You are on page 1of 165

Modeling Dust Formation in Lime Kilns

by

Malahat Fardadi

A thesis submitted in conformity with the requirements


for the degree of Doctor of Philosophy
Department of Mechanical and Industrial Engineering
University of Toronto

Copyright by Malahat Fardadi 2010

Modeling Dust Formation in Lime Kilns


Malahat Fardadi
Doctor of Philosophy
Department of Mechanical and Industrial Engineering
University of Toronto
2010

Abstract
Dusting is one of the major problems in the operation of lime kilns because dust particles
interfere with kiln operation and reduce its efficiency. A numerical model is developed to
predict the rate of dust formation in rotary lime kilns. The model consists of four major
components: 1) a 3D model for the kiln gas, solving fluid flow, heat transfer, and combustion in
the gas region; 2) a 1D model for the kiln bed, solving for variation of the solids composition,
including moisture content, along the kiln; 3) a 3D model to predict the motion of the solids in
the bed, and to estimate the reaction rates; 4) a mathematical model to predict the rate of particle
pickup from the bed. Additionally, motion of dust particles was modeled, for the first time,
using Stochastic Separated Flow model (a Lagrangian approach). The developed model of
particle tracking enables the user to predict distribution of dust particles in the gas section of the
kiln. Different components of the model were validated using experimental data published in the
literature.
The developed model was used to simulate operation of a full-scale lime kiln at typical operating
conditions, i.e. at different fuel and air flow-rates. Dusting signatures were also estimated for
each setting to determine the effect each operating condition has on dusting.

The results

presented in this thesis indicate that dust formation is mainly affected by the kiln gas velocity.
Effect of other operating conditions was found to be negligible within the ranges studied. The
results presented here suggest that dust formation can be controlled by minimizing the input gas
flow rate.

ii

Acknowledgments
I was privileged to work under supervision of Professor Honghi Tran, Professor Pierre Sullivan,
and Professor David Kuhn. Their invaluable advice, enthusiasm and support throughout my
graduate studies provided me with motivation to stay on track and to fulfill my potential. Thank
you for letting me do things my way.
I extend thanks to my advisory committee members, Professor Naser Ashgriz and Professor
Markus Bussmann for their invaluable and helpful suggestions.
I would also like to extend my gratitude to Mrs. Brenda Fung, Mrs. Jho Nazal, Mr. Joe Babtista
and many other administrative staff for service during my PhD.
I would like to thank my great, valuable friends in Mechanical and Industrial Engineering
department for making this journey enjoyable. My special thanks go to Professor Hamid Salimi,
Professor Fardad Azarmi, Dr. Hanif Montazeri, Dr. Babak Emami, Dr. Morteza Eslamian, Pooria
Ferdowsi and many more friends in the department.
I would like to thank my beloved parents, Ghamar and Fereidoun, for whom I have a great deal
of respect. You taught me to work hard and always stand up for my values. Your calm voices
were like opium on all my pains, you were not only my true friends but also precious advisors.
My special thanks go to my great, lovely little sister, Mahshid for all her support, love and
encouragement throughout the difficult times I faced. You were always cheering me up when I
needed a boost. You have been always an invaluable friend and consultant to me.
My sincere appreciation goes to my true love, Amir, because of his unbelievable support during
all my research. His endless support during the past six years has allowed me to chase my
dreams. Without you, your courage, enthusiastic support and unconditional kindness, I could
never complete this work.

iii

Table of Contents:

Acknowledgments ........................................................................................................................ iii


Table of Contents:........................................................................................................................ iv
List of Figures: ............................................................................................................................ vii
List of Tables:.............................................................................................................................. xii
Nomenclature: ............................................................................................................................ xiii
Chapter 1

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

1.1 Background ..........................................................................................................................1


1.2 Objective and Scope ............................................................................................................4
Chapter 2

LITERATURE REVIEW .....................................................................................7

2.1 Literature Review.................................................................................................................7


2.1.1 Experimental Work........................................................................................................7
2.1.2 Numerical Work...........................................................................................................10
2.1.3 Summary ......................................................................................................................14
Chapter 3

GAS FLOW MODELING: THEORY..............................................................15

3.1 Kiln Specification and Operating Conditions ....................................................................16


3.2 Turbulent Fluid Flow .........................................................................................................17
3.3 Combustion and Flame ......................................................................................................18
3.3.1 Rate of Combustion Reaction ......................................................................................20
3.3.2 Flame Length and Craya-Curtet Number ....................................................................22
iv

3.4 Heat Transfer Mechanisms ................................................................................................24


3.4.1 Modeling Conduction ..................................................................................................25
3.4.2 Modeling Convection...................................................................................................27
3.4.3 Modeling Radiative Heat Transfer...............................................................................31
3.4.4 Evaluation of Effective Thermal Resistance................................................................34
Chapter 4

MODELING SOLIDS BED ................................................................................40

4.1 One-Dimensional Model of the Lime Kiln Bed ................................................................43


4.1.1 Simplifying Assumptions ............................................................................................44
4.1.2 One Dimensional Equations of the Bed.......................................................................45
4.1.3 Solution Algorithm ......................................................................................................50
4.2 Three Dimensional Modeling of Solids Flow and Calcination in Bed..............................50
4.2.1 Simplifying Assumptions ............................................................................................52
4.2.2 Governing Equations and Solution Procedure.............................................................54
4.2.3 Boundary Conditions ...................................................................................................56
Chapter 5

MODELING DUSTING......................................................................................58

5.1 Dusting Mechanisms and Pickup Velocity ........................................................................59


5.2 SSF Model .........................................................................................................................67
5.3 Solving Turbulent Fluid Flow............................................................................................68
Chapter 6

RESULTS and DISCUSSION ............................................................................71

6.1 Numerical Model Validation Results.................................................................................71


6.1.1 Non-Premixed Open Flame: Validation of combustion and Radiation .......................72
6.1.2 Natural Gas Fired Furnace: Validating Combustion and Radiation ............................75
6.1.3 Pilot Kiln, Gas Phase ...................................................................................................82
v

6.1.4 Pilot Kiln, Solids Phase ...............................................................................................89


6.2 Full Scale Lime Kiln..........................................................................................................90
6.2.1 Boundary Conditions and Operating Conditions.........................................................91
6.2.2 Gas Phase .....................................................................................................................93
6.2.3 Solids Phase .................................................................................................................98
6.3 Dust Generation Rate.......................................................................................................106
6.3.1 Procedure to Calculate Rate of Dusting.....................................................................106
6.3.2 Rate of Dusting ..........................................................................................................107
6.3.3 Effect of Size Distribution .........................................................................................118
Chapter 7

CONCLUDING

REMARKS,

CONTRIBUTION,

and

PRACTICAL

IMPLICATIONS .......................................................................................................................120
7.1 Conclusion .......................................................................................................................120
7.2 Practical Implications.......................................................................................................121
7.3 Concluding Remarks........................................................................................................122
7.4 Future Work .....................................................................................................................123
REFERENCES...........................................................................................................................124
Appendix A: Benchmarking the heat transfer model ............................................................130
Appendix B: Results of 1D Model ............................................................................................135
Appendix C: Material Properties .............................................................................................147

vi

List of Figures:
Figure 1.1.

Lime reburning process stages [3]

Figure 1.2.

Primary and Secondary Air coming into the kiln [3]

Figure 1.3.

Schematic diagram of lime kiln [3] and the small angle it makes with the
horizontal (dash line).

Figure 3.1.

Cross section of two co-flowing jets.

23

Figure 3.2.

A cross section of a typical kiln, showing heat exchanges between


components

Figure 3.3.

24

Schematics of two layers of thermal resistance between kiln gas and ambient
air

Figure 3.4.

27

Schematics of the benchmark geometry to study internal flow in a pipe.


Dash-lines represent boundaries of the 2D-axisymmetric domain.

Figure 3.5.

Radiative heat transfer [26, 46]. Radiation intensity, I, changes due to


absorption, scattering, and emission by the gas.

Figure 3.6.

29

32

Transverse cross-section of a kiln: showing layers of refractory bricks and


steel. In the picture, ri, rm, ro, Ti, Tm, To, and Ta are defined. Picture is not to
scale.

Figure 3.7.

35

Electrical resistance representation of thermal resistances between different


shell components in a kiln

38

Figure 4.1.

Schematic of the bed inside the kiln

40

Figure 4.2.

Left: Picture of the interior of the kiln taken from the burner window (under
the burner). Right: Schematic of the bed motion in a transverse crosssection of the kiln, showing the angle of repose. Kiln rotation is clockwise.

Direction of gravity is shown by g . The angle of repose is estimated at

r 35 [68].

41
vii

Figure 4.3.

Lime mud. This picture was taken at Espanola Mill in 2008

42

Figure 4.4.

Reburned lime. This picture was taken at Espanola Mill in 2008

42

Figure 4.5.

Axial cross-section of a kiln. One-dimensional discretization diagram of the


bed.

Figure 4.6.

45

Non-Newtonian viscosity relationship used to model the bed

shows

viscosity (Pa.s) versus shear strain rate (1/s). Adopted from [47]. Mounir
[47] has measured viscosity of Hamada sands in several experiments and
calculated a general viscosity-shear relationship.

53

Figure 4.7.

Boundary conditions on the gas phase and the solid lime bed

56

Figure 5.1.

Schematics of the forces applied on a particle sitting on the bed surface [22]

59

Figure 5.2.

Drag Coefficient for a sphere as a function of particle Reynolds Number.


Adopted from [30]

Figure 5.3.

60

SEM photo of CaCO3 particles. Average surface roughness was found to be


about 1 m .

Figure 5.4.

61

Schematics of the velocity profile in a boundary layer region. Particles that


are larger than the boundary layer thickness behave differently than those
smaller than the boundary layer thickness. Particles in the lime kiln bed are
typically much smaller than the boundary layer thickness and follow the
small particle regime [16].

Figure 5.5.

62

Pickup velocity obtained from a force balance correlation (the presented


approach) and its comparison with semi-empirical correlation obtained by
Cabrejos [17] for small non-spherical particles with density of 1300 kg/m3.

Figure 5.6.

Schematics of lifters at a transverse cross-section of a lime kiln and


illustration of saltation mechanism

Figure 5.7.

65

66

Dispersion of copper, glass beads, corn pollen, and hollow glass particles as
a function of residence time. Presented numerical data is in good agreement
with experimental data of SL [62], with errors of less than 10%.
viii

70

Figure 6.1.

Temperature and velocity contours (a)temperature profile (K), (b)velocity


profile (m/s).

Figure 6.2.

73

Comparison between results of the presented numerical work and


experimental measurements reported by Mahmud et. al. [44]. Graphs show
radial profiles of the mean gas temperature at different cross sections. (-)
current numerical results, (o) experimental measurements of Mahmud et. al.
[44].

74

Figure 6.3.

Temperature contour along gas chamber geometry

77

Figure 6.4.

Temperature profile along axial distance from inlet.

(o) experimental

measurements, (--) numerical results of Kontogeorgos et. al. [37], and


present numerical work
Figure 6.5.

77

Radial temperature profiles at distances 50, 150, 250, and 400 mm from
entrance.

(o) experimental measurements, (--) numerical results of

Kontogeorgos et. al. [37], (-) Present Work.


Figure 6.6.

78

Mass Fraction of H2O at radial distances distances 500, 600, 700, and 800
mm from entrance. (--) numerical results of Kontogeorgos et. al. [37], (-)
Present Work

Figure 6.7.

79

Mass Fraction of CO2 at radial distances distances 500, 600, 700, and 800
mm from entrance. (--) numerical results of Kontogeorgos et. al. [37], (-)
Present Work

Figure 6.8.

80

Grid dependence of axial velocity profile of natural gas fired furnace for
coarse, intermediate and fine meshes.

Figure 6.9.

81

Grid dependence of temperature profile of natural gas fired furnace for


coarse, intermediate and fine meshes.

81

Figure 6.10.

Geometry of the pilot kiln at UBC was created and meshed in GAMBIT.

83

Figure 6.11.

Grids used for numerical modeling. Because of symmetry, only half of the
pilot kiln needs be meshed.

84
ix

Figure 6.12.

Comparison between experimental measurements of Alyaser [3] and


numerical modeling results for cases 1, 2, and 3 in Table 6.2.

88

Figure 6.13.

Temperature versus axial position on pilot scale kiln [5]

90

Figure 6.14.

Geometry of the inside shell of the full scale lime kiln used in the presented
work.

Figure 6.15.

91

Meshed Geometry of the kiln: the outside refractory lining (top grid) and the
inside volume within the refractory shell (bottom grid)

Figure 6.16.

Total (top) and Radiative (bottom) heat fluxes to the inner wall of the kiln
for the case of Ct=0.46. Only 40m of the kiln length is shown.

Figure 6.17.

97

Temperature Contour on inner wall (bottom) and outer wall (top) of the lime
kiln for the case of Ct=0.95. Only 40m of the kiln length is shown

Figure 6.18.

93

98

Average gas and bed temperatures along the kiln, calculated based on the
1D model of the kiln. Mud flow rate: 8.5 kg/s, fuel flow rate: 0.58m3/s,
primary air: 0.64m3/s, excess-air: 10%. This is the first case of the operating
conditions in Table 6.3.

Figure 6.19.

100

Average composition of the solids bed and percentage of mass versus


distance from the hot end, along the length of the kiln. Mud flow rate: 8.5
kg/s, fuel flow rate: 0.58m3/s, primary air: 0.64m3/s, excess-air: 10%. This
is the first case of the operating conditions in Table 6.3.

101

Figure 6.20.

Temperature contour in the solids bed section of the lime bed

103

Figure 6.21.

Contours of axial velocity in the bed and the gas. (the contours were
obtained from different simulations and were combined into this figure for
comparison reasons).

Buoyancy causes the tail of the flame to move

upward.
Figure 6.22.

104

Velocity vectors inside the bed show movement of solid lime particles
inside the bed

105

Figure 6.23.

Minimum gas velocity required to pickup a particle as a function of the


particle diameter. Density of particles was assumed to be 2830 kg/m3.

Figure 6.24.

108

Pickup probability of particles of different sizes as a function of average gas


velocity. Turbulent intensity of the gas was assumed to be 15%, which is a
typical value for gas flows. [26]

Figure 6.25.

110

Size distribution (number fraction) of dry lime particles in bed. Based on


measurements using Particle Size Analyzer (MasterSizer S; Malvern
Instruments, UK)

Figure 6.26.

111

Particle pickup rate (in kg/s) evaluated for several ranges of particle sizes
exposed to an ambient air flow velocity between 0 and 12 m/s. Turbulent
intensity was assumed to be 15%, surface area was assumed to be 40m2.
Smaller particles have a higher pickup probability, but contribute little to the
mass flow rate due to their small mass.

113

Figure 6.27.

Percentage of dusting rate versus average kiln gas velocity.

114

Figure 6.28.

Dust generation rate (percentage of feed rate) plotted as a function of gas


velocity for the 12 cases defined in Table 6.3. The curve represents the line
of best power fit: y 1.28 104 x 6.3 . The coefficient of determination ( R 2 ,
a measure of goodness of fit) for this fit is 0.983.

117

Figure 6.29.

Three size distributions tested for rate of pickup.

118

Figure 6.30.

Dust Load, defined as the total rate of pickup divided by the mud flow rate,
plotted as a function of average kiln gas velocity. Three different size
distributions were assumed for the particles in the kiln bed. The curves
represent the power fits. The coefficients of determination ( R 2 ) are 0.983,
0.966, and 0.976 for the fits to data of size distributions 1, 2, and 3,
respectively.

119

xi

List of Tables:
Table 3.1.

Specifications and Operating Conditions of the modeled kiln

16

Table 3.2.

Reaction constants for methane combustion [26]

21

Table 3.3.

Coefficients of absorption correlation in equation (3.29) for water vapor,


CO2, and CH4

Table 5.1.

34

Pickup velocity for different particle sizes. Non-dimensionalized forces on


a stationary particle. Comparison of the forces applied on a stationary 30micron particle as functions of the gas velocity. At gas velocities above 8
m/s, the net force on the particle becomes positive and the particle is
entrained to the gas

Table 5.2.

64

Snyder and Lumley's particles and their properties. pl p d p2 / (18t ) is the


particle response time, or the time required for the particle to adapt to flow
fluctuations. p is particle density, d is the mean particle diameter and t
is turbulence viscosity. Rep is based on mean relative velocity and particle
diameter.

69

Table 6.1.

Input flow conditions adopted from [37]

76

Table 6.2.

Flow rate, and Craya-Curtet number for experiment conducted at UBC pilot
scale kiln [3]

Table 6.3.

83

Temperature contours for twelve cases covering operating conditions of the


Canadian lime kiln. Craya-Curtet numbers cover a range of 0.39 to 0.95,
and excess-air value was varied between 10% and 30%.

Table 6.4.

95

Effect of kilns operating conditions on dust generation rate (percentage of


mud feed rate). The operating conditions of the 12 cases are defined in
Table 6.3.

116

xii

Nomenclature:
A calcination

pre-activation constant of calcination reaction

A combustion

pre-activation constant of calcination reaction

Ac

beds transverse cross-sectional area

AH

Hamaker constant

A
a
b
C
Ct

pre-activation constant
absorption coefficient
surface roughness
concentration

Cd

drag coefficient

c
D

specific heat
internal diameter of cylinder

Craya-Curtet number

i
Deffective

effective mass diffusivity

dp

mean particle diameter


particle diameter

Ecalcination

activation energy of calcination reaction

Ecombustion

activation energy of calcination reaction

Ea

activation energy

E
G

specific energy of mixture


incident radiation

Gw

incident radiation on wall

acceleration due to gravity

HR
H

humidity ratio
heat of reaction

enthalpy

htotal

I (r , s )

coefficient of convective heat transfer

radiation intensity at position r and direction s

I
keff

unity matrix
effective thermal conductivity

kt

turbulent thermal conductivity

kB

thermal conductivity of refractory brick


xiii

kS

thermal conductivity of steel

k
M
m
m

thermal conductivity
feed rate of carbonate calcium at feed end
flow rate of moisture in bed
mass flow rate

n1

rate exponent for fuel

n2

rate exponent for oxidizer

n
Nu
P
Pr
Q

refractive index
Nusselt number
pressure
Prandtl Number
total flow rate in the duct

qr

radiative heat transfer per unit mass, kJ/kg

heat flux

R
Rth

universal gas constant (8.314 J/moloK)


thermal resistance

RaD

Rayleigh Number

Re P

Reynolds number based on mean relative velocity and particle diameter

Re

r
r

Reynolds number
position
radius or radial distance

Si

rate of creation of species i due to combustion or entrance of CO2 due to


calcination
source term in reaction equation

Sh

s'
s
T
t
U amb

direction
scattering direction vector
path length
local temperature in oK
time
ambient gas velocity

Ud

dynamic velocity

Uk

kinetic velocity

uf

instantaneous turbulent flow velocity

uf

time-averaged turbulent flow velocity

xiv

u 'f

fluctuating component of turbulent flow velocity

axial distance

mass fraction of ith species

Greek

degree of conversion

Thermal expansion coefficient

thickness of boundary layer


turbulent dissipation

turbulent kinetic energy


turbulent viscosity

kinematic viscosity, m2/s


density, kg/m3

bulk water density in the bed

particle density

Stephan Boltzmann Constant (5.67210-8 W/m2/K)

fL

shear stress, N/m2


integral timescale

Pl

particle response time

stress tensor
phase function

'

solid angle

angle of repose

wall emissivity (0< <1)

scattering coefficient

xv

Subscript
a
e
f

ambient
effective
fluid

i
o
s
w

inner layer
outer layer
solid/steel
water

Superscript

i
T

species type in combustion (CH4, CO2, H2O, N2 or O2)


transpose

xvi

Chapter 1
INTRODUCTION

1.1 Background
Rotary kilns are used in different areas of chemical and metallurgical industries to directly heat
inorganic materials. The largest units are used in cement and lime production industries. Since
most kiln processes are energy-intensive, it is valuable to optimize kiln operation. Improving
this process requires better understanding of the interaction between the flame, gas phase and
solid bed in the kiln. However, the hostile environment of the kiln prevents direct observation or
measurement. Therefore, numerical modeling of the process can potentially provide a reliable
and effective alternative.
Lime kilns are used in Kraft pulp mills. A typical lime kiln is a long inclined cylinder that
rotates around its axis and operates at slopes from 1.2 to 2.4 and rotational speeds of 0.5 to 1.5

rpm. The lime kiln is basically a heat exchanger and a reactor. It heats the mud from ambient
temperature to calcination temperature using a burner located at the discharge end, of the kiln.
The kiln rotation enhances mixing to ensure uniform heat transfer to the mud [68].
The lime kiln has four stages: drying, heating, calcining, and cooling. Therefore, each lime kiln
can be divided into four functional zones (Figure 1.1) which represent the stages needed for lime
mud (CaCO3) to be converted into reburned lime (CaO).

Figure 1.1. Lime reburning process stages [3]

Lime mud enters the drying zone of the kiln with typically about 75% solid content, defined here
as 75% lime mud and 25% water moisture. The drying zone in lime kilns contains a chain
section where series of heated chains are attached to the kiln wall. These chains are capable of
drying the moist mud up to 95% solids by mass.
In the second stage, dry mud is heated by the hot kiln gas and mixed with lifters and tumblers
that are attached to the kiln wall ensuring uniform heat transfer from hot flue gas into lime mud.
Calcination, an endothermic reaction, takes place in the third zone. Calcination is the process of
converting lime mud (primarily Calcium Carbonate, CaCO3 , generated in the caustisizing plant)
to reburned lime (Calcium Oxide, CaO that will be recycled back to the caustasizing plant). The
following reaction takes place:

CaCO3 CaO CO2

(1.1)

during which CO2 is generated and enters the kiln gas. The minimum temperature required for
the calcination reaction to occur is around 815oC [68]. However, temperatures above 1150oC
result in overburning of the lime mud (reducing its reactivity) and damaging the refractory walls
[68].
Cooling is the final stage of kiln operation, which occurs when lime mud passes under the burner
and move to the discharge end. Reburned lime exits the kiln through the product coolers (as

shown in Figure 1.2), exchanges heat with the secondary air to improve the kilns thermal
efficiency.

Figure 1.2. Primary and Secondary Air coming into the kiln [3]

Typical lime kiln burners have different stages that are used to mix fuel and primary air to
produce the required flame shape and patterns. Some burners are designed to atomize liquid fuel.
Figure 1.2 shows the flows of primary and secondary air streams and combustion fuel.
Secondary air enters the kiln through the coolers where it is preheated to between 177oC and
204oC. Natural gas is one of the most widely used fuels in the pulp and paper industry. Natural
gas, which is liquid free, is easily dispersed in the air/fuel mixture and contains 84.1% (volume)
methane [68].
Installation of brick lining at the inner walls of the kiln shell (in one or more layers) significantly
reduces heat loss through the shell to the outside air. Usually, the kiln designs are such that the
thickness of refractory layer varies along the length of the kiln, according to the amount of
radiation and heat transfer from the hot gas to the surrounding air. For example, in a Canadian
kiln that is 76.2m long and 2.4m in diameter, the typical thickness of the refractory wall is 0.3m
in the burner zone [68].

Dusting is a phenomenon that occurs during kiln operation when dried lime particles are picked
up, entrained and dispersed in the turbulent kiln gas. Figure 1.3 schematically illustrates how
dusting occurs during kiln operation. As the lime mud moves along the kiln from the feed-end
towards the discharge-end, it is dried by the heat produced by the flame. Lime particles on the
bed surface are exposed to the kiln gas. The kiln gas picks up small bed particles from the bed
surface and entrains them into the turbulent kiln gas to form a dust suspension [3].

Figure 1.3. Schematic diagram of lime kiln [3] and the small angle it makes with the horizontal (dash
line).

1.2 Objective and Scope


Dusting is an important phenomenon because dust particles typically constitute 5% to 25% of the
dry mud feed rate [3]. Most entrained dust particles are captured by the precipitators and
scrubbers (located outside the kiln) returned to the kiln; however, dusting reduces kiln capacity
because of the load of the recycled material.

A high dust load increases the burden on

precipitators, scrubbers and pre-coat filters [34, 68]. For kilns equipped with scrubbers, dust is
recycled via pre-coat filters by adding water to the filtered dust material and feeding them back
to the kiln. Dust also leaks from the kiln shields, firing hood and gas ducts, causing a dusty
environment around the kiln. It should be noted that in the calcining zone, where lime forms and
nodulizes, lime nodules agglomerate and grow larger as they move towards the front end of the
kiln. As a result, dust entrainment in this zone is less likely.

A small fraction of dust that originates in the product coolers, obscure the burner flame, causing
pyrometers to indicate lower than actual temperature [68]. This may lead to unstable kiln
operation, particularly when the front end temperature is used as a control parameter.
To reduce dusting or control its adverse effects on kiln operation, it is necessary to understand its
mechanisms. Also, it is necessary to correlate the extent of dusting inside the kiln to the kilns
operating parameters, for optimizing kiln operation.

Such correlations, however, require

knowledge of process parameters inside the kiln, which are mostly not available due to
inaccessibility and hostile conditions inside the kiln. For this reason, dusting mechanisms and
the effect of dusting on kiln operation are not yet fully understood.
A remedy to this problem is the use of numerical modeling as an alternative tool to predict the
process parameters inside the kiln. Modeling dusting (pickup and transport of dust particles in
the kiln) is not possible without a reliable model of fluid flow and heat transfer in the kiln.
Dusting strongly depends on interaction between the gas flow and the mud fine particles. Gas
temperature in particular, is highly dependent on flame characteristics and heat transfer
parameters in the kiln and must be carefully considered. To develop a reliable model that
correctly predicts the kiln characteristics, the model needs to be validated against benchmark
cases and verified against experimentally measured data [26, 46, 68].
The main objective of this study is to numerically model the kiln, to estimate the rate of dusting,
and to investigate the effects of different operating condition on the dust load. In order to
achieve the outlined goals, a numerical model of a full-scale lime-kiln needs to be developed.
This thesis focuses on four major parts required for modeling dust formation:
1) Modeling flow, heat transfer, and combustion in full-scale lime kilns
2) Modeling lime bed including heat transfer and calcination
3) Modeling pickup and dispersion of particles from the bed (dusting).
4) Coupling the three models to fully describe dust formation.

This thesis describes the numerical techniques used to conduct this study, and presents the results
and its practical implications for lime kiln.

Chapter 2
LITERATURE REVIEW

2.1 Literature Review


In this section, research works concerning the subject of this thesis have been reviewed. Since
the focus of this thesis is on the numerical modeling of the lime kiln, among numerous
experimental work only those employed for model validation are reviewed here, as well as the
most relevant research papers on the numerical modeling and simulation.

2.1.1 Experimental Work


In a general study, Mahmud et. al. [44] experimentally and numerically studied a lifted nonpremixed turbulent free jet flame. They measured the flame temperature and the concentration
of the flame gas species. In the numerical part, they particularly modeled the radiation heat
transfer. Validation of their model with the experimental data revealed that the prediction ability
of the model was better close to the burner region (about 5% error), but not satisfactory in the
downstream region (with errors more than 50% in predicting concentrations in some parts of the
domain). Their experimental data for concentration and gas temperature have been used in this
thesis to validate the combustion and heat transfer model.
The most relevant experimental work to this study were those conducted at the University of
British Columbia (UBC) [5, 70]. The experiments were conducted on a pilot rotary lime kiln,
where the operating parameters and heat transfer rates could be measured. The pilot kiln used at
the UBC experiments was 5.5 m in length with an outside diameter (OD) of 0.61 m. It was made
of a steel shell of 7 mm wall thickness, lined with a 3 mm insulating blanket and a 100 mm of
refractory wall. This kiln did not have any additional internal heat transfer devices (such as
chain section, lifters, tumblers) and was filled with various amount of solid content. It had a

rotation speed of 1.5 rpm and was fired with natural gas. Heat transfer rates from gas phase to
two different bed materials, Ottawa sand and limestone, were measured. The net rate of heat
transfer to the bed, kiln gas, refractory wall and the radial heat flux inside the refractory surface
were also measured at various locations. Primary air was supplied through eight nozzles at the
ambient air temperature. Secondary air also entered the kiln from another eight nozzles at the
ambient air temperature. Ottawa sand and limestone entered the kiln from the feed end at room
temperature. Granular solids exited the kiln after being calcined by the kiln hot gas from the
discharge end and took about 1.5 hours for a complete set of data collection. Temperatures in
the gas, bed material and the kiln wall were measured by 66 thermocouples installed in the
rotating section of the kiln. Two major sets of experiments were conducted in this lab on Ottawa
sand and limestone. Coarse Ottawa sand with dp = 2.5mm was fed at a rate of 62 kg/h into the
kiln. Gas firing was at 1.97 L/s, with 25% excess air. Flow at the entry point was turbulent with
Reynolds number about 8000. A model for the rotary kiln heat transfer was also developed by
the UBC researchers. The model accounts for the interaction of all the transport paths and
processes.

The numerical model was employed to explore the effect of various operating

conditions on the optimum operation of a lime kiln. For instance, they found that high rates of
net heat input to the bed material decreases quickly with distance. Also, they concluded that
except in the endothermic reaction zone, the installation of devices that increase heat transfer to
the bed material from the kiln gas, may not lead to an increase in the bed heat transfer. Gas
temperature were determined at several points 2.5 cm off the bed surface and 10 cm off the
diagonally opposite wall surface [5]. In another set of experiments, limestone enters the kiln (2.3
L/s natural gas at 10 pcs excess air, 42 kg/h limestone). It was shown that the net heat input to
the bed significantly increases due to limestone calcination reaction [5, 70]. In the present
research work, their experimental data of heat flux and rate of heat transfer from the gas to the
bed [47] have been used to validate our heat transfer model from the gas to the bed. This has to
be done in order to find the calcination rate within the bed. The calcination experimental data for
model validation were taken from ref. [70]. They measured the local gas, solids, and wall
temperatures and percentage of calcination for a range of operating conditions. They explored
the influence of limestone type, feed rate, rotational speed, inclination angle, and particle size on
calcination and heat flow in the kiln.

They found that the local calcination is dependent

primarily on the solids temperature and therefore on the heat transfer.

Aloqaily [2] experimentally investigated flame patterns (including flame length and shape),
using cold flow modeling techniques. The flame patterns in lime kilns with separate noncondensable gases (NCG) inlet were investigated as a function of the combined Craya-Curtet
number (Ct) and excess air, which were found to be the dominant parameters affecting the flame
length, shape, and temperature. Aloqailys results show that the flame length increases with an
increase in the Ct number and a decrease in the excess air. Aloqaily proposed an empirical
equation for calculating the flame length. However, the work was based on cold flow modeling
techniques and no combustion was involved. In the current research, the developed numerical
model is used to investigate whether Aloqailys observations for a cold flow is still valid for
reacting flow. Aloqaily [2] showed that increasing Ct results in increasing flame length, due to
reduction in rate of mixing between fuel and secondary combustion air.
In contrast to heat transfer and flame experiments that has been studied experimentally rather
extensively, very few experimental work have been performed on dusting. In the only study on
dusting available to the author, elutriation of particles from the kiln (dusting) was studied at UBC
[70] as a function of several processing conditions such as air flow rate, rotational speed,
percentage of solids fill and wall roughness, experimentally. The experimental setup consisted
of a rotating cylinder and a cyclone dust collecting unit. A variable speed motor provided a
continuous range of rotational speed from 0 to 60 RPM. Alumina particles were used as the kiln
bed material and Ottawa sand was used as the coarse material of sizes 0.6 to 0.85 mm in twocomponent experiments. Air was drawn to the system through a centrifugal blower connected to
the cyclone, so that the system operates in an open loop. The air passes through the cylinder
collecting particles and flows to the cyclone wherein the dust particles of up to 10 microns are
removed. Dust sampling along the cylinder wall was done using a sharp-edge circular probe.
The local concentration of particles was measured in the gas phase. The elutriation rate was
affected by cylinder exit geometry, wall roughness and segregation patterns in the bed. Gas
velocity showed a very strong effect, with elutriation rate varying with velocity to a power of
4.6.

An entrainment mechanism has been formulated based primarily on the collision of

particles in the solids bed. It was concluded that the geometry of the kiln exit dam affects dust
carryover into the cleaning equipment. Also, the rotational speed has an adverse effect on
particle entrainment due to the centrifugal force it imparts to bed particles, particularly at the
lower edge of the bed. In their research, however, they did not numerically model the dusting
9

process. The experimental data of ref. [65] is used to validate the dusting numerical model
developed in the current work.

2.1.2 Numerical Work


1.3.2.1 Modeling Heat Transfer
Using FLUENT, Paul [51] developed a three-dimensional numerical model for the bed section in
a lime kiln to predict the temperature inside the bed and the concentration of the lime species. He
solved the energy and concentration equations in the bed and observed a substantial variation in
the temperature and concentration distribution during the calcinations, which could influence the
lime quality. Paul [51] validated his bed model against the experimental data he had taken from
a batch kiln.
In a similar study of modeling heat transfer within the bed, Boateng and co-workers developed a
mathematical model to estimate the rate of heat transfer from the gas to the bed [12]. They
combined a two-dimensional slice of a beds transverse plane with a one dimensional plug flow
model for the rotary kiln. A quasi-three-dimensional rotary kiln model gives the temperature
distribution within the bed, inside the refractory wall and within the gas, at any axial position of
the kiln. The model was set for two main conditions to simulate the bed: mixed bed, typically a
rolling bed with uniform particle size distribution, and segregated bed, with non-uniform particle
sizes. For the condition of mixed bed, particles were assumed to be spherical and uniform in size
(0.569 mm for fine Ottawa sand and 2.25 mm for coarse Ottawa sand). In the current research,
however, the movement of the bed material is considered as a non-Newtonian fluid flow. Here
in the presented study, the results of our modeling work is compared against the work of Paul
[51] and Boateng et. al. [12].
Kontogeorgos et. al. [37] developed a numerical model for the radiation of turbulent nonpremixed natural gas flames confined in axisymmetric chambers. The model predictions were
compared against the experimental data of Stoomer [63]. Using their model, Kontogeorgos et.

al. [37] calculated the absorption coefficient of the combustion species as a function of

10

temperature. The geometry of the chamber of that study is similar to that of a lime kiln. They
modeled primary and secondary air and fuel in a fashion similar to the combustion process that
occurs in a lime kiln. This helps us compare our model results directly with their results. In
addition, in this work, the experimental data of Stoomer [63] have been used for the validation of
the combustion model. However, since the main source of those experimental data is not
available to the author, the data have been taken from ref. [37]. In Stoomers experiments, fuel
is injected from a central jet while primary air enters from an annulus. Secondary air also enters
from second annulus in the direction of primary air, similar to what that occurs in a lime kiln.
Boateng and Barr [14] developed a mathematical model for particle mixing and segregation in a
traverse plane of a rotary kiln. In this model, the bed motion keeps finer particles at the core or
active layer, while coarse particles at the surface in contact with the gas phase. The model
correlates the particle segregation rate to the kiln diameter, bed depth, and rotation rate. The
model has also been used to predict the size of segregated core and the effect of segregation on
material mixing inside the bed. Their model prediction ability was acceptable when compared
with the experimental data (with about 20% error).
In another study, Barr et. al. [6] developed a unified model for heat transfer at a kiln crosssection. That model is the advanced version of their previous model [5]. Using a finitedifference approximation for one-dimensional transient conduction, the heat transfer was solved
within the kiln including the bed. To calculate the energy exchange between the covered wall
and the bed, the finite-difference model was extended into the contacting bed. Their calculations
showed that the temperature of the bed material and the inside refractory surface were coupled,
even in the calcination zone.
Mujumdar and Ranade [49] proposed a CFD based model to predict the key transport
phenomena in rotary cement kilns. They modeled the bed and the gas regions of the rotary kiln
including the swirling airflow produced by kiln burners, coal combustion, gas-phase combustion
of volatile matter and radiative heat transfer in the gas region. These coupled computational
models were able to satisfactorily predict the available data from the industrial kilns with about
6% error. Their modeling approach in a sense is similar to the current approach, as they
considered various effects in the cement kiln; however, the chemical reactions in the bed and

11

combustion in a cement kiln are fundamentally different from the chemical reactions in lime
kilns.
A three dimensional steady state model has been developed at UBC to predict flow and heat
transfer in rotary lime kiln [27]. However, the model does not include the dusting. The model
includes three sub models for turbulent hot flow, the bed, and rotating wall. In that work, a CFD
solver has been developed based on finite volume method. The model includes turbulent gas
phase, buoyancy effects, combustion of natural gas, and radiation. The bed has been set up as
active layer and plug flow, which has been modeled by a finite element technique [27]. Reaction
inside the bed has been modeled as a function of temperature, mass fraction of CaCO3 and
particle size diameter. Refractory wall has also been modeled by setting different thermal
conductivities for different wall materials. Those sub-models have been combined and coupled
to develop a full scale model of a rotary kiln. The model has been validated against the
experiment data at the UBC [3].

1.3.2.1 Modeling Particle Dispersion


Modeling particles in fluid flow requires modeling both the carrier fluid phase and the dispersed
phase (particles). Although in modeling dispersion, the fluid phase is mostly modeled using an
Eulerian approach, the dispersed phase can be modeled in either an Eulerian or a Lagrangian
frame. In an Eulerian approach, where the particle phase is treated as a second continuum phase,
particle properties are averaged over a cell and solved as continuous parameters within the
computational domain [22]. Since the dispersed phase in the Eulerian approach is a secondary
fluid, its equations are similar to the equations of the carrier fluid and easily implemented within
a numerical model. This approach assumes that the dispersed phase is so dense that the particles
are in equilibrium with other local particles while they propagate in the carrier fluid [22]. A
dense particulate flow is defined as one whose motion is determined by inter-particle
interactions. In contrast, a particulate flow whose motion is determined by the forces from the
carrier fluid (drag and lift) is usually referred to as a dilute flow. Since the particulate flow in
this study is not a dense flow, it is more appropriate to use the Lagrangian models for the
numerical investigation.

The Lagrangian approach tracks individual particles or groups of


12

particles as they move within the carrier fluid and is applicable to both dilute and dense flows.
Particle motion is found either by solving the equations of motion for individual particles (such
as in dynamic or Stochastic Separated Flow models) or by considering the propagation of a
package of identical particles (such as in PDFP models). A dynamic model is a type of
Lagrangian modeling technique that solves the flow around the particles, finds the form and
shear drag forces on them, and tracks their trajectories under the influence of such turbulent drag
forces [50, 57]. This method is computationally expensive and is not feasible for the current
study where large number of particles should be modeled.
An alternative is to use the Stochastic Separated Flow (SSF) approach that solves the carrier
phase with a commonly used turbulence model (e.g. ) as if the dispersed phase did not
exist. The model then solves the particles equation of motion and trajectory by determining the
drag force based on an estimated instantaneous fluid velocity at particle location. Finding
trajectories of a large number of particles produces the statistical information required for
characterizing the particle behavior [33]. This method is capable of incorporating inter-particle
forces and has been used to study elliptical particles [11, 55].
In SSF model, the instantaneous fluid velocity ( u f ) is usually estimated by adding a randomly
sampled fluctuating fluid velocity term ( u ' f ) to the average fluid velocity ( u f ) [33].

The third Lagrangian approach is the PDFP model. The PDFP model, similar to SSF model,
solves the fluid flow without considering the effect of dispersed phase. The PDFP model tracks
a group of identical particles that are represented by a probability density function (PDF), usually
a Gaussian. The center of the Gaussian PDF characterizes the average velocity of the particle
package as it evolves in time, while its spatial variance represents the spread of the particles [58,
73, 59]. This approach eliminates the need to calculate thousands of particle trajectories and,
therefore, is computationally efficient. The models in this approach are all based on the original
work of Taylor [66] on dispersion of a fluid particle (i.e. a fluid packet) in turbulent flows.
Taylor assumed a simple shape for the fluid particles fluctuating-velocity-correlation function
and analytically estimated the spread of the particles as a function of time. Although Taylors
model was derived for fluid particles, the motion of a small dispersed particle is not very
different in nature than that of a fluid particle [58], and therefore a similar approach can be used.
13

Assuming a similar velocity correlation function for a dispersed particle, and by estimating the
required timescales, PDFP model finds the time dependence of parcel dispersion (dispersion of a
particle package) in turbulent flows [7,8,36,41,42].
The above mentioned SSF and PDFP models are capable of incorporating momentum coupling
between fluid and dispersed phases by integrating the rate of change in particle momentum in
each computational cell and by adding it as a source term to the fluid momentum equation [57,
59, 21 ].
Most numerical studies on turbulent dispersion are compared with the experimental work of
Snyder and Lumley [62]. In their experiments single particles were consecutively released from
a point source into a uniform homogeneous grid-generated turbulent flow. About 700 particles
were photographed and their trajectories were tracked for determination of particle velocity and
displacement statistics. The experiments were conducted with various particle sizes and masses
in vertical direction. Their experimental data (i.e. the particle displacement statistics that can be
related to the density function of a parcel) have been used as a benchmark for the early stages of
this thesis.

2.1.3 Summary
Our literature review reveals that several aspects or components of the operation of rotary lime
kilns have been modeled, but there is still a need for a comprehensive model that describes all
aspect of kiln operations including dusting, combustion, heat transfer mechanisms, calcinations
reaction and interaction between solid and liquid phases.

In addition, such fundamental

phenomena as particle pickup mechanisms and dust formation have not been investigated
theoretically and numerically before. Hence, the main objective of this work is twofold: (1)
Fundamental study and modeling of dust formation in lime kilns; (2) providing a comprehensive
and full-scale numerical model capable of modeling the entire lime kiln and exploring the effect
of processing parameters on optimum operation of a lime kiln. For instance, from the practical
point of view, it is important to predict that for given mean particle size what would be the
maximum allowable gas velocity without causing the particle pickup and dust formation in the
kiln.

14

Chapter 3
GAS FLOW MODELING: THEORY

This chapter focuses on the modeling of fluid flow, heat transfer, and combustion in the gas
section of lime kilns.
A rotary kiln is basically a counter-current heat exchanger. Energy enters the kiln in the form of
chemical energy of the fuel (usually natural gas with 98% CH4) that is mixed with the input air
stream and combusts. The heat released from the fuel combustion:
is partly transferred to the solids bed (mud);
some is lost through the refractory walls to the ambient air ; and
the rest exits the kiln with the gas stream in the form of heat.

The energy required to dry and calcine the feed material (mud) is supplied by the flame.
Consequently, rates of drying and calcination reactions are primarily determined by the rate of
heat transfer to the bed. This is particularly important because dusting typically occurs in the
regions where the solids bed is dry, but is not calcined [68].
To be able to model dusting, it is therefore required to model the calcination process and heat
transfer rates between the gas and the bed. Thus, it is required to model the gas phase and the
bed, as well as the interaction between the two. Due to the complexity of the physics involved in
lime kilns, modeling the full extent of the kiln processes requires significant simplifying
assumptions and extensive validations to ensure applicability and accuracy.
Therefore, the presented modeling work in this chapter, aimed to model the kiln gas, was first
applied to a number of small test cases for validation of individual models. The model was
validated against three sets of experimental data available in the literature [3, 37, 44], results of
which are presented in Chapter 6.

15

The numerical modeling work was mainly conducted in FLUENT (commercial CFD software,
finite-volume-based, ANSYS, Canonsburg, PA, USA). This chapter presents modeling details of
the flow of combusting hot gas and its related phenomena, namely turbulence, combustion,
radiation, conduction and convection. Interaction between gas, bed, kiln walls, and details of the
bed model will be discussed in Chapter 4.

3.1 Kiln Specification and Operating Conditions


The kiln that is being modeled is similar to a Canadian pulp and paper kiln with the
specifications outlined in Table 3.1.

Table 3.1. Specifications and Operating Conditions of the modeled kiln

Parameter

Value(s)

Length

108 m

Internal Diameter

3.09 m

External Diameter

3.5 m

Fuel Flow Rate


Primary Air flow rate

0.44 m3/s, 0.61 m3/s (2200 to 3000 SCMH)


0.64 m3/s, 1.07 m3/s, 1.41 m3/s, 1.55 m3/s

Excess Air

10%, 20%, 30 %

Mud flow rate

8.5 kg/s

- Mud solid to the kiln

5.95 kg/s ( 70% )

- Water coming in with mud

2.55 kg/s ( 30% )

Fill Rate

12 %

Feed end temperature

300oC

Density of dry lime (CaO)

3345 kg/m3

Density of dry mud (CaCO3)

2830 kg/m3

Bulk density of mud at inlet

670 kg/m3

Refractory Thermal Conductivity

1 Wm-1K-1

Fuel Type

Natural Gas (98% CH4)


5.521107 J/kg

- Gross Heating Value


- Adiabatic Flame Temperature

2327 K

16

3.2 Turbulent Fluid Flow


Flow inside large cylinders and ducts are widely encountered in practical engineering
applications, such as rotary kilns. Assuming that the kiln diameter is about 2 to 5m and mass
flow rate of air/fuel mixture is 12 kg/s, such classical flow typically yields a Reynolds number
between 105 and 106 . Such flows are, therefore, turbulent.
To model turbulent flow in the gas section of a kiln, the model is used, which is one of the
most widely used models of turbulence and has been proven robust and reasonably accurate to
model industrial flows. This model, compared with simpler one-equation models, provides a
better accuracy, but is more computationally expensive.

Compared to more sophisticated

turbulence models, such as second-moment closure models, the model performs faster, but
yet relatively accurately.

The model provides a good balance of accuracy and

affordability. The choice of the turbulence model is also related to the nature of the flow and the
main region of interest in the domain. For example, for the case of boundary layer separation,
the model does not provide an accurate prediction. However, for the case of turbulent
mixing (the main concern in modeling combustion, as is in this thesis), the model has
proven reliable and robust [3, 13, 49].
In the model, flow is assumed to be isotropic, and molecular viscosity is neglected
compared to turbulent viscosity [39]. In this model, the governing equations are averaged over
time and are expressed in terms of time-averaged quantities. The time-average of turbulent flow
velocity ( u ) is u and its fluctuating component is u ' u u . The time-averaged continuity and
momentum equations are expressed as:

ui 0
xi

P

ui u j
g

x j
xi
x j

(3.1)
u u j 2 u

ij k ui ' u j '
i
x j xi 3 xk

(3.2)

The last term of the time-averaged momentum equation is the Reynolds stress term and needs to
be evaluated to close the equations. In model, this term is approximated by the following
relation:

17

u u j
ui ' u j ' t i
x j xi

2
u
ij t k
xk
3

(3.3)

where t is the turbulence viscosity (= C 2 / ), is turbulent kinetic energy, and is


dissipation rate. In model two separate time-averaged transport equations are solved for
turbulence properties, and . The general steady-state equations of turbulence model
are discussed in detail in standard textbooks on turbulence [71] and are summarized here:
uj

u
1

ij i
( t / k )
x j
x j
x j
x j

uj

u
2 1

C 1 ij i C 2

( t / )
x j
x j
x j
x j

(3.4)

(3.5)

where
C 0.09

C 1 1.44

C 2 1.92
1.0

1.3

(3.6)

Because of the nature of turbulence and its associated eddies, turbulence plays a dominant role in
transporting mass, momentum, and heat.

3.3 Combustion and Flame


Combustion is a complex sequence of exothermic chemical reactions between a fuel and an
oxidant that result in production of heat. The fuel used in this research is natural gas (mostly
methane) and the oxidizer is air. The combustion reaction for methane and air can be written as:
CH 4 2O2 2(3.76 N 2 )
CO2 2 H 2O 2(3.76 N 2 )

(3.7)

Enthalpy change of this reaction is 803kJ per mole of CH 4 [5]. Reaction (3.7) is the net sum of
a series of simpler reactions that take place during combustion [13]. In addition, equation (3.7)
18

neglects the reaction of nitrogen with oxygen and NOx formation.

At high temperatures

(typically higher than 1500 C) nitrogen reacts with oxygen and produces NOx compounds,
which are hazardous and are considered as a main source of environmental pollution. NOx
formation is highly dependent on:

Temperature
Nitrogen availability (in the fuel or oxidizer)
Oxygen availability (Stoichiometric condition of the flame and excess air. Excess air is
the ratio of the air flow rate injected to that necessary for stoichiometric combustion.)
Despite its significant environmental effects, NOx production rate is typically very small and has
a negligible effect on the fluid flow in the combustion system [44]. NOx formation is therefore
neglected in this study.
In this study, the reaction of methane (as fuel) and air (as oxidizer) is modeled with species
transport model in FLUENT.

Mixing and transport of chemical species are solved using

conservation equations accounting for the generation/elimination of species in the reaction,


convection/diffusion of species, and heat sources due to the reaction. The steady-state mass
conservation equation based on species transport model is described as:

i
. uY i . Deffective
Y i S i

(3.8)

where superscript i denotes the species type (any of the following 5 species: CH 4 , CO2 , H 2O ,
i
N 2 , or O2 ). Y i is the mass fraction of the ith species and Deffective
is the effective mass diffusivity

of the ith species. The first term on the right hand side of equation (3.8) represents accumulation
of species i due to diffusion. S i represents the rate of creation of species i due to combustion
reaction or entrance of CO2 from the bed due to calcination reaction in bed (discussed in Chapter
4, section 4.1.2).
Conservation of momentum in a steady-state reacting flow is similar to that in non-reacting flows
and is expressed as [23]:


uu P g

(3.9)

where is the stress tensor:


19

effective (u (u )T ) u
3

The energy equation can be written as [23]:

i
u E P keffectiveT u hi Deffective
Y i S h S A
i

(3.10)

where keffective is the effective thermal conductivity of the gas mixture, equal to sum of molecular
and turbulent thermal conductivities [35].

Specific energy of the mixture is defined as

E h P / u 2 / 2 and h is defined for ideal gas as h Y i hi .


i

The first two terms on the right hand side of equation (3.10) are divergences of energy flux due
to conduction and viscous dissipation, respectively. The third term is energy flux due to the

i
Y i hi is significant in methane combustion and
diffusion of species. The term . Deffective
i 1

should not be neglected.


Sh is the net heat of chemical reactions per unit volume (J/kmol/m3), and is discussed for

methane reaction in section 3.3.1.

S A is the net absorbed radiation per unit volume, and is

discussed in section 3.4.3.


In equations (3.8), (3.9), and (3.10), the first terms that involve time derivative (

), can be set
t

equal to zero, because the presented modeling work assumes steady state conditions.

3.3.1 Rate of Combustion Reaction


When fuel and oxidizer molecules are mixed, combustion reaction starts at a rate that depends on
the concentration of the reactants and products and their temperatures. Power law kinetics,
proposed by Lee et. al. [51], expresses the rate of this chemical reaction:

dC fuel
dt

n2 /( n1 n2 )
1 /( n1 n2 )
Coxidizer
Acombustion exp combustion
C nfuel
RT

20

(3.11)

where C, Acombustion and Ecombustion are concentration (kmol/m3), pre-activation constant (1/s), and
activation energy (J/mol), respectively. n1 and n2 are the rate exponents for fuel and oxidizer in
the reaction. R is the universal gas constant and T is the temperature in K. These constants for
methane combustion were taken from [35, 26] and are tabulated in Table 3.2.
Table 3.2. Reaction constants for methane combustion [26]

Parameter

Value

A0

2.1191011 1/s

Ea

2.027108 J/mol

R
n1

8.314 J/mol/K

n2

2/3

1/3

The rate of reaction in Equation (3.11) also determines the amount of heat generated by the
reaction. The source term in the energy equation (3.10) is Sh , which is expressed in the
following form for CH 4 combustion reaction:
Sh

dCCH 4
dt

H combustion

(3.12)

where H combustion is the heat of reaction produced by oxidizing one kmol of CH 4 and CCH 4 is the
concentration of CH 4 calculated based on Equation (3.11).
Equation (3.11) shows that at low temperatures the rate of reaction is close to zero. This implies
that the combustion reaction practically never starts at low temperatures. For this reason, a spark
is always needed to start a flame. When the reaction starts, heat of reaction increases the
temperature, and the reaction rate goes up exponentially. This means that the reaction becomes
self-sustained as long as there is enough fuel and oxidizer.
Methane is sometimes considered as a fast-burning fuel, meaning that its characteristic time of
reaction is much shorter than the characteristic time of mixing. This means that the rate of

21

combustion reaction is primarily determined by how fast the fuel and oxidizer mix.

The

bottleneck of the combustion reaction is therefore the local availability of fuel and oxidizer.
In kiln burners, which are considered non-premixed burners, fuel and oxidizer enter from
separate inlets and are not mixed prior to entering the burner. In such flames, turbulence slowly
mixes fuel and oxidizer in the reaction zone, where they quickly burn due to high temperatures
[43]. In other words, the reaction timescale is much shorter than mixing timescale, and is
therefore considered instantaneous.
The main challenge of modeling fast-burning mixtures is the correct prediction of mixing
behavior of the fuel and oxidizer. Although this approach improves the computational speed, it
is at the cost of sacrificing the solution accuracy.

3.3.2 Flame Length and Craya-Curtet Number


This section discusses an approximate method of estimating the flame length based on a
dimensionless number, Craya-Curtet number.
For non-premixed combustion, where fuel enters the combustor separately from the oxidizer,
combustion usually occurs within a relatively thin layer [53] where fuel and oxidizers are mixed.
The extent of mixing -and therefore the local flame thickness-, depend greatly on the diffusion of
oxidizer into fuel (or vice versa). For fast-reacting species, the rate of reaction primarily depends
on mixing patterns rather than chemistry of species. Therefore, approximate shape of the flame
in a kiln can be estimated based on patterns of fuel-air mixing [1].
For the sake of estimating the flame-length, we assume that methane is fast-burning. Also, for
simplicity, we assume that the gas flow in a kiln (see Figure 1.2) is analogous to the flow of two
co-flowing jets: a central jet exiting the burner tip (consisting of primary air and fuel), and an
annular jet (representing the secondary air flow). For such simple case of two co-flowing ducted
jets (see Figure 3.1), and with an additional iso-thermal assumption, the mixing pattern is shown
to be characterized by the following non-dimensional parameters [9]:

22

Nozzle Reynolds number (Re)


Ratio between nozzle and duct radii ( r1 / r2 )
Craya-Curtet number (Ct).

Figure 3.1. Cross section of two co-flowing jets.

Ct is a similitude parameter that was developed by Beker [9, 10] using a simplified theory of
mixing for two isothermal incompressible confined jets, shown schematically in Figure 3.1. Ct
is defined using the time averaged equations of continuity and momentum for steady state flow
within the jets, whose radial and tangential velocities are neglected:
Ct

Uk

(3.13)

(U d2 0.5U k2 )

where Uk is the kinetic velocity, and Ud is the dynamic velocity, defined as:
Uk

U
2
d

u1r12 u2 r22 r12

(3.14)

r22

u12 r12 u22 r22 r12


2
2

1
u22
2

(3.15)

Kinetic velocity is the velocity obtained if the primary and secondary streams were uniformly
mixed through the kiln. Dynamic velocity is based on the mixing of primary and secondary
streams that would yield the same momentum contained in the two streams less the static
pressure head of the secondary stream.

23

The expression for the Craya-Curtet number [9,10,31] was developed to characterize the mixing
length of the two isothermal jets. Experimental investigation of mixing jets by Aloqaily [1]
showed that mixing length is proportional to the Craya-Curtet number to the power of 0.6 and jet
densities.

3.4 Heat Transfer Mechanisms


In this section, the basics of numerical modeling of heat transfer mechanisms, namely
conduction, convection, and radiation are discussed.
Energy released from the combustion reaction (in the gas phase) is the driving force for the inbed reactions. The energy released from the combustion zone is transferred to the bed through
mechanisms shown schematically in Figure 3.2:
Directly from the combustion zone to the bed (through radiation)
Indirectly from the combustion zone to the bed: (gas to wall by radiation and convection,
then wall to bed by radiation and conduction).
It has been shown in the literature by several researchers [3, 5, 6, 12-1416, 24, 27-29, 49] that the
dominant form of heat transfer to the bed is through indirect heating (from gas-phase via rotary
kiln refractory walls by radiation and convection).

Figure 3.2. A cross section of a typical kiln, showing heat exchanges between components

24

Figure 3.2 also shows details of heat transfer forms between other components in a typical lime
kiln:
1. Heated gas towards the exposed bed (direct radiation and convection)
2. Heated gas towards the exposed refractory walls (radiation and convection)
3. Heated refractories towards the bed and vice versa (radiation)
4. Heated refractories towards other refractories (radiation)
5. Heated refractories rotate constantly and once they sit underneath the bed, theyll
conduct their heat to the bed (conduction)
6. Inner layer of refractory lining towards the outer [steel] shell (conduction)
7. Outer [steel] shell towards ambient air (convection and radiation)
Implementation of these heat transfer mechanisms are discussed in detail in this section of the
thesis.

3.4.1 Modeling Conduction


Conduction is usually referred to an exchange of energy by direct interaction between molecules
of a substance subject to a temperature gradient [13]. In kilns, conduction occurs within the gas,
within the bed, and within the refractory bricks.
The generated heat in the combustion zone (flame) is conducted and convected away to
other gas molecules in the kiln.
The heat absorbed by the exposed bed surface (through radiation and convection) is
conducted towards the middle of the bed
The heat absorbed by the inner walls of the refractory lining is conducted to the outer
shell

25

Typically, conduction in solids or static fluids follows the standard Fourier Law of conduction
[20]:

q k T
(3.16)

where q , k , and T are heat flux, thermal conductivity, and temperature gradient, respectively.

In the gas phase, however, the flow is turbulent, and therefore, mixing plays an important role in
the heat transfer rates. To account for the effect of turbulence, to model conduction heat transfer,
thermal conductivity of the air can be replaced by an effective conductivity: k = kmolecular +
kturbulent. Molecular thermal conductivity (kmolecular) is a material property, and turbulent thermal
conductivity (kturbulent), is estimated as c p t / Prt , where Prt is the turbulent Prandtl number,
equal to 0.85.
In fluids, the net rate of heat addition due to conduction per unit volume of the fluid, in general
terms, is written as:
2T 2T 2T
q k 2T k 2 2 2
y
z
x

(3.17)

In solids, conduction heat transfer across a layer of thickness, t, subject to a temperature


difference of T, is often represented by its analogous electrical form:
Qk

T
t

V
R

(3.18)

According to such analogy, thermal resistance across a layer of thickness t with thermal
conductivity of k is defined as:

Rth

thickness
thermal conductivity

(3.19)

In order to model conduction in the kilns refractory bricks and steel shell, the effective thermal
resistance of the two (or more) layers is calculated as the sum of the two thermal resistances. The
conduction in two layers of refractory and steel (as shown in Figure 3.3) can be written as:

Rth

trefractory
krefractory

tsteel
ksteel

(3.20)

26

Figure 3.3. Schematics of two layers of thermal resistance between kiln gas and ambient air

This effective thermal resistance is implemented into the heat transfer equations to account for
the effect of conduction through the kiln walls in section 3.4.4.

3.4.2 Modeling Convection


Convection, in general terms, refers to heat transfer that occurs between a solid phase and an
adjacent moving fluid. Convection can be described as conduction that is enhanced by the fluid
motion. The rate of convective heat transfer is usually expressed as q h(T f Ts ) , which is
called Newtons law of cooling. In this equation, q is the rate of convective heat transfer per unit
area, h is the coefficient of convection.
The rate of convection can be determined by solving the standard fluid flow equations and the
standard equations of conduction. However, adjacent to the solid boundaries (where convection
occurs), fluid boundary layers are formed. Within boundary layers, the steep changes in the flow
properties produce large strain rates and temperature gradients. Since thickness of the boundary
layers is typically much smaller than the characteristic size of the geometry, producing a proper
mesh for the numerical model is usually challenging.
When boundary layers are not very thin with respect to the size of the system, the rate of
convective heat transfer can be numerically estimated by resolving the temperature profile in the
thermal boundary layer (by properly refining the mesh) and finding the temperature gradients at
the interface. Due to relatively steep temperature gradients in the boundary layer, the convection
coefficient estimated from such approach can be very sensitive to mesh sizes. One remedy is to
use the Enhanced Wall Function feature that reduces the sensitivity of the numerical results to
mesh sizes. In this modeling work, both approaches were employed concurrently for maximum
efficiency and accuracy: Enhanced Wall Function with several layers of mesh refinement.
27

While the wall function reduces the grid sensitivity of the results, mesh refinement ensures that
the wall grids are fine enough to resolve the near-wall log layer.
Convection heat transfer mechanisms occur in two different sections in a kiln:
Gas to the inner-walls (similar to the internal flow problem of flow inside a pipe)
Outer wall to the ambient air surrounding the kiln (natural convection cooling of a hot
rod)
The approach taken for modeling each of the two is discussed in details in sections 3.4.2.1 and
3.4.2.2.
3.4.2.1 Internal Flow Heat Transfer and Performance of Enhanced Wall Function

To be able to model the kiln, the inner geometry has to be meshed and appropriate equations
(continuity, flow, and energy) must be solved with proper boundary conditions. To be able to
properly estimate the rate of convective heat transfer to the walls, the boundary layer (adjacent to
the kiln wall), has to be appropriately resolved.
From a numerical standpoint, one can start with a uniform (or approximately-uniform) mesh, and
then refine the boundary meshes such that the solved temperature profile is resolved. This means
that the steep temperature profile in the boundary layer will have at least 4 or 5 nodes, which can
be used to reliably estimate the rate of heat transfer to the boundary.
To validate this methodology and its applicability to the internal flows, a two-dimensional
axisymmetric benchmark case similar to Figure 3.4 was created and solved numerically. This is
a simple benchmark cases for which a semi-empirical correlation is given as [35]:

h Nu

k
0.3 k
0.023Re0.8
D Pr
D
D

(3.21)

This correlation is valid for fully-developed turbulent flows with 0.6 Pr 160 , and
ReD 10, 000 . For rotary lime kiln operations, typical values of Pr and ReD are 0.7 and 106,

respectively.

28

Figure 3.4. Schematics of the benchmark geometry to study internal flow in a pipe. Dash-lines
represent boundaries of the 2D-axisymmetric domain.

The assumed benchmark geometry was 20m long and 1m in diameter (inner). The geometry was
solved in two dimensions axisymmetrically. The inlet boundary condition was defined as massflow-inlet feeding air at temperature of 330K. The fluid was cooled down by means of a set
heat-flux from the walls of the pipe at rate of q = -10 W/m2. This rate of heat flux is low enough
to ensure that temperature variation does not significantly affect the material properties in
correlation (3.21).
For the specified benchmark, the semi-empirical correlation (3.21) leads to a value of

h 3.4 Wm 2 K 1 , was compared well with the results of the numerical modeling.
From the numerical results it was concluded that:
Predictions of enhanced wall-function compare well with the semi-empirical correlation.
Maximum relative error is about 5% in the range studied; meaning that it produces fairly
acceptable, and roughly speaking, it is not mesh dependent.
Enhanced wall function gives us a better accuracy in the case of capturing heat transfer
near the wall.
Using enhanced wall function comes at the cost of slightly higher computation times.
The results indicate that the approach is well suited for the internal flow modeling and will
therefore be used for kiln modeling.

29

3.4.2.2 External Flow Heat Transfer (from wall to outside air):

Another form of convection in the kiln operation is the convection from the outer wall of the kiln
to the ambient air surrounding the kiln. Since this is a form of natural convection, for the sake of
simplicity, we will approximate this form of convection with natural cooling of a hot rod. This
approximation is valid because:
housing of a typical kiln is much larger than the kiln (or there is no housing)
the ambient air surrounding the kiln is stationary
due to the kiln rotation, the temperature of the outer shell is relatively uniform in the
transverse direction. The temperature uniformity of the outer shell is further enhanced
because of the high thermal conductivity of the steel shell.
Convection heat transfer from outer layer of the kiln to the ambient air can therefore be evaluated
using semi-empirical correlations such as [35]:

0.387 RaD1/ 6
k

ha 0.6

8/
27
D
1 0.559 / Pr 9 /16

105 RaD 1012

(3.22)

where D is the external diameter of the cylinder exposed to free convection and
RaD

g (Ts T ) D 3 c
k

(3.23)

where RaD is named Rayleigh number, which is simply the product of Grashof and Prandtl
numbers ( RaD GrD Pr ) and is based on the diameter of the cylinder.
Appendix A examines the applicability of this correlation to modeling external flows. It also
compares the estimates of the convection heat transfer resulted from numerical modeling and
correlation (3.22).
To apply this correlation to a kiln geometry, the following simplifying assumptions should be
realized:

30

Outer shell temperature of the kiln is uniform at any cross-section (fixed axial distance
from the kiln end).
The room in which the rotary kiln is enclosed is much larger (in height and width) than
the kiln diameter. With this assumption the heat transfer rates can be estimated based on
semi-empirical correlations of natural convection.

3.4.3 Modeling Radiative Heat Transfer


Thermal radiation is a form of energy transfer through electromagnetic waves. Electromagnetic
waves are normally emitted from any material. Thermal emissions are not monochromatic and
usually cover a range of frequencies in the electromagnetic spectrum. However, the thermal
emission curve has a peak whose frequency increases (and wavelength becomes shorter) with
increasing temperature.
When radiation is incident on a homogeneous body, part of the radiation energy reflects off the
body and the rest penetrates into the body and is absorbed, or transmits through the body and
emerges from the other side. When a body is a strong absorber and has a zero reflection, it is
called a radiative black body. A black body emitter at temperature T , emits electromagnetic
waves at intensity of T 4 , where is the Stefan-Boltzmann constant (5.672 10 8 W / m 2 / K 4 ).
In reality, objects are never black and are approximated as gray body radiators that emit at a
rate of w T 4 , where w is the emissivity constant ( 0 w 1 ).
For a fluid medium that absorbs, emits, and scatters electromagnetic waves, the radiative heat

transfer equation at position r and in the direction s (as shown in Figure 3.5) can be written as
[35]:

4
dI r , s


2 T
a s I r , s an
s
ds
4

'

'

I r , s s .s d

'

(3.24)

where s ' is the scattering direction vector, s is path length, a is absorption coefficient, n is


refractive index, s is scattering coefficient, I r , s is radiation intensity at position r and

direction s , and T is the local temperature. is the phase function between radiation in

direction s ' and its scattering in direction s . ' is the solid angle in which scattering occurs.
31


s'

Scattering Addition

d '

Gas Emission

Incoming radiation

I (r , s )

Outgoing radiation
dI
I (r , s ) ds
ds

Absorption Loss
ds

Figure 3.5. Radiative heat transfer [26, 46]. Radiation intensity, I, changes due to absorption,
scattering, and emission by the gas.

Several approaches are used to model radiation in Computational Fluid Dynamics. Optical
thickness, defined as absorption coefficient multiplied by geometrical length scale of the domain
( aL ), is a good indicator of which model is appropriate for modeling radiation. For flow in
combustors, L is defined as the diameter of the combustion chamber. In the current study, the
length scale of the kiln is about 3m and absorption coefficient is of the order of 30m-1 (oxygen at
1000K [46]). Therefore, aL is estimated to be of the order of 102, which means that kiln gas
material in kilns has to be treated as an optically thick medium ( aL 1 ). Therefore, the P1
model of radiation was chosen to model radiation in this modeling study [61].
For optically thick media, such as kiln flames, radiative heat flux can be expressed in terms of
radiation intensity as [61]:
qr

3 a s C s

(3.25)

where a is the absorption coefficient, and G is the incident radiation. Isotropic scattering
condition was considered for this study.
The transport equation for G can be expressed as [61, 71]:
32

1

G a G 4 a T 4 0
3a s

(3.26)

Divergence of radiative heat flux can be found by combining equations (3.25) and (3.26). This
divergence can be used as a heat source/sink in the energy equation of the flow, equation (3.10)
to account for absorption/emission of electromagnetic waves:
S A qr aG 4 a T 4

(3.27)

Equation for the heat flux to a solid boundary [26, 46]:


qr , w

22 w

4 T

Gw

(3.28)

where is wall emissivity and G is wall incident radiation, in equation (3.28).


One of the major problems in modeling radiation is the fact that radiation includes a wide
spectrum of wavelengths while absorption properties are very sensitive to wavelength. This
means that a perfect model of radiation would solve for radiation intensities at different
wavelengths and material properties would be known at different wavelengths over a wide
spectral range. This requires a computing power that was not feasible for this work.
Instead, there are many approximate solution strategies that provide acceptable results. One of
them is P1 which is used in this study. In this model, radiative properties of materials are based
on temperature, rather than wavelength.

It should be noted that spectrum of wavelengths

radiated from or absorbed by a material depends on the temperature of the material. Therefore,
assuming that absorption properties -indirectly- depend on temperature is not a bad assumption,
but treats all wavelengths collectively. Though, it still requires extensive experimental data of
radiative properties. Such data were reported in the literature by Modest [46]. Algebraic fits
were produced by Giordano [28], and the correlations were successfully employed by Giordano
[28] for the case of a non-premixed methane-air flame. Equation (3.29) defines the general form
of such experimentally-obtained correlations by Stoomer and Giordano [28, 63] and Table 3.3
tabulates the coefficients:

33

log 10 a i a bT cT 2 dT 3 eT 4 fT 5

(3.29)

where a , b , c , d , e and f are defined in Table 3.3:


Table 3.3. Coefficients of absorption correlation in equation (3.29) for water vapor, CO2, and CH4

aH 2O

aCO2

aCH 4

2.063558

0.33462

0.86724

1.74981103

5.24171103

2.21413 104

5.19882 107

7.59928 106

7.82762 107

6.49135 1011

4.70275 109

2.83529 1010

1.414311012

3.36283 1014

1.65932 1016

These data were introduced into the FLUENT code using a User Defined Function (UDF): At
each numerical cell in the geometry, partial pressures of each species were found and total
absorption coefficient was derived by adding all partial absorptions.

3.4.4 Evaluation of Effective Thermal Resistance


In this section, a novel approach is presented that can be used to mathematically solve for the
rate of heat transfer between the kiln gas (Ti in Figure 3.6) and the ambient air outside the kiln
(Ta in Figure 3.6).
In this approach, the effective thermal resistance of the multi-layer kiln shell and the thermal
contact resistance between the shell and the ambient air is estimated in the form of a correlation.
The estimated thermal resistance is then used in specifying the thermal boundary condition of the
gas region.

34

This approach eliminates the need to mesh and solve the kiln shells and the ambient air.
Therefore, this approach substantially reduces the computation time, and makes it possible to
reduce the mesh sizes for higher accuracy.
Several articles in the literature [13, 48, 49] have reported on incorporating the heat transfer
coefficients into the boundary condition, however, most of them assume that the wall thickness is
negligible compared to the size of the domain. For the case of the kiln, the wall is relatively
thick (compared to kiln diameter) and the geometries are cylindrical.

Therefore, new

correlations have to be developed.


Q

2 ri

2rm 2ro

Figure 3.6. Transverse cross-section of a kiln: showing layers of refractory bricks and steel. In the
picture, ri, rm, ro, Ti, Tm, To, and Ta are defined. Picture is not to scale.

The following assumptions are made:


It is assumed that there is no heat generation or heat loss within the walls of the kiln.
All thermal resistances between the inner wall of the kiln and the ambient air are
combined into a total heat convection coefficient ( htotal ) which satisfies q htotal (Ti Ta ) ,
where q at inner wall is the net heat flux from the kiln gas to the kiln wall. Ti and Ta are
temperatures of the inner wall of the kiln and ambient air. (see Figure 3.6)

35

Rotation in the lime kiln is about 1.5 rpm, and the tangential wall velocity is much faster
than the axial velocity of the bed. This means that the active/passive thermal layer within
the bed can be neglected, as Boateng describes in [12]. Barr et. al. [5] estimated that the
thickness of active layer to be 20 mm, which is much smaller than the shell thickness.
This also means that at any given time and point on the kiln wall, heat transfer rate can be
described as: q htotal (Ti Ta )
It is assumed that temperature gradients within the kiln shell in the axial/tangential
directions are much smaller than the radial components and are therefore neglected.
It is assumed that htotal is independent of angular location on the kiln. This is partly
justified by the assumptions made in 3.4.2.2 and also the results presented in Appendix
A.
When this proposed correlation is used, there is no need to mesh the kiln wall and the
outer region of the kiln.
Given the above assumptions, Conservation of energy equation in the refractory bricks can be
written as: k 2T 0 , given then above assumptions, this equation in cylindrical coordinate is
reduced to:

d dT
(r
)0
or
dr dr
dT
2 rLk B
Const. Q
dr

qr
dT
Q

ii
dr 2 k B Lr k B r

T (r ) Ti

qi ri r
ln
k B ri

(3.30)

(3.31)

In this equation, Q is the total power of heat transfer to the kiln shell, qi is the heat flux at the
inner layer, L is the kiln length, k B is the thermal conductivity of the brick, and ri is the inner
radius of the kiln.
Using equation (3.31), the interface temperature between refractory and steel ( Tm in Figure 3.6)
can be expressed in terms of the heat flux in the brick, q .

36

Tm Ti

qi ri rm
ln
k B ri

(3.32)
Similarly, same equations can be written for heat conduction within the steel layer, in which kS is
the thermal conductivity of steel:
qr
dT
Q

ii
dr 2 kS Lr kS r
To Tm

T (r ) Tm

qi ri r
ln
kS rm

qi ri ro
ln
k S rm

(3.33)

(3.34)

In addition to the heat conduction within the shell, the convective heat transfer from the outer
shell to the ambient air has to be accounted for. This form of heat transfer can be approximated
by natural cooling of heated rods, characterized by equation(3.22). The heat transfer equation
can be written as:

Q ha 2 ro L Ta To
Ta To

qi ri
ha ro

(3.35)

Combining equations (3.32), (3.34), and (3.35) we get


Ta Ti

qi ri rm qi ri ro qi ri
ln
ln
k B ri kS rm ha ro

(3.36)

37

Figure 3.7. Electrical resistance representation of thermal resistances between different shell
components in a kiln

One may reformulate this last equation into the form of Ta Ti qi Rth,e , where Rth ,e is the
equivalent thermal resistance between the inner layer and the ambient air:
Rth ,e

ri rm ri ro ri
ln ln
k B ri kS rm ha ro

(3.37)

The heat transfer equation at the boundary is:


qi k gas

dTair
dr

qr , w

(3.38)

inner wall

This relationship is the basis for the thermal boundary condition of the kiln gas at the inner wall.
Equation (3.37) for Rth ,e can be implemented into equation (3.38) and the numerical model to
form a Robin-type Boundary Condition:

38

Tair inner wall Tambient

k gas

dTair
qr , w
dr inner wall
Rth ,e

(3.39)

This implementation can be done using the UDF (User-Defined-Function) feature of FLUENT,
the commercial CFD software used for this study. qr , w is defined in equation (3.28).

39

Chapter 4
MODELING SOLIDS BED

This chapter describes how the solids bed is modeled in the overall approach to modeling dusting
in the lime kilns.

feed end
hot end

Figure 4.1. Schematic of the bed inside the kiln

Figure 4.1 and Figure 4.2 illustrate how the lime bed sits inside the lime kiln with respect to the
burner/flame. Motion of the solid particles in the bed is mainly determined by the slope and the
rotational speed of the kiln. As schematically shown in Figure 4.2, solid particles from the
bottom of the bed are pushed up by the clockwise rotation of the kiln wall. Once they reach their
peak, they slide down from the top surface at the angle of repose. Angle of repose is the
maximum angle at which a pile of grains retains its slope. Angle of repose for lime solids is
estimated at 35o [49].


g
Lc

Figure 4.2. Left: Picture of the interior of the kiln taken from the burner window (under the burner).
Right: Schematic of the bed motion in a transverse cross-section of the kiln, showing the angle of

repose. Kiln rotation is clockwise. Direction of gravity is shown by g . The angle of repose is
estimated at r 35 [68].

Lime mud sludge (Figure 4.3) enters the kiln from the feed end and goes through a limereburning process and turns into reburned lime (Figure 4.4) and exits the kiln from the hot end.
This endothermic process receives its required heat from the flame. The lime mud that enters the
kiln contains CaCO3 (about 73% mass), inert materials (about 2%), and water (about 25%) [68].
The mass concentration of CaO is assumed to be 0% at the feed end. As the reburning reaction
takes place, CaO concentration gradually increases to 60% - 98% depending on the operating
conditions and efficiency of the kiln [68]. The degree of conversion at any point in the bed is
accordingly defined as

nCaO
where nCaCO3 and nCaO are the molar concentrations of
nCaO nCaCO3

CaCO3 and CaO.

41

Figure 4.3. Lime mud. This picture was taken at Espanola Mill in 2008

Figure 4.4. Reburned lime. This picture was taken at Espanola Mill in 2008

Since the bed material is what creates dusting in the lime kiln operation and since the focus of
this thesis is to model the dusting process, it is necessary to model the bed and its associated
reactions.
To model the bed of a lime kiln, one should be able to characterize the spatial distribution of the
following:

the bed composition:


o moisture content
o calcination percentage
42

bed temperature
To fully characterize the bed in a numerical model, one should account for different physical
phenomena that occur in the bed, including evaporation, calcination reaction, and heat transfer
mechanisms.
To model the bed, two approaches were taken that are described in this chapter. The first
approach was a one-dimensional model that was used to determine the variation of bed
characteristics (averaged at transverse cross-sections) along the length of the kiln. In the 1D
model, effect of evaporation was considered and curves of bed characteristics along the kiln were
obtained.

Such characteristics included moisture content, bed temperature, and degree of

calcination reaction.
The second approach was to conduct a three-dimensional modeling of the kiln bed. The bed was
solved as a reacting fluid in 3D model. However, for technical reasons, physics of evaporation
could not be incorporated into the numerical model. Therefore, the results obtained from the 1D
bed model was combined with the 3D bed model to fully characterize the bed.

4.1 One-Dimensional Model of the Lime Kiln Bed


This section describes the development of a simplified one-dimensional code that is used to
approximately evaluate the extent of calcination reaction along the kiln.
The bed in a lime kiln consists of the material that is fed into the kiln as a wet mud and is then
dried and transformed into reburned lime as it moves from the feed end towards the hot end. The
solids in the bed move along the bed due to the slight angle it makes with the horizontal (Figure
1.3) and also due to the rotation of the kiln and the induced rotation and tumbling of the solids.
The lime mud that enters the feed end of the kiln is usually composed of 20% to 30% water [68].
As the mud moves along the kiln towards the hot end, it exchanges heat with the hot kiln gas.
First, the mud mixture heats up and the moisture evaporates. Then the dried mixture heats up to

43

a temperature about 800oC [68] to 900oC [70], which is sufficient for the calcination reaction to
take place.
The rotational motion of the bed causes the solids to be mixed and exposed to the hot kiln gas.
This mixing effect is sometimes enhanced by installing curved or straight baffles on the inner
wall of the kiln, which are called lifters or tumblers. They make the bed material to be uniformly
mixed and exposed to the hot kiln gas. Still, as reported by [24], some of the finer/denser
particles may segregate and form a kidney section within the bed that is never exposed to the kiln
gas. Besides, the intense radiation from the flame causes the temperature at the surface of the
bed to be higher than that within the bed.
Despite the reported non-uniformities in the bed temperature, we assume that properties
(temperature and composition) of the bed material are uniform at transverse cross-sections of the
kiln. This assumption is used to simplify the modeling approach, as it reduces the problem to
solving a set of one-dimensional equations in the bed. This set of one-dimensional equations in
the bed can be accompanied by the 3D modeling results of the gas phase to form an approximate,
but quick numerical solution for the lime kiln operation. Such a model will enable us to
approximately determine the extent of calcination reaction and bed temperature along the length
of the kiln.

4.1.1 Simplifying Assumptions


In the one-dimensional modeling approach, several simplifying assumptions and approximations
have to be made:

The bed properties (including temperature and composition) are uniform at any
transverse cross-section of the bed

The bed moves at a uniform axial velocity towards the hot end
The calcination reaction occurs at a rate following Arrhenius rate equation
Mass percentage of inert material in the bed is about 2% and is therefore neglected

44

The first assumption is inherent to any 1D model and can be corrected with 3D modeling.

4.1.2 One Dimensional Equations of the Bed


The developed model is a one-dimensional finite-volume representation of the governing
equations for conservation of mass and energy. These equations are discretized in space along
the kiln length (axial slices) according to the discretization scheme illustrated in Figure 4.5.

Figure 4.5. Axial cross-section of a kiln. One-dimensional discretization diagram of the bed.

The ultimate objective is to obtain expressions for:

Bed temperature, T ( x)
Degree of conversion, ( x)
Flow rate of moisture (water) in bed, mw ( x)
along the length of the kiln, in terms of feed rate of carbonate calcium at the feed end ( M ).
The simplified one-dimensional equations of the bed can be described and discretized as follows:

45

Continuity of Mass:
Material that enters the kiln contains calcium atoms (Ca). Despite all the chemical reactions, the
mass-flow-rate of calcium atoms at any transverse cross-section of the kiln remains constant
along the length of the kiln. Neglecting the inert materials in bed, the total mass-flow-rate of bed
material through the kiln is:

m t m CaCO3 m CaO m w

(4.1)

Assuming that the mass-flow-rate of feed is constant in time and equal to:
feed
M m CaCO
Constant = 5.95 kg/s (according to Table 3.1)
3

(4.2)

and considering the molar mass of contributing species according to:


CaO : 56 gr mole

CaCO3 :100 gr mole


Ca : 40 gr mole

(4.3)

the net mass-flow-rate of calcium atoms at any transverse cross-section can be written in terms
of the feed rate of carbonate calcium (M) as:

m Ca m CaO
m Ca

40
40
40 feed
m CaCO3
Constant along length of the kiln=
m CaCO3
56
100
100

40M
100

(4.4)

Degree of Conversion:
The calcium that enters the kiln in the form of CaCO3 goes through calcination reaction and
converts to CaO. The degree of conversion at any point is defined as the molar concentration of
CaO to the total molar concentration. This is equivalent to mole-flow-rate of CaO to total moleflow-rate:
m CaO
100 m CaO 100 m CaO
56

feed
56 m CaCO
56 M
m CaO m CaCO3
3

56
100

(4.5)

46

The degree of conversion, , is a number that varies between 0 and 1. At the feed end of the
kiln, this number is 0, and as the mud moves along the kiln towards the hot end, it gradually
increases to a number close to 1, a number that depends on kiln efficiency. For typical kilns
at the hot end (or discharge end), is between 0.6 and 0.98 [68].
Combining equations (4.4) and (4.5), mass-flow-rate of CaO and CaCO3 can be written in
terms of M and as:
56

M
m CaO
100

m CaCO (1 ) M
3

The total mass-flow-rate through the bed at any point can be expressed in terms of M, , and
flow rate of water as:
m t m w M (1

56
44
) m w (1
)M
100
100

(4.6)

Rate of Calcination Reaction:


General equation of reaction for species i , according to Arrhenius rate equation is:
Ea
1 d i
Ae RT
i
dt

(4.7)

One can relate the density of species i to its mass-flow-rate in the bed according to:
m i i AcV

i : CaO, CaCO3 , Water

Ac : transverse cross-sectional area

(4.8)

Assuming that Ac is constant or has negligible variation along the length of the kiln:
Ea
1 d i dm i
V dm i
RT
Ae

i dt m i dt m i dx

E
d ln m i
A
exp( a )
dx
V
RT

(4.9)

47

For carbonate calcium species this relationship becomes:


i : CaCO3
A
E
d
ln (1 ) M calcination exp( calcination )
dx
V
RT

A
E
d
ln(1 ) calcination exp( calcination )
dx
V
RT

(4.10)

Equation (4.10) can be discretized in space according to the diagram in Figure 4.5 in the
following form:
1 n 1
A
x
E

ln(
) calcination n 1 exp calcination

RT
n
1 n
V

n 1 1 (1 n ) exp Aevaporation n 1 exp calcination


RTn
V

(4.11)

Rate of evaporation:
Rate of evaporation was estimated based on a rate equation similar to Arrhenius rate equation.
This approach has previously been tested for modeling of evaporation [51]. The constants of the
rate of evaporation were taken from [51]. The rate of evaporation can be written as [13]:
E
d w
evaporation
RT
w Aevaporation e
dt

(4.12)

where Aevaporation and Eevaporation are the pre-activation constant and activation energy of
evaporation rate, the values of which are taken from [51]. T is the local temperature and w is
the bulk density of water in the bed. Equation (4.12) can be written as:

Aevaporation
Eevaporation
d ln m w
exp(
)

dx
V
RT

(4.13)

The discretized form of this equation can be written as:

48

E
evaporation

x
RTn
m w,n 1 m w,n exp Aevaporation n 1 e

(4.14)

Conservation of energy:
Energy is transferred to the bed from the kiln gas in the forms of convection and radiation heat
transfers. In an axial element of the bed, as illustrated in Figure 4.5, energy enters the element
from the gas (denoted by Qg 2b , heat transfer rate per unit length of the kiln), energy leaves the
element towards the adjacent kiln wall ( Qb 2 w ).

The algebraic sum of these energy

inflows/outflows is spent on drying the mud, the chemical reaction in the bed, and also
increasing the bed temperature.
Qg 2b Qb 2 w

m w ( x) m w ( x x)
L f , water
x
T ( x x) T ( x)
m t C
x
m CaCO3 ( x) m CaCO3 ( x x)
H calcination
+
x

(4.15)

In this equation, H calcination is the enthalpy of calcination reaction (equal to 1985 kJ per kg of
CaCO3), L f , water is the latent heat of evaporation of water (2257.2 kJ/kg) [51], Q is the net rate of
heat transfer per unit length of the bed. This relation can be further simplified as:
Qg 2b Qb 2 w

dm w
44
d

dT
)M C
M
H calcination
L f , water m w (1
100
dx
dx

dx

(4.16)

Equation (4.16) is solved in parallel with equations (4.10) and (4.13) to solve for
m w ( x) and ( x) .

49

4.1.3 Solution Algorithm


The equations outlined above for modeling a one-dimensional kiln bed were solved using the
following algorithm:
1. Specify boundary conditions (temperature, composition, mud-flow-rate) at inlet of the
kiln bed
2. From the 3D kiln gas model, import the rate of heat transfer to the bed (radiation and
convection) per unit length (along the length of the kiln).
3. Initialize the solution with approximate profiles of temperature and composition
4. Solve rate equations to determine the rate of evaporation and rate of calcination
reaction in the bed.
5. Solve mass conservation equation to determine new composition profiles.
6. Solve energy-conservation equation in the 1D bed of solids to determine the
temperature profile
7. Compare the resulting profiles of temperature and composition with those of the
previous step. If the new solutions are very different (e.g. residuals are greater than
1%), return to step 4.
8. Export information.

4.2 Three Dimensional Modeling of Solids Flow and


Calcination in Bed
To model the bed in 3D, motion of the bed has to be modeled. The correct prediction of the bed
motion is essential, because the total residence time of bed material significantly affects its
degree of reaction. Since the kiln bed consists of small particles, the most intuitive approach to
modeling the bed is to model the particles individually (or consider their representative
50

particles). However, this approach is impractical for the purpose of this thesis that is concerned
with reaction rates and heat transfer.
Another approach was therefore sought in the presented work: the flow of the solids in the bed
(lime mud) was solved as a fluid motion.

However, the flow of solid particles differs

significantly from that of a Newtonian fluid. Therefore, to correctly predict the general flow
profile of the bed, the solids bed was modeled as a non-Newtonian fluid with a viscosity that is
similar to sand. The non-Newtonian sand viscosity is relatively high when the shear strain is
low, and is relatively low when shear strain is high. Such viscosity profile can approximate the
flow of solids in the bed cross section. Although this assumption may not provide the exact flow
profile in the transverse cross-sections, it approximately accounts for rotation of the solid
particles, their velocity along the kiln, and their residence time inside the kiln.
In typical kiln operations, lime mud enters the bed as a composition of calcium carbonate
(CaCO3) and water. With the help of heat generated by the flame, water evaporates and
calcination reaction occurs, converting calcium carbonate to CaO and CO2. The generated CO2
exits the bed from the bed surface.
To develop a realistic model of the kiln bed, one should account for evaporation and reactions.
Therefore, FLUENT software and its Species Transport Model was used to model the bed
reactions. However, it was impractical to account for the complex flow of vaporized water and
the generated CO2 between the solid particles. To the knowledge of the author, the current
models implemented in FLUENT do not allow for diffusion of one of the species into other
species. This is because only one momentum equation is solved and all species move at the
same velocity. Therefore, it was not possible to account for the motion of the evaporated water
and the generated CO2. Another approach was therefore taken:
For the model of calcination reaction, it was assumed that material properties of CaO (the bed
material after calcination) are the same as those of CaCO3 (the bed material before calcination).
For the modeling purposes, it was assumed that no CO2 is generated.
assumption will not affect the flow of bed or its reactions.

51

This simplifying

4.2.1 Simplifying Assumptions


Motion of solid lime/mud particles in the bed depends greatly on the rotational speed and tilt
angle of the kiln as well as the particle parameters. Accurate modeling of such motion is a great
challenge and is beyond the scope of this work. Therefore a simplified approach was taken and
the bed was modeled as a fluid phase. The viscosity of the bed was assumed to be similar to the
non-Newtonian viscosity of sand dunes. The general regime of the bed motion was compared to
the previous work of Paul et.al. [51] and Boateng [13]. The following simplifying assumptions
were also made in modeling the bed:
Time dependence of the bed motion was neglected and the bed motion and reaction was
modeled as a steady phenomenon.
Bulk density of lime/mud was assumed to be locally uniform: i.e. effect of the air trapped
between the mud lumps was not considered.
Viscosity of the fluid bed was assumed to be non-Newtonian, and similar to that of sand.
The viscosity was considered to be a function of strain-rate magnitude, according to
Figure 4.6.

52

Figure 4.6. Non-Newtonian viscosity relationship used to model the bed shows viscosity (Pa.s)
versus shear strain rate (1/s). Adopted from [47]. Mounir [47] has measured viscosity of Hamada
sands in several experiments and calculated a general viscosity-shear relationship.

It was assumed that the surface of the bed is flat and makes 30o angle with horizontal
[68].
Calcination reaction was modeled by writing a custom designed User-Defined-Function
that solves reaction rate equations.
k Acalcination exp(

The rate constant for this reaction is given by

Ecalcination
) , where Ecalcination is the activation energy (ranges from 180
RT

KJ/mol to 220 KJ/mol) and Acalcination is the pre-activation constant (ranges from 106 to
108) [51]. For this study, Ecalcination and Acalcination were assumed to be 200 KJ/mol and 107
respectively. Although the numerical results and reaction rates are very sensitive to these
numbers, only the above numbers were used due to lack of access to reliable
experimental data.

53

Degree of reaction at the inlet was assumed to be zero and it was calculated for each
point in the bed according to its upstream properties and reaction-rate-equation.
Effect of evaporating water in the bed was neglected.
Heat required for the calcination reaction was added as a negative source term to the
energy equation.
The generated CO2 from the chemical reaction in the bed was considered to leave the bed
from the bed surface and enter the gas at zero velocity. Effect of flow of CO2 gas within
the bed was neglected.
Radiation within the bed was neglected because of the high absorption rates in the solids
region. Radiation equations were therefore not solved within the bed.

4.2.2 Governing Equations and Solution Procedure


The flow of bed material was first solved, while the energy and reaction equations were not
being solved. The general steady-state continuity equation of the following form was solved:

. u 0
(4.17)

along with the following steady-state momentum equation:


uu P g
2


with non Newtonian (u (u )T ) u
3

(4.18)

. Turbulence effects were not considered as the

flow is relatively slow (axial velocity of bed is around 0.03 m/s). Once the above flow equations
are solved, the general form of the solids motion in the bed is obtained.
In the second step, the flow equations were disabled (as the general bed motion was already
solved) and the energy and the reaction rate equations were enabled.
For the model of evaporation, it was assumed that no water enters the bed, and no water
evaporates. Only the effect of evaporation was considered. The effect of water evaporation is:

54

To absorb heat to heat up to 100C


To absorb heat to evaporate
In the presented approach, the removed heat was accounted for by adding an external heat sink to
the domain in the regions where water was present. Since in this approach FLUENT does not
solve for the moisture content, such information was read from the 1D bed model (see section
4.1). The following volumetric heat source (in units of W/m3) was added to the FLUENT model
in the form of a UDF:
m c dT L f dm w
S water w w

Ac dx Ac dx

where L f , cw , and m w are the specific heat, latent heat of evaporation and mass-flow-rate of
water, respectively. dT / dx , m w and dm w / dx are read from the 1D kiln bed model (see section
4.1). The m w and dm w / dx terms are non-zero close to the feed-end and vanish when all water is
evaporated.
The energy equation that was solved in the domain was similar to the general energy equation of
(3.10) with zero time derivatives (steady solution):

u E P keff T u Scalcination S water

(4.19)

In this equation, Scalcination is a heat source term accounting for the heat required for calcination
reaction:

Scalcination H calcination Acalcination e

Ecalcination
RT

(4.20)

In this equation, H calcination is the change in enthalpy per unit mass of CaCO3 and is the local
bulk density of the bed. This term removes heat from each numerical cell.

55

4.2.3 Boundary Conditions


As schematically shown in Figure 4.7, heat is transferred to the top surface of the bed through
convection and radiation. Radiation heat flux at any numerical face on the bed surface is
calculated based on the radiation intensity in an adjacent gas cell using equation (3.28). Since
most of the radiation flux by the gas is absorbed by the outer bed layer, the radiation intensity
within the bed is very small and therefore negligible compared to that in the gas region. This
means that the absorbed radiation flux is mostly conducted towards the bottom of the bed and is
used for evaporation of water and calcination of CaCO3.

Figure 4.7. Boundary conditions on the gas phase and the solid lime bed

The assigned thermal boundary condition at the top surface of the bed was the rate of heat
transfer which was extracted from the 3D kiln gas model (described in section 3.4.2.).
At the bottom of the bed, it was assumed that heat is transferred through the kiln walls
(refractory lining + steel) in the form of convective heat transfer. Due to high absorption

56

coefficient of solids, radiation intensity was neglected in the solids zone and only
convection/conduction was considered.
The momentum boundary condition at the bottom of the bed was assumed to be no-slip with
moving wall. The wall movement included rotation about the kiln axis (representing the kiln
rotation). The momentum boundary condition at the top surface of the bed was defined as slipboundary condition (zero shear-stress) to represent the free movement of bed solids adjacent to
the kiln gas.
In the 3D kiln gas model, CO2 was injected into the kiln gas domain from the top surface of the
bed. It was assumed that the following amount of CO2 per unit length of bed surface is injected
into the gas phase:

dm CO2

E
M CO2 Ac
calcination

Acalcination e RT
dA
M CaCO3 Lc

(4.21)

In equation(4.21), M CO2 and M CaCO3 are molecular mass of CO2 and CaCO3 respectively. Ac
and Lc are the beds cross-sectional area and the width of the exposed bed surface (Figure 4.2).

57

Chapter 5
MODELING DUSTING

Dusting in lime kilns refers to the entrainment of small solid particles from the bed into the kiln
gas. As the solids particles in the bed slide and slowly fall through the kiln, a large number of
lime particles are entrained and dispersed in the turbulent kiln gas to form a dust suspension that
amounts from 5 to 20% of the mud feed rate [3, 68]. Dust particles are mostly dried CaCO3
particles that range in size between 10 to 50 microns [3, 12, 68].
A high dust load increases the burden on precipitators and filters that recycle dust to the
production line. This reduces the overall efficiency of the process. Also, a high dust load
reduces radiative heat transfer to the bed, causing additional reduction in kilns efficiency.
Furthermore, a high dust load interferes with the operation of kiln pyrometers (by obscuring their
field of view), creating problems for the kilns control system.
Dust formation is the result of the interaction between the turbulent gas phase and the solid
particles in the bed. Such interaction, including the pickup and entrainment of particles, greatly
depends on the gas velocity in the gas section and the size of the particles. Particle entrainment
occurs

through

several

mechanisms

that

are

discussed

in

this

chapter.

mathematical/computational model is also presented that can predict the behavior of particles on
the bed surface and their interaction with the gas.
Particles that are picked up and entrained into the gas section will follow distinctive trajectories
under the influence of gas velocity, drag resistance, and gravity. This chapter also presents a
methodology to predict these trajectories and dispersion behavior of turbulent gas using the
Stochastic Separated Flow (SSF) model.
The presented model development will be used in Chapter 6 to predict pickup, entrainment, and
transport of dust particles. The developed model, coupled with the information of the kiln gas
and kiln bed, will also be used to predict extent of dusting and local density of entrained
particles. With such information, one can suggest ways of reducing dust formation in lime kilns.
58

5.1 Dusting Mechanisms and Pickup Velocity


A particle on the surface of the kiln bed experiences several forces, including gravitational force,
buoyancy, adhesion force by other particles, and aerodynamic drag/lift force by the turbulent kiln
jet. Direction of the net force determines whether or not it is entrained into the mainstream air
flow as a dust particle: A high-enough lift force can overcome gravity and adhesion forces
applied on a particle, causing lift off and entrainment to the gas.
Lift Force
Buoyancy

Force

d
Friction

Adhesion
Weight

Figure 5.1. Schematics of the forces applied on a particle sitting on the bed surface [22]

Major forces on a stationary spherical particle are shown in Figure 5.1 and are listed below:
Gravity and Buoyancy: Sum of gravitational force and buoyancy force can be written as:
Fg ( p g ) d p3 g / 6

(5.1)

where and d are density and diameter. Subscripts p and g represent particle and gas phase,
respectively.
Aerodynamic drag force [22]:

2 d p2
1
Fdrag Cd g u g u p
(5.2)
2
4

where u g , u p , d p , and Cd are, respectively, gas velocity, particle velocity, particle diameter,
and drag coefficient (Figure 5.2). Direction of the aerodynamic drag force is opposite to the
relative motion of particle with respect to air.
59

Figure 5.2. Drag Coefficient for a sphere as a function of particle Reynolds Number. Adopted from
[30]

Adhesion force: Adhesion is the intermolecular (Van der Waals) force applied on a particle by

other particles sitting immediately below the bed surface. It can be estimated from the following
correlation [22]:
Fadhesion

AH d p

(5.3)

24( z b) 2

where AH is the Hamaker constant, z is the minimum distance between two surfaces (or mean
free path of the separating fluid), and b is the surface roughness of the lime particles. A and z are
estimated to be 5.0 1020 J and 4.0 1010 m [22]. b is the surface roughness of the particle
surface and is estimated at 1 m based on the image processing of SEM photos of the CaCO3
particles (Figure 5.3).

60

Figure 5.3. SEM photo of CaCO3 particles. Average surface roughness was found to be about 1 m .

Saffman lift force: Saffman force is the aerodynamic lift force applied on a particle that is in a

rotational flow field (flow with lateral velocity gradient). This force can be estimated using the
following relationship [22]:

FSaffman 1.61 g d p ReG u g u p

(5.4)

where ReG is the Reynolds number calculated based on the velocity gradient of the flow field in
which the particle is experiencing the lift force:

ReG

g d p2

du g
dy

(5.5)

where g is the dynamic viscosity of gas.


Magnus force: Magnus force is the aerodynamic lift force developed due to the rotation of

particles at angular velocity p :


FMagnus


1
d p3 g ( u g p ) (u g u p )
8
2

(5.6)

61

Figure 5.4. Schematics of the velocity profile in a boundary layer region. Particles that are larger
than the boundary layer thickness behave differently than those smaller than the boundary layer
thickness. Particles in the lime kiln bed are typically much smaller than the boundary layer thickness
and follow the small particle regime [16].

Figure 5.4 shows a schematic picture of a stationary particle sitting on a solid boundary with a
fluid flow that forms a boundary layer. If the particle is smaller than the boundary layer
thickness, it will experience velocities that are much smaller than the ambient gas velocity.
For kilns operating at typical operating conditions of Table 3.1, particles that are on the bed
surface are typically smaller than the boundary layer thickness. Therefore, the gas velocity they
experience is much less than the average kiln gas velocity.
To relate the gas velocity that a stationary particle experiences with the gas velocity of the kiln
gas, it was assumed that the time-averaged gas velocity profile around the particle (sitting in the
th

vicinity of the wall) follows the 1 7 power law

ug ( y)
U amb

1/ 7

y
. Although the power law is not

valid in the immediate vicinity of the wall, it is estimated that particle centers lie outside viscous
sublayer, and therefore, the gas velocity profile experienced by a particle of diameter d p is:
1/ 7

ug ug ( y)

dp

d
U amb p
2

(5.7)

62

Pickup velocity is defined as the ambient gas velocity at which the aerodynamic lift forces on a
particle balance the downward forces of gravity and adhesion. In other words, pickup velocity is
the minimum ambient gas velocity required for a particle to experience a net upward force.
To find the pickup velocity (the minimum ambient gas velocity required to balance a stationary
particle), the vertical force balance relation ( Fg FBuoyancy Fadhesion FSaffman FMagnus ) was setup
and solved in MATLAB using equations (5.1), (5.3), (5.4), and (5.6):
( p g ) d 3p g / 6

1.61 g d p

AH d p
24( z b) 2

g d p2

du g
dy

(5.8)

1
( u g u p ) d 3p g ( u g p ) (u g u p )
8
2

Equation (5.8) was solved with assumption of stationary particle ( u p 0 , p 0 ) exposed to a


velocity gradient of

du g
dy

thickness, estimated at

dp
2

U dp
amb

7 2

6 / 7

, where is the momentum boundary layer

0.38 Rex 0.2 [19].

Equation (5.8) was solved for different particle diameters and the pickup velocity was calculated.
For comparison reasons, different terms of equation (5.8) were evaluated for different particle
sizes when exposed to their pickup velocity. These terms are reported in Table 5.1 in the form of
a non-dimensionalized ratio: force / weight. For the ranges studied, the Magnus force is much
smaller than other terms and can be neglected. The adhesion force seems to be important when
the particle diameter is smaller than 20 microns. The Saffman force is of the same order of
magnitude as the weight.

63

Table 5.1. Pickup velocity for different particle sizes. Non-dimensionalized forces on a stationary
particle. Comparison of the forces applied on a stationary 30-micron particle as functions of the gas
velocity. At gas velocities above 8 m/s, the net force on the particle becomes positive and the
particle is entrained to the gas

Diameter (m)

Upickup (m/s)

Fadhesion / Weight

FSaffman / Weight

FMagnus / Weight

1.2

0.573

1.57

2.9e-9

10

1.7

0.143

1.14

3.6e-8

15

2.6

0.064

1.06

1.4e-7

20

4.0

0.036

1.03

3.3e-7

25

5.9

0.0229

1.02

6.4e-7

30

8.3

0.0159

1.02

1.1e-6

The solution of equation (5.8) was obtained for different particle sizes and the results are plotted
in Figure 5.5.

The solution is compared with the semi-empirical correlation obtained by

Cabrejos [17]. Despite their similar increasing trends, the semi-empirical correlation predicts
higher pickup velocity. Several factors contribute to the observed difference: Pickup velocity in
[17] is defined as the minimum velocity for horizontal sediment transport, whereas the present
model evaluates the vertical pickup of particles. Additionally, the particles tested in [17] were
non-spherical, while the presented model assumes spherical particles.

64

Figure 5.5. Pickup velocity obtained from a force balance correlation (the presented approach) and
its comparison with semi-empirical correlation obtained by Cabrejos [17] for small non-spherical
particles with density of 1300 kg/m3.

Another mechanism of dusting is saltation, which is related to the tumbling motion of the solids
in the bed. Saltation is schematically illustrated in Figure 5.6. Some of the particles at the
bottom of the bed are lifted due to their contact with the rotary kiln wall or by means of lifters.
Particles picked up by the lifters are pushed up far above the highest point of bed (Figure 5.6).
Once these lifted particles fall back down towards the bed, they interact with the kiln gas. These
particles will follow distinctive trajectories under the influence of local gas velocity, drag, and
gravity. Lighter particles may follow the air streams and leave the kiln from the feed end in the
form of dust particles. On the other hand, heavier particles fail to enter into suspension and
descent towards the bed; either to rebound upon striking the bed, or embed themselves in the bed
and eject other particles.

65

Figure 5.6. Schematics of lifters at a transverse cross-section of a lime kiln and illustration of
saltation mechanism

The forces applied on an ejected particle include gravity and aerodynamic lift forces (i.e. drag,
Magnus, Saffman). These forces are the same as those experienced by a stationary particle on
the bed (as discussed above), except that there is no adhesion force on an in-flight particle
(because of loss of contact). In view of the fact that the adhesion force is generally smaller than
the gravitational force (see Table 5.1), adhesion force will have a negligible effect on the pickup
velocity. This means that particles that are too heavy to be picked up (when stationary), will fall
back down to bed if they are lifted through saltation mechanism. Also, particles that are very
light and are easily picked up (from rest), will be carried by the kiln gas if they are lifted through
saltation mechanism. Therefore, dust formation due to saltation can be correlated to the pickup
of a stationary particle. Saltation mechanism was therefore neglected in this study.

66

5.2 SSF Model


This section focuses on dispersion of the solid particles that are already lifted through pickup
mechanisms and are entrained into the turbulent air flow.
Due to computational limitations, it is not feasible to dynamically solve for trajectories of all
particles as a time-dependent particle-flow problem.

Therefore, representative particle

trajectories are solved. Stochastic Separated Flow modeling technique (SSF) is an alternative to
using dynamic model. In this model, the carrier phase is solved as a steady turbulent flow with a
commonly used turbulence model (e.g. k ). In this approach, the flow is solved separately
from the dispersed phase (i.e. assuming that the particles did not exist). A solution will
only provide average flow properties. Instantaneous velocities are then found by adding random
components (from a Gaussian fluctuating velocity distribution function) to the average flow
velocity. The model then assumes that a particle is injected into the flow and solves the
particles equation of motion (and therefore its trajectory) by determining the drag force based on
an estimated instantaneous fluid velocity at particle location. Finding trajectories of a large
number of particles produces the statistical information required for characterization of the
particle behavior [33].
Estimating the instantaneous flow velocity, which is the main challenge of SSF models, is
performed by sampling a random fluctuating velocity ( u ' f ) and by adding it to the mean flow
velocity ( u f ) [33]. Based on the method of determining u ' f , SSF models are subdivided into
two categories: models based on eddy lifetime, and models based on the time-correlated lifetime
[32].
In the eddy lifetime approach, a mean-zero Gaussian probability distribution function for u ' f is
produced and sampled from at each particle location. The sampled fluctuating fluid velocity acts
on the particle for a duration mimicking the fluid-particle interaction time. During this period,
the particle accelerates and moves with a drag force calculated based on the relative particle-fluid
velocity. The next time step will include a new sampling of u ' f from the distribution function
that has no correlation with u ' f in the previous step. Time correlated lifetime concept is similar

67

to eddy life-time concept, but it accounts for time-correlation between the particle velocity at the
current time and that at the previous time step [33].
Another challenging part of SSF model is to correctly estimate the interaction time between
particles and eddies. This interaction time depends on the average eddy lifetime, which has been
the subject of numerous studies.

Eddy lifetime, which is also known as the integral

timescale, fL , is a time interval during which the fluid velocity is correlated with itself. Semiempirical expressions for fL has been suggested by different researchers [33, 45]: In most
literature it is assume to be fL 0.3 / , where and are the local turbulent kinetic energy
and turbulent dissipation, respectively.

5.3 Solving Turbulent Fluid Flow


To be able to develop a verifiable model, a validation case was set up and modeled in Fluent. A
square channel similar to that in the experiment of Snyder and Lumley (SL) [62] is considered:
The channel is 41cm x 41cm x 4.9m and the air flow velocity is 6.55 m/s. GAMBIT software
was used to divide the domain into 30x30x200 structured cells (maximum aspect ratio of about
2). Turbulent flow was modeled using - model. No-slip boundary condition was chosen for
the walls, mass-flow-inlet for the intake and pressure outlet for the outlet of the channel.
Turbulence boundary conditions were assigned as turbulent intensity of 10% and hydraulic
diameter of 41cm (the model is not significantly affected by these numbers). The grids in the
region of particle flow were refined during simulation using grid adaptation feature of FLUENT
software (only one level of refinement was used). Mesh refinement for the walls were not
considered because the near-wall flow was not of importance in this case.
In Fluent, the average flow velocity and other fluid parameters can be determined and used at
any point in the geometry. The turbulent kinetic energy (), for example, provides information
about the average fluctuating fluid flow velocity: The fluctuating fluid velocity has a mean-zero
Gaussian distribution whose variance is proportional to . This distribution function is sampled
and a random value is generated using Monte-Carlo method. Instantaneous flow velocity is
determined by adding the fluctuating component of velocity to the average velocity. This

68

instantaneous velocity can be used to determine the instantaneous drag force on the local
particle.
Each particle is injected in the domain with a velocity of 6.55 m/s using a User Defined Function
(UDF, a manually written C-script that acts as an interface between the user and Fluents
sophisticated source code). Drag force exerted on the particle is then found from the local
instantaneous velocity. Drag coefficient relation, CD ( Re) , are adapted from [64]. The evaluated
drag force is assumed to last for a period of time mimicking the average time of interaction
between particles and eddies. After this period, drag force will be modified according to newlysampled fluctuating velocities.

The model then solves particles equation of motion and

determines its trajectory.


Seven hundred of each of the four particle types in the experiment of [62] are injected in the
numerical channel and their trajectories are found using SSF model. Particle trajectory data is
exported to MATLAB for post-processing. A MATLAB script is developed to: 1) import and
analyze the information of the 700 or more trajectories, 2) determine the average lateral position.
3) analyze dispersion information of the particles at different times and locations, 4) determine
spatial distribution of particles. Properties of the particles are listed in Table 5.2. The results of
the numerical modeling are compared with experimental data in Figure 5.7. The numerical result
compare experimental results well by about 20% error. In these figures Y2 is the mean squared
displacement of the particles with respect to their average position.
Table 5.2. Snyder and Lumley's particles and their properties. pl p d p2 / (18t ) is the particle
response time, or the time required for the particle to adapt to flow fluctuations. p is particle
density, d is the mean particle diameter and t is turbulence viscosity. Rep is based on mean
relative velocity and particle diameter.

Particle Type

Diameter (m)

p (kg/m3)

pl (ms)

Re p

Hollow Glass

46.5

260

1.7

0.05

Corn Pollen

87.0

1000

23.0

1.16

Solid Glass

87.0

2500

59.0

2.38

Copper

46.5

8900

60.0

1.51

69

Figure 5.7. Dispersion of copper, glass beads, corn pollen, and hollow glass particles as a function of
residence time. Presented numerical data is in good agreement with experimental data of SL [62],
with errors of less than 10%.

70

Chapter 6
RESULTS and DISCUSSION

This chapter discusses the results of the modeling work described earlier in Chapters 3 through 5.
In the first section, the models are validated by comparing the results of the model predictions
against several experimental studies.

The following sections present the results of the bed

model, including its movement and calcination reaction. Final section in this chapter discusses
the numerical results of dust formation, industrial implications and comparison between kiln data
and the numerical results.

6.1 Numerical Model Validation Results


Kiln systems that are being modeled in this thesis are complicated systems that encompass
several underlying reactions and physical phenomena.

Although modeling of separate

phenomena such as heat transfer mechanisms were validated separately, it is required to validate
the numerical model against cases that involve combinations of these physics.

Therefore,

numerical model of heat transfer and combustion is validated in this section against published
experimental results. Validation cases discussed here include an open flame (Mahmud et. al.
[44]), and a ducted flame (Kontogeorgos et. al. [37]).
In addition, a major validation case was run based on experiments conducted at a pilot kiln at the
University of British Columbia (UBC) by Alyaser et.al. [3]. This validation case verifies the
aspects of our modeling approach, from combustion to heat transfer mechanisms.

71

6.1.1 Non-Premixed Open Flame: Validation of combustion and


Radiation
Numerical modeling of radiation and combustion of methane air turbulent flame has been
validated against experimental and numerical results of Mahmud et. al. [44]. The burner used in
this experiment was a straight vertical tube with an internal diameter of 5 mm.

In this

experiment, fuel (99.5% methane) was injected into air. The velocity of fuel was 46.4 m/s
corresponding to Reynolds number of 1.4104. The experimental results presented by Mahmud
et. al. [44] were also accompanied by numerical modeling at the University of Leeds [44].
The experiment of Mahmud et. al. [44] was modeled to validate the numerical modeling
approach used in this thesis. The geometry was constructed and meshed using GAMBIT as 2D
axisymmetric (about 99000 structured grids, maximum aspect ratio of about 2.2), and was then
modeled in FLUENT. The turbulent transport of momentum was modeled using the
model.

Species-Transport model and P1 model were also used to model combustion and

radiation, respectively (see Chapter 3 for details).


Figure 6.1 shows temperature at seven cross sections of the kiln and velocity contours along the
flame path that has been modeled through FLUENT. Figure 6.1a shows temperature contour
along the flame path. The highest temperature occurs between 0.4 and 0.6 m at radial distance,
at which fuel has been burnt completely. Figure 6.1b shows velocity contour along horizontal
cross section of the kiln. Figure 6.1b shows that when fuel and air start to mix up, the velocity is
the highest. Figure 6.2 shows the comparison between experimental results of Mahmud et. al.
[44], which has been shown by circles, and the current numerical model which has been shown
by solid line, predicted radial profiles of the mean gas temperature versus the radial distance.
The graphs show a comparison between the numerical results of the current research and the
experimental results of Mahmud et. al. [44].

72

Velocity (m/s)

Temperature (K)

(a)

(b)

Figure 6.1. Temperature and velocity contours (a)temperature profile (K), (b)velocity profile (m/s).

Figure 6.2 shows that at cross-sections upstream of x=0.5m, the maximum temperature occurs at
a non-axis point, explaining the fact that the hottest part of a non-premixed flame is a layer in
which fuel and oxidizer mix. Downstream of x=0.7m, the combustion is complete, the flame
becomes colder, and flow becomes more uniform.

Figure 6.2 illustrates good comparison

between the presented numerical model and the experimental results of Mahmud et. al. [44].
Although difference of order of 13% exists between the experimental and numerical results, the
numerical model correctly predicts the trends and thus can be reliably used for kiln modeling.

73

Figure 6.2. Comparison between results of the presented numerical work and experimental
measurements reported by Mahmud et. al. [44]. Graphs show radial profiles of the mean gas
temperature at different cross sections. (-) current numerical results, (o) experimental measurements
of Mahmud et. al. [44].

74

6.1.2 Natural Gas Fired Furnace: Validating Combustion and


Radiation
In order to validate the approach taken in this thesis for modeling of combustion, heat transfer
and radiation, a confined combustion chamber [63] was modeled, and the results were compared
against available experimental [63] and current numerical [37] results. The selected chamber has
a separate primary and secondary fuel inlets, and the flame is confined in an open cylinder 990
mm in length and 293 mm in radius. Its geometry is schematically shown in Figure 6.3. Fuel is
injected from a central jet (r =3 mm), while primary air enters from an annulus (45mm outer
diameter, 15mm inner diameter). Secondary air also enters from second annulus in the direction
of primary air. Table 6.1 shows the inlet conditions for fuel, primary and secondary air inlets.
Kontogeorgos et. al. [37] developed a numerical code and compared its results with the series of
experiments conducted by Stoomer et. al. [63]. The numerical code used by Kontogeorgos was
an in-house two-dimensional CFD code, called 2PHASE, which was developed in the laboratory
of Heterogeneous Mixtures and Combustion Systems [37], for modeling reacting, multi-phase
and multi-component flows.

This code was used [37] to simulate the flow and chemical

reactions in the combustor chamber.


The geometry of the experimental setup was numerically constructed and meshed in GAMBIT
(about 127,000 structured mesh, maximum aspect ratio of about 2) and was solved in FLUENT
as a 2D axisymmetric geometry, based on the specified flow conditions. The numerical model
employed Species Transport with Eddy Dissipation model for combustion, and P1 model for
radiation. Radial temperature profiles for four axial locations were numerically predicted and the
results were compared to the measurements of Stoomer et. al. [63].

75

Table 6.1. Input flow conditions adopted from [37]

Inlet Boundary Conditions

Fuel inlet

Primary air

Secondary air

Axial Velocity (m/s)

21.9

4.8

0.3

Radial Velocity (m/s)

Turbulent Kinetic Energy ( m 2 / s 2 )

2.2

0.4

0.0004

Dissipation of Turbulence ( m 2 / s 3 )

500

50

0.0002

Temperature (K)

295

295

295

Kontogeorgos et. al. [37] developed a numerical model, which was used to validate current
numerical results. Figure 6.3 shows the temperature contour in an axial cross-section of the duct.
Figure 6.4 shows the temperature at the axial locations along the centerline of the chamber.
Moving from the fuel inlet to 400 mm downstream, axial temperature increases to about 1400
C. The current numerical results show a good agreement with the experimental measurement of

Stoomer et. al. [63] and the numerical modeling predictions of Kontogeorgos et. al. [37].
As shown in Figure 6.3, temperature levels decrease considerably downstream of the high
temperature flame zone. Radial temperature profiles at four different locations from entrance are
compared against the experimental and numerical results in Figure 6.4 [37].

The present

numerical model for non-premixed, turbulence natural gas flame compares well with the
experimental results. Despite the relatively large difference (about 25%) with the numerical
work of Kontogeorgos et. al. [37], the present model produces less error when compared to the
experimental measurements.
To further validate the combustion model, the mass fraction of H2O and CO2 were compared
against the numerical results of Kontogeorgos et. al. [37]. (Experimental measurements of
concentrations were not taken or reported). Both numerical works show a similar decreasing
trend for mass fractions of H2O and CO2 with increasing radial distance.

76

Temperature (C)

Secondary
Air
Primary Air
Fuel
Primary Air

Secondary
Air

Figure 6.3. Temperature contour along gas chamber geometry

Figure 6.4. Temperature profile along axial distance from inlet. (o) experimental measurements, (--)
numerical results of Kontogeorgos et. al. [37], and present numerical work

77

Figure 6.5 shows radial profile of temperature at four different axial locations in the chamber.
On the axis of flow (r = 0) at the inlet, non-premixed fuel is injected, but concentration of
oxidizer is zero. In the vicinity of the inlet, combustion reaction only occurs in a thin shell,
where fuel and primary-air mix. This shell is represented by the peaks observed in Figure 6.5.
Turbulent mixing thickens this shell further downstream.

Figure 6.5 also shows that the

maximum temperature at 50mm radial distance is about 1400oC, while at 250mm it is about
1600oC.

Figure 6.5. Radial temperature profiles at distances 50, 150, 250, and 400 mm from entrance. (o)
experimental measurements, (--) numerical results of Kontogeorgos et. al. [37], (-) Present Work.

78

Figure 6.6 compares mass fraction of water at four axial distances resulted from the presented
numerical model and that of Kontogeorgos et. al. [37]. Mass fraction of carbon dioxide at three
axial distances decreases in Figure 6.7 when axial distance increases which is consistent with
both numerical models. The difference (about 30%) between the model predictions can be
explained by the fact that simplifying assumptions and approximate material properties were
assumed in both numerical works.

Figure 6.6. Mass Fraction of H2O at radial distances distances 500, 600, 700, and 800 mm from
entrance. (--) numerical results of Kontogeorgos et. al. [37], (-) Present Work

79

Figure 6.7. Mass Fraction of CO2 at radial distances distances 500, 600, 700, and 800 mm from
entrance. (--) numerical results of Kontogeorgos et. al. [37], (-) Present Work

6.1.2.1 Mesh Dependency

To test the grid dependency of the solutions, two additional mesh grids (coarse and fine) were
generated to model the above experiment: coarse mesh of 31,000 elements and a fine mesh of
508,000 elements. The axial velocity and temperature profiles along the axis were compared for
the three different mesh grids and are presented in Figure 6.8 and Figure 6.9. It is observed that
discretization errors are significantly small when intermediate mesh sizes are used.
80

Figure 6.8. Grid dependence of axial velocity profile of natural gas fired furnace for coarse,
intermediate and fine meshes.

Figure 6.9. Grid dependence of temperature profile of natural gas fired furnace for coarse,
intermediate and fine meshes.

81

6.1.3 Pilot Kiln, Gas Phase


To validate the performance of the presented numerical modeling approach in predicting kiln
operations, it is necessary to test its performance in a similar geometry with similar conditions.
Since such validation is not practical in operational kilns, a pilot kiln was used to validate the
numerical model for combustion and heat transfer. Experiments conducted by Alyaser [3] on a
pilot scale kiln at the University of British Columbia (UBC) were modeled in this work. The
pilot scale kiln is 5.5m-long and 0.4 m in diameter. Figure 6.10 shows the geometry used in the
numerical model of the pilot kiln, including the fuel nozzle and primary and secondary air jets.
Burner at the pilot scale kiln uses natural gas which contains 95% methane as fuel and burns it
with air (as oxidant).
At the inlet of the burner in the pilot kiln, fuel is injected and mixed with primary air. Primary
air does not usually provide enough oxygen to combust all of the fuel. Downstream of the
burner, the methane-air mixture is further mixed with secondary air for a complete combustion.
Table 6.2 summarizes the mass-flow-rates of primary and secondary air in three different
experimental cases considered here. The fuel flow rate was 5.1 m3/hr for a fuel velocity of 2.47
m/s.

Temperature measurements inside and outside the pilot kiln were used to validate

predictions of the developed numerical model.

82

Figure 6.10. Geometry of the pilot kiln at UBC was created and meshed in GAMBIT.

Table 6.2. Flow rate, and Craya-Curtet number for experiment conducted at UBC pilot scale kiln [3]

Parameter

Case1

Case2

Case3

Gas Flow Rate (m3/hr)

5.1

5.1

5.1

Primary Air (m3/hr)

46.6

34.9

23.3

Secondary Air (m3/hr)

11.7

23.3

34.9

Primary / Secondary Air

0.8

0.6

0.4

Craya-Curtet Number (Ct)

0.41

0.55

0.80

6.1.3.1 Geometry and the Mesh

Because of the secondary and primary air inlets, the geometry of the kiln cannot be assumed
axisymmetric.

This means that the geometry should be constructed and meshed in three

83

dimensions. GAMBIT was used to construct the geometry of the pilot kiln. Hexagonal and
tetrahedral meshes were used. About 150,000 meshes were used. The maximum skewness was
about 0.81. The contraction/expansion factors used for meshing were between 0.9 and 1.1.
Additionally, FLUENTs grid adaptation feature was used to refine the meshes at the inlets and
close to the wall boundaries, such that they would resolve the steep temperature and velocity
profiles in those regions. The total number of mesh elements was about 250,000, after grid
adaptation.
To study mesh dependence of the solution, two additional mesh grids (coarse and fine) were
created for the geometry and were used in this study. The total number of elements for the
coarse and fine meshes was about 150,000 and 780,000. Velocity profile was looked at for the
three meshes. The velocity profile showed 8% variation from coarse mesh to intermediate mesh,
and 2% from intermediate to fine mesh. Since the computational time of the fine mesh was
excessively high (in excess of two weeks), intermediate mesh size was considered for this study.

Figure 6.11. Grids used for numerical modeling. Because of symmetry, only half of the pilot kiln
needs be meshed.

84

6.1.3.2 Numerical Model

A model was set up to model fluid flow, combustion and heat transfer. Turbulence effects were
modeled using the standard model of turbulence. The main source of energy input to the
kiln is the fuel combustion in the burner region, also known as discharge end (see Figure 1.3).
The generated heat by the flame is conducted and radiated away towards the surrounding gas and
wall boundaries. The methane-air reaction is modeled here as:
CH 4 + 2O 2 CO 2 + 2H 2 O + 802

kJ
mole of CH 4

(6.1)

A species transport model was used to model methane-air mixing and combustion. In this
model, species of the inlet air and fuel are considered and solved separately. These species are
transported through the geometry according to conservation laws (mass, energy, and
momentum). If the temperature is high enough for combustion process in the regions where CH4
and O2 are mixed, these species are replaced with H2O and CO2 molecules (according to
Stoichiometry and reaction rate for the process). This process produces heat. Also, finite-rate
formulation with eddy-dissipation (see section 3.3.1) was used to simulate the finite reaction
rates. This model is shown to yield fast convergence and accurate predictions.

Conduction is modeled using Fouriers law of conduction ( q k T ) and also advection of


fluid according to general energy equation that is solved within the geometry.
The heat produced by the combustion reaction is radiated and also conducted through the kiln
gas to the wall boundaries. The wall boundaries absorb the radiated heat and also conducted heat
and transmit that to the ambient gas (outside the kiln) by means of conduction and radiation.
Convection can be modeled in numerical models by refining the mesh within the formed thermal
boundary layer region. This means that when a hot gas flows in a kiln and forms a thermal
boundary layer over the kiln wall, mesh spacing close to the wall boundaries should be fine
enough to resolve the temperature gradients within the thermal boundary layer.
To test the ability of FLUENT in modeling convection using a fine mesh within the thermal
boundary layer, a series of benchmark cases were studied (Appendix A). It was concluded that

85

when the number of nodes in the boundary layer is more than 8, the prediction of heat flux
through the thermal boundary layer is within the 3% of that expressed by semi-empirical
formulas of h. It is therefore expected that a solution to a complex case (e.g. the kiln) with more
than 8 nodes in its thermal boundary layer will provide an acceptable estimate of heat flux
through the thermal boundary layer.
Because of the fuel that is combusted in the burner region in the form of flame, the maximum
temperature of the kiln gas can reach 2000K. This high-temperature suggests that radiative heat
transfer is an important factor in heat transfer and cannot be neglected [26 ,60]. P1 model of
radiation was used to solve radiative heat intensity inside the kiln. It is relatively efficient in
terms of computational time and it also takes into account effects of absorption/emission by/from
gas as well as wall boundaries.
6.1.3.3 Boundary Conditions

The fuel nozzle inlet was considered as a mass-flow-inlet boundary. Primary and secondary
nozzle inlets were also considered mass-flow-inlet boundaries.

This means that the fluid

velocities at these inlets are set to values that result in the specified mass-flow-rate. Turbulent
intensity of 10% and appropriate hydraulic diameters were assigned to fuel and
primary/secondary air inlets. Symmetry boundary condition that is applied on symmetry planes
(e.g. YZ plane in Figure 6.11) assumes that gradient of all fluid properties are zero in normal
direction (normal to boundary). (Figure 6.2)
The outlet of the kiln was considered as an outflow boundary. This type of boundary condition
is used to model flow exits where the details of the flow temperature, velocity and pressure are
not known prior to solution of the problem. The model extrapolates the required information
from within the domain. This type of boundary condition assumes a zero diffusion flux for all
variables.
6.1.3.4 Results

Figure 6.12 compares numerical predictions of the temperature profiles at the centerline, inner
wall, and outer wall of the pilot kiln to the experimentally measured temperatures [3]. The solid
lines are the numerical results of the current research and the dots are the experimental results of
86

Alyaser [3]. The maximum temperature is roughly 1300oC at the centerline and the temperature
drops downstream to about 500oC.
These comparisons show that the developed numerical model provides a reliable tool to model
heat transfer and combustion, and to estimate temperature profiles in lime kilns.

87

(Case 1)

(Case 2)

(Case 3)
Figure 6.12. Comparison between experimental measurements of Alyaser [3] and numerical
modeling results for cases 1, 2, and 3 in Table 6.2.
88

6.1.4 Pilot Kiln, Solids Phase


The numerical model of the bed developed in the present work (described in detail in Chapter 4)
was validated by modeling the experiments conducted at the University of British Columbia on a
pilot scale kiln [5]. The geometry of this pilot kiln is shown in Figure 6.10. Measurements of
the bed motion, the calcination reaction and the heat transfer from the gas phase to the bed were
compared against numerical predictions. Calcination of lime, which occurs at about 800oC [68]
to 900oC [70], was also included in the model. The numerically obtained temperature profiles at
the axial position along the kiln length are plotted in Figure 6.13.

The experimental

measurements were taken inside the bed and inside the gas, 10 and 25 centimeters off the bed.
This figure clearly shows that the mud temperature increases inside the solid lime bed from the
feed end to the discharge end. This figure shows a good agreement between UBC experiments
and the present numerical results. Although the numerical model produces an error of about
10% in predicting the temperature profile, the trends are fully compatible with the
experimentally observed results.
The modeling approach can therefore be reliably applied to the full scale kiln to study effects of
different operating conditions.

89

Figure 6.13. Temperature versus axial position on pilot scale kiln [5]

6.2 Full Scale Lime Kiln


A Canadian kiln has been used in this study for modeling purposes [1]. This lime kiln is a 108
meter long cylinder and 3.5 meter internal diameter. Secondary air enters the kiln from 9
identical circular inlets, called satellite coolers, with internal diameters of 0.3 m, while primary
air and fuel enter the kiln from a 46 cm internal diameter burner. Figure 6.14 shows the internal
shell of the studied Canadian lime kiln. The geometry was constructed and meshed in GAMBIT
and simulated in FLUENT. This section discusses details of the kiln geometry, boundary
conditions used to model the gas phase and combustion, and the numerical results of this model.
It should be noted that effect of bed is not included in this section and will be discussed in the
next section.

90

Figure 6.14. Geometry of the inside shell of the full scale lime kiln used in the presented work.

6.2.1 Boundary Conditions and Operating Conditions


Some of the operating parameters of the kiln, including the mass-flow-rate of fuel and primary
air, and temperature of the fuel and air at inlets were provided by the kiln operators. These
values are measurements taken on site by the kiln technicians regularly to ensure satisfactory kiln
operation. However, not all required parameters can be measured. For example, the amount of
excess-air, which is typically between 10% and 30%, can not be measured. Usually excess air is
obtained from a simple mass balance inside the kiln by knowing the amount of primary and fuel
flow rate and also the produced lime. In this research, three different operating conditions will
be assumed to account for all possible cases:
Low excess-air value (10%)
Medium excess-air value (20%)
High excess-air value (30%)
A geometry, based on the drawing of the Canadian kiln, was constructed and meshed in
GAMBIT, and was then simulated in FLUENT. Figure 6.15 shows the outer and inner shell
91

geometry of the kiln used in this thesis. About 250,000 hexagonal and tetrahedral meshes were
generated. The maximum skewness was about 0.83. The contraction/expansion factors for
meshing were between 0.9 and 1.1. Additionally, FLUENTs grid adaptation feature was used to
refine the meshes in regions of steep gradients during simulation.
Mass-Flow inlet boundary conditions were assigned to the fuel and air inlets. Mass-flow-rate
of primary air and fuel were used to calculate mass-flow-rate of different species (oxygen,
nitrogen, methane, carbon dioxide, water) at the inlets. Different flow-rates are shown in Table
6.3.
The mass-flow-rate of secondary air was calculated based on the level of excess-air (mass
balance) and species mass-flow-rates were assigned accordingly to the secondary air inlet, which
was assumed to have a mass-flow-inlet boundary condition (specified mass flow-rate).
The outlet of the kiln was set as an outflow boundary condition with specified pressure. Wall
was set to be a no-slip wall boundary which could transfer heat to the ambient air based on
equations (3.37) and (3.39). The governing equations were then solved using second order
discretization schemes. The relaxation factors for the convergence were varied between 0.6 and
1 to create a balance between convergence speed and stability of the solution.
Air and fuel were both assumed to be ideal gases with temperature-dependent properties:
viscosity, specific heat, density, and absorption coefficients were assumed to be functions of
temperature. Pressure functionality was not considered because of the negligible pressure drop
in the kiln, which is much less than atmospheric pressure.

92

Figure 6.15. Meshed Geometry of the kiln: the outside refractory lining (top grid) and the inside
volume within the refractory shell (bottom grid)

As described in detail in section 3.3, all of the heat transfer mechanisms, including convection,
conduction, and radiation were included in the model. The model was then solved and solutions
were extracted, processed and the trends were studied.

6.2.2 Gas Phase


The main objective of this thesis is to model dusting and to determine the effect of different
operating conditions on dusting. Since modeling dusting requires that the carrier phase (kiln gas)
be modeled, it is therefore required that a model of kiln gas be developed. Also, effect of kilns
operating conditions on the kiln gas has to be determined. It is therefore necessary to determine
93

a realistic range for each operating condition. Some of the operating conditions are directly
measured, such as fuel flow rate, inlet gas and fuel temperatures, primary air flow rate, and mud
flow rate.

However, some other parameters are not directly measured and are therefore

estimated. One example is the secondary air flow rate, which cannot be directly measured, but is
estimated based on the measurements of oxygen level at the feed end and the following
relationship for excess-air-ratio, e:
CH4 + (1 + e) 2(O2 + 3.76 N2) CO2 + 2H2O + 2e O2 + 7.52(1+e) N2

(6.2)

Based on the operating conditions of the Canadian kiln, and also based on some estimated values
for excess-air values, 12 cases were selected/constructed such that most typical operating
parameters of kilns are covered. Table 6.3 tabulates these twelve cases. Figures in Table 6.3
show different cases with different Craya-Curtet numbers, primary, secondary and fuel flow
rates.
The numerical solutions confirm dependency of flame length on Craya-Curtet number, and agree
with previous studies by Aloqaily [1] that increasing the Craya-Curtet number increases the
length of the formed flame.
Another observation was that adding the effect of buoyancy increases the temperature of the
upper wall of the kiln, which is consistent with previous observations [2, 27] and well-known
physics of buoyancy. Since the presented model is validated against the experimental results of
the pilot kiln at UBC, it is speculated that the model works well for the case of a full scale lime
kiln (for which experimental results are not available for validation).

94

Table 6.3. Temperature contours for twelve cases covering operating conditions of the Canadian
lime kiln. Craya-Curtet numbers cover a range of 0.39 to 0.95, and excess-air value was varied

Temperature (K)

Fuel (m3/s)

Primary Air (m3/s)

Excess Air (%)

Ct

between 10% and 30%.

0.81 10% 0.64 0.58

0.61 10% 1.07 0.61

0.47 10% 1.55 0.61

0.39 10% 1.41 0.44

0.88 20% 0.64 0.58

0.66 20% 1.07 0.61

0.51 20% 1.55 0.61

0.43 20% 1.41 0.44

95

Table 6.3 Continued

0.95 30% 0.64 0.58

0.72 30% 1.07 0.61

0.55 30% 1.55 0.61

0.46 30% 1.41 0.44

Figure 6.16 shows the heat flux to the inner shell of the kiln. This figure shows that the heat
transfer to the kiln wall is at its maximum in the flame tip region (due to radiation) and also at
the top of the kiln (due to gravity). This figure also shows that the contribution of radiation to
the total heat flux at its maximum point is about 70%, while its contribution diminishes
downstream of the kiln.

96

Rate of Heat Transfer (W/m2)


2

Figure 6.16. Total (top) and Radiative (bottom) heat fluxes to the inner wall of the kiln for the case
of Ct=0.46. Only 40m of the kiln length is shown.

Figure 6.17 shows temperature contours on the outer and inner shell of the kiln. For this model,
wall thickness was constant (except in the dam region) and there was no rotation. Also in Figure
6.17, there is a hot spot on the upper wall of the kiln which is due to flame impingement on the
inner layer of the upper wall. The predicted maximum temperature is moderately higher than
what is observed in actual kiln operations.

This is probably due to the effect of Non

Condensable Gas (NCG) [1] that is usually injected above the fuel inlet. The NCG nozzle is
typically smaller than the burner nozzle, but, the low-temperature high-density NCG can push
the flame downward, preventing it from touching the top wall, and therefore, reducing the
maximum temperature. The main burner itself is also oriented toward the bed in most cases,
which directs it away from the top kiln wall.

97

Figure 6.17. Temperature Contour on inner wall (bottom) and outer wall (top) of the lime kiln for
the case of Ct=0.95. Only 40m of the kiln length is shown

6.2.3 Solids Phase


6.2.3.1 One Dimensional Modeling

The first attempt to modeling the bed of a full scale lime kiln was a 1D modeling approach which
was discussed in detail in section 4.1. This model assumes average properties at bed crosssection and solves the equations of evaporation, calcination, and heat transfer along the length of
the kiln. This model estimates the rate of heat transfer from the kiln gas to the bed based on the
information obtained from the gas phase model (discussed in section 6.2.2). Equations of mass
and energy conservation were solved in one-dimension (along the kiln length) to obtain average
bed properties along the kiln.

98

The one-dimensional modeling is particularly useful because inclusion of vaporization in the 3D


model was very challenging. Therefore, the evaporation phenomenon was only modeled in the
1D model and the moisture content of the bed was fictitiously applied to the 3D model (as a heat
sink for the energy equation).
Figure 6.18 shows the average bed temperature obtained from the 1D kiln model and the average
gas temperature obtained from the 3D CFD model along the kiln. The assumed operating
conditions are those listed in Table 3.1, with fuel flow rate of 0.58m3/s, primary air flow rate of
0.64m3/s, and excess-air value of 10%. (This is the first case out of the 12 cases defined in Table
6.3). This figure shows that the average bed temperature increases from about 20C (at the feed
end, x = 108m) to about 100C in about 12 meters from the feed end. The bed temperature
remains almost constant until all moisture is evaporated (x = 90m). Once the moisture is fully
rejected, the temperature starts to rise to about 850C, where the calcination reaction starts to
occur. In the calcination zone the temperature remains almost constant until the bed solids reach
the flame region and calcination becomes almost complete and the temperature starts to rise
again.

99

Figure 6.18. Average gas and bed temperatures along the kiln, calculated based on the 1D model of
the kiln. Mud flow rate: 8.5 kg/s, fuel flow rate: 0.58m3/s, primary air: 0.64m3/s, excess-air: 10%.
This is the first case of the operating conditions in Table 6.3.

The 1D model was also used to determine the evaporation rate, moisture content, rate of
calcination, and degree of calcination. Figure 6.19 shows the bed composition along the length
of the kiln. Moving from the feed end (x = 108m) towards the hot end (x = 0m), it is observed
that the moisture level remains constant for about 10 meters during the initial heating process,
until the bed temperature reaches 100C, high enough for the evaporation to start. Then the
moisture evaporates (while temperature is constant). In this section the CaCO3 level increases
because the moisture is rejected from the bed material and percentage of the CaCO3 in the solids
bed becomes close to 100%.
Once the moisture is fully evaporated, the bed temperature gradually increases until it reaches
temperature of about 850C, at which the calcination reaction picks up. Since calcination is
100

endothermic and requires heat, the bed temperature does not significantly increase while (or in
the region where) calcination reaction occurs.

Once the bed is fully calcined, the bed

temperature increases again, leading to temperatures above 1000C.


In the configuration studied here the bed is fully calcined in about 100m.

Figure 6.19. Average composition of the solids bed and percentage of mass versus distance from the
hot end, along the length of the kiln. Mud flow rate: 8.5 kg/s, fuel flow rate: 0.58m3/s, primary air:
0.64m3/s, excess-air: 10%. This is the first case of the operating conditions in Table 6.3.

The abovementioned approach was used to predict the bed composition for the remaining 11
cases defined in Table 6.3. Average air temperatures were taken from 3D CFD modeling of the
gas phase. Calcination reaction was modeled in 1D and bed compositions were determined. The
results are summarized in Figure B.4 to Figure B.14 in Appendix B.

101

Although the presented results assume several simplifying assumptions, they can be used to
comparatively study the effect of different operating conditions.
Comparison of data in Figure B.4 to Figure B.14 indicates that variation of kilns operating
conditions does not significantly affect the trend of temperature profile in the kiln. Variation of
operating conditions only causes relatively small changes in the temperature of the gas and the
bed. It is found that using the highest fuel rate (0.61 m3/s) produces a peak temperature of
1554C in the kiln gas, whereas, the lowest fuel rate (0.44 m3/s) produces a peak temperature of
1456C.
The temperature variation of the bed material (due to change of kilns operating conditions) is
less than 55C before completion of the calcination reaction. After completion of the calcination
reaction, the bed temperature may increase by about 100C. This increase will greatly depend on
the location where the reaction is complete. If the calcination reaction is not fully completed
when the bed material discharges, this temperature increase is not observed.
Unlike the temperature profiles that are not significantly affected by the kilns operating
conditions, the bed composition and temperature at the discharge end greatly depend on the
operating condition. This can be explained by the long exposure time of the bed material to the
temperatures inside the kiln: Although temperature difference is small, the reaction rates (that
increase exponentially with temperature) can be very different, and the effect becomes
pronounced due to the long reaction time.

6.2.3.2 Three Dimensional Modeling

After successful validation of the 3D numerical model of the bed against the experimental results
of Barr at UBC (section 6.1.4), the model was applied to the geometry of full-scale kiln (Figure
6.14). This geometry incorporates a rotating wall, air and fuel flow, combustion, a moving lime
mud bed with calcination reaction, and heat transfer mechanisms.
The numerically-obtained temperature contours at different cross sections along the kiln length
are shown in Figure 6.20. This figure clearly shows that the mud temperature increases inside

102

the solid lime bed from the feed end to the discharge end. Calcination reaction has also been
included in this model.

Figure 6.20. Temperature contour in the solids bed section of the lime bed

Figure 6.21 shows contours of axial velocity in several cross-sections of kiln gas close to the hot
end (flame region). This figure shows the effect of flame on the gas velocity profile experienced
in the gas section. It is shown that the bed surface experiences slightly lower than average gas
velocity in the sections immediately downstream of the flame. This is due to the fact that when
the gas is not well mixed, and there is temperature non-uniformity at a given cross-section, the
high temperature gases tend to rise due to buoyancy. This causes the flame to curve up, meaning
that the tail of the flame moves towards the top of the kiln. This also suggests that air-velocity
and air-temperature on the surface of the bed are relatively low.

103

Figure 6.21. Contours of axial velocity in the bed and the gas. (the contours were obtained from
different simulations and were combined into this figure for comparison reasons). Buoyancy causes
the tail of the flame to move upward.

The contours of bed velocity and gas velocity in Figure 6.21 were obtained from different
numerical models and were combined into Figure 6.21 for comparison reasons.

Negative

velocities observed in the bed region occur because the mud and the gas move in opposite
directions.
Velocity vectors inside the bed, as illustrated in Figure 6.22, clearly show the movement of the
solids inside the bed.

104

Velocity (m/s)

g
Figure 6.22. Velocity vectors inside the bed show movement of solid lime particles inside the bed

105

6.3 Dust Generation Rate


The ultimate goal of this thesis is to estimate the rate of dusting, and to estimate the effect
operating conditions have on the rate of dusting.

6.3.1 Procedure to Calculate Rate of Dusting


To estimate the rate of pickup of dust particles from the bed surface, the following procedure
was followed:
1. Information of particles in the kiln bed and gas flow in the kiln gas:
From the 1D kiln bed model, the area of the dusting zone was estimated. Since majority
of dust particles are picked up from the regions with dry CaCO3, dusting zone was
considered to be regions with moisture content less than 1% and degree of reaction less
than 5%.
From the 3D kiln gas model, information of the kiln gas flow in the dusting zone was
extracted: Kiln gas velocity, turbulent kinetic energy, and turbulent dissipation rate were
averaged at transverse cross-sections and were used as functions of axial location.
2. Domain discretization:
The dusting zone of the kiln was discretized using 200 uniform elements of size x and
width Lc as shown in Figure 4.2.
x

3. Rate of pickup from each element:


Size-distribution was divided into 5 ranges of particles. Number of exposed particles
from each range was evaluated based on the surface area of element x and average
particle diameter.
Pickup velocity was obtained for each range of particle size

106

Probability of pickup of an average particle from each range in a turbulent flow field of
average kiln velocity (Ukiln, evaluated locally from 3D kiln gas model) was obtained
Rate of pickup from each element for each range of particle size was obtained from the
following:

all ranges
of particle
size

number of exposed particles in element x, in one range of particle size


probability of pickup average mass of particles in each diameter range
characteristic time of pickup

Characteristic time of pickup was assumed to be the same as the momentum response
time of each particle
3. Total rate of dust pickup:
The total rate of dust pickup was calculated by adding the rate of pickup from all
elements.
These steps are explained in more detail here:

6.3.2 Rate of Dusting


A general model of particle pickup was presented in Chapter 5 using the forces applied on a
stationary particle on a flat bed (Figure 5.1). This model was used to find the minimum ambient
gas velocity that can apply a net upward force on a stationary particle sitting on a flat surface.
In order to estimate the rate of dusting, the force balance equation (5.8) was solved for an
ensemble of cases. This ensemble included a wide range of particle diameters observed in the
dust samples (2 microns to 60 microns), and also a wide range of average overhead gas
velocities, 2 m/s to 40 m/s. It should be noted that typical kiln gas velocity is about 6 m/s.
Density of particles was assumed to be 2830 kg/m3. [68]
The force balance equation was then solved for each particle size. The minimum gas velocity
that initiates particle pickup was then calculated. The correlation between particle diameter and
minimum pickup velocity is plotted in Figure 6.23.

107

Figure 6.23. Minimum gas velocity required to pickup a particle as a function of the particle
diameter. Density of particles was assumed to be 2830 kg/m3.

For each nominal value of the gas velocity, it was assumed that the actual gas velocity can vary
according to a distribution function, whose full-width-half-maximum is 2.35 , where is the
turbulent kinetic energy. was calculated based on a turbulence intensity ratio of 15%, which
is a typical value for gas flows.
The velocity distribution function was sampled 500 times for 5 particle sizes (5, 10, 15, 20 and
30 microns). For each of the sampled velocity values, the force balance relation (5.8) was
evaluated to determine whether the lift force is strong enough to initiate lift-off.
The value of the lift-force is not the only factor that determines whether a particle lifts off. For
momentum exchange to occur, it is necessary for the particle to experience a net upward force

108

for a period of time on the order of its momentum-exchange characteristic time [22]. The
momentum-exchange characteristic time is given by Crowe et. al. [22] as p

pd 2
.
18air

The time-dependent velocity function in a turbulent flow is auto-correlated. The auto-correlation


function, defined as R ( ) u (t )u(t ) , drops exponentially with time. The characteristic
time of the velocity-autocorrelation, c , is estimated to be of the order of c 0.3 / [62],
where and are turbulent kinetic energy and turbulent dissipation rate (found from gas phase
model solved in the FLUENT). Therefore, the probability of the velocity function to maintain a
certain value for a time period of , can be estimated to be e / c .
Thus, the probability of the fluctuating velocity field to maintain a value above the minimum
pickup-velocity is the product of the probability of the instantaneous velocity to exceed the
minimum pickup velocity, and the probability of the velocity field to maintain its value for a
period of p , which is e

p / c

. The resultant probability is plotted for the case of lime particles in

Figure 6.24.

109

Pickup Probability (%)

Average Gas Velocity (m/s)


Figure 6.24. Pickup probability of particles of different sizes as a function of average gas velocity.
Turbulent intensity of the gas was assumed to be 15%, which is a typical value for gas flows. [26]

In order to associate the rate of pickup with the probability of pickup, it is necessary to estimate
the number of particles available for pickup. Since the pickup probability varies with particle
size, particle size distribution affects the pickup behavior and needs to be taken into account. For
this purpose, size distribution of dry lime particles was experimentally measured using a Particle
Size Analyzer (MasterSizer S; Malvern Instruments, UK), and the resultant distribution was
divided into 5 sections. (Figure 6.25)

110

Number fraction (%/m)

Diameter (m)
Figure 6.25. Size distribution (number fraction) of dry lime particles in bed. Based on measurements
using Particle Size Analyzer (MasterSizer S; Malvern Instruments, UK)

It was then assumed that:


The particle size distribution within the bed is the same as that of the particles sitting on
the bed surface.
The characteristic time of pickup is p , the particles momentum-exchange characteristic
time
The particles are picked up layer by layer (packing effect was neglected)

111

Based on the above assumptions, estimate of pickup probability, and also the percentage of
exposed particles on a 40 m2 surface (typical area of dusting zone in kilns), mass flow rate of
pickup was estimated for each range of particle diameter, as shown in Figure 6.26. It is observed
that smaller particles have a higher pickup probability, but contribute little to the mass flow rate
due to their small mass. For example, particles that are smaller than 10m in size (11% of the
total number of particles), are mostly picked up when gas velocity is greater than 6m/s, but their
maximum dust generation rate is around 400gr/s (relatively low).

On the contrary, larger

particles are less likely to be picked up. But once the gas velocity is high enough to pick them
up, they introduce a relatively high mass flow-rate of dust. For example, particles that are
between 10m and 15m (39% of the total number of particles), are picked up at velocities
above 8m/s, but their maximum dust generation rate is 2.6kg/s.

112

Figure 6.26. Particle pickup rate (in kg/s) evaluated for several ranges of particle sizes exposed to an
ambient air flow velocity between 0 and 12 m/s. Turbulent intensity was assumed to be 15%, surface
area was assumed to be 40m2. Smaller particles have a higher pickup probability, but contribute little
to the mass flow rate due to their small mass.

Total dusting rate (Figure 6.27), was then calculated by adding the curves of pickup rates for
different ranges of particle size shown in Figure 6.26.

This curve shows that dusting is

negligible when the gas velocity is below 4 m/s. Beyond 4m/s the rate of dusting exponentially
increases, but the rate of increase diminishes at velocities above 8m/s due to unavailability of
particles.

113

Dust Production Rate / Mass Feed Rate

Gas Velocity (m/s)


Figure 6.27. Percentage of dusting rate versus average kiln gas velocity.

Despite the fact that these results are based on many simplifying assumptions, they can provide a
good understanding of the effect of different operating parameters. These results can be reliably
used for comparative studies.
One of the factors that affect the rate of dusting is the area of the dusting zone. It is known that
most of the dust particles are made of dry CaCO3 particles. It can therefore be assumed that
dusting is mostly initiated in a region where mud is fully dried, but is not yet calcined. In the
calcining zone, where lime forms and nodulizes, lime nodules often agglomerate and grow as
they move towards the front end (burner/hot end) of the kiln. Since pickup of large nodules is
unlikely, dust generation in the calcining zone is unlikely and can be neglected. In the presented
modeling approach, area of the dusting zone was estimated based on the results of the 1D bed

114

model by estimating the length of the bed section with less than 5% moisture and more that 95%
CaCO3.
To study the effect of kilns operating conditions on dusting, the dust generation rate was
estimated for the 12 cases defined in Table 6.3, covering the ranges of kilns typical operating
conditions.

The approach discussed above was applied to each case to estimate the bed

composition profile, length of dusting zone, gas properties in the dusting zone, and dust
generation rate. The results are summarized in Table 6.4.

115

Table 6.4. Effect of kilns operating conditions on dust generation rate (percentage of mud feed

Mean gas temperature


in the dusting zone (K)

Total mass flow rate


of gas (kg/s)

Mean gas velocity in


the dusting zone (m/s)

Dust generation
rate (%)

8.4

5.3

5%

0.61 10% 1.07 0.61

44

818

8.8

5.5

6%

0.47 10% 1.55 0.61

46

834

8.8

5.6

7%

0.39 10% 1.41 0.44

49

798

6.4

3.9

1%

0.88 20% 0.64 0.58

47

817

9.1

5.7

8%

0.66 20% 1.07 0.61

43

821

9.6

6.0

10%

0.51 20% 1.55 0.61

43

825

9.6

6.1

10%

0.43 20% 1.41 0.44

48

795

6.9

4.2

1%

0.95 30% 0.64 0.58

42

830

9.9

6.3

12%

10

0.72 30% 1.07 0.61

43

826

10.3

6.6

15%

11

0.55 30% 1.55 0.61

44

824

10.3

6.6

15%

12

0.46 30% 1.41 0.44

49

802

7.5

4.6

2%

Fuel (m3/s)

822

Primary Air (m3/s)

48

Excess Air

0.81 10% 0.64 0.58

Ct

Case #

Length of dusting
zone (m)

rate). The operating conditions of the 12 cases are defined in Table 6.3.

As discussed above, dust generation rate depends on several factors, including the area of the
dusting zone and the gas velocity in the dusting zone.

However, the range of operating

conditions used in Table 6.4 only covers a small range of values for the length of dusting zone
(from 42m to 49m). Therefore, this study may not fully reveal the impact of length of dusting
zone. (One may model different kiln diameters/lengths to consider a broader range of dusting
zone area, but it is beyond the scope of this study).

116

Table 6.4 clearly shows that the effect of gas velocity dominates other factors. To further
demonstrate this, the dust-load predictions of the 12 cases are plotted in Figure 6.28. The trend
of dust-generation-rate in this figure is similar to that of Figure 6.27. It is clear that, in the range
of operating conditions studied, the rate of dusting is dominantly determined by the mean gas
flow-rate, and effect of other operating conditions are small or negligible. The data points of
Figure 6.28 can be described by a power fit as y 1.28 104 x 6.3 with R 2 (goodness of fit) of
0.983.

Figure 6.28. Dust generation rate (percentage of feed rate) plotted as a function of gas velocity for
the 12 cases defined in Table 6.3. The curve represents the line of best power fit: y 1.28 104 x 6.3 .
The coefficient of determination ( R 2 , a measure of goodness of fit) for this fit is 0.983.

117

6.3.3 Effect of Size Distribution


To determine the effect of particle size distribution on the rate of pickup, three different particle

Number fraction (%/m)

size-distributions were tested and the rate of pickup was calculated.

Diameter (m)
Figure 6.29. Three size distributions tested for rate of pickup.

The procedure outlined in section 6.3.1 was followed for each of the size distributions and the
rate of pickup was determined. Comparisons of the pickup rates for these particle sizes are
presented in Figure 6.30. It is observed that for all distributions studied, the dust load varies as a
power of average gas velocity. The line of best power-fit was calculated for each of the tested
particle size distributions. The coefficients of determination (goodness of fit) for the fits to data
of size distributions 1, 2, and 3 are, respectively, 0.983, 0.966, and 0.976.
As expected, the rate of pickup and dust load decrease when smaller particles are distributed in
the bed. It is observed that the exponent of average velocity in the obtained correlations (slope
of the fits in logarithmic scale) decreases with particle size.
118

Figure 6.30. Dust Load, defined as the total rate of pickup divided by the mud flow rate, plotted as a
function of average kiln gas velocity. Three different size distributions were assumed for the
particles in the kiln bed. The curves represent the power fits. The coefficients of determination ( R 2 )
are 0.983, 0.966, and 0.976 for the fits to data of size distributions 1, 2, and 3, respectively.

119

Chapter 7
CONCLUDING REMARKS, CONTRIBUTION, and
PRACTICAL IMPLICATIONS

7.1 Conclusion
A numerical modeling approach was developed to model full scale lime kilns. The model is
capable of predicting the process parameters of typical rotary lime kilns, and in particular, their
dusting signatures. The developed model consists of three sub-models: a 3D kiln gas model, a
1D/3D kiln bed model, and a model of dusting.
The 3D kiln gas model solves for fluid flow, combustion, and heat transfer in the kiln gas. This
model predicts the rate of heat transfer to the bed, temperature, and velocity profiles in the kiln
gas. The numerical approach in this model was validated against several experiments of flame
combustion, heat transfer, and a scaled pilot kiln.
The 3D kiln bed model approximately solves for the motion of the bed material. It predicts the
rate of calcination reaction based on Arrhenius rate equation in the 3D bed geometry. However,
this model does not simulate the evaporation of moisture and CO2 generation inside the bed.
This is mainly because, to the knowledge of the author, the commercial CFD software used in
this study is not capable of modeling gas/vapor flow between solid particles in the bed.
Therefore, a 1D kiln bed model was developed to predict bed properties (including moisture
content, degree of calcination, and temperature) along the length of the kiln. The 1D bed model
uses the rate of heat transfer (obtained from the 3D kiln gas model) and solves the governing
equations (continuity, energy and rate equations) to determine bed properties.
A model of dusting is also developed that: 1) calculates the minimum ambient gas velocity for
different particle sizes; 2) estimates probability of pickup based on the turbulent fluctuations of
the gas velocity, obtained from the 3D kiln gas model; 3) estimates rate of pickup based on the
size distribution of particles in bed (experimentally measured, or assumed), surface area of the
120

dusting zone (obtained from 1D kiln bed model), probability of the particle pickup, and the
particles characteristic time of pickup.
The developed dusting model was tested for 12 different operating conditions of a kiln and it was
concluded that the rate of dust generation is mainly affected by the kiln gas velocity. Effects of
other operating parameters (such as fuel flow rate, excess air, and primary-to-secondary air ratio)
were minor compared to the effect of overall gas velocity in the kiln gas region. It was found
that the rate of pickup of dust particles (with experimentally measured size distribution) is
proportional to the ambient kiln gas velocity to the power of 6.3.
To study effect of size distribution on the rate of dusting, the dusting rate was also found for few
other (assumed) size distributions. It was observed that the rate of dusting decreases when
average particle size is increased. Similarly, it was found that the dusting rate is proportional to
the gas velocity to the power of a constant factor. This constant factor varies between 5.0 and
8.9, depending on the size distribution.
The presented work suggests that numerical modeling is an effective tool to model combustion,
heat transfer, and radiation in the gas region of rotary kilns. It also suggests that coupling of a
3D gas model with a 1D bed model provides an effective tool to predict bed properties.

7.2 Practical Implications


1) The present work provides a useful tool to predict dust formation and energy consumption in
rotary lime kilns. Understanding factors affecting the extent of dusting is necessary for devising
a control strategy to improve kiln operation.
2) The effect of change of fuel and air flow rates can be analyzed using the gas model which
includes combustion and heat transfer models. For kilns equipped with scrubbers, dust is
recycled with pre-coat filters that bring more water to the kiln thereby greatly affecting the kiln
thermal efficiency. The model can be further developed to predict kilns thermal efficiency.

121

3) Calcination of lime was modeled using a one-dimensional model to predict the evaporation
rate. One-dimensional model coupled with three-dimensional model for the gas is capable of
predicting the calcination rate as well as evaporation rate along the kiln.
4) The developed model of dusting is capable of estimating the rate of dust generated.
5) Dust generation and dispersion was also modeled by randomly distributing particles on the top
surface of the bed. A high dust recycling load will reduce the kiln capacity and increase burden
on precipitators, scrubbers and pre-coat filters. The model is capable of predicting dust load
inside the kiln.

For full applicability of the present models on the industrial scale kilns,

calibration of the model based on industrial experience is required.

7.3 Concluding Remarks


Due to the complex phenomena that occur inside the kiln, the presented modeling approach
assumes several simplifying assumptions and employs several semi-empirical correlations.
Although the presented modeling approach has been extensively validated, it has not been
validated against the actual kiln data. Thus, one should be cautious in using the presented
numbers for direct applications to kilns without validation.
The presented modeling approach can be reliably used for comparative studies. This means that
effect of changing operating conditions of the kiln can be predicted.

This is particularly

important in making currently operational kilns more efficient. It can also be utilized for training
of kiln operators, as well as sensitivity studies and design/operation optimization of rotary kilns.
The presented approach was initially developed to study dust generation and effect of kilns
operating conditions on dusting.

However, it has the potential to be generalized to study

functionality of other process parameters as well.

122

7.4 Future Work


This work can be improved in the following ways:
1) Effect of changing mud-flow-rate and fill-rate (percentage of kilns cross-sectional area
occupied by mud) can be studied. Fill-rate is directly related to the production rate. Low fillrates can lead to over-burning of lime, while high fill-rates may lead to lower exit temperatures
and under-calcination of the product. Estimating the effect of varying fill-rate can be used to
optimize the kiln operation and efficiency. Results of such work can also be used to estimate the
effect of fill-rate and mud-flow-rate on dust load.
2) It will be very useful to add the effect of ring formation to the developed numerical model.
Ring formation reduces the cross-sectional area, increases the local gas velocity, and enhances
dust formation. Therefore, it is essential to take its effect into account. Furthermore, adding the
effect of ring formation will enhance the model prediction, and also, enables the kiln operators to
adjust operating conditions of the kiln for optimum quality and performance.
3) Effect of Non-Condensable-Gas (NCG) was not included in the models. NCG is often used in
kiln operations and its effect needs to be taken into account.
4) Chain section was not considered in the presented model, but its effect in enhancing heat
transfer to mud is considerable. Chain section affects the rate of mud drying, and therefore,
affects the rate of dusting. Addition of a module to take into account the effect of chains will
enhance the model predictions.

123

REFERENCES
1. Aloqaily, A., D.C.S. Kuhn, P.E. Sullivan, H.N. Tran, Flame length in lime kilns with
separate NCG burner, Engineering, Pulping & Environmental Conference Proceedings,
Atlanta (GA), November 2006.
2. Aloqaily, A., A study of the aerodynamics inside rotary kiln with two burners, PhD Thesis,
Department of Chemical Engineering, University of Toronto, 2008.
3. Alyaser, A.H., Fluid flow and combustion in rotary kilns, PhD Thesis, Department of Metals
and Materials Engineering, University of British Columbia, 1998
4. Bagnold, R.A., Physics of blown sands and desert dunes, William Morrow, New York, 1941.
5. Barr, P.V., J.K. Brimacombe, A.P. Watkinson, A heat-transfer model for the rotary kiln: 1.
Pilot kiln trials, Metallurgical Transactions B, Process metallurgy, v.20, n.3, 391-402, 1989.
6. Barr, P.V., J.K. Brimacombe, A.P. Watkinson, A heat-transfer model for the rotary kiln:
2.Development of the cross section model, Metallurgical Transactions B, Process metallurgy,
v.20B, n.3, 403-419, 1989.
7. Baxter, L.L., and P.J. Smith, Turbulent dispersion of particles: the STP model, Energy &
Fuels, v.7, n.6, 852-859, 1993.
8. Baxter, L.L., Turbulent transport of particles, PhD thesis, Brigham Young University, Provo,
Utah, 1989.
9. Beker, H.A., H.C. Hottel, and G.C.Williams, Concentration intermittency in jets, Tenth
International Symposium on Combustion, 1253-1263, 1965.
10. Beker, H.A., H.C. Hottel, and G.C.Williams, Concentration fluctuations in ducted turbulent
jets, Tenth International Symposium on Combustion, P7, 1963.
11. Besnard, D., and F.H. Harlow, Nonspherical particles in two-phase flow, Int. J. Multiphase
Flow, v. 12, n. 6, 891-912, 1986.

124

12. Boateng, A.A., P.V.Barr, A Thermal model for rotary kiln including heat transfer within the
bed, Int. J. Heat Mass Transfer, V39, No10, 2131-2147, 1996
13. Boateng, A.A., Rotary Kilns: Transport phenomena and transport processes, Elsevier, 2008.
14. Boateng, A.A., P.V.Barr, Modeling of particle mixing and segregation in the transverse plane
of a rotary kiln, Chemical Engineering Science, v. 51, n.17, 4167-4181, 1996.
15. Burry, D., and G. Bergeles, Dispersion of particles in anisotropic turbulent flows, Int. J.
Multiphase Flow, v.19, n.4, 651-664, 1993.
16. Cabrejos, F.J., G.E. Klinzing, Incipient motion of solid particles in horizontal pneumatic
conveying, Journal of Powder Technology, v.72, 51-61, 1992.
17. Cabrejos, F.J., G.E. Klinzing, Pickup and saltation mechanisms of solid particles in
horizontal pneumatic transport, Powder Technology, v.79, 173-186, 1994
18. Cammino, A.K., Pierre Begin, Scot, A. Drennan, Case study: Multiple fuel firing, AFRC
Spring meeting, Ottawa, Canada, 1997
19. Cengel Y.A., J. M. Cimbala, Fluid Mehcanics: fundamentals and applications, 2nd edition,
McGraw Hill Book Company, New York, 2006.
20. Cengel Y.A., Heat Transfer: a practical approach, 2nd edition, McGraw Hill Book Company,
New York, 2003.
21. Crowe, C., M. Sharma, and D. Stock, The particle-source-in cell (PSI-CELL) model for gasdroplet flows, J. of Fluids Engineering, v. 99, n.2, 325-332, 1977
22. Crowe, C., M. Sommerfeld, Y. Tsuji, Multiphase Flows with Droplets and Particles, CRC
press, 1998.
23. Currie, I.G., Fundamental Mechanics of Fluids, CRC Press, 2003

125

24. Dhanjal, S.K., P.V. Barr, and A.P. Watkinson, The rotary kiln: An investigation of bed heat
transfer in the transverse plane, Metallurgical and Materials Transactions B, v.35, n.6,
December, 2004
25. Elghobashi, S., Particle-laden turbulent flows: direct simulation and closure models,
J.Applied Scientific Research 48, 301-314, 1991.
26. FLUENT Manual, ANSYS Inc., Canonsburg, Pa, USA, 2008
27. Georgallis, M., P. Nowak, M. Salcudean, and I.S. Gartshore, Mathematical modeling of lime
kiln, Internal Report, Department of Mechanical Engineering, University of British Columbia
28. Giordano,

P.,

D.Lentini,

Combustion-radiation-turbulence

interaction

modeling

in

absorbing/emitting nonpremixed flames,Combustion Science and Technology, v. 172, n.1,


1-22, 2001.
29. Gorog, J.P., T.N. Adams, an J.K. Brimacombe, Heat Transfer from Flames in a Rotary Kiln,
Metallurgical Transactions B, v14B, Spetember 1983, 411-423
30. Graebel, W.P., Engineering Fluid Mechanics, Taylor & Francis Group, 2001
31. Guruz, A.G., H.K.Guruz, S.Osuwan, and F.R. Stewart, Aerodynamics of a confined burning
jet, Combustion Science and Technology, v.9, 103-110, 1974.
32. Hennick, E.A., and M.F. Lightstone, Comparison of stochastic separated flow models for
particle dispersion in turbulent flows, Energy & Fuels, v.14, n.1, 95-103, 2000.
33. Hennick, E.A., A comparison of stochastic separated flow models for particle dispersion in
turbulent flows, MASc. Thesis, University of Toronto, 1998.
34. Hough, G., Chemical recovery in the alkaline pulping processes: a project of the Alkaline
Pulping Committee of the Pulp Manufacture Division, TAPPI, 1985.
35. Incropora, F.P., and D.P. De Witt, Fundamentals of heat and mass transfer, 3rd Edition, John
Whiley, 1990.

126

36. Jain, S., Three-dimensional simulation of turbulent particle dispersion, PhD thesis,
University of Utah, Utah, 1995.
37. Kontogeorgos, D.A., E.P. Keramida, M.A. Founti, Assessment of simplified thermal
radiation models for engineering calculations in natural gas-fired furnace, International
Journal of Heat and Mass Transfer v. 50, n. 25-26, 5260-5268, 2007.
38. Kalman, H., A. Satran, D. Meir, and E. Rabinovich, Pickup (critical) velocity of particles,
Powder technology, v. 160, n.2, 103-113, 2005.
39. Launder, B.E., D.B., Spalding, Lectures in mathematical models of turbulence, Academic
Press, London, 1972.
40. Liakos, H.H., M.A. Founti, and N.C. Markatos, Modeling of stretched natural gas diffusion
flame, Applied Mathematical Modeling, v. 24, n. 5-6, 419-435, 2000.
41. Litchford, R. J., and S. M. Jeng, Efficient statistical transport model for turbulent particle
dispersion in sprays, AIAA Journal, v. 29, n. 9, 1443-1451, 1991.
42. Litchford, R.J., and S. Jeng, Probability density function shape sensitivity in the statistical
modeling of turbulent particle dispersion, AIAA Journal, v. 30, n. 10, 2546-2549, 1992.
43. Magnussen, B.F., B.H. Hjertager, On mathematical modeling of turbulent combustion with
special emphasis on soot formation and combustion, In 16th Symp. (Intl.) on Combustion,
Combustion Inst., 719-729, 1976.
44. Mahmud, T., S.K. Sangha, M. Costa, A. Santos, Experimental and computational study of a
lifted, non-premixed turbulent free jet flame, Fuel, v.86, n. 5-6, 793806, 2007.
45. Milojevic,D., Lagrangian stochastic-deterministic (LSD) predictions of particle dispersion in
turbulence, Particle and Particle Systems Characterization, v.7, n.4, 181-190, 1990.
46. Modest, M.F., Radiative Heat Transfer, McGraw-Hill, New York.
47. Mounir, N., Analysis of Liquefaction Induced Lateral Ground Displacement Using Smoothed
Particle Hydrodynamics, Ph.D. Thesis, University of Tsukuba, Japan, 2006
127

48. Mujumdar, K. S., A. Arora, and V.V. Ranade, Modeling of rotary cement kilns: Application
in Energy Consumption, Ind. Eng. Chem. Res. v.45, 2315-2330, 2006.
49. Mujumdar K.S. and V.V. Ranade, CFD modeling of rotary cement kilns, Asia-Pac. J. Chem.
Eng., v. 3, 106-118, 2008.
50. Patankar, N.A., and D.D. Joseph, Modeling and numerical simulation of particulate flows by
the Eulerian-Lagrangian approach, Int. J. of Multiphase Flow, v. 27, n. 10, 1659-1684, 2001.
51. Paul, M., A three dimensional numerical model to predict temperature and degree of
calcination in the solids bed in lime kilns, Department of Chemical Engineering and Applied
Chemistry, University of Toronto, 2004.
52. Peray, K.E., W.J. Waddell, The Rotary Cement Kiln, Chemical Publishing Co., NY 1972.
53. Poinsot, T. and

D. Veynante, Theoretical and Numerical Combustion, 2nd Ed., R.T.

Edwards, 2005
54. Ranade, V.V., Modeling of rotary cement kiln, Third International Conference on CFD in the
Mineral and Process Industries, CSIRO, Melbourne, Australia, 2003.
55. Rosendahl, L., Using a multi-parameter particle shape description to predict the motion of
non-spherical particle shapes, Applied Mathematical Modelling, v.24, n.1, 11-25, 2000.
56. Salcudean, M.E., P. Sabhapathy, and F. Weinberg, Numerical study of free and forced
convection in the LEC growth of GaAs, J. of Crystal Growth, 94, 522-526, 1989.
57. Sharma, N., and N.A. Patankar, A fast computation technique for the direct numerical
simulation of rigid particulate flows, J. of Comp. Physics 205, 439-457, 2005.
58. Shirolkar, J.S., M.Q. McQuay, Probability density function propagation model for turbulent
particle dispersion, Int. J. Multiphase Flow, v. 24, n.4, 663-678, 1998.
59. Shirolkar, J.S., and M.Q. McQuay,

A PDF propagation approach to model turbulent

dispersion in swirling flows, Eur. J. Mech. B-Fluids, v.20, n.5, 699-726, 2001.

128

60. Siegel, R., J. R. Howell, Thermal Radiation Heat Transfer, Hemisphere Publishing
Corporation, Washington DC, 1992.
61. Siegel, R., J.R. Howell, Thermal Radiation Heat Transfer, second edition, 1981.
62. Snyder, W.H., and J.L. Lumley, Some measurements of particle velocity autocorrelation
function in a turbulent flow, Journal of Fluid Mechanics, v. 48, part 1, 41-71, 1971.
63. Stoomer, P., Turbulence and OH structure in flames, Delft University, Delft, 1995.
64. Subramanian R.Sh., http://www.clarkson.edu/subramanian/ch301/notes/dragsphere.pdf
65. Tackie, E.N., A.P. Watkinson, J.K. Brimacombe, Experimental study of the elutriation of
particles from rotary kilns, The Canadian Journal of Chemical Engineering, v. 67, 1989.
66. Taylor, G.I., Diffusion by continuous movements, Proc. London Math. Soc., v. 20, n. 20,
196-211, 1920.
67. Tennekes, H., and J.L. Lumley, A First Course in Turbulence, The MIT Press, 1972.
68. Venkatesh, V., Lime Reburning: The rotary lime kiln: Chemical recovery in the alkaline
pulping processes,Revised Edition, Co-Edited by Robert P. Green and Gerald Hough, 1992.
69. Walkate, P.J., Boundary layer meteorology, v. 39, 175-190, 1987.
70. Watkinson, A.P., J.K. Brimacombe, Limestone Calcination in a Rotary Kiln, Metallurgical
Transactions B, v. 13B, 1982.
71. Wilcox, D.C., Turbulence modeling for CFD, Second edition, DCW Industries, 2002.
72. Yi J., and M.W. Plesniak, Dispersion of a particle-laden air jet in a confined rectangular
cross-flow, Powder Technology, v.125, n. 2-3, 168-178, 2002.
73. Zhou, Q., M.A. Leschziner, 8th Turbulent Shear Flows Symposium, 1991.

129

Appendix A: Benchmarking the heat transfer model

A.1. Modeling Turbulence Mixing and Reaction Rates

Magnussen et. al. [43] provide a turbulence-chemistry interaction model called eddy dissipation.
In this model, the net rate of production of species i in reaction r, denoted by Ri ,r is given by the
smaller of the two equations below:
Ri , r vi' ,r M ,i A

Y
min ' R
k R vR ,r M , R

Ri , r vi' ,r M ,i AB

PY

k N v ''j ,r M , j
j

(A.1)

(A.2)

where YP is the mass fraction of any product species, P


YR is the mass fraction of a particular reactant, R

A is an empirical constant equal to 4.0


B is an empirical constant equal to 0.5
In equations (A.1) and (A.2), the chemical rate is governed by the large eddy mixing timescale,

/ as described by Eddy break-up model of Spalding [71]. In the mixing controlled regime,
the reaction rate Ri ,r is calculated using Eddy break-up law, whereas in kinetically-controlled
regime, the reaction rate Ri ,r is calculated according to Arrhenius formula [40]. Combustion
proceeds whenever / 0 .
A.2. External Flow Heat Transfer (from wall to outside air):

Another factor that controls heat transfer rates in the kiln is heat convection from the outer wall
of the kiln to the ambient air. Similar to the internal flow heat transfer, modeling this flux is
possible both by fine-meshing the outside geometry of the kiln (and resolving its thermal
130

boundary layer), or by using semi-empirical correlations. However, unlike the case of internal
flow, the external flow is not complex: it is just similar to a simple case of natural convection of
a hot rod placed in static air. For this case, h is already provided by different heat transfer
textbooks [35]:

k
0.387 RaD1/ 6

ha 0.6
8/ 27
9
/16
D
1 0.559 / Pr

105 RaD 1012


(A.3)

where D is the external diameter of the cylinder exposed to free convection and
RaD

g (Ts T ) D 3

(A.4)

As shown in Figure A.1, a two dimensional geometry of external flow around a pipe was created
and meshed. Because of symmetry, only half of the geometry was modeled. Ambient air was
assumed to be at T = 300K and a constant heat flux of q = -10 W/m2 was assumed to heat the
ambient air around the cylinder. Far-air boundaries were set to be at atmospheric pressure. The
domain was meshed and meshes near the cylinder wall were consecutively refined to generate a
range of y+ values between 0.45 and 14.

Figure A.1 shows grid structure and extent of

refinement around the pipe.

131

Symmetry line

Figure A.1.Mesh structure and refinement in the external layer of a pipe.

132

The natural convection flow was then solved within the domain using enhanced wall function.
Figure A.2 shows temperature contours and formation of thermal boundary layer around the
cylinder. Convection heat transfer coefficient, h, was then found using
h

q
10Wm 2

(Tambient Twall ) (300 K Twall )

(A.5)

at different angular locations around the pipe. Figure A.3 plots such h functionality for different
mesh sizes and compares them with semi-empirical correlation (A.3).

Figure A.2. Temperature contour around a circular cylinder.

133

Figure A.3. Convection Coefficient for different mesh sizes.

134

Appendix B: Results of 1D Model


1D modeling approach was used to predict the composition of kiln bed for the last 11 cases
defined in Table 6.3. Average air and bed temperatures were taken from 3D CFD modeling of
the gas phase. Calcination reaction was modeled in 1D and bed compositions were determined.
The results are presented here.

135

Figure B.1. Case 2 of the 12 operating condition cases in Table 6.3.

136

Figure B.2. Case 3 of the 12 operating condition cases in Table 6.3.

137

Figure B.3. Case 4 of the 12 operating condition cases in Table 6.3.

138

Figure B.4. Case 5 of the 12 operating condition cases in Table 6.3.

139

Figure B.5. Case 6 of the 12 operating condition cases in Table 6.3.

140

Figure B.6. Case 7 of the 12 operating condition cases in Table 6.3.

141

Figure B.7. Case 8 of the 12 operating condition cases in Table 6.3.

142

Figure B.8. Case 9 of the 12 operating condition cases in Table 6.3.

143

Figure B.9. Case 10 of the 12 operating condition cases in Table 6.3.

144

Figure B.10. Case 11 of the 12 operating condition cases in Table 6.3.

145

Figure B.11. Case 12 of the 12 operating condition cases in Table 6.3.

146

Appendix C: Material Properties


Thermal Conductivity of Combustion Species
Temperature
o
K

CO2
J/Kg/K

H2O
J/Kg/K

N2
J/Kg/K

O2
J/Kg/K

200

736

1792

1028

890

300

847

1859

1038

917

400

938

1913

1047

944

500

1013

1963

1058

972

700

1128

2076

1096

1029

1000

1235

2295

1170

1092

1200

1277

2437

1204

1112

1500

1325

2617

1243

1140

2000

1373

2841

1285

1181

2500

1399

2995

1309

1215

3000

1414

3099

1322

1245

3500

1425

3171

1332

1271

4000

1438

3224

1341

1295

Properties of Combusting Gas [49]:


Property

Value

Absorption Coefficient (m-1)

a 0.32 0.28e T /1135 K

Isotropic Scattering Coefficient

0.13 m-1

Heat Capacity (J/kg/K)

c 0.106T 1173 , 1550 [29]

Viscosity (kg/m/s)

1.672 106 T 1.058 105

Flame Emissivity

0.25 [29]

Fuel (Natural Gas) Density

0.68 kg/m3 @ 1 atm, 15C

Fuel (Natural Gas) Gross Heating Value

5.521x107 J/kg [29]

Fuel (Natural Gas) Adiabatic Temperature

2327 K [29]

147

Properties of Bed [48]:


Property

Value

Heat Capacity (J/kg/K)

1088

Thermal conductivity (W/m/K)

0.4 [12], 0.5

Bulk Density (kg/m3)

670

Material Density (kg/m3)

2830

Emissivity

0.8 [29], 0.9

Properties of Kiln Walls / Refractory Bricks: [12]


Property

Value

Heat Capacity (J/kg/K)

1000 1200

Thermal conductivity (W/m/K)

0.4, 4.0 [48]

Density (kg/m3)

1334

Refractory Emissivity

0.9 [48], 0.8 [29]

Kiln shell Emissivity

0.78 [48]

Average Heat Transfer Coefficients [29]


- Flame to Solid

h 50 W / m 2 / K

- Flame to Wall

h 20 W / m 2 / K

- Shell to Air

h 10 W / m 2 / K

Calcination Reaction Properties [51]:


Property

Value

Heat of Reaction (kJ/mole)

179.4

Pre-Activation Constant (1/s)

108

Activation Energy (J)

220728

148

Evaporation Reaction Properties [51]:


Property

Value

Heat of Reaction (kJ/mole)

85

Pre-Activation Constant (1/s)

2.55107

Heat of Evaporation (kJ/kg)

2257.2

149

You might also like