Вы находитесь на странице: 1из 148

A STUDY OF THE DISCHARGE COEFFICIENT OF JETS FROM ANGLED SLOTSAND CONICAL ORIFICES

Kevin William Linfield

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Aerospace Science and Engineering
University of Toronto

Copyright @ ) 2000 by Kevin William Linfield

The author has granted a nonexclusive licence aowing the National Library of Canada to reproduce, loan, distniute or se copies of this thesis in microfonn, paper or electronic formats.

L'auteur a accord une licence non exclusive permettant a la Bibliothque nationale du Canada de reproduire, prter, distribuer ou v&e des copies & cette thse sous la fome de microfiche/^ de reproduction sur papier ou sur format lectronique.

The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fiom it may be printed or otherwise reproduced without the author's permission.

L'auteur conserve la propribt du droit d'auteur qui protge cette thse. Ni la thse ni des extraits substantiels de celle-ci ne doivent tre imprims ou autrement reproduits sans son
autorisation.

Abstract
A Study of the Discharge Coefficient of Jets fiom Angled Slots and Conical Orifices Kevin William Linfield
Doctor of Philosophy Graduate Department of Aerospace Science and Engineering University of Toronto

2000
Non-ideal flows through angled orifices occur often in engineering applications such as m a s flow meters, blast-wave simulators, gas propulsion devices and projectile launchers.

s required to For such flows, a correction factor in the form of a discharge coefficient i
facilitate good design and efficient operation. The discharge coefficient depends on many parameters including atmosphenc to stagnation pressure ratio, gas specific heat ratio, exit to charme1 area ratio, orifice shape, wall angle and onfice edge rounding. The dependence on these parameters (except edge rounding) was studied numerically by developing a two-dimensional finite-difference computer program that solves the subsonic Bowfield in the hodograph plane by a relaxation method and the attached supeaonic Bowfield, if present, in the physical plane by a method of characteristics, joined together at the sonic surface by matching stream Function values. The discharge coefficient was also studied experimentally by developing and testing an experimental facility using a new technique based on the partial blowdown of a pressurized vesse1 through a short pipe ending with conical orifices of different axea ratios, wall angles and onfice edge roundings. Numerical and experimental data from these studies and also fiom the Iiterature are compaxed and discussed. A s o h a z e package called the 'Cd Algonthmn was developed to reproduce quickly the combined numericd and experimental discharge coefficients for the entire set of parameters, and it outperforms previous algorithms in accuracy, efficiency

and comprehensiveness. This CaMgonthm uses analytical, numerical and experimental results at the limits for incompressible, cnticai and choked flows, and other Cd values between these limits are reproduced by using piecewise cubic polynomial splines.

Acknowledgements
1wouid like to thank the many people at the University of Toronto Institute for Aerospace Studies who provided assistance, advice, or simply an open ear which enabled me to
complete my thesis. There are too many people to List, but my sincerest thanks go out to them ail.
My professor, Dr. James Gottlieb, provided me with a great deal of valuable advice

on aiI aspects of my thesis. 1thank him for taking the tirne to help teach me how to learn.

H i s willingness to stay late and see me on weekends helped make it t hrough difncult times.
My other D..4.C. cornmittee members, Dr. Phi1 Sullivan and Dr. David Zingg, helped

keep me on track, made sure my content was relevant and provided valuable comments

and discussion. Without the encouragement of my family, 1 never would have started, and without their never ending love and support, 1never would have finished. Thank you. Most of all, 1 thank my wife, Doreen, who has lived with my thesis for far too long. She has given me more support than she will ever realize.

Contents
Abstract
Acknowledgements
Notation
1 Introduction
1.1 Motivation
1.2 1.3

ii
iii
vii
1
1

................................... The Discharge Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . Historical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 Theoretical Determination of the Discharge Coefficient . . . . . .
..... .....

2
3

3
5

1.3.2 Experimentd Determination of the Discharge Coefficient 1.4 Purpose and Objectives . . . . . . . . . . . . . . . . . . . . . .
1.5

6
6 6

Flow Types and Definitions

1.6

......................... 1.5.1 Incompressible and Compressible Flow . . . . . . . . . . . . . . . 1.5.2 Subcritical, Critical, Supercritical. and Choked Flow . . . . . . . Introduction to O r i f i c e Flows . . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 Description of the Flow in the Physical Plane . . . . . . . . . . . 1.6.2 Description of the Flow in the Hodogaph Plane . . . . . . . . . .

7
9
9

10
15

2 Theory for and Testing of CFD Code

2.1 Introduction

2.2

.................................. Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


2.3.1 2.3.2 Derivation of the SubcriticaI Boundary Conditions . . . . . . . . .
2.3.3

15

15
17 17

2.3 Solution Method for Subcritical and Critical Flow

............. Non-dimensionalization of the Stream F'unction Equation . . . . .


Solution to the Subsonic Boundary-Value Problem . . . . . . . . .

19

21

2.3.4 Mapping the Hodograph Solution to the Physicd Plane 2.3.5

......

23
24

A Subcritical Flowfield Solution . . . . . . . . . . . . . . . . . . .

2.4 Supercritical and Choked Flow Regime


2.4.1 Method of Characteristics
2.4.2

...................

25
25
26 27 29

...................... Supersonic Boundaxy Conditions . . . . . . . . . . . . . . . . . .

SolutionoftheInitial-ValueProblem . . . . . . . . . . . . . . . . 2.5 Matching Subcritical and Supercritical Regimes . . . . . . . . . . . . . .


2.4.3

2.6 Detemination of the Discharge Coefficient . . . . . . . . . . . . . . . . . 2.7 Accuracy of the Numericd Solution . . . . . . . . . . . . . . . . . . . . .
2.7.1 Cornparison of Results

31

32

........................

33
46

Experimental Apparatus and Method 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Orifice Flow Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 . 4 pparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Data Acquisition and Reduction . . . . . . . . . . . . . . . . . . . 3.3 Discharge Coefficient Calcuiation and its Error . . . . . . . . . . . . . . . 3.4 Calibration of the Experimentd Apparatus . . . . . . . . . . . . . . . . .
4

46

47

47 50
52

53 56

CdAlgorithm
4.1 Introduction

71
71

4.2

.................................. Generation of the Standard Graph . . . . . . . . . . . . . . . . . . . . .


4.2.1 4.2.2

72
73

Cdl,, for the Standard Graph

...................

3 1

,,

fbr the Standard Graph

..................
.................

74

4.2.4

5 1

for the Standard Graph . . . . . . . . . 78 & v==P r d 4.3 Equivaient CdDistribution Principle . . . . . . . . . . . . . . . . . . . . 78 4.4 Obtaining Cd(-. for Nonstandard Conditions . . . . . . . . . . . . . . . 79
4.2.6

-1

. ,

for the Standard Graph

77

..

and

-1ar

acd

4.4.1
4.4.2

cd(%= for a Non-2ero Area Ratio . . . . . . . . . . . . . . . . .


cd(z1 for a Zero Area Ratio . . . . . . . . . . . . . . . . . . . . .

80
81

4.4.3 4.4.4
4.4.5

C#~Z, for a Non-Zero Area Ratio . . . . . . . . . . . . . . . . . .


for Conicd Orifices with Edge Rounding

..........

82 83
84

Cd(,

foi Experimentd Area

Ratio Effect

.............

5 Results and Discussion

5 . 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 CdAlgolithm Matchhg Theoreticai Results . . . . . . . . . . . . . . . . 5.2.1 Wall Angle Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 AreaRatioEffect . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Specific Heat Ratio Effect . . . . . . . . . . . . . . . . . . . . . . 5 . 2 . 4 Planar and Axisymmetric Flow Effect . . . . . . . . . . . . . . . . 5.2.5 Notes on the Approximations in the CdAlgonthm . . . . . . . . . 5.3 Cd Algorithm Compared to Other Algorithm . . . . . . . . . . . . . . . 5.4 CdAlgorithm Matching Experimental Results . . . . . . . . . . . . . . . 5 . 4 . 1 Edge Rounding Effects . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Experimental Area Ratio Effect . . . . . . . . . . . . . . . . . . . 5.5 Notes on the Experirnental Discharge Coefficient . . . . . . . . . . . . . . 5.5.1 Lip Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2 Viscous Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6 Notes on the Equivalent Cd Distribution Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6 Conclusions

100 . 100
101 101

101 102 102 102 103 105


105

106 107 107 108


109

6.1 Contributions

.................................

119 120

A Derivation of Governing Flow Equations A.1 Subsonic Planar Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2 Subsonic Axisymmetric Flow . . . . . . . . . . . . . . . . . . . . . . . . . A.3 Supersonic Planar Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . A.4 Supersonic Axisymmetric Flow . . . . . . . . . . . . . . . . . . . . . . .

127
127 129 132 135

Notation
Quantities
sound speed sound speed at M = 1 area
area ratio of orifice
A,t -

&hanne1

gas specific heat at constant pressure

gas specific heat at constant volume

contraction coefficient discharge coefficient velocity correction factor negative characteristic constant positive characteristic constant orifice exit diameter Doppler frequency residud at node i, j unit distance mass fiow rate Mach number nondimensional flow velocity a*

grid size

pressure pressure ratio Patm Po

radial physical coordinate (a;xidy symmetric) orifice Lip radius of curvature gas constant

Reynolds number distance from physicd origin temperature component of velocity cornponent of velocity

f l o w velocity
reservoir volume physical coordinate (planar) physical coordinate (planar ) physical coordinate ( a w i a l l ysymmetric)

wall angle equivalent angle normalized wall angle planar or axisymrnetric parameter convergence tolerance Cp gas specific heat ratio cil

laser beam angle laser light wavelength constant (3.1415926 ...) flow angle

stream function
normalized stream function velocity potential

standard deviation density Mach angle (method of characteristics) dynamic viscosity characteristic variable acceleration fator

Vector Quantities-

directionai area

normal direction velocity

Subscripts
atmospheric condition actual value orifice exit condition row/column iodicator ideal d u e quantity at physical coordinate limit jet location condition at lip rnexhum quantity stagnation quantity
quantity dong the sonic line

test section wall quantity one dimensional three dimensionai critical condition choked condition derivative with respect to V derivative with respect to O second derivative with respect to V second derivative with respect to 6 derivative with respect to Me second derivative with respect to M ,

Overheads non-dimensional variable


rate

Chapter 1
Introduction
1.1

Motivation

The measurement of m a s flow rates of liquids and gases through angled rectangular slots and conical orifices shown in figure 1.1 often differ appreciably fkom those predicted by using simple theories for incompressible and compressible fiows. To facilitate the convenient use of simple theory to easily and accurately predict actual m a s f l o w rates, a corrective factor has historically been incorporated in the simple theory. The discharge coefficient

is the required factor. The actual mass flow rate is denoted by to illustrate that its value depends inherently on multidimensional flow features and can be measured experimentally or predicted by advanced analytical and numericd methods. The ideal mass flow rate is the theoretical value obt ained by assuming one-dimensioual inviscid Bow of a fluid through the orifice with the flow at the throat uniform in velocity and parallel to the axis of symrnetry. When the discharge coefficient is avaiiable kom tables, graphs or an algorithm, then k , , = Cd&,. Although some experirnentd and theoretical results exist in the literature, this readily available data does not provide sufncient detail over the full range of ail the geometrical and flow parameters. To remedy this shortcoming, a two-dimensional computational fluid dynamics cornputer program was written to determine the discharge coefficient for planar flow through a dot and axisymmetric flow through a c o n i d orifice as a function of the wali angle, area ratio, pressure ratio and gas specific heat ratio. This code extends the work of others and produces more accurate

results than previously available in the literature. Details of this code are presented in chapter 2. Additionally, a new experimental method based on the partial depressurization of a high-pressure reservoir has been developed and experimental results have been obtained to provide additional data and is detailed in chapter 3. The new experiments

a l l angle, area ratio and orifice edge rounding. The numerical studied the effect of w
and experimental results have been incorporated into a computer procedure which will reproduce the data completely and quickly. This procedure is detailed in chapter 4.
There are a number of practical applications of the discharge coefficient. Tables of

ASME or BS data are used with 90' orifice plates for flow meter applications and the
production, distribution and fiscal metering of gas and light hydrocarbon liquids. In the aerospace industry, turbojet aircraft and rockets use a converging conical orifice or nozzle in their propulsion systems where the discharge coefficient and the coefficient of t h s t are important for obtaining efficient performance at different throttle settings. Another application is active blast wave reflected wave eliminators (RWE) which use rotating louvers at the end of the duct or channel [Guice k Gottiieb, 19871. The rotating louvers produce different wall angles (including angles greater than 90") and changing area ratios which control the amount of physical louver blockage at the end of the channel to minimize the reflected blast wave. In addition, laboratory high-low pressure guns use a perforated plate separating a high pressure chamber from the projectile for delivering

low projectile base pressures for low muzzle velocities with good repeatability, to convert a high pressure gun into a low pressure one [Maillette, 19891.

1 . 2

The Discharge Coefficient

The discharge coefficient depends on a number of factors which are a part of the orifice

n the literature, geometry and the flow conditions. Based on this work and that found i
the most important parameters in determining the discharge coefficient are: wali angle
a,exit-techanne1 area ratio 4, orifice shape X, edge radius of curvatureto-diameter

ratio - gas specific heat ratio y and atmospheric-to-stagnation pressure ratio T . This d' c m be silmmarized as

Other parameters mentioned in the literature include the Iip thickness-to-diameter ratio, gravity (via the Ftoude number), surface tension (through the Weber number) and

Reynolds number. The effects of gravity and the interaction of the jet and the atmosphere are neglected in the derivation of the governing flow equations and it i s assumed

s minimal [Shapiro, 19531. The orifice thickness-to-diameter ratio has that their effect i

[1979],and this eEect is not part of this analysis because the orifice was msumed to have either a sharp or rounded lip. The contraction coefficient has aIso been used in the literature to indicate the deviation between actual and ided flow rates. It is defined by
been detailed in Ward-Smith

where Ajet is the cross sectional minimum fiee jet area (&O known for subsonic flow as the vena contracta, defined in section 1.6.1) and .dkt is the cross sectional exit axea of the orifice. The contraction coefficient, valid for inviscid incompressible and subcritical flow,

has often been used synonymously with the discharge coefficient, but the contraction definition loses meaoing when dealing with supercritical flow. Since there is no find minimum-axea contracted jet for supercriticai flow, a definition involving mass f l o w rates is required.

1 . 3 Historical Background
The contraction property of a jet emanating from an orifice has been known since ancient times where its application was used to meter water fiom aqueducts. In the late 1800s, experirnents and numerical analysis were performed for incompressible f l o w s where it was determined that only about 60% of the expected fluid was discharged. -4 brief summary of some of the important historical background is presented in this section.

1 . 3 . 1

Theoretical Determination of the Discharge Coefficient

The initial t heoretical studies on the discharge coefficient were investigated numericdy

based on complex potentiai theory for planar, incompressible, inviscid flow and are sumrnarized by Gwevich [1965] and Milne-Thomson [1968].Using this method, an expression for the discharge coefficient for planar flow through a slot in an infinite reservoir can be derived: 1 Cdtr=~ = 1 1 ; cot sin ( m u ) d o '

[ (9)

This equation i s readily integated for angles expressed as a radian fraction (eg. t ,) : and for a = $ yields the classic result Cdl,, = 5, and for a Borda mouthpiece with a = r yields Cdl,, = f. A number of more recent authors have used this technique a r t i n to determine the discharge coefficient, including Hunt [1968], Larock [1969]and M
[1977]. Chaplygin [19021 solved the subcritical compressible flowfleld of a 90' slot horn an infinite resenroir using the hodograph transformation and expressed the solution as an infinite series of hypergeometric functions. His work was extended by a number of other authors, most not ably Falkovich [1957]who published detaed analysis involving the entire range of wall angles O < ai < T and area ratios O 5 A, 5 1 for subcritical and critical flows. (The work of Pickett [1989] to produce a practical routine for obtaining discharge coefficients by Falkovich's solution method has been used to produce numerical results throughout this thesis.) The first numericd cdculation on the discharge coefficient who extended Chaplygin's met hod for at the choked pressure ratio was by Frank1 [1947], the case of maximum discharge and obtained the result Cd= 0.85 for a 90" slot f r o m an infinite reservoir. Frnkl's method was extended by Gushchin (19591 to include the effect Due to the complexity of the axially symmetric equations, little has been published on uisymmetric f l o w through a conical orifice in cornparison to planar Bow through a slot. The method of Chaplygin and complex potential theory are not applicable to axially symmetnc flows. Early analysts calculated that the discharge coefficient from an orifice was the sarne as that for a similu dot (including authors such as Rouse & Abd-Fetouh

[1950]who used the electrical-analogy technique), but it was Garabedian [1956] who demonstrated that the discharge coefficient for axisymmetric flow was less than that for planar flow. With the introduction of the digital cornputer, it became practical to convert the partial differentid equations for planar or axsymmetric orifice flow into finite ciifference or finite volume representation and solve these using "modern" CFD methods, For example, the subcntical planar flowfield c m be solved straightfomardIy as a finite-difference boundary-value problem in the hodograph plane. This and its extension to supercritical f l o w was performed by Norwood [1962] and Benson & Pool [1965], who solved the planar flowfield in which the subsonic and supersonic regirnes were calculated separately and a matching solution dong the boundary sonic surface was determined iteratively. The tuciaily symmetric problem was nrst solved by Fenain et al [1974], who determined the

entire flowfield in the hodograph plane, and then by Alder [1976], who solved the suhsonic portion of the flowfield in the hodograph plane and the supersonic portion in the physical plane. Filippov et al [1982]examined the unsteady plana and axisymmetric flow equations for an infinite vesse1 using the finite volume method to solve the choked flodeld through orifices with Merent gas specific heat ratios.

A method to determine the discharge coefncient based on simplifjdng assumptions that permit compressible flows to be determined from incompressible ones using a flow analogy was published by Buckingham [1931]for subcritical flows and then extended into the supercritical range by Cunningham [1950].Both authors assumed that the velocity distribution dong the orifice wall and channel nomal to the direction of flow was the same for both incompressible and compressible flows and then they computed the variation of the discharge coefficient with the pressure across the orifice using one-dimensional theories. The same assumptions for flow from an infinite reservoir were applied by Jobson [1955]. This method offered a quick but approximate means of determining the discharge coefficient, but it was not accurate for al1 orifices. Bragg [1960]extended the work of Jobson and determined a much better flow analogy. He assumed that at any cross section of the approach to the outlet the mass flux at the walls was proportional to the average mass flux through the section. This constant of proportionality was taken to be the sarne in al1 sections and i s independent of the flow rate. With this assumption, Bragg's method provides consistent results for al1 orifices. Unfortunately, none of these methods predicts the choking condition nor are they complete or accurate for al1 configurations.

1 . 3 . 2

Experimental Determination of the Discharge Coefficient

A literature survey revealed that a few experimental methods have been used to measure the discharge coefficient of angled orifices for liquid and gas flows. Most experimentalists like Perry [1949]used a cornpressor to produce a longduration steady air or steam flow through a pipe system that contained the test orifice; measured the mass Bow rate with a calibrated massflow orifice, nozzle or venturi meter placed in series with the test orSce;
calculated the
m a 9 9

flow rate by using simple theory; and then obtained the discharge

coefficient by using equation 1.1. Fenain et al [19?4] replaced the cabrated flow meter with an interferorneter such that the orifice jet could be photographed and the actual

aid of interferograms. Alder [1976] used an indraught method in which a large low-pressure chamber drew air fiom the atmosphere through a
mass 0ow rate evduated with the

pipe system that contained a calibrated f l o w meter and the test orifice. Historicaiiy, the majority of tests were performed with incompressible fluids through axisymmetric orifices due to their simplicity and practical application. The American Society of Mechanical Engineers (ASME) and British Standards (BS) have released detailed analysis on orifice flow meters [ASME, 19611 and their hmdbooks which ded with the variety of parameten that affect orifice flow meters, including diameter ratio, Reynolds number, position of taps, orince thickness, edge rounding and orifice eccentricity

are fiequently referenced. Due to the complexity of designing and implementing planar
flow though a slot that minimizes the effect of not having an infinite width, few experimentalists have tested this setup. The literature shows that Benson & Pool [1965] and Alder [1976] performed some experiments with slots of finite width with limited success.

1 . 4 Purpose and Objectives


The purpose of this thesis is to perform a comprehensive study on the effects of the parameters given in equation 1.2 on the numericd and experimental discharge coefficient. This study culminates with the development of a computer procedure that returns the

value of the discharge coefficient for steady, irrotat ionai, isent ropic compressible flows for planar and axisymmetric orifice flow configurations. This procedure, named the Cd
Algorit hm, is based on t heory, numerical results, empirical relations, as well as experiwere chat it had to be fast, accurate and mental data. The objectives of the CdAlg~nthm complete (for the variables specified in equation 1.2). It must be fast for its application as a subroutine for other codes (especidy computational fluid dynamics program), the determined discharge coefficient must agree with theoretical resuits, and it must provide accurate values over the complete range of the input parameters.

1 . 5
1 . 5 . 1

Flow Types and Definitions


Incompressible and Compressible Flow

The efflux of fluids from an orifice is driven by the pressure clifference across the element. Historicdy, the pressure ratio across an orifice discharging from an infinite reservoir to the atmosphere has been used to determine the fiow regime. The ratio of the atmospheric

pressure pam to the stagnation pressure p, across the orifice is denoted by r , where
Patm p z - . Po

This pressure ratio is used to divide the flowfields into subcritical and supercritical regimes as detailed below. Flows in which the density p i s constant throughout the Bowfield are known as incompressible flows, which can be considered the limiting case when the pressure ratio tends to one and the speed of sound

becomes infinite.

1.5.2 Subcritical, Critical, Supercritical, and Choked Flow


The flow regimes for an orifce are detennined by the pressure ratio, and its limiting value For the regimes depends on the ratio of the gas specific heats. This is defined as the ratio

of the specific heat at constant pressure e, to the specific heat at constant volume c, and is given by

Subcritical flow is flow which is everywhere subsonic. Subsonic f l o w is Elow in which the velocity everywhere (including the fiee jet) is less than the speed of sound, i.e., the Mach number hl = V / a everywhere is l e s than one. The pressure range for subcritical
flow is

> r > r*,


T

(1*6)
equals one, and critical

where the two limiting cases are incompressible flow, where

flow, where r is equd to r.. For subcntical flow, the jet emanating from the orifice tapers

to a minimum area at a theoretical infinite distance fiom the exit plane if the viscous effects of the interaction of the jet with the atmosphme are neglected. Critical flow occurs when the maximum velocity (located on the fiee jet surface) is

sonic or M i , = 1. The pressure ratio at which critical flow is obtained is defined as the
criticai pressure ratio r. and i s given by

At this pressure ratio, the limiting velocity of the jet is sonic, and the rest of the flowfield remains subsonic. Theoreticdy, the location of the sonic condition across the jet occurs a finite distance from the exit plane of the orifice. The distinction between onedimensional c h o h g and multidirnensional choking is very important in detemining the discharge coefficient. The one-dimensional mass flow rate is constant for aIl pressure ratios less than or equal to the critical value (supercritical flow) and hence the flow is one-dimensionaUy choked. The actual mass flow rate increases with increasing stagn* tion pressure (decreasing pressure ratio) beyond the critical value due to the changing shape and location of the multidimensionai sonic surface around the plane of the orifice lip. The pressure ratio must be quite low in order for the position of the sonic surface to stabilize in shape and position and the flow to actually "choke" or reach the condition of maximum discharge. Supercritical flow involves flowfields with subsonic, sonic and supersonic flow regimes with the jet velocity greater than one i.e. R/rjet > 1. For pressure ratios less than F . , the sonic surface which was far away from the onfice exit for critical Bow moves to the exit plane and is curved between the angled wall lip and the axis of symmetry. Supercritical flow occurs for pressure ratios less than the criticai value:

Unlike a converging-diverging (Laval) nozzle, onfice flow with a sonic velocity at the vena contracta is afFected by reductions in the downstream pressure. These reductions cause the vena contracta to open and allow more fiow to pass through the orifice. At a sufiiciently low pressure ratio
p..,

changes in the pressure ratio no longer affect

the shape or position of the sonic d a c e . At this pressure ratio the flow becomes independent of r and the flow is multidunensionaily choked at

This pressure ratio is a function of both the gao specific heat ratio and the wall angle, and i s detailed in section 2.4.2.

1 . 6
1 . 6 . 1

Introduction to Orince Flows


Description of the Flow in the Physical Plane

A side-view of flow through a converging orifice is given in figure 1.2 for subcritical flow from a channel, subcritical flow fkom an i n h i t e reservoir (plenum flow) and supercritical flow from a channel. The flow of fluid through an orifice originates far upstream of the exit and terminates far downstream of the exit. For the subcriticd Bon. through a Channel shom in figure 1.2a, the fluid starts traveling through a symmetricai fnctionless duct with uniform subsonic velocity V, at a distance -w downstream at location A-A'. The flow accelerates d o m the channel and through the convergent section B-B' to C-C' where it exhausts from the onfice with a sharp lip at location C-Cr into an infinite reservoir at a pressure pst,. The exiting fluid forms a jet bounded by a free streamline C-D and C'-Dl at a constant Mach number kljetand pressure pjet. In the limit as the channel walls are located far apart , the angled converging portions of the boundary extend backwards to infinity. This is known as flow from an innite reservoir or plenum flow, as shown in figure 1.2b. In this case, the angled walls C-B and Cl-B' extend to infinity upstream at the same (constant) angle a. For superc~tical flow, the jet initially contracts to a minimum cross sectionai area, As indicated in this figure, the curved sonic surface then expands as seen in figure 1 . 2 ~ . (C-G-Cl) leaves the lip at a right angle and intersects the axis of symmetry also at a right angle. A Prandtl-Meyer expansion fan is centred on the orifice lip and intersects the sonic surface, bringing the flow from subsonic to supersonic velocities [Guderley, 19621. The final characteristic of t his fan denotes the limit of upstream dist urbance propagation and beyond this the flow is truly supersonic. In figure 1.2 the diagrams c m represent axisymmetric fiow through a conical orifice or planar flow through a slot. For axisymmetric flow the upper and lower boundaries (channel, converging section and 6ee jet) are connected in a cylinrical shape while for planar flow the boundaries extend as infinite planes normal to the page.

Vena Contracta
For both gas and liquid jets, the flow contracts with distance tom the exit to a minimum area c d e d the vena contracta, or aerodynamic throat, where the flow becomes parailel to the axis of symmetry after Ieaving the orifice.

T h i s minimum

area is caused by

the inward radial momenturn of the Buid and is known to occur both theoreticdy and experimentally. For the subcriticd case of a subsonic gas jet, the contraction occurs asymptoticdly with distance to a minimum area at an W t e or very large distance from the opening (if the interaction between the jet and the atmosphere i s neglected). For the critical case of a sonic gas jet, the contraction occun asymptotically with distance to a minimum area but this minimum now occurs at a finite distance fiom the jet. For the supercritical case of a supersonic gas jet, the flow b t contracts to a minimum area at a relatively short distance from the opening but then expands to a larger area as the flow accelerates to supersonic speeds.

1 . 6 . 2

Description of the Flow in the Hodograph Plane

The flow velocity V and f l o w angle 0 compose the two variable in the hodograph plane. These variables can be expressed in terms of the axial u and traverse v components of velocity as u = V cos 0 and v = V sine. Solving for V and yields

v2= u2 + v2

and

e = tan-'

(3.

The flow diagrams detailed above from the physical plane are represented in the hodogaph plane in figure 1.3. The entire subsonic physical domain maps into a rectangular region, as seen in figure 1.3a. The flow from infinity (A-A') maps onto a source point

x i s . The channel wall (A-B), which is a line of constant 0, Iies on on the flow velocity a
the 8-axis from the source to the orifice stagnation corner (B), where the flow velocity is

0 (for the lower and upper physical half-plane zero. The flow angle changes fiom O0 to f
respectively) and dong the converging wad the flow angle is constant but the velocity increases to Yet at the lip (C). The jet leaves the orifice at this constant velocity and the

flow angle tends to zero as the jet contracts to a location far downstream and represents the sink (D-E-Dl).
For the fiow rom an inhite reservoir, the source point expands to cover the entire left boundary (far upstream condition) as shown in figure 1.3b. The source is nom from

B-F-B' and originates from all flow angles at a zero flow velocity. The subcritical portion of the supercriticai slot or orifice in the hodograph plane i s shown in figure 1.3~~ where the main ciifference between this figure and the subcriticd s that the sonic surface forms the right boundary at V = V w or M = 1 Uistead one i s not shown in this figure as it of the free jet. The supercritical portion of the flow i

is not easily solve in the hodograph plane but is detailed in Guderley [1962]. For
the supersonic portion of supercritical flow, the V,0 coordinates are replaced by the

hodograph characteristics coordinates [Benson & Pool, 19651 since a point in supersonic flow can only influence the points Iying in its zone of dependence.

Figure 1.1: Flow from an infinite reservoir through a dot and orifice.

channel

G '

fieejet

channel

Figure 1.2: Orifice or dot flow in the physical plane: (a) subcritical slot or orifice with a

ChameI, (b) subcritical s l o t or orince fiom an infinite reservoir, (c) supercritical d o t or orifice with a channel.

t
B*
lower corner

lower waU
- ---

C '
channel

lower tkee jet source

upE'e= corner

upper free jet

1
0

upper wail

lower wall
source
F '

Iower free jet


sink

source

upper fiee jet


upper wall

lower wall

B'
lower corner channel sonic surface

UPpet corner

sonic surface

upper wall

Figure 1.3: O r i f i c e or dot flow in the hodograph plane: (a) subcritical slot or orifice with a channeI, (b) subcritical slot or orihe fiom an infinite reservoir, (c) subcritical portion of supercriticai slot or orifice with a channel.

Chapter 2
Theory for and Testing of CFD Code
2 . 1 Introduction
Cornputer codes have been written in the past to solve subsonic, sonic and supersonic

flowfields associated with orifices and thereby produce the discharge coefficient [Nonvood,

1962) [Alder, 19761. These codes are not avaiiable today, and the numerical data from these codes that are published in the literature are insufficient for the purpose of the present thesis work. Hence, a two-dimensional computational fluid dynamics (CFD) code was Nntt.en specificaliy for this thesis, in order to determine the discharge coefficient for both sharp iipped slots and conicai orifices as a function of the pressure ratio, area ratio, waii angle and the gas specific heat ratio. The governing equations, boundary conditions, solution procedure and some related information (convergence, accuracy, agreement with previous computations) are presented in this chapter.

2.2

Governing Equations

The partial differentid equation for irrotational planar f l o w is given by [Shapiro, 19531

in terms of the stream hnction @ for the hodograph plane. The corresponding partial

l o w is given by [Alder, 19761 differential equation for ;usymmetric f

df'

sin (0) cos ( O )

,( 1

(1 - M2)

+: +&*)

= 0.

The partial differentiai equation for irrotational flow in terms of the velocity potential # is given by [Shapiro, 19531

for planar flow in the physical plane (x and y coordinates). The corresponding equation for avially symmetric flow in the physical plane ( r and r coordinates) is [Shapiro, 19531

The above partial differential equations can be shown to be elliptic for subsonic f l o w , parabolic for sonic flow and hyperbolic for supersonic flow, based on the theory of characteristics (paths of information propagation). Note that the derivation of equations 2.1 to 2.4 is given in appendk A. Lines of constant stream function called streamlines help rnap particle paths in the physicai plane, whereas lines of constant velocity potential intersect these streamlines at right angles. Working with the stream function simplifies the determination of the discharge coefficient as the change in its value defines the mass flow rate through a soc d e d "stream tube". The strearn function equation is used to solve subsonic flowfields

whereas the velocity potentiai equation is used in an alternate f o m as the method of characteristics to solve the supersonic part of the flowfield (if present). The stream function equation (for subsonic Bow) and the velocity potential equation (for supersonic flow) were solved using a onite difference technique. The orifice flowtield was divided into the subsonic and supersonic (if present) portions, and then solved independently. The common boundary between these two regions is the sonic surface and its position was determined iteratively based on certain boundary conditions. The subsonic flowfield was solved as a boundary-value problem in the hodograph plane using a relaxation technique, while the supersonic flowfield was solved as an initial-value problem in the physical plane using the method of characteristics. The CFD code was written with the foiiowing assumptions: the compressible fluid obeys the ideal gas law @ = pRT,

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

1 7

where 73 is the gas constant), the f l o w is irrotational (inMacid and isentropic), the orifice lips are sharp, the f l o w discharges into a still atmosphere with no shear layer or mixing, the upstream flow has a uniform approach velocity and the smdl effect of gravity c m be ignored.

2.3

Solution Method for Subcritical and Critical Flow

Working in the hodograph plane offers a number of advantages over the physical plane for converging orifice flows. Solving the Stream function equation in the hodograph plane removes difficulties with detennining the location of the free streamline (for subcriticd flow) or the sonic surface (for criticai or supercritical f l o w ) in the physical plane, and the physical distances upstream and downstream at infinity map onto a single point or boundary (source and sink). Other advantages include that for planar flow the subsonic hodograph equation (2.1) is a linear partial differential equation, and when representing the hodograph plane as rectangular coordinates (as opposed to the standard polar coordinates) the entire flowfield maps onto a rectangle resulting in the simple application of a

standard grid for the finite difference equations. In the hodograph plane, the subcritical portion of the flowfield is solved as a boundary-value problem and onIy half the flowfield needs to be determined due to symmetry.

2 . 3 . 1

Non-dimensionalization of the Stream h n c tion Equation

The Bow variables presented in equations 2.1 and 2.2 are dimensional. In Systeme International (SI) units these dimensions are

However, it is more convenient to work with non-dimensional variables and a specialized Mach number. The non-dimensional velocity M, = is chosen rather than the Mach

nurnber because hl is a b c t i o n of the sound speed that varies throughout the flodeld.

The constant a* is the speed of sound when M = 1 . The use of M . results in the folIowing set of relationships:

CHAPTER 2. THEORY FOR

AND

TESTING OF CFD CODE

and

1C, (nhere & T o non-dimensianalize the Stream function equations, let iI = ,


is the stream function value of the free surface) and use non-dimensional units in the
dJmax

physical plane. Hence,

When these new E o m of the derivatives are applied to the stream function hodograph equations, one obtains

For planar flow and

2R cos ( O ) (-B:

1 4

RL

+ CI. @ Y + )

n2 sin ( O ) cas (8)


for axially symetnc flow, where

1 ((1 - A) M?

iP:

+
=

qs) =0

2 y+l A= and 1-s h . .

po+mu

pr"ltz.P

'

The determination of the discharge coefficient was sirnplified by pre-specifying the mass fiow rate through the orifice. Define

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

19

for planar and axisymmetric flow respectively, where e is a unit distance. By convention,

l , ,= while = O for axthe m i a i m u stream b c t i o n for planar flow is r isymmetric f l o w ,which for both planar and axisymmetric flows yield 3 = O on the a v i s
of symmetry. By defining the maximum and minimum stream function in this manner,

l o w rate through the orifice has been fkced, and the resultant equations for the m a s f the discharge coefficient depend only on the physical coordinate and the nomalized flow
velocity a t the sharp lip (see section 2.6 for more details).

2.3.2

Derivation of the Subcritical Boundary Conditions

The flows through orifices as viewed in both the physicai and hodograph planes were described in section 1.6. In this section, the stream function conditions bounding the orifice problem are detailed for the three exampies seen in figure 2.1. Note that for axisymmetric f l o w the upper and lower boundaries represent the continuation of the

l o w which the slot is assumed to be infinitely same entity, unlike the case for planar f
wide (into the page). Subcritical boundary conditions for solving the flows through orifices are needed in terms of the stream function for the hodograph plane. Consider the flow through the orifice in figure 2.1a. By convention [Gurevich, 19651,the stream function increases in value from right to left when looking along a streamline in the direction of the f l o w . Hence, the upper boundary (including the channel, angled wall and free jet) is a streamline with
,$ =

or 8 = 1 and the lower boundary is a streamline with @ = qmin or J! = -1 for planar flow. For axisymmetric flow, the upper and lower boundaries are the same physical
@max

entity and hence both shaxe

Due to symmetry, the centreline is a streamline I I ,= with . = O b r planar flow or $ = = O for axisymmetric flow. Far 2 upstream and far downstream (as I or Z tend to ho), au streamlines are parallel to

+ $'min

+=

the centreline and the boundary condition can be derived fkom the basic definition of @. Upstream, the flow is uniform with a velocity V, and a density p,. For planar flow

P and for parailel uniform flow & = O. Substituting 11, = -u and integrating fiom the Po centreline (y = O) to the channe1 wdl yields

FOR AND TESTING OF CFD CODE CHAPTER 2. THEORY

and after normalization

and for uniform flow

= O. Substituting

P er= -ru
Po
n

and integrating from the centreline

( r = O) to the channel yields


Poo

$(r) = -VmPo

TL

2'

M e r normalization this equation becomes

If the angled walls extend to infinity the flow frorn a channel becomes the florv from an infinite reservoir (or plenum flow), as sketched in figure 2.lb. The only difference from the subcritical orifice with a channel is that the far upstream iIr boundary condition is a function of the flow angle. The boundary conditions for the supercriticd orifice are shown in figure 2.1~.For solving the subsonic portion of the f l o w the only difference from the subcritical onfice detailed above is the downstream boundary. The sonic surface is the lirnit for subsonic Bow and the stream function distribution dong it

needed to be determined in combination with the supersonic solution. The boundary stream function values determined above are presented in the hodograph plane in figure 2.2 for the subcritical orifice, the flow fkom an infinite reservoir and the subsonic portion of the supercritical orifice. Note that for the subcritical case the complete boundary is known and that two stream function singularities exist: the source

and sink. For the supercriticd case, the J> distribution dong the sonic surface is initially unknown so values were assumed and then determined iteratively in combination with the supercritical portion of the flow. For the flow from an infinite reservoir the source singularity extends over the entire upstream boundary. The distribution here can be determined mathematically-defme a distance s , where s is large, and use it in conjunction

with the BOWangle. Again, using the dennition of @ for planar flow

Now x = -s cos 0 and y = s sin0 far upstream so & = ssin Ode - cos eds and d y = s cos Mt9 + sin Bds. Substitution into the above equation and simplification yields

Upon integration

+(el
so

Poo = -v,se,

Po

0 4(0) = -= -= Fa for the upper and lower half orifice respectively. For axidy symmetric flow
@mm

w,.

Y w

dql, = $,dz

+ &dr

P = --rvdt
Po

+ -Pr u t
Po

P = --TV
Po

P sin 0dz + -TV cos Odr.


Po

Now z = -s cos 8 and T = s sin 0 so dz = s sin 8d - cos 8ds and dr = s cos 8d8 + sin 8ds. Substitution into the above equation and simplification yields
Pd0 d$ = -v,s2

sin @dB.

Po

Upon integration
@(O) = &v,s2

(1 - cos O ) ,

and after normalkation

It can be seen that the source extends over the upstream boundary and the stream function varies from P ! = O at 6 = 0 to 9 = 1 at B = -a for the upper half orifice.

2 . 3 . 3

Solution to the Subsonic Boundary-Value Problem

The governing equations in section 2.2 were solved as a boundary-value problem by h i t e difference methods and relaxation techniques. Details are given herein. The partid derivatives of @ in section 2.3.1 were replaced with their second-order onite ciifference approximations. The hodograph domain was divided into (N -1) divisions (for N nodes in each orthogonal direction), equdy spaced in M. and 8. AU interior nodes used the standard central ciifference formulas (where i,j refers to the row and column, respectively):

CWTER 2. THEORY FOR AND TESTING OF CFD CODE

whiie the derivatives on the boundaries (which were required to convert fiom the hodo-

graph to the physical plane) used the three-point ,one-sided, second-order, finite-difference equations with similar error terms. Once the finite difference formulas were substituted into the partial differential equation, one obtains a set of linear (for planar f l o w ) or nonlinear (for axisymmetric flow) algebraic finite difference equations which were solved using relaxation met hods. These methods involve splitting the sparse matrix obtained kom the finite differencing and then iterating until a solution is found. They are commonly used for boundary value problems

as they are easy to program, although they may be inefficient for large problems. For was used to force the solving the linear planar problem, successive over relaxation (SOR)
residual to zero, that is,

SOR is an enhancement to the GaussSeidel method which incorporates an acceleration factor w to speed up convergence. For the Gauss-Seidel method, w = 1 while for SOR 1 c w < 2. The value of w used was not constant for the planar case as Chebyshev
acceleration was used to modiQ its value at each iterative pass of the grid [Press et al, 19921. The advantage of Chebyshev acceleration is that the norm of the error is always decreasing with each iteration. For the nonlinear axially symmetric problem, non-linear over relaxation (NLOR)[Ames, 19773 was used, with the acceleration factor fixed at w = 1.6. This value was determined by numencal experimentation. Large values of w (above 1.8) did not remit in the stream function converging at every node, whe w = 1.75 caused the LI norm to oscillate (non-zero amplitude sinusoid). Values of
w

5 1.6 Iead to convergence, and, since it is generally better to overestimate w rather

9 7 7 1 ,this value was used for all of the axisymmetric m. than underestimate it [Ames, 1
For the planar case, setting the initiai value of B for the interior nodes between the

maximum and minimum values converged to a solution, but for the -etric

case

a good initial is required, as is the physicd variable t at each node. The solution to the slot was solved first and then the initial @ values used for a;irisymmetric f l o w were Qijlui = B2-1 phaarw T h i s relationship was based on the upstream channel boundary condition derived in section 2.3.2. The starting values of the physical variables f and i were set to the planar Z and i solution values respectively. Updates to the stream huiction for NLOR were determined by

The stream b c t i o n distribution in the hodogaph plane for planar f l o w was deemed to have been solved once the SOR routine had converged. The criteria used were: the
maximum change of Q between iterations at any interior node was less t han c = 1 x IOwL4, or that the sum of the residuals at al1 interior nodes was less than
E.

For avisymmetric

flow, the partial differential equation is nonlineu and the physical variable i. must be

solved in conjunction with Q. The solution was deemed to have converged once the mavimurn change in the stream function was less than c, the Li n o m of f per node was less than 1 x 10-4 and the change in the axea ratio between NLOR iterations rnust also be less than c.

2.3.4

Mapping the Hodograph Solution to the Physical Plane

Once the subsonic stream function solution was determined, the physical coordinates at each (il.I.,0) grid point were calculated from the hodograph-physical plane relationships.

l o w the equations are For planar f

Po

( N e cos 6 - aM.an 9 ) ,

and for &aUy symmetric f l o w


Po

(4~. Me COS d - Be s i n O) ,

CWTER 2. THEORY FOR AND TESTING OF CFD CODE


Po

24

(OeCOS B

+ !PM.M. sin 8 ) ,

w here

The physical coordinates were determined using a piecewise cubic spline interpolation routine to first fit the hodograph data and then integrate dong lines of constant M.. Lines of constant M. were chosen over lines of constant due to the simpler form of the equations with the derivative -.

ne of constant M. to the the angled wdl, integate dong the wall (line of constant 0): and then integate down the lines of constant 1 C . I from the w a l l to the axi*s and to the
channel (if present) . This integration procedure helped minimize numerical difficult ies

0 6

The integration path was to integrate up one

and increased the accuracy in determining the physical variables near singularities in Q. The physical coordinates throughout the Bowfield are required for the a?<iallysymmetric solution and the coordinates of the sonic surface are required for the starting location of the supercritical met hod of characterist ics solution. n the hodogaph plme is that the far upstream velocity, One disadvantage of working i not the area ratio, must be pre-speciiied. This is significant because if a certain area ratio is desired, an iterative procedure involving multiple solutions of the entire flowfield is required.

2.3.5

A Subcritical Flowfield Solution

Examples of the subcritical flowfield solution in the hodograph and physical planes are given in figures 2.3 and 2.4. Figure 2.3 shows the stream function distribution in the subsonic hodograph plane for a slot with a finite-sized upstream channel and a wall angle of 90. The two iIr singularities are evident on the M, axk, and the lines of constant
@ (incremented by 0.1) travel fiom the source to the sink. The results of mapping the

hodograph plane to the physicd plane for this example are s h u m in figure 2.4. The dot wail and channel are indicated, as are the lines of constant M . (integration lines) throughout the flowfie1d. Recaii that these lines are also h e s of constant gas pressure,

CUPTER 2. THEORY FOR AND TESTING OF CFD CODE

25

temperature and density. The effect of the downstream sink (4 singularity) is seen in the lower nght section of the figure as equal spacing in the hodograph plane results in large ciifferences in the physical coordinates between nodes. A similar pair of graphs for the case of subsonic flow fiom an i n h i t e reservoir is shown in figures 2.5 and 2.6, where it c m be seen that the stream function source in the hodograph plane is now spread over the leR boundary.

2.4

Supercritical and Choked Flow Regime

The supercritical flowfield was solved in the physical plane due to the iteration procedure used to match the conditions along the sonic surface between subsonic and supersonic f l o w . Since only the discharge coefficient for supercritical flow is desired, only that part l o w which has the characteristics iduencing the sonic surface needed to be deterof the f mined. This portion of the flowfield is an initial-value problem starting from the sonic surface and covers the zone of subsonic influence from the Prandtl-Meyer fan emanating from the orifice iip, or from the free jet surface (for non-choked flow). From the solution to the subsonic flowfield, the physical and hodograph coordinates were known along the sonic surface (initial-value condition) and, using the method of characteristics, a grid of characteristic intersections was built out into the domain of influence as seen in figure 2.7.

2.4.1

Method of Characteristics

The method of characteristics for supersonic flow i9 derived corn the velocity potential equations in the physical plane and the calculation procedure follows Shapiro [1953]. The method of characteristic equations are divided into two 'Tamilies" or sets of characteristic equations depending on the choice of the positive or negative sign. The rightninning Mach Iines are the "-" characteristics while the lebrunning Mach lines are the "+"characteristics. The two pairs of equations for planar flow are given by

and where

For axiaIly symmetric f l o w ,the two pairs of equations are


O=-&+
sinpsint9 dF dF -+C-, -=tan(&-p) dz' sin (O - p ) P

and

where Y in appendix A.

sinp sin8 dT dF = tan (0 + p ) . G+, dZ sin (0 p) P is again given by equation 2.27. The derivation of these equations is presented

0 = ~ -

2.4.2

Supersonic Boundary Conditions

The characteristic network for supersonic flow is shown in figure 2.7. The sonic surface is

y) or (2,?) and 8 known f i o m the the initial-value condition with M = 1 and values of (I, subsonic solution. The values of @ = (O) were initidy estimated for the subcritical boundary-value problem, and their proper values were iteratively determined. An important supercritical pressure is the one at which choking occurs. The choked
pressure ratio can be calculated straight-forwardly for two-dimensional planar flow. If the flow exiting the orifice is choked, al1 the characteristics intersecting the sonic surface originate in a Prandtl-Meyer expansion centred at the lip [Guderiey, 19621. Under this condition, the fiow inside an orifice wiil cease to be afTected by lowering the external h i s corresponds to the point H and al1 Qi being coincident pressure (decreasing r ) . T with point L in figure 2.7. Along the characteristic curve LH

and along the other characteristic OH

Subtracting and solving for

yields

At the lip L, the flow angle is 19 = -0, and along the centreline the flow angle equals zero. Since OL is the sonic d a c e , M = 1 and hence uo = = v ( M = 1) = O. When the point H i s coincident with L, the orifice is choked and Ma = M.,. Using these limiting conditions, the above equation reduces to

v a = -a
2

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE


where

and

Hence the choking pressure ratio for planar flow can be calculated by simply knowing the wall angle and the gas specific heat ratio.

l o w case, the entire supercritical flowfield must be solved nuFor the axisymmetric f
mericaily by the method of characteristics to determine the choked pressure ratio. Once solved, the value of M., can be determined from

and the relationship between u and M. Alder [1976] showed that T.. b r axially symrnetric flow is slightly lower than that obtained for planar Bow, with the pressure ratio T.. only differing in the third decimai place. For euample, for a = 90 and 7 = 715, rmlp,, = 0.039 and r . . l ~ = 0.035. For unchoked f l o w , only some of the negative characteristics emanate from the lip and the rest from the free streamline (free jet sudace). The method of detennining the intersection of the negative characteristics and the free jet surface was determined by the free surface point unit process as detaiied in Hodge & Koenig [1995]. Across the free surface, the static pressure is constaot, hence LM and v are constant along the free surface (jet boundary) but the flow angle varies. For the intersection of the negative characteristics with the jet surface,

and then the physical coordinates of the intersection of the jet surface and characteristic can be determined as detailed below.

2.4.3

Solution of the Initial-Value Problem

The method of characteristic equations given in 2 . 4 . 1 were written in finite clifference form after a minor modification. Equation 2.29 has a term with s i n (0 + p) in the denorninator and problems arose when (0 - p) n O or K. An alternate form of this equation can be

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

28

derived by substituting the second portion of equation 2.29 into the first part to yield another fom of the equation for O:

O=V-

sinsinp d i - +C+cos (O p) T

T h i s equation avoids the problems resulting from having a zero (or very s m d number)
for the denominator. The characteristic mesh was started from the sonic surface and extended into the supersonic flowfield. Let any two points on a constant M line be represented by "a" and "b7',where Oa > Ob and the solution point be represented by "c". Then, the method of characteristic equations for planar flow written in finite difEerence f o m are

and

For axisyrnmetric fiow, the finite difference representations axe sin fisin 0 dZ\ cos (6 + p ) i / '

with the two integrals evaluated by using

? a

sin (@a - p a )

8,sin c(, + Fesin sin (Oc - pc)

and the characteristic slopes are

CHAPTER 2. THEORY FOR

AND

TESTMG OF CFD CODE

29

From these sets of equations, a mesh of intersection points between the two famiiies was built up from an initiai-value iine. The average value of 0 between nodes "a" and "c" and between "b" and "c" was determined using the conventional method

8,

e.+e, , 2

abc=- eb+e,
2

but the determination of each intersection point in the supercritical region was made more accurate by using a slightly different method to calculate the "average" p values. Generally, one uses the arithmetic mean between any two points "an and "cn;

Clac

2 for the average value dong the characteristic segment. The integral method using the mean-value theorem for integrals [Salas et al, 19861 combined with p = sin-' (&) produces

Pa + P c

Clac=
which was proven to give significantly more accurate results at reasonable grid sizes due

C f = 1+ region. to the rapid changes in p around the i To verify t hat the solution satisfied the physical boundary conditions, the stream
Eunction was required at the grid points in the supersonic flowfield. From the definition of the stream function, the following equations were derived:
d* =

M@
(1 +
2)

a[cos (0) Q - sin (0) d2j


8 %
[COS (8)

(2.46)

for planar flow and

MF@
d\k =

(1 + 2+,12)

dF - sin (O) d i ]

for axialiy symmetric flow. The trapezoidal method of integration was then used to

obta h the st ream function values dong the positive characteristics.

2.5

Matching Subcritical and Supercritical Regimes

The method of matching the subcriticd and supercritical Bow regimes involves an iterative process which depends on the strearn function vaiue dong the sonic d a c e and

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

30

along the &ee jet surface. The value of Y fiom the Lip aiong the free jet surface must be constant a t rE' = 1, the stream function distribution aiong the sonic surface must strictly increasing fiom the axis of symmetry to the orifice lip, and the stream function value along the sonic surface must satisfy the boundary condition that there is a Prandtl-Meyer expansion fan at the lip. The method used was detailed in Alder [1976] and i s based on perturbing the sonic surface iy values.

For planar flow, the partial dinerential equation i s linear and only one iterative p a s of the sonic surface is required. For axisymmetric flow, the partial dinerential equation is nonlinear and the solution requires two or three iterations until the error vector becomes sufficiently small. For choked flow, at each Qi, 1 5 Mqi 5 khetand Q = 1 in figure 2.7. If the fow is not choked, then if AlQi > Adjet then O = 1 when hl = ib4, almg the characteristic (strearnline boundary condition). For an arbitrary initial q,,, ( O ) , Qoi # 1 (or @ # 1 when b l =b k ) . Therefore, one has a vector of errors

where Nq,=

aqi- 1 and N

is the number of axial or traverse finite difference grid

points (for an N by N grid). Note that the node on the 2- or 2-&us and the node on the p have a fixed value of 9 and hence do not play a role in determining iImnic (O).

Now examine how each Ipi affects each

aqi :

which, in mat*

form for ail the (N - 2) sonic surface points, is

which can be represented by 6iYQ = ZAqp, and can be rearranged to permit the solution of the "correct" sonic surface ik d u e s as

_I_)

_C3

CHAPTER2. THEORY FOR

AND

TESTING OF CFD CODE

31

An example of a "correctn stream function sonic surface distribution for non-choked flow is shown in figure 2.8 where it c m be seen that three of the negative characteristic lines emanate from the s h q Iip and the jet location has been determined. The subcritical portion of the supercritical flowfield in the hodograph plane is shown in figure 2.9, where the strearn function distribution dong the sonic surface needed to be detennined in conjunction with the solution of the supercritical portion, shown in figure 2.10. In this latter figure, the method of characteristics mesh is shown fiom the sonic surface (initialvalue line) to the onfice lip (but note that the mesh shown is simply a connection of the characteristic grid points and hence the indicated slopes dong the sonic surface, which should form cusps, are not fully correct.) Since al1 negative characteristics intenect the lip, this figure indicates that the flow is choked.

2.6

Determinat ion o f the Discharge Coefficient

The discharge coefficient was determined numerically fiom the flowfield solution by calculating the actual and theoretical m a s s flow rates through the orifice from the centreline to the lip and applying equation 1.1. The actual mass flow rate per unit width for planar f l o w through a slot is given by

is the actud m a s flow rate for axiisymmetric flow through a conical orifice. For planar

for a unit width. For axisymmetric flow, A* =

- O = a.P so

m ,

= 2irp,,a.P.

The one-dimensional ideal mass flow rate is given by


hideal
*

= hip w

hp and

mideid

2 =h i p r ~ Fip p

(2.48)

for planar and axial1y symmetric flow, respectively. Combining the above and simplification yie1ds
' I

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE


and

where

2 . 7 Accuracy of the Numerical Solution


The values of the discharge coefficient obtained from the CFD code have been cornpared to the results of others and these cornparisons are detailed here. The accuracy of the strearn function distribution and the mapping fiom the hodograph to physical plane is directly dependent on the grid size, where the total number of points used in the solution is N x N as both hl*and 0 ranges were divided into N - 1equally spaced segments. The finite difference approximations given in section 2.3.3 are of order ( ~ l l . 1 and . ) ~ ( ~ 6 )-4s ~. N -t 00,these error terms tend to zero so it is expected that the results of the numerical solution approach the "true" results as N becomes large. For the computed value of the discharge coefficient one can write

to fit the data, where A and B are convergence constants and Ce, is the value of the discharge coefficient as N -+ W. The CFD code was run for a variety of N values and the resultant data was fit to equation 2.51 by a least-squares fit using the downhillsimplex method [Nelder & Mead, 19651 [Press et al, 19921. The parameters of the fits are presented in table 2.1 for a = 90, 7 = 715 and A, = O, while the gaphs are
presented in section 2.7.1. For the critical and subcritical solutions, the value of B was anticipated to be about -2, showing that the accuracy was proportional to 1 / p , as the finite ciifference equations used are second order accurate. The results hom the choked cases do not necessady produce as large of a convergence factor due to some low-order approximations used in the method of characteristics. Solving the subcritical cases for N 5 200 did not take excessive computer tirne, but the supercritical cases required perturbations of the stream hinction value at every node on the sonic surface, which required the entire flowfield to be solved (N - 2) times. This

means that solving the choked flowfield required very long computer nuis and hence large grid sizes were not aiways practicai. Methods to hprove the accuracy include

CHUTER 2. THEORY FOR AND TESTING OF CFD CODE

Planer incompressible
crit ical choked
- --

1 0.61102 1 0.119 1 -1.91 1 0.74456 1 0.171 1 -1.79 1 0.84923 1 3.03x10-' 1 -0.333


- -

Axisymmetric incompressible 0.59137 critical choked


0.72606 0.83073

8.84
4.41 0.446

-2.65
-2.17

-1.62

Table 2.1: Curve fit parameters for the CFD code resuits. using an unequal number of nodes in each direction, using nodal concentration neax the * line), or solving the flowfield on a corne grid and them mapping the singularities (0 = O solution to a finer grid to improve convergence. These improvements were not undertaken in this study as the discharge coefficients were deemed to be sufficiently accurate with reasonable grid sizes.

2 . 7 . 1

Cornparison of Results

The CFD code was an extension of the method published in Norwood [1962] and Alder [1976],so it was expected that it would reproduce those results well. Since Alder's results were published over twenty years ago, little additional information on the planar and axisymmetric jet problem has been detailed. With the currently available cornputers, significantly larger grids c m be used to obtain more accurate results. Alder's maximum grid size was 24 x 24 (but only 16 x 16 for the choked axisymmetric case) while the maximum grid size used to check the accuracy of the planar incompressible case for this thesis was 1500 x 1500. For most orifice configurations it was detennined that grid sizes of about 100 x 100 were more than sufncient to produce good results for a multitude of purposes.
Incompressible, Criticd and Choked Values
Results ftom the literature have generdy concentrated on the limits of the flow regimes: incompressible, cntical and choked f l o w . The discharge coefficient for these conditions

are now d e d for planar and aiosymmetric flow. The incompressibie flow through a 90 dot from an infinite reservoir can be solved

CHAPTER 2. TKEORY FOR AND TESTING OF CFD CODE


Year

I analytical solution
II Rouse & Abd-Fetouh

1950 1956

I Hunt

II Bloch

1967 1969

I present study
Table 2.2: CdIV, for a 90' slot and conical orifice with A, = 0.
X

using the complex potential method and the exact value of C~IF" = -ru 0.6110154 n+2 is well known. For this case, the error in Cdwas plotted against N on a log-log scale as seen in figure 2.11. Rom this figure, a linear relationship is shown (for the log-log scale)

and a least-squares fit analysis was undertaken where it was determined that -4= 0.154 and B = -1.96 -i. -2 for N > 100 showing that the accuracy is proportionai to 1/w.
The convergence rate of the

CFD code with N for planar flow i s shown in figure 2.12

and for avisyrnmetric flow in figure 2.13 where it can be seen that using N = 100 is more than sufncient to produce three significant figures. A cornparison with the results from
the literature are presented in table 2.2 for the dot and the conicai orifice. Details on the history of the determination of these values was presented in chapter 1, except For Bloch's remit, which was taken fiom Alder [1976]. Unlike the planar incompressible case, there is no known exact value for the critical (or choked) case. The graph of the calculated Cd(,,=, versus N for planar and avisyrnmetric flows through a 90' orifice with 7 = 715 and A, = O at the critical pressure ratio are shown in figures 2.14 and 2.15. Again, it requires less than N = 100 to obtain three decimal accuracy. Selected results from the literature (detailed in chapter 1) are presented in table 2.3. The results published by Kosolapov & Sivoborod [1984] are included for

e r e not very confident about their completeness even though they stated that they w
results since they determined a dinerent value for the discharge coefficient depending on the integration path chosen. The graphs of the calculated d u e s of Cal,oo versus N for planar and a>asymmetric choked Bows through a 90 orifice with y = 715 and A, = O are shown in figures 2.16

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE


Author

II Norwood
-

1 1962 1

0.745 0.76

Kosolapov & Sivoborod 1984

I present study

Pickett

1989 0.744559129 0.74456 0.72606

and 2.17. Simila.to the incompressible and critical cases, N = 100 is sufficient to obtain three significant figures of accuracy. A surnmary of results from the literatuw is presented in table 2.4.
Author

1 Y e a 1 CdI,,,. 1 cdIZ+.,1
ptanar
rl

II Nowood
Aider

I Benson & Pool


Fenain et al

1 1962 1

0.850

1 1965 1
2974 2976

0.853

1
0.837' 0.830

0.849
0.842' 0.84925

I Filippov et al
II present study

1 1982 1
1 1

1
1

0.819'
0.83083

1
11

Table 2.4: Cd(r=r.e for a 90 slot and orince with A, = O and 7 = 715. Note that a superscript ( )' indicates that the results were obtained from reading a curve on a figure. From the cornparisons detailed aboue, it can be concluded that the CFD code produces accurate values of the discharge coefficient for both dots and conicd orifices. Additional results which detail the effects of vuying the area ratio, pressure ratio, wall angle and the gas specific heat ratio are detailed in chapter 5. The results of the CFD code as they were applied in developing the CdAlgorithm are presented in chapter 4.

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE


A
00n
l

channel Y=l

B
free jet Y=l D
...................... . . . . . . . . . * . * . . * . . T -

Y='y@) Y='&)

1 cbmd ................................................................ F........................,,..

: 'A

- e

channel

Figure 2.1: Physical boundary conditions: (a) subcriticd orSce with channel, (b) subcritical orifice nom an innnite reservoir, (c) supercritical orifice with channel. Note that C = -1 for planar flow and C = 1 for anSymmetric flow.

lower wall lower comer

channel
source

':B
B
.-

Y=C
A?-

lower ee jet

Ir-6

s;o
y=o

Y=l A

corner
-

"PF

D ' L D
upper fiee jet

upper wall

lower wall source


lower fiee jet
dv -

sink

source

upper free jet

upper wall

lower wall

B '
lower comer Y = C

channel

Y=C

C'

Y = Y ( ~ )surface

sonic

"PF corner

sonic surface

upper wau

Figure 2-2: Hodograph boundary conditions: (a) subcritical orifice wit h channel, (b) subcritical orifice from an W t e reservoir, (c) subcritical portion of supercritical orifice with channel. Note that C = -1 for planar flow and C = 1 for axisymmetric flow.

M e

Figure 2.3: A subcritical solution in the hodograph plane for flow though a dot.

Figure 2.4: The subcritical solution i n the physical plane for Figure 2.3.

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

M.

Figure 2.5: A subcritical solution in the hodogaph plane for plenum flow.

Figure 2.6: The subcritical solution i n the physical plane for Figure 2 . 5 .

Figure

The supercritical flowfield of an upper-half orifice for arbitrary

Figure 2.8: The supercritical flodeld for the correct

(O).

CHUTER 2. T E ~ O R FOR Y AND TESTTPJG OF CFD CODE

M.

Figure 2.9: The subcritical portion of the supercritical solution in the hodograph plane.

Figure 2 . 1 0 : The supercritical solution in the physical plane for Figure 2.9.

FOR CHAPTER 2. THEORY

AND

TESTWG OF CFD CODE

Figure 2.11: The effect of N on , Cd ,l

for incompressible flow through a ~ l o in t an infinite

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE


0.61 5 1 O.~llO*
a
r r
I

exact mq*~s!!t.........-...; ................. ; ...................... i.......................

' c ~ ~ ~ e s url t

---

1 ..-

0,61108

'

.................... :.. ....; ................ ........................ ; 0-61107 ,

....................... ;. ...................

1 ..........................

..-

0.61 106 -...

i. .......... i...................................... :. . . . . . . . . . . . . . , . . . . . . . . . . . .

.:. . . . . . . . . . . . . . . . .

.-.

op
0.61 105 -...... ............... '- .............................................. i 0.61 104
...................................*..................*..............--..-

- ;. i.. i5 ;; . 0.61103
.,

....

.....

. . . . . .

. . .

-.

.;. . . . . . . . . . . . . . . . . . . . . . .

. . .

0.61 102 0.61 101

. . . . . . . . . . . . . . . . .
'

*,. .
I

. .

;.

. . . .

. . . .

0..

, * * 1 . : 1 * : " , ' 1 : L ~ + : ~ L u . + ~ * ' ~ .


l

200

400

600

800 N

Io00

1200

1400

1600

Figure 2.12: Plruiar incompressible Bow discharge coefficient convergence results.


I I

b l

CFD results le-squares Rt

------8

I L I

Figure 2 . 1 3 : Axisymrnetric incompressible fiow discharge coefficient convergence results.

FOR CHAPTER 2. THEORY

AND

TESTING OF CFD CODE

Figure 2.14: Planar critical f l o w discharge coefficient convergence results.

Figure 2.15: Axisymmetric critical f l o w discharge coefficient convergence results.

CHAPTER 2. THEORY FOR AND TESTING OF CFD CODE

Figure 2.16: Planar choked Bow discharge coefficient convergence results.


r
t '

a . L.

,.
8:

CFD results least-squaresfit

-------

Figure 2 . 1 7 : Axbymmetric cboked 0ow discharge coefficient convergence results.

Chapter 3

Experimental Apparat us and

Met hod
3.1 Introduction
fiperimental data for the discharge coefficient of conical orifices were obtained to check previous data published in the literature, to extend the Cd database with new data
(previously unavailable), and to facilitate a more extensive cornparison with old and

new theoretical and numencal results to assess dserences. New experimentd data were required for these purposes, in particular for the cases of conical orifices with large wall 9 0 1 8 0 ' ) and different degrees of rounding of the orifice lip edges. angles (
A new facility based on a reservoir blowdown method was constructed, tested and used

to produce these new data for this thesis. Over 4000 experimentai nuis were performed

using a total of 21 different orifices. The apparatus was calibrated by comparing new discharge coefficient data with those published in the literature, and also by collecting
new data using a reference conical orifice with a large radius of rounding for which the

discharge coefficient should approach unity. A description of the facility is given first, the theory for the facility operation and the method of determining the discharge coefficient follows, and an assessrnent of facility performance in producing good data is presented

Iast.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

3 . 2
3 . 2 . 1

Orifice Flow Facility


Apparatus

The mechanical aspects of the blowdown facility are shown schematicdy in figure 3.1 and detailed below. The gas reservoir consists of a 1.7 m3 high-pressure tank with a designed pressure range of 0-2 MPa. This tank has special mounts to accommodate a thermocouple and transducer to measure the temperature and pressure of the gas inside the vessel. Pipe connections to the reservoir are also provided for a dust injector at the vessel top and a gas evacuation and gas filling system at the vessel bottom. The main duct from the reservoir contains a smooth aluminum bell-mouthed convergent nozzle, fullported manually-operated ball valve (10 cm intemal diameter), an 18 cm-long smoothly contoured pipe reduction from 10 to 5 cm in diameter, a 17 cm-long test section with two

and an angled conical orifice optical-quality windows for laser Doppler velocimetry (LDV)
attached at the end of the pipe. The bell-rnouthed nozzle eliminates flow separation and l o w to produce reduces radial momentum effects and transverse pressure variations in the f

l o w profile earlier in the duct. The entry nozzle was designed fairly short a more uniform f
to minimize the thickness of the boundary layer. The length of the area reduction section was kept short to minirnize the w d boundary-layer and other stagnation-pressure losses, while the test section is sufficiently long so that the f l o w therein is essentially uniform dong the duct. The test section diameter was measured in four locations with a vernier
4 , . \vas calculated caliper and was determined to be 50.6 mm. Hence, the test section axea m2 or to be 2.01 1x IO-^ m2. The error in this value was estimated to be 6AtS= 8 x

about 0.4%.

The dimensions and types of conical orifices used in the experiments are summarized
in tables 3.1 to 3.3. All orifices were constructed of 6061 aluminum (Al) except for the "sharpn 45' and 90" orifices, which were constructed fkom steel (Fe). Note that many of the 45" o a c e s were designed so they could be instded backwards and thereby becorne 90 orifices. AU orifices were designed with a sharp-as-possible lip, except for those used to study the effects of edge rounding. All exit diameters were designed to be O.5OO-inches except for the orifices with large area ratios. The exit diameters were measured by the
manufacturer with a precision hoie-gauge and the maximum error in the exit area was estimated to be 6Amt = 2 x W 7 m2, or roughly 0.2%.

A dust injection system is implemented on the reservoir to inject I O pm cornstarch

Table 3 .1: Conical orifices with small area ratios and varying wall angles.

Table 3.2: Conical onfices with large area ratios and varying wall angles.

Table 3.3: Conicd orifices with srnidi area ratios and varying radii of cumature and wall angles.

CHAPTER 3. EXFERIMENTAL APPARATUSAND METHOD


particles into the tank just pnor to a test because the

49

LDV system requires that the

flow be seeded. The injection system uses a s m d pressure vesse1 (about 1.2 x 10-3 m3) which contains these particies and the test gas at a pressure greater than that of the reservoir. The injection system is isolated frorn the tank by a small needle-valve which ope- momentarily to permit a small amount of cornstarch to be blown into the reservoir at the commencement of a test.

A high-speed rotary vane vacuum pump is connected to the resenroir to evacuate the s used for tank to a pressure of about 1 kPa before reservoir filling. The vacuum pump i ail tests, even when using air as the test gas so that good-quality dry air is used to avoid problems (eg., condensation in the flow, particle wetting and coagulation).

4 gas inlet system with appropriate valves is used to fil1 the reservoir with the test
gas (eg. air, helium). This same fluid is also used to supply the high-pressure gas for the

dust injection system. The operation of the facility is now described. The large gas reservoir is charged with the test gas to the desired pressure after Iirst evacuating the tank of atmospheric air with the vacuum pump.' Once the tank has been pressurized the gas is permitted to reach room temperature by waiting for heat transfer to subside. Some dust is injected into the reservoir Ma the dust injector, and then the flow through the test orifice commences by rapidly opening the bal1 valve. Subsonic f l o w is initiated in the pipe system via the bell-mouthed nozzle inside the reservoir; this flow passes through the area reduction to the test section, then it passes through the test orifice with angled walls and a reduced

final diameter, and then it enters the atmosphere as a subsonic or supersonic fkee jet. Measurements during this quasi-steady blowdown process include the reservoir pressure with the transducer, reservoir temperature with the t hermocouple and the test-section flow velocity with the LDV system. After a short time (eg., 3 s), and aRer only a partial reservoir blowdown, the f l o w is stopped by rapidly closing the b d valve; this conserves the resenroir gas for subsequent experimental runs. A data acquisition system not s h o w in figure 3.1 is used to record and analyse the measurement data.
?Che mervoir couid be reevacuated and refiIIed to minimhe the effof the residual gas, or altemativeiy, the tank couid be purged by nrnning the vacuum pump whiie slowly aliowing test gas to enter the cesemoir.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

3.2.2

Instrumentation

Measurements of the atmospheric pressure, resemir pressure and temperature, and test section flow velocity were recorded during each partial resenroir blowdown test to determine the actuai and ideal mass flow rates. The equipment used for these measurements

is now described.
The arrangement of the LDV system used for measuring the flow velocity at the test

LDV setup is operated in the dual-beam forward modular scatter mode. -411optical components were standard Thermo-Systems h c . (TSI)
section is detailed in figure 3.2. The optics. The laser was a 4 W (total output) Spectra-Physics model 165 Argon-ion laser operating at a wavelength X of 514.5 nm at 1.0 W. The 1.5-mm-diameter laser b e m
was divided into two equally intense paralle1 beams each 12 mm ofhet from the original

beam position by a pnsm type TSI-9115 beam splitter. The spacing of these beams was adjusted slightly by a a TSI-9113 beam spacer to permit entry into a TSI-9188 beam expander. This beam expander increased both the beam separation and their diameten
by a factor of 2.27. The laser beams then entered a TSI-9119 focusing lens with a 600-

mm focal length (actual focal length 604.4 mm through the optic window) which focused these two beams to intersect at a half-angle K of 2.35O inside the test section. The laser beams passed through a set of opticai-quality glass windows and forrned an ellipsoidal probe volume (with a thickness and diameter of 0.116 mm and a length of 1.83 mm) at the centre of the test section. The interference of the two laser beams produce a probe volume with 18 fringes spread 6.27 pm apart.
Light scattered from a particle passing through this probe was collected in the forward on-axis direction and focused by a TSI-9140 receiving optics assembly with a focal length of 249.1 mm (actual252.1 mm through the optic window) onto a pinhole 0.2 mm in diameter at the entry of a TSI-9160photomuitiplier. The amplified signals from this photomultiplier were sent to a TSI-1980counter (minimum velocity 6.4 x m/s, maximum velocity 636.9 m/s) where they were internally passed through both a high- and low-pass filter. The counter output was the freguency of a vaiidated Doppler b u t in the

h i s signal was connected to the digital input port on the data form of a digital signal. T
acquisition board. Additional details on the LDV setup and application c m be found in Linfield [1992]. The test section velocity equipment error can be determined fiom the effect of chang-

ing one-bit on the

TSI counter system, which yields a value of

bu, = 7 ~ 1 0 m/s ~

(calculated for the most common setup). T h i s error takes into account only the counter signal being transmitted to the data acquisition cornputer, and it does not consider other

S I model effects of the LDV system which are supposed to be handled internally by the T
1980 counter.

LDV measurement of the flow velocity

was chosen over other methods

(eg. hot wire probe, pitot tube) because it offers a number of advantages. These include: no flow disturbances, high spatial and temporal resolution, the f l o w velocity is independent of the fluid thermodynamic properties and no calibration is required. The
disadvantages include potential high cost, the requirement of flow seeding (since LDV actually meanires the velocity of small particles suspended in the fluid) and the use of a transparent medium. Calculations were performed to determine the relaxation process

of the seeded flow based on the method detailed in Rudinger [19801. It was determined
that for a worse-case assumption of zero initial particle velocity, the 10 pm cornstarch particles would reach over 99.9% of the test gas velocity (for the most common setup) within a distance of one centimeter at the test section. Since the test section length is approximately 5 cm, it cm be concluded that the seeding particles do not have a velocity significantly different than that of the test gas.

The pressure of the gas in the reservoir was measured with Endevco model 8510-200 or 8510-15 piezo-electric differential pressure transducen, the former for high pressure tests and the latter for low pressure tests. These gauges offer a linear 300-mV hllscale range with a 3.-mm face diameter, and their high sensitivity combined with high resonance make them suitable for measuring dynarnic pressures. The signal from the gauge was amplified and conditioned by an Endevco model 4428 signal conditioner and was transmitted to one of the analogue-to-digital input ports on the data acquisition board. Calibration of the gauges were performed before and after placement in the reservoir to verify the linear relationship between pressure and voltage. The dominant error for the stagnation pressure gauge was determined to be the change of one-bit on the A/D system, yielding bpo = 0.1943 = 0.19 psi for the high pressure (200 psi) gauge and dp, = 0.01454 sz 0.015 psi for the low pressure (15 psi) gauge, or roughly 0.1% of the full-scale range.
The temperature of the gas in the reservoir was measured with an OMEGA unsheathed fine gauge (0.001 inch) type K thermocouple. An OMEGA model MCJ electronic ice point was used to convert the thermocouple voltage into zero-degree referenced

signals. The electronic ice point provided the electricai equivalent of an ice bath reference at OC with an accuracy of f0.5OC. An amplification circuit was designed, tested

and built based on the LT1037 operationai amplifier due to the very amaii voltages proh i s high-quality operational amplifier bas a very s m d duced by the themocouple. T input offset voltage (less than 10 pV) , which was required due to the range of the output voltage produced by the thermocouple. The resulting analogue output was connected to one of the analogue-tedigital converters on the data acquisition board. This probe was calibrated using a variable temperature water bath and the enor was determined to be 6T0 = 0.5 K , which was greater than the effect of changing one-bit on the .4/D data acquisition system, which would result in an error of dTo = 0.04 K . Hence, the enor in the temperature probe was less than 0.2%. The pressure of the atmosphere into which the jet discharges was measured with a Fisher Scientific NOVA mercury barorneter with a vernier scale. The estimated error bp,,, was f0.1 mmHg or about 0.01%.

3.2.3 Data Acquisition and Reduction


The outputs from the velocity, pressure and temperature measuring devices were connected to the input of a Data Translations DT 2821 combination analogue-to-digital (.\ID)and digital input/output (DIO) board attached to an interna1 expansion slot in a personal computer. This board provides up to sixteen A/D ports, each with a 12-bit resolution, and one 16-bit DI0 port. The registers on the board were automaticaily rnapped into the computer's mernory to permit control of the data acquisition process. A progam was written to control the board, obtain the data, convert the input signals into velocity, temperature and pressure, and to store these flow measurernents on the computer's hard drive. Prograrnming the A/D subsystem of the DT2821 board involved initialization of the board, selection of the operating parameters, selection of the input channels, conversion of the two analogue signais int O dimensional quant ities and t hen the processing of the converted digital data. The flow velocity was determined fiom

where fD is the Doppler fiequency obtained from the output of the LDV counter and df6n, is the fiinge spacing in the LDV probe volume. T h e Doppler fiequency was detennined fiom

where the 12-bit digital mantissa Dm and the 4 b i t exponent n were obtained from the LDV counter, and the fringe spacing was determined fiom X dfringD == 6.27 Pm. 2s i nK A computer program was written to study the one-dimensional isentropic discharge of a high pressure vessel, and a cornparison of the predicted and experimental results indicated that the flow from the vessel codd be assumed to be isentropic for approximately the first five seconds of testing. Hence, a quasi-steady mode1 was used to facilitate a fit of the fiow parameters. The pressure and temperature i n the reservoir and the velocity at the test section were measured with respect to tirne ( t )and then each of these was fitted by using the least-squares method with a third-order polynomial. This can be justified by the foilowing: as described in Chan [1990], the stagnation conditions in the reservoir

can be approximated by

for the case of one-dimensional, choked, isentropic flow, where V, is the volume of the reservoir. Equation 3.3 can be written in the form

and the right-hand side of this equation can be expandeci as a binomial series to give

which was approximated by 1+ q t

+q t 2+ c3t3 for these curve fits. The coefficients q in

front of the power series terms were determined by using a least-squares fitting method

and the measured data. The erron associated with each of the flow variables (T,, po) were chosen to be the maximum between the measurement error detailed previously or the standard deviation obtahed from fitting the polynomial to the experimentai data. The , = Q clt c2t2+ c3fl and t he velocity was fit to a t hird order polynomial in the form u error in the flow velocity was also chosen to be the maximum between the measurement error and the standard deviation of the le&-squares fit.

+ +

3.3

Discharge Coefficient Calculation and its Error

The mass flow rate fiom this facility was determined indirectly by analysing measurements of the reservok temperature and presswe and the f l o w velocity upstrearn of the test

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

54

orifice with respect to time. T h i s measurement technique was chosen to avoid flow losses

a s s flow rate meters used in previous experimentai methods, and and disturbances fkom m
also because mass flow rate meten are normally used only in steady flows. The experimental discharge coefficient was calculated by using equation 1.1. The determination of the actual and ideal mass flow rates are detailed below. For uniform flow across the o r i k e exit;

where the jet conditions are taken as the exit conditions of the orifice flow. This equation can be written in terms of the facility measurements as

where the equation of state for a perfect gas, the definition of Mach number and the isentropic reiationships between static and stagnation fiow variables were used in the derivation. The Mach number Mjetin equation 3.4 is determined from
Po Patm
Po -

Patm

(Recall that in one-dimensional flow, choking occurs at the sonic and Ajet = condition for which the flow Mach number is unity.) The actual mass flow rate is determined from the steady state continuity equation in the test section, that is,
~mud =

m,, = C,pts&uo*

The above equation is a slightly modified form of the continuity equation for uniform l o w in the test section is not perfectly uniform across the duct, an flow. Since the f experimentdy determined correction factor C, is used to account for the waU boundazylayer effect, and t is detailed in section 3.4. The equation to determine the actual m a s l o w quantites is flow rate in terms of the measured f

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

55

which was deriveci from the modifieci continuity equation using the equation of state for a perfect gas, isentropic relationships between static and stagnation f l o w variables and a form of the energy equation with no work or heat transfer. The discharge coefficient can be calculated by taking the ratio of equations 3.6 and
3.4 to yield

wliich is expressed i u terms of experiniental and known quantities.


Root-mean-square (RMS) errors are used to estimate the errors in the ideal and actual flow rates and therefore in the discharge coefficient [Kline& McClintock, 19531.

Individual errors are assumed to exhibit a Gaussian distribution. The RMS error for the ideal mass flow rate @en by equation 3.4 can be s h o w to be

where

and

wit h

The RMS error for the actual mass flow rate given by equation 3.6 is

( ) + (ci)

++ (C22 ) ' ,

(3-9)

APPARATUS AND METHOD CHAPTER 3. EXPEEUMENTAL


Findy, the error in the experimental discharge coefficient given by equation 3.7 is

where C3=CL

+ 1.

3.4

Calibration of the Experimental Apparatus

The experimental apparatus was calibrated by comparing discharge coefficient data from

the present experiments with those from other published data and a reference conical orifice. This is required because the flow velocity at the test section was measured only at the centreline, and the m a s flow rate calculated from this maximum velocity and on the bais of the assumption of uniform flow across the duct is not correct (overpredicted)due to the thin wall boundary layer and imperfections in the bell-mouthed inlet nozzle. The
actuai mass flow rate is obtained from the apparent measured mass f l o w rate m (m & ,apparent )

by introducing a correction factor C, as

This correction factor which was used previously in equations 3.6 and 3.7 was determined by comparing the new test results with experimental values published in Alder's Ph.D. thesis [Alder, 19761 and by measuring the discharge coefficient of a reference orifice with a large radius of curvature. Alder calibrated his apparatus by using a convergent-divergent nozzle as a reference orifice, so it was decided to calibrate the new experimental apparatus against Alder's 90" conicai orince results, and then verify that this correction also applies for the case of a 4 5 O conicai orifice as well as to the reference orifice. If this occurs, then the same value of Cuwill be used to obtain ail experimental discharge coefficient values in this thesis. The actuai mass f l o w rate for the experiments in this thesis i s given by

s expressed as hence the discharge coefficient for this work i

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

57

in which the apparent discharge coefficient is based on the apparent mas flow rate. The corrected discharge coefficient should be equal to the reference discharge coefficient; hence

On rearrangement, t herefore,
C d defines how the correction factor can be detennined. A cornparison of Alder's results

cP-= apparent

( C F ~ with ~ ~the ~ new ~ ~uncorrected ) experimentai results (C,Ppvnt) (with sample error bars) is shown in figure 3.3 for the 90' and 45" sharp steel conical orifices, in order to illustrate clearly that the present uncorrected measurements axe slightly higher than Alder's. The correction factor is determined from the data in figure 3.3 as follows: -4Iderk data points were splined together using a piecewise cubic spline and then the ratio of the value obtained from this spline and the new experimental data (Le.C, for the individual data points) was plotted versus the Reynolds number at the test section; this is shown in figure 3.4 for the 90 orifice. From this figure it can be seen that for Re > 70000 the value of Cu is greater than the average value

C, = 0.971 (dashed line), while for Re < 70000

s generally below this line. The vaiue of Re = 30000 is indicated on this figure the data i
(vertical dotted line) as this was the limiting value of the Reynolds nurnber used in the calibration procedure. This limiting value was chosen based on a variety of previous experimental results in the literature which indicated that somewhere between 1x 104 < Re < 5x 104 the discharge coefficient is no longer a huiction of Reynolds number. For a statistical analysis of this data, the Cuvalues were divided into ten compartments or bins,

and the result of this grouping is shown i n figure 3.5 dong with the normalized Gaussian curve. From this figure it can be seen that the data is bimodal about the average value, and that the data is platykurtic (flat-peaked) compared to a Gaussian distribution. To verie the value of C, obtained above, a similar analysis was perfonned with the discharge coefficient data for the 45' orifice. The ratio of Alder's splined 4 5 O data and the uncorrected new experimental results are shown in figure 3.6 and the average value for this orifice i s Cu = 0.973. Rom this figure it c m be seen that the results of the 4 5 O orifice show less spread about this average vaiue than the 90 results show about its average. The grouped statistical r e d t s and the normaiized Gaussian c m e are show0 in figure 3 . 7 . Rom this figure it can be concluded that the 4 5 O orifice results are more

CHAPTER 3. EXFERIALENTAL APPARATUS AND METHOD

58

leptokurtic (sharply peaked) than the results for the 90 oritice. The application of this correction factor (C, = 0.971) to the apparent discharge coefficients for the 45 and 90 sharp orifices (shown in figure 3.3) produces cotrected values that are now compared to

h i s correction factor brings the new corrected discharge Alder's results in figure 3.8. T coefficientsinto good agreement with Alder's data.
The uncorrected discharge coefficient results obtained f r o m the reference conical onfice with a large radius of c m t u r e (Rld = 0.82) is shown in figure 3.9. From this figure it c m be seen that the experimental discharge coefficient is always greater than one, but the results are fairly independent of the pressure ratio. Using the assumption that
reference sz 1, the Cuvalues for the discharge coefficient for this orifice is constant at Cdl the individual data points are plotted versus the Reynolds number at the test section in figure 3.10 dong with its average C, value of 0.971. From this figure a slight Reynolds

number dependence is seen, but to simplify the calibration procedure a constant value of C, i s asnimed. A statistical analysis of this data was performed as detailed for the comparison with Alder's data and the grouped statistical results and a normalized Gaussian curw are shown in figure 3.11. Rom this analysis, the average value of the correction factor was determined to be Cu= 0.971, verifying the results obtained by the cornparison with Alder's data. The calibration statistics for these three orifices (including the number of data points n and the standard deviation o)are summarized in table 3.4. The values of the correction

nf

Orifice 90' sharp

O
-- - -

l
1

4 5 O sharp

129 0 . 9 7 1 0.007 112 0 . 9 7 3 0.003

Table 3.4: Calibration statistics for the 90 and 45" orifices,

= 0.971 and 0.973) determined for the new apparatus are within the same factor (C,
range as those published by other experimentalists. For example, Thornock & Brown [1972]used a c~nvergin~diverging nozzle to caiibrate their apparatus and obtained a correction factor of 0.985. In this work a correction factor of 0 . 9 nis used to correct aU discharge coefficient data fkom the current experiments. The Reynolds number at the test section wari needed previously for figures 3.4 and

APPARATUS AND METHOD CHAPTER3. EXPERMEWTAL

is the test section viscosity. In t e m s of the rneasured and known quantities, equation 3.15 is where

which was determined from the equation of state for a perfect gas, the energy equation with no work or heat transfer, the definition of sound speed and the isentropic relationships between the static and stagnation flow properties. 12 m/s) For most of the onfice tests the flow velocity was relatively small (uo and therefore the static and stagnation values of the pressure and temperature were almost the same. The viscosity of air at the test section is essentially only a function of temperature and was determined from a modified equation of Zucrow & Hoffman [1976]

with coefficients presented in table 3.5. This equation was stated to be accurate within

Table 3.5: Coefficients for the viscosity equation 3.17.


0.25% in the temperature range 270

< T 5 600 K at atmospheric pressure and within

1% at pressures up to 1.2 MPa at 300 K. For equation 3.17, the test section temperature TU was calcdated using a form of the energy equation with no work or heat transfer

as T , and ut,are known kom the measurements.

Some additional confirmation of the correction factor value is presented here by show-

ing cornparisons of some new corrected discharge coefficient data with corresponding data Erom the Literature. A cornparison of Aider's discharge coefficients for the 15,30 and 60"

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

60

conicd orifices for A, IJ O and 7 = 715, dong with the new experimental results are preh i s figure also features the discharge coefficients for the 45 and 9 0 ' sented in figure 3.12. T orifices which were detded previously. The 90 conicd orifice with A, a O and 7 = 7/5 has been experimentally investigated by many authon including Callaghan & Bowden [1949], Perry [1949], Bachmann, Benedict [1970] and Alder [1976].(Note that Bachmann's resuits were taken from Cunningham (19501). A graph of these authors' discharge coefficients dong with those produced in t his study (including the theoreticai curve) are shown in figure 3.13. From these two figures it can be seen that slight deviations between the results from the various experimentalists are evident. It is suspected that the effect

of using orifices with a non-ideai lip (not perfectly sharp edge) accounts for rnost of these discrepancies, which axe larger than would be explained by experimental error. A slight rounding of the lip edge results in dramatic increases in the discharge coefficient. This also helps explain why virtually d l of the experimental data in figure 3.13 lies above the theoretical results (solid line) for a perfectly sharp orifice edge. Results and discussion of the effect of iip rounding on the discharge coefficient are presented in chapter 5.

CHAPTER 3. EXPERMEBITAL APPARATUS AND METHOD

1
transducer @ ) , thermocouple (T,) dust injectot

Note: diagram is not to scale.

angled
orifice - - ---

free

L
reservoir

bal1 valve gas inlet

test

jet

section

LA
PumP

Figure 3.1: Diagram of the reservoir blowdown facility.

magnified sampling probe volume

photomultiplier with a pinhole receiving optics

Figure 3.2: Configuration of the laser Doppler veloeimetry system.

present study Alder [l976j

Figure 3.3: Cornparison of the new uncorrected experimental results with that of Alder.

experimentai raUo - -+ .-: average value lower Re llmit *--*e*ob

----

50000

1OOOOO
Reynolds number

150000

200000

Figure 3.4: Ratio of Alder's results and the new uncorrected results with varying Reynolds
number deterrnined at the test section for the 90' orifice.

0.95

0.98

0.99

Figure 3.5: Statistical distribution of C . for the calibration of the 90 orifice.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

experimental ratio
I

j-

average value lowar Re limit

-* ------:
D*+oee*o

50000

100000

isoooo

Reynolds nurnber

Figure 3 . 6 : Ratio of Alder's results and the new uncorrected results with varying Reynolds number determined at the test section for the 4 5 O orifice.

Figure 3.7: Statistical distribution of C, for the calibration of the 45' orifice.

Figure 3.8: Cornparison of the new calibrated experimentd data with Alder's results.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

Figure 3.9: Uncorrected discharge coefficient results for the Mly-rounded conical orifice

(Rld = 0.82).

50000

1OOOO

150000 200000 Reynols nurnber

exgerirnental ratio average value 250000 300000

----350000

Figure 3.10: Ratio the expected Cdand the experimental results for a 90" conical orifice

having Rld

CS

0.82 with varying Reynolds number determined at the test section.

Figure 3.11: Statistical distribution of

C' for the calibration of the 90 conical orifice

with R/d = 0.82.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

Figure 3.12: Cornparison of new results and Aider's discharge coefficients for the case of a 5 90, A, sr O and 7 = 715.

CHAPTER 3. EXPERIMENTAL APPARATUS AND METHOD

Linfield

Cailaghm & B O Z ~ Bachrnann Benedict 1970 Alder 17 9761 theory

[1gii

a rn

.... .. .. ... .. , ...... ... .., ... .. .. .. ...

..

.. . . ... ..... , . ... .. . . ... . . ....

..

..

.. .....

. .. .

. . ,,

. ., ..

Figure 3.13: Cornparison of various discharge coefficients for a conical orifice with a =
90, A, x O and 7 = 715.

Chapter 4
Cd Algorithm
4 . 1 Introduction
The discharge coefficient for flows through onfices i s a function of many parameters: pressure ratio, wail angle, area ratio, gas specific heat ratio, orifice edge rounding and geometry (for a dot or conical orifice). These dependencies were set out and observed in chapters 1 , 2 and 3. Hence,

in mathematicai notation. For specific values of these parameters, the discharge coefficient can be obtained by analytical methods (chapter l), numericd methods (chapter 2),

or experimental methods (chapter 3). However, these methods are not practical for engineering usage when Cdvalues are required quickly for a variety of different values of the parameters. As a consequence, a semi-empirical method is developed in this chapter to generate or reproduce discharge coefficient values quickly, efficiently and accurately. This method is embodied in a short software subroutine dubbed the "Cd.4lgorithm7', and it is suitable for many engineering applications/purposes for the entire set of parameters with a full range of parameter values.
The formulation of the Cd Aigorithm is developed as a sequence of 2 steps. The f h t step consists of splining theoretical and numerical data in the form of curves of the discharge coefficient versus the pressure ratio for aiI values of cr and y and the other

h i s is defined parameters fixed at specinc values (A, = O, Rld = 0, and x = planar). T


as "standard condition$'. Hence,

Cd= Cd( t a, , Ar = O, y,R/d = O, x = pianar) ,

and Cdfor this set of conditions will be referred to as standard conditions for a standard graph. The second step consists of using an approximate equivalence principle to extend the splined resdts for the standard graph to include the fuil mriations in the parameters A,, R/d and X. These two steps result in a complete spline fit to the discharge coefficient as set out by equation 4.1. For any particuiar orifice, the standard g a p h is produced by fint determining the discharge coefficient at the incompressible pressure ratio by methods detailed in this chapter. This discharge coefficient is subsequently used as the discharge coefficient for the above-mentioned standard conditions and then the equivdent Cd distribution principle (as detailed in section 4.3) is applied to obtain discharge coefficient values over the entire range of pressure ratios.

4.2

Generation of the Standard Graph

A standard graph of Cdversus (1- r ) for cr = 45,90 and 135" with 7 = 715 is shown in . 0 :1.2,1.4 and 1.6 figure 4.1 For the standard conditions. The standard g a p h for 7 = 1 for the 90 dot for the standard conditions is shown in figure 4.2. To reproduce one line on either of these standard graphs (for Lved values of a and 7), three piecewise cubic polynomial splines are used: one to span the range of subcritical flow, one to span the first portion of the regime of supercritical flow, and the last to span the smdl range of choked flow The three piecewise cubic splines for Cdversus (1-r)-or equally r-for 6xed values of u and 7 are determined completely by specifying the following:
incompressible flow limit,

l o w limit , critical f
lower choked f l o w limit, upper choked flow limit. These Cdvalues and their derivatives can be obtained in t e m s of both a and 7 for the

standard graph, so that any particular iine on the standard g a p h c m be reproduced. Expressions for the discharge coefficient and its derivative are now sought.

4.2.1
tion (-

C&=, for the Standard Graph

For a planar dot from an infmite reservoir the continuity equation and momentum equa-

SAp 6

d A = JA u ,

[# - d

~) can ] be combined to yield

l o w in the axial direction [Bragg, 19601,where .4* is for one-dimensional compressible f


, is far downstream of the slot, as seen in the geometric area of the dot and state ( ) 9 figure 4.3. For ondimeusionai incorupressible f l o w ,p = p, = onst, po - p m = 3p0~u;vaii

and po - p3 = ;pou& Also, using the continuity equation f a downstream and the definition of the discharge coefficient for inviscid incompressible Bow Cd= one can express equation 4.3 as

3,

POQ

uW,drldr = 2

(Cdj,=, ~ 2 - 1) .

(4.4)

This exact equation for incompressible flow can be integrated only if uwdias a function of Adi is known. This relationship is not known, hence, an approximate solution for the integral is sought, such that an approximate expression for Cd(p,, can be obtained from equation 4.4. For a slot from an infinite reservoir, as -+ oc the wall velocity can be expressed

using the ratio of radial and traverse flow areas, in which the wall angle is expressed in radians. This farfield expression is now used as an approximation for the relationship

and Adi in both the near- and farfield. Upon substitution, integration and hirt her mathematical manipulation, a semi-empirical first approximation for CdI ,=, is obtained as
between
uwdl

Cdl,~ =

1+

J-

Although this k t approximation reproduces the known solutions at the two extreme wdl angles of O0 (Cd= 1 ) and 180 (Cd= 1/2), the solution deviates &om the numerical

values (CFDcode results) in the midrange (too low), a s shown in figure 4.4. To try to overcome this deficiency, this first approximation is modified sghtly by changing the dependency on a to give
T

The results of this second expression are &O shown in figure 4 4 , and the results from this equation are now slightly higher than those obtained fiom the CFD code, but the shapes of the fimt and second approximations are both very sirnilar to that of the CFD data. T h i s approximation process leads to the following approximate expression
1

in which the exponent /3 is chosen such that this expression best fits the CFD code

results shown in figure 4.4. A least-squares cime fit to the CFD code data produces a value of B = 1.099. Alternately, the value of /3 could be determined on the basis that

& for

= 90. In this latter case,

which is very similar to the value obtained from the non-linenr curve fit. The results of

= 1.15) are also shown in figure 4.4 and they are very good compared to the CFD code results (largest error in the discharge coefficient is less
this third approximation (with than 1%).This third and final approximation

i s used in the Cd Aigorithm to determine the discharge coefficient for incompressible - in the Iimiting case of incompressible Bow. Note that the discharge coefficient Cdj,-, flow depends on the dot angle but not the gas specific heat ratio.

The slope

t angent-gas t heory [Shapiro, 19531 in conjunction wit h results presented previously in

acd at the incompressible pressure ratio is determined by using results from dr

section 4.2.1 for Cdl,=,. T h i si s possible because an approximate pressuredensity reIationship fiom the tangent-gas theory in the form of a linear relationship between the pressure and the specific volume d o w s one to construct slightly compressible Bow solutions from knom incompressible ones. These resdta for the discharge coefficient for slightiy compressible f l o w me pressure dependent, which can then be differentiated to give

;? ,1

T h e isentropic relationship between the pressure and density p = p, (p/pJT is replaced


by its t angent-gas version [Gurevich, 19651:

This approximate equation of state is central to yielding a much shorter approximate, hodogaph equation in t e m s of the stream b c t i o n (see Gurevich [1965] for more details).

The resulting hodograph equation can be solved more easily and analyticdy by using complex potential theory. Busemann [1937] used such tangent-gas equations to derive
90 c d l t s l

2 Pjet 1+--

for a 90" slot in an infinite reservoir. Busemana's results can be viewed as an extension of

cdlEl from incompressible f l o w to the case of slightly compressible flow.


h

The tangent-gas results are now used to extend equation 4.9 for the discharge coefficient for incompressible f l o w to the case of slightly compressible Bow, that is This is done in one step by writing
1

such that Busemaan's slightly compressible results are recovered for the case of a = r/2.

The density ratio can be replaced in t e m s of the pressure ratio and the gas specific heat ratio by rearranging equation 4.10 to yield Pjet Y 7 =
Po

7 + 1 - o i Po *-

7+1-v

and this last equation is substituted into equation 4.12 to yield

It is assumed that this single correction for slightb compressible f l o w applies over the entire range of a. Results for versus (1 - r ) are shown in figure 4.5, where they are compared to the exact results obtained by Falkovich's method, and these show that
N

this expression is very accurate for 1 2 r 3 0.9.


m now The derivative of the discharge coefficient with respect to the pressure ratio c

be detennined from equation 4.14 to produce

In the incompressible limit

which provides the slope at the incompressible pressure ratio for the standard graph. The values of
1.4 and 1.6 in

from equation 4.16 are plotted versus the wall angle for y = 1.2, dong with the exact results fiorn Fnlkovich's solution. This

cornparison shows that the approximate results for values.

are very close to the exact

4.2.3

for the Standard Graph

The discharge coefficient at the critical pressure ratio was found to be dicectly proportional (linear) with the gas specific heat ratio for a constant wdl angle. The CFD code was run for eleven values of the wall angle and seven values of the specific heat ratio to produce le&-squares fit equations of the form

where P is the normalized wall angle (a/*). The values of A and B were plotted against the wall angle and a polynomial least-squares fit was determined for these coefficients. These fits were forced through the limiting value for an open chamel: for ii = O the slope

B must be zero and the intercept A must be 1. With these limits, the best-fit equations
were determined to be

and

The discharge coefficient at the critical pressure ratio as determined from the CFD code dong with the iinear Ieast-squares fit nes are shown in figure 4.7, where the h e a r relationship between the discharge coefficient and the specific heat ratio is evident for alI values of the wall angle.

The dope -at the critical pressure ratio was detemhed numerically using three-point dr one-sided differencing from three ruos of the subcritical portion of the CFD cornputer code. The dope was then fit to the empirical equation

dCd

which can be d t t e n as the linear equation

where the parameters A and B were calculated to be

and

based on a linear least-squares fit. The results f r o m equation 4.18 dong with the CFD code data is presented in figure 4.8. The least-squares fit i s fairly accurate throughout the range of the wall angle, with the maximum deviation occurring for wall angles greater

than 150. Al1 three Lines shown in the figure have a vertical intercept of 0.1549 as for 6 + O the derivative dso tends to zero but the ratio of the t w o values is not indeterminant.

4~2~5 c d

and Cdl,=O for the Standard Graph

The discharge coefficient at the choked pressure ratio was determined to be directly proportional (iinear) with the gas specific heat ratio for a constant wall angle, sirnilar to that seen at the cntical pressure ratio. The CFD code was run for 21 different wall angles to provide data for a least-squares fit of the equation

Equations for A and B were determuid kom Ieast-squares fits of A versus d and B versus 6. One limit for each fit is aiready known- for an open channel o! = O, Cd = 1

for al1 7, and the dope B = O. With these limits, the least-squares fit equations were determined to be

A = 1 - 0.27026 and

+ 0.1322a2

B = 0.0488ii - 0 . 0 6 1 2 4 ~ +~0 . 0 4 8 0 5 ~ ~ .
Some of the choked discharge coefficients from the CFD code are plotted versus the gas specific heat ratio in figure 4.9 dong with the least-squares equations where the lineax
relationsliip is clearly seen.

For pressure ratios less than the choked pressure ratio (including the theoretical limit
r = 0) the discharge coefficient is no longer a function of the pressure ratio and hence

for O 5 F

5 r*. the discharge coefficient is constant at the choked value:

At pressure ratios less than pressure ratio. Hence

and

r =r**

-1

ap

= .,

for the Standard Graph

p..,

9Ir=,-.

the discharge coefficient is no longer a function of the and are both zero for al1 slot configurations and

9Ir=,

flow conditions:

4.3

Equivalent Cd Distribution Principle

The previous section detailed how the standard graph cm be produced for the standard conditions expressed by equation 4.2. The additionai effects of the area ratio, onfice edge rounding ratio and axkymmetric f l o w are incorporated into the Cd Algonehm by using an approximate equivalence principle. This principle is now set out and explained.

Principle 4.3.1 Equivalent Cd Distribution Principle: The dutribution of Cd versw (1 - F ) for a particukar onfice (with specific values of a, A, and R/d) is equivalent to the distribution of the standard gmph Ca= ( r ,a, A, = 0,7, R/d = 0, x = planar) that shares the same discharge coeficient for incompressble flow.
This principle implies that if two orifices of different physicai geometries have the same discharge coefficient for incompressible flow, they will produce the same standard graph

for cornmon fluid properties. Therefore, the Cd Algorithm concentrates on determining

= 1 for all the various orifice geometries (Ar, Rld) and then using this value to produce the standard graph. I n order to apply the relationships presented in section 4.2, the input parameters must be expressed as standard conditions (Le. A, = O, x = plana, Rld = O). Once the discharge coefficient for incompressible flow i s known for a particular orifice, the wall angle at which these standard conditions apply is required to detennine the spiine parameters. Equation 4.9 c m be rearranged to give
the discharge coefficient at
t

and an iterative process is required to determine the wall angle a for the standard conditions. Preliminary studies showed that the equivalent Cd distribution principle was applica-

ble to the complete range of the discharge coefficient parmeters. A sampie of the error introduced by applying this principle is presented in figure 4.10 comparing the discharge
coefficient at the critical pressure ratio obtained by Falkovich's solution method for the case of a 9 0 ' slot with 7 = 715 and varying the area ratio, and the case of a dot under standard conditions (A, = O) with the wall angle varying. Rom this graph it can be seen that the maximum deviation occun at a large area ratio but the error rapidly decreases

as the area ratio approached one. It can be seen that for this example the equivalent Cd distribution principle produces discharge coefficients which are less than 0.35% large.

4.4

Obtaining CdI V - ,- for Nonstandard Conditions

Using the equivalent Cddistribution principle permits the effects of a l l orifice geometries to be incorporated into the Cd Algorithm. In this section, details on determining the discharge coefficient for incompressible flow to account for the effect of the area ratio, axisymmetric flow through a conical oriftce, edge radius of curvature-to-diameter ratio and the experimentd d e c t of large area ratios are presented.

4.4.1

c&!!!"for a Non-Zero Area Ratio

Potential theory presented in Gurevich [I9651 shows that the discharge coefficient for incompressible Bow with a varying area ratio can be determined by the equations

and

where h is an input parameter used to determine the area ratio. For a given value of a

and A, a value of h is guessed and an iterative process is undertaken to pick a value of h which will yield the chosen value of A, f r o m equation 4.24. Once the second equation is satisfied, the discharge coefficient is determined from equation 4.23. In the case when & = 1 the solution is analytic and is given by

The three integrals in equation 4.24 can be solved by analytical rather than numerical methods. Each integral can be written in the form

where b = (1,1,-2) and a = (h, f , 1). If one lets d = P - where p and q are integer constants and q 2 p, then one can substitute x = 24 and dx = qrq-Ldz which yields
CI -4

where f = af . This integral can be replaced by one of two expressions. If q is even, let
q - p = r and q = 2711, and the integral can be replaced by

2712 %-*

2
k=l

eos ((2k

ii) *r) h

(z2

- 2fz cos

- 1) ) + 12) . ((2k2m
T

(4.26)

If q is odd, let q - p = r and q = 2m + 1, and the integral is

Since the summation in the last two equations is from 1to rn (where m is generally srnail) determining the solution of the three integrals as summations i s quite f't-on the order of a fraction of a second on a curent personal cornputer. For this solution method, the wall angle m u t be converted to a ratio of integen. This process is not overly difncult for most wa angles and a srnall routine was written to convert oi into a rational fraction.
plnar The gmph obtained by applying the above procedure to determine Cd= Cdlr=l as a function of area ratio can be seen in figure 4.11, where the results from the summation equations and the exact results obtained by Falkovich's solution are identical as expected

(since this derivation of the discharge coefficient is exact as it originates from complex potential theory).

4.4.2

for a Zero Area Ratio

For axisymmetric flow through a conicai orifice from an infinite reservoir, determining

the discharge coefficient for incompressible plana flow through a slot presented in section 4.2.1 can be undertaken. For incompressible a,xisjmmetric flow, it can be shown that as Ad, -+ oc the wall velocity can be expressed as

which, following the procedure described in section 4.2.1, yields a first approximation for

As in the planar case, although this equation reproduces the known solution at the two

extreme w d angles of O* (Cd= 1) and 180" (Cd= 1/2), the solution deviates from the numericd values determined fkom the CFD code (approximation is too low), as seen in figure 1.12. The power on the cu term can be written as ,LI which leads to the following semi-ernpiricd expression

A second approximation with @ = 1 produces a much better fit, as seen in figure 4.12, although this curve is slightly lower than the CFD code results for a > 45". A third approximation for B can be determined by matching the discharge coefficient of a 90" a=9O0 orifice. Based on the numerical value of Cdl,, = 0.59137 determined fiom the CFD code, , = 0.94 and so

is the third and final approximation used to determine the discharge coefficient for conical

orifices at the incompressible pressure ratio. The application of ail three approximate equations is seen in figure 4.12, dong with the results fiom the CFD code. From this figure, it can be seen that the final approximate equation is slightIy higher than the numerical results for cr < 9 0 ' and slightly lower than the numerical results For cr > 90".

4.4.3

~~12, f o r a Non-Zero Area Ratio

The discharge coefficient for conical onfices with a non-zero area ratio can be determined from

where ACd is given by

This formulation was chosen based on theory which indicates that for the two extreme wall angles (O and 180) the plana and axisymmetric discharge coefficients are the same. Additionally, the maximum deviation between planar and axisymmetric flow was determined from the CFD code to occur at A, = O and the theoretical minimum occurs at A, = 1 (at which point the discharge coefficient for both planar and axiisymmetric flow i s one). The CFD code was run for both planar and axisymmetric flows for four dinerent area ratios: 0.0,0.25,0.5 and 0.75 and the ciifference between the plana and axisymmetric discharge coefficients were calculateci. These clifferences were fit to a fifth-order

polynomid given by equation 4.33 and the five coefficients of d were determined to be

where the value of A was h e d at zero (since for an open channel a! = O and Cd = 1) and

B was chosen to be a constant based on the numerical results. Values of ACd for the four dBerent area ratios can be seen in figure 4.13 dong with the least-squares fit results. From this figure it can be seen that al1 curves shaxe a similar slope at a = Oo, justifying the choice of B = const. Since the value of the planar discharge coefficient bas been shown to always be greater or equal to that of the axisymmetric discharge coefficient, if the least-squares fit equations chosen to reproduce the data produce ACd < 0, the difference between the two discharge coefficients is chosen to be zero (i.e. ACd = 0).

4.4.4

Cd(,=,for Conical Orifices with Edge Rounding

The discharge coefficient for incompressible flow for conical orifices with edge rounding

is determined by first calculating the "equivalent" angle which is defined as the angle produced by applying equation 4.22 to a known value of the discharge coefficient. Since
%

there is a one-toone correspondence between & and Cd(t,, for the standard conditions, knowing t 5 is equivalent to knowing the discharge coefficient at the incompressible limit. Experiments were performed on orifices with three wall angles (a= 45,90 and 135') and dinerent R/d values as detailed in chapter 3. From these experimental results, the empirical relation

&=a-Acr
was chosen, with al1 angles expressed in radians. The difEerence between the equivalent

angle and the conical orifice angle can be expressed as

where

which was determined by a le&-squared fit. Note that the coefficient in front of the a2term has units of inverse radians as the dimension of A is radians. The least-squares fits using equation 4.35 are shown in figure! 4.14 along with data points obtained Erom the experimental results detailed in chapter 3 and additional data obtained by a summ e r student (Haxmony Smith) in the gasdynamics group at the University of Toronto Institute for Aerospace Studies. The least-squares fit equations were forced through the origin because if R/d = O it is expected that the experimentd and theoretical discharge coefficients are the same, and the upper ASME limit of Rld = 0.82 was used as a data point in the fitting analysis. Once & i s known from the above analysis, the discharge coefficient for incompressible fiow can be determined by the substitution of into equation 4.9.

4.4.5

Cd(,=1 for Experimental Area Ratio Effect

Since the experimental results often differ from the theoretical results significantly, an experimental parameter has been incorporated into the CdAlgorithm to more accurately reproduce the erperirnental results. A similar method to thnt proposed in section 4.4.4 for determining the discharge coefficient of conical orifices with edge rounding was used to determine the effect of large experimental area ratios. As detailed above, the equident angle

can be determined which is subsequently used in equation 4.9 to determine the discharge coefficient for incompressible flow. From the experimentai tests results on conical orifices with large area ratios (detailed in chapter 3), an empiricd relationship was determined to permit the Cd Algorithm to better reproduce the standard graph. Rom the new experimental data the empirical relation

ii=a-~a!
angle and the orifice angle can be expressed as

(4.36)

was chosen, with d angles expressed in radians. The ciifference between the equivalent

where

is a piecewise wali angle fitting function. Note that the first tenu in the equation for

B has the units of radians. An insufncient number of experiments were performed to

v e w this celationship, as evident in figure 4.15. The fitting lines were forced through the ongin as for an infinite reservoir with a perfectly sharp lip, it would be expected that the experirnentai and theoretical discharge coefficients would be identical.

.4gain, once ii is known from the above analysis, the discharge coefficient for incompressible flow can be determined by the substitution of I into equation 4.9.

0.9

0.8

0"
O.?
...............................

pressure ratio locus (rH)


0.6
..............

, ....

.C...*..*.....-L

1350

. . * .....-*

-..-..

................... , . .

. . . . . . . . .

r .

subcriUc&l flow

supeicrlual flow

0.5

Figure 4.1: Example of the standard discharge coefficient graph with 7 = 715 and V wall angles.
0.9
I

~ ~ O U S

pressure mu0 taws (r..)


0.85

0.8

0"

0.75

0.7

0.65

0.6

Figure 4.2: Example of the standard discharge coefficient graph with a = 90' and various specific heat ratios.

Figure 4.3: Notation for incompressible f l o w through a d o t in an infinite reservoir.

first a~proximatioRsecond approximation ------third approximation - - - - - - - CFD code results

Figure 4 . 4 : Approximations and numerical results For the discharge coefficient for planar

flow through a slot at r = 1.

slom from Fdkuvich dope from approximation

-------

20

40

60

80 100 wall angle a (deg)

120

140

160

180

Figure 4.6: Cornparison of methods to obtain the dope

&i at the critical pressure ratio.

.........

Angle

Note: lines &e at every 15'

CF0 code results


least-squares fit
1 t
1
I

1.1

1.2

1.3

? -4

1.5

1.6

1.7

1-8

specific heat ratio y

Figure 4 . 7 : CFD code results and the le&-squares fits for planar flow for

Figure 4 . 8 : CFD code results and the least-squares fits for determinhg planar flow, where W =

for

0.7

CFD code results least-squares fit


1
1.1

r, I
I

1.2

t .3

1-4 1.5 specific heat ratio y

1.6

1.7

1.8

Figure 4 . 9 :

CFD code results and the least-squares fits for planar Bow for

0.4

0.6

area ratio A,

Figure 4 . 1 0 : Error (%) generated from Falkovich's solution rnethod between

C'dl,.

determined by applying the equivdent Cddistribution prnciple (with A, = O and variable

a) and the actual results obtained for a 90 dot with 7 = 715 and varying the area ratio.

sumrnation solution Falkovich's solution

-------

Angle

0.4

0.6

area ratio A,

Figure 4.11: Cornparison of Cdl,, obtained from Falkovich's solution and the summation solution.

first approximation second approximation ------third approximation -- - CFD code results


a

20

40

60

80 100 waJl angle (deg)

120

140

160

180

Figure 4 . 1 2 : Approximation values and numericd results for the discharge coefficient for incompressible Bow through a conicd orifice.

20

40

60

80 100 wall angle (deg)

120

140

160

180

Figure 4 . 1 3 : Differences between the dot and conical orifice Cdl,=, b r different area ratios.

Figure 4.14: Experimental edge-effects results and least-squares fits for the equivaient Cd principle.

0.1

0.2

0 . 3

0.4 area ratio A,

0.5

0.6

0.7

0.8

Figure 4.15: Experimental area effect results and linear equations for the equivalent Cd principle.

Chapter 5

Results and Discussion

Introduction

The CdAlgorithm, which was detailed in chapter 4, provides a quick and accurate method to obtain the standard graph of the discharge coefficient Cdversus (1 - T ) as a function of wall angle, pressure ratio, area ratio, gas specific heat ratio and experimental radius of-curvature ratio. The CdAlgorithm uses the equident Cd distribution principle which was shown previously to produce accurate data. Results and discussion of the discharge
coefficients obtained from the Cd .4lgorithm are presented herein. Initially, cornparisons are made to theory in the form of numerical results from the CFD code, Falkovich's solution method, and some data from the literature. Then, the Cd Algorithm is compared to other approximate methods fiom the literature which produce the discharge coefficient distribution gaph. Due to the number of test cases that codd be chosen, selective cornparisons have been presented with independent variations in the wall angle, gas specific heat ratio, the exit-to-channel area ratio and type of flow (whether the flow is planar or axisymrnetric). Cornparisons are made to the new experiniental discharge coefficient results obtained from the facility described in chapter 3, as weil as some experimentai

discharge coefficients fiom the iiterature. Finaily, an explanation is presented on why the experimental discharge coefficient i s greater than that predicted by theory.

CHAPTER 5.

RESULTS AND

DISCUSSION

5.2

Cd Algorithm Matching Theoretical Results

The Cd Algorithm considen the effects of four variables when reproducing the theoretical Cddistribution Cumes:

Cd = C d (a, Ar, 7, X)

The results and discussions presented in this section are produced by varying one of these parameters while keeping the others constant. In this manner the effect of that one variable can clearly be seen.

5.2.1

Wall Angle Effect

The discharge coefficient produced by the Cd Algonthm with various wall angles for A, = 0, 7 = 715 and x = plan= are shown in figure 5.1 for a full range of a, dong with results from Falkovich's solution method, CFD code results (for critical and choked pressure ratios) and some supercritical results from Alder [1976].From this figure it c m be seen that the CdAlgorithm is very accurate at the incompressible flow limit for al1 r d 1 angles and only a slight deviation fiom the subcntical numerical results is seen at the extreme w a l l angles where the results at the critical pressure ratio appear to be slightly small wall angles high-0.8%-for large wall angles (180) and slightly low-0.2%-for (5 and Io). In the supercritical flow regirne the results from the Cd Mgonthm match the 8 0 ' case) but it produces choked values fairly well (minor deviation are noted for the 1 results that are slightly higher than Alder's data for the 90' slot but very accurate matches the numerical values compared to Aider's 30" results. Overall, the CdAlg~nthm shown on this graph quite well.

5.2.2

Area Ratio Effect

The discharge coefficient r e d t s of varying the area ratio are presented in figure 5.2 for planar flow with a = 90 and y = 715. From this figure it can be seen that for area ratios l e s than or equal to 0.5 the Cd Algorithm matches the numerical results quite weIl. For larger area ratios the CdAlgorithm overpredicts the subcntical values and underpredicts the discharge coefficient at the choking condition. For euample, at A, = 0.95, the subcritical results are 0.5% too hi& while the choked results are 0.6% too
Iow.

CHAPTER 5. RESULTS AND DISCUSSION

5.2.3

Specific Heat Ratio Effect

The discharge coefficients for a 90 slot from an infinite reservoir with varying gas specific heat ratio are presented in figure 5.3. Rom this figure it can be seen that the subcritical portion of the graph is matched quite well, as does the discharge coefficient at the choked pressure ratio (maximumerror 0.2%). The curve for 7 = 1.4 shows slight deviation from Aider's data for supercritical flow (maximum error of 0.5%) but overall the CdAlgorithm matches the numerical results quite well .

5.2.4

Planar and Axisymmetric Flow Effect

The results of the Cd Algorithm are shown for four wail angles in figure 5.4 for planar

and axisymmetric f l o w with Ar = O and y = 715, dong with some discharge coefficient
values kom the CFD code and some from Alder. Rom this figure it cm be seen that the

CdAlgorithm matches both the planar and axisymmetric values in the subcritical flow
range quite well and matches the planar values in the supercritical flow range fairly well. Deviations for axisymrnetric flow are seen in the supercritical and choked flow regimes, where the Cd -4lgorithm produces results which are at most 0.9% high.

5.2.5

Notes on the Approximations in the Cd Algorithm

4 sseen in the previous four figures, the Cd Algorithm reproduces the theoretical d i s charge coefficient distribution for plenar and axisymmetric flow with varying wdl angle,
area ratio and gas specific heat ratio with a maximum error of 0.9% (occurring for a x i symmetric Bow at the choking pressure ratio). From these figures, the confidence level of the various approximations made in chapter 4 can now be considered. The reproduction

of the discharge coefficient Cd(,l and the derivative at the incompressible limit are quite accurate, as detailed in chapter 4 and seen in the previous graphs.

%I,

Use was made of the near relationship between the discharge coefficient and the gas
specific heat ratio at both the critical and choked pressure ratios, but the polynomial
curve fits could be improved or better relationships which would relate the discharge

coefficient at these pressure ratios to the discharge coefficient at the incompressible pres-

sure limit could possibly be determined. However, the polynomial fits are fairly accurate for engineering applications. The derivative of the discharge coefficient at the critical could be improved as it is suspected that this causes most of the pressure ratio SInrn

discrepancies in matching the discharge coefficient values in the supercritical flow regime. Various attempts were made to produce a better relationship that was still fairly simple but the current method was chosen due to its simplicity and reasonable accuracy.

5.3

Cd Algorithm Compared to Other Algorithms

A few approximate methods are available from the literature to obtain the discharge coefficient distribution for the standard graph. Two approximate methods which cover most onfice configurations were detailed briefly in chapter 1: the subcritical method of Buckingham [1931]and its extension into the supercriticai flow regime by Cunningham [1950] was derived for the specinc case of flow fiom a channel with a = 90, while the method of Jobson [1955] (and its modification by Bragg [1960]) was derived by considering the flow from an infinite reservoir. Both of t hese methods require the discharge coefficient at the incompressible limit as an input parameter to produce the graph of Cdversus r for different gas specific heat ratios. A cornparison of the discharge coefficient results from the CdAlgorithm with that obtained by Buckingham's method (and Cunningham's r o m Falkovich's solution and the CFD code are extension), along with some results f shown in figure 5.5 for the 90" slot with 7 = 715 and three different area ratio (including the limiting case of A, = O). From this figure it can be seen that for small area ratios the subcritical results of Falkovich are reproduced well by both algorithms, but Cunningham's supercritical extension does not predict the rnultidimensional choking condition indicated on the figure by the CFD code results. As the area ratio increases, Buckingham's method produces results which are greater than the exact numerical results of Falkovich, and this discrepancy increases as the area ratio increases. The Cd Algorithm is much more consistent in following Falkovich's solution results and predicts the flow regime where the discharge coefficient is no longer a function of the pressure ratio (i.e. choked conditions). Cunningham's supercritical extension does not predict this choking condition for any area ratio due to the one-dimensional flow approximations used in his algorithm. The results of Jobson's method (and its modification by Bragg) are shown in figure 5.6 for slots h m an infinite reservoir with 7 = 715 and various w d angles dong with selected results from the CFD cornputer code a t r = 1, r = r. and r = r,.. Bragg's modification offer a v a t Unprovernent over Jobson's method, the latter of which predicts that the discharge coefficient decreases with increasing pressure ratio for smali wail angles. This deviation was noted by Jobson b e l ,who recommended that his algorithm o d y be used

CHAPTER 5.

RESULTS AND

DISCUSSION

104

when Cd(,l < 0.7. Bragg's modification of Jobson's method corrects this shortcoming,

and it can be seen that his results match the subcritical portion of the graph quite well for al1 wall angles. However, the choked portion of the graph is not predicted due to the one-dimensional assumptions made in his derivation. Compared to the Cd Algorithm,
Bragg's results are slightly higher in the subcnticd range and are significantly higher in the choked flow region.

-4 more recent algorithm for axisymmetric flow through conical orifices has been

published by Parker & Kercher [1991]. Their method i s stated to be vdid only for a = 90, A, 5 0.25, and includes a correction factor for the effects of orifice edge rounding. The results obtained by their method and the Cd Aigorithm are presented in figure 5.7 For

a sharpiipped conical orifice with 7 = 715, A, = O and 0.25, along with data from the CFD code at the incompressible, critical and choked flow pressure ratios. From this figure it can be seen that Parker & Kercher's method is not consistent as the two lines for different area ratios cross in the supercritical flow region and the results for A, = 0.25 deviate significantly nom the numerical values. A cornparison of the results fiom the CdAlgorithm with that obtained by Parker & Kercher's method for conical orifices with edge rounding is shown in figure 5.8 along with new experimental data. Rom this figure it can be seen that the Parker & Kercher's method underpredicts the effect of deviations From a sharp Lip and their method faiis to predict the new experimental data.

The time required to obtain the graphs by any of these approxmate methods is very short on present-day personal cornputers, although this would vary slightly with the programming and compiler options. For example, the time to produce data for 100 lines

on a graph (each line consisting of 100 data pain equally spaced in the pressure ratio) by the Cd Algorithm-including reading the input conditions fiom a file and storing the results on the cornputer's hard dnve-took about one second on a Pentium 233 MHz personal computer. Under the same conditions, a combined Buckingham/Cunningham algorithm took between 9 and 17 seconds (dependhg on the value of the input parameters), Jobson's method required about one second, and Bragg's method required about 87 seconds. Bragg's algorithm requires iterations at each pressure ratio to determine the s the reason this method requires the most t h e . discharge co&cient, which i

CHAPTER 5. BSULTS AND DISCUSSION

5.4

Cd Algorithm Matching Experimental Results

Experimental and theoretical discharge coefficients often m e r appreciably as detailed in the literature [Alder, 19761 [Benson k Pool, 19651. Much of these deviations are a result of the fact that an orifice can not practically have a perfectly sharp edge as some blunting or rounding is always present. To account for the discrepancies between theoretical and experimentai discharge coefficient values, the CdAlgonthm considen the separate effects

of two variables when reproducing the evperimental Cddistribution curves: the efTect of edge rounding in the form of the ratio of the the radius-of-curvature to a i t diameter (for small area ratios) and the effects of a large area ratio. Unfortunately, insufficient data exists to account for both of these parameters simultaneously. The details of the experimental apparatus and method were presented in chapter 3.

5.4.1

Edge Rounding Effects

The equivalent discharge coefficient distribution principle was applied to the series of conical orifices with srnail area ratios (A, n 0.06), 7 = 715, smail edge rounding ratios, and with the wall angle varying every 15' fiorn a = 15' to a = 150' as detailed in table 3.1. The experimental discharge coefficient results and curves obtained from the
CdAlgorithm are shown in figures 5.9 and 5.10. As detailed in chapter 3, oniy some of

the orifices were segmented to determine the radius of cwature and hence Rfd. For the orifices not segmented, values of Rld were assumed as follows: for rr < 90, Rld

= 0.003,

and for a > 90, R / d n 0.005. These values were chosen based on the orifices which were segmented, dong with the orifice manufacturer's comments that a sharp lip was more difficult to create on the orifices with wdl angles greater than 90'. From these two figures it c m be seen that the CdAlgorithm reproduces the experimental data fairly well, although for (1 - r ) > 0.8 the approximate results are generally slightly high (maximum deviation is about 2%), while for (1 - r ) < 0.2 the Cd Algorithm is oRen slightly low (maximum deviation approximately 2%). The reproduction of the 75" and 105' orifices deviate more than the others, which most lkely is a resuit of an incorrect assumed value of Rld. The pressure ratio for the choked flow condition c m be seen in the evperimentd data for orifices with wall angles less than 60'. This value appears to be in good agreement with that predicted by the CdAlgorithm.
A cornparison of the Cd Algorithm with new experimental resuits with 7 = 715,

CHAPTER 5. REsWLTS AND DISCUSSION

106

are presented in figures 5.11 to 5.13 for the 45, 90 and 135* orifices respectively. Rom these figures it is evident that small changes in the lip geometry can have a large effect on the discharge coefficient, and this effect increases with increasing waiI angle. The CdAlgorithm reproduces the experimental data favly well, although it produces discharge coefficients that are often greater than that obtained experimentaliy (maximum deviation is about

4 n 0.06 and various values of R/d (tabulated in table 3.1 and table 3.3)

2.5%). With more experimental data on orifices with different values of Rld, a better relationship to predict the discharge coefficient a t the incompressible pressure ratio for al1 values of Rld could be obtained.

5.4.2

Experimental Area.Ratio Effect

The discharge coefficient produced by the CdAlgorithm is compared to new experimental results in figure 5.14 for the case of 7 = 7/5, A, = 0.7 and varying wall angle. Since these experimental values were used in the development of the Cd Algorithm, the resulting fit must be reasonable. An additional onfice with a = 135' and A, = 0.4 was tested to aid in the validation of the chosen relationship for application in the Cd Algonthm and its discharge coefficient results are shown in figure 5.15. Rom this figure, it can be seen that the CdAlgorithm produces a curve which i s at m o s t 3% high, although a fair amount of scatter is seen in the experimental results due to the higher velocities encountered at the test section resulting from the orifice's large area ratio. Comparing the results of the Cd Algorithm to other experimental results fiom the literature is made difficdt since most of the published data does not detail how sharp the orifice edge was. However, the results from the Cd Algorithm compared to experimental

r o m Thornock & Brown [1972] and Grey & Wilsted [1949] are given in figures 5.16 redts f and 5.17 for 7 = 715 and dinerent waU angles and area ratios. Rom these figures it can
be seen that the CdAigorithm predicts the resdts fairly weLl in most cases, although it
siightly overpredicts the results fkom Thornock & Brown (maximumdeviation is about
2%) -

CHAPTER 5. R~~SULTS AND DISCUSSION

1 0 7

5.5
5.5.1

Notes on the Experimental Discharge Coefficient


Lip Geometry

A tmly sharp lip i s an ideal which cannot be achieved in practice-there is dways some blunting which will affect the point of flow separation. The deviations fiom a sharp lip

can take any form but are classicaily divided into a thickness and/or a radius of c w a t u r e A number of publications on the thick orifice plate have appeared including W d S m i t h (19791 and Deckker & Chang [1965], but little information is available on the eEect of edge rounding for compressible flows. The International Standards Organization (ISO) has defined a sharp lip for an orifice flow meter as one which satisfies

PSO,19801. The only orifices tested which satisfies this criterion were the 4 5 O and 90 steel orifices, which is the reason they were used to calibrate the experimental apparatus. An orifice with inlet corner rounding will have a higher discharge coefficient as a result of reduced flow sepmation at the upstream idet edge. The experimental results performed in this thesis have shown that even a small amount of blunting c m dramatically increase the discharge coefficient. The ASME indicates that the discherge coefficient does not change beyond the upper limit of R/d = 0.82 and suggests that Cd= 0.9975 for this R/d value for inviscid incompressible Bow through a 90' conical orifice [McGreehan 9t Schotsch, 19871. Practically, this definition can be extended to al1 flows with the assumption that Cd = 1 for R / d 2 0.82.

A number of orifices were manufactured with a prespecified radius-of-curvature and after testing was complete the orifices were given to Integrity Testing Laboratory Inc. where they were segmented and a portion of the lip encased in epoxy. This piece was then buffed and photographed under an electron microscope. Rom these pictures an approxirnate radius of c m t u r e wa9 determined. The summary of the orifices used for these tests was preaented in table 3.3.
A number of equations exist in the literatare to predict the effect of edge rounding, although most are empirical and only applicable to 90 orifices for incompressible f l o w . To be complete, both the lower and upper mits must be matched: when Rld + O the sharp-lipped value should be realized, and as R/d + 0.82 the discharge coefficient should tend to one. Alder [19?6] used the r d t s fkom his theoretical solution to derive

CHAPTER 5. RESULTSAND DISCUSSION


the equation

s applicable for supercritical flow in the limit as R/d which i

+ O, where x = O for planar

fiow and x = 1for axidy symmetric flow. Although this equation does consider different

wall angles as weli as both common fiow types, its restrictions on the pressure ratio and the s m d R/d values limit its application. Empincal relations based on experiments for incompressible flow through 90' orifices are given in Crockett & Upp [1973], Benedict et al [1975], and McGreehan & Schotsch [1987], although only the latter's equation

with f = 0.008

+ 0.992 exp (-5.5 9 - 3.5 (ff) ') has the correct limits.

The equivalent

Cddistribution principle detailed in chapter 4 uses a combination of a logaxithmic reiationship between the orifice wall angle and R/d and then requires an iterative procedure to obtain the discharge coefficient at the incompressible pressure ratio as detailed in chapter 4.

5.5.2

Viscous Effects
the discharge coefficient as it can resdt in rotational flow due to

Viscosity can &ect

the thin boundary layer aext to the walls and the interaction of the free jet with the atmosphere. The CFD code assumption of irrotationd flow is therefore not valid euperimentally. For inviscid flow the discharge coefficient is strictly due to the radiai momentum

of the flowing fluid. For viscous flows, the boundary iayer develops in the approach duct and on the orifice w d towards the lip. T h i s thin layer causes a reduction in the mass flow at the boundary and therefore reduces the discharge coefficient compared to inviscid the boundary layer theory (for the same conditions). But as detailed in Alder [1976], thickness is expected to be fauly constmt dong the orince wdl (approxhnately 6 pm for the 90 conical orifice) then drop rapidly to about 1 pm at the orifice lip. This decrease in boundary layer thickness could ound off' the effective orifice lip and may slightly increase the discharge coefficient in a manner similar to the Rld effect. Another effect of viscosity would be a recirculation zone (toroidal vortex) between the channel and the angled wd. Kosolapov & Sivoborod [1984] have indcated that the positive pressure gradients in the immediate vicinity of the stagnation corner cause a separation of the boundary layer. As a resdt of the change in c m t u r e of the streamiines

there i s an effective decrease in the wall angle which would lead to an increase in the experimental discharge coefficient.

5.6

Notes on the Equivalent Cd Distribution

Principle
The assumption that two orifices that have the same value of the discharge coefficient at

the incompressible b i t share the same discharge coefficient distribution on the graph Cd versus (1-r) (Le. the equivalent Cd distribution principle) has been shown in the previous figures to be valid. The discrepancies introduced by this assumption for reproducing the discharge coefficient distribution are not serious for engineering applications since the error compared to the theoretical results i s less than 1%.The equivaient Cddistribution principle is based on assumptions proposed by Jobson [1955] who noted that the discharge distribution with respect to the pressure ratio for any orifice was directly reiated to its discharge coefficient at the incompressible limit. An analytical proof of this concept is left for future endeavors.

Figure 5.1: Cd Algorithm results compared to values obtained by numerical methods for A, = 0,7 = 715, x = planar and various wall angles.

Figure 5.2: CdAlgorithm results compared to values obtained by numerical rnethods for planar flow with a! = 90, y = 715 and various area ratios.

CHUTER 5.

~SULTS AND

DISCUSSION

065
6.5 -.-

-.............. .-..-. ............. .-.-..


I

...........................................

...................................................
1

CdAtgorihm CFD code results Falkovich data Aldets data


0.8

OP

0.4 1-r

0.6

Figure 5.3: CdAlgorithm results compared to values obtained by numerical methods for

planar and axisyrnmetric flow with A, = O, a = 90 and various gas specific heat ratios.
1
l
I

0.95
0.9

.............

. . . . . . .

OP

0.4
1 -r

0.6

0.8

Figure 5-4: CdAlgorithm results compared to values obtained by numerical methods for planar and axisymmetrc flow with A, = 0 , =~ 715 and various wail ongles.

CHAPTER 5. RESULTS AND DISCUSSION


1

. - .

--..

. . . . A . . . . . . . . . . . . . .. . ........................................... . . 0.55 ...................... ...............-....* ......-...

0.5

Buckingham/Cunningham Faikovicti!hta i Cd QOnmm ..--.... CF0 code results

-1

0.2

0.4 t -r

0.6

0.8

Figure 5.5: Cd Algorithm and results from Buckingham/Cunningham7sapproximate method compared for planar flow through a dot, with a = 90 and y = 7/5.
t

0.95
0.9
0.85

0.8

0"

0.75
O.?

0.65
0.6

0.55

Ob

Figure 5.6: Cd Alg~nthm and results from Jobson's and Bragg's approximate method

compared, with A, = O and 7 = 715.

CHAPTER 5. RESULTSAND DISCUSSION

CdAlgorithm Parker & Kerctier CFD code results


0.8

------1

OP

0.4

0.6 t-r

Figure 5.7: Cd Algorithm and results from Parker bc Kercher's approximate method compared with a = 90, y = 715 and various 4.
1
1
I

0.5

i
I 1
t

~~'~lgofithrn
newexperimenaldata Parker & Kefcher
0.8

o ---

0.2

0.4
1 -r

0.6

Figure 5.8: Ca Algorithm and results fiom Parker & Kercher's approximate method compared with A, = 0.06, y = 7/5, a = 90 and various values of Rfd.

0.5

new expedmental results CdAlgorithm 0.8

0.2

0.4
1-r

0.6

1
%

Figure 5.9: Cd Algorithm and new experimental results w i t h y = 715, A ,

0.06,

R/d = 0.003 -t 0.005 and varying the wall angle (15 to 135').

Figure 5.10: CdAlgorithm and new experimental results with y = 715, A, R/d = 0.003 -t 0.005 and varying the waU angIe (30 to 150).

= 0.06,

-.-

0.2

0.4
1-r

0.6

0.8

Figure 5.11: Cd Algorithm and new experimental results with 7 = 715, A,


a = 4 5 O and various values of Rld.
1

= 0.06,

~i~lgorlthrn
OS
1

new experirnentaidata
0.8

0.2

0.4
t -r

0.6

Figure 5.12: Cd Algorithm and new experimental results with 7 = 715, R, a = 90 and various values of R/d.

0.06,

CHAPTER 5.

~SULTS AND

DISCUSSION

new experimental data Algonthm

O -

Figure 5.13: Cd Algorithm and new experimental results with y = 715, 4 , = 0.06, a = 135' and vaxious values of Rld.

CEAPTER 5. RESULTS AND D I S C U S S ~ N

0.2

0.4

0.6

0.8

1-r

Figure 5.14: CdAlg~nthm and new experimental data for y = 7/5, A, = 0.7 and various wnll angles.
new experirnental data Ca Aigorithm
$

-,

0.5

0.2

0.4
l- r

0.6

0.8

Figue 5.15: CdAlgorithm and new experimental data for y = 715,cu = 1 3 5 ' and va.rious area ratios.

CHAPTER 5.

RESULTS AND

DISCUSSION

CdAigorithm

a
1

0.5 L
O

Thomock I Brown [19721


0.6
1- r

01

0.4

0.8

Figure 5.16: Cd Algorithm and expe~mentd results from Thornock & Brown [1972] with

0.9

0.8

3
O.?

0.6

1-r

Figure 5 . 1 7 : Ca Mgonthm and experimental results fiom Grey & Wiisted [1949] with A, = 0 . 6 4 .

Chapter 6

Conclusions
The CFD computer code developed in this thesis solves two-dimensional, finite-difference, steady flows through sharp edged orifices. It produces numerical discharge coefficients for essentially the full range of values of the parameters wall angle, pressure ratio, area ratio

and gas specific heat ratio. For this code, the flowfield is divided into the subsonic regime that is solved via relaxation methods in the hodograph plane, and supersonic regime (if present), which i s solved as an initial value problem via the method of charactenstics in the physical plane. For supercritical flows, an iterative procedure was adopted to match conditions dong the sonic surface. This code improves upon methods previously publisbed. The results are more accurate and discharge coefficient data has been produced for the full range of the various flow parameters, which had been lacking in the literature. The developrnent of a complete code to include the effects of orifice edge rounding remains to be done. Experimental data for the discharge coefficient of conical o d k e s were obtained to extend the Cddatabase with new data, and to provide experimental results for the Cd Algorithm for both its irnplementation and validation. N e w experimentai data were required for these purposes, in particular for the cases of orifices with large wail angles (9-180) and different degrees of rounding of the orifice p edges. A new facility based on a reservoir blowdown technique was designed, constructeci and tested. The facility produced numerous experimentd data for the discharge coefficient that are of equivalent or better quality than produced by previous experimentai methods. This new method also offers advantages. It circumvents previous facility requirements for large flow meters and high-volume cornpressors that produce long-duration steady flows, although a standard and expensive LDV system is required. The new method produces experimentd data

for the discharge coefficient easily and efficiently with short test t h e s (3 s), and the measured data c m be analysed easily with personal computers to quickly yield numerous discharge coefficient values over a wide range of atmospheric-to-reservoirpressure ratios. Experimentally determined discharge coefficients tend to be greater than predicted by sharglipped theoretical or numerical results due to rounding of the orifice lip edge. A procedure called the CdAlgorithm was designed and implemented to produce the discharge coefficient distribution as a function of the pressure ratio for orifices with vazious

wall angles, area ratios, gas specific heat ratios and orifice edge rounding-to-diameter ratios. The CdAlgonthm uses the equivalent discharge coefficient principle to produce
the staadard graph quickly and accurately. The equivalent discharge coefficient principle

bas been demonstrated to be applicable to both slots and conical orifices under steady, compressible flow conditions. Discharge coefficient results obtained by using the Cd Algorithm have been shown to have a maximum error of less thao 1%compared to the theoretical or numerical results, which is considered more t h m sufficiently accurate for engineering usage. The CdAlgonthm is much better than other algorithms that exist in s faster, more the literature for producing the standard discharge coefficient graph. It i accurate, it predicts the rnultidimensional choking limit and is applicable to a wider range of orifices configurations.

6.1 Contributions
a

produced discharge coefficient results numerically that are more accurate than currently available in the literature

produced discharge coefficients numerically for a wider range of the dependent

a i i angles greater than 90') than available in the literature parameters (including w
a designed and implemented a new technique to experimentdy measure the discharge

coefficient of orifices for quasi-steady compressible flows


a

obtained experime~tal data (including the discharge coefficient) over a wide range of pressure ratios for orifices with different wall angles, area ratios and lip radius of
cunmture

performed an experimental investigation into the effect of lip rounding for conical orifices with different waiI angles

developed and implemented a cornputer procedure (called the Cd Algorithm) to produce the discharge coefficient distribution for the standard graph (Cd versus (1 - r))for planar slots and conical orifices as a function of wall angle, gas specific heat ratio, area ratio and iip radius of curvature which is better than existing appr-ate methods as it is faster, more accurate and applicable to a wider range of variables
0

developed an accurate and efficient expression for the determinationof the discharge coefficient and its derivative Nith respect to the pressure ratio at the incompressible pressure ratio for plan= and axisymmetric flows as a function of wall angle

References
Ader, G.M., 1976, '"lkamonic Flow Through SharpLipped Convergent Nozzles", University of Edinburgh Ph.D. thesis.
Ames, W.F., 1977, "Numerical Methods for Partial Differential Equations", Academic Press, New York.

ASME Research Cornmittee, 1961, "Flowmeter Computational Handbook" , .-\merican Society of Mechanicd Engineers, New York.
Benedict, R.P., 1970,"Generalized Contraction Coefficient of an Orifice for Subsonic

and Supercritical F~ows", Tkans. ASME: J Basic Eng., #70-WA/FM-1.


Benedict,

R.P.,Wyler, J.S, Brandt, G.B.,1975, "The Effect of Edge Shaxpness on

the Discharge Coefficient of an Orifice", Tki111.s. ASME: J. Engineering for Power, Oct. 1975, pp. 576-582. Benson,

R.S.,Pool, D.E.,1965, ''The Compressible Flow Discharge Coefficients for

a Two-Dimensionai St", Int. J. Mech. Sci., Vol. 7, pp. 337-353. Bragg, S.L., 1960, "Effect of Compressibility on the Discharge Coefficient of Orifices and Convergent NozzIes", J. Mechanical Engineering Science, Vol. 2 No. 1. p p 35-44. Buckingham, E., 1931, "Note on Contraction Coefficients of Jets of Gu",J. Research, US National Bureau of Standards, pp. 765-775. Busemann, A., 1937, UHodographenmethodeder Gasdynamik" , Ztschr f angew, Math. und. Mech., Band 17, Heft 2, pp. 73-79 (in German). Callaghan, EE., Bowden, D . T . , 1949, UInvestigationof Flow Coefficient of Circdar, Square, and Ellipticd Orifices at High Pressure Ratios", NACA TN1947.

Chan, Y S , 1990, "High-Pressure Reservoir Depressurization of Air with Condensation in Reservoir and Nozzle" , University of Toronto M.A.Sc. thesis. Chaplygin, S.A., 1902, "Gas Jets", Scient& Memoirs, Moscow University, English Translation: NACA TM1063,1944. Crockett, K.A., Upp, E.L., 1973, "The Memurement and Effects of Edge Sharpness on the Flow Coefficients of Standard Orifices" 'Itans. ASME: J. Fluid Eng., June 1973, pp. 271-275. Cunningham, R.G., 1950, "Supercritical Compressible Flow Through a Square-

Edged Orifice", Proc. Midwestern Conference on Fluid Dynamics, Vol. 1, pp. 373-388. Deckker, B.E.L., Chang, Y.F., 1965, "An Investigation of Steady Compressible Flow Through Thick Orifices", Proc. Instn. Mech. Engrs., Vol. 180, pp. 312-323.
Fakovich, S.V., 1957, "On the Theory of Gas Jets", Prikladnaia Matematika i Mekhanika, Vol. 21 No. 4, pp. 459-464 (in Russian). Fenain, M., Dutouquet, L., & Solignac, J.L., 1974, "Calcul des Performances d'une Tuyre Propulsive Convergente. Comparaison avec L"Expriencen, La Recherche -4rospatide, No. 5 pp. 261-276 (in Rench). Filippov, GA., Saltanov, G..4., Sivoborod, V.A., Kosolapov, Y S . , 1982, "Discharge of Gases with Differing hdiabatic Jhponents", Izvestia Akademii Nauk SSSR. Energetika i Transport, Vol. 20, No. 1, pp. 121-126, English Translation: Allerton Press Inc.

Frankl, F . I . , 1947, 'The Flow of a Supersonic Jet fkom a Vesse1 with Plane Walls" Dokl. Akad. Nauk. SSSR Vol. 58, No. 3, pp. 381-384 (in Russian).
Garabedian, P.R.,1956, "The Mathematical Theory of Three-Dimensional Cavities
and Jets", Bulletin of the American Mathematical Society, Vol. 62, pp. 219-235.

Grey, R.E., Wilsted, H.D., 1949, "Performance of Conical Jet Nozzles in Terms of Flow and Velocity Coefficients", NACA Report #933. Guderley,
Gnice,

KG., 1962, The Theory of Tkansonic Flow", Pergamon Press, Oxlord.

R . L . , Gottiieb, J.J., 1987, "Parsive and Active Reflection Eliminators",

Tenth International Symposium on the Military Applications of Blast Simulation, Vol. 1, pp. 114-141. Gurevich, ML, 1965, "The Theory of Jets in Ideal Fluids" , Academic Press, New York. Gushchin, B.A., 1959, "Flow of a Gas Stream fkom a Container of Finite Width at

Maximum Discharge", PMM Vol. 23, No. 4, pp. 770-776, English Translation: J. Applied Math and Mechanics, Vol. 23, pp. 1097-1106.
Hodge, B.K., Koenig, K., 1995, "Compressible Fluid Dynamics with Personal Cornputer Applications", Prentice

H a l l ,New Jersey.

Hunt, B.W., 1968, "Numericd Solution of an Integral Equation for Flow f r o ma Circular Orifice", J. Fluid Mech., Vol. 31, pp. 361-377.

ISO,1980, "Measurement of Fluid Flow by Means of Orifice Plates, Nozzles and Venturi Tubes Inserted in Circular Cross-Section Conduits Running Full", ISO
5167-1980,International Standards Organization, from the book "Measurement of Fluid Flow in Closed Conduits", ISO Standards Handbook 15, 1983, pp. 263-332.
the Flow of a Compressible Fluid Through Orifices", Proc. Jobson, D.A.,1955, Instn. Mech. Engrs., London, Vol. 16, pp. 767-772.

Kline, S.J., McClintock, F.A., 1953, "Describing Uncertainties in Single-Sarnple


Experiments", J. Mech. Eng., Jan. 1953, pp. 3-8. Kosolapov, YS., Sivoborod, V-A-, 1984, 'Theoretical and Numericai Investigation of the Flow of Gas rom Flat Slits and Axisymmetric Openings", Izvestiya Akademii Nauk SSSR, Mekhanika Zhidkosti i Gaza No. 2, pp. 109-115, English Translation: Plenum Publishing Corp. 0015-4628/84/1902-0266.

LU OC^, B.E., 1969, "Jets from Two-Dimensional Symmetric Nozzles of Arbitrary Shape", J. Fluid Mech., Vol. 37, pp. 479-489.

Linfield, K. W., 1992, "Ekperimentd Pressure and Velocity Profiles of Simulated Blast Waves Produced by a Composite Driver", University of Toronto M..4.Sc. thesis.
Maiilette, J., 1989, "Interior Ballistic Anaiysis of the High-Low Pressure Gun",

University of Toronto M.A.Sc. thesis. Two-Dimensional Jet Efflux From a Channel", J. Martin, C.S.,1977, ccAsymmetric Fluid Mech., Vol. 80, pp. 1-15. McGreehan, W.F., Schotsch, M. J., 1987, "Flow Characteristics of Long Orifices with Rotation and Corner Radiusing", ASME 87-GT-162. Milne-Thomson, k Co., London.

LM., 1968, "Theoretical Hydrodynamics", 5" Edition, MacMillan

Nelder, J.A., Mead, R., 1965, "A Simplex Method for Function hlinirnization", Computer Journal, Vol. 7, pp. 308-313.
Norwood, R.E., 1962, 'Swo Dimensional Transonic Gas Jets", Proc 4th US National Congress of AppIied Mechanics, Vol. 2, pp. 1359-1367.

Parker, DM., Kercher, DM., 1991, "An Enhanced Method to Compute the Compressible Discharge Coefficient of Thin and Long Orinces with Inlet Corner Radiusing", Heat Transfer in Gas Turbine Engines, ASME HTD Vol. 188, pp. 53-63.

Perry,J.A., Jr, 1949, "Critical Flow Through SharpEdged Orifices", Trans. ASME, Oct. 1949, Vol. 71, pp. 757-764.
Pickett, J.S., 1989, 'The Generalization and Evaluation of Solutions Occurring in Chaplygin's Ideal Gas+ Jet Theory", University of Toronto M.A.Sc. thesis. Press, W.H., Teukolskyi, S.A., Vetterling, W.T.,Flannery, B.P., 1992, "Numericd Recipes i n FORTRW", 2nd Edition, Cambridge University Press. Rouse, H.,Abd-Fetouh, A., 1950, "Characteristics of Irrotational Flow Through Axisymmetric Orifices", J. App. Mech., Dec. 1950, pp. 121-426.

Salas, S.L.,Hue, E . , Anderson, J.T., 1986, "Calculus One and Several Variables", 5th Edition, John Wiley & Sons, New York.
Shapiro, A.H., 1953, "The Dynamics and Thermodynamics of Compressible Fluid Flow", Volume 1and 2, Ronald Press, New York. Thornock, R . L . , Brown, E.F., 1972, "An Experimental Study of Compressible Flow Through Convergent Conicai Nozzles, Including a CompaRson with Theoretical Results", Tkans. ASME: J. Basic Engineering, Dec. 1972, pp. 926-932.

Ward-Smith,A.J . , 1979, "Critical Flowmetering: The Characteristics of Cyhrical Nozzles with Sharp Upstream Edgesa', Int. J. Heat & Fluid Flow, Vol. 1 No. 3,
Sept. 1979,pp. 123-132.
Zucrow, M. J., Hohan, J.D., 1976, "Gasdynamics"Volume 1and 2, John Wiley &

Sons, New Y o r k .

Appendix A
Derivat ion of Governing Flow
Equat ions
The derivation of the governing equations for compressible and incompressible Bows through two-dimensional slots and onfices with sharp and rounded edges are presented. The governing equations are obtained f i r s t for subsonic planar and axisymmetric flows (for slots and orifices, respectively) for the hodograph plane. Then, the governing equations are derived for supersonic planar and axisymmetric Bows (that occur in parts of slots and orifices, respectively) for the physicd plane. For subsonic flow the governing equation for compressible and incompressible Bows are presented in terms of the strearn hnction, whereas for supersonic compressible flow the goveniing equations become those of the method of characteristics featuring two-dimensional wave motion.

A. 1 Subsonic Planar Flow


The derivation of the governing equations used for solving planar (x, y) compressible and incompressible subsonic flows through slots foilows the conventional approach of Shapiro [1953].The continuity equation

and the condition of irrotationd

allow the definition of the velocity potentia.14;


u=dZ

and v

=&,

and the stream hnction @;


u
Po

and

II=--&,
P

Po

both in the physical plane. Since t$ and 11 are functions of both x and y, one may write
dq5 = &dx

+ @ydy = I.'

[cos (8) dx + sin ( 8 ) dg]

and

The physical coordinates can be eiiminated by first solving for dz and dy from the above equations in terms of d 4 and d@, that is

and

By using the definitions

dx = xvdV

+ zed8

and

dy = yvdV

+ yedO

in the previous two equations and s i m p ~ gone , obtains

Xe

COS (8) po sin (O) =40 --$e v P V

and

The variables x and y are eliminated from these equations by noting that

Mer differentiating equations A.3 to A.6 and using some mat hematical manipdation,
#e = -v$v P
Po

and

t$v

=V

For incompressible flow, the density p is a constant so

d p = u2 (in which a is the speed of sound and p is whereas for compressible flow using d~ the pressure) and using Euler's equation d p = -pVdV for frictionless and irrotational motion yields
q ,j
Po 1 (1 - b ~ *$0, = --)

where the Mach number Ad = -. a To eliminate the velocity potential in terms of the stream function, differentiate again
m d set

and then recall that p and 1 M depend only on V. T h i s procedure results in the linear differentid equation of the stream function in the hodograph plane

for compressible Bow and

v211vv + W v + llee = 0
for incompressible flow, both for the case of planar flow.

A.2

Subsonic Axisymmetric Flow


(2, r)

The derivation of the govenng equations for solving axisymmetric

compressible

and incompressible flows through orifices foilows the approach of Alder 1 1 9 7 6 1 ,which is similar to the derivation of the planar equations, but the equations have additional terrns for axially symmetric flow. The continuity equation

and the condition of irrotational flow

allow the defhition of the velocity potential &;

u = # = and u = & , and stream function $;

in the physical plane. Since # and @ are b c t i o n s of both z and r ,


dd = d:dt

+ 4,dr = V [cos ( 8 )dz + sin ( O ) dr]

and
d+ = &dz

+ &dr = V-Pi [- sin (O) dz + cos ( O ) dr] .


Po

The physical coordinates can be eliminated by first solving for dz and dr, that is

and

By using the definitions

dz = rvdV

+ zedO

and dr = rvdV

+ redO

in the previous two equations and simplifying, one obtains

Ze

cos (0) p, sin (8) =#e - - @O *

Pr

(A.12)
(-4.13)

and

The variables z and r are eliminated fkom these four equations by noting that

After dinerentiating equations A.ll to A.14 and some mathematical manipulation,


$6

PO = -VdV Pl'

$fl and q5v = -

EM2
P

PO re

-;)+pr~+v.

To eliminate the velocity potential in t e m s of the stream hinction, differentiate again, set
and then recall that p and hl depend only on V. T h i s procedure results in the nonhear difkrential equation of the strem function in the hodognph plmc

The derivatives of r can be determined as a function of the hodograph variables and the Stream function as rf? = - Po [COS (O) @O + V sin ( O ) qv] , (A.16) r p r 2 v (A. 17)
'WB --

pr2V
rv =-

cos (el [Vsin (0) qvv + V

(A.18)

and
PO [V$V(cos ( 8 )

pr*V

+ sin (0) 2) + 'r

sin (O) (hl2 - i)]


t

(A.19)

and then upon substitution into equation A.15 yields

sin (0) cos (01

($(1 - IV) ~+ i:

$c*e)]

=O

for compressible flows. The above equation reduces to

1 sin e cos B V*
for incompressible flows.

(-*;

+ +bA)]

=O

A.3

Supersonic Planar Flow

The velocity potential for planar f l o w can be defined from

V=gradd=vt$
or

Substitution of these into the continuity equation

dp = a2 or Recall that d~

Hence,

and

Upon substitution of these two equations into the continuity equation and after simplification produces

or in terms of the velocity components

Following the method presented in Shapiro [1953],equation A.22 and the partial derivatives
d#z = 4n&

+ 4zgdy

and
d& = $,d~

+ $,dy

can be put in the rnatrk form

For supersonic flow, the characteristic curves can be determined if the derivatives of the veiocity components u and v (and hence ail of the Buid properties) are discontinuous. This is handled in the method of characteristics by letting , ,& 4, and #w be discontinuous; for euample,
(1-$)

0
d#,
d#,

(1-$)
O

det

4q, =

dx O

dy

) -9
dy

(1-$1
O dy

&

The determinant in the denominator produces

dy Solving for gives

dx

-UV

faJu2
a 2

+ v2 - a2

- u2

The deterrninant in the numerator yields the equation

Recall that v =

and u = 4%;hence, du = d 4 , and du = d$,. With manipulation of

dy the above equation and substitution for one obtains

dx

-=

dv

du

-uv~a\/u~+tP-a~ v2 - a2

along with Using the hodogaph variables u = V cos(9) , v = V sin(@) and 0 = tan-L 1 and -= dm, the equation for the Mach angle relationships p = s i n d '
dy can be simplified after much mathematical manipulation to yield

(3

dx

Substitution of the hodograph variables and the Mach angle relationships into -, along du with
du = -V sin(0)dB cos(0)dV

du

and
du = V cos(0)dO sin(O)dV,

The above equation can be simplified by int roducing the Mach number relationships 2 V = Ma, d V = a dM + R/I da and = 1+y i c l l to yield

(2)

This 1 s t equation can be integated directly to produce the Prandtl-Meyer equation


d= F/ztan-l 7-1

,/- r + l

* tan-LJMZ-~ +constant.

(A.27)

The above procedure yields two c'families"or sets of characteristic equations. The rightrunning Mach lines ( "-" characteristics)

and the left-ninning Mach lines (

"+" characteristics) are

is the Prandtl-Meyer angle.

A.4

Supersonic Axisymmet ric Flow

The derivation of the equations for axially symmetric f l o w follows the same procedure as that for planar flow. The velocity potential cm be defined frorn

Substitution of these into the continuity equation

yields

d p = a2 or Recail as before that dp

Hence,

and Upon substitution and simplification these equations yield

or in terms of the velocity components

The previous process for supersonic planar f l o w can be repeated for supersonic axisymmetric flow, but the coordinates (x,y) are now replaced by (2,r ) and the additional t e m for a,xially symmetric flow is included. The velocity potentid equation (A.34) and
the partial derivatives
d4z = d Z &

+ dmdr

can be written in niatrix form as

For supersonic flow in the method of characteristics, the derivatives of the velocity corn-

ponents (i.e. &, , # and & . ) may be discontinuous or indeterminant as detailed for planar fiow above; for example

From the determinant in the denominator

Solving for

dr yields dz

The determinant in the numerator produces the equation

and u = &; hence, du = d& and du = d#,, and the above equation d~ can be manipulated and after substituting for - produces
dz

R e c d that v =

Using the hodograph variables u = V cos(0), v = V sin(@) and B = tan-l


the Mach angle relationships p = sin-'

( ) :

dong with

and

1 = d m , the equation for

dr cm be simpMed after much mathematical manipulation to yield dz

which is the same as the planar case. Substitution of the hodograph variables and the du Mach angle relationships into - dong with du

and
du = Vcos(0)de

+ sin(8)dV

yields

The above equation can be simplified by using V = Ma, dli = a d M = 1 + -Af2 to yield 2

(I)~

+ M da and

This last equation can be integrated to produce

Therefore, the right-running Mach lines ("-" characteristics) are determuied with
@=-Y+

sin p sin6 dr - +Csin (O - p ) r

and

and the left-running Mach lines ("+" characteristics) are detennined using
0=uand

sinp sin 0 dr sin (9 p) r

+ - +c+

Вам также может понравиться