 and 5 millivolts, as
well as by the MPO-2 loop oscillograph.
The temperature fluctuations were re-
o
a
corded on the diagram tape of the EPP-
09 potentiometer with a speed of 6 mm/
/sec; the loop oscillograph recording
was on a 35-mm moving picture film
with a speed of 10 mm/sec using the No.
5 loop. , The statistical characteristics
of the temperature fluctuations were cal-
culated automatically by the 46K1 corre-
lometer using special amplifiers with a
transmission band of 0.18-300 cps [5].
The arrangement used for measuring the
temperature fluctuations in the wall is
shown in Fig. 2.
From the common point of the /75

switch [presumably PMT, DBL] the sig-
nal of all the thermocouples was fed
^f through tumbler switch TS-1 either to
the input of amplifier A2 or to the cir-
cuit of potentiometer EPP-09. The
Figure 1. General View of Test thermocouple was connected directly to
Section and the Location of Mic- the input terminals of amplifier Al, by-
rothermocouples. 1) Mixing passing switch PMT. The amplified
Chamber'; 2) Compensator; thermocouple signal could be directed
3) Outer Jacket; 4) Heat-Con- either to the correlometer or to the
ducting Tube; 5) Tubes for Re- MPO-2 loop oscillograph. The magni-
moving the Thermocouple from tude of the temperature fluctuations was
the Section; 6) Thermocouple monitored by means of cathode oscillo-
hood; 7) Thermocouple Beads graph EO-7. In the experiments, along-
[??] in the Wall of the Heat side with measuring the temperature
Conducting Tube. fluctuations, we have also recorded the
73

constant component of all the thermo-

---- couples, by moving tumbler switch
TS1 into the appropriate position, in
PMT which the common point of switch PMT

iermo- * was connected with the circuits of po-
ouples Dewar's tentl ometers EPP-09 and PP-1. All
Eo-7 flask TS-1 the thermocouple connecting lines
âtl^. were screened, the screen was ground-
oscil, ed at the point where the thermocouples
M
Ts-2 Al a2 left the test section.
- H
The goal of the experiment was to
study the behavior of temperature fluc-
`^'^^ tuation in one point in the wall, as well
as the correlations between the fluctu-
MPO-2 of PP-1 ations two points. To study the be-
09 - havior of the low-frequency components
of the process, the temperature fluc-
Figure 2. Measuring Arrangement. tuations were recorded on the diagram
Al and A2 are Amplifiers; MPO-2 is tape of the EPP-09 during 10-30 min-
a Loop Oscillograph; EPP-09 and PP- utes, while for the high-frequency com-
-1 are Potentiometers; 46K1 is a Cor- ponents of the process they were re-
relometer; TS-1, TS-2 and TS-3 are corded on the moving-picture film of
Tumbler Switches; PMT is a Twenty- the MPO-2 oscillograph during 4-5
Point Switch. min. The correlometer was used to
determine the root-mean-square val-
ues of the temperature fluctuations in
two points of the wall o- t and Q and the correlation coefficients in these points
i
t1t^;
Rtitzc r3r tv r h
(2)
The integration time was 5 min; the fluctuation characteristics in each experi-
ment were measured 3-4 times and their average was used in calculations. To
check the changes in the correlation coefficients for symmetry, we have picked
thermocouple No. 5 as the base thermocouple, since it occupies the middle posi-
tion between the nine thermocouples inserted into the tube wall. The average
thermal flux, referred to the inner surface of the heat-conducting pipe, was cal-
culated from the heat balance equation
CcpyAt
q = F watts/m2.
(3)
The experiments were performed for the following Reynolds number ranges:
in the inner tube Re tube= 41, 900-196, 000; in the intertubular space Re i! s 10,700-
74
49, 800. Under these conditions the thermal flux varied within the limits of
q = 170 . 103 -600 . 103 watts/m.
The difference in the average mixed temperatures of the heat transfer
agents in both heat exchanger passages or the total mean temperature difference
between the two liquid metal flows comprised At = 140-250'. If we take into ac- /76
count the temperature drop in the wall, then the total mean temperature differ-
ence between the fluid and the wall for both heat exchanger passages is
dta dot Atatub + ta i. s. = 70=160 C.
Experimental Results
Temperature fluctuations in the wall of a duct is a complex random process

and the statistical characteristics of this process are determined by many factors.
For example, the root-mean-square value of the fluctuations is a factor of the fol-
lowing quantities
at = f (q, Re, Pr, a N, , (D, RC etc. ) . (4)
Our experiments show that the intensity of temperature fluctuations in the wall
increases linearly with an increase in the thermal flow or of the mean temper-
ature difference with a constant heat transfer coefficient. Figure 3 shows the
curve of a t = f(q). The thermal flux varied within the limits of q = (170-600).
-103 watts/m2.
Since the fluctuation intensity depends on the rate of flow (Reynolds num-
ber), i.e., on the degree of turbulization of the flow and on the magnitude of the
mean temperature difference between the fluid and the wall, one should expect
that function ut = f(Re) will have a maximum. This is due to the fact that no
temperature fluctuations in the wall take place in laminar flow, while in a de-
veloped turbulent flow (as a result of reduction in Ot a ) the fluctuation intensity
will also become lower with an increase in Re. Consequently, at should have a
maximum in some range of Reynolds numbers. This assumption was confirmed
by our experiment (Fig. 4). The dimensionless expression T =Qt'wlq. (Dl-d2)'
characterizing the fluctuation intensity, is plotted off on the ordinate axis of Fig.
4a, while the ratio of the fluctuation intensity to the mean temperature differ- /77
ence in the intertubular space
star = TNu i. s. w
21fi.
a i. S.
75
6, is plotted off on the ordinate
axis of Fig. 4b. The above fluc-
tuation intensity maximum is sit-
uated in the range of Rei, s --
18, 000. This apparently is
valid. only for a given heat ex-
changer geometry; the maxi-
mum for other designs may ap-
pear with different Reynolds
numbers. Knowing the magni-
100 200 300 400 500 q m2 tude of u it is possible to esti-
Figure 3. The Intensity of Temperature mate the value of the maximum
Fluctuations as a Function of the Ther- peak to peak amplitude of tem-
mal Flux. o -Re tube = 80,300 (Flow perature fluctuations. For ex-
ample, for a normal distribu-
With Unstable Rc); e - Re tube= 196, 000. tion of fluctuation amplitudes,
the maximum peak-to-peak am-
plitude can be 56t (the probabil-
ity of this occurrence is 1.5 %).
The simplest frequency characteristic of a random process is the average
frequency
nav= ^N r
(5)
where N is the number of temperature fluctuation extrema and T is the period.

The average frequency in our experiments was within the limits of 1-4 cps. It
should be noted that frequency max of temperature fluctuations which makes
the greatest contribution to the intensity at which the process takes place will
be lower than nav'
15
N
0 6
.y

10
b^
Jr 2
10 10 30 40 Reis s10 3 10 20 30 40 Re.
I. S .
a} b)
Figure 4. The temperature Fluctuation Intensity

as a Function of the Reynolds Number. a) Gener-
alization of *= o-A /q D l - d2 'as a Function
of the Reynolds Number; b) of/A t as a
i, s
Function of the Reynolds Number.
76
In analyzing the values of 6t in the wall one must remember that the ther-
mocouples recorded the superposition of two, independent random processes of
temperature fluctuations from the direction of the intertubular space as well as
from the direction of the inner tube. In this sense the ratio v t/Ata is some-
1.s
what arbitrary. However, the thermocouple junctions are situated closer to the
intertubular space, hence the thermocouples read primarily the temperature fluc-
tuations which penetrate the wall from the direction of the intertubular space,
which can be shown by calculating the damping of fluctuations in the wall, using
Eq. (1). It is known, [3], that in turbulent flow of a fluid in a pipe (d = 52, 2 mm, / 78
Re = 105), the temperature fluctuation of maximum intensity have a frequency
n llan = 0.8 cps. Consequently, if we pick nm . = cps as the fundamental fre-
quency, then we get the following pattern of reduction in the amplitude of fluctu-
ations as a function of increase in the wall thickness
X, mm 0.1 0.5 1.0 1.5 2.0 2:5

A= /Am 0.923 0.681 6.445 0.295 0,197 0.133
Figure 5 shows the correlation coefficients over the angle of the heat con-
ducting wall. The two correlation curves, i.e., Rt(-SÔ) =R t (6 ^ ), are in good
agreement, hence we have constructed for each flow regime a joint curve of cor-
relation coefficients, which drop sufficiently rapidly to zero and even become
negative. The shape of the correlation curve depends on the Reynolds number,
The correlations in the wall reflect the dimensions of the turbulent dis-
turbances in the fluid flow region near the wall. Hence, the behavior of the var-
iation in the correlation coefficient can supply information on the structure of
turbulent disturbances near the wall. A sharp drop in Rt(S ) in the wall points
(P
to the fact that the temperature fluctuations are produced by turbulent eddies of
sufficiently small size. The higher the Reynolds number, the faster does the /79
correlation coefficient reach zero, consequently, the smaller the size of the ed-
dies in the region near the wall. The latter, apparently, have a specific orien-
tation: their rotational motion takes place along the axis of the tube. Hence, af-
ter some distance 8p0 , commensurable with the maximum dimensions of the ed-
dies, the correlation coefficients take on negative values (see Fig. 5). When the
Reynolds number becomes smaller the correlation reduces at a lower rate, which
points to the increase in size of the maximum eddies in the region near the wall.
It is interesting to note that there is a maximum to the size of the largest eddies,
which can be seen from the behavior of the variation in 8(P0 as a function of the
Reynolds number (Fig, 6). Eddies- in the region near the wall are of maximal
size when Re s ^ 18, 000, i, e. , when a t has its maximum. This shows that the
main contribution to the intensity of temperature fluctuations is made by-large-
scale turbulent disturbances.
77
Rti "
cx
0, 8
1 y
Q4
0 ,
0,8
0. 2 5
0

0.
3 6'
O.f
0 0.1 0.2 0.389/text 01 0,2 0.360/text

10 15 20258r^,mm 5 10 15 20 25 A,,mm
Figure 5. The Correlation Coefficients Rt(-6d (),

Rt (S^ )(o) as a Function of the Reynolds Number.
O
Curve Curve
Re tube Re tube
1 196 000 4 80 , 300
S 116,000 5 58'500
3 96,700 6 41;900
A sharp drop in the correlation curve is observed in region of large Retube

(80, 300-196, 000) when 6,0 0, which points to the existence of a wide range of
eddy sizes in the region near the wall (see Fig. 5). For Re tube= 41, 900 and
58,500, the correlation curve has a smoother character as 8(P -- 0, which is due
to the predominance of eddies of the same size in the region near the wall, since
the smallest eddies disappear, while the size of the large eddies is reduced.
By analogy with the turbulence scale for temperature fluctuations in the
wall we can introduce integral scales of temperature disturbances which are cal-
culated from the expression
NO
LsW = R (6,) cd%.
(6)
Figure 6 shows the curve of the scale of temperature disturbances in the /80
wall, referred to the external diameter of the heat conducting wall, which dif-
fers little from the length of the circumference at which the centers of the ther-
mocouple junctions are located. Consequently, the magnitude of scale L5
(P
increases with a reduction in Re. However, we have not observed an explicit
78
maximum in the curve of L 6 as
was the case of 6 , since at the
0
x smallest Reynolds numbers the
size of the largest eddies de-
creases, and. the smallest eddies
disappear. Hence, the integral
scale remains approximately con-
stant.
50 100 150
RetubeX 10' 3 It was noted in [2] that the
state of the heat conducting sur-
Figure 6. The Curve of S^p /L face exerts an appreciable effect
ext on the behavior of the temperature
0
= f(Re) (o) and L6 /L ext/ = f(Re) fluctuations in the wall. In our ex-
perimentwe have obtained addition-
(0), the Scale of Temperature Dis- al information on this problem. The
turbances in the Wall, as a Function behavior of temperature fluctuations
of the Reynolds Number. with Re tube=116,000, if there is no
R on the heat conducting surface is
shown in Fig. 7a. When R is pres-
ent, the statistical characteristics of temperature fluctuations change sharply
(the amplitude and frequency of the fluctuations are reduced appreciably). The
appearence of R is accompanied by an increase in the average wall temperature.
For example, for a thermal flux of 5 . 10 5 watts/m2 the wall temperature increased
as a result of the appearance of R by 100-120 C, while the fluctuation intensity
was reduced by a factor of 3-3.5. 'When R disappears, the temperature state of
the wall returns to the previous level and the fluctuation intensity increased to its / 81
previous magnitude. This effect of R c on the temperature fluctuations in the wall
is recorded on the diagram of the EPP-09 instrument and shown in Fig. 7b. When
R is unstable, the temperature fluctuations brought about by the turbulent disturb-
ance [turbulence] have superimposed upon them additional sharp temperature bursts,
which maybe as high as 70'C when q =590, 000 watts/m 2 (see Fig. 7c). Calcu-
lations made using Eq. (1) show that a sufficiently thin oxide film up to 0.1 mm
thick is capable of reducing the temperature fluctuation amplitude at a frequency
of 1 cps several-fold.
Our experiments, which were performed using a heat exchanger of a given
design with fixed dimensions, are [of necessity] partial in character. The am-
plitude and frequency spectrum of temperature fluctuations in the wall will de-
pend appreciably on the dimensions of the passage in which the heat transfer
agent flows. For the same Reynolds numbers we have: the smaller the dimen-
ions of the passage, the higher the frequency and the smaller the amplitude of
the fluctuations. Hence analysis of additional variable thermal stresses which
arise in the wall due to temperature fluctuations should be performed so as to
take into account the specific features of the given heat exchanger.
79
Figure 7. Record of Temperature Fluctuations Made by
the EPP-09 Instrument for Re tube= 116,000, q =590,000
watts /2 in the Absence of Rc (a), With a Stable Rc(b)
and With an Unstable Rc (c).
NOTATION
e length of heat exchanger

Ih.
d1 , d2 internal and external diameter of the heat-conducting tube
D1 internal diameter of the heat exchanger shell
V, tk temperature fluctuations in points i and k of the heat-conducting
1 tube, C
vt , 9 root-mean-square values of temperature fluctuations in the wall,' C
itk
R, , coefficient of correlation between random processes in points i and
title k of the heat conducting wall
G flow rate of heat transfer agent
At heating of the heat transfer agent in the test section, C
characteristic linear dimension
80
LS integral scale of temperature disturbances in the wall with
co respect to angle co
S distance from the point of intersection of the curve of the cross-

^0 correlation function with the abscissa axis
Text length of the circumference along the external diameter
tube and i. s subscripts of parameters pertaining to the tube or to the inter-
tubular space
R thermal contact resistance
c
REFERENCES
1. Subbotin, V. I., et al.: Teploenergetika, No. 3 (1962);

2. Subbotin, V.I. , et al.: Atomnaya energiya, 18, No. 5, 527 (1965).
3. Subbotin, V.I. , et al.: Teplofizika vysokikh temperatur, No. 1 (1964).
4. Sveshnikov, A. A.: Prikladnyye metody teorii sluchaynykh funktsii [Applied
Methods of the Theory of Random Functions]. Sadpromgiz, Leningrad, 1961.
5. Bobkov, V. P., et al.: Teplofizika vysokikh temperatur, No. 4 (1965),
81
HEAT TRANSFER IN TURBULENT FLOW OF FLUIDS IN SQUARE DUCTS
M. Kh. Ibragimov, Yu. M. Rabotyashnik, G. I. Sabelev,
V. I. Sidirov and V. I. Subbotin
The fluid dynamics in rectangular ducts have been extensively analyzed in /82
several publications [1-3]. The problems of heat transfer in turbulent fluid flow
have not been studied to such an extent. While publications on the flow of fluids
in rectangular ducts describe studies not only of the average, but also of local
characteristics of turbulent flow (distribution of velocities, turbulent stresses),
then in the majority of publications devoted to the study of heat transfer in rec-
tangular ducts present data only on average heat transfer coefficients [4-8].
The boundary conditions in many of these studies were close to conditions
of constant wall temperature along the duct perimeter. To calculate the heat
transferred to the heat transfer agents with Pr >_ 1 it is recommended to use the
same formulas which are used for heat transfer in round pipes, and to use the
hydraulic diameter as the characteristic dimension. However, this method is
not always justified. As was shown in [9-11], the use of the hydraulic diam-
eter does not yield satisfactory results for liquid-metal heat-transfer agents.
For example, for liquid metals flowing in rectangular ducts the heat transfer
coefficients calculated on the basis of the hydraulic diameter become smaller
by almost two-fold, if the duct is heated from one side rather than from two. In
calculating high thermal flux heat exchangers it is necessary to know not only the
average heat transfer coefficients, but also the local distribution of heat fluxes
and temperatures on the heat conducting surface and in the duct walls [11, 12].
Below are presented results obtained in the study of heat transfer in a
square duct in turbulent flow of water and mercury. We have studied the tem-
perature fields in the wall and in the fluid, as well as the statistical character-
istics of turbulent temperature fluctuations.
Methods Used in the Experiments
Experiments with heat transfer in water were performed in an open-cir-

cuit setup. The water flow rate was kept constant by using a 125-liter supply
tank. It was provided by a filter for removing impurities. The rate of flow of
water in the circuit was adjusted by a valve and measured by an RS-7 rotameter.
The experiments with mercury were performed in a closed circuit, with the mer-
cury circulated through it by a centrifugal pump. . The mercury flow rate was ad- /83
justed by a valve and was measured by a Venturi tube used in combination with the
DT-50 differential manometer. Brand R-3 mercury, which was filtered before
being circulated, was used in the experiment.
82
The test section consisted of a square duct, installed in the circuit in the
vertical position, with a cross section of 30 x 30 mm, wall thickness of 6 mm
and 1800 mm long (Fig. 1). The experimental section had at its inlet a chamber
for equalizing the velocity distribution, while at the outlet it was provided by a
mixing chamber. In both chambers were installed thermocouples in sleeves,
used for measuring the temperature of the fluid at the inlet and outlet from the
test section. The outlet chamber was provided by a probe for measuring the
temperature in the fluid flow. The inside surface of the duct was machined to
19 mm
finish class05. The distance between walls 2 and 4 (see Fig. lb) was 30+0.
+0.34 -0.08 '
while the distance between walls 1 and 3 was 30-0.10 mm. Wall 4 was used as
the basic wall for measuring the temperature distribution.
The heat was supplied to the test section by four electric heaters of the
plate type, made from 30mm wide 1Kh18N9T steel, conforming to the width of
the flow cross section of the duct, 0.1 mm thick and 1100 mm long (see Fig. 1b).
Each heater was fastened to a square textolite rod and was pressed to the duct
by clamp screws, fastened in steel columns. A layer of mica about 0.1 mm
thick was placed between the heater and the duct wall. The current and voltage
at the heaters were measured. by class 0.5 astatic instruments. The heaters
could be started up in my combinations. The experiments were performed with
the duct heated from',)ur sides, with a heater corner (walls 1 and 4), with heat-
ing two opposite wall s (walls 2 and 4) and with heating one wall (wall 4).
The temperature distribution in wall 4 was measured at three different
distances from the point where heating started: 1000, 1020 and 1040 mm (which
corresponded to distances from the inlet of 1645, 1665 and 1685 mm) by 27 micro-
thermocouples, 9 of which were built into each of the cross sections (see Fig.1c).
These thermocouples were made from copper and. constantan wires 0.09 and 0.10
mm in diameter, coated by Viniflex insulation and placed over the external side
of wall 4 into slots 0.1 mm wide and 0.1-0.15 mm deep at an angle of 75 to the
duct axis. The hot junction of the thermocouples were welded to the bottom of
the slot in such a manner that the junctions of all the nine thermocouples would
be situated in a plane perpendicular to the duct axis. The wall temperature was
measured differentially, referring it to the temperature of the heat transfer agent
leaving the duct. The thermocouple readings were measured by ar R-2/1 potenti-
ometer with a sensitivity limit of 0.01 millivolts. The temperature distribution
in the wall of the square duct was measured under steady conditions in the zone
of stabilized heat transfer. The distance from the duct inlet to the second cross
section (see Fig. 1) is 1/d h = 56, while from the start of the heated section it is /85
1/dh = 34.
The temperature in the water and mercury flows were measured by a ther-
mocouple probe (Fig. 2). The copper-constantan thermocouple was made from
wires 0.09 and 0.10 mm in diameter with Viniflex insulation, welded together
with the capillary. The 0.25 mm diameter thermocouple which is obtained in
this manner was placed at a distance of 5 mm from the edge of the cone. The
probe was moved along the duct cross section by means of two mutually perpen-
dicular micrometer screws, which made it possible to position the probe at any
point in the duct with a precision to within 0. 02 mm. In order to reduce the effect
83
of free play, the probe in our
P7 experiments was moved in one

direction only, i.e., from the
center of the duct toward the
walls.
The temperature in the
water and mercury flows was
measured under steady con-
ditions in the region with ther-
mal and fluid flow stability..
The moving thermocouple was
connected differentially with
a thermocouple placed at the
outlet from the test section.
To obtain average temper-
atures in each point of the
a flow, the thermocouple read-
ings were continuously re-
corded on the diagram tape
of the EPP-09 potentiometer
with a scale of 0-1 m y dur-
2 ing 40-60 secs. The PPTN-1
low-resistance potentiometer,
which provided the compensat-
ing voltage was installed in
series with the EPP-09. The
readings of the thermocouple
which were obtained for a
given probe position were
average with respect to time,
and these quantities were used
to plot graphs of temperature
distributions in the flow in var-
ious planes along lines perpen-
dicular to the wall. The pri-
mary temperature used for
analysis of experimental data
was the mixed average tem-
Figure 1. Schematic Diagram of the Test Sec- perature of the flow in the
tion. 1) Thermocouple Probe; 2) Outlet Pipe; cross section in which the
3) Square Duct; 4) Textolite Bar; 5) Column; movable thermocouple moved.
6) Electric Heater; 7) Thermocouples Built In- The physical parameters of
to the Wall; 8) Inlet Thermocouple; 9) Inlet the water were those in [131,
Pipe; 10) Inlet Chamber. a) Cross Section of while for mercury the param-
Test Section; b) Cross Section of Square Duct eters of [141-.were used.
(1, 2, 3 and 4 are the Numbers of the Walls);
c) Location of Hot Junctions of Thermocouples When measuring the
at the Outer Surface of the Duct (I, II and III temperature fluctuations the
are Sections in Which the Termocouples Have thermocouple signal was fed
Been Inserted); 11-19; 21-29 and 31-39 are to an amplifier with a trans-
Thermocouples Numbers. mission band from 0. 18 to
84
200 cps and an amplification factor of 10 5 , equipped with a noise silencer (damp-
ing factor of 104). The amplified signal was fed either to the MPO-2 loop oscil-
lograph, or to the 46K1 correlometer [15]. The latter could be used to measure
the root-mean-square value of temperature fluctuations, the averaging time of
which is 5 rains. In addition, the thermocouple readings were recorded by the
loop oscillograph on a motion picture film. The time marks were recorded simul-
taneously on the film. The thermocouple readings were recorded for 60-70 secs
in each experiment.
The Heat Transfer and the Temperature Distribution

in the Wall of a Square Duct
The results obtained by measuring the temperature distribution at the out-

er surface of the wall of a square duct are presented in Figs. 3a and b. As can
be seen, the temperature of the outer surface of a heat conducting wall is not con-
stant along the perimeter. The temperature distribution when the duct is heated
from four sides has the form of a symmetrical curve with two maxima, situated
at a distance of about 0.7 y/a from the center line of the duct wall. The shape of
the distributions we have obtained depends, naturally, on the velocity distribution
in the duct, on the shape, dimensions and thermal conductivity of the duct wall.
It can also be seen from Fig. 3 that, when the duct is heated from four sides,
there are observed on the outer surface of the wall temperature minima in zones /88
directly adjoining the duct's corners. These minima are produced by the fact
that the heaters are narrower than the outer duct surface [being heated]. The fact
that a non-heated, heat conducting section was situated at the end boundary of the
heater reduced the thermal resistance of the corner zone and, consequently, re-
sulted in dropping the temperature of the outer surface of the duct. The temper-
ature distribution on the inner, heat conducting wall of the duct, when heated from
four sides, will not be identical with the distribution shown in Figs. 3a and b. The
temperature maxima at the outer surface for the same amount of heating will be
in the corners of the duct. This was substantiated experimentally by measuring
the temperature fields in the fluid in the same square duct.
We note that the temperature distributions obtained in the three cross sec-
tions (see Fig. 3a) are practically identical with one another. The small differ-
ences between them and the slight asymmetricity can be ascribed to bending or
displacement of the heater's band, as well as to the fact that it was not uniformly
pressed to the duct. The shape of temperature curves on heating from one and
two sides is highly affected by the heat flowing along the side, nonheated walls,
which in these cases act as cooling ribs.
Curves obtained on heating from two and one side, shown in Figs. 3a and b
have a temperature minimum in the corner of the duct. In the given duct the lat-
eral, unheated wall had a relatively high thermal conductivity (9 ^ 10 watts/m-de-
gree). But, as is shown in [12], even if the lateral wall has a thermal conductivity
85
2 3 4 of 0.1 watts/m-degrees, the ther-
mal flux to it is so appreciable, that
the temperature of the heated. wall in
a corner of a duct heated from two
or one side is lower than its tem-
2.1 perature in the central region. The
shape of Curve 2 (see Fig. 3a),
drawn for heating the corner, re-
flects the peculiar features of Curves
1 and 3 and is intermediate between
them Curve 2 shows to what dis-
ZZ
tance along the duct's perimeter
the effect of these heat leakages on
27 the wall temperature extends.
\ Up to now we have considered
9
20 only the shape of the temperature
curves. We now compare these
19 70 curves with the average mean tem-
== 11 perature differences, which canbe
18 12 done to estimate the effect of the
i liquid's thermal conductivity on the
_ 13 heat transfer. For the simplest
17 case, when the fluid flows between
two parallel plates, it was shown
by Seban [10) that the heat transfer
14 coefficient will be affected by the
IF 15 heating method (i. e. , whether the
duct is heated from one or two sides)
Figure 2. Construction of the Thermo- for liquids with Pr << 1; if, how-
couple Probe. 1) Clamping Ring; 2) X- ever, Pr S 1, then the heat trans-
axis Guide; 3) Ball; 4) Packing Gland; fer coefficient is practically inde-
5) Y-axis Guide; 6) Sleeve; 7) Microm- pendent of the heating method. Ex-
eter Screw; 8) Mixing Chamber Housing; periments on heat transfer in slot-
9) Clamping Cone; 10) Thermocouple type and rectangular ducts con-
Well; 11) Mixing Unit; 12) Column; firm the fact that the number of
13) Heater-Clamping Screw; 14) Tex- sides heated has an effect on the
tolite Bar; 15) Tube With Thermocouple; heat transfer coefficient in the case
16) Heater; 17) Current-Carrying Heat- of a liquid metal heat transfer agent
er End Pieces; 18) Adjusting Screws; [11]. This effect of the Prandtl num-
19) Bellows-Type Packing; 20) Hinge; ber on the heat transfer coefficient
21) Air-Tight Flange; 22) Top Flange; is a result of the fact that in liquid
23) Cup. metals the temperature distribution /89
varies smoothly over the duct cross
section, andhence, it depends on the
heating conditions of the second wall.
In ordinary, low-conductivity fluids the entire temperature drop which determines
the heat transfer coefficient is concentrated in the layer near the wall and it is al-
most unaffected by the heating conditions at the other wall. Let us now compare
the results obtained with water and mercury heated from one and two sides (see
Figs. 3a and b). While Curves 3 and 4 (water) are practically identical, then Curve 4
86

.a 21 1 1
0,04 2 2
F ^? 4
-^
w 3
`^ 4
3
`' 0,02 ^ I ' II 4 I/!
-1 0 1 -1 0 1 -1 0 yea
a)
03 - '
1 j
Pe=720
0.2
2
d
43
i Pe=990
b^
o. 1
0.2
0.3
e=1140
1
0.2
2- -^- - /A _T- 13
0.1
-1 0 1 -1 0 y/Q
b)
Figure 3, Temperature Distribution on the Outer Sur-

face of a Wall of a Square Duct When Heat is Transfer-
red to Water (Re = 4910) (a) and to Mercury (b), 1) Heat-
ing From Four Sides; 2) Heating a Corner; 3) Heating
From Two Sides; 4) Heating From One Side, The Di-
mensionless Distance From the Center Line of the Duct
Wall has Been Laid off on the Abscissa Axis, While the
Dimensionless Temperature of the Wall has Been Laid
off on the Ordinate Axis. The Roman Numerals Denote
Numbers of Duct Sections (a). All the Curves of (b)
Pertain to the Second Cross Section
(one sided heating) in experiments with mercury lies noticeably higher than Curve
3, It can be seen from Fig. 3b that the average mean temperature difference for
87
heating from one side is higher than that for heating from two sides. Fig.
3a shows that, in experiments with water, the heating method does not have this
effect on the heat transfer, although in some cases one may observe also a drop
in the temperature of the outer, heated wall, produced by increased heat leakage
due to the increased effective length of the cooling rib.
Comparison of Curves 1 and 2 (see Figs. 3a and b) shows that, for heat trans-
fer to mercury, the total mean temperature difference in the corner of a ductheat-
ed at the corner is higher than for heating from four sides. This increase in the
mean temperature difference for two-sided heating as compared with heating from
four sides can be ascribed to the same effect of the heating method on the heattrans-
fer as is observed in slot-type ducts. In this case heating from four sides can be
regarded as twice heating from two sides. Thus, a reduction in the portion of the
heated perimeter on heat transfer in liquid metals increases the mean temperature
difference, while the presence of heat flux leakages over the unheated, conducting
walls can result in reducing it. In some cases these effects may compensate for
one another and the average heat transfer coefficient may not be dependent on that
portion of the heated perimeter.
On Fig. 3b are drawn dashed lines the ordinate of which corresponds to the
average mean temperature difference over the thickness of the duct wall. The dif-
ference between the ordinates of the dashed line and one of the curves for any val-
ue of y/a yields an arbitrary value 1/Nu in the given point on the perimeter.
Figures 4a and b show curves of Nusseltnumbers as a function of the fluid flow
rate- attendant to heat transfer to water and mercury. The values of Nusselt num-
bers were determined from the expression
2q a
Nu= _
y
C t w -' qb _
^%f
The Nusselt numbers obtained should be regarded as [somewhat] arbitrary. Fig-

ure 4a (line 1-2-3) shows the dependence of Nu on the Reynolds number for non-
metal liquids, which is used for calculating the heat transfer coefficients in round
pipes [5]. Segment 1-2 corresponds to transient flow, while Segment 2-3 corre-
sponds to the formula
Nu= 0.02111e"Pr o.43 (Prf/Prw)1.21
The straight line in Fig. 4b was obtained from a formula used for calculating the /90
heat transfer to liquid metals in round pipes [16]
Nu = 5 + 0.025Pe-8.
88
The spread of experimental points in Figs. 4a and b can be attributed to two fac-
tors: effect of the thermal conductivity of the fluid and of heat flux leakages a-
long the duct's wall on the mean temperature difference for different methods
of heating.
Temperature Fields in the Flow of a Fluid.
Results obtained by measuring the temperature in the flow of water and

mercury are shown in Figs. 5-10. It can be seen from these figures that the
temperature distributions for the case of mercury are smoother than for water.
As is the case with pipes, the main part of the temperature , drop is concentrated

in the region near the wall within the limits of y/a = 0.9-1.0. As the Reynolds /92
number becomes larger the width of this region is reduced (see Fig. 5a). For
mercury the effect of the flow velocity on the temperature distributions at
Re = (16-45) 103 is weak (see Fig. 5b).
The shapes of temperature
au
distributions in different duct cross
3 sections for heat transfer to water
a
0 so ^ are not the same. While for the
temperature distribution with x/a =
40 = 0 the minimum is in the center of
2 the duct, then for the remaining
o(D distributions the temperature min-
Lx mo imum is displaced toward the wall
20 and is situated within the limits of
z 8
y/a = 0.6-0.8 (see Fig. 6a). In
1 heat transfer to mercury the tem-
m perature distributions in the [var-
` 4 6 8 10* 2 Re
ious] duct cross sections have no
a)
inflections and the minima are sit-
Hu
uated at y/a = 0. It is remarkable
that in the flow region near the wall
12 the slopes of the temperature curves
differ appreciably in different cross
o sections. The maximum slopes of
10
OD temperature curves in this region
in the case of water are observed in
8
the cross section x/a = 0. 90, and
600 700 800 900 1000 1100 1200 Pe they become smaller as one moved
b) from this cross section to the cen-
ter of the duct and to the corner re-
Figure 4. Heat Transfer in a Square Duct gion (see Fig. 6a). In the case of
to Water (a) and Mercury (b) for Different mercury the maximum temperature
Heating Conditions. o - Heating From Four curve slopes are observed in the
Sides; o - Heating From Two Sides; m - center zone and become smoothly
- Heating From One Side; - Heating the smaller as one comes closer to the
Corner. duct's corner (see Fig. 6b). The
89
12
P4:: 4Re=E
0/O
10
Re=16600
8
7700 (4500
*^ 6
122100
0 92 0.4 0.6 0.8 y/a 0 0,2 0.4 9e A9 y/a

a) b)
Figure 5. Temperature Distributions in a Square Duct

for x/a = 0 and for Heating From Four Sides for Water
(a) and for Mercury (b); t 1 is the Temperature of the
Fluid When y/a = 0, 99.
tf,deg
20 Cf, de g
36
18
82
12
28
8
24
4
j I I I
0 L 0.2 a4 0.6 0.8 y/a 200 0.2 04 0,6 0,8 y/a

a) b)
Figure 6. Temperature Distributions in Different Flow Cross

Sections for Heating From Four Sides for Water (a) (Re = 4910)
and Mercury (b) (Re = 16, 600),
90
reduction in the slope of the temperature curve at x/a = 0 can be attributed to the
effect of secondary flows, directed from the heat-conducting surface to the cen-
ter of the duct [3]. As was shown experimentally,
the secondary flows in a square duct occupy the
entire duct cross section. The fluid is carried by
-Q2 the secondary flows along a diagonal from the cen-
ter to the corner of the duct, and then the second-
ary flow flows along the wall and then from the
02 -0.29 center region of the wall to the center of the duct.
04 0,1^r
When x/a = 0, the direction of the secondary flow
coincides with the direction of the thermal flux.
0. p 02 Consequently, the secondary flow in this point
0,8 011 should assist in reducing the temperature gra-
0.51 dient. The effect of the secondary flows on the
20 -0 2y1a
1 0.8 0,6 0,4 i-
temperature distribution in the flow of water is
particularly noticeable by the distortion of iso-
Figure 7. Isotherms in therms in Figs. 7 and 8. All figures display the
the Flow of Water in a characteristic inflection of the isotherms in the
Square Duct Heated From direction of the duct's corner.
Four Sides for Re = 7700. As can be seen from Fig. 9, the shape of
isotherms in the flow of mercury differs appre-
ciably from the shape of isotherms obtained in
the flow of water. It is obvious that for such
highly-conducting fluids as mercury, the effect of secondary flows on the shape
of temperature fields is much smaller. When the heat is supplied asymmetri-
cally (see Figs. 8 and 9) the temperature distribution in the fluid flow is affected
by heat leakages along the heat conducting walls of the duct. The effective area
of heat transfer is found to be larger' and is not equal to the area of the heater.
The behavior of isotherms near the unheated wall points to the fact that a part of /93
the heat supplied by the heater flows along the conducting walls and is transferred
from them to the flow of the fluid. The quantity of heat which is transferred
to the fluid by the unheated walls, and the shape of the temperature fields depend
on the heat transfer coefficient, dimensions of the duct, thermal conductivity and
wall thickness. In particular, it follows from Fig. 9 that in the case of heating
from one and two sides the temperature of the fluid in the corners of the duct is
lower than the temperature in the middle part of the wall and, consequently, the
fluid and the wall in the corners of the duct are not overheated.
The high thermal conductivity of mercury, secondary flows and thermal
fluxes through the wall result in the fact that the maximum temperature of the
heat conducting surface is situated not in the corner of the duct, but in the cen-
ter of the duct wall. Such a temperature distribution naturally depends on the
boundary conditions at the outer surface. If the heater is wider than the width
of the flow cross section of the duct and the heat transfer at the outer surface
is nonuniform, for example, high heat transfer density in the corner of the duct, /94
then the temperature maximum of the heat conducting surface may be found to be
in the corner of the duct.
We wish to note the following: in a square duct heated from four sides the
isotherm corresponding to the average mixed temperature of mercury in a given
91

cross section is situated at a distance y/a and x/a 0. 7, which coincides with
the location of a similar isotherm in a pipe ( for liquid metals).
Statistical Characteristics of Turbulent Temperature Fluctuations
Figure 10 shows oscillograms of temperature fluctuations in the flow of

water in a duct heated from four sides with a specific thermal flux of q _ 15, 000
watts/m2 and Re = 4910. To the left of
each oscillogram is shown the tempera-
ture scale. To the right are given the
!0 0.8 0.6 0.4 0.2 0 0.2 -0 -0.6 0.8 y/o
coordinates of the probe and its location
08 is noted by a dot. It can be seen from the
oscillogram that the characteristics of the
06 fluctuation curve differ from one another
0.4 in the different points of the duct's cross
02
-0.28 section. First, while in the central zone
of the duct one observes high-frequency
0 0 J a) temperature fluctuations with a relative
02 moderate amplitude (-1 C) (see Fig. 10a), /9_5
o^ 0.37 then in the region adjoining to the ductwall
05 0.52 0 i -0.77 (x/a = 0, y/a = 0. 98), the temperature fluc-
0
8 0.2 tuations are as high as 10 C and their fre-
3.0 ô^ I -022 quency is lower. In the very corner of the
1.O - duct the fluctuation frequency is even lower
Os
-0.17 ( see Fig., 10c). Secondly, the oscillogram
0 Q (see Fig. 10c) does not exhibit high-fre-
- '"s3 quency components of the spectrum. This
08 OS 0.1, 0.2 0 -0.2 y/a shows that only large - scale fluctuations
b) are capable of entering the corner region.
Figure 8, Isotherms in the Flow And, finally, the oscillograms in the cor-
of Water in a Square Duct. a) Heat- ner of the duct are not symmetrical with
ingthe Corner, Re = 4910; b) Heat- respect to the time - average temperature
ing From Two Sides ( the Left Wall in the given point of the flow, which can
is Heated), Re = 4910, be seen from Fig. 10 and from other os-
cillograms which were obtained with high-
er water flow rates. The same asymmetry
has been observed also in experiments with
mercury. It consists in the fact that the
temperature curve has peaks directed primarily in one direction away from the
average temperature. For the region near the wall they are directed in the di-
rection of smaller temperatures, while for the center region they are directed
in the direction of higher temperatures. Apparently these peaks are produced
when the center of a turbulent disturbances passes through the thermocouple. If
the thermocouple is situated in a region adjoining the wall, then the peaks are
produced by cold portions of the fluid which arrives from the center of the duct.
In the central region of the flow the temperature peaks are brought about by the /96
passing of hot portions of the fluid, carried from the wall to the center of the flow
through the thermocouple junction.
92
A more complete picture
of the distribution of tempera-
ture fluctuations in different
cross sections of the square
duct was obtained using the cor-
relometer. The results obtain-
ed by measuring the root-mean-
square values of temperature
fluctuations in water flowingin
0 22 0.4 0.6 u y/a 0 0.2 0.4 0 F 0.8 y/a
a square duct are shown in Figs.
a) b) 11-13a. Figure 11 shows the
Figure 9. Isotherms in the Flow of Mercury curve of a- t , the root-mean-
in a Square Duct. a) Heating From Two Sides, square values of temperature
Re = 28, 400; b) Heating From One Side; fluctuations along lines paral-
Re = 28, 200. lel to axis Y. The parameter
of these curves is the distance
from the center of the duct a-
long axis X.
The distribution of o- t over the cross section of a square duct is shown
more graphically in Fig. 12. The lines of this figure correspond to the locus
of points with a constant root-mean-square temperature fluctuation of the flow.
The parameter of curves in Fig. 12 is o t . As can be seen from Figs. 11 and
12, the distribution of 6t over the duct cross section is quite complex. In the
center part and at the walls at has a minimum; the maxima are situated at the
axial lines of the square at a distance of ---14.5mm (y/a = 0.97) from the cen-
ter of the duct. At the wall o t increases very sharply. As x/a becomes larger
(see Fig. 11) the maxima are reduced smoothly and move away from the wall of
the duct. They move farthest away in the corner region. In the center region
of the flow, when x/a and y/a :5 0.5, the magnitudes of a t are practically con-
stant. In regions with 0.5 < x/a < 0.9 and 0.5 < y/a < 0.9 the curves of constant
values of o t have a peculiar shape: at the axial lines these curves are displaced
toward the center of the duct, while in the corner region they are extended along
the bisectrix into the corner of the duct.
The shape of curves v t = const attests to the presence of secondary flows
in the square duct. Let us examine the distribution of a- t along the line y/a =
= 0.8. The lines of u t = const for x/a = 0.6 are displaced to the center of the
duct. The location of these displacements is the same as the maximum value
of the normal velocity component of the secondary flow [3]. Lines v t = 0.7 and
0.9 for y/a = 0.3 conversely, are displaced toward the wall of the duct, The
region of minimal 6t for x/a -_ 0.8 is identical with the location of the eddy of
the secondary flow, where the normal components of the secondary flow veloc- / 97 .
ity are at minimum. We thus notice a correspondence between the magnitudes
of normal components of secondary flows and the magnitude of o- t'
93
N
o
IM
c
Figure 10. Oscillograms of Temperature Fluctuations

in Water Flowing in a Square Duct, a) x = 0, y = 0;
b) x = 0, y = 14, 75 mm; c) x = 13 mm; y = 14, 4 mm;
Time Scale is 1 sec.
6t C
x/Q=0.975
0,940
2,l
1,3
11
0,835
0,630 Figure 11. Curves of o' t the

0,313 Root-Mean-Square Values of
0 Temperature Fluctuations in
Different Cross Section of a
0 0,2 0.4 0.6 O.S y/a Flow of Water, R=4910.
94
6t
0 0.2 Q^ 0,6 0.8 y/Q
Figure 12. Distribution of

Curves of Constanto-t in the
Cross Section of a Duct in
the Flow of Water,
Re = 4910.
8 0.2 0.4 0.6 0.8 y/a
a)
Figure 13 shows the change in 6t, G ge =77000

vt (determined at the same thermal
9
flux) along the center line of the duct 44600
(x/a = 0) for different flow rates of ^S
300001
water and mercury in a square duct.
As Re is increased from 4910 to
22 000 the central re 'on in which
ot is practically constant, extends 0 0.2 0, 4 0.6 0.8 y/a
b)
from y/a = 0.5 to y/a = 0.9 and the Figure 13. Distributions of 6t in a
magnitude of o-t becomes increas-
ingly uniform over the duct's cross- Square Duct Heated From Four Sides
/98 section. The shape of the curves With a Flow of Water (a) and Mercury
is no longer self-similar. The ab- (b) for Different Reynolds Numbers.
sence of a peak in the u curve in the
region near the wall at Re = 22,100 is produced by the relatively large dimen-
sions of the pickup. The maximum of u is located near the laminar sublayer,
the thickness of which at the above Reynolds number is about 0. lmm. Hence
a thermocouple with an outer diameter of 0.25 mm is situated outside the region
of maximum values of o-t and, in addition, may appreciabley distort these read-
ings.
We wish to note that Figs. 11-13 display only the amplitude characteristics
of the fluctuations. The frequency characteristics of the fluctuations, as was
already shown, differ in different regions in the duct. Judging by the oscillo-
grams, the closer the point to the wall and to the corner of the duct, the great-
er the contribution of low-frequency components. The average frequency of
fluctuations, determined from the oscillograms of Fig. 10 as half of the number
95
=0,18
"'t 1
3
S
S
0,
0 0,2 0.4 0..6 0.8 y/a
Figure 15. Distribution of Lines

0 0,4 0.8 y/a at = const in the Cross Section of
Figure 14. Distributions of the Root- a Square Duct for Mercury,
Mean-Square Value of at, the Temper- Re =30,400.
ature Fluctuation, in Different Cross
Sections in a Flow of Mercury,
Re = 30, 400. of extrema of the temperature
curve per second, was found to
be 6. 7, 2.8 and 0.9 cps for the
central region, region near the
wall and the duct's corner, respectively. The fluctuation characteristics for
mercury were determined using the same equipment as for water.
Figure 14 shows values of a t along lines parallel to axis Y. Figure 15
shows lines of constant at, constructed on the basis of curves of Fig. 14. It
follows from these figures that the root-mean-square values of temperature
fluctuations for mercury are distributed smoother than in experiments with wa-
ter. The maximum of at is also located on the axial lines of the cross section,
but it is displaced appreciably from the wall of the duct to the center, i. e., /99
y for amax for Re = 30, 400 is 9 mm (y/a = 0.6). This change in the distribution
of at with an increase in the thermal conductivity of the liquid [when water is re-
placed by mercury, DBLI agrees with a hypothesis stating that the amplitude of
the turbulent temperature fluctuations is proportional to the product of the mix-
ing path and the gradient of the average temperature [171. In liquids with high
thermal conductivities the temperature gradient exists in the entire interval
from the wall to the center of the duct, while when heat is transferred to water
the main temperature drop is observed near the wall. Hence, the magnitude of
a tmax , which is obtained in experiments with water is located at the wall of the
duct, while that obtained with mercury is displaced to the center of the duct. The
shape of curves of constant at obtained in experiments with mercury differs from
similar curves obtained with water (see Fig. 15). In the mercury experiments
the effect of secondary flows on the at field is also smaller, i.e., curves of at =
= 0. 45, 0. 36, etc., are extended in the direction of the duct's corner. This again
96

fl

r, se cs 7T_ SP.c

a) b)
Figure 16. Normed Autocorrelation Function in the Center

of a Turbulent Flow of Water in a Square Duct Heated From
Four Sides; Re = 4910; a) x/a = 0, y/a = 0; b) x/a = 0;
y/a = 0.8.
confirms the fact that in liquid metals with high thermal conductivity the effect
of secondary flows on the heat transfer becomes lower. The effect of the aver-
age velocity within the given limits of variation An Re does not have a substan-
tial effect on the magnitude of u (see Fig. 13b). Comparing Figs. 11 and 13 we
note that the magnitudes of o-t in the center of the duct in experiments with water
and mercury are close to one another by absolute magnitude. The magnitude of
9t was measured with a specific thermal flux of q = 15, 000 watts/m 2 provided by
heaters on heating from four sides. When the Reynolds number is increased the
magnitude of at over a large part of the duct cross section in the flow of mercury
becomes greater than for water
The magnitude of u and the average frequency of fluctuations do not fully /100
characterize the temperature fluctuations. An important characteristic is the
probability that the fluid in the given point of the flow will have a given temper-
ature. We have used the oscillograms to determined the distribution of the prob-
ability of temperature fluctuation amplitude, which is in good agreement with a
normal distribution of random variables. Since the points examined are situated
in different regions of the duct's cross section (for example, at the wall and at
the corner of the duct) it may be assumed that in the first approximation the prob-
ability distribution of amplitudes of the temperature fluctuations conforms to the
normal distribution for a large part of the cross section of the square duct.
Figures 16a and b show the normed autocorrelation function (correlation
coefficient) in two points of a flow of water, calculated from the expression
t , (0) t (r)
R^ =
s, 2
97
In the main body of the flow (see Fig. 16) the correlation coefficient decreases
rapidly to zero; on further increase in T, the time shift, function R(T) fluctuates
about zero and is gradually damped out. The life of the temperature disturb-
ance, determined by the first intersection between the curve of R and the ab-
scissa axis is 0.35 and 0.55 secs. The life of a turbulent disturbances can be
used to determine its dimensions. The characteristic dimension of a temper-
ature disturbance in the axial direction is L 1/a = T u/a = 6 (when Re = 4910),
i, e., it is appreciably greater than the cross sectional dimensions of the duct.
Results
Our experimental work shows that for accurate calculation of temperature /101
field in rectangular ducts it is insufficient to have relationships used for calcu-
lating heat transfer in round pipes. This is true for liquid metal, as well as for
ordinary, low-thermal-conductivity, heat transfer agents.
The fact that secondary flows have an appreciable effect on the heat trans-
fer process in a square duct was proven experimentally. This was demonstrated
by the characteristic distribution of the wall temperature, by the specific shape
of the mean temperature distributions and of the temperature fluctuation inten-
sity in the fluid flow.
The existing semi-empirical theory of heat transfer do not at all take into
account the contribution of secondary flows, hence, these theories should be cor-
rected accordingly.
The heat transfer in a square duct, in addition to secondary flows, is also
affected by heat fluxes through the wall, which depend on the duct geometry, thick-
ness and thermal conductivity of the wall, which is a characteristic feature of all
nonround channels. Hence, for accurate calculations of nonround ducts from the
point of view of fluid flow as well as from the point of view of heat transfer one
must examine each specific geometry and solve the problem on taking into ac-
count all the factors which affect the temperature field.
NOTATION
tf 1 temperature of fluid at the duct inlet, C
twl temperature at the outer surface of the wall, C
w2 temperature at the inner surface of the wall,' C

G flow rate of heat-transfer agent
98
Qt root-mean-square value of temperature fluctuations, C
x, y current coordinates (the coordinate origin is in the center of
the duct)
a half-width of the flow cross section
b wall thickness
d hydraulic diameter of the duct
REFERENCES
1. Nikuradze J.: Ingenieur-Archiv, 1, 306 (1930).

2. Laufer J.: NACA T N-2133, 1950.
3., Brundrette E. and Baines, W. D.: J of Fluid Mech. , 19, 3 (1964).
4. Bailey, A. and Cope, W. F.: Aeronaut. Res. Comm, (Great Britian),
Techn. Rept. , 43, 199 (1933).
5. Mikheyev, M. A.: Osnoy teploperedachi [Fundamentals of Heat Transfer].
Gosenergoizdat, Moscow, 1949.
6. Kays, W. M.: Techn. Rept. 14 Dept. Mech. Engng. , Stranford University,
Calif., (1959).
7. Tsarenko, N. V.: Izvestiya Kievskogo politekhnecheskogo instituta, 17,
143 (1956),
8. Lubarsky, B. and Kaufman, S.: NACA, Report., No. 1270 (1956),
9. Novikov, I. I. and Voskresenskiy, K. D.: Prikladnaya termodinamika i
teploperedacha [Applied Thermodynamics and Heat Transfer]. Gosatomi-
zdat, Moscow, 1961.
10. Seban, R. A.: Trans. ASME, 72, 789 (1950).
11, Subbotin, V.I. , et al.: Teploenergetika, No. 9 (1961).
12. Novotny, I. L. , et al.: E. R, I. Eckert A. I. Ch. E. Journal, v. 10, No, 4 /102
p. 466-470 (1964).
13. Teplofizicheskiye svoystva veshchesty [Thermophysical Properties of Sub-
stances]. Handbook, under the editorship of N. B. Vargaftik. Gosenergoi-
zdat, Moscow, 1956.
14. Nikol I skiy, N. A. , et al.: In the collection "Voprosy teploobmena" [Prob-
lems of Heat Transfer]. Published by the Academy of Sciences of the
USSR, Moscow, 1959.
15. Bokov, V. P., et al.: Teplofizika vysokikh temperatur, 3, No. 5 (1965).
16. Subbotin, V. I., et al.: Teplofizika vysokikh temperatur, 3, (1965).
17. Subbotin, V.I. , et al, Teploenergetika, No. 3 (1962).
99
FLUID DYNAMICS IN AN ANNULAR SPACE WITH LONGITUDINAL RIBS
Yu. D. Levchenko, V. I. Subbotin, P. A. Ushakov and A. V. Sheynina
To reduce the quantity of the coolant in the active zones of reactors, the /102
coolant annuli should be narrow. However, in this case the production tolerances
for the geometric dimensions become comparable with the dimensions of the cool-
ant annulus proper. This can result in appreciable eccentricities, which cause
redistribution of coolant velocities, and consequently, also ,cause the appearance
of temperature variations along the perimeter of fuel cells. The present paper
presents the study of a narrow (d l/d2 = 0.895) coolant annulus with three longi-
tudinal spacing ribs (Fig. 1). It considers a concentric duct and ducts with rel-
ative eccentricities s = 0.5 and 1 (the limiting case).
The test section was made of a pipe (d 2 = 95mm) made from 1Khl8N9T
steel with a dural rod (d l = 85 mm) placed inside it. The pipe and the rod formed
an annular slot 1900 mm long and with dl/
/d2 = 0.895, S = d2 -d l/2 = 5 mm. Brass
spacing ribs were inserted into longitu-
dinal slots, milled on the surface of the
rod. In the experiments we have used
2 three rods, one of which had identical
ribs, which were used to keep it concen-
4 trical with the outer tube. The dimen-
sions of the ribs of the two remaining /103
rods were selected so as to provide for
Z=0
^V . e>0 eccentricities of 0.55 and S. We have
examined the displacement of the rod in
Figure 1. Cross Section of the direction of one of the ribs only. At
Test Section. 1) Large Rib; the inlet to the test section was placed
2) Wide Annulus Segment; a mixing header with a mesh for equaliz-
3) Narrow Annulus Segment; ing the velocity distribution. Along the
4) Small Rib. surfaces of the central replaceable rods
were milled slots into which six capil-
laries 2 mm in diameter from 1Kh18N9T
steel were inserted and fastened by the
quick-setting "Stynscryl" plastic, after
which the surface was carefully smoothed out. Holes in the capillaries (0.5 mm)
for measuring the static pressure were provided in cross sections situated at
distances of 500 and 1000 mm from the inlet. Static pressure pickups have also
been provided at the outer tube at a distance of 150-200 from one another over the
length of the tube.
100
Moving the inner rod [off-center] (see Fig. 1) forms three cells, two of
which have the same geometry ("narrow" cells), and one has a different cross
section ("wide" cell) The basic parameters of the test section are given in
Table 1.
TABLE 1. Basic Parameters of Test Section

a=0,5 a=1
Designation U
E =
1 I 2 1 2
How cross section 459 346 648 254 878

of cell, mm2
Hydraulic diameter -9.20 7.30 13.0 5.40 17.0
of cell dh, mm
NOTE: The numbers 1 and 2 refer to the narrow and wide
cells, respectively.
The construction of the test stand has been described in [1]. The pressure
drops were measured by U-shaped glass differential manometers using distilled
water. The temperature of the water was varied from 6 to 80C in order to ex-
tend the range of Reynolds numbers, Depending on the pressure drops the dif-
ferential manometers were filled with mercury, tetrabromoethane, carbon tet-
rachloride and chlorobenzene. The rate of flow of the water through the test
section was measured by calibrated venturi tubes. Moderate (less than 2 m3/
/hour) water flow rates were determined by measuring tanks. The water tem-
perature was monitored by copper resistance thermometers and the temperature
of the surrounding air was measured by mercury thermometers.
The friction resistance coefficients X were determined on the basis of the
measured pressure drops for the concentric spaces. For eccentric spaces this
was not possible, since in this case the flow rate distribution among the cells is
unknown. The experiments have shown that for a concentric space the pressure /104
T
drops in the cells do not differ by more than 5%. To reduce the time needed for
measurements, the measurements for small water flow rates were made in one
cell only.
The velocity distributions were measured in cells of eccentric annular spaces.
The experiments were performed on an air stand, the construction of which is de-
scribed in [2]. The local velocities were measured by a glass Pitot tube (with an
external diameter of 0.2 mm and an internal diameter of 0.12.mm) in the duct out-
let cross section. The static pressure was measured through A hole 0.8 mm in
diameter in the wall of the outer tube. The Pitot tubes were positioned and moved
in two mutually perpendicular directions by a coordinate spacer with a sensitivity
of 0, 05 mm. The pressure drops were measured by a U-shaped differential
101
manometer, filled with alcohol. The air flow rate through the model was kept
constant by monitoring it by the pressure drop over a 600 mm long section.
The theoretical limiting error in measuring X, the friction coefficients in
water was 5-10%, which includes the errors in the geometric dimensions, water
flow rates and pressure drops. The accuracy in measuring local velocities in
the air was 1.5-2%, while that of measuring the friction coefficients X was 7%.
The velocity distributions in the cells of eccentric annular spaces were also de-
termined analytically, using the method due to Deissler and Taylor [3]. The or-
iginal data used in the calculations were: working medium - water at 18C, pres-
sure gradient dp/dl = 238 kg/m2 -m).
The measured friction coefficients in the concentric space are shown in
Fig. 2. The experimental data for laminar flow are in good agreement with the /105
theoretical formula obtained for annular passages without ribs [4]-[6]
`di/ R ' ( 1)
where Re = ud h/v and d is the equivalent hydraulic diameter of the passage,
taking into account the ribs;
z
( d 1) In di
In d1
` \ di / z+`11 L \ di )z-1
For a round pipe T = 1, while for a flat slot T = 1.5. For our case
Calculations using Eq. (1), derived for a round pipe, naturally yields values of
X on the lower side. Equation (1) is in good agreement with experimental data
obtained in [4] for d l/d2 = 0, 212-0.838.
Transition from laminar to turbulent flow is observed at Re = 2300-2400.

The experimental data for turbulent flow are in agreement with the well-known
formulas for round pipes due to Blasius (2) and Nikuradse (3)
1= 0.316Re - O - z5 ; ( 2)
I = 0.0032 + 0.221Re -0237. (3)
102
A-10
A,.
z
2
\
1 b,^
^
10 N
a^
\om
' _ 4
3
^-
1 L
10 2 103 10`^ 105 Re
Figure 2. Friction Coefficients for a Concentric Annu-

lar Passage With Longitudinal Ribs. 1) and 2) Calculated
From Eq. (1) for w = 1 and 1.5 Respectively; 3) and 4)
Calculated Using Eqs. (2) and (3) Respectively; the
Circles Denote Experimental Results.
Consequently, if we take the hydraulic diameter as the characteristic dimension,

then the hydraulic resistance in the thin concentric annular space with three lon-
gitudinal ribs can be calculated by using formulas for round tubes in turbulent flow
and by formulas for ribless annular passages in laminar flow.
The characteristic velocity distributions and the isotachs [equal-velocity
lines] obtained experimentally with an eccentricity of a = 1 are shown in Fig. 3.
These data were used for constructing the dimensionless velocity distributions
Cu)-t \ymax^,
where u is the local velocity, u is the average velocity for the cell under study,
y is a coordinate measured from the surface of the rod along a normal and y
is the width of the annulus for a given angle co. max
Comparing the measured distributions of velocity u/u with those calculated

by the Deissler-Taylor method for two eccentricities show that they agree in the
majority of cases. The slight disagreement which is observed near the rib (co =
= 6 0 ) is due to the fact that the calculations for regions near the ribs are not too
accurate.
For each angle cp we found the average velocity /108

b max Ymax
Su
(R I + y) dy u (R I + y) dy
u u
u
U`c \ u )W
= ___p
bmax
_ 0
ymax
4
()
ymaxCRi- 2 )
(R I + y ) dy
0
103
0
9 u, mise c
D) 0j, 45'
SS.
Figure 3. Distribution of Velocities u/u in the Wide (a) and Narrow (b) Cells for e = 1.
The curve of Ucp as a function of co is shown in Fig. 4, from which it follows that
the average velocity in the direction of the normal in the wide cell depends little
on the eccentricity, while
in the region of 20 <cp < 100
it practically does not vary
along the perimeter and is
approximately equal to the
average velocity in the cell.
The picture observed in nar-
't v .y row cells is different. When
^îi^
A P, = 0.5 Ucp varies from 1.2
^ 1
for co = 150 . to 0.7 for co
Ai
= 110, i. e. , by approxi-
mately a function of 1.7.
Velocity Ucp varies even
^ I
more rapidly in the case of
P, = 1. In this case the ex-
0 20 40 M 80 perimental data could be
100 9*
obtained only for a part of
Figure 4. Distribution of Relative Velocity, U ^ the narrow cell (cp <_ 65

over angle co A. and are Experimental Data for since the comparatively
pa sll dimensions of the
the Wide and Narrow Cells, Respectively, when passage made it impos-
e = 0.5; o and Denote the Same for e = 1; the
Dashed Line Shows Theoretical Results. men to perform on of -
mets in the region of min-
imal spaces. Analytical
calculations were also made only for co c 850 . The values of Ucp obtained
theoretically and experimentally are close to one another (see Fig. 4).
The average velocities in the cells found by graphical integration and the
measured pressured gradiants dp/dl were used to calculate the Reynolds num-
bers and the friction coefficients X (Table 2). In was assumed in these calcula-
tions that the pressure gradients in all the cells are .the same (238 kg/m 2 -m). / 109
However, it was established experimentally that the pressure gradients in the
cells differ (see Table 2), which apparently is due to the effect of local resist-
ances to flow at the cell inlets and outlets. This phenomenon is one of the causes
for the differences between the calculated and experimentally obtained values of
u2/u1 , shown in Table 2.
If we make a correction to take into,account the differences in dp/dl, then

the theoretical values of u2/u1 will be: 1.37 for e = 0.5 and 1, 73 for e = 1, Then
the maximum difference between the theoretical and experimentally obtained re-
sults will not exceed 10% even for a =1. . Figure 5 compares the values of X (see /110
Table 2). Lines were drawn through the experimental and theoretical points on
the assumption that for all the cells the friction coefficients are proportional to
Re-0.25. When e = 0.5 the theoretical and experimental values for the wide cell
are in agreement with one another and conform to Eq. (2) for a round pipe, while
105
TABLE 2. Experimental and Theoretical Data
e = 0.5 8= 1.0
Parameters Experi. I Theoret. Experi, f Theoret.
t 2 t l 2 1 I 2 1 I 2
tw, C - I - I 18 I 18 I - I - I 18 18
Lair, 0 C I 32 I 31 I ( I 31 I 31
I
U, m/sec 41.1 ( 55.6 11.03 11.51 135.8 56.61 1.01 11.8.3
f
Pre 3sure gradient, dp/dl, 343 308 238 238 255 236 238 1213
kg/1 n2, m I
I
He 10-3 118.6 144.9 7 .2 ( 18.7 ( 12.1 ( 60 5.12 1 29.5

i
X-103 12.52 12.19 13.27 I 2.7 ( 1.82 i 2. l2 i 2.47 2.38
1
R., do of the average ve- 1.35 1.46 1.58 i.81

lot ,ities in the wide and
na rrow cells, u2/u1
Rat io of the fluid flow 2.52 2.73 5.46 6.26

rat ,s in the cells, Q2/Q1
NOTE: 1 and 2 denote the narrow and wide cells, respectively.
2.10
1F.
1
\
1_
12
r o 0 10= z it 6 8 Ke
Figure 5. Friction Coefficients X for Cells of Eccentric

Annular Spaces.
Calculated Experiment
e=0.5 p p,
a = 1.0 p i O
106
in the narrow cell they are situated by 6% lower. For an eccentricity of E = 1
the data for the wide cell are also in agreement with Eq. (2). At the same time
the data for the narrow cell are by approximately 40% lower than values of ob-
tained by Eq. (2).
Thus, our experiments and calculations have shown that the approximate
method due to Deissler and Taylor [3] can be used for practical calculations of
distributions of velocity and of frictional resistance coefficients of narrow ec-
centric passages.
REFERENCES
1. Sheynina, A. V.: Gidravlicheskoye soprotivleniye puchkov sterzhney v

osevom potoke zhidkosti [Resistance to Flow of Clusters of Rods in an
Axial Flow]. Present collection.
2. Ibragimov, M. Eh. and Isupov, I. A.: Raschet i eksperimental'noye
issledovaniye poley skorosti v kanale slozhnoy geometrii [Calculation and
Theoretical Study of Velocity Fields in an Intricately Shaped Passage].
Present collection.
3. Deisler, R. G. and Taylor, M. F.: Reactor Heat Transfer Conference of
1956, TID-7529/Pt 1/, Book 2, 416, 1957.
4. Koch, R. and Feind, K.: Chem._ Ingr. Techn. , 30, No. 9, 577 (1958).
5. Walker, Y. E., et al.: A. I. Ch. E. Journal, vol. 3, No. 4, 484 (1957).
6. Idel'chik, I. Ye.: Spravochnik po gidravlicheskim soprotivleniyam [Hand-
book on Resistances to Fluid Flows] p. 58. Gosenergoizdat, Moscow-
Leningrad, 1960.
107
EFFECT OF ECCENTRICITY ON HEAT TRANSFER TO LIQUID
METALS IN ANNULAR SPACES
V. I. Subbotin, V. D. Talanov and P. A. Ushakov
The coolant annulus which is formed by a fuel-element rod and the tube in /111
which it is placed is one of the versions of ducts used in reactors with liquid met-
al cooling. When the fuel element is displaced (an eccentricity is formed) the
symmetry of the flow past the fuel rod is disturbed, producing temperature var-
iations along the rod's perimeter, local overheating, additional thermal stresses,
etc.
The use of spacing ribs only partially alleviates the above shortcomings.
Perceptible eccentricities may appear due to geometric tolerances of the pas-
sage even when these ribs are used. In addition, the presence of the ribs pro-
duces additional temperature bursts beneath the ribs and alongside them, simi-
lar to those already observed [1]. The eccentricities have a particularly appre-
ciable effect on thin coolant annuli, in which the geometric tolerances are com-
mensurable with the dimensions of the annulus proper.
The present paper describes the effect of spacing ribs and eccentricities
on the temperature field of the surface of the heat-conducting tube which is cooled
by liquid metal flowing in the narrow annulus (D/d 2 = 1.12). The experiments
were performed with the Na-K alloy (22% Na and 787o K) using a setup consisting
of a closed circuit made from 1Kh18N9T steel with a volute electromagnetic pump
[2], providing an alloy flow rate of about 10 m 3/hour at a pressure of about 4 kg/
2
/cm . The heat was carried away from the stand in a double--tube kind of heat
exchanger, through a double wall with an intermediate contact layer made of lead.
A small part of the coolant was continuously circulated through a cold oxide trap,
which was cooled through a double wall. The oxides content in the circuit was
monitored by a distillation sampler, the sample collected which was subsequently
analyzed.
The test duct (Fig. 1) has been made so as to make it possible to rotate the
center tube through any specified angle directly during the measurements. The
annulus was formed by a jacket with internal diameter D = 76 mm, wall thickness
of 3 mm and an internal tube with d 1 = 2R1 = 64mm and d2 = 2R2 = 68 mm. Thus
the width of the coolant annulus was 4 mm; the electric heater of the inner tube
was made from 3 mm nichrome wire in the form of a spiral. To prevent short-
circuiting the coils, the spiral was first oxidized by heating it in air. In addition,
the nichrome wire was coated by a thin layer of an alundum-orthophosphoric oxide /112
mixture. The spital is insulated from the internal surface of the tube by an or-
ganic silicon compouxid 0.6 mm thick. One end of the heater is attached to the
108
housing, while the other has a
potential over it. The power
which was supplied to the heat-
er from the AOMK-100/0.5
transformer was measured by
class 0.5 meters. This heater
design provides q = const at the
inner surface of the tube. The
heated duct length was 1 0 = 675,
out of a total length I of 755 mm.
The inside tube may be
positioned concentrically or ec-
centrically by means of replace-
able end grids and slat-shaped
centering ribs 4 mm thick in the
following manner. The ribs are /113
placed into longitudinal slots
milled in the outer tube and
are packed by vacuum rubber.
In all the cases the inside tube
is displaced relative to the
jacket in the direction of one
of the ribs. The angles are
read from the axis of this rib.
The flow cross section of the
duct (upon reducing the area
of the ribs) was 856 mm2 , and
the hydraulic diameter was
dh = 7.58 mm.
The temperature of the
Figure 1. Structure of the Experimental Duct. heat-conducting surface was
1) Thermocouple in a Sleeve; 2) Mixing Head- measured by 12 copper Gon-
ers; 3) Jacket; 4) Slat-Shaped Ribs; 5) Nich- stantan thermocouples, situ-
rome Heater; 6) Inside Tube; 7) Thermocouples ated at different distances
for Measuring the Temperature Distribution from the duct inlet. The ther-
of the Coolant at the Duct's Exit, mocouple wires in Viniflex in-
sulation were placed into steel
capillaries with dimensions of
0.5/0.3 mm. The latter to-
gether with the thermocouple wires were inserted into the wall of the inside tube
by calking and metallizing.
The temperature of the Na-K alloy at the duct's inlet and outlet was mea-
sured by thermocouples located in the mixing headers. In addition, at the point
where the coolant left the duct we also measured the temperature distribution of
the coolant along the perimeter of the annulus, for which we have used thermo-
couples in capillaries placed each 15 0 along the perimeter.
109
All the components of the test duct in contact with the coolant were made
from 1Kh18N9T stainless steel. The test duct was placed on the stand in the ver-
tical position, and the liquid metal flowed from the bottom upward. The experi-
ments were performed for coolant temperatures from 40 to 100C. The average
thermal conductivity coefficients were: for the wall X w = 13.2 kcal/m-hour-de-
gree, for the Na-K alloy X f = 19.5 kcal/m-hour-degree. The oxygen content of
-3
the alloy did not exceed 10 % by weight.
As was shown by the experiments, bursts in the temperature of the heat-
conducting surface are observed near the ribs in the concentric annulus. The
dimensionless variations in the temperature of the concentric annulus for the
region with stable heat transfer is shown in Fig. 2a. The magnitude of the var-
iations depends to a large extent on the Peclet number, which was calculated on
the basis of the average velocity and of the d calculated upon subtracting the
rib area. As the Peclet number is made large the region of elevated tempera-
tures near the rib narrows down somewhat. This, apparently, is due to the ef-
fect exerted by the higher turbulization of the flow in high-velocity coolants. In
the range of Peclet numbers used the dependence of the temperature nonuniform-
ity on the Peclet number can be approximately described by a power law. In log-
arithmic coordinates this relationship is expressed by straight lines (see Fig. 2b).
The variations in the temperature along the perimeter of the heat-conduct-
ing surface of eccentric annuli with ribs for Z/dh pt 79 are shown in Fig. 3. The
Peclet numbers have been calculated on the basis of the average coolant velocity
and of the hydraulic diameter of the annulus, upon subtracting the rib area. Tem-
perature bursts at points with co = 120 0 correspond to the rib locations. In the re-
gion with the minimum annulus width ((p = 0) the temperature increase is basic- /114
ally produced not by the ribs, but rather by over-heating of the coolant in the
narrow parts of the eccentric annulus. The temperature nonuniformity in eccen-
tric annuli is much greater than that produced by the ribs.
The temperature distributions for eccentric smooth annuli behave in approx-
imately the same way as the distribution for ribbed annuli, with the exception of
the local temperature bursts beneath the ribs. In the region of low Pe (transient
and laminar flows) the temperature nonuniformity for a smooth passage is less
dependent on the Peclet number than for large Pe. A similar trend was observed
also in experiments in a ribbed annulus with a relative eccentricity of x = 0.5.
Comparing experimental data obtained for ribbed and ribless annuli we see
that the temperature variations over the perimeter of the central heat-conducting
tube for low Peclet numbers (20 < Pe < 80) in both cases is approximately the
same (Fig. 4). As Pe is increased, the temperature variations in the ribless an- /116

nulus reduce much faster than in a ribbed space. Apparently the ribs interfere
with the intermixing of the coolant and make tagential flows of heat in the coolant
more difficult.
The temperature variations over the perimeter of the inner tube is determ-
ined basically by the difference in the amounts of heat absorbed by the coolant in
the narrow and wide parts of the eccentric annuli (Fig. 5). It can be seen that the
110
0.12
0,10
14 o
0041 I I Iv^
0 20 40 Sp, deg
a)
K K
A Figure 2. Stabilized Values of

x' Temperature Nonuniformity in
x'
a Concentric Annulus With Ribs,
LZ_t, o, , o and o are Results of Ex-
10 20 40 60 80100 200 400 Pe periments With Pe = 21, 75, 223
b) and 540, Respectively,
0,4
0.3
'oa
0.1
i
0
0,1
U 6U 7ZU l0, deg
w
8
6
4 i Figure 3, Variations in the Tem-
V perature of a Ribbed Annulus With
2 Cd
an Eccentricity of x = 0.25. , o, o,
o and 0 are Experimental Results
for Pe = 26, 49, 78, 236 and 559,
20 40 6010100 200 400 Pe Respectively.
111
4,0 U local mean temperature
differences between the
2. 2.0 wall and the coolant are
7-0,5 much smaller than the total
0 ô- N
a 0, 8 temperature variation,
^\ o,s and the curves describing
o, 0,41; the temperature distribu-
tions in the fluid and in the
2 wall are approximately
o.
0.08
4^_ 4 ^
equidistant. A similar pat-
tern is observed also in
ribbed annuli.
906 ;06
004
20 40 60 80 100 20J 400 600 Pe The temperature dis-
tributions we have found were
Figure 4. Variations in the Temperature of used to calculate the distri-
Ribbed (a) and Smooth (b) Eccentric Annuli. butions of the thermal fluxes
along the tube perimeter [3]
27. ti T. + ak Cos 4, (1)

k=1
Rw
^N
^N 1\
n
^N
^N
9 _ 1
9
I
k=1
akkek cos k(p, (2)
1\
,N3
t
where Tw = R X f is the
0 p=deg q 2
0 60 120 9,deg 60 1Zp
a) b) dimensionless local tem-
perature of the wall in the
Figure 5. Temperature Distributions in the Wall section at hand, T is the
(Solid Line) and in the Fluid (Dashed Line) in
value of T averaged over
a Ribless Annulus With an Eccentricity x = 0.5 w
and with Z/d h = 75, Obtained for Pe = 27 (a) the perimeter, a k is a co-
and Pe =147 (b). efficient of the Fourier
series describing the tem-
perature distribution, q
and q are the local specific
thermal flux and the flux averaged over the perimeter; e k is a parameter charac-
terizing the design and properties of the fuel rod [3].
Ek
_ '-w 1 bik

1 +tik 11 =RZ.
Ri
(3)
112
Since the heat conductingtube had a relatively small wall thickness (6/11 2 = /117
= R2 -Rl/R2 = 0. 059), which made thermal fluxes over its perimeter difficult,
the variation in the distribution of thermal fluxes q/q was [only] moderate. We
present an example. For a coolant annulus with Pe = 82, 1/d h = 75 and with an
eccentricity of x = 0.5 we found
T,oTw ^ 0.572cos cp 0.217cos 2cp+0,0715cos 3cp+

+0.020 cos 4cp + 0,0177 cos 5cp ; 0,006 cos 6cp;
q -1 (0.024 cos cp 0.036 cos 2cp =, 0,0266 cos 3(p +
4
+ 0.013 cos 4cp + 0.018 cos 5cp 1 0,0086 cos 6W).
When co = 0, the magnitude of q/q differs from unity by approximately 10-13%

and at the remaining angles it does so by not more than 5%. Consequently, the
thermal flux over a major part of the tube's perimeter is close to the average
thermal flux.
The experiments have shown that for eccentricities of x? 0.25 the heat
transfer does not become stabilized as far as 80 d (Fig. 6). It may be as-
sumed that the section of thermal sta-
bilization will become larger with an
increase in Pe. Thus the data obtained
for eccentric annuli pertain [only] to
a x=0.25 the inlet section.
X. The temperature variations in
^3 0 ribbed eccentric annuli is more depend-
ent on the Peclet numbers than in con-
x=0,5 centric annuli. In the first case the
temperature variation is produced pri-
^N marily by the eccentricity, while in the
ô. second case it is due to local temper-
,4^ ature bursts in the vicinity of ribs. It
K, should be emphasized that we are here
comparing the temperature variations
as a function of the Peclet number for
the inlet section in eccentric spaces
20 40 60 l/dT with these variations for a stabilized
region of heat transfer in a concentric
Figure 6. Changes in Temperature passage. It is logical to assume that
Variations Along the Length of Rib- for the stabilized heat transfer region /118
less Eccentric Annuli. 1, 2, 3, 4 and the temperature variation in eccentric
5 are Curves for Pe = 26, 54, 80, 143 annuli will depend less on the Peclet
and 234, Respectively. number than for the inlet section.
113

The curves of Nu as a function of the Peclet number for different Z/d h are
shown in Figs. 7 and 8, The Nusselt numbers were calculated from the expres-
sion
Nu = 4dk
i rf) if
( w
where w and t are the average temperatures of the wall and the fluid for the val-
ue of Z/dh at hand. The Peclet numbers have been calculated on the basis of the
average coolant velocity. The hydraulic diameter was found for the ribbed annuli
upon subtracting the rib area, and for ribless passages it was found from the ex-
pression (dh) sm - d2 d1'
Data on the average heat transferred by concentric annular spaces for sta-
bilized heat transfer are presented in Fig. 9. On the same figure are drawn
curves obtained from Lyon's semiempirical formulas quoted in [4]
Nu=4.9+0.0175Pe 8 (4)
and from those given in [5]
Nu = 5.1 + 0.02 Pe O,8 . ( 5)
The experimental data for a ribless annulus are higher than values obtained from /119
the above formulas by not more than 15% at large Peclet numbers. When Pe is
reduced this difference becomes smaller. Previously published data for mercury /120
are also in satisfactory agreement with Eqs. (4) and (5), [5]. The presence of
ribs reduces the values of Nu calculated by taking into account temperature bursts
beneath the ribs and near them by an average of 25%.
On the basis of the results we have obtained, we may claim the following.
First, the variation in the coolant temperature over the width of a concentric
annulus is small in comparison with the temperature variation over the passage's
perimeter (see Fig. 5). Secondly, for values of I/ dosed in this investigation, the
heat transfer does not become stabilized. Tangential thermal fluxes in a narrow
space with a coolant and in the thin wall of the heat conducting tube are moderate
for the inlet section, which is attested to by the weak change in the thermal flux
along the perimeter of the passage and the close-to-linear dependence of the tem-
perature variation on Z/d h (see Fig. 6). Let us now consider the problem of tem-
perature variations under the following assumption; 1) the heat transfer agent
temperature is constant over the width of the annular space and is equal to the
wall temperature, but it varies over the perimeter, i. e., t = t = t = f(cp);
114
VU
ryU ^ _r 10
8 --
6 - 6
41 4 7 ^
2 2
4 4
l
5
0,8 20 40 60 80 700 200 400 600 Pe
20 40 60 80100 200 400 600 Pe
a) a)
'YU
10
8
2 2^
1 ,. 1_
-t-- I I i I I i t
Zo 40 60 80 700 200 400 600 Pe 40 67 SO 100 900 4017 6O0 A
b) b)
Figure 7. Nusselt Numbers as a Figure 8. Nusselt Numbers as a

Function of Pe for Ribbed Annuli Function of Pe for Ribless Annuli
With an Eccentricity of x = 0.25 With an Eccentricy of x = 0.25 (a)
(a) and x = 0.5 (b). 1, 2, 3, 4 and and x = 0.5 (b). 1, 2, 3, 4 and 5 are
5 are for 1/d h = 6, 13, 26, 52 and for Z/dh = 6,18,25,50 and 75,
79, Respectively. Respectively.
7I I i l i l l l 2 . R Îi^"'

31 1 1 1 1 111

I I I I I I I I
30 40 50 60 70 80 90 100 200 300 400 .00 600 700 900 Pe
Figure 9. Average Heat Transfer in Concentric Annular Spaces

for Stabilized Heat Transfer. and o Denote Experimental
Data for a Ribless and Ribbed Annuli, Respectively; 1 and 2
Correspond to Calculations by Eqs. (4) and (5), Respectively.
115
2) no tangential heat fluxes take place in the fluid or wall; 3) on the heat conduct-
ing surface the specific thermal flux is the same in all points (q = const); 4) the
flow is hydrodynamically stable.
The thermal balance of an elementary volume of the coolant, bounded by
angle dco, height dx and the passage's walls can be written as
gR2dcpdx= at dxUV do Cpy,
where U^ is the average velocity through the elementary area;
r2 Ri
do = 2 dy ; rm = Rz + y,
where y is the distance between the walls of the passage along the normal to the
surface of the inside tube and R2 is the outer radius of the inside tube. Trans-
forming we get
8T _ 2
8X Pe U^ ^
11 ` ( 1 ql
d 2 )L\ Rzl 2 J (6)
where U is the coolant velocity, averaged over the entire passage
T 11f X=x.
qR 2 ' R2
Integrating Eq. (6) and transforming, we get
8 (A 1)2Z U
T Pe [t (2d -) Up ' (7)
where p = y/R2 ; A = D/2R2 is the ratio of the diameters of the outside and in-
side tubes and Z = x/dh is a dimensionless coordinate along the passage.
To find the velocity distribution we assume, by analogy with [6] that the /121
d'Arcy formula is valid for each elementary area dw, if for the characteristic
116

length we take twice the space 2y (local hydraulic diameter), and for the charac-
teristic velocity we take the average velocity (UÔ) in the elementary area
dp _ S pGT
dx Py 2
For the entire annulus dp/dx = ^/2y. 2P T . Assuming that the local and average
coefficients of frictional resistance and ^ are described by the same law, for
example, by the Blasius formula = 0.316/Re0, 25 we will get from the last two
equations
Up \n_ \ n ' (8)
y (A 1
where n = 0.715. If we assume that as this was done in [6], then n = 0.5.
The velocity distribution calculated from Eq. (8) with n = 0.715 is in satisfactory
agreement with experimental data obtained in [7].
Finally, from Eqs. (7) and (8) we get
8z (,4--1)2.7i5
T=
e ,i.7t5 (2+N)( 9)
Despite the large number of assumptions made, analytical calculations using Eq.
(9) yield satisfactory results, which is attested to by Table 1. Consequently, the
assumptions we have made [do] reflect the heat transfer process at the inlet sec- /122
tion of the annular space under study.
The above experimental data were obtained for a uniform heat supply to the
inner surface of the inside tube of the passage (q - const at R1). However, these
data are just as valid for the case of uniform heat generation in the wall of the tube
(qv = const at Rl - r - R2), which was shown, for example, in [3]. In fact, the tem-
perature field of the heat transfer surface does not depend on the method by which
the tube is heated (q = const at R or q v = const in the wall), since the tangential
heat fluxes are determined solely by the relative thickness and thermal conduc-
tivity of the tubes's wall under identical conditions in the fluid. The similarity
parameter s characterizing the properties and dimension of the tube is in both
cases described by Eq. (3).
The inside tube used in the experiments can be regarded as thin-walled [3],
since the magnitude of sk/k depends little on the number of the harmonic (k) of the
117
TABLE 1. Comparison of Results Calculated From
Eq. (9) With Experimental Data for Ribless
Eccentrical Annular Spaces
Eccentricity x I 0.25 I 0.5'
Peclet numbers I 79 1 234 1 568 I 82 1234 1558
imax_tmin Calculated I0,42010,140I0,05S1 1, IS ( 0,4t1 0.t;

W w -
qR^ f
X I
Obtained experi. 0.364I0.112I0.045I 1.30 0.35I 0,15

i
Maximum disagreement, % i 19 I 25 I 29 I 10 10 I 13
TABLE 2. Values of the Similarity Parameter s k for

Xw/A f = 0.677 and lw6A fR2 = 0.0398
Number of I 1 I 2 I 3 ( 4 I 5 6
harmonic, k I
ai edfrom 0,0410 0,0808 0,1205 1 0,1590 0.1965 0.2330
Eqb 1
8h /k I 0.0410 I 0.0404 I 0,0402 ( 0,0398 I 0,0393 I 0.0388

Fourier series, which can be seen from Table 2. Then it is possible to trans-
form parameter e.k to the simpler form
1,8
ek f2 , S = R 2 Ri.
(10)
In our case e k = 0, 398, hence the results obtained in the experiments are valid
for passages in which the parameters of the thin-walled inside tubes satisfy the
conditions
1,6 ti 0,04; Pr '< 1.

fR2
118
The method of approximate thermal simulation presented in [3] makes it possible
to use the data we have obtained for calculating a given group of fuel element rods.
REFERENCES
1. Stein, R. and Begell, W. A. 1. Ch. E. Journal, vol. 4, No. 2, 127 (1958).

2. Marin, N. I., et al.: Atomnaya energiya, 18, 3, 239 (1965),
3. Ushakov, P.A.: Priblizhennoye teplovoye modelirovaniye tsilindricheskikh
teplovydelyayushchikh elementov [ Approximate Thermal Simulation of
Cylindrical Fuel Elements]. First article in Part II of present collection.
4. Harrison, W. and Menke, J.: Trans. ASME, 71, 797 (1949). /123
5. Subbotin, V.I. , et al.: Atomnaya energiya, 9, 4, 310 (1960).
6. Khabakhpasheva, Ye. M. and Korsun, A. S.: In the collection Zhidkiye metally
[Liquid Metals], page 34, Gosatomizdat, Moscow, 1963.
7. Levchenko, Yu. D., et al.: Gidrodinamika v kol'tsevom zazore s prodol'nymi
rebrami [Fluid Dynamics in an Annular Space With Longitudinal Ribs].
,Article immediately preceding this one in the present collection.
119

TRANSFER OF HEAT IN TURBULENT FLOWS OF FLUIDS IN

ANNULAR AND PLANE SPACES
N. I. Buleyev, V.A. Mosolova and L. D. Yel'tsova
This article presents calculations of temperature fields and heat transfer /123
coefficients for turbulent flows of fluids in annular and plane spaces. Use is made
here of integral approximations for turbulent thermal conductivity coefficients [1].
The calculated Nu t and Nut , corresponding to one-sided internal or exter-
nal heating of the fluid flow are represented in the form of functions of param-
eters Re, Pr and O = al /a2 , where al and a2 are the radii of the inner and outer
tubes, forming the space, respectively.
The calculated heat transfer coefficients for several fluids are compared
with the available experimental data.
Statement of the Problem
Let us consider the steady temperature regime in a turbulent flow of a fluid

in a space with the specified thermal fluxes q and q2 at the walls of the passage.
The starting equation of the heat influx in the given case can be written in the
form [1]
8 r 8r { r(t+et) 8T } '
w (])
where
Et ( M o) = 0,20Lo F (^1) fo (4^1) f, (X 4 t1) G (i1) dyl,

(2)
and also
rr L I ar Iwhen y,,> 25
tl =
042LO ' F (1) _ 0 when ' 2"
120

fi (x) _ 2 (1 e- I x l), /124
_ 75 Lo _ LZ I aw
q y* L ar
! 0,8+0,2 / v)"" when Y > 1,
t 1 when v < 1.
Here T is the temperature of the fluid, v, k and s t are the coefficients of mole-
cular viscosity, molecular and turbulent thermal diffusivity respectively and
L is the turbulence scale.
For calculating the coefficient e using Eq. (1) we restrict ourselves by the
dimensionless integration interval -2.4 < 71 < 2.4, i, e., by the interval I r r 0 1
L 0 [1]. The quadrature formulas for s t will have the form
s t (Mo) = 0,20Lo {0,15 [D ( 0,4) + D (0,4)] +

+ 0,23 [D ( 1,2) + D (1,2)1+ 0,12 [D ( 2, 0)+D(2,0)]),
where
D (ii) = F (-t) to ( qT j) fi (%")
To determine L in a cross section of a rectilinear passage of infinite length,

[1] has suggested the approximation
2R
1_ 1 1
L (M) 2 l dy'
o (3)
where I (cp) is the distance between the point in question M to the perimeter of the
passage cross section in the variable direction cp.
In Eqs. (1)-(3) we change to the dimensionless variables
a2 ' a2' y 1-0' U V*
^ = v* , Re =v = 2UD,
a ^ 8 2
ZI 2 ,
_ (4)
v* = a p1 l aaz
l-8
S w,
a
where a = a2 a i , and
121

For a duct with an annular cross section Eq. (3) for the turbulence scale
L(M), making use of the notation of Eq. (4), yields the following results [2]
a2 L ai = B
(1 Y) Y,
5
()

where /125
8 1^ 1 ^ ^ 0 .
H 0
sin x _
E (k, (p) _ ^ V1-0 sin e a da,

0
n/2
H-(k) _ ` ycos2ada
1k2 sin a
0
The values of function B(^, O) are close to unity.

For the case of a plane space of width h, Eq. (3) takes on the simple form
L = ( 1 Y)Y,

(6)
where y = Z/h and l is the distance between variable point M and one of the plates
forming the space.
To eliminate 8T/8z from Eq. (1) we use the balance equation, which is ob-
tained by interating Eq. (1) along the radius from a 1 to a2
cp(a' a , )w aT =2 ( a i g i-I- a 2 g2) (7)
Substituting the value of 8T/8z from the above expression into Eq. (1), and using
the notation of Eq. (4), we get
1 8 k Et 8T _ 2 i +, U
g ay {^' v + v) 8y 1-02
X OgCO U*
y U ( 8)
122
From the point of view of mathematics it is sufficient to examine only the
case of heat supply from one side, since for the case of two-sided heating with
arbitrary relationships between fluxes q and q2 the solution for temperature T
can be represented in the form of a linear superposition of solutions for the case
of only inside and only outside heat supplies.
Equation (8) was solved by a numerical method. Here coefficient was
F
calculated from Eq. (2) using the field of velocities w obtained in [1].
Results of Calculations
Table 1 present the results of calculating the dimensionless temperatures /126
tY cpv * (T I T)
` qi
in the flow of fluid in the space for a given constant thermal flux q 1 at the internal
wall of the space and the dimensionless temperatures
_ cpv * (T2T)
2
q2
for a given thermal flux q2 at the inner wall of the space. Here T 1 and T2 are
the temperatures of the inner and outer walls of the space, respectively.
If we introduce the concepts
2a q l 2^ Pr _ 2a 42
and Nu2 _ 20 Pr
Nu _ X(T2T)
A ( Ti T) V1 *2
for the corresponding cases of heating from one side, then the following can be
said about the results of calculations of Nu t and Nut in annular spaces. First
of all, the calculated Nu 1 are greater than Nut . As parameter O becomes larger, /127
the difference in the calculated values of the above two Nusselt numbers, naturally,
decreases (Table 1). For a specified value of O the relative difference between
Nut and Nut is reduced with an increase in the Reynolds number. There exists
a tendency toward the reduction in the relative difference between Nu l and Nut
with an increase in the Prandtl number.
123
TABLE 1. Calculated Average Temperatures T 1 and T2 and
of Nut and Nu2 in Spaces With the Flow
of Fluid Heated From One Side
8=0.5 1 8=0,8 8=1
Pr I Re 10-3 9.30 146.7' 1219 ( 9.22 47.5 221 114.1 169.6 ( 322
f
'T , 2.24 5,68 9.55 2.44 5.99 10,0

0.025 )^ 72 2,57 6,29 10,4 2.56 6.17 10.2 3,22 7,14 10.9
l Nu2 6.11 9194 24.1 6.21 10.3 24.9 6,89 12,6 32,6
0.8 tY 1 18.2 21.7 24.2 19.6 22,7 24.5

Si ^Y 2 20.4 23,0 25.4 20.8 23,5 25.4 21.5 24,2 26,2
Nu l 27.6 92.2 331 26.0 89.5 332
l Nu 2 24,6 87.0 316 24.5 86.5 320 33.0 117 434
10 { Nu 2 116 490 1930
Figures 1 and 2 show the temperature distributions calculated in fluid flows

in spaces with O = 0, 5-1 at different values of the Reynolds number and with the
heat supplied to the fluid from one side. For the selected dimensionless varia-
bles, a, Nu and Re the calculated Nu2 for the space are lower as compared with
the Nusselt numbers calculated for a pipe with the same Re, particularly numbers
increase rapidly when O is increased from 0 to 0.5. When O is increased from
0.5 to 1, but with the Reynolds number kept constant, the calculated Nu 2 remain
practically constant1
For spaces with O = 0, 5-1 for 5. 10 3 < Re < 5 . 105 the calculated Nu2 differ
from the Nusselt numbers for a pipe by approximately 25% in the case of fluids
1Reference [3] presents calculations for Nut and Nu2 in annular spaces for
fluid flows with low Prandtl numbers, there the values of s needed for the cal-
culations were calculated on the basis of the latest measurements of the fields of
velocities w in the spaces. All that has been said above about Nu t and Nu2 cal-
culated in the present article for fluids with Pr <_ 1 is in satisfactory agreement
with results of calculations of [3].
124
Y/
1.0 ZO
2 2
0.6 Q6
0.2 0.2
AS 0.6 0,7 0.8 0,9 0,6 0.6 0.7 0.2 0.9 f

a) b)
Figure 1. Calculated Temperature Distributions in the Flow of

a Fluid in a Space With the Heat Supplied Through the Inside (a)
and the Outside (b) Wall; O = 0. 5, Pr = 0.025; Curves 1 and 2
are for Re =9.3-10 3 and Re =2.2- 105
with low Prandtl numbers and by 10-15% for gases (Fig. 3). In the case of fluids /128
with Pr = 10 the calculated values of Nu 2 for the space practically do not differ
from the Nusselt numbers for a pipe at the same Re.
The values of Nu2 calculated for liquid metals with O = 0.5-1 can be approx-
imated by the expression

Nu = 5.0+ 0.020Pe O 8 (9)
For gas flowing in spaces with O = 0.5-1 the following expression will be approx-
imately valid

Nu = 0.018Reo. s . pro.4 (10)
The structure of interpolation formulas (9) and (10) has been made traditional
purposely, for convenience in comparing results obtained in the present paper
with results for a pipe and spaces from other publications. Equation (9) yields
somewhat higher values of Nusselt numbers than that given in [4) and of almost
identical with the equation presented in [5].
For known values of functions 'F l (^) and 'F2 (^) the distributions of temper-
ature T in the flow of a fluid heated from both sides with an arbitrary relation-
ship between fluxes Q1 and Q2 can be calculated by the following formula (Fig. 4).
cpv* (T2 T) = 42'Y2 ( )+Qi[ IF i( )'F^.(1)l. (11)
125
Table 2 presents the calculated values of the average dimensionless temperature
AY = cpv* (T 2 T/q2 and of NO = 20 Pr/7 in the flow of a fluid in a plane space
with symmetrical heat supply from two sides. The values of Nu* obtained for the
plane space can be related to the Nusselt numbers obtained on heating from one
side by the following approximate expression

Nu* =8,6 -;- [1.50 0:15 lg (1 + 100Pr)] (Nu 5.0) (12)
which is valid for fluids with Pr = 0-10, When > 10 it may be assumed that Nu*
= Nu.2
91 IP
TABLE 2. Calculated Average Temper-
1
atures * and NO in the Flow of a Fluid in
1.2
a Plane Space With the Fluid Heated Sym-
2 metrically From Two Sides
1.0 104 Re
3 Pr
1.41 1 6,96 1 32,2
0,8
0.025 T 1.89 4.42 , 7.41
Nu * 11.7 20,1 47.8
0.6
0.8 T 16.5 19.3 22.6
Nu * 43.0 147 503
0,4
10 `Y 6967 68.5 69.6
Nu * 128 519 2040
0,2
0 0.2 94 0.6 0.8 Z

Figure 5 compares the calculated /130
Figure 2, Temperature Dis- Nu and NO for flows of liquid metals in
tributions Calculated in the a plane space with experimental data of
Flows of Various Fluids in a [7]. As can be seen from this figure, the
Flat Space (O = 1) With Heat theoretical and experimental results are
Supplied From One Side; Re= in good agreement. The calculated Nus-
= 6,96 . 104 ; Curves 1, 2, 3 are selt numbers for the flow of gas in a plane
for Pr = 0.025, 0, 8 and 10, space are in good agreement with results
Respectively, obtained by Doroshchuk, [8], On the other
"The results provided by Eqs. (9), (10) and, (12) were obtained by the au-
thors earlier, when using the arrangement suggested in [6] for calculating the
velocity and temperature fields in the passage. Equations (10) and (12) are
presented in [7],
126

hu
40
i
20
10
2 l-
8
3 ^
6
4
4 6 8 10 4 2 4 6 8 1,1 2 R2
a)
Nh
406
206
l00
80
A
46
20 1 r r i i i i i t 1
4 6 8 104 2 4 S 8 10 3' 2 Re
b)
Figure 3. Calculated Nusselt Numbers for a Flow of

Fluid in a Space With Heat Supplied From One Side;
Pr = 0. 025 (a), Pr = 0.8 (b). 1) O = 0. 5-1 (Heating
From the Outside; 2) O = 0 (Round Pipe); 3) O = 0.5
(Heating From the Inside); 4) O = 0.2 (Heating From
the Inside).
band, the calculated heat transfer attendant to the flow of fluid in a space with
Pr = 10 are in agreement with measurements made by Mikheyev and others [9] /131
pertaining to the small difference between Nut and Nu t , as well as with respect
to bringing these numbers close to the Nusselt numbers in a pipe for the same
Reynolds numbers.
127
Figure 4. Temperature Distribu-
tions Calculated in the Flow of a
Fluid in a Gap Heated From Two
Sides; O = 0.5; q 1/q2 = 1, Pr =
=0.025; Curves 1 and 2 are for
01 5 0,6 0.7 OR 09 F Re = 9.3 . 10 3 and Re = 2.2 . 105,
9u
20
4
14 3 ^/
10
t .^ 2
6
102 2 6 B 103 2 Pe
Figure 5. Calculated Nusselt Numbers for Flows

of Liquid Metals in a Plane Space; the Experimen-
tal Data are That of [7]. 1) and 2) are Values Ob-
tained Theoretically and Experimentally, Respec-
tively, for Heating From One Side; 3) and 4) are
Values Obtained Theoretically and Experimentally,
Respectively, for Symmetrical Heating
From Two Sides.
REFERENCES
1. Buleyev, N. I.: Teoreticheskaya model' turbulentnogo obmena v trekhmernom

potoke zhidkosti [ Theoretical Model of Turbulent Transfer in a Three-Di-
mensional Flow of Fluids]. Report presented at the 3rd International Con-
ference on Peaceful Uses of Atomic Energy. Geneva, 1964.
2. Buleyev, N. I.: Teoreticheskaya model' mekhanizma turbulentnogo obmena
v potokakh zhidkosti [Theoretical Model of the Mechanism of Turbulent
Transfer in Flows of Fluids]. In the collection Teploperedacha [Heat
Transfer]. Published by the . Academy of Sciences of the USSR, Moscow, 1962.
3. Dwyer, O. E. and Tu, P. S.: Nucl. Sci. Engng, 15, No. 1 (1963).
4. Harrison, W. B. and Menke, J. R.: Trans. ASME, 71, 797 ( 1949),
5. Seban, R.A.: Trans. ASME, 72 9 789 ( 1950).
128
6. Buleyev, N. I.: Raspredeleniye skorosti i tmperatury v turbulentnom potoke
zhidkosti v krugloy trube [Velocity and Temperature Distributions in the
Turbulent Flow of a Fluid in a Pipe]. In the collection Voprosy teploob-
mena [Problems Heat Transfer]. Published by the Academy of Sciences of
the USSR, Moscow, 1959.
7. Subbotin, V. I. , , et al.: Atomnaya energiya, 9, No. 4 (1960).
8. Doroshchuk, V. Ye. and Frid, F. P.: Energomashinostroyeniye, No. 11(1959).
9. Averin, Ye. K. , et al.: In the collection Teploperedacha i teplovoye modeli-
rovaniye [Heat Transfer and Thermal Simulation], page 138. Published by
the Academy of Sciences of the USSR. Moscow, 1959.
129
EXPERIMENTAL STUDY OF HEAT TRANSFER FROM A FLAT HORIZONTAL
SURFACE TO SODIUM BY FREE CONVECTION
A. P. Kudryatsev, D. M. Ovechkin, D. N. Sorokin,
V. I. Subbotin and A. A. Tsyganok
The experiments described in this article were performed in order to com- /131
pare the levels of heat transfer on boiling [1], [2] and free convection. In addi-
tion, they were of independent interest since we know of only two publications /132
devoted to the study of heat transfer from a horizontal surface to liquid metals
under free convection conditions, [3] and [4].
Reference 3 concerns itself with heat transfer from sodium, which gives
up its heat to a flat round plate from stainless steel 203 mm in diameter which
is located approximately 3 mm below the free level of the .sodium, which is con-
tained in a tank 760 mm in diameter. The sodium was heated by furnaces situ-
ated outside the tank. The experiments were performed with sodium tempera-
tures in the range of 140-350C. On the basis of experimental data it was found
in [31 that
Nu = 0.262 (Gr Pr2)0.35. (1)
The characteristic dimension in the above expression is the diameter of the plate.
On the basis of the power of (Gr- Pr 2 ) in the above expression, it is assumed in
[3] that the heat transfer took place in the turbulent region.
Experiments with mercury situated between two horizontal chromium plates
(electroplated) copper plates are described in [4]. Here use was made of plates
134 mm in diameter held at distances of 35 and 66.5 mm from one another. The
lower plate was heated while the top plate was cooled. The data thus obtained
can be expressed thus
Nu =0.069 (Gr Pr) 1 /3 Pr0.074. (2)
The characteristic dimension here was the distance between the plates.
Our experiments were performed with flat horizontal surfaces from
1Kh18N9T stainless steel 38 mm in diameter, which was heated by electron bom-
bardment. The heat transfer surface was placed in a work tank from stainless
steel 165 mm in diameter. The sodium level above the surface was about 150
mm. Chromel-alumel thermocouples, which measured the temperatures of the
heat conducting wall were placed in protective capillaries from 1Khl8N9T
130
stainless steel 0, 8 x 1.5 mm in diameter, which were then covered by molten
copper under a vacuum (thermocouple wire diameter 0.2 mm). The temperature
of the sodium was measured by chromel-alumel thermocouples placed in a sleeve
(thermocouple wire diameter 0.5 mm), the beads of which were located by 60
mm above the heat conducting surface. Such an arrangement is also described
in [5].
We have measured the temperature differences between the wall and the
liquid and the temperature of the sodium, as well as the temperature fields in
the latter using a moving thermocouple. The readings of the thermocouples, in-
cluding the movable one, which was connected differentially with the stationary
thermocouple, were recorded continuously by EPP-09 potentiometers (1 sec)
and were also measured intermittently by the R-2/1 potentiometer. We have
performed four series of test. In all of them the sodium was held under inert
gas pressure of about 2 atm absolute. The regime of the heat transfer agent
(laminar or turbulent) was determined on the basis of temperature fluctuations
recorded by the movable thermocouple and by the thermocouples in the wall.
Results of Experiments for the Turbulent

Region
q 10-6 ( At, tl, I q . 10 6 , Ot, I tl,
k cal/(m2 hr) degs. degs cal/(m2 hr degs. degs.
1 Series I 2 Series
0.77 80,2 300
0,74 89.9 25 2, 1.12 105 324
1,51 140 281 1.53 135 419
1,11 120 294 3 Series
0.74 99.4 518 0.72 806 481
0,69 94.2 572 1 .93 101, 518
1,08 133 634 1.11 121 555
0,71 86.1 660 1.31 138 592
1,52 146 680 1.50 150 621
0.95 114 682
0,70 86.3 694 4 Series
0,70 95.4 699 0.71 74.5 550
1,00 127 719 1.06 107 690
1,43 139 734 1.46 136 733
For thermal fluxes below 0, 7 . 106 kcal/m2-hour neither the movable ther-
mocouple, situated near the wall, nor the thermocouples built into the wall re-
corded any temperature fluctuations. When the heat fluxes were made large,
both the movable and the stationary thermocouples record temperature fluctua-
tions. At a distance of more than approximately 15 mm from the heat conduct-
ing surface no temperature fluctuations have been observed for thermal fluxes of
up to 1.5 . 106 kcal/m2 -hour. The presence of fluctuations shows that the fluid flow
at the heated surface is turbulent. For thermal fluxes of (0.7-0. 9) 10 6 kcal/m2
-hour,wicespndtaofrmhelintub arsfe
131
regime, the Grashof number is about 108,
which agrees with data obtained by Fedynskiy
[6 ], who has found that transition from laminar
to turbulent flow takes place at Gr > 1.5 . 108
irrespective of the Prandtl number. Figure 1
shows a recording of the readings of the mov-
able thermocouple in contact with the wall for
q = 0.95 . 106 kcal/m2 -hour and t1 = 682C.
Figure 2 displays temperature distribu-

U
tions in the sodium obtained by the movable
N thermocouple for three values of the thermal
flux. The dashed line denotes the position of
the center of the movable thermocouple's junc-
tion at the instant it touches the wall. The heat
transfer coefficients a, calculated on the basis
of the temperature differences between the wall
24 C and the liquid and on the basis of the tempera-
ture measured by the movable thermocouple in
the liquid next to the wall, with subsequent ex-
trapolation of the temperature to the wall, are
Figure 1. Record of the in satisfactory agreement. For example, a /134
Readings of the Movable determined by the temperature difference for
Thermocouple, Touching thermal flux values of (0. 35, 0. 95, 1.53)- 106
the Wall. kcal/m-hour comprise 6. 75. 10 3, 8.35-10 3 and
11.3 . 103 , respectively, while cu determined
from the temperature distributions constructed in Fig. 2 comprise, respec-
tively, 7.96-10 3, 8.64-10 3 and 13.3 . 103 kcal/m-hour-degree
It is recommended in [6], on the basis of analysis of differential equations,
that the experimental data on heat transfer by free convection of liquid metals
be processed in the coordinates Nu - Cr- Pr 2 , i. e. , without giving consideration
to the viscosity. Using these coordinates we have constructed (Fig. 3) the ex-
perimental curves for the turbulent region. These points have a scatter of + 15%
about averaging curve 1, which represents the relationship
Nu = 0.38 (Gr Pr 2)1/3. (3)
The characteristic temperature here is the temperature of the fluid far from the
wall. If the wall temperature is taken as the characteristic temperature, then
the results will differ by approximately 4.6%, since the temperature difference
between the wall and the liquid metal in our experiments did not exceed 150C.
The physical properties of sodium used in the calculations are those given in
[7] and [8]. The diameter of the heat conducting surface (38 mm) was taken as
the characteristic dimension.
132
H, r. im
25
20
15 \\\
2
10
3
U
Lj ZO 40
1
60
B 4mm
T 60 100 &'-c
Figure 2. Temperature Distributions for Sodium, Con-

structed on the Basis of Readings of the Movable Ther-
mocouple. The Dashed Line Denotes the Position of the
Center of the Junction of the Movable Thermocouple at
the Instant When the Latter is in Contact With the Wall.
Curve 1 = 0.35 . 10 6 ; Curve 2 = 0.95 . 106;
Curve 3 = 1.53 10 6 kcal/m2-Hour.
No
+15
'O r
6
0
2
S
Figure 3. Experimental
Data on Heat Transfer.
Test Series: - First;
o - Second; n - Third;
(3 - Fourth. 1) Obtain-
3 ed From Eq. (3); 2) Ob-
tained From Eq. (1);
f 6 B ,rPr2X10 9 3) Obtained From Eq. (4).
133

Reference [6] recommends to use the following formulas for heat transfer
calculations in liquid metals by free convection under turbulent and laminar
regimes
Nu = 0.16(\ 1G,--Pr= 11/s

Pr ) (4)
Nu = 0.67 r GrPr2
11/4
I+Pr ) ( 5)
Line 3 of Fig. 3 was obtained by using Eq. (4). If the Nusselt numbers are cal-
culated from Eqs. (4) and (5) at a sodium temperature of 500 0 , then Eq. (4)
yields Nusseltnumbers approximately two-fold smaller than Eq. (5) for a Grashof
number of about 108 (Gr Pr _ 2.5-10 3 ), which corresponds to transition to tur-
bulent flow. Equations (4) and (5) yield identical values of the Nusselt number /136
only when Gr 10 12 . Consequently, Eq. (4) is incorrect.
Equation (5) is in satisfactory agreement with Eq. (3) when Gr. Pr 2 =2.5 . 103,
i. e. , at the start of the turbulent heat-transfer region.. In the interval of
Gr Pr of about from 2.5 10 3 to 8 . 103 Eq. (1) (line 2) yields values of Nu by
about 25% lower than Eq. (3). If the values of o! are calculated from Eq. (2) for
the turbulent region [q > 0.7 10 kcal/m 2 -hour], then they lie by about 20% lower
than the average value of data obtained by us. Equation (5) is in satisfactory
agreement with our experimental points obtained with low thermal fluxes, i. e. ,
in the laminar region (points not shown).
It thus may be assumed that Eqs. (1)-(3) reflect the level of heat transfer
to liquid metals by free convection in the turbulent region, while Eq. (5) is valid
for laminar flow.
REFERENCES
1. Subbotin, V. I., et al.: Teploperedacha v elementakh reaktorov, okhlazhda-

yemykh zhidkim metallom [Heat Transfer in Reactor Elements Cooled by
Liquid Metals]. Report presented at the 3rd International Conference on
the Peaceful Uses of Atomic Energy. Geneva, 1964.
2. Kudryatsev, A. P., et al.: Teplootdacha pri kipenii natriya v bol'shom ob"
yeme [Heat Transfer From Boiling Sodium in a Large Volume], Part III
of present collection.
3. Donald, M. and Conolly T.: Nucl. Sci, Engng, 8, No. 5 (1960).
4. Globe, S. and Dropkin, D.: Trans. ASME, 81, 24-30 (1959).
5. Ivashkevich, A. A., et al.: Ustanovka s elektronnym obogrevom dlya issle-
dovaniya teploobmena pri kipenii metallov v usloviyakh svobodnoy konve-
ktsii [Setup with Electronic Heating for the Study of Heat Transfer From
Boiling Metals By Free Convection]. Part III of present collection.
134
Metal Heat Transfer Agents]. Atomizdat, Moscow, 1958.
7. Vargaftik, N. B.: Spravochnik po teplofiziche skim svoystvam gazov i
zhidkostey [Handbook of Thermophysical Properties of Gases and Liq-
uids]. Fizmatgiz, Moscow, 1963.
8. Zhidkometallicheskiye teplonositeli [Liquid-Metal Heat Transfer Agents].
A. Ye. Sheyndlin, Editor, translated from English. Foreign Lit. Pub.
House, Moscow, 1958.
135
PART TWO
HEAT TRANSFER AND FLUID DYNAMICS OF

INTRICATELY-SHAPED PASSAGES
APPROXIMATE THERMAL SIMULATION OF CYLINDRICAL

FUEL-ELEMENT RODS
P.A. Ushakov
In reactors with liquid-metal cooling use is made most frequently of fuel /137
rods (fuel elements), arranged either in the form of regular triangular or square
lattices, or placed in clusters or one-by-one in separate ducts. The geometryof
. the passages created by the rod clusters or of the individual ducts is responsible
for the appearance of variations in the temperature over the fuel rod periphery.
This variation produces additional thermal stresses, i.e., local overheating,
which reduces the reliability of the fuel elements.
The lack of uniformity in the fuel element temperature, all other conditions
remaining equal, is due to the thermal conductivity and on the dimensions of its /138
component parts, i.e., the fissionable-material core, protective jackets, contact
layers, etc. All this should be taken into account in thermal simulation of
fuel elements.
Let us consider a fuel rod with three shells [jackets]. The more elementary
fuel elements are particular cases of the above. Let us assume that ideal thermal
contacts exist between the shells, between the rod and the first shell, as well as
between the outer shell and the coolant; the physical parameters of the system are
constant and do not depend on the temperature, the fluid is incompressible and does
not change its state of aggregation; the effect of natural convection is negligible and
the heat generation in the rod is uniform.
We now analyze equations for laminar flow of the fluid assuming that the sim-
ilarity criteria which will be thus found are valid also for turbulent flow which, on
the average, is steady with respect to time. On these assumptions the equation
of motion for the fluid
R4 Re V2W grad e,,, (W, grad) W'= 0. (1)
Equation for the heat propagation through the fluid
dg 1
R4 Pe
V 2 Tt (WgradTt).=0. (2)
136

The temperature fields in the fuel rod and in the shells are described, re-
spectively, by the Poisson and Laplace equations
(3)
V2 T, = 0, V2 T 2 = 0, V2T3=0. (4)
The boundary conditions
when =0 T
= 0.
when g =, To = T,; aT _ Xo aT ;
When = ^2 T,=T2; a _ ;
.9T2 (5)
when = g3 T2 = T3; a^ )12 g
when g=1 T3=Tt; T =

Xf Tf . (6)
Here we have used the notation /139
W= l i l f r Rt
_ T t == 11 - Ili is the specific thermal
It ' t 'R =. , 4 , 4 = 4u 2R4
flux at the heat-conducting surface (when ^ = 1), d is the equivalent hydraulic

2
diameter of g a
theand
p assa e u = Gip/pu is the Euler number, which is a func-
tion of Re. The remaining notation is shown in Fig, 1,

Analyzing Eqs. (3) and (4) and boundary conditions (5) we can find a rela-
tionship between the temperature distribution and the distribution of thermal
fluxes over the perimeter of the outer shell [jacket] of the fuel element (^ = 1)
Tw _ Tw + ^' ah cos kcp, (7)

k=1
=1 I kakek cos kcp,

9 k=1 (8)
137
Tw = T 3 (co, 1, Z), where Z is a dimen-
sionless distance along the passage and
co is an angle, measured over the rod's
perimeter. Without detriment to gener-
ality of the problem, we may assume
that the functions being considered are
even and thus to drop the sine terms in
the series.
Parameter ek which characterizes
the thermophysical properties of the fuel
element can be found analytically. The
Fourier series coefficients can be rep-
resented in the form
n
ak = T3 (y, 1, Z) cos ky dy
Figure 1. Schematic of a Fuel Rod 0
[Cross Section] With Three Shields
[Jackets], z "
_ Tf ((p, 1, Z) cos ky dy.
0
The specific thermal flux is related to the temperature gradient of the fluid
9
aT f
e ag 1=g,
hence we can write Eqs, (7) and (8) in the form
T f (cp, 1, Z) = Tf (1, Z) + C T f ((p, 1, Z) cos ky dy ] cos kcp,

h=i 0 (9)
I
k=i
kek [
5 Tf ((p, 1, Z) cos ky dT ] cos ky.
0 (10)
From the above two equations, together with Eqs. (1) and (2) we get the
criterial equation for geometrically similar channels
Tf ( q), t, Z) = f ( y , g , Z, Re, Pe, er,). (11)
138
We now find P, k' The solution of the [above] system of equations for the /140
fuel element is most conveniently represented in the form of a sum of two so-
lutions
T, ((p, ^, Z) _ j (^, Z) + T_ ((P, ^, Z), (12)
where 8 is a solution which takes into account heat generation in the core at con-
stant temperature along the perimeter of the outer shell ^ = 1: e(1, Z) = Tw(Z),
while T is a solution for the case when there is no heat generation in the rod,
but the temperature varies over the perimeter of the outer shell, i, e. , when
= 1: T (cp, 1, Z) _00z a h (Z) Cos ky.
k =1
Solutions for E)Q, Z)
.o = T w - ^3 In 1_( )2 ]
}Z Inzn }1 I i+2 io L ^1
1 = TW -} f In + In z -^ f In ^2
_ (13)
E) 2 = 7 w}g I n - ^2
In
a
03 =TW-I f In
Solutions for T(cp, ^, Z) have the form
To = 1J 11h^ h COS kT;

k=1
T1 = ^J ( A hb k +
YY
Bhsf - h) COs kq);
k=1 (14)
M
^1
T2 =!J (Ch^h + Dh^-h) Cos kT;
h=1
W
W^11 Y _ h)
T3 = 'E (D' W + F h^ COs ky.
k=1
The coefficients found from the boundary conditions are
Hh = N hS2 k b3 ( 1 -^ m 1) ( 1 + m 2) ( 1 + M3);
A h = Nh.2 kS3k (1 -P m2) ( 1 + m3);
Bh = N 1 h 3 km1 ( 1 + m2) ( 1 + m3);
139
Ch = NkS3k ( 1 + m3) (S i kni , m 2'^ S2k)i
DA=N Y k 3 k (Sg 1YYm1 Y
k ^1I S2kYYm2 ) 1 4- m3)i
(YY i m 2 + S2k)]i
E h = Nh gg2kmgg3 (S1 km 1 S2k m 2) ^- S3 k (S1
Fk = Nk 12S3k k (S1 km i F 2 km 2) k 3 km 3 (Si7YtIM2
3b k I- 2It)]r

where /141
Nk =gg k R2 k (( 3 k + m 3) (Si km i + b'km2) -^

+ S3 k (1 + b 2km 3) (î km i m 2 + 52 k )]^
M i ^1 -7^0 , = ^2+a1
m 2=
R2 -X I ,
m
= 1 3- 12 . -
1 <m < 1.
X1+ô r 3 %3'+'X2'
The relative thermal flux at the surface of the outer shell
4 _ X 3 8T 3 1 _-
g Tf 8 I g=1
=1 a,,L a ^f (Eh ^ } Fk -k) cos kcp where = 1.

k=i
Resorting to Eq. (8), we get
f Eh = L 3 (EkFk)
It
a,3t2 k ^M 3 ^9 k )âi knt i L S2km 2) + ^3 k ^1 m 3s3 k )^S2km 1^2 S2k)

Xf g 2 k ^ m 3+ g 3 k ) ( i km i) 2 km 2)+ 3 k ( 1 +m3,3 k ) ($ikmint2 + SZk)
(15)
The requirement that e = idem can be rigorously satisfied for any harmon-
ics (any values of k) only on the condition that simplexes ^1' ^2' 3 9 " 1/^0'-'2/x'1'
2'^'ff X3 have the same values in nature and on the model. This requirement
X 3A
also follows directly from the system of starting equations and boundary conditions
(3-9).
The creation of thermal models of fuel elements with volumetric heat gener-
ation and with ideal thermal contacts between the layers under laboratory conditions
is very difficult. The above difficulties can be eliminated by using approximate
thermal simulation. This will somewhat reduce the accuracy of the results
The substance of the latter consists in replacing the fuel rod with shells by
some. equivalent tube, at the inner surface of which we, in one way or another
140
satisfy the conditions: q = const or t = const along the perimeter To be able to
select the dimensions and the thermal conductivity of the equivalent tube it is nec-
essary to know approximately the behavior of the temperature distribution over
the perimeter of the fuel element under study, ie e> , the fundamental harmonic
of the Fourier series which describes the temperature distribution. In some
cases it is possible to guess this harmonic. For example, for cluster of fuel
rods arranged in a regular triangular lattice the fundamental harmonicJs k o =6,
for a square lattice k0 = 4, while for an eccentric annulus k o = 1, [1-4], When
it is difficult to guess the fundamental harmonic, one must resort to successive
approximations. Let us assume that in experiments with the model we have ob-
tained the distribution of the dimensionless variation in the temperature /142
A7 1 ,. = (T. -- T) }^ art Cos Iccp, (16)

rt ^ i
and on the basis of these data we have calculated the distribution of thermal fluxes
in the model
q 1 ahk h lodCos Iccp.

(E )
(17)
rt_i
Had the distribution of thermal fluxes on the surface of the fuel element been the
same as in the model, then the variation in the fuel element temperature would
have been described by the expression
n
AT.* = (T. Tw)* ^ >] (80MOa Cos Iccp.
(18
In the case V en the effective thermal conductivity in the model is lower

than that of the fuel element, which corresponds to the case of (Ek)mod < (E'k)f, e'
the experimental value of AT will be on a high side. On the other hand, the lack
of uniformity of thermal fluxes over the perimeter will turn out to be weaker for
the model than in the fuel. element, while the lack of uniformity AT* will be on
the lower side as compared with the actual lack of uniformity in the fuel element's
temperature. For the opposite case of (Ek)mod > (Ek)f> e' AT* will yield higher
than actual, while AT will yield lower than actual values of variations in the fuel
element's temperature.
The true value of the variation in the fuel element's temperature will lie
between values obtained on the basis of experiments performed with the model
141
given by Eq, (16) and those calculated on the assumption of q/q = idem, which is
the basis of Eq. (18), Approximately this can be defined as the root-mean-square
value
ATw-}-OTw* 1 n r
(hmodl
(,T-)tme^ 2 =2I akL 1 + (eh)
e) f e.
.J cos ky.
k=1
If the divergence between AT w and AT* exceeds the desirable accuracy, then
the experiments are repeated using a model the parameters of which have been ap-
propriately adjusted. The average temperature of the heat-conducting surface de-
pends much less on the design of the fuel element than the variation in the temper-
ature. This is attested to experimental data presented in [1-3]. Assuming, by
analogy with a round tube, that the Reynolds number for liquid metal can be ap-
proximately eliminated from criterial Eq. (11), we get
Tf ((P, ^, Z) ' f ((P, ^, Z, Pe, Eh,,). (19)
In approximate thermal simulation the goal is to obtain approximate simi-

litude of temperature fields of the fluids, and consequently also, of the heat-con-
duction surfaces. Knowing the temperature field of the heat-conducting surface /143
it is not difficult to find the temperature field inside the fuel element using Eqs.
(13) and (14).
The condition that q = const at the surface to which heat is supplied to the
model can be satisfied, for example, by using electric heaters. In some exper-
iments use was made of heaters made in the form of coils from nichrome wire
and insulated from the tube walls by a thin layer of organic silicon. The condi-
tion that t = const over the inner perimeter can, apparently be created by press-
ing or soldering copper rings inside the tubes.
Let us consider some particular fuel element designs. The value of F.kfor
a single-shell element is obtained from Eq. (15) by setting ^ 1 = ^2 = ^3 = ^0 and
X3 = X w
4 --mtph
X.
Ch Xf 1 + M- 2h (20)
where in Xw-X 0%'w + A 0; ^ 0 = R l/R2 , while R1 and R2 are the inner and outer
radii of the shell, respectively.
If the rod has a thick shell, i.e., if ^2k << 1, then the temperature field
(Tw) of the heat conducting surface is practically independent of the thermal
142

conductivity (k 0 ) of the heat-generating rod (of the boundary conditions at the

inner surface of the shell).
If the thermal conductivities of the rod and the shell are approximately the
same, i. e. , X0 sz^ Xw, then the temperature field of the heat-conducting surface
(Tw ) practically does not depend on the shell thickness, which also follows from
the solution of Eq, (12). In fact, the temperature level of the fuel element is de-
termined by the function 0, while the temperature distribution over the perimeter
is given by the function T, which is independent of the radius of the heat-generat-
ing volume Q 0) .
The above cases are equivalent to a heat-generating rod without a shell or

to a tube with sufficiently thick walls with internal heat supply. The simulation
conditions here become much simpler
(21)
C^f )f. e. Xf /mod' ( oh)mod 1.
It is easy to simulate fuel elements in which the thermal conductivities of the rod
and of the shell differ highly. When X 0 << X (m ^ 1), we get from Eq. (20)
N2w1 ' _2h

1+to h (22)
and when X0 >> X (m L-- -1)
Xw 1+gIh
s h ti
7,f 1goh (23)
These fuel elements are simulated by tubes in which we impose on the inner sur-
face the conditions q = const (22) or t = const (23) over the perimeter.
We shall assume that the fuel element has a very thin shell, i. e., that /144
S << R2 (6 is the shell thickness), while the value of m is moderate. Expanding
20k in a binomial series, we get
8h
o = C1- R2b) 21 , 1-2k R2 -f ...

It follows from Eq. (20) that
m C1-2kR /
N _
Ek
^w 1- 2 w 1-m 0
1. Xf 1+M Xf (24)
Xf I+M C1-2kR
2
i. e. , the effect of the shell can be disregarded. The criteria for shell thinness
in the given case is the fact that depends little on k, the number of the
harmonic. 8
If the thermal conductivities of the rod and the shell of a fuel element with
a thin -walled shell differ, i. e., the fuel elements are equivalent to thin-walled
tubes with q = const or t = const, then
8
I 1-1+2kR
_ 2 N w8
If k XfR2
Ek~ 1+1-2kR (25)
2
for N 0 << X or
1+1-2k a
R2 ;^,,^R2
= 1
Xf 1-1+2k k Jf8 (26)
R2
for X 0 >> Xw. In these cases simulation becomes appreciably simpler and re-
duces to selecting tubes and liquid-metal heat-transfer agents, which satisfy the
conditions
=
RfR = idemand' fa2 = idem
for q ;^:i const and t Rt^ const, respectively, at the inner surface of the tubes.
Method.s of approximate thermal simulation can be used in generalizing ex-
perimental data and for solving problems by the method of thermal-electric anal-
ogy.
Let us consider some examples of approximate simulation. For this we
use experimental data obtained with a Na-K solution (A AA f = 0.695) for rods
packed closely in a square arrangement, [2], an eccentric annulus [4] and for
a rod placed inside a hexagonal jacket [5]. The fuel-element models are tubes
heated from within by electric coils (q ^-_ const at R 1). The values of ^0 for the /147
144
equivalent fuel elements in all the cases were assumed to be 0. 92, which corre-
sponds to a jacket thickness. of 6 = 0.4mm and an outer radius of R 2 = 5mm.
The ratio of thermal conductivities Xw/^ f for fuel elements arranged in a closely
packed square in one case is assumed to be 1.25 while in the other it is 0.416.
These values of the ratio of thermal conductivities are obtained for cases when
the fuel element jacket is made from ferritic-pearlitic steel (X w = 25 kcal/m-
hour-degree), while the coolants are the Na-K alloy (a, f st: 20 kcal/m-hour-de-
gree) or sodium (A f ^ze 60 kcal/m-hour-degree), respectively.
For fuel elements corresponding to annuli and for those located in the cor-
ner of the shield, we have considered two cases
f =0.416 and ^f =0.2.
The value XW/" f = 0.2 is not encountered and is used in the calculations only in
order to extend the range within which the fuel element parameters are varied.
The values of m = X - 0/Xw + X 0 for the equivalent fuel elements were
found from the expression
1myo k0 1
7Lu, _

C ^f i+n ,. 2k0 / f, e. Ekp)mod
with the condition that for dense square packing k 0 = 4, while for the two other
cases k0 = 1.
The main parameters of the models and of the fuel elements being simula-
ted are tabulated in Table 1, which also lists the values of coefficients of the
Fourier series (ak) obtained by analyzing experimental data for the models.
The calculated parameters e and the temperature variations AT and

AT* calculated from Eqs. (16), (18) and (20) are shown in Figs. 2-4. Despite
the fact that the values of (k)mod and (P'. Of. e differ appreciably when k k k0 , /148
the [temperature] variations of the model and the fuel elements in the dense
packing arrangement do not differ significantly (see Fig. 2). A somewhat great-
er difference between ATw and AT* is observed for the annulus (see Fig. 3) and
for the corner rod (see Fig. 4), since , in these cases the values of (8Omod and
(s k) f. e differ more appreciably and, in addition, the character of curves de-
scribing the temperature field is more complicated.
The sought value of temperature variations is estimated as the arithmetic
mean of ATw and AT*.
145
1
2 D
C1
2
w
Oq
1 3 5 7 9 11 k mot+
w 1
r
0,4
/0.3 0 ^
aW
0.2 0, ^
y f ?w QyrÔ ^
Fes;
0,1 Ra
212
î
D Q
0.6
1.2 \ hi
-0.1 v
-0,2 0
-0.3
40 cO
2 ^
0 10 20 30
0,4
Figure 2, Fuel Rods Densely Packed 0 40 80 720 760 cO
in a Square Arrangement. 1 and 2 are
the Values of (:ek)mod/(e'k)f, e and TW Figure 3. Annulus With R0/R2 =1, 11
for Type 1 and 2 Fuel Elements; the With an Eccentricity of 0.5 (for Nota-
Dashed Line Denotes the Value of ATW. tion see Fig. 2).
TABLE 1, Basic Parameters of Models and Fuel Elements
Type of passage al as a3 a4 as
Formed by rods packed

densely in a square 0.362
arrangement
Annulus with an eccen- 0.572 0.217 0,0715 0.02 0.0177

tricity of 0.5
Corner rod, triangular 2.12 1,10 0.27 0.187 0,154

lattice
146

TABLE 1. (Concluded)
Type 1 fuel element) Type 2 fuel element
ae as Pe mod M
I "W I m
f XW
.f I Xf
0.041 100-500 0.773 0.778 1,25 0.246 0,416
0,006 82 0,942 0.970 0,41 F, 0,772 0.20
0.021 23 I 0,933 0,937 0.416 0.722 0.20
The error in determining [these] values
iwm ax_ tw max_ min

tw - tw
(AT.)max = - f und (ATz ) max (OT,)min= Xf,
2
which we denote by N and M respectively, are presented in Table 2,
TABLE 2. Errors in Determining the Var- As can be seen from this

iations in Temperature, % table, the accuracy of simula-
Type 1 fuel Type 2 fuel tion in the above cases does not,
Type of passage element element go past the precision limits of
N M
I N I M
I
experiments of this class. This
Space formed by densely 1 1 2 1
does not, however, mean that
packed rods any fuel element can be simu-
Annulus 2 1 10 5 lated with the same accuracy.
Corner rod 1 6 4 9 All depends on the relationship
between the parameters of the
fuel rod elements and the limi-
tations in the selection of heat
transfer agents and structural
materials for the models.
The methods of approximate thermal simulation presented in this article

are used successfully in the study of active [cores] zones of reactors with liquid-
metal cooling.
147
. 2.s
_w
Y
1, 8
7
0 7 3 5 7 3 11 k
4 180
S
1.^
A
2 ^P a }
1_k Z
d
^ G
v
0.
1.E
-2.5
^l
0 40 80 720 760 Ô
Figure 4. Corner Rod in a Triangu-

lar Lattice With s/2R2 = 1.04 (for No-
tation see Fig. 2).
REFERENCES
1. Subbotin, V. I. et al.: Atomnaya energiya, 9, 6, 461 (1960).

2. Subbotin, V.I. et al.: Atomnaya energiya, 13, 2, 162 (1962).
3, Subbotin, V. I. et al.: Teplos''yem v elementakh reaktorov s zhidkometal-
licheskim okhlazhdeniyem [Heat Removal in Elements of Liquid-Metal-
Cooled Reactors]. Report read to the 3rd International Conference on
Peaceful Uses of Atomic -Energy, Geneva, 1964,
4. Subbotin, V. I. et al.: Vliyaniye ekstsentrisiteta na teploobmen zhidkikh
metallov v kol'tsevom zazore [Effect of the Eccentricity on Heat Transfer to
to Liquid Metals in an Annular Space]. Part I of present collection.
5. Zhukov, A. V. et al.: Eksperimental'noye issledovaniye na modelyakh poley
temperatury teplovydelyayushchikh elementov [Experimental Studies of
Temperature Fields of Fuel Elements, Using Models]. Part II of present
Collection.
148
HEAT TRANSFER FROM LOOSELY-SPACED [FUEL] ROD CLUSTERS TO
LIQUID METAL FLOWING IN THE AXIAL DIRECTION
A. V. Zhukov, V. I. Subbotin and P. A. Ushakov
The present article considers the transfer of heat from loosely-spaced /149
[fuel] rod clusters in which the temperature variation over the rod perimeters
is small and the heat transfer coefficients depend little on the fuel element de-
sign (jacket thickness, thermal conductivity of the fissionable material and the
jacket).
Analyzing the differential equations describing the processes taking place
in the fissionable material, fuel element jackets and the coolant, together with
the appropriate boundary conditions, we can find the needed similarity criteria
which define the temperature fields and the average heat transfer in the rod
clusters.
t qR2"' X t = h C dh , X, (p, S,
Xf , Xff , Re, Pr)
Nu = fa ^S, ?f Re, Pr) .
CI, x, 7f
Parameters ^, X 0 f and Xw/X f which characterize the fuel element design have
their greatest effect on the temperature fields and on the average heat transfer
in densely-packed fuel-element clusters (x = 1), when maximum temperature
variations are observed over the perimeter of the latter, [1-6]. In clusters
with x 1.1 the temperature variations over the perimeter of the fuel elements
are moderate and the above parameters have little effect on the average Nusselt
number. In this case the Nusselt numbers calculated by using the hydraulic di-
ameter depend primarily on the physical properties and the flow regime of the
coolant as well as on the value of x.
Basic information on investigations concerned with heat transfer to liquid
metals in loosely packed rod clusters is presented in Table 1. Figure 1 depicts
the cross sections of clusters used in these studies.
The authors of [7-9] have used seven-tube clusters with spacers
fastened to a cylindrical shell. The tubes were heated internally so as to pro-
vide qw = const at the inner surface. The center tubes from St. 3 or copper
were equipped with thermocouples embedded into longitudinal slots near the heat
conducting surface. These tubes could be rotated through any angle. As a rule,
the total heat supplied to the coolant by the clusters was the quantity measured.
In individual cases the temperature was measured at the outlets from the central
cells of the cluster. The dependence of Nusselt numbers on the relative spacing
149
0
TABLE 1. Characteristics of Test Sections and Experimental Conditions [7-13]
I Reference and year of publication
Designation I [ 7] [8] [9] [10] [10-13]
1956 1 1958 I 1958 I 1959 1962 I 1963 I 1964 , 1961 i 19G1-1964
Coolant Liquid metals with Pr = 0.025-0.036 Na llg Hg, the Na-K

alloy
Material of the heat- Carbon steel 1Kh18N9T Cu Cu, nickel- or

conducting surface S tee 1, St. chromium-plated
3 Steel surfaces
Method of heat Heating by passing current through an inside tubular heater Electric heating by
suPP1Y nichrome coils
amber of thermocouples 4 7 6 5 12 9
>n the center tube
Copper and constant

Description of thermo Chromel-alumel thermocouple wires 0.18-0. 3 mm in diameter in two-
couples and the -I passage porcelain tubes along the inner surface or in metal capillaries thermocouple wires
i n a c mm in diameter
manner in which 1.0 X 0.15 mm in diameter along the outer surface with subsequent
a capillary 1.09 X
they are attached metallizing or electroplating of the surface
X 0.74 mm in diameter
Calculated Calculated Calculat- Measured

vIethod for determining theoretically ed theo-
theoretically
the temperature in the and retically
center cells measured
description of thermo- _ Chromel-alumel thermocouple

couples used for wires 0.18 mm in diameter
measuring the liquid's in capillaries 1.0 X 0.15 mm
temperature in the in diameter
center cells
i
Coolant velocity V, I 0.18-1.0 0.16 0.9 0.0488-
m/sec 2.02 0.59 0.61
Re 10- 3 I I 10-48 ( 3.8-57 4.3-28 I 8-80 I 7.5-200
Peclet number I 500-1800 I 300-1450 30-345 128-172 1190-17001 150-4000
' I
Prandtl number 0.036-0.025 3.10-2 (72-57) X 64.10- 4 0.016 0.02
X 10- 4 0.021
Coolant temperature, 200-250 210-310 220-270 49-127 38

t f, 0C
Thermal flux, q, (50-100).103 36 700 (66.5-100).103

kcal/m2 hour
Metal purity, oxygen was not checked 2.4.10-4 1.7.10_2

content by weight,
d,, mm 1 17 I 7.6
d2, mm 22 ( 12.7 I 12.7
j
I
a= I 1.14 I 1,3 1 1.2 I 1.5 I 1.38 I 1.75
L, mm I 600 68 5 I 1510 1240 I I 1712

10, mm ( 50 0 I 600 1050 700 I 1016 I 1370
Limits of rotation -10 -{ 10 + 10+50 held at 0-360 in steps 0-360 was not rotated
of center tube, cp , +35 ( +70 30 of 30
r de gre es
cn
r
of the rods was not established
[8], [9]. To calculate average
Nusselt numbers in clusters with
x = 1.2-1.75 [8] recommends the
empirical formula
N N eD
o
Nu = 6 + 0.006Pe
for 30 _ Pe < 4000andRe > 10-'. (1)
to t[y N
00 00
CO CO 00
In [10-13] use was
made of clusters consisting of
13 and 19 heated tubes, held in
CO a cylindrical shell. The shell
diameter was selected so as to
keep the hydraulic diameters of
the cluster and of the central
CO
cells equal to one another. Ther-
0 m mocouples
CO were embedded into
longitudinal slots on the surface
of the copper center element.
This element was chromium or
_ nickel plated in order to obtain,
Q U respectively, a mercury-wetted
o
U ^
w o or nonwetted surface. The cool-
z ant temperature in the cells of /154
06 N o the test section was not mea-
w sured, and the center tube was
a
as not rotated during the study of
H temperature fields.
Dwyer and Tu [14]

have analytically calculated the
heat transfer coefficients in rod
n ' clusters with 1.3 :!^- x <_ 3.0. The
method used in calculations was
based on replacement of the hex-
agonal boundary around the rod
o cluster [apparent error, should
c be rod cluster/cluster of rods,
41 5 DBL] by an equivalent circle.
The calculations were thus re-
" " U CO Q duced to solving the problem
A of heat transfer in annular pas-
y sages. The velocity distribution
x was calculated by the method
due to Rothfus [15]. The final
formula used in calculations has
the form [12, 13].
152

Nu = 6.66 + 3,126x + 1.184x 2 + 0.0155 (^ Pe)0.86

102 <Pe<101 ; 1,3<x<3.0. (2)
The effective dissimilarity factor for the turbulent heat transfer and the momen-
tum are determined, according to Dwyer's estimated, by the expression
=1 1.82
g y 1.4
Pr ( v )max
A solution similar to that given in Eq. (2) was obtained by Friedland and
Bonilla [16], using data obtained by Knudsen and Katz [17] on the velocity dis-
tribution in annular passages
Nu= 7+ 3.8x1'52+0,027x'27 ( Pe)o.s^

1.38-x<10. (3)
Equations (2) and (3) yield close results and show that, when the hydraulic diam-
eter is used in analysis, the Nusselt numbers in the clusters depend on the rel-
ative spacing of the rod lattice.
Experimental Clusters and Methods Used in the Experiments
Experiments using mercury were performed on three identical stands

(Fig, 2), while those using the Na-K alloy were performed on the setup described
in [18]. The mercury (brand R3 ) and the Na-K alloy were circulated by centri-
3
fugal pumps with an approximate delivery of up to 10m hour with a pressure of
about 18 kg/cm2 and with a delivery of 150 m 3/hour at a pressure of about 6 kg/
/em2 , respectively, Impurities were continuously removed from the coolants
during operation in the following manner: the mercury was filtered through suede,
while the impurities from the Na-K alloy were removed in cold traps. The mer-
cury flow rate was measured by precalibrated Venturi tubes connected to U-
shaped differential manometers, while the flow rate of the alloy was measured
by electromagnetic flowmeters. The coolant flow rates were adjusted by bel-
lows-type valves. All the components of the test stands were made from
1Kh18 N9 T steel.
Experiments with mercury were performed using clusters with x = 110,
is 20,1 . 30, 1.40 and 1.50. The clusters were made up of seven heated tubes
and twelve rods installed into appropriately shaped displacers [ 1], Experiments /155
with the Na-K alloy were performed using a cluster with x = 1. 15, consisting of
153
37 heated tubes contained in a hexa-
gonal shell [18]. The tubes were
000 heated internally by nichrome coils,
which provided qW - const at their
inner surfaces. The coils were in-
dZ
* sulated from the walls by a layer of
electric insulation - 0.3-0.4 mm
thick. The center tubes were equip-
ped with 12 thermocouples placed in
capillaries from 1Kh18N9T steel
with an outer diameter of 0, 5 mm,
The thermocouple wires were made
from copper and constantan in chlo-
(S'B
_8?'O rovinyl insulation (for experiments
with mercury) and from chromel,
kopel and alumel in alundum and
organic silicon insulation (experi-
ments with the Na-K alloy) . The
Figure 1. Sections Through Clusters
thermocouple wire diameters were
Used in [7-9] (a and b) and [10] and 0.08 mm, and the junctions were
[11] (c and d), -calked at the . tube surface.
The center tube was inserted
l
d , mm 3
d , mm I D, mm S;d: into the cluster by means of a pack-
ing gland, which made it possible to
a I 17 ( 22 92, 63 11,2-1,5 rotate it through any angle without
disturbing the operating mode of the
b I 17 I 22 I 96 ^. 1.13
setup. The thermocouple capillaries
were held hermetically in the pack-
C I 7,6 I 12.7 ( 66 I 1133
ing gland in a round bore made in the
d I 7.6 I 12,7 I 101 I 1,75
wall of the tube and filled with the
AKR,-7 plastic. The coolant temper-
ature was measured at the outlet from
each cell by thermocouples in capil-
laries 0.5 or 0.8 mm in diameter, as
well as in the cluster headers by ther-
mocouples in sleeves.
Table 2 shows the main geometric dimensions of the test clusters. The
clusters had a moderate number of heated tubes, hence, the temperature field of
the coolants over the cluster cross section was nonuniform due to the effect of /156
nonheated displacers and jackets. The experimental data was analyzed on the
basis of the coolant temperatures in the center cells; in this manner they canbe
used for thermal calculations for clusters with a large number of heat-produc-
ing rods (for an infinite lattice). We have also studied the temperature field of
the surface of the center tube with the coolant flowing at different velocities.
Here the temperature of the medium at the inlet into the cluster and the thermal
flux were maintained constant. The tube was rotated from cp = 0 to co = 3600 in
steps of 5-100 for clusters with x = 1.1 and 1. 15 and in steps of 30 and 60 0 for
clusters with x = 1, 2, 1. 3, 1.4 and 1.5. The mean temperature differences be-
tween the wall and the fluid were measured at different distances from the inlet.
154
J
Figure 2. Schematic Diagram of the Setup for Mercury

Experiments. 1) Filter; 2) Montejus; 3) Filter; 4) Pump
Tank; 5) Mercury Tank; 6) Pump; 7) Heat Exchanger;
8) Water Inlet and Outlet; 9) Cooler; 10) Damping Tank;
11) Flowmeter.
The stabilized heat transfer coefficients were calculated from the expression /157
a = For clusters with x = 1.1 the temperature variation over the perimeter
of the center tube was estimated. Here the thermocouple readings were aver-
aged and reduced to the reduced angles co = 0, 5, 10, 20 and 300.
In analyzing the experimental results the values of the physical parameters
of 1Kh18N9T steel, of the mercury and of the Na-K alloy were taken from [19]
and [20]. In some cases temperature fluctuations of the heat-conducting surface
and of the coolant were recorded. For this purpose use was made of the EPP-09
rapid-acting, single-point electron potentiometers with a scale of 1 mV and re-
quiring 1 sec in order for the carriage to traverse the entire scale. Sections
through the clusters are shown in Fig. 3.
Experimental Results
Fluctuations in the temperature of the heat conducting surface and of the

coolant were observed in experiments using clusters with small relative spacings
(x ^_- 1.1). To reduce the resistance to flow, a grid, with a large flow cross sec-
tion, approximately equal to the flow cross section of the cluster, was placed at
the cluster inlet. Here particularly high temperature fluctuations were observed
in the center cells. The amplitudes of these fluctuations became smaller with an
increase in the coolant velocity. The temperature fluctuations in the wall of the
center tube increased as the distance from the start of the heating zone became
larger.
155
TABLE 2, Main Geometric Dimensions of Test Clusters
Designation and
units Numerical values
Relative spacing of rods,` 1.1 1.1 1.15 1.15 1,2 1.3 1,4 1.5
X = S/2112
1
Tube length L, mm ( 770 1495 I 750 1 891 ( 770 1 891 1 770
Distance from the butts of 30 40 50 40 50 40

tubes to the start of the
heated zone., 1st, mm
(
Lingth olf heated zone, 710 445 690 800 700 800 1 700
0, I I I
Outer diameter of tubes, 17.6 17 17.6 16 17.6
d2 = 2R2 , mm
Inner diameter of tubes, 13.6 13 15,4 13.6 12.5 13.6

d l = 2R 1 , mm
f
Flow cross section of the 40.7 52,06 71.5 105.4 117,5 180
cell w, mm2
Flow cross section of the 978 4300 1720 2530 2820 4314
cluster 0, mm2
I-
Hydraulic diameter of a 5,9 7.80 10.3 15,2 18.6 26.1
cell in the cluster,
d = 4w/p, mm
s-^
z
Figure 3. Sections Through Experimental Clusters.
156
Temperature fluctuations observed in the cluster with x = 1.1 are produced
by fluctuations in the flow rates of the coolant in the cells, while the total flow
rate through the cluster remains constant. In addition, the temperature fluctua-
tions are brought about by the temperature gradient over the cluster cross sec-
tions. Experiments show that heating of the shell of the above cluster with x =
= 1.1 practically does not affect the temperature fluctuations in the center cells. /158
The fluctuations in the coolant temperature in extreme cells do increase in this
case.
Thus, the main cause for temperature fluctuations in this case is the lack
of stability in the [individual] flow rates and, consequently, in the quantity of heat
transferred to the coolant in cells of the cluster. After the inlet resistance has
been increased, the temperature fluctuations dropped appreciably.
Another picture was observed in clusters with a less tight spacing of tubes
(x > 1. 2), where the coolant temperature fluctuations were at maximum in the
intermediate, rather than the center cells. Apparently, this is due to intermix-
ing of hot and cold fluid streams within the cluster's cells. In experiments with
a cluster with x = 1.2 the flow cross section of the cluster was by a factor of
2.5-3 larger than the line cross section of the inlet centering grid.
Temperature fluctuations were also studied in experiments with the Na-K
alloy using a cluster with x = 1.24 consisting of 37 heated tubes (Fig. 4). For
the same coolant velocity the fluctuations in the temperature of the heat-conduct-
ing surface increase with an increase in the distance from the start of the heated
section. The behavior of temperature fluctuations as a function of the rate of flow
of the Na-K alloy is remarkable. When the alloy flow is turbulent (Re = 5200) and
q = 15,500 kcal/m2 -hour the temperature fluctuations tmax t min comprised
about 0.3--C, while when Re = 3470 they reached the maximum (approximately
1C). When the velocity was reduced further the fluctuations decreased. Final-
ly, in a purely laminar flow (Re = 740) the fluctuations practically disappeared.
The behavior of temperature fluctuations in the Na-K alloy in the center cells
was similar.
The behavior of changes in the fluctuations shown in Fig. 4 can be explained
as follows. An increase in the fluctuations over the length of the cluster is due to
an increase in the amount of heat transferred to the coolant in the cells. Low-
frequency fluctuations at high Reynolds numbers apparently are produced by a
certain amount of instability in the flow rates of the coolant through the cells
which, in its turn, is brought about by the conditions at the inlet into the cluster,
by the less than uniform operation of the pump, as well as by variations in the
power supply. High-frequency fluctuations with a low amplitude are apparently
related to the turbulent flow velocity fluctuations, which disappear in laminar
flow. Maximum temperature fluctuations naturally take place in transient flow.
Thus, the magnitude of fluctuations in rod clusters depends on the inlet
conditions and on the resistance of the inlet grid. When the inlet is smooth and
the dynamic resistance at the inlet is low as compared with the resistance tothe
flow of fluid of the grid and the cluster,, the temperature fluctuations are prim-
arly determined by the flow turbulence. They become larger with an increase
in the thermal load. In reactors operated under high thermal loads fluctuations
157
01
OD
s
S
v
f 4
I 3
1
_! -I _
1 + ^-1- 1 ^ ^ I Y I ^ _ _ - - -
a b c d e f
Figure 4. Fluctuations in the Wall and Coolant Temperatures in a Cluster With x = 1.24.
w, m/sec Re Pe w, m/sec Re Pe
a 0.348 5200 157 d 0.178 2840 78
b 0.281 4250 125 e 0.047 920 19.3
c 0.224 3470 99 f 0.03 740 11.6
Fluctuations in the wall temperature of the center tube

1 2 3 4 5) fluctuations in the coolant temperature in center
cells.
Z/dh 0 20 27 47.5
in the temperature of the heat-conducting surface and of the coolant may reduce /160
the operating reliability of the fuel elements. From this point of view it is un-
desirable to have flow transition (from turbulent to laminas) of the coolant in a
reactor, for which the temperature fluctuations are at maximum.
A characteristic temperature distribution over the perimeter of center
tubes of clusters with x = 1.1 and 1.5 (original data) is shown in Fig. 5. For
x = 1.1 temperature bursts are observed in narrow spots of the cells, (cO = 0, 60,
1200 , etc,) produced by coolant overheating in these places. However, the ef-
fect of the above variability is compensated for by the effect produced by the
bending of tubes of the cluster, which results in redistribution of the coolant
flow rates among the individual cells, as a result of which the amount of heat
transferred to the coolant in different cells is nonuniform. Hence, the variabil-
ity in the temperature due to the bending of tubes is increased as the section at
hand is taken further away from the inlet and reduction in the coolant velocity
(i. e. , increasing the quantity of heat transfer).
For tightly-packed clusters (x =1-1. 1) the temperature difference ( w tf)
varies.over the perimeter much less than the wall temperature, since the behav-
ior of temperature distributions in the wall and in the coolant are approximately
the same. This follows from analysis of experimental data.
Starting with the condition of AP = idem for all the cells and assuming that
the friction coefficient f R^ Re-0.25 we can write an expression for the rate of
flow by volume of the coolant through a cell of the cluster
3
Cu 1.75
V- 1.25
P1.75
The relative variation of the wall temperature
8tw
N d In (tf+0).
W
Hence, we get
8tw_ 8tf
1 1.
80 814 (4)
Increase in the coolant temperature in a cell when q w = const
3
i-75 P
(tfto)_ply r. Cw) (5)
159
4,v I,mrn tMM
41
L L I I filf 38 650
30
2o^
34 42 JL I
10
4049 4199 38 1;4#4#
2,9 -,4 ::F
32 4; 1 1 1 1 ^
28 330 36 500
24
L
40 40
I IJ- I L I L-1.
--1 .1 J 1
26 250 N 36 400
22
39
24 M M A M3,10
20 N M
20 M
L- L I I I-
22- - 190 32 13,0
78
Iw -7 71
=L 22
t Ht
ff-- --
i
29
24-7 24 A
20-- 21T so
26 ------
0 M WO 190 M 300t0,deg
a
0 M 120 710 240 309 9,deg
b
0

0 120 I'M 240 300 Sp, deg
c
Figure 5. Distribution of Temperature Over the perimeter of Center Tubes

of Clusters. a) and b) for x = 1. 1, q = 50, 000 kcal/m2-hour and W = 10. 4
and 0. 43 m/sec; c) for x = 1.5, q = 43,400 kcal/m2 -hour and W = 0. 46 m/ se c.
Making use of the above expression, we get
St f _ 8 (t f to) 12 1 Nudp 8^
$ $ N 1.75 ' Wh ' Pe ( p 0) ) ; (6)
The variation in the thermal flux over the perimeter of the tubes, which is due
to the difference in the amount of heat imparted to the coolant in the different
cells does not exceed 4 %. Hence, it may be assumed approximately that q = const,
In this case /162
8^
p N8a
^^
V "' a
To find 6 a/a on the basis of experimental results we have assumed that a NW0,450
Then
/
! . 25.0.45
\ 0,45 N P ^` 1.75
aN (
V
We write the relative variation of the mean temperature difference
81D _ _ 8a1,25 . 0.45 dp_66)

$ a = 1.75 C p 0) ' (7)
It follows from the above expression and from Eqs. (4) and (6) that
8t'V N 12 1 Nu
1. (8)
a$ 1.25.0.45 dh Pe +
For a cluster with x = 1. 1, q = 50, 000 kcal/m 2 -hour, W = 0, 43 m/sec and l =

= 400 mm, we get
Nu -2 1 68 btw N 12.0.68
-Fe- `" 10 ; dh 80 1.25.0.45 ' 1=15.
Thus the variation in the wall temperature (6tw) is approximately by a fac-

tor of 15 greater than the variation in the mean temperature differences (6).
161

0.04 For clusters with a large relative

spacing the effect of tube bends is only
moderate, while temperature variations
with a period of 600 , which is due to the
1. 0.02
cell's geometry is negligible (See Fig. 5).
Figure 6a shows a periodic tem- /163
0 perature variation obtained in experi-
ments for a cluster with x = 1.1. The
temperature distribution over the tube
-0,02 perimeter is close to the sinusoidal
twtw ti A cos 6cp.

-9.0^ ^w (9)
qR2
0 70 20 (0, deg
The moderate difference in the val-

ues of A for two clusters with x = 1.1 for
the same coolant velocities are within the
limits of experimental error (Table 3).
The distribution of thermal fluxes
7.05 1.7 775 S1d2 over the tube perimeter can be found by
using the solution to the problem of ther-
mal conductivity of a tube in the cluster
Figure 6. Dimensionless Temper-
presented in [1].
ature Distribution in a Cluster With
x =1.1 (a), and the Effect of the Rod
Spacing on Variations in the Ther-
mal Flux (b). 1, 2, 3 and 4 are for 4 =1-B cos 6y, (10)
4
W=0.12, 0.24, 0.66 and 1.04m/sec
Respectively; 1 = 400 mm; B is a where
Coefficient in Eq. (10).
B=6A
1g"
1+^,2
The temperature variation in a cluster with x = 1.1 is moderate and for the range
of velocities under study is by a factor of 5-10 smaller than in a densely packed
cluster of rods [1, 2]. This is not dangerous from the point of view of fuel el-
ement overheating. For example, for fuel elements, the equivalent thermal con-
ductivity of which is equal to the thermal conductivity of tubes of the clusters un-
der study (k ^ 14 kcal/m-hour-degree), the temperature fluctuation w w
r calculated for q = 106 kcal/m2 7 hour comprises not more than ^ 10-15C. Co- e
efficient B [Eq. (10)] as a function of the relative spacing x is shown in Fig. 6b.
The variation in thermal fluxes and, consequently, also in the temperatures
decreases rapidly with an increase in x. In experiments with clusters with x =
= 1.15 these fluctuations were practically nonexistent even when the coolant flow
was laminar.
162
TABLE 3. Numerical Values of Coefficients A in Eq. (9) for
Clusters With x = 1,1
1 0 =710 mm
W, m /sec I 0.42 0.49 I 0.60 ! 0,85 I 1,04 I 1,2
A I 0.025 I 0,020 I 0,015 I 0.013 I 0.0115 I 0,0101
1p = 445 mm
W, m /sec ! 0.12 I 0.24 I 0.31 I 0.43 1 0.52 ! 0.66 1 0.83 ! 1,04

A I 0.038 I 0.032 I 0.032 I 1
0.026 1 0.0247 0.01751 0.016 I 0.0137
Figure 7 shows a characteristic

distribution of the temperature over
the length of a cluster. The coolant
temperature in the center, intermed-
IA C. C. iate and extreme cells is different. If /164
tf, i. C. the characteristic temperature in this
case is tf, the average coolant temper-
"tf
ature over the entire cluster cross sec-
tf,, e. c. tion, rather than tf,cent,cell' then the
mean temperature differences between
the wall and the coolant will come out
0 200 400 600 Z,mm to be much greater. This again em-
phasizes the importance of taking into
Figure 7. Temperature Distribution account the variation in the coolant
Over the Length of a Cluster With = temperature over the cross section /165
= 1. 1 When W = 0, 42 m/sec and q = of clusters with a moderate number
of rods (tubes). The manner in which
= 25, 200 kcal/m2-hour, t c, c' tfi, c the heat transfer varies over the length
and tf ex are the Average Mixed of the clusters is shown in Fig. 8. Sta-
bilization of heat transfer in loosely-
Temperatures of Mercury at the Inlet packed clusters ensures with lower
From the Center, Intermediate and values of 1/d h' than in tightly-packed
Extreme Cells, Respectively; tf is ones. This can be explained as follows.
the Average Mixed Temperature of First, the formation of the velocity and
Mercury in the Header, the Circles temperature distributions in tightly-
Denote the Temperature of the Wall packed clusters is made difficult by
of the Center Tube, the narrow parts of the cells. Secondly,
163
the hydraulic diameter, which is by
convention used as the characteris-
tic dimension becomes appreciably
larger with an increase in the rela-
tive spacing.
It is interesting to note that,
for the same flow velocity of the
mercury and the same tube diam-
eters, the relative spacing practi-
cally does not affect the magnitude
0 20 40 60 l/dh
of a (Fig. 9). For this reason the
experimental data in terms of sim-
1
Figure 8. Variation in Heat Transfer ilarity criteria calculated on the

basis of the outer diameter of tubes / 166
Over the Length of Clusters. 1) x (the tube diameters in all the clus-
=1.4, 1.5;2)x=1.2, 1.3;3)x=1.1 ters are the same, see Table 1),
(1 0 = 445 and 710 mm) . for 1. 1 < x c 1.5 are in good agree-
ment with one another can be ex-
pressed as (Fig. 10)
Nu = 0.58 Pea' 45, (11)
where d2 , the outer diameter of the tube is taken as the characteristic dimen-
sion. When using this approach in analyzing the experimental data, the data ob-
tained using mercury and the Na-K alloy in developed turbulent flow are in good
agreement with one another. The experimental points can be approximated by
Eq. (11) with a scatter not exceeding 15%.
If, in Eq. (11), we take the hydraulic diameters of the center cell as the
characteristic linear dimension, then it will take on the form
Nu=0.58 (1.1x2-- 1)0.55 .pe0.45

1.1 < x < 1.5; 400 < Pe < 4000. (12)
Figure 11 compares experimental and theoretical data obtained by various

authors, [8], [9] and [13]. The data of [10] obtained with a cluster with x = 1.38
were not used since they should be regarded as qualitative [only]. The above
reference notes an appreciable effect of the cluster's geometry and bends of the
rods on their temperature fields. The reproducibility of experiments when re-
placing the working tubes in the cluster with x = 1.38 was poor. It can be seen
that the experimental data of [3] break up into smooth distributions for each of
the relative rod spacings (x). This behavior of points is not observed in exper-
imental data of [8] and [9].
164
v
Nom^++
Fi
ti
U
Cd
xU
n-
0
0 0.2 0.4 0.6 0.8 1.0 12 Wm/sec
Figure 9. Heat Transfer Coefficients as a Function of the

Mercury Velocity. (3, 9, o., + and Denote Clusters With
x = 1. 1, 1. 2, 1. 3, 1.4 and 1.5 Respectively d 2 ^ 17 mm.
A) kcal/m2-hour-degree.
Nud Figure 11 displays dash-

ed curves of Nu = f(Pe), cal-
culated by Dwyer. The theo-
20 retical curve D for the cluster
with x = 1.75 was taken from
[13], the theoretical curves,
t0
C, B and A for clusters with
x = 1. 5, 1.2 0 and 1.1 have
8 been taken from the article by
6 Dwyer, devoted to analytical
calculation of heat transfer
Figure 10, Curve of Nud = f(Pe )d Obtained to liquid metals flowing in rel-
atively tightly packed rod clus-
by Using the Outer Diameter of Tubes as ters. As can be seen, the re-
the Characteristic Dimension, o - Cluster sults of these calculations show
With x = 1, 15 (d 1 = 15.4 mm, d2 = 17 mm) that the Nusselt numbers de-
pend appreciably on the rela-
for the Na-K alloy; 4, , o, + and are tive spacing-of the rods. Thus,
Clusters With x = 1. 1, 1. 2, 1. 3, 1.4 and
1. 5, Respectively for Mercury. when Pe = 10 3 the calculated
Nusselt numbers in clusters
with x = 1. 1 and 1. 75 differ
by approximately 60 %. The
numerical values of the theoretical relationships Nu = f(Pe) in clusters with x =
= 1.2 and 1.10 lie above the corresponding experimental data for turbulent flow
by 60% on the average.
Of great interest are heat transfer in rod clusters in the laminar flow of
the coolant. The distributions of temperature and of the thermal fluxes over the
perimeter of the fuel elements, as well as the average Nusselt numbers in lam-
iner flow can be calculated analytically. Reference [21] presents a corresponding
solution for the boundary condition w = const over the perimeter of the fuel el-
ements. The Nusselt numbers have been calculated primarily for loosely-spaced
165
U

2 'k^^ 0000
D _Q a a, ^0e^rx's^` ô
_+ o q
0
2 o ^^ oo A ^
0
e 1 0 s +o' o 0 0
4 LA
2 4 6 8 ^p 2 4 6 8 10 z 2 4 6 8 193 2 Pe
Figure 11. Curves of Nu = f(Pe) Using the Hydraulic Diam-

eter as the Characteristic Dimension. 1, 2, 3, 4, 5, 6 and 7
are Calculated Values of Nusselt Numbers for Laminar Flow
in Clusters With x = 1. 1, 1. 15, 1. 2, 1. 3, 1. 4, 1.5 and 1. 75,
Respectively, [4]. EJ and o are Clusters With x =1. 15 (dl =
= 13 mm, d 2 = 17 mm, d 1 = 15.4 mm, d2 = 17 mm, Respec-
tively) for the Na-K Alloy; co, a., o., + and a are Clusters
With x = 1. 1, 1. 2, 1. 3, 1.4 and 1. 5, Respectively for Mer-
cury; e and x Denote a Cluster With x = 1.75 Using Mer-
cury and the_ Na-K Alloy, Respectively, [13]; o and are
Clusters With x = 1.2 and 1, 5, Respectively, for Sodium,
[8] and [9], A, B, C and D are Curves Obtained Theoreti-
cally by Dwyer, [13], for Clusters With x = 1. 1, 1. 2, 1.5
and 1. 75, Respectively.
clusters (x 2!^ 1. 1). Reference [4] gives a general analytical solution of thermal
conductivity and heat transfer equations in the fissionable material, jacket and
coolant for any boundary condition. The Nusselt numbers have been calculated
for clusters with x 1-1.5 for two limiting boundary conditions t w = const and
q = const. This solution proves that the boundary conditions and the relative
w
rod spacing have an appreciable effect on the average Nusselt numbers. The
solid lines in Fig. 11 are for theoretically calculated values of Nusselt numbers
for laminar flow, taken from [4]. The Nusselt numbers for clusters with x = 1.1,
1. 15 and 1.2 were obtained by taking into account the properties of the tubes (wall
thickness and its thermal conductivity), used in the experiments. The values of
Nusselt numbers for these clusters lie between the limiting values of Nusseltnum-
bers calculated on the basis of the boundary conditions q w = const and t = const
over the perimeter of the tubes. Since in clusters with x = 1.2 the properties of
the tube have almost no effect on the temperature fields and on the heat transfer,
the theoretical values of Nusselt numbers for qw , = const and t w = const are equal
to one another.
The average heat transfer in laminar flow of the Na-K alloy was studied us-
ing a cluster with x = 1. 15, consisting of 37 heated tubes 17 x 2 mm in diameter.
166
The methods used for performing the experiments and processing the data were
the same as in the study of heat transfer in turbulent flow. However, in the re-
gion with very low Peclet numbers, when large temperature gradients and, con-
sequently, axial thermal fluxes through the coolant and in the wall were appre-
ciable, appropriate corrections were introduced into the mean temperature differ-
ence between the wall and the coolant for the region with stabilized heat transfer,
using the formula /168
dr
of C 1 + %fwf
AO= Pe (13)
where At is the increase in the coolant temperature, w and wf are the cross sec-
tional areas of the tube and of the coolant flow, respectively.
As can be seen from Fig. 11, the experimental points for laminar flow in
a cluster with x = 1.15 are clustered about Nu ,z:i 5, which does not depend on the
Peclet numbers. This value
of the Nusselt number is shown
in Fig. 12 for comparisonwith
the corresponding results ob-
tained theoretically in [4]. This
figure also displays average val-
ues of Nusselt numbers for lam-
inar flow in clusters with x =
= 1. 2, 1.5 and 1. 75, taken
from [8], [9] and [13]. The
value of Nu for the cluster
with x = 1.5 [9] lies below
the corresponding theoreti-
cally obtained value by ap-
proximately a factor of 2.7.
In clusters with x = 1. 1, 1. 2,
1. 3, 1.4 and 1.5 experimen-
to 1.2 1.4 46 tal data were also obtained on
the average heat transfer for
Figure 12. Comparison of Experimental Data the transient (from turbulent
on Nusselt Numbers With the Analytically Ob- to laminar) flow of mercury
tained Relationships of Nu = f(x) for Laminar (up to Re - 3500). However,
Flow, Taken From [4]. 1) w = const; 2) cLx, these data turned out to be on
the lower side as compared
= const, Over the Perimeter of Fluid Elemem. with theoretically calculated
(for Notation see Fig. 11). Nusselt numbers for laminar
flow.
This may be ascribed to the fact that the design of clusters with x = 1. 1, /169
1. 2, 1. 3, 1.4 and 1.5 did not sufficiently satisfy the features peculiar to
167
heat transfer in laminar flow. - One may note the following features of heat trans-
fer of laminar flow in fuel element clusters: 1) an extended temperature distri-
bution at the outlet of the center cells, due to the insufficient rate of coolant in-
termixing; 2) axial and radial, thermal fluxes, which increase with a reduction in
the Peclet number; 3) reconstruction of the velocity distribution over the cross
section of the cluster in the case of transient flow; 4) presence of large temper-
ature gradients over the heated length, which is responsible for the appreciable
effect of errors in determining the location of the start and end of the heated
section. All this makes difficult the experimental determination of the average
mixed temperature of the coolant in the center cells of the cluster and as the end
results reduces the accuracy of experimental data on the average Nusseltnumbers.
The experimental data obtained with clusters with x = 1. 1, 1. 2, 1. 3, 1.4
and 1.5 in turbulent flow should, when extrapolated to the region of low Peclet
numbers, join the corresponding theoretical values of Nusselt numbers for lam-
inar flow. From this point of view it makes sense to generalize the data presented',
in [5] , which yield continuous curves of Nu = f(Pe) for turbulent, transient and
laminar flows of the coolant.
The empirical formula (12) can be used for calculating Nusselt numbers
for developed turbulent flow. It should not be extrapolated to region where
Pe < 400.
NOTATION
t0 coolant temperature at the inlet to the cluster, C

A-WAw'Xf thermal conductivities of the fissionable material, fuel-
element jacket and the coolant, respectively
e v turbulent viscosity
V volumetric flow rate of coolant
d 2 = 2R2 outer diameter of tube
d 1 =2R 1 inner diameter of tube
S distance between tube centers
P wetted (which is the same as heated) perimeter of the
cell angle of rotation of the center tube
cp deflection angle of central tube
1The design of the clusters was developed to determine temperature field

and heat transfer in developed turbulent flow of the coolant.
168
x = S/2112 relative rod spacing

^ =R I /R 2 relative wall thickness of the tube
REFERENCES
1, Subbotin, V. I. , et al, : Atomnaya energiya, 9, 6, 461 (1960). /170

2e Ushakov, P. A., et al.: Atomnaya energiya, 13, 2, 162 (1962)
3. Subbotin, V. I., et al.: Teplootdacha v elementakh reaktora okhlazhdaye-
mykh zhidkim metallom Heat Transfer in Elements of Liquid-Metal
Cooled Reactors]. Report presented at the 3rd International Conference
on Peaceful Uses of Atomic Energy. Geneva, 1964,
4 Ibragimov, M. Kh. and Zhukov, A. V. : Atomnaya energiya, 1836, 630 (1965),
5 m Ibragimov, M. Kh. and Zhukov, A. V : Atomnaya energiya, 20, 5, 425 (1966).
6. Buleyev, N. I., et al.: Teplofizika vysokikh temperatur, 2, 5, 749 (1964).
7 Borishanskiy, V. M. and Firsova, E. V. : In the collection Zhidkiye Metally
[Liquid Metals]. Gosatomizdat, Moscow, 1963.
8 Borishanskiy, V. M. and Firsova, E. V. : Atomnaya energiya, 14, 6, (1963)
9. Borishanskiy, V. M. and Firsova, E. V. Atomnaya energiya, 16, 5, (1964),
10 Friedland, A. J., et al.: Trans. ASME, III, 526 (1961).
11 Nimmo, B. and Dwyer, O. E: Trans, ASME, 312 (1965),
12, Maresca, M. W. and Dwyer, O, E.: Trans. ASME, 86, 180 (1964),
13. Dwyer, O. E. and Lyon, R. N.: USA report No, 225 presented to the 3rd
International Conference on Peaceful Uses of Atomic Energy, Geneva, 1964,
14. Dwyer, O. E.: T u P a S. Chem Engng Progr, Sympos Series, v56, 30 (1960).
15. Rothfus, R. R. , et al.: A. I. Ch, E. Journal, 4, 240 (1958)
16. Friedland, A. I. and Bonilla, C. F. : A. I. Ch. E. Journal, 7, No, 1 (1961).
17. Knudsen, J. G. and Katz, D. L.: Proceedings of the Midwest Conference
on Fluid Dynamics, University of Illinois, Urbana, III, 1950, 175-203,
18. Zhukov, A. V, et al.: Eks.perimental?noye issledovaniye na modelyakh
poley temperatury teplovydelyayushchikh elementov [Experimental Study
of Temperature Fields of Fuel Elements, Using Models], Article immedi-
ately following this one in the present collection.
19, Neymark, B. Ye.: Teploenergetika, 3, (1955).
20 Nikol I skiy, N. A. , et al : In the collection Voprosy teploobmena [Problems
of Heat Transfer, pp, 193-1991. Published by the Academy of Sciences
of the USSR. 1959.
21, Sparrow, E. M., et al : Trans. ASME, 415 (1961),
169
EXPERIMENTAL STUDY OF TEMPERATURE FIELDS OF FUEL
ELEMENTS, USING MODELS
A. V. Zhukov, L. K. Kudryatseva, Ye. Ya. Sviridenko, V. I. Subbotin,
V. D. Talanov and P. A. Ushakov
The active zones and side screens of reactors frequently consist of hexa- /170
gonal fuel-element holders in which the cylindrical fuel elements are placed in
a regular triangular lattice. To prevent the fuel elements from buckling and to
eliminate the attendant temperature variations, the fuel elements are held at
fixed distances from one another by winding spiral fins on the shell; the height
of the fins is equal to the minimum space between the elements or to half of this
distance. In the first case the fuel elements are in "rib-to-shell" contact, while
in the second the contact is "fin-to-fin. " The peripheral fuel elements contact
by their fins the fuel holder wall in either case.
The reliability of reactor operation depends to a large extent on the condi- /171
tions under which the heat is removed from the fuel-element surfaces, i . e. , on
the temperature regimes of the latter. Heat transfer in clusters of smooth rods,
arranged in a regular, infinite alternating pattern was studied previously for a
wide range of variation in the relative spacings 1 < x < 1.5. The results of these
studies are presented in [1] and [2].
The characteristic features of heat transfer from finned fuel elements held
in hexagonal assemblies are produced by the fact that the peripheral fuel elements
are situated near a flat wall and by the fact that all the fuel elements are finned.
The peripheral elements in the fuel-element holders at hand are not cooled uni-
formly over the perimeter due to the difference in the configuration of the spaces
surrounding them. Various quantities of the coolant flow through the different
spaces andhence, thetemperature of the coolant differs appreciably over the per-
imeter of the fuel elements. The fact that the fuel elements are finned can pro-
mote the intermixing of the coolant over the cross section of the fuel-element
holder, which improves the conditions for heat transfer. On the other hand,the
presence of fins produces local increases in the temperature of fuel-element
cladding.
Liquid metals have relative high thermal conductivity coefficients and low
coefficients of specific heat per unit volume. They are characterized by high
heat-transfer coefficients and high temperature rises in the coolant flowing in
passages. Hence, the temperature distribution in fuel-element surfaces is de-
termined basically by the local temperature gains of the coolant, rather thanby
the local heat-transfer coefficients.
Temperature fields of fuel elements under different conditions of heat
transfer (centrally and peripherally located fuel elements) were investigated at
the Physics and Power Institute. The experiments were performed on a stand
(Fig. 1) consisting of two main loops [one] with sodium and [the other] - with a
170
I
I^
I ^ 2
I' water
-ter J ---
Coolant route
Gas route
Vacuum route
Flowmeter
Vacuum manometer
ca Adjustable shut-off valve
b Potentiometric level meter
Shaft-type level meter
Venturi tube
Thermocouple
Figure lq Schematic Diagram of the Experimental Stand, 1) Vacuum Pump;

2), 3) and 19) Cooler; 4) Measuring Tank; 5) Breathing Tank; 6) Centrifugal
Pump; 7) Sampler; 8) Air Cooler; 9), 10), 11) and 20) Traps; 12) Inter-Cir-
cuit Heat Exchanger; 13) Mixing Tank; 14) Heating Loop; 15) Electromagnet-
ic Pump; 16) Vapor Trap; 17) Gas Bottle; 18) Test Section; 21) Filter;
22) Drain Tank.
sodium-potassium alloy (22% Na + 78% K) s The sodium was circulated by an

electromagnetic, and the sodium-potassium alloy by a centrifugal pump provid-
ing a pressure of about 6 kg/cm2 for a coolant flow rate of about 150 m3/hour.
The impurities were removed from the coolant by cold traps cooled by sodium-
potassium alloy provided from an auxiliary loop. The oxygen content of the
coolants was measured by distilling a coolant sample with subsequent chemical
analysis of the residue. The heat removal in the circuits was brought about in
metal-to-water heat exchangers, built with a double wall with a graphite inter-
layer (for high temperature) or with an interlayer from Wood's metal (for low
temperatures). Bellows-type valves were used in the sodium-potassium loop,
while valves with frozen-sodium seals were used in the sodium loop. All the
components of the stand were made from 1Kh18N9T steel, The experiments
with the sodium-potassium alloy were performed using models of four types.
Model No, 1, consisting of 37 tubes assembled in a triangular array with
a relative spacing x = 1,15 and placed in a hexagonal shell is shown in Figs. 2
and 3. The tubes were heated by nichrome coils so as to give q = const at the
171
4
6
7
.o
9
10
Section AA
Figure 2. Overall View of Model

No. 1 (Smooth Tubes) 1, 2, 5, 7
and 8) Tubes With Thermocouples
and Packing Glands; 3) and 20)
Terminal Blocks; 4) [Retaining]
Plank; 6, 9, 10 and 19 Pack-
ing and Fastening of Thermo-
couples; 11, 12, 13, 23 and
24), Headers With Packings and
Repellers; 14 and 28) Grating
With Thermocouples and Mixing
Header; 15 and 25) Gratings;
16) Shell; 17 and 27) Thermo-
couples; 18) Cooler; 21) Packing
Gland; 22) Connecting Stub;
26) Heated Tubes.
172
11 1
Corner
element
X111
1
1
a
b Side
element
n
0 c
Figure 3. Sections Through Models, a), b) and c)

are Models No. 1 (With Ribbed Tubes), No. 3 and 4,
Respectively; I, II and III are Numbers of Intertub-
ular Spaces [Cells].
inner surface, The center, side and corner tubes were equipped with 12 thermo-
couples each, placed in capillaries with an external diameter of 0.5 or 0.8 mm.
These tubes are held in ribbed holders, in close contact with the tube surfaces.
Using these holders it is possible to rotate the tubes without disturbing the cir-
cuit's mode of operation. The remaining 34 tubes had fins spot welded to them.
The height of the ribs was half the size of the narrow space between the tubes.
In addition, this model was also studied in the finless version.
Model 2 is basically of the same design, except that it consists if 37 smooth
tubes with a relative spacing x = 1, 24. This model was used to study the temper-
ature fields in the center and corner tubes,
Model No, 3, consisting of 19 tubes assembled in a triangular array with
relative spacing x = 1, 04 and placed in a hexagonal shell,is shown in Fig. 3b. /175
The tubes were held at fixed distances by spiral fins, the height of which was
equal to the size of the narrow intertubular space (the ribs were in contact with
the surfaces of neighboring tubes and of the shell walls). The center, a side and
173
a corner tube were made so that they can be rotated. The temperature at the
tubes surfaces was measured by thermocouples in protective capillaries with
an outer diameter of 0, 3 mm,
Model No, 4 (Fig. 3c) consists of seven heated tubes, twelve rods and a
specially-shaped spacers, - placed in a cylindricals shell. The tubes and the
rods are arranged in a regular triangular array with a relative spacing of x =
= 1.04. Three side heated tubes and three displacing rods have three fins each,
while the remaining tubes and spacers are smooth. Thus, the fins are in con-
tact with the surfaces of the smooth tubes and of the spacing rods. The center
tube is made so that it can be rotated. In its surface were embedded 12 thermo-
couples in protective capillaries with an outer diameter of 0, 5 mm.
The temperature of the fluid at the outlet from each cell and in the headers
was measured in each model, for which purpose we used thermocouples in cap-
illaries with an outer diameter of 0.8 or 0. 5. The geometric characteristics
of the models are presented in Table 1,
TABLE 1, Geometic Characteristics of Models
Model
Designation
Ns f X2 I J\, 3 ( \: 4
Outer diameter of tubes d2, rnm 17 17 17.6 17.6

Inner diameter of tubes di, mm 13 13 16,4 13.6
Relative lattice spacing, x 1.15 1.24 1,04 1.04
Heatedlengthlo, mm 690 730 700 700
Distance between end lattices 750 785 770 770
L, trim
Cross sectional area of rib, mm2 1.8x1.2 0;8X0-8 0.8x(1.75
Windind pitch of each rib, mm 750 1 1500 1500
Side Elements
The temperature variations were determined over the perimeter of a side

tube, located near a flat wall. Figure 4 presents experimental results obtained
with model No. 1, The temperature minimum is observed in the region adjoin-
ing the shell (co = 0 0 ), while the maximum took place at co =180 . This behavior
can be attributed to the difference in the temperature gains of the coolant in
spaces II and III, surrounding the side tube (see Fig. 3a and Table 2),
Figure 5 shows the results of approximately calculated relationships be- /178
tween the average velocities (WII/W III )I volume flow rates (VII/VIII) and tem-
perature gains of the coolant (ta ll/Otlfl) in spaces [cells] II and III as a function
of the relative tube spacing for clusters with hexagonal holders (models No, 1
174
TABLE 2. Geometric Characteristics of Spaces [Cells]
I, II and III of Clusters of Smooth Tubes
Model
Designation General expression ,,; I, + X, 2
x=1,15 I x= 1,24
Area of flow cross
section of the cell
I 0I=
z ` 21^ - x 2 -11,mm2 8,676 13,1
^j z3
II Phi=
8z
(4 x2 - 1mm 2 77,60 109
rojjI7 8d--, 2-V3

x 2 -1mm 2
III = C 52,06 79
Wetted perimeter
of cell
2y3 x=, I
I P1= 12 2 10,2 10.5
\
n d2 2x
11 PII = 2 +1 mm 46,2 47.8
III PIII= 22 , mm I 26,7 26,7
Heated perimeter
of cell
I POI = 22 mm I 8.91 8.91
-
II 2 mm I 26,8 i 26,8
Poil =
III + mm I 26,3 I 26,8

POIII = 22
Hydraulic diameter
of cell
/ 21437 X2 -1
d2 (
I dhj ,mm 4.4 3,4
2 3 x+1
d2 (4,2-1
II 2 , mm 9,16 6,71
dhil
TE x+1
2 3
III dhI I I C x 1d2 , mm 11.8 7.80
175
and No. 2 with smooth tubes).
The calculations were made for
turbulent and laminar flows of
the coolant on the assumption
that between the cells the cool-
ant does not become intermixed
and exchange of heat does not
take place [between its individ-
ual streams], while the pressure
drop in all the cells is identical
PL
AP = X TV2 = idem. (1)
h
The calculations were per-

formed for three cases.
1. Turbulent flow with
Re > 105 . In this case it may
be assumed that the friction co-
efficient does not depend on the
geometry of the cells (X = idem),
It follows from Eq. (1) that the
average velocity of the coolant
0 120 240 go, de gre e in the cell is
Figure 4. Temperature Variation in a
Ribbed Side Tube of Model No. 1.. (2)
h
m/sec Re Pe 2. Turbulent flow with

1
p 8.65 77 300 ` 2550

104 -_ Re _5 10 5 . In this case it
may be assumed that
* 2.37 ( 19800 594
I
Q 0.95 8 830 234 0 .3164i WCIo (3)
I t 2,=
o , while
Q I 0.196 I 2 345 ( 45,5
I 0,072 I 878 ( 16,7

3. Laminar flow. In this /179
case
X=-IF '
(4)
Where A is a coefficient which depends on the shape of the passage. From Eq.
(1) we get that
W ^ . (5)
176
WO/W) It is assumed that coefficients
A are the same for the side
and center cells.
i 0
As can be seen from
Fig. 5, the theoretical rela-
tionships between the veloc-
s ities, volume flow rates and
temperature gains of the cool-
ant in cells II and. III depend
to a large extent on the rela-
tive spacing of the rods in the
holder. When x is changed,
the minimum distance between
0. 8 the peripheral elements and the
2 flat wall change also. This pro-
duces a change in the relation-
ship between the flow cross sec-
a1 tion of cells II and III. The lat-
41 1.2 43 1,4 x
a ter circumstance results, in the
final count, to redistribution of
velocities, flow rates and tem-
perature gains of the coolant
in the above cells.
For model No. 1 with
smooth tubes (x = 1.15) the
theoretical estimates are
rrt
so
1, 37;0,75.
3
40
As can be seen, the tempera-
2 ture gain the coolant in cells
II is by 25 % lower than thatin
cell III, which is precisely the
2,0 circumstance responsible for
the temperature variationover
the perimeter of the side tube.
These theoretical estimates
0 should be regarded as approx-
1 1,1 1.2 l 3 74 ^ imate, but useful for qualita-
b tive clarification of the charac-
ter of temperature fields over
the perimeter of the side tube
Figure 5, Caption on Following Page. obtained in the experiments.
177
dtri/d till It can be seen from
Fig. 4 that, at relatively
3 high Peclet numbers, local
distortions of the tempera-
75
ture distribution are observ-
ed, which are due to the pres.
I ence of fins. At low Pe this
I ! f effect disappears, however,
! 2 the overall temperature var-
46 ------ ---- - --- -- ------ iation increases sharply. In
I i the absence of fins the wail
temperature varies over the
perimeter smoothly at all
0.5
Peclet numbers.
I I
i I Analysis of experimen-
tal data shows that for a wide
range of Peclet numbers (100
< Pe < 4000), the presence
17 1.2 7.3 7.4
c
of fins does not sharply re-
duce the temperature varia-
tions in the side-element
Figure 5. Calculated Values of
WII/WIII (a); VII/ wall. Some peculiar features
/VIII (b); AtI1/AtIII (c) for Cells II and III of Mod- ture
of the behavior of tempera-
variations can be ob-
els No. 1 and No. 2 With Smooth Tubes. 1) Tur- served in the laminar and
bulent Flow With Re < 10 5 ; 2) Same as Above transient flows of the cool-
With 104 , Re < 105 ; 3) Laminar Flow. ant. For the model with
smooth tubes in the range
20 < Pe < 80 (800 < Re
< 3000) one observes maxi-
max min
tw /q, ], X f (Fig, 6). For a clus-
ma in the temperature variations (tw
ter of finned tubes these maxima are not as pronounced and are somewhat dis-
placed into the region of lower Peclet numbers. An exception are the initial sec-
tions (1/dh < 30). As the rate of flow becomes lower, the temperature gains, and /181
consequently, also the variations in the temperature of the peripheral elements,
increase. This increase cannot be infinite, due to heat transfer between the cells,
which is proportional to the difference in temperature gains. In low-velocity re-
gions an effect is exerted by natural convection,, which aids in intermixing the
coolant and reducing the temperature variations, hence, a maximum in the tem-
perature variations is observed at certain Peclet numbers. The lower temper-
ature variations observed in finned tubes in this region are attributed to the fact
that the ribs aid in intermixing the coolant.
In laminar flow the effect of the ribs is, apparently, moderate, while the
development of natural convection in a cluster of finned tubes is more difficult
as compared with this phenomenon in a cluster of smooth tubes. Fence, the var-
iations in the temperature of the side finned tube in laminar flow is found to be
somewhat greater than for a smooth tube (see Fig. 6).
178
Ma*min
t
qRZW
dh
2
0
4 '
Z
19.4
4
2 ^
4 10. S
2 I
0 1 10 104 703 4.1103
Figure 6. Curve of the Maximum Variation in the Temperature

of a Side Tube of Model No. 1 as a Function of the Peclet Num-
ber. o and Denote Points Obtained for the Smooth and Ribbed
Tubes, Respectively.
In the experiments (particularly with the cluster of smooth tubes of model

No. 1) we have observed the following: in transient flow the temperature dis-
tribution of the coolant at the outlet from the cluster varies in such a manner /182
that the coolant has its maximum temperature gain not at the center of the clus-
ter, but near its periphery (Fig. 7). Here the average temperature of the peri-
pheral tube exceed the average temperature of the center tube, while the tem-
perature variation over its perimeter preaches a 4naxim.um. These effects dis-
appear at higher and lower Peclet numbers.
It is assumed that the above anomalies in the temperature fields are due
to the redistribution of the rates of flow of the coolant.among the cells of the
cluster attendant to transition from turbulent to laminar flow. This redistribu-
tion is produced by the fact that such a transition takes place primarily in the
peripheral cells, in which the velocity of the coolant (see Fig. 5a) and the hy-
draulic diameters (see Table 2), and consequently, also the Reynolds numbers
are lower than in the center cells. Here the flow in the center zone still remains
turbulent. Using Eq. (1) we can write
W II 1^^TIII dhll (6)

}^.n II
` III y
^hiII
where XtIII and XtII are the coefficients of resistance of cells III (turbulent flow)
and cells II (laminar flow). It can be seen that at the instant of transition from /184
179
000
I 9 Rt -4.) R21A,}
^ 8
T
23 16
-tin)f$RzâJ 0'
i
r2
^ ^ Re=563
^^ t' W=O.f90m /sei 8 Pe=12,9
2. 4 Re =5200
3
8 Pe=130
4
i S
7 G
1,!
1G
i ^ 9
li
6
0,! i
i ^^ ! W=17.132m/s ac
4
L Re=1450
î Pe-39
0 i
1.2 W=J,02m1sec G
Ire=12900
1 '_ Pe=255 1G

V6
/' ^ W=0,185m/sE C
^ ^ Re=2010
0 6 Pe=53.1

_27r^m /sec
^^ I It Q { 1 PrC 12TU0
61
4
0.3 - --
_ W- :slip/sec r
200 400 600 0 1B0 360 20 60 1100 140 0 200 400 600 0 180 360 20 60 100 140
j mm 5o, deg Ry mm 1, mm cp, deg R, mm
Figure 7. Temperature Fields in a Cluster of Smooth Tubes of Model No, 1, and O

Denote the Average Temperatures of the Walls of the Side and Center Elements, Respec-
tively; the Dashed Lines Shown the Maximum and Minimum Temperatures of the Wall of
the Side Tube; 0 Denotes the Temperature of the Coolant Around a Side Tube at the Out-
let From the Cluster; o Denotes the Temperature of the Coolant
Over the Radius of the Cluster,
turbulent to laminar flow an additional velocity drop, by a factor of x't III x t II
takes place in cells II, for which reason the flow of the coolant moves into the
center of the cluster.
The reduction in the average temperature and in the temperature variation
of the peripheral tube toward the end of the heated section (see Fig. 7) is appar-
ently due to the turbulizing effect of the upper butt grid and due to redistribution
of velocities about it.
In the finned tubes cluster all the above effects were less pronounced, since
in this case the redistribution of the flow rates was weaker due to the smaller dif-
ferences in the hydraulic diameters of cells II and III (dh II = 6 mm, d III =6.5 mm).
Figure 8 shows changes in the temperature variations in side tubes of mod-

el No, 1 with ribbed tubes (x = 1.15) as a function of the length 1/dh of the heated
section for the range 200 < Pe < 1000. For this range of changes in Pe the tem-
perature variations became stabilized. In the region of large Peclet numbers
(1000-500) this ensued at length 1/d h 60, while for lower Pe (400-200) this
took place at 1/dh PL, 80-90.
1,2
Pe=200
N 0.8
o^ 300
00
500
C^ 0 4
^k
1000
v
20 30 40 M 60 70 80 11dh
Figure 8. Curves of the Maximum Variation in the

Temperature of a Side Tube of Model No. 1 as a
Function of the Length of the Heated Section.
The temperature variation over the perimeter of a side tube of model No.
3 (x = 1.04) behaves in approximately the same manner as in experiments with
x = 1. 15. However, the temperature maxima in the wall in this case are dis-
placed to the points where the side tube comes close to tubes 3 and 4 ((P = 210
and 150 0 , see Fig. 3b). Temperature minima correspond to cells in which the
coolant is relatively underheated, which is due to the smaller spaces between
the tubes as compared with model No. 1.
For the same Peclet numbers the magnitude of the temperature variation
in the side tube of model No. 3 exceed appreciably the temperature variation in
181
a side tube of model No. 1 for the following reasons. First, the relative spac-
ing of tubes in model No. 3 (x = 1.04) is appreciably smaller than in model No. 1
(x = 115), which interferes with intermixing of the coolant between the cells.
Second, the relative distance from. the peripheral elements to the flat shell in
model No. 3 is greater than in No. 1, Hence, the variations in the coolant flow
rates, and, consequently, also in the temperature between the different cells of
model No, 3 is more appreciable. This corresponds to results of qualitative
calculations shown in Fig. 5. Third, . the relative wall thickness of tubes of
model No. 3 is lower than in model No, la This interferes with tangential heat
fluxes and produces an additional temperature variation as compared with tubes
of model No. 1,
The temperature variations over the heated length of the side tube of model
No 3 did not become stabilized. This circumstance is apparently attributable to
the lesser intermixing of the coolant between cells II and III as compared with
model No. 1,
Corner Elements
The temperature fields of the corner elements were studied using models
No. 1, 2 and 3. The tubes of models No, 1 and 2 were smooth, while those of
model No, 3 were ribbed. The results of measurements are shown in Figs. 9-11.
It can be seen from Fig. 9 that in clusters of smooth tubes with x = 1,15 and 1/d h=
= 76 the maximum temperature in the wall of a corner tube is observed in the re-
gion adjoining the corner of the shell, while the minimum temperature is observed
in the opposite part of the tube's perimeter. There are no explicit temperature
jumps at points where the tube comes closest to the shell.. The temperature be-
haves similarly in the corner element of model No. 2 (x =1.24), Apparently, in
fuel element clusters with x =1.15 and 1, 24 the minimum distance between the
shell and the surface of the corner tube, which is equal in models No, 1 and 2
to half of the space between the tubes (height of the fin) is entirely sufficient to
eliminate appreciable overheating of the coolant. The wall temperature atpoints
close to neighboring tubes does not increase either. As was shown [2], a varia-
tion in the wall temperature virtually does not exist in clusters of fuel elements
arranged in a regular infinite array with a relative spacing x > 1. 1.
We should note the rather insignificant minima in the wall temperature of
the corner tube in the region of co = 90 and 270 0 , which are observed in experi-
ments with model No. 1 at high Reynolds numbers (see Fig. 9), These minima
are due to the fact that the coolant flowing in cell II is cooler than that flowing
in cell III, At low Reynolds (Peclet) numbers (for example, Re = 760) these min-
ima do not exist.
The variations in the temperature of corner tubes of model No. 1 are ap-
preciably lower than the corresponding temperature variation of side tubes in
the range of Re(Pe) under study. The shape of the curve of maximum variation
in the wall temperature of a corner tube (P ax -- tW in/q' R2 )X as a function of
182
Pe is approximately the same /187
as for side tubes. However,
the maxima in the tempera-
ture variations for the cor-
ner tubes are less pronounced.
The manner in which the
1
^N
wall temperature changes along
the perimeter of a corner tube
of model No. 3 for 1/d h = 145
^ 3
is shown in Fig. 10. The wall

,tea
temperature in this case changes
in a different manner than in
corner elements of models No.
1 and 2. The temperature max-
imum is observed at the point
where it is close to tube 1 (cp =
=1800, see Fig. 3b), the tem-
perature minima occur in cells
in which the coolant is relative-
0 60 120 180 240 300 cy, deg
ly underheated. This behavior
is attributable to the fact that
Figure 9. Temperature Variations of a Smooth the relative spacing between
Corner Tube of Model No. 1. the corner elements of model
No. 3 and the shell is by afac-
tor of two greater than in mod-
W, m/sec I Re I Pe
els No. 1 and 2.
2.71 I 28 300 I 616 Figure 11 shows changes

I
in the maximum variations in
O 2,48 I 19 900 I 609 the temperature of corner and
side tubes of model No. 3 as a
I 1.26 I 13 750 ( 283 function of the Peclet number
at 1/d h = 145 and 180. The
I 0.498 I 4 620 I 117
temperature variation of the
I 0.083 760 I 19.4 corner and side tubes for
I
1/dh = 145 is approximately
Q I 0.029 ( 339 I 6.4
the same for the same Peclet
numbers in the range 20 < Pe <
< 200 under study. However,
for the given value of 1/d h thermal stabilization ensues only for the corner tube.
Hence, as l/dh increases, the temperature variation in the side tube is increased
somewhat.
It is seen from Fig. 11 that the curve of temperature variations as a function
of Pe for the corner and side tubes of model No. 3 has no maxima. In a cluster of
fuel elements with x = 1.04 heat transfer between the cells, as well as natural con-
vection are more difficult than in clusters with x =1.15 and 1.24. This is responsible
183

for the continuous increase in

the maximum [temperature]
max min /q-
pin
variation tw(
X R2 )Xf of the corner and side

tubes of model No. 3 with a
reduction in the Peclet num-
^N bers in the range 20 < Pe <
< 200. Maxima in this case
are apparently displaced into
the region of lower Peclet
v numbers.
Center Elements
iL- 240 300 p, deg Study of the tempera-

ature fields of a finned cen-
ter tube of model No. 1 has
Figure 10. Variations in the Temperature of a shown that the wall temper-
Corner Tube of Model No. 3, ature under the fins increases.
As can be seen from Fig. 12,
Pe I Pe
the maxima in the wall tem-
perature which are observed /189
g I 8 I e ( 63
in the vicinity of the fins are
more or less the same for
23 I 103 large Re. As the Reynolds
II II numbers are reduced the
( 32 O, 166 temperature field is no long-
er symmetrical. The tem-
pArature disturbances induc-
ed by the fins extend to an
appreciable part of the tube's
perimeter. This effect is not as yet sufficiently clear. However, it may be as-
sumed that it is due to a certain deviation of the fins from the nominal dimen-
sions, not entirely identical contact between individual fins and the tube surface,
as well as to an increase in the laminar layer near the wall with a reduction in
Re.
Following are the values of the Nusselt numbers averaged over the perim-
eter of the center tube of model No. 1 in the region of stabilized heat transfer:
Pe 1150 1000 835 693 553 228 166 111 54.3 27.3 9.93 7.65 4.71 3.72
Nu 10.2 9.44 8.56 7.22 6.55 5.46 5.68 3.80 5.17 5.07 5.15 4.64 6.00 3.20
184
The above Nusselt and Peclet
numbers have been calculated
by means of the following ex-
pression
IZI
Qc _ 4dh
Nu=
q Cf . 7f) ^T
^ 3 (7)
m Ii' f
Pe = CPy,
If
where d is the hydraulic di- /190

ameter calculated by sub-
20 30 0 60 80 100 150 Pe
tracting the area occupied by
the ribs, tw is the tempera-
Figure 11. Maximum Temperature Variation in ture of tube's surface, aver-
the Corner and Side Tubes of Model No. 3 as a aged over the perimeter and
Function of the Peclet Number. O - Side Tube, calculated by taking into ac-
Z/dh = 180; * - Same as Above for Z/d h = 145; count the temperature in-
crease beneath the ribs; t
o Denotes a Corner Tube With Z/dh = 145.
is the average temperature
of the coolant in the section
at hand and q is the specific
thermal flux, calculated for
a smooth tube.
Figure 13 shows experimental data for stabilized Nusselt numbers for a
model with x = 1.15 (finned tubes), as well as the curve of Nu = f(Pe), calculat-
ed for a cluster with x = 1. 15 (smooth tubes) using a generalized relationship
obtained in [3]. Comparison of experimental data with this curve is rather ar-
bitrary, however, it follows from it that the fins of tubes of the model with x =
= 1.15 does not result in an appreciable change in Nu as compared with the smooth
version of tubes, if the Nusselt and Peclet numbers are calculated from Eqs. (7).
The appreciable scatter in the experimental data for laminar flow should be re-
garded as a natural phenomenon, since the experimental accuracy in this region
is only moderate.
Results of studies of the temperature field of the center tube of model Noe
3 (x = 1.04) are presented in Fig. 14. The temperature distribution over the
perimeter of the center tube is determined primarily by the configuration of the
cells. When fins are positioned in narrow parts of the cells, the additional in-
crease in the temperature variation comprises about 5-10% of the average var-
iation. The distributions of thermal fluxes over the perimeter were found by
solving the problem of thermal conductivity and on the basis of results obtained
by measuring the temperature distribution.
185
The dependence of the tem-
perature variation of the center
element on the Peclet number for
20 < Pe < 300 is described by the
formula
max
tw aw
N At 1.5 Pe-0.32.
c 4R2 (8)
^ ^ n r
q
The values of the Nusselt num- /192
bers in the region with stabil-
V ized heat transfer for the cen-
ter element of model No, 3 are
as follows:
Pe 22 33 66 106 175
Nu 1.26 1.62 2.90 4.60 5.90
0 60 120 180 W 300 9, deg The strong dependence of

c the Nusselt on the Peclet num-
ber (the power of Pe being about
Figure 12, Variation in the Temperature 0. 75, as compared with clusters
of a Finned Center Tube of Model No, 1. of smooth tubes (the Peclet num-
a) W = 1.54 m/sec, Re = 14,400, Pe =366; ber here 0.45) is remarkable and
b) W = 2, 37 m/sec, Re = 20, 600, Pe =601; is apparently brought about by the
c) W =4, 02 m/ sec, Re = 32, 330; Pe =1030. fins.
nu
?0
10
2
I I I I IIIIII I I I I IIIII I I I I IIIII I I I I IIII
r 2 4 6 g in 2 4 6 9 m2 2 4BB -Y/)J 2 4 B PP
Figure 13. Values of Nusselt Numbers in the Stabilized

Heat-Transfer Region of Model No. 1. The Black Dots
Represent Data Obtained on a Model With Finned Tubes;
the Solid Line is the Curve of Nu = f(Pe) for a Cluster of
Smooth Tubes.
Results of studies of the temperature field of the center tube of model No. 4
are presented in Figs. 15 and 16, In the given model, due to alternation of smooth
186

and finned tubes, the sur-

face of the center tube can
be broken up into three re-
gions with identical heat-
transfer conditions (330 <
< cp < 90 0 , 210 < co < 330,
cp <(p< 210 0 ) (see Fig. 3c).
Distributions of tempera-
ture and of thermal fluxes
4- averaged over the range
^N 330 < cp< 900 are shown
in Fig. 15. Temperature
maxima in the narrow parts
of cells differ depending on
whether the neighboring tube
is smooth or ribbed. The
presence of fins on a neigh-
boring tube for a given clus-
ter geometry produces a
certain smoothing of the
temperature variation at
the center tube.
Figure 16 shows the

0 S 10 15 20 25 ^0, deg curve of the maximum var-
iation in the dimensionless
Figure 14. Variation in the Wall Temperature temperature as a function
and in the Thermal Fluxes Over the Perimeter of of Pe (Re); which shows
the Center Tube of Model No. 3. more pronouncedly than in
a cluster of smooth tubes
the effect of the coolantve-
Pe II I Pe
locity on the temperature /193
11
variation in the center tube.
o 1 22 4 1 67
The cause of this are the fins
0 ( 33 II V I 106
which aid in intermixing the
coolant over the cross sec-
A I 63 + 1175 tion of each cell. The in-
termixing effect becomes
greater with an increase
in velocity.
The values of the Nusselt numbers for the region with stable heat transfer
for the center tube of model No. 4 are as follows.
Pe 75 117 178 270 373

Nu 1.1 2.0 2.3 2.9 3.5
187
4
1
4
4A 2 1.3G
3
is
0,01 12G
5
1,10
K10,06
Oc
/ 7.O1
3
0.01
i 0.81
Q0. 0.76
QO 0. ^
90 60 30 0 t0, de g 30 60 30 0 V de g
Figure 15, Variation in the Wall Temperature and in Ther-

mal Fluxes over the Perimeter of the Center Tube of Model
4. 1), 2), 3), 4), 5) and 6) are curves for Pe = 75, 117, 178,
270, 373 and 588, Respectively,
Results
1, On the basis of the measured temperature field of models of fuel ele-

ments, assembled in hexagonal holders, the following conclusion can be made.
The peripheral elements (situated at the flat wall of the shell) can have appre-
ciable temperature variations produced by differences in the geometry of the
surrounding spaces. This circumstance must be taken into account in design-
ing liquid-metal cooled reactors,
2. The variation in the wall temperature of peripheral elements is due
primarily to the different temperature gains of the coolant in spaces surround-
ing the fuel elements.
3. Elements situated near a flat wall at a distance equal to the minimum
spacing between the fuel elements have a relatively greater temperature varia-
tion than elements the distance between which and the flat wall comprises half
of the maximum distance between the fuel elements. In both cases the temper-
ature variation depends substantially on the relative spacing of the fuel elements.
In tightly-packed clusters (x = 1.04) the temperature variation in the peripheral
elements is higher than in the looser clusters (x = 1. 15, 1.24),
4. In the loosely-packed clusters the temperature variation in the side ele-
ments is appreciably higher than in corner elements. In tightly-packed clusters
the variation in the wall temperature of these elements is approximately the same. /194
In both cases the temperature maximum of peripheral elements in turbulent and
188
U 76 laminar flows of the coolant is approx-
imately equal to the maximum temper-
0.72
ature of the center fuel elements.
5. In experiments with x = 1.15
k0, 08
1.5 we have discovered the effect of redis-
Î tribution of the coolant flow rates over
0.04 the cluster's cross section attendantto
the transient flow of the coolant. This
redistribution can result in the fact that
the average wall temperature in the per-
7 8 910 2 2 V 3 4 5902Pe ipheral elements will exceed the average
temperature of the center elements.
6. The temperature variations in
N the walls of side elements become sta-
bilized over the heated length of tight
clusters at appreciably higher values
î of 1/dh than for loosely-packed clus-
ters. The temperature variation in a
90 60 90 0 t0, deg side tube of the cluster with x = 1. 04
does not become stabilized even at a
Figure 16. The Maximum [Temper- distance of 1/dh - p = 180.
ature] Variation in the Center Tube
of Model No. 4 as a Function of Re
and Pe 1) and 2) the Heat-Conduct- 7. The finning of fuel elements
ing Surface is Facing a Ribbed and arranged in regular lattices in hexa-
Smooth Side Tube, respectively, gonal holders reduces the temperature
variations in the peripheral elements
for the case of turbulent coolant flow.
This is not observed for laminar flow,
8. In recalculating the experimental data obtained for specific fuel ele-
ment designs, one must satisfy the conditions for approximate thermal simula-
tion of fuel elements presented in [4].
REFERENCES
1. Subbotin, V. I., et al.: Teplootdacha v elementakh reaktora, okhlazhdaye-

mykh zhidkim metallom [Heat Transfer in Elements of Liquid-Metal Cool-
ed Reactors]. Report presented at the 3rd International Conference on
Peaceful Uses of Atomic Energy, Geneva, 1964.
2, Zhukov, A. V., et al.: Teploobemen pri prodol'nom obtekanii zhidkim metal-
lom razvinutykh puchkov sterzhney [Heat Transfer From Loosely-Packed
(Fuel) Rod Clusters To Liquid Metal Flowing in the Axial Direction]. Sec-
ond Article in Part II of this collection.
3. Ibragimov, M. Kh. and Zhukov, A. V.: Atomnaya energiya, 20, 5, 425 (1966),
4 Ushakov, P. A.: Priblizhennoye teplovoye modelirovaniye tsylindricheskikh
teplovydelyayushchikh elementov [Approximate Thermal Simulation of Cylin-
drical Fuel Elements). First article in Part II of this collection.
189
HEAT TRANSFER IN THE INTERTUBULAR SPACE OF
LIQUID-METAL HEAT EXCHANGERS
P. L. Kirillov and M. Ya. Suvorov
To calculate the heat transfer coefficient in tube-type heat exchangers one /194
must know the laws governing the transfer of heat in the tubes and in the inter-
tubular space (on axial and transverse flow) of the tube clusters. Heat transfer
attendant to the flow of liquid metals in pipes has been studied extensively and at
present there are available recommendations on the calculations which agree with
one another quite satisfactorily. [On the other hand] very little was done to study
the heat transfer attendant to axial flow past tubes, [1-3]. The previously ob- /196
tained formulas for calculations required checking, since the preceding experi-
ments were performed without checking the purity of the metal or under circum-
stances where such control was unsystematic.
The purpose of the present investigation was in experimental study of heat
transfer on axial flow of liquid metal in the intertubular space of a heat exchanger.
Here we have determined the dependence of the average heat transfer coefficient
on the velocity and on the geometric dimensions of the cluster. To generalize the
data obtained, we have used the hydraulic diameter as the characteristic dimension.
The schematic of the stand, consisting of two loops, which was used in the /19,
experiments is shown in Fig. 1. In the first (heating) loop the heat-transfer agent
is sodium, while in the second loop the coolant is an eutectic sodium-potassium
alloy. In the experimental heat exchanger the sodium was piped through the tubes,
while the sodium-potassium alloy was pumped through the intertubular space. The
sodium was pumped by a centrifugal pump with a frozen-sodium seal. The so-
dium was pumped through an electric heater and the heat exchanger. This stand
was used simultaneously to study the heat transfer of liquid metal in round
tubes, for which purpose a special test section was installed. The oxides from
the sodium were removed by a cold trap, while the admixture content was mon-
itored by cork-type oxide [content] indicators. The flow rate of the liquid metal
was measured by magnetic flowmeters, which were calibrated by a measuring
tank. The sodium-potassium alloy was pumped through the circuit by an electro-
magnetic pump through the experimental heat exchanger and an evaporator. The
heat from the evaporator was removed by boiling toluene, the vapor of which con-
densed in the water-cooled condenser. The alloy circuit was provided by a cold
trap, oxide [level] indicator and a tank for calibrating the flowmeters.
The design of the heat exchangers which were tested on the stand is shown
in Fig. 2. The heat exchanger consisted of a cylindrical jacket, with seven tubes
placed inside of it and six displacers, welded to the casing. The geometry of the
inlet and outlet sections of all the heat exchangers is identical. Only their length
and relative spacing were changed, which made it possible to expose the effect of
these factors on the average heat transfer (Table 1). The sodium and the sodium-
potassium alloy were pumped through the heat exchanger only after both metals
190
Figure 1. Schematic Diagram of the Experimental Setup (Sodium Circuit).
1) Loading Tanks; 2) Electromagnetic Pump; 3) Condenser; 4) and 15) Vapor
Traps; 5) and 16) Breathing Tanks; 6) and 18) Cold Traps; 7) and 17) Mea-
suring Tanks; 8), 11), -13, 19 and 21) Oxide [Content] Indicators; 9) Evapora-
tor; 10) Heat Exchanger; 12) Electric Heating Loop; 14) TsN-280 Centrifugal
Pump; 20) Test Section (Tube).
Sec AA
7
11
Sec BB
9
Sec BB
Figure 2, Heat Exchanger. 1) Turbulizer; 2) Connecting

Stub; 3) Thermocouple Hood; 4) and 11) Inlet and Outlet Mix-
ing Headers; 5) Casing; 6) and 7) Displacers; 8) Assembled
Bellows; 9) Spacing Ring; 10) Tube; 12) Ring; 13) Dowel.
TABLE 1. Basic Geometric Characteristics of the

Heat Exchangers
Hydraulic
Diameter Working diameter Tube
x L /dh
of tubes, length, of clus- material
mm mm ter, mm
7.6 x 0.3 425 1.1 3,57 119 1Xh18N9T

7 x 0.2 314 1.14 2,62 120 1Kh18N9T
6.8 x 0.3 500 1.14 2.93 170 EI-211
b.8 x 0.3 660 1,14 2,93 225 EI-211
7x0.2 514 1,25 4,54 113 1Kh18N9T
7 X 0.2 765 1.25 4.54 169 1Khl8N9T
7 x 0.2 1040 1,25 4.54 230 1Kh18N9T
6.8 x 0.3 460 1.4 7,67 60 EI-211
6.8 x 0.3 1300 1.4 7.67 170 EI-211
192
had oxides removed from them by means of cold traps. The oxygen content after
-3
trapping was 5 . 10 -4% and 2 . 10 % by weight in the sodium and sodium-potassium
alloy circuits, respectively. After obtaining a stable thermal state, we have mea- /200
sured the following quantities: 1) flow rates of the sodium and of the sodium -potas-
sium alloy by means of magnetic flowmeters; 2) average temperatures at the in-
let and outlet of the heat-transfer agents. The temperatures were measured by
chromel-alumel thermocouples with wire diameter of 0. 2, placed in capillaries
from stainless steel. The thermocouple emf was measured by the R-2 / 1 poten-
tiometer.
On the basis of the measured flow rates and temperatures at the ends of the
heat exchanger we have determined the heat transfer coefficient
k1 _ Q
n8tLn (1)
Since
k1 _ 1 1
1
dout 1
;1, in+2^ W In d in +a z dout (2)
the heat transfer coefficient in the intertubular space of the heat exchanger
becomes
1
a2 - ou 1
d
1 1
în dout) .
wtaid mn^ i3, J (3)
The coefficient of heat transfer a from the sodium was calculated using the ex-
pression for a round pipe
Nut = 5+ 0.025 Peo R . (4)
In processing the experimental data the characteristic dimension of the in-

tertubular space was the hydraulic diameter for the entire cluster
dh=S.
(5)
The experimental data are presented in Fig. 3a. Curves which can be described
by the expression Nut -f ( Pe2 06 ) can be drawn through the experimental points.
193
Nu
16
At
12
10
0
8
6
o
00 v , p
o v o 4
4
0 0 'p o 8
0
0
2
00
0
1
zuu 300 400 500 600 700 900 900 1000 1100 1,000 Pe
a
soy
501
i on
/ Oc /one
4A
/' o o
/ vt l
v
o^
301
O
t // p Q o/
p

a //
00
1 /
o //
/
[uu JUU fUU SUV 6110 /UU tl011 9UU IUOU 771111 7ZO Fe
b
Figure 3. Experimental Data in Coordinates Nu t = f(Pe2 ) (a);
X 1, mm
N it, 425 A 1.14 314
660
r dh l ( ) lb) V 1.1.125 4 500
5t4
1,14
j L 0.O?7(r 1.1)o, tu j 12b 10/u 1.4u 46u
` n 1.46 I Gnu
194
Figure 3a shows that heat transfer depends, in addition, on the relative spacing
and on the length of the heat exchanger. To generalize the experimental data, on
consideration of these factors, we have selected the following form
Nu, = A{L+B(xm)Ì Pe2.s

(6)
where A, B, m and n are determined graphically. The range within which all
the characteristic quantities were varied in tests with all the heat exchangers
is given in Table 2.
After the constants are determined, the [above] formula takes on the form
Nu, == 8 {"L + 0.027 (x 1.1) 0 '"} Pe26.

(7)
Figure 3b displays the experimental data in the coordinates
Nuz
lg 1 Pe
f( g z)
L +0.027 (x- 1.1)0.46
The scatter of experimental points about the average line is +10%. /202
When operating the heat exchangers we have observed phenomena which re-
quires close attention to the deviation of the aatuab, characteristics of the heat ex-
changers from their design characteristics. Reference is made here to the re-
duction in the average heat transfer coefficient in the heat exchanger after pro-
longed operation, when the oxide content in the sodium increases. The experi-
mental heat exchanger (Z = 314 mm) was placed on a stand and operated for four
months, the first two weeks of which were spent studying the exchanger's char-
acteristics. During the entire time other experiments were performed at the
stand, during which the oxygen content in the sodium circuit varied from 5.10-3
to 5 . 10-4 [by weight], After the experiments were performed the stand was dis-
connected. When it was restarted we have discovered that the trap for removing
oxides from the sodium did not operate (it was completely clogged up by the oxides).
Nevertheless, we read the heat-exchanger characteristics again and discovered in
the process that the heat transfer coefficient has become smaller by a factor of
more than two. By increasing the temperature in the circuit and by removing
oxides from the metal by a new cold trap the heat transfer coefficient was [then]
increased appreciably. However, it still remained by approximately 20 17o lower
than the initial coefficient. We have disconnected then the end flanges of the heat
exchanger and washed the tubes by alcohol and water and dried them. The char-
acteristics of the heat exchanger with a surface thus cleaned returned to their initial
195
to
w
TABLE 2.a Range of Parameters Under Study
Len eat- Velocity, Temperature aZ kcal/m 2 - k kcal/m-

Re!o-4 I I
mm â nentr m/sec ( gain, CC Pe Nu
ho
hour-degree ^ hour-degree
Na I 5.6-13.4 , 3.7-20.8 I 8.43-24.65 ( 585-1420) 11.1-15,3 1 - I -
425
NaK I 2.0-9.9 ( 9-53 I 1.87-10.92 I 207-1060 f 1.5-4.7 9200-26100 56-111
Na I 12 I 13-15 ' 20.8-24.2 I 1200 I 14.2
31!i -
NaK ( 2.87-12.6 33-120 j 2.42-10.28 I 202-914 ( 2.7-7.3 ` 23 '100-61 300 ` 111-188
Na I 12.3-13 16-19 I 21.8-23,55 11163-12271 1.4-14,4
500
NaK I 3,28-11.8 , :35-104 I 3.24-11 6 I 259-936 2. 13 -6.0 2002(X)-45WO I 88.5-138
Na I 1-1.5-11 .1; 15.2-18 I 20-22.8 I '1090-111'2 1:3.7-13.9 -
rsU
NaK I 2,61-10,7 I 38-133 I 2.(;1-1(1,41 I 2 03-8 5 2 I 2. 151-4.78 116600-37 0001 7 7,6-123
514 Na I 9.5-10 I 9.7-12 f 13.6-16 I 935-980 I 12.9-13,151 -
NaK 1
2.0-9.2 1 16.5-65 , 2,96-11.35 1 265-1240 ( 4 62-11,9 21900-561 000 1 1 0 5- 1 80
Na 1 11.6-12.3 9,3--- 11 1 14.99-19.15 9611280 ( 13.15-14.6
765
NaK 1 2.69-8.8 1 18,5- 58 3.42--12 1 359-1-130 1 4.7-9.3 1 22 400-4900() 1 106-163
Na 1 10./1-1(1.6 14.'1-16.2 20.2-22 1 1050-1082 1 13.5-13.651 - 1 -
104U - -
Na IC 1 2.:38 - 7.4 1
1 1.1-16.8 1 3.78-11,7 1 284-876 ( 3.5-7.72 1 2 7 !d10-38 (100 ( '90,() -IM)
Na 1
1:3.5-13,11 1 14,1-15.1 1 24.7-26.42 1 1278-1297 1 Ui.135 -14.7 1 - 1
-
460
NaK I 1.4-5.'1 25.'3-516.5 1 3,7:3-13.28 279-1040 1 7,9-17 1 _24000-50000 1 9-1-144
Na ( 12.5 --- 12-1 .7 1 14 -16.6 1 22,15 --24 1178--1200 14.11-1 !12,51
1 1
N,IK '1.9--4.11 31 -71 1.95- 12.3 1 373-923 I 5.S1 -103 1 157222) :1(14110 I. t3--t12.3
A^
values. Figure 4 shows the variations in Nu during the experiments. A simi-
lar picture was also observed when a 1040 mm long heat exchanger was used.
Nu
6
3 0
4 ^
3
2 s
200 300 400 500 600 700 800 900 Pe
Figure 4. Effect of Impurities in the Metal on the

Heat Transfer Coefficient. o - Pure Metal; A Con-
taminated Metal; - After Filtration of Metal, T =10
Hours at t = 370C; 0 After Filtration, T = 50 Hours
at t = 250C and T = 10 Hours at t = 470C; o - After
Washing the Heat Exchanger, Pure Metal.
Results
On the basis of our experiments we can conclude the following. /203
1. The dependence of the heat transfer coefficient on the velocity in the

intertubular space can be expressed as
a2 W2' 6 or Nu, PeO-6
2. The dimensionless heat transfer coefficient (Nu 2 ) increases with an in-

crease in the relative spacing of the tubes in the cluster.
3. The average heat transfer coefficient for the intertubular cluster [ap-
parent error ] should be space, ] becomes stabilized at relative lengths of
about 300-400.
4. Until more accurate relationships are obtained, using heat exchangers
with a large number of tubes, it is recommended to use the expression
NU2==8Pe`.6{ L _-0.027 (x-1.1)0.,61
197
which is valid in the range Pe = 200-1200
x=1.1-1.4 a L/dh = 60 230;
here the oxygen content of the sodium should be <_ 5 . 10 -3 % by weight and in the
sodium-potassium alloy it should be 2.5-10 -3 1/o by weight.
5. The sodium oxides, by settling on the heat-exchanger walls, can reduce
the heat transfer coefficient several fold.
NOTATION
Bout outer diameter of tube

d.ininner diameter of tube
d hydraulic diameter of cluster
L test length of the tube
F cross sectional area of the cluster
S total wetted perimeter of cluster
x relative spacing
n (this was omitted by translator)
Q1 quantity of heat given up by the sodium
Q2 quantity of heat received by the Na-K alloy
6t average logarithmic temperature difference
k average heat transfer coefficient
al coefficient of heat transfer inside the tube
0'2
average coefficient of heat transfer in the intertubular space
w1 velocity of sodium inside the cluster's tubes
w2 velocity of the sodium-potassium alloy in the interbular space
of the cluster
198
REFERENCES
1. Borishanskiy, V. M. , et al.: In the collection Zhidkiye metally [Liquid

Metals], Gosatomizdat, Moscow, 1963.
2. Schwenk, H. C. and Schannon, R. H.: Power, 99, (11), 92 (1955).
3. Brooks, R. D. and Rosenblatt, A. L.: Mech. World and Engng Rec., 75(5),
363 (1953).
199

ANALYSIS OF THE APPLICABILITY OF AVERAGE HEAT-TRANSFER

COEFFICIENTS IN HEAT-EXCHANGER DESIGN
P. A. Ushakov
This article proves analytically that the values of the average heat trans- /204
fer coefficients for short heat exchangers found by different methods can differ
appreciably from one another. It is also proven that, for approximate design
calculations for these exchangers it is expedient to use the average heat trans-
fer coefficients determined by means of the mean temperature difference be-
tween the wall and the fluid, rather than by integration of local heat transfer
coefficients.
There exist three methods for finding the average heat transfer coefficients
in ducts with q = const, in which the initial sections of thermal and flow stabil-
ization are taken into account. According to the first method the average heat
transfer coefficient is determined by means of the mean temperature difference
between the wall and the fluid
a j =
g
1 1 (twtf)d.
(t+
t 0
(1)
Coefficient al can be used to determine the average temperature of the heat-

conducting surface.
Under the second method of averaging one integrates the local heat trans-
fer coefficients
t t
a 2 = 1 di., dx= 9 1
11
w- t
dx.
0 0 (2)
Coefficient a2 characterizes the average intensity of heat transfer and is not

suitable for determining the average temperature of the surface. In general
cx1 a2.
The heat exchanger method (q ^ const) can be used to find some average
heat transfer coefficients (third method)
1 _ 1_ 1 _RT
(ef) i K (3)
P1
R (4)
6U 0)2 K P2
200
where K is the experimentally determined average heat transfer coefficient
_ q In $1
K= $2 - 01 , (5)
where 0 1 and 02 are the temperature differences between the heat-transfer agents /205
at the inlet and outlet of the heat exchanger, R is the thermal resistance of the
wall separating the fluids, a1 and (32 are the average heat transfer coefficients
found by the first (1) and second (2) methods, respectively.
We now show that the method for averaging the heat transfer coefficient
K(x) corresponds to the formula
K= i0 K (x) dx. (6)
The solution of the heat-balance equation has the form
. t i t 2 = 01 exp ( mpKxx), (7)
atx=Z
02 = 01 exp (mpKl),
where p is the heat-transfer perimeter, t 1 and t2 are the temperatures of the

fluids in the heat exchanger;
1 1
m
- CP1 Yi U 1 - CP V2V2
x
Kx = i K (x) dx; K = Kx_i.
0
X
we note that d (xKx ) = Kxdx + AKx = Kxdx + xd z K (x) dx) = Kxdx K
x dx +
+ K(x)dx = K(x)dx. 0
201

Substituting the above into Eq. (7), we get
4 = 1 (ti -0 K (x) dx= '' exp (mphxx)

^
0
z
0
X
Xd(K.,;x)=mil [exp (nzpK1)- 1)=K '12

-"i
P In
i
Thus the coefficient K obtained in heat-exchanger experiments corresponds to

Eq. (6).
For further analysis we now assume that the local heat transfer coefficients
in the intertubular space of the heat exchanger and in the cluster of heat-generat-
ing rods [fuel elements] of the same geometry (q = const) is the same, i.e., we
disregard the effect of variations in q over the length of the heat exchanger on the
local heat transfer coefficients. We also must know the distribution functions of
the local heat transfer coefficients. We therefore assume that the heat trans-
fer coefficients for counterflow in the intertubular space (8) and in the tubes (9) /206
are described by their corresponding formulas
a _ 1
a 1exp [ n ( xo-f- x )1 ' (8)
P = 1
b 1 ex p [ n ( 1 + x p x)[ ' (9)

where a and b are the stabilized values of heat transfer coefficients, n and x0
are some constants, and- Z is the length of the heat exchanger. This form of
a function is convenient for mathematical analysis. When x 0 = 0 the local co-
efficients at the inlet sections go to infinity, a -as x 0 and 3 -as x ->
1. We introduce the following notation: N = (a/a)x =0 = (p/b)x=1 are the rel-
ative values of the local heat transfer coefficients at the point where the fluid
enters the duct, while M = (a/a)x =1 = (a/b)x =0 are the local heat transfer co-
efficients at the point where the fluid leaves the duct.
Transforming Eqs. (8) and (9), we get
u _ 1
1N exp (nx) (10)
b M-1
1 'q (nx) (11)
202

Specifying values of N and M is equivalent to specification of nl and nx 0 /207
M
n1-1n(M-1N-11
N J nxo=In N
N-1
The initial distribution of heat transfer coefficients in a counterflow heat

exchanger is shown in Fig. 1 for N = 10, M = 1.02 and N - > ^, M = 1. 02. It
can be seen that nZ changes little as N varies from 10 to
The local heat transfer coefficients can be found from Eqs. (10) and (11)
1 a exp (nx)
K (x) = 1 a lei-1 (12)
1 N-1
R f T+ a Aex p ( nx ) T M exp (2nx) N
where A = (1 + Ra + a/b).
According to Eq. (6), the average heat transfer coefficient is determined
by integrating Eq. (12)
g In T
a n1 Vg '

A{-2
M-1
a N-1
b
N-1
M A-1/d-2 b `y
a III-1
13
( )
A-} Va-2 b N A^-2 b A,
it can be seen that always 6 = A2 4 a/b `'Af 1 N `y, i > 0. The average heat trans-
fer coefficients can be found using Eqs. (1), (2), (10) and (11)
al _ q
a I I _ N-3I
M N-1)
1 if) dx 0a
dx Nhlln ( N ) (14)
(cw
l
a2- 1 a dx= 1
a 1 a N
1n ,
1
N-1
In M-1 (15)
203
a. A In the case at hand we will
a' b always have
tp
1712
1
n i
Q
=
I and 172 = 62
6
4 The sought relationship be-

2 tween the average heat trans-
2 -
fer coefficients found by
different methods [Eqs. (1)-
(4)j, has the form
0 1 2 3 17x; n(1 -x)
a
Figure 1. Initial Distribution of Heat Transfer e, _ aof 1
Coefficients Over the Length of a Counterflow -
Heat Exchanger. 1) N = 10; 2) N --> ^. = C a ) 1 C a- Ra )_a
6
(16)
E2- c ef) 2 a)^ ` h - Ra) b (17)
We recall that subscripts 1 and 2 correspond to the different methods for find- /208
ing the average heat transfer coefficients.
Let us consider some limiting cases. When M --1 (the heat exchanger
length goes to infinity) and bounded values of N the quantities
C a )i' \ a )z' C ef)i and C ef^z

go to unity, which is quite understandable from the point of view of physics.
If the thermal resistance of the wall is negligible (Ra = 0), while the heat
transfer coefficients for the fluid in the heat-exchanger tubes are much greater
than those for the intertubular space (a/b = 0) then, as N -:
1) (^/aef)1 0, since aef =K-j--, while the magnitude of cx1 is bounded;
= a2 , although both these quantities go to infinity, /209
2) (^/ ef)2 1, since of
If N -- -, while the values of Ra and a/b do not simultaneously go to zero,
then (&/af)2 -. Thus, the average heat transfer coefficients found by the
first method are always bounded, while those found by the second method go to
infinity when the local heat transfer coefficients in the initial section go to in-
finity (when x = 0, N -> -).
204
The results obtained on the basis of Eqs. (16) and (17) for M = 1.02 and
N = 100 and 10 are shown in Figs. 2a and b. For convenience in comparing
these figures the reciprocals of e 2 , i. e. , 1/e 2 are also given. It can be seen
that in all cases of practical importance (Ra > 0, a/b > 0) the average heat
transfer coefficients found by the first method differ much less from
those found by the second method. of than
It is entirely obvious that the calculations for a heat exchanger can be

performed accurately only when one knows the local heat transfer coefficients
or by using previously known average (a1 or a2 ) and effective L( ef )
coefficients. 1 or ( ef)2]
The question arises as to what will be the errors in calculations for short
heat exchangers using average heat transfer coefficients, i. e., using the formulas /210
R11
1
= + a1 +
K1
R+
RZ 22^2
0 0.4 0.8 1.2 1.6 lâ

a
&Z
Figure 2. Ratio of the Average
0.9 - 123 Heat Transfer Coefficients De-
4 termined by Different Methods
s of M =1.02 and N =100 (a) and
N = 10 (b). 1), 2) and 3) are
6
a7 Values of E 1 ; 4) 5) and 6) are
Values of 1/ s2 for a/b = 0, 0.3
D 0.4 0.8 42 1,6 Rb
b and 1, Respectively.
205
Figure 3. Heat Transfer
Coefficients Found by Dif-
ferent Methods for M =1.02
and N = 100. 1), 2) and 3)
are Values of K1/K; 4), 5)
and 6) are Values of K2/K
for a/b = 0, 0.3 and 1,
D 0,4 0,8 1,2 1,6 Rb Respectively.
Comparison of the values of K1 and K2 with the actual average heat trans-
fer coefficient K, which was performed using Eqs. (14), *(18) and (19), (Fig. 3)
shows that heat exchanger calculations using the average heat transfer coef-
ficients a1 will result in much smaller errors than calculations using
q2 An
exception are the cases when the relative thermal resistance of the walls (Ra)
and the relative heat transfer coefficients (a/b) are at the same time small
quantities.
206
RESISTANCE TO AXIAL FLOW OF FLUIDS PAST ROD CLUSTERS
A. V. Sheynina
Rod clusters in contact with a heat transfer agent flowing along their length /210
are used extensively in all kinds of equipment. It is beyond doubt that the resist-
ance to flow exerted by the clusters and by other intricately-shaped ducts cannot
always be calculated using formulas for round tubes and the hydraulic diameter.
Up to present a fair amount of experimental information has been accumu-
lated on the resistance to flow of rod clusters, [1-11]. Articles on the theo-
retical aspects of this problem have also been published, [12, 131. As a rule,
the experiments were performed in the isothermal flow of water or air. It is
proven in [14] and [15] that the flow of liquid metals obeys the same laws as the
flow of ordinary fluids (water, air, etc.). The use of ordinary fluids substan-
tially simplifies the performance of experiments. The physical properties of
liquid metals depend little on the temperature, hence, the data obtained for iso-
thermal flow of water or air can be used for the isothermal flow of liquid metals.
The available data on the flow resistance of clusters are not always in
agreement with one another. Figures 1 and 2 present experimental data for clus-
ters with triangular and square lattices, as well as curves for a round tube con-
structed on the basis of the Blasius formula.
A, = 0.316 x Re- 0.25 , (1)

and the Nikuradse formula
,% = 0,0032 + 0,221Re- 0237 . (2)
New data for clusters with relative tube spacing S/d = 1, 1. 05 and 1.4 (see /214
Figs, 1 and 2 and Tables land 2) were obtained on, a stand (Fig. 3). The working
medium was distilled water at 18-80C. The necessary pressure was produced
by the KSM-70 pump with a delivery of 70 m3/hour at a pressure of 200 m of wa-
ter. The water flowed through the heat exchanger to a measuring header, con-
sisting of four parallel Venturi tubes with nozzle diameters of 38, 25, 15 and
10 mm. After the measuring header the distilled water was supplied to either
the vertical or the horizontal test sections. Then the water was drained to mea-
suring tanks or to the drain tank. The measuring tanks were used for calibrating
the flowmeters by the volume method. The equipment and parts of the stand were
made from 1Khl8N9T stainless steel, The stand was supplied with distilled water
by the ED-2M electric distillation units. The water was heated by electric heaters
installed in the supply tank. The water temperature in the stand was measured by
copper resistance thermometers combined with an automatically balanced EMV-2
-211A electronic bridge.
207
0
ao
2.10
o
1 ^ 0
o 6 } o
------ OÔw C , tF 0 -0-
4
p
^Q ^V d GpV"Q Q
8
00
4 ^v d ,8^1^ l
17
A 2
q
Q
II'y'V
'^^jv
O.l
N r o 0
70' Z Y 0 0 RC
Figure 1. Friction Coefficients for Clusters With a Triangular Lattice, 1) and 2) Denote Curves
Obtained From Eqs. (1) and (2); the Legend Used is Given in Table 1.

A-102
f0
6 o
(S'
If
Q) q ^' 6B c^1
8
NA
a
12
Q om 1
1
0,8
0.6
103 2 /1 6 B f0 2 y 6 8 70 2 9 6 Re
Figure 2. Friction Coefficients for Clusters With a Square Lattice. 1) and 2) Denote
Curves Obtained From Eqs. (1) and (2); the Legend Used is Given in Table 1.
0
to
TABLE 1. Characteristics of Experimental Clusters
With a Triangular Lattice
u Work- Range 0 Cross sec-
N d dh ^ L L '1 tion of
med-
0. v ium Re 10 t ^
QZ model
[l1 1.0 19 176 7.81 1.81 700 Water 1-20 AV
[21 40 205 20,9 1800 Water 8-630 m
M 1,015 Air 3-30 q
1,05 37 10.,912.76' 2.16 840 Water 3-700 =F^
1.05 37 74.0 3.02 3,00 7000 Water 3-120
[4,j 112 19 12.7 4.9 4.9 620 Water 5-100 Q
[1j 1.13 19 12.0 4.8 5.0 980 Water 14-70
0 00 .
[5] 1,19 102 47 2,68 1840 Water 0.9-71 MM
[6] 1.2 19 11.016'.47 1500 Air 1-40 0
[7J 7.2 7 22.0 12.94 11.0 1050 Water 5-20

(OoD
oos,
1.4 37 70.011.6211.53 840 Water 12-120 O
181 1,46 37 159 19.5 215 1220 Water 40100 n
1.76 9.2
[9j 2.05 39 19.0 MIS 2900 . A it 6-70 Q
2.37 98.6
The pressure drop at the Venturi tubes and in the test sections was mea-
sured by type DT-50 differential manometers with 1300mm long tubes. The
manometers were filled with mercury, tetrabromoethane, carbon tetrachloride, /216
depending on the magnitude of the pressure to be measured. Pressure drops
exceeding 1200mm of Hg were measured by spring-type comparison manom-
eters. The pressure drops were measured along the length of ducts of models
as a function of the average water velocity. It was assumed that the velocity
of the latter is uniform over the model's cross section. The experimental data
were analyzed by using the hydraulic diameter of an infinite cluster.
210
Figure 3. Schematic Diagram of the Experimental Setup. 1) Centrifugal Pump; 2) Heat
Exchanger; 3) Delivery Tank; 4) Measuring Header; 5) Damper; 6) Horizontal Test Section;
7) Electric Distillation Units; S) and 9) Measuring Tanks; 10) Vertical Test Section;
11) Drain Tank.
TABLE 2. Characteristics of Experimental Clusters
With a Square Lattice
v Work- eon Cross sec-
S in
v d
N d
dh., dh,,, L med- Range ' y ^ tion of
tx d ium Re-7q-3 A a model
[101 10 2S 11,4 3.12 312 600 Water 2-200 Q
[111 40 48.2 13,2 13.2 3400 Water 5-100
08.
[51 1.106 100 8 2,88 1840 Water 1-10 0
[41 1.12 49 8,45 51 4,4 5100 gg^^'ggo

8$88
[f1 1.2 49 792 6,6 f6 if Q o88 8$
[l11 1.31 4 2S.6 172 32,0 3400 Water 6-40 q OO

00
In order to check the correctness of the measuring method, control exper-

iments were performed using a round pipe. Here use was made of a seamless,
round brass tube 30 mm in diameter and 3600 mm long. The obtained values of
the friction coefficients were in satisfactory agreement with data for round pipes.
Let us now analyze the experimental data for densely-packed rod
clusters, shown in Fig. 4. The experimental data for single ducts formed by
the triangular and square closely-packed arrangements [2, 111 are in good
agreement with data obtained for closely-packed rod clusters [1] and [10]. The
friction coefficients of triangular and square arrays were close to one another.
Approximately the same follows from the theoretical calculations of [12] and [13].
It can be seen from Fig. 4 that calculations made on the basis of the Deis-
sler-Taylor method give friction coefficients of densely-packed clusters some-
what on the low side. Calculations made by Buleyev in the region 8000 < Re /217
< 50,000 are in satisfactory agreement with experimental results. The experi-
mental data obtained by the present author lie by 30-40% lower than curves ob-
tained from Eqs, (1) and (2) and they can be approximated by an empirical for-
mula which is valid for triangular and square densely-packed arrays
I = 0.68? fr/pipe (3)
where Xfr/pipe are friction coefficients for round pipes calculated from Eqs. (1)
and (2). The experimental data of [11] for a single passage in a dense square lat-
tice are also satisfactorily (to within about 7%) described by Eq. (3)
212
Figure 4. Friction Coefficients for Densely-Packed Rod
Clusters. a) Square Lattice; b) Triangular Lattice; 1) and
2) Calculations on the Basis of Eqs. (1) and (2), Respect-
ively; 3) and 5) Curves Obtained by Buleyev, [12] for S/d =
= 1; 4), 6) Curves Obtained by Deissler, [13] for S/d = 1;
7) for S/d = 1.02. For the Notation Used see
Tables 1 and 2.
The results obtained experimentally by Palmer and Swanson [3] for a single
passage of a triangular cluster with S/d = 1.015 lie close to the curve obtained
from Eq. (1) for round pipes. It thus follows from the experiments that the fric-
tion coefficient becomes greater by about approximately 45% when S/d changes
from 1 to 1. 015. This appreciable increase in the friction coefficient attendant
to a small increase in the space between the rods is also obtained in calculations
by the Deissler-Taylor method [13]. Thus, when S/d changes from 1 to 1. 02,
the value of X increases by approximately 35%. This effect is produced by a
sharp reduction in stagnation zones brought about even by a small increase in
the distance between rods.
Let us consider experimental data obtained for two clusters with S/d =1.05.
The first cluster consisted of a bundle of 37 tubes, surrounded by a plane hexa-
gonal displacer. The size of the hexagon was selected so that the hydraulic diam-
eters of the model and the cell be the same. The second cluster has the same
number of tubes, but it is surrounded by a specially shaped displacer. The ex-
treme cells of the second cluster have the same cross section as the center cells.
The friction coefficient for the cluster with the flat [plane] displacer (Fig. 5) at
Re = (3-30)-10 3 is by approximately 30% lower than for a round pipe. At Re > /218
> 30, 000 the experimental points and the curve for the pipe tend to gradually
come close to one another. The friction coefficients for the cluster with the
4
specially shaped displacer at Re = 10 are practically identical with these co-
efficients for a round pipe.
213
A-10 z

og oo1
o'qw % 00
o
90 ^^'es
2 4 6 S rn 2 4 6 8 ins Ke
Figure 5. Friction Coefficients in Clusters With a Tri-

angular Lattice, S/d = 1. 05. . - Data for a Cluster With
a Specially Shaped Displacer; o - Data for a Cluster With
a Rectangular Shell; the Solid Line was
Obtained From Eq. (1).
It can be seen from Fig. 5 that the friction coefficients for the two clusters
differ appreciably from one another. This can be attributed primarily to the dif-
ferent shape of the cluster cross section, i, e. , effect of the shape of the shell.
Figure 6 presents results obtained on measuring the flow resistance of a
cluster with S/d ;zt; 1. 4. In this cluster a bundle consisting of 37 rods was sur-
rounded by a plane displacer. The hydraulic diameter of the model, taking the
shell into account is the same as the hydraulic diameter of the center cells. The
results thus obtained are compared with those for a cluster with S/d = 1, 46, [8].
The resistances of these clusters differ by approximately 50%. In this case, as
well as for clusters with S/d = 1. 05, the lower coefficient is that for the cluster
with the plane hexagonal shell.
Comparing the results obtained by different authors (see Figs. 1 and 2) we
can note relationships common to all of them. The friction coefficient of a clus-
ter, when using d as the characteristic dimension, depends on the relative lat-
tice spacing S/d, with an increase in which the friction coefficient increases also.
The lowest value is observed in dense clusters (S/d = 1).
The experimental results depend also on the effect of the shape of the dis-
placers on the friction coefficients in the clusters. Comparisons were made of
data for clusters with close values of S/d and with close number of tubes, but
with different shell shapes. The results of this comparison are shown in Fig. 7,
where are given the values of the friction coefficient of certain clusters at Re =
= 40. 10 3 as a function of S/d. It can be seen from Fig. 7 that the friction coef-
ficients depend on the shape of the shell; here X of clusters with a plane shell
is lower than in clusters with a specially shaped displace r. Thus, analysis of /219
experimental data shows that the resistance of the clusters depends on the shape
of the shell. This problem requires special study,
214
A1q
n
Z
8
1n Z 6 ,C 20
Figure 6. Friction Coefficients in Clusters With a Trian-

gle Lattice, S/d 1.4. o - Data Obtained by the Present
Author for S/d = 1.4; n - for S/d = 1. 46, [8]; the Solid
Curve was Obtained From Eq. (1).
R10^
The problems involved in
3,5 determining the flow resistance
3 of clusters with different spac-
2,5
ing devices are more complicat-
ed. In these cases the flow re-
2 sistance is affected by the geom-
etry of the spacing elements
proper. The most extensively
?5
used and simplest method for
1 7.7 7,2 43 1,4 S/d spacing the rods are spiral fins.
A study was made of the flow re-
Figure 7. Friction Coefficients in Clusters sistance of finned clusters, pro-
as a Function of S/d for Re = 40 . 103 , o vided with two and four ribs each
Data for Clusters With Hexagonal Shells; wound on the outer surface of the
rod.
A - With a Specially-Shaped Shell; - for
a DenselJ -Packed Cluster the Dash-Dot
Line was Obtained Theoretically for a The experimental models
Round Pipe, consisted of clusters of tubes
or rods, surrounded by a spec-
ially shaped displacer. The
ribs from the surfaces of ex-
treme tubes, facing the displacers, were removed. In analyzing the experimental
data the characteristic dimension was the hydraulic diameter of a dimensionless
cluster, taking the fins into account. The basic characteristics of the tested cluster
are shown in Table 3.
The most complete tests were performed using clusters with a triangular
lattice with S/d = 1.05. It was discovered for this spacing that the number of
ribs and their lead affect [the friction coefficient]. Clusters with two and four
fins were tested. For clusters with four fins the experiments were performed
with fin winding leads T = 1400, 300 and 70 mm. The friction coefficients for
clusters with T = 1400 and 300 mm differed little from one another (Fig. 8a).
The same was observed for clusters with two ribs. For winding leads exceed-
ing 300 mm (T/d > 20), the difference between the friction coefficients of clusters
215
with and without ribs with S/d = 1. 05 is inconsequential and lies within the limits
of experimental error (Figs. 8a and b).
^ 10'
a
n
0
d 0 a
2 fag s ^.
a
&p
xo ..
7 y a 3
d101
o e b 8
S
o a
o a _
Figure 8. Friction Coefficients for Finned Clusters.

1) Curve Obtained From Eq. (1); 2) S/d = 1. 05 (ribless),
3) S/d = 1.1 (ribless), [1] ; the Notation is That of Table 3.
Similar results were also obtained before. The experiments were per- /220
formed with finned clusters with SO = 1,1 and 1..15. Two clusters with S/d =
= 1. 1 (see Table 3, rows 1 and 2) had fin winding leads T = 400 and 300 mm.
The experimental data for these clusters lie by: 1,0-20% above the Blasius curve
for a round tube (Fig. 8c). Changing the fin lead from T = 400 to T = 300 (T <
< 20d) did not appreciably affect the friction coefficient.
The finned cluster with S/d = 1. 15 (Table 3, row No. 8) consisted of 19
tubes 12 mm in diameter. Each tube was equipped by four fins with a winding
216
TABLE 3. Characteristics of Ribbed Clusters With
a Triangular Lattice
Range .tea Rib
N d d ,^ L
.1 ^ hp shapes
d
Z.a Re10s M a
1 411 19 10,6 3.9 2 300 40-100 300 as @as
2 1.11 19 12.6 3.9 2 300 O 400 0, 60 6

8-100
3 105 37 14.0 2.88 2 1000 2-100 A 1400 0.35 x0.65
4 1.05 37 14,0 2.88 2 1000 2-100 A 300 0.35 0 6S
5 1,05 37 14; 0 276 4 1000 2-100 O 1400 0,35 6S
6 1.05 37 14;0 2,76 4 1000 2-100 q 300 0.35 ^0 05
7 1.05 37 74.0 2,79 4 600 2-100 70 0,35 0.85
8 1,15 19 12,0 4.W 4 800 10-50 Q 700 0,9 00.45
lead of T = 700mm, The friction coefficient for this cluster is higher than for
a round tube by 30 %. The results obtained with a ribbed cluster with the same
value of S/d are being compared (see Fig. 8a) with data for a smooth cluster
(S/d = 1.13) [1].
It can be seen from Fig. 8, that the presence of spacing fins wound spirally,
has little effect on the resistance coefficient at T/d < 20, which is also true with
respect to the effect of the number of ribs on the magnitude of X. When the wind- /221
ing lead is reduced to 70 mm (5d) in the cluster with S/d = 1, 05, the value of X
has increased by more than a factor of two.
The curve of X as a function of T, obtained for clusters with four ribs at
S/d = 1. 05, is shown in Fig. 9 in the coordinates 100 X Re 0.25 = f(T/d), It can
be seen that, when the winding lead is reduced in a range of from 100 to 20 d,
X almost does not change. A sharp increase in l is, [however, ] observed at
low T.
Spacing devices of two more types were also studied. One of them was a
"gimlet"-type insert. The experiments were performed with a cluster with /222
S/d = 1. 13, with 19 rods 13.6 mm in diameter and 1000 mm long, surrounded
217
A70 Z.
Re 0,25 by a specially-shaped displaces.
ô A "gimlet", i. e. , a plate 0.8 mm,
.0 3.6 mm in diameter and with a
twisting lead P = 30 and 50 'mm
0 Circular tube was placed into each cell of the
cluster.
0
- 7 5 70 s0 T/d
In the second case the
Figure 9. The Friction Coefficient of a spacing device was a steel wire,
Finned Cluster as a Function of the Fin on which streamlined sleeves,
Winding Lead, i. e., "beads", were placed at
uniform intervals. Up to 30
Re-10- 1 10 20 40 60 so 1 00 beads were placed in each cell.
notation O -+- * o q The characteristic dimension
used was the hydraulic diam-
eter of the model, without tak-
ing into account the spacing devices. The resistance coefficient was referred
to the arbitrary average velocity of the model without taking into account the
thickness of the latter [beads].
A.10z
As^
M,N;
00 ^^^gn, .a I.,m I' t

o ova_ gya`
0 om
o
1
70 Is' 2 3 4 5 6 7 8 9 10 5 Re
Figure 10. Resistance Coefficient for Clusters With

a Triangular Lattice With S/d = 1. 13, Provided With
Spacing Devices. o - Without Spacing Devices; n -
"Bead"-Type Spacer; L. and s "Gimlet"-Type Spacer,
P = 50 and 30 mm, Respectively; the Solid Line was
Obtained From Eq. (1).
The experimental results which are convenient for practical calculations /223
of hydraulic losses for these two types of inserts, as well as data for a cluster
without spacers are presented in Fig, 10. The change in the twisting lead of the
gimlet from P = 50 to P = 30 did not appreciably affect the resistance coefficient
The use of the "bead"-type insert increased the flow resistance coefficient two-
fold, while the use of the "gimlet"-type spacer increased it by a factor of three
as compared with the cluster without spacers. In both cases the resistance
218
coefficient is some arbitrary quantity, since the calculations were performed
using an average velocity and hydraulic diameter, calculated without taking the
spacing inserts into account. It is difficult to take into account the closing of
the cross section by the inserts, since the configuration of the inserts varies
over the length of the duct.
REFERENCES
1. Subbotin, V.I., et al.: Atomnaya energiya, 9, Issue 4, 308 (1960).

2, Levchenko, Yu. D., et al.: Raspredeleniye skorostey v yacheyke plotno
upakovannogo puchka sterzhney [Velocity Distribution in a Cell Formed
by a Densely-Packed Cluster of Rods]. Article immediately following
the present one.
3. Palmer, L. D,' and Swanson, L. L.: International Development in Heat
Transfer, Boulder, London, III, No. 63, 1961-1962.
4. Le Tourneau, B. , et al.: Trans. ASME, 79, No. 8, 1751 (1957).
5. Wantland, J. L.: Reactor Heat Transfer Conference of 1956, AEC Report
TID, 7529, 525 (1957).
6. Mikhaylov, A. I., et al.: Inzhenerno-fizicheskiy zhurnal, No. 11, 42 (1964).
7. Firsova, E. V. : Inzhenerno-fizicheskiy zhurnal, No. 5, 17 (1963).
8. Miller, P., et al.: A. I. Ch. E. Journal, 2, No. 2, 226 (1956).
9. Salifov, A. P., et al.: Teploenergetika, No. 8, 481 (1954).
10. Ushakov, P.A., et al.: Atomnaya energiya, 13, Issue 2 (1962).
11. Gunn, D. J. and Darling, C. W.: W. Trans. Instn. Chem. Engrs, 41, No. 4,
163 (1963).
12. Buleyev, N. I., et al.: Teplofizika vysokikh temperatur, 2, No. 5, 749 (1964).
13. Deissler, R. G. and Taylor, M. F. : Reactor Heat Transfer Conference of 1956.
TID-7529 (1957).
14. Borishanskiy, V. M., et al.: In the collection Zhidkiye metally [Liquid Met-
als], page 71, Gosatomizdat, Moscow, 1963.
15. Kondrat'yev, N. S.: In the collection Teploperedacha i teplovoye modeliro-
vaniye, page 93. Published by the Academy of Sciences of the USSR,
Moscow, 1959.
219
VELOCITY DISTRIBUTION IN A CELL FORMED BY A
DENSELY-PACKED CLUSTER OF RODS
Yu. D. Levchenko, V. I. Subbotin, P. A. Ushakov
and A. V. Sheynina
Experimental study of the velocity distribution in ducts with a nonround /223

cross section are needed for developing reliable methods for calculating thefluid
dynamics and heat transfer in various heat exchanger devices, including also nu-
clear reactors.
Dense packing of rods in a triangular lattice makes it possible to develop /224
reactor cores with low porosity with respect to the heat-transfer agent (for ex-
ample, for fast-neutron reactors). The closeness of the rods creates flow char-
acteristic similar to those in ducts with extended acute corners. Dense packing
(S/d = 1) is the limiting case of placing rods in a cluster, for which the effects
of velocity and temperature variations over the perimeter are most strongly
pronounced.
Eckert and Irvine [1] have studied the flow in a triangular passage and have
discovered a number of peculiar features. In the flow one can observe a nonuni-
form development of turbulence over the passage's cross section. Transition
from laminar to turbulent flow ensues at lower (than in a round tube) Reynolds
numbers, calculated on the basis of the hydraulic diameter and the average ve-
locity. In one section of the channel it is possible to have, alongside with a tur-
bulent main body of the flow, regions with laminar flow. Changes in the temper-
ature of heat-conducting walls point to the fact that the turbulent heat transfer is
nonisotropic, i, e. , heat transfer in the direction of the bisectrix of the angle
proceeds with difficulty.
A characteristic feature of ducts with sharp corners was the large length
of the fluid-dynamic stabilization section. On the basis of measured velocity
distributions in ducts with nonround cross section, I. Nikuradse [2] has sug-
gested the existence of secondary fluid flows.
Of [special] interest for theoretical studies of fluid dynamics in the flow of
fluids around rod clusters is work performed by Deissler and Taylor [3] and by
Byleyev, [4]. The methods due to Deissler and Taylor are built on simple as-
sumptions, and can be used by any engineer for practical calculations. [On the
other hand] for calculations using the three-dimensional model of turbulent trans-
fer developed by Buleyev one must use computers. In the case at hand bothmeth-
ods yield close results, but the future apparently belongs to methods based on
more general assumptions than those made by Deissler and Taylor.
Despite the great scientific and practical interest, very little is known
about the velocity distributions in ducts with nonround cross section. The staff
220
of the Physics and Power Institute has performed measurements and calculations
of the distribution of velocities and tangential stresses, as well as friction resist-
ance coefficients in a duct simulating a densely-packed rod cluster.
A densely packed rod cluster is the only cluster in which it is possible to
isolate an individual cell in the pure form. Hence the studies were not performed
with the cluster of rods, but in a duct, specially constructed for this purpose, sim-
ulating a cluster of rods. The duct was made from tubes fabricated from 1Khl8N9T
steel 219 x 18 mm in diameter. After turning on a lathe to a diameter of 205 x 11
mm with surface finish V5 the tube was cut lengthwise into three parts. The sec-
tors [thus obtained] were welded together by argon-shielded are welding. The air /225
was supplied from a special pump. The experimental stand used for the experi-
ments is described in [5]. The dynamic pressure was measured by a glass Pitot
tube 0.20 x 0.12 mm in diameter, placed directly at the duct's outlet. The static
pressure was measured through a hole in the duct's wall 0.6 mm in diameter, sit-
uated a distance of 35 mm from the outlet flange. This hole, as well as the others,
was drilled before the duct was welded together, in order to provide for the max-
imum possible precision and quality of finish.
Measurement of velocities in the outlet cross section of a duct has a number
of advantages. The Pitot tube can be brought to the wall at a distance equal to
half its diameter, using the KM-6 cathetrometer [sic] to rigorously control the
distance from the wall, to move it easily into any point of the flow, etc.
The dynamic pressure drop was measured by a glass differential manom-
eter filled with alcohol. The alcohol level in the manometer was measured by
the KM-6 cathetrometer to within 0.05 mm.
The Pitot tubes were made from plain glass, heating it and sharply ex-
truding from tubes 4 mm in diameter. From a large number of tubes which were
thus drawn we have picked only those which had a ratio of the inner to the outer
diameters of 0.595 to 0.605. These tubes had sufficient rigidity and a relatively
moderate inertia. Then these glass capillaries were polished at their ends by
fine emery cloth, so that the end of the capillary was made perpendicular to its
axis.
The Pitot tube was moved by a coordinate spacing device using micrometer
screws with scale divisions of 0. 01 mm. The Pitot tube could be moved in mu-
tually perpendicular directions over distances of 100 x 100 x 25 mm. Due to sym-
metry of the cell produced by triangular packing of the rods, it is sufficient to
obtain the velocity distribution in 1/6 of the cross section, however, in order to
clarify whether the fluid flow was distributed uniformly, we measured the veloc-
ity distribution at Re = 14.2 . 103 in a region embracing half the cross section.
From the dynamic pressure obtained in each point we have calculated the
local velocity
U^C ^/ f air ^1)
221
where h = h(y = h(802 -y airand C is a calibration factor, which was
alc yair)
set equal to unity since the Pitot tube was made according to recommendations
of [6].
The tangential stress at the wall T O was calculated on the basis of velocities
measured by a Pitot tube 0.20 x 0. 12 mm in diameter near the wall of the duct
assuming that the velocity at the wall is zero and that it increases in the region
near the wall in proportion to the distance from the wall; the local tangential /226
stress at the wall can be calculated from the expression
ay_ U
Within the limits of a linear velocity distribution along a perpendicular to the wall,
i.e., for &U/8y = const, the value of T O calculated from the above expressiondoes
not depend on the distance from the wall. Since only the Pitot tube was placed at
large distances, where 8U/8y ^ const, the experimental values of TO were found
to be on the low side. Hence the curve showing the distribution of TO over the
duct's perimeter was drawn using the values of local tangential stresses calcu-
lated from Eq. (2).
In calculating the tangential stress at the wall no corrections were intro-
duced to compensate for the distortion of flow by the Pitot tube, the viscosity
of the fluid and the high velocity gradient. Introduction of these corrections, [7]
has increased the scatter of the experimental points. The average tangential
stresses TO thus obtained differed from the TO determined using the flow resist-
ance of the passage. It is possible that these corrections are applicable only
for large-diameter Pitot tubes and are not suitable for the tubes we have used.
The flow-resistance coefficient was measured using distilled water. The
design of the experimental setup is described in [8]. The pressure drop over
the length of the passage was measured by differential manometers, in which
mercury, tetrabromoethane and carbon tetrachloride were used as the working
media. The friction coefficient l was calculated on the basis of the average ve-
locity, temperature and pressure drop over the length of the passage.
The results of velocity distribution measurements in a passage formed by
a closely packed rod cluster are presented in the form of dimensionless isotachs
U/tJ in Fig. 1. For the same pressure drop over the length of the passage which
was obtained in our experiments, we have calculated the velocity distribution us-
ing the Deissler-Taylor method [3]. There were no qualitative differences be-
tween the two above distributions, for which reason the calculated velocity dis-
tribution is not shown.
Our measurements have shown that the velocity distribution is sufficiently
symmetrical, which justifies performance of measurements only in 1/6th of the
passage cross section.
222
Figure 1. Velocity Distribution (Isotachs) in a Cell Formed
by a Rods of a Closely-Packed Cluster (Experiments
With Air for Re = 14, 200).
For convenience in using the experimental data, Fig. 2 shows the velocity
distribution in the form
UN = f (y/Jmax)
The parameter [here] is angle co, measured from the point where the rod are in
contact. Judging by Fig. 2, the velocity distribution at small angles comes close
to the parabolic, which justifies the assumption of laminar flow in the corners
of the cell. This figure also compares the experimental and theoretical results
and it also displays previously unpublished results obtained by Buleyev, Calcu- /227
lations using the Deissler-Taylor method yields local velocities in the center on
the high side and velocities in the corners of the duct on a low side, which is also
obtained when using the Buleyev method, Consequently, the 'turbulent flow is a
duct formed by closely packed rods of a cluster actually occupies a larger area
than is shown by theoretical methods.
The average velocity obtained from the velocity distribution calculated by
the Deissler-Taylor method is by approximately 10% higher than the average ve-
locity determined in experiments for the same value of the pressure gradient
dp/dz = 5, 22 kg/m2-m.
223
U
Cr 2y -- 28.5.
3
7 . 79.5
X'
/ ,, ^2.,.
ti
i
0,5
10.5
i 3

0 45 Y/Ymax as Y/Ymax
a b
Figure 2. Velocity Distribution Along Normals to the Sur-
face of a Cell Formed by aDensely-Packed Rod Cluster.
1) Experimental Data; 2) Data Obtained Using the Deissler-
Taylor Method; 3) Data Obtained Using Buleyev's Method.
Figure 3 displays the velocity distribution over the line of symmetry of the
passage (along the bisectrix of an angle) obtained theoretically and experimentally.
This figure, as is the case of Fig. 2, shows that the theoretical calculations yield
lower values of velocities near a corner of the duct than the experiments. On the
other hand, the velocities in the center part of the channel, calculated by the Deis-
sler-Taylor method come out higher than actual.
The results of measurements and calculations show that in a ductproduced /229
by a densely-packed rod cluster there exist regions with very low velocities -
"stagnation zones", the boundary of which corresponds approximately to an an-
gle cp = 70.
The velocity distribution in the cell was also measured at Re = 37.5 . 103
(Fig. 4). The authors were alarmed by the large distortions in the isotachs.
Apparently additional work must be done to finally determine the pattern of ve-
locity distribution for a given mode of fluid flow.
Distortion of isotachs are usually attributed to the effect of secondary flows
In the fluid, directed along the bisectrix into the corner of the duct. If subse-
quent studies will confirm the results of preliminary measurements, then the
presence of secondary flows in cells formed by closely-packed rod clusters will
be proven.
Secondary flows can ordinarily be detected in a fluid by measuring the pres-

sure distribution over the duct cross section. However, due to the fact that the
expected pressure drops are small, a special method must be developed for
224
U
U
o
( 0 1
I
o
o^ i
1
I
r^ N 0
o
I
II o
0 0
0..1
I_
I
I
I
I
I
0 0.25 0,50 0.75 1 Y%Ymax

Stagnation
zone
Figure 3. Velocity Distribution Along the Line of
Symmetry of a Cell Formed by a Densely-Packed
Rod Cluster, o - Experimental Data; - Calcu-
lated by Buleyev; the Solid Line Denotes Curve Ob-
tained by Using the Deissler-Taylor Method.
Figure 4, Velocity (Isotach) Distributions in Cells Formed

by a Closely-Packed Rod Cluster for Re = 371500.
measurements of this sort. The authors did not as yet succeed in obtaining sat-
isfactory information using ordinary static pressure probes,
225
r, Figure 5 shows in dimension-
to 0 less coordinates the tangen-
0 tial stress distributions over
^s the duct perimeter; here TO
0 o
o o has been calculated on the ba-
1,2
0 ;013o sis of the measured flow re-
0 0 sistance of the duct
i o
0.8
i
dh, dp
0 TO r 4 * dz (3)
o.^
Rio
the values of TO for co < 9 0 were
10 15 20 25 cp, deg
obtained by extrapolation. In-
Figure 5. Distribution of Tangential Stresses tegrating experimental curve 1

at a Wall Along the Perimeter of a Cell Form- graphically (see Fig. 5), we
have found that
ed by a Densely-Packed Rod Cluster for Re
= 14, 200. o - Experimental Data; - Points
Calculated Using the Measured Velocity Dis- 1
6
tribution and the von Karman Universal J1, 1 6 ' To'To =- 0.96.
Logarithmic Law. u
Thus, the value of TO obtained by measuring the local tangential stresses differs /231
from the value of TO obtained on the basis of the flow resistance by approximately 4 %.
Figure 6 compares the distribution curves over the perimeter of the pas-
sage obtained experimentally and theoretically upon using the Deissler-Taylor
and Buleyev's methods, It can be seen that the theoretical methods yield higher-
than-actual local tangential stresses in the center and lower-than-actual stresses
near the narrow part of the passage. The slight difference between the calculated
TO/T0 obtained by the authors and those taken from [31 is apparently due to sub-
jective errors in graphical solution of the problem and hence is not of importance.
The theoretical method due to Deissler and Taylor is based on the assump-
tion that the distribution along a perpendicular to the wall is governed by the uni-
versal logarithmic law. This hypothesis was checked by measuring the velocity
distribution and the distribution of tangential stresses at awall with Re = 14.2. 103
The experimental points were plotted in Fig. 7 in coordinates U here the /232
local values of T O for the applicable angle were substituted into the expressions
for these two quantities. The scatter of points near the wall (y+ < 30) is due to
the fact that a large contribution in this region is made by errors in determining
the distance to the wall. The experimentally obtained velocity distributions along
a perpendicular to the wall can in the first approximation be described by Von Kdr-
man's logarithmic law.
226
zp ------T --- On the basis of the above
Zp
f ^{ considerations we have calculat-
3 i^ ed the distribution of T O / T 09 the
2
tangential stresses over the pe-
rimeter of the duct, makinguse
1.6, of velocity distributions away
f
^ from the wall, obtained exper-
I ^
imentally and of Von Fhrmants
1.2 universal law, When TO/T0
are calculated by this method
0.8
i
the errors in measuring ve-
locities and distance from the
wall, which are quite appreci-
0.4 able in the region close to the
wall, are eliminated. The val-
10
I
15
I
20
LA
25 tP, deg
ue of T O was determined from
the expression
Figure 6, Tangential Stresses Determined io
Experimentally and Calculated Analytically. Ul P = 5.5 + 5.751g y v P (4)

1) Experimental Results for Re =14,200;
2) Calculated by Buleyev for Re = 9400 and
26, 000; 3) Calculated Using the Deissler- by substituting the values of
Taylor Method for Re = 15,400; 4) Calcula- velocities at a point in the flow
tions Made by Deissler and Taylor for and the distance from the given
20, 000 < Re < 900, 000, [3]. point to the wall. Equation (4)
was solved graphically. The
values of TO for given angles
co calculated on the basis of several magnitudes of velocity and corresponding
distances from the wall were in good agreement.
The results obtained from Eq. (4) were plotted in Fig. 5. They are , in sat-
isfactory agreement with experimentally obtained distributions of T0 . However,
it cannot as yet be claimed that this will also be true for Re > 14.2 . 103 , It is
possible that for higher Reynolds numbers the suggested secondary flows will
affect the distribution of the tangential stresses.
In experiments with distilled water we have measured the pressure drops
over the length of the passage and determined the frictional resistance coeffi-
cients X (Fig, 8). When processing these data using the hydraulic diameter the
experimental points plotted in logarithmic coordinates lie parallel to the straight
linescalculated by the Blasius formula: X = 0.316 Re-0, 25 for Re < 105 and by
the Nikuradse formula: X = 0.0032 + 0.221 Re-0.237 for Re > 105 . The values
of the friction coefficient for the passage at hand are lower than those for round
tubes by approximately 40%. This is in agreement with results obtained pre-
viously in experiments with densely packed rod clusters, [9]. If the diameter
of the circle inscribed into a cell formed by the cluster is taken as the charac-
teristic dimension, then the experimental results are in approximate agreement
with formulas for round tubes, as is also the case in [9] (see Fig. 8).
227
U I --
go
Q3
1 Q3
15 + ^ r
e
10
no
r ^
S o 2 -
3 r
i
Q ^
2
lg Yy
Figure 7. Comparison of Experimental Data on the

Velocity Distribution in Cells Formed by a Densely-
Packed Rod Cluster With the Universal Logarithmic
Law, 1) U+ = 5.5 + 5.75 lg y; 2) U+ _ -3. 05 + 11.53
lg y; 3) U =y+ ; +, s, , q ,', , 7 and o Were Obtained
for co = 10. 5, 13. 5, 16. 5, 19, 5, 22. 5, 25, 5, 28, 5 and
30', Respectively,
It can also be seen from Fig. 8 that calculations using the Deissler-Taylor
method yield values of the frictional resistance coefficient in densely-packed rod
cluster by 7-10% lower [than actual].
In order to find the friction coefficient at Re = 14, 2 . 103 , we have used the
data on velocity distributions, obtained in experiments with air, from which we
calculated the average velocity. The pressure gradient was measured directly
in our experiments. In addition we have calculated the friction coefficient using
the measured distribution of tangential stresses over the passage's perimeter.
As is shown by Fig. 8, the, values of friction coefficients found by the different
methods are in agreement. The theoretical limiting errors in measuring the
velocity distributions comprised about -2.5% and for the tangential stresses
they were about 10 %. The friction resistance coefficients were determined
to within approximately + 7 %.
228
A . 102 00.
6 3 'M' 2 4 6 B ens 2 4 If Ba
Figure 8. Coefficients of Frictional Resistance for a

Cell Formed by a Closely-Packed Rod Cluster. 1) and
2) Were Obtained From Blasius and Nikuradse Formu-
las for Round Pipes; 3) Represents Calculation of
Deissler and Taylor; o and Denote Experiments
Using Water; the Characteristic Dimensions are the
Hydraulic Diameter and the Diameter of the Inscribed
Circle, Respectively; the Cross Denotes Measure-
ments of Velocity and Pressure Drop Distributions in
Air; o Denotes Measurements of the Velocity and Tan-
gential Stresses Distributions in Air; n Denotes Theo-
retical Results Obtained by the Deissler-Taylor Method.
NOTATION
d diameter of rods in the closely-packed cluster

S distance between rod axes in the cluster
d hydraulic diameter of the duct
h dynamic pressure
y distance from the wall, measured along a normal to it
y distance from the wall along the line of symmetry of the
max duct, along the normal to the wall
y' distance from the wall in directions parallel to the axis of
symmetry of the duct_
cp angle, measured from the contact line of the rods
229
X friction coefficient
U+__ U velocity
,4'J T O
P
y_ v
.Tp
distance from wall
P
REFERENCES
1. Eckert, E. R. G., and Irvine, T.F.: Trans. ASME, v. 82, Series C, No. 2 (1960). /234
2. Nikuradze, I.: Ingenieur-Archiv, 1, 30 (1930).
3. Deissler, R. G. and Taylor, M. F. Reactor Heat Transfer Conference of
1956, TID-7529 (Pt. -1), Book 2, 416 (1957).
4. Buleyev, N. I.: Teploperedacha [Heat Transfer], page 64. Published by
the Academy of Sciences of the USSR, Moscow, (1962) .
5. Ibragimov, M. Kh. , et al.: Raschet i eksperimental'noye issledovaniye
poley skorosti v kanale slozhnoy formy [Calculation and Experimental
Study of Velocity Fields in an Intricately-Shaped Passage]. Article
immediately following this.
6. Popov, S. G.: Izmereniye vozdushnykh potokov [Air-Flow Measurements].
Gostekhizdat, Moscow (1947).
7. Wuest, W.: Archiv ftir technisches Messeu, 116, No. 10 (1961).
8. Levchenko, Yu. D., et al.: Gidrodinamika v kol'tsevom zazore s prodol'
nymi rebrami [Fluid Dynamics in an Annular Space with Longitudinal
Ribs]. Part I of present collection.
9. Subbotin, V. I. , et al.: Atomnaya energiya, 10, Issue 4, 308 (1960).
230
CALCULATION AND EXPERIMENTAL STUDY OF VELOCITY FIELDS IN
INTRICATELY-SHAPED PASSAGES
M. Kh. Ibragimov, I. A. Isupov and V. I. Subbotin
Modern heat-engineering facilities and devices use passages of various /234

shapes, in particular, clusters of rods surrounded by a shell. The heat trans-
fer processes in such passages depend substantially on the velocity distribution
in the coolant flow. The present article considers the fluid dynamics of a pas-
sage formed by a cluster of three round rods, surrounded by a triangular shell.
Such a passage combines fluid-dynamic peculiarities in the rod clusters, as well
as the effect of the shell on the fluid-dynamic characteristics. The above flow
characteristics were determined theoretically and experimentally.
Calculating the Velocity Field and the Friction Resistance Coefficients

Deissler and Taylor [1] have proposed an approximate graphical-analytic
method for calculating the velocity fields and the resistance coefficient in intri-
cately-shaped passages. The passage cross section is broken up into a number
of areas, bounded by walls and by velocity-gradient lines. Along with as-
sumptions on the existence of hydrodynamic stability and invariability of the
physical parameter of the fluid, two basic assumptions are made. It is assumed
that, first, the velocity distribution along the normal to the wall in the intricately-
shaped passage is describable by the universal law for a round pipe and, second,
that no transfer of momentum takes place between the fluid volumes selected in
the above manner. The problem is solved by successive approximations.
Deissler and Taylor have used their method for infinite clusters of rods,
where the position of the line of maximum velocity is entirely obvious. In the
passage under study the position of this line for some parts of the passage can- /235
not be specified, due to the shape of the cross section, which introduces addi-
tional difficulties in the calculations.
The calculations were performed for 1/6 of a symmetrical part of the cross
section. In order to increase the accuracy in calculating the areas, the part of
the cross section under study was magnified eight-fold. The perimeter was broken
up into parts equal to the length of arc subtended by 5 angle in the rod. Fig-
ure 1 shows the cross section of the passage and the velocity gradient lines in
the first approximation. The initial position of the maximum velocity line was
taken through points equidistant from the walls. Our experiments show that this
assumption is satisfactorily justified.
The calculations were performed for air under the following conditions:
Re = 62 . 103 , r++ = 2210 (r = dh/2), A I = 3.72 mm, Pbar = 760 mm of Hg, and
231

a b
Figure 1. Shape of Passage's Cross Section (a) and

Diagram of the Gradients (b).
t = 20C Parameter r++ was selected according to a preliminary estimate of

the magnitude of dp/dx, the pressure gradient, on the basis of X calculated by
the Blasius formulas at Re = 62 . 103 . We have assumed that r = d h/2 since, as
was shown by calculations, the selected magnitude of the characteristic dimen-
sion r does not affect the end results of the calculations.
In [1] the magnitude of r was taken as equal to the rod radius. The calcu-
lations were made by specifying values of u++ from 9 to 22. The dimensionless
oA^^ and distances yi were calculated for each value of u++,
velocities u. = ' ju^
V ^lr
Use was made here of the universal velocity distribution due to Von Karman
y + < 5 u T Y+
5<y+<30 cc + 11.51"y+-3.05
Y+ Z 30 u+ = 5.751g u + ;- 5.5. (1)
The values of yi thus calculated were used to determine distances yi /236
Y r
Yi = AAA
r++
01r
From the midpoints of the corresponding intervals were constructed normals,

on which the values of u++ were laid off as functions of distance y.. i
232
In area segments with highly curved boundaries some velocities u ++ cal-
culated for the given area segments were found to be located in neighboring
area segments. The isotach grid thus obtained was compared with the grid of
velocity gradients, their mutual orthogonality was estimated visually and the
gradient lines were adjusted. It came out in the process of calculations that
the maximum velocity line in the wide part of the passage is somewhat displaced
toward the rod. To obtain this line more precisely we have made six approxi-
mations, while two approximations were thought to be sufficient for the remain-
ing part of the passage.
The isotach grid which was finally obtained was used to calculate the av-
^^n11
!J uhf+AAI,
erage velocities u + = k=1 AA1 in each elementary area, the average velocity
1
^N1
GJ u1+'AAZ
u0+
'-' A O = 14.7 over the passage cross section, the Reynolds number /237

Re =u+
0o r++. d = 65 . 103 and the friction resistance coefficient X =2-d h
r ++2
r u0
= 0.018.
Figure 2 shows a theoretical velocity distribution in the form of dimension-
less isotachs u/u0 . It can be seen that the velocity of the fluid changes substan-
tially over the passage cross section. The point of maximum velocity (1) is dis-
placed by 6/y max 12% to the center, while the maximum velocity point in nar-
row parts (in corners of the passage) is displaced toward the rod by 61y 0 5 %.
To determine the validity of the theoretical data obtained and to determine

the boundaries of applicability of the Deissler-Taylor method, we have experi-
mentally studied the velocity distribution, the tangential stresses at the wall and
the position of the maximum velocity line.
Experimental Study of the Fluid-Dynamic Characteristics of a Passage
The experiments were performed with an air stand, which is schematically

depicted in Fig. 3. The air, the flow rate of which was measured by a pre-cali-
brated normal diaphragm, and the temperature of which was determined by a mer-
cury thermometer with a scale division of 0.1C, was supplied to the test section
by a fan with a nominal delivery of 800 m3/hour at a pressure of 750 mm of water.
The length of the test section is about 50d
hl the cross section of its passage is
shown in Fig. 1a. The inner surfaces of the passage were machined to class six
surface finish. The rods could be moved, by screw-type devices, along the bi-
sectrixes of the shell's corners. The length of travel was measured by rulers
with verniers.
233
1.21
Figure 2. Theoretical (Using the Deissler Method) Distri-

bution of Relative Velocities u/u0 ; Re = 65 . 103,
Figure 3. Schematic Diagram of Experimental Setup.

1) Fan; 2) Slide Valve; 3) Differential Manometer; 4) Dia-
phragm; 5) Thermometer; 6) Mesh; 7) Damping Tank;
8) Experimental Passage; 9) Coordinate Spacer;
10) Micromanometer.
The static pressure was measured in five cross sections of the passage /238
through holes in the shell 0.5 mm in diameter (the ratio of the wall thickness of
the shell to the hole diameter was 6). In drilling the holes particular attention
was paid to their perpendicularity to the wall and to removal of burrs. Five
pickups were made in a cross section located at 'a distance of 33 mm from the
outlet, and these were used for measuring the velocity fields. The remaining
cross sections had two pressure pickups each. All the pickups of a given cross
section were connected for more reliable pressure measurement.
234
The local velocities were measured by a total pressure metal tube (Pitot
tube) with a working part diameter of 0.45 x 0.15 mm and length of 12 mm, the
end of the tube was rounded to a half-spherical shape. The pitot tube was sol-
dered into a reducing tube, serving as a holder, 6 mm in diameter which was
fastened in the coordinate spacer, and which could be used to move the holder
in four ways: vertically (accuracy of 0.05 mm), horizontally (0.1 mm), rotation
in the vertical plane (0.5) and rotation about the horizontal axis (0.2).
The velocity distributions were measured in a cross section removed by
10 mm from the outlet end.
The distributions were studied in detail only in 1/6 of a symmetrical part
of the cross section. In the remaining parts only control measurements were
made, which have shown that the flow in the passage is satisfactorily symmet-
rical. The velocity distributions were measured along the normals to the flat
part of the shell wall (see Fig. 1a, coordinate y) and selectively along normals
to the rod surfaces. The pressure drops were measured by a type MMN in-
clined-tube micromanometer.
The velocity fields were measured in passages of two shapes: with S/d =
= 1.15 (Fig. la, initial geometry) and when the rods were moved symmetrically
to S/d = 1.3. In both cases the Reynolds number was the same and equal
. 3
to 39 10 . The flow resistance of the passage was measured in the interval of
. 3 3
Reynolds numbers from 18. 3 10 to 39. 10 with the rods moved symmetrically
by different distances.
The study of tangential stresses at a wall by measuring the velocity dis-
tribution in the viscous sublayer requires the use ofathinner Pitot tube and utili-
zation of more sophisticated experimental methods. In our opinion the method
presented in [2] is simpler. For the case when the Pitot tube lies on the wall
in the dynamic similarity layer, Preston has established a relationship be-
tween the tangential stress at the wall and the dynamic pressure difference
P P0 , measured by the tube
(P
P z)- t FC OA) ' (2)
where F is an empirical function, determined in experiments with a round tube,

while d is the outer diameter of the Pitot tube. All the corrections made nec-
essary by the large velocity gradient at the wall have been included in this func-
tion. Since the dynamic pressure difference P P O is quite high, ordinary in-
struments can be used for these measurements.
In order to be able to use data of [2] care must be taken that the tube, when /239
moved along the perimeter, remains within the limits of the same layer in which
the calibration was performed and that the dimensions of,the tube be similar to
those used in [2]. Hence, we have selected a Pitot tube with a ratio of 0.6between
the inner and outer diameter, when the former is 0.35 mm. The tangential stresses
235

were measured in a section removed by 4 mm from the end of the passage. The
method of electric contact was used to determine that the tube was in contactwith
the wall. Since the experiments have shown that the tangential stresses in sym-
metrical parts are approximately symmetrical, these measurements were made
only in 1/6 of the cross section. The tangential stresses at the wall were
measured at Re = 53.4 . 103 and 35.7 . 103 in the passage with the initial geometry
and with the former Reynolds number only in a passage with rods moved into cor-
ners up to contact with the shell.
The length of the flow stabilization sections was estimated on the basis of
velocity distributions measured at different distances from the inlet. For this
purpose slotted holes, through which were introduced Pitot tubes were made in
the wall of the shell at distances of 30d and 40d from the inlet. After the tube
was inserted, the slot was packed. The velocity distributions were measured
along two normals to the wall of the shell at z/h = 0 and 0.38 in a passage with
the rods moved outward (S/d = 1.3) at Re = 39. 10 3
The experimental data were processed as follows. The local velocities
were calculated by the formula
UP Po) -4-6Po] .2
Y V P '
where 6P0 is a correction produced by measuring the total pressure and the stat-
ic pressure in different cross sections.
We have estimated the effect of the air's humidity and compressibility on
the local velocity being calculated. It was shown that these factors, as well as
changes in the density of alcohol in the manometer produced by the temperature
have little effect. The limiting relative error in measuring the velocity was
about t 3%.
To determine the isotachs we have constructed intermediate graphs of u =
= f(y) for different z. In determining the value of the resistance coefficient A
we have used pressure data obtained with pickups placed at 25, 35 and 50 hydrau-
lic diameters from the inlet. The pressure drop over the length at these distances
is satisfactorily approximated by a straight line. The coefficients were calculat-
ed by the d'Arcy formula. The limiting relative error in measuring /. was about
f 10%.
The tangential stresses at the wall were calculated by a formula presented

in [2]
= 4pv = 2 1(P-- Po) 4-SPOIdh

TO! (3)
dh 4pv-
236
Using the local tangential stresses -r oil we found the relative stresses TOi by /240
graphical integration
6 [(Yrot-Al) rod +(EtoiAOshe11I

trot = Lo (4)
The magnitudes of 70i were compared with the value of T O , calculated from the
expression
dp AO
do= dz Lo (5)
The limiting error in measuring the tangential stresses was about f 7 %.

Figure 4 shows velocity distributions in the form of dimensionless isotachs
u/uO , from which it is seen that there exists a sufficiently large velocity variation
over the passage cross section. In the passage with S/d = 1.15 the velocity at the
center comprises u/u O = 1. 2, in point 1 it is 1. 3, while at the line of maximum
velocity in a corner u/uO = 0.96. When the rods are moved into the corners, the
variation becomes greater. The velocity varies particularly at points where the
flow narrows down. Thus, for S/d = 1.3 in the center of the passage u/u 0 = 1.37,
in point 1 it is 1. 34, while at the line of maximum velocity in a corner u/u 0 =0.55
The position of the maximum velocity line in a majority of points agrees

satisfactorily with the locus of points equidistant from the wall along a normal.
Point 1 deviates most: in a passage With S/d = 1. 15 S/y Omax ^:_- 12 %, while in
a passage with S/d = 1.3 S/yOmax = 10 %. The velocity maximum in this point
is very flat and the above deviations are [therefore only] approximate. The de- /241
viation of maximum velocity lines in passages of both forms decreases rapidly
along the z coordinate and in the corner this line practically coincides with the
midpoint of the space.
.Figure 5 shows the friction coefficient as a function of the Reynolds num-
bers for different rod positions. All the values of X of the passage under study
for any symmetrical positioning of the rods are smaller than those calculated
by the Blasius formula. Within limits of scattering of experimental points, the
expression X = f(Re) is parallel to the Blasius curve.
The passage with the most uniform cross section (S/d = 1.15) has the high-
est flow resistance. Figure 6 shows coefficient X as a function of S/d for several
Re. The reduction in the friction coefficient when the rods are in extreme posi-
tions may be attributed to the presence of stagnation zones.
The curve of the friction coefficient in the triangular passage under study at
S/d 1 is in qualitative agreement with data for triangular rod lattice s in axial flow.
237
1 on
Figure 4. Experimental Distribution of Relative Velocities

u/u0 ; S/d = 1. 15.
A 9`
?.8 0.316
--
3 = Re O.25
2.
?.5 i
?.f ,
177Z^7 -
2 I I
?.2 -I
^1 I - il
2 I î I
1.9 - i i 7
0 I I T
1,5
o JA -
1,4
1.3 It I I
_1
1.2 I J
78 13 20 21 22 23 2r 25 26 27 28 29 30 37 32 33 34 353637 Re70-3
Figure 5. The friction Coefficient as a Function of the

Reynolds Number at Different S/d, o, s, a., and o are
Points for S/d =1.15, 1.09, 1.01, 1.3 and 1.41, Respectively.
Figure 7 compares the velocity distributions at various distances from the

inlet for a passage with S/d = 1.3. The distributions at distances 40 and 30d /243
from the inlet differ by approximately 13 %, while at 50 and 40d they differed by
238
A10
not more than 3%. Conse-
3 quently, at distances of more
2.
than 40d from the inlet the
2 velocity distribution may be
regarded as approximately
stabilized.
The distributions of
tangential stresses at a wall
in a passage with the rods
placed in various positions
is shown in Fig. 8. The
tangential stress T0i/TOi
1 1.1 1,2 1.3 s/d varies in a passage with S1 =
= 1. 15 within the limits of
Figure 6. The Friction Coefficient as a Func- 0. 9-1.26, while in a passage
tion of S/d. 1), 2) and 3) are Curves for Re= with S1 =1.3 it is within the
33 3
= 39. 10 , 28. 7 . 10 and 18.3. 10 , Respect-
limits of 0-2.1. The exper-
ively; s. Denotes Experimental Values, o De- imica nt points are not e
notes Points Obtained From the Expression v alue ns of scattered. The
value the average tan-
0. 316/Re 1/ 4; 0 is for a Rod Lattice With gential stress at the wall,
Re = 40.103, found by Preston's method
(T0i) and those calculated on
the basis of the pressure drop
(T0) are in satisfactory agree-
ment with one another. The maximum-difference between T 0 and T o was 7-15%
for the different rod arrangements in the passage. This circumstance shows
that Preston's method is suitable for determining the tangential stresses at the
wall in such intricately-shaped passages.
The velocity distribution in the passage under study in the coordinates u+,
y+ is in agreement with the universal velocity distribution. Consequently, the /244
universal velocity distribution is practically applicable also in intricately-
shaped passages, if the velocity distribution is measured along a normal to the
passage walls.
Comparison of Theoretical and Experimental Data
The friction coefficients for Re = 65 . 103 obtained theoretically by the Deis-

sler-Taylor method and extrapolations of experimental data were found to be in
satisfactory agreement. Comparison of velocity fields shows that the theoreti-
cally calculated velocities are higher than actual in the wide parts of the cross
section and lower than actual in the narrow parts (Fig. 9). The maximum dif-
ferences in the local velocities comprises about 25%. The distributions were
239
1.25
=0
1

Q75
1 h =0,38
0.5
V u z 0.4 0,6 0.8 y/t,
Figure 7. Velocity Distributions at Different Distances

From the Inlet, Re = 39 . 103 . ., x and o are for x/d h =
= 30, 40 and 50, Respectively.
Zoi x
ZVi x x
x
Z
J
1.'
/ I
30 60 90 120 150 780

0 25 50 75 100 1250 25 50 75 l,
_ Rod Shell
Figure 8. Distribution of Tangential Stresses

T0i/T0i
Over the Perimeter of the Passage. At S/d = 1. 15,
Re = 35.7 . 103 (o.); Re = 53.4 . 103 (o);
at S/d = 1. 41, Re = 53.4. 103 (x).
240
0.2 0.4 0.6 0.8 1 0
Figure 9. Theoretical (Using the Deissler Method)
and Experimental Velocity Distributions. o - Ex-
perimental Data for Re = 39 . 103 ; the Solid Line was
Calculated Theoretically for Re = 65 . 102 ; 1), 2), 3),
4) and 5) are Curves for z/h = 0, 0. 12, 0. 31, 0.42
and 0. 72, 'Respectively.
compared for several values of Re, differing from one another (65 . 103 and 39.103),
however, it is fully justified due to the fact that the dimensionless velocity distri-
bution depends little on the Reynolds number.
The difference between the theoretical and experimental data is a conse-
quence of the assumptions used in the calculations. As is shown by the experi-
ments, as well by data of [3], the universal velocity distribution in a round pipe
describes satisfactorily the velocity distribution even in an intricately-shaped
passage. Hence, this assumption is an experimentally proven fact and is be-
yond doubt. The main source for the above difference is the assumption that
there is no transfer of momentum between elementary volumes. The only time
when there actually will be no such transfer is that when the momentum trans-
fer is produced by the gradient, when the dimensions of the turbulent eddies
(length of the displacement path) is much smaller than the dimensions of the
cell and the entire transfer depends on the velocity gradient du/dz.
However, in addition to momentum transfer produced by the [velocity] /245
gradient, there also exists convective transfer of momentum, which is due to
the motion of large eddies. Such transfer, naturally, will result in mutual
transfer of momentum between the volumes and smoothing out the distribution
Of TO over the passage's perimeter. Convective transfer of momentum be-
tween elementary volumes cannotbe eliminated by choosing these volumes along
velocity gradient lines, as this is -suggested by Deissler and Taylor. Hence,
their basic equation of the force balance in an elementary volume
,roi Al= dX DAL

(6)
241
is invalid in the case of intricately-shaped passages. It follows from analysis
of experimental data, that the greatest imbalance of Eq. (6) is observed in those
parts of the passage where the cross section changes appreciably. Hence, the
more the distance to the maximum-velocity line varies, the more appreciable
is the computed velocity increase [ apparent] shown by theoretical calculations.
(See Figs. 2 and 10). The [apparent] reduction in velocity in those parts of the
cross section where the elementary areas AA do not change, is produced by
a reduction in T0 in these points.
Rod Shell
-10
z,
ze
I ^^^
1,4 li
1,0
0.6
20 49 60 80 100 120 0 20 40 60 Z, mm
Figure 10. Theoretical (Using the Deissler Method) and

Experimental Distributions of Relative Tangential Stresses
at a Wall. - Experimental Points for Re = 35.7 x 10;
the Solid Line was Calculated at Re = 65 . 103.
Figure 10 compares the distribution of tangential stresses T Oi obtained theo-

retically using the Deissler method, with the experimental distribution. The
shape of the theoretical curve of r0i/T0i is steeper. The maximum difference
in values of TO is about 55 %.
Thus, calculations of the velocity field using the Deissler-Taylor method

can yield satisfactory results for passages with a moderately variable cross
section, when the contribution of the convective momentum transfer is not ap-
preciable.
Simplified Method for Calculating Velocity Fields in

Intricately-Shaped Passages
The applicability of the universal velocity distribution for a pipe for calcu- /246
lations involving intricately-shaped passages means that, in order to calculate
the velocity fields in such passages it is only necessary to know the tangential
stresses at the wall.
1 This method was subsequently refined. See Atomnaya energiya, 21, 2 (1966).
242
It follows from the assumption that the fluid dynamics in intricately-shaped
passages are appreciably affected by convective momentum transfer, that [4)
TQ
f2. b0
=f^ (1-e (7)
TO
where fl and f2 are some constants, while y 0 is the average value of y0 over the
entire cross section of the passage. The above expression shows that for low
y01y0 the variations of y0 and TO are proportional. In the limit, as y0 0,
To 0. As y 0/y0 becomes larger the relationship between the variations in y o /247
and TO becomes less pronounced, due to the effect of convective momentum
transfer by large-scale eddies.
Constant fl is determined by the norming condition
'o 01
dl=1. (8)
Lo 0
The value of f2 is an empirical quantity, which is related to the shape of the pas-
sage cross section. To determine, it is necessary to analyze the experimentally-
determined relationship TO /T0 = F(y0 /y0) for different passage shapes. Figure
11 shows that experimental points with a scatter of approximately 20% are grouped
about two averaging curves. Curve 1 characterizes the distribution of T O /TO in a
very irregularly shaped passage, while curve 2 corresponds to a more uniform
cross section.
The distribution of T o/T0 depends perceptibly on the overall irregularity
in the passage's shape. As a parameter which characterizes this irregularity
we can take
a= Yo max = Yo mi n
Yo
In a passage with uniform cross section a = 0, in the passages under study a

= 2.98 (at S/d = 1.3) and a = 1, 39 (for S/d = 1,15), Upon introducing param-
eter a, f2 of Eq. (7) takes on the form
f2 = az3 (9)
for a passage with a uniform cross section a = 0 and T 0/T0 =1,
243
Vol 0 00 o 00 0
0 0 1
0
1.' 0
o Q 2
e
1 A
v
0.11
0 0, 5 1.0 1.5 2.0 45 YIA
Figure 11. The Relative Tangential Stress T O /TOi as a Func-

1y
tion of yO 0 , o - Rod; o - Shell (S/d = 1.41); o - Rod, cp =
= 0-65 (S/d = 1. 15); - Rod, co = 65-180; - Shell.
Comparing values of TO /T 0 obtained experimentally in a passage formed

by closely-packed rods, [3], with those obtained using Eq. (7) we see that the
theoretically and experimentally obtained values of this ratio are in satisfac-
tory agreement. Parameter a in such a passage is 3. 02. The maximum dif-
ference, for y0 /y0 > 0. 3, comprises about 15%. Taking into account the fact
that in the passages under study the magnitude of a varied within sufficiently
wide limits (from 0 to 3), we can assume that Eq. (7) is valid for calculating
TO/T O in other intricately-shaped passages formed by various cylindrical and
flat surfaces.
Knowing the manner in which the tangential stress varies over the perim-
eter of a passage, it is possible to recommend a simplified method for calculat-
ing the velocity field in hydrodynamically stabilized turbulent flow of a fluid.
1. The maximum velocity line is drawn as the locus of points equidis-
tant from'the passage's perimeter, while the latter is broken up into uniform
intervals, through the ends of which normals y Oi are drawn to the maximum
velocity line.
2. The magnitude of r 0 is determined on the basis of the specified flow /249
rate of the fluid through the passage and by selecting a friction coefficient (for
example, using the Blasius formulas).
244
3. Parameters a and Y0 are determined.
4. The tangential stress distribution is calculated using Eqs, (7)-(9).

5, Successive values of y are specified and the velocity distribution at
each normal y0 is calculated, using the universal velocity distribution,
6. The size of elementary areas AAV formed by the wall, normals y 0

and the maximum velocity line is determined.
7. From the isotachs, the average velocity in each elementary area AAi
and the average velocity over the passage cross section are calculated by
graphical integration. u0p
8. The frictional resistance coefficient is determined as
XP= 8-ro
P-uuP .
If the flow rate calculated on the basis of the average velocity differs ap-
preciably from the specified flow rate, then the initial value of T O is adjusted
(taking into account the X which was obtained) and the calculations are repeated.
Using this method we have calculated the velocity at individual normals
in a passage formed by closely-packed rods and in a triangular passage with
S/d = 1.3. Figure 12 compares the theoretically and experimentally obtained
velocity distributions. The greatest difference between these distributions is
about 8% in passages formed by closely-packed rods and about 10% in the tri-
angular passage.
NOTATION
u average local velocity in direction of the flow, m/sec

u0 average velocity in the passage
A passage cross sectional area
AA area of ith elementary area
L0 passage perimeter
d hydraulic diameter 4- A0/L0
245
UP,
2 ^ \
1.0 /
0.3
d
0 0.2 04 0.6 0.i y/yo
a
I/Al
Ile
U
2
O.b
' 3
O.d
2
1
O,y 3
O,c
0 0,2 04 0.6 0.8 y/yo

b
Figure 12. Velocity Distribution in a Passage Formed by

Densely-Packed Rods. Re = 14 10 3 , Ref. [3] (a); and in
a Passage With S/d = 1.3 With Re = 39 . 103 (b), o - Ex-
perimental Points; the Solid Curves Were Obtained Theo-
retically; 1), 2) and 3), are for co = 28.5, 22.5 and 43.5,
Respectively.
246
Al parameter segment
r characteristic dimension
T Oi tangential stress at the wall in the i-th point of the perimeter
T O average tangential stress at the walls of the passage
dp/dx drop in pressure over the length of the passage [pressure gradient]
6 displacement of the position of a point on the maximum velocity
line with respect to the position of a point equidistant from the walls
y
i
distance along a normal to the wall in an ith elementary area
Y O distance along a normal from a wall to the maximum velocity line

P total pressure
P o static pressure.

U+= velocity /250
y^ T
4 P
U++ U
velocity parameter
rdp/dx
r P
TO
^/
V
P distance
r++ rdp/dx
P r parameter
REFERENCES
1. Deissler, R. and Taylor, M.: NACA, TN-4384, 1958.

2. Preston, D. R.: Zh.[?] Mekhanika, No. 6 (34) (1955).
3. Levchenko, Yu. D., et al.: Raspredeleniye skorostey v yacheyke plotno upa-
kovannogo puchka sterzhney [Velocity Distribution in a Cell Formed by a
Densely Packed Cluster of Rods]. Article immediately preceding this one
in the present collection.
4. Ibragimov, M. Kh. , et al.: Atomnaya energiya (in print).
247
PART THREE
BOILING AND CONDENSATION OF METALS
ELECTRONICALLY-HEATED SETUP FOR THE STUDY OF HEAT TRANSFER
FROM BOILING METALS BY FREE CONVECTION
A. A. Ivashkevich, A. P. Kudryavtsev, D. M. Ovechkin, D. N. Sorokin,
In 1957 Lyon and others [1] have published data on heat transfer from boil- /251
ing sodium and from a sodium-potassium alloy in a large volume (heat fluxes of
approximately up to 0.350 . 106 kcal/m2-hour). For sodium with heat fluxes of
about 0.3 . 106 kcal/m2 hour the heat transfer coefficient was about 7.10 kcal/m 2
-hourdeg.[Hwv],noxperimtaldhsbnpuieocrtalh
fluxes qcr. Calculations made using standard relationships which describe ex-
perimental data for qcr for ordinary fluids, yield values of the critical thermal
flux for sodium, at atmospheric pressure, for example, of from 1.5 . 106 to 10.106
kcal/m2 hour. Hence in working out the methods for performing the experiment
being here described particular attention was paid to selection of a method for
heating the heat-conducting wall and for measuring the temperature difference
between the wall and the fluid [metal].
We have examined various methods for heating the heat-conducting wall for
large-volume conditions, and have arrived at the following conclusions.
1. Heating by passing current directly through the test section-is impossible
due to the electrical conductivity of the metallic heat-transfer agents, while radi-
ant heating can produce thermal fluxes only up to (1.5-2) 10 6 kcal/m2-hour.
2. Heat fluxes in excess of 107 kcal/m2 -hour can be obtained by using an
electric arc or gases passed through an arc (plasmatron), as well as high-fre-
quency heating and heating the boiling alkaline metals by a heat-transfer agent
at a temperature higher than that of the metals. However, this creates great
technical difficulties such as, for example, stabilization of the heat supply over
an extended period of time when heating by the electric arc; corrosion of the heat
conducting wall from the hot-gas side when using the plasmatron; determination
of the thermal flux when using high-frequency heating, making of a special reli- /252
able high-temperature circuit when heating by the heat-exchanger method.
3. Electronic heating makes it possible to supply appreciable amounts of
power to a relatively moderately-sized surface and thus to obtain high thermal
fluxes and to measure them with sufficient accuracy. For example, the thermal
fluxes in high-power generator tubes are as high as 86 . 106 kcal/m2 -hour, [2j.
248
In conjunction with this we have selected for our study of critical thermal
fluxes and heat transfer from boiling metals the method of electronic heating the
heat-conducting wall, which was used previously in experiments using water [3].
The substance of the method consists in the fact that the electrons emitted by
the cathode are accelerated by the anode voltage and impinge on the anode. This
means that the electrons are stopped in a very thin layer (the thickness of this
layer is several microns, even at voltages of about 10 kv), in which the kinetic
energy of the electrons is converted into thermal energy.
When determining the power generated at the anode one does not take into
account the energy losses due to x-ray radiation and secondary electrons, and it
is also assumed that the electrons leave the cathode with zero energy. It has been
estimated that the error in determining the anode power produced by the above as-
sumptions is less than 1% for voltages of up to 10 kv. Under electronic heating the
surface of the heat-conducting wall facing the cathode is not subjected to oxidation
and corrosion even at high temperatures as compared, for example, with arc heat-
ing, since it is situated in a vacuum (10 -5 -10 -6 mm of Hg). The high-voltage and
evacuated system which is needed under this method can be made by selecting stand-
ard equipment.
The main part of an electron tube is the cathode. In radio electronics use is
made of purely metallic, film and semiconductor cathodes, which may be divided
into direct- or indirect-heated devices [4]. In the electron tube used in the study
of heat transfer the cathode operates under conditions more difficult than cathode
of radio-electronics tubes. In the case at hand the cathode is required to retain
its emission properties under a relatively rough vacuum in the tube ( 10-4 mm
of Hg) and operate stably at temperatures in excess of 1300-150'0C, since the
temperature of the surface of the heat conducting wall facing the cathode may be
as high as that. In addition, the cathode should retain its emission properties un-
der ionic bombardment and on relatively frequent disassembly of the tube. A tung-
sten cathode satisfies these requirements [4].
The electron tube shown in Fig. 1 consists of a diode, the anode of which
(the tube body) is held under ground potential, while adjustable voltage is supplied
to the cathode. A removable shell, to the top butt of which is welded a flat work- /253
ing section (header), is fastened to the upper part of the tube body. The inner sur-
face of the header is subjected to electron bombardment and radiant heating by the
cathode, which is made from tungsten wire 1 mm in diameter in the form of a flat
spiral with an average outer diameter of 38 mm and a distance of 2 mm between
loops. The cathode is fastened to molybdenum holders by a molybdenum wire 0.2
mm in diameter over a length of 12-15 mm, the latter is fastened in copper
conductors 10, which in their turn are fastened in conductors 25, cemented by the
BF-2 glue to an organic-glass insulator. This design of conductors make it pos-
sible to center the cathode and to adjust its height position. The cathode is pro-
vided by six supports from tantalum foil 0.15 mm thick, which prevents it from
sagging. To prevent scattering of the electrons which travel from the cathode to
the header, the former is surrounded by a screen made from 0.15 mm thick tan-
talum sheet and with an inner diameter of 39 mm. The distance between the head- /254
er and the cathode is approximately 15 mm, while that between the top edge of the
screen and the header is about 5-7 mm. To protect the insulator from overheating
249

L"
1 17-
14-.
16
_l
I !
`73--
12
00
9
I e ^^ 19 24 A
21 r' 20

320 { 23 22 31 30'
j_
R,-32
t- cJ2S Sec, AA
2 , ^
-4204 --
û
L
27 9
28
} ^--i
r j29
Figure 1. Overall View of the Electron Tube.

1) Measuring Tube; 2) Connector Nut; 3) Pressing Sleeve;
4), 6), 9), 24) and 26) Inserts; 5) Ring; 7) Connecting
Sleeve; 8) Housing; 10) Conductor; 11) Screws; 12) Shell;
13) Holder; 14) Insulator; 15) Supporting Bracket of the
Screen; 16) Wire, 0.2 mm in Diameter; 17) Screen;
18) Cathode; 19) Cooling Sleeve; 20) Contact2l) Ring;
22) Insulator; 23) Flange; 25) Conductor; 27) Collecting
Tank; 28) Cooling Cover; 29) Connecting Sleeve; 30) High-
Voltage Cable; 31) Tubular Segment; 32) Sleeve.
the conductor is cooled by transformer oil. A removable container, which serves

to catch the metal in case the header breaks down during operation, is fastened to
the lower part of the tube body. The individual parts of the tube are packed by vac-
uum rubber inserts.
The vacuum in the electron tube is measured by the MM-5 and LT-4M pick-
ups. Various types of headers were used in the experiments (Fig. 2). Type I
headers were used for the study of critical thermal fluxes. The header temper-
ature was measured by chromel-alumel thermocouples using wires 0.2 mm in
diameter, insulated by aluminum oxide. The thermocouples were placed in cap- /255
illaries from 1Kh18N9T steel 0.5 x 0.1 or 0.8 x 0. 15 mm in diameter, which
were forced into holes drilled to a depth of about 15 mm.
Type I headers are not recommended for the study of heat transfer
from boiling fluids, since, for example, headers from stainless steel have a
large error due to the fact that it is impossible to determine precisely the position
250

of the thermocouple bead (A5 =f 0.4

mm). For example, for boiling wa-
^_ GJ3
ter with q = 0.63 . 106 kcal/m2-hour
Thermocouples and a = 31 . 103 kcal/m2-hour-degree
O y Electroplated the maximum error in determining
coating the heat transfer coefficient in a sin-
N y î S
i (Ni or Ni + Cr) gle experiment is about 160%. In or-
der to obtain a higher accuracy, it is
^^^` T ,1 necessary either to appreciably re-
duce the error produced by the unde-
Thermocouples terminacy A5 in determining the po-
sition of the thermocouple bead (which
is impossible at present), or to make
the headers from materials with a
0/:Is higher thermal conductivity. For this
purpose methods were developed and
multilayer headers were produced, in
which the thermocouples which mea-
sure the wall temperature were plot-
gag ted with copper [5] (see Fig. 2, II,
Thermocouples III and IV). Brand M-1 molten cop-
per is poured into a bowl from 1Kh-
X63 18N9T steel under a vacuum of 10-5
I mm of Hg. Before the copper is
poured the thermocouples are placed
in the bowl. It thus becomes pos-
IV sible to embed several thermocou-
^\ \i\
2 38
Thermocouples
ples into one header. Metallograph-
ic examinations have shown that the
Figure 2. Types of Headers Which diffusive coupling between the copper
are Used. and the 1Kh18N9T steel is satisfac-
tory. Hence, we can regard the ther-
mal contact between the steel and the
copper as ideal.
The surface of type II headers was electroplated by a layer of nickel (or nic-
kel followed by chromium) 20-25 micromicrons thick to protect the alkaline metal.
The type III header after potting was stoppered from both sides; here steel with a
thickness of 0.2 mm was left over at the side facing the heat-transfer agent. Be-
fore potting of type IV headers they were topped by a 0.2 mm thick foil from 1Kh-
18N9T steel.
In copper-potted headers the error in determining the temperature differences
is appreciably smaller as compared with steel headers (which is due to the high
thermal conductivity of the copper). In addition, in this case there is practically
no distortion of the temperature field, which does not take place when thermocouples
are embedded in slot. At high thermal fluxes (> 1 . 10 6 kcal/m2 -hour) such head-
ers can operate at approximately the same sodium temperatures as the steel
headers.
251
The electrical system of the tube consists of a high-voltage rectifying de-
vice rated at 20 kwt and of measuring instruments. The above rectifying device
includes: type ATSK-25 adjustable three-phase autotransformer, the TM20/10
step-up transformer, six type TR15/6 thyratrons connected in the Laryonov cir-
cuit, and a filter, consisting of two RL-32.5-0.25 capacitors and of a choke.
The tube heating system was supplied from a specially-made oil-type heat-
ing transformer. The heating power was adjusted by an autotransformer. Volt- /256
age stabilizers have been included in the tube and thyratron heating circuits. The
anode voltage is measured by means of a voltage divider and class 0.5 M 82 mul-
tiscale millivotmeter. The anode voltage was monotored by the S-96 electro-
static kilovoltmeter, while the anode current was measured by class 0.5 m 104
multiscale milliameter. All the additional resistances were measured to within
0.1%. The measuring instruments were protected by R-350 shunts. The
heating current and voltage were measured by class 1.5 control-panel instruments
in the primary circuit of the heating transformer, since the secondary winding is
held at the anode voltage. The rectifier was equipped by operational and pro-
tective shunts, which could be used for shutting off the anode voltage and the tube
heating.
The curve showing the specific emission current j for tungsten as a function /257
of the temperature, constructed on the basis of data of [4] is shown in Fig. 3a. It
can be seen that up to a temperature of 2300K j is practically zero, and that it in-
creases rapidly with a further increase in temperature. In electron tubes used in
radio engineering the operating temperatures of tungsten cathodes are within the
limits of 2450-2650K [4].
Electron tubes are capable of operation in the saturation and in the space
charge modes. When operating under the saturation mode the anode current is
limited by the emissive power of the cathode, which increases with an increase
in the temperature, all other conditions remaining equal. If the emission power
of the cathode exceeds the anode current, then it is limited by the magnitude of
the space charge between the cathode and the anode. In this operating mode the
anode current is a function of Ua/2 [4]
U3/2
1a = 2.33 . 10-sxPa ', (1)
C
where he is the anode to cathode distance, F a is the anode surface area, x is a

coefficient which takes into account nonuniformity in the electron bombardment.
For a plane electrode system x = 1. For the geometry of our tube (a c and hc, see
Fig. 1a) the value of x was close to unity (about 0.98) [4], from which it follows
that heat generation by electron bombardment at the surface of the header facing
the cathode, under the space charge mode, should be close to uniform. Under
the saturation mode the emission current distribution will depend on the cathode
temperature.
252

to
J
NC^
F-i
S
PAY
2
n
2200 2,'00 2600 WO T, X U
Q b
0
N
Cd
V U
O
^ n
U 0 2 G 3U, kv
C d
Figure 3. Electron Tube Characteristics.

a) The Specific Emission as a Function of the Temper-
ature for Tungsten; b) Qualitative Dependence of I a, the
Anode Current on the Anode Voltage U a (Tc < T < T );
1 2 3
c) Qualitative Dependence of the Thermal Flux qa on the
Anode Voltage U (Tc < T < T ); d) Boundary of the
1 2 3
Tube's Operating Modes (Saturation and Space Charge)
as a Function of hc .
The specific thermal flux is calculated from the formula
U aa
r
qa = 0.86 (2)
Th
It follows from the above expression and from Eq. (1) that in the space charge
mode qa ^ Ua/2.
253
Figures 3b and c show the qualitative dependence of I and q on the anode
voltage for operation under the space charge and saturation modes (branches for
T c , T cand T c ). Figure 3d shows h e and q a as a function of Uafor the tube
1 2 3
we have used for I = 2 amps (nominal rectifier current) and r = 1. Figure 3d
can be used to estimate the boundary between the space charge and saturation
operating modes as a function of the cathode to anode distance. For example,
for h e = 20 mm it is possible to operate in the space charge mode at Ua of up to
4.5 kV, producing a thermal flux of approximately up to 7 . 10 kcal/m 2 -hour, The
cathode heating current and voltage have been calculated by the method presented
in [4]. To estimate the variation in the thermal flux at the heat conducting surface
produced by possible nonuniform heat generation at the inner surface of the header,
we have considered the limiting case of heat generation along a coil (x P^:; 1/3), the /258
dimensions of which are equal to those of the cathode's coil. The following as sump-
tions were used in the calculations,
1. The. header is considered to be an infinite plate, and the spiral distribu-
tion is replaced by a stepped distribution. The step lead (la c ) is equal to the cath-
ode coil lead, and the width of the step (b) is equal to the diameter of the cathode
wire (see Figs, la and 4).
2. The thermal conductivity of the plate and the coefficient of heat transfer
from the plate to the fluid are constant. The process of heat transfer in the plate
is described by the expression
d2t 821
a..
2+ ay 2 = 0 (3)
with the boundary conditions
ut (s)
=0
8l a(t,utfl)
Solving Eq. (3) with the above boundary conditions we get an expression for the /259
thermal flux
io n7l _not
^û _
^'ub ^ at n - a an
(A lt o c B"o J ) Cos X. (4)
'IC
n==1
c
Figure 4 shows the variability in the thermal flux at the heat conducting
surface of the header as a function of the header thickness for different values
254
of the thermal conductivity, cal-
culated from Eq. (4), It can be
seen from the figure that even in
this hypothetical case (for head-
ers used in our experiments) the
variability in the thermal flux
from the heat conducting surface
due to the stepped shape of the
heat generation curve will be less
than 0. M These calculations
show that the nonuniform gener-
ation of heat at the inner surface
of the header will be smoothed out
at the heat-conducting surface. To
eliminate variability in heat gen-
eration also at the inner surface,
for example, due to lack of uni-
G.
formity in the cathode tempera-
% 2 3 4 Amm ture when the tube operates un-
der the saturation mode, the work
Figure 4, Variability Aq/q in the Heat should be performed in the space-
Transfer as a Function of Thickness 6 charge mode.
and Thermal Conductivity X of the Ma-
aerial of the Heat Conducting Wall, We have estimated the ther-
mal flux distribution at the heat-
conducting surface of the header
as affected by radial heat losses. This process can be described by the expression
0 2 l 1 at a21
0 (5)
Tr r ar + aye
with the boundary conditions

at _ _ 9r
^J u-0
at a (t tfl)
For headers potted by means of copper the above boundary conditions must be sup- /260
plemented by the requirement of "matching" at r = R (Fig. 5)
Equation (5) with the specified boundary and "matching" conditions, and on
the assumption that a, to and X are constant, was solved numerically for different
255
r
.Q
Figure 5. Distribution of the Thermal Flux at the Heat-

Conducting Surface of the Header.
A - 1 mm; a = 40. 10 3 kcal/m2 -hour-degree; X l = 300
kcal/m-hour-degree; X 2 = 15 kcal/m-hour-degree.
combinations of a, X and A. Figure 5 shows the distribution of the thermal flux

at the heat-conducting surface of a header the working part of which is made from
copper. It can be seen that the thermocouples which measure the wall tempera-
ture are embedded in a region with a uniform flux distribution. Hence, the calcu-
lations show that the thermal flux at the heat-conducting surface of a header at
the region where the thermocouples are embedded can be regarded as uniform.
To check the operating fitness of the tube we have measured the critical
thermal fluxes and the heat transfer from a large volume of boiling water at at-
mospheric pressure. Two tanks, 70 and 200 mm in diameter were used. Con-
densers, heated by ordinary water were placed in their upper part. The shell
(see Fig. 1, item 7) of the electron tube was inserted from the bottom of the tank.
The water, when using these tanks, was maintained at the saturation temperature
by auxiliary heaters.
In experiments for measuring qcr the lamp heating was first switched on
and then the anode power was increased by increasing the anode voltage. At ther-
mal loads close to the critical the anode power increased in moderate steps (by
2-3 % of q) with a holding time at q = const of the order of 2 wins, The production
of critical flux was sharply recorded on the EPP-09 potentiometer due to a sharp /261
increase in the temperature difference between the wall and the fluid. Temperature
control was set at 100-200C above the wall temperature preceding the critical
flux. After each critical flux has been obtained, the heat-conducting surface of the
header was polished by a micronic emery cloth and washed with ethylene alcohol.
The salt-content of the water used in the experiments was not monitored.
Figure 6 shows magnitude of critical thermal fluxes from boiling distilled
and tapwater at surfaces from 1Eh18N9T steel and from copper, The experimental
256
.[ Y p7
N ^' 0 4 o I
CCd
U
x
1 --T 1 e' 4
7e7 e8 i i 8
1_17
^ n __^ I ^ ^ 5s i a
1
0 5 70 75 20 25 30
Number of test
Figure 6. Magnitudes of Critical Thermal Fluxes

From a Large Volume of Boiling Water at
Atmospheric Pressure.
o - Ordinary Drinking Water, Surface of 1Kh18N9T

Steel; m - Distilled Water, Same Surface; - Dis-
tilled Water, Copper Surface.
data obtained using different headers agree with one another and with published
data, [6] with respect to magnitude of the critical thermal flux, as well as with re-
spect to the scatter of points. The identical numbers next to points show that the
experiments were performed with the same header, without dismantling the tube. /262
The volt-ampere characteristics constructed for experiments on q cr shows that
the electron tube operated under the space-charge mode. Figure 7 shows an ex-
ample of the volt-ampere characteristic for experiments on q cr shown in Fig. 6
and designated by numeral 8. It can be seen from Fig. 7 that the volt-ampere
characteristic of the tube did not change during the eXpe imei tu,
The temperature differences over the height of a header were measured in
five headers, using thermocouples, in order to determine the thermal flux and to
compare it with the flux calculated from Eq. (2) (Fig. 8). We have compared rel-
ative, rather than absolute units to eliminate the effect of errors due to imprecise
determination of the distance between the thermocouple beads as well as due to
errors in the thermal conductivity of the header material. The results presented
in Fig. 8 show that the thermal fluxes determined by different methods are in sat-
isfactory agreement.
The feasibility of obtaining high thermal fluxes using the tube was determin-
ed by cooling the test section by running water moving in a slot formed by the head- /263
er and a special jacket. These experiments show that the installation is capable
of producing thermal fluxes of about 20.10 6 kcal/m2 -hour. During a short time
period (about 5 mins) we have obtained a maximum thermal flux (about 36 . 10 6
kcal/m2 -hour), but then the anode transformer was overloaded by approximately
two-fold.
257
.7, 0 Headers type H, III and IV (see
Fig. 2) were used to perform experi-
r;2
ments for measuring the heat trans-
fer coefficient attendant to boiling of
.9 a large volume of distilled water, in
the process of which we have mea-
sured the temperature difference be-
0.6
tween the wall and the liquid using an
R2/1 semiautomatic potentiometer
and we have recorded the data using
the EPP-09 automatic potentiometer.
0 The experimental results are pre-
sented in Fig. 9, which, for compar-
A4
ison, also displays lines obtained
from criterial formulas due to Kru-
zhilin [7] and Kutateladze [8]. These
formulas describe satisfactorily the
I data on heat transfer on nucleate
boiling of water at atmospheric pres-
sure. The figure shows that the data
we have obtained agree with these
formulas. The heat transfer coef-
ficients were calculated from the
I expression
0.2
600 800 7000 1200 1400 U,v
a -At'
where q is the resultant specific /264
Figure 7. Volt-Ampere Characteris-
tics of the Tube Used in Experiments thermal flux, kcal/m2 -hour, while
With Water (Points Correspond to At is the temperature difference be-
Points 8 of Fig. 6). tween the heat conducting surface of
the header and the fluid in the con-
tainer, C.
The resultant specific thermal flux was found as follows
`7 = ` " a nheat'lrad;
q = 0.86 UaIa/Fh is the anode component of the specific thermal flux, where Ua
is the anode voltage, V, I is the anode current, amps, and F is the heat-con-
ducting area of the header, m2; qheat - 0.86 U heat I heat - K K 1 F./Fh is the specific
thermal flux produced by cathode radiation, where Uheat is the heating voltage,
V, Iheat is the heating current, amps; K is a factor which takes into account that
part of the heating power which is radiated to the 'working surface of the header,
Kl is a factor which takes into account losses in the heating transformer and
feeding cable and e is the emissivity of the header; grad - 4. 9- FF d (T1/100)4
h
258
1
Figure 8. Comparison of the Ratio q2/ql of Ther-
mal Fluxes Calculated From Eq, (2) With the Ratio
Ot2/Atl of the Temperature Differences, Measured
Over the Height of the Header and Corresponding to
the Above Thermal Fluxes (the Different Symbols
Denote Points Obtained With Different Headers).
is the specific thermal flux produced by radiation from the inner surface of the
header, where Frad is the radiating surface of the header, m2 , T is the abso-
lute temperature of the radiating surface, K. In experiments with water we have
not taken into account the thermal flux due to radiation from the inner surface of
the header, since the surface temperature (T l) was low.
The temperature difference between the wall and the fluid was calculated as
follows
At = At m At op At,,
where At is the measured temperature difference between the header and the
fluid, degrees; Atcop = q 6cop is the temperature difference in the copper layer,
cop
degrees; 8 cop is the thickness of the copper layer, m; X cop is the thermal
259
- - --------- ---------
N
eo
' z T1 O
9
e

^t
1yâ ^
I
i l
I I
' ^^ I ^ I I
U 1^ 4
_ 0,2
25
+S+ - __^-1^1J!J
8 70 'f kcal/(m2. hr)
Figure 9. Data on Heat Transfer From a Large Volume

of Boiling Distilled Water, P = 1 Atmosphere Absolute
(the Different Symbols Denote Points Obtained in Differ-
ent Series of Experiments).
1) Curve Obtained from the Krushilin Formula; 2) Curve
Obtained from the Kutateladze Formula.
S
conductivity of copper, kcal/m-hour-degree); At . = q An is the temperature differ-
n
ence in the nickel layer for type II header (see Fig. 2); for type III and IV headers
the temperature drop is calculated for a layer of 1Kh18N9T steel; 6 is the thick-
ness of the nickel or steel layer, and X is the thermal conductivity of the nick-
el or steel, kcal/m-hour-degree,
The limiting relative error of a single measurement of the thermal flux q =
6
=10 kcal/m2 -hour is 10 %. The limiting relative error in determining the heat
transfer coefficient A-a/a will depend on the magnitude of the thermal flux. For
example, for boiling water with q = 0.2 . 106 kcal/m2 -hour (q =qn) Aq/a ^ 40%,
while for q = 0.6. 106 kcal/m2 -hour Aa/ a - 25%.
When the thermocouples are embedded into a stainless steel wall (type I
header, Fig. 2) of the same configuration the limiting relative error will be Lrx/aS:Z^
N200 %, for q = 0.2 . 106 kcal/m2 -hour and Aala ^ 240% for q = 0. 6 . 106 kcal/m2 - /265
-hour.
260
This example shows that the use of multilayer header, in which the thermo-
couples are potted with copper (a metal with high thermal conductivity) makes it
possible to substantially reduce the error in determining the heat transfer coef-
ficient.
To study heat transfer from metals boiling in a large container we have de-
vised a setup the schematic of which is shown in Fig. 10.
1 ^^
Figure 10. Schematic Diagram of the Setup Used to Study

the Transfer of Heat From Alkaline Metals Boiling in a
Large Vessel.
1) Gear-Type Oil Pump; 2) Diffusion Pump; 3) Electronic
Tube; 4) Working Tank; 5) Air Valve; 6) Movable Thermo-
couple; 7) Level-Measuring Tank; 8) Pressure Pickup;
9) Distillating Sampler; 10) Cold Oxide Trap; 11) Elec-
tromagnetic Pump; 12) Filling Tank.
-------Metal Lines; -----Gas-Vacuum Lines;
---Oil Lines; -----Water Lines.
Figure 11 depicts the design of the operating tank and the movable thermo-
couple. The electron tube is inserted into the tank from its bottom. The inner
surface of the working section (header) is heated, while the heat transfer agent
boils at the outer surface. Two holes in the lower part of the working tank serve
for filling and draining, as well as for circulating the heat transfer agent through
the working tank when removing the oxides. A ribbed air-cooled condenser, used
for condensing the heat-transfer agent's vapor, is welded to the upper part of the
261
Figure 11. Working Tank With Movable
Thermocouple,
1) Shell withHeader; 2) CoolingSleeve;
3) Working Tank; 4) Rod with Thermo-
couples; 5) Condenser; 6'and32) Guide
Bushings; 7 and 9) Flanges; 8) Copper
Gasket; 10 and 15) Bushings; 11) Guid-
ingTube; 12) Bellows from 1Kh18N9T
Steel; 13) Guide Screw; 14) Connecting
Stub; 16) Connecting Tube; 17) Siloxane-
teflonPacking; 18 and 33) Connector
Nut; 19) Graduated Circle; 20) Selsyn
Jacke; 21) Selsyn; 22, 27 and 29) Bear-
ings; 23) Gear; 26) Pinion; 24) RD-09
Electric Motor; 25) Coupling Sleeve;
28) Housing; 30) Lead Screw; 31) Nut;
34) ClampingSleeve; 35) Siloxane-
teflon Packing; 36)-Connecting Stub;
37) Probe with Stationary Thermocou-
ples; 38, 39 and 40) Electric Heaters
working tank. The necessary temperature in this tank is maintained by three elec-
tric heaters, rated at 14, 3 and 2kW, respectively, supplied from' autotransformers
The movable thermocouple assembly is fastened ,to the condenser flange, It /267
is packed by gasket from annealed copper or nickel. In the rod's grooves are
inserted and metallized four capillaries from 1Kh18N9T steel 0.8 X 0.15 mm in
262
diameter, in which are placed chromel-alumel thermocouples with alundum insu-
lation (wire diameter 0.2 mm). At the end of the rod the thermocouples are bent
away at a right angle from the rod's axis and from one another. The end of the
rod can come into direct contact with the surface, in which the capillaries with
the thermocouples also come into direct contact with the heat-conducting surface.
The end of one of the movable thermocouples is inserted approximately at the cen-
ter of the heat conducting surface. The movable thermocouples, as well as the
thermocouples embedded in the header are connected differentially with the ther-
mocouples held in tube 37.
Tube 16 and rod 4 connected to it are moved vertically by an RD-09 electric
motor. Bellows 12, made from 1Kh18N9T steel can be used to move the movable
thermocouple over a distance of approximately up to 280 mm, without disturbing
the air-tightness of the setup. The position of thermocouples in their vertical mo-
tion is determined by a selsyn indicator. The selsyn pickup is connected by a gear
and pinion to the shaft of the electric motor.
The receiving selsyn together with a three-arrow position indicator is locat-
ed at the control panel. It determines the height position of thermocouples to ap-
proximately within 0.3 mm. The angular position of the movable thermocouples
is varied manually during the work. The angles of rotation are read from gradu-
ated circle 19.
To monitor the temperature over the height of the working tank and the con-
denser, thermocouples are embedded at appropriate points (not shown in the
figure).
All the parts of the stand in contact with the heat-transfer agent are made
from 1Kh18N9T steel and are provided with nichrome electric heaters and ther-
mal insulation. The temperature of the subassemblies and pipelines of the setup
is monitored by chromel-alumel thermocouples.
REFERENCES
1. Lyon, R. and Foust, A.: Katz D. Chem. Eng. Progr. Symposium Series, v.
51, No. 12, 1955.
2. Hickey, I.: J. Appl. Phys. v. 24, No. 10 (1953).
3. Povarin, P. I. and Kulakov, I. G.: Informatsionnoye pis I mo No. 6. Energeti-
cheskiy Institute AN SSSR [Information Bulletin No. 6. Power Institute of
the Academy of Sciences of the USSR]. 1958.
4. Tsarev, B. M.: Raschet i konstruirovaniye elektronnykh lamp [Design and
Construction of Electron Tubes]. Gosenergoizdat, Moscow-Leningrad, 1952.
5. Sorikin, D. N., et al.: Byulleten' isobreteniy, No. 18, 1962.
6. Alad'yev, I. T. and Yashnov, V.I.: In the collection Konvektivnaya teplopere-
dacha v dvukhfaznom i odnofaznom potokaky [Convective Heat Transfer in
Twin-Phase and Single-Phase Flows]. Energiya Publishing House, Moscow-
Leningrad, 1964.
7. Kruzhilin, G. I. and Subbotin, V. I.: In the book Proceedings of the Second In-
ternational Conference on Peaceful Uses of Atomic Energy. Geneva, 1958.
263
Doklady sovetskikh uchonykh [Reports by Soviet Scientists], Vol. 2, p. 134,
Atomizdat, Moscow, 1959.
8. Kutateladze, S. S.: Osnovy teorii teploobmena [Fundamentals of Heat Trans-
fer Theory]. Mashgiz, Moscow-Leningrad, 1962.
264
TRANSFER OF HEAT FROM SODIUM BOILING IN A LARGE VESSEL
A. P. Kudryavtsev, D. M. Ovechkin, D. N. Sorokin,
The experiments were performed at flat horizontal sections which were /268
heated (at a diameter of 38 mm) by electron bombardment. Boiling took place at
a surface made from 1Kh18N9T steel. The experimental setup and the test sec-
tions are described in [1].
In these experiments we measured the temperature difference between the
wall and the fluid, pressure and temperature of the sodium in the vessel in the pow-
er supplied to the test section, and in a number of experiments also the tempera-
ture fields in the sodium, using amovable thermocouple. The indications of thermo-
couples which read the temperature difference between the wall and the fluid, the
fluid temperature and the readings of the movable thermocouple, which was con-
nected differentially with the thermocouple situated in the fluid were continuously
recorded by EPP-09 potentiometers (1 sec) and intermittently measured by the
R2/1 potentiometer. The oxygen content of the sodium was monitored by the meth-
od of distillation at the start and termination of experiments. Before start of tests
the working tank was evacuated to a pressure of about 10 -2 mm of Hg at a temper-
ature of about 300C.
We have performed three series of experiments (a series denotes a set of
experiments performed at one test section without purification and replacement
of the heat transfer agent). For convenience in analysis the data of two series of
experiments are broken up into subseries by the level of t s, which were obtained
in successive periods of time.
Before starting the experiments of series 1 oxides were removed from the /270
sodium in a cold trap. The oxygen content of the sodium did not exceed 10-3%by
weight. In experiments of series 2 and 3 the oxygen content of the sodium was in-
-1
tentionally brought up to approximately 10 % by weight.
Figure 1 shows the curve of a as a function of time for constant q and tfl.
The results of heat-transfer experiments are shown in the form of curves of a
plotted against q in Figs. 3, 4 and 6, in which the points are denoted by numerals
in accordance with the time sequence in which they were obtained. Figures 2 and
5 display the recordings of the temperature difference between the wall and the
fluid for a number of points. The large difference in the values of a in Figs. 3, 4
and 6 for the same q and to is not due to experimental error, but to the existence
of various heat-removal modes. For example, in Fig. 3 for q = 0.82 . 106 kcal/
/m 2 -hour one value of a is 15.6- 103 kcal/m 2-hour-degree, while another is 7.7
103 kcal/m2 -hour degree.
265
o 00
o G
N^
I '
Cd
x VC
o 4 00 ' o _ _ _ _
V O O O pO O
12 z, hours
Figure 1. Variation of a in Time (Test Series 1a) for

6 2
q = 0. 7 , 10 kcal/m -hour and to = 690C.
Analysis of data presented in Figs. 1, 3, 4 and 6 shows that they can be at-
tributed to the following three modes of heat removal, [2].
1. Heat removal from sodium boiling under fully developed conditions. The
wall temperature fluctuations here are moderate or entirely undiscernible (see
point 4, of Figs. 4b and 5a). The curve of a as a function of q under this mode is
the same (with a power of about 2/3) as for nonmetallic liquids. The data obtained
for sodium boiling under the above conditions can be expressed as
a 4q2/3. (1)
The curve obtained using this expression is 1 of Figs. 3, 4 and 6. No discernible /271
dependence of a on s was noticed for fully developed boiling of sodium in the pres-
sure range of the tests (about 0.1-1 absolute atmospheres).
Figure 7 shows a distribution of temperature in sodium, typical for fully de-
veloped boiling, measured by a movable thermocouple which, while at the wall,
registered temperature fluctuations with a frequency of about 1 cps, while not tem-
perature fluctuations were recorded away from the wall. The frequency of temper-
ature fluctuations at the surface of a heat-conducting wall attendant to boiling can
serve, judging by experiments with water of [3] and [4], as an indicator of a kind
of.frequency of break-off of vapor bubbles. The measured fluctuation near the heat
conducting wall comprises about 50% of the mean temperature difference Ot t Tem-
perature fluctuations at the surface of the heat conducting wall with magnitudes com-
mensurable with OZt0? have also been observed in boiling water, [5] and [6]. The thick-
ness of the boundary layer in sodium boiling under, fully developed conditions is about
3-4 mm.
266
2. Heat removal by free con
g) with subse--vection(whubl
quent dissipation of heat by surfac e
evaporation. Here the sodium tem-
perature at the wall may exceed the
saturation temperature by more than
100C. For example, in our exper-
ments (see Fig. 3) there was noboil-
ing at Ats 100 [q ;ze 0.7. 106 m2 -
_ v
-hour]. A similar picture was ob-
served in experiments depicted in
Fig. 4a. We should note that the
relationship between the amount of
superheat of the fluid at the heated
surface, the thermal flux and the
superheating of the fluid in the ves-
sel will depend on the design of the
setup.
4 2
The heat transfer coefficients
for heat removal by free convection
for our conditions are in good agree-
ment with the formula
24 C
u 95g 1 1 3 (2)
(Curve2 in Figs. 3, 4 and 5). The

above expression has been obtained
on the basis of experiments with
free convection using the same set-
up, [7]. Here the sodium was held
under inert-gas pressure (about 2
absolute atmospheres) and the so-
dium temperature at the heat-con-
Figure 2. Recording of the Temperature ducting wall was by several hun-
Difference Between the Wall and Fluid for dred degrees lower than t s . Mea-
Point 4 of Test Series 1a; surements made by a movable ther-
q = 0. 7 , 106 kcal/m2-hour. mocouple at to < is [7], as well as
1) a = 17.7 103 kcal/m2-hour-degree; at to > is show that, in case of free
2) a = 8.3 . 103 kcal/m2 -hour-degree. convection, no temperature fluctu-
ations can be observed at the head-
conducting wall (in the laminar re-
gion) or away from it. The thick-
ness of the thermal boundary layer
(about 15 mm) for convecting heat removal exceeds appreciably that of the bound-
ary layer attendant to developed boiling.
267
3, Heat removal attendant
to unsteady boiling. Under these
conditions the removal of heat is
accompanied by appreciable tem-
perature fluctuations in the wall
(see point 4 in Figs. 1 and 2).
4t The maximum magnitude ( max)
;3'jY 82
Cd
of fluctuations in the wall tem-
U
x 8
perature is equal to the differ-
2 3q S sz ence inAtûnder free convection
(without boiling) and attendant to
Z9
- fully developed boiling.
^
I The magnitudes of the heat

;p s 2 4 6 8 Q+kcal/(m2 hr) transfer coefficients for unsteady
boiling lie between the values of
Figure 3. Experimental Data for Series a for the two preceding heat-re- /273
lb; tfl= 700 12 Co Curve 1 Corresponds movalmodes; here the lesser the
to Eq. (1); Curve 2 Correspondsto Eq. (2). temperature fluctuations in the
heat-conducting wall, the closer
the heat transfer coefficients lie
to cx for developed boiling (Figs.
4 and 5). Experimental data for
unsteady boiling cannot be described by a single power law [as could be done for
the steady case, DBL], since the heat removal conditions for individual experi-
ments vary appreciably, coming close either to heat removal under developed
boiling or by convection without boiling.
Analysis of recordings of temperature fields in sodium, temperature dif-
ferences between the wall and the fluid, as well as direct observations of the
boiling of sodium using a special x-ray installation give grounds for assuming
the following mechanism for heat transfer by unsteady boiling.
After the first boiling up, large bubbles travel through the volume (seconds
or fractions of a second) which absorb the heat of the superheated sodium. As
a result of this boiling 6t and At are reduced to magnitudes corresponding ap-
proximately to developed boiling (see Fig. 2). As a rule, in the very beginning
(from several seconds to several tens of seconds) the formation of vapor bubbles
at the heat-conducting surface takes place in the same manner as in developed
boiling. Then, at At s which becomes established after the boiling up, the vapor /274
formation center ceases to function and the wall must be superheated more (as
compared with developed boiling) in order for a vapor bubble to form in the same
vapor-formation center.
It is certain to a given degree that several neighboring vapor-formation
centers may cease operating simultaneously. The larger the number of neigh-
boring vapor-formation centers which cease to function at a specified At and
268
--^
--- --- i --- --^ I 98
w j
! I
- - - 3 `x19
+17.. 2-
I ' f 15o4
l^
---- ---- -.---.__..__.__ _ _l 1__i 5 19
jS
a
X0,'3 1'r
.y
Zs i
5p12
10 1 1 21 1 1
,, k
4
105 z
4
a b X Q)kcal/(m' . hr) d
,05 RV s o 01
tJ
013
^ 6
v 0 721 1 10
8 of .
^ S
15 122
N 9 _
16
U
.G i
17J
I
i
; r--
12 172y
E
2
f
lo5 2 4 5 B 7p 6 Q,kcal/(m`.hr)
b
Figure 4. (continued on following page)
269
70 S .-
--_J ___!---
8 --- -- - - - -- ^113
10 12
^-
F 611
7p e16
,
ao 4 71
$ 82 - - -
- -- --- i--
61
N ii ^ 17
a 6J
a^
i SZd
- 2' i - -
Cd 9
U 32
1
Z`
2
fJ
4 180
10` 4 4.
1B1
J-
2
105 Z 4 6' X J0'" Q,kcal /(m`% hr)
C
Figure 4. Experimental Data for Test Series 2.

a) to = 700 12C (Test Series 2a) ; b) tfl. = 785
f 8C (Test Series 2b) ; c) to = 860 12C (Test
Series 2c).
and the higher the amount of superheat which is needed to restart them, more
will the head-conducting wall be overheated in the given spot.
Analysis of data shows the following.
1. As the thermal flux increases, tendencies appear to transition from con-
vection (without boiling) to unstable boiling (see Fig. 3, point 4), as well as to
transition from unstable to fully developed boiling. This is precisely the cause /275
for the "large slope_ " (> 0, 7) of the test data pertaining to unstable boiling (see
Fig. 4a).
2. As the sodium temperature rises, the thermal flux at which boiling
starts becomes smaller, and the probability of developed boiling increases. For
example, in test series lb boiling started at q ^-, 0. 8 . 106 kcal/m2 -hour (t fl ti
N 7000 C), while in series 2c there was boiling at q 0.2 . 106 kcal/m2-hour
(t fl^ 8600C).
3. Increase the time of contact (apparently to certain limits) between the
stainless-steel heat-conducting wall and the sodium results, all other conditions
remaining equal, in an increase in the thermal flux needed for maintaining the
boiling. For example, in test series la during the first eight hours we have
270
Figure 5. Recording of the Tem-
perature Difference Between the
Fluid and the Wallin Test Series
2b (a) and 2c (b).
a) q = 2.57 10 6 kcal/m2-hour,
a =107 . 10 3 kcal/m2 hour-degree
(Point 4) and a =59. 7 . 10 3 kcal/m2
-hourdeg(Point10);bq=.47
10 6 kcal/m 2 -hour, a = 25. 6.103
kcal/m 2 -hour-degree, a = 19.4-
-10 3 kcal/m 2 -hour-degree, a =
= 12.1 . 10 3 kcal/m2-hour-degree
and a = 9.7- 10 3 kcal/m 2 -hour-
a b
degree (Points 1, 2, 3 and 4,
Respectively.
105
8
17 0 0 73 '14
6
12 07
v
_s
^ 4 --- 17 i 05
7s
70 -
N y 03
^ 2
i^ 8
xU 2
is 01
6
m4
8
4 b.
Figure 6. Experimental Data of
195 Z 4 6' 8 10 Q,kcal/mZ -hour. Test Series 3; to =700 12C.
271
H, mm^
^J
30-
2J
I
10-
0 0
U
1 N
M
^;e
Figure 7. Temperature Distribution in Sodium for De-

veloped Boiling (Point 15, Test Series 2b).
q =0.7 . 106 kcal/m2-hour; a=38.9-10 3 kcal/m2 hour-degree.
noticed mutual transitions from unstable boiling to convection (without boiling).

During the subsequent nine hours heat removal took place by convection only at
the same q and tfi. In test series 2a boiling took place first at q - 0.5 . 106 kcal/
/m2 -hour, while at the end of the tests there was no boiling at q ^^ 0.7 . 106 kcal/
/m 2 -hour. A similar picture was observed also in series 2c and 3.
272
4. Within the limits of experimental accuracy no effect of oxygen on heat
transfer from boiling sodium was observed ( the oxygen content was varied ap-
-3 %
proximately from 10-1 to 10 by weight).
Measurements of temperature fields in the fluid and in the vapor show that
the degree of superheat of the fluid with respect to the vapor depends on the mode
of heat removal. For example, for point 14 in Fig. 4c pertaining to developed /277
boiling, the superheat 8t comprises about VC, while for points 17 and 18 of the
same test series, pertaining to unstable heating, it is about 15C. The super-
heating of sodium is also manifested by comparison of the experimentally mea-
sured sodium pressures and temperature with the expression for the saturation
pressure as a function of the temperature given by the formula [8]
I abs.at4.521^T^.
g7
(3)
Superheating attendant to boiling is produced by the fact that a part of the

supplied heat is given up to the fluid near the heat conducting wall, and then from
the fluid it is transmitted by vapor bubbles. If, by analogy with the heat transfer
coefficient, we introduce aint' which characterizes the thermal resistance of the
boundary layer near the phase interface and the thermal resistance of the phase
transition, then we can write an approximate expression for the total amount of
supplied heat Q (including the heat supplied by auxiliary heaters)
Q = J dint St dFO. (4)

ref,
For Q and Ps = const St will be the higher, the smaller F o , the phase interface
surface and aint. When there is no boiling, the phase interface is only the free
surface and St will be the highest. In the case of boiling St will be lower, since
F0 is made larger by the vapor bubbles. Under unstable boiling St should be
higher than for developed boiling. Qualitatively the above is substantiated also by
data for water. When water boils under atmospheric pressure the magnitude of
R depends on the surface roughness. The rougher the surface, the lower St,
since a larger number of vapor - formation centers operates at the surface [9].
The magnitude of a nt depends on the level of is as well as on the admixture
content of the sodium, since the latter have an appreciable effect on the thermal
resistance of the phase transition. Approximate estimates of aint made on the
basis of our measurements show that for a free phase interface for sodium in the
range of Ps -- 0.1 . -1 absolute atmospheres alnt is about 104 kcal/m2 hour - degree.
273
Several experiments on transfer of heat from boiling sodium were perform-
ed using test sections from copper, electroplated with nickel or with nickel with
chromium (about 20 microns). At temperatures of about 600C heat removal was
accomplished only at unstable boiling or by convection, with subsequent dissipa-
tion of heat by evaporation from the free surface. At temperatures above 700C
all the three heat-removal modes exist. However, the electroplated coatings
were found to be insufficiently stable to sodium. In certain experiments they
broke down and the sodium came into direct contact with the copper, with the re-
sult that an excessive rough surface was produced. Only developed boiling of so-
dium is observed at the copper surface in the temperature range of approximate-
ly 700-880C. The data obtained in these experiments on heat transfer attend- /278
ant to developed boiling and convection without boiling are in agreement with
Eqs. (1) and (2), respectively, while the values of a for unstable boiling lie be-
tween the a for developed boiling and for convection.
Following are the main features characteristic of sodium boiling in a large
vessel at a surface from stainless steel in a pressure range of about 0.1-1 abso-
lute atmosphere.
1. Feasibility of superheating the fluid at the surface by 100-150C over
is without taking any special measures. ,In this case heat removal maybe accom-
plished by convection (t fl > t s ) of large thermal fluxes (about 10 6 kcal/m2-hour)
due to high mean temperature differences and the relatively high heat transfer
coefficients for sodium.
2. Alongside with ordinary nucleate boiling, unstable boiling may also
exist at fluxes of up to qcr, [7] and [10].
3. Instability of heat transfer, consisting in the multiple change of heat-

removal modes during a relatively short time period (see Figs. 1-3).
In a number of cases water may also be appreciably superheated at the sur-
face. For example, it is pointed out in [11] that, using a clean glass surface,
Harvey superheated water up to 200C, while Kenrick obtained temperatures
of up to 270C. The wetting contact angle between water and pure glass is zero,
[12]. In experiments described in [13], after the system was subjected to a high
pressure it was possible to superheat water to about 160C on a nichrome wire
and up to about 150C on aplatinum wire at atmospheric pressure. The boiling
up was followed by ordinary nucleate boiling. Reference [14 ] notes that atvery
low pressures (for water probably below 0.02 absolute atmospheres) the state
of natural convection is followed by the transient mode, without passing through
the nucleate boiling mode.
In experiments described in [15-17] pertaining to heat transfer from boil-
ing water and a number of organic fluids in large vessels under low absolute pres-
sures (0.1 absolute atmospheres and below) the same dependence of a on q (about
0.7) was obtained as in the case of Ps > 1 absolute atm. The study of heat trans-
fer from water boiling on horizontal tubes in a large vessel at pressures of 0.05-
2. 02 absolute atmospheres and thermal fluxes of (0.013-0.32) 10 6 kcal/m2-hour
274
is reflected in [15]. Reference [16 ] describes experiments on the transfer of heat
from boiling water and ethyl alcohol under natural circulation conditions in the
range Ps = 0. 125-1 absolute atmospheres. Reference [171 presents experimental
data obtained by Braunlish with a large volume of boiling water (PS 0.09-1.03
absolute atmospheres) and those due to Kruder for boiling water (P =0.037-
1.56 absolute atmospheres), methyl (P s = 0.08-1.42 atm abs) and butyl (P s = 0.17-
1.41 atm abs) alcohol in a large vessel.
The authors of the present article have performed experiments on the trans-
fer of heat from boiling water in a pressure range of about 0. 026-1 atm abs and /279
with thermal fluxes in excess of 10 5 kcal/m2 -hour (surface from 1Kh18N9T steel).
The wall thermocouples did not record any temperature fluctuations and boiling at
constant andps did not cease. The heat transfer coefficient was proportional to
q0.7 and ps0.18
The superheating of water described in [11] and [13] is attributable to con-

ditions of originating of the vapor bubbles. It has been shown in a number of the-
oretical and experimental works, [11-13] and [18-22] that, when fluids boil
on solid surfaces, the vapor bubbles originate in depressions filled with vapor or
gas. They primarily originate in depressions with poor wettability. Whenabub-
'ble is formed in a vapor or gas filled depression, the radius of the depression in-
let which is still capable of serving as a vapor-formation center can be estimated
from the expression
' 2A6T8
Pmin= ry" Ate (5)
In the case when the dimensions of vapor nuclei (at dts = 1C) are large or when
the wettability is good, while there are no gas (vapor) [holding] cavities on the
surface, boiling up requires high superheats Ats.
There is no basis for the assumption that conditions needed for origination
of vapor bubbles for metals are different from those for nonmetallic fluids. For
example, in experiments with boiling mercury [23] and [24] the same qualitative
picture was obtained with wetted and unwetted surfaces as for water boiling un-
der similar conditions [25]. In both cases when the surface was not wetted the
formation of the vapor phase at the heat-conducting surface started at a temper-
ature close to ts.
Of great importance for boiling is the stability of the vapor-formation cen-

ter functioning. For example, it is noted in [18] and [21] that the most stable
depressions are those with greatest wetting angle.
The appreciable superheats of sodium at the surface, the existence of dif-
ferent modes of heat removal and the instability of heat. transfer attendant to
275
boiling which were observed in our experiments (P s ft 0.1-1 abs atm) can be ex-
plained, ofi the basis of presently utilized concepts about the origination of vapor
bubbles, by the following main causes.
1. As was already noted in [26], the vapor nuclei of sodium superheated
by 10 C are large. For example, at P s = 1 abs atm and the same dt ,
s Amin' the
radius of the first nucleus, calculated from Eq. (5) is ten-fold greater for sodium
than for water. Experiments with sodium were performed on surfaces on which
water boiled at At = 5C (atmospheric pressure). If it is assumed that the same
vapor- or gas-filled depressions serve as vapor formation centers for sodium al-
so, then according to Eq. (5) they should have corresponding to them superheats
of At s = 200C at ts = 700C and At s = 78C at t s = 800C. However, these values of
At are higher than experimental values, particularly than those for developed /280
boiling. It, therefore, maybe assumedthat relatively larger depressions werethe
[actual] vapor formation centers. At surfaces not specially machined the size
distribution of the depressions should be close to normal and the number of "large"
depressions, which are the predominant vapor-formation centers may be relative-
ly low.
However, this factor alone does not suffice for explaining the experimental
data we have obtained. For example, in test series 3 at t flc^-, 700C and AV--23C
(point 8) the fluid was boiling, while in test series 2c the boiling ceased tempor-
arily (point 18) at a higher temperature ( tflL- 860C) and At c-- 100C. The value
of Amin calculated from Eq. (5) for point 8 of test series 3 exceeded by approx-
imately 20-fold the pmin for point 18 in test series 2c. The increased value of
pmin alone does not suffice to account for the existance of developed and unstable
boiling under identical pressures at surfaces with the same finish.
2. The heated surface is wetted satisfactorily by the sodium. According
to available data, metal surfaces are wetted satisfactorily by alkaline metals at
temperatures of about 500C and above [27].
Sodium is capable of reducing oxides of metals, if the free energy of the
system becomes lower as a result of the reaction
\11 01 ,, - 211'\ a -7 iii-Na,O - Wd
(here M denotes a metal). The temperatures at which our experiments were per-
formed (about 700-900C) are such that sodium reduces oxides of nickel, molyb-
denum and iron. The rate of oxide reduction depends on the surface material and
the temperature level, however we are not aware of quantitative data on this. Ap-
parently the satisfactory wettability of metals by sodium at temperatures above
500C is due to the reduction of oxide films, which is substantiated by experiments
concerned with wettability of oxides of certain metals by sodium (the extreme wet-
ting angle in this case being about 110-140C) [28].
276
When the oxide film of a surface is reduced, the latest parts of the latter
to undergo such a reaction will, most probably, be the oxides in the depressions.
As the wettability is improved, the magnitude of the superheat At needed for in-
itial boiling up and the probability of existence of unstable boiling should increase.
These tendencies were proven experimentally.
When boiling sodium (and other metals) one must determine whether the so-
dium reduces the oxide films at the heat-conducting surface (nongetter surface
or, conversely, the material of the heat-conducting surface reduces the oxides of
the sodium (getter-type surface), Zirconium and titanium can serve as examples
of getter materials for sodium. Boiling of sodium at getter and nongetter sur-
faces may differ appreciably.due to the difference in the wettability of these sur-
faces. All the discussions of this article pertain to nongetter surfaces.
3. High thermal conductivity of sodium. High thermal conductivity pro-
motes superheating of the fluid in the entire volume, on one hand, and increases
the time needed for superheating the fluid at the heat-conducting surface after
break-off of the bubble on the other hand. The latter increases the probability
that the vapor remaining in the depression after the bubble breaks off will be re-
placed by fluid. The time needed for superheating this fluid, estimated on the as-
sumption that the latter is superheated only due to thermal conductivity, is by ap-
proximately 20-fold greater for sodium (700-900C) than for water (100 0 C), if q
and At in both cases are the same. This also proves that the frequency of vapor
bubble breakoff from boiling sodium should be lower than in the case of water.
The same opinion was expressed above on the basis of the measured frequency /281
of temperature fluctuations in developed boiling, determined by a movable ther-
mocouple situated at the wall.
Thus, the relative large size of the depressions which serve as vapor-for-
mation centers for sodium at P s sz^ 0.1-1 abs atm and the good wettability of the
latter are responsible for the high At on initial boiling up, as well as (together
with the high thermal conductivity) for the instability in the functioning of the va-
por-formation centers.
The low values of aint at the free surface of the liquid metal aid in the ex-
istence of unstable boiling in sodium. The fluid in the vessel can be appreciably
superheated only at low oint at the free surface. The magnitude of At on boiling
is reduced due to a reduction in At e, as well as of St. The difference in At for
heat removal by free convection without boiling the heat removal with boiling
(q = const) can depend appreciably on the R attendant to convection without boil-
ing. The larger the ration St/At s , the more probable it is that boiling will cease.
The depression in the functioning of the vapor-formation centers, and the

subsequent cessation of boiling could also have been affected by the pressure surge
accompanying the boiling up. However, comparing the recording of temper-
ature differences between the wall and the fluid and the experimentally observed
time during which the pressure surge took place, we came to the conclusion that
this effect is either entirely absent or insignificant.
277
In closing, let us consider the effect of the oxygen content of sodium on heat
transfer on boiling.
If the oxygen content is within the limits of solubility at boiling temperatures,
then the sodium reduces the oxide film on the surface of the construction materials
noted above. In conjunction with this it may be assumed that the oxygen content of'
the sodium (within the limits of solubility) should not appreciably affect heat trans-
fer attendant to boiling, which is also substantiated by our experiments. If, how-
ever, the oxygen content of the sodium exceeds the solubility at the given temper-
ature; then the oxides may become deposited on the heating surface and thus affect
the heat transfer on boiling.
The subject of heat transfer from sodium boiling in a large vessel is treated /282
in [29 - 321. In [29] and [32] the experiments were performed in a wide range of
thermal fluxes. These data are, as a whole, in satisfactory agreement with
those obtained by us. In [30] and [31] the experiments were performed with low
thermal fluxes and the data thus obtained lie above our data and the data of [29]
and [32]. It should be noted that at low thermal fluxes appreciable errors in de-
termining the heat transfer coefficient are possible due to instability of the boil-
ing process as well as in the determination of the mean temperature difference.
NOTATION
is - saturation temperature, C
t - temperature of heat conducting surface, C
O max - maximum magnitude of the fluctuation in the temperature difference
between the fluid and the wall, C?
Ps - saturation pressure
a nt - heat transfer coefficient at the fluid-vapor interface
r - heat of vaporization
F - surface of the heat conducting surface
F - area of the interface
ID
A = 1/427 - thermal equivalent of work
REFERENCES
1. Ivashkevich, A. A., et al.: Ustanovka s elektronnym obogrevom dlya issledo-

vamiya teploobmena pri kipenii metallov v usloviyakh svobodnoy konvektsii
278
[Electronically-Heated Setup for the Study of Heat Transfer from Boiling
Metals by Free Convection]. First article of Part III of this collection.
2. Subbotin, V. I., et al.: USSR report No. 328, Proceedings of the Third In-
ternational Conference on the Peaceful uses of Atomic Energy, Geneva,
1964, v. 8, N.Y., 1965.
3. Rogers, T. and Mesler, R. A.: I. Ch. E. Journal, v. 10, No. 5, 656 (1964).
4. Torikai, K.: Report of Japan No. 580. Proceedings of the Third Interna-
tional Conference on the Peaceful uses of Atomic Energy, Geneva, 1964,
v. 8, N.Y., 1965.
5. Moore, F. D. and Mesler R. B. A.: I Ch. E. Journal, v. 7, No. 4, 620 (1961).
6. Hsu, Y. Y., et al.: Trans. ASME, Ser. C, No. 3, 254 (1961).
7. Kudryavtsev, A. P., et al. : Eksperimental'noyeissledovaniyeteplootdachiothor-
izontal'noy ploskoy poverkhnosti k natriyu pri svobadnoy konvektsii [Ex-
perimental Study of Heat Transfer from a Horizontal Flat Surface to So-
dium by Free Convection] Part I or II of present collection.
8. Makansi, M. M., et al.: J. Phys. Chem., 50, 40 (1955).
9. Jakob, M. and Fritz, W.: Forschung Gebiete Ingenieur, 2 (434), 1931.
10. Kudryavtsev, A. P., et al.: Eksperimental'noye issledovaniye kriticheskikh
teplovykh potokov pri kipenii natriya v bol'shom ob"yeme [Experimental
Study of Critical Thermal Fluxes in Sodium Boiling in a Large Vessel].
Article immediately following the present.
11. Paese, D. C. and Blinks, L. R.: J. Phys. Colloid. Chem., 51, No. 2,
556 (1947).
12. Harvey, E. N. , et al.: J. Appl. Phys. , 18, No. 2, 162 (1947).
13. Sabersky, R. H. and Gates, C. W.: Jet Propulsion, 25, No. 2, 67 (1952).
14. Schorck, V. E. and Lienhard, I. H.: Trans. ASME, Ser. C, v. 85, No. 3, /283
261 (1963).
15. Minchenko, F. P. and Firsova, E. V.: Article in the collection Voprosy
teplootdachi i gidravliki dvukhfgznykh sistem [Problems of Heat Trans-
fer and Fluid Flow of Two-Phase Systems]. Gosenergoizdat, Moscow,
1961.
16. Kolach, G.A. and Vvedenskiy, Yu. G.: Trudy Moskovskogo energetiche-
skogo instituta. Issure 48, 1963.
17. Kutateladze, S. S. and Borishanskiy, V. M.: Spravochnik po teplopereda-
che [Heat-Transfer Handbook]. Gosenergoizdat, Moscow-Leningrad,
1958.
18. Bankoff, S. G.: Trans. ASME, 79, No. 4, 735 (1957).
19. Corty, C. and Foust, A, S.: Chem. Engng Progr. Sympos. , Ser. 51, No.
17, 1 (1955).
20. Hsu, Y. Y.: Trans. ASME, Ser. C, v. 84, No. 3, 18 (1962).
21. Griffith, P. and Wallis, I.: Chem. Engng Progr. Sympos. , Ser. 56, No.
90, 49 (1960).
22. Clark, H. B., et al.: Chem, Engng Progr. Sympos. , Ser. 55, No. 29,
103 (1959).
23. Lozhkin, A. N. and Kanayev, A.A.: Binarnyye ustanovki[Binary Devices].
Mashgiz, Leningrad, 1946.
24. Bonilla, C. F., et al.: Chem. Engng Progr. Sympos. , Ser, v. 53, No.
20 (1957).
25. Averin, K. Ye.: Izvestiya AN SSSR, OTN, No. 3 (1954).
26. Deyev, V. I. and Solov'yev A. N.: Inzhenerno-fizicheskiy zhurnal, No. 6
(1964).
279
27. Balzhiser R. E. , et al.: Nuclear Science Abstracts, v. 1, No. 9, Abstracts
14033 (1964).
28. Bradhurst, D. H. and Buchanon, A. S.: P. zn. Ehimiya, Ref. 116543 (1962).
29. Noyes, R. S.: Trans. ASME, Ser. C, 85, No. 2, 125 (1963).
30. Lyon, R., et al.: Chem. Engng Progr. Sympos. , Ser, v. 51, No. 7,41 (1955).
31. Borishanskiy, V. M. , et al.: Atomnaya eneriya, 19, Issue 2, 191 (1965).
32. Petukov, B.S., et al.: Third International Heat Transfer Conference, Vol. V.
Chicago, August 7, 1966.
280
EXPERIMENTAL STUDY OF CRITICAL THERMAL FLUXES IN SODIUM
BOILING IN A LARGE VESSEL
V.I. Subbotin and A. A. Tsyganok
Up to present very little was published on critical thermal fluxes in metals /283
boiling in a large vessel. A study [11, was made of the dependence of the criti-
cal thermal flow in a large volume of magnesium boiling on horizontal tube on the
magnesiu m concentration in the mercury. As the magnesium concentration in the
mercury increased, the critical flux also increased and reached a maximum for a
magnesium content of about 0. 03 % by weight. In the pressure range under study
(1-10 abs atm) no dependence of the thermal flux on the pressure was observed.
Reference [2] describes the study of critical thermal fluxes in boiling potas-
sium in a large volume at a horizontal tube in a pressure range of 0. 007-1.55 abs
atm (the magnitude of the critical flux at atmospheric pressure was 1.63 . 106/m2 -
hour). Temperature fluctuations in the volume were observed, as well as super- /284
heating of the potassium above the saturation temperature. Reference [3 ) cites an
empirical equation describing these experimental data
1.66.106. Po,iss, (1)

9cr
where Ps is in kg/cm2.
Noyes [4] has studied critical thermal fluxes in sodium boiling under condi-
tions of natural convection at horizontal molybdenum and stainless steel tubes un-
der pressures of about 0. 03-0.07 kg/cm2 . On the basis of these experimental
data, as well as for data obtained by McAdams for water and some other organic
fluids, Noyes supplements the equation for the boiling crisis under natural con-
vection conditions, obtained on the basis of the fluid-dynamics concept of crisis,
[5], by the factor Pr-O. 245
\
1' )
q cr = 0.144ry" 1 y
( _ J1 Pr- 0,245. (2)
According to the author, this equation is valid for sodium also at higher pressures,
as well as for other metals.
We have performed seven series of experiments at a flat horizontal surface
heated by electron bombardment at a radius of 38 mm. The heat conducting surface
281
was placed in a working tank from 1Kh18N9T steel 165 mm in diameter. The
heat was removed from the condenser by air. The temperature of the heat-con-
ducting wall was measured by chromel-alumel thermocouples, with wires 0.2 mm
in diameter which were covered by aluminum oxide and placed in a protective
capillary 0.8 x 0. 15 mm in diameter made from 1Kh18N9T steel. The capillaries
were placed into holes 0.85 mm in diameter drilled in the wall. The sodium tem-
perature was also measured by chromel-alumel thermocouples (wire diameter 0.5
mm), placed in a sleeve.
The temperature difference between the fluid and the wall and the sodium
temperature were continuously recorded on the EPP-09 automatic potentiome-
ters, in which 1 sec is needed for the carriage to traverse the entire scale. The
crisis was recorded on the basis of a sharp increase in the temperature differ-
ence between the wall and the fluid. The [power] supply was shut off by end
switches, with which the potentiometers were provided.
The first five test series were performed using a surface from the VZh-98
alloy, the sixth series was conducted on a molybdenum surface, while the surface
for the seventh series was made from 1Kh18N9T steel. After the first crisis was
obtained, the load was increased rapidly (approximately up to 80% of the critical
load) and then the power supply was increased by steps of approximately 3 %. The
setup and the methods used for performing experiments for measuring q cr are de-
scribed in detail in [6].
As was already pointed out [7], sodium boiling under natural convection
conditions in a pressure range of about 0.1-1 abs atm can proceed by two modes.
1. Developed boiling mode, under which only insignificant fluctuations in
the wall temperature are recorded by the thermocouples.
2. Unstable boiling mode, under which these temperature fluctuations may /285
be very high (for example, at q ^ 10 6 kcal/m2 -hour the fluctuations may be as
high as 700 C). Depending on the boiling mode preceding the initiation of the cri-
sis, the magnitude of qcr may differ.
Figure 1 shows examples of recorded temperatures of the molybdenum wall

before and during the crisis for points 4 and 8. For point 4 the temperature fluc-
tuations before the crisis were appreciably greater than those for point 8. The
magnitude of the critical thermal flux for point 4 is by approximately 60% lower
than qcr for point 8, although the pressure (P s ) in both cases was approximately
the same. In our experiments on heat transfer from a copper surface the ther-
mal fluxes exceeded the q cr presented in the present article and no crisis ensued
at all, although the sodium pressure in both cases was the same.
The difference in the values of qcr brought about by the mode of boiling
(developed and unstable boiling) preceding the crisis can be attributed to the fact
that the probability of onset of the boiling crisis is the higher, the larger the va-
por bubbles or the vapor regions at the heat conducting wall. This point of view /286
can be supported by data for water and organic fluids, for which q cr increases
282

0
00C --
Figure 1. Record of the Temperature of a Molybdenum

Wall Before and During the Crisis for Point 4 (a)
and Point 8 (b) of Fig. 2.
with underheating [?] of the fluid to is and with an increase in the pressure (up to
1/3 Pd. This can be attributed to the fact that with an increase in the underheat-
ing and in P the maximum diameter of bubbles situated at the wall decreases al-
though the number of vapor-formation centers increases with an increase in Ps
(for a constant thermal flux).
Figure 2 shows values of qcr obtained in boiling sodium. The level of crit-
ical thermal fluxes can, with a scatter of 30%, be described by the formula
4cr = (1.5 -1.3P 5 ) 108, ( 3)
where Ps is in atmospheres. The above empirical expression (line 4) can be used

only in the range of pressures 1.5 . 10-2 -1.2 abs atm. Experimental points obtained
by Noyes [4] are in good agreement with the above expression. However, on the
whole, the agreement between Eqs (2) and (3) is much worse (see Fig. 2). Figure
3 also displays lines obtained from criterial formulas of [5] and [8], which de-
scribed experimental data on qcr for water and organic fluids boiling in a large
vessel
gcr
= 0. 14,
V(
283
^ I ^ (^ I 2 m
Cd I o^G s
_^8^
o 3
i
U L
o
il
O f ! o o O 40
4
B
ô o r---^ e ^ off n
0
I
Û I i i I
I
I I I I 3
^L
y' 0,02 0,04 006(;080,1 0,2 0,4 40 08 1 Ps,abs atm
Figure 2, Experimental Data on Critical Thermal Fluxes:

c, n.. 4, 3, o and [sic] - the V Zh-98 alloy; -in - molybdenum;
o - 1Kh18N9T steel; - data of [4],
Curves Obtained for the formula
I 5
2 2
.3 4
4 3
11.32 ,0.2110.4 (ry %)0 , 36 (y Y")0.49

/cr 4700
2.0.:sIC1 ue'-to.is (5)
In our calculations we have used the values of A, C and v from [9], while /287
r and y" were taken from [10] and y from [11]. The values of v were calculated
from the expression
6 - = 200.12 9.833.10 - 21,
presented in [11] a The values of Ps were obtained from the formula
Ig Pa t,,: !.521 15 Y'2s0 , (6)
also from the above reference.
As can be seen from Fig, 2, the Noyes formula [2] (should be (2), DBL)
[line 2] agrees with Eq. (3) only in a relatively narrow pressure range (about
0.025-0a 1 atm abs), and a difference of about a factor of two is observed at ex-
tremes of the pressure range under study. The values of q cr calculated from
Eq. (4) [line 3 1 lie by approximately a factor of three lower than the experimental
284
data, while those calculated from Eq. (5) [ line 1 ] lie by approximately a factor of
three above the latter.
Formula (4) is based on the fluid dynamics theory of the boiling crisis and
satisfactorily expresses the manner in which qcr depends on o s f r water and or-
ganic fluids. The fact that this formula does not agree with the experimentally
obtained values of q cr does not mean that the nature of the crisis attendant to the
boiling of metal does not obey the laws of fluid dynamics. It is possible that in
boiling metals, due to their high thermal conductivities, a substantial part of the
heat from the heat-conducting surface is transferred by the fluid even at thermal
fluxes close to the critical. This may be one of the factors responsible for the
disagreement between experimentally obtained qcr for sodium and those given by
E q . (4).
It can be seen from Fig. 2 that points obtained on the surface made from
the VZh-98 alloy have a scatter of less than 30%. If an averaging line is drawn
through these points, then its slope will be approximately the same as the slope
of the line obtained from Eq. (1) which describes experimental data for potassium.
On the other hand, the slopes of lines 1, 2 and 3 are much greater. However, the
available data still do not suffice to obtain a generalized expression for q cr as a
factor of Ps , as well as for determining the laws governing the effect of the sur-
face on qcr in boiling metals.
REFERENCES
1. Korneyev, M. I.: Teploenergetika, No. 4 (1955).

2. Balzhiser R. E., et al.: Nuclear Science Abstracts, 18, No. 9, 14033 (1964).
3. Brooks, R. D. and Bonilla, C. F.: Nucleonics, 22(x, 43 (1964),
4. Noyes R. C.: Trans. ASME, 85g (1963).
5. Kutateladze, S. S.: Osnovy teorii teploobmena [Fundamentals of the Heat Trans-
fer theory]. Mashgiz, Moscow, 1962.
6. Ivashkevich, A. A., et al.: Ustanovka s elektronnym obogrevom dlya issledo-
vaniya teploobmena pri kipenii metallov v usloviyakh svobodnoy konvektsii
[Electronically-Heated Setup for the Study of Heat Transfer from Boiling
Metals by Free Convection]. First article of Part III.
7. Subbotin, V.I. , et al.: Report No. 328, presented to the 3rd International
Conference on Peaceful Uses of Atomic Energy, Geneva, 1964.
8. Kruzhilin G. N. and Subbotin, V. I.: In the book Proceedings of the Second /288
International Conference on Peaceful Uses of Atomic Energy. Geneva,
1958. Doklady sovietskikh uchenykh [Reports by Soviet Scientists], Vol.
2, page 134, Atomizdat, Moscow, 1959.
9. Vargaftik, N. B.: Spravochnik po teplofizicheskim svoystvam gazov i zhid-
kostey [Handbook of Thermophysical Properties of Gases and Liquids].
Fizmatgiz, Moscow, 1963.
10. Volyak, L. D.: Inzhenerno-fizicheskiy zhurnal, No. 3 (1962).
285
11. Zhidkometallicheskiye teplonositeli [Liquid-Metal Heat-Transfer Agents].
Collection of articles translated from English under the editorship of
A. Ye. Sheyndlin, Foreign Literature Publishing House, Moscow, 1958.
286
MEAN TEMPERATURE DIFFERENCE IN LOW-PRESSURE VAPOR
GENERATORS AND CONDENSERS
P. L. Kirillov
During the last few years the need has arisen for low-pressure vapor gen- /288
erators. In particular, a number of projects [1] and [2], use is made of vapor
generators for obtaining mercury or potassium vapor, which is then directed in-
to a turbine. The vapor pressure at the turbine inlet can be from several tenths
of an atmosphere to several atmospheres. Here the flow resistance of the vapor
generator is comparable with the absolute presssure of the vapor at the outlet.
This produces an unusual distribution of the mean temperature of the heated heat-
transfer agent (it is cooled, despite the fact that heat is supplied -to it). The tem-
perature distribution of the vaporized heat transfer agent (saturation tempera-
ture) along the length is determined by the changes in pressure which, in their
turn, depend on the flow rate and on the vapor content at each section. This pic-
ture is observed in vapor generators operating at any pressure, but the drop in
the saturation temperature, produced by a reduction in pressure usually does not
exceed several degrees and may be disregarded in calculations. At low pressure
the temperature of the heat-transfer agent may drop by several tens of degrees,
which is appreciable.
According to the above, the temperature of the heat transfer agent over the
length and the mean temperature pressure will not obey the exponential distribu-
tion, meaning that the ordinary expression for the log mean temperature differ-
ence is no longer valid. The purpose of the present paper is to obtain a new ex-
pression for the log mean temperature difference under these conditions.
For the elementary heat transfer surface in a vapor generator we, as us-
ual, have the following equations (Fig. 1)
dQ_ A-( 1 , t ) df, (1)
dQ = c,G, dt, = G;; dq , ( for straight flow) (2)
dQ c,G, dt, = Gz d92 (for counterflow). (3)
Here dq2 = di2 + wdw/g. We restrict ourselves to low velocities, such that /289
wdw/g << di2.
It is obvious that one can assume a constant k for each vapor-generator

section only approximately. However, the manner in which the heat-transfer
287

t t coefficient varies over the length,

particularly, due to changes in
the coefficient of heat transfer
to the two-phase flow, is so com-
plicated, that analytical calcu-
lation in which k appears as a
function of length are as yet
t"e tz
difficult to perform. In addi-
tion, the relationship a(Z) has
o f f o F f not been as yet studied in all
Q b "
the regions of the flow and the
Figure 1. Temperature Variation in the Vapor- available experimental data do
izing Part of the Vapor Generator. not suffice for constructing
an analytical formula, hence,
a) for Straight Flow; b) for Counterflow..
F
lc= ^kdf. (4)
0
The validity of the above expression, which is not generally obvious, was proven
in [3].
The change in the enthalpy dig will be found from known thermodynamic re-
lationships (see [41)
di =T ds+Av dp, (5)
T ds = c x dT ;-r dx, (6)
U" 4
dp = tr v^ dT. (8)

Substituting Eqs. (6)- ( 8) in Eq. ( 5), setting v' v", and disregarding changes /290
in the latent heat of vaporization, we will get for our case
ai, c' dt - ;- r da . (9)
The above expression can also be obtained from a T-s diagram.
;i i3 i ' - -- c'(it rS I'S --r.,.r , _--C'(11 ; - rd.L ,-arlr, (10)
288
whence, if we disregard changes in r, we get Eq. (9) Substituting farther this
expression into Eqs. - (2) and (3), we get
dQ
(11)
c(IQ rdx
(12)
2 2
Now
rdx
dt1dt2=( i
c1G1 1 )dQ
Tc2G2 CZ
Substituting dQ = k ptx df, we get
1G1 +C2G2
d (At.,) _ (--!
0
1) hJt r d f rC2dx . (13)
Here the minus sign pertains to straight flow, while the plus sign is for counter-
flow. Denoting furtherL/c
\ 1 1 G + 1/c 2 O G lk = n, _ r dx = f , we arrive at
the expression c2 Tf) ()

(14)
f
the solution of which is
At, =e nt [ ^ m(f)e" t dj+t] . (15)
We assume in the first approximation that the vapor content varies linearly over
the length, then dx/ df = const and
r dx
M (f)=c2 df =const. (16)
From Eq. ( 15) we get that

At, = Ae- n / (P+ (17)
289
where
r dz
czn df
It is easy to see that, when dx and cp = 0, Eq. (17) yields an exponential tempera=
ture distribution over the length and reduces to the ordinary expression for the
log mean temperature difference.
For straight flow we get (see Fig. 1a) /291
A t., (F) _: tl t" --- At" = ale- F cp, (18)
whence
A ----(At ;- r,,) OFD (19)
At' ._ (t i 12) _ - A Y. (20)
Equating Eqs. (19) and (20), we get
dt'_cP
In = rzF. (21)
Let us now find the mean temperature difference for the entire straight-flow va-
por generator
F
T , (e- nF -1)
0t = F At,, df = F (22)
0
After Eqs, (19) and (21) are substituted here and simple transformations are
made, we get
At"At'
^t = ;ice cp ^' (23)
In At' +
'
For counterflow (see Fig. 1b), we have

t i t2 = A cp, (24)
:le- ,7F cp, (25)
290

whence we get an expression similar to Eq. (21)
In tit2- + ,p =nF
t, _ .t " T (26)
and the average temperature difference
dt^ (tit)(ti t2) ^ (27)

In ti t_+(p
titz+(P
The above expression and Eq. (23) can be written in the general form
Ot = ,tc),t"

ln,t6+(P tP,
(28)
Atm +T
where
_ r dx
(P = 1 2 ^J (29)
C2
ciG! + c2G2 / k
The minus sign is for straight flow, while the plus sign is for counterflow.
The processes - taking place in condensers are similar (Fag. 2). For an el-
ementary heat transfer surface in this case we have /292
dQ - _ r, (t, --- t,), (if, (30)
" 'Q -- --G, di, : C,C, dt, (for straight flow), (31)
----G, df, _ --c,G, (It ., (for counterflow), (32)
Examining Eqs. (5)- ( 8) or a T-s diagram for our case, as this was done above,
we get
dil _ : -- c, dt, r d.r. (33)
291

t Substituting this expression in

Eqs. (31) and (32), we write
r dx (34)
dt t = cjG
('Q c, ,
j
`Z^dQ
F F f dt2 = c2C2
(35)
a b
1 1
d(t' t 2 ). _(e,C,
Figure 2. Temperature Variation in a Condenser. \ _ C2c 2 J
r dx
a) for Straight Flow; b) for Counterflow.. lc, (t, t 2 ) T d f rl (36)
or
d (Atx)
di
_ 1
c,C,
1
^2G2
1/
k' Atx
__ r dx
ridf 37
Denoting here
r dx 1 1
i aff , nl = C
m f (f) _ ^ CZGZ) k1+
CA
we arrive at an equation, entirely analogous to Eq. (14)
d (At x) +n;At
di
x = m, ( f). (38)
Solving it in the method presented above, we get an expression for the log mean
temperature difference in the condenser
t b AtM
At = A
In Atb-}-rpf '(Pi, (39)

where /293
r 1 dx
Cl 1 1 'IT
C c,G;i i CI / h' (40)
The plus sign pertains to straight flow, while the minus sign is for counterflow.
292
REFERENCES
1. Proceedings of 1962 High-Temperature Liquid-Metal Heat Transfer Technol-

ogy Meeting, BNL-756, N.Y., 1962.
2. Davis, H. L.: Nucleonics, 22(x, 34 (1964).
3. Ushakov, P. A.: Analiz primenimosti srednikh koeffitsientov teploobmena
dlya rascheta teploobmennikov [Analysis of the Applicability of Average
Heat Transfer Coefficients for Heat-Exchanger Design]. Part II of pres-
ent collection.
4. Vukalovich, M. P. and Novikov, I. I.: Tekhnicheskaya termodinamika [ Engi-
neering Thermodynamics]. 3rd edition. Gosenergoizdat, Moscow-
Leningrad, 1962.
293
TRANSFER OF HEAT FROM CONDENSING SODIUM VAPOR
N. B. Bakulin, M. N. Ivanovskiy, V. P. Sorokin,
V. I. Subbotin and B. A. Chulkov
The experimental study of heat transfer from condensing vapors of pure al- /293
kaline metals is made difficult by the aggressiveness of the latter, the high tem-
perature level and the need to measure high heat transfer coefficients under the
above conditions. Small admixtures of noncodensing gases, vapor of metals
with low boiling temperatures or contaminant deposition on the condenser wall
can result in appreciable distortions. Instability of vapor formation under cer-
tain vapor-generator operation mode makes measurements difficult.
It is known that steel surfaces are well wetted by sodium and that it under-
goes film condensation. Due to the high thermal conductivity of liquid sodium
(60-80 w/m-degree), the condensate film creates a low thermal resistance R to
the transfer of heat to the condenser wall. Under these conditions heat transfer
is appreciably affected by the following boundary conditions:.
1) The temperature. difference between the vapor and the condensate needed
for bringing about the phase transition results in the appearance of resistance Rph
to the phase transition;
2) When the vapor contains noncondensing gases or other admixtures, a dif-
fusion resistance Rd arises in the vapor near the condensate surface;
3) Contaminant deposits on the condenser wall create a thermal contact re-

sistance Rc. The total thermal resistance attendant to condensation is [thus] com-
posed of the above resistances. (R = R f + Rp , + Rd + Rc). The contribution made /294
by a given resistance to the total resistance depends on the specific conditions of
condensation.
At present very little has been published on the transfer of heat from con-
densing sodium vapor [1-3]. The heat transfer coefficients obtained in [1] and
[2] are appreciably lower than those which can be attributed to the thermal re-
sistance of the condensate film. The measurements in [1] were performed at va-
por temperatures of 630-870C and with temperature differences of 2.7-4.1C be-
tween the vapor and the wall; the heat transfer coefficients were 65-75 kW/m 2
-degr,i.lsthan20%ofericl nts,whakeiocunt
only the resistance of the condensate film. No reasons are given for obtaining
low (as compared with those calculated by the Nusselt theory) heat transfer co-
efficients.
To analyze the effect of boundary conditions on heat transfer attendant to
condensation of pure "stationary" sodium vapor, we have performed experiments
294
on a setup described in [4]. Using a movable thermocouple 0.5 mm in diameter,
we have measured the temperature fields in the vapor, in the liquid sodium filling
the condenser's bath and in the condenser wall, under steady thermal conditions.
To measure the wall temperatures, the movable thermocouple was lowered into
a narrow (1.2 mm) vertical slot milled in the condenser. The thermal fluxes were
calculated on the basis of the measured temperature gradients in the condensate
and at the condenser wall. The readings of thermocouples (as a function of the po-
sition of the thermocouple probe) for several characteristic thermal modes are
shown in Fig. 1. By examining the temperature field in liquid sodium and at the
wall it is possible to determine the presence or absence of a temperature jump at
the liquid-wall interface, i.e., on the magnitude of R c . In the temperature range /295
under study (up to 520 0 C), no R was found between pure sodium and 1Eh18N9T
steel. The table gives experimentally obtained values of temperature jumps at the
vapor-liquid interface, as well as the heat transfer coefficients p h = 1/R cal-
ph'
culated on the basis of these jumps. Figure 2 displays experimental values of ph
as a function of the temperature of the saturated vapor. It can be seen /296
from the figure that p h in the range of parameters under study increases highly
with a rise in the vapor temperature and at 480C it is about 50 kw/m-degree.
Experimental Data on Condensation of Pure Sodium Vapor
Vapor emperature Thermal Heat Transfer Coefficient

Condensation
Temperature Jump, Flux 2 kw /m2 - degree. Factor
C C kw/m 6 Theoretical 17 Experimental
, .0 1.3 986 71 103 1,18

4183.5 3.1 157 50.1 50.7 0.99
451.0 3.6 152 43,0 42.3 0.91
462.0 3.5 131 34.2 37.4 1.(A
464.5 3,4 130 35,9 33.3 1.03
433.5 5.3 117 17.6 22.1 1,t1
412.5 7.6 99 12.1 1:3,0 I.0'c
412.5 7.4 106 12.1 14.3 1.07
391.0 12.3 103 7.7 8.8 1.07
387.5 14.1 114 7.2 ^8.1 1.06
381.0 17.3 120 6.5 6.9 1.03
401.5 8.5 91 9.5 10.7 1.06
431.5 4.1 84 18.9 20.5 1.01
431.0 4.0 72 18.2 18,0 0.99
4 !15.0 3.3 77 24.3 23,3 0.93
445.0 3.2 81 24.3 25.3 1,02
461.0 2.3 86 33.4 27,3 0.90
461,0 2.2 77 33,2 35.0 1.03
411.0 7.1 84 11.7 1t,8 1.00
473.0 2.5 98 41.5 38.8 0.97
599.0 <1.5 195 4170 X130.0 I -
295
mmf
_ L i 44^ 491C ,1 Styaam
{-
i ^' 2 c ondensate
sl --- I. 521 ,
I
1
3 C ondenser wall
t, C
Figure 1. Temperature Fields Measured in

Condensing Sodium.
1) Steam; 2) Condensate; 3) Condenser wall.
Examination of the phase transition

process from the point of gas kinetics
^ 40 makes it possible to obtain theoretical
relationships for ph. The resultantmass
20 flux can be related to the temperature and
is pressure under phase transition conditions
0 1 1 { I { { { { {
by the well-known Hertz-Knudsen formula,
{ {11
380 400 420 440 460 t, C
which is precisely what was done in [5],
but, as is shown [6-8], this formula
Figure 2. Heat Transfer Coefficients does not take into account the increase in
Calculated on the Basis of Measured resultant mass flux produced by the con-
Temperature Jumps at the Vapor- vective motion of the vapor along the nor-
Liquid Interface and Thermal Flux. mal to the phase interface. When the con-
The Curve Represents Values of a h densation factors are close to unity, dis-
p regard of the convective velocity of the
Calculated From Eq. (3) With f = 1. steam produces large errors. The most
general expression for the resultant mass
flux attendant to phase transition was ob-
tained in [6]
(l;jlj2 r; 4c) (2.t0Ta/:,t)^1z (2teRT/\I)1/s

(1)
A correction for the effect of the vapor velocity was obtained in [6]; when consid-
ering the average molecular velocity it is
1/f
1/j 1/2 v/4c
296
In [7] this problem is considered in more detail for the particular case of v/c 1.
On the assumption that the velocity distribution of the molecules evaporating from
the fluid is Maxwellian, the same results were obtained as in [6], i. e., the cor-
rection is 2/(2 - f). Approximate consideration for the case of v/c >> 1 of [8] yields
a correction of about 1.7/(2 - f). For the case of v close to c the solution was in-
conclusive and the calculations are performed by successive approximations.
Using Eq. (1) we can write a formula for p h for vapor condensation at the
condensate surface
9 __ r+cn (To T) Po P
aph ToT (1/f-1/2a/4c)(ToT){ (2:tgRTo/\f)' 1z (2tgRT/Af)' /2 } ( 2)
In the case of TO close to T and v/c 1, we can simplify Eq. (2) and write
u.Ph= 2f \2zgRTo / \ dT /To 2To } ( 3)
In this case ph for condensation does not depend on the thermal flux and is a func-
tion of the vapor pressure (temperature) only.
Using experimental data on temperature jumps at the vapor-condensate in-
terface attendant to condensation of sodium and Eqs. (2) and (3), we can calculate
f for sodium. The results of these calculations are presented in the table and in Fig. /297
3. The physical parameters for sodium vapor in the calculations were taken from
[9]. As can be seen from Fig. 3, the experimental points for f thus found are close
to unity in the entire temperature range. This result is in agreement with the gen-
erally held view that the condensation factors for liquid and solid metals, in which
the vaporizing molecules are identical with the condensating ones should be unity.
However, the experiments which served as the basis for this opinion were per-
formed at low pressures and low phase transition rates. The data obtained by the
present authors give grounds for the assumption that f for pure sodium in the pres-
sure range covered in the experiments does not depend on the pressure and is equal
to unity.
The thermal resistance of the
phase transition on condensation of
sodium can be calculated from for-
f8 --- o -- mulas of the gas kinetic theory set-
ting f = 1. When there are not dif-
0 fusion and thermal contact resist-
0.8 -
ances and the pressures are not too
380 400 420 440 460 t' -C low, the moderate resistance of the
transition results in a high rate of
Figure 3. Condensation Factors Calculated heat transfer from condensing vapor
on the Basis of Experimental Data. of pure sodium.
297
Although the experiments were performed on condensation of "stationary"
vapor, i. e., when the vapor did not move along the surface of condensation, the
thermal resistance of the phase transition will be the same also in the case of
moving vapor. The motion of the vapor along the surface of condensation does
not affect ph, since the longitudinal velocity of vapor in a vapor layer near to the
surface of condensation with a thickness of the order of a mean molecular path is
zero, or is a priori known to be less than the velocity of thermal motion of molecules.
We wish to note that heat transfer from condensing vapors of alkaline metals,
i. e., potassium [41 and sodium, is one of the high-rate processes of heat transfer.
Heat transfer coefficients under appropriate pressures and high-rate condensate
removal can be as high as 50 kw/m2 -degree and higher. However, as was pointed
out by the present authors in [4], even small admixtures of noncondensating gases
and other contaminants can appreciably reduce the rate of heat transfer. The prob-
lems of thermal resistance of the phase transition and the condensation factors that
were here considered may be of interest not only for condensation of metals, but
also in the study of processes of vacuum distillation, boiling, flow of metal vapors
through a nozzle, and other phenomena related to phase transformations.
NOTATION

f - condensation factor /298
v - convective velocity of the vapor in the direction normal to the
condensation surface
c - average thermal velocity of vapor molecules
R - universal gas constant
M - molecular weight
T0 , PO - temperature and pressure of vapor above the condensate, C
T, P - temperature at the condensate surface and the saturated vapor

pressure corresponding to this temperature, C
ph - coefficient of heat transfer attendant to phase transition
G - mass flux attendant to phase transition
298
REFERENCES
1. Misra B. and Bonilla C.: Chem _ . Engng. Progr. Sympos, Ser. 52 (18),
7-21 (1956).
2, Roth, L. T.: Proceeding of 1962 High-Temperature Liquid-Metal Heat Trans-
fer Technology Meeting, BNL-756, N.Y., 1962, 276.
3. Noyes, R. C.: See p. 97 of Ref. [2] above.
4. Subbotin, V. I. , et al.: Teplofizika vysokikh temperatur, No, 4, 617 (1964).
5. Silber, R. C.: Engineering, 161, No. 5 (505).
6. Rish, R.: Hely. phys, acta, 6, 132 (1933).
7. Kucherov, R. Ya. and Rikenglaz, L. E.: Doklady AN SSR, 133, No. 5, page
1130, 1960,
8. Zwick, S. A.: J. Appl. Phys. , 31, No. 10, 1735 (1960).
9. Vargaftik, N. B.: Spravochnik po teplofizicheskim svoystvam gazov i zhidkos-
tey [Handbook of Thermophysical Properties of Gases and Liquids]. Fiz-
matgiz, Moscow, 1963.
299
PART FOUR
TECHNIQUES OF WORK WITH LIQUID METALS
CORROSION RESISTANCE OF CONSTRUCTION MATERIALS TO SODIUM
V. V. Zotov, B. A. Nevzorov and Ye. V. Umnyashkin
Sodium and the eutectic alloy of sodium and potassium are the most exten- /299
sively used liquid-metal coolants. An important problem in their use is the se-
lection of construction materials with sufficient resistance to corrosion in non-
isothermal flow, satisfactory mechanical properties (ultimate strength, relative
elongation, etc.) at elevated temperatures, moderate thermal expansion coeffi-
cient as well as with good workability i. e. , ease of rolling, welding, etc.
In the study of corrosive action of sodium on various metals extensive use
is made of vial tests under isothermal conditions. Since it is impossible to de-
termine the actual rates of certain processes (mass transfer, solubility, effect of
nonmetal admixtures in the liquid metal, etc. ) the vial tests are of value only for
qualitative determination of the corrosion resistance of the material in the given
liquid-metal coolant. In tests in containers (vertical and inclined) the temperature
drops produce convection flows of the liquid metals, which increases the rates of
all corrosion processes. Directed flow of the coolant exists in loops with temper-
ature drops. In this case the flow velocities of the liquid-metal coolant depend on
the temperature difference and can be calculated. Nevertheless, results of cor-
rosion tests obtained in nonisothermal containers and convection loops cannot be
used with sufficient reliability in the design of equipment. Various rotary-type
test machines (with and without interchange of the liquid metal) are also unable to
produce the needed results. And only results of corrosion tests obtained by test-
ing materials in circuits with forced circulation can be used with confidence for /300
design purposes since the circulation loops afford the main parameters, i. e.,
specified coolant temperature and temperature drop, necessary flow velocities
of the liquid metal, nonmetallic admixture content of the latter, as well as a spe-
cific ratio of the coolant volume to the surface area washed by the flow.
Several kinds of corrosion interaction with the liquid metals are observed
when testing materials in circulation loops. The main kind of corrosion for steels
and alloys in liquid metals is by solution of the solid-phase components in the liq-
uid metal. The atoms of the solid substance may dissolve uniformly over the en-
tire washed surface. In this case the solution effect removes the entire surface
layer and no corrosion interaction zones are observed. If, however, the individ-
ual components of the alloy are dissolved selectively, then corrosion zones may
form on the surface in contact with the liquid metal due to the appearance of por-
osity, disintegration and pitting in the surface layer. Frequently corrosion in-
teraction has a clearly expressed intercrystalline character, i. e. , when the cor-
rosion proceeds along the grain boundaries. Intercrystalline corrosion in a liq-
uid sodium medium can extend, for certain metals, to a sufficiently great depth,
which in the final count results in an appreciable reduction in the ultimate strength
and in the relative elongation. The rate of such corrosion is sometimes appreciable,
300
since, while [only] individual atoms are carried away by the liquid metal on uni-
form solution, in intercrystalline corrosion entire crystals may be removed from
the surface of the constructional material. The dissolved components and the cor-
rosion products are carried by the liquid metal into cooler parts of the device,
where they settle out of the solution and crystallize on wall of pipelines, heat ex-
changers and other units, due to supersaturation attendant to the temperature re-
duction. This process is called thermal mass transfer or mass transfer due to a
temperature gradient. Deposition of the transferred metal particles and corrosion
products can result in partial or total stoppering up of ducts.
Another kind of mass transfer is also observed, i. e., concentration mass
transfer, which may take place under isothermal conditions. It comes about in
the case when several metals with different solubilities are in contact with the
liquid metal and when these metals interact with one another, forming interme-
tallic compounds, nitrides, carbides or solid solutions. Concentration mass
transfer is dangerous from the point of view of reducing the plasticity properties
of the metal, since the latter may become embrittled due to the formation of var-
ious compounds.
The rate of corrosion processes may be affected by the following factors: /301
test temperature, temperature gradient in the system, cyclical temperature
changes, ratio of the liquid-metal volume to the surface area of the system, pur-
ity of the liquid metal, its rate of flow, state of the metal surface (quality of fin-
ish, protective coatings, etc.), state of the material (heat treatment, presence
of internal stresses, grain size, etc. ), number of different metals in contactwith
the same liquid metal, chemical composition of the materials of construction, dur-
ation of tests, etc.
Chemical Nature of Oxygen in Liquid Sodium

The corrosion resistance of pure metals, as well as of structural steels and
alloys exposed to the atmosphere and immersed in electrolytic solutions depends to
a large extent on the protective properties of passive films which are oxidic in na-
ture. Many investigators believe that also in sodium the corrosion resistance of
materials depends to a certain extent on the chemical stability of their oxidic films.
By this attribute all the metallic materials can be divided into three classes.
The first class is made up of stainless steel and nickel, i. e. , materials the
oxides or compound oxides of which have a chemical stability in sodium smaller
than or close to the stability of sodium oxide.
The second class embraces materials, the oxides or compounds oxides of
which are more stable than sodium oxide. Here the oxides of these metals are
well coupled to the base; zirconium is presented as an example of these.
The third class contains materials the oxides of which are more stable than
sodium oxide, but they are poorly coupled with the base, these are, for example,
niobium, vanadium, uranium, beryllium, etc.
301
As a rule, breakdown of the oxide films results in the disappearance of the
potential barrier, and thus, in an increase in the rate of corrosion. It is shown in
[1] that the passive film on pure chromium and on the surface of austenitic stain-
less steels is retained in sodium up to 500-540C, and it breaks up rapidly when
the temperature is made higher. It is thus obvious that corrosion of stainless
chromium and chromium-nickel steels in sodium up to 500C depends to a given
extent on the passive film, while on further increase in the temperature the steel
displays its true corrosion resistance, not screened by films which act as
potential barriers.
Many investigators point to the fact that at temperatures above 500C the
corrosion rate of construction materials in sodium (dissolution, frontal and in-
tercrystalline corrosion) is determined primarily by the oxygen content of the so-
dium [1-5]. This is proven experimentally. For example, Bruch [5] has shown /302
that in sodium flowing at 0.4 m/sec at 450C the rate of dissolution of iron in-
creases 16 fold when the oxide content is increased from 0.01 to 0. 04% by weight.
Hempel [6] presents data on the corrosion rate of molybdenum in sodium contain-
ing 0.01 and 0.05% of oxygen by weight at 705C. In the first case the corrosion
rate is equal to a weight loss by the specimens of the order of 0.00014 milligrams/
/cm2 -hour, while in the second case it is 0.0111 milligrams/cm 2-hour.
It is shown in [5] that the presence of oxygen in sodium reduces the temper-
ature at which intercrystalline corrosion starts in stainless steel. Thus brand
304 steel was subjected-to initial intercrystalline corrosion to a depth of one-two
grains when held for 2500 hours at 300C in sodium with an 0.8% oxygen content by
weight. Precisely the same corrosion effect was observed in specimens of this
steel brand when tested in sodium with 0.003% of oxygen by weight at 710C for
3000 hours.
The above facts, which could be continued, show that, when the oxygen con-
tent of sodium is increased, the structural materials are subjected to corrosion
failure at a high rate due to the additional development of some chemical process.
Published data are available on this problem. Horsley [7], referring to [8]
has shown experimentally that when heating sodium containing 5 and 10% of oxygen
by weight in iron crucibles at 800C for seven days ferrites with compositions
Na2 O- FeO and (Na 2 0)2 FeO are formed on the wall (this was shown by x-ray
analysis). When using sodium with a lower oxygen content (0. 005, 0. 01 and 0.16%
by weight) no ferrite containing films were discovered on the crucible walls. Nev-
ertheless, Horsley has suggested that the intensifying effect of oxygen on metal
corrosion in sodium consists in the formation of chemical compounds such as fer-
rites, which are highly soluble in sodium.
Harbourn [9], and Weeks and Klamut [10] in examining the problem of mass
transfer came to the conclusion that oxidation of metal (for example, of iron)
Fe' Na2O > 2Na + FeO; (1)
302
may take place in the hotter part of the circulation system. It is assumed that in
the cooler part of a nonisothermal circulating system the ferric oxide is reduced
by the sodium, yielding the metal
FeO ' 2Na > Fe , Na 2O ' (2)
Hence, the transported masses consist of sintered highly-dispersed iron powder.

To clarify the detail the mechanism of the accelerating effect of oxygen on
corrosion of structural materials in sodium it is interesting to examine the chem-
ical nature of the oxygen dissolved in the sodium. Analyzing elementary physio-
chemical processes of corrosion it was decided on theoretical considerations that
partial dissociation of sodium oxide in the liquid sodium into ions is possible and /303
a suggestion was made that the active form which accelerates the corrosion of met-
als in sodium are the negative oxygen ions. To check this assumption experiment-
ally, direct current was passed through oxygen-containing liquid sodium. This
method was used previously for a number of important studies for determining the
nature of,hydrogen, carbon and sodium dissolved in liquid copper and steel [11), as
well as of carbon in hard steels [12-15).
U-shaped vessels with two hollow spheres at the ends, into which molybdenum
electrodes were soldered, were made from Pyrex. The vessels were filled under
vacuum with sodium, containing from 0.03 to 0, 07% of oxygen by weight. The filled
and soldered vessels were held in a furnace at 300 t 5C and simultaneously with
this 2 amps of direct current was passed through them for 1-6 hours. After termi-
nation of the experiments the sodium (1.7 grams) held in the anode and cathode spheres
was analyzed chemically for oxygen. Results of eight experiments are shown in
Table 1, from which it can be seen that in all cases the oxygen was transported
TABLE 1. Results of Chemical Analysis for Oxygen and Certain

Theoretical Data on Electron Transfer of Oxygen in Sodium
at 300C (With a Current of 2 amps).
Be fore After experiment Bywhatfac- ,Quantity
experiment _ for has the of oxy-
me
at Test Anode Cathode oxygen con- gen.trans
No. dura- heres spheres
p tent in the ported to
No, on 102% milli- _anode sphere anode pe''
p-
hours by grams 2 increased as
102% Milli- 10 % Milli- compared hour
wt Milli
grams grams with cathode
wt grams
1 1 2,4 0,4 3,2 0.55 1,6 0,27 2 0,07

2 2 4.7 0.8 6.3 1,07 3,1 0.53 2 0.07
3 2 3,85 0.65 6.1 1.03 1,6 0,27 3,8 0.09
4 6 5,65 0.96 10.5 1.78 0,86 0,14 12 0,07
5 4 6.4 1.08 9.3 1,58 3.5 0,60 2,7 0.06
6 3 3,75 0,64 6,5 1.1 1 0.17 6.5 0.07
7 4 4.35 0.74 7,6 1,3 1.1 0.18 7 0.07
8 5 7 1,19 9.8 1,7 4.2 0.72 2.3 0,05
303
only to the anode vessels (to the positive pole); here the amount of oxygen trans-
ported per 1 amp-hour was practically constant. Thus the experimental results
show that, at the low oxygen solubilities in sodium at 300C, negatively-charged
oxygen ions travel to the anode.
It stands to reason that at a higher temperature not only the solubility [16]
but the electrical transfer of oxygen ions in the sodium should increase. Exper-
iments were performed at 600C in steel tubes 12 mm in diameter, with a wall /304
thickness of 0, 4 mm and length of 180 mm for 2 and 4 hours. At the middle of
each tube was placed a steel tube segment, which served as an oxygen source.
-3
The tubes were filled under vacuum with sodium containing 6-8 . 10 % of oxygen
by weight, welded up and held in a furnace while direct current at the rate of 10
amps was passed through them [17, 18]. After termination of the experiment
the tubes were frozen and their anode and cathode ends were cut off under vacu-
um. Without removing them from the vacuum chamber, the cut off ends of the
tubes were analyzed chemically for the oxygen content (Table 2). A control test
without passing of current showed that at 600C the inner glass tube interacted
with the sodium and enriched it with oxygen from 0.008% by weight (starting con-
tent) to 0.10-0.11% by weight (by a factor of 13. 75 during 4 hours).
TABLE 2. Results of Chemical Analysis for Oxygen and

Some Theoretical Data on Electron Transfer
of Oxygen in Sodium at 600C,
Oxygen content in Oxygen transported
the ends of tubes, to the anode per 1
by weight amp-hour
Test number Conditions --
anode cathode ( milli-
(^) I (_) % by wt. I grams
I
600 C, 10 a, 2 h 0,32 0,0015 0.0157 1,943

1 (four tubes) 6000 C, 10 a, 4 h 0.57 0.0037 I 0,09-16 2,671
2 (four tubes) 600 C, without cur- 0.11 0,1
3 (two tubes) rent, 4 hours
(control test)
Experiments in which current was passed show that appreciable amounts of

oxygen are carried only to the positive poles of tubes. In experiment No, 1,1.943
milligrams were passed per 1 amp-hour, while in experiment No. 2, 2.674 milli-
grams of oxygen were carried over. This difference can be attributed to the time
difference and to the fact that sodium oxide was supplied by the glass tubes into
sodium not yet saturated with oxygen. The experiments'at 300-600C have shown
that transport of negatively charged oxygen ions is observed to the positive pole;
here 30 times more oxygen is carried at 600C, This leads to the conclusion that
304
the oxygen dissolved in the sodium is in the ion state. But since molarization
takes place simultaneously with the dissociation, we must use the balance equa-
tion
1a202^;a+ 0- 2 , (3)
Apparently, sodium oxide exists in sodium in the stable form as colloidal

and other particles. Most probably the molecular solution of sodium oxide ex-
ists in negligible concentrations, disappears entirely with an increase in temper-
ature due to displacement of the equilibrium to dissociation. The dissociation is
caused by the fact that the neutral atoms of sodium (the solvent) are capable of /305
lightly deforming the outside shell with the valence electron when interacting with
the ion lattice of the sodium oxide. As a result of the formation of induced dipoles,
the sodium atoms solvate the ions of the sodium oxide and thus weaken the forces
of attraction between them. Final dissociation takes place by means of the ther-
mal energy of the medium [19]. On these assumptions the oxygen ion should have
the form of the solvated complex
[ i\a7,, d - 0] - 21
where m is the number of atoms of sodium solvates with induced dipoles.

Postulation of the existence of oxygen in ionic form in liquid sodium makes
possible a new approach in clarifying the corrosion mechanism. For example, the
accelerating effect of oxygen on the process of dissolution of steels in liquid so-'
dium can be attributed to the fact that reactions with the participation of most chem-
ically active ionic form of oxygen (for dissolution of iron) take place in the diffu-
sion layer at the interface
Fe -IL 0 - 2 > FeO + 2e-, (4)
2-L,:a+d-2e-> 2\a o . (5)
These reactions should have high rates. Quantitatively they are limited only by
the rate of supply of oxygen ions. Each act of formation of ferric oxide disturbs
the solubility equilibrium at the given temperature and a corresponding number
of iron ions is transferred from the lattice to the diffusion boundary layer., Sub-
sequently the ferric oxide molecules move away from the boundary layer by dif-
fusion (stage which controls the total dissolution rate) and, by virtue of changes
in the concentration, are reduced by sodium according to the reaction
Fe0 d- 2 \a > Fe + Na 2O. (6)
The above mechanism for the corrosion of iron with participation of oxygen
ions is somewhat similar to processes observed on homogeneous catalysis with
305
formation of intermediate chemical compounds. In our case ferric oxide serves
as such a compound.
Corrosion Resistance of Ferritic-Pearlitic Chromium Steels

Austenitic chromium-nickel steels are sufficiently corrosion resistant in
sodium up to temperatures of 650-700C. But, due to their relatively low ther-
mal conductivity and high thermal expansion coefficient, these steels can crack
under alternating thermal loads. Hence, it is of great interest to examine the cor-
rosion resistance of ferritic-pearlitic nonnickel steels, the chemical composition /306
of which is shown in Table 3, The steels were tested in a sodium stand at temper-
ature of 500, 600 and 700C, in sodium flowing at 5 and 12 m/sec and containing
2-5% of oxygen by weight. Upon metallographic examination of the specimens af-
ter holding in the above sodium flows, no corrosion interaction zones were observ- /307
ed in the overwhelming majority of steels. Only in the EP-78, EP-79, 1Kh1MFS
and 12KhMF steels was there observed an elevated corrosion of grain boundaries
in the surface layer at the surface to a depth of 10-20 at 600C, But the quan-
tity of the pearlitic component in the majority of the tested steels decreased ap-
preciably (as compared with the starting composition) as early as at 500C, while
steels such as EP-78, EP-79, 1Kh2M, 12KhMF and 12Kh2MFSR practically com-
pletely lost their carbon at 600C, which is also substantiated by chemical analysis
(Table 4), Only the EI-802 steel retained its starting structure when held for long /308
time periods in sodium at 700C. The carbon content of the EI-531 and EI-802 steels,
as shown by chemical analysis, practically did not change. Changes in the weight
of specimens of all the pearlitic-class steels tested per unit time were practically
insignificant and varied within the limits of 0-0. 0003 milligrams/cm -hour, de- 2
pending on the test temperature.
TABLE 3. Chemical Composition of Pearlitic and

Chromium Steels
Element content, % by weight

Steel
brand I
Sf I r Mo I S I \U I NV
0.11 0.20 1.03 0.25 0,20

EP 79W 0.16 O.E3 1,10 0,36 0.22 0,33
EP-78 0.13 0.52 1.16 0.50 0.20
iKh1M1FS 11.15 0.52 1.20 1.03 0.26 --
12Kh2MFSR 0.12, 0,53 1.76 0,56 0,24
1Kh2M2F 0,14 0.95 2.19 1.54 0.32
1Kh2M 0.15 0,0S 2,25 0.99
EI-531 0.17 0,57 2,51 0.65 0.31 0.64
1Kh5MS 0,14 1,54 6,32 0,70
EI-802 0.14 0.10 11,6 0.67 0.42 0,30 0,30
EI-993 0,20 0.35 12.4 0.52 0.31 0,52 0.59
EP-38 0.17 1.33 13.6 0.78 0,41 0.37
I
306
TABLE 4, Decarbonization of ' Pearlitic and Chromium Steels
in a Flow of Sodium (Flow Velocity of 5 m/sec)
Carbon content, o by weight
I After holding in sodium at
Steel brand Testhours
time, the temperature, *C
Starting
500 1 000 1 700
12KhW 2000 0.11 0,06 0,03

EP-79 2000 0,16 0.1 0.09
EP-78 2000 0.13 0,06 0.05
iKh1MiFS 2000 0,15 0,06 0,09
12Kh2MFSR 2000 0,14 0,08 0,02
1Kh2M2F 2000 0.14 0,1 0,1
1Kh2M 1800 0.15 0.1 0.06
EI-531 1800 0,17 0,17 0,14
1 KhSMS 2(YJO 0,14 0.14
EI-802 6500 0,14 0,13 0.13
EI-993 6500 0.2 0,14
EP-38 65-00 0,17 0,05
The mechanical properties of steels tested in sodium (Table 5) differ no-

ticeably from the properties of the same steels held at the same temperature in
a container with argon ("without a medium'9. The mechanical properties of EI-
531 and EI-802 steels practically did not change when they were tested in a sodium
flow at temperatures of 600 and 700C, The above shows that changes in the struc-
ture of ferritic-pearlitic chromium steels and in their mechanical properties on
holding in a sodium flow takes place as a result of decarburization of these steels,
the rate of which at a given temperature depends on their alloying-element con-
tent (chromium, niobium, canadium and tungsten), which binds the steels carbon
into stable carbides. The higher the quantity of carbide-forming elements, the
stronger is the carbon bound in complex carbide compounds and the more stable
all the properties of steel subjected to liquid sodium.
The carbon content of the steel should not exceed the quantity needed for
formation of stable carbides with the carbide-forming elements added to the steel.
In the opposite case the steel gives up the excess carbon which will be carried by
the sodium flow to other sections of the apparatus where, as a result of this, part
and components made from austenitic stainless steels may become carbonized,
which may result in embrittlement and failure. Thus, the EI-993 steel contains
in the initial state somewhat more carbon than the EI-802 steel (0, 20 and 0,14%,
respectively), while the content of other elements is approximately the same.
When tested at 700C in the flow of sodium for 6500 hours the EI-993 steel lost
the entire excess carbon, and in the final count it only retained as much of it as
contained in the EI-802 brand, If, however, excess carbide-forming elements
are added to the steel, this can impair the workability of the steel (ease of roll-
ing, welding, etc.). Experiments show that high-chromium pearlitic steels, for
example, EI-802, M13 and others also tend to carbonization in the flow of so-
dium if the latter is contaminated by hydrocarbons to some degree. Carbonization
307
takes place to a great depth and the plasticity of steel then is highly reduced,
which can result in failure of components.
TABLE 5. Mechanical Properties of Ferritic-Pearlitic Steels

Tested in a Sodium Flow (Velocity of 5 m/sec)
fter holding with- After testing in
Temperature i w Time, out a medium sodium
Steel brand I C I hours o ue
k5d mm
12KhMF 5u0 2000 I 51 21 38 29

600 2000 51 25 29 39
EP-79 500 2000 59 27 51 27
600 2000 59 25 43 36
EP-78 500 2000 60 21 45 29
600 2000 52 26 39 36
1Xh1M1FS 500 2000 67 16 52 i8
600 4000 55 17 40 35
12Kh2MMFSR 500 :3600 60 25 52 21
600 3600 55 29 37 33
1Kh2M2F 500 2000 67 16 63 15
600 4000 65 14 53 26
1'Kh2M 500 1800 44 * 30 " 41 31
600 1400 44 * 30 * 34 38
EI-531 500 1800 46 * 27 * 46 27
600 1400 46 * 27 * 44 28
1KhSMS 500 3600 82 20 78 19
600 3600 74 19 66 22
EI-802 500 2000 51 25 48 24
600 2000 52 211 46 25
700 4000 49 22 45 22
EI-993 500 3600 9l 17 93 13
600 3600 87 17 77 17
700 2500 57 22 59 21
700 6500 57 22 57 24
EP-38 500 3600 91 17 93 13
600 3600 90 18 77 18
700 2500 68 12 68 13
700 6500 68 12 67 12
NOTE: The asterisk denotes the mechanical properties of steels

in the initial state, i. e. , after heat treatment; vu is the ultimate
strength of steel and S is the relative elongation.
It can be claimed on the basis of the above data that corrosion processes
in ferritic-pearlitic steels held in sodium are practically independent of the flow
velocity in the velocity range at hand (5-12 m/sec),
No erosion-type failures were observed in steel specimens tested at flow
velocities of 12 m/sec.
308
Corrosion Resistence of Austenitic Stainless Steels
Austentic stainless steels are capable for prolonged operation in sodium at /309
temperatures up to 650-700C; they are sensitive to the oxygen content of the so-
-3
dium. At oxygen contents of up to 5 . 10 % by weight [these] steels are not sub-
ject to appreciable corrosion failures. However, at higher concentration corro-
sion is observed with predominance of intergranular corrosion (Table 6),
TABLE 6 Chemical Composition of Stainless Austenitic

Steels Which Were Tested
Alloying Element, % by weight
Steel
brand C ( Si cr Ni I N.1) d +. Ti I w
I I I
1Kh1ONIOT 0.06 0.43 17.5 10.3 --- - 0.45 -

EI-448 0,08 0.51 16.5 12,5 - 2,01 0.52 -
EI-211 0.06 2,6 19,1 14 - -- - -
EI-888 0.07 3.9 13.3 14 0,6 - - -
EI-847 1 0.07 0,26 15,6 15 0.7 3,04 - -
EI-844 0.02 0. 116 ! 5. 1 15.1 - 3.05 - -
EI-854 0.09 0..',5 14 15.7 1 - - -
EI-850 0.02 4.36 11.8 16 0.53 - - -
EI-403 0.08 0.88 17.9 16.3 1.22 1.22 - -
EI-851 0,i1 0,27 13.8 19.1 1 - - 2.34
EP-71 0.03 0.26 21,6 21.5 - - 1.2 1.29
EP-148 0.1 0.27 21,5 24.t 1,07 2,62 - 1.4
Austenitic stainless steels can become highly carburized when the sodium
is contaminated by carbon-containing compounds or when the liquid metal is si-
multaneously in contact with carbon or pearlitic steels. Here their plasticity is
sharply reduced. Consequently, use should be made of proper combinations of
materials in order to avoid [excessive] oxygen content in the system and contam-
ination of the liquid sodium by hydrocarbons.
In our experiments we have discovered transport of nitrogen from the gas
atmosphere to the surface of the austenitic steel. When nitrogen was used as a
protective gas for the sodium, we have observed nitriding of the ducts of our set-
up which were made from 1Kh18N10T steel. The setup operated with an isother-
mal flow of sodium at a temperature of 450'C for one year. Metallographic ex- /310
amination of the inner surface of ducts showed a narrow (10 micromicrons) light
strip with a high hardness. Structural analysis by x-rays showed presence of
chromium nitrides on the surface. A similar light band was also observed on
specimens from 1Kh18N10T steel, tested on the stand in isothermal flow of so-
dium at 400'C during 1800 hours, A two-phase structure is observed from the
surface to a depth of 5 micromicrons consisting of nitrogen-saturated ferrite,
chromium nitrides and austenite. The specimens increased insignificantly in
weight due to the nitriding,
309
Here the ultimate strength of the steel did not change, but the plasticity re-
duced appreciably. While in the initial state the relative elongation of the steel
was 44%, then after testing in sodium flow it dropped to 34%. The above examples
show that it is not permissible to use nitrogen as a protective gas for sodium,
since this can result in embrittlement of the elements of apparatus due to nitrid-
ing. It may be assumed that at temperatures higher than 400 and 450C the
1Kh18N10T steel would have been nitrided to a greater extent.
Chronium-nickel stainless steels in the form of tubular or flat tension spec-
imens were tested at temperatures of 600 and 700C in sodium flowing at 5 and 12
-3
m/sec with an oxygen content of 3 10-3 - 7 . 10 % by weight. It can be seen from
Table 7 that the weight of all the austenitic steels was reduced in the tests. As
was shown experimentally, the tendency of these steels to mass transfer becomes
apparent when corrosion tests are made in a flow of liquid sodium starting with
600C. When testing at this temperature the specimens changed their weight very
little (0. 0008 milligrams/cm 2 -hour), while in tests at 700C the rate of weightloss
increased to about 0.0030 milligrams/cm 2 -hour (experiment time 4000 hours). The
reduction in the weight of specimens serves as grounds for the conclusion that aus-
tenitic steels have a greater tendency to the transfer of mass than ferritic-pearl-
itic steels, which is due to their nickel content which has an appreciably higher
solubility in sodium than iron and chromium.
Metallographic analysis of austenitic steels shows corrosion interaction of
intercrystalline character (high corrosion of grain boundaries). At 600C it was
observed primarily only in those specimens of these steels which were tested in
sodium containing 7.10 -3% of oxygen by weight, and it is absent at oxygen concen-
-3
trations of 3-4 . 10 % by weight. In the case of the higher oxygen content, the
depth of intercrystalline interaction of EI-211, 1Kh18N10T and EI-448 steels at
600C is as high as 70-150 micromicrons during 5100 hours. Specimens from
1Kh18N10T steel tested in sodium at 700C for 6500 hours had an appreciably
smaller depth of corrosion (20-30 micromicrons) due to the lower oxygen con-
tent of the sodium, while no corrosion interaction zones were observed in spec-
imens from EL-211 steel. The test results show that the corrosion resistance
of austenitic steels depends to a large extent on the oxygen content of the sodium.
All the tested steels of this class are sufficiently corrosion resistant materials
in the flow of liquid sodium at temperatures up to 700C, if the oxygen content
-3
does not exceed 5 4 10 % by weight.
. The mechanical properties of austenitic chromium-nickel steels tested at /311
600C in sodium differed little from the properties of the starting material (after
quench hardening) and from the properties of the same steels held at the same
temperature in an inert-gas atmosphere.
At 700C practically all the austenitic steels exhibit aging tendencies to one
or another degree, which produced some increase in the ultimate strength (6u)
and a reduction in the relative elongation (6). Here the mechanical properties of
specimens tested in sodium are on the level of properties of specimens held in /312
argon. This means that processes which bring about the changes in mechanical
310
TABLE 7 Results of Corrosion Testing and Changes in
Mechanical Properties of Austenitic Chromium-Nickel
Steels in a Flow of Sodium (Velocity 5 m/sec)
i Maxi- Oyge, I After holding After testing
U mum , content wed in sodium
Steel Time, deth
p
brand hours a
a sodium, u t,, ^;, 6u
2I
ôsion 9/6 'by wt ' g/mm 2 g^
E~ micro ns
1 Kh18N10T 600 1500 50 7 69 .31 70 26

600 5100 100 7 69 .,I 75 25
700 2500 20--30 4 66 19 67 27
700 11500 20-30 4 66 19 70 26
EI-448 600 1500 80 7 69 20 75 13
600 5100 150 7 69 20 75 15
EI-211 600 1500 30 7 58 30 67 27
600 5100 70 7 58 30 64 24
700 2500 0 4 46 4 59 5
700 6500 0 4 46 4 58 5
EI-888 700 2500 0 4 73 25 68 30
700 6500 20--30 4 73 25 70 30
EI-847 600 2000 0 3-4 62 45 56 43
600 4000 0 3-4
700 4000 20 4 69 32 68 32
EI-844 600 2500 0 5 55 28 56 36
600 5500 0 5 54 32 58 24
EI-854 600 2.500 40 5 57 24 56 27
600 5500 60 5 55 26 56 17
EI-850 600 2500 50 5 76 28 74 30
600 5500 80 5 66 16 70 17
EI-403 600 2000 0 3-4 69 39 61 38
700 4000 30 4-5 72 29 73 28
EI-851 600 2500 50 5 57 27 58 25
600 5500 50 5 53 15 58 21
EP-71 600 4000 3-4 94 23 93 23
700 4000 15 4-5 79 18 82 23
EP-148 600 4000 3-4 71 28 70 31
700 4000 15 4-5 79 20 79 20
properties are not related to the corrosive action of the sodium. The plastic
properties drop most appreciably in the EI-211 steel, due to intensive separa-
tion of the secondary phase at 700C not only along grain boundaries but also in
the grain proper. In the 1Kh18N10T, EI-847 and EI-888 steels the separation
of new phase components takes place primarily along grain boundaries.
X-ray diffraction studies of steel tested in sodium as well as held in argon,
show that the composition of phases precipitated in the process of aging is differ-
ent for each steel brand and is related to the alloying components, For EI-847
steel which has a high molybdenum content, the precipitate was the Fe 3 Mo inter-
metallic compound (Laves 'phase). The EI-888 steel contains about 4% of silicon
311
and hence, a substantial amount of the complex intermetallic compound Ni 2_(Cr, Fe) Si
is precipitated in this steel in the process of aging.
The amount of TiC in the structure of the 1Kh18N10T steel increased and,
in addition, a noticeable quantity of the sigma phase appeared. Thus, examina-
tion of the mechanical properties, microstructure and data of x-ray diffraction
analysis show that all the austenitic steels are structurally unstable to some ex-
tent at 700C, but the majority of them retain satisfactory mechanical properties
even after exposure of 6500 hours.
Intercrystalline corrosion interac-
tion with sodium reduces the plasticity
2h of austenitic steels. Figure 1 shows the
change in the relative elongation of the
EI-211 steel as a function of the testing
time at 600C and of the oxygen content
7CV0 2000 3000 4000 5060
of the sodium. Despite some scatter
Time, hrs. of the experimental points it can be seen
that the plasticity of austenitic steels de-
Figure 1. Relative Elongation of EI- pends appreciably on the oxygen content
211 Steel Tested in Flow of Sodium at of the sodium.
600C as a Function of the Test Time
and of the Oxygen Content of the Sodium. No effect of the rate of the liquid /313
-3 sodium (12 m/sec) on the corrosion or
1) at 5 . 10 % by Weight;
-3 erosion of stainless chromium-nickel
2) at 4 10 % by Weight. steels was observed.
Corrosion Resistance of Heat-Resistant Alloys With a Nickel Base

Work with liquid sodium at high temperatures requires materials with high
heat resistance. Nickel-base alloys such as Nimonic are heat resistant up to tem-
peratures of 850-900C. We have tested the following nickel-base alloys in the
flow of sodium: EI-437b. EI-559A and K120N60M5V10Yu2T2 and, for comparison,
we have also tested the EI-211 stainless chromium-nickel steel containing about
14% of nickel by weight (Table 8). The corrosion tests were performed at sodium
temperatures and velocities of 700-930C and 2 m/sec. The oxygen content of the
-3
sodium was maintained at 6 . 10 % by weight or less by a cold trap. The testing
time at each temperature was 300 hours.
The results of these experiments are shown in Fig. 2. The experimental
data point to a sharp increase in the rate of mass transfer for Nimonic-type al-
loys at temperatures above 800C. The rate A of mass transfer in the tempera-
ture range of the test can be expressed by the formula
_Q (7)
K=Koe IIT'
312
111? '5J d111? I sJ
1, off, where K is the rate of mass
transfer, milligrams/cm 2 -
hour; K0 is a coefficient, con-
N stant for a given material, which
takes into account the effect of /314
bo various factors on the mass trans-
fer, milligrams/cm.2 -hour, R is
the universal gas constant, which
is 1.987 cal/gram-atom-degree,
T is the temperature, K, and Q
0.90 0,85 0.90 0,95 1,00
TxJO^deg-1 is the heat of dissolution, cal/
/gram-atom (in the graph this
is the tangent of the slope of line
Figure 2. The Rate of Mass Transfer (K) as in K to axis 1/T).
a Function of the Temperature (T, K).
1) The ET-437bAlloy; 2)The Kh20N60M5V10- The K0 and Q calculated
Yu2T2Alloy; 3) The EI-559AAlloy; 4) EI-211 on the basis of experimental data
Steel. are shown in Table 9. The mag-
nitudes of Q are very close for
all the materials tested and ap-
parently are characteristic of
lattices with gamma structures, since we have here the dissolution of nickel-con-
taining materials. The products of mass transfer after the EI-437b alloy has been
held in a flow of sodium at 880C for 190 hours contained 21.5% by weight of chrom-
ium, 72. 0% by weight of nickel and 0.93% by weight of titanium. Initially the EI-
437b alloy was composed of 77% by weight of nickel, while after the test the nickel
content dropped to 59% by weight. Consequently, as they became dissolved from
the surface of the alloy, the nickel atoms, by diffusion from the inner volume of
alloy (which is shown by layer-by-layer chemical analysis) again were supplied /315
to the surface, where they were dissolved by the sodium, etc. This mechanism
of leaching out of nickel from alloys may continue for a long time with attendant
constant reduction in the nickel content. The calculated heat of dissolution Q is
an average for the individual components of the alloy.
All the tested materials can be arranged by their readiness to dissolution
and to mass transfer in the following order: EI-211 steel, the EI-559A, Kh20N-
60M5V10Yu2T2 and EI-437B alloys. It can be seen from this that the rate of mass
transfer is directly proportional to the nickel content. Addition of aluminum to
nickel-base alloys reduced the mass transfer rate. As is shown experimentally,
the EI-559 alloy is more prone to mass transfer than the EI-559A alloy, although
the nickel content in the former is even somewhat lower than in the latter; how-
ever, the aluminum content of the EI-559A alloy is higher, due to which it is less
subjected to mass transfer in sodium than the EI-559. The EI-827 Nimonic-type
alloy with a high nickel content (74% by weight) exibited very high corrosion re-
sistance to the transfer of mass in sodium also due to its elevated aluminum con-
tent. Aluminum binds the nickel into the stable Ni3Al compound and thus reduces
its diffusion mobility and consequently also the tendency to dissolution in sodium.
The mass transfer rate is not only a function of the temperature but also of
the test time. Figure 3 shows that K, the mass transfer rate increases with an
313

TABLE 8. Chemical Composition of Nickel-Base Heat-

Resistant Alloys and of the EI-211 Steel
Content of alloying elements, o by weight
Material
Cr I NI I Uo I Al Fe I Ti I W
EI-437b 0.05 0.38 20.6 75 - 0.35 0.43 2,54 --

Kh20N60MSVlOYu2T2 0.1 --- 20 60 5.0 2.3 - 1.3 10
EI-559 0.1 0.5 18.2 55.9 - 2.3 23 - -
EI-559A 0.02 0.5 16.6 57.8 - 3.3 21 - -
EI-827 0.07 0.4 10 74 6.0 4.3 - 5.0
EI-211 0.16 2.4 20.6 13.8 - -
NOTE: The EI-437b, Kh20N60M5V10Yu2T2 and EI-827

alloys contain about 0.01% by weight of boron.
1) Material; 2) Content of Alloying Elements, % by W eight
3) EI; 4) KHZON6OM5VIOYU212
TABLE 9. Magnitudes of K 0 and Q for the Materials Tested

Material KO X10-5 mg/(cm2 hr)I Q, cal/gram-atom
EI-4376 70.52 39 900

Kh20N60M5VIOYu2T2 80.05 40 000
EI-559A 52.65 40 800
EI-211 9.93 1 41 100
1) Material; 2) Milligrams/cm 2 - hour; 3) Q, cal/gram atom; 4) EI-

4376; 5) KH2ON6OM5VIOYU2; 6) EI-559A; 7) EI-211
increase in the holding time, i.e., the longer the test the higher the loss of weight
per unit time. And only for the EI-211 steel is the rate of mass transfer constant
at 930C. The experiments have shown that mass transfer is reduced appreciably
with improved finish of the surfaces washed by the liquid metal. Nickel has a high-
er solubility in sodium than other components of structural' materials. Due to the
difference in the dissolution rates of the alloy base and of its alloying components,
the surface in contact with the sodium obviously increases, due to the appearance
of rough spots and hence, the rate of mass transfer also increases. Since an in-
crease in the surface results in increasing the mass transfer rate, then the work-
ing surfaces of components and elements in contact with liquid sodium should have
a high finish.
All the experimentally studied nickel-base alloys are, to one degree or another,
prone to mass transfer. The dissolved components are carried by the sodium from
the hot to the cooler zone, where they are saturated and precipitate. This process
is highly dangerous, since the transfer products which form are capable of accum- /316
ulating in the cold zone of the device and to plug up the liquid-metal passages. The
transported masses start to precipitate on the walls of the stand's ducts when the
314

liquid metal's temperature is reduced by as lit-

tle as 20-30C and move along the duct surfaces
N up to temperature zones of 600 C (for a test tem-
perature of 630C). Numerous experiments show
that thickest layers of the transfer products form
5
at points of reduction in duct cross sections (or
in slots), where the rate of flow of the liquid met-
al is increased appreciably. Residues of the trans-
ferred mass are sintered to metal walls at tem-
i00 200 900 V, hours peratures of 600C and above and are held there
very strongly, while at temperatures below 600 C
Figure 3. The Rate of Mass they are quite loose and are poorly bound to the
Transfer (K) as a Function of base.
the Holding Time (T) at 9300
for the Following Nickel-Base
Alloys. Corrosion Resistance of Refractory Metals
1) Kh20N60M5V10Yu2T2;
2) EI-559A; 3) EI-211 steel, Refractory metals such as molybdenum,
niobium, etc. , can be used as structural ma-
terials with liquid sodium. Corrosion studies
of molybdenum and niobium specimens, tested
in a flow of sodium at 5 m/sec and with an ox-
ygen content of 4 . 10-3% by weight during 3600 hours at 600 and 700C show that
molybdenum has satisfactory corrosion resistance under these conditions, while
the niobium specimens failed practically completely as early as at 600C. The
molybdenum had an insignificant weight gain. Similar results were obtained when
testing in a sodium flow at 930C. The molybdenum specimens increased insig-
nificantly in weight, while the niobium specimens, even if they did not fail during
the test (300 hours), exhibited a very high rate of weight loss (about 0.1 milligrams/
/cm 2 -hour) .
A large number of tests performed with molybdenum in liquid sodium heated
to up to 1050C have verified its high corrosion resistance. Results of studies
show that the corrosion resistance of molybdenum does not depend on the rate of
flow of the liquid metal in the velocity range under study (0.2-2.5 m/sec). When
-2
the oxygen content of the sodium was held to 2 10 % by weight, no embrittlement
of molybdenum specimens was observed and their mechanical properties at the end
of the test remained on the initial level, In all cases a slight increase in the weight
of molybdenum specimens was observed, which apparently is due to the formation,
at the surface, of an intermetallic molybdenum-nickel compound such as MoNi (the
ducts and other components of the stand were made from the EI-211 chromium- /317
nickel steel). Metallographic analysis did not show any corrosion zones on molyb-
denum specimens held in a flow of sodium at 600, 700 and 930C.
The corrosion resistance of niobium depends on the oxygen content of the
liquid metal (Fig. 4). The rate of corrosion of niobium in sodium flowing at 2
m/sec at 900'C as a function of the oxygen content can be described by the ex-
-2
pression (in the range of up to 4 . 10 % oxygen by weight)
K=172-10-5-c8' 3-x, (g)
315

where X is the oxygen content

of the sodium, % by weight.
^^
y 3 Prolonged (3600 hours) testing
N 2
{ of niobium even at 600C and
Û
with an oxygen content of
^ 8 4-10-3 % by weight results in
E
x ;
total failure of the specimens.
Absence of a corrosion
i
zone in niobium specimens
!O-d I7 I 1 (which lost up to 80% of their
n 0,01
y ? o
thickness) points to the fact
o 04
Oxygen content o f the niob iu m, that interaction with oxygen re-
Yo by weight. sults in the formation of very
Figure 4. The Corrosion Rate for Niobium loose, poorly bound niobiates,
as a Function of the Oxygen Content of which were practically com-
Sodium at 900C, pletely washed off by the flow
of the liquid metal. Only at the
surface of individual specimens
was there observed a corrosion
layer up to 15 micromicrons deep, which consisted of niobium oxides in very poor
bound with the base metal.
The strength of niobium specimens tested at 900C for 300 hours is reduced
sharply, while their relative elongation increases, which points to start of crys-
tallization. Metallographic tests also show that niobium undergoes partial re-
crystallization in the process of tests, which starts usually from the specimen
surface.
The rate of flow of the liquid sodium affects the corrosion resistance of nio-
bium. After testing at 900C in liquid sodium flowing with velocities of 0.2 (in
convection loops), 1. 0 and 2. 5 m/sec (in a circuit) the rates of weight loss by the /318
niobium (for the same oxygen content) were 0. 04, 0.06 and 0,10 milligrams/cm 2 -
hour, respectively. Obviously, when the flow rate of liquid sodium is increased,
a larger quantity of dissolved oxygen arrives at the surface of niobium specimens,
which is the factor responsible for the increased corrosion rate. In addition, the
increase in the sodium flow velocity should promote a more rapid removal of the
loose products of corrosion which form on the niobium surface.
Comparatively prolonged use of niobium in liquid sodium coolant requires
that the oxygen content of the coolant be held below 1 . 10-3 % by weight.
REFERENCES
1. Manly W. D.: Corrosion, 12, No. 7, 46-52 (1956),

2, Epstein, L.: In the book: Proceedings of the International Conference on
Peaceful Uses of Atomic Energy, Geneva, 1955 (US Publishers McGraw-
Hill for Pergamon Press, or vice versa).
316
3. Hofman, E. E. and Manly, W. D.: Problems in Nuclear Engineering, Lond. ,
1957.
4. Evans, J. W.: Nuclear Engineering, 4, No. 5, 35 (1957).
5. Bruch, E. G. : Nuclear Metallurgy, 4, No. 2, 20 (1956).
6. Hempel, C. A.: Corrosion, v. 14, No. 12 (1958).
7, Horsley, G. W.: J. Iron and Steel Inst. No. 1, 182 (1956).
8. Kelman, Z. R. , et al.: Resistance of Materials to Attack by Liquid Metals.
USA, 1950.
9. Harbourn, B.: Corros. Prevent, and Control., v. 5, No. 1 (1958).
10. Weeks, J. R. and Klamut, C. J.: Corrosion of Reactor Materials, v. 1,
p. 105, 1962, International Atomic Energy Agency, Vienna.
11. Rabkin, M. A.: Zhurnal prikladnoy khimii, XXX, 5, 791 (1957).
12. Prosvirnin, V.I.: Vestnik metallopromyshlennosti, 12 (1937).
13, Lebedev, T. A. and Guterman, V. M.: Trudy TsNIITMASh, H, 1948.
14. Romadin, K. L.: Trudy Voyenno-vozdushnoy inzhenernoy Akademii im.
Zhukovskogo. Issue 157, 1947.
15. Babikova, Yu. F. and Gruzin, P. L.: Fizika metallov i metallovedeniye
[Physics of Metals and Metals Science]. Issue 5, page 255. Metallurgi-
zdat, Sverdlovsk, 1957.
16. Zhidkometallicheskiye teplonosite1i (natriy) [Liquid-Metal Heat-Transfer
Agents (sodium)]. Collection of articles under the editorship of A. Ye.
Sheyndlin. Foreign Literature Publishing House, Moscow, 1959.
17, Nevzorov, B. A.: Zhurnal fizicheskoy khimii. Issue 3, XXXV (1961).
18. Nevzorov, B, A.: Nature of Oxygen in Liquid Sodium. In the transactions
of the Conference: Corrosion of Reactor Materials, Vol. 2. Internation-
al Atomic Energy Agency, 'Vienna, 1962.
19. Nekrasov, B. V.: Kurs obshchey khimii [A Course in General Chemistry].
Goskhimizdat, Moscow, 1954.
317
DETERMINATION OF THE OXYGEN CONTENT OF SODIUM AND OF THE
SODIUM-POTASSIUM ALLOY BY VACUUM DISTILLATION
I. B. Dmitryeva, F. A. Kozlov and E. K. Kuznetsov
Determination of the oxygen content of sodium and of sodium-potassium al- /318

loys is one of the more difficult problems of analysis. The method moÂ exten-
sively used is that of mercury extraction, which is based on separation of oxides /319
from metals by transforming it into an amalgam [1]; however, the sensitivity of
this method which is 0.01% of oxygen by weight does not satisfy the exacting
requirements which are at present being put to purification as well as to corntent-
control of heat transfer agents. The method for determining the oxygen cor, tent
of conventionally used liquid-metal coolants deserving most attention is that of
vacuum distillation, [2], the substance of which consists in separating the metal
from its oxides under a vacuum, which is possible due to the great difference 411
their vapor pressures.
Studies performed under laboratory conditions show that increasing the sen-
sitivity of the above method requires increasing the sample weight which is taken
for analysis, and reducing possible contamination of the sample in the process
of taking it and in preparation for analysis. In order to do this, various versions
of distilling samplers have been devised, which can be used for taking a sample
of the coolant (weighing 100 grams) directly from the circuit and to analyze it by
the vacuum distillation method.. The samples are taken rapidly, the time needed
for this being not more than a minute and a half, which eliminates the possibility
of redistribution of oxides during sampling. The fact that sampling and analysis
are made by the instrument means that the sample will not be contaminated when
it is prepared for analysis. For the sodium-potassium alloy this method of analy-
sis for oxygen is more expedient, since, due to the fact that this alloy is liquid at
room temperature, it is difficult to take a sample for analysis in a different vessel.
Designs of distilling samplers are shown in Fig. 1. The device shown in
Fig. lb can be used to take sequential samples at intervals from several minutes
to several hours with subsequent simultaneous distillation of all the four samples.
This device consists of three valves and a vacuum chamber. That partof itwhich
is shut off by the valves is the vessel from which the samples are taken. In order
that the metal should not get spilled when pouring the sample into the cup held in
the vacuum chamber, a special tube has been provided. It is necessary to care-
fully remove the alkaline metal from the vacuum chamber's surface, in order to
eliminate additional contamination when taking the following sample. To make
the valve more accessible for washing, the tube is removable. The specified dis-
tillation temperature is maintained by an adjustable-heat furnace, and is monitor-
ed by thermocouples. The vapors of liquid-metal coolants condense on the surface
of a baffle plate, which prevents the vapor from getting into the vacuum line, and
at the lower flange. The device is split into two parts so that the cup with the res-
idue can be removed after the distillation. The 'system is made airtight by a vacu-
um-rubber gasket.
318
3
^, ^ i li - .'i E
Et ^. ;^ J
a b
c
Figure 1. Designs of Distilling Samplers for Work With
Sodium (a and b) and With Sodium-Potassium Alloy (c).
1) Liquid-Metal Valves; 2) Drain Tube; 3) Vacuum Chamber
C asing; 4) Baffle Plate; 5) Line to Vacuum Pump; 6) Thrust
Pieces; 7) Packing; 8) Sampling Cup; 9) Furnace; 10) Ther-
mocouple; 11) Packing; 12) Rotary Baffle Plate; 13) Cup
Position Indicator; 14) Rotary Bar; 15) Movable Level Gage.
Keeping in mind the specifics of work with the sodium-potassium alloy, the
above device has been equipped with a rotary baffle plate, which is placed into the
cover of the vacuum chamber through the packing and with a funnel, which protects /320
the cup from metal which may fall into it from condensation surfaces.
Experimental method. Before the vacuum chamber is assembled, its inside
surfaces and the cup are throughly washed with distilled water and alcohol and are
319
dried. After assembly the chamber is washed with argon and evacuated to a res-
idual pressure of 1 . 10-2 mm of Hg. Before sampling the cup is degassed at 500-
550C. Simultaneously with this the rate of flow of the metal through the sampler
is adjusted to 1500 liters/hour.
The device operates under the same temperature mode before the metal is /321
pumped through the sampler as well as during this. The sodium was continuously
circulated through the cold trap. The cold trap temperature varied within not
more than 5% of that specified. The sodium was circulated through the distilling
sampler for not less than 30 minutes.
Before taking each sample the plugging up temperature was determined not
less than twice, using a cock-type oxide-level indicator. The samples were taken
only as long as the differences in the plugging up temperature did not exceed 10C;
when this happened the temperature of the still furnace was lowered, the circula-
tion was shut off and, when it was determined that there is no flow, the sample
was drained, after which the furnace was put on and the specified temperature es-
tablished. The vacuum pump operated continuously during the distillation; after
the latter was finished the furnace was shut off, the distillation device was cooled,
the system was filled with argon and the chamber was dismantled.
To determine the quantity of oxides in the distillation residue the inner sur-
face of the cup was washed with 100 milliliters of distilled water in 10 milliter
portions at room temperature. The solution thus obtained was titrated by sulfuric
acid in the presence of a mixed indicator and the oxygen content of the sodium
was determined. The sample weight was found by collecting the condensate and
weighing it; the error in weight due to loss of metal and additional oxidation of it
in the air did not exceed 5%, which was verified experimentally.
When four samples were distilled simultaneously, the weight of sodium col-
lected in each cup was determined. For this purpose the sodium temperature dur-
ing sampling was determined and, using a level gage (see Fig. lb) the quantity of
sodium in the cup was measured.
Experimental results. Selection of the distillation conditions is a decisive
factor in obtaining the required precision for a given method. These conditions
should ensure complete removal of the metal from the sample and not permit ap-
preciable vaporization of the oxides. On the assumption that the oxides, after
the metal is removed from the sample, are held under the equilibrium pressure
of dissociation and that the gas behaves as an ideal gas, we have estimated the
possible oxide losses from a nickel cup 40 mm in diameter during one hour. It
was found that in a 100 gram sample the oxide losses by evaporation reduce
the results by 10 -5 , 10 -4 and 7 . 10 -4 % by weight for temperatures of 500-600C,
respectively.
The distillation conditions were selected at temperatures of 500, 550 and
600C with a distillation time of 1 to 40 hours, the oxygen concentration varied
here from 0.0005 to 0.02% by weight (Table 1). 'When the samples are distilled
(at temperatures of 500 and 550C) for 3-6 hours, the reproducilibity of results
is satisfactory, which is proven experimentally. When this time was increased /322
(to 25-40 hours at temperatures of 500C and to about 10 hours at temperatures
320
TABLE 1. Results of Experiments for Selecting the
Distillation Conditions
:,ou c sô^ c euo^ c
Distil- Oxygen PIu u Distil- Oxygen Plug- Distil- Oxygen
gPIugn
gp
temp- lation content, g3 gp
temp- lation content, gingup
temp- lation content,
time, o/ b time b era C re, time a by
o^
era C re, hours
weight eraCre, hours weight hours' weight
218 3 ^2.3 2S6 5 5.7 285 1 4,7

2 9 3 1.9 25:3
I
5 5,3 234 3 3.1
1
' 25 3 1.9 2S6 7 4.6 284 5 2
12 .1 1.4 182 10 4.6 289 7 1
212 '4 1
^5 iU 1.'4 320 3.5 10
319 :3,5 8.9
25i 5 5.9 3.14 5 9.6
281 7 4.2 315 6 9
2S6 10 3.8 325 7.5 7,7
335 3.5 20 338 10 8.1
338 3.5 19
335 3.5 14
33 6.5 19
3. 5 7.5 16
335 10 11
33U 25 7.2
336 40 6.9
of 550C), the quality of results became poorer. This can apparently be attributed
to partial vaporization of the metal oxides, as well as interaction between the latter
and the cup's material and diffusion into the walls. When the distillation time was
25-40 hours the inner surface of the cup became covered by a blackish-green film,
which did not dissolve in water. At 600C the oxygen content of the sample became
smaller in proportion to the distillation time.
On the basis of the results obtained and of the theoretical data, we have se-
lected the following distillation conditions for subsequent tests: temperature 500-
550C, time 5 hours, residual pressure 1 . 10 -2 mm of Hg.
When distilling the alloy, the rotary baffle plate was covered for one hour
before the end of distillation, with the result that no alloy drops fell into the cup
when the sampler was dismantled.
When using the four-cup sampler, the samples taken first are stored for a
prolonged time in the vacuum chamber. To check for further oxygen contamination
of the metal [due to the storage time, DBL], we have made special experiments,
which show that metal held in a cup under a vacuum of 10 -2 mm of Hg for 9 hours,
did not become oxidized.
The duration of one analysis was 8 hours when using distilling samplers with /323
one cup and 2-3 hours when a four-cup device was used. The amount of oxygen
removed from the sampler by vacuum distillation was determined by comparing
321
the quantity of oxygen added to the sample to that found in the sample after distil-
lation. Here measures were taken to hold the natural oxygen content of the sam-
ple to a minimum. It was equal to the concentration which is provided by a cold
trap with an inside temperature of 110C (0.0003-0, 0005% by weight).
The oxygen introduced into the liquid metal was measured as follows: before
making the analysis a specified quantity of the alkaline solution of sodium hydro-
xide was placed in the nickel cup; this solution was made to evaporate at 100 C and
was finally dried at 150-200C, The cup with the residue was placed in the vacuum
chamber and a sample of the liquid metal was taken from it. In order that the hy-
droxide should interact with the metal and that the oxides should become distribut-
ed over the entire volume of the sample, the latter was held for 2-3 hours at 1500C
and only then was the distilling device operated under distillation conditions. These
conditions and the subsequent analysis were identical to those described above. Ad-
dition of oxygen to the sample by this method yields oxides and a hydride. The lat-
ter dissociates under distillation conditions and is removed from the sample.
The alkali hydroxides which are added to the sample should be prevented
from being lost or being transformed into carbonates, they also should be dehy-
drated. We have performed special tests which show that, under the vaporization
and drying conditions we have used, no hydroxide losses took place. On the av-
erage 7% of the hydroxide became carbonated when vaporizing in air.
When performing the 'oxygen put in - oxygen found" experiments the cup /324
was not degased, since this can result in loss of the hydroxide. Therefore, we
have performed experiments to determine the control-test value for a degased
-4
nickel cup, which, on the average, was 5 . 10 % of oxygenbyweight, when calculat-
ed for a 100 gram sample. The oxygen content of the sodium poured into the cup
was determined by a cock-type oxide indicator. On the average the percentage of
oxide loss from the sample calculated on the basis of the data of these experiments
was 14% (Table 2).
TABLE 2. Results of "Oxygen Added - Oxygen Found" Tests
Starting quan- Oxygen added, Total amount Quantity of

tity of oxygen, g of oxygen, oxygen found, Oxygen lost,
grams X 10-
gramsX 10-`1 grams X 10-3 grams X 10-3
5 39 39.5 33.7 15
6 39 39.6 34,7 12
4 39 39.4 32.0 19
6 7,8 8.4 7,0 17
6 7.8 8.4 7.0 17
6 1.56 2,26 2.0 11
3 1.56 1.86 1,4 21
4 1.58 1.98 1.6 19
8 0.84 1.64 1,6 2.0
4 0,78 1.18 1,1 7.0
7 0,78 1.48 1.4 5,0
322
To determine the control test value it is necessary to use the distilling sam-
pler to take and analyze a sodium sample with zero oxygen content. We could not
do this, and hence, the control-test value was determined as the quantity of oxides
which form on the surface of the cup when it is held in sodium vapor under the dis-
tillation conditions used by us.
The use of the four-cup distillating sampler (see Fig. lb) made it possible
to determine the control-test value for several cups from different metals simul-
taneously:
Nickel jx 10 -4% by weight) 3.7 0.69 4.7 6.6 3.5 2.2 3.8 1.8
-4
Molybdenum (x 10 % by weight) 1.3 1 0 0.68 1.2 1.8
The lowest control-test value is given by a molybdenum cup (1.0 f 0.4).10 -4% by
-4
weight, and a nickel cup (3.4 1.4) 10 % by weight, calculated for a sample of
100 grams.
Thus, the minimum oxygen concentration determined by the distilling sam-
-4 -4
pler is (2-5)-10 % by weight when using nickel cups and about 1 . 10 % by weight
when using molybdenum cups. The increase in sensitivity is limited by the con-
trol-test value.
The distillation method can be used to determine oxygen concentrations in
the range of 1. 10-4 - 5.10-2 To by weight.
REFERENCES
1. Pepkowitz, Z. P. and Judd, W. C.: Analyt. Chem. 22, No. 10, 1283 (1950).
2. White, I. C.: Nucl. Sci. Abstrs, 15, 8290 (1957).
323
THE COCK-TYPE INDICATOR AS AN INSTRUMENT FOR
DETECTING ADMIXTURES IN SODIUM
F. A. Kozlov, E. K. Kuznetsov and V. I. Subbotin
A large volume of work has been performed lately for developing a device /324
for continuous sampling and monitoring of admixtures in liquid sodium [1], and
[2]. These devices are at present undergoing experimental assimilation and have
not as yet come into extensive practical use. It is precisely for this reason that /325
the cock-type indicator described in [3-5] is practically the only one and the most
reliable instrument which is used extensively in liquid-metal facilities; it can be
used for rapid (30 minutes) and remote determination of the admixture content of
the coolant. The operating principle of this indicator is based on the ability of the
admixtures precipitating from a solution to plug up passages.
The instruments include a special device with a small passage, a heat ex-
changer for cooling the liquid metal, adjusting valve or a special pump for estab-
lishing the required rate of flow through the instrument, a flowmeter, and ther-
mocouple for determining the temperature of the liquid metal flowing through nar-
row holes. It is used to determine the oxygen content of sodium and of the sodium-
potassium alloy; it can also be used to determine other admixtures the solubility
of which changes with the temperature.
Dependence of Readings of the Cock-Type Oxide-Content

Indicator on the Main Parameters
Theoretical analysis. The cock-type indicator does not record the precip-
itation of the first crystals of the oxide, but the time when the resistance of the
device due to oxide precipitation increases to such an extent that the flow rate of
sodium through the device is reduced from the starting value Q 0 to some quantity
Q1 In general we can write
Qi = aQo,
where 0. 97 < a < 1.

The time needed for precipitating the required quantity of oxides (T st) de-
pends on the design of the instrument and conditions of experiment. The experi-
ment is performed with constant reduction of the temperature of the metal flow-
ing through the hole. When the flow rate drops to Q 1, the plugging up tempera-
ture (tst), which is lower than the saturation temperature (t s ) by At, is read.
Obviously, for a given rate of temperature reduction in the washer dt/dT = const
is tp1+ Tpl _ tpl+ At. (1)
d
324
Let us consider how At depends on the main parameters of the instrument
and experimental conditions. We shall assume that:
1) The pressure drop in the instrument during the.experiments remains
constant and is determined from the d'Arcy formula;
2) The entire resistance of the instrument is concentrated in the holes
where the sodium flow has the lowest temperature and where are precipitated
the oxides which increase the flow resistance of the instrument;
3) The instrument has n holes with diameter d 0 and length 1;
4) The flow of sodium through the holes is turbulent and the resistance co- /326
efficient is determined by the Blasius law, while the oxide flow per unit surface
per unit time is given by the expression [6]
5) The flow does not contain crystallization centers moving together with
the sodium.
To reduce the flow rate of metal through the instrument from Q 0 to Q1 , the
passage cross section must be reduced by some quantity S'. Using the d'Arcy
relationship and the Blasius law, it is easy to show that
S' = 0, 7S:^ do (1 a t ''1LJ. (3)
With consideration given to local resistances (narrowing and widening of flow)
S'_- 0.755 do L i VL a) , (4)
where ^ 1 and ^2 are the total resistances of the holes for flow rates Q 6 and Q1.
The flow resistance in the instrument is [actually] concentrated not only in

the holes, where the plugging up takes place, hence, S' shouldbe greater than cal-
culated by Eqs, (3) or (4). This can be taken into account by introducing the factor
co > 1, The larger the resistances other than those of the holes, the greater cp. It
is obvious that
t
a _
1P1
Ty\a .17^^20! (1'
IUdZ(jt
J e100.i1 1 r
325

or, making use of the fact that
and ud
t
=O (5)
we get
tpl
r^( \ 019
S' = :7K, Z dT, (6)
0 J
where
Yâ111,â
K, s^K
s(-t)0.9 7
During time -rp1 the flow rate through the instrument changes little, and hence, /327
we can write
zp1
S K,^J \ dr. (8)
0
In order to evaluate the integral in the above expression we must know the
manner in which the supersaturation and the size of the holes vary with time. The
change in time in the supersaturation is determined by the oxygen solubility in the
sodium as a factor of the temperature and of the rate of temperature drop in the
holes. This relationship is found to be cumbersome. But if it is assumed that
the sodium temperature does not go below the saturation temperature by more
than 50C, then in this temperature range the solubility of oxygen as a function
of the temperature can be written in the form
_ "' (9)
Cs dt (ts t)+
C
where dC s/dt for each oxygen concentration in the sodium has a different value.
Making use of the fact that the temperature reduction proceeds at a constant rate,
we get from Eq. (9) the dependence of supersaturation on the time
dcs (It
Z dt
di - T. (10)
326
The change in the hole size due to oxide deposition can be found from
1 ai.s
which was obtained from conditions of materials balance making use of Eqs. (2),
(5) and (7) Integration of (11) within the limits of d 0 - 0 and 0 - T, respectively,
yields
d (1, 2 _ 229 .K,Qo.s

It ^Zi .i2 ) 1/' s ' ( 12 )
Substituting the above expression and Eq. (10) into (8), integrating and transform-
ing using Eqs. (1) and (3) and the fact that Q = Q 0/2(1 + a), we get
lf0lNa2odu^FenPrll4 rlt {1[i^ ( ^ n i4 /ii ) l 1

1 t ti 8.7 - _
p lc
--- f
( 13
M Na 2 OYNaQo ^^ ^' dt (1 r-n)
In deriving the above no consideration was given to reduction in the supersatur-

ation due to deposition of oxides on the surfaces of the heat exchanger and the
pipeline connecting the heat exchanger with the washer. If we take into account /328
the fact that under actual conditions the sodium flow at these sections is laminar,
then it can be shown that reduction in the supersaturation due to deposition at
these sections will be higher, the greater the value of the expression
Approximate estimates show that the reduction is supersaturation for oxide indi-
cators in which the pipeline connecting the heat exchanger with the washer is not
longer than 0 2 m, while the heat exchanger does not have branched out surface,
is insignificant. Under other conditions this effect may be substantial, which
must be taken into account when designing the device.
Obviously the [postulated] absence of crystallization centers moving to-
gether with the sodium will produce an additional reduction in the supersatura-
tion and will subsequently result in lower-than-actual readings.
It can be seen from Eq. (13) that, in order to make smaller the difference
between actual conditions and the instrument's readings, it is necessary to re-
duce the dimensions of the holes in which the oxides are deposited and also the
327
rate at which the temperature is reduced at the washer, and to increase the rate
of flow of the sodium through the instrument. It also follows from this express-
ion that in order to maintain At constant when the oxygen content is decreased
(reduction in dC/dt) it is necessary to either increase the sodium flow rate through
the instrument or to reduce the cooling rate.
Experimental results. One of the designs of cock-type indicators used is
shown in Fig. 1. The characteristics of holes and heat exchangers of [a number
of] indicators are shown in the Table.
TABLE Characteristics of Holes and Heat Exchangers

of Oxide Indicators
Instru- Holes Heat exchange
ment ' Total (Tube di- assage
num- Dimensions Number cross
area am Yeete,r section, Design
be (
ber mrr ^nm
1 1 Y. 1 16 16 30 x 2.5 0.00049 Rubber tube

2 1x1 16 1 e, 55 x 3 0.00188 a
3 0.52 x 0.52 34 --9.2_ 5.5 / 3 0.00188
4 0. , ; 0. ,1 34 ---2.5 55 3 0.00188 a
5 0.29 % 0.32 64 6 :11 / 2.5 0.00106 Two- pass coil
6 0. 31 ;' 0..',4 16 1. 7 31 / 2.5 0.00106 Same as above
7 1.SL/0..52 16 4.3 31 /2.5 0.00106 n
6 0.67 x 0.87 16 12 :11 / 2.5 0.00106 >
9 I 1 >; 0.75 12 9 20/2 0.0002 Single-pass cell
10 10.52 x 0. 52 34 9.2. 1 0.0002 Same as above
In performing the experiment we have worked out the following method for /329
determining the stopping up temperature. Before starting the test, when the tem-
perature of the liquid metal in the device was at maximum (400-500C) the holes
were washed with sodium (flow rate of 1 m 3/hour) for 15-20 minutes. The exper-
iment was started only when the flow rate of sodium through the device for the
given pressure drop became constant. After washing the holes, the washer was
lowered,, the required flow rate of sodium was established; the fan is started 5-10
minutes later and, adjusting the air flow rate, the required rate of metal temper-
ature reduction in the holes was selected. The tests were run either until the flow
of metal through the device ceased entirely, or until the flow rate became 0,1-0.2
of the initial quantity.
The dependence of the instrument's readings on the sodium flow rate was
studied using nine washers in various combination with heat exchangers (see
Table). For each instrument it was read not less than with two oxygen contents
in the sodium. The absolute level of oxygen concentration in the sodium varied
-2
from 2.5 . 10 % by weight (t PI -- 400C) to 6 . 10-4% by weight (tpl R^ 150C). The
flow rate of sodium through the device was varied from 10 to 800 liters/hour.
328

VQ
Figure 1. Design of a Cock-Type Oxide Indicator.

1) Thermocouple Pocket; 2) Main Valve; 3) Heat Ex-
changer; 4) Flowmeter; 5) Adjusting Valve.
If, in the case of large sodium flow rates, the heat exchanger capacity did
not suffice to reduce.the temperature to the required level, the temperature in the
circuit had to be lowered. Here the temperature drop in the heat exchanger was
not less than 70C. Some results obtained in experiments for determining the de-
pendence of the instrument's reading on the metal flow rate through it, are pre-
sented in Fig. 2.
A fact common to all the experiments is that, when the flow rate of sodium
is reduced below a certain level, the instrument readings become lower and un-
stable. Here the lower the oxygen concentration in the sodium, the higher the
flow rate at which the instrument starts giving unstable readings, Reduction of
hole size for the same heat exchanger design and with the oxygen concentration /330
of sodium remaining constant also results in increasing the flow rates at which
the instrument readings are lower than actual.
Figure 3 presents results characterizing regions of unstable (I) and stable
(II) operation of indicators; the points show the flow rate of sodium through the
instrument for which the indicators give
lower than actual readings for the given
4,,p 1 , i r-- oxygen content of the sodium. It can be
I + stated that region I is a region which the
IJ reduction in the readings is a quantity
which is clearly recorded in experiments.
25u According to this approach [which is based
upon Eq . (13) ] . one would expect that in
201 qo the transition from region I to region II,
the flow rate would increase with o- re-
150 duction in the oxygen concentration in the
I ^
sodium or upon an increase inthe.hole size.
0 101 200 309 400 Qo,1/hr. Similar behavior was observed ex-
perimentally. Quantitative comparison of
Figure 2. Readings of Indicator experiments shows' that, when the oxygen
No. 5 as a Function of the Metal content is reduced from 0.01 to 0.0005%
Flow Rate. by weight, in order for the instrument to
329
tPi"
I operate reliably it is necessary to increase the flow
rate by a factor of 3.2; in our experiments the flow
rate has to be increased 2-3 fold. [Thus] the depend-
1 O \ ence on the hole size is less sharply expressed in the
^ II experiments than was expected.
û
0 When comparing experimentally - and theoretic-

2^0 ^ 1 O ally-obtained data consideration should be given to
instability of results for determining the transition
^J
o o\ boundary between regions I and II. In our opinion
0 200 acv n,,, 1/ hr. this is due to the statistical character of oxide depo-
sition in holes and to some instability of parameters
Figure 3. Regions. of such as the rate of cooling, distribution of flow resist-
Unstable (I) and Stable ance in the device among the holes and the adjusting
(II) Operation of Indi- valve. When the sodium flow rate through the device
cator No. 5. is reduced or when the hole size is increased, the
portion of the resistance devolved upon the adjusting /331
valve increases. This should have resulted in a high-
er co and, consequently, also in a higher rate of re-
duction in the instrument's reading with a reduction in the flow rate than it
would follow from Eq. (13).
It should be noted that data obtained in these experiments cannot be used to
judge about the difference between the saturation and plugging up temperatures
for the stable operation region. Clarification of this problem requires either ac-
cumulation of more statistical data or performance of special experiments.
The experiments for determining the dependence of the indicator's readings
on the size of holes, rate of temperature reduction at the washer, which are de-
scribed below, were performed. with flow rates for which the instrument's oper-
ation was stable (see Fig. 3, region II).
The dependence of the instrument's readings on the hole size was studied
by comparing two instruments simultaneously connected to the circuit. Here the
oxygen concentration was maintained at the specified level. The plugging up tem-
perature was determined at least twice for each instrument and then averaged.
Comparisons were made for instruments 4 and 10, 9 and 2, 5 and 7, 3 and 9 (see
the table); the averaged ratios of plugging up temperatures were 1. 08, 1. 03 and
1, respectively. It can be seen that indicators with smaller holes yields higher
plugging up temperatures, i. e. , the smaller the hole, the lesser the difference
between the plugging up temperature determined by the instrument and the satur-
ation temperature. As the oxygen concentration in sodium was made larger, the
differences in the indicators' reading became smaller. For example, when com-
paring readings of indicators 4 and 10 in the range of t pl = 155-200C, the indica-
tors with smaller holes give readings higher by 11 %, while in the ranges 200-280,
280-320, 320-360 and 360-390C the readings were higher by 10, 8, 6 and 4%
respectively.
Using average ratios of plugging up temperatures and assuming that

1p1 1P1_ tP1
9_1*ip1 tP1'
(15)
_(P
330
, t3 , t5 to t4 This ratio as a
we have calculated the ratios of t2pl pl, t4
pl pl and t9pl pl
function of the hole size is shown in Fig. 4. It can be seen that, as the hole
size is made larger, the instrument reads lower.
The dependence of the instrument's readin gs on the rate of reduction of the
temperature of the metal flowing through the holes was studied for instruments 1,
2, 3 and 9. The oxygen con-
tent of the sodium was varied
from 0. 004 to 0. 026% byweight.
The experiments with each in-
strument were performed for
two oxygen concentrations. The
rate of temperature reduction
was from 0.3 to 37C per min-
ute. Within limits of experi-
mental error no dependence
of the instruments' readings
on the rate of temperature re-
duction of the metal flowing
through the holes was observ-
2 Q4 O'S Q8 d,mm ed in this range of concentra-
tions.
Figure 4. Readings of a Cock-Type Oxide Indi-
cator as a Function of Hole Dimensions. At the same time exper-
iments performed in region I
(see Fig. 3) with differentrates
of temperature reduction at the
washer have shown that, when the rate of temperature reduction is made smaller, /332
the difference between the actual oxygen content and the instrument's readings is
reduced. Moreover, in experiments far from the interface between the regions
in cases when the temperature of the sodium flowing through the holes was re-
duced by 10-15C below the saturation temperature, a gradual reduction in the
flow rate due to deposition of oxides on the holes, was also observed.
Considering experimental results and the possibility of occasional variations
in the flow rate of metal through the instrument, it can be said that in measure-
ments with cock-type indicator it is undesirable to have a higher rate of temper-
ature reduction than 10C/min, since this can result in a random error in deter-
mining the plugging up temperature.
The dependence of the plugging up temperature which is determined by the
cock-type oxide indicator on the hole size made it necessary to perform calibrat-
ing tests.
Calibrating a Cock-Type Oxide Indicator

The No. 4 oxide indicator (see table) was calibrated by two methods:
1) comparing the stopping up temperature with the temperature inside the cold
trap which has operated for a prolonged period at constant temperature and 2)
determining the oxygen concentration in the sodium by a distilling sampler, [7].
331
Calibration by the first method is based on the experimental and logical fact that,
when the liquid metal circulates for a long time through the cold trap, its oxygen
concentration will be determined by the temperature conditions of the trap. Dur-
ing the experiment the temperature in the trap was maintained at the specified
level to within t 10C,
After the entire sodium was passed through the cold trap at least five times
the plugging up temperature determined by the oxide indicator was practically con-
stant and, within limits of experimental error ( 10%) was in agreement with the
temperature in the cold trap.
The use of the second calibration method was as follows. With the oxygen /334
concentration in the circuit and the circulation [rate] of the sodium through the
sampler constant, the plugging up temperature was measured at least twice by the
indicator and it was averaged, Immediately after the last reading was taken on
the oxide indicator, a sample was taken and analyzed (Fig. 5),
The scatter of experimental data on oxygen concentration in excess of 1.10-3%
by weight, as compared with the average, did not exceed 30%. This scatter becomes
larger at higher concentrations, which may be attributed to the fact that the value
of the control test, which determines the sensitivity of the method is, at low con-
centrations, commensurable with the natural oxygen content of the sodium. It is
obvious that, within the above limits of accuracy in determining the plugging up
temperature and in determining the oxygen content of the sodium, the relationship
which is obtained represents the solubility of oxygen in sodium as a function of
temperature. Analyzing it by the method of least square, we get
igC ,s t.i4 T
I 0 . (16)
Comparison of results obtained by the present authors with those of other

workers shows that in temperature ranges below 280C the former yield a higher
oxygen solubility in sodium than the latter. In the temperature range 140-190C
the present results are close to those of [4, 8-10]; in the range of 250-390C
data of [5, 8, 11-131 are practically identical. At hight temperatures [6]
and [11] give oxygen solubilities exceeding data by the present authors 3 and 4
fold, while data of [12] and [13] at these temperatures are in agreement with the
present data.
The relationships presented in [9] and [10] apparently reflect the results of
the same experimental work [4]. However, the results in it were obtained by a
cock indicator with larger holes (1.27 mm), which should give oxygen solubilities
in sodium on the high side (an indicator with such holes gives a lower plugging up
temperature than an indicator with 0, 24 mm holes, by more than 8%).
The results which were obtained show that the cock-type oxide indicator is
a quite complicated instrument, something which is not reflected in published work.
The instrument's indications depend on many parameters. Among them we must
note the flow rate of sodium through the instrument devolved upon one hole. It was
shown that when the flow rate through the instrument is reduced below a certain
332
CO2'
%by
lengtl
0
r
o
10^
r
v n

o
V q
\
0
n
\oo
10-`
- n npe o
-- --
o
Q'
0
10`
0
o ^ o
o
10-f
7.2 7.4
T-1 7.6 1.8 2.0
I
2.2
FTI
2 4' TX;O ' deg 1
Figure 5. Calibrating a Cock-Type Oxide Indicator.

NOTATION: q and - [5]; (D - [12]; - [8]; - [10];
- [7];v - [11]; o and are for data obtained by
the present authors.
value the latter gives lower than actual readings. The larger, the holes, the
higher the metal flow rate which is needed for stable operation. For an indica-
tor with holes of 0, 87 x 0, 87 mm it is about 600 liters/hour. /335
Experimental studies show that, as should have been expected on the basis
of theoretical analysis, even when the flow rate of sodium is high, the instrument's
readings depend on the hole size. A device with holes of 0.24 x 0, 31 mm gives
readings on the average by 10% lower than an instrument with hole size of 1 x 1 mm,
It follows from theoretical analysis that, in order to reduce the error, one
should strive when designing the instrument that the flow resistance be concen-
trated in the holes, while the length of sections, washed with sodium whose tem-
perature is below the saturation temperature, their surface and volume should be
held to a minimum,
333
Calibration of the instrument with holes of 0.24 mm by the first method has
shown that, for large sodium flow rates, the plugging up temperature agrees, with-
in the 10C of experimental error, with the saturation temperature. This cir-
cumstance made is possible to refine the relationship governing the temperature
dependence of oxygen solubility in sodium.
NOTATION
M - molecular (atomic) weight
D - coefficient of diffusion
At - negative deviation in the instrument's readings, C
j - rate of flow of sodium oxide per unit surface per unit time
C s - concentration of oxygen in the sodium, corresponding to the
solubility at the given temperature
Z = C -- C s - supersaturation
d 0 - diameter of holes in the instrument
^ - integral resistance coefficient
a - factor which takes into account stability of flow rate through
the instrument and the sensitivity of the secondary instrument
co - factor taking into account the distribution of flow resistances
in the instrument
Pr = v/D - diffusion Prandtl number
T - time
P, - some constant approximately equal to unity

n - number of holes in instrument
S - surface [area]
S - passage area
V - volume of section up to the washer washed with the super-
saturated solution
334
REFERENCES
1. Blake, L. R.: J. of the Brit. Energy Conf. , 6, 1 (1961).

2. Blake, L. R. and James, A. R.: Nucleonics, 19, 5, 66 (1961).
3. Brugeman, W. H.: A. I. Ch. E. Journal, 2, 153 (1956).
4. Grey, I. L.: Nucleonics, 14, 10, 34 (1956).
Translated from English under the editorship of A. Ye. Sheyndlin, Foreign
Literature Publishing House, Moscow, 1958.
6. Levich, V. G.: Fiziko-khimicheskaya gidrodinamika [Physio-chemical Fluid
Flow] [??]. Fizmatgiz, Moscow, 1959.
7. Dmitriyeva, N. B., et al.: Article preceding this in the present collection. /336
8. Allen, J.: Nucl. Engng, 7, 76, 352 (1962).
9. Sidjel [ ? ?] S. , et al.: Proceedings of the International Conference on Peace-
ful Uses of Atomic Energy, Geneva, 1955, Technology and Engineering
Series. Published in the US by McGraw-Hill for Pergamon Press or
vice versa.
10. Evans, J. W.: Nucl. Engng, 4, 35, 59 (1959).
11. McIntosh, A. B. and Bagley, K. Q.: The Journal British Nuclear Energy
Conference, 3, 1, 15 (1958),
12. Williams, D. D. , et al.: J. Phys. Chem. 63, 1, 78, (1954).
13. Bannykh, Z. S. and Fefelova, I. V.: In Trudy Ural I skogo khimicheskogo
instituta. Issue 5, Goskhimizdat, Leningrad, 1957-1958.
335
EFFECT OF ADMIXTURES ON THE ELECTRIC RESISTANCE
OF MOLTEN LITHIUM
M. N. Arnol'dov, M. N. Ivanovskiy, V.I. Subbotin
and B.A. Shmatko
A number of publications have appeared lately which deal with experimental /336
determination of the electrical resistivity of molten lithium [1], [2]. The devia-
tions between the results of these references exceed the total experimental error,
which may be caused by differences in the chemical composition of the lithium
specimens used.
The basic (quantitatively speaking) admixtures of molten lithium, which in-
teract with it are: nitrogen, oxygen and hydrogen (from vapor in the air). The
quantity of admixtures in the lithium depends on the specific experimental condi-
tions and may turn out to be a random quantity. The admixtures are a source of
additional (admixture-caused) electric resistivity.
Consideration of the electric resistivity of lithium as a function of the ad-
mixture content is of interest also from another point of view. It helps in under-
standing the physical nature for the origin of admixture-caused electric resistance
as a whole and in liquid metals in particular.
Attempts have been made to explain the experimental data on the effect of
metallic admixtures on the electric resistivity of some molten alkaline metals,
based on the development of the well-known Linde law [3], [4]. The authors of
[5], using and developing the work of Friedel and others, [6], have calculated the
magnitude of the electric resistance of molten lithium containing potassium and
sodium admixtures.
The effectiveness of theoretical calculations is appreciably reduced due to
the inavailability of experimental data on which to base the calculations, in partic-
ular, due to inavailability of data on the dependence of the partial volumes dissolved
in the lithium on the concentration of these volumes. Of great importance is the
magnitude of the charge of the admixture ion in the lithium. We have measured /337
the electric resistivity of molten lithium as a function of the temperature and of
the concentration of the oxygen, nitrogen and hydrogen dissolved in it.
Chemical-thermodynamics calculations show that the oxygen, nitrogen and
hydrogen exist in lithium at temperatures of 200-350C mainly in the form of an
oxide, nitride and hydride of lithium, respectively. The existence of admixtures
such as lithium hydroxide or - carbonate under these conditions is impossible from
the point of view of thermodynamics.
The measuring tube into which the lithium being tested was poured, was made
from 1Eh18N9T steel with an inner diameter of 7.95 mm and a wall thickness of
0.25 mm. Before the lithium was poured, its electric resistance in the entire
336
range of temperatures being studied was measured. Thermostatic control of the
measuring tube was obtained by a thermostat with a boiling Dowtherm (E). The
electrical resistance of the tube with the lithium, which was poured into it, was
measured by a double dc bridge. The lithium used in the test was ordinary, with
the following composition:
Element OZ N2 C I EI,
Composition 1.10-2 I 1 . 10- 2 1 5 . 10- } 3 lU-'
1
Element Na K Ca Be n1g I Ag
Composition I 5 . 10 - 1 8.10 -3 6.10 - 3 <10 - 3 3 . 10- 3 3.10- 4
I
Element Cu Ti Ni 1'e 11n Ba
Composition 3 . 10 -3 3 . 10-3 2 . 10- 3 1 . 10 -3 1.10-4 1.10-'-
) J
Element I Al I 7r Cr + Si I
Composition I 3 . 10 - 2 3 . 10- 4 3.10 - 2 I 4. Si
The admixture content in the lithium was measured by introducing gaseous

oxygen, nitrogen and hydrogen beneath the surface of molten lithium in a special
tank. In order to rapidly obtain equilibrium concentrations of the admixtures un-
der test and to [then] hold it constant, the lithium was continuously circulated from
the tank into the tube and back. The quantity of the admixtures added was deter-
mined by measuring their pressure in a tank with known volume, from which the
gaseous admixture was supplied to the lithium tank. The quantity of the admix-
tures added and results of chemical analysis of lithium for the content of the giv-
en admixture were systematically compared. Differences in the quantities com-
pared did not exceed the error inherent in chemical analysis and the measurement
of the admixture quantities added.
The obtained dependence of the electric resistivity of lithium on the content /338
of nonmetallic admixtures (Fig. 1) can be used to explain the differences in ex-
perimental data obtained by various investigators who, most probably, made
measurements with lithium with different composition and content of admixtures.
It follows from Fig. 1 that the dependence of the electric resistivity of lithium
on the admixture content can be satisfactorily described by a linear formula,
which is in agreement with the general viewpoint of the theory of admixture-
caused resistance at low admixture concentrations. Unfortunately, the total in-
availability of data on the dependence of the specific volumes of admixtures and
their partial pressures on the admixture temperature and concentration
makes it impossible to process the data we have obtained in the manner that this
was done in [3-5].
It is remarkable that the electric resistivity of lithium depends on the oxygen
content much less than on the nitrogen and hydrogen contents only in the case when
all these admixtures are nonionized. In the opposite case the oxygen ion, which
has a larger charge than the hydrogen ion, should produce a larger admixture-
caused resistance. Apparently, here the latter is not determined by the difference
in the ionic charges of the admixtures and of the solvent metal, but rather in the
337
difference in dimensions of the admixture and solvent atoms. This difference pro-
duces a nonuniform charge density in the volume of the metal, which is precisely
the cause for electron scattering which produces the admixture-caused electric
resistance [4]. The absence of nitrogen and hydrogen ions in lithium at tempera-
tures up to 350C, or, more precisely, their negligible content, is shown by the
following: the magnitudes of the admixture-caused electric resistivity do not de-
pend on the temperature, i. e., the Matthiessen rule is satisfied.
Figure 2 shows the electric resistivity of lithium as a function of the tem-
perature for different nitrogen and hydrogen concentrations. The differences be-
tween the resistivities of contaminated and pure metals for a given temperature /339
is the admixture-caused electric resistivity. As can be seen from Fig. 2, the
resistivity lines as a function of temperature for different admixtures' are mutu-
ally parallel, which shows that the Matthiessen rule is satisfied.
The portion of admixtures which is ionized in the lithium depends on the
temperature. Hence, if the admixture-caused resistivity is found not to be tem-
perature dependent, this means that the lithium contains no ionized nitrogen or
hydrogen.
J x ohm cm
I I
X !"'ohm cm
i I o
J o
33
I 2 0
O
I 32 - -_---
5 0
7
3
31
O ^
t O
30
I
O
k2
29
280 - 300 320 t, T

I- 0
/,0 e0 0"-U Wt//

Figure 2. Electric Resistivity of Lith-
ium as a Function of the Temperature
Figure 1. The Electric Re-
for Different Nitrogen and Hydrogen
sistivity of Lithium at 300C Contents.
as a Function of the Oxygen, 1), 2) and 3) Changes in Nitrogen Con-
Nitrogen and Hydrogen tent; 4) and 5) Changes in Hydrogen
Content. Content.
338
It should also be noted that the relationships expressed by Figs. 1 and 2
remained unchanged also for a measuring tube held, together with the lithium,
at 300C for 800 hours.
REFERENCES
1. Fridman, J. F., et al.: J. Chem. Phys., 34, No. 3 (1961).

2. Solov'yev, A. N., et al.: Zhurnal prikladnoy mekhaniki i tekhnicheskoy
fiziki. 1, No. 1 (1963).
3. Blatt, F. Phys. Rev. ; 108, 285 (1951).
4. Green, B.: Phys. Rev., 4, 126 (1962).
5. Shpil I rayn, E. E. and Savchenko, V.A.: Teplofizika vysokikh temperatur,
3, No. 1, 1965.
6. Friedel, J.: Philos. Mag. , 7, 43, 153 (1952) .
339
THE USE OF HOT TRAPS FOR REMOVING OXYGEN FROM SODIUM
F. A. Kozlov and E. K. Kuznetsov
INTRODUCTION
One of the main problems arising in using liquid sodium as a reactor cool- /340
ant is that of keeping its oxygen content to a minimum. The required degree of
purification is determined by the construction materials used in the facility, the
working temperature, facility design and other factors. At present extensive use
is made of two methods of oxygen removal: purification by cold and hot traps (get-
ter purification). Both methods differ from one another radically by the nature of
phenomena on which their operation is based, and also by the minimum concentra-
tion of oxygen which is obtained by using these methods.
The solubility of oxygen in sodium drops with a reduction in the temperature,
which is the basic method for oxygen removal by the cold-trap method. The min-
imum possible sodium temperature in cold traps is limited by the 100-150C lev-
el, hence, the minimum oxygen concentration produced by these devices is (1-. 5) 10-3
% by weight, [1-3 1.
The getter purification method uses the ability of some substances to reduce
sodium oxides to pure sodium. The higher the reaction temperature, the higher
the reaction rate. This method ensures a higher degree of oxygen removal than
the cold-trap method. Consequently, if the oxygen concentration in the sodium
should be below that produced by cold traps, the hot trap should be used.
The problems which must be resolved when removing oxygen and other ad-
mixtures by a hot trap are: selection of the getter material, study of the laws gov-
erning the interaction between the getter and the sodium's admixtures, determi-
nation of the maximum oxygen concentration produced by the [given] getter, se-
lection of the hot trap design, developing techniques for obtaining and maintaining
the specified oxygen concentration in the sodium using the hot traps [1, 3 - 6].
The condition characterizing the ability of a metal to reduce sodium oxide
to pure sodium is the reduction in the free energy of reaction
n\a^0 0Te ; TeA,+2nNa. (1)
Figure 1 shows the free energy needed for formation of stoichiometric oxides of
various metals as a function of the temperature, [7]. It can be seen that a num- /341
ber of metals can be used as getters. However, it is most practical to use zir-
conium and its alloys with titanium, since, when reacting with oxygen, these form
a dense film bound to the basic metal of the oxide. The interaction of zirconium
340

C' ' M(
fir
' e
21
101
2I5ry5 4s
141
lift
186
221
251
0 400 100 7200 700 20619 T, 0
Figure 1. The Free Energy of Oxide Formation (per 1

Gram-Mole of Oxygen).
with oxygen is described in [5] and [6]. It was shown experimentally that the in-
crease in the weight of zirconium can be described by an equation such as
B .
exp ( _T) k
oP T (2)
However, the available data are insufficient for resolving the above problems. /342
[For this reason] we have performed experiments to clarify the laws governing
oxygen removal by hot traps with a zirconium getter.
341
Interaction Between Zirconium and Sodium-Held Oxygen
In our experiments we have used arc-molten zirconium iodide foil 0, 1, 0.5
and 0.8 mm thick. Specimens of the needed size were placed in special chambers
which were cut into the circulation circuit of the installation. The chambers were
constructed so that it was possible to test the specimens at three different sodium
velocities. Before starting the test the sodium was pumped through the specimen
chambers for 8-35 hours at temperatures not above 350C, simultaneously remov-
ing oxygen from it by a cold trap. When the oxygen concentration in the circuit
dropped to 0.001-0. 0005% by weight, the temperature in it was raised, and then
the specimens were tested for a specified time.
In the course of experiments the oxygen content of the sodium was maintain-
ed at the specified level by continuously pumping the sodium through the cold trap,
the temperature inside of which was 110-140C. The constancy of oxygen concen-
tration in the circuit was monitored by a cock-type oxide indicator. A number of
specimens was tested in sodium purified by a hot trap.
After terminating the experiments the chambers were cut out, the sodium
residues were dissolved in ethyl alcohol, the foil was washed with water and dried
and then its weight gain was determined. Control experiments showed that this
kind of processing does affect the accuracy of weight gain determination. Individ-
ual specimens were replaced in the chambers for continuation of test, after the
weight gain has been determined. We have performed 113 experiments at 510-
1100C. The time during which the specimens were held in sodium varied from
15 to 435 hours, while the sodium velocity was from 0.05 to 5.7 m/sec.
In calculating coefficients A, B and n of Eq. (2) we have used only those data
for which the weight gain in the roughly purified sodium was less than 17% of the
total weight gain in the tests. A part of the tested specimens was subjected to
chemical, metallographic and x-ray diffraction analysis.
Dependence of the Weight Gain of Zirconium Specimens on

the Time of Holding in Sodium
For zirconium specimens tested in sodium (constant temperature) k of Eq.
(2) is constant. Then, analyzing the experimental results in the form
Ig S_ __ f(lgi),
which get a straight line (Fig 2). In this method of data analysis, we have used /343
the method of least squares to determine the power n for temperatures of 540,
560 and 620C, the average value of which was 0.51.
342
Dependence of Weight Gain of Zirconium Specimens on the Temperature
In the study of temperature dependence of the weight gain in zirconium the
temperature range was extended to 1100Cm However, due to failure of the zir-
conium foil (at high temperatures] we
could only determine the weight gains
tg `S ; 1 of specimens tested at 720C. The
experimental data, which are treated
n j in the form lg( ^ 51) =f(1/T) are
o o I ST
3 _-- g ----- - -- -.,,----.-- shown in Fig. 3. As expected, within
I limits of experimental error, they lie
^`d 1 on a straight line, the equation of which,
found by the method of least squares,
has the form
2
1 2 3 l_q z 4P
1g, - 4.253 6800 Y (3 )
.ST0.61 ,
Figure 2. Weight Gain of the Zir-
conium as a Function of the Hold-
ing Time in Sodium. Changing to Eq. (2), we get
O Ig
l^P^
S
-- +.187 -;- 0.54 1,,^ 2 for 5i0'C;
3-318 + 0.50 I t,, for 6201C.

AP = 1.8.109to.:,l eap r __ 157ii0 1 (4)
1.05 1.1 7,15 42 7.25 110.1
Figure 3, Weight Gain of Zirconium as a Function

of the Temperature.
We now consider the results obtained by testing 0,1 mm thick zirconium /344
specimens at temperatures above 650C in more detail. The weight gains of the
specimens at these temperatures are lower than those which were expected on
the basis of Eq. (4) This reduction may be due to two factors: increase in the
rate of mass transfer for zirconium with an increase in the temperature of the so-
dium, washing the specimens, and reduction in the rate of zirconium penetration
343
by the oxygen, due to the reduced oxygen concentration gradient. For these rea-
sons in calculating n in Eq. (2) we did not take into account results obtained at
675C and above.
Dependence of the Weight Gain of Zirconium Specimens on the Oxygen

Concentration and on the Sodium Velocity
To clarify the qualitative dependence of the weight gain of zirconium speci-
mens on the oxygen concentration of the sodium, the latter were tested at three
different velocities of sodium,
which was purified by a hot
t rap, Th t t s were run a t
a es
NC I
620C for 115 hours (Fig. 4).
`
M
In sodium purified by a hottrap
x the weight gains of the zircon-
ium were found to be 2.5-3 fold
smaller as compared with those
Z obtained in sodium purified by
a cold trap. When the velocity
k2 0.4 0,6 b"/,m/sec
as reduced from 0.2 to 0.02
m/sec the weight gain became
Figure 4. The Weight Gain of Zirconium as smaller by a factor of more
a Function of the Sodium Velocity. than 2.3. We did not succeed
in precisely measuring the
c - Oxygen Removed from the Sodium by a oxygen content of the sodium
Cold Trap; o -Oxygen Removed from the So- in this experiment. It is
dium by a Hot Trap. only known that it was less
than 0. 00030/0 by weight.
Changes in the Chemical Composition of Specimens

Subsequent to Holding in Sodium
Chemical analysis of zirconium specimens held in sodium showed that, with-
in limits of experimental error, the weight gain of the zirconium is determined by
oxygen absorption. No systematic relationships for variations in concentrations
of elements such as carbon, nitrogen, hydrogen, iron, chromium, sodium, cop- /345
per, aluminum, etc., was discovered. At temperatures below 620C no stable
concentration of nickel in the specimens was observed, while at temperatures
higher than this the nickel concentration was higher than for the starting zirconium.
X-ray Diffraction and Metallographic Analyses of Zirconium

Specimens Subsequent to Testing
X-ray diffraction was used to analyze the starting zirconium and specimens
0. 1 mm thick tested at temperatures of 540, 560, 600, 675 and 720C for 250 hours.
Analysis of the x-ray pictures showed that specimens . tested at 600 and 675C
344
contain on their surface a layer of ZrO2 . No zirconium oxides were found
on specimens tested at 540-560C.
On specimens 0.8 mm thick, held in sodium at temperatures of 620 and
720C for 432 and 250 hours, respectively, metallographic analysis has discov-
ered a zirconium oxide film about 0.1 mm thick. The oxide film increases sharp-
ly the microhardness of the specimen surfaces, which may become as high as
800 kg/mm2.
The grain size of crystals of specimens tested in sodium increase in size
on raising the temperature and extending the holding time. It should be noted
that the absorbed oxygen increases the brittleness of the zirconium. From this
point of view the permissible weight gain per unit surface for zirconium sheets
0.1 mm held in sodium purified by a cold trap is about 4%.
Change in the Electric Resistivity of Zirconium Subsequent

to Holding in Sodium
The resistance of the starting zirconium and of specimens held in sodium
was measured by the R-329 double bridge. The experimental data were reduced
to 0C. It was assumed in the calculations that the temperature coefficient of
electric resistivity does not depend on the oxygen concentration and is equal to
4.4 . 10-3/C [8] (Fig. 5). The differences between the electric resistivities of
the zirconium-oxygen alloy and the zirconium oxidized in sodium purified by a
cold trap are apparently due to nonuniform distribution of oxygen over the spec-
imen cross section.
The electric resistivity of specimens from hot traps is in better agreement
with data of [8]. This can be attributed to the fact that the above zirconium was
held longer in sodium with a lower oxygen concentration than the specimens tested
in sodium purified by a cold trap. Under these conditions the rate of oxygen dif-
fusion into the sodium exceeds the rate of formation of the surface oxide layer /346
and, consequently, the oxygen absorbed by the zirconium is more uniformly dis-
tributed over the specimen cross section, forming a structure to zirconium-ox-
ygen alloys. It is obvious that for this very reason the zirconium in hot traps is
more brittle than in cold trap purified sodium.
Removal of Oxygen From Sodium by Hot Traps

On the basis of the above results we have constructed and tested several
hot traps, in which zirconium foil 0.1 and 0.3 mm thick was used as the getter
(Fig. 6). These traps were tested in sodium circuits with a metal volume from
0. 05 to 0.2 m3 at 550-670C. The test time varied between 100 and 1150 hours,
the flow rate of the sodium through the hot traps was 1-2.5 m3 /hour. The so-
dium purity was monitored by a distilling sampler [9].
345
//
//
U
mr, ^
ô
0
x
Q
c
^ o O O
0 1 2 3 f .4plp"x
Figure 5. Electric Resistivity of Zirconium as a Function of

the Oxygen Concentration in Sodium.
.0. - Zirconium Held in Sodium Purified by a Cold Trap; a - Pur-
ified by a Hot Trap; b, a and A. Z irconium from Different Hot
Traps; the Dashed Line Reflects Data of [8]
Sec. AA
Sec. BB
Figure 6. Section Through a Hot Trap.

1) Reducing Cover; 2) Thrust Ring; 3) Packing Ribs;
4) Trap Casing; 5) Pin; 6) Grid;
7) Zirconium; 8) Mesh.
346
Figure 7 shows the manner in which the oxygen concentration in the closed
circulation circuit varied as the sodium was purified by a cold and hot trap. Sim-
ilar results were obtained in other experiments in which oxygen was removed
from sodium by hot traps. A characteristic phenomenon encountered in these
experiments is the increase in the oxygen content of the sodium after the cold
trap is shut off and the temperature is increased. This is due to the following
factors.
\gym
o,
x
a
9 4 O ^ 1 0 I b " 4---. ^C8aA
100 200 300r00 500 z, hours
Figure 7. Change in Oxygen Concentration in a Circuit

When the Sodium is Purified by a Cold (1) and Hot (2) Traps.
A rise in the temperature increases the solubility of oxygen in the sodium, /348
and the rate of dissolution and speeds up the diffusion processes. Hence, oxygen
from remote corners of the circuit, from walls of expansion tanks and the struc-
tural metal is absorbed by the circulating sodium.
Analyzing the sodium for oxygen using a distilling sampler (see Fig. 7) we
see that after purification the oxygen content of the sodium remains approximately
on the same level of (2-5) 10 -4% by weight. However, these numbers do not re-
flect the actual oxygen concentration produced by the hot traps, but point only to
the limiting sensitivity of the distilling sampler. Hence, the results of these ex-
periments cannot be used to judge about the maximum oxygen content of sodium
purified by hot traps. The above is verified by the fact that the weight gain of the
zirconium held in sodium purified by a hot trap and the weight gain of zirconium
from the hot traps is by several factors lower than that of specimens tested in
sodium purified by a cold trap.
The following results were obtained with respect to the mechanical strength
of the zirconium foil.
1. Zirconium foil 0.1 mm thick (weight gains of 1.5-2%) becomes so brit-
tle that it can no longer be used.
2. In 0. 1 mm thick foil tested for 1000 hours at 650-670C the oxide film
peeled and the foil proper failed. Foil 0.3 mm thick tested under the same con-
ditions lost part of its initial elasticity, but it still could be used.
347
Estimating Equilibrium Oxygen Concentrations in Sodium
It was pointed out above that only those materials can be used as getters in
which the free reaction energy (AF) described by Eq. (1) becomes lower. This
condition is necessary, but not sufficient. The second requirement which is pre-
sented to a getter is providing the required oxygen concentration in the liquid met-
al. Exact calculations based on the law of mass action require knowledge of the
activity coefficients for the sodium-held and getter-held oxygen, as well as the
solubilities of the getter proper, as well as of its oxide in sodium [10, 11]. These
data are either inavailable or incomplete, [Therefore] we give below estimates of
the equilibrium concentrations of oxygen, on some assumptions.
For the reaction described by Eq. (1) we write
Ln
aNa'a!Tpbon
AZ
Ku b n = CXp {
a Me , Na20 WT .
Making use of the fact that the change in the isobaric potential AZ for reactions /349
proceeding at constant volume is practically equal to AF [10], and that a Na2 0 =
C Na2 O /C*Na2O [111, we get
^ n / aPjtl 't ^1^TeÔn

CN320 ' C *a20..
V c RT > AI
CSp --- ^ abtC
6
Assuming further that the solubility of the getter and of its oxide in sodium is zero
(the reaction is heterogeneous) and the activity of the sodium (the reaction proceeds
with excess sodium) is equal to unity, we have
_AF (5)
CN a20 = CNa20 CXp RTn } .
The following can be concluded from the results tabulated above. The great-
est degree of oxygen removal can be obtained by using thorium, uranium and a
zirconium-titanium alloy as getter materials. The degree of oxygen removal is
reduced with an increase in temperature. The lowest equilibrium oxygen concen-
trations in sodium are obtained at temperatures of about 400C and lower, but then
the reaction rate becomes lower. Hence, the temperature at which a hot trap should
be operated should be in the range of 550-650C, Oxides of construction materials
(FeO, NiO and others) should become reduced in sodium even when the oxide con-
tent of the latter is above the permissible.
348
TABLE Equilbrium Concentrations of Sodium Oxides for
Different Getters, Calculated From Eq. (5),
Cna 2 p (mole fractions)
Reaction
1100 C I 600 C ( 800, C
Zr +2Na L 0 /_fOz ;-4Na 1.3.10-15 7.2.10- 13 3.9.10-11

Ti-1-2Na 2 0 TiO^+4Na . 10
1.3 10- .
9.1 10- 9 1,1.10 -7
Nb+2Na 2 0 t Nb02+4Na . 5
4.3 10- 3,5.10--5 1.5.10-1
Ta+2Na20 4.2 . 10- 6 . 5

5 5 T11205+4Na 1.15 10- 1.8.10-5
Th+2Na 2 0 ;t ThOL-I-4Na
U+2Na 2 0 _?- UO2-{-4Na
4.3 . 10- 21
9.7 . 10- 17
4.4 . 10- 17
9.5 . 10 -14
1.2.10-14
5.10-12
Calculation and Design of Hot Traps

The calculation and design of a hot trap are closely related to the system in
which it is to operate. If the materials from which the facility is made present
rigorous requirements to the coolant purity, then the hot trap must maintain the
oxygen concentration on a level which is permissible from the point of view of
corrosion of these materials, If this is not necessary, then the rate of oxygen
removal from the coolant should be such that its concentration should not increase
above the saturation concentration at the lowest coolant temperature in the given
facility.
The following must be known when the determining factor in a given system
is the rate of corrosion of construction materials: the manner in which the cor-
rosion rate depends on the temperature, the time and the oxygen concentration;
the permissible time of contact between the structural material and the oxygen-
containing sodium, and the dependence of this time on the oxygen's temperature
and concentration. Many of these data are not known at present. Hence, the cal-
culation techniques will not be presented for a specific facility.
The purpose of these calculations is determining the quantity of getter mate-
rial needed for obtaining the required hot-trap output and for ensuring the stable
operation of the latter and of the facility as a whole during a specified time period.
In performing these calculations one must, in addition to starting data (they are de-
termined by the system which is to be served by the trap) select the getter mate-
rial, the hot trap temperature, the geometric size of the getter and other parameters.
Selection of a Getter and Operating Parameters

Zirconium foil can be used as a getter for removing , oxygen from sodium. On
the basis of experimental data and of thermodynamic calculations, we assume a tem-
perature range of 550-650 0 C for the zirconium getter. The lower temperature limit
349
is bound by the rate of absorption of oxygen by the zirconium (below 550C the re-
action rate is reduced appreciably), while the upper limit is limited by the increase
in the mass transfer of zirconium [6].
The velocity of sodium flowing through the trap is taken so that the flow in
the cells formed by the foil is turbulent. Calculations show that for a hydraulic
diameter of cells of about 10 -3 m the velocity of sodium in the latter should be
0.6 m/sec. If the parameters of the facility (sodium flow rate, pressure drop)
or design considerations require that the veolocity in the trap be lower, this can
be done, but then consideration should be given to the possible reduction in the
oxygen absorption by zirconium.
The thickness of the zirconium foil is determined by the total weight gain of
zirconium per unit surface for the entire time of trap operation. The total weight
gain should be such that no oxide films, which subsequently peel, form on the zir-
conium surface and that the foil does not become too brittle. . The weight gain per
unit area satisfying these conditions will be called the normally permissible weight
gain.
It can be seen from the above that the maximum weight gain permissible per /351
unit area is a function of parameters such as foil thickness, operating tempera-
ture, oxygen concentration in the sodium, etc., i.e.,
,'
Sh i l (1 ' C'
C-
rnax '5)'(6)
At present the form of the above equation is not known precisely, and it has to be
determined experimentally. On the basis of general considerations and of some
experimental data it can be expected that the maximum permissible weight gain
will basically depend on the foil thickness, meaning that it should increase as 6
is made larger. It is shown experimentally that when using foil 1 . 10-4 m thick
the maximum permissible weight gain is limited by 2% or to approximately 0.006
kg/m2 .
Assuming that (AP/S) m,-. is proportional to the foil thickness and disregard-
ing the dependence on the oxygen concentration of the sodium and on the tempera-
ture, Eq. (6) can be written in the form
CS ) max y 06 . i (7)
where we shall assume that K1 = 1.3 when S = 3 . 10 -4 m and that K 1 = 1 when S =

= 1 . 10-4m.
350

Making use of Eq. (4) we can write
15 700 ).
6-2.73.102op'51h',Ii2esp( (8)
T
Depending on the sodium content and on the behavior of flow in the hot trap
(turbulent or laminar flow), the value of K 2 varies from 1 to 0.25-0.5. No ana-
lytical expression has been found to relate K2 to the oxygen concentration. It is
only known on the basis of experimental data that K 2 = 1 when the oxygen concen-
tration is 0. 0005-0. 001% by weight, and K2 = 0.25-0.6 when the sodium is puri-
fied by a hot trap.
Calculating the Weight of Zirconium Without Consideration of the

Rate of Its Interaction With Oxygen
The minimum quantity of zirconium needed for removing the specified quan-
tity of oxygen from the sodium can be calculated from, the reaction equation
Zr- 11 - 0 2 > Zr02.
However, in this case the entire zirconium is transformed into an oxide, which
apparently cannot be obtained in practice. Hence, the calculations should be per- /352
formed using data on the maximum permissible weight gains per unit surface.
The total amount of oxygen to be removed by a hot trap can be expressed as
(w1- 1\'(Cin.--Cfin)P,,a
(9)
Making use of the above expression, we can write the weight of the zirconium in
the form
1:36:Yzi ( C in. -- C fin) P N a
(10)
3 max
200 ( Ay)
It was taken into account in deriving the above that S = 2PZr/yZ 8'
r
It should be noted that calculations for a hot trap using Eq. (10) can be
performed only when the rate of oxygen removal is not of importance.
351
Calculating the Weight of Zirconium for a Given Rate of
Oxygen- Zirconium Reaction
We calculate the weight of zirconium from Eq, (4). In this case we must
make certain assumptions, which are not entirely verified by experimental data.
1. The rate of reaction between zirconium and oxygen is independent of the
sodium velocity.
2. Intermixing of the sodium in the circuit is ideal.
3. The concentration of oxygen in the sodium which at a given instant is
in contact with zirconium in the hot trap is constant.
The equation of material balance for oxygen in a closed system for an infi-
nitesimal time interval has the form
K3P',,âdC=G(CC') dT, (11)
while for the sodium flowing through the hot trap it is expressed as
Sd 4P
G (CC') di ' C S) (12)
100 K4
We shall assume that the oxygen concentration at the hot trap outlet is not
zero (C' /= 0), Solving Eqs. (11) and (12) we then have
We calculate d(OP/S) using the expression
5r __KzAi"expi 7 )_k.K Zi " . (14)
Two cases will be considered,

1. The rate of reaction between zirconium and oxygen in the sodium when /353
C < 0.001% by weight is expressed by the average value of K 2 . Then
1f0l.l', -0-1 dT.

[IC
]S 3'h411,Nj
(15)
352
Integrating the above expression from Cin to Cfin and from ( op - 0) to op, re- T T T
spectively, and recalculating for the weight of zirconium, we get
K3K4'6YZr(Cin,Cfin)1'Na
1) 200kK 2'[ TOP Top T ( 0)"J (16)
2. When the rate of reaction between zirconium and oxygen is a function of

the latter's concentration we have
We perform our calculations on the assumption that K2 = C/CD when C < CO.
Finding (AP/S) for this case, substituting its value into Eq. (13) and integrating,
we get
_ K3h,bY7, r(Cin.Cfi n )1'Na

200-KO [ C fin' T C in.
p (Tp--TO)nJ
(17)
where
k
Ko =
C0 .
It is obvious that Eq. (17) has a physical meaning only when Cfin p > Cin(TOp
- 6) and Cin < C6 . When Cin > C O, then the weight of the zirconium must be
T
determined by simultaneously solving Eqs. (16) and (17).

When the oxygen content of the sodium flowing through the hot trap is ap-
preciable, the zirconium placed at the trap inlet will have a larger weight gain
than that situated far from the inlet. On the assumption that the weight gain due
to absorption of oxygen from the sodium is described by Eq. (14) and that K2 =
= C/CO3 while the concentration of oxygen at the trap inlet is C o , we have ob-
tained an expression which can be used to estimate the specific weight gain of
zirconium foil situated at distance y from the trap inlet, during time T
SP- n exp
CB) (T'
1 T__1 exp ( X) dT,

0
(18)
353

where
100nAexPBT
yr n- IS
X
Co G
Design of Hot Traps

The required quantity of zirconium foil is placed in the hot-trap housing in /354
the form of stacks. One of the placing methods possible is alternating corrugated
and noncorrugated foil sheets. Here it can be assumed that the cells formed by
the foil are equilateral triangles with side length b. Then the volume Vh. t. of a
hot trap can be calculated from the formula
Cb^2 ^
_5
2
1 1. t :- Sh. tbh. t. -- PZr
1/3SbyZr (19)
The total section of the hot trap (She t. ) for a given sodium flow rate through it
can be found from
b+S
GC/
Sh. t. (20)
3600 jl3v Re, , C 1-;- 2 1/35
The inlets of hot traps using zirconium foil must be provided with special
devices, which would be able to hold the particles which are not soluble in sodium,
which form in it. The latter can appear due to failure of zirconium by embrittle-
ment and due to peeling of the oxide film from the zirconium surface. If stainless
steel mesh is used for this purpose, then their design should be as follows: the
flow resistance should be made as low as possible; deposition of mechanical par-
ticles should not rapidly increase the latter. In addition, to prevent the zircon-
ium foil stacks from failing under their own weight and by dynamic loads of the
flow, special dividing grids should be placed between the stacks.
NOTATION
AP/S - increase in the weight of zirconium per unit surface
P - weight
T - time of contact between zirconium and sodium
354
Top - total operating time of the hot trap
TO - time needed for one purification of sodium at the end of the
hot trap's operating life
6 - thickness of zirconium sheet
G - flow rate of sodium
C - oxygen content of sodium
CO - minimum oxygen content of sodium for which no reduction is
observed in the rate of oxygen absorption
C Na2 O - equilibrium concentration of Na2 O in sodium /355
C Na2O - saturation concentration of Na2 O in sodium

A, B, n - coefficients of equation describing the weight gain of zirconium
K2 - coefficient which takes into account reduction in the weight gain
of zirconium when the oxygen concentration is reduced to below
C 0 and the flow turbulence in the hot trap is lessened
K3 - coefficient which takes into account the oxygen flowing in from
dead-end and other sections of the circuit
K4 - coefficient which takes into account the incompleteness of wash-
ing the zirconium surface by sodium
Ka - thermodynamic equilibrium constant .
REFERENCES
1. Hinze, R, B, : Nucl, Engng., 5, 48, 225 (1960).

2 m Faris, F., Johnson, D., et al.: In the book Proceedings of the Second United
Nations International Conference on Peaceful Uses of Atomic Energy,
Geneva, 1958.
3, Allen, J.: Nucl, Engng, , 7, 76 352 (1962),
4 Bawman, F. and Cubicciotti, D.A.: J. Ch. E. Journal, 2, 2, 173 (1956).
5, Carter, R. L., et al.: Proceedings of the Second United Nations International
Conference on the Peaceful Uses of Atomic Energy, held in Geneva. V. 7,
Reactor Technology, Geneva United Nations, 1958, p, 72,
6. Davis, M. and Draycott, A. : Proceedings of the Second United Nations Inter-
national Conference on the Peaceful Uses of Atomic Energy, held in Geneva.
V,7, Reactor Technology, p, 94, Geneva, 1958.
355
7. Yeremenko, V. N.: Titan i yego splavy [Titanium and Its Alloys]. Published
by the Academy of Sciences of the Ukrainian SSR, Kiev, 1960.
8. Tsirkoniy [Zirconium]. Collection of articles, Part II. Translated from
English under the editorship of V. A. Alekseyev. Foreign Literature
Publishing House, Moscow, 1955.
9. Dmitriyeva, I. B. , et al.: Second article, Part IV of present collection.
10. Karapet'yants, M. Kh.: Khimicheskaya termodinamika [Chemical Thermo-
dynamics]. Goskhimizdat, Moscow-Leningrad, 1953.
11. Distefano, J. R. and Hoffman, E. E.: Atomic Energy R eview, 2, 1, 3-34 (1964).
356
MAXIMUM PERMISSIBLE LEAKAGE IN A SODIUM-TO-WATER
STEAM GENERATOR
N. N. Ivanovskiy and F. A. Kozlov
The use of sodium as the coolant in large nuclear power stations makes /355
it necessary to develop efficient and reliable sodium-to-water steam genera-
tors. At present the greatest promise is held out by a single-wall generator,
[1], which ensures higher economy and simplicity than the two-wall design.
However, water leakages in the single-wall device is more dangerous, since
then the water gets into the sodium. One of the major problems which this
creates is that of removing the products of interaction with water from the so- /356
dium. According to [2], the reaction between water with sodium, on excess of
the latter, can be written as
4Na + H,1 0 Naz0 + 2NaH. (1)
Accumulating in the sodium the products of reaction can interfere with the heat
transfer between the water and sodium, stopping up of the circuit's ducts, as
well as increased corrosion of construction materials. Hence, the water must
be .continuously removed from the sodium by a cold trap. As was shown in [3],
this device is capable of removing the products of interaction between water and
sodium from the latter.
Only a part of the liquid metal is circulated through the trap. Hence, de-
pending on the amount of leakage, a given concentration of products of interac-
tion is established in the sodium. Reliable operation of a facility requires that
all the admixtures be in the dissolved state at operating temperatures. This
condition serves as a limit on the amount of leakage above which the facilitywill
be in danger of failure. To determine this maximum leakage we shall now de-
rive an equation which could be used to calculate the change in the oxygen con-
tent of the sodium with time, in the presence of oxygen sources and drains in
the system.
It is shown in [3] that in cases when the minimum temperature in the sys-
tem, when water gets into sodium, exceeds 220C, the sodium oxides will pri-
marily pass into the crystalline phase. Hence, the calculations can be based on
the quantity of oxygen which gets into sodium in the form of H 2 O, while the max-
imum permissible concentration could be the solubility of oxygen in sodium at
minimum sodium temperature in the wall.
On the condition that intermixing in the system is ideal, the equation of
material balance for the oxygen in sodium has the form
Pr a dC = (G, G2) dT. (2)
357
In our case the oxygen is removed by a cold trap. Assuming that the oxygen
concentration of the sodium passing through the cold trap is reduced to C', we
have

G2 = G t ( C C'). (3)
Then Eq. (2) will become

PNadC=[GI G t (C C')l dz. (4)
Integrating the above expression for z = 0 and C = C', we get
tG t
C=C'-{ Gt1 (1 e
(5)
Under these conditions the maximum oxygen content of the sodium will be /357
't
it is reached during a time
SP;^n
> (7)
^G t
Changing to the size of the maximum permissible leakages, we have
G,12o = 1.126 t (Cmax C') (8)
Using Eqs. (7) and (8) for a specific case it is possible to calculate the maxi-
mum permissible leakage of a sodium-to-water steam generator, as well as
the time for reaching the maximum concentration in the circulating sodium.
It can be seen from Eq. (8) that the duration of operation of a steam gen-
erator with a leak will be determined by the thoroughput of the cold trap and
by the magnitude of the leakage. In order that the system be capable of pro-
longed operation with large leakages, cold traps with high thoroughput and ca-
pacity must be designed.
358
NOTATION
PNa - weight of sodium in facility

G - flow rate of sodium through cold trap
G1 - influx of oxygen into sodium by water leakage
G2 - oxygen removed from sodium by the cold trap
C - oxygen content of sodium after -r hours of leakage of steam
generator, parts by weight
C' - oxygen content of the sodium at cold-trap outlet, parts by
weight
G O - maximum permissible leakage of steam generator
2
REFERENCES
1. Poplayskiy, V. M. , et al.: Teploenergetika, No. 7 (1966).
2. Ivanovskiy, N. N. and Kozlov, F. A.: Atomnaya Energiya, 17, 5 (1964),
3. Subbotin, V.I., et al.: AtomnayaEnergiya, 19 (3), 298 (1965).
359
SOME PROBLEMS OF OPERATION AND DESIGN OF
LIQUID-METAL LOOPS
A. V. Drobyshev, V. A. Kurov and I. T. Filipov
The extensive experience accumulated up to now in work with alkaline met- /357
als has not been sufficiently illuminated in available publications [1-4]. Liquid
metal loops are extensively used in the study of compatibility between the mate-
rials of construction and the liquid metal, in the study of heat transfer and fluid /358
dynamics, as well as in testing fuel elements and their mockups outside the
reactor.
The construction material specimens are tested in liquid-metal filled ves-
sels, placed on stationary and rotating supports. Installations with a wide range
of possibilities for this purpose are loops with natural and forced coolant circu-
lation (Figs. 1 and 2). The stands are used for prolonged tests at constant cool-
ant temperature and velocity.
Arrangements used for the study of thermal-flow dynamics which have test
sections of complex design equipped with a larger number of temperature, pres-
sure, flow rate, etc., pickups [5], are frequently made as multi-loop circuits
which, as a rule, operate in a wide flow-rate and temperature range (Fig. 3).
Mockups of fuel elements are tested in similar arrangements, where integrated
data on the specimen behavior can be obtained. To obtain the required number
of thermal cycles (thermal shocks) in a short period of time, use is made of
specially designed setups (Fig. 4). The fuel-element mockup is placed in a so-
dium-filled vessel, which was induction heated by the LA-107 generator and
water cooled; cycle duration is 41 secs.
We shall here consider only loops with forced coolant circulation (see Fig.
3), since they include all the equipment of simpler loops and, in addition, they
also have special equipment such as metal-to-metal high-temperature heat ex-
changers [6], etc. The majority of experiments was performed in setups with
capacities not above 250 liters and liquid-metal flow rates of not more than
20 m3hour. These setups should operate stably and reliably when the flow rate
is varied from zero to nominal. In setting up the thermal balance of the system
(nominal operating mode) one must know precisely the entire assortment of tub-
ing from which the circuit is made. Even insignificant deviations of diameters
have an appreciable effect on redistribution of temperatures and velocities of /360
the liquid metal over the entire circuit. Due to this it is always useful to have
an excess-capacity heater and higher-than-specified pressure head. The tend-
ency to reduce the loop volume and to provide for excess pump head capacity
has lead to the use of centrifugal pumps with n < 50 [7], however, it was found
[8] that the operating reliability of electromagnetic pumps is higher. As a rule,
the major part of the pressure head which is developed by the pump is used up
at the test section (Fig. 5). Hence, the heat-exchanger apparatus of experimental
360
setups is not the most efficient from the point of view
of compactness and of minimum volume of the alkaline
metal, as compared with, for example, slot-type heat
exchangers but has an acceptable flow resistance of
the ducts.
Auxiliary Systems
Preparing the heat-transfer agent. The sodium
is supplied from the plant in special 100 liter contain-
ers, and its oxygen content is up to 0.05% .by weight.
Depending on the test conditions, the sodium can be
used without distillation, as well as with preliminary /361
distillation. In the second case the metal is heated
until it melts and is poured into an evaporating tank.
Distillation takes place at 550-560C with a continu-
ously operating pre-evacuation pump. The sodium
vapor condenses in a condensing tank. The eutectic
sodium-potassium alloy is prepared in a special in-
stallation with a proportioning device [10].
The admixture content of the heat-transfer agent
is monitored by samplers indicators and other devices,
which are installed on the bypass lines of the loop. The
Figure 1. Schematic samples are taken by distilling samplers [11] or using
of a Natural- Circula- sections which are first frozen and then removed [for
tion Loop. analysis]. Satisfactory checking precision is provided
1) Vapor Trap; 2) Com- by the indication method, which is based on the stop-
pensating Tank; 3) Cool- ping up of narrow passages at a specific flow rate and
er; 4) Test Section; temperature.
51 Heater; 6) Drain Tank.
Oxygen and nitrogen are removed from the heat
transfer agent during operation by cold [13] and hot
[14] oxide traps. Under maximum conditions the ox-
-3
ygen content of the liquid metal is held to 1. 10 % by weight when using the cold
-4
and 1 . 10 % by weight when using the hot traps.
The gas-and-vacuum system is used for filling the setup with the liquid /362
metal, for maintaining the operating pressure, sampling and for testing the air-
tightness of the arrangement. Vacuuming is performed through a compensating
tank and a vapor trap (Fig. 6). The diameter of the pipeline connecting the trap
to the tank should be at least 70 mm.- The vapor trap is a pipe 120 mm in diam-
eter, with welded inclined baffles and a layer of stainless chips. In vacuum sys-
tems of setups with liquid-metal volumes of 50-500 liters use is made of tubes
30-40 mm in diameter and not more than 3 m long (from the header to the ves-
sel), and of the VN-1 and VN-2 pumps. Tubes 10-20 mm in diameter were used
in the gas systems.
361
Figure 2. Recuperator-Less Loop.
1) Drain Tank; 2) Pump; 3) Heater; 4) Test Section;
5) Cooler; 6) Vapor Trap; 7) Compensating Tank.
Instrumentation. Standard membrane-type manometers are used as pres-

sure pickups. Compensating pressure pickups have also been developed (Fig. 7).
The level is measured by contact level meters (pin-type, automotive spark plugs,
etc.,), which are based on closing an electric circuit by the liquid metal, or po-
tentiometer-type gages [16]. The temperature was measured by standard ther-
mocouples, as well as by microthermocouples.
Equipment Layout and Safety Measures

The experimental setups are located in bays or halls. The designer must
take into account convenience in operation, installation and dismantling of equip-
ment, location of instrumentation pickups, shutoff and regulating valving, plac-
ing preheaters and thermal insulation and safety requirements. Extensive pipe-
line length increases the liquid-metal volume, flow resistance and heat losses
to the surroundings. More rigorous layout requirements are presented to set-
ups with centrifugal pumps with frozen-sodium seals. In this case the pump is
placed on a platform at a height of 2-2.5 m. The pump seal should be at the el-
evation of the maximum liquid-metal level in the compensating tank, since [other-
wise] metal may spill into the premises or gas may be sucked into the pump
chamber. The compensating tank should contain enough metal to completely fill
the equipment situated above the maximum level. A drainless cold oxide trap
may be placed at any level. The distilling sampler should also be placed in a
safe and convenient place, since it has to be used during the setup's operation.
The still should be rigidly supported, so that, when the latter is disassembled,
the metal drops suspended from the screen will not drop into the cup and thus
distort the measurements. The oxide indicators and the measuring tank are
placed on the setup's platform. Heat exchangers, test section, coolers, heater,
362
Figure 3. Diagram of a Three-Loop Arrangement.
1) Drain Tank of First Loop; 2) Cooler of First Loop; 3) Drain
Tank of Third Loop; 4) Distilling Sampler; 5) Pump of First
Loop; 6) Compensating Tank of First Loop; 7) Recuperator of
First Loop; 8) Measuring Tank; 9) Compensating Tank of Third
Loop 10) Pump of Third Loop; 11) Cold Oxide Trap; 12) Recup-
erator of Second Loop; 13) Cold Oxide Trap; 14) Oxide indi-
cator; 15) Compensating Tank of Second Loop; 16) Pump of
Second Loop; 17) Distilling Sampler; 18) Cooler of Second
Loop; 19) Mixing Header; 20) Drain Tank; 21) Mixing Header;
22) Loop-to-Loop Heat Exchangers 23) Hot Oxide Trap;
24) Heater.
hot traps, etc., are placed at elevation eliminating formation of gas pockets and /363
providing for completer draining of metal. The drain tank is placed in the lowest
point of the loop. For convenience in operating the setup and to reduce its height,
the drain tank is usually placed in a pit (Fig. 8); the pipelines are inclined at3-5 0 .
Expensive equipment (transformers, vacuum pumps) should not be placed beneath
pipelines and equipment with liquid metal. The setup platform should be rigidly
fastened to permanent walls to eliminate vibrations. To protect the service per-
sonnel from possible accidents, the high-temperature sections should be placed
in inert-gas chambers. All the equipment beneath the setup platform is covered
by movable shields.
Pans should be placed at zero elevation to collect metal which may spill
during an accident and to localize a possible fire. Water may be used at alka-
line-metal setups only when absolutely necessary. The water pipelines are
laid below the platform and are insulated. The drain lines of the loop are main-
tained at a temperature higher than the melting temperature of the metal, and
are remotely controlled by electrically-driven valves. The remaining valves
363
To the
i 12-107
generator
E_
__._..._W ate r
Drain
Figure 4. Schematic Diagram of a Setup

for Thermal-Shock Testing of Specimens.
mil
1) valve; 2) specimen; 3) vessel with metal;
4) inductor; 5) insulation; 6) level gage and
thermocouples; 7) casing.
y ;.
1Water
1
1 a
Figure 5, Sections Through Test Sections

of a Setup,
1) casing; 2) spacer; 3) specimens.
are manually operated. The stand is supplied by a signalling and blocking sys-
tem. When metal leakage is discovered, the electric supply is shut off and the
valves of drain lines are opened. The main heater is also shut off automatically
when the temperature of the heat-transfer agent and supply busbars increases
above the permissible value, when there is no liquid metal flowing, no water in
the busbars and when the pump stops suddenly.
Accidents were previously caused by nichrome heaters, wound onto pipe-
lines and supplied with current at 220 V. When using small-diameter tubing
the porcelain buses [insulators] cracked and the heaters were short-circuited
364
Figure 7. Section Through a Compen-
sating Manometer.
1, 2 and 3) Liquid-Metal Inlets; 4) Fast-
I ening Plate; 5) Cover; 6) Bellows 7) Core;
Outlet
8) Casing; 9) Bellows Joint; 10) Connect-
Figure 6. Joining or Nut; 11) Spring; 12) Washer; 13) Rod;
a Gas-and Vacuum 14) Nut; 15) Spool; 16) Sleeve; 17) Con-
Line to the Loop. necting Pipe; 18; 19 and 20) Gas Inlets.
1) Compensating
Tank; 2) Level Gage; by the pipeline. Two methods are now
3) Vapor Trap. used to prevent this, one which consists
in using separating transformers, and
the other in reducing the voltage to ap-
proximately 48 V by type OSU-100/0.5
transformers. Accidents are also caused by centrifugal pumps with frozen-so-
dium shaft seals. The sodium in the space between the casing and the shaft is
frozen by process water. During operation bicarbonates are deposited on the
walls of the cooling jacket, which interferes with removal of heat and thus leads
to splatter of metal. [To remedy this the sodium-held] admixtures are period-
ically removed by a 20% solution of nitric acid. Accidents also happened when
the main heater and pump blocking system were not connected (the heater was
supplied with 15, 000 amps of current and, when the pump jammed, it burned
instantaneously), as well as due to defects in the construction materials, i. e.,
microcracks, micropores, pealing, etc. Accidents may also be produced by
manufacturing defects, such as low quality of jointwelding; usually these are
discovered during the run-in period.
The cost of developing an experimental loop for heat engineering studies /364
at maximum temperature of 500C can be broken up as follows: 1) design work
7%; 2) cost of construction materials 6%; 3) standard process equipment 18%;
365
Figure 8, Equipment Layout for a Three-Loop Setup,
1) Cooler; 2) Pump of First Loop; 3 and 8) Cold Oxide Traps; 4) Hot Oxide Trap; 5) Transformer; 6) Mixing
Header; 7) Compensating Tank; 9) Pump of Second Loop; 10) Cooler; 11) Heat Exchanger; 12) Heat; 13) Re-
cuperator; 14) Inter-Loop Heat Exchanger; 15) Test Section; 16) Drain Tank; 17) Cooler Heating System;
Vacuum Pump.
4 ) standard electrical equipment 12%; 5) instrumentation 12%; 6) making non-
standard process equipment 27%; 7) process equipment installation 7%; 8) elec-
trical equipment installation 6%; 9) adjustment of electrical equipment and of
instrumentation 5%.
REFERENCES
1. Balandin, Yu. F. and Markov, V. G.: Konstruktsionnyye materialy dlya

ustanovok s zhidkometallicheskim teplonositelem [Construction Materials
for Facilities Using Liquid-Metal Heat-Transfer Agents]. Sudpromizdat,
Leningrad, 1961.
2. Andreyev, P.A., et al.: Zhidkometallicheskiye teplonositeli yadernykh
reaktorov [Liquid-Metal Coolants of Nuclear Reactors]. Sudpromgiz,
Leningrad, 1959.
3. Zhidkometallicheskiye teplonositeli (natriy i natrievo-kalievyye splavy)
[Liquid-Metal Coolants (Sodium and Sodium-Potassium Alloys)]. Col-
lection of articles translated from English under the editorship of A. V.
Sheyndlin. Foreign Literature Publishing House, Moscow, 1953.
Metal Heat-Transfer Agents]. Atomizdat, Moscow, 1958.
5. Zhukov, A. V. , et al.: Part II of present collection.
6. Kurov, .V. A. and Menshikov, V. F.: Article of present collection immed-
iatel3lofollowing this.
7. Kuznetsov, V.A. , et al.: Atomnaya Energiya, 7, 11 (1959).
8. Marin, N. I., et al.: Present collection, Part IV.
9. Kays, W. M. and London, A. L.: Compact Heat Exchangers, The National
Press, Palo Alto, California, 1955.
10. Sittig, M.: Natriy [Sodium]. Gosatomizdat, Moscow, 1961. [Most likely
a translation].
11. Dmitrieva, I. B., et al.: Second article, Part IV of present collection.
12. Kozlov, F. A., et al.: Article following [11] in present collection.
13. Kozlov, F. A., et al.: Teplofizika vysokikh temperatur, 3, No. 1, 154 (1965).
14. Kozlov, F. A. and Kuznetsov, E. K.: Part IV of present collection.
15. Turchin, N. M. and Shumskiy, R. V. [sic].: Article following [6] in pres-
ent collection.
16. Kirillov, P. L., et al.: Atomnaya energiya, 9, 3 (1960).
367
HEAT EXCHANGERS OF EXPERIMENTAL LIQUID-METAL LOOPS
V.A. Kurov and V. F. Men'shikov
No systematized material on the above topic has been published in the Sov- /364
iet Union. In the publications available [1] and [2], primary attention is paid
to problems of heat transfer and practically no consideration is given to design
and operation of heat exchangers. In the present paper we are considering the
design and operation of heat exchangers for sodium and its alloys [3 ]. These are
primarily coolers (metal-to-water, metal-to-air), interloop heat exchangers, /365
recuperators 1 and mixing devices.
An efficient heat exchanger should satisfy the following requirements:
1) Reliable operation of facilities under all conditions (sometimes with a
changed heat-transfer surface) during the specified time of operation, sufficient
strength safety factor at the given rates of temperature variations (startup and
shutoffs);
2) inspection and contaminant removal in shortest time;
3) uniform distribution of heat-transfer agent flow (no stagnation zones,
gas pockets, etc. );
4) is provided with facilities for complete draining of heat transfer agents;
5) is accessible for the necessary quality control on manufacture.
The requirements produced by the physical and chemical properties of the
heat transfer agents are well known [4].
As a rule, the designer is limited by the above requirements. However,
among heat exchangers satisfying these requirements preference is given to
those with the highest efficiency and with lower power consumption for pumping
the liquid metal [5]. The above requirements are satisfied by conformance to
the following conditions.
1. The velocity of the liquid metal in the supply stubs is specified as
equal to (or smaller than) its velocity in the working section of the heat exchanger.
Otherwise it is necessary to make a tapered stub with a taper angle of 15 0 or /366_
to place equalization sections into the inlet chambers, to equalize the flow ve-
locity over the cross section.
1 The term recuperators is used here to designate heat exchangers, in the

cavities of which only one heat-transfer agent flows all the time.
368
2. When the heat transfer agent flows into the intertubular space perpen-
dicular to the tubes a baffle and uniform distribution of the liquid metal over the
inlet cross section are provided.
3. To prevent nonproductive flow of the heat transfer agent, spacers are
provided between the casing and the tube cluster.
4. To increase the rate of heat transfer, baffles are placed in the inter-
tubular space (for flow across the tubes).
5. In order to make the tube cluster more rigid (to eliminate vibrations),
the tubes are spaced by screens, which can serve simultaneously as baffles.
6. When the heat-transfer surfaces are coiled, comb-shaped spacers are
used.
In cooler shown in Fig. 1 is provided with a device for uniformly distri-
buting the flow over the tubes when the inlet angle of the chamber is more than
300 . Experience shows that without use of such devices heat transfer does not
take place over the entire surface. Table 1 lists types of heat exchangers.
Metal-to-Water Coolers
The use of water for cooling the above heat transfer agents involves op-
erating difficulties, but makes it possible to develop a small-size device. The
heat transfer surface always has two air-tight walls (made from tubes of differ-
ent diameter) and an adjustable intermediate space. The latter makes it pos- /368
sible to reduce the thermal stresses in the walls which are particularly high
when water is let into a heated up loop. To ensure acceptable heat transfer co-
efficients the intermediate space is filled by an interlay from an alloy such as
Wood's metal, graphite powder, gas, etc. The intactness of walls is monitored
by changes in pressure in the intermediate space and, in case of a gas interlayer,
also by closing of an electric circuit by the liquid metal which penetrates the wall.
It is practically impossible to adjust the capacity by varying the water flow /369
rate, since the outlet temperature is limited, hence, adjustment is obtained by
changing the surface of heat transfer by the following techniques:
1) Sectioning;
2) Moving the water-shell casing.
3) Changing the level of interlayer. This method is preferable, since it
makes it possible to obtain smooth heat removal in a wide range.
When making such a cooler, particular attention is paid to provisions for
checking the welded seams connecting the tubes to tube headers and of the lat-
ter to the casings. On the basis of operating experience we recommend the use
369
1 2 } d
Figure 1. Metal-to-Water Cooler.

1) Flow Equalization Section; 2) Inlet Chamber; 3) Tube Header;
4) Casing; 5) Bayonet Tube; 6) Connecting Stub of Interme-
diate Cavity; 7) Draining Stub.
TABLE 1. Heat Exchanger Types
Without intermediate layer /intermedi

oil- 2 2 Shell-a
Shell-and-tube
pe tube tube tube
U_ With With Ri Lend
-ularW,
'c
?dmOn^Sgb^1-
^= id compen-
sator,
+ ` + + + + + + +
2.M/G + I IM
^ M/W + + + +
D /M-Inter-loo p Met -to-metal heat exchan ger's and recu
perators
2) M/G-Metal -air cooler
31 M/W-Metal-water cooler
of heat exchangers shown in Fig. 2 (for a surface of < 1 m 2 ) and in Figs. 1 and
3 (for a surface of > 1 m2).
370
5
Figure'3. Metal-to-Water
Figure 2. Metal-to-Water Cooler.
Cooler. 1) Casing; 2) Jacket; 3) Shell;
1) Casing; 2) Support; 3) Metal- 4) Spacer; 5) Drain System;
Carrying Coil; 4) Level Gage; 6) Condenser Jacket; 7) Condenser;
5) Water-Carrying Coil. 8) Inlet (Outlet) Water Chamber.
Metal-to-Air Coolers
The main advantages of metal-to-air heat exchangers as compared with
water cooled apparatus is the ability of adjusting the heat removal within a wide
range and greater operating safety. They can also be modified to provide for
preheating. Their disadvantage is the large size. Use is made of coolers with
surfaces of < 3 m2 and >3m 2 (Fig. 4).
371
Inter-Loop Metal-to-Metal Heat
Exchangers and Recu erators
While the purpose served by
coolers and interloop heat exchang-
ers is obvious, the inclusion of re-
cuperators in loop arrangements re-
quires some clarification. Recup-
erators are thermal bridges be-
tween the hot and cold branches of
the same loop (Fig. 5), whichmake
it possible to reduce the power con-
sumption in the heater and to pro-
vide, for a given heat-transfer agent
flow rate, the specified maximum
and minimum temperatures in the
loop. If we disregard heat losses
in the loop, then
Nh=NhNr,
where Nh is the power consumption

in a recuperator-equipped loop, N
is the power consumed by the heater
in a recuperator-less loop and N
is the recuperator rating. The in- /370
clusion of a recuperator in a loop
reduces the required cooler sur-
face. Coolers are structurally
Figure 4. Metal-to-Air Cooler. more complicated than recupera-
1) Jacket; 2) Coils; 3) Spacing System; tors and, in addition, they have a
4) Oultlet Header; 5) Inlet Header. lower heat-transfer coefficient.
The apparent increase complication
of the loop layout [actually] results
in increasing the total heat-transfer
surface and increases the safety in
operating the circuit.
The mean temperature difference between the cold and hot branches of
the recuperator, which is determined by the heater, may be as high as 200C
(heating to 500C) for steady-state operation. The high temperature levels and
large mean temperature differences require close attention to finishing of
structural components which are related to uniform distribution of flows over
the cross section and to compensation of thermal expansion. To protect the
tubes from the effect of transverse flow, the section of the tube cluster ahead
of the inlet stub is protected by a baffling device (Fig. 6), which serves as a
continuation of the casing.
372
Figure 6. Baffling Device
Which is a Continuation of
the Casing.
Figure 5. Sche- In recuperators (Fig. 7) the ring serves as con- /371

matic Diagram tinuation of spacers of all shapes, and also serves for
of a Recuperator- distributing the flow uniformly over the cross section
Equipped Loop, of the intertubular space.
1) Pump; 2) Heat- The tubes are connected to the tube lattice by
er; 3) Working welding only. Tubes with wall thicknesses of more
Section; 4) Re- than 1 mm and with a diameter of up to 30 mm are
cuperator; connected by flashing off the edge of the hole in the
5) Cooler. tube header, while when the wall thickness is less
than 1 mm, this is done by flashing off the "whisker"
in the tube header. Both methods produce a strong
and tight joint when the tubes are made to conform to the hole diameters. In
both cases it is possible to have high tube cluster densities, i. e., spacing of
1.2 d and even less.
The thermal stresses are reduced by creating a stagnant zone of the heat-
transfer agent, using a screen, while in tubes this is obtained by lining and
covering (see Fig. 1). The difference in thermal expansions in the straight
tubes and in the casing was compensated by placing several bellows on the tube
cluster (see Fig. 7 and Table 1). To compensate for thermal expansions of dif-
ferent (not welded together) construction materials in recuperators, the tube /373
cluster was placed into the casing on packing rings. Leakages between differ-
ent spaces comprised less than 1% of the heat-transfer agent flow rate.
Nonconformance to the above requirements results in breakdowns. For
example, the recuperator shown in Fig. 8 failed as a result of improper flow
distribution between the tubes.
373
Rigid-design recuperators are usually made in the manner shown in Fig. 9.
Use is also made of recuperators with compensation on the heat-transfer surface
of less than 0 5 m2 (Fig 10) and more than 0.5 m2 (see Fig. 7 and Table 2).
1 2 3 4 5 6
Figure 7. Schematic Diagram of a Recuperator With

a Compensating Bellows.
1) Bellows; 2) Movable Chamber; 3) Tube; 4) Spacer;
5) Casing; 6) Chamber.
Figure 8. Recuperator After Failing,
Mixers are used in loops for:
1) Heating the entire volume of the heat transfer agent to the specified
temperature, provided that only a part of the flow, which is heated to a higher
temperature, passes through the heater.
374

2) Reducing the heat-transfer agent temperature ahead of the recuperator.

We have used tubular-type and double-tube type recuperators (Figs. 11 and 12).
TABLE 2. Fitness of Materials of Construction at

Different Temperatures
em era- Heat- Oxygen .I Time of

Material turep C transfer Velocity, content, lOperation State
m/sec
' agent hours
1Kh18N10T I 650 I Na I 2 2.10- :3 I 4500I Continues to

operate
1Kh18N10T I G00 ( K 2 ( - 2000 I Same as above

I
iKh18N9T ( 250 I tiaK I 7000 I I

Same as above
I 350
Figure 9. Schematic of a Recuperator.

1) Chamber; 2) Tube-Header Chamber; 3) Casing;
4) Spacer; 5) Tube; 6) Baffling Device.
Figure 10. Schematic

of a Recuperator,
1) Chamber; 2) Casing;
3) Star-Shaped Spacer;
I ^^
4) Tube.
375
Figure 11. Schematic of a Tubular-Type Mixer,
1) Chamber; 2) Tube Header; 3) Casing;
4) Tube; 5) Chamber.
.J
Figure 12. Schematic of a Double-Tube Mixer.

1) Cap; 2) Tube; 3) Casing 4) Ring; 5) Connecting Stub.
REFERENCES
1. Kutateladze, S. S. , et al.: Zhidkometallicheskiye teplonositeli [Liquid-

Metal Heat-Transfer Agents]. Atomizdat, Moscow, 1958.
Collection of articles translated from English under the editorship of
A. V. Sheyndling. Foreign Literature Publishing House. Moscow, 1953.
3. Drobyshev, A. V. , et al.: Article immediately preceding this one in the
present collection.
4. Chechetkin, A. V.: Vysokotemperaturnyye teplonositeli [High-Temperature
Heat-Transfer Agents]. Gosenergoizdat, Moscow, 1962.
5. Andreyev, M. M., et al.: Teploobmennaya apparatura energicheskikh usta-
novok [Heat-Transfer Devices of Power Facilities]. Mashgiz, Moscow, 1963.
376
STUDY OF VELOCITY FIELDS BY THE ELECTROMAGNETIC METHOD
N. N. Turchin and R. V. Shumski.y
Processes of heat transfer are closely related to the fluid dynamics of the /374
heat transfer agent. The velocity fields are studied experimentally primarily by
means of Pitot tubes and of thermoanemometers. At present these devices can-
not be used in liquid metals. The high electrical conductivity of the latter makes
it [however] possible to determine local velocities by measuring the electric po-
tentials which are generated when the liquid moves in a magnetic field. The fea-
sibility of using the electromagnetic method is clarified in [1-3], which consider
a round tube for which the distribution of potentials in axisymmetric flow was ob-
tained analytically. The experimental data are in good agreement with the calcu-
lations.
The present article also presents results of experiments with a round tube,
but it considers a more general case of the motion of a fluid in the annulus formed
by two coaxial tubes with an arbitrary [thermal] conductivity of the walls. It is
assumed that the velocity distribution is axisymmetric [v r = v0 =0, v =V (r)],
z
and that it does not change in a homogeneous magnetic field which is sufficiently /375
large so that edge effects can be disregarded; the secondary induced field is neg-
ligible. Determination of the potential in some point M (Fig. 1) reduces to solv-
ing the following system of differential equations (the subscripts denote the re-
gion for which the equations are written)
C ta Ui == 0, (la)
C"Ua 8 0 - silt U, (lb)

jr
The general form of solution for each of the above is
U 1 __ (A lr+C lr -1 ) sin , (2a)
U2
i Bor-1 1 sill ,
ry
r2 (2b)
U3 = ( A 3r Car-1 ) siii E). (2C)
377

Using the boundary conditions ( equality of potentials and equality of the normal
components of the flux at the interface) and solving simultaneously Eqs. (2), we
can determine the coefficients of these equations. Then the potential field in the
fluid will be expressed as
^1
Un : - B O Sill 0 r -, Ur (11 ;- t o
v 1 '2
-1) r
L l d', ( D3 +1)(1D)
D3
ry r ^
Uo D D . r
i (3)
2 1
,
where
() (D l - 1 ) 1 -IL (12/d2 .
D= (D3
d^, 62
D3 --
Ai ( D3 I- 1 ) (Di-i-1)' v3 1 d3 /d;
v0 is the average flow velocity and v is the electrical conductivity of the medium.
The flow rate is determined by the total potential difference measured at the outer
surface of tubes between points a, and b with coordinates r = r 4 , O = 7r/2
2 d3 d- d2
6U3 B ovo (c1 3 -- d2) (4)
d3 6g
a4 (1;
d4/-^ __L 4
or points a', b' with coordinates r = r l, O = f 7r/2
d l 1. } 3
SU 1 _.= Bovo(d3t12) d3 l
ul ( (5)
die) v2 1 dz l ] 1 nD '
1 i di di
It is easy to show that, when d 2 = 0, Eq. (4) becomes identical with the expres- /376
sion for a round tube presented in [4].
It is also of interest to present a formula for the flow resistance. In the
case of a passage with conducting walls, electromagnetic forces appear in the
fluid and they slow down the motion. The addition flow resistance which is thus
produced is equal to [5]
grad P = Kcr2voBa,
378
where the deceleration coefficient for an annular space
(D3-1)-^_^_ ^3 (DI
A,
K = (D3 Ir,
.' 1)(D3-1)-i
Bi 7z 1 i (1 2/d2
vi 1 di/d2
From Eq. (3) we [now] can get an expression for the potential field in a round
pipe in the form
ll2 =_ Bo sill U (v r ;- Cvo) r; (6)
here v is equal to the average velocity in a passage cross section of radius r. /377
The value of C varied from + 1 when o-w = 0 to -1 for Qw
When the tube
size is selected properly, C may become zero (when (T fl = 1 + d3/d4/1 -
- d3 /d4).
Thus the potential at each point is not determined by the local velocity, but
rather by the average velocity in a cross section whose radius is equal to the cur-
rent radius of the point proper. The potential will determine the local velocity
only in a homogeneous velocity distribution. In other cases the local velocities
can be determined on the basis of the diagram of potentials, using Eq. (3). Fig-
ure 2 shows such a diagram for different values of C and Re. In turbulent flow
the potential diagrams [curves] lie close to one another and when the Reynolds
number is varied from - to 10 4 the signal increases on the average by 10-20% /378
depending on the value of C.
The quantity which depends more substantially on the velocity is the poten-
tial gradient (Fig. 3). Differentiating Eq. (6) over the radius, we get
d1', t -
lip Sill (7) [V, V, C 1,01, (7)
where v is the local velocity. The probe for measuring grad U should be pro-
vided with two electrodes situated at a distance of dr << r 0 from one another. Us-
ing this kind of probe it is possible to simultaneously measure the potential dia-
gram and the gradient. Then it is easy to get from Eqs. (6) and (7) that
? _= Bo siii (^ C L
TZ -r- `d V, CVO (8)
379
Reference [3]presents results of
experiments for measuring the veloc-
ity distribution by both methods. The
electrode was a metal capillary, in-
side which passed an insulated wire,
welded at its end to the capillary. In
this case it is impossible to say with
certainty that the signal measured is
equal to the potential in the fluid point
in contact with the junction. The ve-
locity distribution in a flow of the so-
dium-potassium alloy, flowing in a
stainless steel tube with an inside di-
ameter of 36 mm, was measured by a
probe with two molybdenum wires 0.2
mm in diameter. The electrodes were
soldered to the probe by means of a
glass end piece; the distance between
i 2 them was 0.9 mm (Fig. 4). The mag-
net can be moved along the tube, which
Figure 1. Geometry of Passage. makes it possible to measure the po-
t) Region of the Inner-Tube Wall; tential distribution in the cross section
2) Fluid Region; 3) Region of the at the tube elevation and along the mag-
netic field (Figs. 5 and 6). The poten-
Outer- Tube Wall. tial diagrams obtained in a homogene-
ous magnetic field at Re =7.8.103 , are
shown in Fig. 6b. The experimental
points lie between the curves for the above Reynolds number and for a Reynolds
number going to infinity, respectively. It is difficult to determine the Reynolds
number to which the experimentally obtained potential diagram corresponds. As
can be seen from the figure, the straight line about which the experimental points
are clustered has ahigher slope than the calculated curve for Re = 7.8 . 103 . It can
be concluded from this that either the velocity distribution in the magnetic field /379
has become less steep, or that this effect is due to measurement errors. The
velocity distribution should become reconstructed after entering the magnetic
field by a certain distance.
In the given experiment the potential curves measured in a homogeneous
magnetic field at a distance of 10 r 0 from one another were found to be the same
(the data obtained in the plane with I = 540 are not shown in Fig. 6). The curves
along the magnetic field obtained in other experiments performed for a range of
Reynolds numbers from 3.76 . 103 to 3.04 . 104 and in a range of magnetohydro-
dynamic interaction parameters from 0.36 to 148. Apparently, the flattening
in the velocity distribution, if any, was of the order of magnitude of the experi-
mental error. In addition, as was shown above, the potential curve depends little
on the Reynolds number of turbulent flow. Hence, it is practically very difficult
to use it for determining the shape of the velocity distribution.
380
Figure 2. Diagrams of Potentials
7 in the Fluid Flow.
---Laminar Flow; -. -. -Re = 7.5.103;
-... - Re = 7.5 . 104 ; Re = oo .
/^ ;r
0 t 0.5 0.75
r1r,
1.5
7.25
N
Figure 3. Diagram of the Poten-
0. 7.5 tial Gradient in the Flow of Fluid.
The Solid Line Pertains to Lami-
nar Flow; the Dash-Dot Line is
0.5 for Re = 7.5 . 103 , and the Dashed
Line is for Re = 7.5. 10 4
0.25
381
Unit !!
Figure 4. Design of Probe.

1 and 2) Micrometers; 3) Microscrew; 4) Bellows;
5) Centering Plane; 6) Spring; B) Position of Probe
in Tube Relative to the External Electrodes.
6U
^n ----^- --:iii'' : o.w_
-$m - _;+_ - -- - - ----'
0,75
0.5
0.25 -- 3
0
200 300 400 500 600 700 l
Figure 5. Diagrams for the Potentials 6- U Along a

Tube (Re =2.5- 104 , M = 786).
. and + -outside of the Tube at Electrodes 1-1' and
3-3 i , Respectively; o and A - in the Fluid at a Distance
r/r0 = 0.445 and 0, 885, Respectively; 1) Magnetic In-
duction B(I)/B 0 ; 2) Curve of SU = B(I)r/B0r0.
382

I O .1_2sp -70
o \ +-1=?10 "40
'i\+0
0
i a^
J . 0.4 C.3 02 0,1'1
t ^ I I ,
\ I
Figure 6. Curves of Potentials Over the Height of the Tube

(Re = 7. 8 . 10 3 , M = 577).
a) At the End Section of the Magnetic Field; b) in the Homogeneous Field; c) at the
Starting Section of the Magnetic Field; ---, -. - and -.. - are Curves Calculated
for Re = -, Re = 7.8 . 103 and for Laminar Flow, Respectively.
We did not succeed in experimentally obtaining a relationship which would /381

reliably describe the change in grad U over the tube height, using two electrodes
to measure the signal. The experimental data were highly scattered. Changes
in the effective distance between the electrodes were also observed. If we de-
note the latter by Or, the signal measured at the electrodes by AU and the total
signal which measured outside of the tube by SU, we can show that for a linear
potential curve
AU
-- 0,
Ar --- nt7' o,
where m is a constant, depending on the tube parameters. The ration AU/5U

was measured with the probe in the same position (it was in the center), vary-
ing either the induction while holding the velocity constant, or conversely, vary-
ing the velocity at constant induction. It was found that in the first case Ar re-
mained almost constant, while in the second case it diminished (table).
The potential difference between electrodes 2-2" (see Fig. 4) is calculated
from Eq. (4) for an annulus. According to calculations, the value of the signal /382
here should have been by 1. 5% higher than between electrodes 3-3', which was
verified experimentally.
It was noted in [6] that the edge effects practically do not affect the mag-
nitude of the signal at the magnetic field boundaries. It can be seen from Fig. 5
383
that also in the present experiments the potential field at the outer wall of the
tube, as well as in the fluid flow proper follows changes in the magnetic field.
Values of Ar
Po z -0,357m/ sec= const
R, gauss 1300 2370 4670 5820 5830 5985 6115 7520 7650 8120
6U, mV 1.5 2.65 5.15 5.76 6.41 6.62 7.05 8.25 8.42 9.2
AU, mkV 17.5 33.5 71.5 86 86 88.5 91 114 114 125
Ar, mm I 0.42 0.455 0.5 0.54 0.485 0.48 0.465 0.5 0.49 0.49
73 5780 gausses = const
vo, m/sect 0.121 0.263 0.56 0.667 0.85 1.03 1.09 1.25
SU, mV I 2.16 4.68 9.05 11.87 15.2 18.28 19,55 22.2
AU, MW I 8.5 32 77 110 137 215 200 230
- - Ar,^mm I 0.141 0.246 0.306 ^0.334^-0.;1 4, 0_.41 0.37 0.373
REFERENCES
1. Kolin, A. J.: Appl. Phys, , 15, 151 (1944),

2. Williams, E.: Proc. Phys. Soc., A, 42, 469 ( 1930).
3. Gushchin, G. I., et al.: Voprosy magnitnoy gidrodynamiki [Problems of
Magnetohydrodynamics]. Published by the Academy of Sciences of the
Latvian SSR, Riga, 1963.
4. Turchin, N.M.: Teplofizika vysokikh temperatur, 1, 118 (1963),
5, Turchin, N.M.: Magnitnaya gidrodinamika, 1, 147 ( 1965),
384
ELECTROMAGNETIC A-C PUMP FOR LIQUID METALS
RATED AT 120 m3/hour
Ye. M. Avilova, T. V. Doktorova, N. I. Marin
and V.A. Povsten'
The development of a pump for liquid metals with an output of 120 m 3/hour /382
completes the task of designing a series of electromagnetic volute pumps, work
on which was performed for the last several years in the Physics and Power
Institute. 1 The design of this pump is generally similar to that of other pumps
of this series, but differs by a number of features (Fig. 1).
The stator consists of the casing with a stack of iron, assembled from 0.5
mm thick electrical steel sheets, pressed into it. The iron stack has 72 slots
into which are placed a three-phase winding with organic silicon insulation. The
heat is removed from the stator by water flowing in a jacket welded to the cas-
ing. The cooled wedge-shaped copper inserts limit the thermal flow at the pro-
jecting parts of the winding from the liquid-metal side.
The triple-start screw-thread shaped passage for the liquid metal is form-
ed by two welded shells, to the outer of which were welded ribs along a spiral.
The inner shell protects the armature from the liquid metal and serves to close
the magnetic circuit. Guiding devices, consisting of tapered sections and ribs, /383
welded to them along a spiral, are welded to the ends of the shells. The spiral
angle may vary from 30 0 to 900 . These devices serve to smoothly change the
velocity and direction of the liquid metal.
The thermal resistance between the stator and the liquid metal is produced
by an air gap, in which are placed very thin corrugated metal screens, which re-
duce the thermal radiation.
Unlike other pumps of the series, this one is made with an airtight casing, /384
which prevents the liquid metal from penetrating the premises on failure of the
1.5 mm thick outer shell.
The pump was tested with sodium in order to refine its parameters, as
well as to expose and eliminate design shortcomings. The testing setup consist-
ed of a circulating loop, the flow rate of sodium in which was measured by a mag-
netic flow meter, connected to the PSRI-01 secondary instrument, the pressure
was measured by the MEM-3 manometric pickup, connected to the EPVI-14 sec-
ondary instrument, the temperature of the stator winding was determined by
chromel-kopel thermocouples, placed in a capillary from 1Kh18N9T steel
1Avilova, Ye. M., et al. Magnitnaya gidrodinamika, 1, 110 (1965).
385
0.8 x 0. 1 mm in diameter. The
pump was thus operated for 2500
hours with sodium at 330C; the
winding temperature here did not
exceed 130C. The permissible
winding temperature for prolonged
operation is 180C. The design
delivery of the pump is 150 m 3 /hour
in which case the sodium flows
through the pump at 10 m/sec,
The tests showed that the pump
operates noiselessly and stably
at deliveries of 40-120 m3/hour.
When the output is increased above
120 m3/hour, i. e. , when the so-
dium velocity is increased to 8 m/
/sec, the pump becomes noisywith
the noise increasing with a further
increase in the delivery rate; the
noise intensity could be reduced
only by increasing the argon pres-
sure in the expansion tank of the
loop. The pump was operated at
its top design speed only for a
short time, with the argon pres-
sure maintained at 0.9-1 kg/cm 2,
but even this could not entirely
eliminate the noise. Prolonged
operation with intensive noise is
undesirable, since the noise is
apparently produced by vibration
Figure 1. General View of Pump. of the thin walls of the metal pas-
sages. This operating mode is /385
1 and 9) Guiding Devices; 2) Screen; 3 and regarded as difficult. The noise
8) Wedge-Shaped Copper Inserts; 4) Arma- can be produced by separation of
ture; 5 and 6) Passage-Forming Shells; the metal stream at the pump in-
7) Stator. let. Consequently, it should be
assumed that 120 m3/hour is the
top delivery rate for this pump.
The pump was also unstable at deliveries of 30-40 m 3/hour. The flow var-
iables of the pump's operation fluctuated. The instability of operation at these
flow rates and, consequently, at high .slip, depends on the instability of electric
contact resistances between the spiral ribs of the passage and the liquid metal,
the effect of which increases with an increase in the slip. When the pump is
first started, and the liquid metal does not as yet sufficiently wet the ribs, an
increase in the electric contact resistance in fact reduces the head. However,
the resistance is reduced appreciably after as little as several hours of opera-
tion. The head developed by the pump increases and remains stable on further
operation (Fig. 2).
386
.atrr,
,.a. ,'hr 5"0 700 :', m3/hr

b
Figure 2. The Pressure (a) and Efficiency (b) as a Func-

tion of the Delivery Rate for Different Applied Voltages
and for a Sodium Temperature of 720C.
As can be seen from Fig. 2a, for a given applied voltage, a reduction in
the liquid-metal flow rate is accompanied by an increase in the head. However,
this also increases the stator current, which, as
is known, depends on the applied voltage, slip
(flow rate) and the electrical conductivity of the
liquid metal. Due to limitations imposed by heat-
ing of the windings, the current cannot exceed 250
amps, hence, the pump can operate with maximum
- - output (Q x H)max at different flow rates only when
it is supplied through a voltage regulator. Figure
sco 3 shows the dependence of the output on the flow
o rate at maximum current of 250 amps and at450C.
This characteristic makes it possible to establish /386
optimal operating mode for the pump. For the a-
a 400 --- - bove sodium temperature the pump has the great-
est output for a flow rate of 115 m 3 /hour, the head
then being 5.2 kg/cm 2 . The power consumption
200 at maximum output is 97 kw, the voltage is 350V,
50 100 Q,M 3 /h, efficiency 17.5%a and the power factor is 0.64.
Figure 3. Delivery as a The fact that the characteristics depend on
Function of Sodium Flow the sodium temperature is due to changes in its
Rate at 250 amps and So- electrical conductivity. When the latter increase,
dium Temperature of 720 the stator current must be limited by lowering the
Delivery and output here are head x flow rate of metal.
387
50 100 0, m3/hr
Figure 4. Head as a Function
of the Flow Rate at 250 amps
and Different Sodium
Temperatures.
supplied voltage which, for the same

flow rate, reduces the head, since
the induction of the magnetic field Figure 5, General View of the Pump
drops (Fig. 4). At present the pump Operating at a Thermophysical Stand..
operates successfully at a thermo-
physical stand (Fig. 5), circulating
sodium at 200-250C. The sodium
flow rate at this stand does not exceed 120 m3/hour, It can be regulated smooth-
ly by varying the voltage supplied to the pump.
388
ELECTROMAGNETIC SCREW PUMP FOR PUMPING OF LIQUID METALS
Ye, M. Avilova, Ye. L. Ivanter, N. I. Marin and V. A. Povsten'
Ordinary d-c pumps must be supplied from current sources of 2000-10,000

amps and above, which is very inconvenient under laboratory conditions and hence,
limits their use. [On the other hand] d-c pumps must be used in studies where
fluctuations in the liquid-metal flow rates are not permissible. Since laboratory
setups usually require only moderate metal flow rates, great promise is held by
d-c electromagnetic pumps with a screw-shaped flow passage which, when built
with a multiwinding cross section can be built to use only moderate amounts of
current (100-200 amps). These pumps can be supplied from moderately-sized
standard rectifiers. The flow rate of metal through such a pump is 2 cm3/sec,
the pressure when pumping sodium and using 200 amps of current is 1.3 kg/cm2
and the working temperature of the liquid metal may be as high as 500C (Fig. 1).
The pressure developed by the pump varies somewhat depending on the kind
of metal pumped. At low flow rates it depends primarily on the electrical con-
ductivity of the liquid metal. The higher this conductivity, the higher the pres-
sure produced. The magnetic flux produced by the exciter winding penetrates
the liquid metal flowing through the pump. The d-c current passed through the
liquid metal along the pump axis interacts with the magnetic field of the exciter
winding. As a result the liquid metal is set into rotational motion which is trans-
formed into translation by the spiral passage. Here the pump delivery pressure
is equal to the sum of pressures produced in the individual windings of the pas-
sage. The magnetic field induced by the current flowing through the liquid metal
is compensated by a return conductor, made in the form of a cylinder from St.
3 steel. It is kept in electric contact with the working section along a tapered
surface using a pressing flange. Due to the moderate current use, the working
part of the pump is made as a single entity with the current supplying conduc-
tors. It consists of two coaxial stainless-steel shells. The inner shell has cut
into it a spiral passage for the liquid metal with a cross section of 4 x 4 mm2,
the passage wall thickness is 0.8 mm and the passage has 15 windings. Small
air gaps serve as electric insulation between the working part and the return
conductor and magnetic system. The magnetic system is formed by the outer
cylindrical magnetic circuit (St. 3), inner core, lower flange and gasket, the
latter used for pump-assembly purposes. The total nonmagnetic gap of the sys- /388
tem is 10 mm.
The pump has independent excitation (64 V, 1 amp), which ensures smooth
pressure adjustment by varying the magnetic flux, the exciter winding is a tapped
coil, made from PNSDK high-temperature conductor wire 0.83/1. 08 in diameter.
If the pump must be preheated, the [tapped] sections are connected parallel to a
standard [50 cycle] power source. Heating here takes place by eddy currents in /389
the magnetic system and walls of the flow part.
389
c
c'.
Figure 1. Design of an Electromagnetic Volute Pump.

1) Terminal Block; 2) Exciter Winding; 3) Magnetic Sys-
tem; 4) Pressing Flange; 5) Return Conductor; 6) Heater;
7) Flow Section; 8) Current Buses.
The 200 watt flow section heater, which is a nichrome-wire spiral placed
in a ceramic holder, is used to compensate for the temperature reduction of the
liquid metal flowing through the pump.
The pump was tested at a stand using an In-Ga-Sn alloy with specific weight
of 6.5 grams/ cm3 and electric resistivity of 30 . 10-6 ohm-cm. The relationships
shown in Figs. 2 and 3 were obtained at liquid metal temperatures of 20C,
After testing the pump was installed in a small loop in which it circulated
an alloy of the same composition.
390
f/,atm
x
0
x ^^s
U
e t x
0 0
X xx
0.5
4 x X
0 100 200 300 ,7,1
Figure 2. Pressure Developed by the

Pump as a Function of the Current Flow I
in Flow Section at Zero Flow Rate and 10 Q, cm3/sec
With Constant Excitation.
Figure 3. Pressure Developed by the
Pump as a Function of the Flow Rate
for Different Current Consumption at
It should be remembered when op- the Flow Section and With Constant
erating the pump that the liquid-metal Excitation.
loop to which the pump is connected, is
held under the operating voltage of the
latter, which does not exceed 1 V. The
current dissipated in a long loop with a small diameter is low. However, in the /390
case of large diameters and short loops, the flow section of the pump should be
made with an electric tap in the center, which will divide it into two parallel
branches. Here the ends of the loop will be at approximately the same potentials
and thus current leakage into the loop will be prevented. In addition, when pump-
ing liquid metals at temperatures of 700-800C, to facilitate electrical insulation,
it is easy to make the pump with series excitation. The copper winding of the ex-
citer can be protected by an inert gas, as well as by chromium plating its surface.
The successful operation of this pump, feasibility of usage for pumping met-
als at 700-800C, the moderate current consumption and the simplicity of design
and manufacture show that a d-c volute pump is a convenient device for pumping
liquid metals under laboratory conditions.
391
HYDROSTATIC BEARINGS IN SODIUM PUMPS
N. N. Aristarkhov, Yu. Ye. Bagdasarov, Yu. L. Kulikov,
Ye. N. Larin, M. S. Pinkhasik, A. A. Rineyskiy
and B.S. Tymosh
A major problem in designing a sodium pump is the selection of the pump /390
bearing. Although it is pointed out in [1] that in a cold-trap equipped loop it is
possible to use ordinary roller and ball bearings, such pumps are not available
at present. This may be in part due to difficulties in selecting materials, as
well as due to the fact that the liquid metal, despite the use of filters, may con-
tain a large quantity of oxides, which has a detrimental effect on the operation of
rolling-contact bearings.
Hydrodynamic as well as hydrostatic bearings can be used for work with so-
dium. The difference between them consists in the manner in which pressure is
obtained in the carrying layer of the fluid. In a hydrodynamic bearing the lift in
the layer dividing the rubbing surface is produced by the fluid captured by the ro-
tating shaft and delivered into the radial clearance. The required shaft lift is
produced on certain combinations of the clearance size, shaft rpm and fluid vis-
cosity. Due to the low viscosity of sodium, the hydrodynamic bearings must be
made with very small clearances, and if the sodium contains solid particles (for /391
example, oxides), then this may result in failure of the bearings, which was ac-
tually observed [2].
In the hydrostatic bearing the external load is produced by the fluid pres-
sure in the bearing layer and may be made independent of the shaft rpm by sup-
plying the fluid by an auxiliary pump. These bearings are extensively used in the
chemical industry [3]. They also operate in fluids with poor lubricating proper-
ties and containing solid particles. The clearance of such a bearing is selected
upon due consideration of the size of the solid particles and the expected opera-
tional eccentricity of the pump shaft. By the method of fluid supply to the bearing
chambers hydrostatic bearings are divided into throttle- and slot-type. The op-
erating principle for any of these two types consists in the following: when the
shaft is deflected by unbalanced forces, the pressure in the chambers to which /392
the shaft has come closer increases and that in the oppositely-located chambers
decreases, thus producing a restoring force which returns the shaft to the initial
position. Thus, the rubbing surfaces in these bearings are separated by the lay-
er of fluid. The operating principle of an elementary hydrostatic bearing is shown
in Fig. 1.
Attempts due to use a throttling hydrostatic bearing in sodium pumps did
not produce positive results. The throttling hole' s were plugged up by oxides
and the bearing just became a simple sliding-contact bearing. After obtaining
negative results with three throttling hydrostatic bearings, we have decided that /393
they should not be used, at least in moderate-delivery pumps (up to 200 m3/hour).
392
A slot-type hydrostatic bear-
ing is shown in Fig. 2. The fluid
from the pump flows through slots
between the shaft and the sleeve in-
to holes, from which it is directed
into oppositely-situated chambers
through spiral passages. The pres-
sure in the bearing's chambers var-
ies according to changes in the inlet
and outlet clearances. Thus, for
example, when the shaft moves to
the right, the pressure in the right
chamber increases not only due to
reduction in clearance 7, at the out-
let of this chamber, but also due to
increase in the size of clearance 1
at the inlet of this chamber. Here
the pressure in the left carrying
chamber is reduced with an increase
- in clearance 3 at its outlet and
3 to reduction in clearance 9 at the in-
let into the opposite opening. Since
i during operation the clearances at
the inlet and outlet increase simul-
tanously, the relationship between
throttling of the fluid at the inlet and
Figure 1. Schematic Diagram of a Throt- outlet, as well as the carrying capa-
tling Hydrostatic Bearing, Where 1, 2, city of the bearing, remain unchanged,
3 and 4 are the Lift-Producing Chambers. Consequently, the rate of flow of the
fluid to the bearing will increase with
increased wear of the journals. Di-
rect flow of fluid between the carry-
ing and regulating chambers is prevented by an annular groove, which is located
between these chambers and which is connected to the low-pressure chamber by
special holes. The slot-type bearing has still another property for operationwith
alkaline metals. Unlike the throttling bearing, almost the entire pressure drop
in the slot-type bearing takes place in the clearance between the stationary sleeve
of the bearing and the rotating sleeve, which appreciably reduces the danger of
stopping up this clearance by oxides. The flow cross sections of holes for inter-
nal supply of the liquid metal to the carrying chambers may be quite large. The
conditions of operation of the rubbing pairs in liquid metals are those of a com-
bined effect of the load, high temperature and medium. When the sodium contains
oxygen, metallic compounds may form with materials which can result in failure
of the surface layer.
The operating conditions of a hydrostatic bearing become much more com-
plicated when sodium is supplied to it from the delivery line of its pump. In this
case on starting, stopping and when operating at low pump rpm, the bearing will
operate with dry or semidry friction. This circumstance presents strict require-
ments to materials from which the rubbing pairs of bearings are made. At pres-
ent there is no single approach to selection of these materials, due to lack of
393
Sec. AA
sufficient experimental data. Some
preliminary data on the operating
endurance of rubbing-pair materials
were obtained using laboratory pumps
[2]. It was established that the serv-
ice life of an RF-1 steel - beryllium
bronze pair is reduced sharply when
the temperature of the pumped metal
is increased to 400C. Cast iron
paired with RF-1 steel failed at tem-
peratures of 200-400 C. A pair
made of RF-1 steel - EI-229 steel
operated for about 300 hours at liq- /394
uid-metal temperature of 400C.
Some recommendations on the selec-
tion of rubbing-pair materials are
presented in [4].
Below are presented experi-
mentally obtained data on the oper-
ating endurance of a hydrostatic
bearing with self-reversing slot
throttling for different oxygen con-
tents of the sodium, shaft rpm and
sodium temperatures (table). The
Figure 2. Schematic Diagram of a Hy- experiments were performed at an
drostatic Bearing With Mutually Rever- experimental setup (Fig. 3).
sible Slot Throttling.
1 and 9) Sodium Passages of Bearings; The hydrostatic bearing was
2 and 8) Radial Sodium-Supply Passages; placed on a pump (Fig. 4), which
3 and 7) Discharge Clearances; 4) Spiral was driven by an electric motor
Passages for Leading Sodium in the Carry- rated at 36 kW. The motor has three
ing Chambers; 5) Carrying Chambers of Findings, thus providing three steps
Bearing; 6) Pump Shaft. of rpm adjustment. The pump and
motor shafts were coupled by a slid-
ing adapter, while the torque is
transmitted by a spline. The com-
posite shaft thus produced has two ball bearings and one (lower) hydrostatic bear-
ing, with the liquid sodium supplied to it from the delivery line of the pump along
a pipe provided by a magnetic flowmeter.
The design of the bearing is shown schematically in Fig. 2. It is made
from Kh18 steel, the shaft is made from 4Kh13 steel with a nominal shaft diam-
eter of 55 mm, bearing clearance of 0.35 mm and 12 carrying chambers. The
bearing and shaft were heat treated as follows: heating to 650C, hardening with
final heating in salt bath to 1020C and oil-bath cooling to 290C, temper harden-
ing at 290C and air cooling. The hardness of working surfaces after heat treat-
ment was 52-55 RH.
In the process of tests the operation of the bearing was monitored on the /395
basis of flow rate of sodium through it, by the power consumption of the pump,
394
Figure 3. Schematic Diagram of a Pump Stand.
1) Pump with an Output of up to 150 m 3/hour and Pressure of 3 kg/cm2;
2) Main Loop; 3) Oxide Indicator; 4) Cold Trap.
cu
Characteristics of Various Stages of the Study
I I Sodium flow rat I

Oxygen Pump Test I No. of inthebearing,m /^
Sodium content _hour_ __.
Stage : temper g/res-
pre Pump time, starts
No, tore, C of loop, I de' hours and Atbegin-1 At end
by .kg 2 1 livery
/ cm stops ning of test of
w e i ght I stage stage
O.i^3 I
II 11:10 I 0.025 5 100 I 12:3 i 110 ) S.7 I 9.0
III 3.O I 0.01 0.6 { 40 450 I -- 6 ---2 5 2.9
IV 15S)-550 0.007 1 5 I 100 120 ! 4 1 9.7 ( 9.5 I
TABLE Concluded
Radial wear of working journals Radial wear of working journals
of the shaft, mm of the bearing
1 I 2 I 3 I 4 ( 5 1 I 2 I 3 I /i I 5
0.015 0.035 0.04 0.005 0.0/1 0.0/13 0.015 0.006 0.013 0.001
10.0851 0.01251 0.025 1 0.03 I 0.025 1 0.0.48 I 0.026 1 0.031 1 0.015 1 0.001
0.0351 0.09 1 0.44 1 0.07 1 0.04 1 M14 1 0.504 1 0.051 1 O.OS 1 0.022
0.01 1 0.08 1 0.02 1 0.07 1 0.02 1 0.23 1 0.23 F 0.262 1 0.204 1 0.164
as well as by periodic listening to the bearing's operation. After the pump is

started up, constant flow rate of sodium through the bearing and the loop is im-
mediately established. Periodic listening to the operating pump shows that the
bearing operates smoothly without side noise. The sodium flowed to the bearing
at a rate exceeding the design value by 2-2.5 m3/hour.
It was noted toward the end of the first test, stage that when the pump oper-
ated with a closed delivery flap, ambient noise was. heard in the vicinity of the
hydrostatic bearing It could have been expected that [direct] contact takes place
between the shaft and the bearing, although the pump pressure increased in this
396
Figure 4. Pump With a Hydrostatic Bearing (the Upper Part of
the Electric Motor is Not Shown) .
1) Center Ball Bearing; 2) Polyfluoroethylene Packing; 3) Pump
Shaft; 4) Sodium-Supply Lines to the Bearing; 5) Hydrostatic
Bearing; 6) Impeller; 7) Delivery Pipe.
mode from 5 to 6 kg/cm2 as compared with the nominal pressure, should have
aided in improving the operating conditions for the bearing. Upon analysis we
have arrived at the following hypothesis. It is possible that, when the rate of
flow of sodium through the bearing is increased, it loses its vibration stability,
and this in turn, breaks down the continuity of the fluid film and the shaft comes
into [direct] contact with the bearing. It is probable that a region exists in which
397
the bearing operates properly. We have also considered another case: short-
duration operation of the bearing without sodium supply through it (the sodium
supply lines to the bearing were "frozen"), also with the pump delivery flap
closed. The bearing operated without ambient noise. At the first stage the pump
was started up and operated for 5-10 sec six tidies without supplying sodium to
the bearing. External inspection of the rubbing pairs after 1100 hours of oper-
ation showed that their surface was smooth without scratches. The shaft was /398
worn concentrically at all the working journals, the bearing also increased; here
the wear of the latter was eccentric relative to the center lines. Figure 5 shows
characteristics of the pump, measured at the start and end of the first test stage.
The flow rate of sodium through the bearing increased toward the end of the tests,
which is due to the increased clearances through which the sodium flowed, pro-
duced by wear. The pump characteristic did not change after the 1100 hours of
operation. The radial wear of the rubbing pairs of the bearing for this and all
the subsequent stages is presented in the table.
The second stage consisting in study of operating endurance of the bearing
on cyclical stops and starts. The cycles were repeated after 60 minutes. The
pump was started up 1-4 minutes after stopping. It was noted that after each
start the flow rate of sodium through the bearing was first somewhat higher than
before stopping and, over 10-15 minutes it gradually took on a value close to the
initial magnitude (Fig. 6). The rate of flow of sodium through the bearing in-
creased from cycle to cycle. No defects, except for wear of working surfaces
was discovered when the bearing was dismantled.
The third stage consisted in the study of operating endurance of the bear-
ing at 1/3 of the nominal shaft rpm. It should be noted that the bearing was de-
signed for apressure of 5 kg/cm2 , hence, the results of this stage should be re-
garded as qualitative [only].
;9 4)
0o
Cd
Abb
00
a
v
^m
w ^ Vl
r
U
a
0
a
v
a
2
0 20 ^0 60 80 i00 i20 140
Pump delivery, m3/hr
Figure 5. Characteristics of the Bearings at the 1st Stage.

1) Pump Pressure; 2) Rate of Flow of Sodium through the
Bearing at the End of Tests; 3) Same as Above at the Be-
ginning of Test.
398
After the pump was started up, /400
the bearing operated smoothly, with-
t out ambient noise. The electric sys-
v 7 -
tem for heating the outer part of the
1 I
sodium supply lines to the bearing op-
0
^ s erated unsatisfactorily. After 24 hours
the temperature on the outer part of the
M pipe dropped to 140C and flow of so-
w ^" dium through the bearing ceased (Fig.
4 7). Reheating the entire system to
I
0 3 470C and starting the pump did not
w yield satisfactory results - sodium
O
^ L did not flow through the bearing. It
Pz
was discovered on disassembling
the pump that the sodium supply lines
to the bearing were chocked by oxides.
3 12 16 20 2^ 29 External inspection and blow down
Time, min. by gas showed that the inner sodium
communications holes in the bear-
Figure 6. Rate of Flow of Sodium ing are free of sodium compounds.
Through a Hydrostatic Bearing on No wear of rubbing pairs was ob-
Starting and Stopping. served during this time (Fig. 8).
Inspection of the bearing after op-
eration at reduced shaft rpm show-
ed that the surface of the rubbing
pairs is smooth without scratches
or dents.
At the fourth stage we have investigated the operating endurance of the bear-
ing at sodium temperatures of 530-550C in order to determine the durability of the
rubbing pairs of the bearing. Comparison of the bearing characteristics obtained
at the start and end of this state (Fig. 9) shows that the rate of flow of sodium
through the bearing decreased toward the end of this stage. After the pump was
disassembled we have discovered, in addition to additional wear, also metal peel-
ing. Some holes for supplying sodium to the carrying chambers were partially
covered over by local galling of steel. No peeling or visible workhardening were
observed. It is possible that the plugging up of holes by galling is responsible
for reducing the flow of sodium through the bearing. Control checks of the shaft /401
wear showed that part of the shaft situated outside the bearing, as well as jour-
nals situated in the bearing sleeve had their diameter increased from 54.78 to
54.90 mm, Metallographic analysis showed that the shaft hardness dropped from
55 and 36 RH. Changes in the shaft diameter are due to structural transforma-
tions in the steel. The high shaft wear at this stage is apparently attributable
to the reduction in the steel hardness. No corrosion interaction between the
metal and the liquid sodium was observed (Fig. 10).
When the bearing operates at low rpm the wear of rubbing pairs is at max-
imum. We note that the conditions under which the bearing operates at this stage
are very difficult (the pump pressure dropped nine-fold and reached 0.6 kg/cm 2).
Under these conditions the continuity of the fluid film preventing the rubbing pairs
399
2,4 ,40
7.
280 '
m F
F
cV
H
d ^^
'^ a
a
H
O.t 760
100
0 4 8 12 16
Time, hr
Figure 7, Change in the Rate of Flow of Sodium

Through a Hydrostatic Bearing When the Sodium
Supply Lines are Chocked by Oxides.
1) Sodium Temperature in the Loop; 2) Rate of Flow
of Sodium through Bearing; 3) Temperature on the
Surface of the Whiter Part of the Sodium Supply Line
to the Bearing.
o^
.F J. J
cd
v 1
P N
^ Y i..
Q
n 00
F 0,7 Cd
Figure 8 Characteristics of the
^3 k
?s Pump and Hydrostatic Bearing at
ti the 3rd Test Stage,
0 0'S 1) Pump Pressure; 2) Rate of Flow
0 2
0 4 w of Sodium through the Bearing.
4^
Ca
P4
0 10 20 30 40 50 00
from direct contact may be disturbed. In addition, the bearing is divided into 12
chambers, which practically prevents the formation of a fluid wedge when the
bearing ceases to operate as a hydrostatic bearing,
400
17
eo
^F
N
e^ d
xi W
0
Figure 9. Bearing Char-
v
acteristics at the 4th Stage.
OM
N 1) Pump Pressure; 2) Rate
a of Flow of Sodium through
b c>+ the Bearing at Start of Tests;
3) Same as Above at End of
VOl
0
Tests.
Cd
(4
0 20 40 60 80 100 120
Figure 10. Working Part of the Pump Shaft.
Our studies should be regarded as qualitative, since the wear could have
been affected by the frequent disassembly and assembly of the bearing. Addi-
tional experiments are needed to finally establish the character of wear of the
rubbing pairs and the effect of the bearing's design on the magnitude of wear.
It can, however, be concluded on the basis of our experiments that the operat-
ing endurance of a slot-type hydrostatic bearing in sodium pumps is satisfac-
tory. The rubbing pairs we have used can be used in hydrostatic bearings for
liquid-sodium temperatures of up to 400C.
REFERENCES
1. Palmer, R. and Platt, A.: Fast-Neutron Reactors. [Translated from Eng-

lish and] published by Atomizdat, Moscow, 1963.
401
2. Kirillov, P. L., et al.: Atomnaya energiya, 7, Issue 1 (1959).
3. Chernousov, N. P.: Novyye konstruktsii germeticheskikh khimiko-tekhno-
logicheskikh mashin [New Designs of Leakproof Machines for the Chemical
Industry]. Kimicheskoye mashinostroyeniye, No. 1 (1959).
4. Reactor Handbook, v. IV. Engineering. N.Y. Interscience publishers, 1964.
402
FLOWMETER FOR MEASURING LARGE LIQUID-METAL FLOWS
N. L. Loginov
Permanent-magnetic flowmeters are extensively used to measure the flow of

liquid metal, since their design is simple and reliable; they can withstand elevated
temperatures and aggressive media, are convenient to use and are sufficiently
accurate. These characteristics make these flowmeters suitable for use in liq-
uid-metal cooled nuclear power facilities. However, the use of the general ar-
rangement for magnetic flowmeters (pipeline, placed between the magnet poles)
in larger power facilities, using large-diameter pipelines, for example, of 200
mm and above, is difficult, since in this case the weight, size and cost of the
magnetic system becomes very high.
The device considered in this paper is a flowmeter which operates by mea-
suring the velocity in one point of the flow. Such an instrument is moderate in
size as compared with the pipeline; it is intended for measuring liquid-metal flow /403
rates from 50 to 1000 m3/hour in a pipeline 300 mm in diameter.
The design of the pickup shown in Fig. 1 includes an all-turned casingfrom
1Kh18N9T steel, installed on the pipeline by welding [thus] keeping the loop leak- /404
proof. Inside the casing is placed a holder with the field-producing magnet, also
made from 1Kh18N9T steel. The magnet is made from the Magnico alloy with di-
rectional crystallization. The maximum magnetic induction in the outer plane of
the casing bottom for the given pickup is 840 gausses. The magnet can be used
at metal temperatures of up to 450-500C; it is fastened by a pin in a slot in the
holder. Two stainless steel electrodes are installed on the same holder, and
they are insulated from the latter by asbestos-cement inserts. The working ends
of the electrodes are sharpened and bear on the casing bottom at a distance of 18
mm from one another. Stainless-steel conductors, used for carrying away the
signal are welded to the free ends of the electrodes. The electrodes contact the
pickup's bottom due to the fact that the holder is pressed by a nut and pin, The
ends of the latter enter slots in the housing and keep the holder at a fixed position.
The pickup is designed so that, if necessary, its entire inner part may be
removed for inspection or repair without disturbing the airtightness of the work-
ing loop. Figure 1 shows the directions of the magnetic and electric fields. The
pickup is installed on the operating pipeline is such a manner that its bottom is
located at the pipeline axis, i, e. , in a region with a uniform velocity distribution.
Figure 2 shows a distribution of the magnetic induction vector component
perpendicular to the pickup's bottom as a function of the distance from the mag-
net butt, obtained by the IMI-3 instrument. The highest point of the graph cor-
responds to the position of the outer surface of the pickup bottom. The figure
shows that the magnetic induction drops almost to zero at a distance of about 50
mm from the bottom. It can thus be expected that the presence of moving metal
403
Figure 1. Design of Pickup and its Installa-
tion on an Operating Pipeline.
1) Magnet-Fastening Pin; 2) Magnet; 3) Holder;
4) Single-Piece Turned Casing; 5) Pin; 6) Nut;
7 and 8) Asbestos-Cement Insert.
at distances of more than 50 mm will not appreciably affect the pickup's sensitiv-
ity. Consequently, it can be calibrated on pipelines about 100 mm in diameter,
which can be done using laboratory liquid-metal setups.
It can be assumed that in facilities with a larger diameter the pickup's

readings are proportional to the velocity of metal ' at the pipeline axis. Know-
ing this velocity it is also possible to calculate the rate of flow of metal through
404

the pipeline. Hence, the results of

gauss calibration are represented in the
form of dependence of the output sig-
nal on the velocity, rather than the
flow rate.
For calibration the pickup was
installed on a sodium loop with a 103
mm diameter pipeline, which was
equipped with an ordinary magnetic
flowmeter calibrated by the volume
method and used as a standard. The
pickup emf was measured by a R-2/1
semiautomatic potentiometer.
The rate of flow of metal through
the pipeline was determined by the
-- "standard" magnetic flow meter, then
a 7 2 3 4 5 l,cm the average velocity in the cross sec-
tion where the pickup was located was
Figure 2, The Component of the Mag- calculated and a curve showing the
netic Induction Vector Perpendicular output signal of the pickup as a function of
to the Pickup Bottom as a Function of the velocity was constructed. The results
the Distance. of calibration are shown in the table. The
output signal of the pickup is not too large
but in the operating velocity range (0.2-3 m/sec) it is sufficient for measure-
ments, for example, using the EPP-09 potentiometer with the 0-1 millivolt scale.
Results of Flowmeter Calibration

._
Flow rate Velocity, EMF Flow rate Velocity, F EMF,
M3/hr y m/secs I mV m3/hr m/sec I mV
(I
5.90 0.252 0,046 28.20 1.205 0,276

9.60 0.410 0.084 29.20 1,250 0.287
11.50 0.493 0.102 29.00 1.200 .0,280
13.35 0.570 0.127 30.60 1.310 0,305
16.50 0.708 0.158 34.40 1.470 0.344
22.30 0.955 0.212 34.10 1.470 0.350
35.50 1.520 0.350 42.00 1.800 0.428
25.85 1,100 0.258 49.40 2.100 0.490
405
TIME NEEDED FOR MELTING CYLINDRICAL [METAL] VOLUMES
V. M. Selivanov, N. M. Trebina and V. I. Sharypin
INTRODUCTION
Melting and solidification of metals is accompanied by changes in their /406
state of aggregation, which has a perceptible effect on the temperature field and
on the melting time of the metal. The melting time of metals cannot be deter-
mined analytically since the surface at which melting takes place changes posi-
tion. This problem was solved by various methods in [1-7]. Dusinb.erre [6]
and Eyres [7] have proposed a method for solving this problem using an analog
computer. Baxter [3], who resorted to this technique, has generalized results
on the melting of bodies with an initial temperature equal to the melting temper-
ature for boundary conditions of the 1st and 3rd kinds. The present paper pre-
sents generalizing results for the melting of cylindrical volumes with an initial
temperature different from the melting temperature for different heat-transfer
conditions at the metal surface (boundary conditions of the 1st, 2nd and 3rd kinds).
Statement of Problem
It is required to describe the process of melting,of a cylindrical bodywith
specified thermophysical parameters and initial temperature t 0. The following
are the assumptions made:
1) the material is a pure substance, i. e., melting takes place at a single
specific temperature;
2) the density of the substance does vary on changes in the state of
aggregation;
3) no heat is lost from the ends of cylinders;
4) the thermophysical properties of both phases (liquid and solid) are the
same and do not depend on the temperature.
Solution of the Problem

A schematic depicting the melting of a cylindrical body is shown in Fig. 1.
The presence in the cylinder of liquid and solid phase regions, a moving bound-
ary between them and complex conditions at the phase interface makes it necessary
406
to describe the temperature field of the cylinder by two /407
equations ( one for each phase). We now write an equa-
tion for the thermal conductivity for a cylinder of infin-
ite length with constant physical properties
(1)
It is possible to obtain a single equation applicable toboth

phases. We introduce new variables, i.e., the enthalpy
and specific flux. The enthalpy
t
J (t) _ ^ C (It or dJ (t)
Figure 1. Sche- dt - C (2)
0
matic Showing
the Melting of The specific flux
a Cylindrical
Body With Liq- t
uid (1) and Sol- ^h (t) = ^ 7^ dt or c (t) _ k.
. dt (3)
id (2) Phases. 0
In order that Eq. (1) be applicable in the entire region 0 _< r:5 r 0 it is necessary
to make allowances for the fact that, in the process of reaching the melting tem-
perature, the heat supplied is expended not only for heating up, but also for melt-
ing. This can be done by supplementing Eq. (1) by the expression
E +0

hin c alt = c dt (4)
E-O _ E _0
After substitution of new variables into Eq. (1) and reducing it to the dimension-
less form, we get
OJ* a'- cb * 1 00*

î'
= ^7r'
r2 r -F-^ fir* (0 C r* <,1, i* > 0)
The initial condition i8 J* (r*, 0) = J* i0 , while the boundary conditions at

the surface are
t* (1, i*) -= t*n I kind;

scDF (1* 0 _ + K*
II kind;
ar T **) = q .XP
r

a(pxar *'*^ Bi [t* (1, ti*) m) III kind.
407
The relationship between the dimensionless parameters is shown in Fig. 2. Let
us now consider the case of J* and t*. From the mathematical point of view it
is immaterial how this relationship will be plotted in coordinates t* (J*). On
practical considerations this is done in the manner shown in Fig. 2a. This po-
sition of the zero point makes it possible to solve Eq. (5) on an analog computer
at tml = 0 for melting and solidification, which follows from the relationships ob-
tained for the dimensionless temperature and enthalpy. The slope in the positive
quadrant, which is unity, appreciably facilitates analysis of data obtained from
temperature fields in the melting body, since [then] J* = t*. In Fig. 2a the en-
thalpy J* _ -1 attendant to melting corresponds to the solid phase at the melting
temperature, while J* = 0 corresponds to the liquid phase at the same tempera- /408
tures. The intermediate values of J* correspond to a two-phase system and show
the part devolved upon the solid phase. If J* > 0, then the liquid phase is super-
heated relative to the melting temperature. Similar considerations can be pre-
sented by analyzing the graphical relationships (h*(J*) and (;)*(t*) (see Figs. 2b
and 2c) To solve Eq. (5) on an analog computer it is necessary to [first] obtain
from it a difference-differential equation. In the case of a cylinder it is most
expedient to divide the radius into equal segments relative to r* 2 , so that each
layer so cut out would have the same energy store. Substituting Z* = r* 2 into
Eq. (5) and replacing the variables of the complex function, we get
aJ* a 2 (A* _iE

a(P*
or` ` -'tZx ZZ2
az
` ( 0 C l' -00)
' (6)
We shall consider the enthalpy only in (n + 1)-discrete zones, situated at

distance Z* = i/n (i = 0, 1, 2, ..... n) from the center, at equal intervals. We
get (n + 1) differential equations
*
diM 2rz[(2i=, 1)r1^21i-4i71^2- i (2i-1)^-i]
For i-=n
a) boundary conditions of the 1st kind
(P * = t* _ t*.
n C -- 11
(7)
b) of the 2nd kind
d.ln 80
(0*n-1 71 (L* L i / 2n 1
1))'. K*;
d7
c) of the 3rd kind
*
d^n.._8n2((I,n- 1(I,)-2(2n 1)-Bi(tnt*).
dT*
408
a b c
Figure 2. Functional Relationship Between

Dimensionless Parameters.
^^=Zm1 The above system of equa- /409

20 tions was solved on the EMU-10
10
analog computer, and nomograms
B were constructed on the basis of
6 results obtained. The initial en-
4 ^- C076 thalpy was calculated from the
3 -2
expression
2
1.0
0.s
0,6 zo -- ---
0.4
11.3
0. The melting time is found from
01
0.04 0,080,7 0.2 0.4 0.60.81.0 2 3 4 56 8 70 20 K"
C* - 1p
Figure 3. Nomogram for Calculating the i'nl - a
Melting Time for Cylindrical Volumes
Heated by a Constant Thermal Flux. The nomograms of Figs. 4
serve to calculate the melting
times for cylinders for boundary
conditions of the 1st and 3rd
kinds. Each nomogram is constructed for specific boundary conditions. When
Bi = - (boundary conditions of the 1st kind) it is necessary to assume that the
surface temperature is is equal to the temperature of the medium tm. The nomo-
grams shown in Figs. 3 and 4 can be used to determine the time for melting cy-
lindrical volumes of substances in which the initial enthalpy is
J* C(io tml)
-1 i 3.
P
409
M1
{-Â i
"Hi-L I I I
0.0^ L^-r^ 9, 7 4^ 2 i7.4 iV !'S 2 0.7 0.2 OW,sVIO 2 3 te'
tc*
t' =Fi
2,
ti I
10
0.4
0.1
01 P.2 0.4WtWO 2 3 4 F t; 9.7 0.2 04 0.60.81.0 2 3 tc*,
Figure 4. Nomograms for Calculating the Melting

Time of Cylindrical Volumes Under Boundary
Conditions of the ist and 3rd Kind.
NOTATION
r - current radius of cylinder
r0 - radius of cylinder
depth of molten layer
t melting temperature, OC
ml -
410

is - surface temperature, C /411
tm - temperature of the heating medium, C

p - latent melting temperature
T- time
J - enthalpy
w - specific flux
q - thermal flux density
a - heat transfer coefficient
r )lt
rp P
V.a.v P
(IT ttrp
0
REFERENCES
1. Veynik, A. I.: Teoplobmen mezhdu slitkom i izlozhnitsey [Transfer of Heat

Between An Ingot and Its Mold]. Metallurgizdat, Moscow, 1959.
2. Veynik, A. I.: Priblizhennyy raschet protsessov teploprovodnosti [Approx-
imate Calculation of Heat-Conduction Processes]. Gosenergoizdat, Mos-
cow-Leningrad, 1959.
3. Baxter, D. S.: In the book Transactions of ASME. Translated from English,
84, series C, No. 4, p. 55. Foreign Literature Publishing House Mos-
cow, 1962.
4. Carslaw, H. S. and Jaeger, I. C.: Oxford University Press. Lond. , 1947.
pp. 71,227.
5. London, A. L. and Sebah, R. A.: Trans. ASME, v. 65, p. 771 (1943).
6. Dusinberre, G. M.: Trans. ASME, v. 67, p. 703 (1945).
7. Eyres, N. R.: Philos. Trans. Roy. Soc. , v. 240A, 1946, 1-57.
411

FROZEN SODIUM SEALS

A. V. Drobyshev, Yu. I. Likhachev and I. M. Saratov
INTRODUCTION
At present much attention is given to the development of shaft packings. /411
Experiments have shown that frozen sodium seals in combination with a special
packing gland can compete with bellows- and membrane-type packings, and in
some cases have certain advantages over them. The main advantage of these
packings is the unlimited height of rise and the feasibility of use with any
shaft, rod, etc., diameter. The leak through this kind of seal is exceedingly
small, only about 2-3 grams per day in a rotating shaft, When used with a valve,
flap, etc., rod no leakage is observed at all [l, 2].
To obtain approximate values of the torque and axial load in these seals it
is nedessary to solve the problem of deformations and stresses in a pliable an-
nular layer, included between two rigid cylindrical surfaces. In addition, one /412
must know the mechanical properties of the packing material at operating tem-
peratures. To obtain these data we have made a special device which can be
used to test sodium specimens at different temperatures. The tests were per-
formed on a UM-1800 machine with variable rate of stress application. The
characteristics of the mechanical properties of sodiumwhich were obtained can
be used to determine the expected torque for a rotating shaft, as well as the
shearing force for the case of a sliding shaft (rod).
Study of a Frozen-Sodium Seal

We consider two kinds of packings glands with frozen sodium seals. In the
first case the rod rotates over angle co under the action of a force couple with
moment M, while in the second case it is imparted an axial displacement S by
force P (Fig. 1). The main purpose here is to obtain relationships between M
and co and between P and S. We now consider the strain in an annular layer of
the packing contained between perfectly rigid cylindrical surfaces of a rod ( r =
= a) and the packing gland housing (r = b).
The Rod is Loaded by a Force Couple

The packing material passes the elastic and elastoplastic strain stage quite
rapidly and hence, creep at these stages can be disregarded. Making use of the
axial symmetry of the problem and the sufficiently extended rod length, we can
regard all the components far from the ends as functions of r only, in which case
412

yz0 - y r0 - 0, Since strain in the axial /413

Sec, nn direction is restricted, the axial dis-
placements u z = 0,
Elastic staLre of Dacldna deforma-

tion. At the start of load application
the packing material undergoes elastic
j straining and its state is defined by the

following equations. Equilibrium equa-
tions
ac7r _ ^r csa --
0 (1)
Figure 1. Calculating Scheme of Tr r
Frozen Sodium Seal.
a^Cre
+ r 0.
car (la)
Geometric relationships
_ ou r _ u r _
Er
ar
so
r
sz Buz
^Z 0 ' (2)
aue uo
}'r6 =
ar r (2a)
Physical relationships
ter=
E
[ 6r E^ (^z^ -6 0)^, ^0 = 1, [ 6 0 ^ L (^z-I-^r)^r
z= G [ CF z I t ( 9, +(50)], (3)
tiro
Yr == G (3a)
where , E and G is Poisson's ratio, Young's modulus and the modulus of rigidity.
We must satisfy the boundary conditions
_ l^t
r ---a, U. 0' T r E) = ti
2rca 2l ' (4)
r=b, ur =uo=0. (5)
413
The solution of system of Eqs. (1)-(3) is known, [3]
6r =B-^ ; UO = B 2 ; 1
6Z = 2E0; T ro = 2 ;
(6)
ZZr == 1 [B(1-2[,)r ri,
uo=Dr-2,
(the bar denotes components of the elastic stage). Satisfying boundary conditions
(4) and (5), we get
TO ra2 ,a2 P \ . 7)
7
T re = i2 ' Yr0 = d
r2 e u0 =
C
2G62 r r /)

(
The remaining components are zero. The angle of rotation of the rod is
_ u0(a) __ D1 a2 1

a îrtGa2l C G2 ) (8)
The elastic deformation terminates by appearance of plastic strain near the rod /414
surface, when the von Misses yield criterion is satisfied
X _ 1T1
(9)
where
(I.) - i l) (l rt3 - i' Trz ^' ^,U) ---

.^ '
h
j (lS r lT^^) 2 - ((T. --- IT ^^^- (Qr --
(lU)
is the total tangential stress; t is the yield strength of the material of the pack-
ing at the given temperature. The torque for which plastic strain starts is
-. 2 . > (11)
Here the angle of twist of the shaft
1 2
Tr===2 C b2 1) (12)
414

The elastoplastic stage of strain in the packing. If the material of the pack-
ing can be workhardened, then there appears a certain zone of plastic deforma-
tions a r <_ c, alongside with the elastic zone c < r - b, which is retained. This
zone continues to be described by Eqs. (6), while Eqs. (1), (la), (2) and (2a) re-
main valid in the plastic zone, while the physical relationships, according to the
theory of small elastoplastic deformations, have the form [4):
a) condition of incompressibility of the material

Er-=Ey--E,= 0; (13)
b) condition of similitude and coaxiality of the strain and stress deviators
EiE _ ^;
Gi Gj ZTIj (` ' --r ' z ' (14)
c) power law for the workhardening
1 =KN", (15)
where
I' ^(Er ^)2 (E z E O) ` i - ( E r E z) ? f Z (Yro i Yzo-I YrZ) (16)
is the shearing-strain, while K and n are coefficients characteristic of a given

material at the specified temperature, which are determined experimentally.
In torsional tests of thin-walled tubes
If
N=2.1ftzb,and r --Ra,
where M is the moment twisting the tube, R, 6 and a are the average radius, /415
thickness and the angle of twist per unit length of the tube, respectively (Fig. 2).
It is also required that components o- r, Tro, U and u0 at the boundary between
regions r and c be continuous. Solving Eq. (la) and satisfying boundary condi-
tion (4), we get for the plastic region
n2T
ti3O_rz (17 )
415

Equations (2) and (13) yield

^^
Ur = r e
e O -_ S r _ : ^^ .
r2
using boundary condition (4), we get
IGr == FO = E r = 0. (18)
from which, together with Eq. (14) it follows that a-r =0-01 and further from Eq.
(1) we have
Since P- Z = 0, v Z = 1/2 (ur + v0 ), hence,
6r -.: 6e = az =- B 1 . (19)
From the above expression and from Eqs, (10), (16) and (18) we have
N =Tre, 17 = Yle (20)
Condition N(c) = T T is satisfied at the boundary of the plastic region (r = c),

whence it follows
c=a V
(21)
^T
From Eqs. (2), (15) and (20) we get
du e 118 Ka2nt7L
r2n 1
dr r
solving which we get
Ka 2ntn
it = 2nr
2a-1
+ Dir. (22)
416
Using Eq. ( 5) and the condition that the components are continuous at boundary
r = c, we will get for the components of the elastic region the same expressions
(7), while for the plastic region
6r=66=6z =0,
i[ =
hn
Ln C S T pen)
n a'- n tae T1,U2
^' + 2G y2 7', (23)
the angle of twist of the rod is
K n _ 11f It 1 1 11f
cp
2,a [ i T 2 l) n
(2Za 1 + 2 G ` 2901 ^ T ) ' (24)
If c = b, then the plastic region extends over the entire packing, in which case /416
the rod is loaded by the moment
f, 2:Th 22 1T,,, (25)
and its angle of twist
KT
YO
(26)
In the case of a packing made from an ideal elastoplastic material [see Fig. 2
(dashed line)], it fails at M T such that an annular plastic hinge is formed on the
rod surface and its further twist proceeds without an increase in the moment
[see Fig. 4 (horizontal line)],
Plastic deformation of the packing. Application to the rod of moment M > M0
results in further development of plastic deformations over the entire cross section
of the packing. This stage of deformation may take place quite a long time, hence,
in addition to instantaneous strain, one also must take into account creep strain.
We now consider this stage of deformation, using the following relationships be-
tween the stresses and the rates of strain[ 5 -61
a) conditions of elastic change in the volume
-s -0, ^C-+56+ x__=0, (27)
417
r b) conditions of similititude and coaxiality
N of stress deviators and of rates of plastic and
A
creep strains
z, I
c
O ,
1 j 2rij ' c Z -aj 2rlj 28
Figure 2, The Stress-

Strain Diagram of the c) disregarding plastic components, the
Packing Material N =f (r) rates of strain are now composed from the rates
and the Method for Recon- of plastic strain and of creep strain
structing it into an N =
= f(I'p) Diagram. The gi == Sp -1_ 'SC j _= T14). -j- 1^ j (C, ) _= T, 0, Z); (29)
Straight Line BC is Par-
allel to the Initial Seg-
ment of Curve N = f(I'), d) assuming that workhardening is gov-
erned by a power law, we shall regard N, the
total tangential stress, as being related to the
measure of plastic strain I' p in the following
manner
(30)
where I'p = ^ Hp dt is defined as an integral taken over the entire straining path,
of the intensity of the plastic shear strain rate
lip _ V 2 _ (sr) z {- ( y)'` -!- ( Z) 2 { 2 L(jj O)2 -{- (i^ o) L { (^^ Z) L] , (31)
K1 and n1 are coefficients characteristic of the given material at the specified

temperature, which must be determined experimentally. Using the r - N di-
agram thus obtained it is possible to construct a curve of I' p, making use of the
fact that Pp = Ra - N/G (see Fig. 2);
e) we assume that the intensity of the creep shear strain rate He is a pow- /417
er-law function of the total tangential stress N
H;:K,W.NI", (32)
where Kc ( t) and m are some time function and a coefficient, which are determin-
ed experimentally. The value of He is defined by a relationship similar to (31),
418

but the creep strain rates ^c and 71c are used instead. We now write the known
relationships between the rates of strain, components of the rates of displacement
and the displacements
av vr.
br-
E =- r
ar ' 0 r '
av,
SZ = az ' (33)
aV VO 0
TIrO= Or r-' (33a)
V dur Fl u O
at ,
r V _
O -- dt
r
(111,
VZ
-- dt ' (33b)
the equilibrium Eqs. (1) and (la) are now written for the current coordinates (r,
O, z) of points of the packing. Boundary conditions (4) and (5) remain the same.
Solving Eq. (la) with boundary condition (4) yields
az
T rO = r2 T (t), (34)
where T(t) = M(t)/2?ra2 1 and M(t) is a function of time (Fig. 3). Due to the fact
that deformation in the axial direction is constrained and that the material of
the packing is [was assumed to be] incompressible, we can use Eqs. (4) and (5)
to write
UZ==ur =0, (35)
consequently,
(36)
Vr=V== ^r= tSO=^-- 0.
Usings Eqs. (1), (10), (28), (29) and (36) we cpxi show that
(jr = Q8 =_ (- _ B ( t), N =- TrO (37)
From the above and fin Eqs, (27) and (28) we get
419

M is e.,
2 ,3
HP _ Jj e^ f^` _ 7Ire (39)
/ /^
Using the above expression and Eqs, /418

(29) and (33), we get
av e vo
I î I H" (r , 1) H`
Figure 3. Change in the Moment solving which together with Eq. (5),
Twisting the Rod with Respect to we get
Time for Different Loading Rates
V 1 > V2 > V3 (1, 2 and 3,
r
Respectively). ^. r NF (.r, t> u',r.
r He (-L' t) dx.
b r r
b (40)
In the general case
du t> dire d r au do arro dZ

1 o dt=-dt
-'L --
Jr dt +- dO dt + Os dt
Since r and z are constant in time, while
Ito --- Lie ( r , 1),
then
dr dz au0 (1110 alto
whence
^'6(r, J) dy,
(41)
0
here integration is performed over the entire loadiim path from t = 0 to some t.
Using Eqs. (40) and (41) and making use of the fact that is constant in time,
we write
r I r t
tzr cxx
Ili (x, y) dy,
uo ==I' II'' ( x , J) (ly r lm1
h 0 b 0
420

where ^ Hp (x, y) dy = F p is a measure of the plastic deformation accumulated

over the entire straining path from t = 0 to t. For sufficiently developed strains
I' p - N and r - N are almost identical and we can replace Eq. (30) by Eq. (15).
Using the latter equation together with Eqs, (32), (34) and (37), we get from
Eq. (42)
e
K-'irn
ue===- 2ar 2n-1 1
'
r-1L 1 a 2m l ('
r 2111 T
b2m ( y ) Kc( y ) dy. (43)
b ' n / 277Lr`Lnt-1 C1 J J
The angle of twist of the rod is
Ile (a ) Kiln (t) / 1 _ a2n \ 1

1P 62 lnml^nnz2m X
In J L,1i^

I
1 a2711
a2m A7`
(y) Kc ( y ) dy (44)
v
The first term in the last expression defines the contribution to the rod's angle /419
of twist by instantaneous strains and does not depend on the manner in which the
twisting moment varies with time. The second term defines the contribution
made by creep and depends on the manner in which moment M(t) changes with
time. Consequently, the relationship co - M in the given case is not unique for
different M(t) (Figs. 3 and 4).
Rod Loaded b ann Axial Force
We now consider the elastic and elastoplastic stage of loading, without con-
sideration of creep strain. Making use of the axial symmetry and the large length
of the rod, all the components away from the ends can be regarded as functions of
r only, in which case y r, = y ze = uO = 0.
Elastic stage of load application to rod. To solve this problem we use the
equilibrium Eq. (1) and the equation
i9rrZ r rz _
Fr r ^' (45)
relationship (2) and the condition
Buz
Yrz = Or (46)
421

without taking creeping l

into account ?
Mo
I I
9r ^Po 9
Figure 4. The Moment Twisting the Rod as
a Function of the Latter's Angle of Twist;
Curves 1, 2 and 3 Correspond to Different
Rates of Changes in Moment M of Fig. 3.
Eqs. (3) and the expression
'Vrz = Gz . (47)
The boundary conditions to be satisfied are
r= a, u r = 01 T rz = T = 2n
it ' (48)
r = b, u r = uz 0. (49)
Solving Eq. (45) and satisfying condition (48), we have /420
va
T rz == r (50)
Using the above expression and Eqs. (46) and (47), making use in the process of
Eq. (49), we get
rc z -- G
va
In r
U
(51)
422

The solution of systems (1), (2), (3) is known, [7)
ir r = lr-;=b .r,
i Cr
8 8 =A+B1r 2 , - -A---$lr2,
tfr-:.E L.1 (1 ) B (1 - 20)

(52)
ab E C.1 (2 }^)+ ^ (1-2') ] ,
v_ = 2 1t (1 Ft) EA.
Using Eqs. (48) and (49), we get
U r_ S r = 3-- U r =U 0 ==U z =0. (53)
The axial displacement of the rod
it y
(a) -= 2 1G 1n a (54)
Plastic deformations appear on the rod surface at the load
Pr = 2:tal-t T . (55)
The rod displacement attendant to this is
S T = tT a In b .
b
(56)
Elastoplastic stage of load application to the packing. On further applica-

tion of load P > PT to a packing made from workhardening material there appears
a zone a:5 r:5 c of plastic deformations, along which is also retained zone c:5
<_ r< b of elastic deformation. Equations (50) and (52) are still valid in the lat-
ter zone; the behavior in the plastic region continues to be governed by Eqs. (1),
(2), (45) and (46), while the physical relationships, according to the theory of
small elastoplastic displacements are expressed by Eqs. (13)-(15). In addition
to satisfying Eqs. (48) and (49) it is also required that components 0- r' Trz' u
and u be continuous at r = c, i, e. , at the boundary of the plastic region.
Equation (50) remains valid for the plastic region. 'Using Eqs. (2) and (13),
we get
up =A' r + P9 = 8r =`^' 2 r ( 57 )
423

Satisfying the continuity condition and the boundary conditions for component u r, /421
we get that A l =A = B = 0, i.e.,
( 11r Er_2?G_.11r__!r_'- E)` ` (io- n ' y

r7 O^
rz
It follows from Eqs. (1), (14) and (58) that o- r =% = 6z = B1 , i.e.,
N- -r r,- (59)
At the boundary of the plastic zone N(c) = T V whence
at
T T (60)
Using Eqs. (15), (46), (50), (58) and (59), we get the expression
du
Z Iit"rz'ti
_^^, _ ,_ _ ran
solving which together with the condition that u is continuous.at the boundary of
the plastic zone yields
IiT -inT ^n-lTn-1 Ta

Il _ _ -- --- - 1 ^." bT
111 1
T (61)
tt-1
IiTq. Jill , T P 2."-TWYT
it, (a) _- 111 I, . (62).
2.TI (11 "1) (ZZatcq.) ,a - i l ZZ1G
Plastic strain proceeds over the entire cross section of the pacldng (c = b) at
the load
Po _ 2jTbIT T, (63)
424
the displacement of the rod attendant to this
6
-- 7L^T 1 l_ \ a ) n 1-1 (64)
The carrying ability of a packing made from an ideal elastoplastic material

is exhausted when plastic deformations appear on the surface of the rod, hence,
here the load on the rod cannot exceed P T (horizontal line in Fig. 6).
Plastic deformation of the packing. When a load P > PO is applied to the

rod, plastic straining takes place over the entire cross section of the packing.
We now consider this loading stage so as to include the effect of creep of the ma-
terial of the packing. For this we use the physical relationships (27) and (32).
The equilibrium Eqs. (1) and (45), Eqs. (33) and (33b) and boundary conditions
(48) and (49) retain their validity. Instead of Eq. (33a) we have

11rz ='9V' (65)
solving the above equation together with boundary condition (48), we get /422
a
T rz = ti (t),
where
P (c)
(t) 2nat ' (66)
For an incompressible material, making use of conditions (48) and (49)

away from the ends
U,= 0,
consequently
V,=Ve= br= ê=^ z =0. (67)
It can be shown further that

Qr=6=6z =B(t), N =tirz, (68)
^: = gi = 0, (i = r; 0, z), (69)
Hp = TIP 9 He = rlr: (70)
425

Using Eqs. (29), (65) and (70), we write
OV t) HO (r,
ar z = H P (r, + t),
whence, mating use of (49) we have
r r
V z = Hp (x, t) dx+ ^ H` (x, t) dx.
b b
In general,
duz
VZ = dt
Buz Buz dr auz d9 + 8uz dz
at + ar at + a dt az dt
Since O and r are constant in time and
dr dz auz
u z = u z (r, t), T
o dt = di 8z ,, = 0'
i, e. , vz = auz/at, whence
r t r t
U., = dx
S
H' (x, y) dy + dx Ho (x, y) dy,
b 0 If 0
where Hp (x, y)dy = P p is the plastic deformation accumulated over the entire
0
straining path from t = 0 to t. Using Eqs. (15), (32), (66) and (68) for sufficient- /423
ly developed deformations, we get
.^ a"T" t Ln-t \ am
uz = (n-1)6 11 - 1 C rn - ! ^mlq) tfm_1 X
t
vm -1 I 1
rift
0
426
b = tCZ (a) _ _ k Pit (t) b bn -1

(n-1) (2ttb1)1e Can. -1 1)
I
_ b
(m-1) (2abl)m C bm-1
am-1 1
l
J 0
K, (y)
Pm
(y) dy (72)
72
The first term here represents the contribution of instantaneous strains, while
the second accounts for the creep strain and depends highly on the manner in
which force P varies with time. Figures 5 and 6 show curves of P as a function
of time and the corresponding 6 - P curves.
Results Obtained by Analyzing Stress-Strain Diagrams for Sodium

It follows from the above that the force needed for moving the movable parts
of frozen seals can be calculated on the basis of stress-strain diagrams for the
heat-transfer agent. The available equipment can be used most conveniently for
determining stress-strain diagrams under tensile loads. The large size of spec-
imens, low loads, rapid corrosion in air are the main experimental difficulties
when working with sodium.
The mechanical properties of the sodium were determined on specimens /424
specially cast in a split metal. chill mold. To increase coupling with the sodium,
the butt parts of the chill mold had a rough screw threading on their inner sur-
face; they served as end pieces for the specimen and the load was applied to them
( Fig. 7) .
The sodium specimens were tested in a somewhat modified UM-1800 ma-
chine. To increase the sensitivity of the load indicator, the sodium specimen
was connected directly to the machine's electrotensometric dynamometer. The
machine can produce several rates of deformation, which was utilized. The me-
chanical properties of sodium as a function of the temperature were obtained by
heating the sodium specimens in a bath filled with VM-4 preliminary -vacuum
oil (Fig. 8).
The sodium was tested in tension at three straining rates z = 0. 6, 2.5 and /425
7.2 hour 1 and temperature of 20C, as well as at four temperature levels (14,
20, 42 and 76C) with a straining rate S = 7.2 hour-1 . The changes in the ulti-
mate strength ab , yield strength and the relative elongation at rupture 6,
60.2,
as a function of one of the parameters (s, or T) is shown in Figs. 9 and 10.
The region of uniform straining in sodium specimens tested in tension is,
as can be seen from Fig. 11, moderate and comprises 3-8%. The conduct of the
test segment of the specimen under load differs from that of other materials. The
round cross section of the specimens of plastic materials retains its shape up to /426
427

P P without taking creep- 1 2

1 ;inio account
r I o
//,00. . Po
t o PT
Figure 5. Variation in the

Tensile Load Applied to a 0
YT Yo
Rod in Time for Different
Loading Rates V 1 > V2 > Figure 6. The Load Applied to the Rod as a
> V3 (1, 2 and 3, Function of the Rod's Displacement. Curves 1,
2 and 3 Correspond to the Different Rates of
Respectively). Variation in P Shown in Fig. 5.
z
l----
r
Figure 7. A Tensile Sodium Specimen.
1) Sodium; 2) Pull rod; 3) Aluminum End Piece of the Spec-
imen (Design Length of Specimen 1 0 = 100 mm; Design Dia-
meter d0 = 22 mm).
failure, while the cross section of a round cast sodium specimen loaded above
the ultimate strength changes, taking on outlines which are not uniform over
the test length of the specimen (Fig. 12).
The methods for calculation require that the experimental results be plotted
and used in the form of S - 'I' load diagrams (where S is the true stress and T is-
the relative narrowing down [necking down]). However, it can be seen from our
description of the experiment that it is impossible to obtain such a diagram for
the segment starting with the ultimate strength and continuing to the point of fail-
ure. Hence, to constructthe N(r) diagram the initial segment of the stress -
strain diagram in the above coordinates was approximated by a power law ex-
tended past the ultimate strength [segment], [6] and [7]. In addition, calcula-
tions were performed on the assumption that past the ultimate strength point the
material of the packing is workhardened insignificantly and its stress-deformation
428
Figure 8. Schematic of Setup for Tensile Testing of
Sodium Specimens.
1) Electronic Potentiometer; 2) Liquid Bath; 3) Dyna-
mometer; 4) Thermometer; 5) Specimen; 6) Thermo-
stat; 7) Motor; 8) Reducing Gear; 9) Pickup Selsyn.
120 axis (Fig. 13, dashed lines).

Y, The N(r) diagrams recon-
structed on the basis of the
^s Vo o
load diagrams P (pl) are
U
o shown in Fig. 13.
N
v The experimental re- /427
3 ap sults show that resistance
C.J_______.I. to deformation increases
2 ya with an increase in the [load-
0 1 2 3 4 S a 7 8 ing] rate and reduction in
Rate of strain., hr- 1 temperature. The forces
needed for moving the mov-
Figure 9. Mechanical Properties of Sodium in able parts of a frozen seal
Tension as a Function of the Straining Rate at 20C. will change correspondingly.
Hence, the starting data for
computations should be ob-
tained at deformation rates and temperatures close to those encountered under
actual conditions.
The temperature range in the tensile test covers the temperatures observed
in operating frozen sodium seals. The range of deformation rates is relatively
small.
Data for steel tested with variations in the rate of straining show that re-
sistance to deformation can increase by 2-3 fold when the rate is changed by an
order of 6-8 [8].
429
0
N C7
^' 4 O
U
ni
W
x
N
0
H v
W
^ 2 N
Id
r-1
av
0 20 40 60 80 T, C
Figure 10. Mechanical Properties of Sodium in

Tension as a Function of the Test Temperature for
a Straining Rate of 7.2 hour -1.
ti I 'e=72Y 1
^' _. I Q6
/ Î i--- i- '- ---

a.........-4-T2- -

Li J^
0 20 40 60 80 100 120
Deformation, mm
Figure 11. Load -Deformation Diagram for Sodium

Specimens (Design Length of Specimen 1 0= 100mm,
its Diameter d 0 = 22 mm),
zTy .
Figure 12, Shape of Sodium Specimen After Failure,
430
N, kga USS
Using data obtained in ten-
/cm2 sile tests of sodium one must
J
take into account the increase in
the resistance to straining at de-
formation rates higher than those
used in these tests.
N,
L`
Below we present calcula-
NB 20 0,6 tion of moving forces using a
stress - strain curve for sodium
obtained at T =14 C and s =7.2
n, hoar -1. The most accurate dia-
i grams can be obtained in torsion
iYB 76 2 - __
tests of thin-walled tubular spec-
imens tested at load rates close
0 0.08 0.16 0.24 r to those encountered in practice.
A special facility is being design-
Figure 13. Stress-Strain Diagrams for So- ed for this purpose.
dium for Different Temperatures
and Loading Rates
Results of c alculations. /428
Analysis of stress - strain curves
of the material of the packing ob-
tained in tensile tests at 14 C yields the following constant of the yield curve in
Eq. (15)
h 1.509 . 10 (cm 2 /kg) n 11.85.
starts at stresses of N T = 2.25 kg/cm2 . Assuming that the mod-

Yielding
ulus of rigidity G is approximately equal to 1000 kg/cm 2 , we get that = 0.225.10-2 r
We have calculated the twisting moments, forces and the corresponding an-
gles of twist and displacements without taking into account creep of the material
for a frozen seal with the following dimensions: a = 19 mm, b = 20 mm, L = 120.
mm for torsion and a = 18 mm, b = 18.25 mm and Z = 180 mm for axial displace-
ment. The results obtained are
.1I,.. h 12 kg- cm : (p, 1.097 . 10-'; Pr = 457 kg;

6,- 5.50-10 - 5 cm
iti9 0 = = 678 kg- cm; y,,, = 2.2:2 . 10 -4 ; Po = 465 kg;
6, = 6.02 . 10_ ; cm.
If it is assumed that the material of the seal is not workhardened past the /429
r
ultimate strength > r (see Fig. 13), then the ultimate twisting moment and
axial force can be found as
:lfmax -._ 2^-(rt2 1Xb , P 21jaIA'
431

where N is tangential stress at the instant when the cylindrical specimen ceases
to yield in an orderly manner. For sodium at 14C we have N,o = 2.75 kg/cm2.
Calculations yield Mmax = 750 kg-cm, Pmax = 560 kg. Figures 14 and 15 show
the twisting moment as a function of the angle of twist of the shaft and the axial
force as a function of the rod's displacement.
Study of the Characteristics of Frozen-Sodium Seals

At present many centrifugal pumps in operation use frozen-sodium seals.
Their shaft rpm are as high as 3000 and the sodium temperature in the snail-
case housing is 750C. The
maximum diameter of shafts
of high-temperature pumps is
i"- - - 40 mm [9].
I ^ i
To determine the power
Il losses due to friction a special
installation was constructed
3 1
which is shown schematically
in Fig. 16. Unlike a pump,
Z 15 ;u zu M w W cg y,/cO, this installation did not have
an impeller; it used a d-c mo-
Figure 14. The Twisting Moment as a Function tor rated at 1.6 kW to obtain
of the Rod's Angle of Twist for a variable rpm.
Frozen-Sodium Seal.
The first experiments
1) Calculated Assuming Power-lawWorkharden- were made with water. The
ing; 2) Calculated Assuming No Workhardening power lost by fluid friction at
after Necking; 3) Experimentally Obtained Curve. different rpm was determined.
The power was compared with
the pump operating with and
without water, whereupon the
installation was dried and filled with sodium. The power was then measured
with sodium let in at the freezing section, at the ` same rpm. The results obtained
(with water and sodium) make it possible to establish that fluid friction exists
in the clearance between the shaft and the housing [10]. Calculations were per-
formed to determine the thickness of the fluid film. Use was made of the formu-
la [11]
4P 5s y 2

(73)
We transform this expression
432

-â where Q is the flow rate, m3/

/hour;
2
_ 96 96v96vnD
_
7? 2u'S 2Q
3
1 J 5 10 0 50 t0` iG L9 6^^r
Substituting these values into
Eq. (73), we get
Figure 15 Axial Force on a Rod as a Func-
tion of the Axial Displacement for a Frozen- 96v.iDe Q %SQrvy
Sodium Seal (for Curve Legend see Fig. 14).

Substituting the appropriate /430
numerical values, we get
\ 'S jQ t0
^1',s1) i 1 it ::1 ti '3.1 i i 1()- 2
10 1.7. 10 M.
Thus, the thickness of the film of liquid sodium in the clearance between the hous-
ing and the shaft is about 15-20 micromicrons. This calculation should be regarded
as a qualitative estimate.
The power lost in fluid friction does not exceed 11/o, by absolute magnitude,
of the power needed for pumping the metal and depends primarily on the pump
shaft rpm This relationship, determined experimentally, is parabolic (Fig. 17),
Reference [6) presents a formula for calculating the fluid-friction power loss in
sliding-contact bearings of the segment-type
AN _= 10KSn 0,iixYCP At (74)
It is also possible to determine the friction power using the known Newton /431
formula
dS (75)
This formula can be transformed to the form
folk a11'2HIP
AN -_' =---- --
1205 (76)
433
6
3
Figure 16. Schematic of an Installation for Determining

the Properties of a Frozen-Sodium Seal for a Rotating Shaft.
1) Multipoint Switch; 2) Stroboscope; 3) d-c Motor; 4)
Seal under Test; 5) Test Volume; 6) Differential Mano-
meter; 7) Venturi Nozzle; 8) Expansion Tank; 9) Dewar's
Flask.
which is done as follows: we assume that the distribution of the peripheral velocity
is linear
F==u:tDllTl , (77)
on the other hand,
IV 67D (78)
then
siDn
F îzDII
606 (79)
we define M at the shaft
Al = FR -= F 2 (80)

AN 10jZ1W, (81)
where W is the angular velocity.
434
%1, Wt Substituting these quantities into Eq.
1`
(75), we will get a quantity expressingthe
fluid-friction power in the clearance be-
tween the shaft and pump housing. Com-
paring experimental data with results ob-
20 tained by calculations a divergence of 30%
10
is found in the region of high n. But these
differences are not of principal import-
ance, since, as was shown, the absolute
-- value of the power lost in friction does
not exceed 11/o. These differences may
1
apparently be attributed to the surface
500 600 800 7000 1200 r" 490 n, rpm finish, since it is precisely at high shaft
rpm that these exert their effect.
Figure 17. Power Lost in Fric-
tion in a Packing Seal, where o To determine the starting torque
is Water and a is Sodium. in frozen-sodium seals use was made of
the installation shown in Fig. 16; in which
a lever with a DS-0.2 dynamometer re-
placed the electric motor. When the so-
dium in the seal is liquid, the torque needed for rotating the shaft was, as expect-
ed, practically zero. Reduction in the sodium temperature to below that of solid-
ification sharply increased the required torque. In experiments to determine the
tangential stress for sodium in the frozen-sodium seal the temperature was mea- /432
sured over the height of the valve, which made it possible to determine the size
of the shaft surface which is in contact with the solidifying sodium. It was found
that the tangential stress depends on the temperature and on the starting shaft
rpm. It was also established that increasing the sodium pressure in the loop from
0 to 12 kg/cm2 does not affect the torque of the friction coefficient.
For seals for which calculations were performed above it was found in ex-
periments that Mmax - 800 kg-cm or Mmax/MT = 1.3 and Pmax = 600 kg or
Pmax/P T = 1.3. Comparison of M max and Pmax obtained experimentally and
theoretically shows that they agree satisfactorily (see Figs. 14 and 15).
CONCLUSIONS
1. A method was developed for calculating the startup forces of frozen
metal (such as Na, K, etc.) seals.
2. Stress-strain diagrams for sodium were obtained for different strain-
ing rates and different temperatures.
3. Characteristics of several designs of frozen-sodium seals were studied
experimentally. Comparison of experimental and theoretical results shows sat-
isfactory agreement.
435
REFERENCES
1. Kirillov, P. L., et al.: Atomnaya energiya, 7, 11, (1959).

Collection of articles, translated from English under the Editorship of
A. Ye. Sheyndlin. Foreign Literature Publishing House, Moscow, 1959.
3. Kristianovich, S. A., et al.: Nekotoryye voprosy mekhaniki sploshnoy sredy
[Certain Problems of Mechanics of a Continuous Medium]. Published by
the Academy of Sciences of the USSR. Moscow-Leningrad, 1938.
4. Il'yushin, A. A.: Plastichnost' [Plasticity]. Fizmatgiz, Moscow, 1948.
5. Hill, R.: The Mathematic Theory of Plasticity. Oxford University Press.
1950.
6. Kachanov, L. M.: Teoriya polzuchesti [Theory of Creep]. Fizmatgiz,
Moscow, 1960,
7. Timoshenko, S. P.: Theory of Elasticity. McGraw-Hill Book Company,
Inc. 1934.
8. Nadii, A.: Theory of Flow and Fracture of Solids. McGraw-Hill Book
Company, Inc., 1950.
9. Shaposhnikov, N. A.: Mekhanicheskiye ispytaniya metallov [Mechanical
Testing of Metals]. Mashgiz, Moscow-Leningrad, 1954.
10. Golubev, A. I.: Sovremennyye uplotneniya vrashchayushchikhsya valov
[Present-Day Packings of Rotating Shafts]. Mashiz, Moscow, 1962.
11. Drobyshev, A. V. and Turchin, N. M.: Atomnaya energiya, 10, 386 (1961).
12. Loytsyanskiy, L. G.: Mekhanika zhidkostey i gazov [Mechanics of Liquids
and Gases]. Gostekhizdat, Moscow, 1957.
13. Ayzenshteyn, A. I.: Tsentrobezhnyye nasosy dlya nQftyanoy promyshlen-
nosti [Centrifugal Pumps for the Petroleum Industry]. Gostoptekhizdat,
Moscow, 1957.
436

APPENDICES
I. CALCULATIONS OF HEAT TRANSFER IN LIQUID METALS
These recommendations were compiled by analyzing data obtained in SSSR /433

and in foreign countries up to 1 January 1966.
The formulas presented below are recommended for convective heat trans-
fer calculations for liquid metals containing admixtures below the limit of solu-
bility at the given temperature. They can be used without reservations when the
liquid metal is heated. When it is cooled and in the cases when the admixtures
precipitate from solution, depending on the specific contamination conditions and
other factors which are at present difficult to assess quantitatively, the impuri-
ties may settle on the heat-transfer surfaces and thus reduce the heat transfer
coefficient several fold.
These cases require separate examination. This remark does not pertain
to boiling and condensation processes. It should be remembered in the calcula-
tions that at the beginning (at 200C sometimes for several hundred of hours) the
heat transfer coefficient of the heat-transferring surface may be by a factor of
3-5 lower the design values until wetting of the wall is achieved.
1. Convective Heat Transfer

Round Tube
Range of appliability (Accuracy,

Theoretical formula and boundarylI 9/0
conditions
Nu = 5-rL 0.025 PA S 3.103Re 10^
(Al) 0.004 < Pr ^ 0.04 15
l/d > 30
qw= collst
Nil =- 4.36 (A2)^ I Be < 2.3. 10:1 1 10
Formula (1) yields local heat transfer coefficients and is valid for the re-
gion of steady-state heat transfer, i.e., far from the tube' inlet and from the start
437
of the heated section. To find the average heat transfer coefficient in short tubes
(5 < 1/d < 30) we must introduce the correction
C)().16
:t ^.;
t (A.3)
Studies show that e. l is a function of not only I/d, but also of the Peclet
number, and on the extent of contamination of the tube wall,
Formula (2) is applicable to the case when axial flow of heat over the fluid
and the wall are small and the effect of natural convection can be disregarded.
When axial flow of heat is taken into account, then the heat transfer coefficient
calculated from Eq. (2) should be multiplied by the correction factor 't', which
is found from the expression
LLT-
where
F,c1to
Longitudinal flow, of heat through the fluid and the walls of the heat exchanger
change the temperature fields, which should be taken into account in calculating
the mean temperature drop in the heat exchanger. For equal water equivalents
of the liquid metal, the theoretical temperature drop in the heat exchanger
where
41 h a Ft-- F'L .'
1Kutateladze, S. S. Osnovy teorii teploobmena [Fundamentals of the Theory

of Heat Transfer], Mashgiz, Moscow-Leningrad, 1962.
2 Subbotin, V.I., et al. AtomnayaEnergiya, 13, No. 2 (1962).
438

At is the temperature drop between the heat-transfer agents at infinite distance

from the heat exchanger; K is the heat transfer coefficient; d is the tube diameter
or the equivalent diameter; F w , Ff and F 1 are the cross sectional area of the wall,
of the liquid metal in the tube and of the liquid metal in the intertubular space,
respectively.
Rectangular Duct
Range of Accur-
Conditions Formula applicability racy,
Nu-==4.9-x-0.0175 Pe eB 4 . 10 2 <Pe<4 . 10 3 , t15

Heating from one side, ( A 6 ) a
qw =coast >5
b
Heating from two sides, N - - f 0 ; U.ULS T' 8 4.102 < Pe < 4 . 10 3 , 20

qw = coast (A 7)
va >5
It should be remembered, in calculating heat transfer in rectangular ducts, /435

that the wall temperature is not uniform over the duct's cross section. The dif-
ference between the wall temperature and the average temperature of the heat
transfer agent for corners of ducts can be by 40-50% higher than for their centers.
In all cases of utilization of Eqs. (A.6) and (A,7) the size of the heat-transferring
part facing the liquid-metal flow should not exceed the duct's dimensions.
Concentric and Eccentric Annular Ducts

_(
Range of ^^ Accu-
-Y Conditions ^y ( Formula applicability (racy, o
Heating from one side, Nu 5-1- 0.02 Pee s 250 < Pe < 7.5.10 3 15
qw = const (A 8) d
e =0 1.0 < d < 2,.0
!
Heating from one side,I N11=-=0.016 Peo.3 40 < Pe < 600 I 25

qw = const (A 8a)
e =0.5 100
Heating from two sides,) Nu 10- x -0.025 Pe08 300 < Po < 4 . 103 25
qw = consi (A.9)
e=0 1.05<
di < 2.0
439
Formulas (A.8) and (A.9) are applicable only for a symmetrical annulus.
In an eccentric annulus heat transfer becomes inhomogeneous to a large extent
over the angular dimension as well as over the length. The heat transfer some-
times is not stabilized at all in the case of large Peclet numbers. A certain role
is played by the wall thickness and its thermal conductivity. Estimates can be
obtained from Eq. (A.8a) for an internally heated annular duct with a wall thick-
ness of 15 mm made from 1Kh18N9T steel.
Rod and Tube Cluster

Range of Accuracy,
Conditions Formula I applicability W I
1- 9/0
Continuous rodsacked W \ 0.2 SO < Pe < 600 -1-20
closely in a triangu
lar Nu =0,15 W 1 --
lattice f /- - <16,5
0, 7 < ^`
co
0,3+
0.04 f
Pe Xf
(A10) * -
Rods arranged in a trian- Nu-==0,58 (1.1x 2 --1) 0 . 53 i00<Pe <4.103 20
gular array with P00,45 (All) * 1,1 <X<1,5
spacing x
Continuous packing of i \ u 1. ct ri , t;; ;' ' 102 < P 2 . 10 1 1_20

continuous rods in a (Al2) ?.;r,
rectangular array 0, 5 < - < ^_0
r i
^ >50
r^r
Intertubular space of dl _, ,_ 2oO < Pe < 1200 15
counterflow heat Nil --' S 0, 0^ r X
exchangers 60 <<
< 260
do
X ( X 1,1) 0.' 1 ) 1 Pe0.5
J 1.t<X<1.4
(A13)
Techniques of Heat Transfer Calculations in Rod Clusters, Arranged in a Tri- /43E

angular Array (Axial Flow) for 1:5 x <_ 1.75 and 1 < Pe < 4000.
1. For each specified spacing x Nu st and Nusq is found from Fig. A. 1.
2. Knowing the dimensions and thermal conductivities of fuel elements,
it is possible to calculate the parameters
R i ^w
fie Rf
1 Here a is the heat transfer coefficient, averaged over the angle at a sec-
tion away from the inlet a = c/tw - tf.
440
flu Figure A.2 is then used to find
t tr =f(A^12).
trust-- I-
3. Then one finds
IN = z 11, (IN 11, Nus,j)+Nu BQ . (A. 14)
4. Figures A.3 and A.4

are used to find f1 and f2.
rl_
1 1.7 1,2 1,4
5. Then Nu = f 1 Nu s + /438
1.3 1.5 7,6 7,7 x
+ f2 Pe is calculated, where Nu
Figure A.1.
is a value averaged over the
perimeter and calculated as
Nil - _ On
( 1 is - Yf) ^'1
6. Using Fig, A. 5 the max-

imum temperature. variation over
the perimeter of fuel elements is
determined for the given II I and x
I
yh
Max YrgR^
h
c
Jy
Condensation of Pure Vabor /440
Thermal resistance to heat

transfer is composed of two parts:
of the thermal resistance of the
Figure A.2. condensate film and of the thermal
resistance of phase transition
J- . R . RMI : R ph
(A. 20)
441
9.11
----------
7
it 45' 16' 5.5 ly Se
Figure A,3,
jpu 3
1. Thermal resistance of the condensate film.

a) Laminar flow Re ml = 4G/g < 1600, with R calculated as for ordinary
ml
fluids. Corrections are needed for r/C A Tml < 501
p
b) Turbulent flow for Re ml =- 4G/g > 1600; with R mi found from Fig. A. 6
for a given Re.
Labuntsov, D. A., Teploenergetika, 12, 47 (1956).
442
i ^, j i b j ^ i . I.
^^ z
a
a^ 7
yti ^, - I } I - i ^ I -
0
mi
Figure A.6
- -- x-7
7.
2. Thermal resistance of the phase transition.
7 97
1.7
71S
,
0.2 0,6 W
1 _ 2r = ( r _ Ps
, s (v __ U,)
kph> t I? Y',, l T ^1 s (A. 21)
:1 6'
Figure A.5. :V
Tranverse Flow Past Tubes
Formula Range of applicability Accuracy,
Nu=Pe o.a 102 < Pe <4 . 103 t30

(II.15) s
==1.2
d
1.1 <
rl -<
1.5
l
10
d
443
Natural Convection
Ac-
Conditions Range of
Formula applicability curacy,
9/0
Horizontal lo<( _,rPV2 < lot

cylinder N11==0,53 (GrPr2) 1/4
(A16)
Horizontal Nu ==0,35 (GrPr) 1/3 2.5.103 < GrPr2 < 3.105 5

plastic (A 17)
Combined Free and Forced Convection in a Round Tube
Formula Range Accuracy,

9/0
A (pr) ZIA, (AI8) I

--L- -
d
3 < PC < 150
where A=:0.34 and Pr=0.005
10<Z<30
:1= 0.2't and Pr==0.02
2. Transfer of Heat Attendant to Changes in the State of Aggregation

Boiling
Range of Accuracy,
I Conditions Formula I applicability I %
2/3
Horizontal surface of a==4q (AI9) O.Iatm p 1.2atm, 1. 30 for
a large volume 105 kcal/(m 2. hr) q sodium
q and
cr potassium
444

Critical Thermal Fluxes

e _=
Range of V)
Conditions ry Formula
ormula utilization racy,
I .. r
q cr = (1.5 + 1.3 p) 10 6 For sodium 0.015 atm t^0
Horizontal surface of kcal/(m2 hour)(A22 p 1.2 atm
a large volume 0.12
qcr=1.7 p 6
(10 (A23) For potassium 0.03 atm t 25
kcal/( m2 hour) p 1.4 atm
NOTATION
A - 1/427
B - height of retangular duct
d l and d2 - major and minor diameters of the annulus
d = 4F/P - hydraulic diameter of duct
F - cross sectional area of duct
I - duct length
L = 1/dh - relative length of duct
P - perimeter
R - universal gas constant
S1 and S2 - spacing along the front and into the depth of a
cluster with flow transverse to tubes
ee- correction factor for shorttubes
4 ^, tin_ r out
Z? I/ .P - dimensionless parameter, reflecting the effect
v " of natural convection.
445
II. CORROSION RESISTANCE AND TEMPERATURE LIMIT OF UTILIZATION
OF CERTAIN CONSTRUCTION MATERIALS IN A SODIUM FLOW

1. Pearlitic Chromium Steels /441
Loss of carbon, Loss of carbon,

Temp-
^ , at 600C 'Temp-
j, at 600C
after 2000 hours after 2000 hours
erature in sodium flow, I erature in sodium flow,
Steel limit of the quantit Steel limit of the quantity
of util- of carbon in of util- of carbon in a f
ization surface layer 1 ization surface layer 1 f
mm thick 1.2 ^Yy= - mm thick 1.2^ l
i
EI-802 700 53 I1Kh1M1FS 450 40
EI-993 600 5-20 EP-79 450 5
EP-38 600 5-20 1Kh2M 450 60
1Kh5MS 550 5-20 EP-78 450 60
EI-531 500 20 12KhMF 450 75
1Kh2M2F , 450 30 12Kh2MFSR %5u 85
1 Construction materials can be arbitrarily divided into those

with good corrosion resistance in sodium at the given temper-
ature and an operating time of 2000 hours, if the carbon loss
does not exceed 10%; into materials with satisfactory resist-
ance if the carbon loss comprises 10-30%, and into materials
with poor resistance on carbon losses of above 30%.
2 The experimental data were obtained for an oxygen concen-
-3 %
tration in the sodium of (2-5) 10 by weight.
3 Carbon loss after 6500 in sodium flow.
2, Nickel-Containing Steels and Alloys, Refractory Metals

Rate of corrosion
Temperature mg/(cm2.hr), in
limit of util- a flow of sodium
Material class I Material' C' at 700 0C with an
ization, o o xyge n content of
for 5000 hours 3 5).10-3
work f by weight
Stainless austenitic EI-211 S00 0,0006

steels EI-854 700 -,-O,0llll7
EP-71 700 0.OGk13
EP-148 700 0. Lk)
EI-847 700 0,002
EI-403 700 0,003
1Khl8N10T 700 0,003-0,O)l
EI-888 700 0,003-0,OJ4
EI-448 700 0,003-0.003
EI-844 700 0.003O.Ni
EI-850 700 0.003-0,0:11
EI-851 700 0.003-0.001
446
TABLE (concluded)
Rate of porrosion
Temperature mg/(cm .hr), in
limit of util- I a flow of sodium
Material class MaterialI ization, C, at 700C with an
for 5000 hours oxygen content of
work `y eight]
Nickel-base allows EI-827 SOO 0,003 O,OQ.t

EI-559A 100 o.00 't
EI-559 100 (). 006
Kh20N60M5V10Yu2T2 100 0.005
Molybdenum 100 0.0005
Refractory metals Niobium 1 WO U3
1000 0.00253
By their corrosion. resistance in sodium, the construction

materials canbe divided into those with good corrosion re-
sistance at the given temperature if the rate of corrosion
does not exceed 0.0010 milligrams/cm2 -hour; into those
with satisfactory corrosion resistance, if their corrosion
rate lies in the range from 0. 0010 to 0. 0100 milligrams/
/cm2-hour.
2 The oxygen content of the sodium should be of the order of
10 -4% by weight.
3 Rate of corrosion at 900C>
Translated for the National Aeronautics and Space Administration by Scripta

Technica, Inc., NASw-1694,
447
ABSTRACTS
Study of Heat Transfer in Liquid Metals in Round
Pipes (V. D. Talanov and P. A. Ushakov)
This paper presents experimental heat transfer data for mercury
and sodium-potassium alloy in stainless steel and copper tubes, respectively.
Empirical dimensionless equations proposedby previous investigators were
found to correlate well with the experimental data. The data showed the thermal
contact resistance to be negligible under the experimental conditions, pro-
vided the oxides of the liquid metal are in the dissolved state.
Heat Transfer with Liquid Metal in a Vertical Pipe
with Low Peclet Numbers (N. A. Ampleyev, P. L. Kirillov,
V. I. Subbotin and M. Ya. Suvorov) 8
The paper presents an attempt to clarify the cause for disagreement
inheat transfer data at low Peclet numbers obtained by previous investi-
gators. Difference from previous equations is attributed, in part, to
natural convection effects, end heat losses, entrance conditions. Absence
of thermal resistance effects was confirmed and the effect of 1/D is studied.
Heat Transfer of Sodium and a Sodium-Potassium Alloy

in Short Vertical Pipes by Combined Convection
(L. G. Volchkov, M. K. Gorchakov, P. L. Kirillov and
F. A. Kozlov) 28
Under the experimental conditions, the combined convection heat
transfer was found to be appreciably higher than in forced convection
only. In short ducts, the length was shown to have little effect on heat
transfer, although the effect remains unclear. The temperature field
is shown to agree with theoretical predictions based on the stabilized
laminar flow problem.
Taking into Account Longitudinal Heat Flows when Measuring
the Average Temperature of Fluids in Pipes and Heat
Exchanges for Low Peclet Numbers (M. K. Gorchakov) 40
The author presents a relationship between the actual average
temperature of the fluid and its theoretical value in tubes at constant
heat fluxes and in double pipe heat exchangers with identical flow rates
in both passages. An evaluation is made of the effects of shape of
velocity profile, channel geometry, longitudinal heat conduction, and
nonuniformity of heat flux.
Statistical Characteristics of Turbulent Temperature
Fluctuations in Fluid Flows (V. P. Bobkov, M. Kh. Ibragimov
The article presents experimental data on certain statistical char-
acteristics of temperature fluctuations in turbulent flow of various fluids
with highly different thermal conductivities (water, mercury, lead). The
448
intensity of temperature fluctuations, spatial correlation coefficients,
scales of temperature disturbance, correlation coefficients, probability
distribution and channel radii are considered in evaluating the various ef-
fects on heat transfer and flow.
Statistical Characteristics of Temperature Fluctuations in
Heat Exchanger Walls Attendant to High Thermal Fluxes
(M. Eh. Ibragimov, V. I. Merkulov and V. I. Subbotin) 71
The authors present some statistical characteristics, intensity and
cross-correlation function, of temperature fluctuations in the wall of
a double pipe heat exchanger with the same agent (liquid metal) flowing
through each duct, but at a different temperature. Experimental equip-
ment and collected data are described and analyzed.
Heat Transfer in Turbulent Flow of Fluids in Square Ducts
(M. Eh. Ibragimov, Yu. M. Rabotyashnik, G. I. Sabelev,
V.I. Sidirov and V.I. Subbotin) 82
Results of a heat transfer study in a square duct with water and
mercury are presented. The temperature fields in the wall and in the
fluid have been obtained as well as the statistical characteristics of
turbulent temperature fluctuations. The inapplicability of equations for
round pipes to square ducts is demonstrated. The effect of secondary
flows on heat transfer is experimentally determined. Correction factors
for square ducts are presented.
Fluid Dynamics in an Annular Space with Longitudinal Ribs
(Yu. D. Levchenko, V. I. Subbotin, P. A. Ushakov and
A. V. Sheynina) 100
Results of an experimental heat transfer study with narrow
annuli, both concentric and eccentric, are presented. Experimental
test section construction and collected data are given. The applica-
bility of previously recommended calculation methods for velocity
distribution and friction factors is shown.
Effect of Eccentricity on Heat Transfer to Liquid Metals
in Annular Spaces (V.I. Subbotin, V. D. Talanov and
P. A. Ushakov) 108
The paper describes the effect of spacing ribs and eccentricities
on the temperature field of the surface of the heat-conducting tube
cooled by liquid metal (NaK) flowing in a narrow annulus. Close agree-
ment with other experimental data for eccentric, but ribless, annuli
is demonstrated.
449
Transfer of Heat in Turbulent Flows of Fluids in Annular
and Plane Spaces (N. I. Buleyev, V.A. Mosolova and
L. D. Yel'tsova) 120
Calculations of temperature fields and heat transfer coefficients
for turbulent flow in annular and plane spaces are presented. Use is
made of integral approximations for turbulent coefficients of thermal
conductivity. The calculated heat transfer coefficients for several
fluids are compared with the available experimental data.
Experimental Study of Heat Transfer From a Flat
Horizontal Surface to Sodium by Free Convection
(A. P. Kudryatsev, D. M. Ovechkin, D. N. Sorokin,
The experimental study was carried out to compare the level of
heat transfer in boiling and in free convection of liquid metals. Experi-
mental data were correlated satisfactorily by equations previously pro-
posed by other authors. A different set of heat transfer equations ap-
plies for turbulent and laminar free convection regions.
Approximate Thermal Simulation of Cylindrical Fuel-Element
Rods (P. A. Ushakov) 136
The author uses dimensional analysis and other analytical tools
to arrive at methods of simulation of thermal conditions of fuel rods in
liquid-metal cooled nuclear reactors. The presented methods are
successfully used in simulation of actual reactor cores.
Heat Transfer From Loosely-Spaced I Fuel] Rod Clusters to
Liquid Metal Flowing in the Axial Direction (A. V. Zhukov,
V.I. Subbotin and P.A. Ushakov) 149
The present article considers experimental heat transfer data
obtained with fuel rod clusters to liquid metal in longitudinal (axial)
flow past the rods. It is assumed that temperature variation over rod
perimeter is small and that the heat transfer coefficients depend little
on rod geometry, jacket thickness, thermal conductivity of the fission-
able material and of the jacket. Empirical equations are derived and
data are compared with existing correlations.
Experimental Study of Temperature Fields of Fuel Elements,
Using Models (A. V. Zhukov, L. K. Kudryatseva, Ye. Ya. Sviri-
denko, V.I. Subbotin, V. D. Talanov and P.A. Ushakov) 170
Results of an experimental heat transfer study with simulated
fuel elements are presented and analyzed. Hexagonal housings were
used and the peripheral elements were-found to undergo appreciable
temperature variations. These temperature variations are analyzed
for different packing modes. Transient coolant flow was also studied
with the attendant effects on rod temperature. Tinning with fuel ele-
ments is considered in turbulent and laminar flow.
450
Heat Transfer in the Interbular Space of Liquid-Metal Heat
Exchanges (p. L. Kirillov and M. Ya: Suvorov) 190
The paper presents the data of an experimental heat transfer study
with axial flow of a liquid metal coolant. The correlation for average heat
transfer coefficient as a function coolant velocity and cluster geometry has
been obtained. Effect of tube'oxidation on heat transfer has been studied
and analyzed.
Analysis of the Applicability of Average Heat-Transfer
Coefficients in Heat-Exchanger Design (P. A. Ushakov) 200
This paper compares different available methods of obtaining
average heat transfer coefficients for short heat exchangers. It is
shown that it is preferable to employ average heat transfer coef-
ficients calculated on the basis of average temperature difference
between the wall and the fluid, to the use of integrated values of local
heat transfer coefficients.
Resistance to Axial Flow of Fluids Past Rod Clusters
(A. V. Sheynina) 207
Experimental data on friction factors for flow through rod
clusters are presented, analyzed and compared with existing corre-
lations. Different types of clusters and configurations were studied.
The effect of different spacers and inserts is discussed.
Velocity Distribution in a Cell Formed by a Densely-
Packed Cluster of Rods (Yu. D. Levchenko, V. I. Subbotin,
P. A. Ushakov and A. V. Sheynina) 220
The paper presents a thorough analysis of an experimental study
of xelocity distribution in ducts with noncircular cross-section. Both
velocity and shear stress as well as friction factors in ducts simu-
lating a densely-packed rod bundle.
Calculation and Experimental Study of Velocity Fields in
Intricately-Shaped Passages (M. Eh. Ibragimov, I. A. Isupov
The present article considers the fluid dynamics of a passage
formed by a cluster of three round rods surrounded by a triangular
shell. Such duct combines the fluid flow characteristics of rod
clusters as well as exposes the effect of the shell on the fluid flow.
These characteristics were studied experimentally and considered
theoretically.
Electronically-Heated Setup for the Study of Heat Transfer
From Boiling Metals by Free Convection (A. A. Ivashkevich,
V.I. Subbotin and A.A. Tsyganok) 248
The authors present a discussion of an experimental study of
electronically heated system for pool boiling of liquid metals. Very
high heat fluxes which can be easily measured were obtained.
451
Transfer of Heat From Sodium Boiling in a Large Vessel
(A. P. Kudryavtsev, D. M. Ovechkin, D. N. Sorokin,
Boiling studies with sodium were carried out at flat horizontal
plates heated by electron bombardment. The heaters were made of
stainless steel. Comparison of pool boiling data obtained in previous
experiments shows generally satisfactory agreement. Effect of dis-
solved oxygen on heat transfer is considered.
Experimental Study of Critical Thermal Fluxes in Sodium
Boiling in a Large Vessel (A. P. Kudryavtsev, D. M. Ovechkin,
D. N. Sorokin, V. I. Subbotin and A. A. Tsyganok) 281
Data of an experimental burnout study with liquid sodium on flat
horizontal surface are presented. The obtained data and data of other
researchers are still too scarce, in the opinion of the authors, to
allow the derivation of an empirical generalized correlation.
Mean Temperature Difference in Low-Pressure Vapor
Generators and Condensers (P. L. Kirillov) 287
The paper studies the behavior of liquid metal vapors as a
heat transfer medium. A new expression for the log mean tem-
perature difference is obtained for both parallel and counterflows.
Transfer of Heat From Condensing Sodium Vapor
(N. B. Bakulin, M. N. Ivanovskiy, V. P. Sorokin,
V. I. Subbotin and B. A. Chulkov) 294
The paper analyzes the effect of boundary conditions on heat
transfer in condensation of pure "stationary" sodium vapor. The
temperature field in the vapor in the liquid metal and the wall were
measured. The effect of noncondensing gases and other contaminants
is considered.
Corrosion Resistance of Construction Materials to
Sodium (V. V. Zotov, B. A. Nevzorov and Ye. V. Umnyashkin) 300
This papers reviews the available data on corrosion resistance
of materials of construction to liquid sodium. The chemical nature
of oxygen in liquid sodium is analyzed and classified. Corrosion
resistance of ferritic-pearlitic chromium steels as well as that of
austenitic stainlesss steels is presented and discussed at various thermal
and flow conditions. Also, alloys with a nickel base and refractory metals
are grouped and classified.
452
Determination of the Oxygen Content of Sodium and of the
Sodium-Potassium Alloy by Vacuum Distillation (I. B Dmitryeva,
F. A. Kozlov and E. K. Kuznetsov) 318
An experimental method for determining oxygen content in
liquid sodium is presented. Apparatus for performing the mea-
surements is described as well as analysis of the obtained results.
The Cock-Type Indicator as an Instrument for Detecting
Admixtures in Sodium (F. A. Kozlov, E. K. Kuznetsov and
V. I. Subbotin) 324
The special indicator for detecting admixtures in liquid
metals is described. The operation, construction, calibration and
other aspects of the instrument are given in great detail. Experimental
data obtained with the instrument by different investigators are plotted
and compared.
Effect of Admixtures on the Electric Resistance of Molten
Lithium (M. N. Arnol'dov, M. N. Ivanovskiy, V.I. Subbotin
and B. A. Shmatko) 336
The effect of admixtures, such as nitrogen, oxygen and hydrogen,
on the electrical resistivity of liquid lithium is described. The tem-
perature dependences of the admixture concentration and type of ad-
mixture are presented.
The Use of Hot Traps for Removing Oxygen From Sodium
(F. A. Kozlov and E. K. Kuznetsov) 340
The paper gives a thorough description of theoretical and
practical aspects of the use of hot traps for oxygen removal from
sodium. The reaction of zirconium claddingwith the oxygen is con-
sidered as a function of time of exposure, temperature, concentration,
liquid metal velocity, changes in chemical composition. Methods of
testing the cladding by x-ray diffraction and metallographic analysis
are described along with the changes in its electrical resistivity.
Maximum Permissible Leakage in a Sodium-to-Water Steam
Generator (N. N. Ivanovskiy and F. A. Kozlov) 357
The paper presents a set of conditions for calculating the maxi-
mum permissible leakage in a liquid sodium steam generator. Safety
factors are considered.
Some Problems of Operation and Design of Liquid-Metal
Loops (A. V. Drobyshev, V. A. Kurov and I. T. Filipov) 360
Operation and design problems of liquid metal loops operating
with forced coolant circulation are described. Auxiliary systems,
instrumentation, safety measures, equipment layout and costs are
included in the brief review.
453
Heat Exchangers of Experimental Liquid-Metal Loops
(V. A. Kurov and V. F, Men I shikov) 368
The paper presents design criteria and operating procedures
for heat exchangers associated with liquid metal loops. Liquid-
metal-to-water coolers, liquid-metal-to-air coolers, as well as
liquid metal-to-liquid-metal exchangers and recuperators are
described.
Study of Velocity Fields by the Electromagnetic Method
(N. N. Turchin and R. V. Shumskiy) 377
The present paper presents experimental data on velocity field
in a round tube and an annular channel. Data for laminar and tur-
bulent flows are given and other parameters, such as end effects,
are considered.
Electromagnetic A-C Pump for Liquid Metals Rated at
120 m3/hour (Ye. M. Avilova, T. V. Doktorova, N. I. Marin
and V. A. Povsten') 385
The physical features, design characteristics and performance
data for a 120 m3/hr electromagnetic A-C pump for circulating liquid
metals are presented.
Electromagnetic Screw Pump for Pumping of Liquid Metals
(Ye. M. Avilova, Ye. L. Ivanter, N. I. Marin and V.A. Povsten') 389
The design features and operating characteristics of an electro-
magnetic screw pump are given. The pump is well suited for lab-
oratory-scale tests with liquid metals.
Hydrostatic Bearings in Sodium Pumps (N. N. Aristarkhov,
Yu. Ye. Bagdasarov, Yu. L. Kulikov, Ye. N. Larin,
M. S. Pinkhasik, A. A. Rineyskiy and B. S. Tymosh) 392
The design of hydrostatic bearings for sodium pumps is con-
sidered. Operating principles of the bearings, temperature re-
sistance, pressure, wear and noise were tested in a sodium loop.
Creep was also considered at temperatures up to 400C.
Flowmeter for Measuring Large Liquid-Metal Flows
(N. L. Loginov) 403
The device described in this paper is a flowmeter which operates
by measuring the velocity at one point in the flow. The size of the
instrument is moderate in comparison with the size of the pipeline.
It is intended for operation at 50-1000 m 3 /hour of a liquid metal
flowing in a pipeline 300 mm in diameter. Other operating param-
eters are also given.
454
Time Needed for Melting Cylindrical [ Metal l Volumes
(V. M. Selivanov, N. M. Trebina and V. I. Sharypin) 406
The paper presents theoretical data and nomograms for cal-
culating the time needed for melting of metal cylinders, starting
at a temperature below the metal melting temperature. Cases of
constant temperature, constant heat flux and constant heat transfer
coefficient are considered (boundary conditions of the first, second
and third kind, respectively).
Frozen Sodium Seals (A. V. Drobyshev, Yu. I. Likhachev
and I. M. Saratov) 412
The paper presents the results of a thorough engineering study
of frozen liquid metal seals. A method for calculating the startup
forces of such seals is given. Stress-strain diagrams for sodium
are given for different strain rates and different temperatures.
Characteristics of several designs of frozen-sodium seals were
studied. Comparison of theoretical and experimental data shows
satisfactory agreement.
Appendix I
Calculations of Heat Transfer in Liquid Metals 437
This appendix presents recommendations for calculating heat
transfer to liquid metals based on world-wide data published up to
January 4, 1966. Equations are given for convective heat transfer
in mund tubes, rectangular ducts, concentric and eccentric annuli,
rod and tube clusters. Equations for condensation of liquid metal
vapor are presented. Combined free and forced convection are given.
Appendix II
Corrosion Resistance and Temperature Limit of Utilization
of Certain Construction Materials in a Sodium Flow 446
This appendix gives tables of corrosion resistance to liquid
metals of pearlitic chromium steels and nickel-containing steels
and alloys as well as refractory metals.
455

19690021129

Uploaded by

Copyright:

Available Formats

19690021129

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

19690021129

Uploaded by

Copyright:

Available Formats

TECH LIBRARY F ,N

By P. L. Kirillov, V. I. Subbotin, and P. A. Ushakov, Editors

Translation of "Zhidkiye Metally."

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION

Experimental Study of Heat Transfer From a Flat

Flowmeter for Measuring Large Liquid-Metal Flows

q, qv, Specific heat load.

a Coefficient of volumetric expansion

At Mean temperature difference, C

The production of atomic [nuclear] power may become an important branch

VERIFIED INFORMATION AVAILABLE FROM OTHER SOURCES. la

L^ IS UNSATISFACTORY BECAUSE OF (Check one or more): i

POOR TERMINOLOGY. 0 NUMERICAL INACCURACIES. I

This collection presents results of original studies performed during the

Numbers in the margin indicate pagination in the foreign text.

Nu = 4.3 + 0.025Pe O.8 . (1)

Nu = 5 4 . 0.025Pe O.8 . (2)

Figure 1. Schematic Diagram of the Test Pipe. 1) Thermo-

In order to check the readings of the movable thermocouples, use was

Nu = 5.2+ 0.025Pe- 8 ; (4)

Nu=5+0.014Pe O.S (5)

Figure 2, Temperature Distribution Along

TABLE 1. Heat Transfer Experiments Using a Sodium-Po-

q X 10- 91 ^('wt f)> Re X Pe I Nn Re X 10-3 Pe

17,12 2.9 3.22 80 5.7 62,3 6.0 35.4 797 10,0

10 20 40 60 80 100 200 400 600 800 1000 2000 4000 NO 8000 Pe

134 6.2 692 8.9 1290 11.7 1995 14,5

N. A. Ampleyev, P. L. Kirillov, V. I. Subbotin

In order to evaluate this integral, in addition to the velocity distributionuQ),

s e = f (Re, y). (2)

Other theoretical solutions reduce, in one way or another, to the evaluation of

Nu = 5 + 0.0021 Pe. (4)

Nu=Nuo+AN O.8 (7)

Nu = 4.36 + 0.343Pe0 4 , where Pe* = U*v*d .

It was discovered in [19] by measuring the temperature distribution in the

The disagreements in the thermal balance

t nl tmch. ate t out = tmch.+ tlos

The heat losses were determined from the formula

0< y+<5 u+=y+

or by the power law

The curve of Re cr = f(Z), according to data of [24] is shown in Fig. 4.

Figure 4. The Critical Reynolds Number as a

w i _ Z r ber (Z') i bei (Z) ber (Z) bei (Z') (^7)

Exn= 89 = 0.4v* rod ( 1 s),

Differentiating Eq. (18) we get

Figure 6. Experimental Data Using the Coordinates

Figure 7. Results of Heat Transfer Experiments in a Pipe.

curves and calculated the coefficients s l = Nuloc/NuOD, which characterize the

Figure 8. Characteristic Temperature Distributions Along

l Id Figure 10. Length of the

thors; the Black Circles Are

Figure 11 Change in the Relative Heat Transfer at the In-

Figure 12. Comparison of Experimental Data

e = e q/e M ratio of turbulent transfer coefficients

y = y v*/v distance from wall

1. Lyon, R. N.: Chem. Eng. Progr. 47, 75 (1951).

Central tube diameter, 102 I 68 I 135

Wall material Copper I1Eh18N9Tj Copper

Width of annular space 5.0 6.0 7.5

* [Translator's Note: Also called regenerators].

Nu = 4.36 + 0.343Pe0 4 , where Pe* = Uvd .