Citation
Numerical modeling of natural convection and conduction heat transfer in canned foods with application to on-line process control

Material Information

Title:
Numerical modeling of natural convection and conduction heat transfer in canned foods with application to on-line process control
Series Title:
Numerical modeling of natural convection and conduction heat transfer in canned foods with application to on-line process control
Creator:
Datta, Ashim Kumar
Publisher:
Ashim Kumar Datta
Publication Date:
Language:
English

Subjects

Subjects / Keywords:
Boundary conditions ( jstor )
Cooling ( jstor )
Free convection ( jstor )
Heat transfer ( jstor )
Heating ( jstor )
Isotherms ( jstor )
Liquids ( jstor )
Velocity ( jstor )
Vorticity ( jstor )
Water temperature ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Ashim Kumar Datta. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
000872393 ( alephbibnum )
14514671 ( oclc )

Downloads

This item has the following downloads:


Full Text









NUMERICAL MODELING OF NATURAL CONVECTION
AND CONDUCTION HEAT TRANSFER IN CANNED FOODS
WITH APPLICATION TO ON-LINE PROCESS CONTROL










By

ASHIM KUMAR DATTA


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1985































Dedicated to the memory of my

grandfather and father















ACKNOWLEDGEMENTS


The author is indebted to Dr. Arthur A. Teixeira, his major

professor, for his encouragement, guidance and patience throughout

the course of the work and especially for his assistance in

preparing the manuscript.

The author is also highly grateful to Dr. Tom I-P. Shih, Dr.

Chung K. Hsieh and Dr. Ruey J. Shyu for their time and extremely

valuable suggestions in the study of the natural convection heat

transfer. Further thanks to Dr. Khe V. Chau and Dr. Spyros Svoronos

for their interest in the study and helpful suggestions.

The importance of a good computing facility (number crunching

to be particular) in developing the natural convection model cannot

be over-emphasized. The availability of the Cyber 170 computer

through Dr. John Gerber and the IFAS administration was vital for

that purpose. The author also greatly appreciates help from Dr.

Ellen Chen, who introduced him to the Cyber facility, and Victor,

Jimmy and Margaret for endless support on problems with the Cyber

system.

The mountain of data available from the transient natural

convection heating model would have been of little use without the

plotting capability provided by the excellent graphics software

(PLOTPAK) made available from Dr. William E. Dunn from the








Mechanical and Industrial Engineering Department of the University

of Illinois at Urbana-Champaign. Help by Subhasis Laha in physically

transporting the plotting software to Gainesville is also

appreciated. Availability of a hard copy plotter was the next

necessity and the author appreciates the equipment provided by Dr.

James W. Jones in the Agricultural Engineering Department.

Finally, the author deeply acknowledges the assistance from his

wife Anasua in typing this manuscript, and her encouragement and

perseverence during the study.













TABLE OF CONTENTS


ACKNOWLEDGMENTS .................... ...........................i

LIST OF SYMBOLS................................................ vi

LIST OF TABLES....................................................xi

LIST OF FIGURES............ ................................ xii

ABSTRACT ........................................ .......... xvi

INTRODUCTION...................... .............................

REVIEW OF LITERATURE.............................................5
Computer Control of Food Thermal Processing..................5
Overview of Food Thermal Processing........................5
On-line Control of Thermal Processing....................10
Conduction Heating of Canned Foods.........................12
Studies Based on Analytical Solution.....................12
Numerical Studies .......................................13
Natural Convection Heating in Canned Foods..................14
Previous Works on Canned Liquid Foods..................15
Other Similiar Works on Natural Convection
Heating in Cylindrical Enclosures........................19

PROBLEM FORMULATION...........................................24
Natural Convection Heat Transfer in Canned Foods............24
Governing Equations and Boundary Conditions..............24
Boussinesq Approximation.................................26
Transformation of the PDEs.............................27
Non-dimensionalization of the Variables.................28
Conduction Heat Transfer in Canned Foods...................30
Governing Equation..................................31
Boundary Conditions.............................. .....31
Computer Control of Conduction Heating Process..............32

METHODOLOGY .................................................... 35
Modeling of Natural Convection Heating in Cans..............35
Grid Generation and Grid Stretching......................36
Discretization of the Parabolic
Temperature and Vorticity Equation......................39
Time derivatives .............................. .....42
Linearization. ........................................44









Convection terms......................................45
Diffusion terms ...................................... 50
Discretization of the Elliptic
Stream Function Equation..................................61
Computational Boundary Conditions for the FDEs...........69
Convergence Criteria for Iteration .......................74
Algorithm for Iterative Solution.........................76
Coding of the Computer Program and Hardware..............80
Modeling of Conduction Heating in Cans......................80
Discretization and Solution of the Equation..............81
Coding of the Computer Program..........................82
Experimental Studies ....... ........ ....................83
Use of the Conduction Heating Model for On-line Control.....86

RESULTS AND DISCUSSION.... ......................... ...... ......93
Natural Convection Heating in Canned Foods..................93
Transient Flow Patterns and Temperature Profiles.........94
Start of flow and conduction layer....................95
Radial temperature and velocity profiles.............106
Axial temperature profiles........................... 113
Slowest Heating Zone................ ..... ............. 115
Contrast with Conduction................................120
Assessment of the Numerical Method......................124
Convergence..........................................124
Selection of the grid size...........................125
Selection of the time step...........................130
Convergence of boundary vorticity...................134
A note on the half grid points.......................134
Conduction Heating in Cans.................................138
Analysis of Boundary Conditions........................139
Comparison of the Transient Temperature Values.......... 141
Effect of Sudden Pressure Drops in the Can.............. 141
Performance of the On-line Control Logic...................141

CONCLUSIONS AND RECOMMENDATIONS..............................151
Conclusions................................................151
Recommendations............................................ 153

APPENDIX
A ALTERNATIVE FINITE DIFFERENCING OF
STREAM FUNCTION EQUATION................................154
B ALTERNATIVE FINITE DIFFERENCING OF BOUNDARY VORTICITY...155
C INPUT DATA FILE FOR NATURAL CONVECTION MODEL............156
D ADE FORMULATION FOR THE CONDUCTION EQUATION.............157

LIST OF REFERENCES.......... ..... .............. ................ 161

ADDITIONAL REFERENCES............................. ....... ....... 167

BIOGRAPHICAL SKETCH................................... ....... 172
















LIST OF SYMBOLS


A = constant

a,b,c = coefficients in tri-diagonal matrices

f = inverse slope of time-temperature graph

g = gravity

Fo = sterilization Fo value
H3
Gr = Grashof number = gB (Ti To) -7
h = outside heat transfer coefficient

H = height of cylinder

j = j value (a constant)

JO = Bessel function of order zero
k = thermal conductivity

n = normal direction

nz = number of grid points in z direction

nr = number of grid points in r direction

p = pressure

Pr = Prandtl number = pC /k = v/a

r = non dimensional distance in radial direction

r = distance in radial direction

R = radius of cylinder








= Rayleigh number = Gr*Pr

= time

= temperature at any point inside at time t > 0

= initial uniform temperature inside (at t = 0)

= boundary temperature at time t > 0

= retort temperature at time t
H -
= -u = non-dimensional velocity in vertical direction

= velocity in vertical direction
H-
= -v = non-dimensional velocity in radial direction

= velocity in radial direction

- non-dimensional distance in vertical direction
H
= distance in vertical direction

= Z-value for organism


= thermal diffusivity = k/pCp

= coefficient of thermal expansion

= relaxation parameter for boundary vorticity

= designates a difference when used as a prefix

= convergence criterion for boundary vorticity

= convergence criterion for SOR

= vertical distance in computational plane = n(z)
T-Ti
= non-dimensonal temperature
To-T
= first eigenvalue in z direction

= eigenvalues in z direction

= first eigenvalue in r direction

= eigenvalues in r direction


viii








= viscosity of liquid being heated

= kinematic viscosity of liquid being heated = p/p

= radial distance in computational plane = (r)

= density of liquid being heated

= deformation parameter in grid stretching

= --t = non-dimensional time
H2
= radial coefficients for kinematic consistency

= radial coefficients for kinematic consistency

= non-dimensional stream function = -
Ha
= stream function defined by v = -
r ar r az
= non-dimensional vorticity
av au H2 v U H2 -
az ar a a

= vorticity

= optimum relaxation parameter for SOR


Subscripts

a

center

cool

heat

i

J

m

n

W


= ambient

= geometrical center of can

= cooling cycle

= heating cycle

= grid points in vertical direction

= grid points in radial direction

= grid point on vertical boundary (z = H)

= grid point on radial boundary (F = R)

= Wall








Superscripts

d = design value

f = normal to grid faces

k = iteration counts in SOR

m = iteration count within a time step

n = time step

t = tangential to grid faces
= "half" step in ADI method (same as n+l/2)












LIST OF TABLES


TABLE PAGE

1. Sequence of line types for identification of stream
function contours in figures 10a-o........................96

2. Heating and cooling rates for various process
conditions........................... ..................... 140

3. Adjusted heating times and resulting lethality (Fo)
in response to process deviations using proposed on-
line control logic and method of Giannoni-Succar and
Hayakawa ................................................. 148












LIST OF FIGURES


FIGURE PAGE


1. Isotherms and velocity profiles in a glass bottle at
various heating times -(after Engelman and Sani, 1983).....18

2. Isotherms and velocity profiles in a can at various
heating times (after Sani, 1985).........................20

3. The grid system and the boundary conditions for the
cylindrical can...........................................40

4. Notations of various quantities defined on a non-
uniform grid system in cylindrical geometry...............62

5. Algorithm for the iterative solution of the set of
equations in natural convection...........................77

6. Experimental setup for conduction heating..................84

7. Timings of various real-time computations for on-line
control of conduction-heated food.........................88

8. Flow diagram for computer control of retort opera-
tions with on-line correction of process deviations.......90

9. Isotherms near bottom wall predicted by the
convective model compared with the predictions
considering conduction (equation 83) only.................97

10a. Isotherm and streamlines in a cylindrical can after 1
second of heating .........................................99

10b. Isotherm and streamlines in a cylindrical can after 2
seconds of heating........ .......... ......................99

10c. Isotherm and streamlines in a cylindrical can after 3
seconds of heating.......................................100

lOd. Isotherm and streamlines in a cylindrical can after 4
seconds of heating......................... ......... ..... 100

10e. Isotherm and streamlines in a cylindrical can after 5
seconds of heating.................. ... .. ............ ..101








10f. Isotherm and streamlines in a cylindrical can after 6
seconds of heating....................................... 101

10g. Isotherm and streamlines in a cylindrical can after 7
seconds of heating......................................102

10h. Isotherm and streamlines in a cylindrical can after 8
seconds of heating...................................... 102

10i. Isotherm and streamlines in a cylindrical can after 9
seconds of heating.................... ... ....... ......... 103

10j. Isotherm and streamlines in a cylindrical can after
10 seconds of heating................. ........... ...... 103

10k. Isotherms and streamlines in a cylindrical can after
30 seconds of heating................... ................ 104

101. Isotherms and streamlines in a cylindrical can after
120 seconds of heating.......................... ....... 104

10m. Isotherm and streamlines in a cylindrical can after
300 seconds of heating.............. ..................... 105

1On. Isotherms and streamlines in a cylindrical can after
600 seconds of heating................................... 105

10o. Isotherms and velocity vectors in a cylindrical can
after 1800 seconds of heating...........................107

11. Predicted radial velocity profile at the sidewall at
mid height in a cylindrical can after 30 seconds of
heating ..................................................108

12. Predicted radial temperature profile at the sidewall
at mid height in a cylindrical can after 30 seconds
of heating ...............................................108

13. Isothermal vertical surface, showing boundary layer
with velocity profile u(x,y) and temperature profile
t(x,y) in water (Higgins and Gebhart, 1983)..............110

14. Predicted radial velocity profiles at the sidewall at
mid height in a cylindrical can after various heating
times .......................... ........................... 111

15. Observed radial velocity profiles at the sidewall at
mid height in water in a cylindrical can after
various heating times (from Hiddink, 1975)...............112

16. Comparison of numerically predicted axial temperature
profiles with observed (Hiddink, 1975) values at
various times during heating.............................114

xiii









17. Locations of the slowest heating points in water over
a heating period of 10 minutes in a cylindrical can......116

18. Migration of the slowest heating points in water over
a heating period of 10 minutes in a cylindrical can......116

19. Transient temperature at various locations during
heating of water in a cylindrical can....................119

20. Transient temperature at various locations during
initial times of heating of water in a cylindrical
can.......... ... ..... ..................... .......** 121

21. Temperature at the slowest heating point during
natural convection heating contrasted with that
during conduction heating in a cylindrical can...........122

22. Isotherms during natural convection heating contr-
asted with those during conduction heating in a
cylindrical can..........................................123

23. Diagram of 39(0.9)x39(0.8) grid on the left and
58(0.85)x58(0.76) on the right...........................126

24. Isotherms after 90 seconds of heating computed using
39x39 grid (left half of the figure) compared with
computations using 58x58 grid (right half of the
figure) ..................................................127

25. Streamlines after 90 seconds of heating computed
using 39x39 grid (left half of the figure) compared
with computations using 58x58 grid (right half of the
figure) .................................................. 128

26. Transient slowest heating temperatures computed with
39x39 and 58x58 grid.................. .................... 129

27. Transient temperature at 1/3rd height on the center-
line computed with 39x39 and 58x58 grid...................131

28. Time step history during a typical run of the
numerical heat transfer model for natural convection.....132

29. Isotherms in a cylindrical can calculated with two
different convergence criteria for boundary vorticity
(6 =0.0001 on the left and 6 =0.001 on the right)........135

30. Streamlines in a cylindrical can calculated with two
different convergence criteria for boundary vorticity
(6 =0.0001 on the left and 6 =0.001 on the right)........136








31. Transient center temperature during conduction-
heating of Bentonite in a cylindrical can................142

32. Transient center temperature during conduction-
cooling of Bentonite in a cylindrical can with and
without drop in retort pressure...........................143

33. Computer plot of reference process (no deviations)
showing 66.8 minutes of heating time and an
accomplished Fo of 6.24............ .............. ......144

34. Computer plot of a process that experienced a step
functional drop of 8.30C (150F) in temperature bet-
ween 50th and 60th minute and still maintained an Fo
of 6.02.................................................. 146

35. Computer plot of a process that experienced a linear
drop in temperature between 40th and 60th minute
(from 00C to 11.1C (200F)) and still maintained an
Fo of 6.18..............................................147












Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

NUMERICAL MODELING OF NATURAL CONVECTION
AND CONDUCTION HEAT TRANSFER IN CANNED FOODS
WITH APPLICATION TO ON-LINE PROCESS CONTROL

By

ASHIM KUMAR DATTA

December 1985


Chairman: Arthur A. Teixeira
Major Department: Agricultural Engineering


A numerical model was developed for predicting detailed flow

patterns and temperature profiles during natural convection heating

of liquids in a can (a cylindrical enclosure). The liquid (water)

was initially stagnant at a uniform temperature of 300C, and the

sidewall, top and bottom wall temperatures were suddenly raised to a

temperature of 1210C. A free (thermally insulated) top liquid

surface was also considered. Boussinesq approximation was used and

all other fluid properties were treated as constant.

Finite difference methods were used to solve the governing

equations in axisymmetric cylindrical coordinates. A vorticity-

stream function formulation was used. Parabolic temperature and

vorticity equations were solved with the alternating direction

implicit method and upwinding the convective terms. The elliptic

stream function equation was solved using successive over relaxation








method. Grid stretching was used in both radial and axial

directions. Boundary wall vorticities were used to check for

convergence of the iteration process.

Plots of transient isotherms, streamlines and velocities were

provided. Calculated radial velocities and axial temperature

profiles at different times compared quite well with available

experimental data. Slowest heating points were located in the bottom

15% of the height of the container studied. These slowest heating

points migrated over time within this lower region of the can but no

particular pattern of migration was noted.

Transient can center temperature was also studied for the case

of conduction-heated solid food. An alternating direction explicit

finite difference solution to the conduction equation compared very

well with analytical solutions and experimental results. The ability

of the finite difference model to perform under arbitrary variation

of boundary temperature allowed the model to be incorporated in an

algorithm for on-line control in batch thermal processing of

conduction-heated food. The uniqueness of this control system in

maintaining the level of desired food sterilization for arbitrary

variation in process heating temperature was shown to offer

considerable advantages over other possible methods of on-line

control.


xvii
















INTRODUCTION


Background and Justification

Thermal preservation of food involves application of heat

sufficient to destroy the microorganisms present in food that cause

spoilage. Commercial canning processes provide such heat treatment

to extend the shelf-life of foods. In a typical canning process, the

container is filled with food and sealed. The sealed container is

heated in a closed vessel with steam or hot water long enough to

kill the microorganisms and is then cooled. In a production situa-

tion, the heating medium (steam or hot water) temperature can

sometimes deviate significantly from design values during the

process. Heating the product more than the required amount would un-

necessarily degrade its quality while wasting energy. However,

insufficient heating could lead to serious public health hazards and

cannot be allowed. Thus, food processors always have to meet this

minimum heat treatment (sterilization). Determination of the heating

time to meet the minimum sterilization demands that the actual

transient temperature history of the food during processing be known

predicted with a heat transfer model or actually measured).








Conduction-Heated Foods

Traditionally, for conduction-heated foods, the required

processing time is determined using the analytical solution to the

heat conduction equation. In practice, the constant medium temper-

ature assumed by the analytical solution cannot always be achieved.

Thus, experimental correction factors have been proposed to adjust

the heating time, predicted by analytical solution, when the medium

temperature goes through certain (predefined) types of deviations.

However, deviations in practice are arbitrary, and it is not

conceivable to know the correction factors for all such situations.

Of course, the problem is trivial if the transient food temperature

is directly measured using a thermocouple; but this is both

inconvenient and impractical in a production situation. Better

models for predicting food temperature in response to arbitrary

fluctuations in medium temperatures are therefore needed.

A numerical finite difference model would be able to predict

the food temperature for truly arbitrary variations in the medium

temperature. However, the calculations required are relatively much

more complicated and time consuming. Using a microcomputer to

perform this would have two advantages. In addition to correctly

predicting food temperature for truly arbitrary heating conditions,

it can perform this automatically without any worker supervision.

Introduction of low-cost microcomputers for on-line process control

is becoming commonplace in the 1980s. Food industries are not an

exception to this situation. Use of microcomputers for on-line

control of conduction-heated food is thus quite conceivable.








Convection-Heated Foods

Unlike solid packed foods that are heated by conduction, thin

liquid foods heat through convection. In general such liquid foods

are mechanically agitated to improve the heat transfer. However,

there are situations where the food material cannot be agitated for

several reasons. The food then heats primarily by natural convec-

tion. In-package pasteurization for liquids (e.g. beer) in bottles

or cans that cannot be agitated is an example. Broth, thin soups,

evaporated milk, most fruit and vegetable juices, fruits in syrup or

water, and pureed vegetables are also examples of products where

natural convection heating can occur when no agitation is used.

Flexible pouches that cannot be agitated to keep the package integ-

rity are another important example. Fermentation is another process

where a natural convection heat transfer model (with heat generation

term included) can provide valuable insight.

Just as optimum heating times are required for conduction-

heated foods, they are also needed for natural convection-heated

products. Thus heat transfer models are similarly needed. Since we

are interested in the minimum heat treatment the food receives, we

need to know the point or the region that heats the slowest. Because

convection sets up complex flow patterns, the location where the

temperature is minimum is by no means obvious. Only a study of

complete transient flow patterns and temperature profiles can give

such insight. Previous studies to achieve this objective have not

been successful. The use of more advanced numerical techniques and

better computing equipment would make it possible to study in detail

the natural convection heat transfer in canned liquid foods.








Such a natural convection heat transfer model could also have

several applications outside the food area. It may be used to study

the heating of buildings (without forced circulation), cooling of

gas turbine blades, storage of cryogenic fluids (e.g. liquid rocket

propellants), and the startup of chemical reactors. Since natural

convection heat transfer would involve a much higher degree of math-

ematical complexity than a conduction heating model (and its appli-

cation to on-line control), the primary objective of this study was

directed toward developing a heat transfer model for natural convec-

tion; followed by improvements to existing conduction-heat transfer

models for application to on-line process control as secondary

objectives.


Objectives

The objectives of this study were to

1. develop a mathematical model to predict temperature and

velocity profiles caused by natural convection heat

transfer in a closed cylinder,

2. investigate the need for possible improvements on an

existing conduction heat transfer model to improve its

suitability for use in on-line process control appli-

cations, and

3. incorporate the conduction heating model in an on-line

process control algorithm for thermal processing of

canned food, and compare its performance against other

possible methods of on-line control.














REVIEW OF LITERATURE


This chapter begins with a brief introduction to thermal

processing as applied to the sterilization of canned foods. The need

for computer control of thermal processing is noted, along with the

importance of heat transfer models in such computer control systems.

Previously reported conduction heat transfer models used in studying

sterilization of conduction-heated canned foods are described next.

The few studies attempting to model natural convection heat transfer

in the case of sterilizing canned liquid foods are noted. The few

other works on natural convection heat transfer in liquid inside a

vertical cylindrical enclosure are also reported.


Computer Control of Food Thermal Processing

An overview of thermal processing is presented in this section.

Problems with traditional approaches to processing of conduction-

heated foods are then noted and the need for on-line control is

stressed. Some of the past approaches in thermal processing of con-

duction-heated foods that could also be used on-line are discussed.


Overview of Thermal Processing

Thermal processing consists of heating foods, filled and

hermatically sealed in containers, in pressurized steam retorts at a








constant retort temperature for a prescribed length of time followed

by rapid cooling. These process times are calculated on the basis of

sufficient bacterial inactivation in each container to comply with

public health standards and reduce the probability of spoilage to a

specified minimum. Estimation of bacterial inactivation requires

understanding of thermal death kinetics of bacteria, along with the

temperature history the food (containing the bacteria) experiences

during the process. Following is a brief introduction to the thermal

death kinetics of bacteria (Ball & Olson, 1957; Stumbo, 1965).

Thermal death of bacteria follows first order kinetics. If C is

bacterial concentration (number of bacteria per unit volume of food

material) at time t,




dC 1 (1)
dt (0.4 34)OT


where (0.434)*D is the inverse of the rate constant. If Ci is the

initial concentration and C is concentration after time At

C
Ci At
log = (2)
C


The constant D (known as 0 value), which is the inverse of the rate

constant (multiplied by the factor 0.434), is the time required to

reduce the bacterial population by one log cycle at a specified

temperature. The temperature dependency of the D value is given by






T -T

D = Do 10 Z (3)



where Do is the D value at some reference temperature To. The Z

value describes the temperature dependency of the lethal rate and is

related to the activation energy. The reference temperature To is

generally taken as 1210C (2500F).

From bacteriological considerations, a thermal process is

specified by noting the time (Fo) necessary to reduce the bacterial

population, from the initial (Ci) to some final safe value (C) at

temperature 1210C (2500F). Thus

C.
F0 = D log (4)


When the heating process is not at the constant temperature of

121C, introducing the temperature dependency of Do and equation (2)

for bacterial death,



T-121
Z At
Fo = D 10 z--

T-121
= 10 Z At (5)



At is the time required at temperature T to achieve the designed

bacterial reduction. As the food material is heated, its temperature

T becomes a continuous function of time t and the expression for Fo

(equation 5) becomes








t T-121
Z (6)
F = f 10 dt (6)


When the food being heated has a spatially varying temperature

profile, the chosen temperature T, for thermal process calculation,

is the temperature at the slowest heating point in the food

material. Thus, for conduction heating in a cylindrical can, the

chosen temperature T is at the geometrical center of the can.

The designed bacterial inactivation for a process is specified

by Fd which is the time required to achieve the desired
0
inactivation of bacteria when the same is heated at a constant

temperature (T) of 1210C. When temperature T is a function of time

(T(t)) instead, the objective of thermal processing is to select
process time t for the particular heating condition, such that

the equivalent Fo received (as given by eqn. 6) is greater than or

equal to the design value F. That is,


Fo(t) Fd (7)


Application to conduction-heating. When applying thermal

processing to conduction-heated food in a cylindrical container,

Ball and Olson (1957) used the solution to the heat conduction

equation in cylindrical coordinates.

T-TR -(X2+ 2)at
R E Ampe m -P Jo(mr) Cos ( z) (8a)
TO-TR m p








The solution was first approximated to

T TR -(X2 + X2)cat
S=- A e Jo(xr) Cos(xz) (8b)
T0- TR


when sufficient time has elapsed, so that the first term in the

analytical series solution (equation 8a) dominates. Equation (8b)

was then rearranged to give


j (TR- TO)
t = f log TR T(9)
TR- T


where
S 2.303 (10)
(f a + (10)
and

j (r,z) = A Jo(xr) Cos (xz) (11)


From the transient heat conduction equation (9) which gives

temperature T as a function of time t and the definition of Fo

(equation 6) which gives Fo as a function of temperature T, it is
possible to find Fo as a function of time t. However this cannot be

performed analytically. Upon solving the two equations numerically,

it is possible to know temperature Th achieved at the end of heating

for a given value of Fo. Ball and Olson (1957) tabulated Th-TR

values as a function of F0 for particular values of f and j. Tradi-

tionally, in thermal processing of conduction-heated foods in cylin-
drical containers, these tables are used to find Th-TR for a given

Fo value. Using this (Th-TR) value and equation (9), the process

time t is found.








On-line Control of Thermal Processing

In a production situation, the heating time t specified

(assuming a constant value of TR) to achieve a given Fo value would

need to be adjusted for arbitrary variations in the heating medium

temperature TR from its specified constant value. Researchers in the

past have attempted this in one of the following three ways

Precalculated tabulations. Correction factors are tabulated to

be applied to th for certain possible types of deviations in

TR(t). Commercial retort control systems often use this technique
(LOG-TEC system, 1984). Such tabulations cannot be exhaustive,

since the process temperature TR(t) could go through any random

deviation. Instead of tabulating such correction factors, Giannoni-

Succar and Hayakawa (1982) developed expressions for correction

factors to be applied to th when TR(t) goes through a step drop. In
their work, other types of deviations had to be approximated to

close-fitting step functions, and multiple drops could not be

considered. Getchell (1980) described a control system that would

attempt to maintain the design heating medium temperature and sound
an alarm when critical low or high limits for the temperature are

exceeded. The operator would then attempt to make necessary changes

in th.

Direct Measurement. Temperature at the can center T(t) is

measured on-line. This eliminates the need to assume constant

heating temperature. Measured T(t) is used with equation 1 to stop

the heating cycle at time th such that

F t Fd (12)
Fo(th) F








This approach was used by Mulvaney and Rizvy (1984). They used a

thermocouple-fitted test can in every retort batch to obtain actual

temperature T(t). The system proposed by Navankasattusas and Lund

(1978) also planned to measure can center temperature on-line. In

commercial practice these methods are cost-prohibitive with regard

to production efficiency, and are considered impractical.
Numerical Models. Instead of measuring T(t) at the can center,

T(t) can be predicted for arbitrary variations in TR(t). In

general, analytical solutions for T(t) are of limited use for
arbitary variations of TR(t). Teixeira and Manson (1982) described

a numerical finite difference approximation to the heat conduction

equation for predicting T(t) from TR(t). This is conceptually quite

different from the approaches mentioned earlier. With a finite
difference model, T(t) can be predicted for truly arbitrary

variations in TR(t) and, of course, use of this model eliminates the
need for having a test can with an actual thermocouple inserted in

it. This model was first described and verified against published

procedures by Teixeira et al., 1969; and later verified against

experimental data and other published procedures for time-varying

boundary conditions by Teixeira et al. (1975 a,b).

However, in the proposed retort control algorithm of Teixeira

and Manson (1982), every time a deviation took place, a new value of

th was found through simulation assuming the deviated retort temper-
ature continued for the rest of the heating process. This is unnec-

essary because any further changes in the retort temperature would

make the estimated th value useless and fresh estimations of th








would require repeated time-consuming computer simulations.

Using the general idea of Teixeira and Manson (1982), the

feasibility of several control algorithms that can work on-line with

available computer hardware needed to be investigated, and led to

the following studies as part of the work carried out in this

project:


1) Ways to obtain the transient temperature solution faster,

using other available methods of solving the conduction

equation.

2) The need to incorporate a convection (Robin's) boundary

condition to reflect the true situation, rather than the

temperature specified (Dirichlet) boundary condition used

by Teixeira et al. (1975a,b) in solving the conduction

equation.


Conduction Heating of Canned Foods

Since the pioneering work of Ball and Olson (1957) mentioned

earlier, conduction heating of canned foods has been studied

extensively. Consequently there is a vast amount of literature

existing on the subject.


Studies Based on Analytical Solution

The majority of the works on conduction-heating of canned foods

were based on one of several different forms of the analytical solu-

tion to the heat conduction equation. Also most of these solutions








were derived assuming, among other things, a step change in the

boundary (can wall) temperature from the initial to the heating

temperature and maintaining the boundary heating temperature

constant throughout the process. More general forms of boundary

conditions have also been formulated. Hayakawa (1964) developed the

general form for an analytical solution that could predict the

transient temperature distribution inside a cylindrical can of

conduction-heated food when the boundary (can wall) temperature was

any Laplace-transformable function of time.

In practice, the heating medium temperature, which affects can

boundary temperature, is not under complete control and can go

through unexpected variations from its design value. For such

arbitrary variations in the boundary temperature, analytical

solutions are of limited use.



Numerical Studies

A finite difference analog of the heat conduction equation can

also be used to achieve the transient temperature of food at finite

but small increments of time. Such methods can utilize the actual

boundary temperature "read" for the incremental time and calculate

the interior temperature values based on this boundary temperature.

Teixeira et al. (1969) were the first to use a forward difference

explicit finite differencing method to study the transient

temperature distribution inside the cylindrical can. Use of the

explicit finite differencing severely limited the choice of grid

size and time increment, due to stability requirements for such a








method. In their later work (Teixeira et al., 1975a) an alternating

direction explicit (ADE) method (Saul'yev, 1957; Allada and Quon,

1966) was used that was unconditionally stable but still explicit.

This method was first applied to simulate thermal processing of

canned foods by Manson et al. (1974). Teixeira et al. (1975a,b)

compared outputs from the ADE method for using constant boundary

temperature and sinusoidal time-varying boundary temperature with

the results from corresponding analytical solutions. The analytical

results for constant boundary temperature agreed within 0.30C

(0.50F) with the results from the ADE method, except at interior

points near the surface after small initial time (less than 3

minutes). Similar agreement between the results of the analytical

and the ADE formulation was also obtained when a sinusoidally

varying boundary condition was used.

Teixeira (1971) also compared the ADE finite difference outputs

for conduction heating in a can with experimental data for constant

boundary temperature and for step increases in the boundary tempera-

ture during heating. The agreement was good except at initial times.

In addition to this ADE method, other methods of solution to

the conduction equation needed to be investigated from the point of

view of their suitability of computation in real-time and ability to

accommodate arbitrary boundary temperature variations.


Natural Convection Heating of Canned Foods

Internal (in enclosures) natural convection flow problems have

received much less attention than the external ones. Also, internal









flow problems are generally more complex than the external ones.

Fewer works have been reported on internal flows in cylindrical

coordinates. Of the studies on natural convection heating in

cylindrical enclosures, only four studies could be located where

they were in reference to natural convection heating during the

thermal processing of liquid foods. Such thermal processing is

characterized by a rather large Grashof number, generally small

overall dimension of the container and water as the liquid. Perhaps

the food processing situation is unique to general heat transfer

studies, which would explain the fact that only one such study could

be located in the more common heat transfer literature.



Previous Works on Canned Liquid Foods

The rate of heat transfer in canned liquid foods was studied

using correlations, and slopes of time-temperature curves with

limited success (Blaisdell, 1963; Barreiro-Mendez, 1979). Possible

flow patterns were postulated by Blaisdell (1963) using dye and

aluminum powder tracers.

Stevens (1972) first tried to study detailed temperature and

flow patterns during natural convection heating of canned liquid

foods. He used an explicit finite difference formulation from the

work of Torrance and Rockett (1969). The can was treated as a ver-

tical cylindrical enclosure. The fluid (ethylene glycol) inside the

can was initially at rest at a uniform temperature of 300C. The can

was then placed in hot water bath at about 1000C. The top surface of

the fluid was free (the can had headspace). The actual temperature









variation on the can boundary was noted and was used in the

numerical calculations. Temperature values were recorded at ten

different points inside the container. The recorded temperature

values agreed poorly with numerically calculated temperatures. The

numerical inaccuracy was attributed to insufficient grid points

which could not resolve the details of the flow. Several sources of

experimental error included inaccurate setting of boundary

temperature, non-zero initial velocity field, errors in thermocouple

measuring circuit and inaccurate placement of the thermocouples.

Some of these factors were thought to have caused three dimensional

movements in the experiment and could not be picked up by the two-

dimensional numerical model.

Detailed temperature and velocity profiles during natural

convection heating of liquid foods in a cylindrical can were

extensively investigated by Hiddink (1975). The numerical solution

technique used was the explicit finite difference technique of

Barakat and Clark (1966) discussed later. The initial temperature

was considered uniform throughout. The top, bottom and sidewall were

suddenly raised to the heating temperature. The top surface of the

liquid was considered free (having headspace) and was thus treated

as thermally insulated. Several test fluids of low and high

viscosities (water, sucrose solution and silicone fluid respec-

tively) were used. The numerical results included temperature and

streamline patterns. The overall flow patterns were visualized using

a "particle streak method." In this method small glittering

particles are suspended in the liquid. The particles are illuminated









by a flat narrow slit of light, exposed against a dark background

and photographed. Detailed study of flow in the boundary layer was

observed using a laser-dopper velocimeter which is an improved

version of the "particle-streak method" mentioned above. Temperature

values were measured by inserting thermocouples at several points in

the liquid. The temperature variation on the top free surface was

observed using a Thermovision Infrared camera. Radial velocity

profiles were presented at different heights in the liquid and at

different times during heating. His experiment and other works (Chu

& Goldstein, 1973) had showed convective heat transfer from the

bottom, as expected, when the bottom was also heated. His numerical

results, however, could not pick up the convective eddies generated

by the bottom heating. Thus his numerical results had poor corre-

spondence with his experimental results.

A finite element method was used by Engelman and Sani (1983) to

study in-package pasteurization of fluids (beer) in bottles. In the

process studied, bottles entered the pasteurizer at 1.70C(350F) and

passed through several progressively hotter zones of hot water

spray, which raised the product temperature to 600C(1400F). This

temperature of 600C is maintained in the holding zone. The product

then passed through several progressively cooler zones, which low-

ered the temperature to 70 800F. Since the product was fluid and

the package was not agitated, it was primarily heated through nat-

ural convection. Experimental details for this study was provided by

Brandon et al. (1981). The bottle geometry was complex (figure 1)

and so were the boundary conditions. The bottom of the bottle was











































20 SECONDS


42 SECONDS


Figure Isotherms and velocity profiles in a. glass
bottle at various heating times (after Engelman
and Sani, 1983)









treated as insulated and wall temperature specified as a function of

time. The numerical results were in good agreement with experimental

data.

Sanil (personal communication, 1985) also developed a similar

model for pasteurization of beer in cylindrical metal cans (aspect

ratio =4.2). Water was the test fluid (Pr=6.81). The initial

temperature was 350C and the sidewall boundary condition used was

approximately 600C. The top of the water level was free (the can had

headspace) and was treated as insulated. The temperature at the

bottom of the can was a function of radial distance. Temperature and

velocity profiles from this study are given in figure 2. Further

quantitative details on this study were not available. These results

were not compared with experiments and also not published.


Other Similar Works on Natural Convection Heating in Cylindrical
Enclosures

One of the earliest of the numerical studies on the subject of

natural convection heating in cylindrical enclosures was performed

by Barakat and Clark (1966), who were looking at thermal strati-

fication and associated processes in liquid propellant tanks for

application to space flight. The vertical cylindrical enclosure,

they studied, was filled with water, and arbitrary temperature and

heat flux variation were imposed on its wall, while its bottom was

kept insulated. At the top free surface of the liquid, heat transfer

to the ambient was considered negligible and thus insulated. They


1R. L. Sani, CIRES, Campus Box 449, University of Colorado, Boulder,
Colorado 80309


































asseasa as a I1.1
6 l 8 * * ** -- --

Gsilla.
\ fIlimII s alllII1*



i11 a 6 ) as t i11
(LiI~LaaliIll)i





S\-* |* a a ) \\ .\


IS S S- I s a
allalIilalasIla












Figure 2. Isotherms and velocity profiles in a can at
various heating times (after Sani, 1985)
o













i *:: : : ;illI;
I .. .. . ..

: .: : : : .: ;: : .:: : ;i : : : : : : :


. . . . . . ..
. . .. *... .. . . .. ... il**
. . * ** . . . . . . .
I1 i
S' . * **I( * * * . i
I I * **CII' i I .. t * ** *
* * * i .*....... ..** I
* * * * * e . * * * **
S*j * i* * , * * * * **




i * * s o il* i * * * * *I*
I S S **S * * 1*1 * * ****
*. * *** * * *I ***
I . . . . .. I* . S * ** *

Ie I IIil" ** * * *
S.* * * .. * *
Ir \ B r I I I .





F r ** * *
i* i ** CI is *


aa I I * * * l* I




.i gu ,2 con ,d.,*










Figure 2. contd.







had numerically calculated and plotted the isotherms and streamlines

inside the enclosure. In their experimental work, they recorded

temperature values at ten different points inside the cylindrical

enclosure. The observed temperature values deviated from the

numerical results by 10%. This discrepancy was attributed to the

small non-symmetry in the imposed wall temperature which might have

produced three dimensional effects and, also, to heat losses from

the bottom and top of the container.

Torrance (1968) compared five separate finite difference

formulations for the problem of natural convection in cylindrical

enclosures. The five formulations were respectively by Barakat and

Clark (1966), Wilkes and Churchill (1966), Fromm (1964) and two of

his own. The physical problem chosen was the study of natural

convection flow induced in a vertical cylindrical enclosure by a

small hot spot centrally located on the bottom. One of his two

formulations was explicit and the other was implicit. The explicit

formulation was shown to be numerically superior in many ways to the

other four and it was also reported to have good. agreement with

experimental observations (Torrance and Rockett, 1969; Torrance et

al., 1969) for lower values of Grashof numbers. This was also the

formulation used by Stevens (1972) referred to earlier.

Kee et al. (1976) performed a numerical study of temperature

distribution and streamlines in a closed vertical cylinder

containing uniformly distributed heat generating tritium gas. The

walls were kept isothermal. An upwind method was used for the

convective terms. Numerical temperature values were compared with









experimentally determined temperatures at three axial and two radial

locations. Thermistor thermometers were used for temperature

measurement. The calculated temperature values were reported to be

within the limits of experimental error, over a range of Grashof

numbers.

Evans et al. (1968) measured the transient temperature field

during natural convection of fluid in a vertical cylinder due to

uniform heat flux. The top surface of the liquid was free and no

transfer of heat or mass was assumed at the liquid-vapor interface

at the top. Dye tracers were used to qualitatively observe the flow

patterns. They reported symmetry about the vertical axis in the

temperature field. On the basis of the experimental observations,

the convective liquid flow was modeled in terms of three regions:

1. a mixing region at the top

2. a central core

3. a boundary layer rising at the heated wall

They developed three different equations to describe the three

regions. These equations were solved simultaneously to predict the

overall behavior of the system. They reported good agreement between

the model and their experimental data.

Shyu and Hsieh (1985) made a numerical study of unsteady

natural convection in thermally stratified water in a cylindrical

enclosure. They studied the effect of placement of insulation on

maintainence of the stratification. Transient flow and temperature

fields were provided for the study.
















PROBLEM FORMULATION


The governing equations and boundary conditions for the natural

convection and conduction heating process together with the assump-

tions are first described here. The on-line control problem for the

conduction heating process is then formulated.



Natural Convection Heat Transfer in Canned Foods

The equations governing natural convection heat transfer in

liquid foods together with the boundary conditions are first

presented. The Boussinesq approximation to the set of equations is

then discussed. To further facilitate the solution, new variables

are defined and the governing equations together with the boundary

conditions are transformed into a new set of equations.



Governing Equations and Boundary Conditions

In natural convection heat transfer, the driving force for the

liquid motion is the buoyancy caused by the density variations due

to change in temperature. The partial differential equations (PDEs)

governing such naturally convective motion of fluid in a cylindrical

space are the Navier-Stokes equations in axisymmetric cylindrical

coordinates (Bird et al., 1976) and are described below.









+ Tar
9r


+ u 3T
3z


- 3
r +
DF


K l1 a 3
- ~r ( (r
p r ar


= + {
8z r


u")
av
ai )


0


aT ) 2T
r) + )
ar az2


3 -3u
ar ar


+ 313
3r r 3r


1 (pV) + --(p) =
r ar 8z


The boundary conditions are



sidewall


T=T1


32u
+ +
Dz2


(r )) + }
aZ2


u = 0


v = 0


centerline


O z < H


3T 0
3F


bottom wall


0 < < R


top wall


0 < < R


z= 0



z=H


T=T1



T=T1


=0 = 0


(17a-d)


u =0


u = 0


v = 0


v = 0


Initially the fluid is at rest at uniform temperature


0 < < R


0 < < H


T=TO


u =0


v =0 (18)


Refer to list of symbols for definition of individual symbols.



As expected, the flow field (equations 14 & 15) is coupled with

the temperature field (eq. 13) through the density term. These


a-
+ +
aF


av
p (- +
at


(13)



(14)



(15)



(16)








equations have been approximated for no viscous dissipation. Jaluria

(1980) discussed various situations under which viscous dissipation

can be neglected. All other fluid properties are treated as

constants except variation of density which is discussed below.


Boussinesq Approximation

The solution to the set of equations can be considerably

simplified using the Boussinesq approximation. In the Boussinesq

approximation; the density is treated as constant except in the body

force term. The density difference in the body force term is approx-

imated as

p p = pB (T T ) (19)


and the density variation in the continuity equation is ignored.

This approximation requires both


B (T Ta) << 1 (20)
and ( )T g << 1 (21)


to be true. The second inequality is generally satisfied for

liquids. So is the situation with the first inequality. For

example, for natural convection in water, as in this problem


B = 0.0002 /C
T = wall temperature of the can = 1210C

Ta= initial temperature of liquid in can = 30C





27


Thus,

B (T Ta) = (0.0002)(121-30) = 0.0182 << 1 (22)


Almost all the previous works on natural convection in similar

areas have used the Boussinesq approximation.


Transformation of the PDEs

To further facilitate numerical solution to the set of

equations, the primary variables are transformed to a new set of

variables. Vorticity (J) and stream function ( ) are defined in

terms of the primary variables u (axial velocity) and v (radial

velocity)


:a u ,a(23)
3z aF



u =1a v = (24a,b)
rF Fr a



Using vorticity so defined, the pressure (p) is eliminated from

the u and v momentum equations 14 and 15. These two equations gave

the parabolic transport equation for vorticity Next, using the

definition of (equation 23) and the definition of (equations

24a and 24b), the equation for is derived. This gave the vorticity

stream-function formulation for the natural convection problem.







28

Non-dimensionalization of the Variables

The variables in the equation are then non-dimensionalized as

below.


z
H

H-


T H2t
H2


r
r -
H
v H
a


(25)


T T1
1
eH
T0- T1


H2 -


1
ir~F :


The resulting equations are the governing equations for natural

convection in the form used for the present study.


ae a(ue) 1 (rve) _a2 1 a aeo
T- + z- } = { T-aiZ + r- ) }


(26)


aw a(u) a3(vw) D- a2W a 1 a(rw)
-9T + { +- } =T -GrPrr + Pr{ -yz+ F r ar (7)

(27)


- T + ( T )


1 a
u = -1 F


(29a,b)


1 a2p
r az


(28)












The boundary

variables c.

sidewall11

r=R/H

centerline

r=O

bottom wall

0
top wall

0

y conditions

an be written



0 < z < 1



0 z < 1



z=0



z=l


in the transformed and non-dimensionalized

as


= 0



=a 0
@r= 0


o 0=O



0=0


W = 0


S= 0 u=0



^ 0 .0
S=0 -u = 0
ar


S= 0 u=0



S= 0 u=0


(30)


with the initial conditions
R
0

W = 0


S= 0 u=0


The relative merits of solving the transformed variables (w,p)

as compared to the (u,v,p) system of primary variables were discuss-

ed by Roache (1982). To solve the (u,v,p) system, another elliptic

Poisson equation would need to be formed for the pressure (p) vari-

able, using the two equations for velocity, regardless of whether

transient pressure solutions are desired. Following are the relative

evaluations of the (u,v,p) system and the (w, ) system for two-



1Explicit boundary conditions for vorticities on solid walls are
unavailable.


(31)








dimensional flows when no free surface is present, iterative methods

are used to solve the Poisson equation, and transient pressure solu-

tions are not required (Roache, 1982).

1. The (w,( ) system needs to solve one parabolic equation for

w and one elliptic Poisson equation for i as compared to the

(u,v,p) system which requires the solution of two parabolic

equations for u and v and one elliptic equation for p.

2. With the (0, 4) system, two additional differential equations

(29a and 29b) are needed to get the velocities u and v, but

the u and v momentum transport equations are more complicated

than the w transport equation.

3. At no-slip walls, the boundary conditions for u and v are

known explicitly walll = 0, wall = 0) as compared to wall

which is not known explicitly. This would be a great advantage

for using implicit methods for the (u, v, p) system but non-

linear instabilities of the pressure term prevents the (u, v,

p) system from using implicit methods.

4. The elliptic equation that needs to be solved for the pressure

takes much longer to converge than the elliptic equation for

p (eq. 28).


Conduction Heat Transfer in Canned Foods

The PDE describing conduction heat transfer in a cylindrical

body of solid-packed food material is presented here. The boundary

conditions present during the thermal processing of canned foods are

described next.








Governing Equation

For strict conduction heating (no movement of the material) of

a homogeneous isotropic cylindrical-shaped body of the food materi-

al, the heat transfer is described by the classical heat conduction

equation
ae a2e 1 eo 9a2
TF ar + -F F + z (32)


This is the energy equation (26) approximated for zero velocities

(diffusion term only). This equation is quite well behaved and easy

to solve. The resistance to heat transfer of the metallic walls of

the can was ignored.


Boundary Conditions

Canned foods are generally heated with hot water or steam and

cooled with water. Thus, the heat transfer coefficient at the out-

side of the can wall could be comparatively small (for water) or

quite large (for condensing steam). In general, the conditions at

the boundaries can be written as

sidewall

r =R/H 0 z 1 3 + hH =0
--+ ar o =O
centerline

r = 0 0 < z 1 = 0 (33a-d)
bottom wall
3e hH
0 < r < R/H z = 0 az e= 0

top wall

0 < r < R/H z = 1 + hH = 0
z k








with the initial conditions

0 < r < R/H 0 < z < 1 9 = 1 (34)




Computer Control of Conduction-Heating Processes

In controlling thermal processes, the objective is to meet the

designed level of bacterial sterilization (Fd) for the process,

irrespective of any retort temperature variation TR(t), and with a

minimum of overprocessing. The lethal effects of thermal processing

are achieved during the heating as well as cooling times (th and

tc). Thus, the objective function in thermal processing is to

S T-121 + T-121
minimize F(th +t) = O 10 dt + fth 10 dt (35a)
h

subject to the constraint: F (th+tc) > Fd (35b)



where the integral from equation 6 has been broken into separate

heating and cooling times. Fo(t) is the accumulated sterilization at

any time t. The first integral above is the contribution to the

sterilization Fo(t) from heating, and the second integral is the

contribution to the sterilization during cooling. Temperature T(t)

is taken to be the temperature of the slowest heating point in the

product. This is so that when the design F (F ) is satisfied at
this point, all other points in the product have also been satis-

fied. For a cylindrical can, T(t) is the temperature at the can

geometric center at time t.








This transient temperature T(t) is a function of transient

retort temperature TR(t), the can wall outside heat transfer coeffi-

cient (h), dimensions of the can (R and H), and thermal diffusivity

(a) of the product for a conduction-heating food. Symbolically,


T(t) = f (TR(t), R, H, a) (36)

For conduction heating in a cylindrical can, the function f is the

heat conduction equation in cylindrical coordinates.

For a given can size and product, the two variables that can

possibly be controlled are the retort temperature TR(t) during

heating (saturated steam or hot water under pressure in the retort),

and the time period of heating th. Cooling is generally done using

available water at ambient temperature. This makes the contribution

to Fo from cooling (given by the second integral in equation 35a)
immune from further control once the cooling process is under way.

Also, both the heating and cooling temperatures can only be applied

to the boundary of the product making it a boundary control problem.

In the case of a process deviation, the heating medium tempera-
ture may go through unexpected variations beyond the capacity of the

temperature controller. Thus in reality, only one variable can be

controlled. It is the time period of heating (th). The problem in

thermal processing is to specify th for arbitrary variations in

TR(t) during heating while the process is under way in real-time.

Instead of searching for the proper th through simulation every
time a deviation takes place as suggested by Teixeira and Manson

(1982), Fo(t) can be monitored in real-time using T(t) calculated








from the finite difference approximation to the heat conduction

equation (32) with measured boundary (retort) temperature TR(t). The

heating process could continue until time th such that

Fo(th) > F (12)


thus satisfying the required sterilization Fd for arbitrary varia-

tions in retort temperature TR(t).
However, the contribution to Fo from cooling (the second

integral in equation 35a) should not be neglected for a conduction

heating product. Depending on the can center temperature at the

start of cooling, size of the can, and other factors in a conduction

heating situation, the contribution to Fo from cooling could be as

much as 40% of the total. This is'unlike convection heated products,
where the container is sometimes agitated and the product is

rapidly cooled, making a small contribution to Fo during cooling. To
avoid gross overprocessing of conduction heated product, therefore,

the cooling Fo must be considered. Also, as mentioned earlier, there
is no control over the possible contribution to Fo from cooling once

cooling is under way.

Thus, the contribution from cooling cannot be neglected, it is

not a constant, and it cannot be controlled. Therefore, it can only

be estimated before cooling is actually started. Such estimation of

cooling lethality can be done through simulation of the cooling

cycle assuming a constant cooling water temperature in the retort.

The objective here is to develop an algorithm to perform this simu-

lation in real-time when other control actions are going on.
















METHODOLOGY


Detailed development of the solution techniques for natural

convection and conduction heat transfer models are described in this

chapter. For the natural convection heating problem, formation of

finite difference equations, setting up of computational boundary

conditions, and development of the algorithm for the iterative

solution of the complete problem, are described first. Along the

same lines, solutions of the finite difference analog to the

conduction equation are described next. Experimental study of

conduction heating is then reported. With the thermal models

working, use of the conduction heating model for on-line control of

conduction-heated food is then presented.



Modeling of Natural Convection Heating in Cans

Because of the inevitable coupling of the governing equations,

except for extremely simplified cases, most natural convection flows

of practical interest are too complex to be solved analytically.

Thus numerical computations are generally employed to obtain the

desired solution. The present problem of natural convection in a

cylindrical enclosure is no exception. Numerical solutions to this

and similar problems demand high speed digital computers with large








amounts of memory that were unavailable until the 1960s. Since then

numerical studies of natural convection heating in enclosures are

being reported constantly. Among the numerical methods, finite

difference methods are probably the most popular in computational

fluid dynamics. The same was chosen for the present study.

Several choices were available for the finite-difference

analogs, computational boundary conditions (treatment of boundary

data in the numerical solution, as compared to real physical

boundary values), and the technique and order of solving individual

PDEs in the overall computational cycle. The relative merits of

these choices are discussed, and appropriate selections are

described in this section.



Grid Generation and Grid Stretching

The discrete points at which finite difference equations are to

be solved constitute the grid system for the problem. For computa-

tional efficiency, the number of grid points should be the minimum

that is required to resolve all the significant spatial variations

of the flow. Since this problem involves a rather large step

increase in the wall temperatures together with a high Rayleigh

number (Ra), strong boundary layers are expected to form near the

walls. This is true particularly for the side (vertical) wall. In

order to resolve these boundary layers adequately, a large concen-

tration of grid points would be necessary near the walls. An alge-

braic method was used to generate the nonuniform grid system. The

particular algebraic expression chosen was evaluated for the proper-








ties listed below (Anderson et al., 1984; Ka'lney de Rivas, 1971).

Here (r,z) refers to the physical domain and (,n) refers to the

computational domain with AE and An as constants.


1. The mapping from (,n) to (r,z) must be one to one.

2. Grid lines should be smooth to provide continuous

transformation derivatives. This requires r(E) and z(n)

as well as -r and d should be continuous over the closed

intervals 0
functions defined on a closed interval are bounded, this in

turn implies that dr and dz should be finite over the

whole interval. If dr becomes infinite at some point, the
d
mapping r(S) would give poor resolution near that point,

which cannot be improved by increasing the number of points,

since A r r (-) (AE) Similarly for dz/dn.

3. Close spacing of grids where large numerical errors are ex-
dr 0 at the
pected. For this problem, that means -= 0 at the
r=R/H
sidewall which ensures high resolution of the boundary layer

near r = R/H. Similarly at z = 0 and z = 1.

4. Absence of excessive grid skewness which can sometimes

exaggerate truncation errors (Raithby, 1976).


An algebraic transformation equation suggested by Kublbeck et

al. (1980) for natural convection in cavities was


1 (+ tan[T/2 (2z-1) al) (37)
2 tani/2 a








where a is

stretching.

ultimately

mation z(n)

below.


a deformation parameter that varies the degree of

However, since the computer coding of the problem is

done in the physical domain, the inverse transfor-

is more appropriate in this context which is given


z =1/2{1 + -- tan ((2n -1) tan a)}
Tr a 2


(38)


This concentrates the grids on either end z = 0 and z = 1.

In the r direction, the above transformation is modified to

concentrate grids on one side (side wall, r = R/H) only. Thus


r = 2 tan1 ( tan o) -
r- 2 H


(39)


To check if the transformations indeed satisfy requirements 1

through 4, we first see that both the transformations are one to

one. Next, looking into the transformation derivatives,


dr 2 R tanw/2 a
d 77 IFT 1 + 2 tan2f/2 a
(40a,b)
dz 1 2 2 tan n /2
n w a 1 + (2n 1)2 tan2Tr/2 a


which are continuous in 0 < 5 < 1 and 0 < n < 1 respectively.

Also


dz
Lt d I
a+1 dn z=0


S2 tanira/2
S 1 + tan2Lt
a+1 we 1 + tan2nr/2








=Lt 2tan (1 2) by L'Hospital's rule
a+1


= 0

Similarly


dz 2 tanWra/2
Lt = Lt = 0
a+ z=l a lr ao 1 + tanZna'/2


and
dr 2 tani/2 a
Lt = Lt 1+tan2/2= 0
a 2l ir=R/H = 1 7r (J 1 + tan27r/2 a


Thus the transformations provide closer spacing of grids near z = 0,

1 and r = R/H. Figure 3 shows the grid system with close spacing of

grids near the top, bottom and sidewall. As can be seen, there is no

excessive grid skewness.

Ka'lney de Rivas (1971) also showed that when conditions 2 and

3 are satisfied for any smooth function, "extra truncation errors"

due to the non-uniform grid system are improved from first order to

second order in An and AE Vinokur (1983) also discussed similar

types of stretching functions.



Discretization of the Parabolic Temperature and Vorticity Equations

Of the several finite difference formulations that are

available for any set of PDEs, careful choices needed to be made

depending on many factors including boundary conditions, geometry of

the problem, type of solution (steady-state or transient) desired,

range of parameters involved (particularly the Grashof number in









e .= 1
mj r-2t Z
m, m-1 ,j
m0, ?Z- -Zi2


or amj 0


ei ,0= Oi I
i,0o O1
iO
e~0= ei


= 0-


SIr i 1 ~ I I I I I II


I 111111


j = 0 O0,j=

SO,j
O,j-


I t I IIIIllllll lii


tI IIII IItlll I


i ,n=
e. =
i ,n=

i ,nf


0

1
-24i ,n-I
rn-/2 r n- r-
n12n n-i


j = n


1
-29
S1,j
r~ Z
ii1


The grid system and the boundary conditions
for the cylindrical can


ii


I I I I I


Figure 3.


'"'


SI I I H IIIIIIIIIIIIlIII


.t - It -- I I ill


4++-++1


ilIFill


I I I I








natural convection problems), as well as realistic constraints on

available resources (Roache, 1982). For this study, the transient

temperature and velocity solutions are needed and the range of

Grashof number is 108 1010 Since the temperature and vorticity

equations are both parabolic and are very similar, they will be

discussed simultaneously in this section.

The performance of the various finite difference methods was

noted in terms of stability, truncation error, and conservation

property. Stability was defined as the controlled growth of the

round-off errors (O'Brien et al., 1950). An uncontrolled growth of

the round-off errors leads to instability in the finite difference

methods. The stability requirements often put restrictions on the

time step as a function of grid spacings. Empirical stability re-

striction on grid size of the type Ax<8/jui,jl and Ay< 8/Ivi,jl

were also reported by Torrance (1968). Truncation error is the error

resulting from the discretization of the PDE. Conservation of any

physical entity (e.g. energy, vorticity) within the finite differ-

ence grid system exists if the difference equations for the entity

are summed over all interior grid points and no spurious sources or

sinks of these quantities are found. Torrance (1968) found that

formulations by Barakat and Clark (1966) and by Wilkes and Churchill

(1966) did not conserve energy or vorticity and he did not recommend

the use of their equations. Another important property of finite

difference formulations is the transportive property. A difference

formulation is transportive if the effect of a perturbation in a

transport property is convected only in the direction of velocity.








Time derivatives. The treatment of the time derivative in

equations 26 and 27 can be broadly classified into implicit and

explicit methods. Explicit finite difference methods are generally

easier to understand and program. Explicit methods have been used by

Barakat and Clark (1966), Torrance (1968), Stevens (1972) and

Hiddink (1975) in studying natural convection heating inside a

cylindrical container. For most of the explicit methods, however,

there are stringent restrictions on the allowed time step, from the

stability considerations. In situations where large temperature

gradients, eddies and recirculations are present like in this study

grid points need to be fine enough to resolve the physical phen-

omena. When using most of the explicit methods, such a fine grid

size could severely limit the allowed time step and thus require

prohibitive amounts of computer time.

Implicit finite difference methods involve more than one

advance time-level value in the same equation and thus require

matrix inversion to calculate function values at a new time level.

These methods generally allow larger time steps due to better

stability properties. However, even though a larger time step is

allowed in the implicit method, it requires many more calculations

to solve that step. For an implicit scheme to save computer time in

the overall computation, it should allow several times larger a time

step than that allowed by an explicit method to achieve a given

accuracy critrion.a


aT. I-P. Shih, Mech. Engg. Dept., Univ. of Florida. Personal Comm.








Instead of using a fully implicit set of equations, alternating

direction implicit (ADI) methods were introduced in companion papers

by Peaceman and Rachford (1955) and Douglas (1955). The advantage of

the ADI method over a fully implicit method is that the set of

equations, although implicit, is tridiagonal. Further the ADI method

as applied to linear equations has a formal order of accuracy of

O(At2,Az2,Ar2). The stability of this method is also unconditional

as determined by Von-Neumann stability analysis. For irregularly

shaped regions, programming could get complicated (Roache, 1982) but

for simple rectangular regions, as is the present situation, that is

no particular problem.

The practical advantage of the ADI method over the explicit

method is, however, nothing like that indicated by Von-Neumann

analysis (Roache, 1982). It is unstable both at large time steps

and/or insufficient convergence of boundary vorticity. In general,

many researchers have indicated that ADI methods do allow larger

time steps and faster overall computation, by a factor of two or

more, and furthermore allow second order accuracy in time. The ADI

method is a very popular method in computational fluid dynamics.

While solving the set of PDEs (26-29) Barakat and Clark (1966)

compared explicit and implicit methods for their study and decided

against the use of the implicit method due to unavailability of the

vorticity boundary conditions at the wall (to be explained later).

They allowed vorticity boundary values to lag one time step, neces-

sary for such explicit formulations. The lagging of boundary

vorticity is considered undesirable (Roache, 1982) and iterations








are required because of this situation. The iterations can also

serve the purpose of preserving second order accuracy in time of the

ADI method. The second order accuracy may otherwise deteriorate due

to the presence of non-linear convective terms (discussed below).

Since the iterations are needed anyway, an implicit set of equations

can be solved, instead, utilising the iterations. As noted earlier,

Stevens (1972) and Hiddink (1975) also used the explicit

formulations of Barakat and Clark (1966) to study the present

problem and could not resolve the convective eddies present due to
the bottom heating. Since the ADI method may allow finer grid

spacings with less stricter restrictions on At, it may be able to

pick up the convective eddies present, without taking prohibitive

amounts of computer time. Thus the ADI method was chosen to solve

the temperature and vorticity equation in this study. The ADI

formulation is written below for the temperature equation 26. At the

"half" step


en+1/2 en a(uo) n+1/2 a(rve) n a2 n+ /2 1 n
AT/2 8z ar a z +r ar Ir -r


and at the "full" step

n+1 en+ 1/2 (u) n+ /2 n(rvI+/2 1 n+1
e -o +(ue) n+ a(rve) n al^e 1 n-
A-t/2 + az I + r az= r arar


Linearization. The velocities in nonlinear convective terms

e.g a(ue)In+1/2
e.g. (u) should ideally be evaluated at the same time level
(n+ 1/2 in this case) as the (or ). However, since u and v
(n+ 1/2 in this case) as the 9 (or w). However, since u and v








depends on i which in turn depends on u this would couple all the

equations together and make it practically incomputable. To avoid

this, the nonlinear terms are linearized. Various ways of lineari-

zing these terms have been discussed by researchers (Anderson et

al., 1984; Peyret and Taylor, 1983; Roache, 1982). A very simple and

common strategy in linearizing is to evaluate the velocity coeffi-

cient at n time level. This is known as "lagging" the coefficients.

However in this case, the accuracy in time drops from second order

(for ADI) to first order. Roache (1982) noted that such "lagging" of

the velocities could still be second order accurate if the velo-

cities u and v were slowly varying. An improvement over this

linearization procedure is to update the coefficients u and v as

iterations proceed. The procedure is repeated until the values u and

v converge. One way to achieve this would be to update un+1/2 and

vn+1/2 by solving the stream function equation at the "half" time

step of the ADI (Roache, 1982). Since solution of the Poisson stream

function equation is the most time consuming operation, it was

decided against this updating procedure.

Anderson et al. (1984) noted that "lagging" the coefficients

has been widely and satisfactorily used by previous investigators,

and, recommended the use of the same for its simplicity. Thus the

velocity coefficients u and v were "lagged" in this finite differ-

ence study.

Convection terms. The presence of the Lonvection terms makes

the temperature and the vorticity equation qualitatively different

from other simple parabolic equations. Proper numerical treatment of








the convection term is important. Several different ways of finite

differencing these terms have been presented in the previous works.

These include space-centered differencing and upwind or upstream

differencing. The upwind methods have several desired properties

that are not shared by the other finite difference formulations, as

applied to the convection term. Both the upwind methods are

transportive, meaning any perturbation is carried along only in the

direction of velocity. Both the methods are also conservative so

that they preserve the integral conservation relations in the

continuum equation. The first of the two upwind differencing methods

is as follows (Lilly, 1965; Forsythe and Wasow, 1960; Frankel, 1956)




(u) = uO. uO.
(ue) (u > 0)
az Az
(41)
uOi+1- ueO
u=- i+1 i u(u < 0)
AZ




Accuracy of this method is 0(At,Az). This was used by Barakat and

Clark (1966) to solve equations 26 through 29. The second variation

of this upwind method is to treat the convection term as follows

(Gentry et al., 1966)


(ue) URR L- UL
az AZ


where








OR= i (uR > 0)

= i+ (UR < 0)
(42)

eL = i- (uL > 0)
= eO (uL < 0)


where uR and uL are velocities as explained below. This latter

upwind method also has a formal accuracy of 0(Az). However, it

behaves like it is second-order accurate when the quantity
( 0 or w ) is slowly varying (Roache, 1982). Torrance (1968) demon-

strated that the second upwind method is indeed superior to the
first upwind method (Roache, 1982). The second upwind method was
used in the present formulation.
The velocities uR and uL are at the right and left face
respectively of the grid volume as shown in figure 4. As will be

explained later, a kinematically consistent velocity formulation

(Parmentier and Torrance, 1975) was used. In this formulation,
transport velocities (uj and vij in figure 4) are defined normal
to the grid volumes. Accordingly,


f f
UR= ui,j UL= Ui-l,j

f f
R= i,j L ,j-I



Following is a development of the finite differences for the

convection terms in the temperature equation. In the z direction,









a(ue) a(ue) an
3z an az

9(ue) 1
an az/an

R i,j- L i-1,j 1
An z z
Si+1- i-1
2 An

z +2- z-( UReij uLi-,j
zi+1- zi_1


UR > O, UL > 0 (a)


Similarly


a(uo) 2
az zi+1- z1 (uR i+l,j uLi-1,j)

z 2 z (URi+l,j uLei,j)
i+1 *i--


z i+ z i l i, uL i,j)
^ i+1 1- 1


UR < 0


UL > 0 (b)


UR < 0 UL < 0 (c)


UR > 0 UL < 0 (d)


To avoid the use of FORTRAN IF statements for checking the sign of

UR and uL in the equations (a) through (d), the equations are
combined as below


a(ue) (R-UR ) i+l,j+(UR+uRl-UL+ u)i,j-(uL+ i-1,j 2
t 2 z+ z i
1+1 -


(43)








In the r direction,

3(rve) ag a(rve) 1 (rve)
ar a-r 9 2r/la aE


1
r -1r
j+1 j-1
2A9


vRrj+ I/i j vLrj- 1/i ,j-1


2
rj+l rj1Rrj+ /2 ,j


Simliarly

6(rve)
6r


2
=r+1 2 (vRr1Nj+ 1,j +l rJ. 1


rj+1 rj-1 Rr+ 1/2 i + vLj- ij



rj+12 rj-1vR N -/ v i Lrj. If 1,]


VR < 0 vL > 0


VR < 0 vL < 0


VR > 0 VL < 0


Therefore, combining the equations


a(rvo)
ar

(VRIVR )r l 1/201 j+1+{VR+IVRI )r l/(L-IvL )rr ji j(,L+ vLI )r J/i j-1
rj+l-rj-1


(44)


VR>0


VL>0
VL


- v Lj- 1i ,j-1








Diffusion terms. Central differencing was used for the

diffusion terms in both the temperature and vorticity equation. The

diffusion terms in the temperature equation are finite difference

below.


a ao an a ae
-3 (7) = Tz Tn- a-z
0ae
1 TZ i+ 12,j
Saz/an An


a i
Tz- 1/2,j


9i+1,j 1i,j i-1,j
1 i+1 i zi zi-l
z z. i1 An
2i+1
2An


2 i__ Ei+l, ,j ei-l
z i+ zi1 i+ zi zi zi_-
i+1 i- +1 1 -


(45)


1 a ae 1 ac a ae
ar a(r) F r ar (r-


ae
r-liJ^

A&


ao
r-i/?


11
r ar
-5-


1 1
r rj+1 rj-
j 2A


S ,j+1 ij ,j ,j-1
rJI2 rj+ r 12 r. rj-
4/ rJ+1 r j-12 j -1


2. O e. .
r j(rj+1 r ) /2 j+1 rj/2 r -1


a20
az


(46)







The finite difference formulations for the time and space deriva-

tives are then put together in the ADI formulation. For each of the

temperature and vorticity equations, the equations are written for

the "half" time step and for the "full" time step of the ADI method.

Algebraic simplifications are done to convert the equations at each

of the time steps to a tridiagonal systems of equations. A glimpse

of the tridiagonal matrices for the "half" and "full" step is then

included.

The tridiagonal systems of equations are solved using the back

substitution procedure (Smith, 1978). The time step At and the

mesh sizes Az, Ar were chosen in such a way that the coefficients

(ai, bi, ci in equation 48a) are all positive and the matrices

diagonally dominant (bi > ai + ci). Richtmyer and Morton (1967)

showed that these two conditions are sufficient to keep the round-

off errors down, in the back substitution procedure. Introduction of

the second upwind differencing method satisfies the criterion that

coefficients ai, bi, and ci are all positive (see Kublbeck et al.,

1980). The second requirement of diagonal dominance was not investi-

gated any further in this study. However, at all times whenever the

numerical values of the coefficients ai, bi, ci were visually

checked, the diagonal dominance was found to be satisfied.

ADI formulation for the temperature equation is now developed.

For convenience, the differential equation for temperature is

repeated. The finite differences corresponding to each of the

derivatives in the temperature equation are then assembled in the

same order.








a3 a(ue) 1 a(rve)
TF az +r ar


a2e 1 a a
S= -z +T-F (r-r) }


At the "half" or intermediate step ("n+l/' written as "*" in short)

* n
9. .-e. .
AT/2

{(e-luR)Oi+1,jUn+lu uI +lul),_(j- +lul)i *,j
Zi+l-Zi-_1 I1uR- R i U,


(V nlv n n n ,n+1+(v+i
'r(r j+l-rj


*


n 0 n zn zn
r r j+- rjl1 rj+1 rj rj- r j -1
r. r -r r -r r.-r
3 j+1 j-1 +1 j j j-1


(47a)


At the "full" step
on+le .
0. J -e
1,j 1,3
AT/2

1 n n + n n n n n 0*
z{( l i+1,j + u+ "I- u"+ RuL ) i,j- ul ( L i-'1,j }


(V/20 +l I )r n+1- n r +1 n n n+1
S(v vj+ R Ij RL1 L r ij J 2vl )rj i 2ij-1
+ rj (rj+l-rj-_1


z -z
S2 il~ji
Zi+l-zi L Zi+l- zi


* *
. .- .
1,3 i-1,j
i i-i


n+l n+1l
+ ( 2 r.-0, .
-2J j++1/ j+1 ij
r (rj+l -r ) rj+ -r.
3 j+1 3-1 j+1 3


,n+1 -n+1

-l j j-1
'J-'J-l


(47b)


Vn n ) r .-? n ) -n
-(v L L /r iJ -(vL L rl j -1







After algebraic manipulation, temperature equations at the "half"
time step


-ai i-l ,j+bi i,j-ci i+,j = rhsi


(48a)


where


S zi+1- Zi+l i- Zi-i1


1 + 1/ (n+U n n+1nj) + 1 1 +
S 1/ n 1 1 1
b =-- R- RI) 2 i
i T z +1-z. R R L + i-z-. --zi z-zi

C i : ,)
zi+1 1-1 '+'z +1 i


rhsi = dn +ee +fe
i1 ij-1 1 i,j+l

1/2v n+ vnjl)r -' rj
d = r- r I + 1FTr -r r-r 1
(r+l j+ -1 J 1 j J+ J -1


e =-


vVc nl,_ rj _li( nl n r 1/
R R L- v r i 1 rj + 2+.r
rj rj+, -rj_1 rj ( j +I- j ) I
5 1 r+rj rj-rj


1 -jVR )rj rj +/2
f andR


and


f
uR = ui,j


f
uL = ui
L -1,j


f
R= Vi,j


f
L ,j-1








After algebraic manipulation, temperature equations at the "full"
time step


-a n+l- b l c +l = rhs
-a i,j-1 b1,j j i,j+l rhsj


(48b)


where

1/2(vn +n I )r r
j r.r. rr.r- -r. + r.
n Lin n
S(v +vnR )r -(vL- _V )rv j-/ 1 r lr r -

S. -r.(r + -r ) jr. -r. r.-r.j+1


1I Rn n ]r r
j r. r. -r. r. jl -r. r .1-rj)
j j3+1 -1 J 3+1-1 j +1 3


rhsj = d-l,j + eOj + fi+,j


u l n+ l 1
d = +
Zi+1-zi-1 Zi+l-zi-i)(zZiz-)


1 12 n n i n 1 1 1


(UR-n 1 RI ) i
f R +
zi+l-zi-l (zi+l-Zi-1)(zi+1-zi)



and


f
uR = u j
R = ij
f
uL = ui.lj
L -1,j


f
VR = Vi j

f
L i ,j-1








The tridiagonal matrices for the temperature equations at the
two ADI time steps are given below. At the "half" time step, the
equations are written for a column (going from the bottom wall to
the top wall). Specified boundary temperature is used both at the
top and bottom walls. *-


b1 c1

-a2 b2
-a3


-am-2 bm_2 -cm-2
-am-1 bm-1


rhsl+a1 0,j
rhs2
rhs3


rhsm-2

rhsm-l+am-l m,j
1 /I


(49a)

At the "full" step, equations are written for the rows (from
the centerline to the sidewall). At the centerline, the insulated
boundary condition (equation 75) is used due to symmetry. At the
sidewall, the specified boundary temperature is used.

bl-al cl i,1 rhs1
-a2 b2 -c2 i,2 rhs2
-a3 b3 -c3 ,i3 rhs3




-an-2 bn-2 -cn-2 0i,n-2 rhsn-2
n+
-an-1 bn-1 in-1 rhsn-l+cn-1 i,n
n-1i,n-1 i,n


(49b)








Upwind differencing of the convective terms in the vorticity equation


a(uw) a(uo)
az) = similar to (
az az

(uR-luRl)"i+l,j+ (uR+IURl-uL+IuLI)wi,j (uL+IuL)wi-l,j
zi+1- zi-1


a(vw)
ar


(50)

(vR- IVRI)wi,j+l + (VR+ IVRI VL+ IVLI)wi,j (VL+ IVLI)&i,j-1
rj+1- rj-1
j+1 j-1


(51)

Upwind differencing of the diffusion terms in the vorticity equation


(52)


+2 i+ mij- i-ij
zi+1- z-1 i+1- zi zi- z i-1


a1 a(rM)3



as a a(r))
= 1' aTr ar


1
rj+1- rj-1"
2A9


1 a(rw)
r+ 1/ ar


,j+ 1/?
AE


1 a(r O)
rj._ 1ar i J1/


2 rj+1i,j+l- r mi
rj+ rj-l rj+ 1/2(r+1 r.)
2 j+3+1 J


r.. j- r. l )
- J I/2, j- rj-1


(53)


Central differencing of the buoyancy term


ar r+1- rj1


(54)


a2
az2








ADI for the vorticity equation


a + a(u) + (v+) a2 2 + a 1 a(rw)
S+ I ar } = -GrPr-D + Pr az ar+ r 3r()


At "half" or intermediate step,

* n
13i ,j-Wi ,j
AT/2


Zi+l-Z -1Lj
1_ n nn l vn v nn nI
Iv v *-v l ,

+(r 1rj 1I ,j+R1 R R L vn+ v ,j -(vn j-1
j+1 :-1


n+1 n+1
1 -e P
S-GrPr2( i,j+1 i- 2 Pr
j+1 J-1 i+1 Zi-1
n n
2 Pr rj+1 ij+l- i,
rj+-r ) r j+/2rj+ r )
'j+1 j+-1 + /2 j+1 3


* *
mi+lj-ti ,j
i+l-zi
i+1


* *

i i-i


n n
rj i ,rj-1 "ij-1
rj. /2 r-rj- )


(55a)


At "full" step,

n+1 *
ST/2o-.
AT/2


+ 1n n un+nln (un+unI
Z +- (u I u I )uR i+l,j+(UR URI ULI i,_ UL il,j
i+1 1 l ,
1 n n n+1 n+lnln + n n+1 n lin+l
(rj +l-rj-1 )(vR-v )wij+1 vR R L L ij L L L ij- 1

n+1 n+1 *
-GrPr2 ij 1 i ,j-1 2 Pr Oi+1,j 'ij "Oi, i-1,j
Sr +1- -1 zi+-zi-1 zi+1-zi z -z i-1


n+1 n+1
2 Pr rj+oi j+1- r ij
(rj+lrj-1 r +1/2 (rj+l-r) i


n+1 n+1
rjoij -rj-1 "i ,j-1
r i-/2 r j-1r


(55b)








After algebraic manipulation

*
-ai j+ bimj -cii+,j = rhsi (56a)

where



1/2 (u +jul ) Pr
ai Z i+l-Zi-l zi+lzi-ilzi-zi-)



bi E z -z (un+iui-un+tunI) + rz 1- +z 1
1 A-T zi+ i-1R R L i+- i-1 +i+1 1i


1/2 (uR- R ) Pr
c. +
1 --z -
ci z i+1-z-1 ii+1-z -1)(i+l-zi


(n+l n+1
rhs. = D2. +Ewm .+Fwn -0.5GrPr2 ij+l j
S ,j-1 1,i 3 i j+1 rj+1 -rj


1/2 (v + vjI) Pr rj-1
d = (v -I+ L J-
j+r -1 rj1/2 rj+1 rj_-1 j r-1

n n n nlr
1 i (VR+IVI IvI) Pr rj 1
1 R R L L Pr 1
e T=- 1/2- r -r -r1 r r r rj _j/
S T-/ j+1 j-1 j+ -rj-1 2 rjl-r] rj



1/2 (VRVR Pr rj+1
f=1 +
rj+1-r1_ rj+1l-rj1 Jrj /2 rj+l-r )








After algebraic manipulation

-aj n+l b mn(l n+l
-a ,-1 + b i, + c ,j+ rhs (56b)

where


1/2 (vn+ vj ) Pr rj_1
aj= (r + r. r -r )Lr.-r. )
j+1 j-1 -1/2 j+ j j-1

(v n n n n
1 R+IVR -vL+Iv L + Pr rj 1 +
bj= A + 2(j+lrj-r (r -rj )Irj (rj+r-rj rj/2 1 rj-1r )
j+1 j-1 j+1 1



/2 (VR R Pr rj+
c.= -r +
J r -rj+l rj-r.jlrjl/ rj -rj



(en+1 -n+1
rhs. dnw +en .+fn -0.5GrPr2 J+1- ,J-
-1,j 1, i+1,j r j+1 -r


1/2 (u+I u~n I)
Zi+ -+1_Zi 1 1-1)



z i+1-z -z. zi+1-zi-1)(z-z
i+1 i-i LI) 1-1


1/2 (-n URI Pr
f =-I +
Zi+l-Z i-I zi+l-zi Zi+l-i)








The tridiagonal matrices for the vorticity equations at the two

ADI time steps are given below. At the "half" time step, equations

are written for the columns (from the bottom wall to the top wall).

Both the bottom and top wall vorticity values are used from previous

iteration.


b1 -c1 01,j rhsl+alw ,j

-a2 b2 -C2 "2,j rhs2
- -a3 b3 -c3 m3,j rhs3




-am-2 bm-2 -cm-2 "m-2,j rhsm-2
-am_-1 ,bm- m-l,j rhsm-l+am-l om,j

(57a)


At the "full" time step, equations are written for the rows

(from centerline to sidewall). The centerline boundary vorticity is

specified (=0) and the sidewall boundary vorticity is used from

previous iteration.
n+1+l
b1 -cl i,1 rhsl+al wi,O

-a2 b2 -C2 "i,2 rhs2
-a3 b3 3 ,3 rhs3




-an-2 bn-2 -cn-2 "i,n-2 rhsn-2
n+1
-an-1 bn i,n-1 rhSn-l+cn-I "in

(57b)








Discretization of the Elliptic Stream Function Equation
The stream function equation (28), obtained from the definition
of a u is also a statement of the circulation theorem
f ~z ar
which is

f w dA = J ch (58)
area perimeter
element of area


Thus the difference formulation of the stream function equation has

to satisfy the circulation theorem in addition to conserving mass.
Difference approximations that conserve mass are more difficult to
achieve in a curvilinear coordinate system as compared to the
Cartesian coordinate system.
To satisfy both the circulation theorem and the conservation of
mass, Parmentier and Torrance (1975) introduced the concept of

tangential velocities parallel to the control volume faces and

transport velocities perpendicular to control volume faces (Figure
4). In what follows, their approach has been adapted for the non-
uniform grid system.
The circulation theorem (equation 58) is written for a grid
volume around node (i,j) as shown in figure 4

ij ( 1/2-z 1(rj rj


-v t v t t u
S(vj i-1,j)(j+/2-r j 1/) (u j-uij-1)(zi+1i/2-z1/


t t t t
= 1i,j- i- ,j i,3 i',j-1 (59)
ij zi+/2-z 1/2 r.+ 2- r.
i + W zi 1/21/2



































j+1/2


- -- Zi-/2




I zj-I
r]. l


Notations of various quantities defined on a
non-uniform grid system in cylindrical
geometry


---Zi++I


i-I


0-
ii,'


I -1

r i./2 r] r


Figure 4.


iJ Z---+Z/2



u --
e-e--


rj-







The tangential velocities u ,j and v-,j at the faces of the

grid volume are obtained by finite differencing equation 29




ut ij+ ij
ij rj+l/2 rj+- rj j


(60a,b)


v - ip
,t i+lj i- ,j
i,j r (z i+- zi)


Substituting these definitions in equation (59)




1 i+l,jj 3J i,j-i-,j
I i+ Zi i-



1 ri j+l- ij 1iji,j-1
rI r )I-
rj+ 1/2-r 1/2 rj+ 1/2rj+1-r ) r.1/ -rj -1

(61)


Note that the same finite difference equation (61) can be obtained

from central differencing the stream function equation (28) without

introducing the circulation theorem. This is shown in Appendix A.
After algebraic manipulation,









- "i = Ai-,j+ B i,j + CJi + Di,j-1 + Eij+1


(62)


where


B (z -zi j z i i-i rj { (rj+l-rj) jr jr -j-l


+ 1 -1/22-+ 1 '-/1
C+ r-.zi.1


C [zi+ /2-zi_ 1/2 )Zi+-i )


D (rj+ 1/2-rj. 1/2 Jrj. 1/2 [r-rj-1




E =1
(r+ 1/2-r 1/2 rj+ 1/2 rj+1-r


The grid point velocities are

interpolation of the transport veloc


then achieved
t
ities u. j and
J


th


r. r ? t 1 r1/21 r t
rj +/- j .r /2u 'i rj+ 1/2- rj 1/2 u -1


. i- z l/? vt
vi 1j Zi+l/2- .i1/2 j,


+ i+z 1- zi t
zi+ 1/- zi- 1/-,


rough linear
t
i ,j


(63a)


(63b)


ui j
1 ,3


r i (z / J -zi- 1)









Second, transport velocities are defined normal to the grid volumes
*
in terms of a new set of stream function values i,




f i. ,j .,j-1 ,f i, 1- i-1,j (64a,b)
1, r ( 1/2j 1,j j z 1/2i zi-1/2




The stream function values i,j are defined at the corner of the

grid volumes. These i,j values are related to the previous stream

function values ij through a consistency criterion which says


f
Lt u u = 0
Ar+0 i,
(65a,b)

Lt v v = 0
Ar+O


ti,j was defined as



,J 2 i,j + j) + X ( ,j+1+ i+1,j+1) (66)


then, using definitions of vj and v (equations 64b and

63b), it was shown that the consistency criterion (equation 65b)

require



j + xj = 1 (67)
J J


Similarly, using definitions of uj and u it was shown that
II >J ui j








the weakest condition that satisfies equation 65a was the recursion

relation


1 + (2j 1) xj (68)
j, -1(68)
Xj 2j + 1


Using boundary conditions on the axis, it was found that



X0 = 0.25 (69)


The transport velocities defined this way conserve volume flow

exactly and are consistent with the grid point velocity field as the

mesh is refined.

The transport velocities uj and v are the velocities

used in the energy and vorticity transport equations. The grid point

velocities u. i and vi, are the velocities plotted in the

velocity plots.

To solve for I from the vorticity-stream function equation

61, we note that it is an elliptic Poisson equation as compared to

the parabolic equations for temperature and vorticity. Direct (non-

iterative) methods to solve the equation are available that are

considerably (10 times) faster than an iterative method like the

successive over-relaxation technique (to be discussed later).

However, direct methods usually require large amounts of storage,

and available computer codes are often limited to rectangular or

simple domains or have restrictions on the type of boundary

conditions. Shih (in press) discussed suitability of direct methods








for "large" systems of equations. Direct methods also have round off

error propagation problem. With direct methods, round-off errors can

be incurred at each mathematical operation, and simply accumulate

until final answers are obtained. When iterative techniques are

used, the presence of round-off errors at the end of any given

iteration simply results in those unknowns being somewhat poorer

estimates for the next iteration. For practical purposes, round-off

error in the final converged values of an iterative scheme is only

that accumulated in the final iteration (Hornbeck, 1975). A very

efficient direct method for solving Poisson's equation was described

by Schumann and Sweet (1976) and the code was available from the

National Council of Atmospheric Research (NCAR) software library.

However, on closer look the algorithm (GENBUN) was found to be

limited to coordinates transformed in one direction only. Since in

this study the boundary layers were expected to be present both

along sidewall as well as top and bottom wall, coordinate stretching

were considered necessary in either direction. This made the above

algorithm unsuitable for the present study.

A popular approach to solve such a steady state equation is to

formulate it as an unsteady equation and solve for the asymptotic

steady state solution. ADI methods have been used for this purpose.

These methods were known to be faster than the successive over-

relaxation method (to be discussed later) for a rectangular region

(Birkoff et al., 1962). However, the optimum sequence of parameters


aR. Sweet. U. S. Dept. of Commerce. National Bureau of Standards.
325 Broadway, Boulder, Colorado. Personal Communication.








that makes the ADI method advantageous is not available in general.

Programming the ADI method for a non-rectangular region may get

complicated, and they are not known for sure to be faster (Roache,

1982).

Another popular method to solve the unsteady formulation of the

stream function Poisson equation is the successive over-relaxation

(SOR) method (Frankel, 1950). It is very simple and effective. In

SOR, the set of solutions are over relaxed by a factor 1 < n < 2

to speed up convergence. The exact value is found through simple

numerical experimentation. It is easier to program compared to ADI

methods. Most of the earlier researchers have used the SOR to solve

the stream function equation. In all the five formulations of

natural convection in a cylindrical geometry, studied by Torrance
(1968), SOR was used to solve the stream function equation. However,

it is time consuming and could end up taking 90% of the total time

spent to solve the problem. Nevertheless, SOR was chosen to solve

the stream function equation in the present study for its

simplicity. Following is the development of the SOR formulation.

Rewriting equation 62,



ij = C i-l ,j+Ci+,j+Dij-l+Eij+l+ij


To get the SOR formulation,


-k+1 1 k+1 k Dk+ + k (70)
i,j B -1,j i+l,j ij1 1i,j+l+ i,j








k+l k+ (71)
i,j ,j i,j


Using equation 71, stream function values ,j at all interior

points are updated. The entire sweep is repeated until the

convergence criterion

k+l k
Max ij 1 ,j (72)
i ,J


is satisfied. The value of e chosen for this study was 10-5. In a
*
transient calculation like this study, the optimum Q may change
*
from one time instant to another. However, the optimum Q found

through calculations during the very first time step was used

throughout the entire transient calculations.



Computational Boundary Conditions for the Finite Difference
Equations

A PDE such as equation 27 for w can describe a wide variety

flow situation. The flows (solutions) are distinguished only by the

boundary and initial conditions and flow parameters like the Grashof

number. Thus, specification of computational boundary conditions,

besides affecting numerical stability, greatly affects the accuracy

of the solution (Roache, 1982). Using the boundary conditions

already noted for the partial differential equations in the previous

chapter, the boundary conditions for the finite difference equations

are now derived.

The boundary condition for temperature is rather simple. When

the can wall temperature is specified, the computational wall








temperature value of the fluid next to the wall is made equal to the

specified value. On the adiabatic walls and at the centerline (which
is also considered adiabatic due to symmetry), a Taylor series
expansion gives



1 W + 0 An +1/2 j An2 + O(An3) (73)

ao
where stands for the derivative of temperature in the direction

normal to the wall and An being distance normal to the wall.


for = 0, (74)
for = +20W


OW+1 = W +1/2 820' An2
an2

= e + 0(An2) (75)


Thus if eW+1 = OW is set, wall temperature is evaluated from the
adjacent temperature and it is second order accurate even though
ao is only first order accurate (Roache, 1982).

In the ADI method, boundary conditions are needed at the
computational "half" step. Since wall temperature 0W is not a
function of time, the boundary temperature value at this "half" time

step is simply equal to eW.
Unlike temperature and velocities, vorticity and stream
function are not primary variables. This makes specification of
their boundary values difficult, particularly for vorticity.









Improper specification of boundary vorticity is found to be

destabilizing in general and the same was found in the present

problem. Vorticity is produced at no-slip (wall) boundaries. It is

then diffused and convected into interior points. No explicit

expression for wall vorticity is available. It is related to the

interior stream function values by the vorticity stream-function

equation (equation 28). However, this would make the vorticity

equation implicit with the stream function equation and the number

of simultaneous equations increase drastically. To avoid this,

Barakat and Clark (1966) chose explicit methods over implicit and

used vorticity boundary values calculated at the previous time step.

Wilkes and Churchill (1966) used the ADI method but similarly used

boundary vorticity values from the previous time step. Thus, in

their formulations, wall vorticity values lagged by one time step at

all times.

Roache (1982) discussed the effects of such lagging of boundary

vorticity values. It forces the use of a smaller time step since

larger time steps could make the solution inaccurate and unstable.

Since larger time steps are a major motivation for ADI, updating of

the boundary values through iteration is preferred. From the work of

Torrance (1968), Roache also noted that the convergence of the

boundary vorticity values actually imposed a time-step restriction

of the form AT < a / Az2 where "a" was some number dependent on the

problem and convergence requirements. Thus, due to the implicitness

of the boundary vorticity values, ADI actually has a time-step

restriction like the simplest of the explicit methods. The








definition of vorticity o was used at the boundaries to obtain

expressions for boundary vorticities. At the bottom boundary,


t
SO,j- v,j
O,j (z l zO)


t t
UO,j- UO,j-1
( rj+ 1/-rj 12


- 1,j
r.z z12


since
t
Vo,j


t
uj= 0


_ ( ,j-Oj)
rj(z1- z0)


S0 from eqn 60a
and uj-=j 0 from eqn. 60a


At the top boundary,


t
SVmj- Vm-lj
m,j= (z z


t t
Smj- um,j-1
( rj/- r.j


_- m-l,j
r. (z -z z -z 1
J m m-1 m m-2


since


v .= 0


m- ,j z -
S m ,jm-1,j
Vm-l,j-jr rn(Z-mI)


- im-1,j
- r.(z -z )
I m m-1


utj and u t ,- 0 from eqn. 60a
mij mll=,


(76a)


Ip,j
r z1
^J


(76b)








At the centerline,
t t t
i,0- Vi-l,0 ui,0- Ui,
i,0o- z. i+- z.1, r i/ 0

=0 (76c)

since

t O and v-, from eqn. 60b
1,0 1a1,0

t i 4,I- -i,0 2_ 1
io rl- O) rli

ui,0=0

At the sidewall,
t t t
S,n Vi-l,n i,n- i,n-
i,n= (z+l/2- z l/ (rn- rnl/


= in-1 (76d)
r ni(r -r )(r r
n n n-1 n n

since

Vi, and vi 0 from eqn. 60b
1i ,n 1i- l,n=

t i,n i,n-1 1,n-1
1i,n-1 r/rn- r r r n -rl)
n -/2 n n-1 n_/2 n n-1

u. =0


Identical expressions can be derived using Taylor series expansions

of equation 28 at the walls. This is shown in the appendix B.

These expressions are first order accurate. But it is the only

form consistent with the kinematic velocity formulation (Paramentier

& Torrance, 1975). The second order form using Taylor series

(Jensen, 1959) has also been used. However, sometimes the second









order form destabilized the calculations (Wilkes & Churchill, 1966)

and sometimes it was less accurate than the first order formulation

(Beardsley, 1969).

The vorticity boundary value at the computational intermediate

or "half" step of the ADI method is more difficult to set than the

corresponding temperature value because the vorticity boundary value

is also a function of time. The following simple way to define this

vorticity value on the boundary is often used and gives sufficiently

accurate results (Bontoux et al., 1978)


n+1/2m+1 n n+1m+l
W = 1/2(W+W ) (77)


where

n = wall vorticity value at time step n

m+l
n+1 = most recent estimate of wall vorticity at n+1
step using equation.

m+1
nmW2 = intermediate vorticity value between n and
(n+l)th time level at the (m+l)th iteration.



Israeli (1972) also used this equation to define intermediate (half

step) boundary vorticity though he used a different equation to
nm+l
calculate boundary vorticity wW



Convergence Criteria for Iteration

As discussed earlier, the set of finite difference equations

were solved iteratively until convergence. The measure of conver-









gence could have several possibilities. Ideally, at convergence, all

the variables should have converged. For practical reasons, it would

be sufficient to check on the variable that converged the slowest.

This is so that when this variable has converged, all others would

have converged. However there are no general guidelines to choose

apriori the slowest converging variable. Interior velocity

components were suggested by Roache (1982) as a measure of

convergence. Frequently, researchers in the past have decided on the

vorticity boundary values as the variables to watch for the

iteration convergence of the entire problem. Roache (1978) referred

to tests on boundary vorticities as the most sensitive test for

convergence. The same variables were used to check for iteration

convergence in this study even though they may not have been the

variables that converged the slowest.

For the chosen variable, there are no definitive criteria for

the measure of convergence. A relative error criteria was used for

this study, which is

-n+m+1 ~n+lm
< 6 (78)
-n+1


at all solid boundaries. The boundary vorticities n+1 were found

from the interior stream function value using equation 76.

As an example of other measures of convergence, while studying

natural convection in a rectangular cavity, Kimura and Bejan (1984)

used the following convergence criteria on interior points







E E n+lm+1 n+1i
i j i -ij < 6 (79)
1I1m+11 6
n+1
i j
together with a convergence criterion on the calculated Nusselt

number.

The choice of 6 is also problem dependent and no general

guidelines exist for it either. Even for very small values of 6,

the iteration of a solution may be stopped prematurely for certain

types of solutions which are not uncommon (Roache, 1982). Values

ranging from 103 to 10-4 were used in their study, and the

chosen value was the largest of these values that did not cause any

perceptible changes in the final solution.



Algorithm for the Iterative Solution

The algorithm for the iterative solution of the complete system

of finite difference equations (equations 47,55,61,64) is now

described. The grid system is generated first. The vorticity, stream

function and velocity values are set to zero which is their initial

condition (t = 0) since initially the liquid was stationary with

uniform initial temperature throughout the body of the liquid. The

non-dimensional temperature values are set to one (initial temper-

ature). The temperature boundary condition for t>0 is then enforced.

The set of equations are then solved for finite increments of time

(At) with iterations performed on boundary vorticity at each time

increment. This is described in figure 5 and explained as below. The

selection of time step the At is discussed under the RESULTS

section.








REPEAT OVER n
*n n"
lm
2=1.0

r
ne m 1.0



ITERATE OVER m
n

I r
=.


P +
r e1


n/2m= f, n, un+1m vn+lm, /2m)
Q 1 Q I


n+1m 2 n12m
0 Q f 2( .


n+1m n+1m
u Q v
"a a


n+lm
W )


n+lm
en + 1


n /2mo
wD = 0


2n1/2' f r n n4
n +3
w = f,(w, u
n+lm I
U) =O
r
n+1" f n 1/2

^ n+lm
) = Wr

ITERATE


m
n+lm
-


n+lm
, Q


,n/2m
, p 2


n+1m n+l1
uS vS


n1/2m n+1m
4 w 'a


n+1m
"r


f5 (n+1m
fo 0


(,n+lm
S 7 j,


-n+m+l
= n+
=yw


m
S,, n+
+ (1-y)w
r


UNTIL -

u(vn+1m
(u,v) l


r: solid boundary
Q: interior points
: centerline
I ^ *
m,w,w: temporary
storage


( n+1m
-f r~l!


UNTIL DESIRED TIME REACHED


Figure 5. Algorithm for the iterative solution
of the set of equations


UNTIL -


-n+m+1
n+1
"r
n+lm+l
Wr


n+1m n+lm
W( 5 O









1. The temperature equations (47a and 47b) were solved using

the ADI technique. This gave new temperature values after

incremental time At.

2. The vorticity equations (55a and 55b) were then solved

(also using the ADI technique). The interior vorticity

values calculated were provisional since old values were

used for boundary vorticity.

3. The stream function equations (61) were then solved for

interior stream function values using the point SOR

technique. This required only interior vorticity values.

Previous stream function values were used as an initial

guess.

4. Boundary vorticity values ( ab) at solid walls (top,

bottom and side) are first calculated-using equation 76

based on the newly calculated stream function values.

However, before updating, the boundary vorticity values

(b) are underrelaxed as

m+1 m+1 m
n+1 n+l1 n+ (80)
"b = b + ( y) Wb

m+l
where n+ = boundary vorticity estimated using

equation and 76, with Oy
5. Steps 2 through 4 are repeated until the estimated wb

values (using equation 76) in two consecutive iterations

meets the convergence criterion







~n+lm+1 ~n+lm
ab %b
< 6 (78)
-n+l
Wb

for top, bottom and sidewall. When the convergence

criterion is satisfied, temperature, vorticity and stream

function values are considered the transient solution at

this time instant.

6. The velocities are calculated from the stream function

values using equation 63 and 64

7. Steps 1 through 6 are repeated for every time increment

until the desired time is reached.



The under-relaxation of boundary values at step 4 has often

been considered necessary in the literature (Peyret and Taylor,

1983; Roache, 1978) for convergence of the solution. Under-

relaxation is usually employed to make a nonconvergent iterative

process converge (Hornbeck, 1975). This was found quite true for the

present study. Without under-relaxation, the time step allowed for a

stable solution got severely reduced to the point that solution to

the set of equations became impractical because of the amount of CPU

time required. As the computations proceeded, there were two sets of

values in storage. These were the estimates of the temperature,

vorticity and stream function values at the present time step and

converged temperature, vorticity and velocity values at the previous

time step.








Coding of the Program and the Hardware

The algorithm presented in figure 5 for solving the vorticity-

stream function formulation of the natural convection problem was

coded in FORTRAN 77 with individual modules for solving temperature,

vorticity, stream function, velocity and boundary vorticity values.

Structured programming practices were used. Use of GO TO statements

were avoided except in two situations to simulate a WHILE statement.

The WHILE statement was unavailable in the particular version of

FORTRAN 77. Single precision of the variables was known to carry 14

significant digits for the particular computer and the same was

taken to be sufficient for this problem.

The input data file for a typical run is in Appendix C.

The computer used was a CDC Cyber 170/730, with two 60 bit CPUs

crunching 60 bit words at 400 MHz, 20 parallel processors, and 1091K

of 60 bit word central memory.

The plotting was done using a plotting package from the

University of Illinois1. The convenience offered by the plotting

package cannot be overemphasized. The plots were viewed on a

Tektronix 4006 terminal and hardcopies were made on Tektronix 4662

plotters.



Modeling of Conduction Heating in Cans

The partial differential equation (PDE) governing conduction

heating of a cylindrical-shaped solid was discretized and finite


1W. E. Dunn, Mechanical and Industrial Engineering Department,
University of Illinois at Urbana-Champaign. Personal Communication.








difference using an alternating direction explicit (ADE)

differencing method, which is explicit but still unconditionally

stable. The transient temperature values predicted by performance of

this ADE finite difference method was compared with analytical

solutions and experimental measurements.


Discretization and Solution of the Equation

Keeping in mind that the equation was needed to be solved in

real-time for on-line control purposes and the ADE method was

already tested (Teixeira et al., 1975; Teixeira, 1971), the same was

selected. The ADE formulation was first reported by Saul'yev (1957).

This was later translated and described by Allada and Quon (1966),

and was first introduced for applications to thermal process

simulation by Manson et al. (1974). The method as applied to the

diffusion equation is unconditionally stable (Saul'yev, 1964) for

temperature-specified boundary condition. It has a formal order of

accuracy of O(At2, Ar2, AZ2 ). Being explicit, it does not require

inversion of matrices as would be required by an implicit method. In

a fluid flow problem, the ADE formulation is generally not preferred

since the presence of the convection term (in any difference form)

makes the complete formulation unstable or returns it to simple

explicit stability limits ( a / for a one-

dimensional problem; Roache, 1982). Another shortcoming of this ADE

method is the fact that even though it is an explicit method, it is

actually implicit in boundary conditions. To handle this implicit-

ness, either explicit expressions for the temperature values need to









be derived algebraically for each of the boundaries or the computer

has to solve a simultaneous set of equations at the boundaries (see

Appendix D). For Robin's boundary condition, the application of the

ADE method required separate algebraic expressions for the top, side

and bottom wall boundaries. Separate expressions were also required

for the top and bottom corners and the centerline. This is in sharp

contrast with the situation if the ADI method were used. Adjustments

would be necessary at the boundaries for the ADI method, but for use

of the ADE method it became clumsy (see appendix D).

For temperature specified at the boundary, this is trivial and

no additional manipulations as noted above are needed. Experiments

conducted later showed that because of the low thermal conductivity

of food (which is mostly water), the Biot number (= hR/k) is quite

large and it is the internal conduction resistance that is limiting.

Thus a convection boundary (Robin's) condition can be approximated

by a temperature specified (Dirichlet) boundary condition for this

purpose. The two shortcomings of the ADE formulation are, therefore,

of little consequence in the case of conduction-heated food. The

equations developed for the convection boundary condition are given

in appendix D.



Coding of the Computer Program

The ADE formulation for transient temperature calculation using

temperature specified boundary condition was coded in PASCAL and

used as a subroutine in a real time control algorithm (to be

discussed later).










Experimental Studies

Experiments were performed to record the transient temperature

values inside cylindrical cans during conduction heating. The effect

of different surface heat transfer coefficients, that were obtained

by using different media (steam and water), on the transient

temperature response was also noted.

The experimental work consisted of series of heat penetration

tests on No. 303 cans filled with a 10% bentonite suspension as a

conduction-heating food model system (Niekamp et al., 1984). Eight

cans were fitted with an Ecklund-type copper-constantan thermocouple

mounted mid-way up the can wall, so as to have the junctions located

at the geometric can center. In order to achieve the two different

levels of head-space vacuum, four of the cans were filled at 900C

and the other four cans were filled at 670C. The cans were

identified according to fill temperature and then processed with an

extended steam-cook to stabilize the bentonite suspension. An

experimental retort was used that had been designed and built at the

University of Florida with the capability of using pressurized

water, pure steam, or steam/air mixtures under controlled rates of

recirculation. The retort was fitted with a special rack designed to

hold the eight cans suspended upright in the mainstream of the

retort chamber, as shown schematically in Figure 6.

The rack consisted of a stainless steel plate with eight

cicular holes through which the cans could be inserted. Special

fixtures were welded at the edge of each hole so that the cans could




Full Text
124
natural convection heating, even though most of the core exhibits
uniform radial temperature due to stratification, the symmetry about
the mid height is lost. This is due to the convective loops from the
bottom heating creating isotherms completely different from the con
duction heating. Also formation of a boundary layer at the sidewall
in natural convection heating clearly separates it from the
conduction heating situation.
Assesment of the Numerical Method
Convergence. The solution to the set of finite difference
equations should approach the true solution to the partial differ
ential equations having the same initial and boundary conditions as
the mesh is refined. This is referred to as convergence of the
finite difference solutions. For linear PDEs, Lax's equivalence
theorem states that for a consistent finite difference approximation
to a properly posed initial value problem, stability is the
necessary and sufficient condition for convergence. The same theorem
has not been proven for non-linear PDEs (Anderson et al., 1984).
Thus convergence cannot mathematically be proven for the present
problem.
However, as shown in the previous sections of this chapter, the
axial temperature profiles and the radial velocity profiles compared
very well quantitatively with available experimental data. Several
phenomena revealed by the solution including formation of the tem
perature and velocity boundary layer, stratification, and, Benard
type convection were expected, and, have all been observed by


89
It is unconditionally stable, thus allowing large time steps. It is
also explicit, thus taking small computation time per step as
compared to many implicit methods.
The time increments At, At^, At^ and the finite difference
calculations were adjusted so that the inequality (72) was
satisfied. On a DEC PDP 11/23, the following computation times were
found to be feasible
At 12 seconds
At. =1.5 seconds
Atp = 9 seconds.
The control logic flow diagram for the retort thermal process
is given in figure 8. The input data are first checked against
specifications for the product. If they agree, steam is turned on
and the computer completes the venting cycle. The retort temper
ature comes up and the computer through a controller interface,
attempts to maintain the design retort temperature T^ As the
heating cycle continues, retort temperature TR(t) is read at
intervals of At T(t) is then calculated from TR(t) using the
finite difference approximation of equation 32. Using equation 35a,
F0(t) is then calculated. The input data includes a specified
pheat ^ which is the F0 value normally achieved at a time when
heating would be terminated for a design total F of F ^ When
oo
F0(t) exceeds the value p^eat t the computer also simulates the
cooling cycle in addition to calculating T(t) and F0(t). If the
F0(t) accumulated so far, together with the simulated contribution
from cooling, exceeds the design total FQ value for the process
(Fq) i.e. when


23
experimentally determined temperatures at three axial and two radial
locations. Thermistor thermometers were used for temperature
measurement. The calculated temperature values were reported to be
within the limits of experimental error, over a range of Gras'nof
numbers.
Evans et al. (1968) measured the transient temperature field
during natural convection of fluid in a vertical cylinder due to
uniform heat flux. The top surface of the liquid was free and no
transfer of heat or mass was assumed at the liquid-vapor interface
at the top. Dye tracers were used to qualitatively observe the flow
patterns. They reported symmetry about the vertical axis in the
temperature field. On the basis of the experimental observations,
the convective liquid flow was modeled in terms of three regions:
1. a mixing region at the top
2. a central core
3. a boundary layer rising at the heated wall
They developed three different equations to describe the three
regions. These equations were solved simultaneously to predict the
overall behavior of the system. They reported good agreement between
the model and their experimental data.
Shyu and Hsieh (1985) made a numerical study of unsteady
natural convection in thermally stratified water in a cylindrical
enclosure. They studied the effect of placement of insulation on
maintainence of the stratification. Transient flow and temperature
fields were provided for the study.


76
Z Z
i j
n+1
ij
m+1
0)
z z
i J
n+1
di j
n+1
dij
m+Ti
mi
(79)
together with a convergence criterion on the calculated Nusselt
number.
The choice of 6 is also problem dependent and no general
guidelines exist for it either. Even for very small values of <5,
the iteration of a solution may be stopped prematurely for certain
types of solutions which are not uncommon (Roache, 1982). Values
-3 -4
ranging from 10 to 10 were used in their study, and the
chosen value was the largest of these values that did not cause any
perceptible changes in the final solution.
Algorithm for the Iterative Solution
The algorithm for the iterative solution of the complete system
of finite difference equations (equations 47,55,61,64) is now
described. The grid system is generated first. The vorticity, stream
function and velocity values are set to zero which is their initial
condition (t = 0) since initially the liquid was stationary with
uniform initial temperature throughout the body of the liquid. The
non-dimensional temperature values are set to one (initial temper
ature). The temperature boundary condition for t>0 is then enforced.
The set of equations are then solved for finite increments of time
(At) with iterations performed on boundary vorticity at each time
increment. This is described in figure 5 and explained as below. The
selection of time step the At is discussed under the RESULTS
section.


Figure 17. Locations of the slowest heating points in
water over a heating period of 10 minutes in a
cylindrical can
N
FKMCP
D I L A
Figure 18.
Migration of the slowest heating points in
water over a heating period of 10 minutes in a
cylindrical can


25
3T
3T
+ v
3T
+ U
3r
3T
3z
_ K ,13 3T w 32T >
= TT" (t~ (r ) + )
p p r 3r 3r 3z2
(13)
p
(. +
' 3t
P (
V
3t
+
- 3 3 \ 3p J 1 3 32 1
V+u ) + p| T (rr) + T" I + P9
3r 3z 3z F 3r 3r 3z2
v + G J ) = iE. + (Fv)) + }
3F 3z 3F 3F F 3F 3z2
(14)
(15)
I_i (rpv) + i-(pG) =0
F 3F 3z
(16)
The boundary conditions are
sidewal1
F = R
0 < z < H
T=Ti
O
11
1 3
< 1
II
o
centerline
F = 0
0 < z < H
iI=o
l=.
tl
o
o
It
1 >
3F
3r
bottom wall
(17a-d)
0 < F < R
z = 0
T=Tl
= 0
O
II
1 >
top wall
0 < F < R
z = H
T=Tl
G = 0
o
II
1 >
Initially the
fluid is at rest
at uniform temperature
0 < r < R
0 < z < H
o
i
ii
i
= 0
v = 0 (18)
Refer to list of symbols for definition of individual symbols.
As expected, the flow field (equations 14 & 15) is coupled with
the temperature field (eq. 13) through the density term. These


velocity (mm/s)
Figure 15. Observed radial velocity profiles at the
sidewall at mid height in water in a
cylindrical can after various heating times
(from Hiddink, 1975)
112


79
-n+1
m+1
0).
~n+l
%
m
~n+lm
%
< 6
(78)
for top, bottom and sidewall. When the convergence
criterion is satisfied, temperature, vorticity and stream
function values are considered the transient solution at
this time instant.
6. The velocities are calculated from the stream function
values using equation 63 and 64
7. Steps 1 through 6 are repeated for every time increment
until the desired time is reached.
The under-relaxation of boundary values at step 4 has often
been considered necessary in the literature (Peyret and Taylor,
1983; Roache, 1978) for convergence of the solution. Under
relaxation is usually employed to make a nonconvergent iterative
process converge (Hornbeck, 1975). This was found quite true for the
present study. Without under-relaxation, the time step allowed for a
stable solution got severely reduced to the point that solution to
the set of equations became impractical because of the amount of CPU
time required. As the computations proceeded, there were two sets of
values in storage. These were the estimates of the temperature,
vorticity and stream function values at the present time step and
converged temperature, vorticity and velocity values at the previous
time step.


3
Convection-Heated Foods
Unlike solid packed foods that are heated by conduction, thin
liquid foods heat through convection. In general such liquid foods
are mechanically agitated to improve the heat transfer. However,
there are situations where the food material cannot be agitated for
several reasons. The food then heats primarily by natural convec
tion. In-package pasteurization for liquids (e.g. beer) in bottles
or cans that cannot be agitated is an example. Broth, thin soups,
evaporated milk, most fruit and vegetable juices, fruits in syrup or
water, and pureed vegetables are also examples of products where
natural convection heating can occur when no agitation is used.
Flexible pouches that cannot be agitated to keep the package integ
rity are another important example. Fermentation is another process
where a natural convection heat transfer model (with heat generation
term included) can provide valuable insight.
Just as optimum heating times are required for conduction-
heated foods, they are also needed for natural convection-heated
products. Thus heat transfer models are similarly needed. Since we
are interested in the minimum heat treatment the food receives, we
need to know the point or the region that heats the slowest. Because
convection sets up complex flow patterns, the location where the
temperature is minimum is by no means obvious. Only a study of
complete transient flow patterns and temperature profiles can give
such insight. Previous studies to achieve this objective have not
been successful. The use of more advanced numerical techniques and
better computing equipment would make it possible to study in detail
the natural convection heat transfer in canned liquid foods.


Water
T>30*C
!op-121*C
sldo=121*C
bo!-121*C
R-4.19cm
H=10.67om
R-57,0.76
Z=57,0.85
8-0.0001
-300
260
-220
-180
-140
-100
-60
-20
20
60
Figure lOg. Isotherms and streamlines in a
cylindrical can after 7 seconds of
heating
After 8 sec
Water
r,=30*C
top-121*C
slde=121*C
bot-121*C
R-4.19cm
H=10.67om
R-57,0.76
Z=57,0.85
6-0.0001
-300
-260
-220
-180
-140
-100
-60
-20
20
60
Figure lOh. Isotherms and streamlines in a
cylindrical can after 8 seconds of
heating


9
The solution was first approximated to
V tr
-U2 + X2) at
A e J0(xr) Cos(xz)
(3b)
when sufficient time has elapsed, so that the first term in the
analytical series solution (equation 8a) dominates. Equation (8b)
was then rearranged to give
t = f log
J V T
(9)
where
and
_ 2.303
a (Xz + X2)
(10)
j (r,z) = A J0(xr) Cos (xz) (11)
From the transient heat conduction equation (9) which gives
temperature T as a function of time t and the definition of FQ
(equation 6) which gives F0 as a function of temperature T, it is
possible to find F0 as a function of time t. However this cannot be
performed analytically. Upon solving the two equations numerically,
it is possible to know temperature achieved at the end of heating
for a given value of F0. Ball and Olson (1957) tabulated T^-Tr
values as a function of FQ for particular values of f and j. Tradi
tionally, in thermal processing of conduction-heated foods in cylin
drical containers, these tables are used to find T^-Tr for a given
F0 value. Using this (Th-TR) value and equation (9), the process
time t is found.


125
110
O
o
go
ZJ
i
o
L-
CD
Ct
70
50

*
*
t
30
* 39(0.9)X39(0.B) grid
y 58(0.76)X58(0.58) grid
i i
0 120 240 360 480
Time of heating (seconds)
Figure 26. Transient slowest heating temperatures computed
with 39x39 and 58x58 grid
600
no
<43


Superscripts
d = design value
f = normal to grid faces
k = iteration counts in SOR
m = iteration count within a time step
n = time step
t = tangential to grid faces
* = "half" step in ADI method (same as n+l/2)
x


169
Hunter, J. H. 1982. Field testing of a microcomputer potato storage
control system. ASAE paper no. 82-4031. Presented at the 1982
summer meeting of the ASAE at Madison, Wisconsin.
Kunhardt, E. E., and P. F. Williams. 1985. Direct solution of
Poisson's equation in cylindrically symmetric geometry: A fast
algorithm. 0. Computational Physics 57:403-414.
Larkin, B. K. 1964. Some stable explicit difference approximations
to the diffusion equation. Mathematics of Computation 18(86):
196-202.
Lenz, M. K., and D. B. Lund. 1977. The lethality-Fourier number
method: experimental verification of a model for calculating
temperature profiles and lethality in conduction-heating canned
foods. J. Food Sci. 42(4):989-996.
Lenz, M. K., and D. B. Lund. 1977. The lethal ity-Fourier number
method: experimental verification of a model for calculating
average quality factor retention in conduction heating canned
foods. J. Food Sci. 42(4):997-1001.
Lenz, M. K., and D. B. Lund. 1977. The lethal ity-Fourier number
method: confidence interval for calculated lethality and mass-
average retention of conduction-heating canned foods. J. Food
Sci. 42(4):1002-1007.
Levy, C., and R. G. Hlavacek. 1980. Computer control optimizes
retort cook for yield, quality, safety(canned foods). Food
Processing 41(11):74-77.
Merson, R. L., R. P. Singh, and P. A. Carroad. 1978. An evaluation
of Ball's formula method of thermal process calculations. Food
Tech. 3:66-72.
Miller, C. W. 1977. Effect of a conducting wall on a stratified
fluid in a cylinder. Paper no.77-292. Presented at the AIAA 12th
Thermophysics Conference, Albuquerque, New Mexico.
Mitchell, A. R., and D. F. Griffiths. 1980. The finite difference
method in partial differential equations. John Wiley & Sons.
Olson, F. C. W., and J. M. Jackson. 1942. Heating curves- theory and
practical application. Industrial Engg. Chemistry 34:337.
Ostrach, S. 1972. Natural convection in enclosures. Advances in Heat
Transfer. 8:161-227.
Ozoe, H., K. Yamamoto, S. W. Churchill, and H. Sayama. 1976. Three-
dimensional, numerical analysis of laminar natural convection in
a confined fluid heated from below. J. Heat Transfer, Trans of
the ASME 98:202-207.


TEMPERATURE (C)
TIME (minutes)
Figure 33. Computer plot of reference process (no deviations) showing
66.8 minutes of heating time and an accomplished FQ of 6.24
ACCOMPLISHED F0 (minutes)


TemperatureCc) -n Axial velocity (mm/sec)
igure 11
Predicted radial velocity profile at the
sidewall at mid height in a cylindrical can
after 30 seconds of heating
108
Distance from sidewail(mm)
Figure 12. Predicted radial temperature profile at the
sidewall at mid height in a cylindrical can
after 30 seconds of heating


lOf. Isotherm and streamlines in a cylindrical can after 6
seconds of heating 101
lOg. Isotherm and streamlines in a cylindrical can after 7
seconds of heating 102
lOh. Isotherm and streamlines in a cylindrical can after 8
seconds of heating 102
lOi. Isotherm and streamlines in a cylindrical can after 9
seconds of heating 103
lOj. Isotherm and streamlines in a cylindrical can after
10 seconds of heating 103
10k. Isotherms and streamlines in a cylindrical can after
30 seconds of heating 104
101. Isotherms and streamlines in a cylindrical can after
120 seconds of heating 104
10m. Isotherm and streamlines in a cylindrical can after
300 seconds of heating 105
lOn. Isotherms and streamlines in a cylindrical can after
600 seconds of heating 105
10o. Isotherms and velocity vectors in a cylindrical can
after 1800 seconds of heating 107
11. Predicted radial velocity profile at the sidewall at
mid height in a cylindrical can after 30 seconds of
heating 108
12. Predicted radial temperature profile at the sidewall
at mid height in a cylindrical can after 30 seconds
of heating 108
13. Isothermal vertical surface, showing boundary layer
with velocity profile u(x,y) and temperature profile
t(x,y) in water (Higgins and Gebhart, 1983) 110
14. Predicted radial velocity profiles at the sidewall at
mid height in a cylindrical can after various heating
times Ill
15. Observed radial velocity profiles at the sidewall at
mid height in water in a cylindrical can after
various heating times (from Hiddink, 1975) 112
16. Comparison of numerically predicted axial temperature
profiles with observed (Hiddink, 1975) values at
various times during heating 114
xi i i


106
counterclockwise loop extended over most of the can and represented
the primary phenomena of upward movement of hot liquid at the side-
wall and its downflow at the core. Several small clockwise and
counterclockwise secondary loops resulting from bottom heating
appeared and died down over time. A smaller clockwise loop was per
sistently present near the bottom centerline. This clockwise loop
could also be seen in the numerical results of Sani (figure 2). At a
still later time of t=1800 seconds (calculated with a
20(0.85)x20(0.80) grid), the clockwise loop near the bottom center-
line persists as can be seen in figure 10o.
Radial temperature and velocity profiles. The transient
behaviour of the liquid velocities near the wall can now be
explained by referring to figure 11. The basic concepts involved in
describing the boundary layer formation are the same as for forced
convective flow situations. The effect of viscosity and temperature
are confined to thin region near the wall. However, as mentioned
earlier, the flow field is coupled with the temperature field for
natural convection flow. It is the difference in temperature that
creates the buoyancy force which moves the fluid upward.
Liquid next to the wall experiences the largest temperature
rise compared to the core temperature. Thus the local Rayleigh
number is largest next to the wall. The buoyancy force resulting
from this large temperature difference contributes to largest velo
cities near the wall (liquid on the solid wall is at rest due to the
no slip boundary condition). The liquid rising due to the buoyancy
effect produces viscous drag on adjacent fluid which also moves up.


r
Figure 19. Transient temperature at various locations
during heating of water in a cylindrical can


153
Recommendations
Using the available hardware, the algorithm described for
controlling conduction-heating process needs to be implemented.
Further work on the natural convection heat transfer model can be
pursued in two major areas. These are to improve the model to make
it less CPU time intensive and to extract more information relevant
to food thermal processing. Specifically some of the ideas that can
be tried include
1. further optimization of grid stretching by trying out no
stretching at the top and/or different stretching
parameters
2. updating of velocities every iteration (with possible need
for relaxation of the values)
3. study of single point and integrated lethality
4. change in the thermal properties corresponding to different
types of foods
5. possibility of incorporating variable thermal properties
(e.g. viscosity)
6. effect of different can sizes and/or aspect ratios.


8
T-121
Fo = L
10
dt
(6)
When the food being heated has a spatially varying temperature
profile, the chosen temperature T, for thermal process calculation,
is the temperature at the slowest heating point in the food
material. Thus, for conduction heating in a cylindrical can, the
chosen temperature T is at the geometrical center of the can.
The designed bacterial inactivation for a process is specified
by which is the time required to achieve the desired
inactivation of bacteria when the same is heated at a constant
temperature (T) of 121C. When temperature T is a function of time
(T(t)) instead, the objective of thermal processing is to select
process time t for the particular heating condition, such that
the equivalent FQ received (as given by eqn. 6) is greater than or
equal to the design value F^ That is,
Fo Fo
(7)
Application to conduction-heating. When applying thermal
processing to conduction-heated food in a cylindrical container,
Ball and Olson (1957) used the solution to the heat conduction
equation in cylindrical coordinates.
T-T0 -(\2+ X2)at
= E S A e m p J0(xr) Cos (x_z) (8a)
T mp uv m -p
T0-TR P


136
Water
T,=30*C
top-121*C
Slde-121*C
b0T-121*C
R=*4.19cm
M~10.G7cm
5U-0.0001
5r=0.001
Contours
-105
-85
-65
-46
-26
-7
13
32
52
72
Streamlines in a cylindrical can calculated
with two different convergence criteria for
boundary vorticity (6=0.0001 on the left
and 6=0.001 on the right)
Figure 30.


TABLE OF CONTENTS
ACKNOWLEDGMENTS iii
LIST OF SYMBOLS vii
LIST OF TABLES xi
LIST OF FIGURES xii
ABSTRACT xvi
INTRODUCTION 1
REVIEW OF LITERATURE 5
Computer Control of Food Thermal Processing 5
Overview of Food Thermal Processing 5
On-line Control of Thermal Processing 10
Conduction Heating of Canned Foods 12
Studies Based on Analytical Solution 12
Numerical Studies 13
Natural Convection Heating in Canned Foods 14
Previous Works on Canned Liquid Foods 15
Other Similiar Works on Natural Convection
Heating in Cylindrical Enclosures 19
PROBLEM FORMULATION 24
Natural Convection Heat Transfer in Canned Foods 24
Governing Equations and Boundary Conditions 24
Boussinesq Approximation 26
Transformation of the PDEs 27
Non-dimensionalization of the Variables 28
Conduction Heat Transfer in Canned Foods 30
Governing Equation 31
Boundary Conditions 31
Computer Control of Conduction Heating Process 32
METHODOLOGY 35
Modeling of Natural Convection Heating in Cans 35
Grid Generation and Grid Stretching 36
Discretization of the Parabolic
Temperature and Vorticity Equation 39
Time derivatives 42
Linearization 44
v


ADDITIONAL REFERENCES
Ames, W. F. 1977. Numerical methods for partial differential equa
tions. Academic Press, New York.
Ander, L. 1981. Microprocessors and their use within food process
industries. Brewer 67(799):148-151.
Aziz, K., and J. D. Heliums. 1967. Numerical solution of three-
dimensional equations of motion for natural convection. Physics
of Fluids 10(2):314324.
Barakat, H. Z., and J. A. Clark. 1966b. On the solution of the dif
fusion equations by numerical methods. Journal of Heat Transfer.
Trans, of the ASME 88:421-427.
Bhowmik, S. R., and K. Hayakawa. 1979. A new method for determining
the apparent thermal diffusivity of thermally conductive food.
J. Food Sci. 44(2):469-473.
Blake, K. R., D. Poulikakos, and A. Bejan. 1984. Natural convection
near 4C in a horizontal water layer heated from below. Physics
of Fluids 27(11):26082616.
Briggs, D. G. 1975. Numerical solution of high Raleigh number two-
dimensional free convection in enclosures with an aspect ratio
of one. Advances in computer methods for partial differential
equations. Edited by R. Vichnevetsky. AICA, New Brunswick, New
Jersey.
Brook, R. C. 1981. Use of microcomputers for process control. Food
Tech. 35(10):89-91.
Bryan, K. 1966. A scheme for numerical integration of the equations
of the motion on an irregular grid free of nonlinear instabil
ity. Monthly Weather Review 94(1):39-40.
Chen, C. S., R. D. Carter, W. M. Miller, and T. A. Wheaton. 1980.
Energy performance of a HTST citrus evaporator under digital
computer control. Trans, of the ASAE 24(6):1678-1682.
Chu, H. H-S., S. W. Churchill, and C. V. S. Patterson. 1976. The
effect of heater size, location, aspect ratio, and boundary
conditions on two-dimensional, laminar, natural convection in
rectangular channels. J. Heat Transfer, Trans, of the ASME
98:194-201.
167


7
0 = 0
o
(3)
where D0 is the 0 value at some reference temperature T0. The Z
value describes the temperature dependency of the lethal rate and is
related to the activation energy. The reference temperature T0 is
generally taken as 121C (250F).
From bacteriological considerations, a thermal process is
specified by noting the time (F0) necessary to reduce the bacterial
population, from the initial (C^) to some final safe value (C) at
temperature 121C (250F). Thus
F
o
(4)
When the heating process is not at the constant temperature of
121C, introducing the temperature dependency of 0Q and equation (2)
for bacterial death,
F
o
D 10
T-121
Z At
D
T-121
= 10 1 At
(5)
At is the time required at temperature T to achieve the designed
bacterial reduction. As the food material is heated, its temperature
T becomes a continous function of time t and the expression for F0
(equation 5) becomes


I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is
fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
Arthur A. Teixeira, Chairman
Associate Professor of
Agricultural Engineering
I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is
fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
L- \.V f'(- .
Khe V. Chau
Associate Professor of
Agricultural Engineering
I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is
fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
Mechanical Engineering
I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is
fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
Tom I-P. Sjrih
Assistant/Professor of
Mechanical Engineering


31
Governing Equation
For strict conduction heating (no movement of the material) of
a homogeneous isotropic cylindrical-shaped body of the food materi
al, the heat transfer is described by the classical heat conduction
equation
30 320 1 30 320
3t Ir7 r 3r II7 K
This is the energy equation (26) approximated for zero velocities
(diffusion term only). This equation is quite well behaved and easy
to solve. The resistance to heat transfer of the metallic walls of
the can was ignored.
Boundary Conditions
Canned foods are generally heated with hot water or steam and
cooled with water. Thus, the heat transfer coefficient at the out
side of the can wall could be comparatively small (for water) or
quite large (for condensing steam). In general, the conditions at
the boundaries can be written as
sidewall
r = R/H 0 < z < 1 |£+hH0 = o
center!ine
r = 0
0 < z < 1
_30
3r
= o
(33a-d)
bottom wall
0 < r < R/H z = 0
_30
3Z
hH
0 = 0
top wall
Me-o
0 < r < R/H
z = 1


171
Steele, R. J., and P. W. Board. 1979. Amendments to Ball's formula
method for calculating the lethal value of thermal processes. J.
Food Sci. 44(1):292-293.
Steensen, F. 1980. Why use a microprocessor for process regulation ?
Food Engg. International 5(12):35-38.
Teixeira, A. A. 1978. Conduction-heating considerations in thermal
processing of canned foods. ASME paper no. 78-WA/HT-55.
Presented at the winter annual meeting of the ASME at San
Francisco, California.
Uno, J., and K. Hayakawa. 1980. A method for estimating thermal
diffusivity of heat conduction food in a cylindrical can. J.
Food Sci. 45(3):692-695.
Uno, J., and K. Hayakawa. 1980. Correction factor of come-up heating
based on critical point in a cylindrical can of heat conduction
food. J. Food Sci. 45(4):353-859.
Wilkes, J. 0. 1963. The finite difference computation of natural
convection in an enclosed rectangular cavity. Ph.D. Thesis.
University of Michigan, Ann Arbor.
Wolfshtein, M., R. S. Hirsh, and B. H. Pitts. 1975. Second-order
non-iterative ADI solution of non-linear partial differential
equa-tions. Advances in computer methods for partial differen
tial equations. Edited by R. Vichnevetsky. AICA, New Brunswick,
New Jersey.
Yang, S., and T. I-P. Shih. 1985. An algebraic grid generation
technique for time-varying two-dimensional spacial domains.
Computational Fluid Dynamics Laboratory Report no. 8502.
Department of Mechanical Engineering, University of Florida,
Gainesvil1e.


41
natural convection problems), as well as realistic constraints on
available resources (Roache, 1982). For this study, the transient
temperature and velocity solutions are needed and the range of
Grashof number is 108 1010 Since the temperature and vorticity
equations are both parabolic and are very similar, they will be
discussed simultaneously in this section.
The performance of the various finite difference methods was
noted in terms of stability, truncation error, and conservation
property. Stability was defined as the controlled growth of the
round-off errors (O'Brien et al., 1950). An uncontrolled growth of
the round-off errors leads to instability in the finite difference
methods. The stability requirements often put restrictions on the
time step as a function of grid spacings. Empirical stability re
striction on grid size of the type Ax<8/|u^ j| and Ay< 8/jv-¡j|
were also reported by Torrance (1968). Truncation error is the error
resulting from the discretization of the PDE. Conservation of any
physical entity (e.g. energy, vorticity) within the finite differ
ence grid system exists if the difference equations for the entity
are summed over all interior grid points and no spurious sources or
sinks of these quantities are found. Torrance (1968) found that
formulations by Barakat and Clark (1966) and by Wilkes and Churchill
(1966) did not conserve energy or vorticity and he did not recommend
the use of their equations. Another important property of finite
difference formulations is the transportive property. A difference
formulation is transportive if the effect of a perturbation in a
transport property is convected only in the direction of velocity.


PROBLEM FORMULATION
The governing equations and boundary conditions for the natural
convection and conduction heating process together with the assump
tions are first described here. The on-line control problem for the
conduction heating process is then formulated.
Natural Convection Heat Transfer in Canned Foods
The equations governing natural convection heat transfer in
liquid foods together with the boundary conditions are first
presented. The Boussinesq approximation to the set of equations is
then discussed. To further facilitate the solution, new variables
are defined and the governing equations together with the boundary
conditions are transformed into a new set of equations.
Governing Equations and Boundary Conditions
In natural convection heat transfer, the driving force for the
liquid motion is the buoyancy caused by the density variations due
to change in temperature. The partial differential equations (PDEs)
governing such naturally convective motion of fluid in a cylindrical
space are the Navier-Stokes equations in axisymmetric cylindrical
coordinates (Bird et al., 1976) and are described below.
24


77
REPEAT OVER n
*n n
0) = 0)
m
l.(
M11 m
oj 1 = 1.0
ITERATE OVER m
n *n
o
in n+1
0) = 0)r
-r1" *rira-
G+1"- Ui+1". vT1. e-m)
m
0)
)
1 /El
n/2 =
T
w* na+1)
n-
%
II
f r n n+lm
vn+1
a
, ,m
n+1
IE
>r
0)
n+lm_
'a
f ( n+2" ,,n+l
US1
m
n^/2m_
2= 0
n+1
m
(0=0).
ITERATE
m
,n+l"_ f r n+1 'i
n fsCa J
m
UNTIL
,n+l
-
n+1
m
fl,old
n+lm
^Q,oId
- -i m+1 ^.m
n+1 c ,n+l >
r f7l^ J
< e
n+1
m+1
0).
= yo)T
~n+l
m+1
+ (I-y)oj
n+1
UNTIL
"n+1
m+1
- 0)
0)
< 6
. << m m
<< =feO
UNTIL DESIRED TIME REACHED
T: solid boundary
ft: interior points
: centerline
in ~ .
o),oj,o): temporary
storage
Figure 5. Algorithm for the iterative solution
of the set of equations


150
In this way, products which experience process deviations can
be released for shipment on schedule and supported with all
necessary documentation in compliance with FDA Canned Food
Regulation. Use of these controlled systems in industry should lead
to improved production efficiency without compromising product
quality and safety in canned food operations.


69
,k+l n
*1,j
¡*+*
+ (1fi } *1,j
(71)
Using equation 71, stream function values i|>. at all interior
points are updated. The entire sweep is repeated until the
convergence criterion
Max
i ,j
.k+1 k
*i,j ^ij
< e
(72)
is satisfied. The value of e chosen for this study was 10^. In a
*
transient calculation like this study, the optimum ft may change

from one time instant to another. However, the optimum ft found
through calculations during the very first time step was used
throughout the entire transient calculations.
Computational Boundary Conditions for the Finite Difference
Equations
A PDE such as equation 27 for to can describe a wide variety
flow situation. The flows (solutions) are distinguished only by the
boundary and initial conditions and flow parameters like the Grashof
number. Thus, specification of computational boundary conditions,
besides affecting numerical stability, greatly affects the accuracy
of the solution (Roache, 1982). Using the boundary conditions
already noted for the partial differential equations in the previous
chapter, the boundary conditions for the finite difference equations
are now derived.
The boundary condition for temperature is rather simple. When
the can wall temperature is specified, the computational wall


49
In the r direction,
a(rv0) a? a(rv0) 1___ a(rv0)
ar ar a? ar/as a?
i vRrj+ V?0i ,j ~ vLrj-V?9i,j-l
rj+l~ rJ-l
AS
2AS
Vd>0
V|_>0
j+i
- vLrj-V2ei.j-i]
Simliarly
6(rv0)
6r
VR < 0
VR < 0
vR > 0
Therefore, combining the equations
vL > 0
< 0
VL < 0
3(rv0)
ar =
(VlVl)ri4/,91,.i+l" r...-r. ,
J+l J-l
(44)


71
Improper specification of boundary vorticity is found to be
destabilizing in general and the same was found in the present
problem. Vorticity is produced at no-slip (wall) boundaries. It is
then diffused and convected into interior points. No explicit
expression for wall vorticity is available. It is related to the
interior stream function values by the vorticity stream-function
equation (equation 28). However, this would make the vorticity
equation implicit with the stream function equation and the number
of simultaneous equations increase drastically. To avoid this,
Barakat and Clark (1966) chose explicit methods over implicit and
used vorticity boundary values calculated at the previous time step.
Wilkes and Churchill (1966) used the ADI method but similarly used
boundary vorticity values from the previous time step. Thus, in
their formulations, wall vorticity values lagged by one time step at
all times.
Roache (1982) discussed the effects of such lagging of boundary
vorticity values. It forces the use of a smaller time step since
larger time steps could make the solution inaccurate and unstable.
Since larger time steps are a major motivation for ADI, updating of
the boundary values through iteration is preferred. From the work of
Torrance (1968), Roache also noted that the convergence of the
boundary vorticity values actually imposed a time-step restriction
of the form At < a / Az2 where "a" was some number dependent on the
problem and convergence requirements. Thus, due to the implicitness
of the boundary vorticity values, ADI actually has a time-step
restriction like the simplest of the explicit methods. The


Water
T,=30*C
lop-lEfC
slde121*C
bot-121*C
R-4.1.9cm
Hs*10.67om
R-57,0.76
Z-57,0.85
a-o.oooi
-300
-260
-220
-180
-140
-100
-60
-eo
20
60
Figure lOe. Isotherms and streamlines in a
cylindrical can after 5 seconds of
heating
Figure lOf. Isotherms and streamlines in a
cylindrical can after 6 seconds of
heating


LIST OF TABLES
TABLE PAGE
1. Sequence of line types for identification of stream
function contours in figures lOa-o 96
2. Heating and cooling rates for various process
conditions 140
3. Adjusted heating times and resulting lethality (F0)
in response to process deviations using proposed on
line control logic and method of Giannoni-Succar and
Hayakawa 148
xi


51
The finite difference formulations for the time and space deriva
tives are then put together in the ADI formulation. For each of the
temperature and vorticity equations, the equations are written for
the "half" time step and for the "full" time step of the ADI method.
Algebraic simplifications are done to convert the equations at each
of the time steps to a tridiagonal systems of equations. A glimpse
of the tridiagonal matrices for the "half" and "full" step is then
included.
The tridiagonal systems of equations are solved using the back
substitution procedure (Smith, 1978). The time step At and the
mesh sizes Az, Ar were chosen in such a way that the coefficients
(a7-, b.¡, c.¡ in equation 48a) are all positive and the matrices
diagonally dominant (b1- > ai + c^). Richtmyer and Morton (1967)
showed that these two conditions are sufficient to keep the round
off errors down, in the back substitution procedure. Introduction of
the second upwind differencing method satisfies the criterion that
coefficients a-¡, b-j, and C] are all positive (see Kublbeck et al.,
1980). The second requirement of diagonal dominance was not investi
gated any further in this study. However, at all times whenever the
numerical values of the coefficients an-, b.¡, Cj were visually
checked, the diagonal dominance was found to be satisfied.
ADI formulation for the temperature equation is now developed.
For convenience, the differential equation for temperature is
repeated. The finite differences corresponding to each of the
derivatives in the temperature equation are then assembled in the
same order.


83
Experimental Studies
Experiments were performed to record the transient temperature
values inside cylindrical cans during conduction heating. The effect
of different surface heat transfer coefficients, that were obtained
by using different media (steam and water), on the transient
temperature response was also noted.
The experimental work consisted of series of heat penetration
tests on No. 303 cans filled with a 10% bentonite suspension as a
conduction-heating food model system (Niekamp et al., 1984). Eight
cans were fitted with an Ecklund-type copper-constantan thermocouple
mounted mid-way up the can wall, so as to have the junctions located
at the geometric can center. In order to achieve the two different
levels of head-space vacuum, four of the cans were filled at 90C
and the other four cans were filled at 67C. The cans were
identified according to fill temperature and then processed with an
extended steam-cook to stabilize the bentonite suspension. An
experimental retort was used that had been designed and built at the
University of Florida with the capability of using pressurized
water, pure steam, or steam/air mixtures under controlled rates of
recirculation. The retort was fitted with a special rack designed to
hold the eight cans suspended upright in the mainstream of the
retort chamber, as shown schematically in Figure 6.
The rack consisted of a stainless steel plate with eight
cicular holes through which the cans could be inserted. Special
fixtures were welded at the edge of each hole so that the cans could


165
Schumann, U., and R. A. Sweet. 1976. A direct method for the
solution of Poisson's equation with Neumann boundary conditions
on a staggered grid of arbitrary size. J. of Computational
Physics 20:171-182.
Shih, T. I-P. In Press. Finite-difference methods in computational
fluid dynamics. Prentice Hall, Englewood Cliffs, New Jersey.
Shyu, R. J. 1985. Numerical study of unsteady natural convection in
cylindrical enclosures. Ph.D. Dissertation. University of
Florida.
Shyu, R. J., and C. K. Hsieh. 1985. Unsteady natural convection in
enclosures with stratified medium. ASME paper no. 85-HT-33.
Presented at the National Heat Transfer Conference at Denver,
Colorado. August 4-7, 1985.
Smith, G. D. 1978. Numerical solution of partial differential
equations. Oxford University Press, Oxford, England.
Stevens, P. M. 1972. Lethality calculations, including effects of
product-movement, for convection heating and broken-heating
foods in still-cook retorts. Ph.D. Thesis. University of
Massachusetts, Amherst.
Stumbo, C. R. 1965. Thermobacteriology in food processing. Academic
Press, New York and London.
Teixeira, A. A. 1971. Thermal process optimization through computer
simulation of variable boundary control and container geometry.
Ph.D. Thesis. University of Massachusetts, Amherst.
Teixeira, A. A., A. K. Datta, J. P. Adams, and W. R. Peterson. 1984.
Surface heat transfer considerations during immersion water
cooling of retorted foods. Proceedings of ICEF-3, Dublin,
Ireland. I:(3), pp.23 Elsevier Applied Science Publishers, Ltd.
Barking, Essex, England.
Teixeira, A. A., J. R. Dixon, J. W. Zahradnik, and G. E.
Zinsmeister. 1969. Computer optimization of nutrient retention
in the thermal processing of conduction-heated foods. Food Tech.
23(6):137-142
Teixeira, A. A., K. Dolan, A. K. Datta, and J. P. Adams. 1985.
Retort depressurization effect on cooling rates in conduction
heating canned foods. Trans of the ASAE 28(2):645-648, 656
Teixeira, A. A., and J. E. Manson. 1982. Computer control of batch
retort operations with on-line correction of process deviations.
Food Tech. 36(4):85-90.


81
differenced using an alternating direction explicit (ADE)
differencing method, which is explicit but still unconditionally
stable. The transient temperature values predicted by performance of
this ADE finite difference method was compared with analytical
solutions and experimental measurements.
Discretization and Solution of the Equation
Keeping in mind that the equation was needed to be solved in
real-time for on-line control purposes and the ADE method was
already tested (Teixeira et al., 1975; Teixeira, 1971), the same was
selected. The ADE formulation was first reported by Saul'yev (1957).
This was later translated and described by Aliada and Quon (1966),
and was first introduced for applications to thermal process
simulation by Manson et al. (1974). The method as applied to the
diffusion equation is unconditionally stable (Saul'yev, 1964) for
temperature-specified boundary condition. It has a formal order of
accuracy of 0(At2, Ar2, Az2 ). Being explicit, it does not require
inversion of matrices as would be required by an implicit method. In
a fluid flow problem, the ADE formulation is generally not preferred
since the presence of the convection term (in any difference form)
makes the complete formulation unstable or returns it to simple
explicit stability limits ( <1/2 > u ^ < 1 for a one
dimensional problem; Roache, 1982). Another shortcoming of this ADE
method is the fact that even though it is an explicit method, it is
actually implicit in boundary conditions. To handle this implicit
ness, either explicit expressions for the temperature values need to


92
Corrected process times for each of these three processing
situations are first noted for the on-line computer control
method. These are the times t^ when
rolh
cool^
'ho 'simulated
> F
d
o
is satisfied. The corrected processes as predicted by the CF method
were found by multiplying the normal process time with the
correction factors predicted by that method.
The accomplished FQ values for the processes corrected by the
on-line control system and processes corrected by the CF method were
calculated using equation 35a (the general method). In either case,
the can center temperatures required for use with equation 35a were
predicted using the ADE finite difference analog of equation 32.


94
Throughout this section, reference will be frequently made to
the experimental study of Hiddink (1975) mentioned in the REVIEW OF
LITERATURE section. Since his study involved more than one wall
temperature and other conditions, the particular conditions appli
cable to any situation are noted alongside. Reference will also be
made to the numerical finite element study of Sani on temperature
and velocity profiles during pasteurization of canned beer (also
mentioned in the REVIEW OF LITERATURE section).
Transient Flow Patterns and Temperature Profiles
The liquid adjacent to the sidewall, the top and the bottom
wall recieves heat from the hot walls. As the liquid is heated, it
expands and thus gets lighter. Liquid further from the sidewall is
still at a much lower temperature. The buoyancy force created by the
liquids at different temperatures forces the hot liquid, next to the
wall, upwards. A sharp drop in temperature occurs from the wall to
the core, creating a boundary layer. The hot liquid going up is
interrupted by the top wall and then travels radially towards the
core. The core liquid, being heavier, moves downwards and towards
the wall. Thus a recirculating flow is created with a boundary layer
at the sidewall, core flow around the centerline and a mixing region
at the top. The same qualitative observations were reported in the
experiments of Evans et al. (1968), Brandon et al. (1982), and
Hiddink (1975).
The arrow plots of velocity vectors create a picture that is
very busy and it is often hard to differentiate between the flow


56
Upwind differencing of the convective terms in the vorticity equation
3(uo))
3z
= similiar to
3(u0)
3z
. ^ur~Iur 1 ^Mj+i,j+ tylRl-yluLl)Mi,j (yKP-i-i.j
z.z. .
i+l i-l
(50)
3( Vw)
3r
(vR- IvRpMjj-n + (VR+ |vr| V KlK'.j ~ 1 vl1 ^i, j-l
(51)
r .l1- r. .
J+l J-l
Upwind differencing of the diffusion terms in the vorticity equation
32cj
3Z2
zi+r zt-i
(1).,, .-O). 0). tJ. .
' i+l.j i,J i,.i i-l.J-
zi+r zi
zi- zt-i
(52)
3 c1 3(ru)
3r 3r '
35 3 /-l 3( raj)
IFF IT^r" 3r '
3( roi)
r.r. t '
J+l J-l
lk.
3r
3(ru)
3r
iJ+V? rj-V? r i J- V?
2AC
2
rj+r rj-i
A?
r.,,0). r.a). r.o). r. ,u. ,
f J+l 1,J+1 J 1,J J 1 ,J J-l 1J-ll
1 r i Tp pi p] ip p i J
rj+V2(rj+r V
j-V^rj" rj-i^
(53)
Central differencing of the buoyancy term
30
3r
9i,j+r 9iJ-l
rj+r rj-i
(54)


APPENDIX B
ALTERNATIVE FINITE DIFFERENCING OF BOUNDARY VORTICITY
Boundary conditions for vorticity at bottom wall
Let Az = zj- z0= z1
Using Taylor's series expansion at (OJ)
3^1
+ l a2l>
*1J ^0,j+ 9z|0jjAz T
OJ
AZ2 + 0(AZ3)
Now, using no slip conditions
u0,j 0 => fIfL
U 9 J
vj j = o -> 4j?| =o
0,J r or|0J
and
i^O j= 0 by choice
=> 1 3 2ip
K1J 2 9zz
From equation
-CO
= > (0,
oj
Az2 + 0(az3)
1 H2
OJ
r 9zz
1J
" 2 (-uf
OJ
2^1
- I-iJ,
rAz-
oj 3r'r3pl,o.J r9z'lO,j
=> h.j = \ (-o,j-r)Az2 + (Az3)
-Pi i
, similarly at other boundaries
rz'zV2
155


22
had numerically calculated and plotted the isotherms and streamlines
inside the enclosure. In their experimental work, they recorded
temperature values at ten different points inside the cylindrical
enclosure. The observed temperature values deviated from the
numerical results by 10%. This discrepancy was attributed to the
small non-symmetry in the imposed wall temperature which might have
produced three dimensional effects and, also, to heat losses from
the bottom and top of the container.
Torrance (1968) compared five separate finite difference
formulations for the problem of natural convection in cylindrical
enclosures. The five formulations were respectively by Barakat and
Clark (1966), Wilkes and Churchill (1966), Fromm (1964) and two of
his own. The physical problem chosen was the study of natural
convection flow induced in a vertical cylindrical enclosure by a
small hot spot centrally located on the bottom. One of his two
formulations was explicit and the other was implicit. The explicit
formulation was shown to be numerically superior in many ways to the
other four and it was also reported to have good, agreement with
experimental observations (Torrance and Rockett, 1969; Torrance et
al., 1969) for lower values of Grashof numbers. This was also the
formulation used by Stevens (1972) referred to earlier.
Kee et al. (1976) performed a numerical study of temperature
distribution and streamlines in a closed vertical cylinder
containing uniformly distributed heat generating tritium gas. The
walls were kept isothermal. An upwind method was used for the
convective terms. Numerical temperature values were compared with


72
definition of vorticity to was used at the boundaries to obtain
expressions for boundary vorticities. At the bottom boundary,
to,
- voj"
.J Ul/2- z0)
t t
^rj+1/2~rj ^2
-
1J
r.z- z\.
J 1 72
(76a)
since
'0,j
v0,j=0
rj(zr zoT
r.z.
J 1
Uq j= 0 and Ug ._j = 0 from eqn. 60a
At the top boundary,
t t t
Vm,j~ vm-l,j um,j" um,j-l
r l2m- W lV/2- rJ-W
_ m-1 ,j
r (z -z )(z -z ij
j m m-1 m m-/2
since
(76b)
Vm 1= 0
m 9 j
(.¡-tf , _)
m 9j m-19j
vm-l,j r.(z -z 7)
j m m-1
V n
m-1 ,j
r.(z -z7)
j m m-17
um,j a"d 4-1,j= 0
from eqn. 60a


98
T(z,t) Tx
jj
'0
1
erf (
/4oit
(83)
where erf is the error function.
Isotherms and streamlines during the first ten seconds of
heating are given in figures lOa-j. These figures (lOa-j) represent
results of computations using a 58(0.85)x58(0.76)* grid. Instability
resulted due to the presence of heated bottom layers with colder
liquid above them. This instability in the bottom layer gave rise to
"bursts" of convective cells in a very regular manner. Such
phenomena are referred to as Benard convection. However, because of
the very large initial temperature difference, this convection
occurred very rapidly in the present study (figures lOe-j) as
compared to its much slower occurrence reported by Shyu and Hsieh
(in press). The formation of these convective cells were responsible
for the convective heat transfer from the bottom. In the study of
Hiddink (1975), these were precisely the convective flow patterns
that could not be picked up by his numerical model. Along the side-
wall, hot liquid rises and gets deposited at the top. This explains
the shape of isotherms at the top core region.
Transient isotherms and streamlines at later time instants of
30, 120, 300, and 600 seconds of heating are presented in figure
lOk-n. These results (figures lOk-n) were computed using
39(0.9)x39(0.8) grid. Two main flow loops could be seen. The larger
*The values 0.85 and 0.76 in parenthesis denote the stretching
parameter sigma (see equations 38 and 39) in z and r direction
respectively.


ACKNOWLEDGEMENTS
The author is indebted to Dr. Arthur A. Teixeira, his major
professor, for his encouragement, guidance and patience throughout
the course of the work and especially for his assistance in
preparing the manuscript.
The author is also highly grateful to Dr. Tom I-P. Shih, Dr.
Chung K. Hsieh and Dr. Ruey J. Shyu for their time and extremely
valuable suggestions in the study of the natural convection heat
transfer. Further thanks to Dr. Khe V. Chau and Dr. Spyros Svoronos
for their interest in the study and helpful suggestions.
The importance of a good computing facility (number crunching
to be particular) in developing the natural convection model cannot
be over-emphasized. The availability of the Cyber 170 computer
through Dr. John Gerber and the IFAS adminstration was vital for
that purpose. The author also greatly appreciates help from Dr.
Ellen Chen, who introduced him to the Cyber facility, and Victor,
Jimmy and Margaret for endless support on problems with the Cyber
system.
The mountain of data available from the transient natural
convection heating model would have been of little use without the
plotting capability provided by the excellent graphics software
(PLOTPAK) made available from Dr. William E. Dunn from the


149
The step function and the gradual drop in temperature are just
two examples of process deviations where the on-line computer
control system is able to maintain the desired sterilization. This
control system predicts the can center temperature using a finite
difference analog of equation 32. It then uses this temperature
value with a numerical (trapezoidal) method of integrating equation
6 (the general method). Both these steps place no restriction what
soever on the retort temperature variation. This enables the on-line
control system to maintain the desired sterilization for arbitrary
variations in retort temperature. This is not true for the correc
tion factor approach of Giannoni-Succar and Hayakawa (1982) which
requires the retort temperature drop to be a step function from an
otherwise constant value. Also, it is not apparent how their method
would correct a process with multiple drops, even if the individual
drops were step functional. The numerical method of calculating the
accumulated process F0 is a more direct approach than Ball's formula
method of calculating the process F0 and subsequent correction
factors.
Other benefit areas include real-time operational information
and complete documentation. Figures 33, 34, and 35 show a complete
history of the process including retort temperature, calculated can
center temperature and calculated F0. In addition to providing all
the required documentation for regulatory and quality control com
pliance, the plots will provide visual descriptions of the complete
process history. The graphs can clearly show if the process went
through any deviation and how the required F0 was still maintained.


r%
3
Figure 2. isotherms and velocity profiles in a can at
various heating times (after Sani, 1985) ^
o
yrrr


50
Diffusion terms. Central differencing was used for the
diffusion terms in both the temperature and vorticity equation. The
diffusion terms in the temperature equation are finite differenced
be!ow.
320 3 .30. 3n 3 .30x
TT -ST'S?* = Iz j
oZ
30, _30i
_ i h+VpJ 9z* i-Vftj
3z/3n An
1
0.
1+1,j
zi+l
z
j+r z
2 An
i-1
0. 0. .
_UJ \-A,a
zi zi-l
An
zi+l zi-l
0. 0. .
- i+l,J i,J
zi+l
z.
1
9i ,j ~ 9i-l,j
zi zi-l
(45)
Ii_ (p!,
r 3r 3r;
r 3r 35 *^3F
dQ _30
- i 8r'i,j-1/?
~ r 3r A5
85
0. 0. 0. 0. 1
1 ,J+1 1 ,J 1 ,J 1,J-1
W
TJ ri+l rJ-l
rj+l ~ rj
A5
W
9i,j+i ~ Qi,j p
J^/2 rJ+1 r-
0. 0. .
T,J
jJ/2 rj rj-l
(46)


CONCLUSIONS AND RECOMMENDATIONS
Conclusions
The following conclusions were drawn from this study.
Natural Convection Heating in Cans
1. A numerical model was developed that is capable of predicting
transient flow patterns and temperature profiles during
natural convection heating of liquids in a cylindrical metal
can. Plots of computed transient isotherms, streamlines, and
velocity vectors were provided.
2. Calculated radial velocity profiles and axial temperature
profiles compared well with available experimental data.
3. Qualitatively, the flow inside the can consisted of liquid
rising at the boundary wall, radial flow and mixing near the
top, and uniform core flow downwards at the center.
4. Thermal as well as velocity boundary layers formed at the side
wall.
5. During very small initial times, the heating from the bottom
of the can was restricted to conduction heating.
6. As time passed, magnitudes of the velocities reduced due to
decreasing temperature difference between the fluid and the
boundary walls.
151


Water
Figure 10a. Isotherms and streamlines in a
cylindrical can after 1 second of
heating
Water
T,=30*C
top-121*C
slda=121*C
bot-121*C
R-4.19cm
h=10.67cm
R-57,0.76
Z=57,0.85
fi-0.0001
-300
-260
-220
1B0
-140
-100
-60
-20
20
60
Figure 10b. Isotherms and streamlines in a
cylindrical can after 2 seconds of
heating


31. Transient center temperature during conduction
heating of Bentonite in a cylindrical can 142
32. Transient center temperature during conduction
cooling of Bentonite in a cylindrical can with and
without drop in retort pressure 143
33. Computer plot of reference process (no deviations)
showing 66.8 minutes of heating time and an
accomplished F0 of 6.24 144
34. Computer plot of a process that experienced a step
functional drop of 8.3C (15F) in temperature bet
ween 50th and 60th minute and still maintained an F_
of 6.02 146
35. Computer plot of a process that experienced a linear
drop in temperature between 40th and 60th minute
(from 0C to 11.1C (20F)) and still maintained an
F0 of 6.18 147
xv


44
are required because of this situation. The iterations can also
serve the purpose of preserving second order accuracy in time of the
ADI method. The second order accuracy may otherwise deteriorate due
to the presence of non-linear convective terms (discussed below).
Since the iterations are needed anyway, an implicit set of equations
can be solved, instead, utilising the iterations. As noted earlier,
Stevens (1972) and Hiddink (1975) also used the explicit
formulations of Barakat and Clark (1966) to study the present
problem and could not resolve the convective eddies present due to
the bottom heating. Since the ADI method may allow finer grid
spacings with less stricter restrictions on At, it may be able to
pick up the convective eddies present, without taking prohibitive
amounts of computer time. Thus the ADI method was chosen to solve
the temperature and vorticity equation in this study. The ADI
formulation is written below for the temperature equation 26. At the
"half" step
n+1/2
0 -
3(u9)
At/2
3z
n+ V2
3(rv0)
3r
320
n+V2
+ 1
r
30,
3r^ r3r)
and at the "full" step
3n+1_ 0n+1/2
Ax/2
3(U0)
3Z
n+1/2
3(rv0)
3r
n+1
320
Iz2-
n+V2
r 3rv 3r;
n+1
Linearization. The velocities in nonlinear convective terms
e.g.
3(u0)
3z
n+ V2
should ideally be evaluated at the same time level
(n+ 1/2 in this case) as the 0 (or w). However, since u and v


I certify that I have read this study and that in my opinion it
conforms to acceptable standards of scholarly presentation and is
fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
S
Spyros Svoronos
Assistant Professor of
Chemical Engineering
This dissertation was submitted to the Graduate Faculty of the
College of Engineering and to the Graduate School and was accepted
as partial fulfillment of the requirement for the degree of Doctor
of Philosophy.
Dean, Graduate School


Water
T,-30*C
!op-121*C
slde = l-21*C
bo-121*C
R-4.L9cm
H=*10.67om
R-38,0.80
Z-38.0.90
a-o.oooi
-105
-83
-65
-45
-23
30
50
Figure 10k. Isotherms and streamlines in a
cylindrical can after 30 seconds
of heating
After 30 sec
Water
T,= 30*C
lop-121*C
slda=l£jl C
bot-121*C
R-4.19cm
h=l O.G7om
R-38.0.80
Z=38,0.90
8-0.0001
-40
-25
-20
-15
10
15
20
Figure 101. Isotherms and streamlines in a
cylindrical can after 120 seconds
of heating


109
Further away from the wall, near the core, the effect of positive
velocities upward is not felt. A thermal as well as a hydrodynamic
boundary layer is thus created. Due to recirculation of the water in
the enclosure, negative velocities appear in the core region. Since
the liquid is going up only through a small cross sectional area
(the thickness of the boundary layer) while its downward movement is
over a much larger area at the core, the downward velocity is much
smaller. This gave rise to the velocity profile as in figure 11.
Such velocity profiles were numerically computed by Shyu (1985),
Sani (personal communication, 1985) and experimentally observed by
Hiddink (1975) and Engelman and Sani (1983).
The thermal boundary layer formed is shown in figure 12. It can
be seen that the temperature did not drop continuously with distance
from the sidewall as would be the situation for natural convection
from a heated vertical wall in an infinite medium (see figure 13).
The temperature first dropped and rose somewhat to become the core
temperature. The rise in temperature from the minimum value in the
profile (closer to wall) to a higher value in the core is caused by
the downward flow of hot fluid in the core from the top.
The magnitude of the velocities computed at the boundary layer
(figure 14) compares excellent with the velocities experimentally
observed by Hiddink (1975) as shown in figure 15. For these two
figures the input data used were radius=8.9cm, height of liquid
=18.00cm, top liquid surface free (thermally insulated) and sidewall
and bottom wall temperatures of 107C. The maximum positive veloci
ties in the boundary layer computed in this study at three different


87
thus satisfying the desired sterilization value for arbitrary
variations in retort temperature T^(t). As noted earlier, the
computer needs to simulate cooling and perform other calculations in
real-time.
Figure 7 shows the timings of various calculations that need to
go on in real-time. For the simulation to be feasible in real-time,
the requirement that
Atj + At 2 < At (81)
has to be satisfied. There is some choice over selection of all
three time intervals At^, At2> At However, At cannot be increased
too much. This could make finite difference calculations of T(t)
from T^(t) inaccurate and unstable. Explicit finite difference
methods generally take less computation time as compared to implicit
methods and may be useful in reducing At^, while At2 can be reduced
by selecting finite difference methods that allow larger time steps
without going unstable. This makes it possible to simulate the
complete cooling cycle in a short time. Implicit finite difference
methods generally allow large time steps as compared to explicit
methods. Thus we see that the requirements for reducing At^ and
At2 are conflicting. It was necessary to look for finite differ
ence methods that could take large time increments while using as
little computation time as possible to calculate each step. The
alternating direction explicit (ADE) finite difference method used
by Teixeira and Manson (1982) does offer both of these advantages.


TEMPERATURE (C)
25
TIME (minutes)
Figure 35. Computer plot of a process that experienced a linear drop in
temperature between 40th and 60th minute (from 0C to 11.1C
(20F)) and still maintained an F 0f 6.18
ACCOMPLISHED Fn (minutes)


39
= Lt
a+l
1
2 tan (ttct/2)
by L1Hospital's rule
Smil arly
= 0
Lt
ff+l
dz
cTn
z=l
Lt
a->-l
2
ir a
tanirg/2
1 + tanzira/2
= 0
and
Lt
a+l
dr
r=R/H
Lt
cr+l
2
TT O
tanu/2 g
1 + tanir/2 a
0
Thus the transformations provide closer spacing of grids near z = 0,
1 and r = R/H. Figure 3 shows the grid system with close spacing of
grids near the top, bottom and sidewall. As can be seen, there is no
excessive grid skewness.
Ka'lney de Rivas (1971) also showed that when conditions 2 and
3 are satisfied for any smooth function, "extra truncation errors"
due to the non-uniform grid system are improved from first order to
second order in An and A£ Vinokur (1983) also discussed similar
types of stretching functions.
Discretization of the Parabolic Temperature and Vorticity Equations
Of the several finite difference formulations that are
available for any set of PDEs, careful choices needed to be made
depending on many factors including boundary conditions, geometry of
the problem, type of solution (steady-state or transient) desired,
range of parameters involved (particularly the Grashof number in


64
"i,j = A*1-l .j+ B*1,j + C*1+l,j + ''"i.j-l + E*T,j+l
(62)
where
A =
rjlZlJ/2'ZiVlZlZi-l)
B =-
rj(zf4ziVlzHrz1 + Vzi-i!' rj4/2-rjJ/2 rjJ/2lrj+rrj>+rji/2(rrrj-i>1
C =
WVi-V^i+r^
D =
l-rj+V2rj-V2Jrj-1/2lrj'rj-i)
E =
tVV2-rJ-V2Jr>V2tl>r'V
The grid point velocities are then achieved through linear
interpolation of the transport velocities u^f and v!f .
jJ J
u.. rj rj-fe. ut.+ rjt.h.: ,rj_ ut.
J rj+V2 rj-V2 1,0 rj+V2 rj-v2 J-1
(63a)
v. = J...~ \r.lA. vt + +V2..1.,. vt .
J Zi+V2" Zi-V21,J Zi+V2 z1-V *J
(63b)


53
After algebraic manipulation, temperature equations at the "half"
time step
ai0i-i,j+bi0i,jci0i+l,j = rhsi
(48a)
where
l/2(uL+|uL|) ^ x
a zi+rzi-i zi+rz,--i^zrzi-i)
bi
ci
. -i + & (uMuSi-uMum + i r L_ + i
Ax z_. .,-z. ^ R I Rl L I LW z. .,-z_. ^ z. ,,-z. z.-z_. ;
i+1 i-1
i n
R'IUR
i+1 i-1 i+1 i i i-1
- Vu"-|u"|)
zi+rzi-i
tzi+rzi-iJtzi+rzf)
rhs, de^ J.J 3 fej j+1
' VVi-'j-i-5 rjlrj+r'ViJirj-rj-i)
I_ i.(vRtlVRl|r.1^-(,L-|,Ll)rJJfr 1 ,ri^-
AT 2 rjirj+rrj-i) rjlrj+rrj-i^rj+i"rj rrrj-i
n i n i
f.. fryKil^ + /Jjft
pjrj+i-rj-iJ rjtrj+rrj-iJlrj+rrj)
and
UR =
i ,j
<
70
II
<
s*
C_i.
f
f
i-lj
L" i,j-l


120
more viscous, there may be significant differences between the
slowest heating temperature and the temperature at other locations.
The transient temperatures at different points in the container
experience rapid fluctuations due to changing flow patterns. This
is illustrated in figure 20 by enlarging the time scale.
Contrast with Conduction
For a thin liquid like water, natural convection speeds up the
heat transfer considerably as compared to a solid (having same
thermal diffusivity value) which can heat only by conduction. Figure
21 shows temperature profile obtained in water at the slowest
heating points using the natural convection heating model developed
in this study, and the temperature profile obtained using the ana
lytical heat conduction equation (equation 8a) at the geometrical
center in a solid product having the same thermal diffusivity as
that of water, heated under identical conditions. As can be seen,
the slowest point in the natural convection-heated can reached al
most the heating temperature (121C) in about 10 minutes. However,
the slowest point in the conduction heated food with the same
thermal properties reached only 101C after 1 hour! For liquids more
viscous, the difference in heat transfer may not be so pronounced.
A contrast of the uniformity of temperature distribution in the
conduction and natural convection heating situations is presented in
figure 22. The essential difference is that for a uniform specified
boundary temperature, conduction heating creates completely uniform
isotherms as expected from the conduction equation, whereas, for


166
Teixeira, A. A., C. R. Stumbo, and J. W. Zahradnik. 1975a.
Experimental evaluation of mathematical and computer models for
thermal process evaluation. J. Food Sci. 40:653.
Teixeira, A. A., G. E. Zinsmeister, and J. W. Zahradnik. 1975b.
Computer simulation of variable retort control and container
geometry as a possible means of improving thiamine retention of
thermally processed foods. J. Food Sci. 40:656.
Torrance, K. E. 1968. Comparison of finite-difference computations
of natural convection. Journal of the National Bureau of
Standards 72B(4)-.281-301.
Torrance, K. E., L. Orloff, and J. A. Rockett. 1969. Experiments on
natural convection in enclosures with localized heating from
below. J. Fluid Mech. 36(1):21-31.
Torrance, K. E., and J. A. Rockett. 1969. Numerical study of natural
convection in an enclosure with localised heating from below-
creeping flow to the onset of laminar instability. J. Fluid
Mechanics 36(1):33-54.
Vinokur, M. 1983. On one-dimensional stretching functions for
finite-difference calculations. J. Computational Physics. 50:
215-234.
Wilkes, J. 0., and S. W. Churchil. 1966. The finite-difference
computation of natural convection in a rectangular enclosure.
AIChE Journal 12(1):161-166.


125
previous researchers. These indicate that the solution is reasonably
accurate.
Selection of the grid size. Many grid sizes ranging from 20x20
to 58x58 were tried. Increase in the number of the grid points
resulted in more iterations necessary to reach convergence per time
step. This was in addition to the increase in number of equations to
be solved. Thus computation time increased considerably with smaller
grid size. On the other hand, large grid sizes cannot resolve many
significant flow features. A grid system was chosen that resolved
all the significant flow features related to natural convection
heating of liquid foods and still consumed "reasonable" computer
time.
The isotherms and streamlines computed at 90 seconds for
39(0.9)x39(0.8) grid and 58(0.85)x58(0.76) grid (figure 23) are
shown in figures 24 and 25. The number in parenthesis are the
stretching parameter a in the grid stretching equations 38 and 39.
The streamlines are almost identical and temperatures are predicted
with less than 2% error for most of the enclosure. Thus, refining
grids beyond 39(0.9)x39(0.8) did not bring significant additional
information and the same grid was selected for the present study.
To further emphasize on the fact that use of 39(0.9)x39(0.8)
did not result in loss of any significant information in reference
to natural convection heating of liquid foods, the transient slowest
heating temperatures computed with the two grid sizes of
39(0.9)x39(0.8) and 58(0.85)x58(0.76) are shown in figure 26. The
figure shows that the two grid systems predict almost identical


170
Packer, G. J. K., and J. L. B. Gamlen. 1974. Calculation of
temperature measurement errors in thermocouples in convection
heating cans. J. Food Sci. 39:739-743.
Patankar, S. V. 1980. Numerical heat transfer and fluid flow. Hemi
sphere publishing Corporation, Washington.
Pepper, 0. W., and S. 0. Harris. 1976. Numerical simulation of
natural convection in closed containers by a fully implicit
method. Numerical/Laboratory computer methods in fluid
mechanics. Presented at the winter annual meeting of the ASME,
New York, New York. Dec 5-10.
Pflug, I. J., J. L. Blaisdell, and R. C. Nicholas. 1965. Rate of
heating and location of the slowest heating zone in sweet fresh
cucumber pickles. Food Tech. 19(6):121-126.
Phillips, T. N. 1984. Natural convection in an enclosed cavity. J.
Computational Physics. 54:365-381.
Ramaswamy, H. S., M. A. Tung, and R. Stark. 1983. A method to
measure surface heat transfer from steam/air mixtures in batch
retorts. J. Food Sci. 48:900-904.
Schneider, G. E., A. B. Strong, and M. M. Yovanovich. 1975. Finite
difference modelling of the heat conduction equation in general
orthogonal curvilinear coordinates using Taylor series ex
pansion. Advances in computer methods for partial differential
equations. Edited by R. Vichnevetsky. AICA, New Brunswick, New
Jersey.
Schwind, R. G., and G. C. Vliet. 1964. Observations and inter
pretations of natural convection and stratification in vessels.
Proceedings of the 1964 Heat Transfer and Fluid Mechanics
Institute, Berkeley, California. June 10-12. Edited by W. H.
Giedt and S. Levy. Stanford University Press.
Seki, N., S. Fukusako, and M. Suguwara. 1977. A criterion of onset
free convection in a horizontal melted water layer with free
surface. J. Heat Transfer, Trans, of the ASME 99:92-98.
Shyu, R. J., and C. K. Hsieh. In Press. Numerical simulation of
transient natural convection in enclosures heated from below. To
be presented at the ASME national meeting.
Shyy, W. 1985. A study of finite difference approximations to
steady-state, convection-dominated flow problems. J. Computa
tional Physics 57:415-438.
Steele, D. J. 1980. Microprocessors and their application to the
control of a horizontal batch retort. IFST Proceedings
13(3):183-193.


70
temperature value of the fluid next to the wall is made equal to the
specified value. On the adiabatic walls and at the centerline (which
is also considered adiabatic due to symmetry), a Taylor series
expansion gives
An +V2ISI An2 + 0(An3) (73)
W 9n lw
where stands for the derivative of temperature in the direction
normal to the wall and An being distance normal to the wall.
for f |u 0 (74)
An2
W
= 0W + 0(An2) (75)
JW+1
+1/2
a2e
9n2
Vl = 0W + In
Thus if 0w+j = 0^ is set, wall temperature is evaluated from the
adjacent temperature and it is second order accurate even though
is only first order accurate (Roache, 1982).
In the ADI method, boundary conditions are needed at the
computational "half" step. Since wall temperature 0,^ is not a
function of time, the boundary temperature value at this "half" time
step is simply equal to 0^.
Unlike temperature and velocities, vorticity and stream
function are not primary variables. This makes specification of
their boundary values difficult, particularly for vorticity.


36
amounts of memory that were unavailable until the 1960s. Since then
numerical studies of natural convection heating in enclosures are
being reported constantly. Among the numerical methods, finite
difference methods are probably the most popular in computational
fluid dynamics. The same was chosen for the present study.
Several choices were available for the finite-difference
analogs, computational boundary conditions (treatment of boundary
data in the numerical solution, as compared to real physical
boundary values), and the technique and order of solving individual
PDEs in the overall computational cycle. The relative merits of
these choices are discussed, and appropriate selections are
described in this section.
Grid Generation and Grid Stretching
The discrete points at which finite difference equations are to
be solved constitute the grid system for the problem. For computa
tional efficiency, the number of grid points should be the minimum
that is required to resolve all the significant spatial variations
of the flow. Since this problem involves a rather large step
increase in the wall temperatures together with a high Rayleigh
number (Ra), strong boundary layers are expected to form near the
walls. This is true particularly for the side (vertical) wall. In
order to resolve these boundary layers adequately, a large concen
tration of grid points would be necessary near the walls. An alge
braic method was used to generate the nonuniform grid system. The
particular algebraic expression chosen was evaluated for the proper-


17
by a flat narrow slit of light, exposed against a dark background
and photographed. Detailed study of flow in the boundary layer was
observed using a laser-dopper velocimeter which is an improved
version of the "particle-streak method" mentioned above. Temperature
values were measured by inserting thermocouples at several points in
the liquid. The temperature variation on the top free surface was
observed using a Thermovision Infrared camera. Radial velocity
profiles were presented at different heights in the liquid and at
different times during heating. His experiment and other works (Chu
& Goldstein, 1973) had showed convective heat transfer from the
bottom, as expected, when the bottom was also heated. His numerical
results, however, could not pick up the convective eddies generated
by the bottom heating. Thus his numerical results had poor corre
spondence with his experimental results.
A finite element method was used by Engelman and Sani (1983) to
study in-package pasteurization of fluids (beer) in bottles. In the
process studied, bottles entered the pasteurizer at 1.7C(35F) and
passed through several progressively hotter zones of hot water
spray, which raised the product temperature to 60C(140F). This
temperature of 60C is maintained in the holding zone. The product
then passed through several progressively cooler zones, which low
ered the temperature to 70 80F. Since the product was fluid and
the package was not agitated, it was primarily heated through nat
ural convection. Experimental details for this study was provided by
Brandon et al. (1981). The bottle geometry was complex (figure 1)
and so were the boundary conditions. The bottom of the bottle was


LIST OF REFERENCES
Aliada, S. R., and D. Quon. 1966. A stable, explicit numerical
solution of the conduction equation for multidimensional non-
homogeneous media. Chemical Engineering Progress Symposium
Series 62(64):151-156.
Anderson, D. A., J. C. Tannehill, and R. H. Pletcher. 1984.
Computational fluid mechanics and heat transfer. Hemisphere
Publishing Corporation, Washington, D.C.
Ball, C. 0., and F. C. W. Olson. 1957. Sterilization in food techno
logy. McGraw-Hill Book Company, Inc., New York.
Barakat, H. Z. 1965. Transient natural convection flows in closed
containers. Ph.D. Thesis. University of Michigan.
Barakat, H. Z., and J. A. Clark. 1966. Analytical and experimental
study of transient natural convection flows in partially filled
liquid containers. Proceedings of the Third International Heat
Transfer Conference. The Science Press, Ephrata 2:152-162.
Barreiro-Mendez, J. A. 1979. Prediction of time-temperature rela
tionships and thiamine retention during thermal processing of
canned model systems heated by natural convection. Ph.D. Thesis.
The Louisiana State University and Agricultural and Mechanical
College.
Beardsley, R. C. 1969. Alternative finite difference approximations
for viscous generation of vorticity at a stationary fluid
boundary. Report GRD/69-2, Massachusetts Inst, of Tech.
Bird, R. B., W. E. Stewart, and E. N. Lightfoot. 1976. Transport
Phenomena. John Wiley and Sons, New York.
Birkoff, G., R. S. Varga, and D. Young. 1962. Alternating direction
implicit methods. Advances in Computers. Vol 3. Academic Press,
New York.
Blaisdell, J. L. 1963. Natural convection heating of liquids in
unagitated food containers. Ph.D. Thesis. Michigan State
University.
Bontoux, P., B. Gilly, and B. Roux. 1980. Analysis of the effect of
boundary conditions on numerical stability of solutions of
Navier-Stokes equations. J. of Computational Physics 36:417-427.
161


96
Table 1.
Sequence of line types for identification of stream
function contours in figures lOa-o.


82
be derived algebraically for each of the boundaries or the computer
has to solve a simultaneous set of equations at the boundaries (see
Appendix 0). For Robin's boundary condition, the application of the
ADE method required separate algebraic expressions for the top, side
and bottom wall boundaries. Separate expressions were also required
for the top and bottom corners and the centerline. This is in sharp
contrast with the situation if the ADI method were used. Adjustments
would be necessary at the boundaries for the ADI method, but for use
of the ADE method it became clumsy (see appendix D).
For temperature specified at the boundary, this is trivial and
no additional manipulations as noted above are needed. Experiments
conducted later showed that because of the low thermal conductivity
of food (which is mostly water), the Biot number (= hR/k) is quite
large and it is the internal conduction resistance that is limiting.
Thus a convection boundary (Robin's) condition can be approximated
by a temperature specified (Dirichlet) boundary condition for this
purpose. The two shortcomings of the ADE formulation are, therefore,
of little consequence in the case of conduction-heated food. The
equations developed for the convection boundary condition are given
in appendix D.
Coding of the Computer Program
The ADE formulation for transient temperature calculation using
temperature specified boundary condition was coded in PASCAL and
used as a subroutine in a real time control algorithm (to be
discussed later).


65
Second, transport velocities are defined normal to the grid volumes
*
in terms of a new set of stream function values xp- .
' iJ
* * *
uf .. vf ... Vr ,
J YWT? rj+vJzi+i/2- i-v^
i-lJ
(64a,b)
The stream function values ik. are defined at the corner of the

grid volumes. These tp. values are related to the previous stream
T 5 J
function values ip. through a consistency criterion which says
Lt u. u.
Ar+0
i ,J i ,J
= 0
(65a,b)
Lt vi i vi i
Ar-K)
i|). was defined as
. .
H ,3
3>.
2
2
^ <*1,J + +1+1,j> (66)
then, using definitions of v^ and v. (equations 64b and
63b), it was shown that the consistency criterion (equation 65b)
require
+ xj = 1
(67)
Similarly, using definitions of u- and u.¡ it was shown that
*1 jJ >J


125
Figure 21. Temperature at the slowest heating point
during natural convection heating contrasted
with that during conduction heating in a
cylindrical can


47
0R
0.
i
Vi
0
L
Vi
(uR > 0)
(Ur < 0)
(42)
(uL > 0)
(uL < 0)
where Ur and ul are velocities as explained below. This latter
upwind method also has a formal accuracy of 0(az). However, it
behaves like it is second-order accurate when the quantity
( 0 or in ) is slowly varying (Roache, 1982). Torrance (1968) demon
strated that the second upwind method is indeed superior to the
first upwind method (Roache, 1982). The second upwind method was
used in the present formulation.
The velocities uR and u^ are at the right and left face
respectively of the grid volume as shown in figure 4. As will be
explained later, a kinematically consistent velocity formulation
(Parmentier and Torrance, 1975) was used. In this formulation,
transport velocities (u[ _ and v( in figure 4) are defined normal
to the grid volumes. Accordingly,
UR
UL=
u
f
1-1J
v A.}
L vi,j-l
Following is a development of the finite differences for the
convection terms in the temperature equation. In the z direction,


127
After 90 sec
Figure 24. Isotherms after 90 seconds of heating computed
using 39x39 grid (left half of the figure)
compared with computations using 58x58 grid
(right half of the figure)


6
constant retort temperature for a prescribed length of time followed
by rapid cooling. These process times are calculated on the basis of
sufficient bacterial inactivation in each container to comply with
public health standards and reduce the probability of spoilage to a
specified minimum. Estimation of bacterial inactivation requires
understanding of thermal death kinetics of bacteria, along with the
temperature history the food (containing the bacteria) experiences
during the process. Following is a brief introduction to the thermal
death kinetics of bacteria (Ball & Olson, 1957; Stumbo, 1965).
Thermal death of bacteria follows first order kinetics. If C is
bacterial concentration (number of bacteria per unit volume of food
material) at time t,
dC 1
dt (0.434)Dl
(1)
where (0.434)*D is the inverse of the rate constant. If Cn- is the
initial concentration and C is concentration after time At
log
At
D
(2)
The constant D (known as 0 value), which is the inverse of the rate
constant (multiplied by the factor 0.434), is the time required to
reduce the bacterial population by one log cycle at a specified
temperature. The temperature dependency of the D value is given by


61
Discretization of the Elliptic Stream Function Equation
The stream function equation (28), obtained from the definition
of u = 4^. J is also a statement of the circulation theorem
3z 3r
which is
/ ai dA = / v Jl
(58)
area perimeter
element of area
Thus the difference formulation of the stream function equation has
to satisfy the circulation theorem in addition to conserving mass.
Difference approximations that conserve mass are more difficult to
achieve in a curvilinear coordinate system as compared to the
Cartesian coordinate system.
To satisfy both the circulation theorem and the conservation of
mass, Parmentier and Torrance (1975) introduced the concept of
tangential velocities parallel to the control volume faces and
transport velocities perpendicular to control volume faces (Figure
4). In what follows, their approach has been adapted for the non-
uniform grid system.
The circulation theorem (equation 58) is written for a grid
volume around node (i,j) as shown in figure 4
(59)
Z i + V2 Z i V2 r j + V2 r j V2


11
This approach was used by Mulvaney and Rizvy (1984). They used a
thermocouple-fitted test can in every retort batch to obtain actual
temperature T(t). The system proposed by Navankasattusas and Lund
(1978) also planned to measure can center temperature on-line. In
commercial practice these methods are cost-prohibitive with regard
to production efficiency, and are considered impractical.
Numerical Models. Instead of measuring T(t) at the can center,
T(t) can be predicted for arbitrary variations in TR(t). In
general, analytical solutions for T(t) are of limited use for
arbitary variations of TR(t). Teixeira and Manson (1982) described
a numerical finite difference approximation to the heat conduction
equation for predicting T(t) from TR(t). This is conceptually quite
different from the approaches mentioned earlier. With a finite
difference model, T(t) can be predicted for truly arbitrary
variations in T^(t) and, of course, use of this model eliminates the
need for having a test can with an actual thermocouple inserted in
it. This model was first described and verified against published
procedures by Teixeira et al., 1969; and later verified against
experimental data and other published procedures for time-varying
boundary conditions by Teixeira et al. (1975 a,b).
However, in the proposed retort control algorithm of Teixeira
and Manson (1982), every time a deviation took place, a new value of
tft was found through simulation assuming the deviated retort temper
ature continued for the rest of the heating process. This is unnec
essary because any further changes in the retort temperature would
make the estimated t^ value useless and fresh estimations of t^


y
V
5
P
a
T
4
X
u
)
*

viscosity of liquid being heated
kinematic viscosity of liquid being heated = y/p
radial distance in computational plane = 5(r)
density of liquid being heated
deformation parameter in grid stretching
t = non-dimensional time
H2
radial coefficients for kinematic consistency
radial coefficients for kinematic consistency
non-dimensional stream function = tt-
na
I
r 3r
stream function defined by =
non-dimensional vorticity
9v 9u H2 9v 9u> _
9Z 3r = a -'
3z 3r
vorticity
optimum relaxation parameter for SOR
_ 1 s'f
r 9z
H2
a
Subscripts
a
center
cool
heat
i
j
m
n
W
ambient
geometrical center of can
cooling cycle
heating cycle
grid points in vertical direction
grid points in radial direction
grid point on vertical boundary (z = H)
grid point on radial boundary (r = R)
Wall
ix


148
Table 3. Adjusted heating times and resulting lethality (F0) in
response to process deviations using proposed on-line
control logic and method of Giannoni-Succar and Hayakawa
(1982).
Reference
Process
Step Drop (figure 33)
Drop duration 10 min.
Drop amount 15F
Gradual drop (figure 34
Drop duration 20 min.
Maximum drop 20F
No
deviation
On-line
control
System
Giannoni-S.
&
Hayakawa
0n-line
Control
System
Giannoni-S.*
&
Hayakawa
Time of
Heating
(minutes)
66.8
72.2
72.0
74.6
77.8
Total F0
(minutes)
6.24
6.11
6.02
6.18
7.49
* Treated as step drop of 20F for 20 minutes duration.


14
method. In their later work (Teixeira et al., 1975a) an alternating
direction explicit (ADE) method (Saul'yev, 1957; Aliada and Quon,
1966) was used that was unconditionally stable but still explicit.
This method was first applied to simulate thermal processing of
canned foods by Manson et al. (1974). Teixeira et al. (1975a,b)
compared outputs from the ADE method for using constant boundary
temperature and sinusoidal time-varying boundary temperature with
the results from corresponding analytical solutions. The analytical
results for constant boundary temperature agreed within 0.3C
(0.5F) with the results from the ADE method, except at interior
points near the surface after small initial time (less than 3
minutes). Similar agreement between the results of the analytical
and the ADE formulation was also obtained when a sinusoidally
varying boundary condition was used.
Teixeira (1971) also compared the ADE finite difference outputs
for conduction heating in a can with experimental data for constant
boundary temperature and for step increases in the boundary tempera
ture during heating. The agreement was good except at initial times.
In addition to this ADE method, other methods of solution to
the conduction equation needed to be investigated from the point of
view of their suitability of computation in real-time and ability to
accomodate arbitrary boundary temperature variations.
Natural Convection Heating of Canned Foods
Internal (in enclosures) natural convection flow problems have
received much less attention than the external ones. Also, internal


Water
T,= 30*C
top-121'C
slda=121 *C
bot-121*C
R-4.19cm
h=10.07om
R-3B.0.B0
Z=38,0.90
a-o.ooot
-25
-20
-15
-to
-7
5
10
After 300 sec
Figure 10m. Isotherms and streamlines in a
cylindrical can after 300 seconds
of heating
Water
T,= 30*C
top-121*C
slda = l'21 *C
bot-121*C
R-4.19cm
H=10.07om
R-3B.0.B0
Z=3B,0.90
a-o.oooi
-to
-8
-6
-4
2
4
6
Figure lOn. Isotherms and streamlines in a
cylindrical can after 600 seconds
of heating


Analysis of Boundary Conditions
The inverse slopes of the heat penetration curve obtained
experimentally for various heating and cooling processes are given
in table 2.
The data presented are averages for several similar processes.
The standard deviations for these values among replicate cans and
replicate processes were in the order of 1.5 minutes. The results
show that fill temperature (a measure of head space vacuum) did not
have a significant effect on the heating or cooling rates. This was
expected since even though there was considerable difference in
vacuum produced in the two fill temperatures, the physical dimen
sions of the head spaces were not much different.
The difference in heat transfer rates between heating and
cooling were seen to be more pronounced than the difference between
two rates of heating by steam and hot water respectively. The dif
ference between heat transfer rates during heating and cooling was
explained primarily by changes in the thermal diffusivity of the
product because of considerable difference between the average tem
peratures experienced during heating and cooling (Datta et al.,
1984). The small difference between the rates of heating with steam
as opposed to using hot water was attributed to the much larger
resistance inside the can (large Biot number) due to very low
thermal conductivity of foods, which limits the rate of heat
transfer. This means that the convection boundary condition that
exists in an actual heating or cooling situation can be approximated
by a specified temperature boundary condition for calculation
purposes.


13
were derived assuming, among other things, a step change in the
boundary (can wail) temperature from the initial to the heating
temperature and maintaining the boundary heating temperature
constant throughout the process. More general forms of boundary
conditions have also been formulated. Hayakawa (1964) developed the
general form for an analytical solution that could predict the
transient temperature distribution inside a cylindrical can of
conduction-heated food when the boundary (can wall) temperature was
any Laplace-transformable function of time.
In practice, the heating medium temperature, which affects can
boundary temperature, is not under complete control and can go
through unexpected variations from its design value. For such
arbitrary variations in the boundary temperature, analytical
solutions are of limited use.
Numerical Studies
A finite difference analog of the heat conduction equation can
also be used to achieve the transient temperature of food at finite
but small increments of time. Such methods can utilize the actual
boundary temperature "read" for the incremental time and calculate
the interior temperature values based on this boundary temperature.
Teixeira et al. (1969) were the first to use a forward difference
explicit finite differencing method to study the transient
temperature distribution inside the cylindrical can. Use of the
explicit finite differencing severely limited the choice of grid
size and time increment, due to stability requirements for such a


80
70
60
50
40
30
T
MID POINT ON AXIS
1/3RD POINT ON AXIS
SLOWEST POINT
20 40
Tima of hooting (seconds)
60
Figure 20. Transient temperature at various locations
during initial times of heating of water in a
cylindrical can


130
100
70
40
10
25
RETORT
AN CENTER
- 15
TTTTTTT rTT TTTTJT I I I | I I I l' j T I I I
25 50 75 100 125 150
20
- 10
- 5
0
TIME (minutes)
34. Computer plot of a process that experienced a step functional
drop of 8.3C (15F) in temperature between 50th and 60th
minute and still maintained an Fn of 6.02
cr>
ACCOMPLISHED F0 (minutes)


63
The tangential velocities ik and .= at the faces of the
jJ 1 jJ
grid volume are obtained by finite differencing equation 29
u* .=
i
^i J+l'^iJ
'j+V2 rj+i* rj)
(60a,b)
ViJ:
N+U ~ ^ ,j
rj(-zi+r zi J
Substituting these definitions in equation (59)
). .=
V zi+V2"zi-i/2
i+l,j i j
zi+r
z.
1
,J 1-1, J
z.-z. .
1 1-1
1 r i J+r^i ,j ^i.j^i.j-l 1
rj>Vfrj- Vz Wj+i-pjr rj-i/2irj-rj-i)
(61)
Note that the same finite difference equation (61) can be obtained
from central differencing the stream function equation (28) without
introducing the circulation theorem. This is shown in Appendix A.
After algebraic manipulation,


zi+l
Figure 4.
Notations of various quantities defined on a
non-uniform grid system in cylindrical
geometry


125
110
O
O
(D 90
L_
O
L_
CD
CL
70
£
CD
y 50
*
*
*
* *
* 39(0.93X39(0.8) grid
v 58(0.763X58(0.58) grid
_i i i i
120 240 360 480
Time of heating (seconds)
600
Figure 27. Transient temperature at l/3rd height at the
centerline computed with 39x39 and 58x58 grid


UFDissertations, 08:38 AM 6/10/2008, UF Libraries:Digital Dissertation Project
Page 2 of 2
Internet Distribution Consent Agreement
In reference to the following dissertation:
AUTHOR: Datta, Ashim
TITLE: Numerical modeling of natural convection and conduction heat transfer in
canned foods with application to on-line process control / (record number:
872393)
PUBLICATION DATE: 1985
I,
Aj> k/V K
as copyright holder for the aforementioned
dissertation, hereby grant specific and limited archive and distribution rights to the Board of Trustees of
the University of Florida and its agents. I authorize the University of Florida to digitize and distribute
the dissertation described above for nonprofit, educational purposes via the Internet or successive
technologies.
This is a non-exclusive grant of permissionsTor 5peciEc..aff-line.and.clniline uses for an indefinite term.
Off-line uses shall be limited to those specifically allowed by "Fair-Use" as prescribed by the terms of
United States copyright legislation ~(cf, Title 17, UTS. Code) as well as to the maintenance and
preservation of a digital archive copy. Digitization allows~the University- of Florida to generate image-
and text-based versions as appropriate and to provide aftdlenhance access-using search software.
This grant of permissions prohibits use of the digitized'Versiofls'for commercial use or profit.
tWC-L- k1
Signature of Copyright Holder
Askim L T)xfhA
Printed or Typed Name of Copyright Holder/Licensee
Personal information blurred
ae/n/ee
Date or Signature
Please print, sign and return to:
Cathleen Martyniak
UF Dissertation Project
Preservation Department
University of Florida Libraries
P.O. Box 117007
Gainesville, FL 32611-7007
Printed for Ashim Datta
6/10/2008


30
dimensional flows when no free surface is present, iterative methods
are used to solve the Poisson equation, and transient pressure solu
tions are not required (Roache, 1982).
1. The (u,tji ) system needs to solve one parabolic equation for
a) and one elliptic Poisson equation for ip as compared to the
(u,v,p) system which requires the solution of two parabolic
equations for u and v and one elliptic equation for p.
2. With the (oi, ) system, two additional differential equations
(29a and 29b) are needed to get the velocities u and v, but
the u and v momentum transport equations are more complicated
than the u transport equation.
3. At no-slip walls, the boundary conditions for u and v are
known explicitly (uwan = 0, vwan = 0) as compared to u ^
which is not known explicitly. This would be a great advantage
for using implicit methods for the (u, v, p) system but non
linear instabilities of the pressure term prevents the (u, v,
p) system from using implicit methods.
4. The elliptic equation that needs to be solved for the pressure
takes much longer to converge than the elliptic equation for
4> (eq. 28).
Conduction Heat Transfer in Canned Foods
The POE describing conduction heat transfer in a cylindrical
body of solid-packed food material is presented here. The boundary
conditions present during the thermal processing of canned foods are
described next.


85
be clamped tightly in place once inserted to the desired height. The
holes were located so that the cans stood two abreast in four
staggered rows. This configuration helped to assure that all the
cans received nearly equal exposure to the heating and cooling
media, and there could be no possible movement of the cans during
sudden pressure changes that could occur during initial cooling.
The same eight cans were used repeatedly in four different
processes with two replicates of each. The same retort temperature
(121C) and process time (75 minutes) was used in all four
processes. They differed in the type of heating medium used (pure
steam or pressurized water), and in the retort pressure during
cooling (either held constant at the level required during heating
or allowed to drop suddenly to near atmosphere with the introduction
of cooling water). The flow rate of recirculating cooling water was
adjusted so as to achieve a low mainstream velocity of 0.05 meters
per second between the cans. Transient temperatures and pressures
were recorded with a data logger.
While recording the transient temperature response at the can
center, it was observed that sudden depressurization of the retort,
during start of the cooling cycle, resulted in a very different
center-temperature profile during the early period of cooling. The
can center temperature was found to drop much more rapidly in the
case of sudden depressurization as compared to when the retort
pressure was maintained constant. It was postulated that such rapid
fall in the retort pressure caused cans to expand, thus creating
sudden loss of internal pressure causing the product to "boil"


159
At the top boundary, finite differencing the boundary condition
Qn+1 Qn+1
nz,j nz-lj hHQn+l
Az k nz,j
(95)
and at the sidewall
in+1 en+1
i,nr i ,nr-l hHQn+l
Ar ki ,nr
(96)
The finite differences for the boundary conditions and the governing
equations are now combined. At the upper corner, combining equations
93, 95 and 96
0n+1 = c-f 0n .+ S(-0n .+ 0n 9 .)
nz,nr AU+P+S)1 nz-l,nr-l 1 nz-l,nr-l nz-2,nr-l;
+ pf-0n + 0n 1 + Hf_Qn 'll
nz-l,nr-l unz-l,nr-2J Rl nz-l,nr-2"
. h (P+Q/R)(1+yAz) S(l+YAr)
where A = (1+ -P-Ar) (1+ -r-Az)
K K 1+P+S 1+P+S
Combining equations 93,95 at the top wall,
(97)
,n+l
nz,J A(1+P+S)
- {0n .+ S(-0n .+ 0n 9 .)
Dj.c\ 1 nz-1 ,j nz-1 ,j nz-2,j ;
+ P((l+ Az)0n+1 0n .+ 0n i -i )
u k nz,j+l nz-l,j nz-l,j-l'
+ %-0n ,+ (1+ -r- Az)0n+1 )}
r^ nz-l,j-l v k 1 nz,j+l;j
(98)
where A = 1+ jrAz
1+P+S


APPENDIX A
ALTERNATIVE FINITE DIFFERENCING OF STREAM FUNCTION EQUATION
The stream function equation
_1 32^ j), 1_
' 7 ~Bz7 + 3r^ r 3r J
Using central differencing
B2ip
Bz7
_3 > 3i{>> 3p 3 > 3^\
3z '3z 3z 3n (Bz'
1
3z/3n
3^
3Z
i+V?J i- V? J
An
*i+lJ~*i J ^iJ ~ ^i-lj
z.z.
1+1 1
V zi-i
zi+r zi-i
2 An
2
zi+r zi-i
An
j-^i+lj" *itj *i,j *i-l,j
zi+r zi
zt- zi-i
)
3 /1 3iK
3rV 3r'
35 1_ (I Mi
3r 35 'r Br1
_3t{> 2. 3^1
i W jt.i+ v? ri-v,arli,.i-i/,
r.i+r ri-i 45
2A5
f 1 > J y I ^ ^ 3 T > J
mv 1. ( r r l~
rj+r rj-i v4(rjn- rjr rj-4(rr rj-i>'
Combining the terms, we get equation 61.
154


162
Brandon, H., P. Pelton, and G. Staack. 1981. State-of-the-art
methodology for evaluation of pasteurizer heating and cooling
processes. MBAA Technical Quarterly 19(1):34-40.
Briley, W. R. 1971. A numerical study of laminar seperation bubbles
using the Navier-Stokes equations. J. Fluid Mech. 47(4):713-736.
Chu, T. Y., and R. J. Goldstein. 1973. Turbulent natural convection
in a horizontal layer of water. J. Fluid Mech. 60:141-159.
Datta, A. K., A. A. Teixeira, and J. E. Manson. In Press. Computer-
based retort control logic for precise on-line correction of
process deviations. J. Food Sci.
Douglas, J. Jr. 1955. On the numerical integration of
S2u/5x2 + 62u/6y2 = <5u/6t by implicit methods. J. Society of
Industrial Applied Mathematics 3(1):42-65.
Engelman, M. S., and R. L. Sani. 1983. Finite-element simulation of
an in-package pasteurization process. Numerical Heat Transfer
6:41-54.
Evans, L. B., R. C. Reid, and E. M. Drake. 1968. Transient natural
convection in a vertical cylinder. AIChE Journal 14(2):251-259.
Forsythe, G. E., and W. R. Wasow. 1960. Finite difference methods
for partial differential equations. John Wiley and Sons, New
York.
Frankel, S. P. 1956. Some qualitative comments on stability
considerations in partial differential equations. Proceedings of
the Sixth Symposia in Applied Mathematics, AMS, 6:73-75.
Fromm, J. 1964. The time dependent flow of an incompressible viscous
fluid. Methods of Computational Physics 3:345-382.
Gentry, R. A., R. E. Martin, and B. J. Daly. 1966. An Eulerian
differencing method for unsteady compressible flow problems. J.
Computational Physics 1:87-118.
Getchell, J. R. 1980. Computer process control of retorts. ASAE
paper no. 80-6511. Presented at the winter meeting of the ASAE
at Chicago, Illinois.
Giannoni-Succar, E. B., and K. Hayakawa. 1982. Correction factor for
deviant thermal processes applied to packaged heat conduction
food. J. Food Sci. 47(2):642-646.
Hayakawa, K. 1964. Development of formulae for calculating the
theoretical temperature history and sterilizing value in a
cylindrical can of thermally conductive food during heat
processing. Ph.D. Thesis. Rutgers, The State University.


Water
T,=30*C
op-121*C
Slde121*C
boI-121*C
R-4.19cm
H=10.67om
R-57.0.76
2=57,0.85
8=0.0001
-300
-260
-220
-180
-140
-100
-60
-20
20
60
Figure lOi. Isotherms and streamlines in a
cylindrical can after 9 seconds of
heating
Water
T,= 30'C
top-121'C
slde=121*C
bot=121'C
R=4.19cm
h=10.67cm
R-57,0.76
Z=57,0.85
6=0.0001
-300
-260
-220
-180
-140
-100
-60
-20
20
60
A ter 10 sec
Figure lOj. Isotherms and streamlines in a
cylindrical can after 10 seconds
of heating


78
The temperature equations (47a and 47b) were solved using
the ADI technique. This gave new temperature values after
incremental time At.
The vorticity equations (55a and 55b) were then solved
(also using the ADI technique). The interior vorticity
values calculated were provisional since old values were
used for boundary vorticity.
The stream function equations (61) were then solved for
interior stream function values using the point SOR
technique. This required only interior vorticity values.
Previous stream function values were used as an initial
guess.
Boundary vorticity values ( Z^) at solid walls (top,
bottom and side) are first calculatedusing equation 76
based on the newly calculated stream function values.
However, before updating, the boundary vorticity values
(Zb) are underrelaxed as
n+1
m+1
= Y a).
n+1
m+1
+ (1 y)
n+1
m
(BO)
-n+1
m+1
where = boundary vorticity estimated using
equation and 76, with 0 Steps 2 through 4 are repeated until the estimated ^
values (using equation 76) in two consecutive iterations
meets the convergence criterion


TIME INCREMEMT (seconds)
0.25
T
T
T
T
0.20
TIME OF SIMULATION (seconds)
Figure 28. Time step history during a typical run of the
numerical heat transfer model for natural
convection
GJ
ro


37
ties listed below (Anderson et al., 1984; Ka'lney de Rivas, 1971).
Here (r,z) refers to the physical domain and (5,n) refers to the
computational domain with A£ and An as constants.
1. The mapping from (C,n) to (r,z) must be one to one.
2. Grid lines should be smooth to provide continuous
transformation derivatives. This requires r(s) and z(n)
as well as and ^ should be continuous over the closed
intervals 0 functions defined on a closed interval are bounded, this in
turn implies that
dr
cff
dr
and
dz
should be finite over the
whole interval. If becomes infinite at some point, the
mapping r(c) would give poor resolution near that point,
which cannot be improved by increasing the number of points,
d p
since a r (^|-) (as) Similarly for dz/dn.
3. Close spacing of grids where large numerical errors are ex-
d p i
pected. For this problem, that means -rr\ =0 at the
a5'r=R/H
sidewall which ensures high resolution of the boundary layer
near r = R/H. Similarly at z = 0 and z = 1.
4. Absence of excessive grid skewness which can sometimes
exaggerate truncation errors (Raithby, 1976).
An algebraic transformation equation suggested by Kublbeck et
al. (1980) for natural convection in cavities was
n = (-1+ tan[ir/2 (2z-l) p]j
2l
tann/2 a
(37)


COOLING
WATER
Figure 6.
Experimental setup for conduction heating
00


141
This is important for real-time calculation since the ADE method
becomes considerably more complicated when required to account for
convection boundary condition as compared to the specified temper
ature boundary condition.
Comparison of the Transient Temperature Values
The transient temperature values recorded at the can center as
heating progressed are reported in figure 31. The observed
temperature values are compared with the predicted values using the
analytical solution to (equation 8a) and the ADE formulation to the
conduction equation. As can be seen, there was excellent agreement
between the predicted and observed values.
Effect of Sudden Pressure Drops in the Can
Figure 32 illustrates the effect of sudden depressurization of
the retort on transient temperature values at the can center. This
had been shown to be due to internal product movement from sudden
loss of pressure inside the can (Teixeira et al., 1985) and empha
sizes the importance of maintaining the retort pressure during
cooling.
Performance of the On-line Control Logic
The estimated process times and the accumulated FQ values for
the three process conditions (figures 33, 34 and 35) are presented
in table 3. For the reference process without any deviation (figure
33) the CF method predicted a correction factor close to 1 or no


73
At the centerline,
vi,0- vi-l ,0 ui ,0" ui,0
ui ,0" z. i.- z.
i,- z. i, ri.- 0
i + 72 1 72 72
=0
(76c)
since
vlj g and v^ Q from eqn. 60b
t i.r *1,0 2*i,i
i,0 tl/2(rr 0) r\jV 1
ui,0=0
At the sidewall,
t t
v. v. .
i ,n i-l,n
U). = 1
u u- ,
19 n in1
1>n" Ui+V2- Zi-V^ (v vv)
- Ill,
i ,n-l
r 1 ,(r -r ,) (r ~r 1.)
n-l/2 n n-1 n n-72
(76d)
since
v? n and v^ . = 0 from eqn. 60b
1 ) n i
. ip. ~ip. ip. ,
ut = i,n i ,n-l i ,n-l
i,n-l r i.(r r TJ r i.(r -r 7}
n-V2 n n-1 n-72 n n-1
u. = 0
i ,n
Identical expressions can be derived using Taylor series expansions
of equation 28 at the walls. This is shown in the appendix B.
These expressions are first order accurate. But it is the only
form consistent with the kinematic velocity formulation (Paramentier
& Torrance, 1975). The second order form using Taylor series
(Jensen, 1959) has also been used. However, sometimes the second


133
problem. For purposes of comparison, the explicit formulation of
Barakat (1965) for the set of equations (26-29) was used. In their
formulation, they "lagged" the velocities u and v in the convection
terms and used a backward finite difference for these terms. Their
stability analysis showed the following restrictions on At
At (
i ,J
Az
2
(AZ)2
(Ar):
(86)
Non dimensionalization of the variables used in the present study is
different from that used by Barakat (1965). Accordingly equations
(84-86) have been adjusted to conform to the present study. For Pr >
1, as is the present situation, inequality (85) is more restrictive
than (84).
For a non-uniform coordinate system, Ar and Az to be used are
their smallest values in the grid systems (Ka'lney de Rivas, 1971).
However, the allowed time step At here is also a function of grid
point velocities. Thus, using equation 85, At (and therefore At )
is calculated at all the grid points. For a 39(0.9)x39(0.8) grid
Min At. =
i, j ,J
.07seconds


Ra = Rayleigh number = Gr*Pr
t = time
T = temperature at any point inside at time t > 0
Tg = initial uniform temperature inside (at t = 0)
Tj = boundary temperature at time t > 0
Tr = retort temperature at time t
LI
u = = non-dimensional velocity in vertical direction
a J
= velocity in vertical direction
H -
v = v = non-dimensional velocity in radial direction
v = velocity in radial direction
z = -q- = non-dimensional distance in vertical direction
z = distance in vertical direction
Z = Z-value for organism
a
B
Y
A
6
e
n
0
A
A
-P
A
A
m
thermal diffusivity = k/pCp
coefficient of thermal expansion
relaxation parameter for boundary vorticity
designates a difference when used as a prefix
convergence criterion for boundary vorticity
convergence criterion for SOR
vertical distance in computational plane = n(z)
T-T l
= non-dimensonal temperature
VT,
first eigenvalue in z direction
eigenvalues in z direction
first eigenvalue in r direction
eigenvalues in r direction
vi i i


43
Instead of using a fully implicit set of equations, alternating
direction implicit (ADI) methods were introduced in companion papers
by Peaceman and Rachford (1955) and Douglas (1955). The advantage of
the ADI method over a fully implicit method is that the set of
equations, although implicit, is tridiagonal. Further the ADI method
as applied to linear equations has a formal order of accuracy of
0(At2,Az2,Ar2). The stability of this method is also unconditional
as determined by Von-Neumann stability analysis. For irregularly
shaped regions, programming could get complicated (Roache, 1982) but
for simple rectangular regions, as is the present situation, that is
no particular problem.
The practical advantage of the ADI method over the explicit
method is, however, nothing like that indicated by Von-Neumann
analysis (Roache, 1982). It is unstable both at large time steps
and/or insufficient convergence of boundary vorticity. In general,
many researchers have indicated that ADI methods do allow larger
time steps and faster overall computation, by a factor of two or
more, and furthermore allow second order accuracy in time. The ADI
method is a very popular method in computational fluid dynamics.
While solving the set of PDEs (26-29) Barakat and Clark (1966)
compared explicit and implicit methods for their study and decided
against the use of the implicit method due to unavailability of the
vorticity boundary conditions at the wall (to be explained later).
They allowed vorticity boundary values to lag one time step, neces
sary for such explicit formulations. The lagging of boundary
vorticity is considered undesirable (Roache, 1982) and iterations


91
f0W + (Fo001 ^simulated Fo <82>
is satisfied, the computer turns off the steam and lets in cooling
water. The computer still keeps reading the retort temperature
Tr(t) and calculating T(t) and F0(t). When T(t), the calculated
temperature at the can center, is below a certain specified value,
cooling is ended by stopping the cooling water flow and the water is
drained prior to unloading the retort.
At the end of the process a complete documentation of measured
retort temperature history Tp(t), calculated can center temperature
history T(t) and the accomplished FQ(t) is kept on file and can be
documented in both tabular and graphical form.
The control algorithm presented in figure 8, implemented in
real-time with a computer, can ensure designed sterilization F^ in a
product for arbitrary variations in retort temperature. The
performance of this system for correcting deviations was compared
with process corrections predicted by the correction factor (CF)
method of Giannoni-Succar and Hayakawa (1982). Comparison of the
two methods was based on
1) estimated process times, and
2) estimated accomplished F0 values.
Three different processes were considered:
1) reference process with no deviation,
2) step drop and later step rise in process temperature, and
3) gradual drop and later step rise in temperature.



PAGE 1

180(5,&$/ 02'(/,1* 2) 1$785$/ &219(&7,21 $1' &21'8&7,21 +($7 75$16)(5 ,1 &$11(' )22'6 :,7+ $33/,&$7,21 72 21/,1( 352&(66 &21752/ %\ $6+,0 .80$5 '$77$ $ ',66(57$7,21 35(6(17(' 72 7+( *5$'8$7( 6&+22/ 2) 7+( 81,9(56,7< 2) )/25,'$ ,1 3$57,$/ )8/),//0(17 2) 7+( 5(48,5(0(176 )25 7+( '(*5(( 2) '2&725 2) 3+,/2623+< 81,9(56,7< 2) )/25,'$

PAGE 2

'HGLFDWHG WR WKH PHPRU\ RI P\ JUDQGIDWKHU DQG IDWKHU

PAGE 3

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f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f PDGH DYDLODEOH IURP 'U :LOOLDP ( 'XQQ IURP WKH

PAGE 4

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

PAGE 5

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

PAGE 6

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

PAGE 7

/,67 2) 6<0%2/6 $ FRQVWDQW DEF FRHIILFLHQWV LQ WULGLDJRQDO PDWULFHV I LQYHUVH VORSH RI WLPHWHPSHUDWXUH JUDSK J JUDYLW\ ) VWHULOL]DWLRQ ) YDOXH + *U *UDVKRI QXPEHU JH 7L 7f K RXWVLGH KHDW WUDQVIHU FRHIILFLHQW + KHLJKW RI F\OLQGHU M M YDOXH D FRQVWDQWf -J %HVVHO IXQFWLRQ RI RUGHU ]HUR N WKHUPDO FRQGXFWLYLW\ Q QRUPDO GLUHFWLRQ Q] QXPEHU RI JULG SRLQWV LQ ] GLUHFWLRQ QU QXPEHU RI JULG SRLQWV LQ U GLUHFWLRQ S SUHVVXUH 3U 3UDQGWO QXPEHU \&SN YD U QRQ GLPHQVRQDO GLVWDQFH LQ UDGLDO GLUHFWLRQ U GLVWDQFH LQ UDGLDO GLUHFWLRQ 5 UDGLXV RI F\OLQGHU YLL

PAGE 8

5D 5D\OHLJK QXPEHU *Ur3U W WLPH 7 WHPSHUDWXUH DW DQ\ SRLQW LQVLGH DW WLPH W 7J LQLWLDO XQLIRUP WHPSHUDWXUH LQVLGH DW W f 7M ERXQGDU\ WHPSHUDWXUH DW WLPH W 7U UHWRUW WHPSHUDWXUH DW WLPH W /, X f§ QRQGLPHQVLRQDO YHORFLW\ LQ YHUWLFDO GLUHFWLRQ D YHORFLW\ LQ YHUWLFDO GLUHFWLRQ + Y f§ Y QRQGLPHQVLRQDO YHORFLW\ LQ UDGLDO GLUHFWLRQ Y YHORFLW\ LQ UDGLDO GLUHFWLRQ ] T QRQGLPHQVLRQDO GLVWDQFH LQ YHUWLFDO GLUHFWLRQ ] GLVWDQFH LQ YHUWLFDO GLUHFWLRQ = =YDOXH IRU RUJDQLVP D % < $ H Q $ $ 3 $ $ P WKHUPDO GLIIXVLYLW\ NS&S FRHIILFLHQW RI WKHUPDO H[SDQVLRQ UHOD[DWLRQ SDUDPHWHU IRU ERXQGDU\ YRUWLFLW\ GHVLJQDWHV D GLIIHUHQFH ZKHQ XVHG DV D SUHIL[ FRQYHUJHQFH FULWHULRQ IRU ERXQGDU\ YRUWLFLW\ FRQYHUJHQFH FULWHULRQ IRU 625 YHUWLFDO GLVWDQFH LQ FRPSXWDWLRQDO SODQH Q]f 77 O QRQGLPHQVRQDO WHPSHUDWXUH 97 ILUVW HLJHQYDOXH LQ ] GLUHFWLRQ HLJHQYDOXHV LQ ] GLUHFWLRQ ILUVW HLJHQYDOXH LQ U GLUHFWLRQ HLJHQYDOXHV LQ U GLUHFWLRQ YL L L

PAGE 9

\ 9 3 D 7 ; X f r YLVFRVLW\ RI OLTXLG EHLQJ KHDWHG NLQHPDWLF YLVFRVLW\ RI OLTXLG EHLQJ KHDWHG \S UDGLDO GLVWDQFH LQ FRPSXWDWLRQDO SODQH Uf GHQVLW\ RI OLTXLG EHLQJ KHDWHG GHIRUPDWLRQ SDUDPHWHU LQ JULG VWUHWFKLQJ f§W QRQGLPHQVLRQDO WLPH + UDGLDO FRHIILFLHQWV IRU NLQHPDWLF FRQVLVWHQF\ UDGLDO FRHIILFLHQWV IRU NLQHPDWLF FRQVLVWHQF\ QRQGLPHQVLRQDO VWUHDP IXQFWLRQ WW QD U U VWUHDP IXQFWLRQ GHILQHG E\ QRQGLPHQVLRQDO YRUWLFLW\ Y X B + } Y X! B = f U D f n ] U YRUWLFLW\ RSWLPXP UHOD[DWLRQ SDUDPHWHU IRU 625 B VnI U ] + D 6XEVFULSWV D FHQWHU FRRO KHDW L M P Q : DPELHQW JHRPHWULFDO FHQWHU RI FDQ FRROLQJ F\FOH KHDWLQJ F\FOH JULG SRLQWV LQ YHUWLFDO GLUHFWLRQ JULG SRLQWV LQ UDGLDO GLUHFWLRQ JULG SRLQW RQ YHUWLFDO ERXQGDU\ ] +f JULG SRLQW RQ UDGLDO ERXQGDU\ U 5f :DOO L[

PAGE 10

6XSHUVFULSWV G GHVLJQ YDOXH I QRUPDO WR JULG IDFHV N LWHUDWLRQ FRXQWV LQ 625 P LWHUDWLRQ FRXQW ZLWKLQ D WLPH VWHS Q WLPH VWHS W WDQJHQWLDO WR JULG IDFHV r KDOI VWHS LQ $', PHWKRG VDPH DV QOf [

PAGE 11

/,67 2) 7$%/(6 7$%/( 3$*( 6HTXHQFH RI OLQH W\SHV IRU LGHQWLILFDWLRQ RI VWUHDP IXQFWLRQ FRQWRXUV LQ ILJXUHV O2DR +HDWLQJ DQG FRROLQJ UDWHV IRU YDULRXV SURFHVV FRQGLWLRQV $GMXVWHG KHDWLQJ WLPHV DQG UHVXOWLQJ OHWKDOLW\ )f LQ UHVSRQVH WR SURFHVV GHYLDWLRQV XVLQJ SURSRVHG RQn OLQH FRQWURO ORJLF DQG PHWKRG RI *LDQQRQL6XFFDU DQG +D\DNDZD [L

PAGE 12

/,67 2) ),*85(6 ),*85( 3$*( ,VRWKHUPV DQG YHORFLW\ SURILOHV LQ D JODVV ERWWOH DW YDULRXV KHDWLQJ WLPHV DIWHU (QJHOPDQ DQG 6DQL f ,VRWKHUPV DQG YHORFLW\ SURILOHV LQ D FDQ DW YDULRXV KHDWLQJ WLPHV DIWHU 6DQL f 7KH JULG V\VWHP DQG WKH ERXQGDU\ FRQGLWLRQV IRU WKH F\OLQGULFDO FDQ 1RWDWLRQV RI YDULRXV TXDQWLWLHV GHILQHG RQ D QRQ XQLIRUP JULG V\VWHP LQ F\OLQGULFDO JHRPHWU\ $OJRULWKP IRU WKH LWHUDWLYH VROXWLRQ RI WKH VHW RI HTXDWLRQV LQ QDWXUDO FRQYHFWLRQ ([SHULPHQWDO VHWXS IRU FRQGXFWLRQ KHDWLQJ 7LPLQJV RI YDULRXV UHDOWLPH FRPSXWDWLRQV IRU RQOLQH FRQWURO RI FRQGXFWLRQKHDWHG IRRG )ORZ GLDJUDP IRU FRPSXWHU FRQWURO RI UHWRUW RSHUDn WLRQV ZLWK RQOLQH FRUUHFWLRQ RI SURFHVV GHYLDWLRQV ,VRWKHUPV QHDU ERWWRP ZDOO SUHGLFWHG E\ WKH FRQYHFWLYH PRGHO FRPSDUHG ZLWK WKH SUHGLFWLRQV FRQVLGHULQJ FRQGXFWLRQ HTXDWLRQ f RQO\ D ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQG RI KHDWLQJ E ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ F ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2G ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2H ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ [L L

PAGE 13

O2I ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2J ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2K ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2L ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2M ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ N ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ P ,VRWKHUP DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ O2Q ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ R ,VRWKHUPV DQG YHORFLW\ YHFWRUV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ 3UHGLFWHG UDGLDO YHORFLW\ SURILOH DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ 3UHGLFWHG UDGLDO WHPSHUDWXUH SURILOH DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ ,VRWKHUPDO YHUWLFDO VXUIDFH VKRZLQJ ERXQGDU\ OD\HU ZLWK YHORFLW\ SURILOH X[\f DQG WHPSHUDWXUH SURILOH W[\f LQ ZDWHU +LJJLQV DQG *HEKDUW f 3UHGLFWHG UDGLDO YHORFLW\ SURILOHV DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU YDULRXV KHDWLQJ WLPHV ,OO 2EVHUYHG UDGLDO YHORFLW\ SURILOHV DW WKH VLGHZDOO DW PLG KHLJKW LQ ZDWHU LQ D F\OLQGULFDO FDQ DIWHU YDULRXV KHDWLQJ WLPHV IURP +LGGLQN f &RPSDULVRQ RI QXPHULFDOO\ SUHGLFWHG D[LDO WHPSHUDWXUH SURILOHV ZLWK REVHUYHG +LGGLQN f YDOXHV DW YDULRXV WLPHV GXULQJ KHDWLQJ [L L L

PAGE 14

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n DVWHG ZLWK WKRVH GXULQJ FRQGXFWLRQ KHDWLQJ LQ D F\OLQGULFDO FDQ 'LDJUDP RI f[f JULG RQ WKH OHIW DQG f[f RQ WKH ULJKW ,VRWKHUPV DIWHU VHFRQGV RI KHDWLQJ FRPSXWHG XVLQJ [ JULG OHIW KDOI RI WKH ILJXUHf FRPSDUHG ZLWK FRPSXWDWLRQV XVLQJ [ JULG ULJKW KDOI RI WKH ILJXUHf 6WUHDPOLQHV DIWHU VHFRQGV RI KHDWLQJ FRPSXWHG XVLQJ [ JULG OHIW KDOI RI WKH ILJXUHf FRPSDUHG ZLWK FRPSXWDWLRQV XVLQJ [ JULG ULJKW KDOI RI WKH ILJXUHf 7UDQVLHQW VORZHVW KHDWLQJ WHPSHUDWXUHV FRPSXWHG ZLWK [ DQG [ JULG 7UDQVLHQW WHPSHUDWXUH DW OUG KHLJKW RQ WKH FHQWHU OLQH FRPSXWHG ZLWK [ DQG [ JULG 7LPH VWHS KLVWRU\ GXULQJ D W\SLFDO UXQ RI WKH QXPHULFDO KHDW WUDQVIHU PRGHO IRU QDWXUDO FRQYHFWLRQ ,VRWKHUPV LQ D F\OLQGULFDO FDQ FDOFXODWHG ZLWK WZR GLIIHUHQW FRQYHUJHQFH FULWHULD IRU ERXQGDU\ YRUWLFLW\ RQ WKH OHIW DQG RQ WKH ULJKWf 6WUHDPOLQHV LQ D F\OLQGULFDO FDQ FDOFXODWHG ZLWK WZR GLIIHUHQW FRQYHUJHQFH FULWHULD IRU ERXQGDU\ YRUWLFLW\ RQ WKH OHIW DQG RQ WKH ULJKWf [L Y

PAGE 15

7UDQVLHQW FHQWHU WHPSHUDWXUH GXULQJ FRQGXFWLRQn KHDWLQJ RI %HQWRQLWH LQ D F\OLQGULFDO FDQ 7UDQVLHQW FHQWHU WHPSHUDWXUH GXULQJ FRQGXFWLRQn FRROLQJ RI %HQWRQLWH LQ D F\OLQGULFDO FDQ ZLWK DQG ZLWKRXW GURS LQ UHWRUW SUHVVXUH &RPSXWHU SORW RI UHIHUHQFH SURFHVV QR GHYLDWLRQVf VKRZLQJ PLQXWHV RI KHDWLQJ WLPH DQG DQ DFFRPSOLVKHG ) RI &RPSXWHU SORW RI D SURFHVV WKDW H[SHULHQFHG D VWHS IXQFWLRQDO GURS RI r& r)f LQ WHPSHUDWXUH EHWn ZHHQ WK DQG WK PLQXWH DQG VWLOO PDLQWDLQHG DQ )B RI &RPSXWHU SORW RI D SURFHVV WKDW H[SHULHQFHG D OLQHDU GURS LQ WHPSHUDWXUH EHWZHHQ WK DQG WK PLQXWH IURP r& WR r& r)ff DQG VWLOO PDLQWDLQHG DQ ) RI [Y

PAGE 16

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f 7KH OLTXLG ZDWHUf ZDV LQLWLDOO\ VWDJQDQW DW D XQLIRUP WHPSHUDWXUH RI r& DQG WKH VLGHZDOO WRS DQG ERWWRP ZDOO WHPSHUDWXUHV ZHUH VXGGHQO\ UDLVHG WR D WHPSHUDWXUH RI r& $ IUHH WKHUPDOO\ LQVXODWHGf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

PAGE 17

PHWKRG *ULG VWUHWFKLQJ ZDV XVHG LQ ERWK UDGLDO DQG D[LDO GLUHFWLRQV %RXQGDU\ ZDOO YRUWLFLWLHV ZHUH XVHG WR FKHFN IRU FRQYHUJHQFH RI WKH LWHUDWLRQ SURFHVV 3ORWV RI WUDQVLHQW LVRWKHUPV VWUHDPOLQHV DQG YHORFLWLHV ZHUH SURYLGHG &DOFXODWHG UDGLDO YHORFLWLHV DQG D[LDO WHPSHUDWXUH SURILOHV DW GLIIHUHQW WLPHV FRPSDUHG TXLWH ZHOO ZLWK DYDLODEOH H[SHULPHQWDO GDWD 6ORZHVW KHDWLQJ SRLQWV ZHUH ORFDWHG LQ WKH ERWWRP b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

PAGE 18

,1752'8&7,21 %DFNJURXQG DQG -XVWLILFDWLRQ 7KHUPDO SUHVHUYDWLRQ RI IRRG LQYROYHV DSSOLFDWLRQ RI KHDW VXIILFLHQW WR GHVWUR\ WKH PLFURRUJDQLVPV SUHVHQW LQ IRRG WKDW FDXVH VSRLODJH &RPPHUFLDO FDQQLQJ SURFHVVHV SURYLGH VXFK KHDW WUHDWPHQW WR H[WHQG WKH VKHOIOLIH RI IRRGV ,Q D W\SLFDO FDQQLQJ SURFHVV WKH FRQWDLQHU LV ILOOHG ZLWK IRRG DQG VHDOHG 7KH VHDOHG FRQWDLQHU LV KHDWHG LQ D FORVHG YHVVHO ZLWK VWHDP RU KRW ZDWHU ORQJ HQRXJK WR NLOO WKH PLFURRUJDQLVPV DQG LV WKHQ FRROHG ,Q D SURGXFWLRQ VLWXDn WLRQ WKH KHDWLQJ PHGLXP VWHDP RU KRW ZDWHUf WHPSHUDWXUH FDQ VRPHWLPHV GHYLDWH VLJQLILFDQWO\ IURP GHVLJQ YDOXHV GXULQJ WKH SURFHVV +HDWLQJ WKH SURGXFW PRUH WKDQ WKH UHTXLUHG DPRXQW ZRXOG XQn QHFHVVDULO\ GHJUDGH LWV TXDOLW\ ZKLOH ZDVWLQJ HQHUJ\ +RZHYHU LQVXIILFLHQW KHDWLQJ FRXOG OHDG WR VHULRXV SXEOLF KHDOWK KD]DUGV DQG FDQQRW EH DOORZHG 7KXV IRRG SURFHVVRUV DOZD\V KDYH WR PHHW WKLV PLQLPXP KHDW WUHDWPHQW VWHULOL]DWLRQf 'HWHUPLQDWLRQ RI WKH KHDWLQJ WLPH WR PHHW WKH PLQLPXP VWHULOL]DWLRQ GHPDQGV WKDW WKH DFWXDO WUDQVLHQW WHPSHUDWXUH KLVWRU\ RI WKH IRRG GXULQJ SURFHVVLQJ EH NQRZQ SUHGLFWHG ZLWK D KHDW WUDQVIHU PRGHO RU DFWXDOO\ PHDVXUHGf

PAGE 19

&RQGXFWLRQ+HDWHG )RRGV 7UDGLWLRQDOO\ IRU FRQGXFWLRQKHDWHG IRRGV WKH UHTXLUHG SURFHVVLQJ WLPH LV GHWHUPLQHG XVLQJ WKH DQDO\WLFDO VROXWLRQ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ ,Q SUDFWLFH WKH FRQVWDQW PHGLXP WHPSHUn DWXUH DVVXPHG E\ WKH DQDO\WLFDO VROXWLRQ FDQQRW DOZD\V EH DFKLHYHG 7KXV H[SHULPHQWDO FRUUHFWLRQ IDFWRUV KDYH EHHQ SURSRVHG WR DGMXVW WKH KHDWLQJ WLPH SUHGLFWHG E\ DQDO\WLFDO VROXWLRQ ZKHQ WKH PHGLXP WHPSHUDWXUH JRHV WKURXJK FHUWDLQ SUHGHILQHGf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

PAGE 20

&RQYHFWLRQ+HDWHG )RRGV 8QOLNH VROLG SDFNHG IRRGV WKDW DUH KHDWHG E\ FRQGXFWLRQ WKLQ OLTXLG IRRGV KHDW WKURXJK FRQYHFWLRQ ,Q JHQHUDO VXFK OLTXLG IRRGV DUH PHFKDQLFDOO\ DJLWDWHG WR LPSURYH WKH KHDW WUDQVIHU +RZHYHU WKHUH DUH VLWXDWLRQV ZKHUH WKH IRRG PDWHULDO FDQQRW EH DJLWDWHG IRU VHYHUDO UHDVRQV 7KH IRRG WKHQ KHDWV SULPDULO\ E\ QDWXUDO FRQYHFn WLRQ ,QSDFNDJH SDVWHXUL]DWLRQ IRU OLTXLGV HJ EHHUf LQ ERWWOHV RU FDQV WKDW FDQQRW EH DJLWDWHG LV DQ H[DPSOH %URWK WKLQ VRXSV HYDSRUDWHG PLON PRVW IUXLW DQG YHJHWDEOH MXLFHV IUXLWV LQ V\UXS RU ZDWHU DQG SXUHHG YHJHWDEOHV DUH DOVR H[DPSOHV RI SURGXFWV ZKHUH QDWXUDO FRQYHFWLRQ KHDWLQJ FDQ RFFXU ZKHQ QR DJLWDWLRQ LV XVHG )OH[LEOH SRXFKHV WKDW FDQQRW EH DJLWDWHG WR NHHS WKH SDFNDJH LQWHJn ULW\ DUH DQRWKHU LPSRUWDQW H[DPSOH )HUPHQWDWLRQ LV DQRWKHU SURFHVV ZKHUH D QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU PRGHO ZLWK KHDW JHQHUDWLRQ WHUP LQFOXGHGf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

PAGE 21

6XFK D QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU PRGHO FRXOG DOVR KDYH VHYHUDO DSSOLFDWLRQV RXWVLGH WKH IRRG DUHD ,W PD\ EH XVHG WR VWXG\ WKH KHDWLQJ RI EXLOGLQJV ZLWKRXW IRUFHG FLUFXODWLRQf FRROLQJ RI JDV WXUELQH EODGHV VWRUDJH RI FU\RJHQLF IOXLGV HJ OLTXLG URFNHW SURSHOODQWVf DQG WKH VWDUWXS RI FKHPLFDO UHDFWRUV 6LQFH QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU ZRXOG LQYROYH D PXFK KLJKHU GHJUHH RI PDWKn HPDWLFDO FRPSOH[LW\ WKDQ D FRQGXFWLRQ KHDWLQJ PRGHO DQG LWV DSSOLn FDWLRQ WR RQOLQH FRQWUROf WKH SULPDU\ REMHFWLYH RI WKLV VWXG\ ZDV GLUHFWHG WRZDUG GHYHORSLQJ D KHDW WUDQVIHU PRGHO IRU QDWXUDO FRQYHFn WLRQ IROORZHG E\ LPSURYHPHQWV WR H[LVWLQJ FRQGXFWLRQKHDW WUDQVIHU PRGHOV IRU DSSOLFDWLRQ WR RQOLQH SURFHVV FRQWURO DV VHFRQGDU\ REMHFWLYHV 2EMHFWLYHV 7KH REMHFWLYHV RI WKLV VWXG\ ZHUH WR GHYHORS D PDWKHPDWLFDO PRGHO WR SUHGLFW WHPSHUDWXUH DQG YHORFLW\ SURILOHV FDXVHG E\ QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU LQ D FORVHG F\OLQGHU LQYHVWLJDWH WKH QHHG IRU SRVVLEOH LPSURYHPHQWV RQ DQ H[LVWLQJ FRQGXFWLRQ KHDW WUDQVIHU PRGHO WR LPSURYH LWV VXLWDELOLW\ IRU XVH LQ RQOLQH SURFHVV FRQWURO DSSOLn FDWLRQV DQG LQFRUSRUDWH WKH FRQGXFWLRQ KHDWLQJ PRGHO LQ DQ RQOLQH SURFHVV FRQWURO DOJRULWKP IRU WKHUPDO SURFHVVLQJ RI FDQQHG IRRG DQG FRPSDUH LWV SHUIRUPDQFH DJDLQVW RWKHU SRVVLEOH PHWKRGV RI RQOLQH FRQWURO

PAGE 22

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n GXFWLRQKHDWHG IRRGV WKDW FRXOG DOVR EH XVHG RQOLQH DUH GLVFXVVHG 2YHUYLHZ RI 7KHUPDO 3URFHVVLQJ 7KHUPDO SURFHVVLQJ FRQVLVWV RI KHDWLQJ IRRGV ILOOHG DQG KHUPDWLFDOO\ VHDOHG LQ FRQWDLQHUV LQ SUHVVXUL]HG VWHDP UHWRUWV DW D

PAGE 23

FRQVWDQW UHWRUW WHPSHUDWXUH IRU D SUHVFULEHG OHQJWK RI WLPH IROORZHG E\ UDSLG FRROLQJ 7KHVH SURFHVV WLPHV DUH FDOFXODWHG RQ WKH EDVLV RI VXIILFLHQW EDFWHULDO LQDFWLYDWLRQ LQ HDFK FRQWDLQHU WR FRPSO\ ZLWK SXEOLF KHDOWK VWDQGDUGV DQG UHGXFH WKH SUREDELOLW\ RI VSRLODJH WR D VSHFLILHG PLQLPXP (VWLPDWLRQ RI EDFWHULDO LQDFWLYDWLRQ UHTXLUHV XQGHUVWDQGLQJ RI WKHUPDO GHDWK NLQHWLFV RI EDFWHULD DORQJ ZLWK WKH WHPSHUDWXUH KLVWRU\ WKH IRRG FRQWDLQLQJ WKH EDFWHULDf H[SHULHQFHV GXULQJ WKH SURFHVV )ROORZLQJ LV D EULHI LQWURGXFWLRQ WR WKH WKHUPDO GHDWK NLQHWLFV RI EDFWHULD %DOO t 2OVRQ 6WXPER f 7KHUPDO GHDWK RI EDFWHULD IROORZV ILUVW RUGHU NLQHWLFV ,I & LV EDFWHULDO FRQFHQWUDWLRQ QXPEHU RI EDFWHULD SHU XQLW YROXPH RI IRRG PDWHULDOf DW WLPH W G& B GW f n f'O f ZKHUH fr' LV WKH LQYHUVH RI WKH UDWH FRQVWDQW ,I &Q LV WKH LQLWLDO FRQFHQWUDWLRQ DQG & LV FRQFHQWUDWLRQ DIWHU WLPH $W ORJ $W f 7KH FRQVWDQW NQRZQ DV YDOXHf ZKLFK LV WKH LQYHUVH RI WKH UDWH FRQVWDQW PXOWLSOLHG E\ WKH IDFWRU f LV WKH WLPH UHTXLUHG WR UHGXFH WKH EDFWHULDO SRSXODWLRQ E\ RQH ORJ F\FOH DW D VSHFLILHG WHPSHUDWXUH 7KH WHPSHUDWXUH GHSHQGHQF\ RI WKH YDOXH LV JLYHQ E\

PAGE 24

R f ZKHUH LV WKH YDOXH DW VRPH UHIHUHQFH WHPSHUDWXUH 7 7KH = YDOXH GHVFULEHV WKH WHPSHUDWXUH GHSHQGHQF\ RI WKH OHWKDO UDWH DQG LV UHODWHG WR WKH DFWLYDWLRQ HQHUJ\ 7KH UHIHUHQFH WHPSHUDWXUH 7 LV JHQHUDOO\ WDNHQ DV r& r)f )URP EDFWHULRORJLFDO FRQVLGHUDWLRQV D WKHUPDO SURFHVV LV VSHFLILHG E\ QRWLQJ WKH WLPH )f QHFHVVDU\ WR UHGXFH WKH EDFWHULDO SRSXODWLRQ IURP WKH LQLWLDO &Af WR VRPH ILQDO VDIH YDOXH &f DW WHPSHUDWXUH r& r)f 7KXV ) R f :KHQ WKH KHDWLQJ SURFHVV LV QRW DW WKH FRQVWDQW WHPSHUDWXUH RI r& LQWURGXFLQJ WKH WHPSHUDWXUH GHSHQGHQF\ RI 4 DQG HTXDWLRQ f IRU EDFWHULDO GHDWK ) R 7 = $W 7 $W f $W LV WKH WLPH UHTXLUHG DW WHPSHUDWXUH 7 WR DFKLHYH WKH GHVLJQHG EDFWHULDO UHGXFWLRQ $V WKH IRRG PDWHULDO LV KHDWHG LWV WHPSHUDWXUH 7 EHFRPHV D FRQWLQRXV IXQFWLRQ RI WLPH W DQG WKH H[SUHVVLRQ IRU ) HTXDWLRQ f EHFRPHV

PAGE 25

7 )R / GW f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f RI r& :KHQ WHPSHUDWXUH 7 LV D IXQFWLRQ RI WLPH 7Wff LQVWHDG WKH REMHFWLYH RI WKHUPDO SURFHVVLQJ LV WR VHOHFW SURFHVV WLPH W IRU WKH SDUWLFXODU KHDWLQJ FRQGLWLRQ VXFK WKDW WKH HTXLYDOHQW )4 UHFHLYHG DV JLYHQ E\ HTQ f LV JUHDWHU WKDQ RU HTXDO WR WKH GHVLJQ YDOXH )A 7KDW LV )R\ )R f $SSOLFDWLRQ WR FRQGXFWLRQKHDWLQJ :KHQ DSSO\LQJ WKHUPDO SURFHVVLQJ WR FRQGXFWLRQKHDWHG IRRG LQ D F\OLQGULFDO FRQWDLQHU %DOO DQG 2OVRQ f XVHG WKH VROXWLRQ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ LQ F\OLQGULFDO FRRUGLQDWHV 77 ? ;fDW f§ ( 6 $ H P fS -[Uf &RV [B]f Df 7 B PS XY P n S 775 3

PAGE 26

7KH VROXWLRQ ZDV ILUVW DSSUR[LPDWHG WR 9 WU 8 ;f DW $ H f -[Uf &RV[]f Ef ZKHQ VXIILFLHQW WLPH KDV HODSVHG VR WKDW WKH ILUVW WHUP LQ WKH DQDO\WLFDO VHULHV VROXWLRQ HTXDWLRQ Df GRPLQDWHV (TXDWLRQ Ef ZDV WKHQ UHDUUDQJHG WR JLYH W I ORJ 9 9 9 7 f ZKHUH DQG B D ;] ;f f M U]f $ -[Uf &RV []f f )URP WKH WUDQVLHQW KHDW FRQGXFWLRQ HTXDWLRQ f ZKLFK JLYHV WHPSHUDWXUH 7 DV D IXQFWLRQ RI WLPH W DQG WKH GHILQLWLRQ RI )4 HTXDWLRQ f ZKLFK JLYHV ) DV D IXQFWLRQ RI WHPSHUDWXUH 7 LW LV SRVVLEOH WR ILQG ) DV D IXQFWLRQ RI WLPH W +RZHYHU WKLV FDQQRW EH SHUIRUPHG DQDO\WLFDOO\ 8SRQ VROYLQJ WKH WZR HTXDWLRQV QXPHULFDOO\ LW LV SRVVLEOH WR NQRZ WHPSHUDWXUH DFKLHYHG DW WKH HQG RI KHDWLQJ IRU D JLYHQ YDOXH RI ) %DOO DQG 2OVRQ f WDEXODWHG 7A7U YDOXHV DV D IXQFWLRQ RI )4 IRU SDUWLFXODU YDOXHV RI I DQG M 7UDGLn WLRQDOO\ LQ WKHUPDO SURFHVVLQJ RI FRQGXFWLRQKHDWHG IRRGV LQ F\OLQn GULFDO FRQWDLQHUV WKHVH WDEOHV DUH XVHG WR ILQG 7A7U IRU D JLYHQ ) YDOXH 8VLQJ WKLV 7K75f YDOXH DQG HTXDWLRQ f WKH SURFHVV WLPH W LV IRXQG

PAGE 27

2QOLQH &RQWURO RI 7KHUPDO 3URFHVVLQJ ,Q D SURGXFWLRQ VLWXDWLRQ WKH KHDWLQJ WLPH W VSHFLILHG DVVXPLQJ D FRQVWDQW YDOXH RI 75f WR DFKLHYH D JLYHQ ) YDOXH ZRXOG QHHG WR EH DGMXVWHG IRU DUELWUDU\ YDULDWLRQV LQ WKH KHDWLQJ PHGLXP WHPSHUDWXUH 75 IURP LWV VSHFLILHG FRQVWDQW YDOXH 5HVHDUFKHUV LQ WKH SDVW KDYH DWWHPSWHG WKLV LQ RQH RI WKH IROORZLQJ WKUHH ZD\V 3UHFDOFXODWHG WDEXODWLRQV &RUUHFWLRQ IDFWRUV DUH WDEXODWHG WR EH DSSOLHG WR WA IRU FHUWDLQ SRVVLEOH W\SHV RI GHYLDWLRQV LQ 75Wf &RPPHUFLDO UHWRUW FRQWURO V\VWHPV RIWHQ XVH WKLV WHFKQLTXH /2*7(& V\VWHP f 6XFK WDEXODWLRQV FDQQRW EH H[KDXVWLYH VLQFH WKH SURFHVV WHPSHUDWXUH 75Wf FRXOG JR WKURXJK DQ\ UDQGRP GHYLDWLRQ ,QVWHDG RI WDEXODWLQJ VXFK FRUUHFWLRQ IDFWRUV *LDQQRQL 6XFFDU DQG +D\DNDZD f GHYHORSHG H[SUHVVLRQV IRU FRUUHFWLRQ IDFWRUV WR EH DSSOLHG WR WK ZKHQ 75Wf JRHV WKURXJK D VWHS GURS ,Q WKHLU ZRUN RWKHU W\SHV RI GHYLDWLRQV KDG WR EH DSSUR[LPDWHG WR FORVHILWWLQJ VWHS IXQFWLRQV DQG PXOWLSOH GURSV FRXOG QRW EH FRQVLGHUHG *HWFKHOO f GHVFULEHG D FRQWURO V\VWHP WKDW ZRXOG DWWHPSW WR PDLQWDLQ WKH GHVLJQ KHDWLQJ PHGLXP WHPSHUDWXUH DQG VRXQG DQ DODUP ZKHQ FULWLFDO ORZ RU KLJK OLPLWV IRU WKH WHPSHUDWXUH DUH H[FHHGHG 7KH RSHUDWRU ZRXOG WKHQ DWWHPSW WR PDNH QHFHVVDU\ FKDQJHV LQ WK 'LUHFW 0HDVXUHPHQW 7HPSHUDWXUH DW WKH FDQ FHQWHU 7Wf LV PHDVXUHG RQOLQH 7KLV HOLPLQDWHV WKH QHHG WR DVVXPH FRQVWDQW KHDWLQJ WHPSHUDWXUH 0HDVXUHG 7Wf LV XVHG ZLWK HTXDWLRQ WR VWRS WKH KHDWLQJ F\FOH DW WLPH WA VXFK WKDW : )R f

PAGE 28

7KLV DSSURDFK ZDV XVHG E\ 0XOYDQH\ DQG 5L]Y\ f 7KH\ XVHG D WKHUPRFRXSOHILWWHG WHVW FDQ LQ HYHU\ UHWRUW EDWFK WR REWDLQ DFWXDO WHPSHUDWXUH 7Wf 7KH V\VWHP SURSRVHG E\ 1DYDQNDVDWWXVDV DQG /XQG f DOVR SODQQHG WR PHDVXUH FDQ FHQWHU WHPSHUDWXUH RQOLQH ,Q FRPPHUFLDO SUDFWLFH WKHVH PHWKRGV DUH FRVWSURKLELWLYH ZLWK UHJDUG WR SURGXFWLRQ HIILFLHQF\ DQG DUH FRQVLGHUHG LPSUDFWLFDO 1XPHULFDO 0RGHOV ,QVWHDG RI PHDVXULQJ 7Wf DW WKH FDQ FHQWHU 7Wf FDQ EH SUHGLFWHG IRU DUELWUDU\ YDULDWLRQV LQ 75Wf ,Q JHQHUDO DQDO\WLFDO VROXWLRQV IRU 7Wf DUH RI OLPLWHG XVH IRU DUELWDU\ YDULDWLRQV RI 75Wf 7HL[HLUD DQG 0DQVRQ f GHVFULEHG D QXPHULFDO ILQLWH GLIIHUHQFH DSSUR[LPDWLRQ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ IRU SUHGLFWLQJ 7Wf IURP 75Wf 7KLV LV FRQFHSWXDOO\ TXLWH GLIIHUHQW IURP WKH DSSURDFKHV PHQWLRQHG HDUOLHU :LWK D ILQLWH GLIIHUHQFH PRGHO 7Wf FDQ EH SUHGLFWHG IRU WUXO\ DUELWUDU\ YDULDWLRQV LQ 7AWf DQG RI FRXUVH XVH RI WKLV PRGHO HOLPLQDWHV WKH QHHG IRU KDYLQJ D WHVW FDQ ZLWK DQ DFWXDO WKHUPRFRXSOH LQVHUWHG LQ LW 7KLV PRGHO ZDV ILUVW GHVFULEHG DQG YHULILHG DJDLQVW SXEOLVKHG SURFHGXUHV E\ 7HL[HLUD HW DO DQG ODWHU YHULILHG DJDLQVW H[SHULPHQWDO GDWD DQG RWKHU SXEOLVKHG SURFHGXUHV IRU WLPHYDU\LQJ ERXQGDU\ FRQGLWLRQV E\ 7HL[HLUD HW DO DEf +RZHYHU LQ WKH SURSRVHG UHWRUW FRQWURO DOJRULWKP RI 7HL[HLUD DQG 0DQVRQ f HYHU\ WLPH D GHYLDWLRQ WRRN SODFH D QHZ YDOXH RI WIW ZDV IRXQG WKURXJK VLPXODWLRQ DVVXPLQJ WKH GHYLDWHG UHWRUW WHPSHUn DWXUH FRQWLQXHG IRU WKH UHVW RI WKH KHDWLQJ SURFHVV 7KLV LV XQQHFn HVVDU\ EHFDXVH DQ\ IXUWKHU FKDQJHV LQ WKH UHWRUW WHPSHUDWXUH ZRXOG PDNH WKH HVWLPDWHG WA YDOXH XVHOHVV DQG IUHVK HVWLPDWLRQV RI WA

PAGE 29

ZRXOG UHTXLUH UHSHDWHG WLPHFRQVXPLQJ FRPSXWHU VLPXODWLRQV 8VLQJ WKH JHQHUDO LGHD RI 7HL[HLUD DQG +DQVRQ f WKH IHDVLELOLW\ RI VHYHUDO FRQWURO DOJRULWKPV WKDW FDQ ZRUN RQOLQH ZLWK DYDLODEOH FRPSXWHU KDUGZDUH QHHGHG WR EH LQYHVWLJDWHG DQG OHG WR WKH IROORZLQJ VWXGLHV DV SDUW RI WKH ZRUN FDUULHG RXW LQ WKLV SURMHFW f :D\V WR REWDLQ WKH WUDQVLHQW WHPSHUDWXUH VROXWLRQ IDVWHU XVLQJ RWKHU DYDLODEOH PHWKRGV RI VROYLQJ WKH FRQGXFWLRQ HTXDWLRQ f 7KH QHHG WR LQFRUSRUDWH D FRQYHFWLRQ 5RELQnVf ERXQGDU\ FRQGLWLRQ WR UHIOHFW WKH WUXH VLWXDWLRQ UDWKHU WKDQ WKH WHPSHUDWXUH VSHFLILHG 'LULFKOHWf ERXQGDU\ FRQGLWLRQ XVHG E\ 7HL[HLUD HW DO DEf LQ VROYLQJ WKH FRQGXFWLRQ HTXDWLRQ &RQGXFWLRQ +HDWLQJ RI &DQQHG )RRGV 6LQFH WKH SLRQHHULQJ ZRUN RI %DOO DQG 2OVRQ f PHQWLRQHG HDUOLHU FRQGXFWLRQ KHDWLQJ RI FDQQHG IRRGV KDV EHHQ VWXGLHG H[WHQVLYHO\ &RQVHTXHQWO\ WKHUH LV D YDVW DPRXQW RI OLWHUDWXUH H[LVWLQJ RQ WKH VXEMHFW 6WXGLHV %DVHG RQ $QDO\WLFDO 6ROXWLRQ 7KH PDMRULW\ RI WKH ZRUNV RQ FRQGXFWLRQKHDWLQJ RI FDQQHG IRRGV ZHUH EDVHG RQ RQH RI VHYHUDO GLIIHUHQW IRUPV RI WKH DQDO\WLFDO VROXn WLRQ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ $OVR PRVW RI WKHVH VROXWLRQV

PAGE 30

ZHUH GHULYHG DVVXPLQJ DPRQJ RWKHU WKLQJV D VWHS FKDQJH LQ WKH ERXQGDU\ FDQ ZDLOf WHPSHUDWXUH IURP WKH LQLWLDO WR WKH KHDWLQJ WHPSHUDWXUH DQG PDLQWDLQLQJ WKH ERXQGDU\ KHDWLQJ WHPSHUDWXUH FRQVWDQW WKURXJKRXW WKH SURFHVV 0RUH JHQHUDO IRUPV RI ERXQGDU\ FRQGLWLRQV KDYH DOVR EHHQ IRUPXODWHG +D\DNDZD f GHYHORSHG WKH JHQHUDO IRUP IRU DQ DQDO\WLFDO VROXWLRQ WKDW FRXOG SUHGLFW WKH WUDQVLHQW WHPSHUDWXUH GLVWULEXWLRQ LQVLGH D F\OLQGULFDO FDQ RI FRQGXFWLRQKHDWHG IRRG ZKHQ WKH ERXQGDU\ FDQ ZDOOf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f ZHUH WKH ILUVW WR XVH D IRUZDUG GLIIHUHQFH H[SOLFLW ILQLWH GLIIHUHQFLQJ PHWKRG WR VWXG\ WKH WUDQVLHQW WHPSHUDWXUH GLVWULEXWLRQ LQVLGH WKH F\OLQGULFDO FDQ 8VH RI WKH H[SOLFLW ILQLWH GLIIHUHQFLQJ VHYHUHO\ OLPLWHG WKH FKRLFH RI JULG VL]H DQG WLPH LQFUHPHQW GXH WR VWDELOLW\ UHTXLUHPHQWV IRU VXFK D

PAGE 31

PHWKRG ,Q WKHLU ODWHU ZRUN 7HL[HLUD HW DO Df DQ DOWHUQDWLQJ GLUHFWLRQ H[SOLFLW $'(f PHWKRG 6DXOn\HY $OLDGD DQG 4XRQ f ZDV XVHG WKDW ZDV XQFRQGLWLRQDOO\ VWDEOH EXW VWLOO H[SOLFLW 7KLV PHWKRG ZDV ILUVW DSSOLHG WR VLPXODWH WKHUPDO SURFHVVLQJ RI FDQQHG IRRGV E\ 0DQVRQ HW DO f 7HL[HLUD HW DO DEf FRPSDUHG RXWSXWV IURP WKH $'( PHWKRG IRU XVLQJ FRQVWDQW ERXQGDU\ WHPSHUDWXUH DQG VLQXVRLGDO WLPHYDU\LQJ ERXQGDU\ WHPSHUDWXUH ZLWK WKH UHVXOWV IURP FRUUHVSRQGLQJ DQDO\WLFDO VROXWLRQV 7KH DQDO\WLFDO UHVXOWV IRU FRQVWDQW ERXQGDU\ WHPSHUDWXUH DJUHHG ZLWKLQ r& r)f ZLWK WKH UHVXOWV IURP WKH $'( PHWKRG H[FHSW DW LQWHULRU SRLQWV QHDU WKH VXUIDFH DIWHU VPDOO LQLWLDO WLPH OHVV WKDQ PLQXWHVf 6LPLODU DJUHHPHQW EHWZHHQ WKH UHVXOWV RI WKH DQDO\WLFDO DQG WKH $'( IRUPXODWLRQ ZDV DOVR REWDLQHG ZKHQ D VLQXVRLGDOO\ YDU\LQJ ERXQGDU\ FRQGLWLRQ ZDV XVHG 7HL[HLUD f DOVR FRPSDUHG WKH $'( ILQLWH GLIIHUHQFH RXWSXWV IRU FRQGXFWLRQ KHDWLQJ LQ D FDQ ZLWK H[SHULPHQWDO GDWD IRU FRQVWDQW ERXQGDU\ WHPSHUDWXUH DQG IRU VWHS LQFUHDVHV LQ WKH ERXQGDU\ WHPSHUDn WXUH GXULQJ KHDWLQJ 7KH DJUHHPHQW ZDV JRRG H[FHSW DW LQLWLDO WLPHV ,Q DGGLWLRQ WR WKLV $'( PHWKRG RWKHU PHWKRGV RI VROXWLRQ WR WKH FRQGXFWLRQ HTXDWLRQ QHHGHG WR EH LQYHVWLJDWHG IURP WKH SRLQW RI YLHZ RI WKHLU VXLWDELOLW\ RI FRPSXWDWLRQ LQ UHDOWLPH DQG DELOLW\ WR DFFRPRGDWH DUELWUDU\ ERXQGDU\ WHPSHUDWXUH YDULDWLRQV 1DWXUDO &RQYHFWLRQ +HDWLQJ RI &DQQHG )RRGV ,QWHUQDO LQ HQFORVXUHVf QDWXUDO FRQYHFWLRQ IORZ SUREOHPV KDYH UHFHLYHG PXFK OHVV DWWHQWLRQ WKDQ WKH H[WHUQDO RQHV $OVR LQWHUQDO

PAGE 32

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f 3RVVLEOH IORZ SDWWHUQV ZHUH SRVWXODWHG E\ %ODLVGHOO f XVLQJ G\H DQG DOXPLQXP SRZGHU WUDFHUV 6WHYHQV f ILUVW WULHG WR VWXG\ GHWDLOHG WHPSHUDWXUH DQG IORZ SDWWHUQV GXULQJ QDWXUDO FRQYHFWLRQ KHDWLQJ RI FDQQHG OLTXLG IRRGV +H XVHG DQ H[SOLFLW ILQLWH GLIIHUHQFH IRUPXODWLRQ IURP WKH ZRUN RI 7RUUDQFH DQG 5RFNHWW f 7KH FDQ ZDV WUHDWHG DV D YHUn WLFDO F\OLQGULFDO HQFORVXUH 7KH IOXLG HWK\OHQH JO\FROf LQVLGH WKH FDQ ZDV LQLWLDOO\ DW UHVW DW D XQLIRUP WHPSHUDWXUH RI r& 7KH FDQ ZDV WKHQ SODFHG LQ KRW ZDWHU EDWK DW DERXW r& 7KH WRS VXUIDFH RI WKH IOXLG ZDV IUHH WKH FDQ KDG KHDGVSDFHf 7KH DFWXDO WHPSHUDWXUH

PAGE 33

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f 7KH QXPHULFDO VROXWLRQ WHFKQLTXH XVHG ZDV WKH H[SOLFLW ILQLWH GLIIHUHQFH WHFKQLTXH RI %DUDNDW DQG &ODUN f GLVFXVVHG ODWHU 7KH LQLWLDO WHPSHUDWXUH ZDV FRQVLGHUHG XQLIRUP WKURXJKRXW 7KH WRS ERWWRP DQG VLGHZDOO ZHUH VXGGHQO\ UDLVHG WR WKH KHDWLQJ WHPSHUDWXUH 7KH WRS VXUIDFH RI WKH OLTXLG ZDV FRQVLGHUHG IUHH KDYLQJ KHDGVSDFHf DQG ZDV WKXV WUHDWHG DV WKHUPDOO\ LQVXODWHG 6HYHUDO WHVW IOXLGV RI ORZ DQG KLJK YLVFRVLWLHV ZDWHU VXFURVH VROXWLRQ DQG VLOLFRQH IOXLG UHVSHFn WLYHO\f ZHUH XVHG 7KH QXPHULFDO UHVXOWV LQFOXGHG WHPSHUDWXUH DQG VWUHDPOLQH SDWWHUQV 7KH RYHUDOO IORZ SDWWHUQV ZHUH YLVXDOL]HG XVLQJ D SDUWLFOH VWUHDN PHWKRG ,Q WKLV PHWKRG VPDOO JOLWWHULQJ SDUWLFOHV DUH VXVSHQGHG LQ WKH OLTXLG 7KH SDUWLFOHV DUH LOOXPLQDWHG

PAGE 34

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t *ROGVWHLQ f KDG VKRZHG FRQYHFWLYH KHDW WUDQVIHU IURP WKH ERWWRP DV H[SHFWHG ZKHQ WKH ERWWRP ZDV DOVR KHDWHG +LV QXPHULFDO UHVXOWV KRZHYHU FRXOG QRW SLFN XS WKH FRQYHFWLYH HGGLHV JHQHUDWHG E\ WKH ERWWRP KHDWLQJ 7KXV KLV QXPHULFDO UHVXOWV KDG SRRU FRUUHn VSRQGHQFH ZLWK KLV H[SHULPHQWDO UHVXOWV $ ILQLWH HOHPHQW PHWKRG ZDV XVHG E\ (QJHOPDQ DQG 6DQL f WR VWXG\ LQSDFNDJH SDVWHXUL]DWLRQ RI IOXLGV EHHUf LQ ERWWOHV ,Q WKH SURFHVV VWXGLHG ERWWOHV HQWHUHG WKH SDVWHXUL]HU DW r&r)f DQG SDVVHG WKURXJK VHYHUDO SURJUHVVLYHO\ KRWWHU ]RQHV RI KRW ZDWHU VSUD\ ZKLFK UDLVHG WKH SURGXFW WHPSHUDWXUH WR r&r)f 7KLV WHPSHUDWXUH RI r& LV PDLQWDLQHG LQ WKH KROGLQJ ]RQH 7KH SURGXFW WKHQ SDVVHG WKURXJK VHYHUDO SURJUHVVLYHO\ FRROHU ]RQHV ZKLFK ORZn HUHG WKH WHPSHUDWXUH WR r r) 6LQFH WKH SURGXFW ZDV IOXLG DQG WKH SDFNDJH ZDV QRW DJLWDWHG LW ZDV SULPDULO\ KHDWHG WKURXJK QDWn XUDO FRQYHFWLRQ ([SHULPHQWDO GHWDLOV IRU WKLV VWXG\ ZDV SURYLGHG E\ %UDQGRQ HW DO f 7KH ERWWOH JHRPHWU\ ZDV FRPSOH[ ILJXUH f DQG VR ZHUH WKH ERXQGDU\ FRQGLWLRQV 7KH ERWWRP RI WKH ERWWOH ZDV

PAGE 35

6(&21'6 6(&21'6 )LJXUH ,VRWKHUPV DQG YHORFLW\ SURILOHV LQ D JODVV ERWWOH DW YDULRXV KHDWLQJ WLPHV DIWHU (QJHOPDQ DQG 6DQL f

PAGE 36

WUHDWHG DV LQVXODWHG DQG ZDOO WHPSHUDWXUH VSHFLILHG DV D IXQFWLRQ RI WLPH 7KH QXPHULFDO UHVXOWV ZHUH LQ JRRG DJUHHPHQW ZLWK H[SHULPHQWDO GDWD 6DQLr SHUVRQDO FRPPXQLFDWLRQ f DOVR GHYHORSHG D VLPLODU PRGHO IRU SDVWHXUL]DWLRQ RI EHHU LQ F\OLQGULFDO PHWDO FDQV DVSHFW UDWLR f :DWHU ZDV WKH WHVW IOXLG 3U f 7KH LQLWLDO WHPSHUDWXUH ZDV r& DQG WKH VLGHZDOO ERXQGDU\ FRQGLWLRQ XVHG ZDV DSSUR[LPDWHO\ r& 7KH WRS RI WKH ZDWHU OHYHO ZDV IUHH WKH FDQ KDG KHDGVSDFHf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f ZKR ZHUH ORRNLQJ DW WKHUPDO VWUDWLn ILFDWLRQ DQG DVVRFLDWHG SURFHVVHV LQ OLTXLG SURSHOODQW WDQNV IRU DSSOLFDWLRQ WR VSDFH IOLJKW 7KH YHUWLFDO F\OLQGULFDO HQFORVXUH WKH\ VWXGLHG ZDV ILOOHG ZLWK ZDWHU DQG DUELWUDU\ WHPSHUDWXUH DQG KHDW IOX[ YDULDWLRQ ZHUH LPSRVHG RQ LWV ZDOO ZKLOH LWV ERWWRP ZDV NHSW LQVXODWHG $W WKH WRS IUHH VXUIDFH RI WKH OLTXLG KHDW WUDQVIHU WR WKH DPELHQW ZDV FRQVLGHUHG QHJOLJLEOH DQG WKXV LQVXODWHG 7KH\ r5 / 6DQL &,5(6 &DPSXV %R[ 8QLYHUVLW\ RI &RORUDGR %RXOGHU &RORUDGR

PAGE 37

Ub )LJXUH LVRWKHUPV DQG YHORFLW\ SURILOHV LQ D FDQ DW YDULRXV KHDWLQJ WLPHV DIWHU 6DQL f A R \UUU

PAGE 38

)LJXUH FRQWG

PAGE 39

KDG QXPHULFDOO\ FDOFXODWHG DQG SORWWHG WKH LVRWKHUPV DQG VWUHDPOLQHV LQVLGH WKH HQFORVXUH ,Q WKHLU H[SHULPHQWDO ZRUN WKH\ UHFRUGHG WHPSHUDWXUH YDOXHV DW WHQ GLIIHUHQW SRLQWV LQVLGH WKH F\OLQGULFDO HQFORVXUH 7KH REVHUYHG WHPSHUDWXUH YDOXHV GHYLDWHG IURP WKH QXPHULFDO UHVXOWV E\ b 7KLV GLVFUHSDQF\ ZDV DWWULEXWHG WR WKH VPDOO QRQV\PPHWU\ LQ WKH LPSRVHG ZDOO WHPSHUDWXUH ZKLFK PLJKW KDYH SURGXFHG WKUHH GLPHQVLRQDO HIIHFWV DQG DOVR WR KHDW ORVVHV IURP WKH ERWWRP DQG WRS RI WKH FRQWDLQHU 7RUUDQFH f FRPSDUHG ILYH VHSDUDWH ILQLWH GLIIHUHQFH IRUPXODWLRQV IRU WKH SUREOHP RI QDWXUDO FRQYHFWLRQ LQ F\OLQGULFDO HQFORVXUHV 7KH ILYH IRUPXODWLRQV ZHUH UHVSHFWLYHO\ E\ %DUDNDW DQG &ODUN f :LONHV DQG &KXUFKLOO f )URPP f DQG WZR RI KLV RZQ 7KH SK\VLFDO SUREOHP FKRVHQ ZDV WKH VWXG\ RI QDWXUDO FRQYHFWLRQ IORZ LQGXFHG LQ D YHUWLFDO F\OLQGULFDO HQFORVXUH E\ D VPDOO KRW VSRW FHQWUDOO\ ORFDWHG RQ WKH ERWWRP 2QH RI KLV WZR IRUPXODWLRQV ZDV H[SOLFLW DQG WKH RWKHU ZDV LPSOLFLW 7KH H[SOLFLW IRUPXODWLRQ ZDV VKRZQ WR EH QXPHULFDOO\ VXSHULRU LQ PDQ\ ZD\V WR WKH RWKHU IRXU DQG LW ZDV DOVR UHSRUWHG WR KDYH JRRG DJUHHPHQW ZLWK H[SHULPHQWDO REVHUYDWLRQV 7RUUDQFH DQG 5RFNHWW 7RUUDQFH HW DO f IRU ORZHU YDOXHV RI *UDVKRI QXPEHUV 7KLV ZDV DOVR WKH IRUPXODWLRQ XVHG E\ 6WHYHQV f UHIHUUHG WR HDUOLHU .HH HW DO f SHUIRUPHG D QXPHULFDO VWXG\ RI WHPSHUDWXUH GLVWULEXWLRQ DQG VWUHDPOLQHV LQ D FORVHG YHUWLFDO F\OLQGHU FRQWDLQLQJ XQLIRUPO\ GLVWULEXWHG KHDW JHQHUDWLQJ WULWLXP JDV 7KH ZDOOV ZHUH NHSW LVRWKHUPDO $Q XSZLQG PHWKRG ZDV XVHG IRU WKH FRQYHFWLYH WHUPV 1XPHULFDO WHPSHUDWXUH YDOXHV ZHUH FRPSDUHG ZLWK

PAGE 40

H[SHULPHQWDOO\ GHWHUPLQHG WHPSHUDWXUHV DW WKUHH D[LDO DQG WZR UDGLDO ORFDWLRQV 7KHUPLVWRU WKHUPRPHWHUV ZHUH XVHG IRU WHPSHUDWXUH PHDVXUHPHQW 7KH FDOFXODWHG WHPSHUDWXUH YDOXHV ZHUH UHSRUWHG WR EH ZLWKLQ WKH OLPLWV RI H[SHULPHQWDO HUURU RYHU D UDQJH RI *UDVnQRI QXPEHUV (YDQV HW DO f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f PDGH D QXPHULFDO VWXG\ RI XQVWHDG\ QDWXUDO FRQYHFWLRQ LQ WKHUPDOO\ VWUDWLILHG ZDWHU LQ D F\OLQGULFDO HQFORVXUH 7KH\ VWXGLHG WKH HIIHFW RI SODFHPHQW RI LQVXODWLRQ RQ PDLQWDLQHQFH RI WKH VWUDWLILFDWLRQ 7UDQVLHQW IORZ DQG WHPSHUDWXUH ILHOGV ZHUH SURYLGHG IRU WKH VWXG\

PAGE 41

352%/(0 )2508/$7,21 7KH JRYHUQLQJ HTXDWLRQV DQG ERXQGDU\ FRQGLWLRQV IRU WKH QDWXUDO FRQYHFWLRQ DQG FRQGXFWLRQ KHDWLQJ SURFHVV WRJHWKHU ZLWK WKH DVVXPSn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f JRYHUQLQJ VXFK QDWXUDOO\ FRQYHFWLYH PRWLRQ RI IOXLG LQ D F\OLQGULFDO VSDFH DUH WKH 1DYLHU6WRNHV HTXDWLRQV LQ D[LV\PPHWULF F\OLQGULFDO FRRUGLQDWHV %LUG HW DO f DQG DUH GHVFULEHG EHORZ

PAGE 42

7 ‘7 Y 7 8 U 7 ] B 7 Z 7 77 Wa U f§ f f§ f S S U U U ] f S  n W 3 9 W ? B S f 9f§X f§ f B S_ 7 f UUf 7 3 U ] ] ) U U ] Y - f L( )Yff ` ) ] ) ) ) ) ] f f ,BL USYf LS*f ) ) ] f 7KH ERXQGDU\ FRQGLWLRQV DUH VLGHZDO ) 5 ] + 7 7L 2 ,, R FHQWHUOLQH ) ] + L, R O WO R R ,W ) U ERWWRP ZDOO DGf ) 5 ] 7 7O 2 ,, WRS ZDOO ) 5 ] + 7 7O R ,, ,QLWLDOO\ WKH IOXLG LV DW UHVW DW XQLIRUP WHPSHUDWXUH U 5 ] + R Lf§ LL Lf§ Y f 5HIHU WR OLVW RI V\PEROV IRU GHILQLWLRQ RI LQGLYLGXDO V\PEROV $V H[SHFWHG WKH IORZ ILHOG HTXDWLRQV t f LV FRXSOHG ZLWK WKH WHPSHUDWXUH ILHOG HT f WKURXJK WKH GHQVLW\ WHUP 7KHVH

PAGE 43

HTXDWLRQV KDYH EHHQ DSSUR[LPDWHG IRU QR YLVFRXV GLVVLSDWLRQ -DOXULD f GLVFXVVHG YDULRXV VLWXDWLRQV XQGHU ZKLFK YLVFRXV GLVVLSDWLRQ FDQ EH QHJOHFWHG $OO RWKHU IOXLG SURSHUWLHV DUH WUHDWHG DV FRQVWDQWV H[FHSW YDULDWLRQ RI GHQVLW\ ZKLFK LV GLVFXVVHG EHORZ %RXVVLQHVT $SSUR[LPDWLRQ 7KH VROXWLRQ WR WKH VHW RI HTXDWLRQV FDQ EH FRQVLGHUDEO\ VLPSOLILHG XVLQJ WKH %RXVVLQHVT DSSUR[LPDWLRQ ,Q WKH %RXVVLQHVT DSSUR[LPDWLRQ WKH GHQVLW\ LV WUHDWHG DV FRQVWDQW H[FHSW LQ WKH ERG\ IRUFH WHUP 7KH GHQVLW\ GLIIHUHQFH LQ WKH ERG\ IRUFH WHUP LV DSSUR[n LPDWHG DV S S S 7 7 f f D D DQG WKH GHQVLW\ YDULDWLRQ LQ WKH FRQWLQXLW\ HTXDWLRQ LV LJQRUHG 7KLV DSSUR[LPDWLRQ UHTXLUHV ERWK 7 7 f m f }n f7 J m WR EH WUXH 7KH VHFRQG LQHTXDOLW\ LV JHQHUDOO\ VDWLVILHG IRU OLTXLGV 6R LV WKH VLWXDWLRQ ZLWK WKH ILUVW LQHTXDOLW\ )RU H[DPSOH IRU QDWXUDO FRQYHFWLRQ LQ ZDWHU DV LQ WKLV SUREOHP r& 7 ZDOO WHPSHUDWXUH RI WKH FDQ r& 7D LQLWLDO WHPSHUDWXUH RI OLTXLG LQ FDQ r&

PAGE 44

7KXV 7 7 f ff m f FO $OPRVW DOO WKH SUHYLRXV ZRUNV RQ QDWXUDO FRQYHFWLRQ LQ VLPLOLDU DUHDV KDYH XVHG WKH %RXVVLQHVT DSSUR[LPDWLRQ 7UDQVIRUPDWLRQ RI WKH 3'(V 7R IXUWKHU IDFLOLWDWH QXPHULFDO VROXWLRQ WR WKH VHW RI HTXDWLRQV WKH SULPDU\ YDULDEOHV DUH WUDQVIRUPHG WR D QHZ VHW RI YDULDEOHV 9RUWLFLW\ WRf DQG VWUHDP IXQFWLRQ LMM f DUH GHILQHG LQ WHUPV RI WKH SULPDU\ YDULDEOHV X D[LDO YHORFLW\f DQG Y UDGLDO YHORFLW\f B Y ] U f X U r U U B U ] DEf 8VLQJ YRUWLFLW\ VR GHILQHG WKH SUHVVXUH Sf LV HOLPLQDWHG IURP WKH X DQG Y PRPHQWXP HTXDWLRQV DQG 7KHVH WZR HTXDWLRQV JDYH WKH SDUDEROLF WUDQVSRUW HTXDWLRQ IRU YRUWLFLW\ 1H[W XVLQJ WKH GHILQLWLRQ RI P HTXDWLRQ f DQG WKH GHILQLWLRQ RI LM HTXDWLRQV D DQG Ef WKH HTXDWLRQ IRU LS LV GHULYHG 7KLV JDYH WKH YRUWLFLW\ VWUHDPIXQFWLRQ IRUPXODWLRQ IRU WKH QDWXUDO FRQYHFWLRQ SUREOHP

PAGE 45

1RQGLPHQVLRQDOL]DWLRQ RI WKH 9DULDEOHV 7KH YDULDEOHV LQ WKH HTXDWLRQ DUH WKHQ QRQGLPHQVLRQDOL]HG DV EHRZ ] X D W f§ W + 7 7 7R 7 + 2f Lf D r 7* r D f 7KH UHVXOWLQJ HTXDWLRQV DUH WKH JRYHUQLQJ HTXDWLRQV IRU QDWXUDO FRQYHFWLRQ LQ WKH IRUP XVHG IRU WKH SUHVHQW VWXG\ DH U DXO DUYf D 7" L f6U c B A U p nL L U DUAUDU f Rf 7U ^ DXFf DYWf B M QUU D U O DUX!f A f]f 7Ua ` f*U3Ub 3U^ JM" JMU& fS f f O DAMM A f6U f f L_! U O] B DL! 7: Y DEf

PAGE 46

7KH ERXQGDU\ FRQGLWLRQV LQ WKH WUDQVIRUPHG DQG QRQGLPHQVLRQDO L]HG YDULDEOHV FDQ EH ZULWWHQ DV VLGHZDOO U 5+ ] R X Y FHQWHUOLQH U ERWWRP ZDOO ] Le R U X 2f H Q R 8 U Y UMM ] R LL X Y WRS ZDOO US = X Y f ZLWK WKH LQLWLDO FRQGLWLRQV U ] R L_f X Y f 7KH UHODWLYH PHULWV RI VROYLQJ WKH WUDQVIRUPHG YDULDEOHV DV FRPSDUHG WR WKH XYSf V\VWHP RI SULPDU\ YDULDEOHV ZHUH GLVFXVVn HG E\ 5RDFKH f 7R VROYH WKH XYSf V\VWHP DQRWKHU HOOLSWLF 3RLVVRQ HTXDWLRQ ZRXOG QHHG WR EH IRUPHG IRU WKH SUHVVXUH Sf YDULn DEOH XVLQJ WKH WZR HTXDWLRQV IRU YHORFLW\ UHJDUGOHVV RI ZKHWKHU WUDQVLHQW SUHVVXUH VROXWLRQV DUH GHVLUHG )ROORZLQJ DUH WKH UHODWLYH HYDOXDWLRQV RI WKH XYSf V\VWHP DQG WKH f V\VWHP IRU WZR Af([SOLFLW ERXQGDU\ FRQGLWLRQV IRU YRUWLFLWLHV RQ VROLG ZDOOV DUH XQDYDLODEOH

PAGE 47

GLPHQVLRQDO IORZV ZKHQ QR IUHH VXUIDFH LV SUHVHQW LWHUDWLYH PHWKRGV DUH XVHG WR VROYH WKH 3RLVVRQ HTXDWLRQ DQG WUDQVLHQW SUHVVXUH VROXn WLRQV DUH QRW UHTXLUHG 5RDFKH f 7KH XWML f V\VWHP QHHGV WR VROYH RQH SDUDEROLF HTXDWLRQ IRU Df DQG RQH HOOLSWLF 3RLVVRQ HTXDWLRQ IRU LS DV FRPSDUHG WR WKH XYSf V\VWHP ZKLFK UHTXLUHV WKH VROXWLRQ RI WZR SDUDEROLF HTXDWLRQV IRU X DQG Y DQG RQH HOOLSWLF HTXDWLRQ IRU S :LWK WKH RL M!f V\VWHP WZR DGGLWLRQDO GLIIHUHQWLDO HTXDWLRQV D DQG Ef DUH QHHGHG WR JHW WKH YHORFLWLHV X DQG Y EXW WKH X DQG Y PRPHQWXP WUDQVSRUW HTXDWLRQV DUH PRUH FRPSOLFDWHG WKDQ WKH X WUDQVSRUW HTXDWLRQ $W QRVOLS ZDOOV WKH ERXQGDU\ FRQGLWLRQV IRU X DQG Y DUH NQRZQ H[SOLFLWO\ XZDQ YZDQ f DV FRPSDUHG WR X A ZKLFK LV QRW NQRZQ H[SOLFLWO\ 7KLV ZRXOG EH D JUHDW DGYDQWDJH IRU XVLQJ LPSOLFLW PHWKRGV IRU WKH X Y Sf V\VWHP EXW QRQn OLQHDU LQVWDELOLWLHV RI WKH SUHVVXUH WHUP SUHYHQWV WKH X Y Sf V\VWHP IURP XVLQJ LPSOLFLW PHWKRGV 7KH HOOLSWLF HTXDWLRQ WKDW QHHGV WR EH VROYHG IRU WKH SUHVVXUH WDNHV PXFK ORQJHU WR FRQYHUJH WKDQ WKH HOOLSWLF HTXDWLRQ IRU HT f &RQGXFWLRQ +HDW 7UDQVIHU LQ &DQQHG )RRGV 7KH 32( GHVFULELQJ FRQGXFWLRQ KHDW WUDQVIHU LQ D F\OLQGULFDO ERG\ RI VROLGSDFNHG IRRG PDWHULDO LV SUHVHQWHG KHUH 7KH ERXQGDU\ FRQGLWLRQV SUHVHQW GXULQJ WKH WKHUPDO SURFHVVLQJ RI FDQQHG IRRGV DUH GHVFULEHG QH[W

PAGE 48

*RYHUQLQJ (TXDWLRQ )RU VWULFW FRQGXFWLRQ KHDWLQJ QR PRYHPHQW RI WKH PDWHULDOf RI D KRPRJHQHRXV LVRWURSLF F\OLQGULFDOVKDSHG ERG\ RI WKH IRRG PDWHULn DO WKH KHDW WUDQVIHU LV GHVFULEHG E\ WKH FODVVLFDO KHDW FRQGXFWLRQ HTXDWLRQ B W ,U U U ,, f 7KLV LV WKH HQHUJ\ HTXDWLRQ f DSSUR[LPDWHG IRU ]HUR YHORFLWLHV GLIIXVLRQ WHUP RQO\f 7KLV HTXDWLRQ LV TXLWH ZHOO EHKDYHG DQG HDV\ WR VROYH 7KH UHVLVWDQFH WR KHDW WUDQVIHU RI WKH PHWDOOLF ZDOOV RI WKH FDQ ZDV LJQRUHG %RXQGDU\ &RQGLWLRQV &DQQHG IRRGV DUH JHQHUDOO\ KHDWHG ZLWK KRW ZDWHU RU VWHDP DQG FRROHG ZLWK ZDWHU 7KXV WKH KHDW WUDQVIHU FRHIILFLHQW DW WKH RXWn VLGH RI WKH FDQ ZDOO FRXOG EH FRPSDUDWLYHO\ VPDOO IRU ZDWHUf RU TXLWH ODUJH IRU FRQGHQVLQJ VWHDPf ,Q JHQHUDO WKH FRQGLWLRQV DW WKH ERXQGDULHV FDQ EH ZULWWHQ DV VLGHZDOO U 5+ ] _eK+ R FHQWHULQH U ] B U R DGf ERWWRP ZDOO U 5+ ] B = K+ WRS ZDOO 0HR U 5+ ]

PAGE 49

ZLWK WKH LQLWLDO FRQGLWLRQV U 5+ ] f &RPSXWHU &RQWURO RI &RQGXFWLRQ+HDWLQJ 3URFHVVHV ,Q FRQWUROOLQJ WKHUPDO SURFHVVHV WKH REMHFWLYH LV WR PHHW WKH GHVLJQHG OHYHO RI EDFWHULDO VWHULOL]DWLRQ )Af IRU WKH SURFHVV LUUHVSHFWLYH RI DQ\ UHWRUW WHPSHUDWXUH YDULDWLRQ 75Wf DQG ZLWK D PLQLPXP RI RYHUSURFHVVLQJ 7KH OHWKDO HIIHFWV RI WKHUPDO SURFHVVLQJ DUH DFKLHYHG GXULQJ WKH KHDWLQJ DV ZHOO DV FRROLQJ WLPHV WA DQG WFf 7KXV WKH REMHFWLYH IXQFWLRQ LQ WKHUPDO SURFHVVLQJ LV WR W -LOO W W ,G PLQLPL]H )WKWFA A A GW F = GW Df XK VXEMHFW WR WKH FRQVWUDLQW )WKWFf AR Ef ZKHUH WKH LQWHJUDO IURP HTXDWLRQ KDV EHHQ EURNHQ LQWR VHSDUDWH KHDWLQJ DQG FRROLQJ WLPHV )4Wf LV WKH DFFXPXODWHG VWHULOL]DWLRQ DW DQ\ WLPH W 7KH ILUVW LQWHJUDO DERYH LV WKH FRQWULEXWLRQ WR WKH VWHULOL]DWLRQ )Wf IURP KHDWLQJ DQG WKH VHFRQG LQWHJUDO LV WKH FRQWULEXWLRQ WR WKH VWHULOL]DWLRQ GXULQJ FRROLQJ 7HPSHUDWXUH 7Wf LV WDNHQ WR EH WKH WHPSHUDWXUH RI WKH VORZHVW KHDWLQJ SRLQW LQ WKH SURGXFW 7KLV LV VR WKDW ZKHQ WKH GHVLJQ )T)Tf LV VDWLVILHG DW WKLV SRLQW DOO RWKHU SRLQWV LQ WKH SURGXFW KDYH DOVR EHHQ VDWLVn ILHG )RU D F\OLQGULFDO FDQ 7Wf LV WKH WHPSHUDWXUH DW WKH FDQ JHRPHWULF FHQWHU DW WLPH W

PAGE 50

7KLV WUDQVLHQW WHPSHUDWXUH 7Wf LV D IXQFWLRQ RI WUDQVLHQW UHWRUW WHPSHUDWXUH 75Wf WKH FDQ ZDOO RXWVLGH KHDW WUDQVIHU FRHIILn FLHQW Kf GLPHQVLRQV RI WKH FDQ 5 DQG +f DQG WKHUPDO GLIIXVLYLW\ Df RI WKH SURGXFW IRU D FRQGXFWLRQKHDWLQJ IRRG 6\PEROLFDOO\ 7Wf I 75Wf 5 + Df f )RU FRQGXFWLRQ KHDWLQJ LQ D F\OLQGULFDO FDQ WKH IXQFWLRQ I LV WKH KHDW FRQGXFWLRQ HTXDWLRQ LQ F\OLQGULFDO FRRUGLQDWHV )RU D JLYHQ FDQ VL]H DQG SURGXFW WKH WZR YDULDEOHV WKDW FDQ SRVVLEO\ EH FRQWUROOHG DUH WKH UHWRUW WHPSHUDWXUH 75Wf GXULQJ KHDWLQJ VDWXUDWHG VWHDP RU KRW ZDWHU XQGHU SUHVVXUH LQ WKH UHWRUWf DQG WKH WLPH SHULRG RI KHDWLQJ WA &RROLQJ LV JHQHUDOO\ GRQH XVLQJ DYDLODEOH ZDWHU DW DPELHQW WHPSHUDWXUH 7KLV PDNHV WKH FRQWULEXWLRQ WR ) IURP FRROLQJ JLYHQ E\ WKH VHFRQG LQWHJUDO LQ HTXDWLRQ Df LPPXQH IURP IXUWKHU FRQWURO RQFH WKH FRROLQJ SURFHVV LV XQGHU ZD\ $OVR ERWK WKH KHDWLQJ DQG FRROLQJ WHPSHUDWXUHV FDQ RQO\ EH DSSOLHG WR WKH ERXQGDU\ RI WKH SURGXFW PDNLQJ LW D ERXQGDU\ FRQWURO SUREOHP ,Q WKH FDVH RI D SURFHVV GHYLDWLRQ WKH KHDWLQJ PHGLXP WHPSHUDn WXUH PD\ JR WKURXJK XQH[SHFWHG YDULDWLRQV EH\RQG WKH FDSDFLW\ RI WKH WHPSHUDWXUH FRQWUROOHU 7KXV LQ UHDOLW\ RQO\ RQH YDULDEOH FDQ EH FRQWUROOHG ,W LV WKH WLPH SHULRG RI KHDWLQJ WAf 7KH SUREOHP LQ WKHUPDO SURFHVVLQJ LV WR VSHFLI\ WA IRU DUELWUDU\ YDULDWLRQV LQ 75Wf GXULQJ KHDWLQJ ZKLOH WKH SURFHVV LV XQGHU ZD\ LQ UHDOWLPH ,QVWHDG RI VHDUFKLQJ IRU WKH SURSHU WA WKURXJK VLPXODWLRQ HYHU\ WLPH D GHYLDWLRQ WDNHV SODFH DV VXJJHVWHG E\ 7HL[HLUD DQG 0DQVRQ f )Wf FDQ EH PRQLWRUHG LQ UHDOWLPH XVLQJ 7Wf FDOFXODWHG

PAGE 51

IURP WKH ILQLWH GLIIHUHQFH DSSUR[LPDWLRQ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ f ZLWK PHDVXUHG ERXQGDU\ UHWRUWf WHPSHUDWXUH 7AWf 7KH KHDWLQJ SURFHVV FRXOG FRQWLQXH XQWLO WLPH WA VXFK WKDW )R9 )R WKXV VDWLVI\LQJ WKH UHTXLUHG VWHULOL]DWLRQ IRU DUELWUDU\ YDULDn WLRQV LQ UHWRUW WHPSHUDWXUH 7UWf +RZHYHU WKH FRQWULEXWLRQ WR ) IURP FRROLQJ WKH VHFRQG LQWHJUDO LQ HTXDWLRQ Df VKRXOG QRW EH QHJOHFWHG IRU D FRQGXFWLRQ KHDWLQJ SURGXFW 'HSHQGLQJ RQ WKH FDQ FHQWHU WHPSHUDWXUH DW WKH VWDUW RI FRROLQJ VL]H RI WKH FDQ DQG RWKHU IDFWRUV LQ D FRQGXFWLRQ KHDWLQJ VLWXDWLRQ WKH FRQWULEXWLRQ WR )4 IURP FRROLQJ FRXOG EH DV PXFK DV b RI WKH WRWDO 7KLV LV XQOLNH FRQYHFWLRQ KHDWHG SURGXFWV ZKHUH ‘ WKH FRQWDLQHU LV VRPHWLPHV DJLWDWHG DQG WKH SURGXFW LV UDSLGO\ FRROHG PDNLQJ D VPDOO FRQWULEXWLRQ WR ) GXULQJ FRROLQJ 7R DYRLG JURVV RYHUSURFHVVLQJ RI FRQGXFWLRQ KHDWHG SURGXFW WKHUHIRUH WKH FRROLQJ ) PXVW EH FRQVLGHUHG $OVR DV PHQWLRQHG HDUOLHU WKHUH LV QR FRQWURO RYHU WKH SRVVLEOH FRQWULEXWLRQ WR )4 IURP FRROLQJ RQFH FRROLQJ LV XQGHU ZD\ 7KXV WKH FRQWULEXWLRQ IURP FRROLQJ FDQQRW EH QHJOHFWHG LW LV QRW D FRQVWDQW DQG LW FDQQRW EH FRQWUROOHG 7KHUHIRUH LW FDQ RQO\ EH HVWLPDWHG EHIRUH FRROLQJ LV DFWXDOO\ VWDUWHG 6XFK HVWLPDWLRQ RI FRROLQJ OHWKDOLW\ FDQ EH GRQH WKURXJK VLPXODWLRQ RI WKH FRROLQJ F\FOH DVVXPLQJ D FRQVWDQW FRROLQJ ZDWHU WHPSHUDWXUH LQ WKH UHWRUW 7KH REMHFWLYH KHUH LV WR GHYHORS DQ DOJRULWKP WR SHUIRUP WKLV VLPXn ODWLRQ LQ UHDOWLPH ZKHQ RWKHU FRQWURO DFWLRQV DUH JRLQJ RQ

PAGE 52

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

PAGE 53

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f DQG WKH WHFKQLTXH DQG RUGHU RI VROYLQJ LQGLYLGXDO 3'(V LQ WKH RYHUDOO FRPSXWDWLRQDO F\FOH 7KH UHODWLYH PHULWV RI WKHVH FKRLFHV DUH GLVFXVVHG DQG DSSURSULDWH VHOHFWLRQV DUH GHVFULEHG LQ WKLV VHFWLRQ *ULG *HQHUDWLRQ DQG *ULG 6WUHWFKLQJ 7KH GLVFUHWH SRLQWV DW ZKLFK ILQLWH GLIIHUHQFH HTXDWLRQV DUH WR EH VROYHG FRQVWLWXWH WKH JULG V\VWHP IRU WKH SUREOHP )RU FRPSXWDn WLRQDO HIILFLHQF\ WKH QXPEHU RI JULG SRLQWV VKRXOG EH WKH PLQLPXP WKDW LV UHTXLUHG WR UHVROYH DOO WKH VLJQLILFDQW VSDWLDO YDULDWLRQV RI WKH IORZ 6LQFH WKLV SUREOHP LQYROYHV D UDWKHU ODUJH VWHS LQFUHDVH LQ WKH ZDOO WHPSHUDWXUHV WRJHWKHU ZLWK D KLJK 5D\OHLJK QXPEHU 5Df VWURQJ ERXQGDU\ OD\HUV DUH H[SHFWHG WR IRUP QHDU WKH ZDOOV 7KLV LV WUXH SDUWLFXODUO\ IRU WKH VLGH YHUWLFDOf ZDOO ,Q RUGHU WR UHVROYH WKHVH ERXQGDU\ OD\HUV DGHTXDWHO\ D ODUJH FRQFHQn WUDWLRQ RI JULG SRLQWV ZRXOG EH QHFHVVDU\ QHDU WKH ZDOOV $Q DOJHn EUDLF PHWKRG ZDV XVHG WR JHQHUDWH WKH QRQXQLIRUP JULG V\VWHP 7KH SDUWLFXODU DOJHEUDLF H[SUHVVLRQ FKRVHQ ZDV HYDOXDWHG IRU WKH SURSHU

PAGE 54

WLHV OLVWHG EHORZ $QGHUVRQ HW DO .DnOQH\ GH 5LYDV f +HUH U]f UHIHUV WR WKH SK\VLFDO GRPDLQ DQG Qf UHIHUV WR WKH FRPSXWDWLRQDO GRPDLQ ZLWK $e DQG $Q DV FRQVWDQWV 7KH PDSSLQJ IURP &Qf WR U]f PXVW EH RQH WR RQH *ULG OLQHV VKRXOG EH VPRRWK WR SURYLGH FRQWLQXRXV WUDQVIRUPDWLRQ GHULYDWLYHV 7KLV UHTXLUHV UVf DQG ]Qf DV ZHOO DV DQG A VKRXOG EH FRQWLQXRXV RYHU WKH FORVHG LQWHUYDOV U5+ DQG ]O UHVSHFWLYHO\ 6LQFH FRQWLQXRXV IXQFWLRQV GHILQHG RQ D FORVHG LQWHUYDO DUH ERXQGHG WKLV LQ WXUQ LPSOLHV WKDW GU FII GU DQG G] VKRXOG EH ILQLWH RYHU WKH ZKROH LQWHUYDO ,I EHFRPHV LQILQLWH DW VRPH SRLQW WKH PDSSLQJ UFf ZRXOG JLYH SRRU UHVROXWLRQ QHDU WKDW SRLQW ZKLFK FDQQRW EH LPSURYHG E\ LQFUHDVLQJ WKH QXPEHU RI SRLQWV G S VLQFH D U r A_f DVf 6LPLODUO\ IRU G]GQ &ORVH VSDFLQJ RI JULGV ZKHUH ODUJH QXPHULFDO HUURUV DUH H[ G S L SHFWHG )RU WKLV SUREOHP WKDW PHDQV UU? DW WKH DnU 5+ VLGHZDOO ZKLFK HQVXUHV KLJK UHVROXWLRQ RI WKH ERXQGDU\ OD\HU QHDU U 5+ 6LPLODUO\ DW ] DQG ] $EVHQFH RI H[FHVVLYH JULG VNHZQHVV ZKLFK FDQ VRPHWLPHV H[DJJHUDWH WUXQFDWLRQ HUURUV 5DLWKE\ f $Q DOJHEUDLF WUDQVIRUPDWLRQ HTXDWLRQ VXJJHVWHG E\ .XEOEHFN HW DO f IRU QDWXUDO FRQYHFWLRQ LQ FDYLWLHV ZDV Q  WDQ>LU ]Of S@M O WDQQ D f

PAGE 55

ZKHUH D LV D GHIRUPDWLRQ SDUDPHWHU WKDW YDULHV WKH GHJUHH RI VWUHWFKLQJ +RZHYHU VLQFH WKH FRPSXWHU FRGLQJ RI WKH SUREOHP LV XOWLPDWHO\ GRQH LQ WKH SK\VLFDO GRPDLQ WKH LQYHUVH WUDQVIRUn PDWLRQ ]Qf LV PRUH DSSURSULDWH LQ WKLV FRQWH[W ZKLFK LV JLYHQ EHORZ ] ^ WDQAAQ f WDQ Df` f 7KLV FRQFHQWUDWHV WKH JULGV RQ HLWKHU HQG ] DQG ] ,Q WKH U GLUHFWLRQ WKH DERYH WUDQVIRUPDWLRQ LV PRGLILHG WR FRQFHQWUDWH JULGV RQ RQH VLGH VLGH ZDOO U 5+f RQO\ 7KXV 5 + f 7R FKHFN LI WKH WUDQVIRUPDWLRQV LQGHHG VDWLVI\ UHTXLUHPHQWV WKURXJK ZH ILUVW VHH WKDW ERWK WKH WUDQVIRUPDWLRQV DUH RQH WR RQH 1H[W ORRNLQJ LQWR WKH WUDQVIRUPDWLRQ GHULYDWLYHV GU 5 Ge LW D ,7 WDPU D WDQLU D DEf G] WDQ WW D GQ LW D Q Of WDQLU D ZKLFK DUH FRQWLQXRXV LQ e DQG Q UHVSHFWLYHO\

PAGE 56

/W DO WDQ WWFWf E\ /+RVSLWDOnV UXOH 6PLO DUO\ /W IIO G] F7Q ] O /W D!O LU D WDQLUJ WDQ]LUD DQG /W DO GU U 5+ /W FUO 77 2 WDQX J WDQLU D 7KXV WKH WUDQVIRUPDWLRQV SURYLGH FORVHU VSDFLQJ RI JULGV QHDU ] DQG U 5+ )LJXUH VKRZV WKH JULG V\VWHP ZLWK FORVH VSDFLQJ RI JULGV QHDU WKH WRS ERWWRP DQG VLGHZDOO $V FDQ EH VHHQ WKHUH LV QR H[FHVVLYH JULG VNHZQHVV .DnOQH\ GH 5LYDV f DOVR VKRZHG WKDW ZKHQ FRQGLWLRQV DQG DUH VDWLVILHG IRU DQ\ VPRRWK IXQFWLRQ H[WUD WUXQFDWLRQ HUURUV GXH WR WKH QRQXQLIRUP JULG V\VWHP DUH LPSURYHG IURP ILUVW RUGHU WR VHFRQG RUGHU LQ $Q DQG $e 9LQRNXU f DOVR GLVFXVVHG VLPLODU W\SHV RI VWUHWFKLQJ IXQFWLRQV 'LVFUHWL]DWLRQ RI WKH 3DUDEROLF 7HPSHUDWXUH DQG 9RUWLFLW\ (TXDWLRQV 2I WKH VHYHUDO ILQLWH GLIIHUHQFH IRUPXODWLRQV WKDW DUH DYDLODEOH IRU DQ\ VHW RI 3'(V FDUHIXO FKRLFHV QHHGHG WR EH PDGH GHSHQGLQJ RQ PDQ\ IDFWRUV LQFOXGLQJ ERXQGDU\ FRQGLWLRQV JHRPHWU\ RI WKH SUREOHP W\SH RI VROXWLRQ VWHDG\VWDWH RU WUDQVLHQWf GHVLUHG UDQJH RI SDUDPHWHUV LQYROYHG SDUWLFXODUO\ WKH *UDVKRI QXPEHU LQ

PAGE 57

)LJXUH 7KH JULG V\VWHP DQG WKH ERXQGDU\ FRQGLWLRQV IRU WKH F\OLQGULFDO FDQ

PAGE 58

QDWXUDO FRQYHFWLRQ SUREOHPVf DV ZHOO DV UHDOLVWLF FRQVWUDLQWV RQ DYDLODEOH UHVRXUFHV 5RDFKH f )RU WKLV VWXG\ WKH WUDQVLHQW WHPSHUDWXUH DQG YHORFLW\ VROXWLRQV DUH QHHGHG DQG WKH UDQJH RI *UDVKRI QXPEHU LV 6LQFH WKH WHPSHUDWXUH DQG YRUWLFLW\ HTXDWLRQV DUH ERWK SDUDEROLF DQG DUH YHU\ VLPLODU WKH\ ZLOO EH GLVFXVVHG VLPXOWDQHRXVO\ LQ WKLV VHFWLRQ 7KH SHUIRUPDQFH RI WKH YDULRXV ILQLWH GLIIHUHQFH PHWKRGV ZDV QRWHG LQ WHUPV RI VWDELOLW\ WUXQFDWLRQ HUURU DQG FRQVHUYDWLRQ SURSHUW\ 6WDELOLW\ ZDV GHILQHG DV WKH FRQWUROOHG JURZWK RI WKH URXQGRII HUURUV 2n%ULHQ HW DO f $Q XQFRQWUROOHG JURZWK RI WKH URXQGRII HUURUV OHDGV WR LQVWDELOLW\ LQ WKH ILQLWH GLIIHUHQFH PHWKRGV 7KH VWDELOLW\ UHTXLUHPHQWV RIWHQ SXW UHVWULFWLRQV RQ WKH WLPH VWHS DV D IXQFWLRQ RI JULG VSDFLQJV (PSLULFDO VWDELOLW\ UHn VWULFWLRQ RQ JULG VL]H RI WKH W\SH $[_XA M_ DQG $\ MYcM_ ZHUH DOVR UHSRUWHG E\ 7RUUDQFH f 7UXQFDWLRQ HUURU LV WKH HUURU UHVXOWLQJ IURP WKH GLVFUHWL]DWLRQ RI WKH 3'( &RQVHUYDWLRQ RI DQ\ SK\VLFDO HQWLW\ HJ HQHUJ\ YRUWLFLW\f ZLWKLQ WKH ILQLWH GLIIHUn HQFH JULG V\VWHP H[LVWV LI WKH GLIIHUHQFH HTXDWLRQV IRU WKH HQWLW\ DUH VXPPHG RYHU DOO LQWHULRU JULG SRLQWV DQG QR VSXULRXV VRXUFHV RU VLQNV RI WKHVH TXDQWLWLHV DUH IRXQG 7RUUDQFH f IRXQG WKDW IRUPXODWLRQV E\ %DUDNDW DQG &ODUN f DQG E\ :LONHV DQG &KXUFKLOO f GLG QRW FRQVHUYH HQHUJ\ RU YRUWLFLW\ DQG KH GLG QRW UHFRPPHQG WKH XVH RI WKHLU HTXDWLRQV $QRWKHU LPSRUWDQW SURSHUW\ RI ILQLWH GLIIHUHQFH IRUPXODWLRQV LV WKH WUDQVSRUWLYH SURSHUW\ $ GLIIHUHQFH IRUPXODWLRQ LV WUDQVSRUWLYH LI WKH HIIHFW RI D SHUWXUEDWLRQ LQ D WUDQVSRUW SURSHUW\ LV FRQYHFWHG RQO\ LQ WKH GLUHFWLRQ RI YHORFLW\

PAGE 59

7LPH GHULYDWLYHV 7KH WUHDWPHQW RI WKH WLPH GHULYDWLYH LQ HTXDWLRQV DQG FDQ EH EURDGO\ FODVVLILHG LQWR LPSOLFLW DQG H[SOLFLW PHWKRGV ([SOLFLW ILQLWH GLIIHUHQFH PHWKRGV DUH JHQHUDOO\ HDVLHU WR XQGHUVWDQG DQG SURJUDP ([SOLFLW PHWKRGV KDYH EHHQ XVHG E\ %DUDNDW DQG &ODUN f 7RUUDQFH f 6WHYHQV f DQG +LGGLQN f LQ VWXG\LQJ QDWXUDO FRQYHFWLRQ KHDWLQJ LQVLGH D F\OLQGULFDO FRQWDLQHU )RU PRVW RI WKH H[SOLFLW PHWKRGV KRZHYHU WKHUH DUH VWULQJHQW UHVWULFWLRQV RQ WKH DOORZHG WLPH VWHS IURP WKH VWDELOLW\ FRQVLGHUDWLRQV ,Q VLWXDWLRQV ZKHUH ODUJH WHPSHUDWXUH JUDGLHQWV HGGLHV DQG UHFLUFXODWLRQV DUH SUHVHQW OLNH LQ WKLV VWXG\ JULG SRLQWV QHHG WR EH ILQH HQRXJK WR UHVROYH WKH SK\VLFDO SKHQn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

PAGE 60

,QVWHDG RI XVLQJ D IXOO\ LPSOLFLW VHW RI HTXDWLRQV DOWHUQDWLQJ GLUHFWLRQ LPSOLFLW $',f PHWKRGV ZHUH LQWURGXFHG LQ FRPSDQLRQ SDSHUV E\ 3HDFHPDQ DQG 5DFKIRUG f DQG 'RXJODV f 7KH DGYDQWDJH RI WKH $', PHWKRG RYHU D IXOO\ LPSOLFLW PHWKRG LV WKDW WKH VHW RI HTXDWLRQV DOWKRXJK LPSOLFLW LV WULGLDJRQDO )XUWKHU WKH $', PHWKRG DV DSSOLHG WR OLQHDU HTXDWLRQV KDV D IRUPDO RUGHU RI DFFXUDF\ RI $W$]$Uf 7KH VWDELOLW\ RI WKLV PHWKRG LV DOVR XQFRQGLWLRQDO DV GHWHUPLQHG E\ 9RQ1HXPDQQ VWDELOLW\ DQDO\VLV )RU LUUHJXODUO\ VKDSHG UHJLRQV SURJUDPPLQJ FRXOG JHW FRPSOLFDWHG 5RDFKH f EXW IRU VLPSOH UHFWDQJXODU UHJLRQV DV LV WKH SUHVHQW VLWXDWLRQ WKDW LV QR SDUWLFXODU SUREOHP 7KH SUDFWLFDO DGYDQWDJH RI WKH $', PHWKRG RYHU WKH H[SOLFLW PHWKRG LV KRZHYHU QRWKLQJ OLNH WKDW LQGLFDWHG E\ 9RQ1HXPDQQ DQDO\VLV 5RDFKH f ,W LV XQVWDEOH ERWK DW ODUJH WLPH VWHSV DQGRU LQVXIILFLHQW FRQYHUJHQFH RI ERXQGDU\ YRUWLFLW\ ,Q JHQHUDO PDQ\ UHVHDUFKHUV KDYH LQGLFDWHG WKDW $', PHWKRGV GR DOORZ ODUJHU WLPH VWHSV DQG IDVWHU RYHUDOO FRPSXWDWLRQ E\ D IDFWRU RI WZR RU PRUH DQG IXUWKHUPRUH DOORZ VHFRQG RUGHU DFFXUDF\ LQ WLPH 7KH $', PHWKRG LV D YHU\ SRSXODU PHWKRG LQ FRPSXWDWLRQDO IOXLG G\QDPLFV :KLOH VROYLQJ WKH VHW RI 3'(V f %DUDNDW DQG &ODUN f FRPSDUHG H[SOLFLW DQG LPSOLFLW PHWKRGV IRU WKHLU VWXG\ DQG GHFLGHG DJDLQVW WKH XVH RI WKH LPSOLFLW PHWKRG GXH WR XQDYDLODELOLW\ RI WKH YRUWLFLW\ ERXQGDU\ FRQGLWLRQV DW WKH ZDOO WR EH H[SODLQHG ODWHUf 7KH\ DOORZHG YRUWLFLW\ ERXQGDU\ YDOXHV WR ODJ RQH WLPH VWHS QHFHVn VDU\ IRU VXFK H[SOLFLW IRUPXODWLRQV 7KH ODJJLQJ RI ERXQGDU\ YRUWLFLW\ LV FRQVLGHUHG XQGHVLUDEOH 5RDFKH f DQG LWHUDWLRQV

PAGE 61

DUH UHTXLUHG EHFDXVH RI WKLV VLWXDWLRQ 7KH LWHUDWLRQV FDQ DOVR VHUYH WKH SXUSRVH RI SUHVHUYLQJ VHFRQG RUGHU DFFXUDF\ LQ WLPH RI WKH $', PHWKRG 7KH VHFRQG RUGHU DFFXUDF\ PD\ RWKHUZLVH GHWHULRUDWH GXH WR WKH SUHVHQFH RI QRQOLQHDU FRQYHFWLYH WHUPV GLVFXVVHG EHORZf 6LQFH WKH LWHUDWLRQV DUH QHHGHG DQ\ZD\ DQ LPSOLFLW VHW RI HTXDWLRQV FDQ EH VROYHG LQVWHDG XWLOLVLQJ WKH LWHUDWLRQV $V QRWHG HDUOLHU 6WHYHQV f DQG +LGGLQN f DOVR XVHG WKH H[SOLFLW IRUPXODWLRQV RI %DUDNDW DQG &ODUN f WR VWXG\ WKH SUHVHQW SUREOHP DQG FRXOG QRW UHVROYH WKH FRQYHFWLYH HGGLHV SUHVHQW GXH WR WKH ERWWRP KHDWLQJ 6LQFH WKH $', PHWKRG PD\ DOORZ ILQHU JULG VSDFLQJV ZLWK OHVV VWULFWHU UHVWULFWLRQV RQ $W LW PD\ EH DEOH WR SLFN XS WKH FRQYHFWLYH HGGLHV SUHVHQW ZLWKRXW WDNLQJ SURKLELWLYH DPRXQWV RI FRPSXWHU WLPH 7KXV WKH $', PHWKRG ZDV FKRVHQ WR VROYH WKH WHPSHUDWXUH DQG YRUWLFLW\ HTXDWLRQ LQ WKLV VWXG\ 7KH $', IRUPXODWLRQ LV ZULWWHQ EHORZ IRU WKH WHPSHUDWXUH HTXDWLRQ $W WKH KDOI VWHS Q Xf $W ] Q 9 UYf U Q9 U UA UUf DQG DW WKH IXOO VWHS QB Q $[ 8f = Q UYf U Q ,] Q9 U UY U Q /LQHDUL]DWLRQ 7KH YHORFLWLHV LQ QRQOLQHDU FRQYHFWLYH WHUPV HJ Xf ] Q 9 VKRXOG LGHDOO\ EH HYDOXDWHG DW WKH VDPH WLPH OHYHO Q LQ WKLV FDVHf DV WKH RU Zf +RZHYHU VLQFH X DQG Y

PAGE 62

GHSHQGV RQ LM! ZKLFK LQ WXUQ GHSHQGV RQ WR WKLV ZRXOG FRXSOH DOO WKH HTXDWLRQV WRJHWKHU DQG PDNH LW SUDFWLFDOO\ LQFRPSXWDEOH 7R DYRLG WKLV WKH QRQOLQHDU WHUPV DUH OLQHDUL]HG 9DULRXV ZD\V RI OLQHDULn ]LQJ WKHVH WHUPV KDYH EHHQ GLVFXVVHG E\ UHVHDUFKHUV $QGHUVRQ HW DO 3H\UHW DQG 7D\ORU 5RDFKH f $ YHU\ VLPSOH DQG FRPPRQ VWUDWHJ\ LQ OLQHDUL]LQJ LV WR HYDOXDWH WKH YHORFLW\ FRHIILn FLHQW DW Q WLPH OHYHO 7KLV LV NQRZQ DV ODJJLQJ WKH FRHIILFLHQWV +RZHYHU LQ WKLV FDVH WKH DFFXUDF\ LQ WLPH GURSV IURP VHFRQG RUGHU IRU $',f WR ILUVW RUGHU 5RDFKH f QRWHG WKDW VXFK ODJJLQJ RI WKH YHORFLWLHV FRXOG VWLOO EH VHFRQG RUGHU DFFXUDWH LI WKH YHORn FLWLHV X DQG Y ZHUH VORZO\ YDU\LQJ $Q LPSURYHPHQW RYHU WKLV OLQHDUL]DWLRQ SURFHGXUH LV WR XSGDWH WKH FRHIILFLHQWV X DQG Y DV LWHUDWLRQV SURFHHG 7KH SURFHGXUH LV UHSHDWHG XQWLO WKH YDOXHV X DQG Y FRQYHUJH 2QH ZD\ WR DFKLHYH WKLV ZRXOG EH WR XSGDWH XQA DQG YQ9 E\ VROYLQJ WKH VWUHDP IXQFWLRQ HTXDWLRQ DW WKH KDOI WLPH VWHS RI WKH $', 5RDFKH f 6LQFH VROXWLRQ RI WKH 3RLVVRQ VWUHDP IXQFWLRQ HTXDWLRQ LV WKH PRVW WLPH FRQVXPLQJ RSHUDWLRQ LW ZDV GHFLGHG DJDLQVW WKLV XSGDWLQJ SURFHGXUH $QGHUVRQ HW DO f QRWHG WKDW ODJJLQJ WKH FRHIILFLHQWV KDV EHHQ ZLGHO\ DQG VDWLVIDFWRULO\ XVHG E\ SUHYLRXV LQYHVWLJDWRUV DQG UHFRPPHQGHG WKH XVH RI WKH VDPH IRU LWV VLPSOLFLW\ 7KXV WKH YHORFLW\ FRHIILFLHQWV X DQG Y ZHUH ODJJHG LQ WKLV ILQLWH GLIIHUn HQFH VWXG\ &RQYHFWLRQ WHUPV 7KH SUHVHQFH RI WKH FRQYHFWLRQ WHUPV PDNHV WKH WHPSHUDWXUH DQG WKH YRUWLFLW\ HTXDWLRQ TXDOLWDWLYHO\ GLIIHUHQW IURP RWKHU VLPSOH SDUDEROLF HTXDWLRQV 3URSHU QXPHULFDO WUHDWPHQW RI

PAGE 63

WKH FRQYHFWLRQ WHUP LV LPSRUWDQW 6HYHUDO GLIIHUHQW ZD\V RI ILQLWH GLIIHUHQFLQJ WKHVH WHUPV KDYH EHHQ SUHVHQWHG LQ WKH SUHYLRXV ZRUNV 7KHVH LQFOXGH VSDFHFHQWHUHG GLIIHUHQFLQJ DQG XSZLQG RU XSVWUHDP GLIIHUHQFLQJ 7KH XSZLQG PHWKRGV KDYH VHYHUDO GHVLUHG SURSHUWLHV WKDW DUH QRW VKDUHG E\ WKH RWKHU ILQLWH GLIIHUHQFH IRUPXODWLRQV DV DSSOLHG WR WKH FRQYHFWLRQ WHUP %RWK WKH XSZLQG PHWKRGV DUH WUDQVSRUWnYH PHDQLQJ DQ\ SHUWXUEDWLRQ LV FDUULHG DORQJ RQO\ LQ WKH GLUHFWLRQ RI YHORFLW\ %RWK WKH PHWKRGV DUH DOVR FRQVHUYDWLYH VR WKDW WKH\ SUHVHUYH WKH LQWHJUDO FRQVHUYDWLRQ UHODWLRQV LQ WKH FRQWLQXXP HTXDWLRQ 7KH ILUVW RI WKH WZR XSZLQG GLIIHUHQFLQJ PHWKRGV LV DV IROORZV /LOO\ )RUV\WKH DQG :DVRZ )UDQNHO f DXHf f XL XLO ] $] XLU XL $= X f X f f $FFXUDF\ RI WKLV PHWKRG LV $W$]f 7KLV ZDV XVHG E\ %DUDNDW DQG &ODUN f WR VROYH HTXDWLRQV WKURXJK 7KH VHFRQG YDULDWLRQ RI WKLV XSZLQG PHWKRG LV WR WUHDW WKH FRQYHFWLRQ WHUP DV IROORZV *HQWU\ HW DO f Xf = 9U 9O $= ZKHUH

PAGE 64

5 L 9L / 9L X5 f 8U f f X/ f X/ f ZKHUH 8U DQG XO DUH YHORFLWLHV DV H[SODLQHG EHORZ 7KLV ODWWHU XSZLQG PHWKRG DOVR KDV D IRUPDO DFFXUDF\ RI D]f +RZHYHU LW EHKDYHV OLNH LW LV VHFRQGRUGHU DFFXUDWH ZKHQ WKH TXDQWLW\ RU LQ f LV VORZO\ YDU\LQJ 5RDFKH f 7RUUDQFH f GHPRQn VWUDWHG WKDW WKH VHFRQG XSZLQG PHWKRG LV LQGHHG VXSHULRU WR WKH ILUVW XSZLQG PHWKRG 5RDFKH f 7KH VHFRQG XSZLQG PHWKRG ZDV XVHG LQ WKH SUHVHQW IRUPXODWLRQ 7KH YHORFLWLHV X5 DQG XA DUH DW WKH ULJKW DQG OHIW IDFH UHVSHFWLYHO\ RI WKH JULG YROXPH DV VKRZQ LQ ILJXUH $V ZLOO EH H[SODLQHG ODWHU D NLQHPDWLFDOO\ FRQVLVWHQW YHORFLW\ IRUPXODWLRQ 3DUPHQWLHU DQG 7RUUDQFH f ZDV XVHG ,Q WKLV IRUPXODWLRQ WUDQVSRUW YHORFLWLHV X> Bf DQG Y LQ ILJXUH f DUH GHILQHG QRUPDO WR WKH JULG YROXPHV $FFRUGLQJO\ 85 8/ X I Y $` ‘ / YLMO )ROORZLQJ LV D GHYHORSPHQW RI WKH ILQLWH GLIIHUHQFHV IRU WKH FRQYHFWLRQ WHUPV LQ WKH WHPSHUDWXUH HTXDWLRQ ,Q WKH ] GLUHFWLRQ

PAGE 65

Xf B DXf Q ] f UL ] B Xf Uf r ]Q 8Q 8 5 L@ / LOM $Q ]LU ]LL $Q ]L8n ] X5M X/OMf f5! 8/! 6L PLOLDUO\ Xf ] ]n9[ ]Ln.HLOM 9LMf 85 8/ 2 f§ IXQ X f 8R 8L ]LU ]LL 5 / / ]f§7f§ A5LM X/HLMf 8U 8/ Df Ef Ff Gf 7R DYRLG WKH XVH RI )2575$1 ,) VWDWHPHQWV IRU FKHFNLQJ WKH VLJQ RI XU DQG XO LQ WKH HTXDWLRQV Df WKURXJK Gf WKH HTXDWLRQV DUH FRPELQHG DV EHORZ Xf 85a,85 ) LL M9,XUa9.f Ma9XOfWLM U =L ‘ ]LO f

PAGE 66

,Q WKH U GLUHFWLRQ DUYf B D" DUYf B BBB DUYf DU DU D" f DUDV D" L Y5UM 9"L M a Y/UM9"LMO UMOa U-O $6 $6 9G! 9_B! ML Y/UM9HLML@ 6LPOLDUO\ UYf U 95 95 Y5 7KHUHIRUH FRPELQLQJ WKH HTXDWLRQV Y/ 9/ UYf DU 9O9mOfULLO9O95OfUL9O9/OfULALM9O9/Ofn_-MO UU -O -O f

PAGE 67

'LIIXVLRQ WHUPV &HQWUDO GLIIHUHQFLQJ ZDV XVHG IRU WKH GLIIXVLRQ WHUPV LQ ERWK WKH WHPSHUDWXUH DQG YRUWLFLW\ HTXDWLRQ 7KH GLIIXVLRQ WHUPV LQ WKH WHPSHUDWXUH HTXDWLRQ DUH ILQLWH GLIIHUHQFHG EHRZ B Q [ 77 f 67n6"r ,] M R= BL B L ffK9S]r L9IWM ]Q $Q M ]LO ] MU ] $Q L B8?$D ]L ]LO $Q ]LO ]LO LOL]LO ] L M a LOM ]L ]LO f ,LB Sp U U n U U U rA) G4 B L UnLM" a U U $ f B : 7ULO U-O UMO a UM $ : LML a 4LM S -A UU 7MUM UMO f

PAGE 68

7KH ILQLWH GLIIHUHQFH IRUPXODWLRQV IRU WKH WLPH DQG VSDFH GHULYDn WLYHV DUH WKHQ SXW WRJHWKHU LQ WKH $', IRUPXODWLRQ )RU HDFK RI WKH WHPSHUDWXUH DQG YRUWLFLW\ HTXDWLRQV WKH HTXDWLRQV DUH ZULWWHQ IRU WKH KDOI WLPH VWHS DQG IRU WKH IXOO WLPH VWHS RI WKH $', PHWKRG $OJHEUDLF VLPSOLILFDWLRQV DUH GRQH WR FRQYHUW WKH HTXDWLRQV DW HDFK RI WKH WLPH VWHSV WR D WULGLDJRQDO V\VWHPV RI HTXDWLRQV $ JOLPSVH RI WKH WULGLDJRQDO PDWULFHV IRU WKH KDOI DQG IXOO VWHS LV WKHQ LQFOXGHG 7KH WULGLDJRQDO V\VWHPV RI HTXDWLRQV DUH VROYHG XVLQJ WKH EDFN VXEVWLWXWLRQ SURFHGXUH 6PLWK f 7KH WLPH VWHS $W DQG WKH PHVK VL]HV $] $U ZHUH FKRVHQ LQ VXFK D ZD\ WKDW WKH FRHIILFLHQWV D Ec Fc LQ HTXDWLRQ Df DUH DOO SRVLWLYH DQG WKH PDWULFHV GLDJRQDOO\ GRPLQDQW E DL FAf 5LFKWP\HU DQG 0RUWRQ f VKRZHG WKDW WKHVH WZR FRQGLWLRQV DUH VXIILFLHQW WR NHHS WKH URXQGn RII HUURUV GRZQ LQ WKH EDFN VXEVWLWXWLRQ SURFHGXUH ,QWURGXFWLRQ RI WKH VHFRQG XSZLQG GLIIHUHQFLQJ PHWKRG VDWLVILHV WKH FULWHULRQ WKDW FRHIILFLHQWV Dc EM DQG &@ DUH DOO SRVLWLYH VHH .XEOEHFN HW DO f 7KH VHFRQG UHTXLUHPHQW RI GLDJRQDO GRPLQDQFH ZDV QRW LQYHVWLn JDWHG DQ\ IXUWKHU LQ WKLV VWXG\ +RZHYHU DW DOO WLPHV ZKHQHYHU WKH QXPHULFDO YDOXHV RI WKH FRHIILFLHQWV DQ Ec &M ZHUH YLVXDOO\ FKHFNHG WKH GLDJRQDO GRPLQDQFH ZDV IRXQG WR EH VDWLVILHG $', IRUPXODWLRQ IRU WKH WHPSHUDWXUH HTXDWLRQ LV QRZ GHYHORSHG )RU FRQYHQLHQFH WKH GLIIHUHQWLDO HTXDWLRQ IRU WHPSHUDWXUH LV UHSHDWHG 7KH ILQLWH GLIIHUHQFHV FRUUHVSRQGLQJ WR HDFK RI WKH GHULYDWLYHV LQ WKH WHPSHUDWXUH HTXDWLRQ DUH WKHQ DVVHPEOHG LQ WKH VDPH RUGHU

PAGE 69

$ U 8f UYf B U UB A W L a:a S U L 7] U 6Unn9 $W WKH KDOI RU LQWHUPHGLDWH VWHS Q9n ZULWWHQ DV LQ VKRUWf r Q L L $W Y.L YA :L UMOUMUUMLf r r r r Q U L ]LU]LL ]LUL n ]U]LL n -LUUUMLfO ? W BQ BQ ? UMY" AM MU7 M UL \ML MaM MLA UMU UM UU -O Df $W WKH IXOO VWHS Rrr L L $[ L A m9 .,.A AXU .O &A1O.X } Q L Q0 fQO Q Q0 ML Q L Q0 ML Q QQ fQO Y.LALLYLY5L!M9LUY,YO,L FYOLr/LfSMLLLML UMWUMUUML r r O ‘ B -M ]LU]LL ]LU]L ]Lf]LO QO ML Q MQ U ULU 9fA M -,7L -_7L 1 -_7L L ? 9" A MO7 B UM 9" A } a }M UMOnUM UU -O Ef

PAGE 70

$IWHU DOJHEUDLF PDQLSXODWLRQ WHPSHUDWXUH HTXDWLRQV DW WKH KDOI WLPH VWHS fDLLLMELLMfFLLOM UKVL Df ZKHUH OX/_X/_f A [ Df ]LU]LL ]LU]LA]U]LLf EL FL L f§tf§ X0X6LX0XP f§Lf§ U f§/B f§f§ L $[ ]B ] A 5 5O / /: ] ]B A ] ] ]]B L L L Q 5n,85 L L L L L L 9X_X_f ]LU]LL W]LU]LL-W]LU]If UKV GHA -IHM M n 99LnML UMOUMUn9L-LUMUMLf ,B LY5WO95O_UA/_/OfU--IU ULA $7 UMLUMUUMLf UMOUMUUMLAUMLUM UUUML Q L Q L I IU\.LOA -MIW SMUMLUMLUMWUMUUML-OUMUUMf DQG 85 L M ,, Vr &BL I I LOM / LMO

PAGE 71

$IWHU DOJHEUDLF PDQLSXODWLRQ WHPSHUDWXUH HTXDWLRQV DW WKH IXOO WLPH VWHS f Q1 M N QIO $Q B D kf f L E f F L UKV f LMM LM LMO Ef ZKHUH DM Q Q 9£\_Y/OfULL ULA UMUMUUML-UMfUMLf BL A UM E L7 UMUMUUMLf UMWUMUUMLf ULfUM UUUML 9AOYJOfA UO+O f:Y 9!UUMLQn!UUMf UKVM GrOM HHILOM 9.+?,f ]LU]LL ]LU]LLL8U]LLf L O-]f§X6LX6LX0X>Q Of§f§Of§ f§rf§ f $W ]LQ]B 5 5, / /1 ] ] A ]] ]] ‘L L L L L L L L 98U_XUf ]LU]LL I L]LU]LL.]LU]9f DQG X 95 I L X / X I 9 I L -O

PAGE 72

7KH WULGLDJRQDO PDWULFHV IRU WKH WHPSHUDWXUH HTXDWLRQV DW WKH WZR $', WLPH VWHSV DUH JLYHQ EHORZ $W WKH KDOI WLPH VWHS WKH HTXDWLRQV DUH ZULWWHQ IRU D FROXPQ JRLQJ IURP WKH ERWWRP ZDOO WR WKH WRS ZDOOf 6SHFLILHG ERXQGDU\ WHPSHUDWXUH LV XVHG ERWK DW WKH WRS DQG ERWWRP ZDOOV f EO &O W! F E F UKVLDLk!M UK6 UKV fDP EP fFP DPO EPOO PM P M UKVPB UKVPODPOPM Df $W WKH IXOO VWHS HTXDWLRQV DUH ZULWWHQ IRU WKH URZV IURP WKH FHQWHUOLQH WR WKH VLGHZDOOf $W WKH FHQWHUOLQH WKH LQVXODWHG ERXQGDU\ FRQGLWLRQ HTXDWLRQ f LV XVHG GXH WR V\PPHWU\ $W WKH VLGHZDOO WKH VSHFLILHG ERXQGDU\ WHPSHUDWXUH LV XVHG EODO & D E F E nF 9 UKVA rL UKV" L UKV EQ FQ *L Q UKVQB BDQO EQO SQO LUKVQOFQO4AQ Ef

PAGE 73

8SZLQG GLIIHUHQFLQJ RI WKH FRQYHFWLYH WHUPV LQ WKH YRUWLFLW\ HTXDWLRQ XRff ] VLPLOLDU WR Xf ] AXUa,XU A0MLM W\Of5O\OX/Of0LM \.3LLM ]] LO LO f 9Zf U Y5 ,Y5S0MMQ 95 _YU_ 9 .O.nM a YO AL MO f U O U -O -O 8SZLQG GLIIHUHQFLQJ RI WKH GLIIXVLRQ WHUPV LQ WKH YRUWLFLW\ HTXDWLRQ FM = ]LU ]WL f 2f f Wn LOM LLL LO]LU ]L ]L ]WL f F UXf U U n O UDMf ,)) ,7AU U n URLf UU W n -O -O ON U UXf U L-9" UM9" U L 9" $& UMU UML $" Uf UDf URf U X I -O B -O }-OO U L 7S f§ SL S@ m} LS f§ S L UM9UMU 9 M9AUM UMLA f &HQWUDO GLIIHUHQFLQJ RI WKH EXR\DQF\ WHUP U LMU L-O UMU UML f

PAGE 74

$', IRU WKH YRUWLFLW\ HTXDWLRQ e r I : *U3UA 3U^ t A A ,n0 f $W KDOI RU LQWHUPHGLDWH VWHS Df f§ WR $W UU Q L Q0 r U QL QL QL QQ r QL QQ r L =+U]LO I9On-5OKOM9O85_X/OX/OKM.On-/OKOM UU Q L QL Q O, QAL Q QAL QL Q U QLQL f! Q ^Y5_Y5OffLMOL9OY5_Y/OY/OffLM &9,YO'f0 r f r r f f f& ‘ IOQO BQO r B*U3U f§,VLWOf§OLLOOf§f SU ^ LLO}MaKLL M B fL -nALOM M U r ] f f§= ] m rr ] f ] f ] f L OL UMUUML ]LU]LLO ]LU]L Q 3U OUMUUML$W IXOO VWHS Q r -œ $ UML PLMUUM 0LWM M U f§ Uf OUMO f Q Q U WR U L Rf 7 -fO }UM9AUMLDf UU Q L QL r / Q/L QL QAL QL A r U QL QQ r L ,AMLX5_X5PLMM^Y_X5_X/_X/_f}LMMYOX/'fLOMO Q L Q0 Q U QL QW Q / Q L A Q Q QL Q UMWUUM7!I9Un 9U, ffL MO 9OYU,YOYO, ff YO9+ MOf r r f f IOQO Q r r B*U3U UL MOaL MO B3U U0OIM0LWB -=O]8L UMUUML ]LU]LL ]LU]L ]L]LL Q Q Q Q 3U UUMOXL MOfUM0L M UMfLM UMO XLMO L UMUUMLf UM Y OUMLSMM UMYO UMnUMLEf

PAGE 75

$IWHU DOJHEUDLF PDQLSXODWLRQ r r r Df Z L f EFRr f H&R L OOM L LM Xf UKV f L LOM L ZKHUH Df 9 X_X_f 3U ]LU]LL LU]LL-L]U]LLf EL LB OMOf§ X0X-,X0X3,O f§ N $W ]] A 5 5 / B1 ]] A ]] ]] fLO L LO L LO L L L FL 9 XUnXU,f 3U ]LU]LL L]LU]LLA]LU]LA UKV IHA HQO f 'Rf (Z )Fc *U3U L-O LL-O UMOUMO G 9Y-_Y_f 3U U M UMUSMLM UMALU-LUML.UUUML....Q f‘IF9N U/OU -O OUMUUMLfW9WUMUUMf A 95a95 ,f f UMUUML LUMWUUML!M-nMUn9 3UMMO

PAGE 76

$IWHU DOJHEUDLF PDQLSXODWLRQ Q N Q f Q B A D WR f 'f F UKV L!-O M L M M LMO M Ef ZKHUH OK Y_YO_f
PAGE 77

7KH WULGLDJRQDO PDWULFHV IRU WKH YRUWLFLW\ HTXDWLRQV DW WKH WZR $', WLPH VWHSV DUH JLYHQ EHORZ $W WKH KDOI WLPH VWHS HTXDWLRQV DUH ZULWWHQ IRU WKH FROXPQV IURP WKH ERWWRP ZDOO WR WKH WRS ZDOOf %RWK WKH ERWWRP DQG WRS ZDOO YRUWLFLW\ YDOXHV DUH XVHG IURP SUHYLRXV LWHUDWLRQ EO FL E & E & DP EP fDPO frL f OUKVODOZM fUM &0 UKV fM f§ UKV fFP 9M UKVP EPO UEVPODPO fPM Df $W WKH IXOO WLPH VWHS HTXDWLRQV DUH ZULWWHQ IRU WKH URZV IURP FHQWHUOLQH WR VLGHZDOOf 7KH FHQWHUOLQH ERXQGDU\ YRUWLFLW\ LV VSHFLILHG f DQG WKH VLGHZDOO ERXQGDU\ YRUWLFLW\ LV XVHG IURP SUHYLRXV LWHUDWLRQ
PAGE 78

'LVFUHWL]DWLRQ RI WKH (OOLSWLF 6WUHDP )XQFWLRQ (TXDWLRQ 7KH VWUHDP IXQFWLRQ HTXDWLRQ f REWDLQHG IURP WKH GHILQLWLRQ RI X A B - LV DOVR D VWDWHPHQW RI WKH FLUFXODWLRQ WKHRUHP ] U ZKLFK LV DL G$ Y -O f DUHD SHULPHWHU HOHPHQW RI DUHD 7KXV WKH GLIIHUHQFH IRUPXODWLRQ RI WKH VWUHDP IXQFWLRQ HTXDWLRQ KDV WR VDWLVI\ WKH FLUFXODWLRQ WKHRUHP LQ DGGLWLRQ WR FRQVHUYLQJ PDVV 'LIIHUHQFH DSSUR[LPDWLRQV WKDW FRQVHUYH PDVV DUH PRUH GLIILFXOW WR DFKLHYH LQ D FXUYLOLQHDU FRRUGLQDWH V\VWHP DV FRPSDUHG WR WKH &DUWHVLDQ FRRUGLQDWH V\VWHP 7R VDWLVI\ ERWK WKH FLUFXODWLRQ WKHRUHP DQG WKH FRQVHUYDWLRQ RI PDVV 3DUPHQWLHU DQG 7RUUDQFH f LQWURGXFHG WKH FRQFHSW RI WDQJHQWLDO YHORFLWLHV SDUDOOHO WR WKH FRQWURO YROXPH IDFHV DQG WUDQVSRUW YHORFLWLHV SHUSHQGLFXODU WR FRQWURO YROXPH IDFHV )LJXUH f ,Q ZKDW IROORZV WKHLU DSSURDFK KDV EHHQ DGDSWHG IRU WKH QRQ XQLIRUP JULG V\VWHP 7KH FLUFXODWLRQ WKHRUHP HTXDWLRQ f LV ZULWWHQ IRU D JULG YROXPH DURXQG QRGH LMf DV VKRZQ LQ ILJXUH f = L 9 = L 9 U M 9 U M 9

PAGE 79

]LO )LJXUH 1RWDWLRQV RI YDULRXV TXDQWLWLHV GHILQHG RQ D QRQXQLIRUP JULG V\VWHP LQ F\OLQGULFDO JHRPHWU\

PAGE 80

7KH WDQJHQWLDO YHORFLWLHV LN DQG DW WKH IDFHV RI WKH ‘ MMJULG YROXPH DUH REWDLQHG E\ ILQLWH GLIIHUHQFLQJ HTXDWLRQ Xr L AL -OnALnM9 UMLr UMf DEf 9Lf18 a A M UM]LU ]L 6XEVWLWXWLQJ WKHVH GHILQLWLRQV LQ HTXDWLRQ f f 9 ]L9]LL LOM L M ]LU ] ]] U fffL -UAL M ALMALMO UM!9IUM 9] :MLSMU UMLLUMUMLf f 1RWH WKDW WKH VDPH ILQLWH GLIIHUHQFH HTXDWLRQ f FDQ EH REWDLQHG IURP FHQWUDO GLIIHUHQFLQJ WKH VWUHDP IXQFWLRQ HTXDWLRQ f ZLWKRXW LQWURGXFLQJ WKH FLUFXODWLRQ WKHRUHP 7KLV LV VKRZQ LQ $SSHQGL[ $ $IWHU DOJHEUDLF PDQLSXODWLRQ

PAGE 81

f LM $rO M %rM &rOM nfnLMO (r7MO f ZKHUH $ UMO=O-n=L9O=Of=LOf % UM]I]L9O]+U] 9]LLn UMUMUM-OUMUUM!UMLUUUML! & :9L9ALUA OUM9UM9-UMOUMnUMLf ( W99U-9-U!9WO!Un9 7KH JULG SRLQW YHORFLWLHV DUH WKHQ DFKLHYHG WKURXJK OLQHDU LQWHUSRODWLRQ RI WKH WUDQVSRUW YHORFLWLHV XAI DQG YI ‘ Mn }X UM f UMIH XW UMWK UMB XW f‘UM9f UM9 UM9f UMY ffDf Y -a ?UO$ YW 9 YW ff=L9 =L9=L9 ]9 rEf

PAGE 82

6HFRQG WUDQVSRUW YHORFLWLHV DUH GHILQHG QRUPDO WR WKH JULG YROXPHV r LQ WHUPV RI D QHZ VHW RI VWUHDP IXQFWLRQ YDOXHV [S n Lr r r r XI YI 9U r f‘<:7" UMY-]LL LYA LODEf 7KH VWUHDP IXQFWLRQ YDOXHV LN DUH GHILQHG DW WKH FRUQHU RI WKH JULG YROXPHV 7KHVH WS YDOXHV DUH UHODWHG WR WKH SUHYLRXV VWUHDP 7 IXQFWLRQ YDOXHV LS WKURXJK D FRQVLVWHQF\ FULWHULRQ ZKLFK VD\V /W X X $U L L DEf /W YL L f YL L $U.f L_f f ZDV GHILQHG DV + A rM! f WKHQ XVLQJ GHILQLWLRQV RI YA f DQG Y HTXDWLRQV E DQG Ef LW ZDV VKRZQ WKDW WKH FRQVLVWHQF\ FULWHULRQ HTXDWLRQ Ef UHTXLUH [M f 6LPLODUO\ XVLQJ GHILQLWLRQV RI X f DQG Xc f LW ZDV VKRZQ WKDW r Mr !-

PAGE 83

WKH ZHDNHVW FRQGLWLRQ WKDW VDWLVILHV HTXDWLRQ D ZDV WKH UHFXUVLRQ UHODWLRQ M f ;M/ =Mf)7 f 8VLQJ ERXQGDU\ FRQGLWLRQV RQ WKH D[LV LW ZDV IRXQG WKDW ;R f 7KH WUDQVSRUW YHORFLWLHV GHILQHG WKLV ZD\ FRQVHUYH YROXPH IORZ H[DFWO\ DQG DUH FRQVLVWHQW ZLWK WKH JULG SRLQW YHORFLW\ ILHOG DV WKH PHVK LV UHILQHG 7KH WUDQVSRUW YHORFLWLHV XW DQG YW DUH WKH YHORFLWLHV ‘ Mn XVHG LQ WKH HQHUJ\ DQG YRUWLFLW\ WUDQVSRUW HTXDWLRQV 7KH JULG SRLQW YHORFLWLHV X DQG Y DUH WKH YHORFLWLHV SORWWHG LQ WKH L MM L YHORFLW\ SORWV 7R VROYH IRU A IURP WKH YRUWLFLW\VWUHDP IXQFWLRQ HTXDWLRQ ZH QRWH WKDW LW LV DQ HOOLSWLF 3RLVVRQ HTXDWLRQ DV FRPSDUHG WR WKH SDUDEROLF HTXDWLRQV IRU WHPSHUDWXUH DQG YRUWLFLW\ 'LUHFW QRQn LWHUDWLYHf PHWKRGV WR VROYH WKH HTXDWLRQ DUH DYDLODEOH WKDW DUH FRQVLGHUDEO\ WLPHVf IDVWHU WKDQ DQ LWHUDWLYH PHWKRG OLNH WKH VXFFHVVLYH RYHUUHOD[DWLRQ WHFKQLTXH WR EH GLVFXVVHG ODWHUf +RZHYHU GLUHFW PHWKRGV XVXDOO\ UHTXLUH ODUJH DPRXQWV RI VWRUDJH DQG DYDLODEOH FRPSXWHU FRGHV DUH RIWHQ OLPLWHG WR UHFWDQJXODU RU VLPSOH GRPDLQV RU KDYH UHVWULFWLRQV RQ WKH W\SH RI ERXQGDU\ FRQGLWLRQV 6KLK LQ SUHVVf GLVFXVVHG VXLWDELOLW\ RI GLUHFW PHWKRGV

PAGE 84

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f $ YHU\ HIILFLHQW GLUHFW PHWKRG IRU VROYLQJ 3RLVVRQnV HTXDWLRQ ZDV GHVFULEHG E\ 6FKXPDQQ DQG 6ZHHW f DQG WKH FRGH ZDV DYDLODEOH IURP WKH 1DWLRQDO &RXQFLO RI $WPRVSKHULF 5HVHDUFK 1&$5f VRIWZDUH OLEUDU\ +RZHYHU RQ FORVHU ORRN WKH DOJRULWKP *(1%81f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f IRU D UHFWDQJXODU UHJLRQ %LUNRII HW DO f +RZHYHU WKH RSWLPXP VHTXHQFH RI SDUDPHWHUV D5 6ZHHW 8 6 'HSW RI &RPPHUFH 1DWLRQDO %XUHDX RI 6WDQGDUGV %URDGZD\ %RXOGHU &RORUDGR 3HUVRQDO &RPPXQLFDWLRQ

PAGE 85

WKDW PDNHV WKH $', PHWKRG DGYDQWDJHRXV LV QRW DYDLODEOH LQ JHQHUDO 3URJUDPPLQJ WKH $', PHWKRG IRU D QRQUHFWDQJXODU UHJLRQ PD\ JHW FRPSOLFDWHG DQG WKH\ DUH QRW NQRZQ IRU VXUH WR EH IDVWHU 5RDFKH f $QRWKHU SRSXODU PHWKRG WR VROYH WKH XQVWHDG\ IRUPXODWLRQ RI WKH VWUHDP IXQFWLRQ 3RLVVRQ HTXDWLRQ LV WKH VXFFHVVLYH RYHUUHOD[DWLRQ 625f PHWKRG )UDQNHO f ,W LV YHU\ VLPSOH DQG HIIHFWLYH ,Q 625 WKH VHW RI VROXWLRQV DUH RYHU UHOD[HG E\ D IDFWRU Q WR VSHHG XS FRQYHUJHQFH 7KH H[DFW YDOXH LV IRXQG WKURXJK VLPSOH QXPHULFDO H[SHULPHQWDWLRQ ,W LV HDVLHU WR SURJUDP FRPSDUHG WR $', PHWKRGV 0RVW RI WKH HDUOLHU UHVHDUFKHUV KDYH XVHG WKH 625 WR VROYH WKH VWUHDP IXQFWLRQ HTXDWLRQ ,Q DOO WKH ILYH IRUPXODWLRQV RI QDWXUDO FRQYHFWLRQ LQ D F\OLQGULFDO JHRPHWU\ VWXGLHG E\ 7RUUDQFH f 625 ZDV XVHG WR VROYH WKH VWUHDP IXQFWLRQ HTXDWLRQ +RZHYHU LW LV WLPH FRQVXPLQJ DQG FRXOG HQG XS WDNLQJ b RI WKH WRWDO WLPH VSHQW WR VROYH WKH SUREOHP 1HYHUWKHOHVV 625 ZDV FKRVHQ WR VROYH WKH VWUHDP IXQFWLRQ HTXDWLRQ LQ WKH SUHVHQW VWXG\ IRU LWV VLPSOLFLW\ )ROORZLQJ LV WKH GHYHORSPHQW RI WKH 625 IRUPXODWLRQ 5HZULWLQJ HTXDWLRQ 7R JHW WKH 625 IRUPXODWLRQ n ) $rMFrLOMrL -O(rMOfOM f f

PAGE 86

NO Q rM f r crr fIL ` rM f 8VLQJ HTXDWLRQ VWUHDP IXQFWLRQ YDOXHV L_! DW DOO LQWHULRU SRLQWV DUH XSGDWHG 7KH HQWLUH VZHHS LV UHSHDWHG XQWLO WKH FRQYHUJHQFH FULWHULRQ 0D[ L M N N rLMf AL}M H f LV VDWLVILHG 7KH YDOXH RI H FKRVHQ IRU WKLV VWXG\ ZDV fA ,Q D r WUDQVLHQW FDOFXODWLRQ OLNH WKLV VWXG\ WKH RSWLPXP IW PD\ FKDQJH IURP RQH WLPH LQVWDQW WR DQRWKHU +RZHYHU WKH RSWLPXP IW IRXQG WKURXJK FDOFXODWLRQV GXULQJ WKH YHU\ ILUVW WLPH VWHS ZDV XVHG WKURXJKRXW WKH HQWLUH WUDQVLHQW FDOFXODWLRQV &RPSXWDWLRQDO %RXQGDU\ &RQGLWLRQV IRU WKH )LQLWH 'LIIHUHQFH (TXDWLRQV $ 3'( VXFK DV HTXDWLRQ IRU WR FDQ GHVFULEH D ZLGH YDULHW\ IORZ VLWXDWLRQ 7KH IORZV VROXWLRQVf DUH GLVWLQJXLVKHG RQO\ E\ WKH ERXQGDU\ DQG LQLWLDO FRQGLWLRQV DQG IORZ SDUDPHWHUV OLNH WKH *UDVKRI QXPEHU 7KXV VSHFLILFDWLRQ RI FRPSXWDWLRQDO ERXQGDU\ FRQGLWLRQV EHVLGHV DIIHFWLQJ QXPHULFDO VWDELOLW\ JUHDWO\ DIIHFWV WKH DFFXUDF\ RI WKH VROXWLRQ 5RDFKH f 8VLQJ WKH ERXQGDU\ FRQGLWLRQV DOUHDG\ QRWHG IRU WKH SDUWLDO GLIIHUHQWLDO HTXDWLRQV LQ WKH SUHYLRXV FKDSWHU WKH ERXQGDU\ FRQGLWLRQV IRU WKH ILQLWH GLIIHUHQFH HTXDWLRQV DUH QRZ GHULYHG 7KH ERXQGDU\ FRQGLWLRQ IRU WHPSHUDWXUH LV UDWKHU VLPSOH :KHQ WKH FDQ ZDOO WHPSHUDWXUH LV VSHFLILHG WKH FRPSXWDWLRQDO ZDOO

PAGE 87

WHPSHUDWXUH YDOXH RI WKH IOXLG QH[W WR WKH ZDOO LV PDGH HTXDO WR WKH VSHFLILHG YDOXH 2Q WKH DGLDEDWLF ZDOOV DQG DW WKH FHQWHUOLQH ZKLFK LV DOVR FRQVLGHUHG DGLDEDWLF GXH WR V\PPHWU\f D 7D\ORU VHULHV H[SDQVLRQ JLYHV $Q 9,6, $Q $Qf f : Q OZ ZKHUH VWDQGV IRU WKH GHULYDWLYH RI WHPSHUDWXUH LQ WKH GLUHFWLRQ QRUPDO WR WKH ZDOO DQG $Q EHLQJ GLVWDQFH QRUPDO WR WKH ZDOO IRU I _X f $Q : : $Qf f -: DH Q 9O : ,Q 7KXV LI ZM A LV VHW ZDOO WHPSHUDWXUH LV HYDOXDWHG IURP WKH DGMDFHQW WHPSHUDWXUH DQG LW LV VHFRQG RUGHU DFFXUDWH HYHQ WKRXJK LV RQO\ ILUVW RUGHU DFFXUDWH 5RDFKH f ,Q WKH $', PHWKRG ERXQGDU\ FRQGLWLRQV DUH QHHGHG DW WKH FRPSXWDWLRQDO KDOI VWHS 6LQFH ZDOO WHPSHUDWXUH A LV QRW D IXQFWLRQ RI WLPH WKH ERXQGDU\ WHPSHUDWXUH YDOXH DW WKLV KDOI WLPH VWHS LV VLPSO\ HTXDO WR A 8QOLNH WHPSHUDWXUH DQG YHORFLWLHV YRUWLFLW\ DQG VWUHDP IXQFWLRQ DUH QRW SULPDU\ YDULDEOHV 7KLV PDNHV VSHFLILFDWLRQ RI WKHLU ERXQGDU\ YDOXHV GLIILFXOW SDUWLFXODUO\ IRU YRUWLFLW\

PAGE 88

,PSURSHU VSHFLILFDWLRQ RI ERXQGDU\ YRUWLFLW\ LV IRXQG WR EH GHVWDELOL]LQJ LQ JHQHUDO DQG WKH VDPH ZDV IRXQG LQ WKH SUHVHQW SUREOHP 9RUWLFLW\ LV SURGXFHG DW QRVOLS ZDOOf ERXQGDULHV ,W LV WKHQ GLIIXVHG DQG FRQYHFWHG LQWR LQWHULRU SRLQWV 1R H[SOLFLW H[SUHVVLRQ IRU ZDOO YRUWLFLW\ LV DYDLODEOH ,W LV UHODWHG WR WKH LQWHULRU VWUHDP IXQFWLRQ YDOXHV E\ WKH YRUWLFLW\ VWUHDPIXQFWLRQ HTXDWLRQ HTXDWLRQ f +RZHYHU WKLV ZRXOG PDNH WKH YRUWLFLW\ HTXDWLRQ LPSOLFLW ZLWK WKH VWUHDP IXQFWLRQ HTXDWLRQ DQG WKH QXPEHU RI VLPXOWDQHRXV HTXDWLRQV LQFUHDVH GUDVWLFDOO\ 7R DYRLG WKLV %DUDNDW DQG &ODUN f FKRVH H[SOLFLW PHWKRGV RYHU LPSOLFLW DQG XVHG YRUWLFLW\ ERXQGDU\ YDOXHV FDOFXODWHG DW WKH SUHYLRXV WLPH VWHS :LONHV DQG &KXUFKLOO f XVHG WKH $', PHWKRG EXW VLPLODUO\ XVHG ERXQGDU\ YRUWLFLW\ YDOXHV IURP WKH SUHYLRXV WLPH VWHS 7KXV LQ WKHLU IRUPXODWLRQV ZDOO YRUWLFLW\ YDOXHV ODJJHG E\ RQH WLPH VWHS DW DOO WLPHV 5RDFKH f GLVFXVVHG WKH HIIHFWV RI VXFK ODJJLQJ RI ERXQGDU\ YRUWLFLW\ YDOXHV ,W IRUFHV WKH XVH RI D VPDOOHU WLPH VWHS VLQFH ODUJHU WLPH VWHSV FRXOG PDNH WKH VROXWLRQ LQDFFXUDWH DQG XQVWDEOH 6LQFH ODUJHU WLPH VWHSV DUH D PDMRU PRWLYDWLRQ IRU $', XSGDWLQJ RI WKH ERXQGDU\ YDOXHV WKURXJK LWHUDWLRQ LV SUHIHUUHG )URP WKH ZRUN RI 7RUUDQFH f 5RDFKH DOVR QRWHG WKDW WKH FRQYHUJHQFH RI WKH ERXQGDU\ YRUWLFLW\ YDOXHV DFWXDOO\ LPSRVHG D WLPHVWHS UHVWULFWLRQ RI WKH IRUP $W D $] ZKHUH D ZDV VRPH QXPEHU GHSHQGHQW RQ WKH SUREOHP DQG FRQYHUJHQFH UHTXLUHPHQWV 7KXV GXH WR WKH LPSOLFLWQHVV RI WKH ERXQGDU\ YRUWLFLW\ YDOXHV $', DFWXDOO\ KDV D WLPHVWHS UHVWULFWLRQ OLNH WKH VLPSOHVW RI WKH H[SOLFLW PHWKRGV 7KH

PAGE 89

GHILQLWLRQ RI YRUWLFLW\ WR ZDV XVHG DW WKH ERXQGDULHV WR REWDLQ H[SUHVVLRQV IRU ERXQGDU\ YRUWLFLWLHV $W WKH ERWWRP ERXQGDU\ WR YR}M r8O ]f W W AUMaUM A } U] ]? Df VLQFH nM YM UM]U ]R7 U] 8T M DQG 8J BM IURP HTQ D $W WKH WRS ERXQGDU\ W W W 9PMa YPOM XPM XPMO fffU OP : O9 U-: B P M U ] ] f] ] LM M P P P P VLQFH Ef 9P P M cWI} ‘ Bff P M PM YPOMf U] ] f M P P 9 Q P M U] ]f M P P XPM DG M IURP HTQ D

PAGE 90

$W WKH FHQWHUOLQH YL YLO XL XL XL ] L ] L ] L UL L Ff VLQFH YOM J DQG YA 4 IURP HTQ E fW B ALU r B rLL L WOUU f U?M9 XL $W WKH VLGHZDOO W W Y Y L Q LOQ 8f f§ Xf Xf ‘ Q L}Qf§ !Q 8L9 =L9A n Y YYf ,OO L QO U U U f U aU ff QO Q Q Q Q Gf VLQFH Y" Q DQG YA f IURP HTQ E f Q L LS aLS LS XW LQ L QO B L QO LQO U LU U 7U LU U ` Q9 Q Q Q Q Q X L Q ,GHQWLFDO H[SUHVVLRQV FDQ EH GHULYHG XVLQJ 7D\ORU VHULHV H[SDQVLRQV RI HTXDWLRQ DW WKH ZDOOV 7KLV LV VKRZQ LQ WKH DSSHQGL[ % 7KHVH H[SUHVVLRQV DUH ILUVW RUGHU DFFXUDWH %XW LW LV WKH RQO\ IRUP FRQVLVWHQW ZLWK WKH NLQHPDWLF YHORFLW\ IRUPXODWLRQ 3DUDPHQWLHU t 7RUUDQFH f 7KH VHFRQG RUGHU IRUP XVLQJ 7D\ORU VHULHV -HQVHQ f KDV DOVR EHHQ XVHG +RZHYHU VRPHWLPHV WKH VHFRQG

PAGE 91

RUGHU IRUP GHVWDELOL]HG WKH FDOFXODWLRQV :LONHV t &KXUFKLOO f DQG VRPHWLPHV LW ZDV OHVV DFFXUDWH WKDQ WKH ILUVW RUGHU IRUPXODWLRQ %HDUGVOH\ f 7KH YRUWLFLW\ ERXQGDU\ YDOXH DW WKH FRPSXWDWLRQDO LQWHUPHGLDWH RU KDOI VWHS RI WKH $2, PHWKRG LV PRUH GLIILFXOW WRVHW WKDQ WKH FRUUHVSRQGLQJ WHPSHUDWXUH YDOXH EHFDXVH WKH YRUWLFLW\ ERXQGDU\ YDOXH LV DOVR D IXQFWLRQ RI WLPH 7KH IROORZLQJ VLPSOH ZD\ WR GHILQH WKLV YRUWLFLW\ YDOXH RQ WKH ERXQGDU\ LV RIWHQ XVHG DQG JLYHV VXIILFLHQWO\ DFFXUDWH UHVXOWV %RQWRX[ HW DO f QYU Z Q fZ P f ZKHUH Q P ZDOO YRUWLFLW\ YDOXH DW WLPH VWHS Q PRVW UHFHQW HVWLPDWH RI ZDOO YRUWLFLW\ DW Q VWHS XVLQJ HTXDWLRQ Q f: P LQWHUPHGLDWH YRUWLFLW\ YDOXH EHWZHHQ Q DQG QOfWK WLPH OHYHO DW WKH POfWK LWHUDWLRQ ,VUDHOL f DOVR XVHG WKLV HTXDWLRQ WR GHILQH LQWHUPHGLDWH KDOI VWHSf ERXQGDU\ YRUWLFLW\ WKRXJK KH XVHG D GLIIHUHQW HTXDWLRQ WR PO FDOFXODWH ERXQGDU\ YRUWLFLW\ A &RQYHUJHQFH &ULWHULD IRU ,WHUDWLRQ $V GLVFXVVHG HDUOLHU WKH VHW RI ILQLWH GLIIHUHQFH HTXDWLRQV ZHUH VROYHG LWHUDWLYHO\ XQWLO FRQYHUJHQFH 7KH PHDVXUH RI FRQYHU

PAGE 92

JHQFH FRXOG KDYH VHYHUDO SRVVLELOLWLHV ,GHDOO\ DW FRQYHUJHQFH DOO WKH YDULDEOHV VKRXOG KDYH FRQYHUJHG )RU SUDFWLFDO UHDVRQV LW ZRXOG EH VXIILFLHQW WR FKHFN RQ WKH YDULDEOH WKDW FRQYHUJHG WKH VORZHVW 7KLV LV VR WKDW ZKHQ WKLV YDULDEOH KDV FRQYHUJHG DOO RWKHUV ZRXOG KDYH FRQYHUJHG +RZHYHU WKHUH DUH QR JHQHUDO JXLGHOLQHV WR FKRRVH DSULRUL WKH VORZHVW FRQYHUJLQJ YDULDEOH ,QWHULRU YHORFLW\ FRPSRQHQWV ZHUH VXJJHVWHG E\ 5RDFKH f DV D PHDVXUH RI FRQYHUJHQFH )UHTXHQWO\ UHVHDUFKHUV LQ WKH SDVW KDYH GHFLGHG RQ WKH YRUWLFLW\ ERXQGDU\ YDOXHV DV WKH YDULDEOHV WR ZDWFK IRU WKH LWHUDWLRQ FRQYHUJHQFH RI WKH HQWLUH SUREOHP 5RDFKH f UHIHUUHG WR WHVWV RQ ERXQGDU\ YRUWLFLWLHV DV WKH PRVW VHQVLWLYH WHVW IRU FRQYHUJHQFH 7KH VDPH YDULDEOHV ZHUH XVHG WR FKHFN IRU LWHUDWLRQ FRQYHUJHQFH LQ WKLV VWXG\ HYHQ WKRXJK WKH\ PD\ QRW KDYH EHHQ WKH YDULDEOHV WKDW FRQYHUJHG WKH VORZHVW )RU WKH FKRVHQ YDULDEOH WKHUH DUH QR GHILQLWLYH FULWHULD IRU WKH PHDVXUH RI FRQYHUJHQFH $ UHODWLYH HUURU FULWHULD ZDV XVHG IRU WKLV VWXG\ ZKLFK LV aQOPA aQOP f f Q P f f aQ DW DOO VROLG ERXQGDULHV 7KH ERXQGDU\ YRUWLFLWLHV R! ZHUH IRXQG IURP WKH LQWHULRU VWUHDP IXQFWLRQ YDOXH XVLQJ HTXDWLRQ $V DQ H[DPSOH RI RWKHU PHDVXUHV RI FRQYHUJHQFH ZKLOH VWXG\LQJ QDWXUDO FRQYHFWLRQ LQ D UHFWDQJXODU FDYLW\ .LPXUD DQG %HMDQ f XVHG WKH IROORZLQJ FRQYHUJHQFH FULWHULD RQ LQWHULRU SRLQWV

PAGE 93

= = L M Q LM P f ] ] L Q GL M Q GLM P7L PL f WRJHWKHU ZLWK D FRQYHUJHQFH FULWHULRQ RQ WKH FDOFXODWHG 1XVVHOW QXPEHU 7KH FKRLFH RI LV DOVR SUREOHP GHSHQGHQW DQG QR JHQHUDO JXLGHOLQHV H[LVW IRU LW HLWKHU (YHQ IRU YHU\ VPDOO YDOXHV RI WKH LWHUDWLRQ RI D VROXWLRQ PD\ EH VWRSSHG SUHPDWXUHO\ IRU FHUWDLQ W\SHV RI VROXWLRQV ZKLFK DUH QRW XQFRPPRQ 5RDFKH f 9DOXHV UDQJLQJ IURP WR ZHUH XVHG LQ WKHLU VWXG\ DQG WKH FKRVHQ YDOXH ZDV WKH ODUJHVW RI WKHVH YDOXHV WKDW GLG QRW FDXVH DQ\ SHUFHSWLEOH FKDQJHV LQ WKH ILQDO VROXWLRQ $OJRULWKP IRU WKH ,WHUDWLYH 6ROXWLRQ 7KH DOJRULWKP IRU WKH LWHUDWLYH VROXWLRQ RI WKH FRPSOHWH V\VWHP RI ILQLWH GLIIHUHQFH HTXDWLRQV HTXDWLRQV f LV QRZ GHVFULEHG 7KH JULG V\VWHP LV JHQHUDWHG ILUVW 7KH YRUWLFLW\ VWUHDP IXQFWLRQ DQG YHORFLW\ YDOXHV DUH VHW WR ]HUR ZKLFK LV WKHLU LQLWLDO FRQGLWLRQ W f VLQFH LQLWLDOO\ WKH OLTXLG ZDV VWDWLRQDU\ ZLWK XQLIRUP LQLWLDO WHPSHUDWXUH WKURXJKRXW WKH ERG\ RI WKH OLTXLG 7KH QRQGLPHQVLRQDO WHPSHUDWXUH YDOXHV DUH VHW WR RQH LQLWLDO WHPSHUn DWXUHf 7KH WHPSHUDWXUH ERXQGDU\ FRQGLWLRQ IRU W! LV WKHQ HQIRUFHG 7KH VHW RI HTXDWLRQV DUH WKHQ VROYHG IRU ILQLWH LQFUHPHQWV RI WLPH $Wf ZLWK LWHUDWLRQV SHUIRUPHG RQ ERXQGDU\ YRUWLFLW\ DW HDFK WLPH LQFUHPHQW 7KLV LV GHVFULEHG LQ ILJXUH DQG H[SODLQHG DV EHORZ 7KH VHOHFWLRQ RI WLPH VWHS WKH $W LV GLVFXVVHG XQGHU WKH 5(68/76 VHFWLRQ

PAGE 94

5(3($7 29(5 Q rQ Q f f P O 0 P RM ,7(5$7( 29(5 P Q B rQ f fR LQ Q f fU U rULUD m* 8L Y7f HPf P f f (O Q 7 Zr }QDff Q b ,, I U Q QOP YQf D P Q ,( !U f QOPB nD I Q QO f 86 P QAPB fQ P f ,7(5$7( P QOB I U Q nL AQ IV&pD P 817,/ QO Q P IOROG QOP A4R,G L P AP Q B r F QO fU IOA H Q P f \Rf7 aQO P ,\fRM Qf 817,/ Q P f f P r P }f IH2 817,/ '(6,5(' 7,0( 5($&+(' 7 VROLG ERXQGDU\ IW LQWHULRU SRLQWV p FHQWHUOLQH LQ a r RfRMRf WHPSRUDU\ VWRUDJH )LJXUH $OJRULWKP IRU WKH LWHUDWLYH VROXWLRQ RI WKH VHW RI HTXDWLRQV

PAGE 95

7KH WHPSHUDWXUH HTXDWLRQV D DQG Ef ZHUH VROYHG XVLQJ WKH $', WHFKQLTXH 7KLV JDYH QHZ WHPSHUDWXUH YDOXHV DIWHU LQFUHPHQWDO WLPH $W 7KH YRUWLFLW\ HTXDWLRQV D DQG Ef ZHUH WKHQ VROYHG DOVR XVLQJ WKH $', WHFKQLTXHf 7KH LQWHULRU YRUWLFLW\ YDOXHV FDOFXODWHG ZHUH SURYLVLRQDO VLQFH ROG YDOXHV ZHUH XVHG IRU ERXQGDU\ YRUWLFLW\ 7KH VWUHDP IXQFWLRQ HTXDWLRQV f ZHUH WKHQ VROYHG IRU LQWHULRU VWUHDP IXQFWLRQ YDOXHV XVLQJ WKH SRLQW 625 WHFKQLTXH 7KLV UHTXLUHG RQO\ LQWHULRU YRUWLFLW\ YDOXHV 3UHYLRXV VWUHDP IXQFWLRQ YDOXHV ZHUH XVHG DV DQ LQLWLDO JXHVV %RXQGDU\ YRUWLFLW\ YDOXHV =Af DW VROLG ZDOOV WRS ERWWRP DQG VLGHf DUH ILUVW FDOFXODWHGfXVLQJ HTXDWLRQ EDVHG RQ WKH QHZO\ FDOFXODWHG VWUHDP IXQFWLRQ YDOXHV +RZHYHU EHIRUH XSGDWLQJ WKH ERXQGDU\ YRUWLFLW\ YDOXHV =Ef DUH XQGHUUHOD[HG DV Q P < Df Q P \f Q P %2f Q P ZKHUH ERXQGDU\ YRUWLFLW\ HVWLPDWHG XVLQJ HTXDWLRQ DQG ZLWK \O 6WHSV WKURXJK DUH UHSHDWHG XQWLO WKH HVWLPDWHG A YDOXHV XVLQJ HTXDWLRQ f LQ WZR FRQVHFXWLYH LWHUDWLRQV PHHWV WKH FRQYHUJHQFH FULWHULRQ

PAGE 96

Q P f aQO b P aQOP b f IRU WRS ERWWRP DQG VLGHZDOO :KHQ WKH FRQYHUJHQFH FULWHULRQ LV VDWLVILHG WHPSHUDWXUH YRUWLFLW\ DQG VWUHDP IXQFWLRQ YDOXHV DUH FRQVLGHUHG WKH WUDQVLHQW VROXWLRQ DW WKLV WLPH LQVWDQW 7KH YHORFLWLHV DUH FDOFXODWHG IURP WKH VWUHDP IXQFWLRQ YDOXHV XVLQJ HTXDWLRQ DQG 6WHSV WKURXJK DUH UHSHDWHG IRU HYHU\ WLPH LQFUHPHQW XQWLO WKH GHVLUHG WLPH LV UHDFKHG 7KH XQGHUUHOD[DWLRQ RI ERXQGDU\ YDOXHV DW VWHS KDV RIWHQ EHHQ FRQVLGHUHG QHFHVVDU\ LQ WKH OLWHUDWXUH 3H\UHW DQG 7D\ORU 5RDFKH f IRU FRQYHUJHQFH RI WKH VROXWLRQ 8QGHUn UHOD[DWLRQ LV XVXDOO\ HPSOR\HG WR PDNH D QRQFRQYHUJHQW LWHUDWLYH SURFHVV FRQYHUJH +RUQEHFN f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

PAGE 97

&RGLQJ RI WKH 3URJUDP DQG WKH +DUGZDUH 7KH DOJRULWKP SUHVHQWHG LQ ILJXUH IRU VROYLQJ WKH YRUWLFLW\ VWUHDP IXQFWLRQ IRUPXODWLRQ RI WKH QDWXUDO FRQYHFWLRQ SUREOHP ZDV FRGHG LQ )2575$1 ZLWK LQGLYLGXDO PRGXOHV IRU VROYLQJ WHPSHUDWXUH YRUWLFLW\ VWUHDP IXQFWLRQ YHORFLW\ DQG ERXQGDU\ YRUWLFLW\ YDOXHV 6WUXFWXUHG SURJUDPPLQJ SUDFWLFHV ZHUH XVHG 8VH RI *2 72 VWDWHPHQWV ZHUH DYRLGHG H[FHSW LQ WZR VLWXDWLRQV WR VLPXODWH D :+,/( VWDWHPHQW 7KH :+,/( VWDWHPHQW ZDV XQDYDLODEOH LQ WKH SDUWLFXODU YHUVLRQ RI )2575$1 6LQJOH SUHFLVLRQ RI WKH YDULDEOHV ZDV NQRZQ WR FDUU\ VLJQLILFDQW GLJLWV IRU WKH SDUWLFXODU FRPSXWHU DQG WKH VDPH ZDV WDNHQ WR EH VXIILFLHQW IRU WKLV SUREOHP 7KH LQSXW GDWD ILOH IRU D W\SLFDO UXQ LV LQ $SSHQGL[ & 7KH FRPSXWHU XVHG ZDV D &'& &\EHU ZLWK WZR ELW &38V FUXQFKLQJ ELW ZRUGV DW 0+] SDUDOOHO SURFHVVRUV DQG RI ELW ZRUG FHQWUDO PHPRU\ 7KH SORWWLQJ ZDV GRQH XVLQJ D SORWWLQJ SDFNDJH IURP WKH 8QLYHUVLW\ RI ,OOLQRLVr 7KH FRQYHQLHQFH RIIHUHG E\ WKH SORWWLQJ SDFNDJH FDQQRW EH RYHUHPSKDVL]HG 7KH SORWV ZHUH YLHZHG RQ D 7HNWURQL[ WHUPLQDO DQG KDUGFRSLHV ZHUH PDGH RQ 7HNWURQL[ SORWWHUV 0RGHOLQJ RI &RQGXFWLRQ +HDWLQJ LQ &DQV 7KH SDUWLDO GLIIHUHQWLDO HTXDWLRQ 32(f JRYHUQLQJ FRQGXFWLRQ KHDWLQJ RI D F\OLQGULFDOVKDSHG VROLG ZDV GLVFUHWL]HG DQG ILQLWH r: ( 'XQQ 0HFKDQLFDO DQG ,QGXVWULDO (QJLQHHULQJ 'HSDUWPHQW 8QLYHUVLW\ RI ,OOLQRLV DW 8UEDQD&KDPSDLJQ 3HUVRQDO &RPPXQLFDWLRQ

PAGE 98

GLIIHUHQFHG XVLQJ DQ DOWHUQDWLQJ GLUHFWLRQ H[SOLFLW $'(f GLIIHUHQFLQJ PHWKRG ZKLFK LV H[SOLFLW EXW VWLOO XQFRQGLWLRQDOO\ VWDEOH 7KH WUDQVLHQW WHPSHUDWXUH YDOXHV SUHGLFWHG E\ SHUIRUPDQFH RI WKLV $'( ILQLWH GLIIHUHQFH PHWKRG ZDV FRPSDUHG ZLWK DQDO\WLFDO VROXWLRQV DQG H[SHULPHQWDO PHDVXUHPHQWV 'LVFUHWL]DWLRQ DQG 6ROXWLRQ RI WKH (TXDWLRQ .HHSLQJ LQ PLQG WKDW WKH HTXDWLRQ ZDV QHHGHG WR EH VROYHG LQ UHDOWLPH IRU RQOLQH FRQWURO SXUSRVHV DQG WKH $'( PHWKRG ZDV DOUHDG\ WHVWHG 7HL[HLUD HW DO 7HL[HLUD f WKH VDPH ZDV VHOHFWHG 7KH $'( IRUPXODWLRQ ZDV ILUVW UHSRUWHG E\ 6DXOn\HY f 7KLV ZDV ODWHU WUDQVODWHG DQG GHVFULEHG E\ $OLDGD DQG 4XRQ f DQG ZDV ILUVW LQWURGXFHG IRU DSSOLFDWLRQV WR WKHUPDO SURFHVV VLPXODWLRQ E\ 0DQVRQ HW DO f 7KH PHWKRG DV DSSOLHG WR WKH GLIIXVLRQ HTXDWLRQ LV XQFRQGLWLRQDOO\ VWDEOH 6DXOn\HY f IRU WHPSHUDWXUHVSHFLILHG ERXQGDU\ FRQGLWLRQ ,W KDV D IRUPDO RUGHU RI DFFXUDF\ RI $W $U $] f %HLQJ H[SOLFLW LW GRHV QRW UHTXLUH LQYHUVLRQ RI PDWULFHV DV ZRXOG EH UHTXLUHG E\ DQ LPSOLFLW PHWKRG ,Q D IOXLG IORZ SUREOHP WKH $'( IRUPXODWLRQ LV JHQHUDOO\ QRW SUHIHUUHG VLQFH WKH SUHVHQFH RI WKH FRQYHFWLRQ WHUP LQ DQ\ GLIIHUHQFH IRUPf PDNHV WKH FRPSOHWH IRUPXODWLRQ XQVWDEOH RU UHWXUQV LW WR VLPSOH H[SOLFLW VWDELOLW\ OLPLWV ‘ X A IRU D RQHn GLPHQVLRQDO SUREOHP 5RDFKH f $QRWKHU VKRUWFRPLQJ RI WKLV $'( PHWKRG LV WKH IDFW WKDW HYHQ WKRXJK LW LV DQ H[SOLFLW PHWKRG LW LV DFWXDOO\ LPSOLFLW LQ ERXQGDU\ FRQGLWLRQV 7R KDQGOH WKLV LPSOLFLWn QHVV HLWKHU H[SOLFLW H[SUHVVLRQV IRU WKH WHPSHUDWXUH YDOXHV QHHG WR

PAGE 99

EH GHULYHG DOJHEUDLFDOO\ IRU HDFK RI WKH ERXQGDULHV RU WKH FRPSXWHU KDV WR VROYH D VLPXOWDQHRXV VHW RI HTXDWLRQV DW WKH ERXQGDULHV VHH $SSHQGL[ f )RU 5RELQnV ERXQGDU\ FRQGLWLRQ WKH DSSOLFDWLRQ RI WKH $'( PHWKRG UHTXLUHG VHSDUDWH DOJHEUDLF H[SUHVVLRQV IRU WKH WRS VLGH DQG ERWWRP ZDOO ERXQGDULHV 6HSDUDWH H[SUHVVLRQV ZHUH DOVR UHTXLUHG IRU WKH WRS DQG ERWWRP FRUQHUV DQG WKH FHQWHUOLQH 7KLV LV LQ VKDUS FRQWUDVW ZLWK WKH VLWXDWLRQ LI WKH $', PHWKRG ZHUH XVHG $GMXVWPHQWV ZRXOG EH QHFHVVDU\ DW WKH ERXQGDULHV IRU WKH $', PHWKRG EXW IRU XVH RI WKH $'( PHWKRG LW EHFDPH FOXPV\ VHH DSSHQGL[ 'f )RU WHPSHUDWXUH VSHFLILHG DW WKH ERXQGDU\ WKLV LV WULYLDO DQG QR DGGLWLRQDO PDQLSXODWLRQV DV QRWHG DERYH DUH QHHGHG ([SHULPHQWV FRQGXFWHG ODWHU VKRZHG WKDW EHFDXVH RI WKH ORZ WKHUPDO FRQGXFWLYLW\ RI IRRG ZKLFK LV PRVWO\ ZDWHUf WKH %LRW QXPEHU K5Nf LV TXLWH ODUJH DQG LW LV WKH LQWHUQDO FRQGXFWLRQ UHVLVWDQFH WKDW LV OLPLWLQJ 7KXV D FRQYHFWLRQ ERXQGDU\ 5RELQnVf FRQGLWLRQ FDQ EH DSSUR[LPDWHG E\ D WHPSHUDWXUH VSHFLILHG 'LULFKOHWf ERXQGDU\ FRQGLWLRQ IRU WKLV SXUSRVH 7KH WZR VKRUWFRPLQJV RI WKH $'( IRUPXODWLRQ DUH WKHUHIRUH RI OLWWOH FRQVHTXHQFH LQ WKH FDVH RI FRQGXFWLRQKHDWHG IRRG 7KH HTXDWLRQV GHYHORSHG IRU WKH FRQYHFWLRQ ERXQGDU\ FRQGLWLRQ DUH JLYHQ LQ DSSHQGL[ &RGLQJ RI WKH &RPSXWHU 3URJUDP 7KH $'( IRUPXODWLRQ IRU WUDQVLHQW WHPSHUDWXUH FDOFXODWLRQ XVLQJ WHPSHUDWXUH VSHFLILHG ERXQGDU\ FRQGLWLRQ ZDV FRGHG LQ 3$6&$/ DQG XVHG DV D VXEURXWLQH LQ D UHDO WLPH FRQWURO DOJRULWKP WR EH GLVFXVVHG ODWHUf

PAGE 100

([SHULPHQWDO 6WXGLHV ([SHULPHQWV ZHUH SHUIRUPHG WR UHFRUG WKH WUDQVLHQW WHPSHUDWXUH YDOXHV LQVLGH F\OLQGULFDO FDQV GXULQJ FRQGXFWLRQ KHDWLQJ 7KH HIIHFW RI GLIIHUHQW VXUIDFH KHDW WUDQVIHU FRHIILFLHQWV WKDW ZHUH REWDLQHG E\ XVLQJ GLIIHUHQW PHGLD VWHDP DQG ZDWHUf RQ WKH WUDQVLHQW WHPSHUDWXUH UHVSRQVH ZDV DOVR QRWHG 7KH H[SHULPHQWDO ZRUN FRQVLVWHG RI VHULHV RI KHDW SHQHWUDWLRQ WHVWV RQ 1R FDQV ILOOHG ZLWK D b EHQWRQLWH VXVSHQVLRQ DV D FRQGXFWLRQKHDWLQJ IRRG PRGHO V\VWHP 1LHNDPS HW DO f (LJKW FDQV ZHUH ILWWHG ZLWK DQ (FNOXQGW\SH FRSSHUFRQVWDQWDQ WKHUPRFRXSOH PRXQWHG PLGZD\ XS WKH FDQ ZDOO VR DV WR KDYH WKH MXQFWLRQV ORFDWHG DW WKH JHRPHWULF FDQ FHQWHU ,Q RUGHU WR DFKLHYH WKH WZR GLIIHUHQW OHYHOV RI KHDGVSDFH YDFXXP IRXU RI WKH FDQV ZHUH ILOOHG DW r& DQG WKH RWKHU IRXU FDQV ZHUH ILOOHG DW r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

PAGE 101

&22/,1* :$7(5 )LJXUH ([SHULPHQWDO VHWXS IRU FRQGXFWLRQ KHDWLQJ

PAGE 102

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r&f DQG SURFHVV WLPH PLQXWHVf ZDV XVHG LQ DOO IRXU SURFHVVHV 7KH\ GLIIHUHG LQ WKH W\SH RI KHDWLQJ PHGLXP XVHG SXUH VWHDP RU SUHVVXUL]HG ZDWHUf DQG LQ WKH UHWRUW SUHVVXUH GXULQJ FRROLQJ HLWKHU KHOG FRQVWDQW DW WKH OHYHO UHTXLUHG GXULQJ KHDWLQJ RU DOORZHG WR GURS VXGGHQO\ WR QHDU DWPRVSKHUH ZLWK WKH LQWURGXFWLRQ RI FRROLQJ ZDWHUf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

PAGE 103

UHVXOWLQJ LQ SURGXFW PRYHPHQW DQG PXFK IDVWHU FRROLQJ UDWHV 'HWDLOV RI WKH ZRUN WKDW VKRZHG GLUHFW HYLGHQFH RI LQWHUQDO SURGXFW PRYHPHQW GXULQJ VXGGHQ GHSUHVVXUL]DWLRQ RI FRQGXFWLRQKHDWHG IRRGV DUH SUHVHQWHG HOVHZKHUH 7HL[HLUD HW DO f 7KH WUDQVLHQW WHPSHUDWXUH UHVSRQVH UHFRUGHG DW WKH FDQ FHQWHU GXULQJ FRQGXFWLRQ KHDWLQJ ZDV FRPSDUHG ZLWK WKH WHPSHUDWXUH DW WKH FDQ FHQWHU SUHGLFWHG E\ WKH $'( DQDORJ WR WKH KHDW FRQGXFWLRQ HTXDWLRQ HTXDWLRQ f 7KH VDPH ZDV DOVR FRPSDUHG ZLWK WKH DQDO\WLFDO VROXWLRQ WR WKH HTXDWLRQ KHDW FRQGXFWLRQ HTXDWLRQf 7R VWXG\ WKH HIIHFW RI VXUIDFH KHDW WUDQVIHU FRHIILFLHQWV WKH WUDQVLHQW WHPSHUDWXUH GDWD ZHUH SORWWHG RQ VHPLORJ SDSHU ZLWK 7 75f 7 75f RQ WKH ORJ VFDOH DQG WLPH RQ WKH OLQHDU VFDOH )URP HTXDWLRQ WKH LQYHUVH VORSH RI WKLV JUDSK LV WKH IYDOXH 7KH GLIIHUHQW LQYHUVH VORSHV IYDOXHVf WKXV FDOFXODWHG IRU KHDWLQJ ZLWK VWHDP DQG KHDWLQJ DQG FRROLQJ ZLWK ZDWHU ZHUH FRPSDUHG $ KLJKHU VXUIDFH KHDW WUDQVIHU FRHIILFLHQW ZDV H[SHFWHG WR UHVXOW LQ ORZHU IYDOXHV VWHHSHU VORSHVf DQG YLFH YHUVD 8VH RI &RQGXFWLRQ +HDWLQJ 0RGHO IRU 2QOLQH &RQWURO 7KH ILQLWH GLIIHUHQFH PRGHO GHYHORSHG IRU SUHGLFWLQJ WHPSHUn DWXUH 7Wf ZDV LQFRUSRUDWHG LQ WKH VRIWZDUH IRU WKH RQOLQH FRQWURO 8VLQJ WKH WHPSHUDWXUH DW WKH FDQ FHQWHU 7Wf )Wf ZDV PRQLWRUHG LQ UHDOWLPH ZLWK PHDVXUHG FDQ ERXQGDU\ UHWRUWf WHPSHUDWXUH 75Wf 7KH SURFHVV ZRXOG FRQWLQXH XQWLO WLPH W5 VXFK WKDW )R\ )RG f

PAGE 104

WKXV VDWLVI\LQJ WKH GHVLUHG VWHULOL]DWLRQ YDOXH IRU DUELWUDU\ YDULDWLRQV LQ UHWRUW WHPSHUDWXUH 7AWf $V QRWHG HDUOLHU WKH FRPSXWHU QHHGV WR VLPXODWH FRROLQJ DQG SHUIRUP RWKHU FDOFXODWLRQV LQ UHDOWLPH )LJXUH VKRZV WKH WLPLQJV RI YDULRXV FDOFXODWLRQV WKDW QHHG WR JR RQ LQ UHDOWLPH )RU WKH VLPXODWLRQ WR EH IHDVLEOH LQ UHDOWLPH WKH UHTXLUHPHQW WKDW $WM $W $W f KDV WR EH VDWLVILHG 7KHUH LV VRPH FKRLFH RYHU VHOHFWLRQ RI DOO WKUHH WLPH LQWHUYDOV $WA $W! $W +RZHYHU $W FDQQRW EH LQFUHDVHG WRR PXFK 7KLV FRXOG PDNH ILQLWH GLIIHUHQFH FDOFXODWLRQV RI 7Wf IURP 7AWf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n HQFH PHWKRGV WKDW FRXOG WDNH ODUJH WLPH LQFUHPHQWV ZKLOH XVLQJ DV OLWWOH FRPSXWDWLRQ WLPH DV SRVVLEOH WR FDOFXODWH HDFK VWHS 7KH DOWHUQDWLQJ GLUHFWLRQ H[SOLFLW $'(f ILQLWH GLIIHUHQFH PHWKRG XVHG E\ 7HL[HLUD DQG 0DQVRQ f GRHV RIIHU ERWK RI WKHVH DGYDQWDJHV

PAGE 105

&DOFXODWH 7Wf XSGDWH IAWf 6LPXODWH FRROLQJ WR HVWLPDWH )&rr $W + W PHDVXUH 7fWf W$W 7 PHDVXUH 75W$Wf ZKHUH $W 7LPH WDNHQ WR FDOFXODWH 7Wf IURP PHDVXUHG 7fWf XVLQJ ILQLWH GLIIHUHQFH DSSUR[LPDWLRQ WR HTXDWLRQ $W 7LPH WDNHQ WR VLPXODWH FRROLQJ DVVXPLQJ LW VWDUWV DW  WLPH LQVWDQW W DQG $W 7LPH LQWHUYDO EHWZHHQ UHDGLQJ UHWRUW WHPSHUDWXUHV 75Wf )LJXUH 7LPLQJV RI YDULRXV UHDO WLPH FRPSXWDWLRQV IRU RQOLQH FRQWURO RI FRQGXFWLRQKHDWHG IRRG

PAGE 106

,W LV XQFRQGLWLRQDOO\ VWDEOH WKXV DOORZLQJ ODUJH WLPH VWHSV ,W LV DOVR H[SOLFLW WKXV WDNLQJ VPDOO FRPSXWDWLRQ WLPH SHU VWHS DV FRPSDUHG WR PDQ\ LPSOLFLW PHWKRGV 7KH WLPH LQFUHPHQWV $W $WA $WA DQG WKH ILQLWH GLIIHUHQFH FDOFXODWLRQV ZHUH DGMXVWHG VR WKDW WKH LQHTXDOLW\ f ZDV VDWLVILHG 2Q D '(& 3'3 WKH IROORZLQJ FRPSXWDWLRQ WLPHV ZHUH IRXQG WR EH IHDVLEOH $W VHFRQGV $W VHFRQGV $WS VHFRQGV 7KH FRQWURO ORJLF IORZ GLDJUDP IRU WKH UHWRUW WKHUPDO SURFHVV LV JLYHQ LQ ILJXUH 7KH LQSXW GDWD DUH ILUVW FKHFNHG DJDLQVW VSHFLILFDWLRQV IRU WKH SURGXFW ,I WKH\ DJUHH VWHDP LV WXUQHG RQ DQG WKH FRPSXWHU FRPSOHWHV WKH YHQWLQJ F\FOH 7KH UHWRUW WHPSHUn DWXUH FRPHV XS DQG WKH FRPSXWHU WKURXJK D FRQWUROOHU LQWHUIDFH DWWHPSWV WR PDLQWDLQ WKH GHVLJQ UHWRUW WHPSHUDWXUH 7A $V WKH KHDWLQJ F\FOH FRQWLQXHV UHWRUW WHPSHUDWXUH 75Wf LV UHDG DW LQWHUYDOV RI $W 7Wf LV WKHQ FDOFXODWHG IURP 75Wf XVLQJ WKH ILQLWH GLIIHUHQFH DSSUR[LPDWLRQ RI HTXDWLRQ 8VLQJ HTXDWLRQ D )Wf LV WKHQ FDOFXODWHG 7KH LQSXW GDWD LQFOXGHV D VSHFLILHG SKHDW A ZKLFK LV WKH ) YDOXH QRUPDOO\ DFKLHYHG DW D WLPH ZKHQ KHDWLQJ ZRXOG EH WHUPLQDWHG IRU D GHVLJQ WRWDO ) RI ) A :KHQ RR )Wf H[FHHGV WKH YDOXH SAHDW W WKH FRPSXWHU DOVR VLPXODWHV WKH FRROLQJ F\FOH LQ DGGLWLRQ WR FDOFXODWLQJ 7Wf DQG )Wf ,I WKH )Wf DFFXPXODWHG VR IDU WRJHWKHU ZLWK WKH VLPXODWHG FRQWULEXWLRQ IURP FRROLQJ H[FHHGV WKH GHVLJQ WRWDO )4 YDOXH IRU WKH SURFHVV )Tf LH ZKHQ

PAGE 107

)LJXUH )ORZ GLDJUDP IRU FRPSXWHU FRQWURO RI UHWRUW RSHUDWLRQV ZLWK RQOLQH FRUUHFWLRQ RI SURFHVV GHYLDWLRQV

PAGE 108

I: )R AVLPXODWHG } )R LV VDWLVILHG WKH FRPSXWHU WXUQV RII WKH VWHDP DQG OHWV LQ FRROLQJ ZDWHU 7KH FRPSXWHU VWLOO NHHSV UHDGLQJ WKH UHWRUW WHPSHUDWXUH 7UWf DQG FDOFXODWLQJ 7Wf DQG )Wf :KHQ 7Wf WKH FDOFXODWHG WHPSHUDWXUH DW WKH FDQ FHQWHU LV EHORZ D FHUWDLQ VSHFLILHG YDOXH FRROLQJ LV HQGHG E\ VWRSSLQJ WKH FRROLQJ ZDWHU IORZ DQG WKH ZDWHU LV GUDLQHG SULRU WR XQORDGLQJ WKH UHWRUW $W WKH HQG RI WKH SURFHVV D FRPSOHWH GRFXPHQWDWLRQ RI PHDVXUHG UHWRUW WHPSHUDWXUH KLVWRU\ 7SWf FDOFXODWHG FDQ FHQWHU WHPSHUDWXUH KLVWRU\ 7Wf DQG WKH DFFRPSOLVKHG )4Wf LV NHSW RQ ILOH DQG FDQ EH GRFXPHQWHG LQ ERWK WDEXODU DQG JUDSKLFDO IRUP 7KH FRQWURO DOJRULWKP SUHVHQWHG LQ ILJXUH LPSOHPHQWHG LQ UHDOWLPH ZLWK D FRPSXWHU FDQ HQVXUH GHVLJQHG VWHULOL]DWLRQ )A LQ D SURGXFW IRU DUELWUDU\ YDULDWLRQV LQ UHWRUW WHPSHUDWXUH 7KH SHUIRUPDQFH RI WKLV V\VWHP IRU FRUUHFWLQJ GHYLDWLRQV ZDV FRPSDUHG ZLWK SURFHVV FRUUHFWLRQV SUHGLFWHG E\ WKH FRUUHFWLRQ IDFWRU &)f PHWKRG RI *LDQQRQL6XFFDU DQG +D\DNDZD f &RPSDULVRQ RI WKH WZR PHWKRGV ZDV EDVHG RQ f HVWLPDWHG SURFHVV WLPHV DQG f HVWLPDWHG DFFRPSOLVKHG ) YDOXHV 7KUHH GLIIHUHQW SURFHVVHV ZHUH FRQVLGHUHG f UHIHUHQFH SURFHVV ZLWK QR GHYLDWLRQ f VWHS GURS DQG ODWHU VWHS ULVH LQ SURFHVV WHPSHUDWXUH DQG f JUDGXDO GURS DQG ODWHU VWHS ULVH LQ WHPSHUDWXUH

PAGE 109

&RUUHFWHG SURFHVV WLPHV IRU HDFK RI WKHVH WKUHH SURFHVVLQJ VLWXDWLRQV DUH ILUVW QRWHG IRU WKH RQOLQH FRPSXWHU FRQWURO PHWKRG 7KHVH DUH WKH WLPHV WA ZKHQ UROK FRROA nKR nVLPXODWHG ) G R LV VDWLVILHG 7KH FRUUHFWHG SURFHVVHV DV SUHGLFWHG E\ WKH &) PHWKRG ZHUH IRXQG E\ PXOWLSO\LQJ WKH QRUPDO SURFHVV WLPH ZLWK WKH FRUUHFWLRQ IDFWRUV SUHGLFWHG E\ WKDW PHWKRG 7KH DFFRPSOLVKHG )4 YDOXHV IRU WKH SURFHVVHV FRUUHFWHG E\ WKH RQOLQH FRQWURO V\VWHP DQG SURFHVVHV FRUUHFWHG E\ WKH &) PHWKRG ZHUH FDOFXODWHG XVLQJ HTXDWLRQ D WKH JHQHUDO PHWKRGf ,Q HLWKHU FDVH WKH FDQ FHQWHU WHPSHUDWXUHV UHTXLUHG IRU XVH ZLWK HTXDWLRQ D ZHUH SUHGLFWHG XVLQJ WKH $'( ILQLWH GLIIHUHQFH DQDORJ RI HTXDWLRQ

PAGE 110

5(68/76 $1' ',6&866,21 5HVXOWV IRU WKH QXPHULFDO QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU PRGHO DUH SUHVHQWHG DQG GLVFXVVHG ILUVW 1H[W QXPHULFDO DQG H[SHULn PHQWDO UHVXOWV DUH SUHVHQWHG IRU FRQGXFWLRQ KHDW WUDQVIHU LQ FDQQHG IRRGV )LQDOO\ SHUIRUPDQFH RI WKH RQOLQH FRQWURO ORJLF XVLQJ ILQLWH GLIIHUHQFH FRQGXFWLRQ KHDW WUDQVIHU PRGHOf LV GHPRQVWUDWHG IRU FRQGXFWLRQKHDWHG IRRGV 1DWXUDO &RQYHFWLRQ +HDWLQJ LQ &DQQHG )RRGV 7KH UHVXOWV RI WKH QXPHULFDO PRGHO IRU QDWXUDO FRQYHFWLRQ KHDW WUDQVIHU DUH GHVFULEHG KHUH ZLWK ZDWHU DV D PRGHO OLTXLG IRRG 8QOHVV RWKHUZLVH QRWHG WKH LQSXW GDWD ILOH IRU WKH UHVXOWV LV JLYHQ LQ $SSHQGL[ & 'HYHORSPHQW RI WKH IORZ SDWWHUQV DQG WHPSHUDWXUH SURILOHV DUH ILUVW GLVFXVVHG DQG GLDJUDPV RI WKH VDPH DUH JLYHQ DW YDULRXV WLPHV GXULQJ KHDWLQJ 7KH GLDJUDPV DUH TXDOLWDWLYHO\ DQG TXDQWLWDWLYHO\ FRPSDUHG ZLWK SUHYLRXV QXPHULFDO VWXGLHV DQG DYDLODEOH H[SHULPHQWDO GDWD /RFDWLRQ DQG PLJUDWLRQ RI WKH VORZHVW KHDWLQJ SRLQWV LQVLGH WKH FDQ DUH WKHQ GLVFXVVHG $OVR YDULDWLRQ RI WKH VORZHVW KHDWLQJ WHPSHUDWXUH ZLWK WLPH ZDV FRPSDUHG ZLWK WKDW RI WUDQVLHQW WHPSHUDWXUH DW RWKHU ORFDWLRQV )LQDOO\ FRPPHQWV DUH LQFOXGHG RQ WKH QXPHULFDO DVSHFWV RI WKH PRGHO

PAGE 111

7KURXJKRXW WKLV VHFWLRQ UHIHUHQFH ZLOO EH IUHTXHQWO\ PDGH WR WKH H[SHULPHQWDO VWXG\ RI +LGGLQN f PHQWLRQHG LQ WKH 5(9,(: 2) /,7(5$785( VHFWLRQ 6LQFH KLV VWXG\ LQYROYHG PRUH WKDQ RQH ZDOO WHPSHUDWXUH DQG RWKHU FRQGLWLRQV WKH SDUWLFXODU FRQGLWLRQV DSSOLn FDEOH WR DQ\ VLWXDWLRQ DUH QRWHG DORQJVLGH 5HIHUHQFH ZLOO DOVR EH PDGH WR WKH QXPHULFDO ILQLWH HOHPHQW VWXG\ RI 6DQL RQ WHPSHUDWXUH DQG YHORFLW\ SURILOHV GXULQJ SDVWHXUL]DWLRQ RI FDQQHG EHHU DOVR PHQWLRQHG LQ WKH 5(9,(: 2) /,7(5$785( VHFWLRQf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f %UDQGRQ HW DO f DQG +LGGLQN f 7KH DUURZ SORWV RI YHORFLW\ YHFWRUV FUHDWH D SLFWXUH WKDW LV YHU\ EXV\ DQG LW LV RIWHQ KDUG WR GLIIHUHQWLDWH EHWZHHQ WKH IORZ

PAGE 112

SDWWHUQV DFFXUDWHO\ 6WUHDPOLQHV RQ WKH RWKHU KDQG FDQ SURYLGH DOO WKH HVVHQWLDO GHWDLOV RI WKH IORZ SDWWHUQV ZLWKRXW WKH FOXPVLQHVV 7KLV LV SDUWLFXODUO\ WUXH LQ WKH LQLWLDO WLPHV ZKHQ YHORFLWLHV DUH PDQ\ WLPHV ODUJHU QHDU WKH ERXQGDU\ OD\HU WKDQ DW WKH FRUH 7KLV FUHDWHG D VFDOLQJ SUREOHP IRU YHFWRU SORWV $OVR LQ WKH ERXQGDU\ OD\HU JULG SRLQWV DUH YHU\ FORVH WRJHWKHU 7KXV IRU PRVW RI WKH ILJXUHV VWUHDPOLQHV DUH SUHVHQWHG LQVWHDG RI YHORFLW\ YHFWRUV $W PXFK ODWHU WLPHV W VHFRQGVf ZKHQ WKH IORZ DSSURDFKHG VWHDG\ VWDWH DQG WKH YHORFLW\ PDJQLWXGHV EHFDPH FRPSDUDEOH LQ WKH FRUH DQG LQ WKH ERXQGDU\ OD\HU YHORFLW\ SORWV ZRXOG EH YLVXDOO\ PRUH PHDQLQJIXO VHH ILJXUH Rf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f f UHJLRQ 2]LVLN f

PAGE 113

7DEOH 6HTXHQFH RI OLQH W\SHV IRU LGHQWLILFDWLRQ RI VWUHDP IXQFWLRQ FRQWRXUV LQ ILJXUHV O2DR

PAGE 114

6,'(:$// %27720 )LJXUH ,VRWKHUPV QHDU ERWWRP ZDOO SUHGLFWHG E\ WKH FRQYHFWLYH PRGHO FRPSDUHG ZLWK WKH SUHGLFWLRQV FRQVLGHULQJ FRQGXFWLRQ HTXDWLRQ\M RQO\

PAGE 115

7]Wf 7[ f§Mf§M n HUI ‘ RLW f ZKHUH HUI LV WKH HUURU IXQFWLRQ ,VRWKHUPV DQG VWUHDPOLQHV GXULQJ WKH ILUVW WHQ VHFRQGV RI KHDWLQJ DUH JLYHQ LQ ILJXUHV O2DM 7KHVH ILJXUHV O2DMf UHSUHVHQW UHVXOWV RI FRPSXWDWLRQV XVLQJ D f[fr JULG ,QVWDELOLW\ UHVXOWHG GXH WR WKH SUHVHQFH RI KHDWHG ERWWRP OD\HUV ZLWK FROGHU OLTXLG DERYH WKHP 7KLV LQVWDELOLW\ LQ WKH ERWWRP OD\HU JDYH ULVH WR EXUVWV RI FRQYHFWLYH FHOOV LQ D YHU\ UHJXODU PDQQHU 6XFK SKHQRPHQD DUH UHIHUUHG WR DV %HQDUG FRQYHFWLRQ +RZHYHU EHFDXVH RI WKH YHU\ ODUJH LQLWLDO WHPSHUDWXUH GLIIHUHQFH WKLV FRQYHFWLRQ RFFXUUHG YHU\ UDSLGO\ LQ WKH SUHVHQW VWXG\ ILJXUHV O2HMf DV FRPSDUHG WR LWV PXFK VORZHU RFFXUUHQFH UHSRUWHG E\ 6K\X DQG +VLHK LQ SUHVVf 7KH IRUPDWLRQ RI WKHVH FRQYHFWLYH FHOOV ZHUH UHVSRQVLEOH IRU WKH FRQYHFWLYH KHDW WUDQVIHU IURP WKH ERWWRP ,Q WKH VWXG\ RI +LGGLQN f WKHVH ZHUH SUHFLVHO\ WKH FRQYHFWLYH IORZ SDWWHUQV WKDW FRXOG QRW EH SLFNHG XS E\ KLV QXPHULFDO PRGHO $ORQJ WKH VLGH ZDOO KRW OLTXLG ULVHV DQG JHWV GHSRVLWHG DW WKH WRS 7KLV H[SODLQV WKH VKDSH RI LVRWKHUPV DW WKH WRS FRUH UHJLRQ 7UDQVLHQW LVRWKHUPV DQG VWUHDPOLQHV DW ODWHU WLPH LQVWDQWV RI DQG VHFRQGV RI KHDWLQJ DUH SUHVHQWHG LQ ILJXUH O2NQ 7KHVH UHVXOWV ILJXUHV O2NQf ZHUH FRPSXWHG XVLQJ f[f JULG 7ZR PDLQ IORZ ORRSV FRXOG EH VHHQ 7KH ODUJHU r7KH YDOXHV DQG LQ SDUHQWKHVLV GHQRWH WKH VWUHWFKLQJ SDUDPHWHU VLJPD VHH HTXDWLRQV DQG f LQ ] DQG U GLUHFWLRQ UHVSHFWLYHO\

PAGE 116

:DWHU )LJXUH D ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQG RI KHDWLQJ :DWHU 7 r& WRSr& VOGD r& ERWr& 5FP K FP 5 = IL f§% )LJXUH E ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 117

:DWHU 7 n& WRSr& VOGH r& ERWr& 5RP + RP 5 = DRRRRL )LJXUH F ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ :DWHU 7 r& WRSLHLr& VOGH r& ERLn& 5FP K FP 5 DRRRRL $IWHU VHF )LJXUH O2G ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ R 2

PAGE 118

:DWHU 7 r& ORSO(I& VOGHrr& ERWr& 5FP +VrRP 5 = DRRRRL HR )LJXUH O2H ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ )LJXUH O2I ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 119

:DWHU 7!r& RSr& VOGR r& ERr& 5FP + RP 5 = f§ )LJXUH O2J ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ $IWHU VHF :DWHU U r& WRSr& VOGH r& ERWr& 5FP + RP 5 = )LJXUH O2K ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 120

:DWHU 7 r& }RSr& 6OGH}r& ER,r& 5FP + RP 5 )LJXUH O2L ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ :DWHU 7 n& WRSn& VOGH r& ERW n& 5 FP K FP 5 = $ WHU VHF )LJXUH O2M ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 121

:DWHU 7r& RSr& VOGH Or& ERr& 5/FP + rRP 5 = DRRRRL )LJXUH N ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ $IWHU VHF :DWHU 7 r& ORSr& VOGD OeMO f& ERWr& 5FP K O 2*RP 5 = )LJXUH ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 122

:DWHU 7 r& WRSn& VOGD r& ERWr& 5FP K RP 5%% = DRRRRW WR $IWHU VHF )LJXUH P ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ :DWHU 7 r& WRSr& VOGD On r& ERWr& 5FP + RP 5%% = % DRRRRL WR )LJXUH O2Q ,VRWKHUPV DQG VWUHDPOLQHV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 123

FRXQWHUFORFNZLVH ORRS H[WHQGHG RYHU PRVW RI WKH FDQ DQG UHSUHVHQWHG WKH SULPDU\ SKHQRPHQD RI XSZDUG PRYHPHQW RI KRW OLTXLG DW WKH VLGH ZDOO DQG LWV GRZQIORZ DW WKH FRUH 6HYHUDO VPDOO FORFNZLVH DQG FRXQWHUFORFNZLVH VHFRQGDU\ ORRSV UHVXOWLQJ IURP ERWWRP KHDWLQJ DSSHDUHG DQG GLHG GRZQ RYHU WLPH $ VPDOOHU FORFNZLVH ORRS ZDV SHUn VLVWHQWO\ SUHVHQW QHDU WKH ERWWRP FHQWHUOLQH 7KLV FORFNZLVH ORRS FRXOG DOVR EH VHHQ LQ WKH QXPHULFDO UHVXOWV RI 6DQL ILJXUH f $W D VWLOO ODWHU WLPH RI W VHFRQGV FDOFXODWHG ZLWK D f[f JULGf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n FLWLHV QHDU WKH ZDOO OLTXLG RQ WKH VROLG ZDOO LV DW UHVW GXH WR WKH QR VOLS ERXQGDU\ FRQGLWLRQf 7KH OLTXLG ULVLQJ GXH WR WKH EXR\DQF\ HIIHFW SURGXFHV YLVFRXV GUDJ RQ DGMDFHQW IOXLG ZKLFK DOVR PRYHV XS

PAGE 124

77f777/ 9 r r rrrAr‹ 9 / -X U r& W 9 rm LF L 8IL ? YW L A L O LO OOOO8) AA B :DWHU 7 r& 7RSO6Of& 6OG r& RUr& 5FP + RP 1 0 $IWHU VHF )LJXUH R ,VRWKHUPV DQG YHORFLW\ YHFWRUV LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 125

7HPSHUDWXUH&Ff Q $[LDO YHORFLW\ PPVHFf LJXUH 3UHGLFWHG UDGLDO YHORFLW\ SURILOH DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ 'LVWDQFH IURP VLGHZDLOPPf )LJXUH 3UHGLFWHG UDGLDO WHPSHUDWXUH SURILOH DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU VHFRQGV RI KHDWLQJ

PAGE 126

)XUWKHU DZD\ IURP WKH ZDOO QHDU WKH FRUH WKH HIIHFW RI SRVLWLYH YHORFLWLHV XSZDUG LV QRW IHOW $ WKHUPDO DV ZHOO DV D K\GURG\QDPLF ERXQGDU\ OD\HU LV WKXV FUHDWHG 'XH WR UHFLUFXODWLRQ RI WKH ZDWHU LQ WKH HQFORVXUH QHJDWLYH YHORFLWLHV DSSHDU LQ WKH FRUH UHJLRQ 6LQFH WKH OLTXLG LV JRLQJ XS RQO\ WKURXJK D VPDOO FURVV VHFWLRQDO DUHD WKH WKLFNQHVV RI WKH ERXQGDU\ OD\HUf ZKLOH LWV GRZQZDUG PRYHPHQW LV RYHU D PXFK ODUJHU DUHD DW WKH FRUH WKH GRZQZDUG YHORFLW\ LV PXFK VPDOOHU 7KLV JDYH ULVH WR WKH YHORFLW\ SURILOH DV LQ ILJXUH 6XFK YHORFLW\ SURILOHV ZHUH QXPHULFDOO\ FRPSXWHG E\ 6K\X f 6DQL SHUVRQDO FRPPXQLFDWLRQ f DQG H[SHULPHQWDOO\ REVHUYHG E\ +LGGLQN f DQG (QJHOPDQ DQG 6DQL f 7KH WKHUPDO ERXQGDU\ OD\HU IRUPHG LV VKRZQ LQ ILJXUH ,W FDQ EH VHHQ WKDW WKH WHPSHUDWXUH GLG QRW GURS FRQWLQXRXVO\ ZLWK GLVWDQFH IURP WKH VLGHZDOO DV ZRXOG EH WKH VLWXDWLRQ IRU QDWXUDO FRQYHFWLRQ IURP D KHDWHG YHUWLFDO ZDOO LQ DQ LQILQLWH PHGLXP VHH ILJXUH f 7KH WHPSHUDWXUH ILUVW GURSSHG DQG URVH VRPHZKDW WR EHFRPH WKH FRUH WHPSHUDWXUH 7KH ULVH LQ WHPSHUDWXUH IURP WKH PLQLPXP YDOXH LQ WKH SURILOH FORVHU WR ZDOOf WR D KLJKHU YDOXH LQ WKH FRUH LV FDXVHG E\ WKH GRZQZDUG IORZ RI KRW IOXLG LQ WKH FRUH IURP WKH WRS 7KH PDJQLWXGH RI WKH YHORFLWLHV FRPSXWHG DW WKH ERXQGDU\ OD\HU ILJXUH f FRPSDUHV H[FHOOHQW ZLWK WKH YHORFLWLHV H[SHULPHQWDOO\ REVHUYHG E\ +LGGLQN f DV VKRZQ LQ ILJXUH )RU WKHVH WZR ILJXUHV WKH LQSXW GDWD XVHG ZHUH UDGLXV FP KHLJKW RI OLTXLG FP WRS OLTXLG VXUIDFH IUHH WKHUPDOO\ LQVXODWHGf DQG VLGHZDOO DQG ERWWRP ZDOO WHPSHUDWXUHV RI r& 7KH PD[LPXP SRVLWLYH YHORFLn WLHV LQ WKH ERXQGDU\ OD\HU FRPSXWHG LQ WKLV VWXG\ DW WKUHH GLIIHUHQW

PAGE 127

)LJXUH ,VRWKHUPDO YHUWLFDO VXUIDFH VKRZLQJ ERXQGDU\ OD\HU ZLWK YHORFLW\ SURILOH X[\f DQG WHPSHUDWXUH SURILOH W[\f LQ ZDWHU +LJJLQV DQG *HEKDUW f

PAGE 128

$[LDO YHORFLW\ $[LDO YHORFLW\ $[LDO YHORFLW\PPVHFf )LJXUH 3UHGLFWHG UDGLDO YHORFLW\ SURILOHV DW WKH VLGHZDOO DW PLG KHLJKW LQ D F\OLQGULFDO FDQ DIWHU YDULRXV KHDWLQJ WLPHV

PAGE 129

YHORFLW\ PPVf )LJXUH 2EVHUYHG UDGLDO YHORFLW\ SURILOHV DW WKH VLGHZDOO DW PLG KHLJKW LQ ZDWHU LQ D F\OLQGULFDO FDQ DIWHU YDULRXV KHDWLQJ WLPHV IURP +LGGLQN f

PAGE 130

WLPHV RI DQG VHFRQGV DUH DQG PPVHFRQG UHn VSHFWLYHO\ 7KH REVHUYHG PD[LPXP SRVLWLYH YHORFLWLHV DW WKH VDPH WLPH LQVWDQWV E\ +LGGLQN f ZHUH DQG PPVHF UHVSHFn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b RI WKH UDGLXV DQG DJUHHV ZLWK REVHUYDWLRQV RI +LGGLQN f 6K\X f DOVR UHSRUWHG ERXQGDU\ OD\HU WKLFNQHVV RI b RI UDGLXV EXW KLV FRPSXWHG YHORFLWLHV ZHUH PXFK VPDOOHU DQG FRXOG EH DWWULEXWHG WR D PXFK VPDOOHU WHPSHUDWXUH GLIIHUHQFH LPSRVHG RQ WKH ZDOO LQ KLV FDVH $[LDO WHPSHUDWXUH SURILOHV 7HPSHUDWXUH YDOXHV DORQJ WKH D[LV RI WKH F\OLQGULFDO FDQ DW GLIIHUHQW WLPHV DUH SUHVHQWHG LQ ILJXUH 7KH D[LDO WHPSHUDWXUH SURILOHV ZHUH DOVR DYDLODEOH IURP WKH H[SHULPHQWDO VWXG\ RI +LGGLQN f 7KHVH DUH SORWWHG DORQJVLGH 7KH LQSXW GDWD IRU ILJXUH DUH UDGLXV FP KHLJKW RI OLTXLG FP WRS VXUIDFH IUHH DQG VLGHZDOO DQG ERWWRP ZDOO WHPSHUDWXUHV RI r& ,W FDQ EH VHHQ WKDW WKH QXPHULFDO PRGHO SUHGLFWV WKH REn VHUYHG YDOXHV ZLWK DERXW b GLVFUHSDQF\ H[FHSW DW VPDOO WLPHV DIWHU KHDWLQJ 4XDOLWDWLYHO\ WRR WKH SUHGLFWHG DQG REVHUYHG VKDSHV RI WKH WHPSHUDWXUH SURILOHV DW GLIIHUHQW WLPHV DUH YHU\ VLPLODU

PAGE 131

r / r fr f§ r Y aa Af§ 7HPSHUDLXUHrFf )LJXUH &RPSDULVRQ RI QXPHULFDOO\ SUHGLFWHG D[LDO WHPSHUDWXUH SURILOHV ZLWK REVHUYHG +LGGLQN f YDOXHV DW YDULRXV WLPHV GXULQJ KHDWLQJ

PAGE 132

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f DUH RI XWPRVW LPSRUWDQFH 7KH OLTXLG DQG WKHUHIRUH WKH EDFWHULD FDUULHG ZLWK LWf DW WKHVH SRLQWV H[SHULn HQFHV PXFK OHVV H[SRVXUH WR KHDW OHVV VWHULOL]DWLRQf WKDQ WKH UHVW RI WKH SURGXFW 6HYHUDO VWXGLHV LQ WKH SDVW %ODLVGHOO +LGGLQN 6WHYHQV f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

PAGE 133

)LJXUH /RFDWLRQV RI WKH VORZHVW KHDWLQJ SRLQWV LQ ZDWHU RYHU D KHDWLQJ SHULRG RI PLQXWHV LQ D F\OLQGULFDO FDQ 1 k ).0&3 / $ )LJXUH 0LJUDWLRQ RI WKH VORZHVW KHDWLQJ SRLQWV LQ ZDWHU RYHU D KHDWLQJ SHULRG RI PLQXWHV LQ D F\OLQGULFDO FDQ

PAGE 134

JHRPHWULF FHQWHU WKH FROGHVW DW DOO WLPHV VHH ILJXUH f 7UDGLn WLRQDOO\ LQ QDWXUDO FRQYHFWLRQKHDWLQJ SURGXFWV WKH SRLQW DW OUG WKH KHLJKW RI WKH FRQWDLQHU RQ LWV D[LV ZDV FRQVLGHUHG WR EH WKH VORZHVW KHDWLQJ SRLQW DW DOO WLPHV 7KLV VWXG\ VKRZV WKDW QRW RQO\ LV WKHUH QR VLQJOH SRLQW WKDW LV FROGHVW DW DOO WLPHV WKH FOXVWHU RI SRLQWV RU WKH VORZHVW KHDWLQJ ]RQHf LV ORFDWHG DZD\ IURP WKH FHQWHUOLQH 7KH FORFNZLVH ORRS IRUPHG DW WKH ERWWRP VHH ILJXUH f IRUFHV OLTXLG XSZDUGV QHDU WKH FHQWHUOLQH 7KLV FORFNZLVH ORRS ZDV DOVR FOHDUO\ VHHQ LQ WKH FDOFXODWLRQV RI 6DQL ILJXUH f %HFDXVH RI WKH KRW OLTXLG JRLQJ XS QHDU WKH FHQWHUOLQH DW WKH ERWWRP WKH WHPSHUDWXUH LV PXFK KLJKHU LQ WKLV UHJLRQ VHH WKH LVRWKHUPV LQ ILJXUHV DQG QHDU ERWWRPf DQG LW LV XQOLNHO\ WKDW WKH ORZHVW WHPSHUDWXUH ZRXOG RFFXU QHDU WKH ERWWRP FHQWHUOLQH $OVR WKHUH LV QR REYLRXV UHDVRQ ZK\ RQH SDUWLFXODU SRLQW ZRXOG VWD\ WKH VORZHVW DW DOO WLPHV +LGGLQN f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n EHWLFDO RUGHU %HWZHHQ WZR FRQVHFXWLYH OHWWHUV RI WKH DOSKDEHW WKH

PAGE 135

HODSVHG WLPH LV VHFRQGV )RU FODULW\ WZR RU PRUH FROG SRLQWV DW WKH VDPH ORFDWLRQ DUH GUDZQ VOLJKWO\ VKLIWHG $V FDQ EH VHHQ WKH PRYHPHQW LV TXLWH UDQGRP DQG QR SDUWLFXODU RUGHU LV DSSDUHQW EXW UHPDLQV ZHOO ZLWKLQ WKH ORZHVW b RI WKH FDQ KHLJKW 7UDQVLHQW WHPSHUDWXUH DW WKH VORZHVW KHDWLQJ SRLQWV 7KH WUDQn VLHQW YDOXHV RI WKH WHPSHUDWXUH DW WKH VORZHVW KHDWLQJ SRLQWV ZHUH WKHQ LQYHVWLJDWHG $Q DYHUDJH RI WKH WHPSHUDWXUHV DW WKUHH VORZHVW KHDWLQJ SRLQWV QHDUHVW WR WKH ERWWRPf ZDV FRQVLGHUHG IRU WKLV SXUSRVH 7KLV DYHUDJH WHPSHUDWXUH WKHUHIRUH GLG QRW FRUUHVSRQG WR DQ\ RQH SDUWLFXODU SK\VLFDO ORFDWLRQ 7KH DYHUDJH VORZHVW KHDWLQJ WHPSHUDWXUH YDOXH ZDV SORWWHG ZLWK WLPH ILJXUH f )RU FRPSDULVRQ WUDQVLHQW WHPSHUDWXUH DW WKH PLG KHLJKW DQG OUG IURP WKH ERWWRPf KHLJKW RQ WKH D[LV LV DOVR SORWWHG 7KH OUG KHLJKW RQ WKH D[LV KDV WUDGLWLRQDOO\ EHHQ WDNHQ WR EH WKH VORZHVW KHDWLQJ SRLQW 4XDOLWDWLYHO\ WKH WUDQVLHQW WHPSHUDWXUH FXUYHV ZHUH DV H[SHFWHG 7KH VORZHVW KHDWLQJ WHPSHUDWXUH E\ GHILQLWLRQ VWD\HG FRQVLVWHQWO\ ORZHU WKDQ WKH RWKHU WZR WHPSHUDWXUHV $OVR WKH VORZHVW KHDWLQJ WHPSHUDWXUH ZDV ULVLQJ FRQVLVWHQWO\ VLQFH KHDW ZDV FRQVWDQWO\ HQWHULQJ WKH FDQ 7KH WUDQVLHQW WHPSHUDWXUH DW WKH OUG KHLJKW VWD\HG FORVHU WKDQ WKH PLG SRLQW WR WKH VORZHVW KHDWLQJ WHPn SHUDWXUH $OWKRXJK WKHUH DUH GHILQLWH GLIIHUHQFHV EHWZHHQ WKH WHPn SHUDWXUHV DW DOO WLPHV WKH PDJQLWXGH RI WKHVH GLIIHUHQFHV LV SUREDEO\ QRW VXIILFLHQWO\ JUHDW WR ZDUUDQW DQ\ FKDQJH IURP WKH WUDn GLWLRQDO SUDFWLFH RI XVLQJ WKH OUG KHLJKW ORFDWLRQ +RZHYHU ZDWHU GXH WR LWV ORZ YLVFRVLW\ KHDWV UDWKHU TXLFNO\ DQG WKLV H[SODLQV WKH OLWWOH YDULDWLRQ RI WHPSHUDWXUH ZLWKLQ WKH FRQWDLQHU )RU D OLTXLG

PAGE 136

U )LJXUH 7UDQVLHQW WHPSHUDWXUH DW YDULRXV ORFDWLRQV GXULQJ KHDWLQJ RI ZDWHU LQ D F\OLQGULFDO FDQ

PAGE 137

PRUH YLVFRXV WKHUH PD\ EH VLJQLILFDQW GLIIHUHQFHV EHWZHHQ WKH VORZHVW KHDWLQJ WHPSHUDWXUH DQG WKH WHPSHUDWXUH DW RWKHU ORFDWLRQV 7KH WUDQVLHQW WHPSHUDWXUHV DW GLIIHUHQW SRLQWV LQ WKH FRQWDLQHU H[SHULHQFH UDSLG IOXFWXDWLRQV GXH WR FKDQJLQJ IORZ SDWWHUQV 7KLV LV LOOXVWUDWHG LQ ILJXUH E\ HQODUJLQJ WKH WLPH VFDOH &RQWUDVW ZLWK &RQGXFWLRQ )RU D WKLQ OLTXLG OLNH ZDWHU QDWXUDO FRQYHFWLRQ VSHHGV XS WKH KHDW WUDQVIHU FRQVLGHUDEO\ DV FRPSDUHG WR D VROLG KDYLQJ VDPH WKHUPDO GLIIXVLYLW\ YDOXHf ZKLFK FDQ KHDW RQO\ E\ FRQGXFWLRQ )LJXUH VKRZV WHPSHUDWXUH SURILOH REWDLQHG LQ ZDWHU DW WKH VORZHVW KHDWLQJ SRLQWV XVLQJ WKH QDWXUDO FRQYHFWLRQ KHDWLQJ PRGHO GHYHORSHG LQ WKLV VWXG\ DQG WKH WHPSHUDWXUH SURILOH REWDLQHG XVLQJ WKH DQDn O\WLFDO KHDW FRQGXFWLRQ HTXDWLRQ HTXDWLRQ Df DW WKH JHRPHWULFDO FHQWHU LQ D VROLG SURGXFW KDYLQJ WKH VDPH WKHUPDO GLIIXVLYLW\ DV WKDW RI ZDWHU KHDWHG XQGHU LGHQWLFDO FRQGLWLRQV $V FDQ EH VHHQ WKH VORZHVW SRLQW LQ WKH QDWXUDO FRQYHFWLRQKHDWHG FDQ UHDFKHG DOn PRVW WKH KHDWLQJ WHPSHUDWXUH r&f LQ DERXW PLQXWHV +RZHYHU WKH VORZHVW SRLQW LQ WKH FRQGXFWLRQ KHDWHG IRRG ZLWK WKH VDPH WKHUPDO SURSHUWLHV UHDFKHG RQO\ r& DIWHU KRXU )RU OLTXLGV PRUH YLVFRXV WKH GLIIHUHQFH LQ KHDW WUDQVIHU PD\ QRW EH VR SURQRXQFHG $ FRQWUDVW RI WKH XQLIRUPLW\ RI WHPSHUDWXUH GLVWULEXWLRQ LQ WKH FRQGXFWLRQ DQG QDWXUDO FRQYHFWLRQ KHDWLQJ VLWXDWLRQV LV SUHVHQWHG LQ ILJXUH 7KH HVVHQWLDO GLIIHUHQFH LV WKDW IRU D XQLIRUP VSHFLILHG ERXQGDU\ WHPSHUDWXUH FRQGXFWLRQ KHDWLQJ FUHDWHV FRPSOHWHO\ XQLIRUP LVRWKHUPV DV H[SHFWHG IURP WKH FRQGXFWLRQ HTXDWLRQ ZKHUHDV IRU

PAGE 138

7 0,' 32,17 21 $;,6 5' 32,17 21 $;,6 6/2:(67 32,17 7LPD RI KRRWLQJ VHFRQGVf )LJXUH 7UDQVLHQW WHPSHUDWXUH DW YDULRXV ORFDWLRQV GXULQJ LQLWLDO WLPHV RI KHDWLQJ RI ZDWHU LQ D F\OLQGULFDO FDQ

PAGE 139

)LJXUH 7HPSHUDWXUH DW WKH VORZHVW KHDWLQJ SRLQW GXULQJ QDWXUDO FRQYHFWLRQ KHDWLQJ FRQWUDVWHG ZLWK WKDW GXULQJ FRQGXFWLRQ KHDWLQJ LQ D F\OLQGULFDO FDQ

PAGE 140

:DWHU 7 r& RSO6Or& VOGH r& ERWr& 5FP +rrRP U}UF WRULWLrF ERULHLr& 5RIIO +DOrP $IWHU VHF )LJXUH ,VRWKHUPV GXULQJ QDWXUDO FRQYHFWLRQ KHDWLQJ FRQWUDVWHG ZLWK WKRVH GXULQJ FRQGXFWLRQ KHDWLQJ LQ D F\OLQGULFDO FDQ

PAGE 141

QDWXUDO FRQYHFWLRQ KHDWLQJ HYHQ WKRXJK PRVW RI WKH FRUH H[KLELWV XQLIRUP UDGLDO WHPSHUDWXUH GXH WR VWUDWLILFDWLRQ WKH V\PPHWU\ DERXW WKH PLG KHLJKW LV ORVW 7KLV LV GXH WR WKH FRQYHFWLYH ORRSV IURP WKH ERWWRP KHDWLQJ FUHDWLQJ LVRWKHUPV FRPSOHWHO\ GLIIHUHQW IURP WKH FRQn GXFWLRQ KHDWLQJ $OVR IRUPDWLRQ RI D ERXQGDU\ OD\HU DW WKH VLGHZDOO LQ QDWXUDO FRQYHFWLRQ KHDWLQJ FOHDUO\ VHSDUDWHV LW IURP WKH FRQGXFWLRQ KHDWLQJ VLWXDWLRQ $VVHVPHQW RI WKH 1XPHULFDO 0HWKRG &RQYHUJHQFH 7KH VROXWLRQ WR WKH VHW RI ILQLWH GLIIHUHQFH HTXDWLRQV VKRXOG DSSURDFK WKH WUXH VROXWLRQ WR WKH SDUWLDO GLIIHUn HQWLDO HTXDWLRQV KDYLQJ WKH VDPH LQLWLDO DQG ERXQGDU\ FRQGLWLRQV DV WKH PHVK LV UHILQHG 7KLV LV UHIHUUHG WR DV FRQYHUJHQFH RI WKH ILQLWH GLIIHUHQFH VROXWLRQV )RU OLQHDU 3'(V /D[nV HTXLYDOHQFH WKHRUHP VWDWHV WKDW IRU D FRQVLVWHQW ILQLWH GLIIHUHQFH DSSUR[LPDWLRQ WR D SURSHUO\ SRVHG LQLWLDO YDOXH SUREOHP VWDELOLW\ LV WKH QHFHVVDU\ DQG VXIILFLHQW FRQGLWLRQ IRU FRQYHUJHQFH 7KH VDPH WKHRUHP KDV QRW EHHQ SURYHQ IRU QRQOLQHDU 3'(V $QGHUVRQ HW DO f 7KXV FRQYHUJHQFH FDQQRW PDWKHPDWLFDOO\ EH SURYHQ IRU WKH SUHVHQW SUREOHP +RZHYHU DV VKRZQ LQ WKH SUHYLRXV VHFWLRQV RI WKLV FKDSWHU WKH D[LDO WHPSHUDWXUH SURILOHV DQG WKH UDGLDO YHORFLW\ SURILOHV FRPSDUHG YHU\ ZHOO TXDQWLWDWLYHO\ ZLWK DYDLODEOH H[SHULPHQWDO GDWD 6HYHUDO SKHQRPHQD UHYHDOHG E\ WKH VROXWLRQ LQFOXGLQJ IRUPDWLRQ RI WKH WHPn SHUDWXUH DQG YHORFLW\ ERXQGDU\ OD\HU VWUDWLILFDWLRQ DQG %HQDUG W\SH FRQYHFWLRQ ZHUH H[SHFWHG DQG KDYH DOO EHHQ REVHUYHG E\

PAGE 142

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f[f JULG DQG f[f JULG ILJXUH f DUH VKRZQ LQ ILJXUHV DQG 7KH QXPEHU LQ SDUHQWKHVLV DUH WKH VWUHWFKLQJ SDUDPHWHU D LQ WKH JULG VWUHWFKLQJ HTXDWLRQV DQG 7KH VWUHDPOLQHV DUH DOPRVW LGHQWLFDO DQG WHPSHUDWXUHV DUH SUHGLFWHG ZLWK OHVV WKDQ b HUURU IRU PRVW RI WKH HQFORVXUH 7KXV UHILQLQJ JULGV EH\RQG f[f GLG QRW EULQJ VLJQLILFDQW DGGLWLRQDO LQIRUPDWLRQ DQG WKH VDPH JULG ZDV VHOHFWHG IRU WKH SUHVHQW VWXG\ 7R IXUWKHU HPSKDVL]H RQ WKH IDFW WKDW XVH RI f[f GLG QRW UHVXOW LQ ORVV RI DQ\ VLJQLILFDQW LQIRUPDWLRQ LQ UHIHUHQFH WR QDWXUDO FRQYHFWLRQ KHDWLQJ RI OLTXLG IRRGV WKH WUDQVLHQW VORZHVW KHDWLQJ WHPSHUDWXUHV FRPSXWHG ZLWK WKH WZR JULG VL]HV RI f[f DQG f[f DUH VKRZQ LQ ILJXUH 7KH ILJXUH VKRZV WKDW WKH WZR JULG V\VWHPV SUHGLFW DOPRVW LGHQWLFDO

PAGE 143

, )LJXUH 'LDJUDP RI f[f JULG RQ WKH OHIW DQG f[f RQ WKH ULJKW

PAGE 144

$IWHU VHF )LJXUH ,VRWKHUPV DIWHU VHFRQGV RI KHDWLQJ FRPSXWHG XVLQJ [ JULG OHIW KDOI RI WKH ILJXUHf FRPSDUHG ZLWK FRPSXWDWLRQV XVLQJ [ JULG ULJKW KDOI RI WKH ILJXUHf

PAGE 145

:DWHU 7 r& WRSL& VGH r& ERW r& 5FP + FP O U 2I &RQWRXUV 6WUHDPOLQHV DIWHU VHFRQGV RI KHDWLQJ FRPSXWHG XVLQJ [ JULG OHIW KDOI RI WKH ILJXUHf FRPSDUHG ZLWK FRPSXWDWLRQV XVLQJ [ JULG ULJKW KDOI RI WKH ILJXUHf )LJXUH

PAGE 146

2 R JR =Lf§ R / &' &W A r r W r f;%f JULG \ f;f JULG L L 7LPH RI KHDWLQJ VHFRQGVf )LJXUH 7UDQVLHQW VORZHVW KHDWLQJ WHPSHUDWXUHV FRPSXWHG ZLWK [ DQG [ JULG QR

PAGE 147

VORZHVW KHDWLQJ WHPSHUDWXUHV DW DOO WLPHV 7KH JULG VL]HV DOVR SUHGLFW LGHQWLFDO WUDQVLHQW WHPSHUDWXUHV DW OUG KHLJKW RQ WKH FHQn WHUOLQH DV VKRZQ LQ ILJXUH 6HOHFWLRQ RI WKH WLPH VWHS $ FRPSUHKHQVLYH VWDELOLW\ DQDn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f LV GHFUHDVHG E\ VHFRQG 7KH WLPH VWHS KLVWRU\ IRU D W\SLFDO UXQ LV JLYHQ LQ ILJXUH $ VLPLODU DOJRULWKP IRU DGMXVWLQJ WKH WLPH VWHS WR OLPLW WKH QXPEHU RI YRUWLFLW\ LWHUDWLRQV ZDV XVHG E\ .HH f IURP WKH ZRUN RI %ULOH\ f 6LQFH LPSOLFLW PHWKRGV JHQHUDOO\ FRQVXPH PRUH &38 WLPH SHU WLPH LQFUHPHQW LW ZDV GHFLGHG WR LQYHVWLJDWH KRZ WKH WLPH VWHSV XVHG LQ WKH $', PHWKRG FRPSDUH ZLWK WLPH VWHSV WKDW FRXOG KDYH EHHQ DFKLHYHG XVLQJ D SRVVLEOH H[SOLFLW ILQLWH GLIIHUHQFH IRUPXODWLRQ IRU WKH

PAGE 148

2 2 /B 2 /B &' &/ e &' \f§ r r r r r r r ;f JULG Y ;f JULG BL L L L 7LPH RI KHDWLQJ VHFRQGVf )LJXUH 7UDQVLHQW WHPSHUDWXUH DW OUG KHLJKW DW WKH FHQWHUOLQH FRPSXWHG ZLWK [ DQG [ JULG

PAGE 149

7,0( ,1&5(0(07 VHFRQGVf 7 7 7 7 7,0( 2) 6,08/$7,21 VHFRQGVf )LJXUH 7LPH VWHS KLVWRU\ GXULQJ D W\SLFDO UXQ RI WKH QXPHULFDO KHDW WUDQVIHU PRGHO IRU QDWXUDO FRQYHFWLRQ *UR

PAGE 150

SUREOHP )RU SXUSRVHV RI FRPSDULVRQ WKH H[SOLFLW IRUPXODWLRQ RI %DUDNDW f IRU WKH VHW RI HTXDWLRQV f ZDV XVHG ,Q WKHLU IRUPXODWLRQ WKH\ ODJJHG WKH YHORFLWLHV X DQG Y LQ WKH FRQYHFWLRQ WHUPV DQG XVHG D EDFNZDUG ILQLWH GLIIHUHQFH IRU WKHVH WHUPV 7KHLU VWDELOLW\ DQDO\VLV VKRZHG WKH IROORZLQJ UHVWULFWLRQV RQ $W $W L $] $=f $Uf f 1RQ GLPHQVLRQDOL]DWLRQ RI WKH YDULDEOHV XVHG LQ WKH SUHVHQW VWXG\ LV GLIIHUHQW IURP WKDW XVHG E\ %DUDNDW f $FFRUGLQJO\ HTXDWLRQV f KDYH EHHQ DGMXVWHG WR FRQIRUP WR WKH SUHVHQW VWXG\ )RU 3U DV LV WKH SUHVHQW VLWXDWLRQ LQHTXDOLW\ f LV PRUH UHVWULFWLYH WKDQ f )RU D QRQXQLIRUP FRRUGLQDWH V\VWHP $U DQG $] WR EH XVHG DUH WKHLU VPDOOHVW YDOXHV LQ WKH JULG V\VWHPV .DnOQH\ GH 5LYDV f +RZHYHU WKH DOORZHG WLPH VWHS $W KHUH LV DOVR D IXQFWLRQ RI JULG SRLQW YHORFLWLHV 7KXV XVLQJ HTXDWLRQ $W DQG WKHUHIRUH $W f LV FDOFXODWHG DW DOO WKH JULG SRLQWV )RU D f[f JULG 0LQ $W L M VHFRQGV

PAGE 151

7KH DFWXDO WLPH VWHS XVHG ZDV VHFRQGV ZKLFK LV WZLFH WKH SRVVLEOH WLPH VWHS IRU WKLV H[SOLFLW PHWKRG &RQYHUJHQFH RI ERXQGDU\ YRUWLFLW\ ,QVXIILFLHQW FRQYHUJHQFH RI ERXQGDU\ YRUWLFLW\ FDQ OHDG WR LQDFFXUDF\ DQG LQVWDELOLW\ LQ WKH VROXWLRQ &RQYHUJHQFH FULWHULD 6f RI fA DQG A ZHUH XVHG IRU WKH ERXQGDU\ YRUWLFLWLHV LQ WZR VHWV RI FRPSXWDWLRQ ZLWK DOO RWKHU LQSXW YDOXHV XQFKDQJHG 7KH WHPSHUDWXUH DQG VWUHDP IXQFWLRQ YDOXHV REWDLQHG IURP WKHVH WZR FRPSXWDWLRQV DUH SUHVHQWHG LQ ILJXUHV DQG $V FDQ EH VHHQ WKHUH LV YHU\ OLWWOH GLIIHUHQFH EHWZHHQ XVLQJ WKH WZR GLIIHUHQW FULWHULD IRU FRQYHUJHQFH 7R EH RQ WKH FRQVHUYDWLYH VLGH A ZDV XVHG LQ DOO WKH FDOFXODWLRQV $ QRWH RQ WKH KDOI JULG SRLQWV &RQIXVLRQ H[LVWV RYHU HYDOXDWLRQ RI KDOI JULG SRLQWV DW WKH FRQWURO YROXPH IDFHV IRU QRQ XQLIRUP JULG V\VWHP ,I ZH XVH WKH KDOI SRLQWV LQ WKH FRPSXWDWLRQDO Q f GRPDLQ DV LV WKH FDVH LI ZH GHULYH LQ WKH IROORZLQJ ZD\ UYHf B Q UYHf B UYHf ML U U Q f Q U f UM U UMO 7KHQ UMO VKRXOG EH GHILQHG E\ U f

PAGE 152

:DIHU 7 r& WRSr& VGD r& ER7r& 5FP + FP /5 r6 /= 4 55 5= $IWHU VHF ,VRWKHUPV LQ D F\OLQGULFDO FDQ FDOFXODWHG ZLWK WZR GLIIHUHQW FRQYHUJHQFH FULWHULD IRU ERXQGDU\ YRUWLFLW\ RQ WKH OHIW DQG RQ WKH ULJKWf )LJXUH

PAGE 153

:DWHU 7 r& WRSr& 6OGHr& E7r& 5 rFP 0a*FP 8 U &RQWRXUV 6WUHDPOLQHV LQ D F\OLQGULFDO FDQ FDOFXODWHG ZLWK WZR GLIIHUHQW FRQYHUJHQFH FULWHULD IRU ERXQGDU\ YRUWLFLW\ RQ WKH OHIW DQG RQ WKH ULJKWf )LJXUH

PAGE 154

+RZHYHU LQ WKH ZRUN RQ NLQHPDWLFDOO\ FRQVLVWHQW YHORFLWLHV E\ 3DUPHQWLHU DQG 7RUUDQFH f WKH KDOI SRLQWV ZHUH PLG SRLQWV EHWZHHQ JULGV LQ SK\VLFDO GRPDLQ LH WKH\ DUH GHILQHG E\ !Y ‘ 99 UMO! f EXW WKH HTXDWLRQV GHYHORSHG LQ WKHLU VWXG\ ZHUH PHDQW IRU DQ XQLIRUP JULG V\VWHP )RU DQ XQLIRUP JULG V\VWHP D LV ]HUR DQG WKH WZR GHILQLWLRQV RI UMO PHUH LQWR RQHr ,W LV LQWHUHVWLQJ WR VHH WKDW DSSDUHQWO\ WKH VDPH HTXDWLRQ f FDQ EH GHULYHG ZLWK WKLV GHILQLWLRQ f RI UMO E\ ZULWLQJ UY4f B 95UM 9"rL Ma 9/UM 9AL -O Y U U U M 9L ‘ UM9 $Uf f AY5UM\"M M Y/ULYLLLU U Q -O -O XVLQJ HTXDWLRQ 7KH FDWFK LV WKDW HTXDWLRQ QRZ WXUQV RXW WR EH ILUVW RUGHU DFFXn UDWH GXH WR VNHZHG FHQWUDO GLIIHUHQFLQJ LQVWHDG RI EHLQJ VHFRQG RUGHU DFFXUDWH IRU UHJXODU FHQWUDO GLIIHUHQFLQJ DV LQ HTXDWLRQ 7KXV XVH RI UM GHIOfQHG E\ HTXDWLRQ WRJHWKHU ZLWK HTXDWLRQ ZRXOG GURS WKH RUGHU RI DFFXUDF\ +RZHYHU LW LV H[SHFWHG WKDW ZLWK VXFK D GHILQLWLRQ WUDQVSRUW YHORFLWLHV FRXOG EH D EHWWHU UHSn UHVHQWDWLYH IRU WKH IOX[HV DW WKH LQWHUIDFH ,W LV XQFOHDU IURP WKHLU ZRUN KRZ VRPH RI WKH HTXDWLRQV IRU FDOFXODWLQJ YHORFLWLHV

PAGE 155

ZRXOG FKDQJH LQ FDVH RI D QRQXQLIRUP FRRUGLQDWH V\VWHP 3DUPHQWLHU UHFRPPHQGHG WKDW HTXDWLRQ EH XVHG HYHQ LQ D QRQXQLIRUP JULG V\VWHP 7KLV PHDQV JULG IDFHV ZRXOG VWLOO EH KDOI ZD\ EHWZHHQ JULG SRLQWV LQ D QRQXQLIRUP FRRUGLQDWH V\VWHP 7KH UHFXUVLRQ UHODWLRQ RI ; HTXDWLRQ f ZRXOG SUREDEO\ FKDQJH ZLWK D QRQXQLIRUP mFRRUGLQDWH V\VWHP +RZHYHU LQ WKLV VWXG\ WKH VDPH [ HTXDWLRQ f ZDV XVHG DV LQ WKH FDVH RI D XQLIRUP FRRUGLQDWH V\VWHP $QRWKHU LQWHUHVWLQJ SRLQW QRWHG E\ 3DUPHQWLHUr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}A( 0 3DUPHQWLHU %R[ QR %URRN 6WUHHW %URZQ 8QLY 3URYLGHQFH 5, 3HUVRQDO &RPPXQLFDWLRQV

PAGE 156

$QDO\VLV RI %RXQGDU\ &RQGLWLRQV 7KH LQYHUVH VORSHV RI WKH KHDW SHQHWUDWLRQ FXUYH REWDLQHG H[SHULPHQWDOO\ IRU YDULRXV KHDWLQJ DQG FRROLQJ SURFHVVHV DUH JLYHQ LQ WDEOH 7KH GDWD SUHVHQWHG DUH DYHUDJHV IRU VHYHUDO VLPLODU SURFHVVHV 7KH VWDQGDUG GHYLDWLRQV IRU WKHVH YDOXHV DPRQJ UHSOLFDWH FDQV DQG UHSOLFDWH SURFHVVHV ZHUH LQ WKH RUGHU RI PLQXWHV 7KH UHVXOWV VKRZ WKDW ILOO WHPSHUDWXUH D PHDVXUH RI KHDG VSDFH YDFXXPf GLG QRW KDYH D VLJQLILFDQW HIIHFW RQ WKH KHDWLQJ RU FRROLQJ UDWHV 7KLV ZDV H[SHFWHG VLQFH HYHQ WKRXJK WKHUH ZDV FRQVLGHUDEOH GLIIHUHQFH LQ YDFXXP SURGXFHG LQ WKH WZR ILOO WHPSHUDWXUHV WKH SK\VLFDO GLPHQn VLRQV RI WKH KHDG VSDFHV ZHUH QRW PXFK GLIIHUHQW 7KH GLIIHUHQFH LQ KHDW WUDQVIHU UDWHV EHWZHHQ KHDWLQJ DQG FRROLQJ ZHUH VHHQ WR EH PRUH SURQRXQFHG WKDQ WKH GLIIHUHQFH EHWZHHQ WZR UDWHV RI KHDWLQJ E\ VWHDP DQG KRW ZDWHU UHVSHFWLYHO\ 7KH GLIn IHUHQFH EHWZHHQ KHDW WUDQVIHU UDWHV GXULQJ KHDWLQJ DQG FRROLQJ ZDV H[SODLQHG SULPDULO\ E\ FKDQJHV LQ WKH WKHUPDO GLIIXVLYLW\ RI WKH SURGXFW EHFDXVH RI FRQVLGHUDEOH GLIIHUHQFH EHWZHHQ WKH DYHUDJH WHPn SHUDWXUHV H[SHULHQFHG GXULQJ KHDWLQJ DQG FRROLQJ 'DWWD HW DO f 7KH VPDOO GLIIHUHQFH EHWZHHQ WKH UDWHV RI KHDWLQJ ZLWK VWHDP DV RSSRVHG WR XVLQJ KRW ZDWHU ZDV DWWULEXWHG WR WKH PXFK ODUJHU UHVLVWDQFH LQVLGH WKH FDQ ODUJH %LRW QXPEHUf GXH WR YHU\ ORZ WKHUPDO FRQGXFWLYLW\ RI IRRGV ZKLFK OLPLWV WKH UDWH RI KHDW WUDQVIHU 7KLV PHDQV WKDW WKH FRQYHFWLRQ ERXQGDU\ FRQGLWLRQ WKDW H[LVWV LQ DQ DFWXDO KHDWLQJ RU FRROLQJ VLWXDWLRQ FDQ EH DSSUR[LPDWHG E\ D VSHFLILHG WHPSHUDWXUH ERXQGDU\ FRQGLWLRQ IRU FDOFXODWLRQ SXUSRVHV

PAGE 157

7DEOH +HDWLQJ DQG FRROLQJ UDWHV IRU YDULRXV SURFHVV FRQGLWLRQV I YDOXHV PLQXWHVf IURP HTQ 6ORSHV RI VHPL ORJ WLPHWHPS JUDSKf 6WHDP :DWHU :DWHU )LOO WHPSHUDWXUH &RRN &RRN &RRO r& r)f r& r)f

PAGE 158

7KLV LV LPSRUWDQW IRU UHDOWLPH FDOFXODWLRQ VLQFH WKH $'( PHWKRG EHFRPHV FRQVLGHUDEO\ PRUH FRPSOLFDWHG ZKHQ UHTXLUHG WR DFFRXQW IRU FRQYHFWLRQ ERXQGDU\ FRQGLWLRQ DV FRPSDUHG WR WKH VSHFLILHG WHPSHUn DWXUH ERXQGDU\ FRQGLWLRQ &RPSDULVRQ RI WKH 7UDQVLHQW 7HPSHUDWXUH 9DOXHV 7KH WUDQVLHQW WHPSHUDWXUH YDOXHV UHFRUGHG DW WKH FDQ FHQWHU DV KHDWLQJ SURJUHVVHG DUH UHSRUWHG LQ ILJXUH 7KH REVHUYHG WHPSHUDWXUH YDOXHV DUH FRPSDUHG ZLWK WKH SUHGLFWHG YDOXHV XVLQJ WKH DQDO\WLFDO VROXWLRQ WR HTXDWLRQ Df DQG WKH $'( IRUPXODWLRQ WR WKH FRQGXFWLRQ HTXDWLRQ $V FDQ EH VHHQ WKHUH ZDV H[FHOOHQW DJUHHPHQW EHWZHHQ WKH SUHGLFWHG DQG REVHUYHG YDOXHV (IIHFW RI 6XGGHQ 3UHVVXUH 'URSV LQ WKH &DQ )LJXUH LOOXVWUDWHV WKH HIIHFW RI VXGGHQ GHSUHVVXUL]DWLRQ RI WKH UHWRUW RQ WUDQVLHQW WHPSHUDWXUH YDOXHV DW WKH FDQ FHQWHU 7KLV KDG EHHQ VKRZQ WR EH GXH WR LQWHUQDO SURGXFW PRYHPHQW IURP VXGGHQ ORVV RI SUHVVXUH LQVLGH WKH FDQ 7HL[HLUD HW DO f DQG HPSKDn VL]HV WKH LPSRUWDQFH RI PDLQWDLQLQJ WKH UHWRUW SUHVVXUH GXULQJ FRROLQJ 3HUIRUPDQFH RI WKH 2QOLQH &RQWURO /RJLF 7KH HVWLPDWHG SURFHVV WLPHV DQG WKH DFFXPXODWHG )4 YDOXHV IRU WKH WKUHH SURFHVV FRQGLWLRQV ILJXUHV DQG f DUH SUHVHQWHG LQ WDEOH )RU WKH UHIHUHQFH SURFHVV ZLWKRXW DQ\ GHYLDWLRQ ILJXUH f WKH &) PHWKRG SUHGLFWHG D FRUUHFWLRQ IDFWRU FORVH WR RU QR

PAGE 159

&DQ FHQWHU WHPSHUDWXUHrFf )LJXUH 7UDQVLHQW FHQWHU WHPSHUDWXUH GXULQJ FRQGXFWLRQn KHDWLQJ RI %HQWRQLWH LQ D F\OLQGULFDO FDQ e UR

PAGE 160

&$1 &(17(5 7(03(5$785( rFf )LJXUH 7UDQVLHQW FHQWHU WHPSHUDWXUH GXULQJ FRQGXFWLRQn FRROLQJ RI %HQWRQLWH LQ D F\OLQGULFDO FDQ ZLWK DQG ZLWKRXW GURS LQ UHWRUW SUHVVXUH

PAGE 161

7(03(5$785( r&f 7,0( PLQXWHVf )LJXUH &RPSXWHU SORW RI UHIHUHQFH SURFHVV QR GHYLDWLRQVf VKRZLQJ PLQXWHV RI KHDWLQJ WLPH DQG DQ DFFRPSOLVKHG )4 RI $&&203/,6+(' ) PLQXWHVf

PAGE 162

FRUUHFWLRQ 7KH RQOLQH FRPSXWHU PHWKRG DOVR PDLQWDLQHG WKH SURFHVV WLPH DQG WKH DFFXPXODWHG ) FORVH WR WKHLU GHVLJQ YDOXHV 1H[W D SURFHVV ZLWK DQ LGHDOL]HG VWHSIXQFWLRQDO GURS ZDV FRQVLGHUHG 7KLV LV VKRZQ LQ ILJXUH 7KH UHWRUW WHPSHUDWXUH XQGHUZHQW DQ LGHDOL]HG VWHS GURS RI r& r)f IRU D SHULRG RI PLQXWHV IURP WKH WK PLQXWH WR WKH WK PLQXWH 7KH &) PHWKRG SUHGLFWHG DQ H[WHQGHG KHDWLQJ WLPH RI PLQXWHV UHVXOWLQJ LQ D WRWDO DFFXPXODWHG )4 YDOXH RI 7KH RQOLQH FRQWURO ORJLF H[WHQGHG WKH KHDWLQJ WLPH WR PLQXWHV UHVXOWLQJ LQ D WRWDO DFFXPXODWHG ) YDOXH RI 7KXV ERWK PHWKRGV SUHGLFWHG YHU\ VLPLODU H[WHQVLRQV WR WKH KHDWLQJ WLPH DQG ZHUH DEOH WR PDLQWDLQ WKH UHTXLUHG WRWDO )4 RI $ SURFHVV WKDW H[SHULHQFHG D JUDGXDO GURS OLQHDU UDPSf DIWHU DERXW PLQXWHV LQWR WKH FRRN F\FOH ZDV FRQVLGHUHG QH[W ILJXUH f 7KH OLQHDU GHFUHDVH LQ WHPSHUDWXUH FRQWLQXHG IRU PLQXWHV ZLWK D UDSLG UHWXUQ WR VHW SRLQW 7KH RQOLQH FRPSXWHU PHWKRG RI FRQWURO H[WHQGHG WKH KHDWLQJ WLPH WR PLQXWHV LQ RUGHU WR VWLOO PDLQWDLQ DQ ) RI 6LQFH WKH &) PHWKRG FDQ RQO\ EH DSSOLHG WR VWHS IXQFWLRQ GHYLDWLRQV WKH JUDGXDO GURS LQ WKLV FDVH ZDV WUHDWHG DV D VWHS GURS RI r) IRU WKH PLQXWH GXUDWLRQ 7KLV UHSUHVHQWHG D PRUH VHYHUH GHYLDWLRQ WKDQ WKH DFWXDO UDPS IXQFWLRQ DQG WKH SUHn GLFWHG H[WHQVLRQ LQ SURFHVV WLPH ZRXOG EH H[SHFWHG WR RYHUFRPSHQVDWH IRU WKH DFWXDO ORVV LQ OHWKDOLW\ EXW ZRXOG EH VXUH WR SUHGLFW D VDIH SURFHVV 7KH UHVXOWV LQ 7DEOH VKRZ KRZ WKH &) PHWKRG SUHGLFWHG D SURFHVV WLPH RI QHDUO\ PLQXWHV WKUHH PLQXWHV ORQJHU WKDQ QHFHVn VDU\f UHVXOWLQJ LQ D ILQDO ) RI PLQXWHV QHDUO\ b RYHUSURFHVn VLQJ FRPSDUHG WR WKH WDUJHW )4 RI PLQXWHVf

PAGE 163

5(7257 $1 &(17(5 7777777 U77 7777-7 , , On M 7 , f§ 7,0( PLQXWHVf &RPSXWHU SORW RI D SURFHVV WKDW H[SHULHQFHG D VWHS IXQFWLRQDO GURS RI r& r)f LQ WHPSHUDWXUH EHWZHHQ WK DQG WK PLQXWH DQG VWLOO PDLQWDLQHG DQ )Q RI FU! $&&203/,6+(' ) PLQXWHVf

PAGE 164

7(03(5$785( r&f 7,0( PLQXWHVf )LJXUH &RPSXWHU SORW RI D SURFHVV WKDW H[SHULHQFHG D OLQHDU GURS LQ WHPSHUDWXUH EHWZHHQ WK DQG WK PLQXWH IURP r& WR r& r)ff DQG VWLOO PDLQWDLQHG DQ ) I $&&203/,6+(' )Q PLQXWHVf

PAGE 165

7DEOH $GMXVWHG KHDWLQJ WLPHV DQG UHVXOWLQJ OHWKDOLW\ )f LQ UHVSRQVH WR SURFHVV GHYLDWLRQV XVLQJ SURSRVHG RQOLQH FRQWURO ORJLF DQG PHWKRG RI *LDQQRQL6XFFDU DQG +D\DNDZD f 5HIHUHQFH 3URFHVV 6WHS 'URS ILJXUH f 'URS GXUDWLRQ PLQ 'URS DPRXQW r) *UDGXDO GURS ILJXUH 'URS GXUDWLRQ PLQ 0D[LPXP GURS r) 1R GHYLDWLRQ 2QOLQH FRQWURO 6\VWHP *LDQQRQL6 t +D\DNDZD QOLQH &RQWURO 6\VWHP *LDQQRQL6r t +D\DNDZD 7LPH RI +HDWLQJ PLQXWHVf 7RWDO ) PLQXWHVf r 7UHDWHG DV VWHS GURS RI r) IRU PLQXWHV GXUDWLRQ

PAGE 166

7KH VWHS IXQFWLRQ DQG WKH JUDGXDO GURS LQ WHPSHUDWXUH DUH MXVW WZR H[DPSOHV RI SURFHVV GHYLDWLRQV ZKHUH WKH RQOLQH FRPSXWHU FRQWURO V\VWHP LV DEOH WR PDLQWDLQ WKH GHVLUHG VWHULOL]DWLRQ 7KLV FRQWURO V\VWHP SUHGLFWV WKH FDQ FHQWHU WHPSHUDWXUH XVLQJ D ILQLWH GLIIHUHQFH DQDORJ RI HTXDWLRQ ,W WKHQ XVHV WKLV WHPSHUDWXUH YDOXH ZLWK D QXPHULFDO WUDSH]RLGDOf PHWKRG RI LQWHJUDWLQJ HTXDWLRQ WKH JHQHUDO PHWKRGf %RWK WKHVH VWHSV SODFH QR UHVWULFWLRQ ZKDWn VRHYHU RQ WKH UHWRUW WHPSHUDWXUH YDULDWLRQ 7KLV HQDEOHV WKH RQOLQH FRQWURO V\VWHP WR PDLQWDLQ WKH GHVLUHG VWHULOL]DWLRQ IRU DUELWUDU\ YDULDWLRQV LQ UHWRUW WHPSHUDWXUH 7KLV LV QRW WUXH IRU WKH FRUUHFn WLRQ IDFWRU DSSURDFK RI *LDQQRQL6XFFDU DQG +D\DNDZD f ZKLFK UHTXLUHV WKH UHWRUW WHPSHUDWXUH GURS WR EH D VWHS IXQFWLRQ IURP DQ RWKHUZLVH FRQVWDQW YDOXH $OVR LW LV QRW DSSDUHQW KRZ WKHLU PHWKRG ZRXOG FRUUHFW D SURFHVV ZLWK PXOWLSOH GURSV HYHQ LI WKH LQGLYLGXDO GURSV ZHUH VWHS IXQFWLRQDO 7KH QXPHULFDO PHWKRG RI FDOFXODWLQJ WKH DFFXPXODWHG SURFHVV ) LV D PRUH GLUHFW DSSURDFK WKDQ %DOOnV IRUPXOD PHWKRG RI FDOFXODWLQJ WKH SURFHVV ) DQG VXEVHTXHQW FRUUHFWLRQ IDFWRUV 2WKHU EHQHILW DUHDV LQFOXGH UHDOWLPH RSHUDWLRQDO LQIRUPDWLRQ DQG FRPSOHWH GRFXPHQWDWLRQ )LJXUHV DQG VKRZ D FRPSOHWH KLVWRU\ RI WKH SURFHVV LQFOXGLQJ UHWRUW WHPSHUDWXUH FDOFXODWHG FDQ FHQWHU WHPSHUDWXUH DQG FDOFXODWHG ) ,Q DGGLWLRQ WR SURYLGLQJ DOO WKH UHTXLUHG GRFXPHQWDWLRQ IRU UHJXODWRU\ DQG TXDOLW\ FRQWURO FRPn SOLDQFH WKH SORWV ZLOO SURYLGH YLVXDO GHVFULSWLRQV RI WKH FRPSOHWH SURFHVV KLVWRU\ 7KH JUDSKV FDQ FOHDUO\ VKRZ LI WKH SURFHVV ZHQW WKURXJK DQ\ GHYLDWLRQ DQG KRZ WKH UHTXLUHG ) ZDV VWLOO PDLQWDLQHG

PAGE 167

,Q WKLV ZD\ SURGXFWV ZKLFK H[SHULHQFH SURFHVV GHYLDWLRQV FDQ EH UHOHDVHG IRU VKLSPHQW RQ VFKHGXOH DQG VXSSRUWHG ZLWK DOO QHFHVVDU\ GRFXPHQWDWLRQ LQ FRPSOLDQFH ZLWK )'$ &DQQHG )RRG 5HJXODWLRQ 8VH RI WKHVH FRQWUROOHG V\VWHPV LQ LQGXVWU\ VKRXOG OHDG WR LPSURYHG SURGXFWLRQ HIILFLHQF\ ZLWKRXW FRPSURPLVLQJ SURGXFW TXDOLW\ DQG VDIHW\ LQ FDQQHG IRRG RSHUDWLRQV

PAGE 168

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

PAGE 169

$ VORZHVW KHDWLQJ UHJLRQ ZDV ORFDWHG DZD\ IURP WKH FHQWHUOLQH LQ D GRQXWVKDSHG UHJLRQ QHDU WKH ERWWRP b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

PAGE 170

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f VWXG\ RI VLQJOH SRLQW DQG LQWHJUDWHG OHWKDOLW\ FKDQJH LQ WKH WKHUPDO SURSHUWLHV FRUUHVSRQGLQJ WR GLIIHUHQW W\SHV RI IRRGV SRVVLELOLW\ RI LQFRUSRUDWLQJ YDULDEOH WKHUPDO SURSHUWLHV HJ YLVFRVLW\f HIIHFW RI GLIIHUHQW FDQ VL]HV DQGRU DVSHFW UDWLRV

PAGE 171

$33(1',; $ $/7(51$7,9( ),1,7( ',))(5(1&,1* 2) 675($0 )81&7,21 (48$7,21 7KH VWUHDP IXQFWLRQ HTXDWLRQ B A Mf B n f a%] UA U U 8VLQJ FHQWUDO GLIIHUHQFLQJ %LS %] B L^!! B S A? ] n] f ] Q %]n ]Q A = L9"L 9" }$Q rLO}-arL B AL}a ALOM ]] 9 ]LL ]LU ]LL $Q ]LU ]LL $Q MALOM rLWM rLMf rLOM ]LU ]L ]W ]LL f L. U9 Un B 0L U nU %U BW^! A L : MWL Y" ULYDUOLLL ULU ULL $ I \ A A } 7 PY U U Oa UMU UML YUMQ UMU UMUU UML!n &RPELQLQJ WKH WHUPV ZH JHW HTXDWLRQ

PAGE 172

$33(1',; % $/7(51$7,9( ),1,7( ',))(5(1&,1* 2) %281'$5< 9257,&,7< %RXQGDU\ FRQGLWLRQV IRU YRUWLFLW\ DW ERWWRP ZDOO /HW $] ]M ] ] 8VLQJ 7D\ORUnV VHULHV H[SDQVLRQ DW 2-f A O DO! rAM ]_MM$] 7 2$= $=f 1RZ XVLQJ QR VOLS FRQGLWLRQV XM ‘ I,I/ 8 YM M R  M"_ R U RU_DQG LA2 M E\ FKRLFH 3L LS .]] )URP HTXDWLRQ &2 RM $] D]f + 2U ]] XI 2fA ,LU$] RM UnUSORU]nO2M KM ? fRMUf$] r$]f f3L L f§ VLPLODUO\ DW RWKHU ERXQGDULHV U]n]9

PAGE 173

$33(1',; & ,1387 '$7$ ),/( )25 1$785$/ &219(&7,21 +($7 75$16)(5 02'(/ 5$',86 2) &
PAGE 174

$33(1',; $'( )2508/$7,21 )25 7+( &21'8&7,21 (48$7,21 $OWHUQDWLQJ GLUHFWLRQ H[SOLFLW IRUPXODWLRQ IRU WKH FRQGXFWLRQ HTXDWLRQ LV QRWHG EHORZ )RU SRLQWV RWKHU WKDQ FHQWHU DQG ZLWK LM LQFUHDVLQJ PRYLQJ DZD\ IURP FHQWHUf B BB [ nG]7 -U U U Hr` H" $W L\X H" HQ H!" LOM $]  I "A "r W-O ,W} 7Wf§ U Q Q rLMO rLWMO $U "O L af§^" 6"L f f "` ff 'L& 9 OOMn 36 SHLMU pLM HLMLfULMU VMLf 4fQ ML f ZKHUH 3 $[$U 4 $[U$U 6 $[$] )RU SRLQWV DW FHQWHU M f IRU XVH ZKLOH L LQFUHDVLQJ B D fD [ aG= U QQO QQ 4Q QQ 4Q QQO IUQ fQ QQO fQO LR HLR BLLRn LR HLR 9LR LU HLR VLR 9Lƒ DW $[ Q L 36 EIHXRV+LR P HLLRf3VLU H"R OL!f f ZKHUH R"A A XVHG GXH WR FHQWHUOLQH V\PPHWU\

PAGE 175

)RU SRLQWV RWKHU WKDQ FHQWHU DQG ZLWK L M GHFUHDVLQJ PRYLQJ WRZDUGV FHQWHUf GH B BB W a ] U U U H2W KL $W " "A 4@ I 4A A !-f $= HQ L }-O Q L -O $U H"+ nOS6 E^V"MV"LU HO f&M! r S8U X 8! f )RU SRLQWV DW FHQWHU M f IRU XVH ZKLOH L M GHFUHDVLQJ W -]7 ), QO Q Q Q n B t Q Rn Q L AL Q Rn" Q AL k" Lf B O LO L L$W $] D) Q rL OH"RVLLR VP LLRf3LU OR Q 36 f

PAGE 176

$W WKH WRS ERXQGDU\ ILQLWH GLIIHUHQFLQJ WKH ERXQGDU\ FRQGLWLRQ 4Q B 4Q Q]M Q]OM B K+4QO $] f N Q]M f DQG DW WKH VLGHZDOO LQ HQ LQU L QUO B K+4QO $U f NL QU f 7KH ILQLWH GLIIHUHQFHV IRU WKH ERXQGDU\ FRQGLWLRQV DQG WKH JRYHUQLQJ HTXDWLRQV DUH QRZ FRPELQHG $W WKH XSSHU FRUQHU FRPELQLQJ HTXDWLRQV DQG Q FI Q 6Q Q f Q]QU $836f Q]OQUO Q]OQUO Q]QUO SIQ Q +IB4Q nOO Q]OQUO XQ]OQU5O Q]OQU K 345f\$]f 6O<$Uf ZKHUH $ 3$Uf U$]f . 36 36 &RPELQLQJ HTXDWLRQV DW WKH WRS ZDOO f QO Q]$36f ^Q 6Q Q f 'MF? Q] M Q] M Q]M 3O $]fQ Q Q L f L f X N n Q]MO Q]OM Q]OMOn bQ U $]fQ ff` UA Q]OMO Y N Q]MOM f ZKHUH $ MU$] f§ 36

PAGE 177

&RPELQLQJ HTXDWLRQV DQG DW WKH WRS ZDOO RQ WKH FHQWHUOLQH Q Q] $36f ^ Q Q] 6 Q Q] Q Q] f 3 \ $]fQ m $ Q]O Q] ]OO} mKHUH $ 8 eD] &RPELQLQJ HTXDWLRQV DQG DW WKH VLGHZDOO f L QU ^k L 6O $Uf"r 4Q LQUO A N n LOQU L QUO $O36f 3c A S BB ‘ A L QUO LQUn 5 LQU Q pLOQUOA f` f ZKHUH $ ^9 (TXDWLRQV DQG DUH WKHQ XVHG IRU WKH LQWHULRU SRLQWV 0RYLQJ RXWZDUGV D FRPSOHWHO\ GLIIHUHQW DOJHEUDLFDOO\f VHW RI HTXDWLRQV DUH XVHG )RU WKH LQWHULRU SRLQWV HTXDWLRQV DQG DUH XVHG $W WKH ERXQGDU\ HTXDWLRQ LV XVHG IRU WKH VLGHZDOO Q QO OQUO rQU 77I7U $W WKH WRS ZDOO HTXDWLRQ LV XVHG Q QL B Q]OM Q]DQG ILQDOO\ XVLQJ HTXDWLRQ IRU WKH XSSHU FRUQHU $Q $Q K W $U Q]QU Q]QU N

PAGE 178

/,67 2) 5()(5(1&(6 $OLDGD 6 5 DQG 4XRQ $ VWDEOH H[SOLFLW QXPHULFDO VROXWLRQ RI WKH FRQGXFWLRQ HTXDWLRQ IRU PXOWLGLPHQVLRQDO QRQ KRPRJHQHRXV PHGLD &KHPLFDO (QJLQHHULQJ 3URJUHVV 6\PSRVLXP 6HULHV f $QGHUVRQ $ & 7DQQHKLOO DQG 5 + 3OHWFKHU &RPSXWDWLRQDO IOXLG PHFKDQLFV DQG KHDW WUDQVIHU +HPLVSKHUH 3XEOLVKLQJ &RUSRUDWLRQ :DVKLQJWRQ '& %DOO & DQG ) & : 2OVRQ 6WHULOL]DWLRQ LQ IRRG WHFKQRn ORJ\ 0F*UDZ+LOO %RRN &RPSDQ\ ,QF 1HZ
PAGE 179

%UDQGRQ + 3 3HOWRQ DQG 6WDDFN 6WDWHRIWKHDUW PHWKRGRORJ\ IRU HYDOXDWLRQ RI SDVWHXUL]HU KHDWLQJ DQG FRROLQJ SURFHVVHV 0%$$ 7HFKQLFDO 4XDUWHUO\ f %ULOH\ : 5 $ QXPHULFDO VWXG\ RI ODPLQDU VHSHUDWLRQ EXEEOHV XVLQJ WKH 1DYLHU6WRNHV HTXDWLRQV )OXLG 0HFK f &KX 7 < DQG 5 *ROGVWHLQ 7XUEXOHQW QDWXUDO FRQYHFWLRQ LQ D KRUL]RQWDO OD\HU RI ZDWHU )OXLG 0HFK 'DWWD $ $ $ 7HL[HLUD DQG ( 0DQVRQ ,Q 3UHVV &RPSXWHU EDVHG UHWRUW FRQWURO ORJLF IRU SUHFLVH RQOLQH FRUUHFWLRQ RI SURFHVV GHYLDWLRQV )RRG 6FL 'RXJODV -U 2Q WKH QXPHULFDO LQWHJUDWLRQ RI 6X[ X\ XW E\ LPSOLFLW PHWKRGV 6RFLHW\ RI ,QGXVWULDO $SSOLHG 0DWKHPDWLFV f (QJHOPDQ 0 6 DQG 5 / 6DQL )LQLWHHOHPHQW VLPXODWLRQ RI DQ LQSDFNDJH SDVWHXUL]DWLRQ SURFHVV 1XPHULFDO +HDW 7UDQVIHU (YDQV / % 5 & 5HLG DQG ( 0 'UDNH 7UDQVLHQW QDWXUDO FRQYHFWLRQ LQ D YHUWLFDO F\OLQGHU $,&K( -RXUQDO f )RUV\WKH ( DQG : 5 :DVRZ )LQLWH GLIIHUHQFH PHWKRGV IRU SDUWLDO GLIIHUHQWLDO HTXDWLRQV -RKQ :LOH\ DQG 6RQV 1HZ
PAGE 180

+LGGLQN 1DWXUDO FRQYHFWLRQ KHDWLQJ RI OLTXLGV ZLWK UHIHUHQFH WR VWHULOL]DWLRQ RI FDQQHG IRRGV $JULFXOWXUDO 5HVHDUFK 5HSRUWV &HQWHU IRU $JULFXOWXUDO 3XEOLVKLQJ DQG 'RFXPHQWDWLRQ :DJHQLQJHQ 1HWKHUODQGV +LJJLQV 0 DQG *HEKDUW 7KH VWDELOLW\ RI YHUWLFDO EXR\DQF\LQGXFHG IORZ LQ FROG ZDWHU +HDW 7UDQVIHU 7UDQV RI WKH $60( +RUQEHFN 5 : 1XPHULFDO PHWKRGV 4XDQWXP 3XEOLVKHUV 1HZ
PAGE 181

0XOYDQH\ 6 DQG 6 6 + 5L]YL $ PLFURFRPSXWHU FRQWUROOHU IRU UHWRUWV 7UDQV RI WKH $6$( f 1DYDQNDVDWWXVDV 6 DQG /XQG 0RQLWRULQJ DQG FRQWUROOLQJ WKHUPDO SURFHVVHV E\ RQOLQH PHDVXUHPHQW RI DFFRPSOLVKHG OHWKDOLW\ )RRG 7HFK f 1LHNDPS $ 8QNOHVED\ 1 8QNOHVED\ DQG 0 (OOHUVLHFN 7KHUPDO SURSHUWLHV RI EHQWRQLWHZDWHU GLVSHUVLRQV XVHG IRU PRGHOOLQJ IRRGV )RRG 6FL 2n%ULHQ * 0 $ +\PDQ DQG 6 .DSODQ $ VWXG\ RI QXPHULFDO VROXWLRQ RI SDUWLDO GLIIHUHQWLDO HTXDWLRQV 0DWKHPDWLFV DQG 3K\VLFV 2]LVLN 0 1 +HDW FRQGXFWLRQ -RKQ :LOH\ DQG 6RQV 1HZ
PAGE 182

6FKXPDQQ 8 DQG 5 $ 6ZHHW $ GLUHFW PHWKRG IRU WKH VROXWLRQ RI 3RLVVRQn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
PAGE 183

7HL[HLUD $ $ & 5 6WXPER DQG : =DKUDGQLN D ([SHULPHQWDO HYDOXDWLRQ RI PDWKHPDWLFDO DQG FRPSXWHU PRGHOV IRU WKHUPDO SURFHVV HYDOXDWLRQ )RRG 6FL 7HL[HLUD $ $ ( =LQVPHLVWHU DQG : =DKUDGQLN E &RPSXWHU VLPXODWLRQ RI YDULDEOH UHWRUW FRQWURO DQG FRQWDLQHU JHRPHWU\ DV D SRVVLEOH PHDQV RI LPSURYLQJ WKLDPLQH UHWHQWLRQ RI WKHUPDOO\ SURFHVVHG IRRGV )RRG 6FL 7RUUDQFH ( &RPSDULVRQ RI ILQLWHGLIIHUHQFH FRPSXWDWLRQV RI QDWXUDO FRQYHFWLRQ -RXUQDO RI WKH 1DWLRQDO %XUHDX RI 6WDQGDUGV %f 7RUUDQFH ( / 2UORII DQG $ 5RFNHWW ([SHULPHQWV RQ QDWXUDO FRQYHFWLRQ LQ HQFORVXUHV ZLWK ORFDOL]HG KHDWLQJ IURP EHORZ )OXLG 0HFK f 7RUUDQFH ( DQG $ 5RFNHWW 1XPHULFDO VWXG\ RI QDWXUDO FRQYHFWLRQ LQ DQ HQFORVXUH ZLWK ORFDOLVHG KHDWLQJ IURP EHORZ FUHHSLQJ IORZ WR WKH RQVHW RI ODPLQDU LQVWDELOLW\ )OXLG 0HFKDQLFV f 9LQRNXU 0 2Q RQHGLPHQVLRQDO VWUHWFKLQJ IXQFWLRQV IRU ILQLWHGLIIHUHQFH FDOFXODWLRQV &RPSXWDWLRQDO 3K\VLFV :LONHV DQG 6 : &KXUFKLO 7KH ILQLWHGLIIHUHQFH FRPSXWDWLRQ RI QDWXUDO FRQYHFWLRQ LQ D UHFWDQJXODU HQFORVXUH $,&K( -RXUQDO f

PAGE 184

$'',7,21$/ 5()(5(1&(6 $PHV : ) 1XPHULFDO PHWKRGV IRU SDUWLDO GLIIHUHQWLDO HTXDn WLRQV $FDGHPLF 3UHVV 1HZ
PAGE 185

&OHYHU 5 DQG &DWWRQ 6WHDG\ QDWXUDO FRQYHFWLRQ LQ D YHUWLFDO F\OLQGHU DW ODUJH 3UDQGWO QXPEHU 3URFHHGLQJV RI WKH +HDW 7UDQVIHU DQG )OXLG 0HFKDQLFV ,QVWLWXWH HGLWHG E\ 5 /DQGLV DQG +RUGHPDQQ &XQVROR DQG 3 2UODQGL $FFXUDF\ LQ QRQ RUWKRJRQDO JULG UHIHUHQFH V\VWHPV 1XPHULFDO PHWKRGV LQ ODPLQDU DQG WXUEXOHQW IORZ (GLWHG E\ & 7D\ORU 0RUJDQ DQG & $ %UHEELD -RKQ :LOH\ DQG 6RQV 1HZ
PAGE 186

+XQWHU + )LHOG WHVWLQJ RI D PLFURFRPSXWHU SRWDWR VWRUDJH FRQWURO V\VWHP $6$( SDSHU QR 3UHVHQWHG DW WKH VXPPHU PHHWLQJ RI WKH $6$( DW 0DGLVRQ :LVFRQVLQ .XQKDUGW ( ( DQG 3 ) :LOOLDPV 'LUHFW VROXWLRQ RI 3RLVVRQnV HTXDWLRQ LQ F\OLQGULFDOO\ V\PPHWULF JHRPHWU\ $ IDVW DOJRULWKP &RPSXWDWLRQDO 3K\VLFV /DUNLQ % 6RPH VWDEOH H[SOLFLW GLIIHUHQFH DSSUR[LPDWLRQV WR WKH GLIIXVLRQ HTXDWLRQ 0DWKHPDWLFV RI &RPSXWDWLRQ f /HQ] 0 DQG % /XQG 7KH OHWKDOLW\)RXULHU QXPEHU PHWKRG H[SHULPHQWDO YHULILFDWLRQ RI D PRGHO IRU FDOFXODWLQJ WHPSHUDWXUH SURILOHV DQG OHWKDOLW\ LQ FRQGXFWLRQKHDWLQJ FDQQHG IRRGV )RRG 6FL f /HQ] 0 DQG % /XQG 7KH OHWKDO LW\)RXULHU QXPEHU PHWKRG H[SHULPHQWDO YHULILFDWLRQ RI D PRGHO IRU FDOFXODWLQJ DYHUDJH TXDOLW\ IDFWRU UHWHQWLRQ LQ FRQGXFWLRQ KHDWLQJ FDQQHG IRRGV )RRG 6FL f /HQ] 0 DQG % /XQG 7KH OHWKDO LW\)RXULHU QXPEHU PHWKRG FRQILGHQFH LQWHUYDO IRU FDOFXODWHG OHWKDOLW\ DQG PDVV DYHUDJH UHWHQWLRQ RI FRQGXFWLRQKHDWLQJ FDQQHG IRRGV )RRG 6FL f /HY\ & DQG 5 +ODYDFHN &RPSXWHU FRQWURO RSWLPL]HV UHWRUW FRRN IRU \LHOG TXDOLW\ VDIHW\FDQQHG IRRGVf )RRG 3URFHVVLQJ f 0HUVRQ 5 / 5 3 6LQJK DQG 3 $ &DUURDG $Q HYDOXDWLRQ RI %DOOnV IRUPXOD PHWKRG RI WKHUPDO SURFHVV FDOFXODWLRQV )RRG 7HFK 0LOOHU & : (IIHFW RI D FRQGXFWLQJ ZDOO RQ D VWUDWLILHG IOXLG LQ D F\OLQGHU 3DSHU QR 3UHVHQWHG DW WKH $,$$ WK 7KHUPRSK\VLFV &RQIHUHQFH $OEXTXHUTXH 1HZ 0H[LFR 0LWFKHOO $ 5 DQG ) *ULIILWKV 7KH ILQLWH GLIIHUHQFH PHWKRG LQ SDUWLDO GLIIHUHQWLDO HTXDWLRQV -RKQ :LOH\ t 6RQV 2OVRQ ) & : DQG 0 -DFNVRQ +HDWLQJ FXUYHV WKHRU\ DQG SUDFWLFDO DSSOLFDWLRQ ,QGXVWULDO (QJJ &KHPLVWU\ 2VWUDFK 6 1DWXUDO FRQYHFWLRQ LQ HQFORVXUHV $GYDQFHV LQ +HDW 7UDQVIHU 2]RH +
PAGE 187

3DFNHU DQG / % *DPOHQ &DOFXODWLRQ RI WHPSHUDWXUH PHDVXUHPHQW HUURUV LQ WKHUPRFRXSOHV LQ FRQYHFWLRQ KHDWLQJ FDQV )RRG 6FL 3DWDQNDU 6 9 1XPHULFDO KHDW WUDQVIHU DQG IOXLG IORZ +HPLn VSKHUH SXEOLVKLQJ &RUSRUDWLRQ :DVKLQJWRQ 3HSSHU : DQG 6 +DUULV 1XPHULFDO VLPXODWLRQ RI QDWXUDO FRQYHFWLRQ LQ FORVHG FRQWDLQHUV E\ D IXOO\ LPSOLFLW PHWKRG 1XPHULFDO/DERUDWRU\ FRPSXWHU PHWKRGV LQ IOXLG PHFKDQLFV 3UHVHQWHG DW WKH ZLQWHU DQQXDO PHHWLQJ RI WKH $60( 1HZ
PAGE 188

6WHHOH 5 DQG 3 : %RDUG $PHQGPHQWV WR %DOOnV IRUPXOD PHWKRG IRU FDOFXODWLQJ WKH OHWKDO YDOXH RI WKHUPDO SURFHVVHV )RRG 6FL f 6WHHQVHQ ) :K\ XVH D PLFURSURFHVVRU IRU SURFHVV UHJXODWLRQ )RRG (QJJ ,QWHUQDWLRQDO f 7HL[HLUD $ $ &RQGXFWLRQKHDWLQJ FRQVLGHUDWLRQV LQ WKHUPDO SURFHVVLQJ RI FDQQHG IRRGV $60( SDSHU QR :$+7 3UHVHQWHG DW WKH ZLQWHU DQQXDO PHHWLQJ RI WKH $60( DW 6DQ )UDQFLVFR &DOLIRUQLD 8QR DQG +D\DNDZD $ PHWKRG IRU HVWLPDWLQJ WKHUPDO GLIIXVLYLW\ RI KHDW FRQGXFWLRQ IRRG LQ D F\OLQGULFDO FDQ )RRG 6FL f 8QR DQG +D\DNDZD &RUUHFWLRQ IDFWRU RI FRPHXS KHDWLQJ EDVHG RQ FULWLFDO SRLQW LQ D F\OLQGULFDO FDQ RI KHDW FRQGXFWLRQ IRRG )RRG 6FL f :LONHV 7KH ILQLWH GLIIHUHQFH FRPSXWDWLRQ RI QDWXUDO FRQYHFWLRQ LQ DQ HQFORVHG UHFWDQJXODU FDYLW\ 3K' 7KHVLV 8QLYHUVLW\ RI 0LFKLJDQ $QQ $UERU :ROIVKWHLQ 0 5 6 +LUVK DQG % + 3LWWV 6HFRQGRUGHU QRQLWHUDWLYH $', VROXWLRQ RI QRQOLQHDU SDUWLDO GLIIHUHQWLDO HTXDWLRQV $GYDQFHV LQ FRPSXWHU PHWKRGV IRU SDUWLDO GLIIHUHQn WLDO HTXDWLRQV (GLWHG E\ 5 9LFKQHYHWVN\ $,&$ 1HZ %UXQVZLFN 1HZ -HUVH\
PAGE 189

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

PAGE 190

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n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

PAGE 191

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

PAGE 192

8)'LVVHUWDWLRQV $0 8) /LEUDULHV'LJLWDO 'LVVHUWDWLRQ 3URMHFW 3DJH RI ,QWHUQHW 'LVWULEXWLRQ &RQVHQW $JUHHPHQW ,Q UHIHUHQFH WR WKH IROORZLQJ GLVVHUWDWLRQ $87+25 'DWWD $VKLP 7,7/( 1XPHULFDO PRGHOLQJ RI QDWXUDO FRQYHFWLRQ DQG FRQGXFWLRQ KHDW WUDQVIHU LQ FDQQHG IRRGV ZLWK DSSOLFDWLRQ WR RQOLQH SURFHVV FRQWURO UHFRUG QXPEHU f 38%/,&$7,21 '$7( $M! N9 f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aFI 7LWOH 876 &RGHf DV ZHOO DV WR WKH PDLQWHQDQFH DQG SUHVHUYDWLRQ RI D GLJLWDO DUFKLYH FRS\ 'LJLWL]DWLRQ DOORZVaWKH 8QLYHUVLW\ RI )ORULGD WR JHQHUDWH LPDJH DQG WH[WEDVHG YHUVLRQV DV DSSURSULDWH DQG WR SURYLGH DIWGOHQKDQFH DFFHVVXVLQJ VHDUFK VRIWZDUH 7KLV JUDQW RI SHUPLVVLRQV SURKLELWV XVH RI WKH GLJLWL]HGn9HUVLRIOVnIRU FRPPHUFLDO XVH RU SURILW W:&/ N ‘ 6LJQDWXUH RI &RS\ULJKW +ROGHU $VNLP / 7f[IK$ 3ULQWHG RU 7\SHG 1DPH RI &RS\ULJKW +ROGHU/LFHQVHH 3HUVRQDO LQIRUPDWLRQ EOXUUHG DHQHH 'DWH RU 6LJQDWXUH 3OHDVH SULQW VLJQ DQG UHWXUQ WR &DWKOHHQ 0DUW\QLDN 8) 'LVVHUWDWLRQ 3URMHFW 3UHVHUYDWLRQ 'HSDUWPHQW 8QLYHUVLW\ RI )ORULGD /LEUDULHV 32 %R[ *DLQHVYLOOH )/ 3ULQWHG IRU $VKLP 'DWWD DNGO#FRPHOOHGX!


140
Table 2. Heating and cooling rates for various process conditions.
f values (minutes) from eqn. 9
(Slopes of semi log time-temp graph)
Steam Water Water
Fill temperature Cook Cook Cool
44.8 44.8 51.2
43.8 44.7 50.9
67C (153F)
90C (194F)


160
Combining equations 94 and 95, at the top wall on the centerline
0
n+1
nz,0
1
A(1+2P+S)
{0
n
nz-1,0
+
S(-0
n
nz-1,0
+
0
n
nz-2,0
)
+ 2P((1+ y Az)0n+1
, 0 A+
nz,l nz-1,0
z-l.-l
here A = U £az -
Combining equations 93 and 96 at the sidewall
(99)
i ,nr
1 {" i+ S((l+ -Ar)0?*! Qn. ,+
1 i,nr-l ^ k i+l,nr i ,nr-l
A(l+P+S)
+ P(-0¡ 1+ 0^ p __ <
^ i ,nr-l i,nr-2' R1- i,nr-
n
i-l,nr-l^
)} (100)
where A = 1+ {V -
Equations 93 and 94 are then used for the interior points. Moving
outwards, a completely different (algebraically) set of equations are
used. For the interior points, equations 91 and 92 are used. At the
boundary, equation 96 is used for the sidewall
0n+1
0n+l + ,l,nr-l
,*nr TTfTr
At the top wall, equation 95 is used
0n+1
0n+i nz-l,j
nz-J
and finally, using equation 96 for the upper corner
An+1 An+1 / h ,
0 =0 / 1+ t Ar
nz,nr nz-1,nr k


where a is a deformation parameter that varies the degree of
stretching. However, since the computer coding of the problem is
ultimately done in the physical domain, the inverse transfor
mation z(n) is more appropriate in this context which is given
below.
z =1/2 {1 + tan^^n -1) tan a)}
(38)
This concentrates the grids on either end z = 0 and z = 1.
In the r direction, the above transformation is modified to
concentrate grids on one side (side wall, r = R/H) only. Thus
R
H
(39)
To check if the transformations indeed satisfy requirements 1
through 4, we first see that both the transformations are one to
one. Next, looking into the transformation derivatives,
dr 2 R
d£ it a IT
tamr/2 a
1 + 52 tan2ir/2 a
(40a,b)
dz 1 2
2 tan tt a/2
dn ? it a
1 + (2n l)2 tan2ir/2 a
which are continuous in 0 < £ < 1 and 0 < n < 1 respectively.


95
patterns accurately. Streamlines, on the other hand, can provide all
the essential details of the flow patterns without the clumsiness.
This is particularly true in the initial times when velocities are
many times larger near the boundary layer than at the core. This
created a scaling problem for vector plots. Also, in the boundary
layer, grid points are very close together. Thus, for most of the
figures, streamlines are presented instead of velocity vectors. At
much later times (t=1800 seconds) when the flow approached steady
state, and the velocity magnitudes became comparable in the core and
in the boundary layer, velocity plots would be visually more
meaningful (see figure 10o).
The magnitudes of the streamlines in each figure are noted on
the right side of each figure under "5 =0.0001". To read the value
that corresponds to a particular stream function contour, the line
type of the contour is noted. Next, the order of that line type in
table 1 is noted counting from the top. The value of the contour is
the number in the same order from the top in the list accompanying
each figure.
Start of flow and the conduction layer. For very small times
after heating starts, the fluid nearest to the bottom wall is
restricted to being conduction-heated. Figure 9 shows the area near
the bottom wall of the can. Further from the sidewall, the isotherms
in this figure compares very well with temperatures predicted at the
same height using the conduction equation in a semi-infinite
(0 < z < ) region (Ozisik, 1980).


26
equations have been approximated for no viscous dissipation. Jaluria
(1980) discussed various situations under which viscous dissipation
can be neglected. All other fluid properties are treated as
constants except variation of density which is discussed below.
Boussinesq Approximation
The solution to the set of equations can be considerably
simplified using the Boussinesq approximation. In the Boussinesq
approximation, the density is treated as constant except in the body
force term. The density difference in the body force term is approx
imated as
p p = p8 (T T ) (19)
a a
and the density variation in the continuity equation is ignored.
This approximation requires both
8 (T T ) 1 (20)
"'! ( )T g 5 1 <21>
to be true. The second inequality is generally satisfied for
liquids. So is the situation with the first inequality. For
example, for natural convection in water, as in this problem
8 = 0.0002 /C
T = wall temperature of the can = 121C
Ta= initial temperature of liquid in can = 30C


Figure 13. Isothermal vertical surface, showing boundary
layer with velocity profile u(x,y). and
temperature profile t(x,y) in water (Higgins
and Gebhart, 1983).


135
Wafer
T,=30*C
top-121*C
s!da=121*C
boT-121*C
R-4.19cm
H=10.67cm
LR=*3S,0.80
LZ=38,0.9Q
5-0.0001
RR-38,0.80
RZ-38,0.90
5=0.001
After 30 sec
Isotherms in a cylindrical can calculated with
two different convergence criteria for
boundary vorticity (<5=0.0001 on the left
and <5=0.001 on the right)
Figure 29.


METHODOLOGY
Detailed development of the solution techniques for natural
convection and conduction heat transfer models are described in this
chapter. For the natural convection heating problem, formation of
finite difference equations, setting up of computational boundary
conditions, and development of the algorithm for the iterative
solution of the complete problem, are described first. Along the
same lines, solutions of the finite difference analog to the
conduction equation are described next. Experimental study of
conduction heating is then reported. With the thermal models
working, use of the conduction heating model for on-line control of
conduction-heated food is then presented.
Modeling of Natural Convection Heating in Cans
Because of the inevitable coupling of the governing equations,
except for extremely simplified cases, most natural convection flows
of practical interest are too complex to be solved analytically.
Thus numerical computations are generally employed to obtain the
desired solution. The present problem of natural convection in a
cylindrical enclosure is no exception. Numerical solutions to this
and similar problems demand high speed digital computers with large
35


80
Coding of the Program and the Hardware
The algorithm presented in figure 5 for solving the vorticity-
stream function formulation of the natural convection problem was
coded in FORTRAN 77 with individual modules for solving temperature,
vorticity, stream function, velocity and boundary vorticity values.
Structured programming practices were used. Use of GO TO statements
were avoided except in two situations to simulate a WHILE statement.
The WHILE statement was unavailable in the particular version of
FORTRAN 77. Single precision of the variables was known to carry 14
significant digits for the particular computer and the same was
taken to be sufficient for this problem.
The input data file for a typical run is in Appendix C.
The computer used was a CDC Cyber 170/730, with two 60 bit CPUs
crunching 60 bit words at 400 MHz, 20 parallel processors, and 1091K
of 60 bit word central memory.
The plotting was done using a plotting package from the
University of Illinois*. The convenience offered by the plotting
package cannot be overemphasized. The plots were viewed on a
Tektronix 4006 terminal and hardcopies were made on Tektronix 4662
plotters.
Modeling of Conduction Heating in Cans
The partial differential equation (POE) governing conduction
heating of a cylindrical-shaped solid was discretized and finite
*W. E. Dunn, Mechanical and Industrial Engineering Department,
University of Illinois at Urbana-Champaign. Personal Communication.


66
the weakest condition that satisfies equation 65a was the recursion
relation
1 + (2j 1) Xj.-L
ZjFT
(68)
Using boundary conditions on the axis, it was found that
Xo = 0.25 (69)
The transport velocities defined this way conserve volume flow
exactly and are consistent with the grid point velocity field as the
mesh is refined.
The transport velocities ut and vt are the velocities
jJ '
used in the energy and vorticity transport equations. The grid point
velocities u. and v. are the velocities plotted in the
i jj i
velocity plots.
To solve for ^ from the vorticity-stream function equation
61, we note that it is an elliptic Poisson equation as compared to
the parabolic equations for temperature and vorticity. Direct (non
iterative) methods to solve the equation are available that are
considerably (10 times) faster than an iterative method like the
successive over-relaxation technique (to be discussed later).
However, direct methods usually require large amounts of storage,
and available computer codes are often limited to rectangular or
simple domains or have restrictions on the type of boundary
conditions. Shih (in press) discussed suitability of direct methods


Figure 2. contd.


method. Grid stretching was used in both radial and axial
directions. Boundary wall vorticities were used to check for
convergence of the iteration process.
Plots of transient isotherms, streamlines and velocities were
provided. Calculated radial velocities and axial temperature
profiles at different times compared quite well with available
experimental data. Slowest heating points were located in the bottom
15% of the height of the container studied. These slowest heating
points migrated over time within this lower region of the can but no
particular pattern of migration was noted.
Transient can center temperature was also studied for the case
of conduction-heated solid food. An alternating direction explicit
finite difference solution to the conduction equation compared very
well with analytical solutions and experimental results. The ability
of the finite difference model to perform under arbitrary variation
of boundary temperature allowed the model to be incorporated in an
algorithm for on-line control in batch thermal processing of
conduction-heated food. The uniqueness of this control system in
maintaining the level of desired food sterilization for arbitrary
variation in process heating temperature was shown to offer
considerable advantages over other possible methods of on-line
xvi i
control.


130
slowest heating temperatures at all times. The grid sizes also
predict identical transient temperatures at l/3rd height on the cen
terline as shown in figure 27.
Selection of the time step. A comprehensive stability ana
lysis, that included all the finite difference equations and the
boundary conditions, was not practicable. It was found that larger
time increments required more iterations per time step that could
actually increase the computation time for the overall problem. A
strategy was taken whereby the computer automatically adjusted the
time increment to restrict the number of iterations to about 15.
This was done in the following manner
1. If the number of iterations during a particular time
step is less than 15 and decreased from the number of
iterations required in the previous time step, the time
step is increased by 0.05 seconds.
2. Otherwise the time step (At) is decreased by 0.05
second.
The time step history for a typical run is given in figure 28. A
similar algorithm for adjusting the time step to limit the number of
vorticity iterations was used by Kee (1974) from the work of Briley
(1971).
Since implicit methods generally consume more CPU time per time
increment, it was decided to investigate how the time steps used in
the ADI method compare with time steps that could have been achieved
using a possible explicit finite difference formulation for the


APPENDIX D
ADE FORMULATION FOR THE CONDUCTION EQUATION
Alternating direction explicit formulation for the conduction
equation is noted below. For points other than center and with i,j
increasing ( moving away from center)
_30 320 320 _1_30
3x = 'dzT + Jr7 + r 9r
e"*}. e? .
1,J 1 ,3
At
01?
i+yu
- e? en+1+ e>?+1
1 ,J
i-l,j
Az
f: 0? 0?+^+ 0?+* .
1tJ+l ItJ 1.1 TtJ-1
2
+ I
r
0n 0n+1
ij+l itj-l
2Ar
0,?+l
i ,3
~{0? .+S(01?.-i - 0? + 0?+} )
DiC V 1+l.J 1,3 1-l.j'
1+P+S
+p(ei,j+r i,j+ eij-i)+r(8i,j+r sj-i)!
Q,n
ji+1
(91)
where P=Ax/Ar2, Q=Ax/2rAr, S=Ax/Az2
For points at center (j=0), for use while i increasing
80 a20 a20
3x ~dZ7 3r2
nn+l nn Qn nn Qn+1, nn+l 9frn n nn+l, n+l >
9i.o- ei,o _9i+i.o' 9i,o- ei.o+ Vi.o 2(0i,r ei,o- si,o+ ViJ
! 2 + at2
Ax
0n+1
i,0
1+2P+S
bfeuo+s<0Hi,o- 6m+ ei-i,o)+2P(si,r e?,o+ li>)
(92)
where o!?+^= 0? ^ used due to centerline symmetry
157


27
Thus,
6 (T T ) = (0.0002)(121-30) = 0.0182 1 (22)
cl
Almost all the previous works on natural convection in similiar
areas have used the Boussinesq approximation.
Transformation of the PDEs
To further facilitate numerical solution to the set of
equations, the primary variables are transformed to a new set of
variables. Vorticity (to) and stream function ( ijj ) are defined in
terms of the primary variables u (axial velocity) and v (radial
velocity)
- 3v 3
3z 3r
(23)
u
1
- r *
r 3r
1_
r 3z
(24a,b)
Using vorticity so defined, the pressure (p) is eliminated from
the u and v momentum equations 14 and 15. These two equations gave
the parabolic transport equation for vorticity Next, using the
definition of m (equation 23) and the definition of ij/ (equations
24a and 24b), the equation for ip is derived. This gave the vorticity
stream-function formulation for the natural convection problem.


145
correction. The on-line computer method also maintained the process
time and the accumulated F0 close to their design values. Next, a
process with an idealized step-functional drop was considered. This
is shown in figure 34. The retort temperature underwent an idealized
step drop of 8.3C (15F) for a period of 10 minutes from the 50th
minute to the 60th minute. The CF method predicted an extended
heating time of 72 minutes, resulting in a total accumulated FQ
value of 6.02. The on-line control logic extended the heating time
to 72.2 minutes, resulting in a total accumulated F0 value of
6.11. Thus, both methods predicted very similar extensions to the
heating time and were able to maintain the required total FQ of 6.0.
A process that experienced a gradual drop (linear ramp) after
about 40 minutes into the cook cycle, was considered next (figure
35). The linear decrease in temperature continued for 20 minutes
with a rapid return to set point. The on-line computer method of
control extended the heating time to 74.6 minutes in order to still
maintain an F0 of 6.18. Since the CF method can only be applied to
step function deviations, the gradual drop in this case was treated
as a step drop of 20F for the 20-minute duration. This represented
a more severe deviation than the actual ramp function, and the pre
dicted extension in process time would be expected to overcompensate
for the actual loss in lethality but would be sure to predict a safe
process. The results in Table 3 show how the CF method predicted a
process time of nearly 78 minutes (three minutes longer than neces
sary) resulting in a final F0 of 7.5 minutes (nearly 20% overproces
sing compared to the target FQ of 6.0 minutes).


18
20 SECONDS
42 SECONDS
Figure 1
Isotherms and velocity profiles in a. glass
bottle at various heating times (after Engelman
and Sani, 1983)


113
times of 30, 240, and 450 seconds are 35, 16 and 12 mm/second re
spectively. The observed maximum positive velocities at the same
time instants by Hiddink (1975) were 32, 18 and 12 mm/sec respec
tively. Thus the computed and observed values have good agreement.
As time passes, the enclosed liquid gets warmer so that the local
Grashof number at wall drops, which means that the buoyancy force
drops. Thus, as the figures show, the maximum velocities at the
boundary layer reduced with time in both computed and observed
values. The computed and observed profiles also agree in predicting
the thickening of the boundary layer as time passes due to further
penetration of heat. The thickness of the boundary layer was within
5-10% of the radius and agrees with observations of Hiddink (1975).
Shyu (1985) also reported boundary layer thickness of 5-10% of
radius but his computed velocities were much smaller and could be
attributed to a much smaller temperature difference imposed on the
wall in his case.
Axial temperature profiles. Temperature values along the axis
of the cylindrical can at different times are presented in figure
15. The axial temperature profiles were also available from the
experimental study of Hiddink (1975). These are plotted alongside.
The input data for figure 16 are radius=8.9cm, height of liquid
=18.0cm, top surface free and sidewall and bottom wall temperatures
of 120C. It can be seen that the numerical model predicts the ob
served values with about 5% discrepancy except at small times after
heating. Qualitatively, too, the predicted and observed shapes of
the temperature profiles at different times are very similar.


67
for "large" systems of equations. Direct methods also have round off
error propagation problem. With direct methods, round-off errors can
be incurred at each mathematical operation, and simply accumulate
until final answers are obtained. When iterative techniques are
used, the presence of round-off errors at the end of any given
iteration simply results in those unknowns being somewhat poorer
estimates for the next iteration. For practical purposes, round-off
error in the final converged values of an iterative scheme is only
that accumulated in the final iteration (Hornbeck, 1975). A very
efficient direct method for solving Poisson's equation was described
by Schumann and Sweet (1976) and the code was available from the
National Council of Atmospheric Research (NCAR) software library.
However, on closer look the algorithm (GENBUN) was found to be
limited to coordinates transformed in one direction only3. Since in
this study the boundary layers were expected to be present both
along sidewall as well as top and bottom wall, coordinate stretching
were considered necessary in either direction. This made the above
algorithm unsuitable for the present study.
A popular approach to solve such a steady state equation is to
formulate it as an unsteady equation and solve for the asymptotic
steady state solution. ADI methods have been used for this purpose.
These methods were known to be faster than the successive over-
relaxation method (to be discussed later) for a rectangular region
(Birkoff et al., 1962). However, the optimum sequence of parameters
aR. Sweet. U. S. Dept, of Commerce. National Bureau of Standards.
325 Broadway, Boulder, Colorado. Personal Communication.


APPENDIX C
INPUT DATA FILE FOR NATURAL CONVECTION HEAT TRANSFER MODEL
4.19
10.67
981
0.0002
0.0014612
0.0089
1
30.0
121.00
38
38
10000
1.82
0.00001
0.0001
2
1
1
1
19
38
1
1
1
0
0.5
40
120200
2
2
0.80
0.90
RADIUS OF CYLINDER IN CM(=1.65 INCHES)
HEIGHT OF CYLINDER IN CM(=4.20 INCHES)
G(ACCN. DUE TO GRAVITY) IN CM/SEC SQ.
VOLUMETRIC COEFF. OF WATER IN /DEG C AT 27 DEG C
THERMAL DIFFUSIVITY OF WATER IN CM SQ/SEC AT 25 DEG C
VISCOSITY OF WATER IN GM/CM/SEC AT 25 DEG SEC
DENSITY OF WATER IN GM/CC
INITIAL UNIFORM TEMPERATURE INSIDE CAN IN DEG C
WALL TEMPERATURES AT TIME GREATER THAN ZERO IN DEG C
NUM8ER OF GRID ELEMENTS ALONG HEIGHT
NUMBER OF GRID ELEMENTS ALONG RADIUS
INTERVAL OF PRINT IN MILLISECONDS
OPTIMUM RELAXATION PARAMETER FOR SOR
CONVERGENCE CRITERION FOR STREAM FUNCTION
CONVERGENCE CRITERION FOR VORTICITY
CENTERLINE BOUNDARY IS INSULATED
SIDEWALL BOUNDARY IS AT HEATING TEMPERATURE
TOP BOUNDARY IS AT HEATING TEMPERATURE
BOTTOM BOUNDARY IS AT HEATING TEMPERATURE
NUMBER OF RADIAL GRID POINTS TO PRINT
NUMBER OF Z GRID POINTS TO PRINT
STRETCHING IN Z DIRECTION
STRETCHING IN R DIRECTION
PRINT GRID COORDINATES
DO NOT FIND OPTIMUM OMEGA
VALUE OF RELAXATION PARAMETER FOR BOUNDARY VORTICITY
MAXIMUM ITERATIONS PERMITTED IN SOR
FINAL TIME IN MILLISECONDS
ONE-SIDED STRETCHING IN R DIRECTION
BOTH SIDED STRETCHING IN Z DIRECTION
VALUE OF SIGMA (EQN. 39) FOR STRETCHING R DIRECTION
VALUE OF SIGMA (EQN. 38) FOR STRETCHING Z DIRECTION
156


75
gence could have several possibilities. Ideally, at convergence, all
the variables should have converged. For practical reasons, it would
be sufficient to check on the variable that converged the slowest.
This is so that when this variable has converged, all others would
have converged. However there are no general guidelines to choose
apriori the slowest converging variable. Interior velocity
components were suggested by Roache (1982) as a measure of
convergence. Frequently, researchers in the past have decided on the
vorticity boundary values as the variables to watch for the
iteration convergence of the entire problem. Roache (1978) referred
to tests on boundary vorticities as the most sensitive test for
convergence. The same variables were used to check for iteration
convergence in this study even though they may not have been the
variables that converged the slowest.
For the chosen variable, there are no definitive criteria for
the measure of convergence. A relative error criteria was used for
this study, which is
~n+lm+^ ~n+lm
0) 0)
-n+1
m
0)
< 6
(78)
~n+1
at all solid boundaries. The boundary vorticities o> 1 were found
from the interior stream function value using equation 76.
As an example of other measures of convergence, while studying
natural convection in a rectangular cavity, Kimura and Bejan (1984)
used the following convergence criteria on interior points


Can center temperature(c)
Figure 31. Transient center temperature during conduction
heating of Bentonite in a cylindrical can £
ro


12
would require repeated time-consuming computer simulations.
Using the general idea of Teixeira and Hanson (1982), the
feasibility of several control algorithms that can work on-line with
available computer hardware needed to be investigated, and led to
the following studies as part of the work carried out in this
project:
1) Ways to obtain the transient temperature solution faster,
using other available methods of solving the conduction
equation.
2) The need to incorporate a convection (Robin's) boundary
condition to reflect the true situation, rather than the
temperature specified (Dirichlet) boundary condition used
by Teixeira et al. (1975a,b) in solving the conduction
equation.
Conduction Heating of Canned Foods
Since the pioneering work of Ball and Olson (1957) mentioned
earlier, conduction heating of canned foods has been studied
extensively. Consequently there is a vast amount of literature
existing on the subject.
Studies Based on Analytical Solution
The majority of the works on conduction-heating of canned foods
were based on one of several different forms of the analytical solu
tion to the heat conduction equation. Also most of these solutions


RESULTS AND DISCUSSION
Results for the numerical natural convection heat transfer
model are presented and discussed first. Next, numerical and experi
mental results are presented for conduction heat transfer in canned
foods. Finally, performance of the on-line control logic (using
finite difference conduction heat transfer model) is demonstrated
for conduction-heated foods.
Natural Convection Heating in Canned Foods
The results of the numerical model for natural convection heat
transfer are described here with water as a model liquid food.
Unless otherwise noted, the input data file for the results is given
in Appendix C. Development of the flow patterns and temperature
profiles are first discussed and diagrams of the same are given at
various times during heating. The diagrams are qualitatively and
quantitatively compared with previous numerical studies and
available experimental data. Location and migration of the slowest
heating points inside the can are then discussed. Also, variation of
the slowest heating temperature with time was compared with that of
transient temperature at other locations. Finally, comments are
included on the numerical aspects of the model.
93


* L 1 * v ~~ ^
30 50 70 90 110 120
Temperaiure(c)
Figure 16.
Comparison of numerically predicted axial
temperature profiles with observed (Hiddink,
1975) values at various times during heating


117
geometric center the coldest at all times (see figure 22). Tradi
tionally, in natural convection-heating products, the point at l/3rd
the height of the container on its axis was considered to be the
slowest heating point at all times. This study shows that, not only
is there no single point that is coldest at all times, the cluster
of points (or the slowest heating zone) is located away from the
centerline. The clockwise loop formed at the bottom (see figure 10)
forces liquid upwards near the centerline. This clockwise loop was
also clearly seen in the calculations of Sani (figure 2). Because of
the hot liquid going up near the centerline at the bottom, the
temperature is much higher in this region (see the isotherms in
figures 2 and 10 near bottom) and it is unlikely that the lowest
temperature would occur near the bottom centerline. Also, there is
no obvious reason why one particular point would stay the slowest at
all times. Hiddink (1975) also noted the same from his experimental
observations. He did not observe any well defined slowest heating
point. Instead, he referred to a "slowest heating region" near the
bottom of the container and reported no stagnant fluid that stays
consistently at one slowest heating point. The coldest region found
in this study, therefore, was a donut-shaped region near the bottom
of the container, away from the sidewall and centerline, as shown in
figure 17.
Migration of the slowest heating points. Figure 18 shows how
the coldest point in the can moved as heating proceeded. The order
of movement is shown in figure 18 by following the letters in alpha
betical order. Between two consecutive letters of the alphabet, the


90
Figure 8, Flow diagram for computer control of retort
operations with on-line correction of process
deviations


134
The actual time step used was 0.15 seconds which is twice the
possible time step for this explicit method.
Convergence of boundary vorticity. Insufficient convergence of
boundary vorticity can lead to inaccuracy and instability in the
solution. Convergence criteria (S) of 10^ and 10"^ were used for
the boundary vorticities in two sets of computation with all other
input values unchanged. The temperature and stream function values
obtained from these two computations are presented in figures 29 and
30. As can be seen, there is very little difference between using
the two different criteria for convergence. To be on the
conservative side, <5 =10"^ was used in all the calculations.
A note on the half grid points. Confusion exists over
evaluation of half grid points at the control volume faces for non-
uniform grid system. If we use the half points in the computational
(n 5) domain, as is the case if we derive in the following way:
9(rve) 9n 9(rve) 9(rve) 9ji
9r 9r 9n 9n 9r
(87)
r.j, r.
>1 rj-l
Then rj+l/2 should be defined by
r
(38)


54
After algebraic manipulation, temperature equations at the "full"
time step
nN+1 j. k nfl+1 An+1 _
-a . i+ b .0- - c.0- -,i = rhs
J ijJ-1 j i,j J i,j+l J
(48b)
where
aj
1,, n n.
Vy|vLl)ri.i/;, r.i-^
rj(rj+rrj-iJ(rjrj-i)
,= 1_+i ^ + rj=b
3 iT 2rj(rj+rrj-i) rjtrj+rrj-i) r+irj rrrj-i
V^-lvgl)^ rlHl/
Wv V>rrj-in'>rrj)
rhsj d9*-l,j + ee1,J + f6i+l,j
VKH\I) 1
zi+rzi-i +( zi+rzi-iiUrzi-i)
= i lJz(uS+iuSi-uMu[1n l-(l +* )
At z.in-z_. 1 R I RI L I LN z. ,-z. ^ z..,-z. z.-z. ;
i+1 "i-1
i+1 "i-1 "i+1 "i "i "i-1
-V2(Ur-|ur1)
zi+rzi-i
f =
+ izi+rzi.iKzi+rzV)
and
u
VR =
f
i J
u
L
u
f
1-1J
V
f
i J-l


CAN CENTER TEMPERATURE (c)
143
Figure 32.
Transient center temperature during conduction
cooling of Bentonite in a cylindrical can with
and without drop in retort pressure


28
Non-dimensionalization of the Variables
The variables in the equation are then non-dimensionalized as
be!ow.
z =
u =
5
a
t =
t
H2
0 =
T T1
To- T1
H2 -
O) = (i)
a
* = TG *
a
(25)
The resulting equations are the governing equations for natural
convection in the form used for the present study.
ae r a(u0l 1 a(rv0) , a20
T? + i + 7 Sr ¡ = (
, _1 ^ r 9 'i i
+ r ar^rar J J
(26)
3o)
Tr
+ {
a(uc) a(vt) j. nr,r a r l a(ru>) ^ ,
3z + -Tr~ } GrPr% + Pr{ -gj? + -gjrC 7 5p- ) J
- )
l a^jj ^
7Sr J
(27)
(28)
1 8i|>
r lz
_ 1 a
" TW
v
(29a,b)


17. Locations of the slowest heating points in water over
a heating period of 10 minutes in a cylindrical can 116
13. Migration of the slowest heating points in water over
a heating period of 10 minutes in a cylindrical can 116
19. Transient temperature at various locations during
heating of water in a cylindrical can 119
20. Transient temperature at various locations during
initial times of heating of water in a cylindrical
can 121
21. Temperature at the slowest heating point during
natural convection heating contrasted with that
during conduction heating in a cylindrical can 122
22. Isotherms during natural convection heating contr
asted with those during conduction heating in a
cylindrical can 123
23. Diagram of 39(0.9)x39(0.8) grid on the left and
58(0.85)x58(0.76) on the right 126
24. Isotherms after 90 seconds of heating computed using
39x39 grid (left half of the figure) compared with
computations using 58x58 grid (right half of the
figure) 127
25. Streamlines after 90 seconds of heating computed
using 39x39 grid (left half of the figure) compared
with computations using 58x58 grid (right half of the
figure) 128
26. Transient slowest heating temperatures computed with
39x39 and 58x58 grid .....129
27. Transient temperature at l/3rd height on the center-
line computed with 39x39 and 58x58 grid 131
28. Time step history during a typical run of the
numerical heat transfer model for natural convection 132
29. Isotherms in a cylindrical can calculated with two
different convergence criteria for boundary vorticity
(6 =0.0001 on the left and 6 =0.001 on the right) 135
30. Streamlines in a cylindrical can calculated with two
different convergence criteria for boundary vorticity
(6 =0.0001 on the left and 6 =0.001 on the right) 136
xi v


164
Mulvaney, S. J., and S. S. H. Rizvi. 1984. A microcomputer
controller for retorts. Trans, of the ASAE 27(6):1964-1969.
Navankasattusas, S., and 0. 8. Lund. 1978. Monitoring and
controlling thermal processes by on-line measurement of
accomplished lethality. Food Tech. 32(3):79-83.
Niekamp, A., K. Unklesbay, N. Unklesbay, and M. Ellersieck. 1984.
Thermal properties of bentonite-water dispersions used for
modelling foods. J. Food Sci. 49:28-31.
O'Brien, G. G., M. A. Hyman, and S. Kaplan. 1950. A study of
numerical solution of partial differential equations. J.
Mathematics and Physics 29:223-251.
Ozisik, M. N. 1980. Heat conduction. John Wiley and Sons, New York.
Parmentier, E. M., and K. E. Torrance. 1975. Kinematically consis
tent velocity fields for hydrodynamic calculations in curvi
linear coordinates. Journal of Computational Physics 19:404-417.
Peaceman, D. W., and H. H. Rachford, Jr. 1955. The numerical
solution of parabolic and elliptic differential equations. J.
Society Industrial Applied Mathematics 3(1):28-41.
Peyret, R., and T. D. Taylor. 1983. Computational methods for fluid
flow. Springer-Verlag, New York.
Raithby, G. 0. 1976. Skew upstream differencing schemes for problems
involving fluid flow. Comput. Methods Appl. Mech. Engg. 9:153-
164.
Richtmyer, R. D., and K. W. Morton. 1967. Difference methods for
initial-value problems. Interscience Publishers, Inc., New York.
Roache, P. 1978. Semidirect calculation of steady two- and three-
dimensional flows. Numerical methods in laminar and turbulent
flow. Proceedings of the 1st international conference at
University College Station, Swansea, U.K. Edited by C. Taylor,
K. Morgan and C. A. Brebbia. John Wiley and Sons, New York.
Roache, P. 1982. Computational fluid dynamics. Hermosa Publishers,
Albuquerque, New Mexico.
Saul'yev, V. K. 1957. A method of numerical solution for the
diffusion equation. Dokl. Akad. Nauk SSSR. 115(6):1077-1079 and
117(1):36-39.
Saul'yev, V. K. 1964. Integration of equations of the parabolic type
by the method of nets. Translated from Russian by G. J. Tee.
Pergamon Press, New York.


15
flow problems are generally more complex than the external ones.
Fewer works have been reported on internal flows in cylindrical
coordinates. Of the studies on natural convection heating in
cylindrical enclosures, only four studies could be located where
they were in reference to natural convection heating during the
thermal processing of liquid foods. Such thermal processing is
characterized by a rather large Grashof number, generally small
overall dimension of the container and water as the liquid. Perhaps
the food processing situation is unique to general heat transfer
studies, which would explain the fact that only one such study could
be located in the more common heat transfer literature.
Previous Works on Canned Liquid Foods
The rate of heat transfer in canned liquid foods was studied
using correlations, and slopes of time-temperature curves with
limited success (Blaisdell, 1963; Barreiro-Mendez, 1979). Possible
flow patterns were postulated by Blaisdell (1963) using dye and
aluminum powder tracers.
Stevens (1972) first tried to study detailed temperature and
flow patterns during natural convection heating of canned liquid
foods. He used an explicit finite difference formulation from the
work of Torrance and Rockett (1969). The can was treated as a ver
tical cylindrical enclosure. The fluid (ethylene glycol) inside the
can was initially at rest at a uniform temperature of 30C. The can
was then placed in hot water bath at about 100C. The top surface of
the fluid was free (the can had headspace). The actual temperature


19
treated as insulated and wall temperature specified as a function of
time. The numerical results were in good agreement with experimental
data.
Sani* (personal communication, 1985) also developed a similar
model for pasteurization of beer in cylindrical metal cans (aspect
ratio =4.2). Water was the test fluid (Pr=6.81), The initial
temperature was 35C and the sidewall boundary condition used was
approximately 60C. The top of the water level was free (the can had
headspace) and was treated as insulated. The temperature at the
bottom of the can was a function of radial distance. Temperature and
velocity profiles from this study are given in figure 2. Further
quantitative details on this study were not available. These results
were not compared with experiments and also not published.
Other Similar Works on Natural Convection Heating in Cylindrical
Enclosures
One of the earliest of the numerical studies on the subject of
natural convection heating in cylindrical enclosures was performed
by Barakat and Clark (1966), who were looking at thermal strati
fication and associated processes in liquid propellant tanks for
application to space flight. The vertical cylindrical enclosure,
they studied, was filled with water, and arbitrary temperature and
heat flux variation were imposed on its wall, while its bottom was
kept insulated. At the top free surface of the liquid, heat transfer
to the ambient was considered negligible and thus insulated. They
*R. L. Sani, CIRES, Campus Box 449, University of Colorado, Boulder,
Colorado 80309


After algebraic manipulation
, n+1 k n+1 n+1 ^
-a -to. + D.). + c.,1 = rhs.
J i>J-l j i ,j j i,j+l j
(56b)
where
lh (v|vl|)
Y_tVHr WVhW
Pr rj-l
r n, i ni n, i ni >,
1 (VIvrK+IvlN
bj' 4? +
* vR|~VivlI J Pr rj r 1 1
2trj+rrj-iJ lrj+i'ri-i) rj-ty> trj+i'rj) Wi
Cj
WKl) ....
rj+i"rj-i trj+rrj-i>jJ/2 lrj+rrj)
Pr rj+l
rhs = do)1? .+e)? .+fajr?,1 .
J i-l,J i,j l+l,j
-0.5GrPr2
(Gn+l+1-0n+l .)
^ iJ+l
rj+rrj-i
V2(uMuP|) Pr
zi+rzi-i C zi+rzi-iH*i-zi-i)
e
1_
At
i-l
Pr
zi+rzi-i
zi+rzi
zi-zi-i
f
zi+rzi-i

115
Slowest Heating Zone
In the absence of agitation, there is very little mixing of the
liquid during natural convection heating. Consequently, the liquid
does not experience the same time-temperature history at all points
inside. Points at the top stay at a consistently higher temperature
than at the bottom. However, at the very bottom, since the bottom
wall is heated, the temperature is higher again. Somewhere in
between the liquid stays consistently at a lower temperature than
the rest of it. From a food processing point of view, these slower
heating points (or regions) are of utmost importance. The liquid
(and therefore the bacteria carried with it) at these points experi
ences much less exposure to heat (less sterilization) than the rest
of the product. Several studies in the past (Blaisdell, 1963;
Hiddink, 1975: Stevens, 1971) attempted to locate these slowest
heating points under various situations. For this study, these
slowest points were investigated with respect to locations,
migration, and temperature values at different times during heating.
Location of the slowest heating points. To find the locations
of the transient slowest heating points, the lowest temperature at
any time instant can be located. When more than one location was
found with the same lowest temperature, the point closest to the
bottom, and closest to the centerline was chosen. In figure 17,
these points are plotted for every 30 seconds during the heating.
The figure points out that the same physical location in the
container does not stay the coldest at all times. This is in
contrast to conduction-heating where symmetrical heating makes the


107
TT3TTT223L
V * ***^*-
V L- Ju r.
120.99C
t
V !* ic i Ufi
\ vt i ///^
I i l
il
llllUF
/
120^96^0 _
Water
T,=30*C
Top-lSlC
Sld8=121*C
3or-121*C
R-4.19cm
H=10.67om
N=19
M = 19
5=0.0001
After 1600 sec
Figure 10o. Isotherms and velocity vectors in a
cylindrical can after 1800 seconds of heating


Axial velocity Axial velocity Axial velocity(mm/sec)
40
111
Figure 14. Predicted radial velocity profiles at the
sidewall at mid height in a cylindrical can
after various heating times


68
that makes the ADI method advantageous is not available in general.
Programming the ADI method for a non-rectangular region may get
complicated, and they are not known for sure to be faster (Roache,
1982).
Another popular method to solve the unsteady formulation of the
stream function Poisson equation is the successive over-relaxation
(SOR) method (Frankel, 1950). It is very simple and effective. In

SOR, the set of solutions are over relaxed by a factor 1 < n < 2
to speed up convergence. The exact value is found through simple
numerical experimentation. It is easier to program compared to ADI
methods. Most of the earlier researchers have used the SOR to solve
the stream function equation. In all the five formulations of
natural convection in a cylindrical geometry, studied by Torrance
(1968), SOR was used to solve the stream function equation. However,
it is time consuming and could end up taking 90% of the total time
spent to solve the problem. Nevertheless, SOR was chosen to solve
the stream function equation in the present study for its
simplicity. Following is the development of the SOR formulation.
Rewriting equation 62,
To get the SOR formulation,
'.0
F (A*1-,j+c*i+l,j+0*i J-l+E*1,j+l+l,j ) (70)


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
NUMERICAL MOOELING OF NATURAL CONVECTION
AND CONDUCTION HEAT TRANSFER IN CANNED FOODS
WITH APPLICATION TO ON-LINE PROCESS CONTROL
By
ASHIM KUMAR DATTA
December 1985
Chairman: Arthur A. Teixeira
Major Department: Agricultural Engineering
A numerical model was developed for predicting detailed flow
patterns and temperature profiles during natural convection heating
of liquids in a can (a cylindrical enclosure). The liquid (water)
was initially stagnant at a uniform temperature of 30C, and the
sidewall, top and bottom wall temperatures were suddenly raised to a
temperature of 121C. A free (thermally insulated) top liquid
surface was also considered. Boussinesq approximation was used and
all other fluid properties were treated as constant.
Finite difference methods were used to solve the governing
equations in axisymmetric cylindrical coordinates. A vorticity-
stream function formulation was used. Parabolic temperature and
vorticity equations were solved with the alternating direction
implicit method and upwinding the convective terms. The elliptic
stream function equation was solved using successive over relaxation
xvi


60
The tridiagonal matrices for the vorticity equations at the two
ADI time steps are given below. At the "half" time step, equations
are written for the columns (from the bottom wall to the top wall).
Both the bottom and top wall vorticity values are used from previous
iteration.
bl -ci
-32 b2 -C2
-33 b3 -C3
-am-2 bm-2
am-l
()*i
l.J
rhsl+alw0,j
r-j
CM
3
rhs2
3,j

rhs3
cm-2
V2,j
rhsm-2

bm-l
rbsm-l+am-l m,j
(57a)
At the "full" time step, equations are written for the rows
(from centerline to sidewall). The centerline boundary vorticity is
specified (=0) and the sidewall boundary vorticity is used from
previous iteration.
Yl+l
bl "C1
i.l
~a2 b2 c2
1,2
a3 b3 c3
wi ,3
=
I -an-2 bn-2 _cn-2
i ,n-2
l an-l bn-l
l
. ,, n+1
rhsl+al rhs rhs-:
rhsn_2
(57b)


Water
T,=30'C
top-121*C
slde=121*C
bot-121*C
R-4.19om
H=10.67om
R-57.0.76
Z=57,0.85
a-o.oooi
-300
-260
-220
-180
-140
-100
-60
-20
20
6
Figure 10c. Isotherms and streamlines in a
cylindrical can after 3 seconds of
heating
Water
T,= 30*C
top-iei*C
slde=121*C
boi-121'C
R-4.19cm
h=10.07cm
R-57.0.76
2=57,0.85
a-o.oooi
-300
-260
-220
-180
-140
-100
-60
-20
20
60
After 4 sec
Figure lOd. Isotherms and streamlines in a
cylindrical can after 4 seconds of
heating o
O


32
with the initial conditions
0 < r < R/H 0 < z < 1 0=1 (34)
Computer Control of Conduction-Heating Processes
In controlling thermal processes, the objective is to meet the
designed level of bacterial sterilization (F^) for the process,
irrespective of any retort temperature variation TR(t), and with a
minimum of overprocessing. The lethal effects of thermal processing
are achieved during the heating as well as cooling times (t^ and
tc). Thus, the objective function in thermal processing is to
t Jill1 t +t Id21
minimize F0(th+tc^ = ^ 10 ^ dt + c 10 Z dt (35a)
uh
subject to the constraint: F0(th+tc) > ^o (35b)
where the integral from equation 6 has been broken into separate
heating and cooling times. FQ(t) is the accumulated sterilization at
any time t. The first integral above is the contribution to the
sterilization F0(t) from heating, and the second integral is the
contribution to the sterilization during cooling. Temperature T(t)
is taken to be the temperature of the slowest heating point in the
product. This is so that when the design Fq(Fq) is satisfied at
this point, all other points in the product have also been satis
fied. For a cylindrical can, T(t) is the temperature at the can
geometric center at time t.


29
The boundary conditions in the transformed and non-dimensional ized
variables can be written as
sidewall 1
r=R/H
0 < z < 1
0=0
o
11
u=0
v=0
centerline
r=0
bottom wall
0 < z < 1
i£ = o
9r u
O) = 0
-e-
n
o
U 0
3r
v=0
0 z=0
0=0
o
ii
u=0
v=0
top wall
0 Z=1
0=0
4> = 0
u=0
v=0
(30)
with the initial conditions
0 0 < z < 1
0=1
(o = 0
i|) = 0
u=0
v=0
(31)
The relative merits of solving the transformed variables
as compared to the (u,v,p) system of primary variables were discuss
ed by Roache (1982). To solve the (u,v,p) system, another elliptic
Poisson equation would need to be formed for the pressure (p) vari
able, using the two equations for velocity, regardless of whether
transient pressure solutions are desired. Following are the relative
evaluations of the (u,v,p) system and the ) system for two-
^Explicit boundary conditions for vorticities on solid walls are
unavailable.


45
depends on ij> which in turn depends on to this would couple all the
equations together and make it practically incomputable. To avoid
this, the nonlinear terms are linearized. Various ways of lineari
zing these terms have been discussed by researchers (Anderson et
al., 1984; Peyret and Taylor, 1983; Roache, 1982). A very simple and
common strategy in linearizing is to evaluate the velocity coeffi
cient at n time level. This is known as "lagging" the coefficients.
However in this case, the accuracy in time drops from second order
(for ADI) to first order. Roache (1982) noted that such "lagging" of
the velocities could still be second order accurate if the velo
cities u and v were slowly varying. An improvement over this
linearization procedure is to update the coefficients u and v as
iterations proceed. The procedure is repeated until the values u and
v converge. One way to achieve this would be to update un+^2 and
vn+V2 by solving the stream function equation at the "half" time
step of the ADI (Roache, 1982). Since solution of the Poisson stream
function equation is the most time consuming operation, it was
decided against this updating procedure.
Anderson et al. (1984) noted that "lagging" the coefficients
has been widely and satisfactorily used by previous investigators,
and, recommended the use of the same for its simplicity. Thus the
velocity coefficients u and v were "lagged" in this finite differ
ence study.
Convection terms. The presence of the convection terms makes
the temperature and the vorticity equation qualitatively different
from other simple parabolic equations. Proper numerical treatment of


16
variation on the can boundary was noted and was used in the
numerical calculations. Temperature values were recorded at ten
different points inside the container. The recorded temperature
values agreed poorly with numerically calculated temperatures. The
numerical inaccuracy was attributed to insufficient grid points
which could not resolve the details of the flow. Several sources of
experimental error included inaccurate setting of boundary
temperature, non-zero initial velocity field, errors in thermocouple
measuring circuit and inaccurate placement of the thermocouples.
Some of these factors were thought to have caused three dimensional
movements in the experiment and could not be picked up by the two-
dimensional numerical model.
Detailed temperature and velocity profiles during natural
convection heating of liquid foods in a cylindrical can were
extensively investigated by Hiddink (1975). The numerical solution
technique used was the explicit finite difference technique of
Barakat and Clark (1966) discussed later. The initial temperature
was considered uniform throughout. The top, bottom and sidewall were
suddenly raised to the heating temperature. The top surface of the
liquid was considered free (having headspace) and was thus treated
as thermally insulated. Several test fluids of low and high
viscosities (water, sucrose solution and silicone fluid respec
tively) were used. The numerical results included temperature and
streamline patterns. The overall flow patterns were visualized using
a "particle streak method." In this method small glittering
particles are suspended in the liquid. The particles are illuminated


REVIEW OF LITERATURE
This chapter begins with a brief introduction to thermal
processing as applied to the sterilization of canned foods. The need
for computer control of thermal processing is noted, along with the
importance of heat transfer models in such computer control systems.
Previously reported conduction heat transfer models used in studying
sterilization of conduction-heated canned foods are described next.
The few studies attempting to model natural convection heat transfer
in the case of sterilizing canned liquid foods are noted. The few
other works on natural convection heat transfer in liquid inside a
vertical cylindrical enclosure are also reported.
Computer Control of Food Thermal Processing
An overview of thermal processing is presented in this section.
Problems with traditional approaches to processing of conduction-
heated foods are then noted and the need for on-line control is
stressed. Some of the past approaches in thermal processing of con
duction-heated foods that could also be used on-line are discussed.
Overview of Thermal Processing
Thermal processing consists of heating foods, filled and
hermatically sealed in containers, in pressurized steam retorts at a
5


58
After algebraic manipulation
* *
-a w. i + b-co* -e-Co.
i l-l,j i i,j
. u).,, = rhs
i i+l,j i
where
(56a)
V2 (u"+|u"|) Pr
31 = zi+rzi-i +12i+rzi-iJizrzi-i)
bi
i_ + ljl (uMuJI-uMuPIl + ( 1 + -k 1
At z.,,-z. ^ R I R! L I !_N z..,-z- ^ z.,,-z. z.-z. J
i+l i-1
i+l "i-1 "i+l "i "i "i-1
ci
V2 (ur'1urI) + Pr
zi+rzi-i izi+rzi-i^zi+rzi^
rhs. =
fe^1 -en.+l. )
Do)9 ,+Ew; .+Fc!¡ .-0.5GrPr2
i,J-l i,J i,J+l rj+l"rj-l
d =
V2(vJ+|v"|)
Pr r
j-1
(rj+rpj-ij + rj^irJ+i-rj-iKrrrj-iJ
1 ,.K+KK+Kn
-fc-Vk
r.Ll-r. ,
J+l J-1
lrj+rrj-i)tV/2trj+rrj) +
^2 (VR~1VR I)
rj+rrj-i irjtrrj-i>jJ/21'j+r'V
Prjj+l


Water
T,= 30*C
1op-lSl*C
slde=121*C
bot-121*C
R-4.19cm
H**10.67om
r,- tor-iti*c
bor-iei*C
R-4.10offl
Hal0.07*m
After 300 sec
Figure 22. Isotherms during natural convection heating
contrasted with those during conduction
heating in a cylindrical can


LIST OF FIGURES
FIGURE PAGE
1. Isotherms and velocity profiles in a glass bottle at
various heating times (after Engelman and Sani, 1983) 18
2. Isotherms and velocity profiles in a can at various
heating times (after Sani, 1985) 20
3. The grid system and the boundary conditions for the
cylindrical can 40
4. Notations of various quantities defined on a non-
uniform grid system in cylindrical geometry 62
5. Algorithm for the iterative solution of the set of
equations in natural convection 77
6. Experimental setup for conduction heating 84
7. Timings of various real-time computations for on-line
control of conduction-heated food 88
8. Flow diagram for computer control of retort opera
tions with on-line correction of process deviations 90
9. Isotherms near bottom wall predicted by the
convective model compared with the predictions
considering conduction (equation 83) only 97
10a. Isotherm and streamlines in a cylindrical can after 1
second of heating 99
10b. Isotherm and streamlines in a cylindrical can after 2
seconds of heating 99
10c. Isotherm and streamlines in a cylindrical can after 3
seconds of heating 100
lOd. Isotherm and streamlines in a cylindrical can after 4
seconds of heating 100
lOe. Isotherm and streamlines in a cylindrical can after 5
seconds of heating 101
xi i


46
the convection term is important. Several different ways of finite
differencing these terms have been presented in the previous works.
These include space-centered differencing and upwind or upstream
differencing. The upwind methods have several desired properties
that are not shared by the other finite difference formulations, as
applied to the convection term. Both the upwind methods are
transport!'ve, meaning any perturbation is carried along only in the
direction of velocity. Both the methods are also conservative so
that they preserve the integral conservation relations in the
continuum equation. The first of the two upwind differencing methods
is as follows (Lilly, 1965; Forsythe and Wasow, 1960; Frankel, 1956)
a(ue) u6i u9i-l
9z Az
u9i+r u6i
AZ
(u > 0)
(u < 0)
(41)
Accuracy of this method is 0(At,Az). This was used by Barakat and
Clark (1966) to solve equations 26 through 29. The second variation
of this upwind method is to treat the convection term as follows
(Gentry et al., 1966)
9(u0)
3Z
Vr
Vl
AZ
where


BIOGRAPHICAL SKETCH
Ashim Kumar Datta was born in the city of Hooghly and raised in
the city of Krishnagar in the state of West Bengal in India. He
attended the Collegiate school in Krishnagar where he received the
Higher Secondary degree in the Fall of 1974. He completed his
Bachelor of Technology in agricultural engineering from the Indian
Institute of Technology at Kharagpur in the Fall of 1979. He then
attended one year of post-graduate studies in statistical quality
control and operations research from the Indian Statistical
Institute, Calcutta. Later, he proceeded to receive his Master of
Science degree in agricultural engineering from the University of
Illinois at Urbana-Champaign in the Summer of 1982. He hopes to
receive his Doctor of Philosophy degree in the Fall of 1985, majoring
in agricultural engineering, from the University of Florida at
Gainesville. He is married to Anasua Datta.
172


138
would change in case of a non-uniform coordinate system. Parmentier3
recommended that equation 89 be used even in a non-uniform grid
system. This means grid faces would still be half way between grid
points in a non-uniform coordinate system. The recursion relation of
X- (equation 68) would probably change with a non-uniform
J
coordinate system. However, in this study, the same x- (equation
J
68) was used as in the case of a uniform coordinate system. Another
interesting point noted by Parmentier*5 was that the need for using
kinematic consistency criterion was, less critical for fine grid
system and vice versa.
Conduction Heating in Cans
The importance of surface heat transfer coefficients during the
actual processing of conduction-heated food is described first in
this section. The temperature values actually measured in the can
center are then compared with the corresponding values predicted by
the ADE finite difference method and the analytical solution. After
thus confirming the accuracy of the finite difference method of
solution, a discussion is presented on how drastically such
prediction using finite differences would be wrong in case there is
sudden pressure drop during the process.
a^E. M. Parmentier, Box no. 1846, 324 Brook Street, Brown Univ.,
Providence, RI 02912. Personal Communications.


55
The tridiagonal matrices for the temperature equations at the
two ADI time steps are given below. At the "half" time step, the
equations are written for a column (going from the bottom wall to
the top wall). Specified boundary temperature is used both at the
top and bottom walls. #
bl Cl
-32 t>2 "c2
-93 b3 "c3
rhsi+ai0>j
rhS2
rhs3
am-2 bm-2 cm-2
"am-l bm-ll
m-2,j
m-1 ,j
rhsm_2
rhsm-l+am-l0m,j/
(49a)
At the "full" step, equations are written for the rows (from
the centerline to the sidewall). At the centerline, the insulated
boundary condition (equation 75) is used due to symmetry. At the
sidewall, the specified boundary temperature is used.
bl"al C1
-a2 b2 -c2
- -83 b3
'c3
V
rhs^
i ,2
rhs?
0i ,3
rhs3
bn-2 "cn-2
Gi ,n-2
rhsn_2
_an-l bn-l
p,n-l
irhsn-l+cn-lQ^n
(49b)


52
30 A r 3(U0), 1 3(rv0) r 320 ,13 r_30 ^ ,
3t i ~W~ + p 3r i = 1 Tz2 + r Sr'-'V J >
At the "half" or intermediate step ("n+V' written as in short)
* n
0. .-0. .
i ,3 i ,3
At/2
. (vKi v^,,- -Wi
rjlrj+rrj-i)
* * *
0 0.,. .-0. 0. .-0. n .
2 r 1+1, J 1,3 1,J 1-1,J i
zi+rzi-i zi+r2i zrzi-i '
Jir:+rrj-i)l
(\ t _n _n \
rj+v? ^9j j+r0T j ri- yj6i ,j~9j ,j-i^
rj+r rj
r.-r. ,
0 J-l
(47a)
At the "full" step
o"+*-0* .
i ,J i ,J
Ax/2
"i +1
^ V KIK+1.J ^ur+ Kl- <+ C^NlK-u
, n i nM n+l n. nM ji+1 n i nM ji+1 n. nn n+l
< vKi^i,3+i+(vivRi>jVi.r(vIvlI,i -cvl+i*Li)pji/ii,j-i
rjtrj+rrj-i-1
*
0.l1 .-0. 0. .-0. .
1+1,J 1 ,J 1,3 1-1,Jj
zi+rzi-i zi+rzi
zizi-l
,n+l ji+1
n+1 jn+1
+ r. (r.,i-r. V)^
J 3+1 3-1
j+
/ JITi J|Ti N / J|Ti .11 i I \
V? ^ ,j+l"T ,J rj- V? ^1 3 ~ 1 j-!
rj+l'rj
r.-r. ,
J J-l
(47b)


48
3(u0) a(u0) 3n
3z 3ri 3z
_ 3(u0) 1
3r) 3z/3n
Un0. U. 0. 4 .
R i,] L i-l,j
An
zi+r zi-i
2 An
ziU' z
uR01,j uL01-l,j) R>0- UL>0
Si miliarly
3(u0)
3z
z'Vx-- zi'J1 Kei+l,j Vi-1,j) UR < 0 UL > 0
O
- fun0-,-. u, 0- 1 Uo < 0 Ui < 0
zi+r zi-i 1R 1 1,3 L K L
z2--T ^R6i,j uLei,j) Ur > 0 UL<0
(a)
(b)
(c)
(d)
To avoid the use of FORTRAN IF statements for checking the sign of
ur and ul in the equations (a) through (d), the equations are
combined as below
3(u0) (UR~IUR
5F" "
i+i j+(VIur1~VK1)0 j~(V1ul1)9t-ij
r
Zi+1 zi-l
(43)


Dedicated to the memory of my
grandfather and father


158
For points other than center and with i, j decreasing (moving towards
center)
de 320 320 _1_30
3t ~ 3z2 + 3r2 + r 3r
eO+t
0?
hi -
At
0? 01? 0?+^+ Q"+] .
1-1,J 1 ,J 1 ,J 1+1, J
2
+
f. 01? Q^1.- 0^1 ,
1>J1 1.J 1,J 1,J+1
AZ2
en. ..
i J-l
0
n+1
i J+l
2Ar
e?H =
',J l+p+S
b{s?.j+s(9?-i.r e"l Cj>
* p<0U-r 9u+ 1
(93)
For points at center (j=0), for use while i ,j decreasing
30 320 320
3t = JzT + 1FI
,n+l
,n
,n
,n
0': &: n- o'.1 n- 0. i+ 0^i n 2 (0? o'? n- 0^+i+ ? i)
1,0 1,0 1-1,0 1,0 l ,0 1+1,0 1 i,-l i,0 1,0 i,1J
At Az2 aF2
0n+1
i,0
le?,o+s(0i-i,o- sm+ 9i+i,o)+2P(0i,r lo+ 0>
n+1
1+2P+S
(94)


88
Calculate T(t)
8 update f^(t)
Simulate cooling
to estimate F0C0*
At
2
H
t 1
measure
T(t)
t+At
T
measure
TR(t+At)
where
At. = Time taken to calculate T(t) from measured T(t)
1 using finite difference approximation to equation 4,
At2 3 Time taken to simulate cooling assuming it starts at
time instant t, and
At = Time interval between reading retort temperatures
TR(t).
Figure 7.
Timings of various real time computations for
on-line control of conduction-heated food


74
order form destabilized the calculations (Wilkes & Churchill, 1966)
and sometimes it was less accurate than the first order formulation
(Beardsley, 1969).
The vorticity boundary value at the computational intermediate
or "half" step of the AOI method is more difficult to.set than the
corresponding temperature value because the vorticity boundary value
is also a function of time. The following simple way to define this
vorticity value on the boundary is often used and gives sufficiently
accurate results (Bontoux et al., 1978)
n+vr1
w
1/2 (
n+1
w
m+1
(77)
where
n+1
m+1
wall vorticity value at time step n
most recent estimate of wall vorticity at n+1
step using equation.
n/2"
W
m+1
intermediate vorticity value between n and
(n+l)th time level at the (m+l)th iteration.
Israeli (1972) also used this equation to define intermediate (half
step) boundary vorticity though he used a different equation to
+,m+l
calculate boundary vorticity ^ 1
Convergence Criteria for Iteration
As discussed earlier, the set of finite difference equations
were solved iteratively until convergence. The measure of conver-


163
Hiddink, J. 1975. Natural convection heating of liquids, with
reference to sterilization of canned foods. Agricultural
Research Reports #839. Center for Agricultural Publishing and
Documentation, Wageningen, Netherlands.
Higgins, J. M. and 8. Gebhart. 1983. The stability of vertical
buoyancy-induced flow in cold water. J. Heat Transfer, Trans, of
the ASME 105:767-773.
Hornbeck, R. W. 1975. Numerical methods. Quantum Publishers, New
York.
Israeli, M. 1972. On the evaluation of iteration parameters for the
boundary vorticity. Studies in Applied Mathematics LI(1):67-71
Jaluria, Y. 1980. Natural convection heat and mass transfer.
Pergamon Press, Oxford, England.
Jensen, V. G. 1959. Viscous flow round a sphere at low Reynolds
numbers (<40). Proceedings of the Royal Society of London,
Series A, 249:346-366.
Ka'lney de Rivas, E. 1971. On the use of nonuniform grids in finite-
difference equations. J. of Computational Physics 10:202-210.
Kee, R. J. 1974. A numerical study of natural convection inside a
horizontal cylinder with asymmetric boundary conditions. Ph.D.
Thesis. University of California, Davis.
Kee, R. J., C. S. Landram, and J. C. Miles. 1976. Natural convection
of a heat-generating fluid within closed vertical cylinders and
spheres. J. Heat Transfer, Trans, of the ASME 98:55-61.
Kimura, S., and A. Bejan. 1984. The boundary layer natural
convection regime in a rectangular cavity with uniform heat flux
from the side. J. Heat Transfer, Trans, of the ASME 106:98-103.
Kublbeck K., G. P. Merker, and J. Straub. 1980. Advanced numerical
computation of two-dimensional time-dependent free convection in
cavities. International Journal of Heat and Mass Transfer
23:203-216.
Lilly, D. K. 1965. On the computational stability of numerical
solutions of time-dependent nonlinear geophysical fluid dyna
mics problems. Monthly Weather Review 93(1):11-26.
LOG-TEC CCS-8 Retort Management System. 1984. Central Analytical
Laboratories, Inc., 2600 Marietta Street, Kenner, Louisiana.
Manson, J. E., C. R. Stumbo, and J. W. Zahradink. 1974. Evaluation
of thermal processes for conduction heating foods in pear-shaped
containers. J. Food Sci. 39:276-281.


ADI for the vorticity equation
£ f + 1W ( -GrPr^ + Pr{ & ^ I I'M ) (
At "half" or intermediate step,
a). to. .
1 ,J 1 ,J
At/2
, 1 rr n i nM ,r n.i ni n.i nn ( n.i nn i
ZHrzi-l f(Vl'JRlh+l,j+(VlUR|-uL+luLlh,j-K+l'JLlh-l,j !
1 rr n i ni >, n lI- n^i n< n^i ,ni >, n r n.i.ni > n
{(vR-|vRl)i,j+l+iVlvR|-vL+lvLl)i,j -CVIvlD1,M
*
-0).
* *
0). C
fln+l _n+l *
= _GrPr2 (Isitlliill) + 2 pr { ii+lj~hii ,j i J'^i-l.j j
r *< z < Z > < z 4 ** z z -z 4
i l-i
rj+rrj-i
zi+rzi-il zi+rzi
n
2 Pr
lrj+rrj-iJ
At "full" step,
n+1 *
(J .
1 ,J 1 ,J
A/2
rj+i mi,j+rrj Mitj
j+
r. . r.)
1/2lrj+l
n n
r. to. .-r. i o). 1
J T ,J Jl 1J-1
rj-V2^-rj-iJ
(55a)
1 rr n i ni-, L(- nLi ni n^i ni ^ r n.i nn i
+ I-^-j-i((uR-|uRm+ijj+{v|uR|-uL+|uL|)ijj-(vluLDi-l,jl
1 ,, n i nM n+1 ,r n,i nt n L ( n i ^ n+1 ( n ni-, n+1 ,
+(rjtrrj-T>f(Vr'1 VrI )i .j+l (VlvrI-vlvlI )1,3 -(vlVJ H ,j-l)
* *
0). -0).
fln+l n+1 *
= _GrPr2 r9i ,j+l~9i ,j-l 3 + 2_Pr rM1+lfj;.MitJ JZlzUi
rj+rrj-i zi+rzi-i zi+rzi zi-zi-i
n+1 n+1 n+1 n+1
2 Pr rrj+lui ,j+lrjMi ,j rji,j "rj-l ui,j-l i
(rj+rrj-i) 1 rj+ v2 lrj+i-pjj" rj-v2l rj'rj-iJ
(55b)


168
Clever, R., and I. Catton. 1972. Steady natural convection in a
vertical cylinder at large Prandtl number. Proceedings of the
1972 Heat Transfer and Fluid Mechanics Institute, edited by R.
8. Landis and G. J. Hordemann.
Cunsolo, D., and P. Orlandi. 1978. Accuracy in non orthogonal grid
reference systems. Numerical methods in laminar and turbulent
flow. Edited by C. Taylor, K. Morgan, and C. A. Brebbia. John
Wiley and Sons, New York.
Dickerson, R. W. Jr. 1969. Simplified equations for calculating
lethality of the heating and cooling phases of thermal
inactivation determinations. Food Tech. 23(3):108-111.
Evans, L. B., and N. E. Stefany. 1966. An experimental study of
transient heat transfer to liquids in cylindrical enclosures.
Chemical Engg. Progress Symposium Series 62(64):209-215.
Fairweather, G., and A. R. Mitchell. 1967. A new computational
procedure for ADI methods. SIAM J. of Numerical Analysis
4(2):163-170.
Fitzpatrick, J. A. 1980. The application of computers and
electronics to process control in Thames (sugar) refinery. The
International Sugar Journal 82(980):231-236.
Harrell, R. C., G. A. Kranzler, and D. C. Davis. 1981. Micro
processor control of a solar fruit juice pasteurizer. ASAE paper
no. 81-5030. Paper presented at the 1981 summer meeting of the
ASAE at Orlando, Florida.
Hayakawa, K. 1974. Response charts for estimating temperatures in
cylindrical cans of solid food subjected to time variable
processing temperatures. J. Food Sci. 39:1090-1098.
Hayakawa, K. 1978. A critical review of mathematical procedures for
determining proper heat sterilization processes. Food Tech.
32(3):59-65.
Heliums, J. 0. 1961. Finite difference computation of natural
convection heat transfer. Ph.D. Thesis. University of Michigan.
Ann Arbor.
Hess, C. F., and C. W. Miller. 1979. Natural convection in a
vertical cylinder subject to constant heat flux. Int. J. Heat
Mass Transfer. 22:421-430.
Hogue, R. L., and V. A. Jones. 1982. Hardware recommendations for
computerized steam injection UHT processing. Paper no. 82-
6017. Presented at the 1982 summer meeting of the ASAE at
Madison, Wisconsin.


42
Time derivatives. The treatment of the time derivative in
equations 26 and 27 can be broadly classified into implicit and
explicit methods. Explicit finite difference methods are generally
easier to understand and program. Explicit methods have been used by
Barakat and Clark (1966), Torrance (1968), Stevens (1972) and
Hiddink (1975) in studying natural convection heating inside a
cylindrical container. For most of the explicit methods, however,
there are stringent restrictions on the allowed time step, from the
stability considerations. In situations where large temperature
gradients, eddies and recirculations are present like in this study
grid points need to be fine enough to resolve the physical phen
omena. When using most of the explicit methods, such a fine grid
size could severely limit the allowed time step and thus require
prohibitive amounts of computer time.
Implicit finite difference methods involve more than one
advance time-level value in the same equation and thus require
matrix inversion to calculate function values at a new time level.
These methods generally allow larger time steps due to better
stability properties. However, even though a larger time step is
allowed in the implicit method, it requires many more calculations
to solve that step. For an implicit scheme to save computer time in
the overall computation, it should allow several times larger a time
step than that allowed by an explicit method to achieve a given
accuracy critrion.3
aT. I-P. Shih, Mech. Engg. Dept., Univ. of Florida. Personal Comm.


137
However, in the work on kinematically consistent velocities by
Parmentier and Torrance (1975) the half points were mid points
between grids in physical domain, i.e. they are defined by
>v2
V2(V rj+l>
(89)
but the equations developed in their study were meant for an uniform
grid system. For an uniform grid system, a is zero and the two
definitions of rj+l/2 mer9e into one* It is interesting to see that
apparently the same equation (87) can be derived with this
definition (89) of rj+l/,, by writing
3(rvQ) VRrj+ V?i ,j~ VLrj- V^i J-l +
0. v, r.
3r
r.
j+ Vi
rj-V2
0( Ar)
(90)
2^vRr.j+y?6j ,.j- vLri-v,3i..i-iJ
r.,.- r. n
J+l J-l
using equation 89
The catch is that equation 90 now turns out to be first order accu
rate due to "skewed" central differencing instead of being second
order accurate for regular central differencing as in equation 87.
Thus, use of rj+1/2 deflned by equation 89 together with equation
87 would drop the order of accuracy. However, it is expected that
with such a definition, transport velocities could be a better rep
resentative for the fluxes at the interface. It is unclear, from
their 1975 work how some of the equations for calculating velocities


2
Conduction-Heated Foods
Traditionally, for conduction-heated foods, the required
processing time is determined using the analytical solution to the
heat conduction equation. In practice, the constant medium temper
ature assumed by the analytical solution cannot always be achieved.
Thus, experimental correction factors have been proposed to adjust
the heating time, predicted by analytical solution, when the medium
temperature goes through certain (predefined) types of deviations.
However, deviations in practice are arbitrary, and it is not
conceivable to know the correction factors for all such situations.
Of course, the problem is trivial if the transient food temperature
is directly measured using a thermocouple; but this is both
inconvenient and impractical in a production situation. Better
models for predicting food temperature in response to arbitrary
fluctuations in medium temperatures are therefore needed.
A numerical finite difference model would be able to predict
the food temperature for truly arbitrary variations in the medium
temperature. However, the calculations required are relatively much
more complicated and time consuming. Using a microcomputer to
perform this would have two advantages. In addition to correctly
predicting food temperature for truly arbitrary heating conditions,
it can perform this automatically without any worker supervision.
Introduction of low-cost microcomputers for on-line process control
is becoming commonplace in the 1980s. Food industries are not an
exception to this situation. Use of microcomputers for on-line
control of conduction-heated food is thus quite conceivable.


33
This transient temperature T(t) is a function of transient
retort temperature TR(t), the can wall outside heat transfer coeffi
cient (h), dimensions of the can (R and H), and thermal diffusivity
(a) of the product for a conduction-heating food. Symbolically,
T(t) = f (TR(t), R, H, a) (36)
For conduction heating in a cylindrical can, the function f is the
heat conduction equation in cylindrical coordinates.
For a given can size and product, the two variables that can
possibly be controlled are the retort temperature TR(t) during
heating (saturated steam or hot water under pressure in the retort),
and the time period of heating t^. Cooling is generally done using
available water at ambient temperature. This makes the contribution
to F0 from cooling (given by the second integral in equation 35a)
immune from further control once the cooling process is under way.
Also, both the heating and cooling temperatures can only be applied
to the boundary of the product making it a boundary control problem.
In the case of a process deviation, the heating medium tempera
ture may go through unexpected variations beyond the capacity of the
temperature controller. Thus in reality, only one variable can be
controlled. It is the time period of heating (t^). The problem in
thermal processing is to specify t^ for arbitrary variations in
TR(t) during heating while the process is under way in real-time.
Instead of searching for the proper t^ through simulation every
time a deviation takes place as suggested by Teixeira and Manson
(1982), F0(t) can be monitored in real-time using T(t) calculated


SIDEWALL
97
BOTTOM
Figure 9. Isotherms near bottom wall predicted by the
convective model compared with the predictions
considering conduction (equationy<83j only
/


NUMERICAL MODELING OF NATURAL CONVECTION
AND CONDUCTION HEAT TRANSFER IN CANNED FOODS
WITH APPLICATION TO ON-LINE PROCESS CONTROL
By
ASHIM KUMAR DATTA
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1985


10
On-line Control of Thermal Processing
In a production situation, the heating time t specified
(assuming a constant value of TR) to achieve a given F0 value would
need to be adjusted for arbitrary variations in the heating medium
temperature TR from its specified constant value. Researchers in the
past have attempted this in one of the following three ways
Precalculated tabulations. Correction factors are tabulated to
be applied to t^ for certain possible types of deviations in
TR(t). Commercial retort control systems often use this technique
(LOG-TEC system, 1984). Such tabulations cannot be exhaustive,
since the process temperature TR(t) could go through any random
deviation. Instead of tabulating such correction factors, Giannoni-
Succar and Hayakawa (1982) developed expressions for correction
factors to be applied to th when TR(t) goes through a step drop. In
their work, other types of deviations had to be approximated to
close-fitting step functions, and multiple drops could not be
considered. Getchell (1980) described a control system that would
attempt to maintain the design heating medium temperature and sound
an alarm when critical low or high limits for the temperature are
exceeded. The operator would then attempt to make necessary changes
in th.
Direct Measurement. Temperature at the can center T(t) is
measured on-line. This eliminates the need to assume constant
heating temperature. Measured T(t) is used with equation 1 to stop
the heating cycle at time t^ such that
W > Fo
(12)


Mechanical and Industrial Engineering Department of the University
of Illinois at Urbana-Champaign. Help by Subhasis Laha in physically
transporting the plotting software to Gainesville is also
appreciated. Availability of a hard copy plotter was the next
necessity and the author appreciates the equipment provided by Dr.
James W. Jones in the Agricultural Engineering Department.
Finally, the author deeply acknowledges the assistance from his
wife Anasua in typing this manuscript, and her encouragement and
perseverence during the study.


LIST OF SYMBOLS
A = constant
a,b,c = coefficients in tri-diagonal matrices
f = inverse slope of time-temperature graph
g = gravity
F0 = sterilization F0 value
H 3
Gr = Grashof number = ge (Ti T0)
h = outside heat transfer coefficient
H = height of cylinder
j = j value (a constant)
Jg = Bessel function of order zero
k = thermal conductivity
n = normal direction
nz = number of grid points in z direction
nr = number of grid points in r direction
p = pressure
Pr = Prandtl number = yCp/k = v/a
r = = non dimensonal distance in radial direction
r = distance in radial direction
R = radius of cylinder
vii


86
resulting in product movement and much faster cooling rates. Details
of the work that showed direct evidence of internal product movement
during sudden depressurization of conduction-heated foods are
presented elsewhere (Teixeira et al., 1985).
The transient temperature response recorded at the can center
during conduction heating was compared with the temperature at the
can center predicted by the ADE analog to the heat conduction
equation (equation 32). The same was also compared with the
analytical solution to the equation 32 (heat conduction equation).
To study the effect of surface heat transfer coefficients, the
transient temperature data were plotted on semilog paper with
( T TR) /( T0- TR) on the log scale and time on the linear scale.
From equation 9, the inverse slope of this graph is the f-value. The
different inverse slopes (f-values) thus calculated, for heating
with steam and heating and cooling with water, were compared. A
higher surface heat transfer coefficient was expected to result in
lower f-values (steeper slopes) and vice versa.
Use of Conduction Heating Model for On-line Control
The finite difference model developed for predicting temper
ature T(t) was incorporated in the software for the on-line control.
Using the temperature at the can center T(t), F0(t) was monitored in
real-time with measured can boundary (retort) temperature TR(t). The
process would continue until time tR such that
Fo Fod
(7)


128
Water
T,=30*C
top-12i#C
s!de=121*C
bot=121*C
R-4.19cm
H=10.07cm
6l-0.0001
6r=0.00 Of
Contours
-45
-35
-30
-22
25
32
40
45
200
200
Streamlines after 90 seconds of heating
computed using 39x39 grid (left half of the
figure) compared with computations using 58x53
grid (right half of the figure)
Figure 25.


4
Such a natural convection heat transfer model could also have
several applications outside the food area. It may be used to study
the heating of buildings (without forced circulation), cooling of
gas turbine blades, storage of cryogenic fluids (e.g. liquid rocket
propellants), and the startup of chemical reactors. Since natural
convection heat transfer would involve a much higher degree of math
ematical complexity than a conduction heating model (and its appli
cation to on-line control), the primary objective of this study was
directed toward developing a heat transfer model for natural convec
tion; followed by improvements to existing conduction-heat transfer
models for application to on-line process control as secondary
objectives.
Objectives
The objectives of this study were to
1. develop a mathematical model to predict temperature and
velocity profiles caused by natural convection heat
transfer in a closed cylinder,
2. investigate the need for possible improvements on an
existing conduction heat transfer model to improve its
suitability for use in on-line process control appli
cations, and
3. incorporate the conduction heating model in an on-line
process control algorithm for thermal processing of
canned food, and compare its performance against other
possible methods of on-line control.


152
7. A slowest heating region was located away from the centerline
in a donut-shaped region near the bottom 15% of the height of
the container studied.
8. The slowest heating points also migrated within this region
but with no particular pattern of migration.
9. The transient temperature values at the slowest heating points
were close to, but different from the transient temperature at
l/3rd height on the axis of the container.
Conduction Heating Model and On-line Control
1. A real-time retort control algorithm was developed. The
algorithm is capable of continious calculations of
accomplished sterilization, on-line correction of arbitrary
deviations in retort temperature and printed documentation.
2. A numerical ADE conduction heat transfer model was adapted for
use with the retort control algorithm. The large Biot number
associated with low thermal diffusivity of food permitted the
use of temperature specified boundary condition at the can
wall.
3. A numerical ADE conduction heat transfer model was also
written for use with finite heat transfer coefficient at the
can wal1.
4. For the numerical heat transfer models to predict temperatures
accurately, the pressure in the retort should be carefully
controlled and sudden drops during cooling avoided.


INTRODUCTION
Background and Justification
Thermal preservation of food involves application of heat
sufficient to destroy the microorganisms present in food that cause
spoilage. Commercial canning processes provide such heat treatment
to extend the shelf-life of foods. In a typical canning process, the
container is filled with food and sealed. The sealed container is
heated in a closed vessel with steam or hot water long enough to
kill the microorganisms and is then cooled. In a production situa
tion, the heating medium (steam or hot water) temperature can
sometimes deviate significantly from design values during the
process. Heating the product more than the required amount would un
necessarily degrade its quality while wasting energy. However,
insufficient heating could lead to serious public health hazards and
cannot be allowed. Thus, food processors always have to meet this
minimum heat treatment (sterilization). Determination of the heating
time to meet the minimum sterilization demands that the actual
transient temperature history of the food during processing be known
predicted with a heat transfer model or actually measured).
1


126
I
Figure 23. Diagram of 39(0.9)x39(0.8) grid on the left
and 58(0.85)x58(0.76) on the right


Figure 3. The grid system and the boundary conditions
for the cylindrical can


34
from the finite difference approximation to the heat conduction
equation (32) with measured boundary (retort) temperature T^(t). The
heating process could continue until time t^ such that
Fo Fo <12>
thus satisfying the required sterilization for arbitrary varia
tions in retort temperature Tr(t).
However, the contribution to F0 from cooling (the second
integral in equation 35a) should not be neglected for a conduction
heating product. Depending on the can center temperature at the
start of cooling, size of the can, and other factors in a conduction
heating situation, the contribution to FQ from cooling could be as
much as 40% of the total. This is unlike convection heated products,
where the container is sometimes agitated and the product is
rapidly cooled, making a small contribution to F0 during cooling. To
avoid gross overprocessing of conduction heated product, therefore,
the cooling F0 must be considered. Also, as mentioned earlier, there
is no control over the possible contribution to FQ from cooling once
cooling is under way.
Thus, the contribution from cooling cannot be neglected, it is
not a constant, and it cannot be controlled. Therefore, it can only
be estimated before cooling is actually started. Such estimation of
cooling lethality can be done through simulation of the cooling
cycle assuming a constant cooling water temperature in the retort.
The objective here is to develop an algorithm to perform this simu
lation in real-time when other control actions are going on.


Convection terms 45
Diffusion terms 50
Discretization of the Elliptic
Stream Function Equation 61
Computational Boundary Conditions for the FDEs 69
Convergence Criteria for Iteration 74
Algorithm for Iterative Solution 76
Coding of the Computer Program and Hardware...... 80
Modeling of Conduction Heating in Cans 80
Discretization and Solution of the Equation 81
Coding of the Computer Program 82
Experimental Studies 83
Use of the Conduction Heating Model for On-line Control 86
RESULTS AND DISCUSSION 93
Natural Convection Heating in Canned Foods 93
Transient Flow Patterns and Temperature Profiles 94
Start of flow and conduction layer ....95
Radial temperature and velocity profiles 106
Axial temperature profiles ...113
Slowest Heating Zone 115
Contrast with Conduction 120
Assessment of the Numerical Method 124
Convergence 124
Selection of the grid size 125
Selection of the time step 130
Convergence of boundary vorticity 134
A note on the half grid points 134
Conduction Heating in Cans 138
Analysis of Boundary Conditions 139
Comparison of the Transient Temperature Values 141
Effect of Sudden Pressure Drops in the Can. 141
Performance of the On-line Control Logic 141
CONCLUSIONS AND RECOMMENDATIONS 151
Conclusions 151
Recommendations 153
APPENDIX
A ALTERNATIVE FINITE DIFFERENCING OF
STREAM FUNCTION EQUATION 154
B ALTERNATIVE FINITE DIFFERENCING OF BOUNDARY VORTICITY...155
C INPUT DATA FILE FOR NATURAL CONVECTION MODEL 156
D ADE FORMULATION FOR THE CONDUCTION EQUATION 157
LIST OF REFERENCES 161
ADDITIONAL REFERENCES... 167
BIOGRAPHICAL SKETCH 172
vi


118
elapsed time is 30 seconds. For clarity, two or more cold points at
the same location are drawn slightly shifted. As can be seen, the
movement is quite random and no particular order is apparent; but
remains well within the lowest 15% of the can height.
Transient temperature at the slowest heating points. The tran
sient values of the temperature at the slowest heating points were
then investigated. An average of the temperatures at three slowest
heating points (nearest to the bottom) was considered for this
purpose. This average temperature, therefore, did not correspond to
any one particular physical location. The average slowest heating
temperature value was plotted with time (figure 19). For comparison,
transient temperature at the mid height and l/3rd (from the bottom)
height on the axis is also plotted. The l/3rd height on the axis has
traditionally been taken to be the slowest heating point.
Qualitatively, the transient temperature curves were as
expected. The slowest heating temperature, by definition, stayed
consistently lower than the other two temperatures. Also, the
slowest heating temperature was rising consistently since heat was
constantly entering the can. The transient temperature at the l/3rd
height stayed closer than the mid point to the slowest heating tem
perature. Although, there are definite differences between the tem
peratures at all times, the magnitude of these differences is
probably not sufficiently great to warrant any change from the tra
ditional practice of using the l/3rd height location. However water,
due to its low viscosity, heats rather quickly and this explains the
little variation of temperature within the container. For a liquid