3. Gamson, B. W.

, George Thodos, and

0. A. Hougen, Trans. Am. Inst. Chem.
Engrs., 39, 1 (1943).
4. Gilliland, E. H., and T. K. Sherwood,
Ind. Eng. Chem., 26, 516 (1934).
5. Glaser, M. B., and George Thodos,
[Link].E. Journal, 4, 63 ( 1958).
6. Lee, C. Y.,and C. R. Wilke, fnd. Eng.
Chem., 47, 1253 (1955).
7. Lynch, E. J., and C. R. Wilke, Uni-
versity of California Radiation Labora-
tory, UCRL-8601 (January 14, 1959).
8. Maisel, D. S., and T. K. Sherwood,
00
Chem. Eng. Progr., 46, 131 (1950).
NR,
9. McAdams, W. H., “Heat Transmission,”
Fig. 4. Comparison of j-factor for single spheres resulting from 2 ed., p. 236, McGraw-Hill, New York
this study with other available date. ( 1942).
t = dry-bulb temperature, O F . LITERATURE CITED [Link], G. C., Sc. D. thesis, Mass.
t, = surface temperature, “F. Inst. Technol., Cambridge (1942).
1. Baumeister, E. B., and C. 0. Bennett,
t;, = wet-bulb temperature, OF. [Link].E. Journal, 4, 69 (1958).
P = viscosity, 1 b . h . ft. 2. Chilton, T. H., and A. P. Colburn, Manuscript received October 26, 1959; revision
received April 7 , 1960; paper accepted April 8,
P = density, lb./cu. ft. lnd. Eng. Chem., 26, 1183 (1934). 1960.

Size Distribution of Droplets from Centrifugal

Spray Nozzles
PAUL A. NELSON and WILLIAM F. STEVENS
Northwestern Technological Institute, Evanston, Illinois

Methods for expressing, measuring, and correlating drop-size distribution data for centrifugal
spray nozzles are discussed.
A method for collecting spray droplets in liquid nitrogen is described which is rapid and
efficient for most sprayed liquids which freeze above -2OOC. Comprehensive correlations for
drop-size distributions are reported based on 114 runs performed with the liquid nitogen method.

Studies of the kinetics of evaporation has not been derived as yet. The best will not fit the distribution. Three s u e
or combustion of droplets issuing from that can be done with experimental characteristics which have been suc-
a spray nozzle require data on the size data is to express the distribution cessfully used are the log of the drop
distribution of the droplets. Unfortu- graphically or to specify a distribution diameter ( 6 ) , the upper limit charac-
nately these data are not easily obtained function and the values of the parame- teristic ( l o ) , and the square-root of
because the droplets are very minute ters in the distribution equation which the drop diameter ( 1 5 ) .
and difficult to sample; also it is diffi- will most closely represent the data. The log-normal distribution and the
cult to express and correlate drop-size Several of the distribution functions square-root normal distribution have
data because the drops issuing from a which have been proposed are based two parameters, a mean and a standard
spray nozzle are nonuniform in size. on the normal distribution, and only deviation, which can be adjusted to fit
The purpose of this paper is to re- this type will be discussed: a particular set of data. The upper-
port on methods of measuring, express- - limit distribution has a third parameter,
ing and correlating drop-size data du
f,(y) =-=-
1
e
--
(Y-UP

2$
the maximum stable drop size, which
which the writers found to be success- permits more flexibility in fitting exper-
ful in a recent study of grooved-core
Vdy s$G imental data.
centrifugal spray nozzles. (1) From a theoretical standpoint the
Equation (1) is the normal distribu- upper-limit distribution has another
D I S T R I B U T I O N FUNCTIONS tion equation expressed in terms of the advantage over the log-normal distri-
volume of droplets. bution and the square-root normal dis-
A theoretical distribution function The size characteristic is a function tribution in that it places reasonable
for expressing drop-size distributions of the drop diameter. However if the limits on the minimum and maximum
Paul A. Nelson is with the Chemical Engineer- drop diameter itself is chosen as the drop size. However despite the fact
ing Division, Argonne National Laboratory, Le-
Inont, Illinois. size characteristic, experimental data that the extreme ends of the distribu-

Page 80 [Link].E. Journal March, 1961

tion curve derived from the square-root as a freezing medium for sprays was Some of the finest frozen droplets were
normal function lead to physical ab- A. P. R. Choudhury at Northwestern Uni- washed through all of the sieves into the
surdities, the square-root normal dis- versity ( I ) who found that the frozen container below by the flow of liquid ni-
tribution has been found to be useful particles dried almost immediately when trogen. This material was recovered by
in expressing data from centrifugal the liquid nitrogen was decanted, perinit- filtering the liquid nitrogen which had been
pressure nozzles within two standard ting dry screening in a cold room. How- collected through a tared filter paper. The
ever it was not possible to collect sprayed filter paper and contents were later
deviations of the square-root mean. I n liquids with densities less than 1.2. weighed with the bottom pan.
Figure 1 the three distributions are In order to remove the density restric- After the spray had been turned off
compared. The upper-limit function tion special equipment was designed and and the inside of the collector thoroughly
and the square-root-normal function built for freezing the droplets. A diagram rinsed with liquid nitrogen, the sieves were
approximate a typical experimental of the spray collector is shown in Figure 3. transferred to an insulated container with
drop-size distribution for a centrifugal The spray collector was an insulated cham- enough liquid nitrogen to keep the drop-
nozzle. However the log-normal distri- ber having an inside diameter of approx- lets frozen during a half-hour shaking
bution deviates quite far from the imately 20 in. Liquid nitrogen sprayed period. The equipment employed for shak-
other functions which represent actual from perforated copper tubing near the ing was rebuilt to accommodate the insu-
top of the chamber at a rate of 3 to 5 lated sieve container.
data more closely. liters/min. and flowed down the inside At the end of the shaking period the
The square-root-normal distribution surface of the collector. The surface of sieve container was removed to a cold box
was found to be useful in expressing the collector within the spray pattern of a shown in Figure 4. The 2-ft. cubical cold
the data obtained in this investigation, centrifugal nozzle was completely covered chamber was cooled by cold air circu-
in addition to the many times it has with a flowing film of liquid nitrogen. lated by a fan through a dry ice chamber.
been used previously by other investi- Droplets sprayed into the chamber were The sieves were weighed in this cold box
gators (1,2,8,15). Therefore a method at least partially frozen by the cold at- on a triple beam balance which was sup-
was sought and developed for deter- mosphere evolved from the evaporating ported on top of the cold chamber by a
nitrogen and completely frozen and car- sheet of clear plastic.
mining the Sauter mean diameter from ried out of the chamber by the liquid ni- The main features of this method are
the volume median diameter (the trogen running down the walls of the the following:
square of the square-root-normal mean) collector. The droplets were screened from 1. Essentially all of the spray is col-
and the square-root-normal standard the liquid nitrogen flowing from the col- lected, thus minimizing sampling errors.
deviation. The Sauter mean diameter is lector by nine sieves which were 8 in. The average recovery of material spra ed
of interest because the surface-to-vol- in diameter by 1 in. high. During the test was 91% for the 111 runs for which &ta
ume ratio can be calculated directly gaseous nitrogen at room temperature was were available.
from the Sauter mean by the following admitted above the nozzle at a rate slow 2. The only restriction on the material
relationship : enough not to affect atomization, in order sprayed is that its melting point must be
to push back the cold atmosphere from high enough to permit screening and
the area around the nozzle. Temperature weighing. Runs were made with materials
measurements indicated that the proper- melting as low as -20°C. with ease.
ties of the gas in the immediate area of 3. Shattering of the droplets as they
the spray nozzle were those of the gas strike the collecting medium is minimized
admitted above the nozzle ( 11). As soon because the particles are partially frozen
as the collector had been cooled to the in the cold atmosphere inside the collect-
temperature of boiling liquid nitrogen, the ing chamber before they strike the liquid
solenoid feed valve was operated and the nitrogen surface.
The desired relationship was derived 4. The screening technique is limited
nozzle began to spray. The first and last
from the square-root-normal distribu- part of the spray was prevented from go- to sprays with a volume median drop di-
tion function and Equation ( 2 ) : ing into the collector by a drawer which ameter greater than 38/r, although finer
slid on tracks just beneath the nozzzle. droplets could be photographed and meas-
The cut-off drawer operated a switch ured.
- &
.
which actuated an electric timer, thus 5. The method is relativeIy rapid and in-
XU8 = expensive. It was possible for a single op-
S"' 1 1 -1
-e ' dz
$z
permitting calculation of the amounts
sprayed from knowledge of the flow rate. erator to make three to four runs per day.

'1) [ sz+ 1 2

dx TABLE1. NOZZLECHARACTERISTICS
(3)
Spraying System Co., Type SL Spray Drying Nozzle
The integral was evaluated-graphi-
Capacity,
gal./hr .
/-
tally for several values of s/V x as de- Core of water
scribed in detail by Nelson (11) . The Orifice
Diameter, Flow area, Spray at 1000
results are plotted in Figure 2. From No. in. NO. sq. in. angle Ib./sq. in.
this graph the Sauter mean diameter
can be obtained from the volume 80 0.0135 10 3.6 x 56 4.0
median diameter and the square-root- 16 7.7 52 5.1
normal standard deviation.
74 0.0225 20 12.4 62 9.9
67 0.032 20 12.4 74 14.1
EXPERIMENTAL PROCEDURE AND 16.7
17 15.4 70
EQUIPMENT 28 64 24.5
21
To obtain the experimental data of this 17 15.4 81 22.2
58 0.042
investigation spray droplets were frozen
by a technique employing liquid nitrogen. 54 0.055 17 15.4 91 29.6
Others have used various frozen-drop 21 28 85 45.3
methods with liquids freezing both above 27 48 73 70.2
( 4 , 5 ) and below ( 1 , 7, 16) room temper-
ature. The first one to use liquid nitrogen 46 0.081 27 48 84 112

Vol. 7 , No. 1 [Link].E. Journal Page 81

 tion of fuel oils having viscosities I.o

Square-root -normal distribution

greater than 10 centipoises.
Upper-limit disiribuiion
98
Log-normal distribution
EXPERIMENTAL RESULTS
95 0.9
E 90 The most important evidence of the
0
method's validity are microphotographs
which show that the screened particles
E 6 0 are round and fairly uniform in size. 3€
$ 5 0 Y 0.8
40 One of the pictures, taken by the
2 % method of Remus ( 1 2 ) , of particles
- 20
5 10
screened through the 74-p sieve and
0
collected on the 6 2 - p sieve is shown in
5 07
Figure 5. On a weight basis there is
2
I
only a small percentage of undersized
particles. The fuzzy appearance of the
0.1
particles is due to water vapor that was 0 0.1 0.2 0.3
I
observed to condense before the picture
was taken. No broken particles or ag- 7?
Drop Size, Wrons
glomerations of smaller particles were Fig. 2. Relationship between the Sauter mean
Fig. 1. Comparison of square-root-normal, up-
evident in any of the microphotographs diameter and the volume median diameter and
per-limit, and log-normal distributions. of screened particles. square-root-normal standard deviation.
For the 111 runs for which recovery
data were available the average re- vs. the square root of the sieve size on
covery was 91.1% of the material normal probability paper as shown in
CHOlC,E OF NOZZLES AND
sprayed. One of the most common Figure 6. In all cases a straight line
MATERIALS INVESTIGATED causes of low recovery was loss of was obtained within the accuracy of
A wide range of nozzle sizes, spray- material from between the sieves dur- the data. The volume median was ob-
ing pressures, and liquid properties ing shaking due to inadequate clamp- tained by squaring the value of the
were covered in the 114 runs made by ing of the sieves. Complete experimen- square-root of the drop diameter at 50
this technique. The nozzle chosen is of tal and calculated data for the 114 cum. wt. % . The square-root-normal
the grooved-core type with interchange- successful runs have been tabulated standard deviation was determined
able orifice inserts. and are on file with the American from the graph by subtracting the
The manufacturer's data (14) for Documentation Institute. * square-root-normal mean from the
the orifice inserts and cores employed For all of the materials and nozzle square root of the drop size at one
in this study is shown in Table 1. combinations investigated the data ap- standard deviation above the mean
The type of nozzle chosen for this peared to fit the square-root-normal which corresponds to 84.13 cum. wt.
study has been employed by other in- distribution function. This was deter- %.
vestigators who showed that it has mined by plotting the cumulative mass It is known that sieves finer than 200
many good properties. Mclrvine (9) Q Tabular material has been deposited as docu- mesh are not very accurate, but in this
showed that for the viscosity range 1 ment 6530 with the American Documentation investigation it was necessary to use
Institute, Photoduplication Service, Library of
to 10 centipoises, viscosity does not Congress Washingon 25 D. C., and may be sieves ranging down to 400 mesh.
obtained' for $2.50 for ihotoprints or $1.75 for
appreciably affect nozzle capacity, 35-mm. microfilm. However it may be observed from
cone angle, or air core diameter. Dar-
nell ( 2 ) showed that the capacity is
proportional to the square root of the
pressure and that the diameter of the Solenoid
/Feed Volve
air core is independent of pressure.
Thus it was possible to vary the axial
velocity without changing the cone
angle or the air core diameter. Switch Operoting
The axial velocity was varied by
changing the pressure on the feed tank
between pressure limits of 100 and
1,500 lb./sq. in. gauge. The pressure
was measurable to within 1% by ac-
curate gauges.
On the basis of their physical prop-
erties the following seven liquids were
chosen for spraying: cyclohexane, n- lnsulotion -.-
octyl alcohol, carbon tetrachloride,
water, nitrobenzene, aniline, and 1,1,2,
2-tetrabromoethane. The melting points
of these materials vary from -22.6' to
6.5"C. The viscosity of these liquids
varies tenfold, the density varies four-
fold, and the surface tension threefold. '--Sieve Support
The properties of the materials sprayed
covered ranges which include almost Receiver
all materials presently being sprayed
with pressure nozzles with the excep- Fig. 3. Diagram of spray collector.

Page 82 [Link].E. Journal March, 1961

Figure 6 that the inaccuracies of the /Triple Beom Bolance rings placed beneath the nozzle (2,
sieves were partially compensated for 13, 1 5 ) . However at pressures above
by drawing a straight line through the
points. The points for a given sieve are
Piano Wire --. 200 Ib./sq. in. gauge it is doubtful that
such data would correIate with the
either consistently above or consist- flow characteristics in the nozzle be-
ently below the lines of best fit. For cause of air turbulence which disperses
instance the points corresponding to the droplets. I t is believed that a better
the 230-mesh sieve were always above index of the true V J V , ratio is the
the curve of best fit, and the points maximum cone angle which can be
corresponding to the 325-mesh sieve readily measured photographically or
were always below. It was assumed with calipers; that is
that the deviations of the data from the
lines of best fit were due mainly to in- Orn=f(x)
V,
accuracies in the sieves.
Only a few duplicate runs were
made during the investigation, so little Fig. 4. Diagram of cold chamber for weighing
The first group in Equation ( 5 ) ap-
information was gained from these runs sieves.
pears to have little effect on the air
concerning the reproducibility of the core to orifice ratio as indicated by ex-
data. Howeyer graphical comparisons perimental data. For instance Darnell
of runs made at varying pressures with The effects of air resistance inside (2) showed that beyond pressures
the same nozzle and material indicate the core, surface tension, and wettabil- where the cone angle is independent of
that, under these conditions, the stand- ity of the fluid for the nozzle were pressure change the air core diameter
ard deviation for the volumemedian neglected. When combined by dimen- was also unaffected by pressure change.
diameter and the square-root-normal Also the data of McIrvine (9) indi-
standard deviation were approximately cated only a slight decrease in the cone
3 and 8 % respectively. angle with increasing viscosity. There-
fore the Reynolds number was ne-
CORRELATION OF THE DROP-SIZE glected. In addition the roughness
DATA group e / D , was neglected in order to
simplify the relationship.
The data obtained in this investiga- Since the other dimensionless groups
tion were correlated by dimensional were eliminated, D J D , was plotted
analysis because the problem of liquid against 8, (assumed to be a function
breakup from a centrifugal pressure of V , / V , ) as shown in Figure 7 for
nozzle is very diacult to approach the data of Darnell ( 2 ) and McIrvine
theoretically. Fluid flow through cen- ( 9 ) . The boundary conditions at air
trifugal nozzles has not yet been ade- Fig. 5. Microscopic view of screened particles.
core to orifice diameter ratios of zero
quately treated, but even if the veloc- and one were helpful in determining
ity and pressure distributions in the sional analysis these variables lead to the general shape of the curve. Each
nozzle were known, it would still be point was calculated by averaging the
the dimensionless groups shown in the
difficult to relate them to the drop-size following relationship: data for D J D , at six different pres-
distribution obtained. sures, 100, 400, 700, 1,000, 1,300, and
The basic assumption in the correla- 1,600 lb./sq. in. gauge except for the
tion of the data is that the flow charac- smallest nozzles of 0.0135- and 0.020-
teristics of the liquid flowing through in. diameter for which the D J D , ratio
the orifice can be defined by variables The velocity ratio V J V , has been was averaged for the pressures 700,
which do not include nozzle dimen- obtained through measurement of the 1,000, 1,300, and 1,600 lb./sq. in.
sions related to a particular design; mass of spray collected in concentric gauge and for 400, 700, 1,000, 1,300,
that is all flow characteristics peculiar and 1,600 Ib./sq. in. gauge respec-
to a nozzle design are damped out by 99.9 tively (11) . When one considers the
the high turbulence in the flow through difficulty in measuring the data and the
the orifice. This assumption was used 99 fact that the roughness group was
to correlate the air core diameter data 98 neglected, the correlation is satisfactory
of Darnell (2) and McIrvine (9) in 95 and tends to corroborate the assump-
order to indicate the feasibility of this
approach for the problem of correlat-
-
i:
90
80
tion that the characteristics of the flow
through the orifice are determined by
ing the drop-size data. 270 the variables in Equation ( 4 ) .
The diameter of the air core formed %
60
50
It should be noted that this result
at the axis of the nozzle is affected by 3 40 does not disagree with the theoretical
the variables shown in the following ._
P 30 work of those who relate the air core
relationship: 20
diameter to nozzle dimensions only
[Doumas and Laster ( 3 ) ] , for the
5 cone angle is itself a function of the
The geometry and dimensions pecu- nozzle dimensions only, over a wide
I
liar to each particular centrifugal noz- range of liquid properties as demon-
zle and the pressure drop through the strated by McIrvine (9).
0.1
nozzle have not been neglected be- It was assumed that the same type
cause these characteristics will be re- Square Root of Sieve Size, (Microns)!+
of approach would be successful for
flected in the average tangential veloc- correlating the drop-size distribution
ity, the average axial velocity, the Fig. 6. Drop-size data illustrating the good fit data obtained in this investigation by
roughness, and the orifice diameter. to the square-root-normal distribution. the frozen-drop technique which was

Vol. 7 , No. 1 [Link].E. Journal Page 83

 1.0 Y = - 0.0811Z2 + 0.1242- 0.186
0.9
where
08
07 Y = logl- X
0.6
DO
3
Do
0.5
0.4
0.3
0.2
Z = log, [ N R e ( N,,
N,,
Y'' (;>TI
01 The average deviation of the experi-
n mental volume median diameter from
0 20 40 60 80 100 120 140 160 this curve was 8.25% for a range of
Maximum Care Angle, ern volume median diameters of from 39.6
to 1 6 2 . 5 ~ The
. deviations did not ap-
Fig. 7. Ratio of air core to orifice diameter vs. the maximum
cone angle.
pear to be related to the orifice diame-
ter, the spray cone angle, the particular
liquid sprayed, or the pressure drop
through the nozzle. Each particular
previously described. In this case a facturer's data also were used for the nozzle gave a CUwe which was slightly
separate correlation must be derived orifice diameter. The axial velocity was different from the curve for h e entire
for each parameter in the distribution calculated from measured flow num- data, but the difference could not be
function describing the data. Also an- hers for each liquid and nozzle cornhi- related to the nozzle properties of ori-
other variable must be considered, the nation according to the following equa- fice diameter and spray cone angle.
surface tension of the liquid sprayed, tion: Consequently the deviations from the
because the surface involved in drop
v. = FnVT curve of Figure 8 appeared to be due
formation is much greater than that in-
valved in formation of the air core. For
a distribution function expressed by
-
, D o 2 [ 1-
4
I'>'( ( 13) to experimental error, slight irregulari-
ties in the nozzles, the neglecting of
pertinent variables from the dimen-
parameters, such as a mean Or The D,/D, ratio was calculated from sional analysis, and the approximation
median and a standard deviation, two through Darners data (2). made in plotting a four-dimensional
independent from H~~~~~~it could have been obtained relationship [Equation (12)] on a
this approach:
-x = fW,, V,, D,, p, p , CT, e ) (7)
from Figure 7 with about the same
results.
two-dimensional graph.
The data for the volume median
By trial-and-error plotting of the drop diameter for water sprays did not
'=g("$ V 1 7 D 0 7 p 7 P , C T , e ) ( 8 ) data it was shown that the volume agree with the data for the organic
By dimensional analysis the number median drop diameters obtained by liquids. If plotted on Figure 8 the data
of independent variables can be re- spraying six organic liquids with eleven for water would lie 30 to 50% above
duced from seven to four: different nozzle orifice and core insert the curve of best fit for the organic
-X DoVapV2D.p Vt
combinations over a pressure range of liquids. There is a possibility that the
100 to 1,500 Ib./sq. in. gauge could be water did not wet the nozzle wall and
c ' V.' 2
Do1 correlated according to the relationship that this caused the difference. The
(9) of Equation ( 9 ) . The results are shown data obtained in thirteen runs with
in Figure 8. The equation of the curve water were plotted with the data ob-
-=$[-,---
S

DO'
0.V.p
P
V."Dop Vt e
u ' V,'D.
] of best fit is tained by Darnell (2) with the same

(10)
where t has a value such that s/D.' is 4
dimensionless. These equations were
3
simplified by neglecting the roughness
factor POUP e/D,. By reasoning in a 2
manner similar to that leading UP to
Equation (6), the group V , / V , was
assumed to be a function of [v,/v.].
only, which is defined by 10-1
-
X I 3
DO

2 (11) 6

4
BY making these simplifications and Carbon tetrachl
considering the square-root-normal dis- 3
tribution one obtains
2 A Nitrobenze
L

X
D" 10-2
(12) 103 2 3 4 5 6 7 8910' 2 J 4 5 6 789105

In this investigation the manufac-

turer's data for the maximum cone
angle, listed in Table 1, were employed Fig. 8. Correlation of volume medion diameter for liquids other than
to calculate (Vt/V,) +. The manufac- water.

Page 84 [Link].E. Journal March, 1961

 range of square-root-normal standard
deviations of from 0.80 to 3 . 4 4 ~ ~ 'It.
is apparent that the data points are not
Do. in-4, degrees Do, in-@,degrees
as well distributed about the curve of
best fit as for the correlations of the
0 ,0135-52 A ,040-79 median. For instance the points repre-
0 .0235-62 A ,040-83 senting carbon tetrachloride lie above
b ,042-81 0 ,040-73
the curve, while those representing cy-
clohexane lie below the curve, with
few exceptions. However it is very dif-
ficult to determine the standard devia-
tion accurately. Slight discrimination
against a size class of particles, say the
smaller particles, by the sampling or
measuring techniques will not greatly
affect the general shape of the distri-
bution curve or the value of the me-
dian, but it may have a drastic effect
on the standard deviation.
It would have been possible to sub-
stitute the superficial average velocity
for the calculated average velocity be-
103 2 3 4 5 6 7 8 9 104 2 3 4 5 6 789105 cause V,' is a function of variables
which have already been considered:
V,' = f(V,, D o , D,) . A rough analysii
Fig. 9. Correlation of volume median diameter for water.
of the data has shown that they could
have been correlated with dimension-
less groups based on the superficial vel-
type of nozzle, as shown in Figure 9. ocity with results as good as were ob-
Darnell's data were obtained by photo- tained with the calculated average
graphing droplets collected in an im- velocity (but with different correlation
miscible liquid at 10 in. below the Within the accuracy of the data the ef- curves resulting).
nozzle. fect of ( V , / V a ) *
was negligible.
The equation for the curve of best The equation for the curve of best EFFECT OF OTHER VARIABLES
fit for the water data is similar to that fit is
obtained for the organic liquids: Y = 0.150Z2- 0.3592 0.986 + It was assumed that the properties
of the gas medium have little effect on
(17)
Y = - 0.1442' + 0.7022 - 1.260 where
atomization at injection pressures be-
low 1,500 lb./sq. in. gauge. Most of
(15) the runs were conducted with nitrogen
where
w gas surrounding the nozzle but two
X runs were conducted at 1,200 lb./sq.
Y = log, - in. injection pressure in which helium
DO
gas was admitted through the gas inlet
The average deviation of the data above the nozzle. A comparison of
from this equation is 13.0% for a these runs with a similar run made

The average deviation of the data from 102

this equation is 6.7%.
8
Adopting the same simplifications as
were utilized for obtaining Equation 6
(12) from Equation (9) one obtains
for the square-root-normal distribution 4
from Equation (10)
3

& b?
(16)
By trial-and-error plotting of the 10'
data it was shown that all of the 8
standard deviation data of this investi-
gation (for both organic materials and 6
water) fit one form of this equation.
The results are shown in Figure 10 in 4
which 3
3 4 5 6 7 8910' 2 3 4 5 6 78910' 2 3 4 5 6

was plotted against Fig .lo. Correlation of square-root-normal standard deviation.

Val. 7 , No. 1 [Link].E. Joumal Page 85

with a nitrogen atmosphere indicated = number of droplets
y, = vx- = size characteristic for
that the properties of the atmosphere = total number of droplets in square-root-normal distribu-
had no effect on atomization for cen- liquid sprayed tion
trifugal nozzles spraying at moderate = Reynolds number, D,V,p/p = In x/xm -x = size character-
y.
pressures. = Weber number, V."D,p/a istic in upper-limit distribu-
A little information was obtained = radial distance from center tion
of orifice -
during the investigation about the ef- - /w
fect of nozzle roughness on atomization, = air core radius z = vx-vx/s,
but it was not extensive enough to in- = orifice radius Z = various groups of independ-
clude in the general correlations. For = standard deviation, and ent variables as defined in
instance it was noted that corrosion of square-root-normal standard Equations (14), (15), and
the nozzle parts which caused rough- deviation (17)
ness resulted in larger mean drop sizes. = surface to volume ratio of a
A nozzle that was unintentionally pol- spray Greek Letters
ished with grinding compound pro- = a constant in Equation (10) em = maximum cone angle deter-
duced a finer spray. Also use of a which is adjusted so that mined by measuring with
tungsten carbide orifice insert which s/D.' is dimensionless calipers directly on the spray
appeared to be smoother than the = volume of droplets or by measuring a photo-
standard steel insert resulted in a finer = axial velocity at some radius, graph of the spray
spray (11). r, within the orifice of the P = density of liquid sprayed
In conclusion the method of correla- nozzle Q- = surface tension of liquid
tion utilized in this investigation ap- = tangential velocity at some sprayed
pears to be valid, because satisfactory radius, T , within the orifice P = viscosity of liquid sprayed
correlations were obtained for widely of the nozzle + = functional notation
varying spraying conditions. The cor- = total volume of droplets in
liquid sprayed
+ = functional notation
relations obtained from the drop-size
data for the particular design of = mass-average axial velocity LITERATURE CITED
grooved-core nozzles utilized in this through the orifice,
investigation may be fairly accurate for volumetric flow rate 1. Choudhury, A. P. R., Ph.D. thesis,
other hollow-core nozzles. This is > Northwestern Univ., Evanston, IUi-
area of flow nois ( 1955).
reasonable because the correlation
variables are properties and dimensions 2. Darnell, W. H., Ph.D. thesis, Univ.
of the spra itself as it issues from the Wisc., Madison ( 1953).
E
nozzle, rat er than characteristics of
the spraying system (such as pressure
= superficial mass-average axial
3. Doumas, M., and R. Laster, Chem.
Eng. Progr., 49, 518 ( 1953).
velocity through the orifice, 4. Joyce, J. R., J. Inst. Fuels, 22, 150
drop or internal nozzle dimensions).
volumetric flow rate (1949).
area of orifice ' 5. - , Tech. Report N o . I.C.T. 17,
ACKNOWLEDGMENT The Shell Petroleum Co., Ltd., Nor-
This work was sponsored by Project man House, Strand, London (1950).
SQUID, which is supported by the Office 6. Kottler, F., J . Franklin Inst., 250, 339,
of Naval Research, Department of the = mass-average tangential vel- 419 (1950).
Navy, under contract N(onr)-1858(2 5 ) 7. Longwell, J. P., Doctoral dissertation,
NR-098-038. ocity through the orifice,
p 2 u ,rdr Mass. Inst. Technol., Cambridge
Ro ( 1943).
NOTATION 8. Marshall, W . R., Jr., Chem. Eng. Progr.
Monograph Ser. N o . 2, 50 (1954).
Do = air core diameter 9. McIrvine, J. S., M.S. thesis, Univ.
D, = nozzle orifice diameter Wisc., Madison ( 1953).
e -
- orifice roughness factor, 10. Mugele, R. A., and H. D. Evans, Ind.
= droplet diameter
length dimensions = unspecified mean value of x Eng. Chem., 43, 1317 (1951).
f = functional notation 11. Nelson, P. A., Ph.D. thesis, North-
f.b) = dn/Ndx = frequency func- = volume median drop diame- western Univ., Evanston, Illinois
tion, rate of change of the ter, that drop diameter at (1958).
number of droplets with re- which half of the volume of 12. Remus, G. A. P., M.S. thesis, North-
spect to the droplet diameter the spray is contained in western Univ., Evanston, Illinois
x , per unit number of drop-
lets
drops
-
having diameters
smaller than x and half of
( 1954).
13. Rupe, J., "Third Symposium on Com-
fw(y) = dn/Vdy = frequency func- bustion and Flame and Explosion
tion, rate of change of the the volume of the spray is Phenomena," Williams and Wilkins
volume of droplets with re- contained in drops having Co. (1949).
M

spect to the size characteris- 14. Spraying Systems Co. Catalog No. 24,
diameters larger than x Bellwood, Illinois.
tic, y, per unit volume of
= maximum stable drop diame-
droplets 15. Tate. R. W.. and W. R. Marshall. Tr..
F, = flow number, ratio of volu- ter Chem. En;. Progr., 49, 161, ' i 2 6
= minimum drop diameter (1953).
metric ffow rate to the square-
root of the pressure drop = Sauter mean diameter 16. Taylor, E. H., and D. B. Harmon, Jr.,
through the nozzle (the flow = size characteristic Ind. Eng. Chem., 46, 1455 (1954).
number is usually a constant = mean value of y
for a particular nozzle and a = In x = size characteristic for
particular liquid) log-normal distribution

Page 86 [Link].E. Journal March, 1961