DESIGN OF MEDIUM PRESSURE NOZZLES FOR COOLING

TOWERS

by

John Edward Thacker

Thesis presented in partial fulfilment of the requirements for the degree of Master of
Engineering at the University of Stellenbosch.

Thesis Supervisor: Prof. P.l Erens

Department of Mechanical Engineering

University of Stellenbosch

May 1997
 Stellenbosch University http://scholar.sun.ac.za

DECLARATION

I, the undersigned, hereby declare that the work contained in this thesis is my own original
work and has not previously, in part or its entirety, been submitted at any university for a
degree .

(Signature of candidate)

day of . 1997
 Stellenbosch University http://scholar.sun.ac.za

ABSTRACT

This project concerns the investigation of parameters controlling the behaviour of full-cone
spray nozzles of the type used in cooling towers.

In the present study large medium pressure hollow and full cone nozzles were investigated. A
literature survey provided insight into the relationships between the nozzle dimensions and
their spray characteristics, while equations found in the literature were used to correlate the
experimental data.

It was found that the spray cone angle of hollow cone nozzles could be manipulated by using
rounded orifice outlets and this finding lead to the development of a uniquely profiled outlet
that actually produces a square spray pattern.

More experimental work was done to determine the relationship between the central jet of a
full-cone nozzle and the other major nozzle dimensions. These results were then correlated and
formulated into a set of guidelines for designing full-cone nozzles.

11
 Stellenbosch University http://scholar.sun.ac.za

OPSOMMING

Hierdie projek behels 'n studie van belangrike parameters in volkegel sproeimondstukke soos
gebruik in koeltorings.

In die huidige studie word groot mediumdruk holkegel en volkegel sproeimondstukke

ondersoek. 'n Literatuurstudie het die nodige insig verskaf omtrent die verwantskap tussen
mondstuk dimensies en hul spuitkarakteristieke, terwyl vergelykings uit die literatuur gebruik is
om die eksperimentele data te korreleer.

Dit was gevind dat die sproeir kegelhoek van die holkegelmondstuk verander kon word deur
gebruik te maak van geronde uitlate. Afleidings wat gemaak is het gely tot die ontwikkeling
van 'n unieke geprofielde uitlaat wat 'n vierkantige sproeipatroon gelewer het.

Bykomstige eksperimentele werk is gedoen om die verwantskap tussen die sentralestraal van 'n
volkegelmondstuk en die ander hoof mondstukdimensies te bepaal. Die reultate is verwerk om
riglyne vir die ontwerp van vierkantige patroon volkegel mondstukke daar te stel.

III
 Stellenbosch University http://scholar.sun.ac.za

ACKNOWLEDGEMENTS

I would like to express my sincere appreciation to the following persons and institutions:

Prof P.l Erens for his guidance, patience and encouragement;

Boet, George and Scotty for their patience and technical assistance;

Industrial Water Cooling(Pty)Ltd for their interest and financial support;

Foundation for Research and Development for their financial support.

IV
 Stellenbosch University http://scholar.sun.ac.za

TABLE OF CONTENTS

DECLARATION .
ABSTRACT. ....................................................... 11

OPSOMMING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. III

ACKNOWLEDGEMENTS. ............................................ IV

TABLE OF CONTENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. V

NOMENCLATURE. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . .. Vlll

1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1

2. LITERATURE SURVEY. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . .. 4
2.1 Pressure-Swirl Atomisers 4

2.2 Water Distribution and Cooling Tower Performance 10

2.3 Drop Size Distribution 11
2.4 Summary Conclusions 12

2.5 Design Objectives for a Sound Nozzle 12

3. THEORETICAL TREATMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 14
3. 1 The Swirler 14
3.2 The Hollow Cone Nozzle 16
3.2.1 Spray Cone Angle 16
3.2.2 Discharge Coefficient 19
3.2.3 Film Thickness 20

4. APPARATUS AND TEST PROCEDURE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 23

4.1 Test Facility 24
4.2 Prototype Nozzle 26
4.3 Measurement of Characteristics 28
4.3.1 Spray Cone Angle 28
4.3.2 Film Thickness 28
4.3.3 Spray Distribution 30

V
 Stellenbosch University http://scholar.sun.ac.za

5. EXPERIMENTAL WORK 32
5.1 The Swirler 32
5.2 Hollow Cone Nozzle Tests 34
5.2.1 Spray Cone Angle 34
5.2.2 Discharge Coefficient 35
5.2.3 Film Thickness 37
5.2.4 Spray Distribution 38
5.3 Full Cone Nozzle Tests 41
5.3.1 Influence of Central Jet 42
5.3.2 Comparison of two FCNs 46

6. DISCUSSION OF EXPERlMENTAL RESULTS 49

6.1 The Swirler 49

6.2 HCN Characteristics 51

6.3 FCN Characteristics 52

7. NOZZLE MANUFACTURE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 54
7.1 Plastic Injection Moulding 54
7.2 Proposed Nozzle Design 55

8. CONCLUSIONS AND RECOMMENDATIONS. . . . . . . . . . . . . . . . . . . . . . . . . .. 57

REFERENCES " R1

APPENDICES

A - PHYSICAL PROPERTIES OF WATER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Al

B - SAMPLE CALCULATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. B1
B. 1 Calculation of Mass Flow Rate B1
B.2 Calculation of Swirl Port Velocity B3
B.3 Calculation of Spray Cone Angle B4
B.4 Calculation of Discharge Coefficient B8

VI
 Stellenbosch University http://scholar.sun.ac.za

B. 5 Film Thickness Calculations B 10

B. 6 Calculation of Rainfall Intensity B 12
B.7 Calculation of Swirler Pressure Drop in HCN B 12

C-TABULATEDEXPERIMENTALDATA Cl
D - DESIGNING A FCN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Dl

VB
 Stellenbosch University http://scholar.sun.ac.za

NOMENCLATURE

List of Sym bois

A Area [m2]
AR Aspect ratio [-]
B[ Coefficient in Equation (B-6) [-]
Bt Coefficient in Equation (B-9) [-]
Bcr Coefficient in Equation (B-8) [-]
b Width of spray [m]
C Discharge Coefficient [-]
d Diameter [m]
E Flow coefficient [-]
F Flow loss coefficient [-]
FN Flow number [m2]
g Gravitational constant [m2/s]
h Height [m]
I Rainfall intensity [kg/m2s]
K Nozzle parameter [-]
Length [m]
m Mass flow rate [kg/s]
m Fluid flow index [-]
n Number of data points [-]
P Pressure [N/m2]
ci Volume flow rate [m3/s]
Re Reynolds number [-]
r Radius [m]
s Swath width [m]
t Film thickness [m]
U Total velocity [m/s]
Up Pressure velocity [m/s]
V Dimensionless axial velocity [-]
v Axial velocity [m/s]

Vlll
 Stellenbosch University http://scholar.sun.ac.za

W Dimensionless tangential velocity [-]

w Tangential velocity [m/s]
x Ratio of air-core area to orifice area [-]
y Nozzle constant of Taylor[48TAI] [-]

List of Greek Symbols

a Half spin chamber angle [0]

~ Ratio of flow meter orifice diameter to pipe diameter [-]
LlP Pressure drop [Pa]
Llt Time lapse [s]

<Pn Cone angle on rounded nozzle exit [0]

y Port helical exit angle [0]
'A Wall friction coefficient [-]
~ Absolute viscosity [kg/ms]
v Kinematic viscosity [m2/s]
8 Half spray cone angle [0]
p Density [kg/m3]
a Dependent variable in Equation (B- 7) [-]
Q Rotational constant [m2/s]

List of Subscripts

a Ambient conditions
c Core
e Entry
ep Predicted entry
fm Flow meter
gi Generalised inlet
Indexing variable
Hg Mercury

J Jet

IX
 Stellenbosch University http://scholar.sun.ac.za

m Mean of inlet and outlet values

n Nozzle
0 Orifice
p Swirl port
r Frictional
rs Rain sampler
s Swirl chamber
sc Sudden contraction
se Sudden expansion
sw Swirler

List of Dimensionless Groups

c- q Discharge coefficient
- ~2i1PnP

Ap
K = -- Nozzle parameter
dsdo

pvd
Re = -- Reynolds number
/l

b
s= - Swath coefficient
2h
v
V =- Dimensionless axial velocity
Do

w=~ Dimensionless tangential velocity

U o

x= Ac Ratio of air-core area to orifice area

Ao

x=~ Ratio of axial velocity to pressure velocity

Up

Q
y = -- Nozzle constant of Taylor [48TAl]
ro Up

r
z=~ Ratio of air-core radius to orifice radius
ro

x
 Stellenbosch University http://scholar.sun.ac.za

List of Acronyms

ABS Acry 1onitrile-butadiene- styrene

AVG Standard average
FCN Full-cone nozzle
FN Flow number
HCN Hollow cone nozzle
LHS Left hand side
NC Numerical control
PA Polyamide
PBTP Polybutylene terephthalate
PC Polycarbonate
PE Polyethylene
PIM Plastic injection moulding
POM Polyoxymethylene
PP Polypropylene
PVC Polyvinylchloride
RHS Right hand side
RID Rainfall intensity distribution
STD Standard deviation
UV Ultra violet

Xl
 Stellenbosch University http://scholar.sun.ac.za

CHAPTER ONE

INTRODUCTION

This thesis is about designing new more effective nozzles for water distribution in cooling
towers. The nozzles are not necessarily revolutionary, but their distribution is seen to be
predictable and acceptable for application to the task of wetting cooling tower fill packs. The
thesis title, "The Design of Medium Pressure Nozzles for Water Distribution in Cooling
Towers", clearly defines the content of this thesis.

The word design implies a definite need for the development of better cooling tower nozzles.
There are three reasons for the implementation of this project. First, the cooling tower industry
in South Mrica is dependent on imported brands of nozzle for use in its towers. The cost of
these nozzles is becoming increasingly high and it is often difficult to obtain nozzles at short
notice. Manufacturing nozzles locally would alleviate both these problems.

Second there is the problem of questionable nozzle performance, i.e. poor distribution. This
problem can be addressed by studying existing nozzles and by looking for ways to optimise
nozzle performance. The ultimate goal of this project is to provide a blueprint for designing
more effective nozzles.

Third there is the problem of cooling tower performance - discussed in the literature survey.
The cooling capacity of evaporative cooling towers is adversely affected by maldistribution of
water to the fill pack. The design of more effective nozzles will alleviate the problem of poor
water distribution and hence increase tower performance.

The question of operating pressure is a key issue as far as nozzles are concerned. Not only are
the pumping costs proportional to nozzle supply pressure, but atomisation too is very much a
function of nozzle injection pressure - according to Lefebvre [89LEl] drop diameter is
inversely proportional to the cube root of pressure. For these reasons it has been necessary to
limit the scope of this thesis to medium pressure nozzles only.

Low-pressure nozzles typically require pressures of between 5 kPa and 15 kPa. Applications
for these nozzles are; water distribution in large natural draught wet cooling towers and
 Stellenbosch University http://scholar.sun.ac.za

irrigation. Medium pressure refers to the range 20 kPa to 100 kPa. It is in this range that

nozzles for mechanical draught cooling towers see Figure 1-2, operate. High-pressure nozzles
are smaller in physical dimension than other nozzles. They operate at pressures anywhere

between 700 kPa and 2000 kPa. These nozzles produce very fine sprays - droplets are in the

order of 10 /-lm in diameter. A typical application for these nozzles is in the combustion of
liquid fuels.

The question anses as to which type of

nozzle is most suitable for the task of wetting

cooling tower fill packs? The answer to this
question is the pressure-swirl or simplex
atomiser. There are two basic types of
simplex atomiser and they are classified
HOLLOW CONE SOLID CONE according to their spray patterns. Firstly the
hollow-cone nozzle, HCN, as the name
Figure 1 - 1: Spray patterns of simplex
implies, produces a spray in which most of
atomisers.
the droplets are concentrated at the outer
edge of a conical spray sheet. The traditional garden-variety sprinkler is the perfect example of
a hollow cone nozzle. Secondly the solid, or full-cone nozzle, FCN, produces the same
conical-shaped spray as the HCN, but it is now made up of a fairly uniform distribution of
droplets throughout its volume. The
MOTOR
two spray structures are depicted in
INDUCED DRAUGHT FAN
Figure 1-1.
DRIFT ELIMINATOR

NOZZLE

Obviously it is the FCN which must

HOT WATER INLET

be used for the uniform distribution

FILL PACK

of water over a large surface area AIR INLETS

such as a cooling tower fill pack.

However, since the FCN is simply the
COLD WATER OUTLET
combination of a swirl atomiser and
an axially aligned central jet, see
Figure 2-3, it was necessary to study

the swirl atomiser independently from Figure 1 - 2: Mechanical draft cooling tower.
 Stellenbosch University http://scholar.sun.ac.za

the FCN. Much of the theoretical and experimental work was, therefore, concerned with the
analysis of the HCN so that a more complete understanding of the FCN could be gained.

The aim of this thesis is to enable uniform water distribution in cooling towers. Evaporative
cooling towers are widely used in industry, and where natural draught cooling towers might be
limited for use in power generation, industrial type mechanical draught cooling towers of the
type depicted in Figure 1-2 are used extensively in many other industrial processes.

The first essential task for this project was to conduct a literature survey on hollow and full
cone nozzles to become familiar with their operation and characteristics, while at the same time
noting the relationship between the major nozzle dimensions and the flow characteristics of
such nozzles. Empirical formulae and experimental correlations were also duly recorded for
use in correlating any new experimental data.

Next an experimental investigation ofHCNs had to be conducted to determine the relationship

between the major nozzle dimensions, namely orifice diameter, do, swirl chamber diameter, ds,
and swirl port area, Ap, and the three fundamental flow characteristics, discharge coefficient,

Cn, spray cone angle, 28, and film thickness, 1. The emphasis of these experiments fell on
determining accurate correlations that could be used to predict any of these characteristics for
any flow condition that might arise inside a medium pressure cooling tower.

The next objective was to build and test various FCN configurations to observe and quantify
the effect of the central jet on radial spray distribution and rainfall intensity. The scope of the
experimental research was geared towards understanding FCNs under typical operating
conditions, i.e. nozzle injection pressures, Pn, of between 25 and 65 kPa. This experimental
data could then be correlated to reveal formulas for predicting rainfall intensity distribution,
RID, as a function of injection pressure and total mass flow rate, m.

Finally and most importantly a means of producing a square spray had to be investigated and
mathematically recorded for future reproduction. Ultimately the experimental and theoretical
analysis was to culminate in a comprehensive design code for designing FCNs for specific
operating conditions including: nozzle injection pressure, Pn, installation height, h, spray area
and required rainfall intensity, I.
 Stellenbosch University http://scholar.sun.ac.za

CHAPTER TWO

LITERATURE SURVEY

The purpose of this chapter is not to delve into the complex nozzle theories of the various
authors, but rather to summarise the findings relevant to medium pressure nozzle design. A
more detailed reproduction of the formulas relating to nozzle characteristics and performance
is given in Chapter 3.

2.1 Pressure-swirl atomisers

The simplest form of HCN is the simplex swirl

atomiser, see Figure 2-1. Liquid is fed through
SWIRL
tangential ports into the swirl chamber. The high CHAMBER

angular velocity of the fluid in the swirl chamber SPIN

CHAMBER
results in the formation of an air-cored vortex,
ORIFICE
which is concentric with the axis of the nozzle.
The outlet from the nozzle is the final orifice, the
diameter of which is less than that of the swirl
chamber. Conservation of linear and angular
28
momentum requires that the fluid accelerates as it
flows, under both axial and radial forces, along
the nozzle and out through the final orifice. The

spray cone angle, 28, is determined by the axial

and tangential components of velocity at the
nozzle exit.

Figure 2 - 1: Sectional view of a swirl

Taylor [48TA1] made a simple analysis of the
atomiser.
simplex swirl atomiser. His assumption of inviscid

irrotational flow simplified the problem allowing him to use the Bernoulli equation in obtaining
a solution. He proposes that the tangential velocity is inversely proportional to radius and that
the constant of proportionality, n, the circulation constant, is dependent on the fluid entry
conditions.
 Stellenbosch University http://scholar.sun.ac.za

1 2

0.8 1.6
~
-~ 0.6
I

1.2
0 ~

u
U
c - I

:::.G
NO.4 0.8
~

0.2 0.4

o o
o 0.2 0.4 0.6 0.8 1
y [-]
I-x-z-c n -cas8 KI
Figure 2 - 2: Relationship between K and the flow variables o/Taylor [48TAI].

The results of his analysis are shown graphically in Figure 2-2. The variables depicted here
include K, the nozzle parameter, x, the ratio of axial velocity within the orifice, vo, to pressure
velocity, Up,z, the ratio of air-core radius to orifice radius and y, the nozzle constant. If either
of the values of K or y are known for a particular HCN then this theory can be used to
calculate the spray cone angle, 28 and the nozzle discharge coefficient, en. Unfortunately this
theory is not well suited for application to real fluids because of the presence of boundary
layers which develop along the inside of the swirl chamber and within the orifice. He [mally
concludes his paper with the following sentence relating to the question of boundary layers;
"For this reason perfect fluid theory has no application in the hydrodynamics of swirl
atomisers."

In his next article Taylor [50TAI] examines the boundary layer which develops inside the
conical spin chamber. He makes use of Pohlhausen's momentum integrals through the
boundary layer in order to determine the velocity distribution and thickness of the boundary
layer. Many authors have subsequently used the same approach to solve this unique problem.
This analysis makes it possible to estimate the boundary layer thickness at the [mal orifice and
 Stellenbosch University http://scholar.sun.ac.za

also determine the flow paths of the fluid in the boundary layer. Comparing the boundary layer
thickness to the film thickness gives one an indication of the usefulness of the inviscid theory.
According to Taylor [50TAl] "a condition can arise in which practically the whole of the
outflow from the orifice is fed by a boundary layer current close to the surface of the swirl

chamber." Chapter 3 contains a worked example, from this paper, which can be used to
estimate the boundary layer thickness at the final orifice.

In their design of experimental swirl atomisers Dombrowski and Hasson [69DO 1] recognised
the importance of the ratio of swirl port length, Ip, to swirl port diameter, dp. At low ratios the

flow tends to diffuse from the port into the swirl chamber, while at higher values the frictional
losses become too large. Results from their pilot experiments indicate an optimum value for
Ip/ dp of 3. Their work was aimed at determining the effect of the variation of the ratios of swirl
chamber diameter, ds, to orifice diameter, do, and orifice length, 1 to orifice diameter, do, on
0,

spray cone angle, 28, and nozzle discharge coefficient, Cn. Their results show that discharge

coefficient is increased by an increase in djdo but decreases with increases in la/do. The spray
cone angle is decreased when either of the two ratios is increased. A point of theoretical
importance discovered by them is that for any given value of la/do a unique relationship exists
between discharge coefficient and spray cone angle.

Som [83 SO 1] investigated the discharge coefficient and spray cone angle of the swirl pressure
atomiser for a non-Newtonian fluid. His analysis pointed to the following as the independent
control parameters;

• the generalised inlet Reynolds number, Regi, based on the tangential velocity at the nozzle
inlet, the diameter of the swirl chamber and the apparent viscosity of the fluid
• the flow behaviour index number of the fluid, m

• the ratio of swirl chamber length to diameter, Ijds

• the half spin chamber angle, a

• and the ratio of swirl chamber diameter to orifice diameter, djdo.

The results of his work reveal the dependence of discharge coefficient and spray cone angle on
the generalised inlet Reynolds number and the flow behaviour index of the fluid. This
dependence is especially significant at lower Reynolds numbers, i.e. Regi < 60 000. The
 Stellenbosch University http://scholar.sun.ac.za

relationships between the geometry of the nozzle and the spray characteristics are the same as
that disclosed by Dombrowski and Hasson [69D01] and Rizk and Lefebvre [85RI1].

Walzel [84WA1] discusses the various designs of single substance pressure atomisers and is
particularly interested in their relevant sheet numbers. The sheet number is important for
predicting drop formation and is the ratio of the product of the sheet length and thickness to
the area of the final orifice. HCNs produce the smallest sheet numbers and thus generate the

smallest drops, which makes them very attractive in industry. According to him larger spray
angles can be generated either by increasing the swirl chamber diameter or by the addition of a
rounded profile at the nozzle exit which deflects the streamlines thereby producing a larger
spray angle. This phenomena of deflecting streamlines along a convex solid body is known as
the Coanda effect, after the Rumanian aeronautical engineer Henri Coanda (1885 - 1972).

Walzel [84WA1] classifies the FCN as a

turbulence nozzle, for which atomisation
does not begin until a critical Weber
number is exceeded. Drops form as a result
of turbulence in the central jet, sheet
SPIRAL
CENTRAL
break-up of the fluid annulus and also on JET
PORT

account of the relative velocity between

the water jet and annulus. FCNs serve for
uniform spraying of large surfaces when
small drops are not required. The FCN is
the combination of a HCN and a jet
atomiser, see Figure 2-3.
V SHEET BREAK-UP

- .... .... __.. _ ~ J .• 0 ••• •• •••

Rizk and Lefebvre [85RI1] analysed the

simplex atomiser giving equations for flow Figure 2 - 3: Sectional view of a FeN.

number and discharge coefficient and also

deriving a relatively simple equation for film thickness. Their findings concerning the effects of
varying nozzle dimensions on nozzle flow characteristics can be summarised as follows:
 Stellenbosch University http://scholar.sun.ac.za

• Film thickness increases with increases in swirl port area, Ap, orifice diameter, do and swirl
chamber length, Is, but decreases with increases in swirl chamber diameter, ds, and orifice
length, 10.

• Flow number, FN, increases with increases in swirl port area, Ap, and orifice diameter, do,
but decreases with increases in swirl chamber diameter, ds. Flow number is not affected by
changes in orifice length, 10, and swirl chamber length, Is.

• The variation of X, the ratio of air-core area to orifice area, is such that it increases when
injection pressure, P n, is increased but decreases when K, the nozzle parameter, is
increased.

• Spray cone angle, 28, is increased when the ratio dido is decreased.

Horvay and Leuckel [86HO 1] made use of the N avier - Stokes equations in their analysis of
swirl pressure nozzles. They derived theoretical equations for the spray cone angle, air-core
radius, re, and the tangential, Uo, and axial, Vo, velocities at the nozzle exit. They also derived
similar equations experimentally.

Prywer and Kulesza [87PRl] conducted experiments with full cone nozzles. They investigated
the effect of the ratio of inlet port area to central jet area, ApiAj, on the uniformity of the spray
distribution; best results were achieved with values of this ratio of about 4,3. This means that
for a nozzle with four swirl ports the most uniform distribution can be expected when the ports
and the central jet each comprise roughly twenty percent of the total flow area.

The most comprehensive source of nozzle information is Lefebvre's [89LEl] Atomization

and Sprays. The sections relating to swirl atomisers incorporate nearly all of the information
contained in this literature study, which makes a summary of this work redundant. This book is
important reading for anyone interested in the field of fluid atomisers. Documented are all the
effects of nozzle dimensions, fluid properties and operating conditions on spray characteristics
such as discharge coefficient, pattemation and atomisation.

Dahl and Muschelknautz [92DA 1] made use of cyclone theory in their analysis of the HeN.
The inclusion of a wall friction coefficient in their calculations makes it possible to determine
both axial and tangential velocity at the nozzle exit and also the frictional pressure loss through
the nozzle. They estimate that roughly 80 percent of the total pressure drop across a swirl
 Stellenbosch University http://scholar.sun.ac.za

atomiser is used to accelerate the fluid, while the remainder is dissipated through wall friction.
These figures are in agreement with data obtained in this thesis.

Chen et al. [93CH1] studied the circumferential liquid distribution of simplex atomisers.
Uniformity in the context of this paper is a measure of the symmetry of the distribution. The
most uniform distributions were obtained with nozzles having an orifice length to diameter
ratio of between 1,5 and 2, although viscous fluids showed better uniformity at la/do equal to
0,5. The results also showed that spray uniformity was improved markedly at higher pressures.
The number of inlet ports had little effect on the uniformity of distribution so long as the ports
were spaced evenly around the entry to the swirl chamber.

Dumouchel et al. [93DU1] conducted a numerical analysis of the viscous flow in a swirl
atomiser. The main input parameter for their computer program was a Reynolds number that is
based on the inlet port diameter and the radial inlet velocity. The flow was simulated for

Reynolds numbers less than 150, however it was found that for Re ~ 100 the velocity field was
independent of Re. Their analysis focuses on determining the effects of varying nozzle

geometry in order to predict the spray cone angle. They concluded that the characteristics of
the conical sheet produced at the nozzle exit are indeed functions of nozzle geometry, a finding
that is in agreement with experimental data.

Koo and Kuhlman [93KO 1] developed a theoretical spray performance of swirl-type nozzles
by solving boundary flows in the swirl chamber of the nozzle using the integral momentum
method. Their investigation revealed the essential structures of the discharge coefficient and
the swath coefficient, s, which is the ratio of height of the nozzle to the width of the spray and
is an indication of the spread of the spray. They also define several parameters, which are
different from those used by previous authors.

In their next paper, Koo and Kuhlman [93K02] design, test and calibrate an experimental
nozzle using the theoretical spray performance theory, which they had previously been
developing. The results are characterised by two flow regimes, namely turbulent and laminar.
This difference in flow was evident in the swath coefficient, which showed two completely
separate but similar trends.

..
 Stellenbosch University http://scholar.sun.ac.za

10

2.2 Water Distribution and Cooling Tower Performance

Kranc [83KR1] studied the effect of non-uniform water distribution on counterflow cooling

tower performance. There are two factors which influence the cost of operation of a cooling
tower: pumping head which is a function of the distribution system and the thermal

performance which is a function of the packing geometry, water distribution and air-side
pressure drop. Flooded areas represent excess pump work and reduced cooling due to
minimised evaporation while dry areas correspond to under performance, even though lower

water temperatures are achieved. The problem of flooding is compounded when the total flow

rate is increased in order to improve coverage of dry areas. He uses the Merkel equation to
compute the design performance of the tower. Performance of a tower with non-uniform water
distribution is calculated relative to the same tower operating with a uniform water
distribution. Results from this work show that some degree of non-uniformity can be tolerated
within a tower without seriously affecting the thermal performance of that tower, while the
negative effect of greater non-uniformities is evident.

Schultz [87SC 1] investigated the effect of uneven water distribution on the performance of
counterflow cooling towers. Having measured the actual distributions of existing nozzles, he
then used this data in a computer simulation program to predict the cooling capacity of the
tower. Results showed reductions in cooling tower performance of between 54 and 64 percent.
These high estimates can be attributed to the simulation program neglecting the effects of
redistribution of water by the fill pack and break-up of water in flooded regions of the tower by
the counter flowing air.

Bellagamba et al. [88BE 1] investigated the distribution of water in cooling towers by

commercially available spray nozzles. Their focus is on low-pressure nozzles used in the
distribution of water to splash packs in large natural draft cooling towers. Here distribution is a
key element in the performance of the whole tower and in particular within the rain zone,
which is the region between the splash pack and the nozzles. According to them tower
performance is affected by uniformity of spray, air-side pressure drop and heat transfer
occurring in the rain zone, which in turn is dependent on drop size.
 Stellenbosch University http://scholar.sun.ac.za

11

Kranc [93KRI] presents a method for estimating the performance of a counterflow cooling
tower with a regular fill. The model utilised is different from his previous work in that
redistribution of water, due to interaction with the structured fill, is taken into account. Three

models of water distribution through the tower are discussed in the paper. In the first instance
the ideal case of uniform flow, the second an initially non-uniform distribution which persists
throughout the tower, and finally an initially non-uniform distribution which is redistributed by
the fill. Results of this analysis show that regions of non-uniformity persist for some depth into

the tower and that performance is significantly degraded. The introduction of a redistributing
fill is beneficial to tower performance.

2.3 Drop Size Distribution

Drop size distribution from HCNs is not actually within the scope of this thesis. It is only in
large natural draught type cooling towers that the size of drops in the rain zone can affect
overall tower performance. In smaller mechanical draught type cooling towers it is rather the
rainfall intensity distribution, RID, that is important and not the drop size distribution. The
following article, which concerns drop formation in FCNs, has been included to illustrate the
relative complexity of this atomisation process.

Sada et al. [78SAI] conducted experimental work with FCNs in order to determine the
relationship between drop size distribution and injection pressure. They found that the
geometric mean diameter of drops decreased when injection pressure was increased, but that
this was only significant at the centre of the spray. They also found that the drop size
distribution varied widely with respect to position within the spray. At the centre of the spray
the drop size distribution could be expressed as logarithmic normal, while at greater local
radial positions the drop size distribution contained two peaks. The peaks are indicative of two

mechanisms of drop formation - turbulence atomisation in the centre and sheet break-up at the
perimeter. This double peak distribution is such that the geometric mean diameter for the first
peak decreases as the radial distance from the spray centre increases while the logarithmic
standard deviation for the second peak decreases with radial distance from the centre, this
implies that the second peak becomes more acute at the environs of the spray.
 Stellenbosch University http://scholar.sun.ac.za

12

2.4 Summary Conclusions

Cooling tower performance is dependent on uniform water distribution for optimum

performance. This finding supports the motion to design more effective sprayers. Improved

FCN s are the solution to the problem of uneven water distribution in cooling towers. Although
there is an abundance of literature on the subject of HCNs, many of the experimental
correlations are only valid for small high-pressure nozzles. Consequently it will be necessary to

calibrate large HCNs at medium pressures so that accurate correlations can be obtained. The

nozzle parameters that most affect the spray characteristics are; K, dJdo, la/do and IJds. The
effect of some form of Reynolds number, related to operating conditions, will have to be
investigated because this effect is not evident with high-pressure nozzles.

2.5 Design Objectives for a Sound Nozzle

The type of nozzle developed here is designed for use in specific towers, for this reason the
experimental test conditions were selected to approximate typical cooling tower conditions.
There are many parameters which affect the spray characteristics of both hollow and full cone
nozzles, however, an experimental investigation, more detailed than the one described within
these pages, would only have complicated the design process without necessarily yielding a
better solution.

Taking into account the ultimate goal for this project, i.e. to produce a practical cooling tower
nozzle, it was necessary to limit the extent of the study of HCN s so that time could be more
appropriately spent on investigating FCNs. However, since roughly 80 percent, following the
advice ofPrywer and Kulesza [87PR1], of the water flowing through a FCN is forced through
the swirl ports, this water will behave within the nozzle primarily as though it were inside a
HCN. Therefore, it is necessary to examine those specific HCNs from which the full cone
cooling tower nozzles will assume their dimensions and many flow characteristics.

Restrictions on the operating conditions of cooling tower nozzles include; an installation height
of between 17 cm and 40 cm above the cooling tower fill pack and an injection pressure of
 Stellenbosch University http://scholar.sun.ac.za

13

between 25 kPa and 90 kPa. Operating pressure is an important consideration in mechanical

draft cooling towers since if the pressure is too high tiny droplets may form and be drawn with

the air stream out of the top of the tower without actually doing any cooling. A supply

pressure of 35 kPa is adequate for sustaining the atomisation process without causing undue
water loss.

Rainfall intensity, I, spray area and airflow rates are the three controllable variables affecting

the thermal performance of cooling towers. The easiest variable to change in an existing tower

is water flow rate, since packing is often standard and the airflow rate is coupled to the

performance of the fan. Typical values for rainfall intensity of industrial type cooling towers are

between 0,5 and 8 kg/m2s. The ideal result for this project would be to yield a formula for

designing nozzles for specific rainfall intensities.

Finally the crux of this thesis; "What is required from a cooling tower nozzle?" In point form
the prerequisites of a good nozzle design are:

• uniform distribution in a square spray pattern

• no clogging under normal operating conditions

• minimal pressure losses

• predictable characteristics for design purposes

• simplicity of design and ease of manufacture

• competitive low cost production.

 Stellenbosch University http://scholar.sun.ac.za

CHAPTER THREE

THEORETICAL TREATMENT

3.1 The Swirler

In order to predict the flow rate through the swirler it is necessary to estimate the pressure loss
coefficient. Approximate analyses of the geometry of the two swirlers will clearly outline the
advantages of streamlining the swirl ports. Figure 3-1 illustrates the two basic swirl port
geometries.

There are three features of the sharp swirler,

STREAMLINED SHARP
which will result in significant pressure losses
namely, the sudden contraction, the protruding
sharp lip, which may also cause flow separation
within the swirl port, and the change in flow
direction. The streamlined swirler will have
smaller entry losses, no separation and smaller

losses from the gradual directional change. Figure 3 _1: Swirl port geometries.
However, frictional losses may well be greater
than those of the sharp swirler because of the increased port surface area.

Neglecting gravitational acceleration the following energy balance equation expressed in terms
of pressure and pressure losses describes the flow through a swirl port:

(3 -1)

where the subscripts 1 and 2 indicate static pressure upstream and downstream of the swirler
and the Fj' s are the individual dynamic head loss coefficients. The loss coefficient due to
sudden contraction, Fsc, can be approximated with the following equation derived from
White's [86WH1] empirical formula for flow losses due to sudden contraction in a pipe:

Fsc ~
d2) '
0,42 ( 1- d>

where by substituting port area, Ap, for d]2 and swirler area, As, for d/ we obtain:
 Stellenbosch University http://scholar.sun.ac.za

15

(3-2)

This equation can be used to estimate the entry loss coefficient for the sharp swider. Based on

the ratio of the swirl port area, Ap = 932 mm2, to the swirl chamber area, As = 3664 mm2, this
equation yields a loss coefficient of 0,31. The values given above are the actual flow areas of

the two tested swirlers. For well-rounded entrances, such as those found in the streamlined

swirler, White [86WHl] suggests that the loss coefficient could be in the region of 0,05.

The next loss coefficient to be estimated is for the overhanging sharp lip. In the absence of any

other significant data and as a crude approximation for this loss coefficient it was decided to

model this geometry as a three-quarters-open gate valve. The loss coefficient could then be

taken from Appendix C of Foust et al. [60FO 1]. The equivalent length in pipe diameters of a

three-quarters-open gate valve is 35. For a schedule 40 pipe with inside diameter of 19 mm

(0,75 inches) this is equivalent to a loss coefficient of 0,85, which if nothing else at least gives

an estimate for the magnitude of the losses associated with the overhanging lip of the sharp-

edged swider.

The third loss coefficient for both swirlers, taken from Appendix C of Foust et al. [60FO 1], is

that of a 45° standard elbow, which has an equivalent length in pipe diameters of 16, or a loss

coefficient of 0,3 5. This coefficient can be used to estimate the losses due to the directional

change in the liquid brought about by the individual swirl ports. By summing up all these

individual loss coefficients, we are now in a position to predict the total loss coefficient for

each of the two swirlers.

F or the sharp swider: For the streamlined swider:

IFi = 0,31 + 0,85 + 0,35 IFi = 0,05 + 0,35

= 1,51. = 0,4.
Rearranging Equation (3-1) gives the unknown port velocity, De:

(3-3)

where ~p sw is the pressure drop across the swirler.

Substituting for the two loss coefficients yields:

 Stellenbosch University http://scholar.sun.ac.za

16

for the sharp swider: and for the streamlined swider:

U, = O,631~211:~

From this estimation of the two loss coefficients it is expected that the fluid velocity through
the streamlined swider will be 33,9 percent greater than through the ports of the sharp swider.
Alternatively, if these swiders were to operate at the same flow rate, i.e. port exit velocities

equal, then the pressure drop across the streamlined swider would be 44,2 percent less than the
pressure drop across the sharp swider.

3.2 The Hollow Cone Nozzle

The following sections describe some of the theories and relevant formulas pertaining to swirl-
pressure atomisers. The idea here is not to make a complete analysis of the HeN but rather to
select from the literature those calculation methods that can be readily applied in predicting the
spray characteristics of simplex atomisers.

3.2.1 Spray Cone Angle

The spray cone angle is defined as the angle between the tangents to the spray envelope at the
nozzle exit. Neglecting radial velocity, which is an order of magnitude less than the tangential
and axial velocities, the spray cone angle is usually expressed as:

(3-4)

where Ua is the total velocity of the spray sheet and v is the mean axial velocity of the liquid
leaving the nozzle.

The first estimate of e comes from Taylor [48T AI]. The equation for cos e is depicted
graphically in Figure 2-2 but is reproduced here for clarity:

cos e = x +
yJiz 2
0 ~
[1(1)
"27 - ]
1 + loge z . (3-5)
(1-z")2
 Stellenbosch University http://scholar.sun.ac.za

17

The relationship between the variables comes from the Bernoulli equation at the core, which
can be written in the form

(3-6)

here z = ~ and rc is the radius of the air-core within the orifice,

ro

x = ~, Vo being the axial velocity in the orifice and Up the so called pressure velocity
Up

which is defined thus:

U
p
= ~2!l.P
__p n and finally

The equation for z in terms ofy is:

z'<+~~>Y;' (3-7)

which is derived through the continuity equation. The nozzle parameter, K is related to y and
the nozzle discharge coefficient, Cn by the following formula:
nCn
K=- (3-8)
4y .

The cyclone theory of Dahl and Muschelknautz [92DAl] provides a practical method for
determining the axial and tangential velocity components, immediately before and after the final
orifice. From these velocities the other discharge characteristics such as spray angle are easily
determined. The first equation they give is that of tangential velocity within the orifice:

U~e
o
r
w=------ (3-9)
o 1+ ",,_A_R_U_e ~.

2 q ~~
AR is the frictional area of the swirl ports and swirl chamber combined and"" is the wall friction
coefficient dependent on the mean nozzle Reynolds number. The Reynolds number is defined
thus:
 Stellenbosch University http://scholar.sun.ac.za

18

(3-10)

Dahl and Muschelknautz [92DA1] provide a graph, somewhat like a Moody diagram, for

estimating A. For Reynolds numbers between 100000 and 300 000 A varies between 0,007 and
0,003 and has a near linear relationship with Rem on the log-log scale. According to Dahl and
Muschelknautz [92DAl] the axial velocity in the outlet depends on the centrifugal forces of

the rotational flow. Under the influence of friction, they give the tangential velocity profile as:

Therefore, by integrating this profile and equating the static pressure near the air core to zero,
the following relationship between dimensionless tangential and axial velocity within the orifice
is found.

(V _1)n+l
W 2 =2V ~a_~_
(3-11)
o a V n-l
a

with

W=~ (3-12)
U'a

(3-13)

and

(3-14)

When leaving the nozzle the liquid near the wall of the orifice accelerates in the direction of the
static pressure gradient, while the liquid near the air core is at ambient static pressure and
therefore does not accelerate. By integration of the radial pressure gradient within the orifice
and conversion to dimensionless form the liquid velocity after leaving the nozzle can be
calculated as:

(3-15)

The dimensionless mean axial velocity past the outlet of the nozzle is then the root mean
square of Va and Va:

(3-16)

The film thickness in the final orifice is calculated from the continuity equation:
 Stellenbosch University http://scholar.sun.ac.za

19

t ~ ro - ~r'0 - q . (3-17)
rrvo

Finally the total velocity of the liquid sheet past the nozzle can be calculated by vectorial
addition of the tangential and axial velocity components:

U
a
= -2
V +w
0
2(
-- ro
r-t
J n
(3-18)
o

The spray cone angle can now be calculated from Equation (3-4):

e= cos'( ~.J. (3-19)

Som [83 SO 1] provides an equation for spray cone angle. The numerical values in the equation
have been evaluated from experimental results even though the form of the equation was
derived theoretically using the Pohlhausen momentum integral method. The equation is:
0,3153( )01949
_ 77,260 ( do / Is ) 2a' [_ (-8,695XI0-5)Regi]
8- ()00611 1 e , (3-20)
1s /d s '

where Regi is the generalised inlet Reynolds number based on the tangential inlet velocity and
swirl chamber diameter.

3.2.2 Discharge Coefficient

The discharge coefficient based on mass flow of a simplex atomiser is defined as the ratio of
actual flow rate to theoretical maximum flow rate through the final orifice. In equation form:

c n =~ , , (3-21)
mmax

where

mmax = pAoUp (3-22)

m
,', C n
=
pA
(2L1P
__ n
J O
,5 .
(3-23)

o p

Numerous formulas have been proposed for the prediction of discharge coefficient of HCNs.
The equations that follow are some of the more practical formulas found in the literature. The
first equation is from Taylor's [48TAl] inviscid analysis of the HCN:
 Stellenbosch University http://scholar.sun.ac.za

20

(3-24)

Jones [82J01] provides an empirical equation containing all the key nozzle dimensions:

C = 0 45 ( PU p
d )
0
-0,02( I )
_0
-0,03(
_s
I ) 0,05
K 0,52
(d )
_s
0,23
(3-25)
n' !l do ds do

Som [83 SO 1] derived the basic form of his equation analytically and then adjusted the
constants and exponents according to experimental data:

3,377( n )1,520 (Is j dJ 0,117

(3-26)
Cn = ( Re )0,347(
do Ids
)1,166(
2a
)0,274'
gi

By far the simplest and potentially most accurate equation for discharge coefficient is from
Rizk and Lefebvre [85RI1]:

Cn = 0,35K' 05 ( ds j do ) 0,25 . (3-27)

3.2.3 Film Thickness

Taylor's [48TA1] analysis lead to an equation for z, Equation (3-7). This equation leads
directly to the film thickness. Due to the nature of the equations it is impossible to obtain an
explicit formula for z in terms of K. However, if K is known then z can either be read off
Figure 2-2 or solved iteratively:

(3-28)

On the topic of film thickness, now is the time to investigate the paper of Taylor [50TA1],
which deals with the question of boundary layer thickness. What follows is a short sample
calculation in which the boundary layer thickness in the final orifice is estimated for a nozzle
with rs = 34 mm, ro = 16 mm and ,1Pn = 28 kPa.

2 x 28000
Up =
998
= 7,483mjs.
 Stellenbosch University http://scholar.sun.ac.za

21

From Table 1 in Taylor [50TA1] take R] = ~ and read offo]:

rs

16
R} =-=047
34 '
".0] = 2,58.

The boundary layer thickness is:

(3-29)

but since .Q is less than roUp

o
->-
0] v
(3-30)
ro - Rj roUp sina

Therefore, substituting in the known values with a = 90°

o 2,58 1 x 10-6
-~--
ro 0,47 0,016 x 7,483 x sin90'
= 0,0159.

In this case, therefore, the boundary layer only extends inwards to a distance of one-sixtieth the
radius of the orifice, or in physical measurement only 0,25 mm. For the same nozzle the film
thickness, t, in the final orifice could be expected to be around 4 mm. Therefore it would seem
from this example that the boundary layer in the final orifice could be ignored in a theoretical
analysis of the swirl pressure atomiser.

Another equation for film thickness, which was seen earlier, IS that of Dahl and
Muschelknautz [92DA1]. We record it here again for convenience:

t ~ ra - ~r a '- q . (3-31)
rev a

Two more equations for film thickness from Rizk and Lefebvre [85RI1] round off this chapter
on the theoretical analysis of simplex atomisers. The first equation, which follows from a fairly
complex derivation, is implicit and therefore needs to be solved through iteration:
, 1560FN ~ 1+ X
(3-32)
t" = ~p!1Pn do (1- X)2 .

The coefficient, 1560, is dependent on the flow path of the liquid in the swirl chamber and on
the pressure drop across the final orifice - here it was derived from experimental data. The
 Stellenbosch University http://scholar.sun.ac.za

22

second equation is appreciably simpler than Equation (3-32) while still retaining all the key
parameters that have been shown to influence film thickness:
O

t = 3,66[d FNIl-
o ,25
(3-33)
JpM\
The flow number which is not a dimensionless number, but which rather has the units of length
squared, is defined thus:

(3-34)
 Stellenbosch University http://scholar.sun.ac.za

CHAPTER FOUR

APPARATUS AND TEST PROCEDURE

From the literature study it became apparent that contemporary swirl atomiser theory would
not be directly applicable to the question of analysing and predicting medium pressure nozzle
characteristics. In pursuit of some mathematical model for medium pressure nozzles it was,
therefore, necessary to gather experimental data through testing of prototype nozzles. In this
chapter the experimental techniques and the physical apparatus used in the project are
discussed.

Figure 4 - J: PholOgraph (?! lesl rig.

The main test facility. see Figure 4-1, was originally designed and built by Lake [93LA I] for an
undergraduate thesis in which he measured the rainfall distribution of several commercially
available FCNs. The variable nozzle. see Figure 4-3, which was also designed and built by him
allows nozzle components to be easily changed between tests. PVC components are placed
inside the main housing together with the necessary spacers so that the various parameters
affecting distribution might be investigated. His work, which involved much experimental
 Stellenbosch University http://scholar.sun.ac.za

24

testing, showed that the water distribution from commercial nozzles was often not uniform and
this finding prompted the present investigation.

4.1 Test Facility

In this section the operation of the test facility is discussed while mention is made of the
various components and flow metering devices. Figure 4-2 shows all the essential elements of
the test facility. Water is supplied to the nozzle by pumps which draw their water from an
underground reservoir. Test conditions are controlled by manipulating the control valve and
the by-pass valve. The flow rate is measured by means of a mercury manometer coupled to an
orifice plate which was designed, manufactured and installed in accordance with BS 1042. The

orifice plate has D and lhD taps and has plate has a diameter ratio, ~, of 0,68 with the supply
pipe which has an inside diameter of 50 ffiffi. The nozzle supply pressure is measured with a
mercury manometer, the open end of which is connected to a reservoir with a large surface
area, the surface of which can be positioned horizontal to the nozzle pressure tapping, thereby
negating the need for a zero reading during testing.

ORIFICE
CONTROL
PLATE VALVE
RESERVOIR

SUPPLY FROM
PUMPS
NOZZLE

ADJUSTABLE
HEIGHT
BLEED TO MAIN TANK

SUPPLY
PRESSURE
MANOMETER

RAIN SAMPLER

FLOWMETER
MANOMETER

RETURN FLOW

Figure 4 - 2: Layout of test facility.

 Stellenbosch University http://scholar.sun.ac.za

25

Figure 4-1 is a photograph of the test rig. The

rotameter visible on the right of the photograph
can only be used to measure flow rates less than
3,5 kg/s and was not used during the course of
these experiments. The rain sampler, another of
Lake's [93 LA I] designs, is an array of 45 mm
diameter cylinders, vertically aligned and equi-
spaced at 8 em intervals, see Figure 4-9. He
originally used the sampler to measure distribution
in a polar co-ordinate system, however, it was
found to be more convenient to obtain rainfall
intensity distribution, RID, data in a Cartesian co-
ordinate system. Consequently modifications were
made to the test facility to enable collection rainfall
intensity data in this format. One quadrant of the
spray is sampled, which is sufficient to characterise
the particular spray distribution. A flexible plastic
pipe between the nozzle and the rest of the pipe
network allows for easy adjustment of the nozzle Figure 4 - 4: Prototype nozzle.
installation height. h.

After passmg through the nozzle and

over the rain sampler, the water is
collected in the sump from where it then
drains back into the underground
reservoir. The inlet water temperature
was measured with an alcohol
thermometer.

Figure 4 - 3: PhotoKraph (?llested .~lI'irlers.

 Stellenbosch University http://scholar.sun.ac.za

26

4.2 Prototype Nozzle

The most striking internal feature of the HCN is the

swirler. Most HCNs make use of tangential inlet ports,
which means that the axial velocity component is almost
nil at the top of the swirl chamber. HCN s however,

usually make use of spiral inlets, which are produced by

cutting helical grooves through
Preliminary tests on a reproduction
the central
of a commercially
core.
Flow
direction
1I 1

WAS .--7I)M--""
available swirler showed significant pressure losses
~o

across the swirler. Therefore, it was decided to design a

streamlined swirler. Figure 4-3 shows a prototype nozzle Figure 4-5: Sharp-edged orifice.
with all its appendages ready for assembly. The nozzle is
assembled within the stainless steel housing by compressing the various components together
between a stopper rim, in the top of the housing, and the threaded end cap. The tested swirlers
can be seen in Figure 4-4. The object on the left is a wooden model of the streamlined swirler
while on the right is the plastic model of a commercial type swirler.

Most of the experimental work was carried out with nozzles using a sharp-edged orifice of the
type depicted in Figure 4-5. The six sharp-edged orifices manufactured had the following
diameters: 21,6 mm, 29,9 mm, 32,4 mm, 37,2 mm, 42,3 mm and 45,6 mm.

Figure 4-6 shows a typical rounded

exit that was used to investigate the
Coanda effect. Three of this type of
exit were manufactured, all having a
throat diameter, do, of 37,2 mm but

having exit angles, <Pn, of 90°, 110°

and 130°. The Coanda effect was
investigated by measuring the change
in spray cone angle that resulted from
Figure 4 - 6: Rounded nozzle exit. the use of the rounded exit profiles as
 Stellenbosch University http://scholar.sun.ac.za

27

Figure 4 - 7: Isometric view of cutter path for machining profiled outlet.

compared to the spray cone angle generated by the equivalent sharp-edged orifice. Figure 4-7
shows the result of a simulation of the NC codes for cutting the prototype profiled exit. A
mathematical description of the profile is given in Appendix D.

The streamlined swider has a central cavity in which

the central jet of the FCN is positioned, there is also a SWIRLER

CORE
solid plug which was used to seal off the cavity in the
event of HCN testing. Figure 4-7 shows the central

jet located inside the swider. The diameter, dj, and S~:~~E
protrusion length, Ij, of the central jet can be adjusted
by inter-changing the lower inserts, which fit firmly
into the main sleeve. Eight lower inserts were LOWER
INSERT

manufactured with the following jet diameters, dj: ~

12,5 mm, 13,25 mm, 14 mm, 15 mm, 15,5 mm,

16 mm, 17 mm and 18 mm. A diffuser type central jet Figure 4 _ 8: Central jet.
was also built and tested. This jet had a 12 mm inlet
diameter and a 17 mm outlet diameter and was
machined out to a linear 6° taper.
 Stellenbosch University http://scholar.sun.ac.za

28

4.3 Measurement of Characteristics

The following paragraphs describe the vanous methods employed III obtaining nozzle
characteristic data. Some of the methods are unorthodox but still effective under the
circumstances. For example, it was initially expected that the film thickness probe would
precisely indicate the interface between the air-core and liquid annulus meet within the orifice.
However, after readings had been taken, the exact position was difficult to locate due to the
insensitivity of the mercury manometer to the small pressures present in the water near the
inner surface of the annulus. This procedure did however, provide informative graphs of the
axial velocity profiles within the orifice.

4.3.1 Spray Cone Angle

A fundamental characteristic of the HeN that requires measuring is the spray cone angle, 28.

The angle in question is subtended by the tangents to the un-atornised liquid sheet and only
exists for a short distance beyond the final orifice. The task of measuring this angle is,
therefore, more perplexing than it might seem. Previous workers have made use of the method
of shadow graphing to record spray angles, however, this technique is ineffective for thicker
spray sheets generated by medium pressure swirl atomisers.

In this project two methods were employed to measure the spray cone angle. The first method
involved measuring the angle with a pair of dividers and protractor, and was used in pilot
experiments to verify the claim that spray angle is independent of injection pressure. The
second method was photographic and involved using a camera, mounted with lens
perpendicular to the axis of the nozzle and parallel to the plane of the orifice, to photograph
the spray so that the spray cone angle could then be measured with a protractor.

4.3.2 Film Thickness

The film thickness within the orifice was measured with the aid of a small probe mounted on a
traverse, see Figure 4-8. The probe consists of a thin pointed tube, with an inlet diameter of
 Stellenbosch University http://scholar.sun.ac.za

29

0,3 mm, connected to a mercury manometer. Since the probe is located at the nozzle exit the
static pressure of the water is assumed to be zero. This means that the pressure reading on the

manometer is the fluid dynamic pressure, lhpvo2, and can be used to calculate the axial velocity.

The film thickness can be calculated indirectly from the axial velocity profile or, directly
determined during the experiment as the point where the manometer reading becomes zero. To
measure the axial velocity profile the probe is moved incrementally across the nozzle exit while

the manometer readings are recorded at each increment. It is from these values that the axial
velocities are calculated. The film thickness is now the value of the radius at which the axial
velocity becomes equal to zero.

I NOZZLE

........

TRAVERSE MANOMETER

Figure 4 - 9: Axial velocity probe.

A photographic method was also implemented to measure the film thickness. This was done by
placing the camera directly below the centre of the nozzle. The camera was isolated from the
spray by a clear perspex sheet placed just above the lens. A macro lens was used so that the
photographs taken would have a narrow focal depth and therefore sharply focused on the
annulus of fluid within the orifice. A calibration photograph was also taken of the orifice,
before the water supply was turned on, so that the measurements taken from the photographs
could be accurately scaled. A spotlight reflecting off a parabolic mirror, below the camera,
provided sufficient lighting for the photographs. The air-core diameter was then measured to
scale off these photographs and the film thickness calculated accordingly. A sample
photograph can be seen in Chapter 5.
 Stellenbosch University http://scholar.sun.ac.za

30

4.3.3 Spray Distribution

The specific apparatus needed to measure a spray distribution are the rain sampler connected

to collecting troughs, in this case 5f! plastic bottles were used, a scale to weigh each bottle and

a stopwatch. A test is conducted at a pre-selected pressure with the nozzle positioned at a

specified height above the sampler. Before commencing a test all the troughs and bottles are
emptied and a cover is placed over the rain sampler. Now the water supply is turned on and the
valves are adjusted until the desired pressure registers on the manometer. When the flow has
stabilised the stopwatch is started and simultaneously the rain sampler cover is removed. The

length of time required to fill one 5f! bottle is dependent on the specific local flow rates at each

sampler. Fortunately the water inside the 5f! bottles is visible through the plastic, which

simplifies the task of deciding when to stop the test, and also gives a good visual indication of
the uniformity of the distribution.

80 mm 45 mm

0 CJ
•••

0 E
E
C>
C>
N

0 0 0 ~

CENTRE PIN

Figure 4 - 10: Rain sampler.

 Stellenbosch University http://scholar.sun.ac.za

31

The test is completed by replacing the cover on the rain sampler and recording the duration of

the test. The rainfall intensity is calculated from the mass of water collected in each sampler,
the duration of the test and the inlet area of the sampler. In the case of an asymmetric nozzle,
i.e. one producing a square pattern, the above process is repeated several times. For each new

set of readings the rain sampler is brought forward, parallel to the previous position, until the
rainfall intensity has been recorded for the entire wetted area beneath the nozzle. With the aid

of a spreadsheet program the results can then be displayed either as three dimensional relief
plots or as contour plots of the rainfall intensity, see Chapter 5 for characteristic samples.
 Stellenbosch University http://scholar.sun.ac.za

CHAPTER FIVE

EXPERIMENTAL WORK

5.1 The Swirler

The main objective for testing the two swirlers was to evaluate
the theoretical analysis of each swirl port geometry, which was
done with loss coefficients in Chapter 3. The swirler
configuration for these tests is shown in Figure 5-1. This
configuration ensures that the swirler being tested will
discharge at atmospheric pressure so that the pressure drop
Palm
across the swirler can be exactly calculated. The swirlers are
tested at a range of pressures while the corresponding flow
rates are calculated from the pressure drop across the orifice Figure 5 - 1: Swirler test
plate. The results of these experiments are shown graphically configuration.

in Figure 5-2 and Figure 5-3 where the actual port velocity is calculated from the continuity
equation. An example of how the actual velocities and loss coefficients are calculated is given
in Appendix B.

9
8
7
,---, 6
"-'
E 5
>-.
.•...•
'[5 4
0
d3
> 3
•
2
•
1 ...
0
0 5 10 15 20 25 30 35 40
llPsw [kPa]

.•. Ue {sharp} • Ue {streamlined} - Up {theory}

Figure 5 - 2: Comparison of swirl port velocities.

 Stellenbosch University http://scholar.sun.ac.za

33

2.5

1.5
,........,
'--'
~
I
• • •
v.:l
•
0.5 • •• • ••
,
0
0 5 10 15 20 25 30 35 40
L1Psw (kPa]

I. Sharp • Streamlined I
Figure 5 - 3: Comparison ofswirler loss coefficients.

100
• • 35
90 •
• • 30
80
• •
70 • • 25
• • ,........,
,........,
ce 60 • ~
:g • • 0
20 '--'
•...
•
••
'--' 50 l::
Q)

0-;" u
40 • 15 •...
• • Q)
0-;
30 • • 10
20 • •
••
10 •• 5

0 0
0 1 2 3 4 5
ill (kg/s]

I • Streamlined • Sharp -- % Difference I

Figure 5 - 4: Comparison ofHCNs with different swirlers.

The next experiment was aimed at determining the effect of swirler design on the required
operating pressure of complete HeNs. This was achieved by assembling two identical nozzles,
i.e. K = 0,421, with the two different swirlers and then measuring their respective mass flow
 Stellenbosch University http://scholar.sun.ac.za

34

rates over a range of injection pressures. Results from this experiment show the significant
reduction in operating pressure, which is achieved through the use of a streamlined swirler.
Figure 5-4 shows the comparative supply pressures of the two nozzles required to maintain the
various flow rates.

5.2 Hollow Cone Nozzle Tests

Having established that the new streamlined swirler considerably reduces the operating
pressure of the HCN, it was decided to abandon the sharp edged swirler in all further tests.
Thus, having decided only to use the new swirler in all remaining experimental work it was
therefore necessary to determine all the relevant spray characteristics of the new HCN.

5.2.1 Spray Cone Angle

The first spray characteristic to be measured

was the spray cone angle, 28. Figure 5-5

shows the liquid spray sheet issuing from a
nozzle with K = 0,632. The spray cone
angles of the nozzles tested were measured
off photographs like this one. Visual
observations and measurements taken with a
pair of dividers, using the method described
in Chapter 4, showed that 28 was dependent
on injection pressure for pressures less than
Figure 5 - 5: Photograph of spray cone.
15 kPa. At these low pressures the spray
cone IS not yet fully developed and the spray sheet tends to be unstable. Increasing the
pressure from 15 kPa to 25 kPa increases the spray cone angle and removes the instability
Increasing the injection pressure further above 25 kPa has no affect on the spray cone angle
With this in mind it was decided to measure the spray angle at an arbitrary pressure of 28
centimetres of mercury, roughly 35 kPa. In Figure 5-6 the spray cone angle data is shown
graphically together with angles predicted by the equations presented in Chapter 3. Sample
calculations implementing the various formulas are given in Appendix B.
 Stellenbosch University http://scholar.sun.ac.za

35

90
80
70
60 •
b' 50
'--'
a:>
N 40
30
20
10
o
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K [-]

I • Measured -Equ.(3-5) -Equ.(3-19) -Equ.(3-20) I

Figure 5 - 6: Spray cone angle versus nozzle parameter, K.

5.2.2 Discharge Coefficient

The data on discharge coefficient was collected from seven nozzles, which varied in geometry

0.7
0.6
0.5
~0.4
c:
u 0.3
0.2
0.1
o
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K [-]
• Measured - Equ.(3-24) - Equ.(3-25) - Equ.(3-26)
-Equ.(3-27) -Equ.(5-1)

Figure 5 - 7: Comparison of experimental and theoretical discharge coefficients.

 Stellenbosch University http://scholar.sun.ac.za

36

only with respect to orifice diameter. Measurements were taken at four different injection
pressures ranging from 12 kPa to 77 kPa. The objective for this experiment was to determine
whether or not discharge coefficient is affected by variations in injection pressure and if so
whether it would be necessary to incorporate this dependence in correlating the data.
Figure 5-7 shows the results from the test at 28 kPa and also shows the comparative
predictions for discharge coefficient by the formulas presented in Chapter 3. Equation (5-1)
was derived directly from Equation (3-27) - the two equations differing only by a factor. The
derivation of Equation (5-1) is discussed in Appendix B, for now it is sufficient merely to
present it here:

C n' = 0 578KO,5(dso./d )0,25 (5-1)

0.7

0.6

0.5
....... 0.4
I
•......••
C
u 0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K [-]
o 12 kPa 0 28 kPa t;,. 49 kPa o 77 kPa
-12kPa:Equ.(5-2) -77kPa:Equ.(5-2) -Equ.(5-1)

Figure 5 - 8: Discharge coefficient versus K.

Figure 5-8 shows the complete collection of discharge coefficient data from all seven
nozzles - the dependence of Cn on injection pressure is clearly evident in this figure.
Equation (5-2), which was also derived from Equation (3-27), takes the variation of discharge
coefficient with injection pressure into account with the use of a Reynolds number term which
replaces the coefficient in Equation (3-27). The equation looks as follows:

C
n'
= 1 468(R edo )-0,067Ko,5(d so./d )0,125 (5-2)
 Stellenbosch University http://scholar.sun.ac.za

37

The Reynolds number, Redo, is based on the pressure velocity, Up, and the orifice diameter of
the nozzle. In spite of these changes Equation (5-2) still retains many similarities with
Equation (3-27). The only real difference between these two equations is the exponent of one

eighth on the (d) do) term. This exponent was changed to compensate for the do term which

is hidden in the Reynolds number, however, the combined exponent for do in Equation (5-2) is
still approximately -0,7 as opposed to -0,75 in Equation (3-27). The exponent that does change
is the one on ds, which now has a value of -0,325 as opposed to -0,25 previously. The
derivation of Equation (5-2) is given in Appendix B.

5.2.3 Film Thickness

As was explained in Chapter 4, two methods were used to measure the film thickness within
the final orifice. The traverse mounted probe proved to be the more accurate and reliable
means of measuring the film thickness. Figure 5-9 shows the axial velocity profiles of a HCN
with do = 37,2 mm at four different injection pressures. The similarity between these profiles is
an indication of the reliability of this method of measurement, however, the imperceptible
difference in the air-core diameters, which may be due to the relatively large probe diameter,

12

10
.---,
'" 8
6
'----'
>.
.•..... 6
"(3
0
v
> 4

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
(ro - r)/ro [-]
I Pn [kPa]: II--B- 18 -----tr--- 36 ~ 73 --a-- 1221

Figure 5 - 9: Axial velocity profiles, K = 0,367.

 Stellenbosch University http://scholar.sun.ac.za

38

may also cause this method to appear inaccurate. However, measurements of the film
thickness of five different orifice diameters taken with the probe all indicate that film
thickness is not affected by variation in injection pressure. Therefore, it was concluded that for
the range of pressures tested film thickness could be assumed to be a constant for each nozzle.
With this in mind it was decided to implement a photographic method of film thickness
measurement .

Figure 5-10 shows the photograph of the air-core

of a HCN with do = 29,9 mm - not all the
photographs taken are as clear as this one.
Consequently slides were produced so that
measurements of the air-core diameters could be
taken from magnified images projected onto a
screen. This, however, did not Improve the
accuracy of the method and it is proposed that
perhaps a thermal imaging camera would result in
a better contrast between water and air and
therefore yield more accurate measurements of
film thickness.

Figure 5 - 10: Photograph a/air-core.

Figure 5-1 1 compares the measured data on film thickness with the various correlations and
theoretical formulas given in Chapter 3. Due to the similarity of the gradients of the series
labelled prohe and Equation (3-33) it was decided to use this equation as a basis for
correlating the film thickness. As with Equation (5-1), this equation differs from its
predecessor by a factor only:
lU':<
_
t - 9,81 IJPM\
I
L
d"FN~
P L1P" I. ~
I
I .
(5-3)

The method of determining the coefficient, 9,81, is given in Appendix B.

5.2A Spray Distribution

Experiments concerning the radial spray distribution of HCNs were done to obtain a more
complete understanding of the operation ofHCNs. Although the system of measurement which
 Stellenbosch University http://scholar.sun.ac.za

39

18
16
20 I
14
,...., 12
~ 10
•.....•
.•..8 o
6
4

o2 J -----+-----+-----+--------'------'----'-------1
f-,

o 0.1 0.2 0.3 0.4 0.5 0.6 0.7

K [-]
t,. Photographic 0 Probe - Equ.(3-28) - Equ.(3-31)
- Equ.(3-32) - Equ.(3-33) -Equ.(5-3)

Figure 5 - 11: Film thickness versus nozzle parameter, K.

only takes selected rainfall intensity data points at one position beneath the spray may not be a
true representation of the distribution, this is sufficient to obtain a visual record of the
distribution and adequately characterise the spray. The data points in all of the following
graphs have been fitted with cubic splines so that the general trend can be more readily
identified.

16
14
Pn: 35 kPa
12
h : 29 cm
~10
S
~8
...:.::
:::6
4
2
o
o 10 20 30 40 50 60
r [cm]

I do [mm]:1-32,4 -37,2 -45,61

Figure 5-12: Affect of varying orifice diameter.

 Stellenbosch University http://scholar.sun.ac.za

40

The ftrst experiments involved measuring the radial distribution of HCNs using the sharp-
edged oriftce so that the effect of increasing oriftce diameter could be investigated.
Figure 5-12 shows how the radial distribution moves outwards as oriftce diameter is increased,
this is to be expected since the spray cone angle is also greater for larger oriftce diameters.

14
12 K : 0,367
......, 10 Pn: 35 kPa
rn
N h : 38 cm
8
~
~
-
•.....• 6
4

2
0
0 10 20 30 40 50 60
r [cm]

I Exit angle: i- Sharp-edged - 90° - 110° - 130° I

Figure 5 - 13: Effect of rounding the nozzle exit.

The next experimental work was aimed at investigating the effect of rounding the fmal oriftce
on radial distribution. Although the ratio of oriftce diameter to swirl chamber diameter appears
to govern the radial distribution, rounding the fmal oriftce can dramatically alter this
distribution. Rounded outlets, see Figure 4-6, were constructed to cover a range of exit angles
in order that the practical limits of the Coanda effect could be determined. The effect of
rounding as compared to not rounding the nozzle exit is shown graphically in Figure 5-13.
Notice how the exit rounded to 130° seems only to flatten out the distribution without actually
increasing the radius of the spray.

Integration of this knowledge concerning the Coanda effect lead to the design and manufacture
of a non-axisymmetric ally proftled outlet to produce a square spray pattern. The proftle makes
use of a varying exit angle to induce the spray further outwards into the comers of the
projected square spray area. Figure 5-14 is a contour plot of the rainfall intensity measured
beneath the HCN using this proftled outlet. Although the distribution intensity was only
 Stellenbosch University http://scholar.sun.ac.za

41

measured over one quadrant of the spray, this distribution has been assumed to prevail in the
others quadrants also so that the complete distribution could be plotted for visual emphasis.
Figure 5-14 illustrates the squaring effect, which can be achieved with the use of a profiled
outlet. Appendix C, which describes the process of designing a FCN, contains a detailed
description of the geometry of such a profiled outlet.

[I: [kglrn2s] ~rID0-4 4-8 08-12 0 12-161

Figure 5 - 14: RID from nozzle with profiled exit.

5.3 Full Cone Nozzle Tests

Having tested the HCN and obtained useful correlations for all the necessary characteristics,
the next step was to test the FCN for a better understanding of its operation and distribution
characteristics. There are two aspects of the central jet, which affect distribution uniformity;
they are central jet diameter, dj, and central jet protrusion length, ~. See Figure 4-7 for a recap
of where these dimensions are measured on the nozzle. The graphs in the following sections
are an attempt to illustrate the affect of the various nozzle elements on the radial distribution
of the FCN. Where it is sufficient merely to show the distribution with a single line graph that
 Stellenbosch University http://scholar.sun.ac.za

42
has been done, however, in the event of non-symmetrical distributions, contour intensity
graphs similar to Figure 5-14have been used.

25

Pn: 35 kPa
20
K : 0,367
,....., h : 29 cm
N'" 15
lj : 0 mm
~
~ ~n: sharp
~ 10

0
0 10 20 30 40 50 60
r [cm]
I dj[mm]: 1-16 -17 -18 -121

Figure 5 -15: Measured RlDsfrom FeNs with sharp-edged orifice.

5.3.1 Influence of Central Jet

Many combinations of jet diameter and orifice diameter were tested during the course of this
project. Much of the data is of little significance, however, because of the irregular and uneven
distributions. Figure 5-15 shows some of the arbitrary distributions obtained from nozzles
using only the sharp-edged exit profile. Notice how these distributions are all hollow in the
centre. This is due to the strong influence of the rotating liquid sheet through which most of
the water exits the nozzle.
 Stellenbosch University http://scholar.sun.ac.za

43

18
16 Pn: 35 kPa
14 h : 29 cm
......,12 lj : 11 mm
CJl
'" !Pn: 130°
~ 10
=. 8
- 6
4
2
o
o 10 20 30 40 50 60
r [cm]

Idj I do [mm]: 1-16137 -15137 -15132 -141321

Figure 5 -16: Measured distributions from FeNs with various K values.

The remainder of the graphs in this section depict distributions from nozzles using the profiled
exit, !Pn = 130°. Notice, in Figure 5-16, how the curves for the nozzle with do = 32 mm have
lower average rainfall intensities. This is due to a reduction in outlet flow area, which causes a
greater pressure loss across the orifice, resulting in smaller mass flow rates through the nozzle.

7
6 Pn: 35 kPa
K : 0,367
......, 5 h :29cm
CJl

t 4 lj : 11,5 mm

-
='3
2
!Pn : 130°

1
0
0 10 20 30 40 50 60 70 80
r [cm]
Idj [mm]: 1-15 -15,5 -161

Figure 5 -17: Effect ojvaryingjet diameter.

The first aspect of the central jet to be investigated was central jet diameter, dj, which was
found to significantly affect the radial distribution of FeNs. Figure 5-17 contains selected data
which illustrates the effect of jet diameter, dj, on radial distribution. The data used in this
 Stellenbosch University http://scholar.sun.ac.za

44
illustration are from experiments conducted with nozzles using a jet protrusion length, lj, of
13 mm. The effect of varying the jet diameter, dj, can be summarised, in terms of the ideal
limit which would result in a perfectly flat and uniform curve, as follows:

• Increasing jet diameter, dj, beyond the ideal limit causes the rotating liquid sheet to become
entrained with the jet thus inducing the distribution to peak in the centre.

• Decreasing jet diameter, dj, marginally below the ideal limit causes the jet to become
diffused by the rotating liquid sheet resulting in a depression at the centre of the
distribution.

• Decreasing jet diameter, dj, substantially below the ideal limit causes the jet to stream
through the air-core, i.e. not interfering with the rotating liquid sheet, resulting m an
extreme peak at the centre of the distribution.

• The ideal jet diameter, dj, results in a distribution that is uniform or flat on top.

While experimenting with HCNs it became apparent that the discharge coefficient of the HCN
could be increased by inserting a solid core into the centre of the nozzle immediately below
the swirler. Not only did this solid core increase the flow rate, but also there was in fact an
optimal protrusion length, which resulted in a maximum discharge coefficient for each nozzle.
The discovery of this phenomenon prompted an investigation into the effect of jet protrusion
length, lj, on FeN performance. Figure 5-18 shows the effect of protrusion length, lj, on radial
distribution. The effects of varying the jet protrusion length, lj, can be summarised as follows:

• Increasing jet protrusion length, lj, has the same effect as increasing jet diameter, dj, namely

8
7 Po: 35 kPa
6 K : 0,367
'Cii'5 h : 29 cm
dj : 15,5mm
~4
~ cPo : 1300
•....•3
2
1
o
o 10 20 30 40 50 60 70 80
r [cm]
lj [mm]: 1- 10,5 -13 -11,5 [

Figure 5 - 18: Effect of varying protrusion length.

 Stellenbosch University http://scholar.sun.ac.za

45

that the distribution will peak in the centre.

• Decreasing jet protrusion length, lj, has the same effect as decreasing jet diameter, dj,
namely that the distribution will dip in the centre.
• As with jet diameter there is also an ideal protrusion length that will result in a uniform
spray distribution, however, this limit is dependent on both these quantities and cannot be
realised if either is outside of the range for such distributions.

Having gained an understanding of the effect of central jet geometry on FCN distribution, it
was then necessary to determine experimentally the relationship between central jet geometry
and orifice diameter. Ideally this work should have been done with a range of profiled nozzle
outlets, however, due to the expense of producing such a range of profiled outlets it was
decided to calibrate jet diameter, dj, and orifice diameter, do, for the sharp-edged orifice
nozzles only, and then to extrapolate the results to include profiled exits. The task of selecting
the jet diameter proved to be more difficult than was originally anticipated due to the fact that
jet inserts had only been manufactured at diameter intervals of 0,5 mm. After much
experimentation, mostly unrecorded visual testing, a series of uniform distributions were
obtained from which the following relationship was wrested:
dj = 0,0193XO,117. (5-4)

This correlation was determined from experimental data for nozzles with a jet protrusion
length, lj, of 9 mm and operating at an injection pressure, Pn, of 35 kPa. In this equation X is
the ratio of air-core area to orifice area and is a function of film thickness.

Having found a correlation for calculating the correct diameter of jet for a gIven orifice
diameter, the next experiment was aimed at determining the relationship between the flow rate
through the central jet and the flow rate through the swirler. This relationship is necessary for
accurately calculating the flow rate through complete FCNs. Figure 5-19 shows the
experimental set-up that enabled the jet flow rate to be measured independently of the swirler
flow rate. The water flowing through the jet was separated from the rest, which naturally
flowed through the swirl ports, collected in a large drum and then simply weighed on a scale.
The flow rate was then calculated by dividing the mass of ware collected by the time in
seconds that the drum took to fill. This was the only practical means of determining the
relationship between the central jet flow rate and the combined swirl port flow rate. Testing
was conducted over a range of flow rates and with jet diameters of 16, 17 and 18 mm.
 Stellenbosch University http://scholar.sun.ac.za

46

The anticipated result for this experiment was

that the flow rate through the jet would be
proportional to the ratio of jet area to inlet port
area, Ap - this was indeed found to be so. The
following correlation, which relates jet flow rate
to swirl port flow rate, was determined from the
experimental results:
A
q.J = 0,989-
A
J
qP . (5-5)
P

5.3.2 Comparison of Two FCNs

Until now, most the experiments concermng

distribution have been conducted on nozzles
Figure 5 - 19: Configuration for determining
that produce axially symmetric sprays. For
central jet flow rate.
practical reasons, however, cooling towers are
constructed with square cross-sections, which means that nozzles producing a circular spray
pattern cannot be effective in wetting the pack uniformly. In this section the prototype FeN
that has been used throughout this project was tested with the profiled exit, see Appendix D,
and compared to a commercially available cooling tower nozzle of similar dimensions. The
critical dimensions of the prototype nozzle are do = 37,2 mm, dj = 15,5 mm and lj = 5,5 mm
and of the commercial nozzle are: do = 37 mm, dj = 17 mm and lj = 0 mm. Both nozzles have a
swirl chamber diameter of approximately 68 mm. Figures 5-20 and 5-21 show sample
three-dimensional relief plots of the distribution intensity of the two nozzles.

Figures 5-22 and 5-23 again show the distribution intensity data of a these two nozzles, only
this time as contour plots. The distribution uniformity and spray pattern is readily identified in
each of these figures. As opposed to being compared at a given injection pressure, Pn, of
35 kPa, the two nozzles were tested at an equivalent mass flow rate, m, of 4,68 kg/so The
reason for this is that the two nozzles have different discharge coefficients resulting from the
different swirler designs and the only fair way to compare the rainfall distributions for such
nozzles is at equal mass flow rates.
 Stellenbosch University http://scholar.sun.ac.za

47

._-~.

I[kglrrfs]:
18
BJ15-18
......, /
.12-15
en 12' /
N

09-12
~ 9
~ 06-9
-
•......
6
3
o ~ 3-6
1&10-3
0
0
r [em]

r [em]

Figure 5 - 20: 3-D RID of prototype FCN with profiled outlet.

II[kglrrfs] :
~ 15-18
......, .12-15
N
en
09-12
~ 06-9
-
•......
o 3-6
@]0-3

r [em]

r [em]
--------------------------

Figure 5 - 21: 3-D RID of commercial FCN

 Stellenbosch University http://scholar.sun.ac.za

48

16
24
32
40
48
o 8 16 24 32 40 48
r [em]

II [kglrris]: ~0-3 .. 3-6 06-9 09-12 .12-15 ~ 15-18[

Figure 5 - 22: RID from prototype FCN with profiled outlet.

o 8 16 24 32 40
r [em]

[£~~rrisD []j] 0-3 3-6 06-9 0 9~12.12-15 ~15-181

Figure 5 - 23: RID from a commercial FCN

 Stellenbosch University http://scholar.sun.ac.za

CHAPTER SIX

DISCUSSION OF EXPERIMENTAL RESULTS

6.1 The Swirler

The most notable observation concerning the streamlined swirler is the significant reduction in
required operating pressure compared with the sharp-edged swirler. Figure 5-2 compares the
performance of two HCN s using different swirlers but having identical K values of 0,421 and
although this is only one test case, the trend is indicative of all HCNs using a more streamlined
swirler. At a supply pressure of35 kPa the nozzle utilising the streamlined swirler shows a gain
in flow rate of roughly 25 percent compared with the other nozzle. As the pressure increases
the percentage difference in flow rate approaches a constant value of 31,5 percent. This
levelling off is the result of the swirler loss coefficients approaching their asymptotic values as
the pressure drop across the swirlers increases. At a flow rate of 2,5 kgls the nozzle using the
streamlined swirler requires only 70 percent of the pressure required for the nozzle using the
sharp-edged swirler. This reduction in operating pressure is of primary importance to the
overall efficiency of the nozzle, especially when considering the long-term costs of cooling
tower operation.

Under normal operating conditions a HCN will exhibit two distinct pressure drops - one
across the swirler and the second across the final orifice. The method of predicting swirler loss
coefficients has proved to be an effective means of analysing the flow through the swirler,
therefore, by expanding this analysis it is also possible to estimate the individual pressure drops
across the swirler and the orifice. This is done by modifying the swirler loss coefficient to
include an exit loss coefficient, which can be attributed to the sudden expansion of the flow
into the swirl chamber. This extra term can be estimated with the following equation, modified
from the original White [86WH1] equation:

F ~ A
1
se ~ ( -
-p
A
J2 ' (6-1)
s

where Fse is the pipe loss coefficient for sudden expansion. Substitute the same values used in
Equation (3-2) namely, Ap = 932 mm2 and As = 3664 mm2, into Equation (6-1) to obtain a
value of 0,556 for Fse,which is valid for both sharp-edged and streamlined swirlers.
 Stellenbosch University http://scholar.sun.ac.za

50

The variation of experimental loss coefficient data with nozzle pressure drop, see Figure 5-3,
is included in this analysis through the use of the following approximate power law curve fit
equations for the two swirlers:

• streamlined: Fsw ~ 13, 3LW sw -1,097 + 0556

, (6-2)
-0,942
• and sharp: Fsw ~ 27,2LWsw + 0,556, (6-3)

where Fsw is the total loss coefficient for swirlers with submerged exits, i.e. under normal
operating conditions. Rearranging Equation (3-1) yields the following equation for pressure
drop across the swirler:
2
LWsw = 1/2 pU e (Fsw + 1), (6-4)

where swirl port velocity, Ue, can be calculated from one of the correlations for discharge
coefficient namely Equation (5-1) or Equation (5-2).

50

Sharp
40
Streamlined

10

o
o 20 40 60 80 100
LWn [kPa]
K: ~ 0,367 -fr- 0,367 ~ 0,456 -B- 0,456 --+--- 0,632 --+-- 0,632

Figure 6 -1: Swirler pressure drop in a HeN

Figure 6-1 shows the results of this analysis for nozzles with different K values and different
swirlers. The sample calculations can be seen in Appendix B.7. Referring back to Figure 5-1 it
becomes apparent why the discharge coefficient falls away when the nozzle parameter, K,
exceeds 0,5. This is due to the high swirler loss coefficients that occur at low swirler pressure
drops. Minimum swirler pressure drops of 10 kPa for the streamlined swirler and 15 kPa for
the sharp swirler are required if the loss coefficients are to approach their asymptotic values.
 Stellenbosch University http://scholar.sun.ac.za

51

For the nozzles with K = 0,632 the swirler pressure drop does not exceed either of these
minimum values even at a nozzle pressure drop of 80 kPa, therefore, at injection pressures of
less 80 kPa these nozzles will incur high irreversible losses through the swirler.

6.2 HCN Characteristics

Spray cone angle is the only characteristic that was measured and found to be in good
agreement with the theoretical treatment of Chapter 3. Unfortunately, due to the distortion of
this angle that is caused by the presence of the central jet in the FCN, knowledge of this angle
is of little use in designing FCNs. This does not mean that the spray cone angle of a FCN
cannot be accurately predicted, but rather that it can be, especially when a rounded exit profile
with known exit angle is attached to the final orifice. In the case of a symmetrical FCN the

spray cone angle can be expected to be within 5 degrees of the nozzle exit angle, <Pn.

The discharge coefficients of the tested nozzles show both a steeper gradient and higher all-
round values compared with the estimates from the equations in Chapter 3. There are two
reasons for the increased discharge coefficient. Firstly, streamlining the swirl ports reduced the
required operating pressure by minimising the irreversible losses. Secondly, these higher values
are due to the use of spiralled inlet ports as opposed to tangential inlets for which the
theoretical formulas were derived.

There are two factors that are related to port entry angle and which greatly affect discharge
coefficient. They are viscous dissipation, which takes place along the length of the swirl
chamber, and back pressure, which is dependent on the ratio of axial to tangential velocity at
the entry to the swirl chamber. For tangential inlet nozzles the velocity of the liquid entering
the swirl chamber is almost purely tangential. This means that in order for the nozzle to
discharge any liquid at all, some of the kinetic energy in the swirling liquid must be used to
increase the axial velocity component. The conversion of tangential velocity to axial velocity is
not in itself a loss, but the viscous losses associated with this process are significant.

Tangential inlet nozzles have lower discharge coefficients due to the greater back pressure,
which is experienced at the swirl port exits. Back pressure exists because of the radial pressure
 Stellenbosch University http://scholar.sun.ac.za

52

gradient that is set up by the swirling liquid. The higher the ratio of tangential to axial velocity
at the swirler exit the greater will be the radial pressure gradient and the greater will be the
back pressure also. This increases the required injection pressure and reduces the discharge
coefficient. The fact that testing was done at relatively low injection pressures is also a factor
that has contributed to the increased discharge coefficients for the experimental nozzles. This is
apparent from the dependence of discharge on injection pressure and is evident in Figure 5-7.
Equation (5-1) and Equation (5-2) have been shown to satisfactorily correlate the measured
discharge coefficient data.

Film thickness showed little variation with injection pressure. In spite of this finding the

equation used to correlate film thickness contains a pressure term. However, due to the nature
of this equation, i.e. taking the fourth root of the various arguments, the influence of pressure
is kept to a minimum.

Radial distribution from HCN s is affected by orifice diameter and orifice rounding and is also a
function of the height of the nozzle above the plane of measurement. Judging from Figure 5-13
there appears to be a limit to the effectiveness of rounding of the nozzle outlet. The 90° and
110° exit profiles move the distribution radially outwards while the 130° profile only succeeds
in flattening out the distribution without actually shifting it any further outwards than was
achieved with the 110° profile. Figure 5-14 shows that a square distribution can be created if
the right exit profile is used.

6.3 FCN Characteristics

Jet diameter and jet protrusion length are two parameters that have been shown to influence
the distribution uniformity of FCNs. The general trend and relationship between these two
dimensions and the orifice diameter have already been discussed in Chapter 5, however, there
are some finer points on central jet selection which have not yet been discussed.

The first point concerns jet diameter. If a distribution is observed to have a peak of higher
intensity halfway between the centre of the spray and the perimeter, as shown by curve labelled
dj = 16 mm in Figure 5-17, this is an indication that the jet diameter is too large for the nozzle
 Stellenbosch University http://scholar.sun.ac.za

53

orifice diameter. Such a distribution cannot be made uniform by altering Ij because decreasing Ij
will only amplify the trough at the centre and increasing Ijwill only exaggerate the bulge, which

is already present in the distribution.

The indication of dj being too small is evident from a distribution that exhibits one peak in the

centre and another towards the edge of the spray. A distribution such as described here cannot

be made uniform by adjusting the protrusion length. The explanation for this is that increasing

Ij will only amplify the central peak and decreasing Ij will only magnify the peripheral peak. The
three curves in Figure 5-18 are a good illustration of this phenomenon, the distribution with

Ij = 11,5 mm is, however, very close to being perfectly uniform.

The ideal jet diameter will cause the distribution to peak in the centre when Ij is too large and
cause the distribution to peak at the perimeter when Ij is too small. However, as Ij is gradually
decreased from being too large the distribution will flatten out in the centre and build up on the

perimeter. This process will result in a distribution that seesaws from a peak in the centre to a

peak at the perimeter. The ideal protrusion length, which corresponds to the ideal jet diameter,

will ultimately lead to a perfectly uniform distribution. The correlation obtained for dj as a

function of X, Equation (5-4), is a necessary and very useful tool for designing FCNs. Even

though this is strictly only valid for sharp-edged orifices, the mere fact that such a relationship

exists at all is useful knowledge. In Appendix D this equation was adapted by adding the

difference between dj obtained from Equation (5-4) and dj determined experimentally with the

FCN using the profiled exit.

Figure 5-20 and Figure 5-21 compares three-dimensional graphs of one quadrant of the spray

distributions of the prototype and commercial FCNs. From these graphs it would appear that

the prototype nozzle delivers a much more even distribution. The large peak in the centre of

the distribution of the commercial nozzle is an unwanted characteristic. In Figures 5-22 and

5-23 the same data used in Figure 5-20 and Figure 5-21 is used to generate contour plots of

the two distributions. These figures clearly illustrate the improvement in distribution uniformity

that is achieved with the prototype nozzle. Both distributions have some semblance to the ideal

square spray pattern, though it is difficult to determine which spray pattern is more square.
 Stellenbosch University http://scholar.sun.ac.za

CHAPTER SEVEN

NOZZLE MANUFACTURE

There are several manufacturing techniques that could be used in the production of cooling
tower nozzles. There are numerous casting methods like the lost wax process, centrifugal die
casting and permanent mould injection die casting. Casting is the most suitable tool for the

manufacture of full cone nozzles, however, it may also be possible, if not too expensive, to
machine the nozzle housings.

7.1 Plastic Injection Moulding

The manufacture of full cone nozzles for cooling towers lends itself to plastic injection
moulding (PIM) and is the logical choice for such an assignment. The choice of plastic is the
most important consideration in the design process. There are many excellent plastics available

today, all with their own unique properties and limitations. Some of the properties that are
listed for plastics today are heat, chemical and UV resistance, tensile strength, hardness,
dielectric strength and flammability.

The choice of plastic for a cooling tower nozzle is governed by the operating conditions inside
the cooling tower. These conditions include: moderate pressures, possibly high temperatures,
60°C is regarded as high for plastic components, a certain amount of vibration. Often the
water is highly chlorinated or being treated with some other chemicals making it essential to
know the chemical resistance of the plastic materia!. The degree of vibration may promote the
selection of an amorphous plastic above a plastic that has a crystalline structure.

There are many plastics suitable for use in a cooling tower environment, especially since
chlorinated water is not normally used in cooling towers. The list of available plastics includes
ABS, PA, POM, PC, PP, PVC, PBTP and PE. The choice of plastic should not be governed
solely by cost but should rather be a decision based on functional requirements, e.g. if the

maximum operating temperature, under the above conditions, of plastic A is 50°C and of

plastic B is 120°C then plastic B should be chosen, even if plastic B is more expensive than
 Stellenbosch University http://scholar.sun.ac.za

55

plastic A. The three plastics that go highly recommended for this application are ABS, POM
and PC.

Once a suitable material has been found it is time to design the mould. Injection moulding and
especially mould design is a highly specialised industry where years of experience is a decided
advantage. For the inexperienced designer a sound mould cavity can be produced if common
sense is used and if cognisance is taken of three important aspects of mould design.

First it is important to remember that, as a general rule, the maXImum section thickness
permissible with plastics is :f:4 mm. A section that is too thick will result in the formation of

voids or will warp, buckle and shrink. Second a well designed cooling and heating system is an
essential part of the mould. Poor heating or too much cooling, anywhere in the mould, can
cause premature freezing of the molten plastic which may result in thermal stresses or even
fractures within the component. The cooling system ensures that component surfaces cool
evenly thus leaving the surfaces smooth and flush. Thirdly the venting system should be
designed in conjunction with the other elements of the mould and should not just be added to
the mould as an afterthought. Sound venting ensures that the mould is completely filled during
the injection process and therefore improves the quality of the final product.

Another aspect of mould design is the question of mould material. Moulds can either be made
from high quality wear resistant steel or from case-hardenable steel. Steel moulds are cheaper
to produce but can themselves only produce a limited number of castings, :f:10 000. Two more
factors which dramatically affect production costs are the specified dimensional tolerances and
surface finish. A tolerance of :f:0,1 mm is three times more expensive to produce than a

tolerance of :f:0,5 mm, while a specified surface roughness of 0,25 ~m/m could result In

production costs being up to ten times that required to manufacture to a finish of 6 ~m/m.

7.2 Proposed Nozzle Design

The nozzle design presented here is a modular construction, which allows nozzle components
to be interchanged to precipitate alternative performance characteristics. The fundamental
element in the nozzle is the swider. One swider is the basis for a range of nozzles, i.e. for a
 Stellenbosch University http://scholar.sun.ac.za

56

given swirl chamber diameter one swirler is manufactured while two or three final orifices can
be inter-changed within the swirler to produce different nozzles. Each orifice is produced
together with a central jet, which also fits into the single swirler.

Each nozzle is designed according to the requirements of its specific application. Typical
requirements for a cooling tower nozzle will be rainfall intensity and nozzle installation data
such as nozzle height above fill pack and the area requiring to be wetted. Appendix D contains

a complete worked example on how to design a full cone nozzle. As far as actual
manufacturing is concerned the following modular construction is proposed:

• A basic housing is used for assembling between

~
\ :t 3° j

two and three nozzles with different orifice diameters,

STOPPER RIM
FOR SWIRLER see Figure 7-1 for a diagram of the proposed housing
design.
STOPPER RIM
/ FOR ORIFICE

Figure 7 - 1: Nozzle housing.

CENTRAL JET ----.0
• The swirler is produced with a central cavity
SWIRLER
into which various central jets fit firmly. The
cavity is the same shape as the one depicted in
STOPPER RIM
Figure 4-7, except that a taper must be included FOR CENTRAL JET

to facilitate the injection moulding. Figure 7-2 is

a sectional view of the swirler and central jet.

• The orifice and rounded outlet will be

manufactured as one unit. The orientation of the Figure 7 - 2: Swirler and central jet.

orifice within the nozzle is also important due to the fact that the square spray must be
orientated to coincide with the tower fill pack. Figure 7-3 shows a cross-section of the
proposed moulded orifice.

• The central jet is moulded to the exact length and diameter

necessary to produce a uniform spray with the corresponding
orifice. The central jet can either be glued or clipped into the
central cavity of the swirler.

Figure 7 - 3: Moulded orifice.

 Stellenbosch University http://scholar.sun.ac.za

CHAPTER EIGHT

CONCLUSIONS AND RECOMMENDATIONS

The primary objective for this project has been to design, test and correlate the behaviour of a
large pressure jet atomiser for water distribution in a square spray pattern for use in industrial
type cooling towers. This objective lead to theoretical and experimental examination of hollow
and full-cone simplex swirl atomisers.

In many respects, however, it must be said that this thesis has only just scratched the surface as
far as swirl spray atomisers are concerned. For instance, it is probable that the swirl port exit

angle, y, exerts a significant influence on the performance characteristics of the HeN and yet in
this study only one swirler outlet angle of 41,50 was used throughout all of the experimental
work.

Two more parameters, which have remained unchanged throughout this project, are inlet port
area and number of ports. These parameters have a significant effect on spray uniformity. The
balance between frictional loss caused by too many ports and spray non-uniformity caused by
too few ports is an aspect, which has not been investigated. Also, the relationship between

swirl port exit angle, y, and swirl port length, lp, is one that may well hold significant influence
on the length and shape of the swirl chamber. The purpose of the swirl chamber is to damp out
the separate streams from the swirl ports and thus ensure that a uniform liquid annulus will

form in the final orifice. However, if y could be chosen so that liquid would enter the swirl

chamber with the exact velocity components needed at the exit to produce a specific spray
cone angle and if the swirler could be constructed to deliver a semi-continuous liquid sheet into
the swirl chamber, then swirl chambers could be designed a fifth of their current length. This
would dramatically increase discharge coefficients allowing researchers to focus their efforts
on refining the drop size distribution of sprays.

Experimental testing of existing cooling tower nozzles revealed that the performance of these
nozzles was less than optimal, both with respect to distribution uniformity and required
operating pressure. The claim that these nozzles produced a square spray pattern was also
found wanting. The geometry of some of the swirlers in these nozzles was found to be of a
rather crude design and it was clearly evident that the sharp edges and overhanging lips would
 Stellenbosch University http://scholar.sun.ac.za

58

contribute to excessively high flow losses. Consequently it was decided to initiate this project
by redesigning the swirler.

The new swirler was designed to have the same port area, Ap, and port entry angle, y, as the
sharp-edged swirler so that the behaviour of the two swirlers could be evaluated and compared
through experimental testing. A theoretical analysis of the swirl port geometries lead to crude
if not effective prediction of the loss coefficients associated with the ports. The loss coefficient
predicted by this method more than three times greater for the sharp-edged swirler than for the
streamlined swirler, a finding which was later validated through experimental work.

The testing of complete HCN s containing the different swirlers showed that required operating
pressure can be significantly reduced for a given nozzle if a streamlined swirler is used. Having
observed the advantages of using a streamlined swirler it was decided to complete the
remaining experiments with nozzles embodying only the streamlined swirler.

Tests were conducted to determine the effect of the variation of orifice diameter, do, and
nozzle supply pressure, Pn, on the performance characteristics of the HCN. These experiments

showed that spray cone angle, 28, and film thickness, t, are not affected by variation in
injection pressure but that discharge coefficient, Cn, has a marked dependence on the orifice
Reynolds, Redo. Equations from the work of Rizk and Lefebvre [85RIl] proved useful in
obtaining correlations for film thickness and discharge coefficient.

By rounding the nozzle outlet it was found that the radial spray distribution could be flattened
out, which is beneficial to the rainfall intensity distribution of the FCN. Knowledge gained from
experiments done to quantify the effect of rounding the orifice outlet was used to design a
profiled nozzle outlet that actually produces a square spray pattern.

FCN tests were then done to determine the effect of jet diameter, dj, jet protrusion length, lj,
and orifice diameter, do, on the distribution uniformity of the tested nozzles. Results showed
that a unique relationship exists between jet diameter and orifice diameter for any given jet
protrusion length. This relationship was used to develop a code for designing FCNs for specific
applications. Finally a complete FCN was tested and found to produce spray with uniform
rainfall intensity distribution in the desired square spray pattern.
 Stellenbosch University http://scholar.sun.ac.za

R1

REFERENCES

[48T AI] Taylor, G.I., The Mechanics of Swirl Atomizers, Proc. 7th Intern. Congr. Appl.
Mech., Vol. 2, pp. 280 -285, 1948.

[50TA1] Taylor, G.I., The Boundary Layer in the Converging Nozzle of a Swirl Atomizer,

Quart. Journal of Mechanics and Applied Mathematics, Vol. 3, Pt. 2, pp. 129-139,
1950.

[60F01] Foust, AS., Wenzel, L.A, Clump, C.W., Maus, L. and Andersen, L.B., Principles
of Unit Operations, John Wiley & Sons, New York, pp. 541-543, 1960.

[69D01] Dombrowski, N. and Hasson, D., The Flow Characteristics of Swirl (Centrifugal)

Spray Pressure Nozzles, AIChE Journal, Vol. 15, July, pp. 604-611, 1969.

[78SA1] Sada, E., Takahashi, K., Morikawa, K. and Ito, S., Drop Size Distribution by Full

Cone Nozzle, Canadian J. Chern. Eng., Vol. 56, pp. 455-459, August 1978.

[81BS 1] British Standards Institution, Methods of measurement of Fluid flow in closed

conduits, BS 1042: Section 1.1, pp. 33-39, 1981.

[82JOl] Jones, AR., Design of Large Pressure Jet Atomiser for Power Plant, ICLASS-82,
pp. 181-185,1982.

[83KR1] Krane, S.c., The Effect of Non-uniform Water Distribution on Cooling Tower
Performance, J. Energy, Vol. 7, no. 6, pp. 636-639, 1983.

[83S01] Sam, S.K., Theoretical and Experimental Studies on the Coefficient of Discharge
and Spray Cone Angle of a Swirl Spray Pressure Nozzle Using a Power-Law Non-
Newtonian Fluid, J. of Non-Newtonian Fluid Mechanics, Vol. 12, pp. 39-68, 1980.

[84WAl] Walzel, P., Design of Single Substance Pressure Atomizers, German Chern. Eng.
Vol. 7, pp. 1-12, 1984.
 Stellenbosch University http://scholar.sun.ac.za

R2

[85RI1] Rizk, N.K. and Lefebvre, A.H., Internal Flow Characteristics of Simplex Swirl

Atomizers, 1. Propulsion, Vol. 1, NO.3, pp. 193-199, May-June, 1985.

[85S01] Som, S.K. and Biswas, G., Dispersion of Spray from Swirl Nozzles, Department of
Mech. Eng., Indian Institute of Technology, Kharagpur 721302 (India), pp. 191-200,
1985.

[86HO 1] Horvay, M. and Leuckel, W., Experimental and Theoretical Investigation of Swirl
Nozzles for Pressure-Jet Atomization, Germ. Chern. Eng., Vol. 9. pp. 276-283,
1986.

[86WH1] White, F.M., Fluid Mechanics, International Edition, McGraw-Hill Book Co.,
Singapore, pp. 335-336, 1986.

[87PR1] Prywer, J and Kulesza, 1., Jet-Centrifugal Atomizer in Application to Spray-

Evaporative Condenser, Proc. 17th Intern. Congr. Refrig., Vienna, Austria, Vol. B,
pp. 775-781, August 1987.

[87SC1] Schultz, T., The Estimated Effect of Uneven Water Distribution on the
Performance of Cooling Tower Fills, B. Eng. Thesis, University of Stellenbosch,
October 1989.

[88BE1] Bellagamba, B., Dinelli, G., Tognotti, L., Zanelli, S., Water Distribution in Cooling
Towers: Characterisation of Industrial Spray Nozzles, Proc. Intern. Cooling Tower
Conference (Proc. April 1989) Pisa, Italy, pp. 1.4: 1-10,1988.

[89KR1] Kroger, D.G., Cooling Tower Performance and Design, Department of Mech.
Eng., University of Stellenbosch, Stellenbosch, pp. A.9, 1989.

[89LE 1] Lefebvre, A.H., Atomization and Sprays, Hemisphere Publishing Corporation, New
York, pp. 5; 112-117; 165-189; 281-300, 1989.
 Stellenbosch University http://scholar.sun.ac.za

R3
[91MUl]Muschelknautz, E. and Trefz, M., VDI-Wiirmeatlas, VDI-Verlag, Dusseldorf, 6.
Aufl., pp. Ljl-Lj9, 1991.

[92DAl] Dahl, H.D. and Muschelknautz, E., Atomization of Liquids and Suspensions with

Hollow Cone Nozzles, Chern. Eng. Technol. 15, pp. 224-231,1992.

[93CH1] Chen, S.K., Lefebvre, A.H. and Rollbuhler, 1., Factors Influencing the
circumferential Liquid Distribution from Pressure-Swirl Atomizers, Trans. of
ASME, Vol. 115, pp. 447-452, 1993.

[93DUl] Dumouchel, c., Bloor, M.I.G, Dombrowski, N., Ingham, D.B. and Ledoux, M.,

Viscous Flow in a Swirl Atomizer, Chem. Eng. Sci., Vol. 48, NO.1, pp. 81-87,
1993.

[93K01] Koo, YM. and Kuhlman, D.K., Theoretical Spray Performance of Swirl-Type
Nozzles, Trans. of ASAE, Vol. 36, pp. 671-678, May-June, 1993.

[93K02] Koo, YM. and Kuhlman, D.K., Spray Performance Models of a Designed Swirl

Nozzle, Trans. of ASAE, Vol. 36, pp. 679-684, May-June, 1993.

[93KR1] Kranc, S.c., Performance of Counterflow Cooling Towers with Structured

Packings and Maldistributed Water Flow, Numerical Heat Transfer, Part A, Vol.
23, pp. 115-27, 1993.

[93LA1] Lake, P., Spray Distributions under Medium Pressure Nozzlesfor Cooling Towers,
B. Eng. Thesis, University of Stellenbosch, November 1993.
 Stellenbosch University http://scholar.sun.ac.za

APPENDIX A

PHYSICAL PROPERTIES OF WATER

The following two equations for calculating water density and absolute viscosity are taken
from Kroger [83KRl]'

Saturated water liquid from 273,15 K to 380 K.

Density:

(A-I)

where:
a = 1,49343xl0-3

b = -3,7164xl0-6

C = 7,09782xl0-9
d = -1 ,90321xl0-20

Absolute Viscosity:

f.l = ax 10b/(T-cl: kg/ms (A-2)

where:
a = 2,414xl0-s

b = 247,8
c = 140
 Stellenbosch University http://scholar.sun.ac.za

APPENDIX 8

SAMPLE CALCULATIONS

The purpose of this appendix is to show the reader how the various formulas and calculation
methods were implemented in obtaining the data for the numerous graphs of Chapter 5 and

Chapter 6. An attempt has been made to present the calculations in the order in which the
relevant equations appear in the thesis. Unless otherwise stated the following conditions will
apply to all the sample calculations:

• water temperature: 30 °C (~ 303 K)

• supply pressure manometer reading: 25 cm Hg

• flow meter manometer reading: 14,5 cm Hg

• Ap: 932 mm2; ds: 68,3 mm; do: 37,2 mm; Is: 68,3 mm; 1 3 mm; dp: 16 mm; Ip: 72 mm
0:

Refer to Figure 2-1 for a recap of where these various dimension apply.

B.1 Calculation of Mass Flow Rate

The most important calculation is to calculate the actual mass flow rate from the reading on the
flow meter manometer. To do this we must first calculate the water density and viscosity from
the equations given in Appendix A.

The density of water at 303 K is calculated from Equation (A-I):

P = (1,49343 X 10-3 - 3,7164 xI 0-6 x 303 + 7,09782 x 10-9 x 3032 -1,90321 X 10-20 x 3036 rl

= 995,7 kg/m 3
.

The absolute viscosity of water at 303 K is calculated from Equation (A-2):

( 247,8 )
l.l = 2,414 X 10-5 x 10 303-140

= 7,998 X 10-4 kg/ms.

The pressure drop across the orifice is calculated from the following equation:

(B-1)

where g is taken as 9,796 mls2 and PHg is taken as 13 550 kg/m3.

Substituting into Equation (B-1) we obtain the flow meter pressure drop in Pascal:
 Stellenbosch University http://scholar.sun.ac.za

B2

i1Pfm = (13550 - 995,7)x9,796xO,145

= 17 832 Pa.

The flow rate can now be determined iteratively with the following routine.
1. Guess the pipe Reynolds number for the flow leading up to the flow meter orifice.
2. Calculate the discharge coefficient, Cd, from the Stoltz equation:

(B-2)

3. Calculate the mass flow rate from the following equation:

(B-3)

where At refers to the throat area of the orifice plate.

4. Calculate a new pipe Reynolds number based on the calculated flow rate and use this
Reynolds number to re-calculate Cd from Equation (B-2).

5. Repeat until the Reo at step 2. and the Reo at step 4. differ only by a selected tolerance.

For this sample calculation we shall use an already converged solution for Reo as the initial
guess in 1. Therefore, substituting Reo = 118 988 into Equation (B-2) we obtain:

Cd = 0,5959 + 0,0312 x 0,682,1 - 0,184 X 0,688 + 0,0029 X 0,682.5

6 ]0'75
10 +
[ 118988

+ 0,039 x 0,68 4( 1- 0,68 4)-1 - 0,0337 x 0,47 x 0,68 3

= 0,61245.
Calculate the flow rate from Equation (B-3):

2x17832
ri1 = 0,61245 x 995,7 x rcO,0172 -----
995, 7( 1- 0,684)
= 3,737 kg/so
The new Reo is obtained from the following equation:

Reo = 4 rh (B-4)
rcD~

Substituting into Equation (B-4) we obtain:

 Stellenbosch University http://scholar.sun.ac.za

B3

Re = 4 x 3,737
4
D 7t 0,05 x 7,998 x 10-

= 118982,

which only differs from the initial guess by 0,005 percent. Therefore, the calculated mass flow
rate of 3,737 kg/s is correct.

B.2 Calculation of Swirl Port Velocity

The velocity through a port is calculated from the measured mass flows, while the predicted
velocities are calculated from the estimated loss coefficients using Equation (3-3). For an
example of how these values are calculated take the actual test conditions:

• the water temperature is 55° C, therefore from Equation (A-I) p = 985,8 kg/m3

• the pressure drop across the swider is 25,23 kPa

• the total mass flow rate is 5,65 kg/so

The predicted port velocity for the streamlined swider is calculated from Equation (3-3):

2 x 25230
U
ep
= 985,8x(0,4+1]
= 6,047 mis,
while the actual measured port velocity is calculated from the continuity equation:
m
U =-
e pAp
5,65
=
985,8 x 932 x 10-6
= 6,149 m/s.
Rearranging Equation (3-1) the actual loss coefficient can also be calculated:
2~
LF j = pu:; -1

2 x 25230
=------1 2
985,8 x 6,149
= 0,354,
which is close to the original estimate.
 Stellenbosch University http://scholar.sun.ac.za

B4

B.3 Calculation of Spray Cone Angle

First the nozzle supply pressure is calculated from Equation (B-1):

i1Pn = (13 550 - 995,7)x9,796xO,28

= 34 435 Pa.

The first calculation of spray cone angle is from the formula of Taylor [48TA1]. Due to the
nature of the relationship between K, y and Cn, it is first necessary to iterate to find the value of
y that yields a Cn from Equation (3-24) which will then correspond to the correct value of K.
In this example the value ofK is:
932
K=---
68,3 x 37,2
= 0,367.

In order to shorten this process we shall demonstrate that a value of y = 0,5904 will yield the
above K. Start by calculating Z2 from Equation (3-7) by substituting y = 0,5904:

= 0,59042 + 0,59044 0,59042

Z2
---+
4 16 2
= 0,5136
:. z = 0,7167.

Substitute into Equation (3-24) and obtain the unknown discharge coefficient:

0,59042
Cd = (1-0,5136) 1-
0,5136
= 0,2757.

Now substitute the above values into Equation (3-8):

K = nO,2757
4 x 0,5904
= 0,367,

which is the value calculated earlier.

The value ofx is calculated by manipulating Equation (3-6):

2
:. x = 1_ 0,5904
0,5136
= 0,5668.
 Stellenbosch University http://scholar.sun.ac.za

B5

The spray cone angle is now calculated by substituting for x, y, z and Z2 into Equation (3-5):

0,5904.J2 x 0,5136 [ 1 (1) ]

cos8 = 0,5668 + 3 - -- -1 + loge 0,7167
(1-0,5136)2 2 0,5136
= 0,744
:.28 = 2 cos-l (0,744)
= 83,8 0.

The method of Dahl and Muschelknautz [92DA1] required iteration in order to calculate Wo
from Equation (3-9), this was due to the fact that the wall friction coefficient is a function of
the mean Reynolds number which in turn is also related to woo The final iteration yielded a
value for Woof 5,298 mis, which was based on the actual measured mass flow rate through the
nozzle. First the port entry velocity, Ue, is again calculated from continuity:
3,737
U =------
e 995,74 x 932 x 10-6
= 4,027 m/s.

Next Remis calculated as follows:

p(UeSin~)+Wo)( 7)
Rem =
~

995, { 4,027sin( 4~Y)+ 5,298)( 0,03415; 0,0186)

=
7,998 x 10-4
= 176911,

where Uesin(y) is the tangential component of the port entry velocity. The value of A which
corresponds to this Reynolds number is 0,0034. The frictional area is calculated as follows:

re(ds2 - do2)
AT ~ red)s + 4relpdp + 4

re(0,06832 -0,03722)
= reO,0683x 0,0683+4reO,072 x 0,016+---- ----
4
2
=0,03171m .

The volume flow rate and entry radius are calculated as follows:
 Stellenbosch University http://scholar.sun.ac.za

B6

. m
q=-
p
3,737
---
995,7
= 0,003753 m3/s

and

=34,15-8
= 26,15 mm.

Substitute the relevant values into Equation (3-9) to obtain Wo:

4027 26,15
, 186
wo= '
1+ 0,0034 0,03171 x 4,027 26,15
2 0,003753 18,6
= 5,298 m/s.

The average velocity through the final orifice is calculated from Equation (3-14):

U = 0,003753
o 7l:0,01862
= 3,453 m/s.

Converting Wo to dimensionless form:

w = Wo
o Ua
5,298
=
3,453
= 1,534

and solving Equation (3-11) with n = 1 we obtain Va = 1,807, or va = 6,24 mls.

Next the dimensionless axial velocity leaving the nozzle is calculated from Equation (3-15):

V2=18072+~15342[(
a' 1 '
1,807
1,807 - 1
)1_1]
... Va = 2,486

and the dimensionless mean axial velocity is obtained from Equation (3-16) as follows:

1,8072 + 2,4862
V=
2
= 2,173.
 Stellenbosch University http://scholar.sun.ac.za

B7

The mean axial velocity in metres per second is:

v=VxUo
= 2,173 x 3,4S3
= 7,SOS m/s.

From Equation (3-17) we obtain the film thickness

t= 0,0186- 001862 _ 0,0037S3

, n 6,24
= 0,0062 m

which when substituted into Equation (3-18) yields the total velocity of the liquid sheet:

U =
a
7 SOS2+S 2982(
, ,
0,0186
0,0186-0,00617
)1
= 9,916 m/s.

Finally the spray cone angle is obtained from Equation (3-19):

28 = 2 cos-1 (7 ,SOS)
9,916
=81,62'.

In order to use the correlations of Som [83S01], Equation (3-20) and Equation (3-26) the
generalised inlet Reynolds number, Regi, must first be calculated from the following equation:
pU e sin y ds
Reg; =---- (B-S)
I.l

Therefore, substituting into Equation (B-S) we obtain:

99S,7 x 4,027 x sine41,S') x 0,0683

Re . = ------------
gl 7,998 X 10-4
= 226892,

which when substituted into Equation (3-20) yields:

0,31S3( )0,1949
28 = _7_7_,2_6_0_3_7_,2_6_8,_3
( / ) n__ [I_ e(-S,69S 10-5)226S92]
X

(68,3/68,3) 0,0611

= 79,74'.
 Stellenbosch University http://scholar.sun.ac.za

B8

BA Calculation of Discharge Coefficient

The experimental testing of nozzle discharge coefficient was done at nozzle supply pressures
other than 34,4 kPa and since Figure 5-6 compares the test data at 28 kPa with the various

other correlations from Chapter 3 it is appropriate to provide sample calculations at this

pressure. The mass flow rate through the nozzle with K = 0,367 at a pressure of 28 kPa is
3,33 kg/so The actual discharge coefficient is calculated from Equation (3-23) as follows:
3,33
... Cn =
2 x 28000
995,7nO,01862
995,7
= 0,4125.

The prediction of Cn from Equation (3-24) has already been calculated on page B3, the value
obtained was 0,2757. The first sample calculation to be done here will be for Equation (3-25).
First the pressure velocity, Up, is calculated:

2 x 28000
Up =
995,7
= 7,46m/s,

which when substituted together with the other relevant values into Equation (3-25) yields:

995 7 x 7 46 x 0 0372) -0,02( 3 ) -0,03(68 3) 0.05 (68 3) 0,23

C = 045 ( ' " 4 -- --'- 0367°,52 -'-
n' 7,998 xl 0- 37,2 68,3 , 37,2
= 0,2567.
From Equation (3-26) we obtain a value for en of:
117
C = 3,377(1)1.520(68,3/68,3r
n (226892 r
347
(37,2/68,3 t
66 (n)0.274

= 0,0694,
while the correlation of Rizk and Lefebvre [85RI1], Equation (3-27), yields the following
solution for discharge coefficient:
25
Cn = 0,35 x 0,3670,5(68,3/37,2r
= 0,2467.

Equation (5-1) which was modelled on Equation (3-27) predicts Cn as follows:

 Stellenbosch University http://scholar.sun.ac.za

B9

Cn = 0,578 x 0,367o.5(68,3/37,2f'25
= 0,4075,
which is within 1,23 percent of the measured value. The coefficient of this equation was
calculated from the following equation:

(B-6)

where n is the total number of data points and Bf is the unknown coefficient. Bf is in effect the
average of all the individual constants which could be calculated for each measured discharge
coefficient, Cni. In calculating Bf, the data from the nozzle with K = 0,632 was excluded for
two reasons. Firstly, looking at the general trend of the data in Figure 5-6 and Figure 5-7 it
appears that discharge coefficient falls away for nozzles with K values in excess of 0,51, while

for K values less than 0,51 the trend is more linear. Secondly, from a design perspective there
is no problem in only correlating Cn for lesser values of K since it is in this range that cooling
tower nozzles are usually designed. Therefore, by substituting the data, obtained at the four
injection pressures, from the remaining six nozzles, into Equation (B-5) the value of Bf = 0,578
was obtained.

The orifice Reynolds number at 28 kPa is calculated based on the pressure velocity:
pUpdo
Red =---
o 11
995,7 x 7,46 x 0,0372
= 7,998 x 10-4
= 345499.
This Reynolds number is used to calculate Cn from Equation (5-2):

Cn = 1,468 x 345499-0,0670,367°,5 (68,3/37,2f'125

= 0,4082,
which is within 1,06 percent of the measured value.

The exponent of Redo in Equation (5-2) was chosen in such a way as to yield a minimum value
for the following dimensionless equation:

(B-7)
 Stellenbosch University http://scholar.sun.ac.za

BI0

where STD refers to standard deviation and AVG is the average. The values of Bcri are
calculated from an equation not unlike Equation (B-5):

B - C_n_i---- (B-8)
cri - Re .-0,067 K.O,5(d jd)0,125 .
dOl 1 S 01

The coefficient in Equation (5-2) is the average of all the Bcri values. Once again only the data
from nozzles with K values less than 0,51 was used in obtaining this equation.

B.5 Film Thickness Calculations

The sample calculations for film thickness which coincide with the formulas of Chapter 3 are as
follows. Firstly we shall note that the measured film thickness is 13,8 mm. From
Equation (3-28), which is a prediction oft using Taylor's [48TAl] theory:
t= 0,0186(1-0,7167)
= 0,0053 m (or 5,3 mm).

From Equation (3-31) we obtain:

t = 00186 - 001862 _ 0,003753

, , 7t 6,24

= 0,0062 m (or 6,2 mm).

The flow number, FN, which must be calculated ill order to use Equation (3-32) and
Equation (3-33) is calculated from Equation (3-34):

FN = 3,33
J995,7 x 28000
= 631 X 10-6 m 2.

The value oft that satisfies Equation (3-32) is t = 0,0052 m, to demonstrate this substitute for t
and calculate both the left hand side, LHS, and right hand side, RHS, of the equation.

LHS = 0,00522
5 2
=27xlO-
, m .

To calculate the RHS the ratio of air-core area to orifice area, X, must first be calculated:
 Stellenbosch University http://scholar.sun.ac.za

B11

X=(d -2t)2
o

da2
(0,0372 - 2 X 0,0052r
=--------
0,03722
= 0,5217.
Substituting into the RHS of Equation (3-32) yields:
6
1560 x 631 x 10- x 7,998 10- X
4
1+ 0,5217
RHS = ----;:::=====----------~
.J995,7 x 28000 x 0,0372 (1- 0,5217)2
= 27, X 10-5 m 2 .

Since the LHS and RHS are equal, t = 0,0052 m is the solution to Equation (3-32).

The film thickness obtained from Equation (3-33) is not much different from the value obtained
above:
~~
t = 3 66
, [ 0,0372 x 631 x 10-6 x 7,998
------;:::=====----
~995, 7 x 28000
X 10-4
]
= 0,0050 m (or 5,0 mm).

Equation (5-3), which was derived from Equation (3-33) gives the following estimate for t:
0,25
0,0372 x 631 x 10-6 x 7,998 X 10-4
t = 9 81 ------;:::=====----
[
, ~995,7 x 28000 ]

= 0,0135 m (or 13,5 mm),

which only differs from the measured value by 2,4 percent. The coefficient of this equation was
calculated from the following equation:

B, ~ i[ 1=1
01
t,
dFN11 lr-"
r,,)n, (B-9)

~pM>n

where Bt is the unknown coefficient and n is the total number of data points. For this
calculation there were only five data points which resulted in a value for Bt of 9,81.
 Stellenbosch University http://scholar.sun.ac.za

B12

B.6 Calculation of Rainfall Intensity

The calculation of rainfall intensity, I, is a basic operation which involves the area of the
sampler cylinder, Ars, and the mass flow rate of water. Take for example a distribution test
which lasted 3 minutes and the mass of water collected in one particular sampler was 2,34 kg.
Since rainfall intensity is traditionally measured in terms of kilograms per square metre per

second, I must be calculated as follows. First the mass flow rate through the sampler, mrs' IS

calculated:

mrs
m
rs
= L1t

= 2,34
3 X 60
= 0,013 kg/s,

from which the intensity for this specific sampler can then be calculated:

0,013
- nO,0452/4
= 8,17 kg/m 2S.

B.7 Calculation of Swirler Pressure Drop in HCN

The swirler pressure drop is calculated from Equation (6-5), however, due to the implicit

nature of this equation, Fsw being itself a function of Msw, iteration is required to solve it. Take
for example a nozzle using the sharp swirler and operating at a supply pressure of 50 kPa.
Characteristic dimensions are: ds = 68,3 mm2, do = 37,2 mm and Ap = 932 mm2, therefore

K = 0,367. The solution is L1Psw = 28,984 kPa. To demonstrate this, calculate Ue and Fsw and
substitute into Equation (6-5). From the definition of Cn:
ill = pUeAp
= pUpAoCn
. UpAoCn
.. Ue=--- (B-10)
Ap

where Cn IS calculated from Equation (5-2). Therefore, assuming p = 1000 kg/m3 and
I.l = 0,001 kg/ms calculate Up and Redo:
 Stellenbosch University http://scholar.sun.ac.za

B13

2 x 50000
D =
p 1000
= 10 m/s

and

1000 x 10 x 0,0372
Re =------
do 0 ,001
= 372000.
Calculating Cn:

Cn = 1,468 x 372000- 0 067

, 0,367°,5 (68,3/37,2t
125

= 0,4062
and then De:
2
10 x nO,0372 x 0,4062
:. U = --------
e 932 X 10-6
= 4,734 m/s.

From Equation (6-4):

F sw' = 272 x 29894 -0,942 + 0556

,

= 1,664,

which when substituted into Equation (6-5) yields:

llPsw = 1/2 x 4,734 (1,664

2
+ 1)
= 29894 Pa,
which is the same value used to initiate this example.
 Stellenbosch University http://scholar.sun.ac.za

APPENDIX C

EXPERIMENTAL DATA
Conditions:
Pn [kPa] ill [kg/s] h [em] do [mm] d [mm] 1 [mm]
79.3 3.93 26.5 37.2 o o

~]I
Iv [em]
0
I
8
I
16
I
24
I
32
I
40
I
48
I
0 0.25 0.50 1.82 5.85 11.19 5.66 0.88
8 0.44 0.82 2.70 6.79 11.51 4.53 0.75
16 1.89 2.77 5.47 9.75 10.19 2.33 0.57
24 6.22 7.48 9.68 11.51 4.84 1.13 0.19
32 11.25 11.57 9.62 4.28 1.32 0.38 --
40 5.03 3.96 2.20 0.94 0.38 0.13 --
48 0.88 0.75 0.50 0.25 0.06 -- --
Table C - 1: RID [kg/m2s} of HCN with sharp-edged S'>virler.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] dj [mm] Ii [mm] ~n [0]
61.5 3.87 26.5 34 37.2 0 0 sharp

0 8 16 24 32 40 48 56
Iv~ reml
0 0.00 0.00 0.08 0.75 2.06 1.43 0.31 0.07
8 0.00 0.02 0.21 0.99 2.27 1.19 0.27 0.06
16 0.17 0.25 0.72 1.74 2.26 0.74 0.18 0.05
24 0.96 1.14 1.77 2.33 1.18 0.34 0.10 0.03
32 2.29 2.35 2.10 1.15 0.44 0.15 0.05 0.02
40 1.18 0.95 0.64 0.33 0.15 0.06 0.02 0.01
48 0.28 0.23 0.17 0.10 0.05 0.02 0.01 0.01
56 0.07 0.06 0.04 0.03 0.02 0.01 -- --
Table C - 2: RID [kg/m2s} of HCN with streamlined S'>virler.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] d [mm] 1 [mm] <Pn [0]
30.7 3.74 29 24.5 37.2 16 0 sharp
0 8 16 24 32 40 48
~]
y [em]
0 9.64 10.53 10.22 8.65 12.21 8.49 0.73
8 10.37 10.53 9.59 8.44 13.20 6.97 0.52
16 8.96 8.96 8.38 8.38 12.63 4.51 0.31
24 7.18 7.44 7.60 8.23 10.16 2.72 0.21
32 11.27 10.37 9.48 8.38 6.71 1.52 0.10
40 6.18 5.13 4.14 3.25 2.10 0.58 0.05
48 0.52 0.47 0.42 0.42 0.31 0.16 --
Table C - 3: RID [kg/m2s} of FCN with sharp-edged orifice.
 Stellenbosch University http://scholar.sun.ac.za
C2

Conditions:
Po [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] d [mm] 1; [mm] ~o [0]
30.7 5.07 27 30.5 37.2 13.5 10 130

0 8 16 24 32 40 48
~]

0 8.75 8.54 7.65 6.97 7.81 8.02 2.93

8 9.43 8.33 7.60 7.60 8.49 7.49 2.67
16 7.65 6.92 6.81 8.28 9.12 5.71 1.47
24 6.34 6.08 7.34 9.75 8.02 2.57 3.67
32 7.44 7.65 9.01 8.02 2.93 0.58 --
40 7.81 7.49 5.45 2.10 0.47 -- --
48 2.83 1.78 0.84 0.21 -- -- --
Table C - 4: RID [kg/m2sj of FCN with protruding central jet.

Conditions:
Po [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] dj [mm] Ii [mm] ~o [0]
24.6 4.76 27 30 37.2 14.5 10 130
0 8 16 24 32 40 48
~I
Iy [em]
0 7.71 8.47 8.34 7.42 7.84 6.71 1.89
8 9.35 8.84 8.09 7.67 8.22 6.04 1.76
16 9.18 7.63 7.55 8.59 8.51 4.15 0.84
24 7.63 7.17 7.92 9.26 6.12 1.38 0.21
32 8.63 8.59 8.89 5.95 1.72 0.25 --
40 7.00 6.04 3.60 1.09 0.25 0.13 --
48 1.72 1.05 0.38 0.08 -- -- --
Table C - 5: RID [kg/m2sj of FCN with protruding central jet.

Conditions:
Po [kPa] ill [kg/s] h [em] Tw [0C] do [mm] dj [mm] 1; [mm] <Do [0]
61.5 6.63 26.5 34 37.2 12 to 17 0 130
0 8 16 24 32 40 48 56 64
I~
Iyreml
0 1.77 1.61 1.59 1.88 2.21 1.74 0.88 0.36 0.15
8 1.50 1.46 1.63 1.92 2.16 1.63 0.82 0.34 0.13
16 1.62 1.61 1.82 2.12 2.09 1.37 0.64 0.27 0.11
24 2.01 1.97 2.09 2.13 1.67 0.97 0.44 0.19 0.09
32 2.19 2.12 2.00 1.67 1.09 0.60 0.29 0.13 0.05
40 1.67 1.51 1.29 0.95 0.58 0.32 0.16 0.08 0.04
48 0.83 0.74 0.60 0.43 0.28 0.16 0.09 0.05 0.02
56 0.37 0.32 0.26 0.19 0.12 0.08 0.04 0.02 0.01
64 0.15 0.13 0.10 0.08 0.06 0.04 0.02 0.02 0.01

Table C - 6: RID [kg/m2sj of FCN with diffusing central jet.

 Stellenbosch University http://scholar.sun.ac.za
C3

Conditions:
II Pn [kPa] I ill [kg/s] h [em] Tw [0C] I do [mm] d [mm] 1 [rum] <Pn [0]
II 27.1 I 4.60 26.5 29 I 37.2 17 0 130

~]II 0 8 16 24 32 40 48 56
Iv [em]
0 5.24 4.85 4.82 5.74 6.76 5.89 2.88 1.15
8 4.93 4.69 5.00 6.03 6.84 5.71 2.67 1.02
16 5.08 5.00 5.55 6.68 6.99 4.82 2.17 0.84
24 5.84 5.84 6.42 7.02 5.95 3.56 1.57 0.60
32 6.55 6.55 6.65 5.92 4.01 2.07 0.89 0.37
40 6.03 5.66 4.74 3.46 1.94 1.07 0.47 0.13
48 3.27 2.93 2.44 1.65 1.00 0.52 0.21 0.08
56 1.34 1.18 0.94 0.65 0.37 0.21 0.08 --
Table C - 7: RID {kglm2sJ of FCN with diffusing central jet.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [0C] do [mm] dj [mm] Ii [mm] <Pn [0]
105.8 8.44 26.5 28.5 37.2 12 to 17 0 130
0 8 16 24 32 40 48 56 64
Iy[em]
~
0 5.99 5.83 6.92 9.47 11.86 10.10 5.49 2.47 1.01
8 5.99 6.04 7.29 10.23 12.20 9.81 5.11 2.22 0.92
16 7.13 7.42 9.05 11.69 12.03 8.30 4.19 1.80 0.75
24 9.89 10.19 11.53 12.32 10.40 6.33 3.10 1.43 0.59
32 12.49 12.28 11.99 10.23 7.17 4.11 2.10 0.96 0.46
40 10.61 9.98 8.55 6.37 4.02 2.31 1.22 0.63 0.29
48 5.95 5.37 4.40 3.31 2.18 1.26 0.71 0.38 0.21
56 2.64 2.43 1.93 1.43 0.96 0.63 0.38 0.25 0.13
64 1.13 0.96 0.92 0.71 0.50 0.34 0.21 0.17 --
2
Table C - 8: RID {kglm sJ of FCN with diffusing central jet.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [0C] do [mm] I d [mm] Ii [mm] <Pn [0]
32.0 4.6 26.5 31 37.2 I 12 to 17 0 sharp
0 8 16 24 32 40 48 56 64
v[em]
~
0 3.56 2.93 2.81 2.91 3.40 3.94 3.77 2.62 1.32
8 3.25 2.72 2.64 2.96 3.35 3.92 3.86 2.58 1.30
16 3.12 2.87 2.93 3.25 3.69 4.00 3.42 2.12 1.03
24 3.31 3.19 3.40 3.77 4.09 3.84 2.87 1.63 0.78
32 3.90 3.81 4.07 4.21 4.05 3.31 2.12 1.11 0.52
40 4.38 4.36 4.30 3.98 3.16 2.16 1.26 0.63 0.27
48 4.17 3.94 3.48 2.83 2.03 1.30 0.75 0.38 0.13
56 2.56 2.35 2.05 1.57 1.09 0.69 0.38 0.19 --
64 1.11 1.05 0.88 0.67 0.44 0.23 0.13 -- --
2
Table C - 9: RID {kglm sJ of FCN with diffusing central jet.
 Stellenbosch University http://scholar.sun.ac.za
C4

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] d [mm] Ij [mm] ~n [0] II
61.5 6.92 26.5 34 37.2 17 0 sharp II
0 8 16 24 32 40 48 56 64
~
0 3.56 2.42 1.83 1.84 1.94 1.39 0.63 0.25 0.08
8 2.53 2.01 1.80 1.81 1.79 1.23 0.54 0.21 0.08
16 2.21 1.96 1.85 1.88 1.62 0.95 0.42 0.16 0.06
24 2.21 2.04 1.94 1.71 1.20 0.65 0.28 0.12 0.05
32 1.98 1.80 1.60 1.20 0.74 0.37 0.17 0.08 0.04
40 1.17 1.07 0.86 0.61 0.35 0.19 0.10 0.05 0.02
48 0.52 0.48 0.38 0.27 0.16 0.09 0.05 0.02 --
56 0.21 0.189 0.14 0.12 0.07 0.04 0.03 0.02 --
64 0.08 0.072 0.06 0.05 0.03 0.02 0.02 -- --
Table C - 10: RID [kg/m2!>} of FCN with enlarged central jet.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] dj [mm] Ij [mm] ~n [0]
92.2 8.44 26.5 28 37.2 17 0 130

0 8 16 24 32 40 48 56
~]
v[em]
0 25.90 20.62 15.03 14.71 15.03 9.37 3.52 1.19
8 17.61 14.71 13.02 13.64 13.71 8.68 3.46 1.26
16 14.21 12.76 12.83 13.46 12.14 6.98 2.77 1.01
24 14.08 12.95 12.89 12.32 9.12 4.78 1.95 0.75
32 13.90 13.27 11.82 9.31 5.78 2.77 1.26 0.50
40 9.37 8.55 7.04 4.78 2.70 1.45 0.69 0.25
48 4.65 4.15 3.27 2.33 1.45 0.82 0.38 0.19
56 1.82 1.57 1.32 0.94 0.63 0.38 0.19 --
Table C - 11: RID [kg/m2s} of FCN with enlarged central jet.

Conditions:
Pn [kPa] ill [kg/s] h [em] Tw [Dc] do [mm] dj [mm] 1j [mm] ~n [0]
65.2 6.59 29 27 37.2 17 0 profiled
0 8 16 24 32 40 48
y~ [em]
0 17.50 15.82 12.31 10.53 13.89 7.55 1.31
8 15.35 14.09 11.37 10.95 12.99 4.77 0.94
16 11.47 10.43 9.54 10.79 12.05 3.93 0.58
24 10.37 9.43 9.59 11.42 9.54 2.52 0.37
32 12.52 11.42 10.32 9.43 4.98 1.26 0.26
40 5.45 5.50 3.88 2.67 1.15 0.37 0.05
48 1.00 1.05 0.73 0.58 0.26 0.10 0.89

Table C - 12: RID [kg/m2s} of commercialzy available FCN

 Stellenbosch University http://scholar.sun.ac.za
C5

Pn [kPa] 55.3 30.7 30.7 55.3 30.7 30.7 30.7 30.7 30.7 46.7
In [kg/s] 5.69 4.23 5.06 6.72 5.3 5.26 4.91 5.56 6.71 8.17
h [em] 40.5 40.5 40.5 40.5 40.5 40.5 40.5 40.5 40.5 40.5
Tw [Dc] 35 35 35 35 35 35 35 35 31 31
do [mm] 32 32 37 37 37 37 37 37 45 45
d; [mm] 14 14 16 16 16 17 16 16 18 18
Ii [mm] 14.5 14.5 0 0 10 10 10 10 0 0
r [em] I [kglm2s]
0 3.88 3.00 2.44 2.14 5.87 1.91 1.72 1.87 3.04 3.67
8 3.53 2.72 3.01 2.64 5.87 2.49 2.13 2.01 3.02 3.49
16 3.44 2.62 4.36 3.64 4.00 3.55 2.96 1.91 2.72 3.16
24 3.72 2.83 5.06 4.16 2.69 3.66 3.27 1.65 3.74 4.22
32 2.45 1.77 3.10 2.75 1.59 2.03 2.00 1.99 3.20 3.77
40 0.79 0.48 0.96 0.94 0.57 0.62 0.61 2.88 1.34 1.80
48 0.11 0.09 0.20 0.22 0.13 0.10 0.12 0.56 0.33 0.50

Table C - 13: RID from FCNs with sharp-edged orifices.

In [kg/s] 5.48 5.66 5.66 5.59 5.68 5.43 5.48 5.66 5.74 5.38
di [mm] 15 15.5 15.5 15.5 16 15 15.5 15 16 15
I [mm] 3 12 13 11.5 11.5 11.5 13 13 13 11
~n [01 sharp 110 130 130 130 130 130 130 130 130
r [em] I [kg/m2s]
0 2.88 1.67 1.69 1.26 1.40 1.45 1.71 1.78 1.89 1.14
8 2.81 1.75 1.72 1.30 1.59 1.48 1.71 1.72 1.89 1.22
16 2.96 1.79 1.70 1.46 1.71 1.52 1.78 1.62 1.94 1.35
24 3.18 1.56 1.54 1.48 1.56 1.49 1.57 1.45 1.69 1.35
32 2.28 1.55 1.35 1.40 1.41 1.39 1.38 1.24 1.41 1.32
40 0.83 2.24 1.31 1.39 1.41 1.38 1.29 1.28 1.28 1.33
48 0.45 1.03 1.13 1.18 1.21 1.21 1.10 0.31 1.02 0.31
56 -- 0.14 0.63 0.65 0.61 0.68 0.64 0.63 0.57 0.64
II 64 -- -- 0.19 0.18 0.14 0.18 0.17 0.15 0.17 0.18

Table C -14: RID from FCNs with: do = 37,2 mm, Pn = 30 kPa, h = 29 cm and Tw = 27 dc.

Pn [kPa] 30.7 30.7 30.7 55.3 88.5 55.3 55.3 55.3 55.3
In [kg/s] 4.2 4.45 3.28 3.86 4.9 5.67 5.82 5.96 5.61
d; [mm] 13.25 14 0 0 0 16 17 18 12 to 17
I [mm] 3 3 0 0 0 0 0 0 0
I r [em] II I [kg/m2s] I
0 1.40 0.99 0.01 0.02 0.02 2.91 3.55 5.02 2.12
8 1.45 1.02 0.04 0.03 0.04 3.27 4.02 5.20 2.55
16 2.26 1.04 0.23 0.26 0.28 4.13 4.84 5.63 3.78
24 3.60 1.04 2.37 2.53 2.91 4.14 3.85 3.76 4.22
32 1.95 1.03 2.81 3.36 4.37 1.99 1.70 1.44 2.14
40 0.42 1.06 0.48 0.63 0.89 0.50 0.46 0.37 0.57
48 0.04 0.82 0.05 0.07 0.09 0.10 0.10 0.07 0.31
56 -- 0.39 -- -- -- 0.01 0.01 0.01 0.02

Table C - 15: RID from FCNs and HCNs with: sharp-edged orifices, do = 32 mm, h = 40,5 cm and Tw = 28°C.
 Stellenbosch University http://scholar.sun.ac.za
C6

Pn [kPa] 39.3 55.3 88.5 30.7 55.3 30.7 55.3 30.7 55.3 30.7
ill [kg/s] 4.01 4.69 5.94 3.49 4.56 3.49 4.56 3.49 4.56 3.49
~n [0] sharp sharp sharp 90 90 110 110 130 130 sharp
I r [em] II I [kg/m2s] I
0 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.01
8 0.03 0.03 0.03 0.01 0.02 0.02 0.02 0.02 0.02 0.03
16 0.17 0.20 0.21 0.13 0.19 0.15 0.14 0.14 0.15 0.25
24 1.21 1.29 1.43 0.59 0.88 0.58 0.77 0.59 0.79 4.45
32 3.55 3.85 4.52 2.16 2.26 1.33 1.48 1.20 1.54 1.97
40 1.41 1.87 2.60 2.19 3.53 2.40 2.24 1.69 1.90 0.21
48 0.19 0.29 0.45 0.04 0.15 0.52 0.99 0.88 1.30 0.01
56 -- 0.04 0.06 -- -- 0.01 0.07 0.19 0.33

Table C -16: RID from HCNs with: do = 37,2 mm, h = 38 cm and Tw = 29°C.

Pn [kPa] 24.6 39.3 59 81.1 108.2 92.2 61.5 30.7

ill [kg/s] 4.17 5.3 6.45 7.55 8.69 8.05 6.22 4.67
h [em] 43 43 43 43 43 34 34 34
Tw [0C] 35 35 33 27 27 29 29 29
I r [em] I [kg/m2s] I
0 1.49 1.80 2.01 2.37 2.72 3.99 3.24 2.58
8 " 1.39 1.69 1.91 2.25 2.59 3.71 2.95 2.35
16 1.23 1.51 1.71 1.97 2.33 3.07 2.46 1.92
24 1.00 1.24 1.44 1.66 1.92 2.52 1.99 1.57
32 0.93 1.10 1.29 1.42 1.71 2.70 2.19 1.95
40 1.34 1.34 1.48 1.60 1.90 3.01 2.54 1.96
48 1.31 1.57 1.71 1.88 2.13 1.43 0.98 0.47
56 0.38 0.77 1.11 1.32 1.51 0.32 0.20 0.08
64 0.06 0.18 0.33 0.44 0.55 0.05 0.03 0.01

Table C - 17: RID from commercial FCN.

Pn K [-]
[kPa] 0.299 0.323 0.367 0.421 0.457 0.509 0.632 0.719
12 0.364 0.391 0.425 0.467 0.496 0.525 0.619 0.631
28 0.351 0.372 0.412 0.453 0.473 0.519 0.592 0.609
49 0.347 0.368 0.407 0.444 0.470 0.506 0.577 0.589
77 0.345 0.364 0.402 0.440 0.467 0.502 0.562 0.590

Table C - 18: A1easured discharge coefficients, C, for HCNs with streamlined swirler and Tw = 25,5 0C.

Pn d [mm]
I [kPa] I 12 to 17 16 17 18
11.6 0.939 1.081 0.989 1.283
25.2 1.302 1.551 1.424 1.779
44.3 1.669 2.004 1.831 2.310
60.3 1.752 2.371 2.119 2.687
92.2 1.920 2.906 2.634 3.256
113.1 2.073 3.411 3.048 3.838

Table C - 19: JvJeasured mass flow rates, ill {kg/s}, through central jets inserted into streamlined swirler.
 Stellenbosch University http://scholar.sun.ac.za
C7

ill K [-]
I [kg/s] I 0.299 0.367 0.457
2.8 80 85 90
3.8 82 86 94
5.3 84 87 96
6.9 86 87 93

Table C - 20: Measured spray cone angle, 28 [7

Pn ill [kg/s] I
I [kPa]
II
I streamlined I sharp I
3.7 1.24 1.12
12.6 1.84 1.70
24.1 2.45 2.15
43.4 3.28 2.78
56.6 3.78 3.13
78.9 4.40 3.59
99.6 4.97 3.99

Table C - 21: Ai/easured mass flow rates for HCN with different swirlers but K = 0,367.

11j [mm] II 15 I 30 I 45 I
c;J1[mm]
32.4 51.0
3.8 4.9
51.0
6.2
50.4
3.8
86.7
ill [kg/s]
4.9
86.1
6.2
86.1
3.8
140.2
4.9
137.7
6.2
140.2
I

37.2 35.1 33.2 38.7 60.3 56.6 65.2 97.8 92.8 107.6
45.5 21.3 22.1 20.3 36.3 35.7 35.1 59.0 58.4 57.2

Table C - 22: Ai/easured injection pressure, Pn [kPaJ, for HCNs with protruding solid core.

Pn [kPa] 121.8 72.6 36.3 18.4

(ro - r) [mm] Yzpvoz [kPa]
0 0 0 0 0
0.60 49.91 31.72 15.98 8.11
1.48 35.78 21.02 9.84 5.04
2.36 26.92 16.11 7.13 3.44
3.24 26.68 15.61 7.13 3.44
4.56 29.51 17.21 7.99 3.69
5.88 31.60 18.44 8.61 4.06
6.76 29.75 17.33 8.24 3.93
7.64 24.83 14.63 7.01 4.06
8.52 18.56 11.19 5.29 2.70
9.40 12.54 7.38 3.57 1.84
10.28 6.64 4.67 2.09 1.11
11.16 3.57 2.09 0.98 0.49
12.04 1.60 1.11 0.49 0.25
12.92 0.61 0.37 0.12 0.00
13.80 0.25 0.09 0.00 --
14.50 0.06 0.00 -- --
15.21 0.00 -- -- --
Table C - 23: Measured dynamic pressure distribution in outlet plane of HCN with do = 37,2 mm.
 Stellenbosch University http://scholar.sun.ac.za
C8

do [mm]
42.3 45.5 37.2 I 32.4 I 21.6
(ro - r) lhpv02 (ro - r) lhpv02 (ro - r) Y2PV02 (ro - r) lhpv02 (ro - r) lhpv02=
[mm] [kPa] [mm] [kPa] [mm] [kPa] [mm] [kPa] [mm] [kPal
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0.18 1.23 0.7 1.97 0.60 8.11 0.44 14.76 0.09 13.16
0.35 11.31 1.14 1.35 1.48 5.04 0.88 18.69 0.26 24.23
0.79 16.36 1.58 1.41 2.36 3.44 1.32 16.23 0.70 23.61
1.23 13.65 2.46 1.11 3.24 3.44 1.76 14.14 1.14 21.15
2.11 9.96 3.34 1.35 4.56 3.69 2.20 12.91 1.58 20.29
2.99 8.61 4.22 1.78 5.88 4.06 2.64 12.42 2.02 19.43
3.87 8.73 5.1 2.83 6.76 3.93 3.08 12.05 2.46 19.43
4.75 8.85 5.98 3.57 7.64 4.06 3.52 11.62 2.90 19.06
5.63 9.22 6.86 4.80 8.52 2.70 4.40 11.56 3.34 18.57
6.51 9.35 7.74 5.17 9.40 1.84 5.28 11.19 3.78 18.08
7.39 9.41 9.06 5.29 10.28 1.11 6.16 10.70 4.22 17.46
8.27 8.92 10.38 3.94 11.16 0.49 7.04 9.47 4.66 17.83
9.15 8.24 11.26 2.89 12.04 0.25 7.92 8.24 5.10 17.46
10.03 6.89 12.14 1.78 12.92 0.00 8.80 6.15 5.98 15.86
10.91 5.29 13.02 1.05 -- -- 9.68 4.49 6.42 11.19
11.79 3.57 13.9 0.68 -- -- 10.56 2.34 6.86 8.61
12.67 1.84 14.34 0.49 -- -- 11.00 1.84 7.30 6.03
13.55 0.98 14.78 0.31 -- -- 11.88 1.05 7.74 1.48
14.43 0.37 15.22 0.12 -- -- 12.76 0.43 8.18 0.43
15.31 0.12 15.66 0.06 -- -- 13.64 0.12 8.62 0.06
16.19 0.00 16.1 0.00 -- -- 14.52 0.00 9.06 0.00

Table C - 24: lvfeasured dynamic pressure distribution in outlet plane o/various HeNs with Pn = 30,7 kPa.
 Stellenbosch University http://scholar.sun.ac.za

APPENDIX D

DESIGNING A FeN

This appendix describes a step by step method for designing specific cooling tower nozzles.

The correlations used in this procedure are the same correlations that are given elsewhere in

this thesis, except that they have been rewritten to expose the hidden nozzle dimensions.

As demonstration a numerical example is given to

show how this design code is implemented.

Assume that the following constraints apply:

• water temperature = 45°C

• required intensity, I = 4 kg/m2s

• spray area, b = 1,2 m

• nozzle height, h = 0,35 m

• ~Pn = 35 kPa.
Figure D - 1 : Nozzle arrangement.

Since discharge coefficient is directly proportional to Ap0,5 and inversely proportional to ds,0,25

it was decided to use the relationship between Ap and ds from the prototype nozzle as a basis

for designing nozzles of other sizes. The following equation can be used to obtain ds from Ap

or Vice versa:

d s -- 3 , 6 x A p 0,57 . (D-l)

The steps thatfollollJ are the recommended means of designing a FeN.

1 Calculate the required volume flow rate through the nozzle that will give the desired rainfall

intensity for the area to be wetted.

A=bxb

= 1,2 x 1,2

= 1,44 m2

m=IxA

= 4 x 1,44

= 5,76 kg/so
 Stellenbosch University http://scholar.sun.ac.za

D2

From Equation (A-I) p = 990,4 kg/m3

:. q = m/p
= 5,76/990,4
= 5,82 x 10-3 m3/s

2 Now calculate the major nozzle dimensions, Ap, ds and do, that will satisfy the
recommendation that K be less than 0,5. The process of selecting these dimensions may require
iteration because the swirler flow coefficient, Esw, is unknown. The swirler flow coefficient is
defined as the percentage of the total mass flow rate passing through the swirl ports. Assume a
value for either Ap or ds and calculate the other from Equation (D-l), then solve
Equation (D-2) to obtain a value for do. Alternatively values for Ap and do can be read directly
off Figure D-2, which is a series of solutions of Equation (D-2) for which the nozzle
parameter, K, has a value of 0,4. These curves can also be interpolated to find solutions for
intermediate rainfall intensities or spray areas. These nozzle dimensions were calculated to

yield a K of 0,4, however, this may not always be the most suitable choice. The maximum
recommended K value for any specific injection pressure should be read off Figure 6-1.

60 3000

50 2500

40 2000
r--I

'6 N

8
S 30 1500 S
<-
20 1000

10 500

o o
o 1 2 3 4 5 6 7 8

d I [kg/m2s] A
---e- 1,44 m2 ---e- 1,0 m2 ---e- 0,64 m21-fr-- 1,44 m2 -fr-- 1,0 m2 -fr-- 0,64 m2

Figure D - 2: Chart for selecting major nozzle dimensions.

 Stellenbosch University http://scholar.sun.ac.za

D3

Equation (D-2) is the closed form solution to the combination of Equation (3-23) and
Equation (5-1). Equating these two equations and inserting the unknown swirler flow
coefficient, Esw, the following equation is derived:

eliminating Cn to reveal
0.8
1558E rhd 0.25
d-' sw S
(D-2)
o - [ ~pflPnAp
~

Assuming a value for Esw of 0,82, which is in line with the recommendation of Prywer and
Kulesza [87PR1], and a swirl port area, Ap, of 1300 mm2 the major nozzle dimensions are
calculated as follows:
0 57
ds = 3,6 x ( 1300 x 10-6 ) •

= 0,0815m
and
~8
d = 1,558 x 0,82 x 5,76 x 0,0815°,25
o [ ~990,4 x 35000 x 1300 x 10-6 ]

= 0,0411m,
which yields a K value of 0,388. The anticipated nozzle discharge coefficient cab be calculated
from Equation (5-1):
0 25
Cn = 0,578 X 0,388°,5( 0,0815/0,0411 ) '
= 0,427.

Alternatively Figure 6-1 can be used to estimate a maximum permissible K for the given flow
conditions. In this instance the nozzle supply pressure, Pn, of 35 kPa corresponds to a K value
of approximately 0,47. This is correct under the assumption that streamlined swirl ports will be

used and a minimum swirler pressure drop, flPsw,of 10 kPa will be observed. A smaller K may
always be used, so long as the diameter ratio, ds/do remains greater than 2, In order to use
Figure D-2 K must be taken as 0,4, which together with the desired rainfall intensity of
 Stellenbosch University http://scholar.sun.ac.za

D4

4 kg/m2s and a wetted area of 1,44 m2 leads to the following approximate solution for the three
unknowns:

do ~ 40,5 mm,

Ap ~ 1300 mm2

and using the relationship between K, do, ds and Ap:

ds = Ap/doK
= 1300/(40,5 x 0,4)
= 80,2mm.

There is no difference between the equations used here and those used previously. The small
discrepancy in the calculated dimensions is as a result of having read from the graphical
solution.

3 Calculate the central jet diameter from Equation (D-3), which is simply Equation (5-4) that
has been modified by adding an additional 1,2 mm. This correction compensates for the fact
that Equation (5-4) is a correlation for nozzles with sharp-edged orifices, but nozzles with
profiled exits are requiring to be designed. Therefore:
d. = 0 0193XO,lJ7+ 00012 (D-3)
J' "

where X is calculated from the following equation derived from Equation (5-3):

d . ]0'25}! J2
X= [{ do-19,62 [ ~~~ Ida (D-4)

Substitute into Equation (D-4):

0,0411XO,82X5,82XlO-3 X5,91XlO-4]O'25}) J2
X= [{ 0,0411-19,62 [ . 0,0411
35000

= 0,127

and then into Equation (D-3) to obtain dj:

dj = 0,0193 X0,127°,117+0,0012
= 0,0164 m.

4 Now check that the total flow rate through the swirler and central jet agrees with the
required flow rate. Calculate the jet flow rate from the swirl port flow rate using
Equation (5-5). Substituting into this equation yields:
 Stellenbosch University http://scholar.sun.ac.za

D5

rc/4 x 0 0 164 2
.. = 0989 ' 0 82 .
qJ ' 1300 x 10-6 , q
= 0,132q.

The total flow rate through this nozzle is 95,2 percent of the required flow rate, which is
suitably accurate for the purpose of nozzle design. However, had the total flow rate at this
point been 15 percent out, then it would have been necessary to return to step 2 to re-estimate
the swider flow coefficient and then recalculate the major nozzle dimensions. The solution
obtained above is accurate enough and no further iteration shall be performed here.

5 The final step in this design process is to outline the

profiled outlet needed to produce a square spray pattern.
The geometry of the following profile is calculated with the
intent that the profile be manufactured on a 3-axis NC
machine. It is recommended that the outlet be machined
with circular cutter paths that vary in height to produce the
profile, see Figure 4- 7.

• The first thing to note about the profile is that it will be

repetitive through 45°, therefore, it is only necessary to

determine cutter positions for the first 45° and for as

many number of circular cuts necessary to give the
Figure D - 3: Geometry o/profiled
desired finish.
outlet.
• Calculate the required mInImUm and maXImum spray
angles for the nozzle, these are 80 and 845 in Figure D-l.

• The equation for the required spray angle, i.e. the angle at which the spray should exit the
nozzle in order to wet the full spray area, at an arbitrary value of exis:

8i = tan -l( ---- b

2 h coscx
)
j
(D-5)

• The half nozzle angle, <Pn, is calculated from the following equation:

l/,j., _ ( 8 1 - 80) 12' 8 (D-6)

/2'1'n; - 8 -8 x +;,
45 0
 Stellenbosch University http://scholar.sun.ac.za

D6

where the extra 12° is used to exaggerate the Coanda effect, which would otherwise have

little effect on producing a square spray pattern. Alternatively this extra 120 could be viewed
as a means of deepening the grooves in the exit profile.

• The outer radius of the profiled outlet, R, has a value 1,6 times the radius of the orifice, ro.
This is in keeping with the prototype designs tested throughout this project, however,
values in excess of this may also prove useful.

• The radius of the ball-nosed cutter used to machine the outlet is rh. The more pronounced a
profile, the smaller will have to be the cutter in order to machine the sharper curves.

• The width of the flat section of the profile is designated as If in Figure D-3.

• The radius of curvature, r8, to which the flat section is tangent, is calculated from the
following equation:

(D-7)

• Finally, the cutter height, Zh,can now be calculated from the following equations:

(D-8)

and for rei> R -If:

(D-9)

There is one last alteration, which must be done to the co-

ordinates calculated with the above equations before they
can be included in the NC codes for machining the profile,

and that is to translate these values to between 30° and

40° in the direction in which the liquid will be spinning

when it exits the nozzle. This translation is vital to the
success of the nozzle because of the fact that water
leaving the nozzle does not exit in a purely radial
direction, which is due to the significant tangential Figure D- 4: Translation of cutter

component of velocity. Without translating the co- co-ordinates.

ordinates for the groove, the spray is likely to separate from the profiled exit causmg
instabilities and unpredictable spray cone angles. This translation is illustrated in Figure D-4.
 Stellenbosch University http://scholar.sun.ac.za

D7

The final consideration, which specifically concerns nozzle assembly in a cooling tower, is the
orientation of the profiled exit relative to the desired orientation of the square spray pattern.
This particular aspect was overlooked in this project even though it is certain that this global
would affect the spray orientation. However, the single profiled exit that was tested, was found
to spray squarely onto the projected spray area when the line drawn between opposing groove

exits was orientated at an angle of 12° relative to the axis of the supply pipe. Figure D-5,

shows the orientation of the tested profile against the supply pipe. The 12° angle discussed
here may be used as a guide, however, it is recommended that individual outlets be tested
experimentally to determine their required orientation.

GROOVES

SUPPLY PIPE

Figure D - 5: Outlet orientation.